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 if (ckWARN(WARN_MISC))
4009 Perl_warner(aTHX_ packWARN(WARN_MISC),
4010 "Undefined value assigned to typeglob");
4013 GV *gv = gv_fetchsv(sstr, GV_ADD, SVt_PVGV);
4014 if (dstr != (const SV *)gv) {
4016 gp_free(MUTABLE_GV(dstr));
4017 GvGP(dstr) = gp_ref(GvGP(gv));
4021 else if (sflags & SVp_POK) {
4025 * Check to see if we can just swipe the string. If so, it's a
4026 * possible small lose on short strings, but a big win on long ones.
4027 * It might even be a win on short strings if SvPVX_const(dstr)
4028 * has to be allocated and SvPVX_const(sstr) has to be freed.
4029 * Likewise if we can set up COW rather than doing an actual copy, we
4030 * drop to the else clause, as the swipe code and the COW setup code
4031 * have much in common.
4034 /* Whichever path we take through the next code, we want this true,
4035 and doing it now facilitates the COW check. */
4036 (void)SvPOK_only(dstr);
4039 /* If we're already COW then this clause is not true, and if COW
4040 is allowed then we drop down to the else and make dest COW
4041 with us. If caller hasn't said that we're allowed to COW
4042 shared hash keys then we don't do the COW setup, even if the
4043 source scalar is a shared hash key scalar. */
4044 (((flags & SV_COW_SHARED_HASH_KEYS)
4045 ? (sflags & (SVf_FAKE|SVf_READONLY)) != (SVf_FAKE|SVf_READONLY)
4046 : 1 /* If making a COW copy is forbidden then the behaviour we
4047 desire is as if the source SV isn't actually already
4048 COW, even if it is. So we act as if the source flags
4049 are not COW, rather than actually testing them. */
4051 #ifndef PERL_OLD_COPY_ON_WRITE
4052 /* The change that added SV_COW_SHARED_HASH_KEYS makes the logic
4053 when PERL_OLD_COPY_ON_WRITE is defined a little wrong.
4054 Conceptually PERL_OLD_COPY_ON_WRITE being defined should
4055 override SV_COW_SHARED_HASH_KEYS, because it means "always COW"
4056 but in turn, it's somewhat dead code, never expected to go
4057 live, but more kept as a placeholder on how to do it better
4058 in a newer implementation. */
4059 /* If we are COW and dstr is a suitable target then we drop down
4060 into the else and make dest a COW of us. */
4061 || (SvFLAGS(dstr) & CAN_COW_MASK) != CAN_COW_FLAGS
4066 (sflags & SVs_TEMP) && /* slated for free anyway? */
4067 !(sflags & SVf_OOK) && /* and not involved in OOK hack? */
4068 (!(flags & SV_NOSTEAL)) &&
4069 /* and we're allowed to steal temps */
4070 SvREFCNT(sstr) == 1 && /* and no other references to it? */
4071 SvLEN(sstr) && /* and really is a string */
4072 /* and won't be needed again, potentially */
4073 !(PL_op && PL_op->op_type == OP_AASSIGN))
4074 #ifdef PERL_OLD_COPY_ON_WRITE
4075 && ((flags & SV_COW_SHARED_HASH_KEYS)
4076 ? (!((sflags & CAN_COW_MASK) == CAN_COW_FLAGS
4077 && (SvFLAGS(dstr) & CAN_COW_MASK) == CAN_COW_FLAGS
4078 && SvTYPE(sstr) >= SVt_PVIV && SvTYPE(sstr) != SVt_PVFM))
4082 /* Failed the swipe test, and it's not a shared hash key either.
4083 Have to copy the string. */
4084 STRLEN len = SvCUR(sstr);
4085 SvGROW(dstr, len + 1); /* inlined from sv_setpvn */
4086 Move(SvPVX_const(sstr),SvPVX(dstr),len,char);
4087 SvCUR_set(dstr, len);
4088 *SvEND(dstr) = '\0';
4090 /* If PERL_OLD_COPY_ON_WRITE is not defined, then isSwipe will always
4092 /* Either it's a shared hash key, or it's suitable for
4093 copy-on-write or we can swipe the string. */
4095 PerlIO_printf(Perl_debug_log, "Copy on write: sstr --> dstr\n");
4099 #ifdef PERL_OLD_COPY_ON_WRITE
4101 if ((sflags & (SVf_FAKE | SVf_READONLY))
4102 != (SVf_FAKE | SVf_READONLY)) {
4103 SvREADONLY_on(sstr);
4105 /* Make the source SV into a loop of 1.
4106 (about to become 2) */
4107 SV_COW_NEXT_SV_SET(sstr, sstr);
4111 /* Initial code is common. */
4112 if (SvPVX_const(dstr)) { /* we know that dtype >= SVt_PV */
4117 /* making another shared SV. */
4118 STRLEN cur = SvCUR(sstr);
4119 STRLEN len = SvLEN(sstr);
4120 #ifdef PERL_OLD_COPY_ON_WRITE
4122 assert (SvTYPE(dstr) >= SVt_PVIV);
4123 /* SvIsCOW_normal */
4124 /* splice us in between source and next-after-source. */
4125 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4126 SV_COW_NEXT_SV_SET(sstr, dstr);
4127 SvPV_set(dstr, SvPVX_mutable(sstr));
4131 /* SvIsCOW_shared_hash */
4132 DEBUG_C(PerlIO_printf(Perl_debug_log,
4133 "Copy on write: Sharing hash\n"));
4135 assert (SvTYPE(dstr) >= SVt_PV);
4137 HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)))));
4139 SvLEN_set(dstr, len);
4140 SvCUR_set(dstr, cur);
4141 SvREADONLY_on(dstr);
4145 { /* Passes the swipe test. */
4146 SvPV_set(dstr, SvPVX_mutable(sstr));
4147 SvLEN_set(dstr, SvLEN(sstr));
4148 SvCUR_set(dstr, SvCUR(sstr));
4151 (void)SvOK_off(sstr); /* NOTE: nukes most SvFLAGS on sstr */
4152 SvPV_set(sstr, NULL);
4158 if (sflags & SVp_NOK) {
4159 SvNV_set(dstr, SvNVX(sstr));
4161 if (sflags & SVp_IOK) {
4162 SvIV_set(dstr, SvIVX(sstr));
4163 /* Must do this otherwise some other overloaded use of 0x80000000
4164 gets confused. I guess SVpbm_VALID */
4165 if (sflags & SVf_IVisUV)
4168 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_NOK|SVp_NOK|SVf_UTF8);
4170 const MAGIC * const smg = SvVSTRING_mg(sstr);
4172 sv_magic(dstr, NULL, PERL_MAGIC_vstring,
4173 smg->mg_ptr, smg->mg_len);
4174 SvRMAGICAL_on(dstr);
4178 else if (sflags & (SVp_IOK|SVp_NOK)) {
4179 (void)SvOK_off(dstr);
4180 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_IVisUV|SVf_NOK|SVp_NOK);
4181 if (sflags & SVp_IOK) {
4182 /* XXXX Do we want to set IsUV for IV(ROK)? Be extra safe... */
4183 SvIV_set(dstr, SvIVX(sstr));
4185 if (sflags & SVp_NOK) {
4186 SvNV_set(dstr, SvNVX(sstr));
4190 if (isGV_with_GP(sstr)) {
4191 /* This stringification rule for globs is spread in 3 places.
4192 This feels bad. FIXME. */
4193 const U32 wasfake = sflags & SVf_FAKE;
4195 /* FAKE globs can get coerced, so need to turn this off
4196 temporarily if it is on. */
4198 gv_efullname3(dstr, MUTABLE_GV(sstr), "*");
4199 SvFLAGS(sstr) |= wasfake;
4202 (void)SvOK_off(dstr);
4204 if (SvTAINTED(sstr))
4209 =for apidoc sv_setsv_mg
4211 Like C<sv_setsv>, but also handles 'set' magic.
4217 Perl_sv_setsv_mg(pTHX_ SV *const dstr, register SV *const sstr)
4219 PERL_ARGS_ASSERT_SV_SETSV_MG;
4221 sv_setsv(dstr,sstr);
4225 #ifdef PERL_OLD_COPY_ON_WRITE
4227 Perl_sv_setsv_cow(pTHX_ SV *dstr, SV *sstr)
4229 STRLEN cur = SvCUR(sstr);
4230 STRLEN len = SvLEN(sstr);
4231 register char *new_pv;
4233 PERL_ARGS_ASSERT_SV_SETSV_COW;
4236 PerlIO_printf(Perl_debug_log, "Fast copy on write: %p -> %p\n",
4237 (void*)sstr, (void*)dstr);
4244 if (SvTHINKFIRST(dstr))
4245 sv_force_normal_flags(dstr, SV_COW_DROP_PV);
4246 else if (SvPVX_const(dstr))
4247 Safefree(SvPVX_const(dstr));
4251 SvUPGRADE(dstr, SVt_PVIV);
4253 assert (SvPOK(sstr));
4254 assert (SvPOKp(sstr));
4255 assert (!SvIOK(sstr));
4256 assert (!SvIOKp(sstr));
4257 assert (!SvNOK(sstr));
4258 assert (!SvNOKp(sstr));
4260 if (SvIsCOW(sstr)) {
4262 if (SvLEN(sstr) == 0) {
4263 /* source is a COW shared hash key. */
4264 DEBUG_C(PerlIO_printf(Perl_debug_log,
4265 "Fast copy on write: Sharing hash\n"));
4266 new_pv = HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr))));
4269 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4271 assert ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS);
4272 SvUPGRADE(sstr, SVt_PVIV);
4273 SvREADONLY_on(sstr);
4275 DEBUG_C(PerlIO_printf(Perl_debug_log,
4276 "Fast copy on write: Converting sstr to COW\n"));
4277 SV_COW_NEXT_SV_SET(dstr, sstr);
4279 SV_COW_NEXT_SV_SET(sstr, dstr);
4280 new_pv = SvPVX_mutable(sstr);
4283 SvPV_set(dstr, new_pv);
4284 SvFLAGS(dstr) = (SVt_PVIV|SVf_POK|SVp_POK|SVf_FAKE|SVf_READONLY);
4287 SvLEN_set(dstr, len);
4288 SvCUR_set(dstr, cur);
4297 =for apidoc sv_setpvn
4299 Copies a string into an SV. The C<len> parameter indicates the number of
4300 bytes to be copied. If the C<ptr> argument is NULL the SV will become
4301 undefined. Does not handle 'set' magic. See C<sv_setpvn_mg>.
4307 Perl_sv_setpvn(pTHX_ register SV *const sv, register const char *const ptr, register const STRLEN len)
4310 register char *dptr;
4312 PERL_ARGS_ASSERT_SV_SETPVN;
4314 SV_CHECK_THINKFIRST_COW_DROP(sv);
4320 /* len is STRLEN which is unsigned, need to copy to signed */
4323 Perl_croak(aTHX_ "panic: sv_setpvn called with negative strlen");
4325 SvUPGRADE(sv, SVt_PV);
4327 dptr = SvGROW(sv, len + 1);
4328 Move(ptr,dptr,len,char);
4331 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4336 =for apidoc sv_setpvn_mg
4338 Like C<sv_setpvn>, but also handles 'set' magic.
4344 Perl_sv_setpvn_mg(pTHX_ register SV *const sv, register const char *const ptr, register const STRLEN len)
4346 PERL_ARGS_ASSERT_SV_SETPVN_MG;
4348 sv_setpvn(sv,ptr,len);
4353 =for apidoc sv_setpv
4355 Copies a string into an SV. The string must be null-terminated. Does not
4356 handle 'set' magic. See C<sv_setpv_mg>.
4362 Perl_sv_setpv(pTHX_ register SV *const sv, register const char *const ptr)
4365 register STRLEN len;
4367 PERL_ARGS_ASSERT_SV_SETPV;
4369 SV_CHECK_THINKFIRST_COW_DROP(sv);
4375 SvUPGRADE(sv, SVt_PV);
4377 SvGROW(sv, len + 1);
4378 Move(ptr,SvPVX(sv),len+1,char);
4380 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4385 =for apidoc sv_setpv_mg
4387 Like C<sv_setpv>, but also handles 'set' magic.
4393 Perl_sv_setpv_mg(pTHX_ register SV *const sv, register const char *const ptr)
4395 PERL_ARGS_ASSERT_SV_SETPV_MG;
4402 =for apidoc sv_usepvn_flags
4404 Tells an SV to use C<ptr> to find its string value. Normally the
4405 string is stored inside the SV but sv_usepvn allows the SV to use an
4406 outside string. The C<ptr> should point to memory that was allocated
4407 by C<malloc>. The string length, C<len>, must be supplied. By default
4408 this function will realloc (i.e. move) the memory pointed to by C<ptr>,
4409 so that pointer should not be freed or used by the programmer after
4410 giving it to sv_usepvn, and neither should any pointers from "behind"
4411 that pointer (e.g. ptr + 1) be used.
4413 If C<flags> & SV_SMAGIC is true, will call SvSETMAGIC. If C<flags> &
4414 SV_HAS_TRAILING_NUL is true, then C<ptr[len]> must be NUL, and the realloc
4415 will be skipped. (i.e. the buffer is actually at least 1 byte longer than
4416 C<len>, and already meets the requirements for storing in C<SvPVX>)
4422 Perl_sv_usepvn_flags(pTHX_ SV *const sv, char *ptr, const STRLEN len, const U32 flags)
4427 PERL_ARGS_ASSERT_SV_USEPVN_FLAGS;
4429 SV_CHECK_THINKFIRST_COW_DROP(sv);
4430 SvUPGRADE(sv, SVt_PV);
4433 if (flags & SV_SMAGIC)
4437 if (SvPVX_const(sv))
4441 if (flags & SV_HAS_TRAILING_NUL)
4442 assert(ptr[len] == '\0');
4445 allocate = (flags & SV_HAS_TRAILING_NUL)
4447 #ifdef Perl_safesysmalloc_size
4450 PERL_STRLEN_ROUNDUP(len + 1);
4452 if (flags & SV_HAS_TRAILING_NUL) {
4453 /* It's long enough - do nothing.
4454 Specfically Perl_newCONSTSUB is relying on this. */
4457 /* Force a move to shake out bugs in callers. */
4458 char *new_ptr = (char*)safemalloc(allocate);
4459 Copy(ptr, new_ptr, len, char);
4460 PoisonFree(ptr,len,char);
4464 ptr = (char*) saferealloc (ptr, allocate);
4467 #ifdef Perl_safesysmalloc_size
4468 SvLEN_set(sv, Perl_safesysmalloc_size(ptr));
4470 SvLEN_set(sv, allocate);
4474 if (!(flags & SV_HAS_TRAILING_NUL)) {
4477 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4479 if (flags & SV_SMAGIC)
4483 #ifdef PERL_OLD_COPY_ON_WRITE
4484 /* Need to do this *after* making the SV normal, as we need the buffer
4485 pointer to remain valid until after we've copied it. If we let go too early,
4486 another thread could invalidate it by unsharing last of the same hash key
4487 (which it can do by means other than releasing copy-on-write Svs)
4488 or by changing the other copy-on-write SVs in the loop. */
4490 S_sv_release_COW(pTHX_ register SV *sv, const char *pvx, SV *after)
4492 PERL_ARGS_ASSERT_SV_RELEASE_COW;
4494 { /* this SV was SvIsCOW_normal(sv) */
4495 /* we need to find the SV pointing to us. */
4496 SV *current = SV_COW_NEXT_SV(after);
4498 if (current == sv) {
4499 /* The SV we point to points back to us (there were only two of us
4501 Hence other SV is no longer copy on write either. */
4503 SvREADONLY_off(after);
4505 /* We need to follow the pointers around the loop. */
4507 while ((next = SV_COW_NEXT_SV(current)) != sv) {
4510 /* don't loop forever if the structure is bust, and we have
4511 a pointer into a closed loop. */
4512 assert (current != after);
4513 assert (SvPVX_const(current) == pvx);
4515 /* Make the SV before us point to the SV after us. */
4516 SV_COW_NEXT_SV_SET(current, after);
4522 =for apidoc sv_force_normal_flags
4524 Undo various types of fakery on an SV: if the PV is a shared string, make
4525 a private copy; if we're a ref, stop refing; if we're a glob, downgrade to
4526 an xpvmg; if we're a copy-on-write scalar, this is the on-write time when
4527 we do the copy, and is also used locally. If C<SV_COW_DROP_PV> is set
4528 then a copy-on-write scalar drops its PV buffer (if any) and becomes
4529 SvPOK_off rather than making a copy. (Used where this scalar is about to be
4530 set to some other value.) In addition, the C<flags> parameter gets passed to
4531 C<sv_unref_flags()> when unrefing. C<sv_force_normal> calls this function
4532 with flags set to 0.
4538 Perl_sv_force_normal_flags(pTHX_ register SV *const sv, const U32 flags)
4542 PERL_ARGS_ASSERT_SV_FORCE_NORMAL_FLAGS;
4544 #ifdef PERL_OLD_COPY_ON_WRITE
4545 if (SvREADONLY(sv)) {
4547 const char * const pvx = SvPVX_const(sv);
4548 const STRLEN len = SvLEN(sv);
4549 const STRLEN cur = SvCUR(sv);
4550 /* next COW sv in the loop. If len is 0 then this is a shared-hash
4551 key scalar, so we mustn't attempt to call SV_COW_NEXT_SV(), as
4552 we'll fail an assertion. */
4553 SV * const next = len ? SV_COW_NEXT_SV(sv) : 0;
4556 PerlIO_printf(Perl_debug_log,
4557 "Copy on write: Force normal %ld\n",
4563 /* This SV doesn't own the buffer, so need to Newx() a new one: */
4566 if (flags & SV_COW_DROP_PV) {
4567 /* OK, so we don't need to copy our buffer. */
4570 SvGROW(sv, cur + 1);
4571 Move(pvx,SvPVX(sv),cur,char);
4576 sv_release_COW(sv, pvx, next);
4578 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4584 else if (IN_PERL_RUNTIME)
4585 Perl_croak(aTHX_ "%s", PL_no_modify);
4588 if (SvREADONLY(sv)) {
4590 const char * const pvx = SvPVX_const(sv);
4591 const STRLEN len = SvCUR(sv);
4596 SvGROW(sv, len + 1);
4597 Move(pvx,SvPVX(sv),len,char);
4599 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4601 else if (IN_PERL_RUNTIME)
4602 Perl_croak(aTHX_ "%s", PL_no_modify);
4606 sv_unref_flags(sv, flags);
4607 else if (SvFAKE(sv) && SvTYPE(sv) == SVt_PVGV)
4614 Efficient removal of characters from the beginning of the string buffer.
4615 SvPOK(sv) must be true and the C<ptr> must be a pointer to somewhere inside
4616 the string buffer. The C<ptr> becomes the first character of the adjusted
4617 string. Uses the "OOK hack".
4618 Beware: after this function returns, C<ptr> and SvPVX_const(sv) may no longer
4619 refer to the same chunk of data.
4625 Perl_sv_chop(pTHX_ register SV *const sv, register const char *const ptr)
4631 const U8 *real_start;
4635 PERL_ARGS_ASSERT_SV_CHOP;
4637 if (!ptr || !SvPOKp(sv))
4639 delta = ptr - SvPVX_const(sv);
4641 /* Nothing to do. */
4644 /* SvPVX(sv) may move in SV_CHECK_THINKFIRST(sv), but after this line,
4645 nothing uses the value of ptr any more. */
4646 max_delta = SvLEN(sv) ? SvLEN(sv) : SvCUR(sv);
4647 if (ptr <= SvPVX_const(sv))
4648 Perl_croak(aTHX_ "panic: sv_chop ptr=%p, start=%p, end=%p",
4649 ptr, SvPVX_const(sv), SvPVX_const(sv) + max_delta);
4650 SV_CHECK_THINKFIRST(sv);
4651 if (delta > max_delta)
4652 Perl_croak(aTHX_ "panic: sv_chop ptr=%p (was %p), start=%p, end=%p",
4653 SvPVX_const(sv) + delta, ptr, SvPVX_const(sv),
4654 SvPVX_const(sv) + max_delta);
4657 if (!SvLEN(sv)) { /* make copy of shared string */
4658 const char *pvx = SvPVX_const(sv);
4659 const STRLEN len = SvCUR(sv);
4660 SvGROW(sv, len + 1);
4661 Move(pvx,SvPVX(sv),len,char);
4664 SvFLAGS(sv) |= SVf_OOK;
4667 SvOOK_offset(sv, old_delta);
4669 SvLEN_set(sv, SvLEN(sv) - delta);
4670 SvCUR_set(sv, SvCUR(sv) - delta);
4671 SvPV_set(sv, SvPVX(sv) + delta);
4673 p = (U8 *)SvPVX_const(sv);
4678 real_start = p - delta;
4682 if (delta < 0x100) {
4686 p -= sizeof(STRLEN);
4687 Copy((U8*)&delta, p, sizeof(STRLEN), U8);
4691 /* Fill the preceding buffer with sentinals to verify that no-one is
4693 while (p > real_start) {
4701 =for apidoc sv_catpvn
4703 Concatenates the string onto the end of the string which is in the SV. The
4704 C<len> indicates number of bytes to copy. If the SV has the UTF-8
4705 status set, then the bytes appended should be valid UTF-8.
4706 Handles 'get' magic, but not 'set' magic. See C<sv_catpvn_mg>.
4708 =for apidoc sv_catpvn_flags
4710 Concatenates the string onto the end of the string which is in the SV. The
4711 C<len> indicates number of bytes to copy. If the SV has the UTF-8
4712 status set, then the bytes appended should be valid UTF-8.
4713 If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<dsv> if
4714 appropriate, else not. C<sv_catpvn> and C<sv_catpvn_nomg> are implemented
4715 in terms of this function.
4721 Perl_sv_catpvn_flags(pTHX_ register SV *const dsv, register const char *sstr, register const STRLEN slen, const I32 flags)
4725 const char * const dstr = SvPV_force_flags(dsv, dlen, flags);
4727 PERL_ARGS_ASSERT_SV_CATPVN_FLAGS;
4729 SvGROW(dsv, dlen + slen + 1);
4731 sstr = SvPVX_const(dsv);
4732 Move(sstr, SvPVX(dsv) + dlen, slen, char);
4733 SvCUR_set(dsv, SvCUR(dsv) + slen);
4735 (void)SvPOK_only_UTF8(dsv); /* validate pointer */
4737 if (flags & SV_SMAGIC)
4742 =for apidoc sv_catsv
4744 Concatenates the string from SV C<ssv> onto the end of the string in
4745 SV C<dsv>. Modifies C<dsv> but not C<ssv>. Handles 'get' magic, but
4746 not 'set' magic. See C<sv_catsv_mg>.
4748 =for apidoc sv_catsv_flags
4750 Concatenates the string from SV C<ssv> onto the end of the string in
4751 SV C<dsv>. Modifies C<dsv> but not C<ssv>. If C<flags> has C<SV_GMAGIC>
4752 bit set, will C<mg_get> on the SVs if appropriate, else not. C<sv_catsv>
4753 and C<sv_catsv_nomg> are implemented in terms of this function.
4758 Perl_sv_catsv_flags(pTHX_ SV *const dsv, register SV *const ssv, const I32 flags)
4762 PERL_ARGS_ASSERT_SV_CATSV_FLAGS;
4766 const char *spv = SvPV_const(ssv, slen);
4768 /* sutf8 and dutf8 were type bool, but under USE_ITHREADS,
4769 gcc version 2.95.2 20000220 (Debian GNU/Linux) for
4770 Linux xxx 2.2.17 on sparc64 with gcc -O2, we erroneously
4771 get dutf8 = 0x20000000, (i.e. SVf_UTF8) even though
4772 dsv->sv_flags doesn't have that bit set.
4773 Andy Dougherty 12 Oct 2001
4775 const I32 sutf8 = DO_UTF8(ssv);
4778 if (SvGMAGICAL(dsv) && (flags & SV_GMAGIC))
4780 dutf8 = DO_UTF8(dsv);
4782 if (dutf8 != sutf8) {
4784 /* Not modifying source SV, so taking a temporary copy. */
4785 SV* const csv = newSVpvn_flags(spv, slen, SVs_TEMP);
4787 sv_utf8_upgrade(csv);
4788 spv = SvPV_const(csv, slen);
4791 /* Leave enough space for the cat that's about to happen */
4792 sv_utf8_upgrade_flags_grow(dsv, 0, slen);
4794 sv_catpvn_nomg(dsv, spv, slen);
4797 if (flags & SV_SMAGIC)
4802 =for apidoc sv_catpv
4804 Concatenates the string onto the end of the string which is in the SV.
4805 If the SV has the UTF-8 status set, then the bytes appended should be
4806 valid UTF-8. Handles 'get' magic, but not 'set' magic. See C<sv_catpv_mg>.
4811 Perl_sv_catpv(pTHX_ register SV *const sv, register const char *ptr)
4814 register STRLEN len;
4818 PERL_ARGS_ASSERT_SV_CATPV;
4822 junk = SvPV_force(sv, tlen);
4824 SvGROW(sv, tlen + len + 1);
4826 ptr = SvPVX_const(sv);
4827 Move(ptr,SvPVX(sv)+tlen,len+1,char);
4828 SvCUR_set(sv, SvCUR(sv) + len);
4829 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4834 =for apidoc sv_catpv_mg
4836 Like C<sv_catpv>, but also handles 'set' magic.
4842 Perl_sv_catpv_mg(pTHX_ register SV *const sv, register const char *const ptr)
4844 PERL_ARGS_ASSERT_SV_CATPV_MG;
4853 Creates a new SV. A non-zero C<len> parameter indicates the number of
4854 bytes of preallocated string space the SV should have. An extra byte for a
4855 trailing NUL is also reserved. (SvPOK is not set for the SV even if string
4856 space is allocated.) The reference count for the new SV is set to 1.
4858 In 5.9.3, newSV() replaces the older NEWSV() API, and drops the first
4859 parameter, I<x>, a debug aid which allowed callers to identify themselves.
4860 This aid has been superseded by a new build option, PERL_MEM_LOG (see
4861 L<perlhack/PERL_MEM_LOG>). The older API is still there for use in XS
4862 modules supporting older perls.
4868 Perl_newSV(pTHX_ const STRLEN len)
4875 sv_upgrade(sv, SVt_PV);
4876 SvGROW(sv, len + 1);
4881 =for apidoc sv_magicext
4883 Adds magic to an SV, upgrading it if necessary. Applies the
4884 supplied vtable and returns a pointer to the magic added.
4886 Note that C<sv_magicext> will allow things that C<sv_magic> will not.
4887 In particular, you can add magic to SvREADONLY SVs, and add more than
4888 one instance of the same 'how'.
4890 If C<namlen> is greater than zero then a C<savepvn> I<copy> of C<name> is
4891 stored, if C<namlen> is zero then C<name> is stored as-is and - as another
4892 special case - if C<(name && namlen == HEf_SVKEY)> then C<name> is assumed
4893 to contain an C<SV*> and is stored as-is with its REFCNT incremented.
4895 (This is now used as a subroutine by C<sv_magic>.)
4900 Perl_sv_magicext(pTHX_ SV *const sv, SV *const obj, const int how,
4901 const MGVTBL *const vtable, const char *const name, const I32 namlen)
4906 PERL_ARGS_ASSERT_SV_MAGICEXT;
4908 SvUPGRADE(sv, SVt_PVMG);
4909 Newxz(mg, 1, MAGIC);
4910 mg->mg_moremagic = SvMAGIC(sv);
4911 SvMAGIC_set(sv, mg);
4913 /* Sometimes a magic contains a reference loop, where the sv and
4914 object refer to each other. To prevent a reference loop that
4915 would prevent such objects being freed, we look for such loops
4916 and if we find one we avoid incrementing the object refcount.
4918 Note we cannot do this to avoid self-tie loops as intervening RV must
4919 have its REFCNT incremented to keep it in existence.
4922 if (!obj || obj == sv ||
4923 how == PERL_MAGIC_arylen ||
4924 how == PERL_MAGIC_symtab ||
4925 (SvTYPE(obj) == SVt_PVGV &&
4926 (GvSV(obj) == sv || GvHV(obj) == (const HV *)sv
4927 || GvAV(obj) == (const AV *)sv || GvCV(obj) == (const CV *)sv
4928 || GvIOp(obj) == (const IO *)sv || GvFORM(obj) == (const CV *)sv)))
4933 mg->mg_obj = SvREFCNT_inc_simple(obj);
4934 mg->mg_flags |= MGf_REFCOUNTED;
4937 /* Normal self-ties simply pass a null object, and instead of
4938 using mg_obj directly, use the SvTIED_obj macro to produce a
4939 new RV as needed. For glob "self-ties", we are tieing the PVIO
4940 with an RV obj pointing to the glob containing the PVIO. In
4941 this case, to avoid a reference loop, we need to weaken the
4945 if (how == PERL_MAGIC_tiedscalar && SvTYPE(sv) == SVt_PVIO &&
4946 obj && SvROK(obj) && GvIO(SvRV(obj)) == (const IO *)sv)
4952 mg->mg_len = namlen;
4955 mg->mg_ptr = savepvn(name, namlen);
4956 else if (namlen == HEf_SVKEY) {
4957 /* Yes, this is casting away const. This is only for the case of
4958 HEf_SVKEY. I think we need to document this abberation of the
4959 constness of the API, rather than making name non-const, as
4960 that change propagating outwards a long way. */
4961 mg->mg_ptr = (char*)SvREFCNT_inc_simple_NN((SV *)name);
4963 mg->mg_ptr = (char *) name;
4965 mg->mg_virtual = (MGVTBL *) vtable;
4969 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
4974 =for apidoc sv_magic
4976 Adds magic to an SV. First upgrades C<sv> to type C<SVt_PVMG> if necessary,
4977 then adds a new magic item of type C<how> to the head of the magic list.
4979 See C<sv_magicext> (which C<sv_magic> now calls) for a description of the
4980 handling of the C<name> and C<namlen> arguments.
4982 You need to use C<sv_magicext> to add magic to SvREADONLY SVs and also
4983 to add more than one instance of the same 'how'.
4989 Perl_sv_magic(pTHX_ register SV *const sv, SV *const obj, const int how,
4990 const char *const name, const I32 namlen)
4993 const MGVTBL *vtable;
4996 PERL_ARGS_ASSERT_SV_MAGIC;
4998 #ifdef PERL_OLD_COPY_ON_WRITE
5000 sv_force_normal_flags(sv, 0);
5002 if (SvREADONLY(sv)) {
5004 /* its okay to attach magic to shared strings; the subsequent
5005 * upgrade to PVMG will unshare the string */
5006 !(SvFAKE(sv) && SvTYPE(sv) < SVt_PVMG)
5009 && how != PERL_MAGIC_regex_global
5010 && how != PERL_MAGIC_bm
5011 && how != PERL_MAGIC_fm
5012 && how != PERL_MAGIC_sv
5013 && how != PERL_MAGIC_backref
5016 Perl_croak(aTHX_ "%s", PL_no_modify);
5019 if (SvMAGICAL(sv) || (how == PERL_MAGIC_taint && SvTYPE(sv) >= SVt_PVMG)) {
5020 if (SvMAGIC(sv) && (mg = mg_find(sv, how))) {
5021 /* sv_magic() refuses to add a magic of the same 'how' as an
5024 if (how == PERL_MAGIC_taint) {
5026 /* Any scalar which already had taint magic on which someone
5027 (erroneously?) did SvIOK_on() or similar will now be
5028 incorrectly sporting public "OK" flags. */
5029 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
5037 vtable = &PL_vtbl_sv;
5039 case PERL_MAGIC_overload:
5040 vtable = &PL_vtbl_amagic;
5042 case PERL_MAGIC_overload_elem:
5043 vtable = &PL_vtbl_amagicelem;
5045 case PERL_MAGIC_overload_table:
5046 vtable = &PL_vtbl_ovrld;
5049 vtable = &PL_vtbl_bm;
5051 case PERL_MAGIC_regdata:
5052 vtable = &PL_vtbl_regdata;
5054 case PERL_MAGIC_regdatum:
5055 vtable = &PL_vtbl_regdatum;
5057 case PERL_MAGIC_env:
5058 vtable = &PL_vtbl_env;
5061 vtable = &PL_vtbl_fm;
5063 case PERL_MAGIC_envelem:
5064 vtable = &PL_vtbl_envelem;
5066 case PERL_MAGIC_regex_global:
5067 vtable = &PL_vtbl_mglob;
5069 case PERL_MAGIC_isa:
5070 vtable = &PL_vtbl_isa;
5072 case PERL_MAGIC_isaelem:
5073 vtable = &PL_vtbl_isaelem;
5075 case PERL_MAGIC_nkeys:
5076 vtable = &PL_vtbl_nkeys;
5078 case PERL_MAGIC_dbfile:
5081 case PERL_MAGIC_dbline:
5082 vtable = &PL_vtbl_dbline;
5084 #ifdef USE_LOCALE_COLLATE
5085 case PERL_MAGIC_collxfrm:
5086 vtable = &PL_vtbl_collxfrm;
5088 #endif /* USE_LOCALE_COLLATE */
5089 case PERL_MAGIC_tied:
5090 vtable = &PL_vtbl_pack;
5092 case PERL_MAGIC_tiedelem:
5093 case PERL_MAGIC_tiedscalar:
5094 vtable = &PL_vtbl_packelem;
5097 vtable = &PL_vtbl_regexp;
5099 case PERL_MAGIC_hints:
5100 /* As this vtable is all NULL, we can reuse it. */
5101 case PERL_MAGIC_sig:
5102 vtable = &PL_vtbl_sig;
5104 case PERL_MAGIC_sigelem:
5105 vtable = &PL_vtbl_sigelem;
5107 case PERL_MAGIC_taint:
5108 vtable = &PL_vtbl_taint;
5110 case PERL_MAGIC_uvar:
5111 vtable = &PL_vtbl_uvar;
5113 case PERL_MAGIC_vec:
5114 vtable = &PL_vtbl_vec;
5116 case PERL_MAGIC_arylen_p:
5117 case PERL_MAGIC_rhash:
5118 case PERL_MAGIC_symtab:
5119 case PERL_MAGIC_vstring:
5122 case PERL_MAGIC_utf8:
5123 vtable = &PL_vtbl_utf8;
5125 case PERL_MAGIC_substr:
5126 vtable = &PL_vtbl_substr;
5128 case PERL_MAGIC_defelem:
5129 vtable = &PL_vtbl_defelem;
5131 case PERL_MAGIC_arylen:
5132 vtable = &PL_vtbl_arylen;
5134 case PERL_MAGIC_pos:
5135 vtable = &PL_vtbl_pos;
5137 case PERL_MAGIC_backref:
5138 vtable = &PL_vtbl_backref;
5140 case PERL_MAGIC_hintselem:
5141 vtable = &PL_vtbl_hintselem;
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 if (ckWARN(WARN_MISC))
5242 Perl_warner(aTHX_ packWARN(WARN_MISC), "Reference is already weak");
5246 Perl_sv_add_backref(aTHX_ tsv, sv);
5252 /* Give tsv backref magic if it hasn't already got it, then push a
5253 * back-reference to sv onto the array associated with the backref magic.
5256 /* A discussion about the backreferences array and its refcount:
5258 * The AV holding the backreferences is pointed to either as the mg_obj of
5259 * PERL_MAGIC_backref, or in the specific case of a HV that has the hv_aux
5260 * structure, from the xhv_backreferences field. (A HV without hv_aux will
5261 * have the standard magic instead.) The array is created with a refcount
5262 * of 2. This means that if during global destruction the array gets
5263 * picked on first to have its refcount decremented by the random zapper,
5264 * it won't actually be freed, meaning it's still theere for when its
5265 * parent gets freed.
5266 * When the parent SV is freed, in the case of magic, the magic is freed,
5267 * Perl_magic_killbackrefs is called which decrements one refcount, then
5268 * mg_obj is freed which kills the second count.
5269 * In the vase of a HV being freed, one ref is removed by
5270 * Perl_hv_kill_backrefs, the other by Perl_sv_kill_backrefs, which it
5275 Perl_sv_add_backref(pTHX_ SV *const tsv, SV *const sv)
5280 PERL_ARGS_ASSERT_SV_ADD_BACKREF;
5282 if (SvTYPE(tsv) == SVt_PVHV) {
5283 AV **const avp = Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5287 /* There is no AV in the offical place - try a fixup. */
5288 MAGIC *const mg = mg_find(tsv, PERL_MAGIC_backref);
5291 /* Aha. They've got it stowed in magic. Bring it back. */
5292 av = MUTABLE_AV(mg->mg_obj);
5293 /* Stop mg_free decreasing the refernce count. */
5295 /* Stop mg_free even calling the destructor, given that
5296 there's no AV to free up. */
5298 sv_unmagic(tsv, PERL_MAGIC_backref);
5302 SvREFCNT_inc_simple_void(av); /* see discussion above */
5307 const MAGIC *const mg
5308 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
5310 av = MUTABLE_AV(mg->mg_obj);
5314 sv_magic(tsv, MUTABLE_SV(av), PERL_MAGIC_backref, NULL, 0);
5315 /* av now has a refcnt of 2; see discussion above */
5318 if (AvFILLp(av) >= AvMAX(av)) {
5319 av_extend(av, AvFILLp(av)+1);
5321 AvARRAY(av)[++AvFILLp(av)] = sv; /* av_push() */
5324 /* delete a back-reference to ourselves from the backref magic associated
5325 * with the SV we point to.
5329 S_sv_del_backref(pTHX_ SV *const tsv, SV *const sv)
5336 PERL_ARGS_ASSERT_SV_DEL_BACKREF;
5338 if (SvTYPE(tsv) == SVt_PVHV && SvOOK(tsv)) {
5339 av = *Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5340 /* We mustn't attempt to "fix up" the hash here by moving the
5341 backreference array back to the hv_aux structure, as that is stored
5342 in the main HvARRAY(), and hfreentries assumes that no-one
5343 reallocates HvARRAY() while it is running. */
5346 const MAGIC *const mg
5347 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
5349 av = MUTABLE_AV(mg->mg_obj);
5353 Perl_croak(aTHX_ "panic: del_backref");
5355 assert(!SvIS_FREED(av));
5358 /* We shouldn't be in here more than once, but for paranoia reasons lets
5360 for (i = AvFILLp(av); i >= 0; i--) {
5362 const SSize_t fill = AvFILLp(av);
5364 /* We weren't the last entry.
5365 An unordered list has this property that you can take the
5366 last element off the end to fill the hole, and it's still
5367 an unordered list :-)
5372 AvFILLp(av) = fill - 1;
5378 Perl_sv_kill_backrefs(pTHX_ SV *const sv, AV *const av)
5380 SV **svp = AvARRAY(av);
5382 PERL_ARGS_ASSERT_SV_KILL_BACKREFS;
5383 PERL_UNUSED_ARG(sv);
5385 assert(!svp || !SvIS_FREED(av));
5387 SV *const *const last = svp + AvFILLp(av);
5389 while (svp <= last) {
5391 SV *const referrer = *svp;
5392 if (SvWEAKREF(referrer)) {
5393 /* XXX Should we check that it hasn't changed? */
5394 SvRV_set(referrer, 0);
5396 SvWEAKREF_off(referrer);
5397 SvSETMAGIC(referrer);
5398 } else if (SvTYPE(referrer) == SVt_PVGV ||
5399 SvTYPE(referrer) == SVt_PVLV) {
5400 /* You lookin' at me? */
5401 assert(GvSTASH(referrer));
5402 assert(GvSTASH(referrer) == (const HV *)sv);
5403 GvSTASH(referrer) = 0;
5406 "panic: magic_killbackrefs (flags=%"UVxf")",
5407 (UV)SvFLAGS(referrer));
5415 SvREFCNT_dec(av); /* remove extra count added by sv_add_backref() */
5420 =for apidoc sv_insert
5422 Inserts a string at the specified offset/length within the SV. Similar to
5423 the Perl substr() function. Handles get magic.
5425 =for apidoc sv_insert_flags
5427 Same as C<sv_insert>, but the extra C<flags> are passed the C<SvPV_force_flags> that applies to C<bigstr>.
5433 Perl_sv_insert_flags(pTHX_ SV *const bigstr, const STRLEN offset, const STRLEN len, const char *const little, const STRLEN littlelen, const U32 flags)
5438 register char *midend;
5439 register char *bigend;
5443 PERL_ARGS_ASSERT_SV_INSERT_FLAGS;
5446 Perl_croak(aTHX_ "Can't modify non-existent substring");
5447 SvPV_force_flags(bigstr, curlen, flags);
5448 (void)SvPOK_only_UTF8(bigstr);
5449 if (offset + len > curlen) {
5450 SvGROW(bigstr, offset+len+1);
5451 Zero(SvPVX(bigstr)+curlen, offset+len-curlen, char);
5452 SvCUR_set(bigstr, offset+len);
5456 i = littlelen - len;
5457 if (i > 0) { /* string might grow */
5458 big = SvGROW(bigstr, SvCUR(bigstr) + i + 1);
5459 mid = big + offset + len;
5460 midend = bigend = big + SvCUR(bigstr);
5463 while (midend > mid) /* shove everything down */
5464 *--bigend = *--midend;
5465 Move(little,big+offset,littlelen,char);
5466 SvCUR_set(bigstr, SvCUR(bigstr) + i);
5471 Move(little,SvPVX(bigstr)+offset,len,char);
5476 big = SvPVX(bigstr);
5479 bigend = big + SvCUR(bigstr);
5481 if (midend > bigend)
5482 Perl_croak(aTHX_ "panic: sv_insert");
5484 if (mid - big > bigend - midend) { /* faster to shorten from end */
5486 Move(little, mid, littlelen,char);
5489 i = bigend - midend;
5491 Move(midend, mid, i,char);
5495 SvCUR_set(bigstr, mid - big);
5497 else if ((i = mid - big)) { /* faster from front */
5498 midend -= littlelen;
5500 Move(big, midend - i, i, char);
5501 sv_chop(bigstr,midend-i);
5503 Move(little, mid, littlelen,char);
5505 else if (littlelen) {
5506 midend -= littlelen;
5507 sv_chop(bigstr,midend);
5508 Move(little,midend,littlelen,char);
5511 sv_chop(bigstr,midend);
5517 =for apidoc sv_replace
5519 Make the first argument a copy of the second, then delete the original.
5520 The target SV physically takes over ownership of the body of the source SV
5521 and inherits its flags; however, the target keeps any magic it owns,
5522 and any magic in the source is discarded.
5523 Note that this is a rather specialist SV copying operation; most of the
5524 time you'll want to use C<sv_setsv> or one of its many macro front-ends.
5530 Perl_sv_replace(pTHX_ register SV *const sv, register SV *const nsv)
5533 const U32 refcnt = SvREFCNT(sv);
5535 PERL_ARGS_ASSERT_SV_REPLACE;
5537 SV_CHECK_THINKFIRST_COW_DROP(sv);
5538 if (SvREFCNT(nsv) != 1) {
5539 Perl_croak(aTHX_ "panic: reference miscount on nsv in sv_replace()"
5540 " (%" UVuf " != 1)", (UV) SvREFCNT(nsv));
5542 if (SvMAGICAL(sv)) {
5546 sv_upgrade(nsv, SVt_PVMG);
5547 SvMAGIC_set(nsv, SvMAGIC(sv));
5548 SvFLAGS(nsv) |= SvMAGICAL(sv);
5550 SvMAGIC_set(sv, NULL);
5554 assert(!SvREFCNT(sv));
5555 #ifdef DEBUG_LEAKING_SCALARS
5556 sv->sv_flags = nsv->sv_flags;
5557 sv->sv_any = nsv->sv_any;
5558 sv->sv_refcnt = nsv->sv_refcnt;
5559 sv->sv_u = nsv->sv_u;
5561 StructCopy(nsv,sv,SV);
5563 if(SvTYPE(sv) == SVt_IV) {
5565 = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
5569 #ifdef PERL_OLD_COPY_ON_WRITE
5570 if (SvIsCOW_normal(nsv)) {
5571 /* We need to follow the pointers around the loop to make the
5572 previous SV point to sv, rather than nsv. */
5575 while ((next = SV_COW_NEXT_SV(current)) != nsv) {
5578 assert(SvPVX_const(current) == SvPVX_const(nsv));
5580 /* Make the SV before us point to the SV after us. */
5582 PerlIO_printf(Perl_debug_log, "previous is\n");
5584 PerlIO_printf(Perl_debug_log,
5585 "move it from 0x%"UVxf" to 0x%"UVxf"\n",
5586 (UV) SV_COW_NEXT_SV(current), (UV) sv);
5588 SV_COW_NEXT_SV_SET(current, sv);
5591 SvREFCNT(sv) = refcnt;
5592 SvFLAGS(nsv) |= SVTYPEMASK; /* Mark as freed */
5598 =for apidoc sv_clear
5600 Clear an SV: call any destructors, free up any memory used by the body,
5601 and free the body itself. The SV's head is I<not> freed, although
5602 its type is set to all 1's so that it won't inadvertently be assumed
5603 to be live during global destruction etc.
5604 This function should only be called when REFCNT is zero. Most of the time
5605 you'll want to call C<sv_free()> (or its macro wrapper C<SvREFCNT_dec>)
5612 Perl_sv_clear(pTHX_ register SV *const sv)
5615 const U32 type = SvTYPE(sv);
5616 const struct body_details *const sv_type_details
5617 = bodies_by_type + type;
5620 PERL_ARGS_ASSERT_SV_CLEAR;
5621 assert(SvREFCNT(sv) == 0);
5622 assert(SvTYPE(sv) != SVTYPEMASK);
5624 if (type <= SVt_IV) {
5625 /* See the comment in sv.h about the collusion between this early
5626 return and the overloading of the NULL and IV slots in the size
5629 SV * const target = SvRV(sv);
5631 sv_del_backref(target, sv);
5633 SvREFCNT_dec(target);
5635 SvFLAGS(sv) &= SVf_BREAK;
5636 SvFLAGS(sv) |= SVTYPEMASK;
5641 if (PL_defstash && /* Still have a symbol table? */
5648 stash = SvSTASH(sv);
5649 destructor = StashHANDLER(stash,DESTROY);
5651 /* A constant subroutine can have no side effects, so
5652 don't bother calling it. */
5653 && !CvCONST(destructor)
5654 /* Don't bother calling an empty destructor */
5655 && (CvISXSUB(destructor)
5656 || CvSTART(destructor)->op_next->op_type != OP_LEAVESUB))
5658 SV* const tmpref = newRV(sv);
5659 SvREADONLY_on(tmpref); /* DESTROY() could be naughty */
5661 PUSHSTACKi(PERLSI_DESTROY);
5666 call_sv(MUTABLE_SV(destructor), G_DISCARD|G_EVAL|G_KEEPERR|G_VOID);
5672 if(SvREFCNT(tmpref) < 2) {
5673 /* tmpref is not kept alive! */
5675 SvRV_set(tmpref, NULL);
5678 SvREFCNT_dec(tmpref);
5680 } while (SvOBJECT(sv) && SvSTASH(sv) != stash);
5684 if (PL_in_clean_objs)
5685 Perl_croak(aTHX_ "DESTROY created new reference to dead object '%s'",
5687 /* DESTROY gave object new lease on life */
5693 SvREFCNT_dec(SvSTASH(sv)); /* possibly of changed persuasion */
5694 SvOBJECT_off(sv); /* Curse the object. */
5695 if (type != SVt_PVIO)
5696 --PL_sv_objcount; /* XXX Might want something more general */
5699 if (type >= SVt_PVMG) {
5700 if (type == SVt_PVMG && SvPAD_OUR(sv)) {
5701 SvREFCNT_dec(SvOURSTASH(sv));
5702 } else if (SvMAGIC(sv))
5704 if (type == SVt_PVMG && SvPAD_TYPED(sv))
5705 SvREFCNT_dec(SvSTASH(sv));
5708 /* case SVt_BIND: */
5711 IoIFP(sv) != PerlIO_stdin() &&
5712 IoIFP(sv) != PerlIO_stdout() &&
5713 IoIFP(sv) != PerlIO_stderr())
5715 io_close(MUTABLE_IO(sv), FALSE);
5717 if (IoDIRP(sv) && !(IoFLAGS(sv) & IOf_FAKE_DIRP))
5718 PerlDir_close(IoDIRP(sv));
5719 IoDIRP(sv) = (DIR*)NULL;
5720 Safefree(IoTOP_NAME(sv));
5721 Safefree(IoFMT_NAME(sv));
5722 Safefree(IoBOTTOM_NAME(sv));
5725 /* FIXME for plugins */
5726 pregfree2((REGEXP*) sv);
5730 cv_undef(MUTABLE_CV(sv));
5733 if (PL_last_swash_hv == (const HV *)sv) {
5734 PL_last_swash_hv = NULL;
5736 Perl_hv_kill_backrefs(aTHX_ MUTABLE_HV(sv));
5737 hv_undef(MUTABLE_HV(sv));
5740 if (PL_comppad == MUTABLE_AV(sv)) {
5744 av_undef(MUTABLE_AV(sv));
5747 if (LvTYPE(sv) == 'T') { /* for tie: return HE to pool */
5748 SvREFCNT_dec(HeKEY_sv((HE*)LvTARG(sv)));
5749 HeNEXT((HE*)LvTARG(sv)) = PL_hv_fetch_ent_mh;
5750 PL_hv_fetch_ent_mh = (HE*)LvTARG(sv);
5752 else if (LvTYPE(sv) != 't') /* unless tie: unrefcnted fake SV** */
5753 SvREFCNT_dec(LvTARG(sv));
5755 if (isGV_with_GP(sv)) {
5756 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
5757 && HvNAME_get(stash))
5758 mro_method_changed_in(stash);
5759 gp_free(MUTABLE_GV(sv));
5761 unshare_hek(GvNAME_HEK(sv));
5762 /* If we're in a stash, we don't own a reference to it. However it does
5763 have a back reference to us, which needs to be cleared. */
5764 if (!SvVALID(sv) && (stash = GvSTASH(sv)))
5765 sv_del_backref(MUTABLE_SV(stash), sv);
5767 /* FIXME. There are probably more unreferenced pointers to SVs in the
5768 interpreter struct that we should check and tidy in a similar
5770 if ((const GV *)sv == PL_last_in_gv)
5771 PL_last_in_gv = NULL;
5777 /* Don't bother with SvOOK_off(sv); as we're only going to free it. */
5780 SvOOK_offset(sv, offset);
5781 SvPV_set(sv, SvPVX_mutable(sv) - offset);
5782 /* Don't even bother with turning off the OOK flag. */
5785 SV * const target = SvRV(sv);
5787 sv_del_backref(target, sv);
5789 SvREFCNT_dec(target);
5791 #ifdef PERL_OLD_COPY_ON_WRITE
5792 else if (SvPVX_const(sv)) {
5795 PerlIO_printf(Perl_debug_log, "Copy on write: clear\n");
5799 sv_release_COW(sv, SvPVX_const(sv), SV_COW_NEXT_SV(sv));
5801 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
5805 } else if (SvLEN(sv)) {
5806 Safefree(SvPVX_const(sv));
5810 else if (SvPVX_const(sv) && SvLEN(sv))
5811 Safefree(SvPVX_mutable(sv));
5812 else if (SvPVX_const(sv) && SvREADONLY(sv) && SvFAKE(sv)) {
5813 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
5822 SvFLAGS(sv) &= SVf_BREAK;
5823 SvFLAGS(sv) |= SVTYPEMASK;
5825 if (sv_type_details->arena) {
5826 del_body(((char *)SvANY(sv) + sv_type_details->offset),
5827 &PL_body_roots[type]);
5829 else if (sv_type_details->body_size) {
5830 my_safefree(SvANY(sv));
5835 =for apidoc sv_newref
5837 Increment an SV's reference count. Use the C<SvREFCNT_inc()> wrapper
5844 Perl_sv_newref(pTHX_ SV *const sv)
5846 PERL_UNUSED_CONTEXT;
5855 Decrement an SV's reference count, and if it drops to zero, call
5856 C<sv_clear> to invoke destructors and free up any memory used by
5857 the body; finally, deallocate the SV's head itself.
5858 Normally called via a wrapper macro C<SvREFCNT_dec>.
5864 Perl_sv_free(pTHX_ SV *const sv)
5869 if (SvREFCNT(sv) == 0) {
5870 if (SvFLAGS(sv) & SVf_BREAK)
5871 /* this SV's refcnt has been artificially decremented to
5872 * trigger cleanup */
5874 if (PL_in_clean_all) /* All is fair */
5876 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
5877 /* make sure SvREFCNT(sv)==0 happens very seldom */
5878 SvREFCNT(sv) = (~(U32)0)/2;
5881 if (ckWARN_d(WARN_INTERNAL)) {
5882 #ifdef DEBUG_LEAKING_SCALARS_FORK_DUMP
5883 Perl_dump_sv_child(aTHX_ sv);
5885 #ifdef DEBUG_LEAKING_SCALARS
5888 #ifdef DEBUG_LEAKING_SCALARS_ABORT
5889 if (PL_warnhook == PERL_WARNHOOK_FATAL
5890 || ckDEAD(packWARN(WARN_INTERNAL))) {
5891 /* Don't let Perl_warner cause us to escape our fate: */
5895 /* This may not return: */
5896 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
5897 "Attempt to free unreferenced scalar: SV 0x%"UVxf
5898 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
5901 #ifdef DEBUG_LEAKING_SCALARS_ABORT
5906 if (--(SvREFCNT(sv)) > 0)
5908 Perl_sv_free2(aTHX_ sv);
5912 Perl_sv_free2(pTHX_ SV *const sv)
5916 PERL_ARGS_ASSERT_SV_FREE2;
5920 if (ckWARN_d(WARN_DEBUGGING))
5921 Perl_warner(aTHX_ packWARN(WARN_DEBUGGING),
5922 "Attempt to free temp prematurely: SV 0x%"UVxf
5923 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
5927 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
5928 /* make sure SvREFCNT(sv)==0 happens very seldom */
5929 SvREFCNT(sv) = (~(U32)0)/2;
5940 Returns the length of the string in the SV. Handles magic and type
5941 coercion. See also C<SvCUR>, which gives raw access to the xpv_cur slot.
5947 Perl_sv_len(pTHX_ register SV *const sv)
5955 len = mg_length(sv);
5957 (void)SvPV_const(sv, len);
5962 =for apidoc sv_len_utf8
5964 Returns the number of characters in the string in an SV, counting wide
5965 UTF-8 bytes as a single character. Handles magic and type coercion.
5971 * The length is cached in PERL_MAGIC_utf8, in the mg_len field. Also the
5972 * mg_ptr is used, by sv_pos_u2b() and sv_pos_b2u() - see the comments below.
5973 * (Note that the mg_len is not the length of the mg_ptr field.
5974 * This allows the cache to store the character length of the string without
5975 * needing to malloc() extra storage to attach to the mg_ptr.)
5980 Perl_sv_len_utf8(pTHX_ register SV *const sv)
5986 return mg_length(sv);
5990 const U8 *s = (U8*)SvPV_const(sv, len);
5994 MAGIC *mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_utf8) : NULL;
5996 if (mg && mg->mg_len != -1) {
5998 if (PL_utf8cache < 0) {
5999 const STRLEN real = Perl_utf8_length(aTHX_ s, s + len);
6001 /* Need to turn the assertions off otherwise we may
6002 recurse infinitely while printing error messages.
6004 SAVEI8(PL_utf8cache);
6006 Perl_croak(aTHX_ "panic: sv_len_utf8 cache %"UVuf
6007 " real %"UVuf" for %"SVf,
6008 (UV) ulen, (UV) real, SVfARG(sv));
6013 ulen = Perl_utf8_length(aTHX_ s, s + len);
6014 if (!SvREADONLY(sv)) {
6016 mg = sv_magicext(sv, 0, PERL_MAGIC_utf8,
6017 &PL_vtbl_utf8, 0, 0);
6025 return Perl_utf8_length(aTHX_ s, s + len);
6029 /* Walk forwards to find the byte corresponding to the passed in UTF-8
6032 S_sv_pos_u2b_forwards(const U8 *const start, const U8 *const send,
6035 const U8 *s = start;
6037 PERL_ARGS_ASSERT_SV_POS_U2B_FORWARDS;
6039 while (s < send && uoffset--)
6042 /* This is the existing behaviour. Possibly it should be a croak, as
6043 it's actually a bounds error */
6049 /* Given the length of the string in both bytes and UTF-8 characters, decide
6050 whether to walk forwards or backwards to find the byte corresponding to
6051 the passed in UTF-8 offset. */
6053 S_sv_pos_u2b_midway(const U8 *const start, const U8 *send,
6054 const STRLEN uoffset, const STRLEN uend)
6056 STRLEN backw = uend - uoffset;
6058 PERL_ARGS_ASSERT_SV_POS_U2B_MIDWAY;
6060 if (uoffset < 2 * backw) {
6061 /* The assumption is that going forwards is twice the speed of going
6062 forward (that's where the 2 * backw comes from).
6063 (The real figure of course depends on the UTF-8 data.) */
6064 return sv_pos_u2b_forwards(start, send, uoffset);
6069 while (UTF8_IS_CONTINUATION(*send))
6072 return send - start;
6075 /* For the string representation of the given scalar, find the byte
6076 corresponding to the passed in UTF-8 offset. uoffset0 and boffset0
6077 give another position in the string, *before* the sought offset, which
6078 (which is always true, as 0, 0 is a valid pair of positions), which should
6079 help reduce the amount of linear searching.
6080 If *mgp is non-NULL, it should point to the UTF-8 cache magic, which
6081 will be used to reduce the amount of linear searching. The cache will be
6082 created if necessary, and the found value offered to it for update. */
6084 S_sv_pos_u2b_cached(pTHX_ SV *const sv, MAGIC **const mgp, const U8 *const start,
6085 const U8 *const send, const STRLEN uoffset,
6086 STRLEN uoffset0, STRLEN boffset0)
6088 STRLEN boffset = 0; /* Actually always set, but let's keep gcc happy. */
6091 PERL_ARGS_ASSERT_SV_POS_U2B_CACHED;
6093 assert (uoffset >= uoffset0);
6095 if (SvMAGICAL(sv) && !SvREADONLY(sv) && PL_utf8cache
6096 && (*mgp || (*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
6097 if ((*mgp)->mg_ptr) {
6098 STRLEN *cache = (STRLEN *) (*mgp)->mg_ptr;
6099 if (cache[0] == uoffset) {
6100 /* An exact match. */
6103 if (cache[2] == uoffset) {
6104 /* An exact match. */
6108 if (cache[0] < uoffset) {
6109 /* The cache already knows part of the way. */
6110 if (cache[0] > uoffset0) {
6111 /* The cache knows more than the passed in pair */
6112 uoffset0 = cache[0];
6113 boffset0 = cache[1];
6115 if ((*mgp)->mg_len != -1) {
6116 /* And we know the end too. */
6118 + sv_pos_u2b_midway(start + boffset0, send,
6120 (*mgp)->mg_len - uoffset0);
6123 + sv_pos_u2b_forwards(start + boffset0,
6124 send, uoffset - uoffset0);
6127 else if (cache[2] < uoffset) {
6128 /* We're between the two cache entries. */
6129 if (cache[2] > uoffset0) {
6130 /* and the cache knows more than the passed in pair */
6131 uoffset0 = cache[2];
6132 boffset0 = cache[3];
6136 + sv_pos_u2b_midway(start + boffset0,
6139 cache[0] - uoffset0);
6142 + sv_pos_u2b_midway(start + boffset0,
6145 cache[2] - uoffset0);
6149 else if ((*mgp)->mg_len != -1) {
6150 /* If we can take advantage of a passed in offset, do so. */
6151 /* In fact, offset0 is either 0, or less than offset, so don't
6152 need to worry about the other possibility. */
6154 + sv_pos_u2b_midway(start + boffset0, send,
6156 (*mgp)->mg_len - uoffset0);
6161 if (!found || PL_utf8cache < 0) {
6162 const STRLEN real_boffset
6163 = boffset0 + sv_pos_u2b_forwards(start + boffset0,
6164 send, uoffset - uoffset0);
6166 if (found && PL_utf8cache < 0) {
6167 if (real_boffset != boffset) {
6168 /* Need to turn the assertions off otherwise we may recurse
6169 infinitely while printing error messages. */
6170 SAVEI8(PL_utf8cache);
6172 Perl_croak(aTHX_ "panic: sv_pos_u2b_cache cache %"UVuf
6173 " real %"UVuf" for %"SVf,
6174 (UV) boffset, (UV) real_boffset, SVfARG(sv));
6177 boffset = real_boffset;
6181 utf8_mg_pos_cache_update(sv, mgp, boffset, uoffset, send - start);
6187 =for apidoc sv_pos_u2b
6189 Converts the value pointed to by offsetp from a count of UTF-8 chars from
6190 the start of the string, to a count of the equivalent number of bytes; if
6191 lenp is non-zero, it does the same to lenp, but this time starting from
6192 the offset, rather than from the start of the string. Handles magic and
6199 * sv_pos_u2b() uses, like sv_pos_b2u(), the mg_ptr of the potential
6200 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6201 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
6206 Perl_sv_pos_u2b(pTHX_ register SV *const sv, I32 *const offsetp, I32 *const lenp)
6211 PERL_ARGS_ASSERT_SV_POS_U2B;
6216 start = (U8*)SvPV_const(sv, len);
6218 STRLEN uoffset = (STRLEN) *offsetp;
6219 const U8 * const send = start + len;
6221 const STRLEN boffset = sv_pos_u2b_cached(sv, &mg, start, send,
6224 *offsetp = (I32) boffset;
6227 /* Convert the relative offset to absolute. */
6228 const STRLEN uoffset2 = uoffset + (STRLEN) *lenp;
6229 const STRLEN boffset2
6230 = sv_pos_u2b_cached(sv, &mg, start, send, uoffset2,
6231 uoffset, boffset) - boffset;
6245 /* Create and update the UTF8 magic offset cache, with the proffered utf8/
6246 byte length pairing. The (byte) length of the total SV is passed in too,
6247 as blen, because for some (more esoteric) SVs, the call to SvPV_const()
6248 may not have updated SvCUR, so we can't rely on reading it directly.
6250 The proffered utf8/byte length pairing isn't used if the cache already has
6251 two pairs, and swapping either for the proffered pair would increase the
6252 RMS of the intervals between known byte offsets.
6254 The cache itself consists of 4 STRLEN values
6255 0: larger UTF-8 offset
6256 1: corresponding byte offset
6257 2: smaller UTF-8 offset
6258 3: corresponding byte offset
6260 Unused cache pairs have the value 0, 0.
6261 Keeping the cache "backwards" means that the invariant of
6262 cache[0] >= cache[2] is maintained even with empty slots, which means that
6263 the code that uses it doesn't need to worry if only 1 entry has actually
6264 been set to non-zero. It also makes the "position beyond the end of the
6265 cache" logic much simpler, as the first slot is always the one to start
6269 S_utf8_mg_pos_cache_update(pTHX_ SV *const sv, MAGIC **const mgp, const STRLEN byte,
6270 const STRLEN utf8, const STRLEN blen)
6274 PERL_ARGS_ASSERT_UTF8_MG_POS_CACHE_UPDATE;
6280 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, (MGVTBL*)&PL_vtbl_utf8, 0,
6282 (*mgp)->mg_len = -1;
6286 if (!(cache = (STRLEN *)(*mgp)->mg_ptr)) {
6287 Newxz(cache, PERL_MAGIC_UTF8_CACHESIZE * 2, STRLEN);
6288 (*mgp)->mg_ptr = (char *) cache;
6292 if (PL_utf8cache < 0) {
6293 const U8 *start = (const U8 *) SvPVX_const(sv);
6294 const STRLEN realutf8 = utf8_length(start, start + byte);
6296 if (realutf8 != utf8) {
6297 /* Need to turn the assertions off otherwise we may recurse
6298 infinitely while printing error messages. */
6299 SAVEI8(PL_utf8cache);
6301 Perl_croak(aTHX_ "panic: utf8_mg_pos_cache_update cache %"UVuf
6302 " real %"UVuf" for %"SVf, (UV) utf8, (UV) realutf8, SVfARG(sv));
6306 /* Cache is held with the later position first, to simplify the code
6307 that deals with unbounded ends. */
6309 ASSERT_UTF8_CACHE(cache);
6310 if (cache[1] == 0) {
6311 /* Cache is totally empty */
6314 } else if (cache[3] == 0) {
6315 if (byte > cache[1]) {
6316 /* New one is larger, so goes first. */
6317 cache[2] = cache[0];
6318 cache[3] = cache[1];
6326 #define THREEWAY_SQUARE(a,b,c,d) \
6327 ((float)((d) - (c))) * ((float)((d) - (c))) \
6328 + ((float)((c) - (b))) * ((float)((c) - (b))) \
6329 + ((float)((b) - (a))) * ((float)((b) - (a)))
6331 /* Cache has 2 slots in use, and we know three potential pairs.
6332 Keep the two that give the lowest RMS distance. Do the
6333 calcualation in bytes simply because we always know the byte
6334 length. squareroot has the same ordering as the positive value,
6335 so don't bother with the actual square root. */
6336 const float existing = THREEWAY_SQUARE(0, cache[3], cache[1], blen);
6337 if (byte > cache[1]) {
6338 /* New position is after the existing pair of pairs. */
6339 const float keep_earlier
6340 = THREEWAY_SQUARE(0, cache[3], byte, blen);
6341 const float keep_later
6342 = THREEWAY_SQUARE(0, cache[1], byte, blen);
6344 if (keep_later < keep_earlier) {
6345 if (keep_later < existing) {
6346 cache[2] = cache[0];
6347 cache[3] = cache[1];
6353 if (keep_earlier < existing) {
6359 else if (byte > cache[3]) {
6360 /* New position is between the existing pair of pairs. */
6361 const float keep_earlier
6362 = THREEWAY_SQUARE(0, cache[3], byte, blen);
6363 const float keep_later
6364 = THREEWAY_SQUARE(0, byte, cache[1], blen);
6366 if (keep_later < keep_earlier) {
6367 if (keep_later < existing) {
6373 if (keep_earlier < existing) {
6380 /* New position is before the existing pair of pairs. */
6381 const float keep_earlier
6382 = THREEWAY_SQUARE(0, byte, cache[3], blen);
6383 const float keep_later
6384 = THREEWAY_SQUARE(0, byte, cache[1], blen);
6386 if (keep_later < keep_earlier) {
6387 if (keep_later < existing) {
6393 if (keep_earlier < existing) {
6394 cache[0] = cache[2];
6395 cache[1] = cache[3];
6402 ASSERT_UTF8_CACHE(cache);
6405 /* We already know all of the way, now we may be able to walk back. The same
6406 assumption is made as in S_sv_pos_u2b_midway(), namely that walking
6407 backward is half the speed of walking forward. */
6409 S_sv_pos_b2u_midway(pTHX_ const U8 *const s, const U8 *const target,
6410 const U8 *end, STRLEN endu)
6412 const STRLEN forw = target - s;
6413 STRLEN backw = end - target;
6415 PERL_ARGS_ASSERT_SV_POS_B2U_MIDWAY;
6417 if (forw < 2 * backw) {
6418 return utf8_length(s, target);
6421 while (end > target) {
6423 while (UTF8_IS_CONTINUATION(*end)) {
6432 =for apidoc sv_pos_b2u
6434 Converts the value pointed to by offsetp from a count of bytes from the
6435 start of the string, to a count of the equivalent number of UTF-8 chars.
6436 Handles magic and type coercion.
6442 * sv_pos_b2u() uses, like sv_pos_u2b(), the mg_ptr of the potential
6443 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6448 Perl_sv_pos_b2u(pTHX_ register SV *const sv, I32 *const offsetp)
6451 const STRLEN byte = *offsetp;
6452 STRLEN len = 0; /* Actually always set, but let's keep gcc happy. */
6458 PERL_ARGS_ASSERT_SV_POS_B2U;
6463 s = (const U8*)SvPV_const(sv, blen);
6466 Perl_croak(aTHX_ "panic: sv_pos_b2u: bad byte offset");
6470 if (SvMAGICAL(sv) && !SvREADONLY(sv) && PL_utf8cache
6471 && (mg = mg_find(sv, PERL_MAGIC_utf8))) {
6473 STRLEN * const cache = (STRLEN *) mg->mg_ptr;
6474 if (cache[1] == byte) {
6475 /* An exact match. */
6476 *offsetp = cache[0];
6479 if (cache[3] == byte) {
6480 /* An exact match. */
6481 *offsetp = cache[2];
6485 if (cache[1] < byte) {
6486 /* We already know part of the way. */
6487 if (mg->mg_len != -1) {
6488 /* Actually, we know the end too. */
6490 + S_sv_pos_b2u_midway(aTHX_ s + cache[1], send,
6491 s + blen, mg->mg_len - cache[0]);
6493 len = cache[0] + utf8_length(s + cache[1], send);
6496 else if (cache[3] < byte) {
6497 /* We're between the two cached pairs, so we do the calculation
6498 offset by the byte/utf-8 positions for the earlier pair,
6499 then add the utf-8 characters from the string start to
6501 len = S_sv_pos_b2u_midway(aTHX_ s + cache[3], send,
6502 s + cache[1], cache[0] - cache[2])
6506 else { /* cache[3] > byte */
6507 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + cache[3],
6511 ASSERT_UTF8_CACHE(cache);
6513 } else if (mg->mg_len != -1) {
6514 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + blen, mg->mg_len);
6518 if (!found || PL_utf8cache < 0) {
6519 const STRLEN real_len = utf8_length(s, send);
6521 if (found && PL_utf8cache < 0) {
6522 if (len != real_len) {
6523 /* Need to turn the assertions off otherwise we may recurse
6524 infinitely while printing error messages. */
6525 SAVEI8(PL_utf8cache);
6527 Perl_croak(aTHX_ "panic: sv_pos_b2u cache %"UVuf
6528 " real %"UVuf" for %"SVf,
6529 (UV) len, (UV) real_len, SVfARG(sv));
6537 utf8_mg_pos_cache_update(sv, &mg, byte, len, blen);
6543 Returns a boolean indicating whether the strings in the two SVs are
6544 identical. Is UTF-8 and 'use bytes' aware, handles get magic, and will
6545 coerce its args to strings if necessary.
6551 Perl_sv_eq(pTHX_ register SV *sv1, register SV *sv2)
6560 SV* svrecode = NULL;
6567 /* if pv1 and pv2 are the same, second SvPV_const call may
6568 * invalidate pv1, so we may need to make a copy */
6569 if (sv1 == sv2 && (SvTHINKFIRST(sv1) || SvGMAGICAL(sv1))) {
6570 pv1 = SvPV_const(sv1, cur1);
6571 sv1 = newSVpvn_flags(pv1, cur1, SVs_TEMP | SvUTF8(sv2));
6573 pv1 = SvPV_const(sv1, cur1);
6581 pv2 = SvPV_const(sv2, cur2);
6583 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
6584 /* Differing utf8ness.
6585 * Do not UTF8size the comparands as a side-effect. */
6588 svrecode = newSVpvn(pv2, cur2);
6589 sv_recode_to_utf8(svrecode, PL_encoding);
6590 pv2 = SvPV_const(svrecode, cur2);
6593 svrecode = newSVpvn(pv1, cur1);
6594 sv_recode_to_utf8(svrecode, PL_encoding);
6595 pv1 = SvPV_const(svrecode, cur1);
6597 /* Now both are in UTF-8. */
6599 SvREFCNT_dec(svrecode);
6604 bool is_utf8 = TRUE;
6607 /* sv1 is the UTF-8 one,
6608 * if is equal it must be downgrade-able */
6609 char * const pv = (char*)bytes_from_utf8((const U8*)pv1,
6615 /* sv2 is the UTF-8 one,
6616 * if is equal it must be downgrade-able */
6617 char * const pv = (char *)bytes_from_utf8((const U8*)pv2,
6623 /* Downgrade not possible - cannot be eq */
6631 eq = (pv1 == pv2) || memEQ(pv1, pv2, cur1);
6633 SvREFCNT_dec(svrecode);
6643 Compares the strings in two SVs. Returns -1, 0, or 1 indicating whether the
6644 string in C<sv1> is less than, equal to, or greater than the string in
6645 C<sv2>. Is UTF-8 and 'use bytes' aware, handles get magic, and will
6646 coerce its args to strings if necessary. See also C<sv_cmp_locale>.
6652 Perl_sv_cmp(pTHX_ register SV *const sv1, register SV *const sv2)
6656 const char *pv1, *pv2;
6659 SV *svrecode = NULL;
6666 pv1 = SvPV_const(sv1, cur1);
6673 pv2 = SvPV_const(sv2, cur2);
6675 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
6676 /* Differing utf8ness.
6677 * Do not UTF8size the comparands as a side-effect. */
6680 svrecode = newSVpvn(pv2, cur2);
6681 sv_recode_to_utf8(svrecode, PL_encoding);
6682 pv2 = SvPV_const(svrecode, cur2);
6685 pv2 = tpv = (char*)bytes_to_utf8((const U8*)pv2, &cur2);
6690 svrecode = newSVpvn(pv1, cur1);
6691 sv_recode_to_utf8(svrecode, PL_encoding);
6692 pv1 = SvPV_const(svrecode, cur1);
6695 pv1 = tpv = (char*)bytes_to_utf8((const U8*)pv1, &cur1);
6701 cmp = cur2 ? -1 : 0;
6705 const I32 retval = memcmp((const void*)pv1, (const void*)pv2, cur1 < cur2 ? cur1 : cur2);
6708 cmp = retval < 0 ? -1 : 1;
6709 } else if (cur1 == cur2) {
6712 cmp = cur1 < cur2 ? -1 : 1;
6716 SvREFCNT_dec(svrecode);
6724 =for apidoc sv_cmp_locale
6726 Compares the strings in two SVs in a locale-aware manner. Is UTF-8 and
6727 'use bytes' aware, handles get magic, and will coerce its args to strings
6728 if necessary. See also C<sv_cmp>.
6734 Perl_sv_cmp_locale(pTHX_ register SV *const sv1, register SV *const sv2)
6737 #ifdef USE_LOCALE_COLLATE
6743 if (PL_collation_standard)
6747 pv1 = sv1 ? sv_collxfrm(sv1, &len1) : (char *) NULL;
6749 pv2 = sv2 ? sv_collxfrm(sv2, &len2) : (char *) NULL;
6751 if (!pv1 || !len1) {
6762 retval = memcmp((void*)pv1, (void*)pv2, len1 < len2 ? len1 : len2);
6765 return retval < 0 ? -1 : 1;
6768 * When the result of collation is equality, that doesn't mean
6769 * that there are no differences -- some locales exclude some
6770 * characters from consideration. So to avoid false equalities,
6771 * we use the raw string as a tiebreaker.
6777 #endif /* USE_LOCALE_COLLATE */
6779 return sv_cmp(sv1, sv2);
6783 #ifdef USE_LOCALE_COLLATE
6786 =for apidoc sv_collxfrm
6788 Add Collate Transform magic to an SV if it doesn't already have it.
6790 Any scalar variable may carry PERL_MAGIC_collxfrm magic that contains the
6791 scalar data of the variable, but transformed to such a format that a normal
6792 memory comparison can be used to compare the data according to the locale
6799 Perl_sv_collxfrm(pTHX_ SV *const sv, STRLEN *const nxp)
6804 PERL_ARGS_ASSERT_SV_COLLXFRM;
6806 mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_collxfrm) : (MAGIC *) NULL;
6807 if (!mg || !mg->mg_ptr || *(U32*)mg->mg_ptr != PL_collation_ix) {
6813 Safefree(mg->mg_ptr);
6814 s = SvPV_const(sv, len);
6815 if ((xf = mem_collxfrm(s, len, &xlen))) {
6817 #ifdef PERL_OLD_COPY_ON_WRITE
6819 sv_force_normal_flags(sv, 0);
6821 mg = sv_magicext(sv, 0, PERL_MAGIC_collxfrm, &PL_vtbl_collxfrm,
6835 if (mg && mg->mg_ptr) {
6837 return mg->mg_ptr + sizeof(PL_collation_ix);
6845 #endif /* USE_LOCALE_COLLATE */
6850 Get a line from the filehandle and store it into the SV, optionally
6851 appending to the currently-stored string.
6857 Perl_sv_gets(pTHX_ register SV *const sv, register PerlIO *const fp, I32 append)
6862 register STDCHAR rslast;
6863 register STDCHAR *bp;
6868 PERL_ARGS_ASSERT_SV_GETS;
6870 if (SvTHINKFIRST(sv))
6871 sv_force_normal_flags(sv, append ? 0 : SV_COW_DROP_PV);
6872 /* XXX. If you make this PVIV, then copy on write can copy scalars read
6874 However, perlbench says it's slower, because the existing swipe code
6875 is faster than copy on write.
6876 Swings and roundabouts. */
6877 SvUPGRADE(sv, SVt_PV);
6882 if (PerlIO_isutf8(fp)) {
6884 sv_utf8_upgrade_nomg(sv);
6885 sv_pos_u2b(sv,&append,0);
6887 } else if (SvUTF8(sv)) {
6888 SV * const tsv = newSV(0);
6889 sv_gets(tsv, fp, 0);
6890 sv_utf8_upgrade_nomg(tsv);
6891 SvCUR_set(sv,append);
6894 goto return_string_or_null;
6899 if (PerlIO_isutf8(fp))
6902 if (IN_PERL_COMPILETIME) {
6903 /* we always read code in line mode */
6907 else if (RsSNARF(PL_rs)) {
6908 /* If it is a regular disk file use size from stat() as estimate
6909 of amount we are going to read -- may result in mallocing
6910 more memory than we really need if the layers below reduce
6911 the size we read (e.g. CRLF or a gzip layer).
6914 if (!PerlLIO_fstat(PerlIO_fileno(fp), &st) && S_ISREG(st.st_mode)) {
6915 const Off_t offset = PerlIO_tell(fp);
6916 if (offset != (Off_t) -1 && st.st_size + append > offset) {
6917 (void) SvGROW(sv, (STRLEN)((st.st_size - offset) + append + 1));
6923 else if (RsRECORD(PL_rs)) {
6931 /* Grab the size of the record we're getting */
6932 recsize = SvUV(SvRV(PL_rs)); /* RsRECORD() guarantees > 0. */
6933 buffer = SvGROW(sv, (STRLEN)(recsize + append + 1)) + append;
6936 /* VMS wants read instead of fread, because fread doesn't respect */
6937 /* RMS record boundaries. This is not necessarily a good thing to be */
6938 /* doing, but we've got no other real choice - except avoid stdio
6939 as implementation - perhaps write a :vms layer ?
6941 fd = PerlIO_fileno(fp);
6942 if (fd == -1) { /* in-memory file from PerlIO::Scalar */
6943 bytesread = PerlIO_read(fp, buffer, recsize);
6946 bytesread = PerlLIO_read(fd, buffer, recsize);
6949 bytesread = PerlIO_read(fp, buffer, recsize);
6953 SvCUR_set(sv, bytesread + append);
6954 buffer[bytesread] = '\0';
6955 goto return_string_or_null;
6957 else if (RsPARA(PL_rs)) {
6963 /* Get $/ i.e. PL_rs into same encoding as stream wants */
6964 if (PerlIO_isutf8(fp)) {
6965 rsptr = SvPVutf8(PL_rs, rslen);
6968 if (SvUTF8(PL_rs)) {
6969 if (!sv_utf8_downgrade(PL_rs, TRUE)) {
6970 Perl_croak(aTHX_ "Wide character in $/");
6973 rsptr = SvPV_const(PL_rs, rslen);
6977 rslast = rslen ? rsptr[rslen - 1] : '\0';
6979 if (rspara) { /* have to do this both before and after */
6980 do { /* to make sure file boundaries work right */
6983 i = PerlIO_getc(fp);
6987 PerlIO_ungetc(fp,i);
6993 /* See if we know enough about I/O mechanism to cheat it ! */
6995 /* This used to be #ifdef test - it is made run-time test for ease
6996 of abstracting out stdio interface. One call should be cheap
6997 enough here - and may even be a macro allowing compile
7001 if (PerlIO_fast_gets(fp)) {
7004 * We're going to steal some values from the stdio struct
7005 * and put EVERYTHING in the innermost loop into registers.
7007 register STDCHAR *ptr;
7011 #if defined(VMS) && defined(PERLIO_IS_STDIO)
7012 /* An ungetc()d char is handled separately from the regular
7013 * buffer, so we getc() it back out and stuff it in the buffer.
7015 i = PerlIO_getc(fp);
7016 if (i == EOF) return 0;
7017 *(--((*fp)->_ptr)) = (unsigned char) i;
7021 /* Here is some breathtakingly efficient cheating */
7023 cnt = PerlIO_get_cnt(fp); /* get count into register */
7024 /* make sure we have the room */
7025 if ((I32)(SvLEN(sv) - append) <= cnt + 1) {
7026 /* Not room for all of it
7027 if we are looking for a separator and room for some
7029 if (rslen && cnt > 80 && (I32)SvLEN(sv) > append) {
7030 /* just process what we have room for */
7031 shortbuffered = cnt - SvLEN(sv) + append + 1;
7032 cnt -= shortbuffered;
7036 /* remember that cnt can be negative */
7037 SvGROW(sv, (STRLEN)(append + (cnt <= 0 ? 2 : (cnt + 1))));
7042 bp = (STDCHAR*)SvPVX_const(sv) + append; /* move these two too to registers */
7043 ptr = (STDCHAR*)PerlIO_get_ptr(fp);
7044 DEBUG_P(PerlIO_printf(Perl_debug_log,
7045 "Screamer: entering, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7046 DEBUG_P(PerlIO_printf(Perl_debug_log,
7047 "Screamer: entering: PerlIO * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7048 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7049 PTR2UV(PerlIO_has_base(fp) ? PerlIO_get_base(fp) : 0)));
7054 while (cnt > 0) { /* this | eat */
7056 if ((*bp++ = *ptr++) == rslast) /* really | dust */
7057 goto thats_all_folks; /* screams | sed :-) */
7061 Copy(ptr, bp, cnt, char); /* this | eat */
7062 bp += cnt; /* screams | dust */
7063 ptr += cnt; /* louder | sed :-) */
7068 if (shortbuffered) { /* oh well, must extend */
7069 cnt = shortbuffered;
7071 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
7073 SvGROW(sv, SvLEN(sv) + append + cnt + 2);
7074 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
7078 DEBUG_P(PerlIO_printf(Perl_debug_log,
7079 "Screamer: going to getc, ptr=%"UVuf", cnt=%ld\n",
7080 PTR2UV(ptr),(long)cnt));
7081 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* deregisterize cnt and ptr */
7083 DEBUG_P(PerlIO_printf(Perl_debug_log,
7084 "Screamer: pre: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7085 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7086 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7088 /* This used to call 'filbuf' in stdio form, but as that behaves like
7089 getc when cnt <= 0 we use PerlIO_getc here to avoid introducing
7090 another abstraction. */
7091 i = PerlIO_getc(fp); /* get more characters */
7093 DEBUG_P(PerlIO_printf(Perl_debug_log,
7094 "Screamer: post: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7095 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7096 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7098 cnt = PerlIO_get_cnt(fp);
7099 ptr = (STDCHAR*)PerlIO_get_ptr(fp); /* reregisterize cnt and ptr */
7100 DEBUG_P(PerlIO_printf(Perl_debug_log,
7101 "Screamer: after getc, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7103 if (i == EOF) /* all done for ever? */
7104 goto thats_really_all_folks;
7106 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
7108 SvGROW(sv, bpx + cnt + 2);
7109 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
7111 *bp++ = (STDCHAR)i; /* store character from PerlIO_getc */
7113 if (rslen && (STDCHAR)i == rslast) /* all done for now? */
7114 goto thats_all_folks;
7118 if ((rslen > 1 && (STRLEN)(bp - (STDCHAR*)SvPVX_const(sv)) < rslen) ||
7119 memNE((char*)bp - rslen, rsptr, rslen))
7120 goto screamer; /* go back to the fray */
7121 thats_really_all_folks:
7123 cnt += shortbuffered;
7124 DEBUG_P(PerlIO_printf(Perl_debug_log,
7125 "Screamer: quitting, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7126 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* put these back or we're in trouble */
7127 DEBUG_P(PerlIO_printf(Perl_debug_log,
7128 "Screamer: end: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7129 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7130 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7132 SvCUR_set(sv, bp - (STDCHAR*)SvPVX_const(sv)); /* set length */
7133 DEBUG_P(PerlIO_printf(Perl_debug_log,
7134 "Screamer: done, len=%ld, string=|%.*s|\n",
7135 (long)SvCUR(sv),(int)SvCUR(sv),SvPVX_const(sv)));
7139 /*The big, slow, and stupid way. */
7140 #ifdef USE_HEAP_INSTEAD_OF_STACK /* Even slower way. */
7141 STDCHAR *buf = NULL;
7142 Newx(buf, 8192, STDCHAR);
7150 register const STDCHAR * const bpe = buf + sizeof(buf);
7152 while ((i = PerlIO_getc(fp)) != EOF && (*bp++ = (STDCHAR)i) != rslast && bp < bpe)
7153 ; /* keep reading */
7157 cnt = PerlIO_read(fp,(char*)buf, sizeof(buf));
7158 /* Accomodate broken VAXC compiler, which applies U8 cast to
7159 * both args of ?: operator, causing EOF to change into 255
7162 i = (U8)buf[cnt - 1];
7168 cnt = 0; /* we do need to re-set the sv even when cnt <= 0 */
7170 sv_catpvn(sv, (char *) buf, cnt);
7172 sv_setpvn(sv, (char *) buf, cnt);
7174 if (i != EOF && /* joy */
7176 SvCUR(sv) < rslen ||
7177 memNE(SvPVX_const(sv) + SvCUR(sv) - rslen, rsptr, rslen)))
7181 * If we're reading from a TTY and we get a short read,
7182 * indicating that the user hit his EOF character, we need
7183 * to notice it now, because if we try to read from the TTY
7184 * again, the EOF condition will disappear.
7186 * The comparison of cnt to sizeof(buf) is an optimization
7187 * that prevents unnecessary calls to feof().
7191 if (!(cnt < (I32)sizeof(buf) && PerlIO_eof(fp)))
7195 #ifdef USE_HEAP_INSTEAD_OF_STACK
7200 if (rspara) { /* have to do this both before and after */
7201 while (i != EOF) { /* to make sure file boundaries work right */
7202 i = PerlIO_getc(fp);
7204 PerlIO_ungetc(fp,i);
7210 return_string_or_null:
7211 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7217 Auto-increment of the value in the SV, doing string to numeric conversion
7218 if necessary. Handles 'get' magic.
7224 Perl_sv_inc(pTHX_ register SV *const sv)
7233 if (SvTHINKFIRST(sv)) {
7235 sv_force_normal_flags(sv, 0);
7236 if (SvREADONLY(sv)) {
7237 if (IN_PERL_RUNTIME)
7238 Perl_croak(aTHX_ "%s", PL_no_modify);
7242 if (SvAMAGIC(sv) && AMG_CALLun(sv,inc))
7244 i = PTR2IV(SvRV(sv));
7249 flags = SvFLAGS(sv);
7250 if ((flags & (SVp_NOK|SVp_IOK)) == SVp_NOK) {
7251 /* It's (privately or publicly) a float, but not tested as an
7252 integer, so test it to see. */
7254 flags = SvFLAGS(sv);
7256 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
7257 /* It's publicly an integer, or privately an integer-not-float */
7258 #ifdef PERL_PRESERVE_IVUV
7262 if (SvUVX(sv) == UV_MAX)
7263 sv_setnv(sv, UV_MAX_P1);
7265 (void)SvIOK_only_UV(sv);
7266 SvUV_set(sv, SvUVX(sv) + 1);
7268 if (SvIVX(sv) == IV_MAX)
7269 sv_setuv(sv, (UV)IV_MAX + 1);
7271 (void)SvIOK_only(sv);
7272 SvIV_set(sv, SvIVX(sv) + 1);
7277 if (flags & SVp_NOK) {
7278 const NV was = SvNVX(sv);
7279 if (NV_OVERFLOWS_INTEGERS_AT &&
7280 was >= NV_OVERFLOWS_INTEGERS_AT && ckWARN(WARN_IMPRECISION)) {
7281 Perl_warner(aTHX_ packWARN(WARN_IMPRECISION),
7282 "Lost precision when incrementing %" NVff " by 1",
7285 (void)SvNOK_only(sv);
7286 SvNV_set(sv, was + 1.0);
7290 if (!(flags & SVp_POK) || !*SvPVX_const(sv)) {
7291 if ((flags & SVTYPEMASK) < SVt_PVIV)
7292 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV ? SVt_PVIV : SVt_IV));
7293 (void)SvIOK_only(sv);
7298 while (isALPHA(*d)) d++;
7299 while (isDIGIT(*d)) d++;
7301 #ifdef PERL_PRESERVE_IVUV
7302 /* Got to punt this as an integer if needs be, but we don't issue
7303 warnings. Probably ought to make the sv_iv_please() that does
7304 the conversion if possible, and silently. */
7305 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
7306 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
7307 /* Need to try really hard to see if it's an integer.
7308 9.22337203685478e+18 is an integer.
7309 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
7310 so $a="9.22337203685478e+18"; $a+0; $a++
7311 needs to be the same as $a="9.22337203685478e+18"; $a++
7318 /* sv_2iv *should* have made this an NV */
7319 if (flags & SVp_NOK) {
7320 (void)SvNOK_only(sv);
7321 SvNV_set(sv, SvNVX(sv) + 1.0);
7324 /* I don't think we can get here. Maybe I should assert this
7325 And if we do get here I suspect that sv_setnv will croak. NWC
7327 #if defined(USE_LONG_DOUBLE)
7328 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",
7329 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7331 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_inc punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
7332 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7335 #endif /* PERL_PRESERVE_IVUV */
7336 sv_setnv(sv,Atof(SvPVX_const(sv)) + 1.0);
7340 while (d >= SvPVX_const(sv)) {
7348 /* MKS: The original code here died if letters weren't consecutive.
7349 * at least it didn't have to worry about non-C locales. The
7350 * new code assumes that ('z'-'a')==('Z'-'A'), letters are
7351 * arranged in order (although not consecutively) and that only
7352 * [A-Za-z] are accepted by isALPHA in the C locale.
7354 if (*d != 'z' && *d != 'Z') {
7355 do { ++*d; } while (!isALPHA(*d));
7358 *(d--) -= 'z' - 'a';
7363 *(d--) -= 'z' - 'a' + 1;
7367 /* oh,oh, the number grew */
7368 SvGROW(sv, SvCUR(sv) + 2);
7369 SvCUR_set(sv, SvCUR(sv) + 1);
7370 for (d = SvPVX(sv) + SvCUR(sv); d > SvPVX_const(sv); d--)
7381 Auto-decrement of the value in the SV, doing string to numeric conversion
7382 if necessary. Handles 'get' magic.
7388 Perl_sv_dec(pTHX_ register SV *const sv)
7396 if (SvTHINKFIRST(sv)) {
7398 sv_force_normal_flags(sv, 0);
7399 if (SvREADONLY(sv)) {
7400 if (IN_PERL_RUNTIME)
7401 Perl_croak(aTHX_ "%s", PL_no_modify);
7405 if (SvAMAGIC(sv) && AMG_CALLun(sv,dec))
7407 i = PTR2IV(SvRV(sv));
7412 /* Unlike sv_inc we don't have to worry about string-never-numbers
7413 and keeping them magic. But we mustn't warn on punting */
7414 flags = SvFLAGS(sv);
7415 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
7416 /* It's publicly an integer, or privately an integer-not-float */
7417 #ifdef PERL_PRESERVE_IVUV
7421 if (SvUVX(sv) == 0) {
7422 (void)SvIOK_only(sv);
7426 (void)SvIOK_only_UV(sv);
7427 SvUV_set(sv, SvUVX(sv) - 1);
7430 if (SvIVX(sv) == IV_MIN) {
7431 sv_setnv(sv, (NV)IV_MIN);
7435 (void)SvIOK_only(sv);
7436 SvIV_set(sv, SvIVX(sv) - 1);
7441 if (flags & SVp_NOK) {
7444 const NV was = SvNVX(sv);
7445 if (NV_OVERFLOWS_INTEGERS_AT &&
7446 was <= -NV_OVERFLOWS_INTEGERS_AT && ckWARN(WARN_IMPRECISION)) {
7447 Perl_warner(aTHX_ packWARN(WARN_IMPRECISION),
7448 "Lost precision when decrementing %" NVff " by 1",
7451 (void)SvNOK_only(sv);
7452 SvNV_set(sv, was - 1.0);
7456 if (!(flags & SVp_POK)) {
7457 if ((flags & SVTYPEMASK) < SVt_PVIV)
7458 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV) ? SVt_PVIV : SVt_IV);
7460 (void)SvIOK_only(sv);
7463 #ifdef PERL_PRESERVE_IVUV
7465 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
7466 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
7467 /* Need to try really hard to see if it's an integer.
7468 9.22337203685478e+18 is an integer.
7469 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
7470 so $a="9.22337203685478e+18"; $a+0; $a--
7471 needs to be the same as $a="9.22337203685478e+18"; $a--
7478 /* sv_2iv *should* have made this an NV */
7479 if (flags & SVp_NOK) {
7480 (void)SvNOK_only(sv);
7481 SvNV_set(sv, SvNVX(sv) - 1.0);
7484 /* I don't think we can get here. Maybe I should assert this
7485 And if we do get here I suspect that sv_setnv will croak. NWC
7487 #if defined(USE_LONG_DOUBLE)
7488 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",
7489 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7491 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_dec punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
7492 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7496 #endif /* PERL_PRESERVE_IVUV */
7497 sv_setnv(sv,Atof(SvPVX_const(sv)) - 1.0); /* punt */
7501 =for apidoc sv_mortalcopy
7503 Creates a new SV which is a copy of the original SV (using C<sv_setsv>).
7504 The new SV is marked as mortal. It will be destroyed "soon", either by an
7505 explicit call to FREETMPS, or by an implicit call at places such as
7506 statement boundaries. See also C<sv_newmortal> and C<sv_2mortal>.
7511 /* Make a string that will exist for the duration of the expression
7512 * evaluation. Actually, it may have to last longer than that, but
7513 * hopefully we won't free it until it has been assigned to a
7514 * permanent location. */
7517 Perl_sv_mortalcopy(pTHX_ SV *const oldstr)
7523 sv_setsv(sv,oldstr);
7525 PL_tmps_stack[++PL_tmps_ix] = sv;
7531 =for apidoc sv_newmortal
7533 Creates a new null SV which is mortal. The reference count of the SV is
7534 set to 1. It will be destroyed "soon", either by an explicit call to
7535 FREETMPS, or by an implicit call at places such as statement boundaries.
7536 See also C<sv_mortalcopy> and C<sv_2mortal>.
7542 Perl_sv_newmortal(pTHX)
7548 SvFLAGS(sv) = SVs_TEMP;
7550 PL_tmps_stack[++PL_tmps_ix] = sv;
7556 =for apidoc newSVpvn_flags
7558 Creates a new SV and copies a string into it. The reference count for the
7559 SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
7560 string. You are responsible for ensuring that the source string is at least
7561 C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
7562 Currently the only flag bits accepted are C<SVf_UTF8> and C<SVs_TEMP>.
7563 If C<SVs_TEMP> is set, then C<sv2mortal()> is called on the result before
7564 returning. If C<SVf_UTF8> is set, then it will be set on the new SV.
7565 C<newSVpvn_utf8()> is a convenience wrapper for this function, defined as
7567 #define newSVpvn_utf8(s, len, u) \
7568 newSVpvn_flags((s), (len), (u) ? SVf_UTF8 : 0)
7574 Perl_newSVpvn_flags(pTHX_ const char *const s, const STRLEN len, const U32 flags)
7579 /* All the flags we don't support must be zero.
7580 And we're new code so I'm going to assert this from the start. */
7581 assert(!(flags & ~(SVf_UTF8|SVs_TEMP)));
7583 sv_setpvn(sv,s,len);
7584 SvFLAGS(sv) |= (flags & SVf_UTF8);
7585 return (flags & SVs_TEMP) ? sv_2mortal(sv) : sv;
7589 =for apidoc sv_2mortal
7591 Marks an existing SV as mortal. The SV will be destroyed "soon", either
7592 by an explicit call to FREETMPS, or by an implicit call at places such as
7593 statement boundaries. SvTEMP() is turned on which means that the SV's
7594 string buffer can be "stolen" if this SV is copied. See also C<sv_newmortal>
7595 and C<sv_mortalcopy>.
7601 Perl_sv_2mortal(pTHX_ register SV *const sv)
7606 if (SvREADONLY(sv) && SvIMMORTAL(sv))
7609 PL_tmps_stack[++PL_tmps_ix] = sv;
7617 Creates a new SV and copies a string into it. The reference count for the
7618 SV is set to 1. If C<len> is zero, Perl will compute the length using
7619 strlen(). For efficiency, consider using C<newSVpvn> instead.
7625 Perl_newSVpv(pTHX_ const char *const s, const STRLEN len)
7631 sv_setpvn(sv, s, len || s == NULL ? len : strlen(s));
7636 =for apidoc newSVpvn
7638 Creates a new SV and copies a string into it. The reference count for the
7639 SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
7640 string. You are responsible for ensuring that the source string is at least
7641 C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
7647 Perl_newSVpvn(pTHX_ const char *const s, const STRLEN len)
7653 sv_setpvn(sv,s,len);
7658 =for apidoc newSVhek
7660 Creates a new SV from the hash key structure. It will generate scalars that
7661 point to the shared string table where possible. Returns a new (undefined)
7662 SV if the hek is NULL.
7668 Perl_newSVhek(pTHX_ const HEK *const hek)
7678 if (HEK_LEN(hek) == HEf_SVKEY) {
7679 return newSVsv(*(SV**)HEK_KEY(hek));
7681 const int flags = HEK_FLAGS(hek);
7682 if (flags & HVhek_WASUTF8) {
7684 Andreas would like keys he put in as utf8 to come back as utf8
7686 STRLEN utf8_len = HEK_LEN(hek);
7687 const U8 *as_utf8 = bytes_to_utf8 ((U8*)HEK_KEY(hek), &utf8_len);
7688 SV * const sv = newSVpvn ((const char*)as_utf8, utf8_len);
7691 Safefree (as_utf8); /* bytes_to_utf8() allocates a new string */
7693 } else if (flags & (HVhek_REHASH|HVhek_UNSHARED)) {
7694 /* We don't have a pointer to the hv, so we have to replicate the
7695 flag into every HEK. This hv is using custom a hasing
7696 algorithm. Hence we can't return a shared string scalar, as
7697 that would contain the (wrong) hash value, and might get passed
7698 into an hv routine with a regular hash.
7699 Similarly, a hash that isn't using shared hash keys has to have
7700 the flag in every key so that we know not to try to call
7701 share_hek_kek on it. */
7703 SV * const sv = newSVpvn (HEK_KEY(hek), HEK_LEN(hek));
7708 /* This will be overwhelminly the most common case. */
7710 /* Inline most of newSVpvn_share(), because share_hek_hek() is far
7711 more efficient than sharepvn(). */
7715 sv_upgrade(sv, SVt_PV);
7716 SvPV_set(sv, (char *)HEK_KEY(share_hek_hek(hek)));
7717 SvCUR_set(sv, HEK_LEN(hek));
7730 =for apidoc newSVpvn_share
7732 Creates a new SV with its SvPVX_const pointing to a shared string in the string
7733 table. If the string does not already exist in the table, it is created
7734 first. Turns on READONLY and FAKE. If the C<hash> parameter is non-zero, that
7735 value is used; otherwise the hash is computed. The string's hash can be later
7736 be retrieved from the SV with the C<SvSHARED_HASH()> macro. The idea here is
7737 that as the string table is used for shared hash keys these strings will have
7738 SvPVX_const == HeKEY and hash lookup will avoid string compare.
7744 Perl_newSVpvn_share(pTHX_ const char *src, I32 len, U32 hash)
7748 bool is_utf8 = FALSE;
7749 const char *const orig_src = src;
7752 STRLEN tmplen = -len;
7754 /* See the note in hv.c:hv_fetch() --jhi */
7755 src = (char*)bytes_from_utf8((const U8*)src, &tmplen, &is_utf8);
7759 PERL_HASH(hash, src, len);
7761 /* The logic for this is inlined in S_mro_get_linear_isa_dfs(), so if it
7762 changes here, update it there too. */
7763 sv_upgrade(sv, SVt_PV);
7764 SvPV_set(sv, sharepvn(src, is_utf8?-len:len, hash));
7772 if (src != orig_src)
7778 #if defined(PERL_IMPLICIT_CONTEXT)
7780 /* pTHX_ magic can't cope with varargs, so this is a no-context
7781 * version of the main function, (which may itself be aliased to us).
7782 * Don't access this version directly.
7786 Perl_newSVpvf_nocontext(const char *const pat, ...)
7792 PERL_ARGS_ASSERT_NEWSVPVF_NOCONTEXT;
7794 va_start(args, pat);
7795 sv = vnewSVpvf(pat, &args);
7802 =for apidoc newSVpvf
7804 Creates a new SV and initializes it with the string formatted like
7811 Perl_newSVpvf(pTHX_ const char *const pat, ...)
7816 PERL_ARGS_ASSERT_NEWSVPVF;
7818 va_start(args, pat);
7819 sv = vnewSVpvf(pat, &args);
7824 /* backend for newSVpvf() and newSVpvf_nocontext() */
7827 Perl_vnewSVpvf(pTHX_ const char *const pat, va_list *const args)
7832 PERL_ARGS_ASSERT_VNEWSVPVF;
7835 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
7842 Creates a new SV and copies a floating point value into it.
7843 The reference count for the SV is set to 1.
7849 Perl_newSVnv(pTHX_ const NV n)
7862 Creates a new SV and copies an integer into it. The reference count for the
7869 Perl_newSViv(pTHX_ const IV i)
7882 Creates a new SV and copies an unsigned integer into it.
7883 The reference count for the SV is set to 1.
7889 Perl_newSVuv(pTHX_ const UV u)
7900 =for apidoc newSV_type
7902 Creates a new SV, of the type specified. The reference count for the new SV
7909 Perl_newSV_type(pTHX_ const svtype type)
7914 sv_upgrade(sv, type);
7919 =for apidoc newRV_noinc
7921 Creates an RV wrapper for an SV. The reference count for the original
7922 SV is B<not> incremented.
7928 Perl_newRV_noinc(pTHX_ SV *const tmpRef)
7931 register SV *sv = newSV_type(SVt_IV);
7933 PERL_ARGS_ASSERT_NEWRV_NOINC;
7936 SvRV_set(sv, tmpRef);
7941 /* newRV_inc is the official function name to use now.
7942 * newRV_inc is in fact #defined to newRV in sv.h
7946 Perl_newRV(pTHX_ SV *const sv)
7950 PERL_ARGS_ASSERT_NEWRV;
7952 return newRV_noinc(SvREFCNT_inc_simple_NN(sv));
7958 Creates a new SV which is an exact duplicate of the original SV.
7965 Perl_newSVsv(pTHX_ register SV *const old)
7972 if (SvTYPE(old) == SVTYPEMASK) {
7973 if (ckWARN_d(WARN_INTERNAL))
7974 Perl_warner(aTHX_ packWARN(WARN_INTERNAL), "semi-panic: attempt to dup freed string");
7978 /* SV_GMAGIC is the default for sv_setv()
7979 SV_NOSTEAL prevents TEMP buffers being, well, stolen, and saves games
7980 with SvTEMP_off and SvTEMP_on round a call to sv_setsv. */
7981 sv_setsv_flags(sv, old, SV_GMAGIC | SV_NOSTEAL);
7986 =for apidoc sv_reset
7988 Underlying implementation for the C<reset> Perl function.
7989 Note that the perl-level function is vaguely deprecated.
7995 Perl_sv_reset(pTHX_ register const char *s, HV *const stash)
7998 char todo[PERL_UCHAR_MAX+1];
8000 PERL_ARGS_ASSERT_SV_RESET;
8005 if (!*s) { /* reset ?? searches */
8006 MAGIC * const mg = mg_find((const SV *)stash, PERL_MAGIC_symtab);
8008 const U32 count = mg->mg_len / sizeof(PMOP**);
8009 PMOP **pmp = (PMOP**) mg->mg_ptr;
8010 PMOP *const *const end = pmp + count;
8014 SvREADONLY_off(PL_regex_pad[(*pmp)->op_pmoffset]);
8016 (*pmp)->op_pmflags &= ~PMf_USED;
8024 /* reset variables */
8026 if (!HvARRAY(stash))
8029 Zero(todo, 256, char);
8032 I32 i = (unsigned char)*s;
8036 max = (unsigned char)*s++;
8037 for ( ; i <= max; i++) {
8040 for (i = 0; i <= (I32) HvMAX(stash); i++) {
8042 for (entry = HvARRAY(stash)[i];
8044 entry = HeNEXT(entry))
8049 if (!todo[(U8)*HeKEY(entry)])
8051 gv = MUTABLE_GV(HeVAL(entry));
8054 if (SvTHINKFIRST(sv)) {
8055 if (!SvREADONLY(sv) && SvROK(sv))
8057 /* XXX Is this continue a bug? Why should THINKFIRST
8058 exempt us from resetting arrays and hashes? */
8062 if (SvTYPE(sv) >= SVt_PV) {
8064 if (SvPVX_const(sv) != NULL)
8072 if (GvHV(gv) && !HvNAME_get(GvHV(gv))) {
8074 Perl_die(aTHX_ "Can't reset %%ENV on this system");
8077 # if defined(USE_ENVIRON_ARRAY)
8080 # endif /* USE_ENVIRON_ARRAY */
8091 Using various gambits, try to get an IO from an SV: the IO slot if its a
8092 GV; or the recursive result if we're an RV; or the IO slot of the symbol
8093 named after the PV if we're a string.
8099 Perl_sv_2io(pTHX_ SV *const sv)
8104 PERL_ARGS_ASSERT_SV_2IO;
8106 switch (SvTYPE(sv)) {
8108 io = MUTABLE_IO(sv);
8111 if (isGV_with_GP(sv)) {
8112 gv = MUTABLE_GV(sv);
8115 Perl_croak(aTHX_ "Bad filehandle: %s", GvNAME(gv));
8121 Perl_croak(aTHX_ PL_no_usym, "filehandle");
8123 return sv_2io(SvRV(sv));
8124 gv = gv_fetchsv(sv, 0, SVt_PVIO);
8130 Perl_croak(aTHX_ "Bad filehandle: %"SVf, SVfARG(sv));
8139 Using various gambits, try to get a CV from an SV; in addition, try if
8140 possible to set C<*st> and C<*gvp> to the stash and GV associated with it.
8141 The flags in C<lref> are passed to sv_fetchsv.
8147 Perl_sv_2cv(pTHX_ SV *sv, HV **const st, GV **const gvp, const I32 lref)
8153 PERL_ARGS_ASSERT_SV_2CV;
8160 switch (SvTYPE(sv)) {
8164 return MUTABLE_CV(sv);
8171 if (isGV_with_GP(sv)) {
8172 gv = MUTABLE_GV(sv);
8181 SV * const *sp = &sv; /* Used in tryAMAGICunDEREF macro. */
8183 tryAMAGICunDEREF(to_cv);
8186 if (SvTYPE(sv) == SVt_PVCV) {
8187 cv = MUTABLE_CV(sv);
8192 else if(isGV_with_GP(sv))
8193 gv = MUTABLE_GV(sv);
8195 Perl_croak(aTHX_ "Not a subroutine reference");
8197 else if (isGV_with_GP(sv)) {
8199 gv = MUTABLE_GV(sv);
8202 gv = gv_fetchsv(sv, lref, SVt_PVCV); /* Calls get magic */
8208 /* Some flags to gv_fetchsv mean don't really create the GV */
8209 if (!isGV_with_GP(gv)) {
8215 if (lref && !GvCVu(gv)) {
8219 gv_efullname3(tmpsv, gv, NULL);
8220 /* XXX this is probably not what they think they're getting.
8221 * It has the same effect as "sub name;", i.e. just a forward
8223 newSUB(start_subparse(FALSE, 0),
8224 newSVOP(OP_CONST, 0, tmpsv),
8228 Perl_croak(aTHX_ "Unable to create sub named \"%"SVf"\"",
8229 SVfARG(SvOK(sv) ? sv : &PL_sv_no));
8238 Returns true if the SV has a true value by Perl's rules.
8239 Use the C<SvTRUE> macro instead, which may call C<sv_true()> or may
8240 instead use an in-line version.
8246 Perl_sv_true(pTHX_ register SV *const sv)
8251 register const XPV* const tXpv = (XPV*)SvANY(sv);
8253 (tXpv->xpv_cur > 1 ||
8254 (tXpv->xpv_cur && *sv->sv_u.svu_pv != '0')))
8261 return SvIVX(sv) != 0;
8264 return SvNVX(sv) != 0.0;
8266 return sv_2bool(sv);
8272 =for apidoc sv_pvn_force
8274 Get a sensible string out of the SV somehow.
8275 A private implementation of the C<SvPV_force> macro for compilers which
8276 can't cope with complex macro expressions. Always use the macro instead.
8278 =for apidoc sv_pvn_force_flags
8280 Get a sensible string out of the SV somehow.
8281 If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<sv> if
8282 appropriate, else not. C<sv_pvn_force> and C<sv_pvn_force_nomg> are
8283 implemented in terms of this function.
8284 You normally want to use the various wrapper macros instead: see
8285 C<SvPV_force> and C<SvPV_force_nomg>
8291 Perl_sv_pvn_force_flags(pTHX_ SV *const sv, STRLEN *const lp, const I32 flags)
8295 PERL_ARGS_ASSERT_SV_PVN_FORCE_FLAGS;
8297 if (SvTHINKFIRST(sv) && !SvROK(sv))
8298 sv_force_normal_flags(sv, 0);
8308 if (SvREADONLY(sv) && !(flags & SV_MUTABLE_RETURN)) {
8309 const char * const ref = sv_reftype(sv,0);
8311 Perl_croak(aTHX_ "Can't coerce readonly %s to string in %s",
8312 ref, OP_NAME(PL_op));
8314 Perl_croak(aTHX_ "Can't coerce readonly %s to string", ref);
8316 if ((SvTYPE(sv) > SVt_PVLV && SvTYPE(sv) != SVt_PVFM)
8317 || isGV_with_GP(sv))
8318 Perl_croak(aTHX_ "Can't coerce %s to string in %s", sv_reftype(sv,0),
8320 s = sv_2pv_flags(sv, &len, flags);
8324 if (s != SvPVX_const(sv)) { /* Almost, but not quite, sv_setpvn() */
8327 SvUPGRADE(sv, SVt_PV); /* Never FALSE */
8328 SvGROW(sv, len + 1);
8329 Move(s,SvPVX(sv),len,char);
8331 SvPVX(sv)[len] = '\0';
8334 SvPOK_on(sv); /* validate pointer */
8336 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
8337 PTR2UV(sv),SvPVX_const(sv)));
8340 return SvPVX_mutable(sv);
8344 =for apidoc sv_pvbyten_force
8346 The backend for the C<SvPVbytex_force> macro. Always use the macro instead.
8352 Perl_sv_pvbyten_force(pTHX_ SV *const sv, STRLEN *const lp)
8354 PERL_ARGS_ASSERT_SV_PVBYTEN_FORCE;
8356 sv_pvn_force(sv,lp);
8357 sv_utf8_downgrade(sv,0);
8363 =for apidoc sv_pvutf8n_force
8365 The backend for the C<SvPVutf8x_force> macro. Always use the macro instead.
8371 Perl_sv_pvutf8n_force(pTHX_ SV *const sv, STRLEN *const lp)
8373 PERL_ARGS_ASSERT_SV_PVUTF8N_FORCE;
8375 sv_pvn_force(sv,lp);
8376 sv_utf8_upgrade(sv);
8382 =for apidoc sv_reftype
8384 Returns a string describing what the SV is a reference to.
8390 Perl_sv_reftype(pTHX_ const SV *const sv, const int ob)
8392 PERL_ARGS_ASSERT_SV_REFTYPE;
8394 /* The fact that I don't need to downcast to char * everywhere, only in ?:
8395 inside return suggests a const propagation bug in g++. */
8396 if (ob && SvOBJECT(sv)) {
8397 char * const name = HvNAME_get(SvSTASH(sv));
8398 return name ? name : (char *) "__ANON__";
8401 switch (SvTYPE(sv)) {
8416 case SVt_PVLV: return (char *) (SvROK(sv) ? "REF"
8417 /* tied lvalues should appear to be
8418 * scalars for backwards compatitbility */
8419 : (LvTYPE(sv) == 't' || LvTYPE(sv) == 'T')
8420 ? "SCALAR" : "LVALUE");
8421 case SVt_PVAV: return "ARRAY";
8422 case SVt_PVHV: return "HASH";
8423 case SVt_PVCV: return "CODE";
8424 case SVt_PVGV: return (char *) (isGV_with_GP(sv)
8425 ? "GLOB" : "SCALAR");
8426 case SVt_PVFM: return "FORMAT";
8427 case SVt_PVIO: return "IO";
8428 case SVt_BIND: return "BIND";
8429 case SVt_REGEXP: return "REGEXP";
8430 default: return "UNKNOWN";
8436 =for apidoc sv_isobject
8438 Returns a boolean indicating whether the SV is an RV pointing to a blessed
8439 object. If the SV is not an RV, or if the object is not blessed, then this
8446 Perl_sv_isobject(pTHX_ SV *sv)
8462 Returns a boolean indicating whether the SV is blessed into the specified
8463 class. This does not check for subtypes; use C<sv_derived_from> to verify
8464 an inheritance relationship.
8470 Perl_sv_isa(pTHX_ SV *sv, const char *const name)
8474 PERL_ARGS_ASSERT_SV_ISA;
8484 hvname = HvNAME_get(SvSTASH(sv));
8488 return strEQ(hvname, name);
8494 Creates a new SV for the RV, C<rv>, to point to. If C<rv> is not an RV then
8495 it will be upgraded to one. If C<classname> is non-null then the new SV will
8496 be blessed in the specified package. The new SV is returned and its
8497 reference count is 1.
8503 Perl_newSVrv(pTHX_ SV *const rv, const char *const classname)
8508 PERL_ARGS_ASSERT_NEWSVRV;
8512 SV_CHECK_THINKFIRST_COW_DROP(rv);
8513 (void)SvAMAGIC_off(rv);
8515 if (SvTYPE(rv) >= SVt_PVMG) {
8516 const U32 refcnt = SvREFCNT(rv);
8520 SvREFCNT(rv) = refcnt;
8522 sv_upgrade(rv, SVt_IV);
8523 } else if (SvROK(rv)) {
8524 SvREFCNT_dec(SvRV(rv));
8526 prepare_SV_for_RV(rv);
8534 HV* const stash = gv_stashpv(classname, GV_ADD);
8535 (void)sv_bless(rv, stash);
8541 =for apidoc sv_setref_pv
8543 Copies a pointer into a new SV, optionally blessing the SV. The C<rv>
8544 argument will be upgraded to an RV. That RV will be modified to point to
8545 the new SV. If the C<pv> argument is NULL then C<PL_sv_undef> will be placed
8546 into the SV. The C<classname> argument indicates the package for the
8547 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
8548 will have a reference count of 1, and the RV will be returned.
8550 Do not use with other Perl types such as HV, AV, SV, CV, because those
8551 objects will become corrupted by the pointer copy process.
8553 Note that C<sv_setref_pvn> copies the string while this copies the pointer.
8559 Perl_sv_setref_pv(pTHX_ SV *const rv, const char *const classname, void *const pv)
8563 PERL_ARGS_ASSERT_SV_SETREF_PV;
8566 sv_setsv(rv, &PL_sv_undef);
8570 sv_setiv(newSVrv(rv,classname), PTR2IV(pv));
8575 =for apidoc sv_setref_iv
8577 Copies an integer into a new SV, optionally blessing the SV. The C<rv>
8578 argument will be upgraded to an RV. That RV will be modified to point to
8579 the new SV. The C<classname> argument indicates the package for the
8580 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
8581 will have a reference count of 1, and the RV will be returned.
8587 Perl_sv_setref_iv(pTHX_ SV *const rv, const char *const classname, const IV iv)
8589 PERL_ARGS_ASSERT_SV_SETREF_IV;
8591 sv_setiv(newSVrv(rv,classname), iv);
8596 =for apidoc sv_setref_uv
8598 Copies an unsigned integer into a new SV, optionally blessing the SV. The C<rv>
8599 argument will be upgraded to an RV. That RV will be modified to point to
8600 the new SV. The C<classname> argument indicates the package for the
8601 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
8602 will have a reference count of 1, and the RV will be returned.
8608 Perl_sv_setref_uv(pTHX_ SV *const rv, const char *const classname, const UV uv)
8610 PERL_ARGS_ASSERT_SV_SETREF_UV;
8612 sv_setuv(newSVrv(rv,classname), uv);
8617 =for apidoc sv_setref_nv
8619 Copies a double into a new SV, optionally blessing the SV. The C<rv>
8620 argument will be upgraded to an RV. That RV will be modified to point to
8621 the new SV. The C<classname> argument indicates the package for the
8622 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
8623 will have a reference count of 1, and the RV will be returned.
8629 Perl_sv_setref_nv(pTHX_ SV *const rv, const char *const classname, const NV nv)
8631 PERL_ARGS_ASSERT_SV_SETREF_NV;
8633 sv_setnv(newSVrv(rv,classname), nv);
8638 =for apidoc sv_setref_pvn
8640 Copies a string into a new SV, optionally blessing the SV. The length of the
8641 string must be specified with C<n>. The C<rv> argument will be upgraded to
8642 an RV. That RV will be modified to point to the new SV. The C<classname>
8643 argument indicates the package for the blessing. Set C<classname> to
8644 C<NULL> to avoid the blessing. The new SV will have a reference count
8645 of 1, and the RV will be returned.
8647 Note that C<sv_setref_pv> copies the pointer while this copies the string.
8653 Perl_sv_setref_pvn(pTHX_ SV *const rv, const char *const classname,
8654 const char *const pv, const STRLEN n)
8656 PERL_ARGS_ASSERT_SV_SETREF_PVN;
8658 sv_setpvn(newSVrv(rv,classname), pv, n);
8663 =for apidoc sv_bless
8665 Blesses an SV into a specified package. The SV must be an RV. The package
8666 must be designated by its stash (see C<gv_stashpv()>). The reference count
8667 of the SV is unaffected.
8673 Perl_sv_bless(pTHX_ SV *const sv, HV *const stash)
8678 PERL_ARGS_ASSERT_SV_BLESS;
8681 Perl_croak(aTHX_ "Can't bless non-reference value");
8683 if (SvFLAGS(tmpRef) & (SVs_OBJECT|SVf_READONLY)) {
8684 if (SvIsCOW(tmpRef))
8685 sv_force_normal_flags(tmpRef, 0);
8686 if (SvREADONLY(tmpRef))
8687 Perl_croak(aTHX_ "%s", PL_no_modify);
8688 if (SvOBJECT(tmpRef)) {
8689 if (SvTYPE(tmpRef) != SVt_PVIO)
8691 SvREFCNT_dec(SvSTASH(tmpRef));
8694 SvOBJECT_on(tmpRef);
8695 if (SvTYPE(tmpRef) != SVt_PVIO)
8697 SvUPGRADE(tmpRef, SVt_PVMG);
8698 SvSTASH_set(tmpRef, MUTABLE_HV(SvREFCNT_inc_simple(stash)));
8703 (void)SvAMAGIC_off(sv);
8705 if(SvSMAGICAL(tmpRef))
8706 if(mg_find(tmpRef, PERL_MAGIC_ext) || mg_find(tmpRef, PERL_MAGIC_uvar))
8714 /* Downgrades a PVGV to a PVMG.
8718 S_sv_unglob(pTHX_ SV *const sv)
8723 SV * const temp = sv_newmortal();
8725 PERL_ARGS_ASSERT_SV_UNGLOB;
8727 assert(SvTYPE(sv) == SVt_PVGV);
8729 gv_efullname3(temp, MUTABLE_GV(sv), "*");
8732 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
8733 && HvNAME_get(stash))
8734 mro_method_changed_in(stash);
8735 gp_free(MUTABLE_GV(sv));
8738 sv_del_backref(MUTABLE_SV(GvSTASH(sv)), sv);
8742 if (GvNAME_HEK(sv)) {
8743 unshare_hek(GvNAME_HEK(sv));
8745 isGV_with_GP_off(sv);
8747 /* need to keep SvANY(sv) in the right arena */
8748 xpvmg = new_XPVMG();
8749 StructCopy(SvANY(sv), xpvmg, XPVMG);
8750 del_XPVGV(SvANY(sv));
8753 SvFLAGS(sv) &= ~SVTYPEMASK;
8754 SvFLAGS(sv) |= SVt_PVMG;
8756 /* Intentionally not calling any local SET magic, as this isn't so much a
8757 set operation as merely an internal storage change. */
8758 sv_setsv_flags(sv, temp, 0);
8762 =for apidoc sv_unref_flags
8764 Unsets the RV status of the SV, and decrements the reference count of
8765 whatever was being referenced by the RV. This can almost be thought of
8766 as a reversal of C<newSVrv>. The C<cflags> argument can contain
8767 C<SV_IMMEDIATE_UNREF> to force the reference count to be decremented
8768 (otherwise the decrementing is conditional on the reference count being
8769 different from one or the reference being a readonly SV).
8776 Perl_sv_unref_flags(pTHX_ SV *const ref, const U32 flags)
8778 SV* const target = SvRV(ref);
8780 PERL_ARGS_ASSERT_SV_UNREF_FLAGS;
8782 if (SvWEAKREF(ref)) {
8783 sv_del_backref(target, ref);
8785 SvRV_set(ref, NULL);
8788 SvRV_set(ref, NULL);
8790 /* You can't have a || SvREADONLY(target) here, as $a = $$a, where $a was
8791 assigned to as BEGIN {$a = \"Foo"} will fail. */
8792 if (SvREFCNT(target) != 1 || (flags & SV_IMMEDIATE_UNREF))
8793 SvREFCNT_dec(target);
8794 else /* XXX Hack, but hard to make $a=$a->[1] work otherwise */
8795 sv_2mortal(target); /* Schedule for freeing later */
8799 =for apidoc sv_untaint
8801 Untaint an SV. Use C<SvTAINTED_off> instead.
8806 Perl_sv_untaint(pTHX_ SV *const sv)
8808 PERL_ARGS_ASSERT_SV_UNTAINT;
8810 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
8811 MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
8818 =for apidoc sv_tainted
8820 Test an SV for taintedness. Use C<SvTAINTED> instead.
8825 Perl_sv_tainted(pTHX_ SV *const sv)
8827 PERL_ARGS_ASSERT_SV_TAINTED;
8829 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
8830 const MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
8831 if (mg && (mg->mg_len & 1) )
8838 =for apidoc sv_setpviv
8840 Copies an integer into the given SV, also updating its string value.
8841 Does not handle 'set' magic. See C<sv_setpviv_mg>.
8847 Perl_sv_setpviv(pTHX_ SV *const sv, const IV iv)
8849 char buf[TYPE_CHARS(UV)];
8851 char * const ptr = uiv_2buf(buf, iv, 0, 0, &ebuf);
8853 PERL_ARGS_ASSERT_SV_SETPVIV;
8855 sv_setpvn(sv, ptr, ebuf - ptr);
8859 =for apidoc sv_setpviv_mg
8861 Like C<sv_setpviv>, but also handles 'set' magic.
8867 Perl_sv_setpviv_mg(pTHX_ SV *const sv, const IV iv)
8869 PERL_ARGS_ASSERT_SV_SETPVIV_MG;
8875 #if defined(PERL_IMPLICIT_CONTEXT)
8877 /* pTHX_ magic can't cope with varargs, so this is a no-context
8878 * version of the main function, (which may itself be aliased to us).
8879 * Don't access this version directly.
8883 Perl_sv_setpvf_nocontext(SV *const sv, const char *const pat, ...)
8888 PERL_ARGS_ASSERT_SV_SETPVF_NOCONTEXT;
8890 va_start(args, pat);
8891 sv_vsetpvf(sv, pat, &args);
8895 /* pTHX_ magic can't cope with varargs, so this is a no-context
8896 * version of the main function, (which may itself be aliased to us).
8897 * Don't access this version directly.
8901 Perl_sv_setpvf_mg_nocontext(SV *const sv, const char *const pat, ...)
8906 PERL_ARGS_ASSERT_SV_SETPVF_MG_NOCONTEXT;
8908 va_start(args, pat);
8909 sv_vsetpvf_mg(sv, pat, &args);
8915 =for apidoc sv_setpvf
8917 Works like C<sv_catpvf> but copies the text into the SV instead of
8918 appending it. Does not handle 'set' magic. See C<sv_setpvf_mg>.
8924 Perl_sv_setpvf(pTHX_ SV *const sv, const char *const pat, ...)
8928 PERL_ARGS_ASSERT_SV_SETPVF;
8930 va_start(args, pat);
8931 sv_vsetpvf(sv, pat, &args);
8936 =for apidoc sv_vsetpvf
8938 Works like C<sv_vcatpvf> but copies the text into the SV instead of
8939 appending it. Does not handle 'set' magic. See C<sv_vsetpvf_mg>.
8941 Usually used via its frontend C<sv_setpvf>.
8947 Perl_sv_vsetpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
8949 PERL_ARGS_ASSERT_SV_VSETPVF;
8951 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
8955 =for apidoc sv_setpvf_mg
8957 Like C<sv_setpvf>, but also handles 'set' magic.
8963 Perl_sv_setpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
8967 PERL_ARGS_ASSERT_SV_SETPVF_MG;
8969 va_start(args, pat);
8970 sv_vsetpvf_mg(sv, pat, &args);
8975 =for apidoc sv_vsetpvf_mg
8977 Like C<sv_vsetpvf>, but also handles 'set' magic.
8979 Usually used via its frontend C<sv_setpvf_mg>.
8985 Perl_sv_vsetpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
8987 PERL_ARGS_ASSERT_SV_VSETPVF_MG;
8989 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
8993 #if defined(PERL_IMPLICIT_CONTEXT)
8995 /* pTHX_ magic can't cope with varargs, so this is a no-context
8996 * version of the main function, (which may itself be aliased to us).
8997 * Don't access this version directly.
9001 Perl_sv_catpvf_nocontext(SV *const sv, const char *const pat, ...)
9006 PERL_ARGS_ASSERT_SV_CATPVF_NOCONTEXT;
9008 va_start(args, pat);
9009 sv_vcatpvf(sv, pat, &args);
9013 /* pTHX_ magic can't cope with varargs, so this is a no-context
9014 * version of the main function, (which may itself be aliased to us).
9015 * Don't access this version directly.
9019 Perl_sv_catpvf_mg_nocontext(SV *const sv, const char *const pat, ...)
9024 PERL_ARGS_ASSERT_SV_CATPVF_MG_NOCONTEXT;
9026 va_start(args, pat);
9027 sv_vcatpvf_mg(sv, pat, &args);
9033 =for apidoc sv_catpvf
9035 Processes its arguments like C<sprintf> and appends the formatted
9036 output to an SV. If the appended data contains "wide" characters
9037 (including, but not limited to, SVs with a UTF-8 PV formatted with %s,
9038 and characters >255 formatted with %c), the original SV might get
9039 upgraded to UTF-8. Handles 'get' magic, but not 'set' magic. See
9040 C<sv_catpvf_mg>. If the original SV was UTF-8, the pattern should be
9041 valid UTF-8; if the original SV was bytes, the pattern should be too.
9046 Perl_sv_catpvf(pTHX_ SV *const sv, const char *const pat, ...)
9050 PERL_ARGS_ASSERT_SV_CATPVF;
9052 va_start(args, pat);
9053 sv_vcatpvf(sv, pat, &args);
9058 =for apidoc sv_vcatpvf
9060 Processes its arguments like C<vsprintf> and appends the formatted output
9061 to an SV. Does not handle 'set' magic. See C<sv_vcatpvf_mg>.
9063 Usually used via its frontend C<sv_catpvf>.
9069 Perl_sv_vcatpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9071 PERL_ARGS_ASSERT_SV_VCATPVF;
9073 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9077 =for apidoc sv_catpvf_mg
9079 Like C<sv_catpvf>, but also handles 'set' magic.
9085 Perl_sv_catpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
9089 PERL_ARGS_ASSERT_SV_CATPVF_MG;
9091 va_start(args, pat);
9092 sv_vcatpvf_mg(sv, pat, &args);
9097 =for apidoc sv_vcatpvf_mg
9099 Like C<sv_vcatpvf>, but also handles 'set' magic.
9101 Usually used via its frontend C<sv_catpvf_mg>.
9107 Perl_sv_vcatpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9109 PERL_ARGS_ASSERT_SV_VCATPVF_MG;
9111 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9116 =for apidoc sv_vsetpvfn
9118 Works like C<sv_vcatpvfn> but copies the text into the SV instead of
9121 Usually used via one of its frontends C<sv_vsetpvf> and C<sv_vsetpvf_mg>.
9127 Perl_sv_vsetpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
9128 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
9130 PERL_ARGS_ASSERT_SV_VSETPVFN;
9133 sv_vcatpvfn(sv, pat, patlen, args, svargs, svmax, maybe_tainted);
9137 S_expect_number(pTHX_ char **const pattern)
9142 PERL_ARGS_ASSERT_EXPECT_NUMBER;
9144 switch (**pattern) {
9145 case '1': case '2': case '3':
9146 case '4': case '5': case '6':
9147 case '7': case '8': case '9':
9148 var = *(*pattern)++ - '0';
9149 while (isDIGIT(**pattern)) {
9150 const I32 tmp = var * 10 + (*(*pattern)++ - '0');
9152 Perl_croak(aTHX_ "Integer overflow in format string for %s", (PL_op ? OP_NAME(PL_op) : "sv_vcatpvfn"));
9160 S_F0convert(NV nv, char *const endbuf, STRLEN *const len)
9162 const int neg = nv < 0;
9165 PERL_ARGS_ASSERT_F0CONVERT;
9173 if (uv & 1 && uv == nv)
9174 uv--; /* Round to even */
9176 const unsigned dig = uv % 10;
9189 =for apidoc sv_vcatpvfn
9191 Processes its arguments like C<vsprintf> and appends the formatted output
9192 to an SV. Uses an array of SVs if the C style variable argument list is
9193 missing (NULL). When running with taint checks enabled, indicates via
9194 C<maybe_tainted> if results are untrustworthy (often due to the use of
9197 Usually used via one of its frontends C<sv_vcatpvf> and C<sv_vcatpvf_mg>.
9203 #define VECTORIZE_ARGS vecsv = va_arg(*args, SV*);\
9204 vecstr = (U8*)SvPV_const(vecsv,veclen);\
9205 vec_utf8 = DO_UTF8(vecsv);
9207 /* XXX maybe_tainted is never assigned to, so the doc above is lying. */
9210 Perl_sv_vcatpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
9211 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
9219 static const char nullstr[] = "(null)";
9221 bool has_utf8 = DO_UTF8(sv); /* has the result utf8? */
9222 const bool pat_utf8 = has_utf8; /* the pattern is in utf8? */
9224 /* Times 4: a decimal digit takes more than 3 binary digits.
9225 * NV_DIG: mantissa takes than many decimal digits.
9226 * Plus 32: Playing safe. */
9227 char ebuf[IV_DIG * 4 + NV_DIG + 32];
9228 /* large enough for "%#.#f" --chip */
9229 /* what about long double NVs? --jhi */
9231 PERL_ARGS_ASSERT_SV_VCATPVFN;
9232 PERL_UNUSED_ARG(maybe_tainted);
9234 /* no matter what, this is a string now */
9235 (void)SvPV_force(sv, origlen);
9237 /* special-case "", "%s", and "%-p" (SVf - see below) */
9240 if (patlen == 2 && pat[0] == '%' && pat[1] == 's') {
9242 const char * const s = va_arg(*args, char*);
9243 sv_catpv(sv, s ? s : nullstr);
9245 else if (svix < svmax) {
9246 sv_catsv(sv, *svargs);
9250 if (args && patlen == 3 && pat[0] == '%' &&
9251 pat[1] == '-' && pat[2] == 'p') {
9252 argsv = MUTABLE_SV(va_arg(*args, void*));
9253 sv_catsv(sv, argsv);
9257 #ifndef USE_LONG_DOUBLE
9258 /* special-case "%.<number>[gf]" */
9259 if ( !args && patlen <= 5 && pat[0] == '%' && pat[1] == '.'
9260 && (pat[patlen-1] == 'g' || pat[patlen-1] == 'f') ) {
9261 unsigned digits = 0;
9265 while (*pp >= '0' && *pp <= '9')
9266 digits = 10 * digits + (*pp++ - '0');
9267 if (pp - pat == (int)patlen - 1) {
9275 /* Add check for digits != 0 because it seems that some
9276 gconverts are buggy in this case, and we don't yet have
9277 a Configure test for this. */
9278 if (digits && digits < sizeof(ebuf) - NV_DIG - 10) {
9279 /* 0, point, slack */
9280 Gconvert(nv, (int)digits, 0, ebuf);
9282 if (*ebuf) /* May return an empty string for digits==0 */
9285 } else if (!digits) {
9288 if ((p = F0convert(nv, ebuf + sizeof ebuf, &l))) {
9289 sv_catpvn(sv, p, l);
9295 #endif /* !USE_LONG_DOUBLE */
9297 if (!args && svix < svmax && DO_UTF8(*svargs))
9300 patend = (char*)pat + patlen;
9301 for (p = (char*)pat; p < patend; p = q) {
9304 bool vectorize = FALSE;
9305 bool vectorarg = FALSE;
9306 bool vec_utf8 = FALSE;
9312 bool has_precis = FALSE;
9314 const I32 osvix = svix;
9315 bool is_utf8 = FALSE; /* is this item utf8? */
9316 #ifdef HAS_LDBL_SPRINTF_BUG
9317 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
9318 with sfio - Allen <allens@cpan.org> */
9319 bool fix_ldbl_sprintf_bug = FALSE;
9323 U8 utf8buf[UTF8_MAXBYTES+1];
9324 STRLEN esignlen = 0;
9326 const char *eptr = NULL;
9327 const char *fmtstart;
9330 const U8 *vecstr = NULL;
9337 /* we need a long double target in case HAS_LONG_DOUBLE but
9340 #if defined(HAS_LONG_DOUBLE) && LONG_DOUBLESIZE > DOUBLESIZE
9348 const char *dotstr = ".";
9349 STRLEN dotstrlen = 1;
9350 I32 efix = 0; /* explicit format parameter index */
9351 I32 ewix = 0; /* explicit width index */
9352 I32 epix = 0; /* explicit precision index */
9353 I32 evix = 0; /* explicit vector index */
9354 bool asterisk = FALSE;
9356 /* echo everything up to the next format specification */
9357 for (q = p; q < patend && *q != '%'; ++q) ;
9359 if (has_utf8 && !pat_utf8)
9360 sv_catpvn_utf8_upgrade(sv, p, q - p, nsv);
9362 sv_catpvn(sv, p, q - p);
9371 We allow format specification elements in this order:
9372 \d+\$ explicit format parameter index
9374 v|\*(\d+\$)?v vector with optional (optionally specified) arg
9375 0 flag (as above): repeated to allow "v02"
9376 \d+|\*(\d+\$)? width using optional (optionally specified) arg
9377 \.(\d*|\*(\d+\$)?) precision using optional (optionally specified) arg
9379 [%bcdefginopsuxDFOUX] format (mandatory)
9384 As of perl5.9.3, printf format checking is on by default.
9385 Internally, perl uses %p formats to provide an escape to
9386 some extended formatting. This block deals with those
9387 extensions: if it does not match, (char*)q is reset and
9388 the normal format processing code is used.
9390 Currently defined extensions are:
9391 %p include pointer address (standard)
9392 %-p (SVf) include an SV (previously %_)
9393 %-<num>p include an SV with precision <num>
9394 %<num>p reserved for future extensions
9396 Robin Barker 2005-07-14
9398 %1p (VDf) removed. RMB 2007-10-19
9405 n = expect_number(&q);
9412 argsv = MUTABLE_SV(va_arg(*args, void*));
9413 eptr = SvPV_const(argsv, elen);
9419 if (ckWARN_d(WARN_INTERNAL))
9420 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
9421 "internal %%<num>p might conflict with future printf extensions");
9427 if ( (width = expect_number(&q)) ) {
9442 if (plus == '+' && *q == ' ') /* '+' over ' ' */
9471 if ( (ewix = expect_number(&q)) )
9480 if ((vectorarg = asterisk)) {
9493 width = expect_number(&q);
9499 vecsv = va_arg(*args, SV*);
9501 vecsv = (evix > 0 && evix <= svmax)
9502 ? svargs[evix-1] : &PL_sv_undef;
9504 vecsv = svix < svmax ? svargs[svix++] : &PL_sv_undef;
9506 dotstr = SvPV_const(vecsv, dotstrlen);
9507 /* Keep the DO_UTF8 test *after* the SvPV call, else things go
9508 bad with tied or overloaded values that return UTF8. */
9511 else if (has_utf8) {
9512 vecsv = sv_mortalcopy(vecsv);
9513 sv_utf8_upgrade(vecsv);
9514 dotstr = SvPV_const(vecsv, dotstrlen);
9521 else if (efix ? (efix > 0 && efix <= svmax) : svix < svmax) {
9522 vecsv = svargs[efix ? efix-1 : svix++];
9523 vecstr = (U8*)SvPV_const(vecsv,veclen);
9524 vec_utf8 = DO_UTF8(vecsv);
9526 /* if this is a version object, we need to convert
9527 * back into v-string notation and then let the
9528 * vectorize happen normally
9530 if (sv_derived_from(vecsv, "version")) {
9531 char *version = savesvpv(vecsv);
9532 if ( hv_exists(MUTABLE_HV(SvRV(vecsv)), "alpha", 5 ) ) {
9533 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
9534 "vector argument not supported with alpha versions");
9537 vecsv = sv_newmortal();
9538 scan_vstring(version, version + veclen, vecsv);
9539 vecstr = (U8*)SvPV_const(vecsv, veclen);
9540 vec_utf8 = DO_UTF8(vecsv);
9552 i = va_arg(*args, int);
9554 i = (ewix ? ewix <= svmax : svix < svmax) ?
9555 SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
9557 width = (i < 0) ? -i : i;
9567 if ( ((epix = expect_number(&q))) && (*q++ != '$') )
9569 /* XXX: todo, support specified precision parameter */
9573 i = va_arg(*args, int);
9575 i = (ewix ? ewix <= svmax : svix < svmax)
9576 ? SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
9578 has_precis = !(i < 0);
9583 precis = precis * 10 + (*q++ - '0');
9592 case 'I': /* Ix, I32x, and I64x */
9594 if (q[1] == '6' && q[2] == '4') {
9600 if (q[1] == '3' && q[2] == '2') {
9610 #if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
9621 #if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
9622 if (*(q + 1) == 'l') { /* lld, llf */
9648 if (!vectorize && !args) {
9650 const I32 i = efix-1;
9651 argsv = (i >= 0 && i < svmax) ? svargs[i] : &PL_sv_undef;
9653 argsv = (svix >= 0 && svix < svmax)
9654 ? svargs[svix++] : &PL_sv_undef;
9665 uv = (args) ? va_arg(*args, int) : SvIV(argsv);
9667 (!UNI_IS_INVARIANT(uv) && SvUTF8(sv)))
9669 eptr = (char*)utf8buf;
9670 elen = uvchr_to_utf8((U8*)eptr, uv) - utf8buf;
9684 eptr = va_arg(*args, char*);
9686 elen = strlen(eptr);
9688 eptr = (char *)nullstr;
9689 elen = sizeof nullstr - 1;
9693 eptr = SvPV_const(argsv, elen);
9694 if (DO_UTF8(argsv)) {
9695 STRLEN old_precis = precis;
9696 if (has_precis && precis < elen) {
9697 STRLEN ulen = sv_len_utf8(argsv);
9698 I32 p = precis > ulen ? ulen : precis;
9699 sv_pos_u2b(argsv, &p, 0); /* sticks at end */
9702 if (width) { /* fudge width (can't fudge elen) */
9703 if (has_precis && precis < elen)
9704 width += precis - old_precis;
9706 width += elen - sv_len_utf8(argsv);
9713 if (has_precis && precis < elen)
9720 if (alt || vectorize)
9722 uv = PTR2UV(args ? va_arg(*args, void*) : argsv);
9743 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
9752 esignbuf[esignlen++] = plus;
9756 case 'h': iv = (short)va_arg(*args, int); break;
9757 case 'l': iv = va_arg(*args, long); break;
9758 case 'V': iv = va_arg(*args, IV); break;
9759 default: iv = va_arg(*args, int); break;
9762 iv = va_arg(*args, Quad_t); break;
9769 IV tiv = SvIV(argsv); /* work around GCC bug #13488 */
9771 case 'h': iv = (short)tiv; break;
9772 case 'l': iv = (long)tiv; break;
9774 default: iv = tiv; break;
9777 iv = (Quad_t)tiv; break;
9783 if ( !vectorize ) /* we already set uv above */
9788 esignbuf[esignlen++] = plus;
9792 esignbuf[esignlen++] = '-';
9836 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
9847 case 'h': uv = (unsigned short)va_arg(*args, unsigned); break;
9848 case 'l': uv = va_arg(*args, unsigned long); break;
9849 case 'V': uv = va_arg(*args, UV); break;
9850 default: uv = va_arg(*args, unsigned); break;
9853 uv = va_arg(*args, Uquad_t); break;
9860 UV tuv = SvUV(argsv); /* work around GCC bug #13488 */
9862 case 'h': uv = (unsigned short)tuv; break;
9863 case 'l': uv = (unsigned long)tuv; break;
9865 default: uv = tuv; break;
9868 uv = (Uquad_t)tuv; break;
9877 char *ptr = ebuf + sizeof ebuf;
9878 bool tempalt = uv ? alt : FALSE; /* Vectors can't change alt */
9884 p = (char *)((c == 'X') ? PL_hexdigit + 16 : PL_hexdigit);
9890 esignbuf[esignlen++] = '0';
9891 esignbuf[esignlen++] = c; /* 'x' or 'X' */
9899 if (alt && *ptr != '0')
9908 esignbuf[esignlen++] = '0';
9909 esignbuf[esignlen++] = c;
9912 default: /* it had better be ten or less */
9916 } while (uv /= base);
9919 elen = (ebuf + sizeof ebuf) - ptr;
9923 zeros = precis - elen;
9924 else if (precis == 0 && elen == 1 && *eptr == '0'
9925 && !(base == 8 && alt)) /* "%#.0o" prints "0" */
9928 /* a precision nullifies the 0 flag. */
9935 /* FLOATING POINT */
9938 c = 'f'; /* maybe %F isn't supported here */
9946 /* This is evil, but floating point is even more evil */
9948 /* for SV-style calling, we can only get NV
9949 for C-style calling, we assume %f is double;
9950 for simplicity we allow any of %Lf, %llf, %qf for long double
9954 #if defined(USE_LONG_DOUBLE)
9958 /* [perl #20339] - we should accept and ignore %lf rather than die */
9962 #if defined(USE_LONG_DOUBLE)
9963 intsize = args ? 0 : 'q';
9967 #if defined(HAS_LONG_DOUBLE)
9976 /* now we need (long double) if intsize == 'q', else (double) */
9978 #if LONG_DOUBLESIZE > DOUBLESIZE
9980 va_arg(*args, long double) :
9981 va_arg(*args, double)
9983 va_arg(*args, double)
9988 /* nv * 0 will be NaN for NaN, +Inf and -Inf, and 0 for anything
9989 else. frexp() has some unspecified behaviour for those three */
9990 if (c != 'e' && c != 'E' && (nv * 0) == 0) {
9992 /* FIXME: if HAS_LONG_DOUBLE but not USE_LONG_DOUBLE this
9993 will cast our (long double) to (double) */
9994 (void)Perl_frexp(nv, &i);
9995 if (i == PERL_INT_MIN)
9996 Perl_die(aTHX_ "panic: frexp");
9998 need = BIT_DIGITS(i);
10000 need += has_precis ? precis : 6; /* known default */
10005 #ifdef HAS_LDBL_SPRINTF_BUG
10006 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
10007 with sfio - Allen <allens@cpan.org> */
10010 # define MY_DBL_MAX DBL_MAX
10011 # else /* XXX guessing! HUGE_VAL may be defined as infinity, so not using */
10012 # if DOUBLESIZE >= 8
10013 # define MY_DBL_MAX 1.7976931348623157E+308L
10015 # define MY_DBL_MAX 3.40282347E+38L
10019 # ifdef HAS_LDBL_SPRINTF_BUG_LESS1 /* only between -1L & 1L - Allen */
10020 # define MY_DBL_MAX_BUG 1L
10022 # define MY_DBL_MAX_BUG MY_DBL_MAX
10026 # define MY_DBL_MIN DBL_MIN
10027 # else /* XXX guessing! -Allen */
10028 # if DOUBLESIZE >= 8
10029 # define MY_DBL_MIN 2.2250738585072014E-308L
10031 # define MY_DBL_MIN 1.17549435E-38L
10035 if ((intsize == 'q') && (c == 'f') &&
10036 ((nv < MY_DBL_MAX_BUG) && (nv > -MY_DBL_MAX_BUG)) &&
10037 (need < DBL_DIG)) {
10038 /* it's going to be short enough that
10039 * long double precision is not needed */
10041 if ((nv <= 0L) && (nv >= -0L))
10042 fix_ldbl_sprintf_bug = TRUE; /* 0 is 0 - easiest */
10044 /* would use Perl_fp_class as a double-check but not
10045 * functional on IRIX - see perl.h comments */
10047 if ((nv >= MY_DBL_MIN) || (nv <= -MY_DBL_MIN)) {
10048 /* It's within the range that a double can represent */
10049 #if defined(DBL_MAX) && !defined(DBL_MIN)
10050 if ((nv >= ((long double)1/DBL_MAX)) ||
10051 (nv <= (-(long double)1/DBL_MAX)))
10053 fix_ldbl_sprintf_bug = TRUE;
10056 if (fix_ldbl_sprintf_bug == TRUE) {
10066 # undef MY_DBL_MAX_BUG
10069 #endif /* HAS_LDBL_SPRINTF_BUG */
10071 need += 20; /* fudge factor */
10072 if (PL_efloatsize < need) {
10073 Safefree(PL_efloatbuf);
10074 PL_efloatsize = need + 20; /* more fudge */
10075 Newx(PL_efloatbuf, PL_efloatsize, char);
10076 PL_efloatbuf[0] = '\0';
10079 if ( !(width || left || plus || alt) && fill != '0'
10080 && has_precis && intsize != 'q' ) { /* Shortcuts */
10081 /* See earlier comment about buggy Gconvert when digits,
10083 if ( c == 'g' && precis) {
10084 Gconvert((NV)nv, (int)precis, 0, PL_efloatbuf);
10085 /* May return an empty string for digits==0 */
10086 if (*PL_efloatbuf) {
10087 elen = strlen(PL_efloatbuf);
10088 goto float_converted;
10090 } else if ( c == 'f' && !precis) {
10091 if ((eptr = F0convert(nv, ebuf + sizeof ebuf, &elen)))
10096 char *ptr = ebuf + sizeof ebuf;
10099 /* FIXME: what to do if HAS_LONG_DOUBLE but not PERL_PRIfldbl? */
10100 #if defined(HAS_LONG_DOUBLE) && defined(PERL_PRIfldbl)
10101 if (intsize == 'q') {
10102 /* Copy the one or more characters in a long double
10103 * format before the 'base' ([efgEFG]) character to
10104 * the format string. */
10105 static char const prifldbl[] = PERL_PRIfldbl;
10106 char const *p = prifldbl + sizeof(prifldbl) - 3;
10107 while (p >= prifldbl) { *--ptr = *p--; }
10112 do { *--ptr = '0' + (base % 10); } while (base /= 10);
10117 do { *--ptr = '0' + (base % 10); } while (base /= 10);
10129 /* No taint. Otherwise we are in the strange situation
10130 * where printf() taints but print($float) doesn't.
10132 #if defined(HAS_LONG_DOUBLE)
10133 elen = ((intsize == 'q')
10134 ? my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, nv)
10135 : my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, (double)nv));
10137 elen = my_sprintf(PL_efloatbuf, ptr, nv);
10141 eptr = PL_efloatbuf;
10149 i = SvCUR(sv) - origlen;
10152 case 'h': *(va_arg(*args, short*)) = i; break;
10153 default: *(va_arg(*args, int*)) = i; break;
10154 case 'l': *(va_arg(*args, long*)) = i; break;
10155 case 'V': *(va_arg(*args, IV*)) = i; break;
10158 *(va_arg(*args, Quad_t*)) = i; break;
10165 sv_setuv_mg(argsv, (UV)i);
10166 continue; /* not "break" */
10173 && (PL_op->op_type == OP_PRTF || PL_op->op_type == OP_SPRINTF)
10174 && ckWARN(WARN_PRINTF))
10176 SV * const msg = sv_newmortal();
10177 Perl_sv_setpvf(aTHX_ msg, "Invalid conversion in %sprintf: ",
10178 (PL_op->op_type == OP_PRTF) ? "" : "s");
10179 if (fmtstart < patend) {
10180 const char * const fmtend = q < patend ? q : patend;
10182 sv_catpvs(msg, "\"%");
10183 for (f = fmtstart; f < fmtend; f++) {
10185 sv_catpvn(msg, f, 1);
10187 Perl_sv_catpvf(aTHX_ msg,
10188 "\\%03"UVof, (UV)*f & 0xFF);
10191 sv_catpvs(msg, "\"");
10193 sv_catpvs(msg, "end of string");
10195 Perl_warner(aTHX_ packWARN(WARN_PRINTF), "%"SVf, SVfARG(msg)); /* yes, this is reentrant */
10198 /* output mangled stuff ... */
10204 /* ... right here, because formatting flags should not apply */
10205 SvGROW(sv, SvCUR(sv) + elen + 1);
10207 Copy(eptr, p, elen, char);
10210 SvCUR_set(sv, p - SvPVX_const(sv));
10212 continue; /* not "break" */
10215 if (is_utf8 != has_utf8) {
10218 sv_utf8_upgrade(sv);
10221 const STRLEN old_elen = elen;
10222 SV * const nsv = newSVpvn_flags(eptr, elen, SVs_TEMP);
10223 sv_utf8_upgrade(nsv);
10224 eptr = SvPVX_const(nsv);
10227 if (width) { /* fudge width (can't fudge elen) */
10228 width += elen - old_elen;
10234 have = esignlen + zeros + elen;
10236 Perl_croak_nocontext("%s", PL_memory_wrap);
10238 need = (have > width ? have : width);
10241 if (need >= (((STRLEN)~0) - SvCUR(sv) - dotstrlen - 1))
10242 Perl_croak_nocontext("%s", PL_memory_wrap);
10243 SvGROW(sv, SvCUR(sv) + need + dotstrlen + 1);
10245 if (esignlen && fill == '0') {
10247 for (i = 0; i < (int)esignlen; i++)
10248 *p++ = esignbuf[i];
10250 if (gap && !left) {
10251 memset(p, fill, gap);
10254 if (esignlen && fill != '0') {
10256 for (i = 0; i < (int)esignlen; i++)
10257 *p++ = esignbuf[i];
10261 for (i = zeros; i; i--)
10265 Copy(eptr, p, elen, char);
10269 memset(p, ' ', gap);
10274 Copy(dotstr, p, dotstrlen, char);
10278 vectorize = FALSE; /* done iterating over vecstr */
10285 SvCUR_set(sv, p - SvPVX_const(sv));
10293 /* =========================================================================
10295 =head1 Cloning an interpreter
10297 All the macros and functions in this section are for the private use of
10298 the main function, perl_clone().
10300 The foo_dup() functions make an exact copy of an existing foo thingy.
10301 During the course of a cloning, a hash table is used to map old addresses
10302 to new addresses. The table is created and manipulated with the
10303 ptr_table_* functions.
10307 * =========================================================================*/
10310 #if defined(USE_ITHREADS)
10312 /* XXX Remove this so it doesn't have to go thru the macro and return for nothing */
10313 #ifndef GpREFCNT_inc
10314 # define GpREFCNT_inc(gp) ((gp) ? (++(gp)->gp_refcnt, (gp)) : (GP*)NULL)
10318 /* Certain cases in Perl_ss_dup have been merged, by relying on the fact
10319 that currently av_dup, gv_dup and hv_dup are the same as sv_dup.
10320 If this changes, please unmerge ss_dup.
10321 Likewise, sv_dup_inc_multiple() relies on this fact. */
10322 #define sv_dup_inc(s,t) SvREFCNT_inc(sv_dup(s,t))
10323 #define sv_dup_inc_NN(s,t) SvREFCNT_inc_NN(sv_dup(s,t))
10324 #define av_dup(s,t) MUTABLE_AV(sv_dup((const SV *)s,t))
10325 #define av_dup_inc(s,t) MUTABLE_AV(SvREFCNT_inc(sv_dup((const SV *)s,t)))
10326 #define hv_dup(s,t) MUTABLE_HV(sv_dup((const SV *)s,t))
10327 #define hv_dup_inc(s,t) MUTABLE_HV(SvREFCNT_inc(sv_dup((const SV *)s,t)))
10328 #define cv_dup(s,t) MUTABLE_CV(sv_dup((const SV *)s,t))
10329 #define cv_dup_inc(s,t) MUTABLE_CV(SvREFCNT_inc(sv_dup((const SV *)s,t)))
10330 #define io_dup(s,t) MUTABLE_IO(sv_dup((const SV *)s,t))
10331 #define io_dup_inc(s,t) MUTABLE_IO(SvREFCNT_inc(sv_dup((const SV *)s,t)))
10332 #define gv_dup(s,t) MUTABLE_GV(sv_dup((const SV *)s,t))
10333 #define gv_dup_inc(s,t) MUTABLE_GV(SvREFCNT_inc(sv_dup((const SV *)s,t)))
10334 #define SAVEPV(p) ((p) ? savepv(p) : NULL)
10335 #define SAVEPVN(p,n) ((p) ? savepvn(p,n) : NULL)
10337 /* clone a parser */
10340 Perl_parser_dup(pTHX_ const yy_parser *const proto, CLONE_PARAMS *const param)
10344 PERL_ARGS_ASSERT_PARSER_DUP;
10349 /* look for it in the table first */
10350 parser = (yy_parser *)ptr_table_fetch(PL_ptr_table, proto);
10354 /* create anew and remember what it is */
10355 Newxz(parser, 1, yy_parser);
10356 ptr_table_store(PL_ptr_table, proto, parser);
10358 parser->yyerrstatus = 0;
10359 parser->yychar = YYEMPTY; /* Cause a token to be read. */
10361 /* XXX these not yet duped */
10362 parser->old_parser = NULL;
10363 parser->stack = NULL;
10365 parser->stack_size = 0;
10366 /* XXX parser->stack->state = 0; */
10368 /* XXX eventually, just Copy() most of the parser struct ? */
10370 parser->lex_brackets = proto->lex_brackets;
10371 parser->lex_casemods = proto->lex_casemods;
10372 parser->lex_brackstack = savepvn(proto->lex_brackstack,
10373 (proto->lex_brackets < 120 ? 120 : proto->lex_brackets));
10374 parser->lex_casestack = savepvn(proto->lex_casestack,
10375 (proto->lex_casemods < 12 ? 12 : proto->lex_casemods));
10376 parser->lex_defer = proto->lex_defer;
10377 parser->lex_dojoin = proto->lex_dojoin;
10378 parser->lex_expect = proto->lex_expect;
10379 parser->lex_formbrack = proto->lex_formbrack;
10380 parser->lex_inpat = proto->lex_inpat;
10381 parser->lex_inwhat = proto->lex_inwhat;
10382 parser->lex_op = proto->lex_op;
10383 parser->lex_repl = sv_dup_inc(proto->lex_repl, param);
10384 parser->lex_starts = proto->lex_starts;
10385 parser->lex_stuff = sv_dup_inc(proto->lex_stuff, param);
10386 parser->multi_close = proto->multi_close;
10387 parser->multi_open = proto->multi_open;
10388 parser->multi_start = proto->multi_start;
10389 parser->multi_end = proto->multi_end;
10390 parser->pending_ident = proto->pending_ident;
10391 parser->preambled = proto->preambled;
10392 parser->sublex_info = proto->sublex_info; /* XXX not quite right */
10393 parser->linestr = sv_dup_inc(proto->linestr, param);
10394 parser->expect = proto->expect;
10395 parser->copline = proto->copline;
10396 parser->last_lop_op = proto->last_lop_op;
10397 parser->lex_state = proto->lex_state;
10398 parser->rsfp = fp_dup(proto->rsfp, '<', param);
10399 /* rsfp_filters entries have fake IoDIRP() */
10400 parser->rsfp_filters= av_dup_inc(proto->rsfp_filters, param);
10401 parser->in_my = proto->in_my;
10402 parser->in_my_stash = hv_dup(proto->in_my_stash, param);
10403 parser->error_count = proto->error_count;
10406 parser->linestr = sv_dup_inc(proto->linestr, param);
10409 char * const ols = SvPVX(proto->linestr);
10410 char * const ls = SvPVX(parser->linestr);
10412 parser->bufptr = ls + (proto->bufptr >= ols ?
10413 proto->bufptr - ols : 0);
10414 parser->oldbufptr = ls + (proto->oldbufptr >= ols ?
10415 proto->oldbufptr - ols : 0);
10416 parser->oldoldbufptr= ls + (proto->oldoldbufptr >= ols ?
10417 proto->oldoldbufptr - ols : 0);
10418 parser->linestart = ls + (proto->linestart >= ols ?
10419 proto->linestart - ols : 0);
10420 parser->last_uni = ls + (proto->last_uni >= ols ?
10421 proto->last_uni - ols : 0);
10422 parser->last_lop = ls + (proto->last_lop >= ols ?
10423 proto->last_lop - ols : 0);
10425 parser->bufend = ls + SvCUR(parser->linestr);
10428 Copy(proto->tokenbuf, parser->tokenbuf, 256, char);
10432 parser->endwhite = proto->endwhite;
10433 parser->faketokens = proto->faketokens;
10434 parser->lasttoke = proto->lasttoke;
10435 parser->nextwhite = proto->nextwhite;
10436 parser->realtokenstart = proto->realtokenstart;
10437 parser->skipwhite = proto->skipwhite;
10438 parser->thisclose = proto->thisclose;
10439 parser->thismad = proto->thismad;
10440 parser->thisopen = proto->thisopen;
10441 parser->thisstuff = proto->thisstuff;
10442 parser->thistoken = proto->thistoken;
10443 parser->thiswhite = proto->thiswhite;
10445 Copy(proto->nexttoke, parser->nexttoke, 5, NEXTTOKE);
10446 parser->curforce = proto->curforce;
10448 Copy(proto->nextval, parser->nextval, 5, YYSTYPE);
10449 Copy(proto->nexttype, parser->nexttype, 5, I32);
10450 parser->nexttoke = proto->nexttoke;
10453 /* XXX should clone saved_curcop here, but we aren't passed
10454 * proto_perl; so do it in perl_clone_using instead */
10460 /* duplicate a file handle */
10463 Perl_fp_dup(pTHX_ PerlIO *const fp, const char type, CLONE_PARAMS *const param)
10467 PERL_ARGS_ASSERT_FP_DUP;
10468 PERL_UNUSED_ARG(type);
10471 return (PerlIO*)NULL;
10473 /* look for it in the table first */
10474 ret = (PerlIO*)ptr_table_fetch(PL_ptr_table, fp);
10478 /* create anew and remember what it is */
10479 ret = PerlIO_fdupopen(aTHX_ fp, param, PERLIO_DUP_CLONE);
10480 ptr_table_store(PL_ptr_table, fp, ret);
10484 /* duplicate a directory handle */
10487 Perl_dirp_dup(pTHX_ DIR *const dp)
10489 PERL_UNUSED_CONTEXT;
10496 /* duplicate a typeglob */
10499 Perl_gp_dup(pTHX_ GP *const gp, CLONE_PARAMS *const param)
10503 PERL_ARGS_ASSERT_GP_DUP;
10507 /* look for it in the table first */
10508 ret = (GP*)ptr_table_fetch(PL_ptr_table, gp);
10512 /* create anew and remember what it is */
10514 ptr_table_store(PL_ptr_table, gp, ret);
10517 /* ret->gp_refcnt must be 0 before any other dups are called. We're relying
10518 on Newxz() to do this for us. */
10519 ret->gp_sv = sv_dup_inc(gp->gp_sv, param);
10520 ret->gp_io = io_dup_inc(gp->gp_io, param);
10521 ret->gp_form = cv_dup_inc(gp->gp_form, param);
10522 ret->gp_av = av_dup_inc(gp->gp_av, param);
10523 ret->gp_hv = hv_dup_inc(gp->gp_hv, param);
10524 ret->gp_egv = gv_dup(gp->gp_egv, param);/* GvEGV is not refcounted */
10525 ret->gp_cv = cv_dup_inc(gp->gp_cv, param);
10526 ret->gp_cvgen = gp->gp_cvgen;
10527 ret->gp_line = gp->gp_line;
10528 ret->gp_file_hek = hek_dup(gp->gp_file_hek, param);
10532 /* duplicate a chain of magic */
10535 Perl_mg_dup(pTHX_ MAGIC *mg, CLONE_PARAMS *const param)
10537 MAGIC *mgret = NULL;
10538 MAGIC **mgprev_p = &mgret;
10540 PERL_ARGS_ASSERT_MG_DUP;
10542 for (; mg; mg = mg->mg_moremagic) {
10544 Newx(nmg, 1, MAGIC);
10546 mgprev_p = &(nmg->mg_moremagic);
10548 /* There was a comment "XXX copy dynamic vtable?" but as we don't have
10549 dynamic vtables, I'm not sure why Sarathy wrote it. The comment dates
10550 from the original commit adding Perl_mg_dup() - revision 4538.
10551 Similarly there is the annotation "XXX random ptr?" next to the
10552 assignment to nmg->mg_ptr. */
10555 /* FIXME for plugins
10556 if (nmg->mg_type == PERL_MAGIC_qr) {
10557 nmg->mg_obj = MUTABLE_SV(CALLREGDUPE((REGEXP*)nmg->mg_obj, param));
10561 if(nmg->mg_type == PERL_MAGIC_backref) {
10562 /* The backref AV has its reference count deliberately bumped by
10565 = SvREFCNT_inc(av_dup_inc((const AV *) nmg->mg_obj, param));
10568 nmg->mg_obj = (nmg->mg_flags & MGf_REFCOUNTED)
10569 ? sv_dup_inc(nmg->mg_obj, param)
10570 : sv_dup(nmg->mg_obj, param);
10573 if (nmg->mg_ptr && nmg->mg_type != PERL_MAGIC_regex_global) {
10574 if (nmg->mg_len > 0) {
10575 nmg->mg_ptr = SAVEPVN(nmg->mg_ptr, nmg->mg_len);
10576 if (nmg->mg_type == PERL_MAGIC_overload_table &&
10577 AMT_AMAGIC((AMT*)nmg->mg_ptr))
10579 AMT * const namtp = (AMT*)nmg->mg_ptr;
10580 sv_dup_inc_multiple((SV**)(namtp->table),
10581 (SV**)(namtp->table), NofAMmeth, param);
10584 else if (nmg->mg_len == HEf_SVKEY)
10585 nmg->mg_ptr = (char*)sv_dup_inc((const SV *)nmg->mg_ptr, param);
10587 if ((nmg->mg_flags & MGf_DUP) && nmg->mg_virtual && nmg->mg_virtual->svt_dup) {
10588 CALL_FPTR(nmg->mg_virtual->svt_dup)(aTHX_ nmg, param);
10594 #endif /* USE_ITHREADS */
10596 /* create a new pointer-mapping table */
10599 Perl_ptr_table_new(pTHX)
10602 PERL_UNUSED_CONTEXT;
10604 Newx(tbl, 1, PTR_TBL_t);
10605 tbl->tbl_max = 511;
10606 tbl->tbl_items = 0;
10607 Newxz(tbl->tbl_ary, tbl->tbl_max + 1, PTR_TBL_ENT_t*);
10611 #define PTR_TABLE_HASH(ptr) \
10612 ((PTR2UV(ptr) >> 3) ^ (PTR2UV(ptr) >> (3 + 7)) ^ (PTR2UV(ptr) >> (3 + 17)))
10615 we use the PTE_SVSLOT 'reservation' made above, both here (in the
10616 following define) and at call to new_body_inline made below in
10617 Perl_ptr_table_store()
10620 #define del_pte(p) del_body_type(p, PTE_SVSLOT)
10622 /* map an existing pointer using a table */
10624 STATIC PTR_TBL_ENT_t *
10625 S_ptr_table_find(PTR_TBL_t *const tbl, const void *const sv)
10627 PTR_TBL_ENT_t *tblent;
10628 const UV hash = PTR_TABLE_HASH(sv);
10630 PERL_ARGS_ASSERT_PTR_TABLE_FIND;
10632 tblent = tbl->tbl_ary[hash & tbl->tbl_max];
10633 for (; tblent; tblent = tblent->next) {
10634 if (tblent->oldval == sv)
10641 Perl_ptr_table_fetch(pTHX_ PTR_TBL_t *const tbl, const void *const sv)
10643 PTR_TBL_ENT_t const *const tblent = ptr_table_find(tbl, sv);
10645 PERL_ARGS_ASSERT_PTR_TABLE_FETCH;
10646 PERL_UNUSED_CONTEXT;
10648 return tblent ? tblent->newval : NULL;
10651 /* add a new entry to a pointer-mapping table */
10654 Perl_ptr_table_store(pTHX_ PTR_TBL_t *const tbl, const void *const oldsv, void *const newsv)
10656 PTR_TBL_ENT_t *tblent = ptr_table_find(tbl, oldsv);
10658 PERL_ARGS_ASSERT_PTR_TABLE_STORE;
10659 PERL_UNUSED_CONTEXT;
10662 tblent->newval = newsv;
10664 const UV entry = PTR_TABLE_HASH(oldsv) & tbl->tbl_max;
10666 new_body_inline(tblent, PTE_SVSLOT);
10668 tblent->oldval = oldsv;
10669 tblent->newval = newsv;
10670 tblent->next = tbl->tbl_ary[entry];
10671 tbl->tbl_ary[entry] = tblent;
10673 if (tblent->next && tbl->tbl_items > tbl->tbl_max)
10674 ptr_table_split(tbl);
10678 /* double the hash bucket size of an existing ptr table */
10681 Perl_ptr_table_split(pTHX_ PTR_TBL_t *const tbl)
10683 PTR_TBL_ENT_t **ary = tbl->tbl_ary;
10684 const UV oldsize = tbl->tbl_max + 1;
10685 UV newsize = oldsize * 2;
10688 PERL_ARGS_ASSERT_PTR_TABLE_SPLIT;
10689 PERL_UNUSED_CONTEXT;
10691 Renew(ary, newsize, PTR_TBL_ENT_t*);
10692 Zero(&ary[oldsize], newsize-oldsize, PTR_TBL_ENT_t*);
10693 tbl->tbl_max = --newsize;
10694 tbl->tbl_ary = ary;
10695 for (i=0; i < oldsize; i++, ary++) {
10696 PTR_TBL_ENT_t **curentp, **entp, *ent;
10699 curentp = ary + oldsize;
10700 for (entp = ary, ent = *ary; ent; ent = *entp) {
10701 if ((newsize & PTR_TABLE_HASH(ent->oldval)) != i) {
10703 ent->next = *curentp;
10713 /* remove all the entries from a ptr table */
10716 Perl_ptr_table_clear(pTHX_ PTR_TBL_t *const tbl)
10718 if (tbl && tbl->tbl_items) {
10719 register PTR_TBL_ENT_t * const * const array = tbl->tbl_ary;
10720 UV riter = tbl->tbl_max;
10723 PTR_TBL_ENT_t *entry = array[riter];
10726 PTR_TBL_ENT_t * const oentry = entry;
10727 entry = entry->next;
10732 tbl->tbl_items = 0;
10736 /* clear and free a ptr table */
10739 Perl_ptr_table_free(pTHX_ PTR_TBL_t *const tbl)
10744 ptr_table_clear(tbl);
10745 Safefree(tbl->tbl_ary);
10749 #if defined(USE_ITHREADS)
10752 Perl_rvpv_dup(pTHX_ SV *const dstr, const SV *const sstr, CLONE_PARAMS *const param)
10754 PERL_ARGS_ASSERT_RVPV_DUP;
10757 SvRV_set(dstr, SvWEAKREF(sstr)
10758 ? sv_dup(SvRV_const(sstr), param)
10759 : sv_dup_inc(SvRV_const(sstr), param));
10762 else if (SvPVX_const(sstr)) {
10763 /* Has something there */
10765 /* Normal PV - clone whole allocated space */
10766 SvPV_set(dstr, SAVEPVN(SvPVX_const(sstr), SvLEN(sstr)-1));
10767 if (SvREADONLY(sstr) && SvFAKE(sstr)) {
10768 /* Not that normal - actually sstr is copy on write.
10769 But we are a true, independant SV, so: */
10770 SvREADONLY_off(dstr);
10775 /* Special case - not normally malloced for some reason */
10776 if (isGV_with_GP(sstr)) {
10777 /* Don't need to do anything here. */
10779 else if ((SvREADONLY(sstr) && SvFAKE(sstr))) {
10780 /* A "shared" PV - clone it as "shared" PV */
10782 HEK_KEY(hek_dup(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)),
10786 /* Some other special case - random pointer */
10787 SvPV_set(dstr, (char *) SvPVX_const(sstr));
10792 /* Copy the NULL */
10793 SvPV_set(dstr, NULL);
10797 /* duplicate a list of SVs. source and dest may point to the same memory. */
10799 S_sv_dup_inc_multiple(pTHX_ SV *const *source, SV **dest,
10800 SSize_t items, CLONE_PARAMS *const param)
10802 PERL_ARGS_ASSERT_SV_DUP_INC_MULTIPLE;
10804 while (items-- > 0) {
10805 *dest++ = sv_dup_inc(*source++, param);
10811 /* duplicate an SV of any type (including AV, HV etc) */
10814 Perl_sv_dup(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
10819 PERL_ARGS_ASSERT_SV_DUP;
10823 if (SvTYPE(sstr) == SVTYPEMASK) {
10824 #ifdef DEBUG_LEAKING_SCALARS_ABORT
10829 /* look for it in the table first */
10830 dstr = MUTABLE_SV(ptr_table_fetch(PL_ptr_table, sstr));
10834 if(param->flags & CLONEf_JOIN_IN) {
10835 /** We are joining here so we don't want do clone
10836 something that is bad **/
10837 if (SvTYPE(sstr) == SVt_PVHV) {
10838 const HEK * const hvname = HvNAME_HEK(sstr);
10840 /** don't clone stashes if they already exist **/
10841 return MUTABLE_SV(gv_stashpvn(HEK_KEY(hvname), HEK_LEN(hvname), 0));
10845 /* create anew and remember what it is */
10848 #ifdef DEBUG_LEAKING_SCALARS
10849 dstr->sv_debug_optype = sstr->sv_debug_optype;
10850 dstr->sv_debug_line = sstr->sv_debug_line;
10851 dstr->sv_debug_inpad = sstr->sv_debug_inpad;
10852 dstr->sv_debug_cloned = 1;
10853 dstr->sv_debug_file = savepv(sstr->sv_debug_file);
10856 ptr_table_store(PL_ptr_table, sstr, dstr);
10859 SvFLAGS(dstr) = SvFLAGS(sstr);
10860 SvFLAGS(dstr) &= ~SVf_OOK; /* don't propagate OOK hack */
10861 SvREFCNT(dstr) = 0; /* must be before any other dups! */
10864 if (SvANY(sstr) && PL_watch_pvx && SvPVX_const(sstr) == PL_watch_pvx)
10865 PerlIO_printf(Perl_debug_log, "watch at %p hit, found string \"%s\"\n",
10866 (void*)PL_watch_pvx, SvPVX_const(sstr));
10869 /* don't clone objects whose class has asked us not to */
10870 if (SvOBJECT(sstr) && ! (SvFLAGS(SvSTASH(sstr)) & SVphv_CLONEABLE)) {
10875 switch (SvTYPE(sstr)) {
10877 SvANY(dstr) = NULL;
10880 SvANY(dstr) = (XPVIV*)((char*)&(dstr->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
10882 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
10884 SvIV_set(dstr, SvIVX(sstr));
10888 SvANY(dstr) = new_XNV();
10889 SvNV_set(dstr, SvNVX(sstr));
10891 /* case SVt_BIND: */
10894 /* These are all the types that need complex bodies allocating. */
10896 const svtype sv_type = SvTYPE(sstr);
10897 const struct body_details *const sv_type_details
10898 = bodies_by_type + sv_type;
10902 Perl_croak(aTHX_ "Bizarre SvTYPE [%" IVdf "]", (IV)SvTYPE(sstr));
10917 assert(sv_type_details->body_size);
10918 if (sv_type_details->arena) {
10919 new_body_inline(new_body, sv_type);
10921 = (void*)((char*)new_body - sv_type_details->offset);
10923 new_body = new_NOARENA(sv_type_details);
10927 SvANY(dstr) = new_body;
10930 Copy(((char*)SvANY(sstr)) + sv_type_details->offset,
10931 ((char*)SvANY(dstr)) + sv_type_details->offset,
10932 sv_type_details->copy, char);
10934 Copy(((char*)SvANY(sstr)),
10935 ((char*)SvANY(dstr)),
10936 sv_type_details->body_size + sv_type_details->offset, char);
10939 if (sv_type != SVt_PVAV && sv_type != SVt_PVHV
10940 && !isGV_with_GP(dstr))
10941 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
10943 /* The Copy above means that all the source (unduplicated) pointers
10944 are now in the destination. We can check the flags and the
10945 pointers in either, but it's possible that there's less cache
10946 missing by always going for the destination.
10947 FIXME - instrument and check that assumption */
10948 if (sv_type >= SVt_PVMG) {
10949 if ((sv_type == SVt_PVMG) && SvPAD_OUR(dstr)) {
10950 SvOURSTASH_set(dstr, hv_dup_inc(SvOURSTASH(dstr), param));
10951 } else if (SvMAGIC(dstr))
10952 SvMAGIC_set(dstr, mg_dup(SvMAGIC(dstr), param));
10954 SvSTASH_set(dstr, hv_dup_inc(SvSTASH(dstr), param));
10957 /* The cast silences a GCC warning about unhandled types. */
10958 switch ((int)sv_type) {
10968 /* FIXME for plugins */
10969 re_dup_guts((REGEXP*) sstr, (REGEXP*) dstr, param);
10972 /* XXX LvTARGOFF sometimes holds PMOP* when DEBUGGING */
10973 if (LvTYPE(dstr) == 't') /* for tie: unrefcnted fake (SV**) */
10974 LvTARG(dstr) = dstr;
10975 else if (LvTYPE(dstr) == 'T') /* for tie: fake HE */
10976 LvTARG(dstr) = MUTABLE_SV(he_dup((HE*)LvTARG(dstr), 0, param));
10978 LvTARG(dstr) = sv_dup_inc(LvTARG(dstr), param);
10980 if(isGV_with_GP(sstr)) {
10981 GvNAME_HEK(dstr) = hek_dup(GvNAME_HEK(dstr), param);
10982 /* Don't call sv_add_backref here as it's going to be
10983 created as part of the magic cloning of the symbol
10985 /* Danger Will Robinson - GvGP(dstr) isn't initialised
10986 at the point of this comment. */
10987 GvSTASH(dstr) = hv_dup(GvSTASH(dstr), param);
10988 GvGP(dstr) = gp_dup(GvGP(sstr), param);
10989 (void)GpREFCNT_inc(GvGP(dstr));
10991 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
10994 IoIFP(dstr) = fp_dup(IoIFP(dstr), IoTYPE(dstr), param);
10995 if (IoOFP(dstr) == IoIFP(sstr))
10996 IoOFP(dstr) = IoIFP(dstr);
10998 IoOFP(dstr) = fp_dup(IoOFP(dstr), IoTYPE(dstr), param);
10999 /* PL_parser->rsfp_filters entries have fake IoDIRP() */
11000 if(IoFLAGS(dstr) & IOf_FAKE_DIRP) {
11001 /* I have no idea why fake dirp (rsfps)
11002 should be treated differently but otherwise
11003 we end up with leaks -- sky*/
11004 IoTOP_GV(dstr) = gv_dup_inc(IoTOP_GV(dstr), param);
11005 IoFMT_GV(dstr) = gv_dup_inc(IoFMT_GV(dstr), param);
11006 IoBOTTOM_GV(dstr) = gv_dup_inc(IoBOTTOM_GV(dstr), param);
11008 IoTOP_GV(dstr) = gv_dup(IoTOP_GV(dstr), param);
11009 IoFMT_GV(dstr) = gv_dup(IoFMT_GV(dstr), param);
11010 IoBOTTOM_GV(dstr) = gv_dup(IoBOTTOM_GV(dstr), param);
11011 if (IoDIRP(dstr)) {
11012 IoDIRP(dstr) = dirp_dup(IoDIRP(dstr));
11015 /* IoDIRP(dstr) is already a copy of IoDIRP(sstr) */
11018 IoTOP_NAME(dstr) = SAVEPV(IoTOP_NAME(dstr));
11019 IoFMT_NAME(dstr) = SAVEPV(IoFMT_NAME(dstr));
11020 IoBOTTOM_NAME(dstr) = SAVEPV(IoBOTTOM_NAME(dstr));
11023 /* avoid cloning an empty array */
11024 if (AvARRAY((const AV *)sstr) && AvFILLp((const AV *)sstr) >= 0) {
11025 SV **dst_ary, **src_ary;
11026 SSize_t items = AvFILLp((const AV *)sstr) + 1;
11028 src_ary = AvARRAY((const AV *)sstr);
11029 Newxz(dst_ary, AvMAX((const AV *)sstr)+1, SV*);
11030 ptr_table_store(PL_ptr_table, src_ary, dst_ary);
11031 AvARRAY(MUTABLE_AV(dstr)) = dst_ary;
11032 AvALLOC((const AV *)dstr) = dst_ary;
11033 if (AvREAL((const AV *)sstr)) {
11034 dst_ary = sv_dup_inc_multiple(src_ary, dst_ary, items,
11038 while (items-- > 0)
11039 *dst_ary++ = sv_dup(*src_ary++, param);
11041 items = AvMAX((const AV *)sstr) - AvFILLp((const AV *)sstr);
11042 while (items-- > 0) {
11043 *dst_ary++ = &PL_sv_undef;
11047 AvARRAY(MUTABLE_AV(dstr)) = NULL;
11048 AvALLOC((const AV *)dstr) = (SV**)NULL;
11049 AvMAX( (const AV *)dstr) = -1;
11050 AvFILLp((const AV *)dstr) = -1;
11054 if (HvARRAY((const HV *)sstr)) {
11056 const bool sharekeys = !!HvSHAREKEYS(sstr);
11057 XPVHV * const dxhv = (XPVHV*)SvANY(dstr);
11058 XPVHV * const sxhv = (XPVHV*)SvANY(sstr);
11060 Newx(darray, PERL_HV_ARRAY_ALLOC_BYTES(dxhv->xhv_max+1)
11061 + (SvOOK(sstr) ? sizeof(struct xpvhv_aux) : 0),
11063 HvARRAY(dstr) = (HE**)darray;
11064 while (i <= sxhv->xhv_max) {
11065 const HE * const source = HvARRAY(sstr)[i];
11066 HvARRAY(dstr)[i] = source
11067 ? he_dup(source, sharekeys, param) : 0;
11072 const struct xpvhv_aux * const saux = HvAUX(sstr);
11073 struct xpvhv_aux * const daux = HvAUX(dstr);
11074 /* This flag isn't copied. */
11075 /* SvOOK_on(hv) attacks the IV flags. */
11076 SvFLAGS(dstr) |= SVf_OOK;
11078 hvname = saux->xhv_name;
11079 daux->xhv_name = hek_dup(hvname, param);
11081 daux->xhv_riter = saux->xhv_riter;
11082 daux->xhv_eiter = saux->xhv_eiter
11083 ? he_dup(saux->xhv_eiter,
11084 (bool)!!HvSHAREKEYS(sstr), param) : 0;
11085 /* backref array needs refcnt=2; see sv_add_backref */
11086 daux->xhv_backreferences =
11087 saux->xhv_backreferences
11088 ? MUTABLE_AV(SvREFCNT_inc(
11089 sv_dup_inc((const SV *)saux->xhv_backreferences, param)))
11092 daux->xhv_mro_meta = saux->xhv_mro_meta
11093 ? mro_meta_dup(saux->xhv_mro_meta, param)
11096 /* Record stashes for possible cloning in Perl_clone(). */
11098 av_push(param->stashes, dstr);
11102 HvARRAY(MUTABLE_HV(dstr)) = NULL;
11105 if (!(param->flags & CLONEf_COPY_STACKS)) {
11109 /* NOTE: not refcounted */
11110 CvSTASH(dstr) = hv_dup(CvSTASH(dstr), param);
11112 if (!CvISXSUB(dstr))
11113 CvROOT(dstr) = OpREFCNT_inc(CvROOT(dstr));
11115 if (CvCONST(dstr) && CvISXSUB(dstr)) {
11116 CvXSUBANY(dstr).any_ptr =
11117 sv_dup_inc((const SV *)CvXSUBANY(dstr).any_ptr, param);
11119 /* don't dup if copying back - CvGV isn't refcounted, so the
11120 * duped GV may never be freed. A bit of a hack! DAPM */
11121 CvGV(dstr) = (param->flags & CLONEf_JOIN_IN) ?
11122 NULL : gv_dup(CvGV(dstr), param) ;
11123 PAD_DUP(CvPADLIST(dstr), CvPADLIST(sstr), param);
11125 CvWEAKOUTSIDE(sstr)
11126 ? cv_dup( CvOUTSIDE(dstr), param)
11127 : cv_dup_inc(CvOUTSIDE(dstr), param);
11128 if (!CvISXSUB(dstr))
11129 CvFILE(dstr) = SAVEPV(CvFILE(dstr));
11135 if (SvOBJECT(dstr) && SvTYPE(dstr) != SVt_PVIO)
11141 /* duplicate a context */
11144 Perl_cx_dup(pTHX_ PERL_CONTEXT *cxs, I32 ix, I32 max, CLONE_PARAMS* param)
11146 PERL_CONTEXT *ncxs;
11148 PERL_ARGS_ASSERT_CX_DUP;
11151 return (PERL_CONTEXT*)NULL;
11153 /* look for it in the table first */
11154 ncxs = (PERL_CONTEXT*)ptr_table_fetch(PL_ptr_table, cxs);
11158 /* create anew and remember what it is */
11159 Newx(ncxs, max + 1, PERL_CONTEXT);
11160 ptr_table_store(PL_ptr_table, cxs, ncxs);
11161 Copy(cxs, ncxs, max + 1, PERL_CONTEXT);
11164 PERL_CONTEXT * const ncx = &ncxs[ix];
11165 if (CxTYPE(ncx) == CXt_SUBST) {
11166 Perl_croak(aTHX_ "Cloning substitution context is unimplemented");
11169 switch (CxTYPE(ncx)) {
11171 ncx->blk_sub.cv = (ncx->blk_sub.olddepth == 0
11172 ? cv_dup_inc(ncx->blk_sub.cv, param)
11173 : cv_dup(ncx->blk_sub.cv,param));
11174 ncx->blk_sub.argarray = (CxHASARGS(ncx)
11175 ? av_dup_inc(ncx->blk_sub.argarray,
11178 ncx->blk_sub.savearray = av_dup_inc(ncx->blk_sub.savearray,
11180 ncx->blk_sub.oldcomppad = (PAD*)ptr_table_fetch(PL_ptr_table,
11181 ncx->blk_sub.oldcomppad);
11184 ncx->blk_eval.old_namesv = sv_dup_inc(ncx->blk_eval.old_namesv,
11186 ncx->blk_eval.cur_text = sv_dup(ncx->blk_eval.cur_text, param);
11188 case CXt_LOOP_LAZYSV:
11189 ncx->blk_loop.state_u.lazysv.end
11190 = sv_dup_inc(ncx->blk_loop.state_u.lazysv.end, param);
11191 /* We are taking advantage of av_dup_inc and sv_dup_inc
11192 actually being the same function, and order equivalance of
11194 We can assert the later [but only at run time :-(] */
11195 assert ((void *) &ncx->blk_loop.state_u.ary.ary ==
11196 (void *) &ncx->blk_loop.state_u.lazysv.cur);
11198 ncx->blk_loop.state_u.ary.ary
11199 = av_dup_inc(ncx->blk_loop.state_u.ary.ary, param);
11200 case CXt_LOOP_LAZYIV:
11201 case CXt_LOOP_PLAIN:
11202 if (CxPADLOOP(ncx)) {
11203 ncx->blk_loop.oldcomppad
11204 = (PAD*)ptr_table_fetch(PL_ptr_table,
11205 ncx->blk_loop.oldcomppad);
11207 ncx->blk_loop.oldcomppad
11208 = (PAD*)gv_dup((const GV *)ncx->blk_loop.oldcomppad,
11213 ncx->blk_format.cv = cv_dup(ncx->blk_format.cv, param);
11214 ncx->blk_format.gv = gv_dup(ncx->blk_format.gv, param);
11215 ncx->blk_format.dfoutgv = gv_dup_inc(ncx->blk_format.dfoutgv,
11228 /* duplicate a stack info structure */
11231 Perl_si_dup(pTHX_ PERL_SI *si, CLONE_PARAMS* param)
11235 PERL_ARGS_ASSERT_SI_DUP;
11238 return (PERL_SI*)NULL;
11240 /* look for it in the table first */
11241 nsi = (PERL_SI*)ptr_table_fetch(PL_ptr_table, si);
11245 /* create anew and remember what it is */
11246 Newxz(nsi, 1, PERL_SI);
11247 ptr_table_store(PL_ptr_table, si, nsi);
11249 nsi->si_stack = av_dup_inc(si->si_stack, param);
11250 nsi->si_cxix = si->si_cxix;
11251 nsi->si_cxmax = si->si_cxmax;
11252 nsi->si_cxstack = cx_dup(si->si_cxstack, si->si_cxix, si->si_cxmax, param);
11253 nsi->si_type = si->si_type;
11254 nsi->si_prev = si_dup(si->si_prev, param);
11255 nsi->si_next = si_dup(si->si_next, param);
11256 nsi->si_markoff = si->si_markoff;
11261 #define POPINT(ss,ix) ((ss)[--(ix)].any_i32)
11262 #define TOPINT(ss,ix) ((ss)[ix].any_i32)
11263 #define POPLONG(ss,ix) ((ss)[--(ix)].any_long)
11264 #define TOPLONG(ss,ix) ((ss)[ix].any_long)
11265 #define POPIV(ss,ix) ((ss)[--(ix)].any_iv)
11266 #define TOPIV(ss,ix) ((ss)[ix].any_iv)
11267 #define POPBOOL(ss,ix) ((ss)[--(ix)].any_bool)
11268 #define TOPBOOL(ss,ix) ((ss)[ix].any_bool)
11269 #define POPPTR(ss,ix) ((ss)[--(ix)].any_ptr)
11270 #define TOPPTR(ss,ix) ((ss)[ix].any_ptr)
11271 #define POPDPTR(ss,ix) ((ss)[--(ix)].any_dptr)
11272 #define TOPDPTR(ss,ix) ((ss)[ix].any_dptr)
11273 #define POPDXPTR(ss,ix) ((ss)[--(ix)].any_dxptr)
11274 #define TOPDXPTR(ss,ix) ((ss)[ix].any_dxptr)
11277 #define pv_dup_inc(p) SAVEPV(p)
11278 #define pv_dup(p) SAVEPV(p)
11279 #define svp_dup_inc(p,pp) any_dup(p,pp)
11281 /* map any object to the new equivent - either something in the
11282 * ptr table, or something in the interpreter structure
11286 Perl_any_dup(pTHX_ void *v, const PerlInterpreter *proto_perl)
11290 PERL_ARGS_ASSERT_ANY_DUP;
11293 return (void*)NULL;
11295 /* look for it in the table first */
11296 ret = ptr_table_fetch(PL_ptr_table, v);
11300 /* see if it is part of the interpreter structure */
11301 if (v >= (void*)proto_perl && v < (void*)(proto_perl+1))
11302 ret = (void*)(((char*)aTHX) + (((char*)v) - (char*)proto_perl));
11310 /* duplicate the save stack */
11313 Perl_ss_dup(pTHX_ PerlInterpreter *proto_perl, CLONE_PARAMS* param)
11316 ANY * const ss = proto_perl->Isavestack;
11317 const I32 max = proto_perl->Isavestack_max;
11318 I32 ix = proto_perl->Isavestack_ix;
11331 void (*dptr) (void*);
11332 void (*dxptr) (pTHX_ void*);
11334 PERL_ARGS_ASSERT_SS_DUP;
11336 Newxz(nss, max, ANY);
11339 const I32 type = POPINT(ss,ix);
11340 TOPINT(nss,ix) = type;
11342 case SAVEt_HELEM: /* hash element */
11343 sv = (const SV *)POPPTR(ss,ix);
11344 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11346 case SAVEt_ITEM: /* normal string */
11347 case SAVEt_SV: /* scalar reference */
11348 sv = (const SV *)POPPTR(ss,ix);
11349 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11352 case SAVEt_MORTALIZESV:
11353 sv = (const SV *)POPPTR(ss,ix);
11354 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11356 case SAVEt_SHARED_PVREF: /* char* in shared space */
11357 c = (char*)POPPTR(ss,ix);
11358 TOPPTR(nss,ix) = savesharedpv(c);
11359 ptr = POPPTR(ss,ix);
11360 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11362 case SAVEt_GENERIC_SVREF: /* generic sv */
11363 case SAVEt_SVREF: /* scalar reference */
11364 sv = (const SV *)POPPTR(ss,ix);
11365 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11366 ptr = POPPTR(ss,ix);
11367 TOPPTR(nss,ix) = svp_dup_inc((SV**)ptr, proto_perl);/* XXXXX */
11369 case SAVEt_HV: /* hash reference */
11370 case SAVEt_AV: /* array reference */
11371 sv = (const SV *) POPPTR(ss,ix);
11372 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11374 case SAVEt_COMPPAD:
11376 sv = (const SV *) POPPTR(ss,ix);
11377 TOPPTR(nss,ix) = sv_dup(sv, param);
11379 case SAVEt_INT: /* int reference */
11380 ptr = POPPTR(ss,ix);
11381 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11382 intval = (int)POPINT(ss,ix);
11383 TOPINT(nss,ix) = intval;
11385 case SAVEt_LONG: /* long reference */
11386 ptr = POPPTR(ss,ix);
11387 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11389 case SAVEt_CLEARSV:
11390 longval = (long)POPLONG(ss,ix);
11391 TOPLONG(nss,ix) = longval;
11393 case SAVEt_I32: /* I32 reference */
11394 case SAVEt_I16: /* I16 reference */
11395 case SAVEt_I8: /* I8 reference */
11396 case SAVEt_COP_ARYBASE: /* call CopARYBASE_set */
11397 ptr = POPPTR(ss,ix);
11398 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11400 TOPINT(nss,ix) = i;
11402 case SAVEt_IV: /* IV reference */
11403 ptr = POPPTR(ss,ix);
11404 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11406 TOPIV(nss,ix) = iv;
11408 case SAVEt_HPTR: /* HV* reference */
11409 case SAVEt_APTR: /* AV* reference */
11410 case SAVEt_SPTR: /* SV* reference */
11411 ptr = POPPTR(ss,ix);
11412 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11413 sv = (const SV *)POPPTR(ss,ix);
11414 TOPPTR(nss,ix) = sv_dup(sv, param);
11416 case SAVEt_VPTR: /* random* reference */
11417 ptr = POPPTR(ss,ix);
11418 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11419 ptr = POPPTR(ss,ix);
11420 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11422 case SAVEt_GENERIC_PVREF: /* generic char* */
11423 case SAVEt_PPTR: /* char* reference */
11424 ptr = POPPTR(ss,ix);
11425 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11426 c = (char*)POPPTR(ss,ix);
11427 TOPPTR(nss,ix) = pv_dup(c);
11429 case SAVEt_GP: /* scalar reference */
11430 gp = (GP*)POPPTR(ss,ix);
11431 TOPPTR(nss,ix) = gp = gp_dup(gp, param);
11432 (void)GpREFCNT_inc(gp);
11433 gv = (const GV *)POPPTR(ss,ix);
11434 TOPPTR(nss,ix) = gv_dup_inc(gv, param);
11437 ptr = POPPTR(ss,ix);
11438 if (ptr && (((OP*)ptr)->op_private & OPpREFCOUNTED)) {
11439 /* these are assumed to be refcounted properly */
11441 switch (((OP*)ptr)->op_type) {
11443 case OP_LEAVESUBLV:
11447 case OP_LEAVEWRITE:
11448 TOPPTR(nss,ix) = ptr;
11451 (void) OpREFCNT_inc(o);
11455 TOPPTR(nss,ix) = NULL;
11460 TOPPTR(nss,ix) = NULL;
11463 hv = (const HV *)POPPTR(ss,ix);
11464 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
11466 TOPINT(nss,ix) = i;
11469 c = (char*)POPPTR(ss,ix);
11470 TOPPTR(nss,ix) = pv_dup_inc(c);
11472 case SAVEt_STACK_POS: /* Position on Perl stack */
11474 TOPINT(nss,ix) = i;
11476 case SAVEt_DESTRUCTOR:
11477 ptr = POPPTR(ss,ix);
11478 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
11479 dptr = POPDPTR(ss,ix);
11480 TOPDPTR(nss,ix) = DPTR2FPTR(void (*)(void*),
11481 any_dup(FPTR2DPTR(void *, dptr),
11484 case SAVEt_DESTRUCTOR_X:
11485 ptr = POPPTR(ss,ix);
11486 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
11487 dxptr = POPDXPTR(ss,ix);
11488 TOPDXPTR(nss,ix) = DPTR2FPTR(void (*)(pTHX_ void*),
11489 any_dup(FPTR2DPTR(void *, dxptr),
11492 case SAVEt_REGCONTEXT:
11495 TOPINT(nss,ix) = i;
11498 case SAVEt_AELEM: /* array element */
11499 sv = (const SV *)POPPTR(ss,ix);
11500 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11502 TOPINT(nss,ix) = i;
11503 av = (const AV *)POPPTR(ss,ix);
11504 TOPPTR(nss,ix) = av_dup_inc(av, param);
11507 ptr = POPPTR(ss,ix);
11508 TOPPTR(nss,ix) = ptr;
11511 ptr = POPPTR(ss,ix);
11514 ((struct refcounted_he *)ptr)->refcounted_he_refcnt++;
11515 HINTS_REFCNT_UNLOCK;
11517 TOPPTR(nss,ix) = ptr;
11519 TOPINT(nss,ix) = i;
11520 if (i & HINT_LOCALIZE_HH) {
11521 hv = (const HV *)POPPTR(ss,ix);
11522 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
11525 case SAVEt_PADSV_AND_MORTALIZE:
11526 longval = (long)POPLONG(ss,ix);
11527 TOPLONG(nss,ix) = longval;
11528 ptr = POPPTR(ss,ix);
11529 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11530 sv = (const SV *)POPPTR(ss,ix);
11531 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11534 ptr = POPPTR(ss,ix);
11535 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11536 longval = (long)POPBOOL(ss,ix);
11537 TOPBOOL(nss,ix) = (bool)longval;
11539 case SAVEt_SET_SVFLAGS:
11541 TOPINT(nss,ix) = i;
11543 TOPINT(nss,ix) = i;
11544 sv = (const SV *)POPPTR(ss,ix);
11545 TOPPTR(nss,ix) = sv_dup(sv, param);
11547 case SAVEt_RE_STATE:
11549 const struct re_save_state *const old_state
11550 = (struct re_save_state *)
11551 (ss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
11552 struct re_save_state *const new_state
11553 = (struct re_save_state *)
11554 (nss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
11556 Copy(old_state, new_state, 1, struct re_save_state);
11557 ix -= SAVESTACK_ALLOC_FOR_RE_SAVE_STATE;
11559 new_state->re_state_bostr
11560 = pv_dup(old_state->re_state_bostr);
11561 new_state->re_state_reginput
11562 = pv_dup(old_state->re_state_reginput);
11563 new_state->re_state_regeol
11564 = pv_dup(old_state->re_state_regeol);
11565 new_state->re_state_regoffs
11566 = (regexp_paren_pair*)
11567 any_dup(old_state->re_state_regoffs, proto_perl);
11568 new_state->re_state_reglastparen
11569 = (U32*) any_dup(old_state->re_state_reglastparen,
11571 new_state->re_state_reglastcloseparen
11572 = (U32*)any_dup(old_state->re_state_reglastcloseparen,
11574 /* XXX This just has to be broken. The old save_re_context
11575 code did SAVEGENERICPV(PL_reg_start_tmp);
11576 PL_reg_start_tmp is char **.
11577 Look above to what the dup code does for
11578 SAVEt_GENERIC_PVREF
11579 It can never have worked.
11580 So this is merely a faithful copy of the exiting bug: */
11581 new_state->re_state_reg_start_tmp
11582 = (char **) pv_dup((char *)
11583 old_state->re_state_reg_start_tmp);
11584 /* I assume that it only ever "worked" because no-one called
11585 (pseudo)fork while the regexp engine had re-entered itself.
11587 #ifdef PERL_OLD_COPY_ON_WRITE
11588 new_state->re_state_nrs
11589 = sv_dup(old_state->re_state_nrs, param);
11591 new_state->re_state_reg_magic
11592 = (MAGIC*) any_dup(old_state->re_state_reg_magic,
11594 new_state->re_state_reg_oldcurpm
11595 = (PMOP*) any_dup(old_state->re_state_reg_oldcurpm,
11597 new_state->re_state_reg_curpm
11598 = (PMOP*) any_dup(old_state->re_state_reg_curpm,
11600 new_state->re_state_reg_oldsaved
11601 = pv_dup(old_state->re_state_reg_oldsaved);
11602 new_state->re_state_reg_poscache
11603 = pv_dup(old_state->re_state_reg_poscache);
11604 new_state->re_state_reg_starttry
11605 = pv_dup(old_state->re_state_reg_starttry);
11608 case SAVEt_COMPILE_WARNINGS:
11609 ptr = POPPTR(ss,ix);
11610 TOPPTR(nss,ix) = DUP_WARNINGS((STRLEN*)ptr);
11613 ptr = POPPTR(ss,ix);
11614 TOPPTR(nss,ix) = parser_dup((const yy_parser*)ptr, param);
11618 "panic: ss_dup inconsistency (%"IVdf")", (IV) type);
11626 /* if sv is a stash, call $class->CLONE_SKIP(), and set the SVphv_CLONEABLE
11627 * flag to the result. This is done for each stash before cloning starts,
11628 * so we know which stashes want their objects cloned */
11631 do_mark_cloneable_stash(pTHX_ SV *const sv)
11633 const HEK * const hvname = HvNAME_HEK((const HV *)sv);
11635 GV* const cloner = gv_fetchmethod_autoload(MUTABLE_HV(sv), "CLONE_SKIP", 0);
11636 SvFLAGS(sv) |= SVphv_CLONEABLE; /* clone objects by default */
11637 if (cloner && GvCV(cloner)) {
11644 mXPUSHs(newSVhek(hvname));
11646 call_sv(MUTABLE_SV(GvCV(cloner)), G_SCALAR);
11653 SvFLAGS(sv) &= ~SVphv_CLONEABLE;
11661 =for apidoc perl_clone
11663 Create and return a new interpreter by cloning the current one.
11665 perl_clone takes these flags as parameters:
11667 CLONEf_COPY_STACKS - is used to, well, copy the stacks also,
11668 without it we only clone the data and zero the stacks,
11669 with it we copy the stacks and the new perl interpreter is
11670 ready to run at the exact same point as the previous one.
11671 The pseudo-fork code uses COPY_STACKS while the
11672 threads->create doesn't.
11674 CLONEf_KEEP_PTR_TABLE
11675 perl_clone keeps a ptr_table with the pointer of the old
11676 variable as a key and the new variable as a value,
11677 this allows it to check if something has been cloned and not
11678 clone it again but rather just use the value and increase the
11679 refcount. If KEEP_PTR_TABLE is not set then perl_clone will kill
11680 the ptr_table using the function
11681 C<ptr_table_free(PL_ptr_table); PL_ptr_table = NULL;>,
11682 reason to keep it around is if you want to dup some of your own
11683 variable who are outside the graph perl scans, example of this
11684 code is in threads.xs create
11687 This is a win32 thing, it is ignored on unix, it tells perls
11688 win32host code (which is c++) to clone itself, this is needed on
11689 win32 if you want to run two threads at the same time,
11690 if you just want to do some stuff in a separate perl interpreter
11691 and then throw it away and return to the original one,
11692 you don't need to do anything.
11697 /* XXX the above needs expanding by someone who actually understands it ! */
11698 EXTERN_C PerlInterpreter *
11699 perl_clone_host(PerlInterpreter* proto_perl, UV flags);
11702 perl_clone(PerlInterpreter *proto_perl, UV flags)
11705 #ifdef PERL_IMPLICIT_SYS
11707 PERL_ARGS_ASSERT_PERL_CLONE;
11709 /* perlhost.h so we need to call into it
11710 to clone the host, CPerlHost should have a c interface, sky */
11712 if (flags & CLONEf_CLONE_HOST) {
11713 return perl_clone_host(proto_perl,flags);
11715 return perl_clone_using(proto_perl, flags,
11717 proto_perl->IMemShared,
11718 proto_perl->IMemParse,
11720 proto_perl->IStdIO,
11724 proto_perl->IProc);
11728 perl_clone_using(PerlInterpreter *proto_perl, UV flags,
11729 struct IPerlMem* ipM, struct IPerlMem* ipMS,
11730 struct IPerlMem* ipMP, struct IPerlEnv* ipE,
11731 struct IPerlStdIO* ipStd, struct IPerlLIO* ipLIO,
11732 struct IPerlDir* ipD, struct IPerlSock* ipS,
11733 struct IPerlProc* ipP)
11735 /* XXX many of the string copies here can be optimized if they're
11736 * constants; they need to be allocated as common memory and just
11737 * their pointers copied. */
11740 CLONE_PARAMS clone_params;
11741 CLONE_PARAMS* const param = &clone_params;
11743 PerlInterpreter * const my_perl = (PerlInterpreter*)(*ipM->pMalloc)(ipM, sizeof(PerlInterpreter));
11745 PERL_ARGS_ASSERT_PERL_CLONE_USING;
11747 /* for each stash, determine whether its objects should be cloned */
11748 S_visit(proto_perl, do_mark_cloneable_stash, SVt_PVHV, SVTYPEMASK);
11749 PERL_SET_THX(my_perl);
11752 PoisonNew(my_perl, 1, PerlInterpreter);
11758 PL_savestack_ix = 0;
11759 PL_savestack_max = -1;
11760 PL_sig_pending = 0;
11762 Zero(&PL_debug_pad, 1, struct perl_debug_pad);
11763 # else /* !DEBUGGING */
11764 Zero(my_perl, 1, PerlInterpreter);
11765 # endif /* DEBUGGING */
11767 /* host pointers */
11769 PL_MemShared = ipMS;
11770 PL_MemParse = ipMP;
11777 #else /* !PERL_IMPLICIT_SYS */
11779 CLONE_PARAMS clone_params;
11780 CLONE_PARAMS* param = &clone_params;
11781 PerlInterpreter * const my_perl = (PerlInterpreter*)PerlMem_malloc(sizeof(PerlInterpreter));
11783 PERL_ARGS_ASSERT_PERL_CLONE;
11785 /* for each stash, determine whether its objects should be cloned */
11786 S_visit(proto_perl, do_mark_cloneable_stash, SVt_PVHV, SVTYPEMASK);
11787 PERL_SET_THX(my_perl);
11790 PoisonNew(my_perl, 1, PerlInterpreter);
11796 PL_savestack_ix = 0;
11797 PL_savestack_max = -1;
11798 PL_sig_pending = 0;
11800 Zero(&PL_debug_pad, 1, struct perl_debug_pad);
11801 # else /* !DEBUGGING */
11802 Zero(my_perl, 1, PerlInterpreter);
11803 # endif /* DEBUGGING */
11804 #endif /* PERL_IMPLICIT_SYS */
11805 param->flags = flags;
11806 param->proto_perl = proto_perl;
11808 INIT_TRACK_MEMPOOL(my_perl->Imemory_debug_header, my_perl);
11810 PL_body_arenas = NULL;
11811 Zero(&PL_body_roots, 1, PL_body_roots);
11813 PL_nice_chunk = NULL;
11814 PL_nice_chunk_size = 0;
11816 PL_sv_objcount = 0;
11818 PL_sv_arenaroot = NULL;
11820 PL_debug = proto_perl->Idebug;
11822 PL_hash_seed = proto_perl->Ihash_seed;
11823 PL_rehash_seed = proto_perl->Irehash_seed;
11825 #ifdef USE_REENTRANT_API
11826 /* XXX: things like -Dm will segfault here in perlio, but doing
11827 * PERL_SET_CONTEXT(proto_perl);
11828 * breaks too many other things
11830 Perl_reentrant_init(aTHX);
11833 /* create SV map for pointer relocation */
11834 PL_ptr_table = ptr_table_new();
11836 /* initialize these special pointers as early as possible */
11837 SvANY(&PL_sv_undef) = NULL;
11838 SvREFCNT(&PL_sv_undef) = (~(U32)0)/2;
11839 SvFLAGS(&PL_sv_undef) = SVf_READONLY|SVt_NULL;
11840 ptr_table_store(PL_ptr_table, &proto_perl->Isv_undef, &PL_sv_undef);
11842 SvANY(&PL_sv_no) = new_XPVNV();
11843 SvREFCNT(&PL_sv_no) = (~(U32)0)/2;
11844 SvFLAGS(&PL_sv_no) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
11845 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
11846 SvPV_set(&PL_sv_no, savepvn(PL_No, 0));
11847 SvCUR_set(&PL_sv_no, 0);
11848 SvLEN_set(&PL_sv_no, 1);
11849 SvIV_set(&PL_sv_no, 0);
11850 SvNV_set(&PL_sv_no, 0);
11851 ptr_table_store(PL_ptr_table, &proto_perl->Isv_no, &PL_sv_no);
11853 SvANY(&PL_sv_yes) = new_XPVNV();
11854 SvREFCNT(&PL_sv_yes) = (~(U32)0)/2;
11855 SvFLAGS(&PL_sv_yes) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
11856 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
11857 SvPV_set(&PL_sv_yes, savepvn(PL_Yes, 1));
11858 SvCUR_set(&PL_sv_yes, 1);
11859 SvLEN_set(&PL_sv_yes, 2);
11860 SvIV_set(&PL_sv_yes, 1);
11861 SvNV_set(&PL_sv_yes, 1);
11862 ptr_table_store(PL_ptr_table, &proto_perl->Isv_yes, &PL_sv_yes);
11864 /* create (a non-shared!) shared string table */
11865 PL_strtab = newHV();
11866 HvSHAREKEYS_off(PL_strtab);
11867 hv_ksplit(PL_strtab, HvTOTALKEYS(proto_perl->Istrtab));
11868 ptr_table_store(PL_ptr_table, proto_perl->Istrtab, PL_strtab);
11870 PL_compiling = proto_perl->Icompiling;
11872 /* These two PVs will be free'd special way so must set them same way op.c does */
11873 PL_compiling.cop_stashpv = savesharedpv(PL_compiling.cop_stashpv);
11874 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_stashpv, PL_compiling.cop_stashpv);
11876 PL_compiling.cop_file = savesharedpv(PL_compiling.cop_file);
11877 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_file, PL_compiling.cop_file);
11879 ptr_table_store(PL_ptr_table, &proto_perl->Icompiling, &PL_compiling);
11880 PL_compiling.cop_warnings = DUP_WARNINGS(PL_compiling.cop_warnings);
11881 if (PL_compiling.cop_hints_hash) {
11883 PL_compiling.cop_hints_hash->refcounted_he_refcnt++;
11884 HINTS_REFCNT_UNLOCK;
11886 PL_curcop = (COP*)any_dup(proto_perl->Icurcop, proto_perl);
11887 #ifdef PERL_DEBUG_READONLY_OPS
11892 /* pseudo environmental stuff */
11893 PL_origargc = proto_perl->Iorigargc;
11894 PL_origargv = proto_perl->Iorigargv;
11896 param->stashes = newAV(); /* Setup array of objects to call clone on */
11898 /* Set tainting stuff before PerlIO_debug can possibly get called */
11899 PL_tainting = proto_perl->Itainting;
11900 PL_taint_warn = proto_perl->Itaint_warn;
11902 #ifdef PERLIO_LAYERS
11903 /* Clone PerlIO tables as soon as we can handle general xx_dup() */
11904 PerlIO_clone(aTHX_ proto_perl, param);
11907 PL_envgv = gv_dup(proto_perl->Ienvgv, param);
11908 PL_incgv = gv_dup(proto_perl->Iincgv, param);
11909 PL_hintgv = gv_dup(proto_perl->Ihintgv, param);
11910 PL_origfilename = SAVEPV(proto_perl->Iorigfilename);
11911 PL_diehook = sv_dup_inc(proto_perl->Idiehook, param);
11912 PL_warnhook = sv_dup_inc(proto_perl->Iwarnhook, param);
11915 PL_minus_c = proto_perl->Iminus_c;
11916 PL_patchlevel = sv_dup_inc(proto_perl->Ipatchlevel, param);
11917 PL_localpatches = proto_perl->Ilocalpatches;
11918 PL_splitstr = proto_perl->Isplitstr;
11919 PL_minus_n = proto_perl->Iminus_n;
11920 PL_minus_p = proto_perl->Iminus_p;
11921 PL_minus_l = proto_perl->Iminus_l;
11922 PL_minus_a = proto_perl->Iminus_a;
11923 PL_minus_E = proto_perl->Iminus_E;
11924 PL_minus_F = proto_perl->Iminus_F;
11925 PL_doswitches = proto_perl->Idoswitches;
11926 PL_dowarn = proto_perl->Idowarn;
11927 PL_doextract = proto_perl->Idoextract;
11928 PL_sawampersand = proto_perl->Isawampersand;
11929 PL_unsafe = proto_perl->Iunsafe;
11930 PL_inplace = SAVEPV(proto_perl->Iinplace);
11931 PL_e_script = sv_dup_inc(proto_perl->Ie_script, param);
11932 PL_perldb = proto_perl->Iperldb;
11933 PL_perl_destruct_level = proto_perl->Iperl_destruct_level;
11934 PL_exit_flags = proto_perl->Iexit_flags;
11936 /* magical thingies */
11937 /* XXX time(&PL_basetime) when asked for? */
11938 PL_basetime = proto_perl->Ibasetime;
11939 PL_formfeed = sv_dup(proto_perl->Iformfeed, param);
11941 PL_maxsysfd = proto_perl->Imaxsysfd;
11942 PL_statusvalue = proto_perl->Istatusvalue;
11944 PL_statusvalue_vms = proto_perl->Istatusvalue_vms;
11946 PL_statusvalue_posix = proto_perl->Istatusvalue_posix;
11948 PL_encoding = sv_dup(proto_perl->Iencoding, param);
11950 sv_setpvs(PERL_DEBUG_PAD(0), ""); /* For regex debugging. */
11951 sv_setpvs(PERL_DEBUG_PAD(1), ""); /* ext/re needs these */
11952 sv_setpvs(PERL_DEBUG_PAD(2), ""); /* even without DEBUGGING. */
11955 /* RE engine related */
11956 Zero(&PL_reg_state, 1, struct re_save_state);
11957 PL_reginterp_cnt = 0;
11958 PL_regmatch_slab = NULL;
11960 /* Clone the regex array */
11961 /* ORANGE FIXME for plugins, probably in the SV dup code.
11962 newSViv(PTR2IV(CALLREGDUPE(
11963 INT2PTR(REGEXP *, SvIVX(regex)), param))))
11965 PL_regex_padav = av_dup_inc(proto_perl->Iregex_padav, param);
11966 PL_regex_pad = AvARRAY(PL_regex_padav);
11968 /* shortcuts to various I/O objects */
11969 PL_ofsgv = gv_dup(proto_perl->Iofsgv, param);
11970 PL_stdingv = gv_dup(proto_perl->Istdingv, param);
11971 PL_stderrgv = gv_dup(proto_perl->Istderrgv, param);
11972 PL_defgv = gv_dup(proto_perl->Idefgv, param);
11973 PL_argvgv = gv_dup(proto_perl->Iargvgv, param);
11974 PL_argvoutgv = gv_dup(proto_perl->Iargvoutgv, param);
11975 PL_argvout_stack = av_dup_inc(proto_perl->Iargvout_stack, param);
11977 /* shortcuts to regexp stuff */
11978 PL_replgv = gv_dup(proto_perl->Ireplgv, param);
11980 /* shortcuts to misc objects */
11981 PL_errgv = gv_dup(proto_perl->Ierrgv, param);
11983 /* shortcuts to debugging objects */
11984 PL_DBgv = gv_dup(proto_perl->IDBgv, param);
11985 PL_DBline = gv_dup(proto_perl->IDBline, param);
11986 PL_DBsub = gv_dup(proto_perl->IDBsub, param);
11987 PL_DBsingle = sv_dup(proto_perl->IDBsingle, param);
11988 PL_DBtrace = sv_dup(proto_perl->IDBtrace, param);
11989 PL_DBsignal = sv_dup(proto_perl->IDBsignal, param);
11990 PL_dbargs = av_dup(proto_perl->Idbargs, param);
11992 /* symbol tables */
11993 PL_defstash = hv_dup_inc(proto_perl->Idefstash, param);
11994 PL_curstash = hv_dup(proto_perl->Icurstash, param);
11995 PL_debstash = hv_dup(proto_perl->Idebstash, param);
11996 PL_globalstash = hv_dup(proto_perl->Iglobalstash, param);
11997 PL_curstname = sv_dup_inc(proto_perl->Icurstname, param);
11999 PL_beginav = av_dup_inc(proto_perl->Ibeginav, param);
12000 PL_beginav_save = av_dup_inc(proto_perl->Ibeginav_save, param);
12001 PL_checkav_save = av_dup_inc(proto_perl->Icheckav_save, param);
12002 PL_unitcheckav = av_dup_inc(proto_perl->Iunitcheckav, param);
12003 PL_unitcheckav_save = av_dup_inc(proto_perl->Iunitcheckav_save, param);
12004 PL_endav = av_dup_inc(proto_perl->Iendav, param);
12005 PL_checkav = av_dup_inc(proto_perl->Icheckav, param);
12006 PL_initav = av_dup_inc(proto_perl->Iinitav, param);
12008 PL_sub_generation = proto_perl->Isub_generation;
12009 PL_isarev = hv_dup_inc(proto_perl->Iisarev, param);
12011 /* funky return mechanisms */
12012 PL_forkprocess = proto_perl->Iforkprocess;
12014 /* subprocess state */
12015 PL_fdpid = av_dup_inc(proto_perl->Ifdpid, param);
12017 /* internal state */
12018 PL_maxo = proto_perl->Imaxo;
12019 if (proto_perl->Iop_mask)
12020 PL_op_mask = SAVEPVN(proto_perl->Iop_mask, PL_maxo);
12023 /* PL_asserting = proto_perl->Iasserting; */
12025 /* current interpreter roots */
12026 PL_main_cv = cv_dup_inc(proto_perl->Imain_cv, param);
12028 PL_main_root = OpREFCNT_inc(proto_perl->Imain_root);
12030 PL_main_start = proto_perl->Imain_start;
12031 PL_eval_root = proto_perl->Ieval_root;
12032 PL_eval_start = proto_perl->Ieval_start;
12034 /* runtime control stuff */
12035 PL_curcopdb = (COP*)any_dup(proto_perl->Icurcopdb, proto_perl);
12037 PL_filemode = proto_perl->Ifilemode;
12038 PL_lastfd = proto_perl->Ilastfd;
12039 PL_oldname = proto_perl->Ioldname; /* XXX not quite right */
12042 PL_gensym = proto_perl->Igensym;
12043 PL_preambleav = av_dup_inc(proto_perl->Ipreambleav, param);
12044 PL_laststatval = proto_perl->Ilaststatval;
12045 PL_laststype = proto_perl->Ilaststype;
12048 PL_ors_sv = sv_dup_inc(proto_perl->Iors_sv, param);
12050 /* interpreter atexit processing */
12051 PL_exitlistlen = proto_perl->Iexitlistlen;
12052 if (PL_exitlistlen) {
12053 Newx(PL_exitlist, PL_exitlistlen, PerlExitListEntry);
12054 Copy(proto_perl->Iexitlist, PL_exitlist, PL_exitlistlen, PerlExitListEntry);
12057 PL_exitlist = (PerlExitListEntry*)NULL;
12059 PL_my_cxt_size = proto_perl->Imy_cxt_size;
12060 if (PL_my_cxt_size) {
12061 Newx(PL_my_cxt_list, PL_my_cxt_size, void *);
12062 Copy(proto_perl->Imy_cxt_list, PL_my_cxt_list, PL_my_cxt_size, void *);
12063 #ifdef PERL_GLOBAL_STRUCT_PRIVATE
12064 Newx(PL_my_cxt_keys, PL_my_cxt_size, const char *);
12065 Copy(proto_perl->Imy_cxt_keys, PL_my_cxt_keys, PL_my_cxt_size, char *);
12069 PL_my_cxt_list = (void**)NULL;
12070 #ifdef PERL_GLOBAL_STRUCT_PRIVATE
12071 PL_my_cxt_keys = (const char**)NULL;
12074 PL_modglobal = hv_dup_inc(proto_perl->Imodglobal, param);
12075 PL_custom_op_names = hv_dup_inc(proto_perl->Icustom_op_names,param);
12076 PL_custom_op_descs = hv_dup_inc(proto_perl->Icustom_op_descs,param);
12078 PL_profiledata = NULL;
12080 PL_compcv = cv_dup(proto_perl->Icompcv, param);
12082 PAD_CLONE_VARS(proto_perl, param);
12084 #ifdef HAVE_INTERP_INTERN
12085 sys_intern_dup(&proto_perl->Isys_intern, &PL_sys_intern);
12088 /* more statics moved here */
12089 PL_generation = proto_perl->Igeneration;
12090 PL_DBcv = cv_dup(proto_perl->IDBcv, param);
12092 PL_in_clean_objs = proto_perl->Iin_clean_objs;
12093 PL_in_clean_all = proto_perl->Iin_clean_all;
12095 PL_uid = proto_perl->Iuid;
12096 PL_euid = proto_perl->Ieuid;
12097 PL_gid = proto_perl->Igid;
12098 PL_egid = proto_perl->Iegid;
12099 PL_nomemok = proto_perl->Inomemok;
12100 PL_an = proto_perl->Ian;
12101 PL_evalseq = proto_perl->Ievalseq;
12102 PL_origenviron = proto_perl->Iorigenviron; /* XXX not quite right */
12103 PL_origalen = proto_perl->Iorigalen;
12104 #ifdef PERL_USES_PL_PIDSTATUS
12105 PL_pidstatus = newHV(); /* XXX flag for cloning? */
12107 PL_osname = SAVEPV(proto_perl->Iosname);
12108 PL_sighandlerp = proto_perl->Isighandlerp;
12110 PL_runops = proto_perl->Irunops;
12112 PL_parser = parser_dup(proto_perl->Iparser, param);
12114 /* XXX this only works if the saved cop has already been cloned */
12115 if (proto_perl->Iparser) {
12116 PL_parser->saved_curcop = (COP*)any_dup(
12117 proto_perl->Iparser->saved_curcop,
12121 PL_subline = proto_perl->Isubline;
12122 PL_subname = sv_dup_inc(proto_perl->Isubname, param);
12125 PL_cryptseen = proto_perl->Icryptseen;
12128 PL_hints = proto_perl->Ihints;
12130 PL_amagic_generation = proto_perl->Iamagic_generation;
12132 #ifdef USE_LOCALE_COLLATE
12133 PL_collation_ix = proto_perl->Icollation_ix;
12134 PL_collation_name = SAVEPV(proto_perl->Icollation_name);
12135 PL_collation_standard = proto_perl->Icollation_standard;
12136 PL_collxfrm_base = proto_perl->Icollxfrm_base;
12137 PL_collxfrm_mult = proto_perl->Icollxfrm_mult;
12138 #endif /* USE_LOCALE_COLLATE */
12140 #ifdef USE_LOCALE_NUMERIC
12141 PL_numeric_name = SAVEPV(proto_perl->Inumeric_name);
12142 PL_numeric_standard = proto_perl->Inumeric_standard;
12143 PL_numeric_local = proto_perl->Inumeric_local;
12144 PL_numeric_radix_sv = sv_dup_inc(proto_perl->Inumeric_radix_sv, param);
12145 #endif /* !USE_LOCALE_NUMERIC */
12147 /* utf8 character classes */
12148 PL_utf8_alnum = sv_dup_inc(proto_perl->Iutf8_alnum, param);
12149 PL_utf8_alnumc = sv_dup_inc(proto_perl->Iutf8_alnumc, param);
12150 PL_utf8_ascii = sv_dup_inc(proto_perl->Iutf8_ascii, param);
12151 PL_utf8_alpha = sv_dup_inc(proto_perl->Iutf8_alpha, param);
12152 PL_utf8_space = sv_dup_inc(proto_perl->Iutf8_space, param);
12153 PL_utf8_cntrl = sv_dup_inc(proto_perl->Iutf8_cntrl, param);
12154 PL_utf8_graph = sv_dup_inc(proto_perl->Iutf8_graph, param);
12155 PL_utf8_digit = sv_dup_inc(proto_perl->Iutf8_digit, param);
12156 PL_utf8_upper = sv_dup_inc(proto_perl->Iutf8_upper, param);
12157 PL_utf8_lower = sv_dup_inc(proto_perl->Iutf8_lower, param);
12158 PL_utf8_print = sv_dup_inc(proto_perl->Iutf8_print, param);
12159 PL_utf8_punct = sv_dup_inc(proto_perl->Iutf8_punct, param);
12160 PL_utf8_xdigit = sv_dup_inc(proto_perl->Iutf8_xdigit, param);
12161 PL_utf8_mark = sv_dup_inc(proto_perl->Iutf8_mark, param);
12162 PL_utf8_toupper = sv_dup_inc(proto_perl->Iutf8_toupper, param);
12163 PL_utf8_totitle = sv_dup_inc(proto_perl->Iutf8_totitle, param);
12164 PL_utf8_tolower = sv_dup_inc(proto_perl->Iutf8_tolower, param);
12165 PL_utf8_tofold = sv_dup_inc(proto_perl->Iutf8_tofold, param);
12166 PL_utf8_idstart = sv_dup_inc(proto_perl->Iutf8_idstart, param);
12167 PL_utf8_idcont = sv_dup_inc(proto_perl->Iutf8_idcont, param);
12169 /* Did the locale setup indicate UTF-8? */
12170 PL_utf8locale = proto_perl->Iutf8locale;
12171 /* Unicode features (see perlrun/-C) */
12172 PL_unicode = proto_perl->Iunicode;
12174 /* Pre-5.8 signals control */
12175 PL_signals = proto_perl->Isignals;
12177 /* times() ticks per second */
12178 PL_clocktick = proto_perl->Iclocktick;
12180 /* Recursion stopper for PerlIO_find_layer */
12181 PL_in_load_module = proto_perl->Iin_load_module;
12183 /* sort() routine */
12184 PL_sort_RealCmp = proto_perl->Isort_RealCmp;
12186 /* Not really needed/useful since the reenrant_retint is "volatile",
12187 * but do it for consistency's sake. */
12188 PL_reentrant_retint = proto_perl->Ireentrant_retint;
12190 /* Hooks to shared SVs and locks. */
12191 PL_sharehook = proto_perl->Isharehook;
12192 PL_lockhook = proto_perl->Ilockhook;
12193 PL_unlockhook = proto_perl->Iunlockhook;
12194 PL_threadhook = proto_perl->Ithreadhook;
12195 PL_destroyhook = proto_perl->Idestroyhook;
12197 #ifdef THREADS_HAVE_PIDS
12198 PL_ppid = proto_perl->Ippid;
12202 PL_last_swash_hv = NULL; /* reinits on demand */
12203 PL_last_swash_klen = 0;
12204 PL_last_swash_key[0]= '\0';
12205 PL_last_swash_tmps = (U8*)NULL;
12206 PL_last_swash_slen = 0;
12208 PL_glob_index = proto_perl->Iglob_index;
12209 PL_srand_called = proto_perl->Isrand_called;
12211 if (proto_perl->Ipsig_pend) {
12212 Newxz(PL_psig_pend, SIG_SIZE, int);
12215 PL_psig_pend = (int*)NULL;
12218 if (proto_perl->Ipsig_name) {
12219 Newx(PL_psig_name, 2 * SIG_SIZE, SV*);
12220 sv_dup_inc_multiple(proto_perl->Ipsig_name, PL_psig_name, 2 * SIG_SIZE,
12222 PL_psig_ptr = PL_psig_name + SIG_SIZE;
12225 PL_psig_ptr = (SV**)NULL;
12226 PL_psig_name = (SV**)NULL;
12229 /* intrpvar.h stuff */
12231 if (flags & CLONEf_COPY_STACKS) {
12232 /* next allocation will be PL_tmps_stack[PL_tmps_ix+1] */
12233 PL_tmps_ix = proto_perl->Itmps_ix;
12234 PL_tmps_max = proto_perl->Itmps_max;
12235 PL_tmps_floor = proto_perl->Itmps_floor;
12236 Newx(PL_tmps_stack, PL_tmps_max, SV*);
12237 sv_dup_inc_multiple(proto_perl->Itmps_stack, PL_tmps_stack, PL_tmps_ix,
12240 /* next PUSHMARK() sets *(PL_markstack_ptr+1) */
12241 i = proto_perl->Imarkstack_max - proto_perl->Imarkstack;
12242 Newxz(PL_markstack, i, I32);
12243 PL_markstack_max = PL_markstack + (proto_perl->Imarkstack_max
12244 - proto_perl->Imarkstack);
12245 PL_markstack_ptr = PL_markstack + (proto_perl->Imarkstack_ptr
12246 - proto_perl->Imarkstack);
12247 Copy(proto_perl->Imarkstack, PL_markstack,
12248 PL_markstack_ptr - PL_markstack + 1, I32);
12250 /* next push_scope()/ENTER sets PL_scopestack[PL_scopestack_ix]
12251 * NOTE: unlike the others! */
12252 PL_scopestack_ix = proto_perl->Iscopestack_ix;
12253 PL_scopestack_max = proto_perl->Iscopestack_max;
12254 Newxz(PL_scopestack, PL_scopestack_max, I32);
12255 Copy(proto_perl->Iscopestack, PL_scopestack, PL_scopestack_ix, I32);
12257 /* NOTE: si_dup() looks at PL_markstack */
12258 PL_curstackinfo = si_dup(proto_perl->Icurstackinfo, param);
12260 /* PL_curstack = PL_curstackinfo->si_stack; */
12261 PL_curstack = av_dup(proto_perl->Icurstack, param);
12262 PL_mainstack = av_dup(proto_perl->Imainstack, param);
12264 /* next PUSHs() etc. set *(PL_stack_sp+1) */
12265 PL_stack_base = AvARRAY(PL_curstack);
12266 PL_stack_sp = PL_stack_base + (proto_perl->Istack_sp
12267 - proto_perl->Istack_base);
12268 PL_stack_max = PL_stack_base + AvMAX(PL_curstack);
12270 /* next SSPUSHFOO() sets PL_savestack[PL_savestack_ix]
12271 * NOTE: unlike the others! */
12272 PL_savestack_ix = proto_perl->Isavestack_ix;
12273 PL_savestack_max = proto_perl->Isavestack_max;
12274 /*Newxz(PL_savestack, PL_savestack_max, ANY);*/
12275 PL_savestack = ss_dup(proto_perl, param);
12279 ENTER; /* perl_destruct() wants to LEAVE; */
12281 /* although we're not duplicating the tmps stack, we should still
12282 * add entries for any SVs on the tmps stack that got cloned by a
12283 * non-refcount means (eg a temp in @_); otherwise they will be
12286 for (i = 0; i<= proto_perl->Itmps_ix; i++) {
12287 SV * const nsv = MUTABLE_SV(ptr_table_fetch(PL_ptr_table,
12288 proto_perl->Itmps_stack[i]));
12289 if (nsv && !SvREFCNT(nsv)) {
12291 PL_tmps_stack[++PL_tmps_ix] = SvREFCNT_inc_simple(nsv);
12296 PL_start_env = proto_perl->Istart_env; /* XXXXXX */
12297 PL_top_env = &PL_start_env;
12299 PL_op = proto_perl->Iop;
12302 PL_Xpv = (XPV*)NULL;
12303 my_perl->Ina = proto_perl->Ina;
12305 PL_statbuf = proto_perl->Istatbuf;
12306 PL_statcache = proto_perl->Istatcache;
12307 PL_statgv = gv_dup(proto_perl->Istatgv, param);
12308 PL_statname = sv_dup_inc(proto_perl->Istatname, param);
12310 PL_timesbuf = proto_perl->Itimesbuf;
12313 PL_tainted = proto_perl->Itainted;
12314 PL_curpm = proto_perl->Icurpm; /* XXX No PMOP ref count */
12315 PL_rs = sv_dup_inc(proto_perl->Irs, param);
12316 PL_last_in_gv = gv_dup(proto_perl->Ilast_in_gv, param);
12317 PL_defoutgv = gv_dup_inc(proto_perl->Idefoutgv, param);
12318 PL_chopset = proto_perl->Ichopset; /* XXX never deallocated */
12319 PL_toptarget = sv_dup_inc(proto_perl->Itoptarget, param);
12320 PL_bodytarget = sv_dup_inc(proto_perl->Ibodytarget, param);
12321 PL_formtarget = sv_dup(proto_perl->Iformtarget, param);
12323 PL_restartop = proto_perl->Irestartop;
12324 PL_in_eval = proto_perl->Iin_eval;
12325 PL_delaymagic = proto_perl->Idelaymagic;
12326 PL_dirty = proto_perl->Idirty;
12327 PL_localizing = proto_perl->Ilocalizing;
12329 PL_errors = sv_dup_inc(proto_perl->Ierrors, param);
12330 PL_hv_fetch_ent_mh = NULL;
12331 PL_modcount = proto_perl->Imodcount;
12332 PL_lastgotoprobe = NULL;
12333 PL_dumpindent = proto_perl->Idumpindent;
12335 PL_sortcop = (OP*)any_dup(proto_perl->Isortcop, proto_perl);
12336 PL_sortstash = hv_dup(proto_perl->Isortstash, param);
12337 PL_firstgv = gv_dup(proto_perl->Ifirstgv, param);
12338 PL_secondgv = gv_dup(proto_perl->Isecondgv, param);
12339 PL_efloatbuf = NULL; /* reinits on demand */
12340 PL_efloatsize = 0; /* reinits on demand */
12344 PL_screamfirst = NULL;
12345 PL_screamnext = NULL;
12346 PL_maxscream = -1; /* reinits on demand */
12347 PL_lastscream = NULL;
12350 PL_regdummy = proto_perl->Iregdummy;
12351 PL_colorset = 0; /* reinits PL_colors[] */
12352 /*PL_colors[6] = {0,0,0,0,0,0};*/
12356 /* Pluggable optimizer */
12357 PL_peepp = proto_perl->Ipeepp;
12358 /* op_free() hook */
12359 PL_opfreehook = proto_perl->Iopfreehook;
12361 PL_stashcache = newHV();
12363 PL_watchaddr = (char **) ptr_table_fetch(PL_ptr_table,
12364 proto_perl->Iwatchaddr);
12365 PL_watchok = PL_watchaddr ? * PL_watchaddr : NULL;
12366 if (PL_debug && PL_watchaddr) {
12367 PerlIO_printf(Perl_debug_log,
12368 "WATCHING: %"UVxf" cloned as %"UVxf" with value %"UVxf"\n",
12369 PTR2UV(proto_perl->Iwatchaddr), PTR2UV(PL_watchaddr),
12370 PTR2UV(PL_watchok));
12373 PL_registered_mros = hv_dup_inc(proto_perl->Iregistered_mros, param);
12375 /* Call the ->CLONE method, if it exists, for each of the stashes
12376 identified by sv_dup() above.
12378 while(av_len(param->stashes) != -1) {
12379 HV* const stash = MUTABLE_HV(av_shift(param->stashes));
12380 GV* const cloner = gv_fetchmethod_autoload(stash, "CLONE", 0);
12381 if (cloner && GvCV(cloner)) {
12386 mXPUSHs(newSVhek(HvNAME_HEK(stash)));
12388 call_sv(MUTABLE_SV(GvCV(cloner)), G_DISCARD);
12394 if (!(flags & CLONEf_KEEP_PTR_TABLE)) {
12395 ptr_table_free(PL_ptr_table);
12396 PL_ptr_table = NULL;
12400 SvREFCNT_dec(param->stashes);
12402 /* orphaned? eg threads->new inside BEGIN or use */
12403 if (PL_compcv && ! SvREFCNT(PL_compcv)) {
12404 SvREFCNT_inc_simple_void(PL_compcv);
12405 SAVEFREESV(PL_compcv);
12411 #endif /* USE_ITHREADS */
12414 =head1 Unicode Support
12416 =for apidoc sv_recode_to_utf8
12418 The encoding is assumed to be an Encode object, on entry the PV
12419 of the sv is assumed to be octets in that encoding, and the sv
12420 will be converted into Unicode (and UTF-8).
12422 If the sv already is UTF-8 (or if it is not POK), or if the encoding
12423 is not a reference, nothing is done to the sv. If the encoding is not
12424 an C<Encode::XS> Encoding object, bad things will happen.
12425 (See F<lib/encoding.pm> and L<Encode>).
12427 The PV of the sv is returned.
12432 Perl_sv_recode_to_utf8(pTHX_ SV *sv, SV *encoding)
12436 PERL_ARGS_ASSERT_SV_RECODE_TO_UTF8;
12438 if (SvPOK(sv) && !SvUTF8(sv) && !IN_BYTES && SvROK(encoding)) {
12452 Passing sv_yes is wrong - it needs to be or'ed set of constants
12453 for Encode::XS, while UTf-8 decode (currently) assumes a true value means
12454 remove converted chars from source.
12456 Both will default the value - let them.
12458 XPUSHs(&PL_sv_yes);
12461 call_method("decode", G_SCALAR);
12465 s = SvPV_const(uni, len);
12466 if (s != SvPVX_const(sv)) {
12467 SvGROW(sv, len + 1);
12468 Move(s, SvPVX(sv), len + 1, char);
12469 SvCUR_set(sv, len);
12476 return SvPOKp(sv) ? SvPVX(sv) : NULL;
12480 =for apidoc sv_cat_decode
12482 The encoding is assumed to be an Encode object, the PV of the ssv is
12483 assumed to be octets in that encoding and decoding the input starts
12484 from the position which (PV + *offset) pointed to. The dsv will be
12485 concatenated the decoded UTF-8 string from ssv. Decoding will terminate
12486 when the string tstr appears in decoding output or the input ends on
12487 the PV of the ssv. The value which the offset points will be modified
12488 to the last input position on the ssv.
12490 Returns TRUE if the terminator was found, else returns FALSE.
12495 Perl_sv_cat_decode(pTHX_ SV *dsv, SV *encoding,
12496 SV *ssv, int *offset, char *tstr, int tlen)
12501 PERL_ARGS_ASSERT_SV_CAT_DECODE;
12503 if (SvPOK(ssv) && SvPOK(dsv) && SvROK(encoding) && offset) {
12514 offsv = newSViv(*offset);
12516 mXPUSHp(tstr, tlen);
12518 call_method("cat_decode", G_SCALAR);
12520 ret = SvTRUE(TOPs);
12521 *offset = SvIV(offsv);
12527 Perl_croak(aTHX_ "Invalid argument to sv_cat_decode");
12532 /* ---------------------------------------------------------------------
12534 * support functions for report_uninit()
12537 /* the maxiumum size of array or hash where we will scan looking
12538 * for the undefined element that triggered the warning */
12540 #define FUV_MAX_SEARCH_SIZE 1000
12542 /* Look for an entry in the hash whose value has the same SV as val;
12543 * If so, return a mortal copy of the key. */
12546 S_find_hash_subscript(pTHX_ const HV *const hv, const SV *const val)
12549 register HE **array;
12552 PERL_ARGS_ASSERT_FIND_HASH_SUBSCRIPT;
12554 if (!hv || SvMAGICAL(hv) || !HvARRAY(hv) ||
12555 (HvTOTALKEYS(hv) > FUV_MAX_SEARCH_SIZE))
12558 array = HvARRAY(hv);
12560 for (i=HvMAX(hv); i>0; i--) {
12561 register HE *entry;
12562 for (entry = array[i]; entry; entry = HeNEXT(entry)) {
12563 if (HeVAL(entry) != val)
12565 if ( HeVAL(entry) == &PL_sv_undef ||
12566 HeVAL(entry) == &PL_sv_placeholder)
12570 if (HeKLEN(entry) == HEf_SVKEY)
12571 return sv_mortalcopy(HeKEY_sv(entry));
12572 return sv_2mortal(newSVhek(HeKEY_hek(entry)));
12578 /* Look for an entry in the array whose value has the same SV as val;
12579 * If so, return the index, otherwise return -1. */
12582 S_find_array_subscript(pTHX_ const AV *const av, const SV *const val)
12586 PERL_ARGS_ASSERT_FIND_ARRAY_SUBSCRIPT;
12588 if (!av || SvMAGICAL(av) || !AvARRAY(av) ||
12589 (AvFILLp(av) > FUV_MAX_SEARCH_SIZE))
12592 if (val != &PL_sv_undef) {
12593 SV ** const svp = AvARRAY(av);
12596 for (i=AvFILLp(av); i>=0; i--)
12603 /* S_varname(): return the name of a variable, optionally with a subscript.
12604 * If gv is non-zero, use the name of that global, along with gvtype (one
12605 * of "$", "@", "%"); otherwise use the name of the lexical at pad offset
12606 * targ. Depending on the value of the subscript_type flag, return:
12609 #define FUV_SUBSCRIPT_NONE 1 /* "@foo" */
12610 #define FUV_SUBSCRIPT_ARRAY 2 /* "$foo[aindex]" */
12611 #define FUV_SUBSCRIPT_HASH 3 /* "$foo{keyname}" */
12612 #define FUV_SUBSCRIPT_WITHIN 4 /* "within @foo" */
12615 S_varname(pTHX_ const GV *const gv, const char gvtype, PADOFFSET targ,
12616 const SV *const keyname, I32 aindex, int subscript_type)
12619 SV * const name = sv_newmortal();
12622 buffer[0] = gvtype;
12625 /* as gv_fullname4(), but add literal '^' for $^FOO names */
12627 gv_fullname4(name, gv, buffer, 0);
12629 if ((unsigned int)SvPVX(name)[1] <= 26) {
12631 buffer[1] = SvPVX(name)[1] + 'A' - 1;
12633 /* Swap the 1 unprintable control character for the 2 byte pretty
12634 version - ie substr($name, 1, 1) = $buffer; */
12635 sv_insert(name, 1, 1, buffer, 2);
12639 CV * const cv = find_runcv(NULL);
12643 if (!cv || !CvPADLIST(cv))
12645 av = MUTABLE_AV((*av_fetch(CvPADLIST(cv), 0, FALSE)));
12646 sv = *av_fetch(av, targ, FALSE);
12647 sv_setpvn(name, SvPV_nolen_const(sv), SvCUR(sv));
12650 if (subscript_type == FUV_SUBSCRIPT_HASH) {
12651 SV * const sv = newSV(0);
12652 *SvPVX(name) = '$';
12653 Perl_sv_catpvf(aTHX_ name, "{%s}",
12654 pv_display(sv,SvPVX_const(keyname), SvCUR(keyname), 0, 32));
12657 else if (subscript_type == FUV_SUBSCRIPT_ARRAY) {
12658 *SvPVX(name) = '$';
12659 Perl_sv_catpvf(aTHX_ name, "[%"IVdf"]", (IV)aindex);
12661 else if (subscript_type == FUV_SUBSCRIPT_WITHIN) {
12662 /* We know that name has no magic, so can use 0 instead of SV_GMAGIC */
12663 Perl_sv_insert_flags(aTHX_ name, 0, 0, STR_WITH_LEN("within "), 0);
12671 =for apidoc find_uninit_var
12673 Find the name of the undefined variable (if any) that caused the operator o
12674 to issue a "Use of uninitialized value" warning.
12675 If match is true, only return a name if it's value matches uninit_sv.
12676 So roughly speaking, if a unary operator (such as OP_COS) generates a
12677 warning, then following the direct child of the op may yield an
12678 OP_PADSV or OP_GV that gives the name of the undefined variable. On the
12679 other hand, with OP_ADD there are two branches to follow, so we only print
12680 the variable name if we get an exact match.
12682 The name is returned as a mortal SV.
12684 Assumes that PL_op is the op that originally triggered the error, and that
12685 PL_comppad/PL_curpad points to the currently executing pad.
12691 S_find_uninit_var(pTHX_ const OP *const obase, const SV *const uninit_sv,
12697 const OP *o, *o2, *kid;
12699 if (!obase || (match && (!uninit_sv || uninit_sv == &PL_sv_undef ||
12700 uninit_sv == &PL_sv_placeholder)))
12703 switch (obase->op_type) {
12710 const bool pad = (obase->op_type == OP_PADAV || obase->op_type == OP_PADHV);
12711 const bool hash = (obase->op_type == OP_PADHV || obase->op_type == OP_RV2HV);
12714 int subscript_type = FUV_SUBSCRIPT_WITHIN;
12716 if (pad) { /* @lex, %lex */
12717 sv = PAD_SVl(obase->op_targ);
12721 if (cUNOPx(obase)->op_first->op_type == OP_GV) {
12722 /* @global, %global */
12723 gv = cGVOPx_gv(cUNOPx(obase)->op_first);
12726 sv = hash ? MUTABLE_SV(GvHV(gv)): MUTABLE_SV(GvAV(gv));
12728 else /* @{expr}, %{expr} */
12729 return find_uninit_var(cUNOPx(obase)->op_first,
12733 /* attempt to find a match within the aggregate */
12735 keysv = find_hash_subscript((const HV*)sv, uninit_sv);
12737 subscript_type = FUV_SUBSCRIPT_HASH;
12740 index = find_array_subscript((const AV *)sv, uninit_sv);
12742 subscript_type = FUV_SUBSCRIPT_ARRAY;
12745 if (match && subscript_type == FUV_SUBSCRIPT_WITHIN)
12748 return varname(gv, hash ? '%' : '@', obase->op_targ,
12749 keysv, index, subscript_type);
12753 if (match && PAD_SVl(obase->op_targ) != uninit_sv)
12755 return varname(NULL, '$', obase->op_targ,
12756 NULL, 0, FUV_SUBSCRIPT_NONE);
12759 gv = cGVOPx_gv(obase);
12760 if (!gv || (match && GvSV(gv) != uninit_sv))
12762 return varname(gv, '$', 0, NULL, 0, FUV_SUBSCRIPT_NONE);
12765 if (obase->op_flags & OPf_SPECIAL) { /* lexical array */
12768 AV *av = MUTABLE_AV(PAD_SV(obase->op_targ));
12769 if (!av || SvRMAGICAL(av))
12771 svp = av_fetch(av, (I32)obase->op_private, FALSE);
12772 if (!svp || *svp != uninit_sv)
12775 return varname(NULL, '$', obase->op_targ,
12776 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
12779 gv = cGVOPx_gv(obase);
12784 AV *const av = GvAV(gv);
12785 if (!av || SvRMAGICAL(av))
12787 svp = av_fetch(av, (I32)obase->op_private, FALSE);
12788 if (!svp || *svp != uninit_sv)
12791 return varname(gv, '$', 0,
12792 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
12797 o = cUNOPx(obase)->op_first;
12798 if (!o || o->op_type != OP_NULL ||
12799 ! (o->op_targ == OP_AELEM || o->op_targ == OP_HELEM))
12801 return find_uninit_var(cBINOPo->op_last, uninit_sv, match);
12805 if (PL_op == obase)
12806 /* $a[uninit_expr] or $h{uninit_expr} */
12807 return find_uninit_var(cBINOPx(obase)->op_last, uninit_sv, match);
12810 o = cBINOPx(obase)->op_first;
12811 kid = cBINOPx(obase)->op_last;
12813 /* get the av or hv, and optionally the gv */
12815 if (o->op_type == OP_PADAV || o->op_type == OP_PADHV) {
12816 sv = PAD_SV(o->op_targ);
12818 else if ((o->op_type == OP_RV2AV || o->op_type == OP_RV2HV)
12819 && cUNOPo->op_first->op_type == OP_GV)
12821 gv = cGVOPx_gv(cUNOPo->op_first);
12825 == OP_RV2HV ? MUTABLE_SV(GvHV(gv)) : MUTABLE_SV(GvAV(gv));
12830 if (kid && kid->op_type == OP_CONST && SvOK(cSVOPx_sv(kid))) {
12831 /* index is constant */
12835 if (obase->op_type == OP_HELEM) {
12836 HE* he = hv_fetch_ent(MUTABLE_HV(sv), cSVOPx_sv(kid), 0, 0);
12837 if (!he || HeVAL(he) != uninit_sv)
12841 SV * const * const svp = av_fetch(MUTABLE_AV(sv), SvIV(cSVOPx_sv(kid)), FALSE);
12842 if (!svp || *svp != uninit_sv)
12846 if (obase->op_type == OP_HELEM)
12847 return varname(gv, '%', o->op_targ,
12848 cSVOPx_sv(kid), 0, FUV_SUBSCRIPT_HASH);
12850 return varname(gv, '@', o->op_targ, NULL,
12851 SvIV(cSVOPx_sv(kid)), FUV_SUBSCRIPT_ARRAY);
12854 /* index is an expression;
12855 * attempt to find a match within the aggregate */
12856 if (obase->op_type == OP_HELEM) {
12857 SV * const keysv = find_hash_subscript((const HV*)sv, uninit_sv);
12859 return varname(gv, '%', o->op_targ,
12860 keysv, 0, FUV_SUBSCRIPT_HASH);
12864 = find_array_subscript((const AV *)sv, uninit_sv);
12866 return varname(gv, '@', o->op_targ,
12867 NULL, index, FUV_SUBSCRIPT_ARRAY);
12872 (o->op_type == OP_PADAV || o->op_type == OP_RV2AV)
12874 o->op_targ, NULL, 0, FUV_SUBSCRIPT_WITHIN);
12879 /* only examine RHS */
12880 return find_uninit_var(cBINOPx(obase)->op_first, uninit_sv, match);
12883 o = cUNOPx(obase)->op_first;
12884 if (o->op_type == OP_PUSHMARK)
12887 if (!o->op_sibling) {
12888 /* one-arg version of open is highly magical */
12890 if (o->op_type == OP_GV) { /* open FOO; */
12892 if (match && GvSV(gv) != uninit_sv)
12894 return varname(gv, '$', 0,
12895 NULL, 0, FUV_SUBSCRIPT_NONE);
12897 /* other possibilities not handled are:
12898 * open $x; or open my $x; should return '${*$x}'
12899 * open expr; should return '$'.expr ideally
12905 /* ops where $_ may be an implicit arg */
12909 if ( !(obase->op_flags & OPf_STACKED)) {
12910 if (uninit_sv == ((obase->op_private & OPpTARGET_MY)
12911 ? PAD_SVl(obase->op_targ)
12914 sv = sv_newmortal();
12915 sv_setpvs(sv, "$_");
12924 match = 1; /* print etc can return undef on defined args */
12925 /* skip filehandle as it can't produce 'undef' warning */
12926 o = cUNOPx(obase)->op_first;
12927 if ((obase->op_flags & OPf_STACKED) && o->op_type == OP_PUSHMARK)
12928 o = o->op_sibling->op_sibling;
12932 case OP_ENTEREVAL: /* could be eval $undef or $x='$undef'; eval $x */
12934 case OP_CUSTOM: /* XS or custom code could trigger random warnings */
12936 /* the following ops are capable of returning PL_sv_undef even for
12937 * defined arg(s) */
12956 case OP_GETPEERNAME:
13004 case OP_SMARTMATCH:
13013 /* XXX tmp hack: these two may call an XS sub, and currently
13014 XS subs don't have a SUB entry on the context stack, so CV and
13015 pad determination goes wrong, and BAD things happen. So, just
13016 don't try to determine the value under those circumstances.
13017 Need a better fix at dome point. DAPM 11/2007 */
13023 GV * const gv = gv_fetchpvs(".", GV_NOTQUAL, SVt_PV);
13024 if (gv && GvSV(gv) == uninit_sv)
13025 return newSVpvs_flags("$.", SVs_TEMP);
13030 /* def-ness of rval pos() is independent of the def-ness of its arg */
13031 if ( !(obase->op_flags & OPf_MOD))
13036 if (SvROK(PL_rs) && uninit_sv == SvRV(PL_rs))
13037 return newSVpvs_flags("${$/}", SVs_TEMP);
13042 if (!(obase->op_flags & OPf_KIDS))
13044 o = cUNOPx(obase)->op_first;
13050 /* if all except one arg are constant, or have no side-effects,
13051 * or are optimized away, then it's unambiguous */
13053 for (kid=o; kid; kid = kid->op_sibling) {
13055 const OPCODE type = kid->op_type;
13056 if ( (type == OP_CONST && SvOK(cSVOPx_sv(kid)))
13057 || (type == OP_NULL && ! (kid->op_flags & OPf_KIDS))
13058 || (type == OP_PUSHMARK)
13062 if (o2) { /* more than one found */
13069 return find_uninit_var(o2, uninit_sv, match);
13071 /* scan all args */
13073 sv = find_uninit_var(o, uninit_sv, 1);
13085 =for apidoc report_uninit
13087 Print appropriate "Use of uninitialized variable" warning
13093 Perl_report_uninit(pTHX_ const SV *uninit_sv)
13097 SV* varname = NULL;
13099 varname = find_uninit_var(PL_op, uninit_sv,0);
13101 sv_insert(varname, 0, 0, " ", 1);
13103 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
13104 varname ? SvPV_nolen_const(varname) : "",
13105 " in ", OP_DESC(PL_op));
13108 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
13114 * c-indentation-style: bsd
13115 * c-basic-offset: 4
13116 * indent-tabs-mode: t
13119 * ex: set ts=8 sts=4 sw=4 noet: