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 dont 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 *_allocated size
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 bit 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. For
800 example, if xpv_allocated is the same structure as XPV then the two
801 OFFSETs sum to zero, and the pointer is unchanged. If the allocated
802 structure is smaller (no initial NV actually allocated) then the net
803 effect is to subtract the size of the NV from the pointer, to return a
804 new pointer as if an initial NV were actually allocated.
806 This is the same trick as was used for NV and IV bodies. Ironically it
807 doesn't need to be used for NV bodies any more, because NV is now at
808 the start of the structure. IV bodies don't need it either, because
809 they are no longer allocated.
811 In turn, the new_body_* allocators call S_new_body(), which invokes
812 new_body_inline macro, which takes a lock, and takes a body off the
813 linked list at PL_body_roots[sv_type], calling S_more_bodies() if
814 necessary to refresh an empty list. Then the lock is released, and
815 the body is returned.
817 S_more_bodies calls get_arena(), and carves it up into an array of N
818 bodies, which it strings into a linked list. It looks up arena-size
819 and body-size from the body_details table described below, thus
820 supporting the multiple body-types.
822 If PURIFY is defined, or PERL_ARENA_SIZE=0, arenas are not used, and
823 the (new|del)_X*V macros are mapped directly to malloc/free.
829 For each sv-type, struct body_details bodies_by_type[] carries
830 parameters which control these aspects of SV handling:
832 Arena_size determines whether arenas are used for this body type, and if
833 so, how big they are. PURIFY or PERL_ARENA_SIZE=0 set this field to
834 zero, forcing individual mallocs and frees.
836 Body_size determines how big a body is, and therefore how many fit into
837 each arena. Offset carries the body-pointer adjustment needed for
838 *_allocated body types, and is used in *_allocated macros.
840 But its main purpose is to parameterize info needed in
841 Perl_sv_upgrade(). The info here dramatically simplifies the function
842 vs the implementation in 5.8.7, making it table-driven. All fields
843 are used for this, except for arena_size.
845 For the sv-types that have no bodies, arenas are not used, so those
846 PL_body_roots[sv_type] are unused, and can be overloaded. In
847 something of a special case, SVt_NULL is borrowed for HE arenas;
848 PL_body_roots[HE_SVSLOT=SVt_NULL] is filled by S_more_he, but the
849 bodies_by_type[SVt_NULL] slot is not used, as the table is not
852 PTEs also use arenas, but are never seen in Perl_sv_upgrade. Nonetheless,
853 they get their own slot in bodies_by_type[PTE_SVSLOT =SVt_IV], so they can
854 just use the same allocation semantics. At first, PTEs were also
855 overloaded to a non-body sv-type, but this yielded hard-to-find malloc
856 bugs, so was simplified by claiming a new slot. This choice has no
857 consequence at this time.
861 struct body_details {
862 U8 body_size; /* Size to allocate */
863 U8 copy; /* Size of structure to copy (may be shorter) */
865 unsigned int type : 4; /* We have space for a sanity check. */
866 unsigned int cant_upgrade : 1; /* Cannot upgrade this type */
867 unsigned int zero_nv : 1; /* zero the NV when upgrading from this */
868 unsigned int arena : 1; /* Allocated from an arena */
869 size_t arena_size; /* Size of arena to allocate */
877 /* With -DPURFIY we allocate everything directly, and don't use arenas.
878 This seems a rather elegant way to simplify some of the code below. */
879 #define HASARENA FALSE
881 #define HASARENA TRUE
883 #define NOARENA FALSE
885 /* Size the arenas to exactly fit a given number of bodies. A count
886 of 0 fits the max number bodies into a PERL_ARENA_SIZE.block,
887 simplifying the default. If count > 0, the arena is sized to fit
888 only that many bodies, allowing arenas to be used for large, rare
889 bodies (XPVFM, XPVIO) without undue waste. The arena size is
890 limited by PERL_ARENA_SIZE, so we can safely oversize the
893 #define FIT_ARENA0(body_size) \
894 ((size_t)(PERL_ARENA_SIZE / body_size) * body_size)
895 #define FIT_ARENAn(count,body_size) \
896 ( count * body_size <= PERL_ARENA_SIZE) \
897 ? count * body_size \
898 : FIT_ARENA0 (body_size)
899 #define FIT_ARENA(count,body_size) \
901 ? FIT_ARENAn (count, body_size) \
902 : FIT_ARENA0 (body_size)
904 /* A macro to work out the offset needed to subtract from a pointer to (say)
911 to make its members accessible via a pointer to (say)
921 #define relative_STRUCT_OFFSET(longer, shorter, member) \
922 (STRUCT_OFFSET(shorter, member) - STRUCT_OFFSET(longer, member))
924 /* Calculate the length to copy. Specifically work out the length less any
925 final padding the compiler needed to add. See the comment in sv_upgrade
926 for why copying the padding proved to be a bug. */
928 #define copy_length(type, last_member) \
929 STRUCT_OFFSET(type, last_member) \
930 + sizeof (((type*)SvANY((const SV *)0))->last_member)
932 static const struct body_details bodies_by_type[] = {
933 { sizeof(HE), 0, 0, SVt_NULL,
934 FALSE, NONV, NOARENA, FIT_ARENA(0, sizeof(HE)) },
936 /* The bind placeholder pretends to be an RV for now.
937 Also it's marked as "can't upgrade" to stop anyone using it before it's
939 { 0, 0, 0, SVt_BIND, TRUE, NONV, NOARENA, 0 },
941 /* IVs are in the head, so the allocation size is 0.
942 However, the slot is overloaded for PTEs. */
943 { sizeof(struct ptr_tbl_ent), /* This is used for PTEs. */
944 sizeof(IV), /* This is used to copy out the IV body. */
945 STRUCT_OFFSET(XPVIV, xiv_iv), SVt_IV, FALSE, NONV,
946 NOARENA /* IVS don't need an arena */,
947 /* But PTEs need to know the size of their arena */
948 FIT_ARENA(0, sizeof(struct ptr_tbl_ent))
951 /* 8 bytes on most ILP32 with IEEE doubles */
952 { sizeof(NV), sizeof(NV), 0, SVt_NV, FALSE, HADNV, HASARENA,
953 FIT_ARENA(0, sizeof(NV)) },
955 /* 8 bytes on most ILP32 with IEEE doubles */
956 { sizeof(xpv_allocated),
957 copy_length(XPV, xpv_len)
958 - relative_STRUCT_OFFSET(xpv_allocated, XPV, xpv_cur),
959 + relative_STRUCT_OFFSET(xpv_allocated, XPV, xpv_cur),
960 SVt_PV, FALSE, NONV, HASARENA, FIT_ARENA(0, sizeof(xpv_allocated)) },
963 { sizeof(xpviv_allocated),
964 copy_length(XPVIV, xiv_u)
965 - relative_STRUCT_OFFSET(xpviv_allocated, XPVIV, xpv_cur),
966 + relative_STRUCT_OFFSET(xpviv_allocated, XPVIV, xpv_cur),
967 SVt_PVIV, FALSE, NONV, HASARENA, FIT_ARENA(0, sizeof(xpviv_allocated)) },
970 { sizeof(XPVNV), copy_length(XPVNV, xiv_u), 0, SVt_PVNV, FALSE, HADNV,
971 HASARENA, FIT_ARENA(0, sizeof(XPVNV)) },
974 { sizeof(XPVMG), copy_length(XPVMG, xmg_stash), 0, SVt_PVMG, FALSE, HADNV,
975 HASARENA, FIT_ARENA(0, sizeof(XPVMG)) },
978 { sizeof(struct regexp_allocated), sizeof(struct regexp_allocated),
979 + relative_STRUCT_OFFSET(struct regexp_allocated, regexp, xpv_cur),
980 SVt_REGEXP, FALSE, NONV, HASARENA,
981 FIT_ARENA(0, sizeof(struct regexp_allocated))
985 { sizeof(XPVGV), sizeof(XPVGV), 0, SVt_PVGV, TRUE, HADNV,
986 HASARENA, FIT_ARENA(0, sizeof(XPVGV)) },
989 { sizeof(XPVLV), sizeof(XPVLV), 0, SVt_PVLV, TRUE, HADNV,
990 HASARENA, FIT_ARENA(0, sizeof(XPVLV)) },
992 { sizeof(xpvav_allocated),
993 copy_length(XPVAV, xmg_stash)
994 - relative_STRUCT_OFFSET(xpvav_allocated, XPVAV, xav_fill),
995 + relative_STRUCT_OFFSET(xpvav_allocated, XPVAV, xav_fill),
996 SVt_PVAV, TRUE, NONV, HASARENA, FIT_ARENA(0, sizeof(xpvav_allocated)) },
998 { sizeof(xpvhv_allocated),
999 copy_length(XPVHV, xmg_stash)
1000 - relative_STRUCT_OFFSET(xpvhv_allocated, XPVHV, xhv_fill),
1001 + relative_STRUCT_OFFSET(xpvhv_allocated, XPVHV, xhv_fill),
1002 SVt_PVHV, TRUE, NONV, HASARENA, FIT_ARENA(0, sizeof(xpvhv_allocated)) },
1005 { sizeof(xpvcv_allocated), sizeof(xpvcv_allocated),
1006 + relative_STRUCT_OFFSET(xpvcv_allocated, XPVCV, xpv_cur),
1007 SVt_PVCV, TRUE, NONV, HASARENA, FIT_ARENA(0, sizeof(xpvcv_allocated)) },
1009 { sizeof(xpvfm_allocated), sizeof(xpvfm_allocated),
1010 + relative_STRUCT_OFFSET(xpvfm_allocated, XPVFM, xpv_cur),
1011 SVt_PVFM, TRUE, NONV, NOARENA, FIT_ARENA(20, sizeof(xpvfm_allocated)) },
1013 /* XPVIO is 84 bytes, fits 48x */
1014 { sizeof(xpvio_allocated), sizeof(xpvio_allocated),
1015 + relative_STRUCT_OFFSET(xpvio_allocated, XPVIO, xpv_cur),
1016 SVt_PVIO, TRUE, NONV, HASARENA, FIT_ARENA(24, sizeof(xpvio_allocated)) },
1019 #define new_body_type(sv_type) \
1020 (void *)((char *)S_new_body(aTHX_ sv_type))
1022 #define del_body_type(p, sv_type) \
1023 del_body(p, &PL_body_roots[sv_type])
1026 #define new_body_allocated(sv_type) \
1027 (void *)((char *)S_new_body(aTHX_ sv_type) \
1028 - bodies_by_type[sv_type].offset)
1030 #define del_body_allocated(p, sv_type) \
1031 del_body(p + bodies_by_type[sv_type].offset, &PL_body_roots[sv_type])
1034 #define my_safemalloc(s) (void*)safemalloc(s)
1035 #define my_safecalloc(s) (void*)safecalloc(s, 1)
1036 #define my_safefree(p) safefree((char*)p)
1040 #define new_XNV() my_safemalloc(sizeof(XPVNV))
1041 #define del_XNV(p) my_safefree(p)
1043 #define new_XPVNV() my_safemalloc(sizeof(XPVNV))
1044 #define del_XPVNV(p) my_safefree(p)
1046 #define new_XPVAV() my_safemalloc(sizeof(XPVAV))
1047 #define del_XPVAV(p) my_safefree(p)
1049 #define new_XPVHV() my_safemalloc(sizeof(XPVHV))
1050 #define del_XPVHV(p) my_safefree(p)
1052 #define new_XPVMG() my_safemalloc(sizeof(XPVMG))
1053 #define del_XPVMG(p) my_safefree(p)
1055 #define new_XPVGV() my_safemalloc(sizeof(XPVGV))
1056 #define del_XPVGV(p) my_safefree(p)
1060 #define new_XNV() new_body_type(SVt_NV)
1061 #define del_XNV(p) del_body_type(p, SVt_NV)
1063 #define new_XPVNV() new_body_type(SVt_PVNV)
1064 #define del_XPVNV(p) del_body_type(p, SVt_PVNV)
1066 #define new_XPVAV() new_body_allocated(SVt_PVAV)
1067 #define del_XPVAV(p) del_body_allocated(p, SVt_PVAV)
1069 #define new_XPVHV() new_body_allocated(SVt_PVHV)
1070 #define del_XPVHV(p) del_body_allocated(p, SVt_PVHV)
1072 #define new_XPVMG() new_body_type(SVt_PVMG)
1073 #define del_XPVMG(p) del_body_type(p, SVt_PVMG)
1075 #define new_XPVGV() new_body_type(SVt_PVGV)
1076 #define del_XPVGV(p) del_body_type(p, SVt_PVGV)
1080 /* no arena for you! */
1082 #define new_NOARENA(details) \
1083 my_safemalloc((details)->body_size + (details)->offset)
1084 #define new_NOARENAZ(details) \
1085 my_safecalloc((details)->body_size + (details)->offset)
1088 S_more_bodies (pTHX_ const svtype sv_type)
1091 void ** const root = &PL_body_roots[sv_type];
1092 const struct body_details * const bdp = &bodies_by_type[sv_type];
1093 const size_t body_size = bdp->body_size;
1096 const size_t arena_size = Perl_malloc_good_size(bdp->arena_size);
1097 #if defined(DEBUGGING) && !defined(PERL_GLOBAL_STRUCT_PRIVATE)
1098 static bool done_sanity_check;
1100 /* PERL_GLOBAL_STRUCT_PRIVATE cannot coexist with global
1101 * variables like done_sanity_check. */
1102 if (!done_sanity_check) {
1103 unsigned int i = SVt_LAST;
1105 done_sanity_check = TRUE;
1108 assert (bodies_by_type[i].type == i);
1112 assert(bdp->arena_size);
1114 start = (char*) Perl_get_arena(aTHX_ arena_size, sv_type);
1116 end = start + arena_size - 2 * body_size;
1118 /* computed count doesnt reflect the 1st slot reservation */
1119 #if defined(MYMALLOC) || defined(HAS_MALLOC_GOOD_SIZE)
1120 DEBUG_m(PerlIO_printf(Perl_debug_log,
1121 "arena %p end %p arena-size %d (from %d) type %d "
1123 (void*)start, (void*)end, (int)arena_size,
1124 (int)bdp->arena_size, sv_type, (int)body_size,
1125 (int)arena_size / (int)body_size));
1127 DEBUG_m(PerlIO_printf(Perl_debug_log,
1128 "arena %p end %p arena-size %d type %d size %d ct %d\n",
1129 (void*)start, (void*)end,
1130 (int)bdp->arena_size, sv_type, (int)body_size,
1131 (int)bdp->arena_size / (int)body_size));
1133 *root = (void *)start;
1135 while (start <= end) {
1136 char * const next = start + body_size;
1137 *(void**) start = (void *)next;
1140 *(void **)start = 0;
1145 /* grab a new thing from the free list, allocating more if necessary.
1146 The inline version is used for speed in hot routines, and the
1147 function using it serves the rest (unless PURIFY).
1149 #define new_body_inline(xpv, sv_type) \
1151 void ** const r3wt = &PL_body_roots[sv_type]; \
1152 xpv = (PTR_TBL_ENT_t*) (*((void **)(r3wt)) \
1153 ? *((void **)(r3wt)) : more_bodies(sv_type)); \
1154 *(r3wt) = *(void**)(xpv); \
1160 S_new_body(pTHX_ const svtype sv_type)
1164 new_body_inline(xpv, sv_type);
1170 static const struct body_details fake_rv =
1171 { 0, 0, 0, SVt_IV, FALSE, NONV, NOARENA, 0 };
1174 =for apidoc sv_upgrade
1176 Upgrade an SV to a more complex form. Generally adds a new body type to the
1177 SV, then copies across as much information as possible from the old body.
1178 You generally want to use the C<SvUPGRADE> macro wrapper. See also C<svtype>.
1184 Perl_sv_upgrade(pTHX_ register SV *const sv, svtype new_type)
1189 const svtype old_type = SvTYPE(sv);
1190 const struct body_details *new_type_details;
1191 const struct body_details *old_type_details
1192 = bodies_by_type + old_type;
1193 SV *referant = NULL;
1195 PERL_ARGS_ASSERT_SV_UPGRADE;
1197 if (new_type != SVt_PV && SvIsCOW(sv)) {
1198 sv_force_normal_flags(sv, 0);
1201 if (old_type == new_type)
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 IoPAGE_LEN(sv) = 60;
1435 if (old_type < SVt_PV) {
1436 /* referant will be NULL unless the old type was SVt_IV emulating
1438 sv->sv_u.svu_rv = referant;
1442 Perl_croak(aTHX_ "panic: sv_upgrade to unknown type %lu",
1443 (unsigned long)new_type);
1446 if (old_type_details->arena) {
1447 /* If there was an old body, then we need to free it.
1448 Note that there is an assumption that all bodies of types that
1449 can be upgraded came from arenas. Only the more complex non-
1450 upgradable types are allowed to be directly malloc()ed. */
1452 my_safefree(old_body);
1454 del_body((void*)((char*)old_body + old_type_details->offset),
1455 &PL_body_roots[old_type]);
1461 =for apidoc sv_backoff
1463 Remove any string offset. You should normally use the C<SvOOK_off> macro
1470 Perl_sv_backoff(pTHX_ register SV *const sv)
1473 const char * const s = SvPVX_const(sv);
1475 PERL_ARGS_ASSERT_SV_BACKOFF;
1476 PERL_UNUSED_CONTEXT;
1479 assert(SvTYPE(sv) != SVt_PVHV);
1480 assert(SvTYPE(sv) != SVt_PVAV);
1482 SvOOK_offset(sv, delta);
1484 SvLEN_set(sv, SvLEN(sv) + delta);
1485 SvPV_set(sv, SvPVX(sv) - delta);
1486 Move(s, SvPVX(sv), SvCUR(sv)+1, char);
1487 SvFLAGS(sv) &= ~SVf_OOK;
1494 Expands the character buffer in the SV. If necessary, uses C<sv_unref> and
1495 upgrades the SV to C<SVt_PV>. Returns a pointer to the character buffer.
1496 Use the C<SvGROW> wrapper instead.
1502 Perl_sv_grow(pTHX_ register SV *const sv, register STRLEN newlen)
1506 PERL_ARGS_ASSERT_SV_GROW;
1508 if (PL_madskills && newlen >= 0x100000) {
1509 PerlIO_printf(Perl_debug_log,
1510 "Allocation too large: %"UVxf"\n", (UV)newlen);
1512 #ifdef HAS_64K_LIMIT
1513 if (newlen >= 0x10000) {
1514 PerlIO_printf(Perl_debug_log,
1515 "Allocation too large: %"UVxf"\n", (UV)newlen);
1518 #endif /* HAS_64K_LIMIT */
1521 if (SvTYPE(sv) < SVt_PV) {
1522 sv_upgrade(sv, SVt_PV);
1523 s = SvPVX_mutable(sv);
1525 else if (SvOOK(sv)) { /* pv is offset? */
1527 s = SvPVX_mutable(sv);
1528 if (newlen > SvLEN(sv))
1529 newlen += 10 * (newlen - SvCUR(sv)); /* avoid copy each time */
1530 #ifdef HAS_64K_LIMIT
1531 if (newlen >= 0x10000)
1536 s = SvPVX_mutable(sv);
1538 if (newlen > SvLEN(sv)) { /* need more room? */
1539 #ifndef Perl_safesysmalloc_size
1540 newlen = PERL_STRLEN_ROUNDUP(newlen);
1542 if (SvLEN(sv) && s) {
1543 s = (char*)saferealloc(s, newlen);
1546 s = (char*)safemalloc(newlen);
1547 if (SvPVX_const(sv) && SvCUR(sv)) {
1548 Move(SvPVX_const(sv), s, (newlen < SvCUR(sv)) ? newlen : SvCUR(sv), char);
1552 #ifdef Perl_safesysmalloc_size
1553 /* Do this here, do it once, do it right, and then we will never get
1554 called back into sv_grow() unless there really is some growing
1556 SvLEN_set(sv, Perl_safesysmalloc_size(s));
1558 SvLEN_set(sv, newlen);
1565 =for apidoc sv_setiv
1567 Copies an integer into the given SV, upgrading first if necessary.
1568 Does not handle 'set' magic. See also C<sv_setiv_mg>.
1574 Perl_sv_setiv(pTHX_ register SV *const sv, const IV i)
1578 PERL_ARGS_ASSERT_SV_SETIV;
1580 SV_CHECK_THINKFIRST_COW_DROP(sv);
1581 switch (SvTYPE(sv)) {
1584 sv_upgrade(sv, SVt_IV);
1587 sv_upgrade(sv, SVt_PVIV);
1591 if (!isGV_with_GP(sv))
1598 Perl_croak(aTHX_ "Can't coerce %s to integer in %s", sv_reftype(sv,0),
1602 (void)SvIOK_only(sv); /* validate number */
1608 =for apidoc sv_setiv_mg
1610 Like C<sv_setiv>, but also handles 'set' magic.
1616 Perl_sv_setiv_mg(pTHX_ register SV *const sv, const IV i)
1618 PERL_ARGS_ASSERT_SV_SETIV_MG;
1625 =for apidoc sv_setuv
1627 Copies an unsigned integer into the given SV, upgrading first if necessary.
1628 Does not handle 'set' magic. See also C<sv_setuv_mg>.
1634 Perl_sv_setuv(pTHX_ register SV *const sv, const UV u)
1636 PERL_ARGS_ASSERT_SV_SETUV;
1638 /* With these two if statements:
1639 u=1.49 s=0.52 cu=72.49 cs=10.64 scripts=270 tests=20865
1642 u=1.35 s=0.47 cu=73.45 cs=11.43 scripts=270 tests=20865
1644 If you wish to remove them, please benchmark to see what the effect is
1646 if (u <= (UV)IV_MAX) {
1647 sv_setiv(sv, (IV)u);
1656 =for apidoc sv_setuv_mg
1658 Like C<sv_setuv>, but also handles 'set' magic.
1664 Perl_sv_setuv_mg(pTHX_ register SV *const sv, const UV u)
1666 PERL_ARGS_ASSERT_SV_SETUV_MG;
1673 =for apidoc sv_setnv
1675 Copies a double into the given SV, upgrading first if necessary.
1676 Does not handle 'set' magic. See also C<sv_setnv_mg>.
1682 Perl_sv_setnv(pTHX_ register SV *const sv, const NV num)
1686 PERL_ARGS_ASSERT_SV_SETNV;
1688 SV_CHECK_THINKFIRST_COW_DROP(sv);
1689 switch (SvTYPE(sv)) {
1692 sv_upgrade(sv, SVt_NV);
1696 sv_upgrade(sv, SVt_PVNV);
1700 if (!isGV_with_GP(sv))
1707 Perl_croak(aTHX_ "Can't coerce %s to number in %s", sv_reftype(sv,0),
1712 (void)SvNOK_only(sv); /* validate number */
1717 =for apidoc sv_setnv_mg
1719 Like C<sv_setnv>, but also handles 'set' magic.
1725 Perl_sv_setnv_mg(pTHX_ register SV *const sv, const NV num)
1727 PERL_ARGS_ASSERT_SV_SETNV_MG;
1733 /* Print an "isn't numeric" warning, using a cleaned-up,
1734 * printable version of the offending string
1738 S_not_a_number(pTHX_ SV *const sv)
1745 PERL_ARGS_ASSERT_NOT_A_NUMBER;
1748 dsv = newSVpvs_flags("", SVs_TEMP);
1749 pv = sv_uni_display(dsv, sv, 10, 0);
1752 const char * const limit = tmpbuf + sizeof(tmpbuf) - 8;
1753 /* each *s can expand to 4 chars + "...\0",
1754 i.e. need room for 8 chars */
1756 const char *s = SvPVX_const(sv);
1757 const char * const end = s + SvCUR(sv);
1758 for ( ; s < end && d < limit; s++ ) {
1760 if (ch & 128 && !isPRINT_LC(ch)) {
1769 else if (ch == '\r') {
1773 else if (ch == '\f') {
1777 else if (ch == '\\') {
1781 else if (ch == '\0') {
1785 else if (isPRINT_LC(ch))
1802 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
1803 "Argument \"%s\" isn't numeric in %s", pv,
1806 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
1807 "Argument \"%s\" isn't numeric", pv);
1811 =for apidoc looks_like_number
1813 Test if the content of an SV looks like a number (or is a number).
1814 C<Inf> and C<Infinity> are treated as numbers (so will not issue a
1815 non-numeric warning), even if your atof() doesn't grok them.
1821 Perl_looks_like_number(pTHX_ SV *const sv)
1823 register const char *sbegin;
1826 PERL_ARGS_ASSERT_LOOKS_LIKE_NUMBER;
1829 sbegin = SvPVX_const(sv);
1832 else if (SvPOKp(sv))
1833 sbegin = SvPV_const(sv, len);
1835 return SvFLAGS(sv) & (SVf_NOK|SVp_NOK|SVf_IOK|SVp_IOK);
1836 return grok_number(sbegin, len, NULL);
1840 S_glob_2number(pTHX_ GV * const gv)
1842 const U32 wasfake = SvFLAGS(gv) & SVf_FAKE;
1843 SV *const buffer = sv_newmortal();
1845 PERL_ARGS_ASSERT_GLOB_2NUMBER;
1847 /* FAKE globs can get coerced, so need to turn this off temporarily if it
1850 gv_efullname3(buffer, gv, "*");
1851 SvFLAGS(gv) |= wasfake;
1853 /* We know that all GVs stringify to something that is not-a-number,
1854 so no need to test that. */
1855 if (ckWARN(WARN_NUMERIC))
1856 not_a_number(buffer);
1857 /* We just want something true to return, so that S_sv_2iuv_common
1858 can tail call us and return true. */
1862 /* Actually, ISO C leaves conversion of UV to IV undefined, but
1863 until proven guilty, assume that things are not that bad... */
1868 As 64 bit platforms often have an NV that doesn't preserve all bits of
1869 an IV (an assumption perl has been based on to date) it becomes necessary
1870 to remove the assumption that the NV always carries enough precision to
1871 recreate the IV whenever needed, and that the NV is the canonical form.
1872 Instead, IV/UV and NV need to be given equal rights. So as to not lose
1873 precision as a side effect of conversion (which would lead to insanity
1874 and the dragon(s) in t/op/numconvert.t getting very angry) the intent is
1875 1) to distinguish between IV/UV/NV slots that have cached a valid
1876 conversion where precision was lost and IV/UV/NV slots that have a
1877 valid conversion which has lost no precision
1878 2) to ensure that if a numeric conversion to one form is requested that
1879 would lose precision, the precise conversion (or differently
1880 imprecise conversion) is also performed and cached, to prevent
1881 requests for different numeric formats on the same SV causing
1882 lossy conversion chains. (lossless conversion chains are perfectly
1887 SvIOKp is true if the IV slot contains a valid value
1888 SvIOK is true only if the IV value is accurate (UV if SvIOK_UV true)
1889 SvNOKp is true if the NV slot contains a valid value
1890 SvNOK is true only if the NV value is accurate
1893 while converting from PV to NV, check to see if converting that NV to an
1894 IV(or UV) would lose accuracy over a direct conversion from PV to
1895 IV(or UV). If it would, cache both conversions, return NV, but mark
1896 SV as IOK NOKp (ie not NOK).
1898 While converting from PV to IV, check to see if converting that IV to an
1899 NV would lose accuracy over a direct conversion from PV to NV. If it
1900 would, cache both conversions, flag similarly.
1902 Before, the SV value "3.2" could become NV=3.2 IV=3 NOK, IOK quite
1903 correctly because if IV & NV were set NV *always* overruled.
1904 Now, "3.2" will become NV=3.2 IV=3 NOK, IOKp, because the flag's meaning
1905 changes - now IV and NV together means that the two are interchangeable:
1906 SvIVX == (IV) SvNVX && SvNVX == (NV) SvIVX;
1908 The benefit of this is that operations such as pp_add know that if
1909 SvIOK is true for both left and right operands, then integer addition
1910 can be used instead of floating point (for cases where the result won't
1911 overflow). Before, floating point was always used, which could lead to
1912 loss of precision compared with integer addition.
1914 * making IV and NV equal status should make maths accurate on 64 bit
1916 * may speed up maths somewhat if pp_add and friends start to use
1917 integers when possible instead of fp. (Hopefully the overhead in
1918 looking for SvIOK and checking for overflow will not outweigh the
1919 fp to integer speedup)
1920 * will slow down integer operations (callers of SvIV) on "inaccurate"
1921 values, as the change from SvIOK to SvIOKp will cause a call into
1922 sv_2iv each time rather than a macro access direct to the IV slot
1923 * should speed up number->string conversion on integers as IV is
1924 favoured when IV and NV are equally accurate
1926 ####################################################################
1927 You had better be using SvIOK_notUV if you want an IV for arithmetic:
1928 SvIOK is true if (IV or UV), so you might be getting (IV)SvUV.
1929 On the other hand, SvUOK is true iff UV.
1930 ####################################################################
1932 Your mileage will vary depending your CPU's relative fp to integer
1936 #ifndef NV_PRESERVES_UV
1937 # define IS_NUMBER_UNDERFLOW_IV 1
1938 # define IS_NUMBER_UNDERFLOW_UV 2
1939 # define IS_NUMBER_IV_AND_UV 2
1940 # define IS_NUMBER_OVERFLOW_IV 4
1941 # define IS_NUMBER_OVERFLOW_UV 5
1943 /* sv_2iuv_non_preserve(): private routine for use by sv_2iv() and sv_2uv() */
1945 /* For sv_2nv these three cases are "SvNOK and don't bother casting" */
1947 S_sv_2iuv_non_preserve(pTHX_ register SV *const sv
1955 PERL_ARGS_ASSERT_SV_2IUV_NON_PRESERVE;
1957 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));
1958 if (SvNVX(sv) < (NV)IV_MIN) {
1959 (void)SvIOKp_on(sv);
1961 SvIV_set(sv, IV_MIN);
1962 return IS_NUMBER_UNDERFLOW_IV;
1964 if (SvNVX(sv) > (NV)UV_MAX) {
1965 (void)SvIOKp_on(sv);
1968 SvUV_set(sv, UV_MAX);
1969 return IS_NUMBER_OVERFLOW_UV;
1971 (void)SvIOKp_on(sv);
1973 /* Can't use strtol etc to convert this string. (See truth table in
1975 if (SvNVX(sv) <= (UV)IV_MAX) {
1976 SvIV_set(sv, I_V(SvNVX(sv)));
1977 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
1978 SvIOK_on(sv); /* Integer is precise. NOK, IOK */
1980 /* Integer is imprecise. NOK, IOKp */
1982 return SvNVX(sv) < 0 ? IS_NUMBER_UNDERFLOW_UV : IS_NUMBER_IV_AND_UV;
1985 SvUV_set(sv, U_V(SvNVX(sv)));
1986 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
1987 if (SvUVX(sv) == UV_MAX) {
1988 /* As we know that NVs don't preserve UVs, UV_MAX cannot
1989 possibly be preserved by NV. Hence, it must be overflow.
1991 return IS_NUMBER_OVERFLOW_UV;
1993 SvIOK_on(sv); /* Integer is precise. NOK, UOK */
1995 /* Integer is imprecise. NOK, IOKp */
1997 return IS_NUMBER_OVERFLOW_IV;
1999 #endif /* !NV_PRESERVES_UV*/
2002 S_sv_2iuv_common(pTHX_ SV *const sv)
2006 PERL_ARGS_ASSERT_SV_2IUV_COMMON;
2009 /* erm. not sure. *should* never get NOKp (without NOK) from sv_2nv
2010 * without also getting a cached IV/UV from it at the same time
2011 * (ie PV->NV conversion should detect loss of accuracy and cache
2012 * IV or UV at same time to avoid this. */
2013 /* IV-over-UV optimisation - choose to cache IV if possible */
2015 if (SvTYPE(sv) == SVt_NV)
2016 sv_upgrade(sv, SVt_PVNV);
2018 (void)SvIOKp_on(sv); /* Must do this first, to clear any SvOOK */
2019 /* < not <= as for NV doesn't preserve UV, ((NV)IV_MAX+1) will almost
2020 certainly cast into the IV range at IV_MAX, whereas the correct
2021 answer is the UV IV_MAX +1. Hence < ensures that dodgy boundary
2023 #if defined(NAN_COMPARE_BROKEN) && defined(Perl_isnan)
2024 if (Perl_isnan(SvNVX(sv))) {
2030 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2031 SvIV_set(sv, I_V(SvNVX(sv)));
2032 if (SvNVX(sv) == (NV) SvIVX(sv)
2033 #ifndef NV_PRESERVES_UV
2034 && (((UV)1 << NV_PRESERVES_UV_BITS) >
2035 (UV)(SvIVX(sv) > 0 ? SvIVX(sv) : -SvIVX(sv)))
2036 /* Don't flag it as "accurately an integer" if the number
2037 came from a (by definition imprecise) NV operation, and
2038 we're outside the range of NV integer precision */
2042 SvIOK_on(sv); /* Can this go wrong with rounding? NWC */
2044 /* scalar has trailing garbage, eg "42a" */
2046 DEBUG_c(PerlIO_printf(Perl_debug_log,
2047 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (precise)\n",
2053 /* IV not precise. No need to convert from PV, as NV
2054 conversion would already have cached IV if it detected
2055 that PV->IV would be better than PV->NV->IV
2056 flags already correct - don't set public IOK. */
2057 DEBUG_c(PerlIO_printf(Perl_debug_log,
2058 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (imprecise)\n",
2063 /* Can the above go wrong if SvIVX == IV_MIN and SvNVX < IV_MIN,
2064 but the cast (NV)IV_MIN rounds to a the value less (more
2065 negative) than IV_MIN which happens to be equal to SvNVX ??
2066 Analogous to 0xFFFFFFFFFFFFFFFF rounding up to NV (2**64) and
2067 NV rounding back to 0xFFFFFFFFFFFFFFFF, so UVX == UV(NVX) and
2068 (NV)UVX == NVX are both true, but the values differ. :-(
2069 Hopefully for 2s complement IV_MIN is something like
2070 0x8000000000000000 which will be exact. NWC */
2073 SvUV_set(sv, U_V(SvNVX(sv)));
2075 (SvNVX(sv) == (NV) SvUVX(sv))
2076 #ifndef NV_PRESERVES_UV
2077 /* Make sure it's not 0xFFFFFFFFFFFFFFFF */
2078 /*&& (SvUVX(sv) != UV_MAX) irrelevant with code below */
2079 && (((UV)1 << NV_PRESERVES_UV_BITS) > SvUVX(sv))
2080 /* Don't flag it as "accurately an integer" if the number
2081 came from a (by definition imprecise) NV operation, and
2082 we're outside the range of NV integer precision */
2088 DEBUG_c(PerlIO_printf(Perl_debug_log,
2089 "0x%"UVxf" 2iv(%"UVuf" => %"IVdf") (as unsigned)\n",
2095 else if (SvPOKp(sv) && SvLEN(sv)) {
2097 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2098 /* We want to avoid a possible problem when we cache an IV/ a UV which
2099 may be later translated to an NV, and the resulting NV is not
2100 the same as the direct translation of the initial string
2101 (eg 123.456 can shortcut to the IV 123 with atol(), but we must
2102 be careful to ensure that the value with the .456 is around if the
2103 NV value is requested in the future).
2105 This means that if we cache such an IV/a UV, we need to cache the
2106 NV as well. Moreover, we trade speed for space, and do not
2107 cache the NV if we are sure it's not needed.
2110 /* SVt_PVNV is one higher than SVt_PVIV, hence this order */
2111 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2112 == IS_NUMBER_IN_UV) {
2113 /* It's definitely an integer, only upgrade to PVIV */
2114 if (SvTYPE(sv) < SVt_PVIV)
2115 sv_upgrade(sv, SVt_PVIV);
2117 } else if (SvTYPE(sv) < SVt_PVNV)
2118 sv_upgrade(sv, SVt_PVNV);
2120 /* If NVs preserve UVs then we only use the UV value if we know that
2121 we aren't going to call atof() below. If NVs don't preserve UVs
2122 then the value returned may have more precision than atof() will
2123 return, even though value isn't perfectly accurate. */
2124 if ((numtype & (IS_NUMBER_IN_UV
2125 #ifdef NV_PRESERVES_UV
2128 )) == IS_NUMBER_IN_UV) {
2129 /* This won't turn off the public IOK flag if it was set above */
2130 (void)SvIOKp_on(sv);
2132 if (!(numtype & IS_NUMBER_NEG)) {
2134 if (value <= (UV)IV_MAX) {
2135 SvIV_set(sv, (IV)value);
2137 /* it didn't overflow, and it was positive. */
2138 SvUV_set(sv, value);
2142 /* 2s complement assumption */
2143 if (value <= (UV)IV_MIN) {
2144 SvIV_set(sv, -(IV)value);
2146 /* Too negative for an IV. This is a double upgrade, but
2147 I'm assuming it will be rare. */
2148 if (SvTYPE(sv) < SVt_PVNV)
2149 sv_upgrade(sv, SVt_PVNV);
2153 SvNV_set(sv, -(NV)value);
2154 SvIV_set(sv, IV_MIN);
2158 /* For !NV_PRESERVES_UV and IS_NUMBER_IN_UV and IS_NUMBER_NOT_INT we
2159 will be in the previous block to set the IV slot, and the next
2160 block to set the NV slot. So no else here. */
2162 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2163 != IS_NUMBER_IN_UV) {
2164 /* It wasn't an (integer that doesn't overflow the UV). */
2165 SvNV_set(sv, Atof(SvPVX_const(sv)));
2167 if (! numtype && ckWARN(WARN_NUMERIC))
2170 #if defined(USE_LONG_DOUBLE)
2171 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%" PERL_PRIgldbl ")\n",
2172 PTR2UV(sv), SvNVX(sv)));
2174 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"NVgf")\n",
2175 PTR2UV(sv), SvNVX(sv)));
2178 #ifdef NV_PRESERVES_UV
2179 (void)SvIOKp_on(sv);
2181 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2182 SvIV_set(sv, I_V(SvNVX(sv)));
2183 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
2186 NOOP; /* Integer is imprecise. NOK, IOKp */
2188 /* UV will not work better than IV */
2190 if (SvNVX(sv) > (NV)UV_MAX) {
2192 /* Integer is inaccurate. NOK, IOKp, is UV */
2193 SvUV_set(sv, UV_MAX);
2195 SvUV_set(sv, U_V(SvNVX(sv)));
2196 /* 0xFFFFFFFFFFFFFFFF not an issue in here, NVs
2197 NV preservse UV so can do correct comparison. */
2198 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
2201 NOOP; /* Integer is imprecise. NOK, IOKp, is UV */
2206 #else /* NV_PRESERVES_UV */
2207 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2208 == (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT)) {
2209 /* The IV/UV slot will have been set from value returned by
2210 grok_number above. The NV slot has just been set using
2213 assert (SvIOKp(sv));
2215 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2216 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2217 /* Small enough to preserve all bits. */
2218 (void)SvIOKp_on(sv);
2220 SvIV_set(sv, I_V(SvNVX(sv)));
2221 if ((NV)(SvIVX(sv)) == SvNVX(sv))
2223 /* Assumption: first non-preserved integer is < IV_MAX,
2224 this NV is in the preserved range, therefore: */
2225 if (!(U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))
2227 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);
2231 0 0 already failed to read UV.
2232 0 1 already failed to read UV.
2233 1 0 you won't get here in this case. IV/UV
2234 slot set, public IOK, Atof() unneeded.
2235 1 1 already read UV.
2236 so there's no point in sv_2iuv_non_preserve() attempting
2237 to use atol, strtol, strtoul etc. */
2239 sv_2iuv_non_preserve (sv, numtype);
2241 sv_2iuv_non_preserve (sv);
2245 #endif /* NV_PRESERVES_UV */
2246 /* It might be more code efficient to go through the entire logic above
2247 and conditionally set with SvIOKp_on() rather than SvIOK(), but it
2248 gets complex and potentially buggy, so more programmer efficient
2249 to do it this way, by turning off the public flags: */
2251 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK);
2255 if (isGV_with_GP(sv))
2256 return glob_2number(MUTABLE_GV(sv));
2258 if (!(SvFLAGS(sv) & SVs_PADTMP)) {
2259 if (!PL_localizing && ckWARN(WARN_UNINITIALIZED))
2262 if (SvTYPE(sv) < SVt_IV)
2263 /* Typically the caller expects that sv_any is not NULL now. */
2264 sv_upgrade(sv, SVt_IV);
2265 /* Return 0 from the caller. */
2272 =for apidoc sv_2iv_flags
2274 Return the integer value of an SV, doing any necessary string
2275 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2276 Normally used via the C<SvIV(sv)> and C<SvIVx(sv)> macros.
2282 Perl_sv_2iv_flags(pTHX_ register SV *const sv, const I32 flags)
2287 if (SvGMAGICAL(sv) || (SvTYPE(sv) == SVt_PVGV && SvVALID(sv))) {
2288 /* FBMs use the same flag bit as SVf_IVisUV, so must let them
2289 cache IVs just in case. In practice it seems that they never
2290 actually anywhere accessible by user Perl code, let alone get used
2291 in anything other than a string context. */
2292 if (flags & SV_GMAGIC)
2297 return I_V(SvNVX(sv));
2299 if (SvPOKp(sv) && SvLEN(sv)) {
2302 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2304 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2305 == IS_NUMBER_IN_UV) {
2306 /* It's definitely an integer */
2307 if (numtype & IS_NUMBER_NEG) {
2308 if (value < (UV)IV_MIN)
2311 if (value < (UV)IV_MAX)
2316 if (ckWARN(WARN_NUMERIC))
2319 return I_V(Atof(SvPVX_const(sv)));
2324 assert(SvTYPE(sv) >= SVt_PVMG);
2325 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
2326 } else if (SvTHINKFIRST(sv)) {
2330 SV * const tmpstr=AMG_CALLun(sv,numer);
2331 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2332 return SvIV(tmpstr);
2335 return PTR2IV(SvRV(sv));
2338 sv_force_normal_flags(sv, 0);
2340 if (SvREADONLY(sv) && !SvOK(sv)) {
2341 if (ckWARN(WARN_UNINITIALIZED))
2347 if (S_sv_2iuv_common(aTHX_ sv))
2350 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"IVdf")\n",
2351 PTR2UV(sv),SvIVX(sv)));
2352 return SvIsUV(sv) ? (IV)SvUVX(sv) : SvIVX(sv);
2356 =for apidoc sv_2uv_flags
2358 Return the unsigned integer value of an SV, doing any necessary string
2359 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2360 Normally used via the C<SvUV(sv)> and C<SvUVx(sv)> macros.
2366 Perl_sv_2uv_flags(pTHX_ register SV *const sv, const I32 flags)
2371 if (SvGMAGICAL(sv) || (SvTYPE(sv) == SVt_PVGV && SvVALID(sv))) {
2372 /* FBMs use the same flag bit as SVf_IVisUV, so must let them
2373 cache IVs just in case. */
2374 if (flags & SV_GMAGIC)
2379 return U_V(SvNVX(sv));
2380 if (SvPOKp(sv) && SvLEN(sv)) {
2383 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2385 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2386 == IS_NUMBER_IN_UV) {
2387 /* It's definitely an integer */
2388 if (!(numtype & IS_NUMBER_NEG))
2392 if (ckWARN(WARN_NUMERIC))
2395 return U_V(Atof(SvPVX_const(sv)));
2400 assert(SvTYPE(sv) >= SVt_PVMG);
2401 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
2402 } else if (SvTHINKFIRST(sv)) {
2406 SV *const tmpstr = AMG_CALLun(sv,numer);
2407 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2408 return SvUV(tmpstr);
2411 return PTR2UV(SvRV(sv));
2414 sv_force_normal_flags(sv, 0);
2416 if (SvREADONLY(sv) && !SvOK(sv)) {
2417 if (ckWARN(WARN_UNINITIALIZED))
2423 if (S_sv_2iuv_common(aTHX_ sv))
2427 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2uv(%"UVuf")\n",
2428 PTR2UV(sv),SvUVX(sv)));
2429 return SvIsUV(sv) ? SvUVX(sv) : (UV)SvIVX(sv);
2435 Return the num value of an SV, doing any necessary string or integer
2436 conversion, magic etc. Normally used via the C<SvNV(sv)> and C<SvNVx(sv)>
2443 Perl_sv_2nv(pTHX_ register SV *const sv)
2448 if (SvGMAGICAL(sv) || (SvTYPE(sv) == SVt_PVGV && SvVALID(sv))) {
2449 /* FBMs use the same flag bit as SVf_IVisUV, so must let them
2450 cache IVs just in case. */
2454 if ((SvPOKp(sv) && SvLEN(sv)) && !SvIOKp(sv)) {
2455 if (!SvIOKp(sv) && ckWARN(WARN_NUMERIC) &&
2456 !grok_number(SvPVX_const(sv), SvCUR(sv), NULL))
2458 return Atof(SvPVX_const(sv));
2462 return (NV)SvUVX(sv);
2464 return (NV)SvIVX(sv);
2469 assert(SvTYPE(sv) >= SVt_PVMG);
2470 /* This falls through to the report_uninit near the end of the
2472 } else if (SvTHINKFIRST(sv)) {
2476 SV *const tmpstr = AMG_CALLun(sv,numer);
2477 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2478 return SvNV(tmpstr);
2481 return PTR2NV(SvRV(sv));
2484 sv_force_normal_flags(sv, 0);
2486 if (SvREADONLY(sv) && !SvOK(sv)) {
2487 if (ckWARN(WARN_UNINITIALIZED))
2492 if (SvTYPE(sv) < SVt_NV) {
2493 /* The logic to use SVt_PVNV if necessary is in sv_upgrade. */
2494 sv_upgrade(sv, SVt_NV);
2495 #ifdef USE_LONG_DOUBLE
2497 STORE_NUMERIC_LOCAL_SET_STANDARD();
2498 PerlIO_printf(Perl_debug_log,
2499 "0x%"UVxf" num(%" PERL_PRIgldbl ")\n",
2500 PTR2UV(sv), SvNVX(sv));
2501 RESTORE_NUMERIC_LOCAL();
2505 STORE_NUMERIC_LOCAL_SET_STANDARD();
2506 PerlIO_printf(Perl_debug_log, "0x%"UVxf" num(%"NVgf")\n",
2507 PTR2UV(sv), SvNVX(sv));
2508 RESTORE_NUMERIC_LOCAL();
2512 else if (SvTYPE(sv) < SVt_PVNV)
2513 sv_upgrade(sv, SVt_PVNV);
2518 SvNV_set(sv, SvIsUV(sv) ? (NV)SvUVX(sv) : (NV)SvIVX(sv));
2519 #ifdef NV_PRESERVES_UV
2525 /* Only set the public NV OK flag if this NV preserves the IV */
2526 /* Check it's not 0xFFFFFFFFFFFFFFFF */
2528 SvIsUV(sv) ? ((SvUVX(sv) != UV_MAX)&&(SvUVX(sv) == U_V(SvNVX(sv))))
2529 : (SvIVX(sv) == I_V(SvNVX(sv))))
2535 else if (SvPOKp(sv) && SvLEN(sv)) {
2537 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2538 if (!SvIOKp(sv) && !numtype && ckWARN(WARN_NUMERIC))
2540 #ifdef NV_PRESERVES_UV
2541 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2542 == IS_NUMBER_IN_UV) {
2543 /* It's definitely an integer */
2544 SvNV_set(sv, (numtype & IS_NUMBER_NEG) ? -(NV)value : (NV)value);
2546 SvNV_set(sv, Atof(SvPVX_const(sv)));
2552 SvNV_set(sv, Atof(SvPVX_const(sv)));
2553 /* Only set the public NV OK flag if this NV preserves the value in
2554 the PV at least as well as an IV/UV would.
2555 Not sure how to do this 100% reliably. */
2556 /* if that shift count is out of range then Configure's test is
2557 wonky. We shouldn't be in here with NV_PRESERVES_UV_BITS ==
2559 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2560 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2561 SvNOK_on(sv); /* Definitely small enough to preserve all bits */
2562 } else if (!(numtype & IS_NUMBER_IN_UV)) {
2563 /* Can't use strtol etc to convert this string, so don't try.
2564 sv_2iv and sv_2uv will use the NV to convert, not the PV. */
2567 /* value has been set. It may not be precise. */
2568 if ((numtype & IS_NUMBER_NEG) && (value > (UV)IV_MIN)) {
2569 /* 2s complement assumption for (UV)IV_MIN */
2570 SvNOK_on(sv); /* Integer is too negative. */
2575 if (numtype & IS_NUMBER_NEG) {
2576 SvIV_set(sv, -(IV)value);
2577 } else if (value <= (UV)IV_MAX) {
2578 SvIV_set(sv, (IV)value);
2580 SvUV_set(sv, value);
2584 if (numtype & IS_NUMBER_NOT_INT) {
2585 /* I believe that even if the original PV had decimals,
2586 they are lost beyond the limit of the FP precision.
2587 However, neither is canonical, so both only get p
2588 flags. NWC, 2000/11/25 */
2589 /* Both already have p flags, so do nothing */
2591 const NV nv = SvNVX(sv);
2592 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2593 if (SvIVX(sv) == I_V(nv)) {
2596 /* It had no "." so it must be integer. */
2600 /* between IV_MAX and NV(UV_MAX).
2601 Could be slightly > UV_MAX */
2603 if (numtype & IS_NUMBER_NOT_INT) {
2604 /* UV and NV both imprecise. */
2606 const UV nv_as_uv = U_V(nv);
2608 if (value == nv_as_uv && SvUVX(sv) != UV_MAX) {
2617 /* It might be more code efficient to go through the entire logic above
2618 and conditionally set with SvNOKp_on() rather than SvNOK(), but it
2619 gets complex and potentially buggy, so more programmer efficient
2620 to do it this way, by turning off the public flags: */
2622 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK);
2623 #endif /* NV_PRESERVES_UV */
2626 if (isGV_with_GP(sv)) {
2627 glob_2number(MUTABLE_GV(sv));
2631 if (!PL_localizing && !(SvFLAGS(sv) & SVs_PADTMP) && ckWARN(WARN_UNINITIALIZED))
2633 assert (SvTYPE(sv) >= SVt_NV);
2634 /* Typically the caller expects that sv_any is not NULL now. */
2635 /* XXX Ilya implies that this is a bug in callers that assume this
2636 and ideally should be fixed. */
2639 #if defined(USE_LONG_DOUBLE)
2641 STORE_NUMERIC_LOCAL_SET_STANDARD();
2642 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2nv(%" PERL_PRIgldbl ")\n",
2643 PTR2UV(sv), SvNVX(sv));
2644 RESTORE_NUMERIC_LOCAL();
2648 STORE_NUMERIC_LOCAL_SET_STANDARD();
2649 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 1nv(%"NVgf")\n",
2650 PTR2UV(sv), SvNVX(sv));
2651 RESTORE_NUMERIC_LOCAL();
2660 Return an SV with the numeric value of the source SV, doing any necessary
2661 reference or overload conversion. You must use the C<SvNUM(sv)> macro to
2662 access this function.
2668 Perl_sv_2num(pTHX_ register SV *const sv)
2670 PERL_ARGS_ASSERT_SV_2NUM;
2675 SV * const tmpsv = AMG_CALLun(sv,numer);
2676 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
2677 return sv_2num(tmpsv);
2679 return sv_2mortal(newSVuv(PTR2UV(SvRV(sv))));
2682 /* uiv_2buf(): private routine for use by sv_2pv_flags(): print an IV or
2683 * UV as a string towards the end of buf, and return pointers to start and
2686 * We assume that buf is at least TYPE_CHARS(UV) long.
2690 S_uiv_2buf(char *const buf, const IV iv, UV uv, const int is_uv, char **const peob)
2692 char *ptr = buf + TYPE_CHARS(UV);
2693 char * const ebuf = ptr;
2696 PERL_ARGS_ASSERT_UIV_2BUF;
2708 *--ptr = '0' + (char)(uv % 10);
2717 =for apidoc sv_2pv_flags
2719 Returns a pointer to the string value of an SV, and sets *lp to its length.
2720 If flags includes SV_GMAGIC, does an mg_get() first. Coerces sv to a string
2722 Normally invoked via the C<SvPV_flags> macro. C<sv_2pv()> and C<sv_2pv_nomg>
2723 usually end up here too.
2729 Perl_sv_2pv_flags(pTHX_ register SV *const sv, STRLEN *const lp, const I32 flags)
2739 if (SvGMAGICAL(sv)) {
2740 if (flags & SV_GMAGIC)
2745 if (flags & SV_MUTABLE_RETURN)
2746 return SvPVX_mutable(sv);
2747 if (flags & SV_CONST_RETURN)
2748 return (char *)SvPVX_const(sv);
2751 if (SvIOKp(sv) || SvNOKp(sv)) {
2752 char tbuf[64]; /* Must fit sprintf/Gconvert of longest IV/NV */
2757 ? my_snprintf(tbuf, sizeof(tbuf), "%"UVuf, (UV)SvUVX(sv))
2758 : my_snprintf(tbuf, sizeof(tbuf), "%"IVdf, (IV)SvIVX(sv));
2760 Gconvert(SvNVX(sv), NV_DIG, 0, tbuf);
2767 #ifdef FIXNEGATIVEZERO
2768 if (len == 2 && tbuf[0] == '-' && tbuf[1] == '0') {
2774 SvUPGRADE(sv, SVt_PV);
2777 s = SvGROW_mutable(sv, len + 1);
2780 return (char*)memcpy(s, tbuf, len + 1);
2786 assert(SvTYPE(sv) >= SVt_PVMG);
2787 /* This falls through to the report_uninit near the end of the
2789 } else if (SvTHINKFIRST(sv)) {
2793 SV *const tmpstr = AMG_CALLun(sv,string);
2794 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2796 /* char *pv = lp ? SvPV(tmpstr, *lp) : SvPV_nolen(tmpstr);
2800 if ((SvFLAGS(tmpstr) & (SVf_POK)) == SVf_POK) {
2801 if (flags & SV_CONST_RETURN) {
2802 pv = (char *) SvPVX_const(tmpstr);
2804 pv = (flags & SV_MUTABLE_RETURN)
2805 ? SvPVX_mutable(tmpstr) : SvPVX(tmpstr);
2808 *lp = SvCUR(tmpstr);
2810 pv = sv_2pv_flags(tmpstr, lp, flags);
2823 SV *const referent = SvRV(sv);
2827 retval = buffer = savepvn("NULLREF", len);
2828 } else if (SvTYPE(referent) == SVt_REGEXP) {
2829 REGEXP * const re = (REGEXP *)MUTABLE_PTR(referent);
2834 /* If the regex is UTF-8 we want the containing scalar to
2835 have an UTF-8 flag too */
2841 if ((seen_evals = RX_SEEN_EVALS(re)))
2842 PL_reginterp_cnt += seen_evals;
2845 *lp = RX_WRAPLEN(re);
2847 return RX_WRAPPED(re);
2849 const char *const typestr = sv_reftype(referent, 0);
2850 const STRLEN typelen = strlen(typestr);
2851 UV addr = PTR2UV(referent);
2852 const char *stashname = NULL;
2853 STRLEN stashnamelen = 0; /* hush, gcc */
2854 const char *buffer_end;
2856 if (SvOBJECT(referent)) {
2857 const HEK *const name = HvNAME_HEK(SvSTASH(referent));
2860 stashname = HEK_KEY(name);
2861 stashnamelen = HEK_LEN(name);
2863 if (HEK_UTF8(name)) {
2869 stashname = "__ANON__";
2872 len = stashnamelen + 1 /* = */ + typelen + 3 /* (0x */
2873 + 2 * sizeof(UV) + 2 /* )\0 */;
2875 len = typelen + 3 /* (0x */
2876 + 2 * sizeof(UV) + 2 /* )\0 */;
2879 Newx(buffer, len, char);
2880 buffer_end = retval = buffer + len;
2882 /* Working backwards */
2886 *--retval = PL_hexdigit[addr & 15];
2887 } while (addr >>= 4);
2893 memcpy(retval, typestr, typelen);
2897 retval -= stashnamelen;
2898 memcpy(retval, stashname, stashnamelen);
2900 /* retval may not neccesarily have reached the start of the
2902 assert (retval >= buffer);
2904 len = buffer_end - retval - 1; /* -1 for that \0 */
2912 if (SvREADONLY(sv) && !SvOK(sv)) {
2915 if (flags & SV_UNDEF_RETURNS_NULL)
2917 if (ckWARN(WARN_UNINITIALIZED))
2922 if (SvIOK(sv) || ((SvIOKp(sv) && !SvNOKp(sv)))) {
2923 /* I'm assuming that if both IV and NV are equally valid then
2924 converting the IV is going to be more efficient */
2925 const U32 isUIOK = SvIsUV(sv);
2926 char buf[TYPE_CHARS(UV)];
2930 if (SvTYPE(sv) < SVt_PVIV)
2931 sv_upgrade(sv, SVt_PVIV);
2932 ptr = uiv_2buf(buf, SvIVX(sv), SvUVX(sv), isUIOK, &ebuf);
2934 /* inlined from sv_setpvn */
2935 s = SvGROW_mutable(sv, len + 1);
2936 Move(ptr, s, len, char);
2940 else if (SvNOKp(sv)) {
2942 if (SvTYPE(sv) < SVt_PVNV)
2943 sv_upgrade(sv, SVt_PVNV);
2944 /* The +20 is pure guesswork. Configure test needed. --jhi */
2945 s = SvGROW_mutable(sv, NV_DIG + 20);
2946 /* some Xenix systems wipe out errno here */
2948 if (SvNVX(sv) == 0.0)
2949 my_strlcpy(s, "0", SvLEN(sv));
2953 Gconvert(SvNVX(sv), NV_DIG, 0, s);
2956 #ifdef FIXNEGATIVEZERO
2957 if (*s == '-' && s[1] == '0' && !s[2]) {
2969 if (isGV_with_GP(sv)) {
2970 GV *const gv = MUTABLE_GV(sv);
2971 const U32 wasfake = SvFLAGS(gv) & SVf_FAKE;
2972 SV *const buffer = sv_newmortal();
2974 /* FAKE globs can get coerced, so need to turn this off temporarily
2977 gv_efullname3(buffer, gv, "*");
2978 SvFLAGS(gv) |= wasfake;
2980 assert(SvPOK(buffer));
2982 *lp = SvCUR(buffer);
2984 return SvPVX(buffer);
2989 if (flags & SV_UNDEF_RETURNS_NULL)
2991 if (!PL_localizing && !(SvFLAGS(sv) & SVs_PADTMP) && ckWARN(WARN_UNINITIALIZED))
2993 if (SvTYPE(sv) < SVt_PV)
2994 /* Typically the caller expects that sv_any is not NULL now. */
2995 sv_upgrade(sv, SVt_PV);
2999 const STRLEN len = s - SvPVX_const(sv);
3005 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
3006 PTR2UV(sv),SvPVX_const(sv)));
3007 if (flags & SV_CONST_RETURN)
3008 return (char *)SvPVX_const(sv);
3009 if (flags & SV_MUTABLE_RETURN)
3010 return SvPVX_mutable(sv);
3015 =for apidoc sv_copypv
3017 Copies a stringified representation of the source SV into the
3018 destination SV. Automatically performs any necessary mg_get and
3019 coercion of numeric values into strings. Guaranteed to preserve
3020 UTF8 flag even from overloaded objects. Similar in nature to
3021 sv_2pv[_flags] but operates directly on an SV instead of just the
3022 string. Mostly uses sv_2pv_flags to do its work, except when that
3023 would lose the UTF-8'ness of the PV.
3029 Perl_sv_copypv(pTHX_ SV *const dsv, register SV *const ssv)
3032 const char * const s = SvPV_const(ssv,len);
3034 PERL_ARGS_ASSERT_SV_COPYPV;
3036 sv_setpvn(dsv,s,len);
3044 =for apidoc sv_2pvbyte
3046 Return a pointer to the byte-encoded representation of the SV, and set *lp
3047 to its length. May cause the SV to be downgraded from UTF-8 as a
3050 Usually accessed via the C<SvPVbyte> macro.
3056 Perl_sv_2pvbyte(pTHX_ register SV *const sv, STRLEN *const lp)
3058 PERL_ARGS_ASSERT_SV_2PVBYTE;
3060 sv_utf8_downgrade(sv,0);
3061 return lp ? SvPV(sv,*lp) : SvPV_nolen(sv);
3065 =for apidoc sv_2pvutf8
3067 Return a pointer to the UTF-8-encoded representation of the SV, and set *lp
3068 to its length. May cause the SV to be upgraded to UTF-8 as a side-effect.
3070 Usually accessed via the C<SvPVutf8> macro.
3076 Perl_sv_2pvutf8(pTHX_ register SV *const sv, STRLEN *const lp)
3078 PERL_ARGS_ASSERT_SV_2PVUTF8;
3080 sv_utf8_upgrade(sv);
3081 return lp ? SvPV(sv,*lp) : SvPV_nolen(sv);
3086 =for apidoc sv_2bool
3088 This function is only called on magical items, and is only used by
3089 sv_true() or its macro equivalent.
3095 Perl_sv_2bool(pTHX_ register SV *const sv)
3099 PERL_ARGS_ASSERT_SV_2BOOL;
3107 SV * const tmpsv = AMG_CALLun(sv,bool_);
3108 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
3109 return (bool)SvTRUE(tmpsv);
3111 return SvRV(sv) != 0;
3114 register XPV* const Xpvtmp = (XPV*)SvANY(sv);
3116 (*sv->sv_u.svu_pv > '0' ||
3117 Xpvtmp->xpv_cur > 1 ||
3118 (Xpvtmp->xpv_cur && *sv->sv_u.svu_pv != '0')))
3125 return SvIVX(sv) != 0;
3128 return SvNVX(sv) != 0.0;
3130 if (isGV_with_GP(sv))
3140 =for apidoc sv_utf8_upgrade
3142 Converts the PV of an SV to its UTF-8-encoded form.
3143 Forces the SV to string form if it is not already.
3144 Will C<mg_get> on C<sv> if appropriate.
3145 Always sets the SvUTF8 flag to avoid future validity checks even
3146 if the whole string is the same in UTF-8 as not.
3147 Returns the number of bytes in the converted string
3149 This is not as a general purpose byte encoding to Unicode interface:
3150 use the Encode extension for that.
3152 =for apidoc sv_utf8_upgrade_nomg
3154 Like sv_utf8_upgrade, but doesn't do magic on C<sv>
3156 =for apidoc sv_utf8_upgrade_flags
3158 Converts the PV of an SV to its UTF-8-encoded form.
3159 Forces the SV to string form if it is not already.
3160 Always sets the SvUTF8 flag to avoid future validity checks even
3161 if all the bytes are invariant in UTF-8. If C<flags> has C<SV_GMAGIC> bit set,
3162 will C<mg_get> on C<sv> if appropriate, else not.
3163 Returns the number of bytes in the converted string
3164 C<sv_utf8_upgrade> and
3165 C<sv_utf8_upgrade_nomg> are implemented in terms of this function.
3167 This is not as a general purpose byte encoding to Unicode interface:
3168 use the Encode extension for that.
3172 The grow version is currently not externally documented. It adds a parameter,
3173 extra, which is the number of unused bytes the string of 'sv' is guaranteed to
3174 have free after it upon return. This allows the caller to reserve extra space
3175 that it intends to fill, to avoid extra grows.
3177 Also externally undocumented for the moment is the flag SV_FORCE_UTF8_UPGRADE,
3178 which can be used to tell this function to not first check to see if there are
3179 any characters that are different in UTF-8 (variant characters) which would
3180 force it to allocate a new string to sv, but to assume there are. Typically
3181 this flag is used by a routine that has already parsed the string to find that
3182 there are such characters, and passes this information on so that the work
3183 doesn't have to be repeated.
3185 (One might think that the calling routine could pass in the position of the
3186 first such variant, so it wouldn't have to be found again. But that is not the
3187 case, because typically when the caller is likely to use this flag, it won't be
3188 calling this routine unless it finds something that won't fit into a byte.
3189 Otherwise it tries to not upgrade and just use bytes. But some things that
3190 do fit into a byte are variants in utf8, and the caller may not have been
3191 keeping track of these.)
3193 If the routine itself changes the string, it adds a trailing NUL. Such a NUL
3194 isn't guaranteed due to having other routines do the work in some input cases,
3195 or if the input is already flagged as being in utf8.
3197 The speed of this could perhaps be improved for many cases if someone wanted to
3198 write a fast function that counts the number of variant characters in a string,
3199 especially if it could return the position of the first one.
3204 Perl_sv_utf8_upgrade_flags_grow(pTHX_ register SV *const sv, const I32 flags, STRLEN extra)
3208 PERL_ARGS_ASSERT_SV_UTF8_UPGRADE_FLAGS_GROW;
3210 if (sv == &PL_sv_undef)
3214 if (SvREADONLY(sv) && (SvPOKp(sv) || SvIOKp(sv) || SvNOKp(sv))) {
3215 (void) sv_2pv_flags(sv,&len, flags);
3217 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3221 (void) SvPV_force(sv,len);
3226 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3231 sv_force_normal_flags(sv, 0);
3234 if (PL_encoding && !(flags & SV_UTF8_NO_ENCODING)) {
3235 sv_recode_to_utf8(sv, PL_encoding);
3236 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3240 if (SvCUR(sv) > 0) { /* Assume Latin-1/EBCDIC */
3241 /* This function could be much more efficient if we
3242 * had a FLAG in SVs to signal if there are any variant
3243 * chars in the PV. Given that there isn't such a flag
3244 * make the loop as fast as possible (although there are certainly ways
3245 * to speed this up, eg. through vectorization) */
3246 U8 * s = (U8 *) SvPVX_const(sv);
3247 U8 * e = (U8 *) SvEND(sv);
3249 STRLEN two_byte_count = 0;
3251 if (flags & SV_FORCE_UTF8_UPGRADE) goto must_be_utf8;
3253 /* See if really will need to convert to utf8. We mustn't rely on our
3254 * incoming SV being well formed and having a trailing '\0', as certain
3255 * code in pp_formline can send us partially built SVs. */
3259 if (NATIVE_IS_INVARIANT(ch)) continue;
3261 t--; /* t already incremented; re-point to first variant */
3266 /* utf8 conversion not needed because all are invariants. Mark as
3267 * UTF-8 even if no variant - saves scanning loop */
3273 /* Here, the string should be converted to utf8, either because of an
3274 * input flag (two_byte_count = 0), or because a character that
3275 * requires 2 bytes was found (two_byte_count = 1). t points either to
3276 * the beginning of the string (if we didn't examine anything), or to
3277 * the first variant. In either case, everything from s to t - 1 will
3278 * occupy only 1 byte each on output.
3280 * There are two main ways to convert. One is to create a new string
3281 * and go through the input starting from the beginning, appending each
3282 * converted value onto the new string as we go along. It's probably
3283 * best to allocate enough space in the string for the worst possible
3284 * case rather than possibly running out of space and having to
3285 * reallocate and then copy what we've done so far. Since everything
3286 * from s to t - 1 is invariant, the destination can be initialized
3287 * with these using a fast memory copy
3289 * The other way is to figure out exactly how big the string should be
3290 * by parsing the entire input. Then you don't have to make it big
3291 * enough to handle the worst possible case, and more importantly, if
3292 * the string you already have is large enough, you don't have to
3293 * allocate a new string, you can copy the last character in the input
3294 * string to the final position(s) that will be occupied by the
3295 * converted string and go backwards, stopping at t, since everything
3296 * before that is invariant.
3298 * There are advantages and disadvantages to each method.
3300 * In the first method, we can allocate a new string, do the memory
3301 * copy from the s to t - 1, and then proceed through the rest of the
3302 * string byte-by-byte.
3304 * In the second method, we proceed through the rest of the input
3305 * string just calculating how big the converted string will be. Then
3306 * there are two cases:
3307 * 1) if the string has enough extra space to handle the converted
3308 * value. We go backwards through the string, converting until we
3309 * get to the position we are at now, and then stop. If this
3310 * position is far enough along in the string, this method is
3311 * faster than the other method. If the memory copy were the same
3312 * speed as the byte-by-byte loop, that position would be about
3313 * half-way, as at the half-way mark, parsing to the end and back
3314 * is one complete string's parse, the same amount as starting
3315 * over and going all the way through. Actually, it would be
3316 * somewhat less than half-way, as it's faster to just count bytes
3317 * than to also copy, and we don't have the overhead of allocating
3318 * a new string, changing the scalar to use it, and freeing the
3319 * existing one. But if the memory copy is fast, the break-even
3320 * point is somewhere after half way. The counting loop could be
3321 * sped up by vectorization, etc, to move the break-even point
3322 * further towards the beginning.
3323 * 2) if the string doesn't have enough space to handle the converted
3324 * value. A new string will have to be allocated, and one might
3325 * as well, given that, start from the beginning doing the first
3326 * method. We've spent extra time parsing the string and in
3327 * exchange all we've gotten is that we know precisely how big to
3328 * make the new one. Perl is more optimized for time than space,
3329 * so this case is a loser.
3330 * So what I've decided to do is not use the 2nd method unless it is
3331 * guaranteed that a new string won't have to be allocated, assuming
3332 * the worst case. I also decided not to put any more conditions on it
3333 * than this, for now. It seems likely that, since the worst case is
3334 * twice as big as the unknown portion of the string (plus 1), we won't
3335 * be guaranteed enough space, causing us to go to the first method,
3336 * unless the string is short, or the first variant character is near
3337 * the end of it. In either of these cases, it seems best to use the
3338 * 2nd method. The only circumstance I can think of where this would
3339 * be really slower is if the string had once had much more data in it
3340 * than it does now, but there is still a substantial amount in it */
3343 STRLEN invariant_head = t - s;
3344 STRLEN size = invariant_head + (e - t) * 2 + 1 + extra;
3345 if (SvLEN(sv) < size) {
3347 /* Here, have decided to allocate a new string */
3352 Newx(dst, size, U8);
3354 /* If no known invariants at the beginning of the input string,
3355 * set so starts from there. Otherwise, can use memory copy to
3356 * get up to where we are now, and then start from here */
3358 if (invariant_head <= 0) {
3361 Copy(s, dst, invariant_head, char);
3362 d = dst + invariant_head;
3366 const UV uv = NATIVE8_TO_UNI(*t++);
3367 if (UNI_IS_INVARIANT(uv))
3368 *d++ = (U8)UNI_TO_NATIVE(uv);
3370 *d++ = (U8)UTF8_EIGHT_BIT_HI(uv);
3371 *d++ = (U8)UTF8_EIGHT_BIT_LO(uv);
3375 SvPV_free(sv); /* No longer using pre-existing string */
3376 SvPV_set(sv, (char*)dst);
3377 SvCUR_set(sv, d - dst);
3378 SvLEN_set(sv, size);
3381 /* Here, have decided to get the exact size of the string.
3382 * Currently this happens only when we know that there is
3383 * guaranteed enough space to fit the converted string, so
3384 * don't have to worry about growing. If two_byte_count is 0,
3385 * then t points to the first byte of the string which hasn't
3386 * been examined yet. Otherwise two_byte_count is 1, and t
3387 * points to the first byte in the string that will expand to
3388 * two. Depending on this, start examining at t or 1 after t.
3391 U8 *d = t + two_byte_count;
3394 /* Count up the remaining bytes that expand to two */
3397 const U8 chr = *d++;
3398 if (! NATIVE_IS_INVARIANT(chr)) two_byte_count++;
3401 /* The string will expand by just the number of bytes that
3402 * occupy two positions. But we are one afterwards because of
3403 * the increment just above. This is the place to put the
3404 * trailing NUL, and to set the length before we decrement */
3406 d += two_byte_count;
3407 SvCUR_set(sv, d - s);
3411 /* Having decremented d, it points to the position to put the
3412 * very last byte of the expanded string. Go backwards through
3413 * the string, copying and expanding as we go, stopping when we
3414 * get to the part that is invariant the rest of the way down */
3418 const U8 ch = NATIVE8_TO_UNI(*e--);
3419 if (UNI_IS_INVARIANT(ch)) {
3420 *d-- = UNI_TO_NATIVE(ch);
3422 *d-- = (U8)UTF8_EIGHT_BIT_LO(ch);
3423 *d-- = (U8)UTF8_EIGHT_BIT_HI(ch);
3430 /* Mark as UTF-8 even if no variant - saves scanning loop */
3436 =for apidoc sv_utf8_downgrade
3438 Attempts to convert the PV of an SV from characters to bytes.
3439 If the PV contains a character that cannot fit
3440 in a byte, this conversion will fail;
3441 in this case, either returns false or, if C<fail_ok> is not
3444 This is not as a general purpose Unicode to byte encoding interface:
3445 use the Encode extension for that.
3451 Perl_sv_utf8_downgrade(pTHX_ register SV *const sv, const bool fail_ok)
3455 PERL_ARGS_ASSERT_SV_UTF8_DOWNGRADE;
3457 if (SvPOKp(sv) && SvUTF8(sv)) {
3463 sv_force_normal_flags(sv, 0);
3465 s = (U8 *) SvPV(sv, len);
3466 if (!utf8_to_bytes(s, &len)) {
3471 Perl_croak(aTHX_ "Wide character in %s",
3474 Perl_croak(aTHX_ "Wide character");
3485 =for apidoc sv_utf8_encode
3487 Converts the PV of an SV to UTF-8, but then turns the C<SvUTF8>
3488 flag off so that it looks like octets again.
3494 Perl_sv_utf8_encode(pTHX_ register SV *const sv)
3496 PERL_ARGS_ASSERT_SV_UTF8_ENCODE;
3499 sv_force_normal_flags(sv, 0);
3501 if (SvREADONLY(sv)) {
3502 Perl_croak(aTHX_ "%s", PL_no_modify);
3504 (void) sv_utf8_upgrade(sv);
3509 =for apidoc sv_utf8_decode
3511 If the PV of the SV is an octet sequence in UTF-8
3512 and contains a multiple-byte character, the C<SvUTF8> flag is turned on
3513 so that it looks like a character. If the PV contains only single-byte
3514 characters, the C<SvUTF8> flag stays being off.
3515 Scans PV for validity and returns false if the PV is invalid UTF-8.
3521 Perl_sv_utf8_decode(pTHX_ register SV *const sv)
3523 PERL_ARGS_ASSERT_SV_UTF8_DECODE;
3529 /* The octets may have got themselves encoded - get them back as
3532 if (!sv_utf8_downgrade(sv, TRUE))
3535 /* it is actually just a matter of turning the utf8 flag on, but
3536 * we want to make sure everything inside is valid utf8 first.
3538 c = (const U8 *) SvPVX_const(sv);
3539 if (!is_utf8_string(c, SvCUR(sv)+1))
3541 e = (const U8 *) SvEND(sv);
3544 if (!UTF8_IS_INVARIANT(ch)) {
3554 =for apidoc sv_setsv
3556 Copies the contents of the source SV C<ssv> into the destination SV
3557 C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3558 function if the source SV needs to be reused. Does not handle 'set' magic.
3559 Loosely speaking, it performs a copy-by-value, obliterating any previous
3560 content of the destination.
3562 You probably want to use one of the assortment of wrappers, such as
3563 C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3564 C<SvSetMagicSV_nosteal>.
3566 =for apidoc sv_setsv_flags
3568 Copies the contents of the source SV C<ssv> into the destination SV
3569 C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3570 function if the source SV needs to be reused. Does not handle 'set' magic.
3571 Loosely speaking, it performs a copy-by-value, obliterating any previous
3572 content of the destination.
3573 If the C<flags> parameter has the C<SV_GMAGIC> bit set, will C<mg_get> on
3574 C<ssv> if appropriate, else not. If the C<flags> parameter has the
3575 C<NOSTEAL> bit set then the buffers of temps will not be stolen. <sv_setsv>
3576 and C<sv_setsv_nomg> are implemented in terms of this function.
3578 You probably want to use one of the assortment of wrappers, such as
3579 C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3580 C<SvSetMagicSV_nosteal>.
3582 This is the primary function for copying scalars, and most other
3583 copy-ish functions and macros use this underneath.
3589 S_glob_assign_glob(pTHX_ SV *const dstr, SV *const sstr, const int dtype)
3591 I32 mro_changes = 0; /* 1 = method, 2 = isa */
3593 PERL_ARGS_ASSERT_GLOB_ASSIGN_GLOB;
3595 if (dtype != SVt_PVGV) {
3596 const char * const name = GvNAME(sstr);
3597 const STRLEN len = GvNAMELEN(sstr);
3599 if (dtype >= SVt_PV) {
3605 SvUPGRADE(dstr, SVt_PVGV);
3606 (void)SvOK_off(dstr);
3607 /* FIXME - why are we doing this, then turning it off and on again
3609 isGV_with_GP_on(dstr);
3611 GvSTASH(dstr) = GvSTASH(sstr);
3613 Perl_sv_add_backref(aTHX_ MUTABLE_SV(GvSTASH(dstr)), dstr);
3614 gv_name_set(MUTABLE_GV(dstr), name, len, GV_ADD);
3615 SvFAKE_on(dstr); /* can coerce to non-glob */
3618 if(GvGP(MUTABLE_GV(sstr))) {
3619 /* If source has method cache entry, clear it */
3621 SvREFCNT_dec(GvCV(sstr));
3625 /* If source has a real method, then a method is
3627 else if(GvCV((const GV *)sstr)) {
3632 /* If dest already had a real method, that's a change as well */
3633 if(!mro_changes && GvGP(MUTABLE_GV(dstr)) && GvCVu((const GV *)dstr)) {
3637 if(strEQ(GvNAME((const GV *)dstr),"ISA"))
3640 gp_free(MUTABLE_GV(dstr));
3641 isGV_with_GP_off(dstr);
3642 (void)SvOK_off(dstr);
3643 isGV_with_GP_on(dstr);
3644 GvINTRO_off(dstr); /* one-shot flag */
3645 GvGP(dstr) = gp_ref(GvGP(sstr));
3646 if (SvTAINTED(sstr))
3648 if (GvIMPORTED(dstr) != GVf_IMPORTED
3649 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3651 GvIMPORTED_on(dstr);
3654 if(mro_changes == 2) mro_isa_changed_in(GvSTASH(dstr));
3655 else if(mro_changes) mro_method_changed_in(GvSTASH(dstr));
3660 S_glob_assign_ref(pTHX_ SV *const dstr, SV *const sstr)
3662 SV * const sref = SvREFCNT_inc(SvRV(sstr));
3664 const int intro = GvINTRO(dstr);
3667 const U32 stype = SvTYPE(sref);
3669 PERL_ARGS_ASSERT_GLOB_ASSIGN_REF;
3672 GvINTRO_off(dstr); /* one-shot flag */
3673 GvLINE(dstr) = CopLINE(PL_curcop);
3674 GvEGV(dstr) = MUTABLE_GV(dstr);
3679 location = (SV **) &GvCV(dstr);
3680 import_flag = GVf_IMPORTED_CV;
3683 location = (SV **) &GvHV(dstr);
3684 import_flag = GVf_IMPORTED_HV;
3687 location = (SV **) &GvAV(dstr);
3688 import_flag = GVf_IMPORTED_AV;
3691 location = (SV **) &GvIOp(dstr);
3694 location = (SV **) &GvFORM(dstr);
3697 location = &GvSV(dstr);
3698 import_flag = GVf_IMPORTED_SV;
3701 if (stype == SVt_PVCV) {
3702 /*if (GvCVGEN(dstr) && (GvCV(dstr) != (const CV *)sref || GvCVGEN(dstr))) {*/
3703 if (GvCVGEN(dstr)) {
3704 SvREFCNT_dec(GvCV(dstr));
3706 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3709 SAVEGENERICSV(*location);
3713 if (stype == SVt_PVCV && (*location != sref || GvCVGEN(dstr))) {
3714 CV* const cv = MUTABLE_CV(*location);
3716 if (!GvCVGEN((const GV *)dstr) &&
3717 (CvROOT(cv) || CvXSUB(cv)))
3719 /* Redefining a sub - warning is mandatory if
3720 it was a const and its value changed. */
3721 if (CvCONST(cv) && CvCONST((const CV *)sref)
3723 == cv_const_sv((const CV *)sref)) {
3725 /* They are 2 constant subroutines generated from
3726 the same constant. This probably means that
3727 they are really the "same" proxy subroutine
3728 instantiated in 2 places. Most likely this is
3729 when a constant is exported twice. Don't warn.
3732 else if (ckWARN(WARN_REDEFINE)
3734 && (!CvCONST((const CV *)sref)
3735 || sv_cmp(cv_const_sv(cv),
3736 cv_const_sv((const CV *)
3738 Perl_warner(aTHX_ packWARN(WARN_REDEFINE),
3741 ? "Constant subroutine %s::%s redefined"
3742 : "Subroutine %s::%s redefined"),
3743 HvNAME_get(GvSTASH((const GV *)dstr)),
3744 GvENAME(MUTABLE_GV(dstr)));
3748 cv_ckproto_len(cv, (const GV *)dstr,
3749 SvPOK(sref) ? SvPVX_const(sref) : NULL,
3750 SvPOK(sref) ? SvCUR(sref) : 0);
3752 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3753 GvASSUMECV_on(dstr);
3754 if(GvSTASH(dstr)) mro_method_changed_in(GvSTASH(dstr)); /* sub foo { 1 } sub bar { 2 } *bar = \&foo */
3757 if (import_flag && !(GvFLAGS(dstr) & import_flag)
3758 && CopSTASH_ne(PL_curcop, GvSTASH(dstr))) {
3759 GvFLAGS(dstr) |= import_flag;
3764 if (SvTAINTED(sstr))
3770 Perl_sv_setsv_flags(pTHX_ SV *dstr, register SV* sstr, const I32 flags)
3773 register U32 sflags;
3775 register svtype stype;
3777 PERL_ARGS_ASSERT_SV_SETSV_FLAGS;
3782 if (SvIS_FREED(dstr)) {
3783 Perl_croak(aTHX_ "panic: attempt to copy value %" SVf
3784 " to a freed scalar %p", SVfARG(sstr), (void *)dstr);
3786 SV_CHECK_THINKFIRST_COW_DROP(dstr);
3788 sstr = &PL_sv_undef;
3789 if (SvIS_FREED(sstr)) {
3790 Perl_croak(aTHX_ "panic: attempt to copy freed scalar %p to %p",
3791 (void*)sstr, (void*)dstr);
3793 stype = SvTYPE(sstr);
3794 dtype = SvTYPE(dstr);
3796 (void)SvAMAGIC_off(dstr);
3799 /* need to nuke the magic */
3803 /* There's a lot of redundancy below but we're going for speed here */
3808 if (dtype != SVt_PVGV) {
3809 (void)SvOK_off(dstr);
3817 sv_upgrade(dstr, SVt_IV);
3821 sv_upgrade(dstr, SVt_PVIV);
3824 goto end_of_first_switch;
3826 (void)SvIOK_only(dstr);
3827 SvIV_set(dstr, SvIVX(sstr));
3830 /* SvTAINTED can only be true if the SV has taint magic, which in
3831 turn means that the SV type is PVMG (or greater). This is the
3832 case statement for SVt_IV, so this cannot be true (whatever gcov
3834 assert(!SvTAINTED(sstr));
3839 if (dtype < SVt_PV && dtype != SVt_IV)
3840 sv_upgrade(dstr, SVt_IV);
3848 sv_upgrade(dstr, SVt_NV);
3852 sv_upgrade(dstr, SVt_PVNV);
3855 goto end_of_first_switch;
3857 SvNV_set(dstr, SvNVX(sstr));
3858 (void)SvNOK_only(dstr);
3859 /* SvTAINTED can only be true if the SV has taint magic, which in
3860 turn means that the SV type is PVMG (or greater). This is the
3861 case statement for SVt_NV, so this cannot be true (whatever gcov
3863 assert(!SvTAINTED(sstr));
3869 #ifdef PERL_OLD_COPY_ON_WRITE
3870 if ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS) {
3871 if (dtype < SVt_PVIV)
3872 sv_upgrade(dstr, SVt_PVIV);
3880 sv_upgrade(dstr, SVt_PV);
3883 if (dtype < SVt_PVIV)
3884 sv_upgrade(dstr, SVt_PVIV);
3887 if (dtype < SVt_PVNV)
3888 sv_upgrade(dstr, SVt_PVNV);
3892 const char * const type = sv_reftype(sstr,0);
3894 Perl_croak(aTHX_ "Bizarre copy of %s in %s", type, OP_NAME(PL_op));
3896 Perl_croak(aTHX_ "Bizarre copy of %s", type);
3900 /* case SVt_BIND: */
3903 if (isGV_with_GP(sstr) && dtype <= SVt_PVGV) {
3904 glob_assign_glob(dstr, sstr, dtype);
3907 /* SvVALID means that this PVGV is playing at being an FBM. */
3911 if (SvGMAGICAL(sstr) && (flags & SV_GMAGIC)) {
3913 if (SvTYPE(sstr) != stype) {
3914 stype = SvTYPE(sstr);
3915 if (isGV_with_GP(sstr) && stype == SVt_PVGV && dtype <= SVt_PVGV) {
3916 glob_assign_glob(dstr, sstr, dtype);
3921 if (stype == SVt_PVLV)
3922 SvUPGRADE(dstr, SVt_PVNV);
3924 SvUPGRADE(dstr, (svtype)stype);
3926 end_of_first_switch:
3928 /* dstr may have been upgraded. */
3929 dtype = SvTYPE(dstr);
3930 sflags = SvFLAGS(sstr);
3932 if (dtype == SVt_PVCV || dtype == SVt_PVFM) {
3933 /* Assigning to a subroutine sets the prototype. */
3936 const char *const ptr = SvPV_const(sstr, len);
3938 SvGROW(dstr, len + 1);
3939 Copy(ptr, SvPVX(dstr), len + 1, char);
3940 SvCUR_set(dstr, len);
3942 SvFLAGS(dstr) |= sflags & SVf_UTF8;
3946 } else if (dtype == SVt_PVAV || dtype == SVt_PVHV) {
3947 const char * const type = sv_reftype(dstr,0);
3949 Perl_croak(aTHX_ "Cannot copy to %s in %s", type, OP_NAME(PL_op));
3951 Perl_croak(aTHX_ "Cannot copy to %s", type);
3952 } else if (sflags & SVf_ROK) {
3953 if (isGV_with_GP(dstr) && dtype == SVt_PVGV
3954 && SvTYPE(SvRV(sstr)) == SVt_PVGV && isGV_with_GP(SvRV(sstr))) {
3957 if (GvIMPORTED(dstr) != GVf_IMPORTED
3958 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3960 GvIMPORTED_on(dstr);
3965 glob_assign_glob(dstr, sstr, dtype);
3969 if (dtype >= SVt_PV) {
3970 if (dtype == SVt_PVGV && isGV_with_GP(dstr)) {
3971 glob_assign_ref(dstr, sstr);
3974 if (SvPVX_const(dstr)) {
3980 (void)SvOK_off(dstr);
3981 SvRV_set(dstr, SvREFCNT_inc(SvRV(sstr)));
3982 SvFLAGS(dstr) |= sflags & SVf_ROK;
3983 assert(!(sflags & SVp_NOK));
3984 assert(!(sflags & SVp_IOK));
3985 assert(!(sflags & SVf_NOK));
3986 assert(!(sflags & SVf_IOK));
3988 else if (dtype == SVt_PVGV && isGV_with_GP(dstr)) {
3989 if (!(sflags & SVf_OK)) {
3990 if (ckWARN(WARN_MISC))
3991 Perl_warner(aTHX_ packWARN(WARN_MISC),
3992 "Undefined value assigned to typeglob");
3995 GV *gv = gv_fetchsv(sstr, GV_ADD, SVt_PVGV);
3996 if (dstr != (const SV *)gv) {
3998 gp_free(MUTABLE_GV(dstr));
3999 GvGP(dstr) = gp_ref(GvGP(gv));
4003 else if (sflags & SVp_POK) {
4007 * Check to see if we can just swipe the string. If so, it's a
4008 * possible small lose on short strings, but a big win on long ones.
4009 * It might even be a win on short strings if SvPVX_const(dstr)
4010 * has to be allocated and SvPVX_const(sstr) has to be freed.
4011 * Likewise if we can set up COW rather than doing an actual copy, we
4012 * drop to the else clause, as the swipe code and the COW setup code
4013 * have much in common.
4016 /* Whichever path we take through the next code, we want this true,
4017 and doing it now facilitates the COW check. */
4018 (void)SvPOK_only(dstr);
4021 /* If we're already COW then this clause is not true, and if COW
4022 is allowed then we drop down to the else and make dest COW
4023 with us. If caller hasn't said that we're allowed to COW
4024 shared hash keys then we don't do the COW setup, even if the
4025 source scalar is a shared hash key scalar. */
4026 (((flags & SV_COW_SHARED_HASH_KEYS)
4027 ? (sflags & (SVf_FAKE|SVf_READONLY)) != (SVf_FAKE|SVf_READONLY)
4028 : 1 /* If making a COW copy is forbidden then the behaviour we
4029 desire is as if the source SV isn't actually already
4030 COW, even if it is. So we act as if the source flags
4031 are not COW, rather than actually testing them. */
4033 #ifndef PERL_OLD_COPY_ON_WRITE
4034 /* The change that added SV_COW_SHARED_HASH_KEYS makes the logic
4035 when PERL_OLD_COPY_ON_WRITE is defined a little wrong.
4036 Conceptually PERL_OLD_COPY_ON_WRITE being defined should
4037 override SV_COW_SHARED_HASH_KEYS, because it means "always COW"
4038 but in turn, it's somewhat dead code, never expected to go
4039 live, but more kept as a placeholder on how to do it better
4040 in a newer implementation. */
4041 /* If we are COW and dstr is a suitable target then we drop down
4042 into the else and make dest a COW of us. */
4043 || (SvFLAGS(dstr) & CAN_COW_MASK) != CAN_COW_FLAGS
4048 (sflags & SVs_TEMP) && /* slated for free anyway? */
4049 !(sflags & SVf_OOK) && /* and not involved in OOK hack? */
4050 (!(flags & SV_NOSTEAL)) &&
4051 /* and we're allowed to steal temps */
4052 SvREFCNT(sstr) == 1 && /* and no other references to it? */
4053 SvLEN(sstr) && /* and really is a string */
4054 /* and won't be needed again, potentially */
4055 !(PL_op && PL_op->op_type == OP_AASSIGN))
4056 #ifdef PERL_OLD_COPY_ON_WRITE
4057 && ((flags & SV_COW_SHARED_HASH_KEYS)
4058 ? (!((sflags & CAN_COW_MASK) == CAN_COW_FLAGS
4059 && (SvFLAGS(dstr) & CAN_COW_MASK) == CAN_COW_FLAGS
4060 && SvTYPE(sstr) >= SVt_PVIV && SvTYPE(sstr) != SVt_PVFM))
4064 /* Failed the swipe test, and it's not a shared hash key either.
4065 Have to copy the string. */
4066 STRLEN len = SvCUR(sstr);
4067 SvGROW(dstr, len + 1); /* inlined from sv_setpvn */
4068 Move(SvPVX_const(sstr),SvPVX(dstr),len,char);
4069 SvCUR_set(dstr, len);
4070 *SvEND(dstr) = '\0';
4072 /* If PERL_OLD_COPY_ON_WRITE is not defined, then isSwipe will always
4074 /* Either it's a shared hash key, or it's suitable for
4075 copy-on-write or we can swipe the string. */
4077 PerlIO_printf(Perl_debug_log, "Copy on write: sstr --> dstr\n");
4081 #ifdef PERL_OLD_COPY_ON_WRITE
4083 if ((sflags & (SVf_FAKE | SVf_READONLY))
4084 != (SVf_FAKE | SVf_READONLY)) {
4085 SvREADONLY_on(sstr);
4087 /* Make the source SV into a loop of 1.
4088 (about to become 2) */
4089 SV_COW_NEXT_SV_SET(sstr, sstr);
4093 /* Initial code is common. */
4094 if (SvPVX_const(dstr)) { /* we know that dtype >= SVt_PV */
4099 /* making another shared SV. */
4100 STRLEN cur = SvCUR(sstr);
4101 STRLEN len = SvLEN(sstr);
4102 #ifdef PERL_OLD_COPY_ON_WRITE
4104 assert (SvTYPE(dstr) >= SVt_PVIV);
4105 /* SvIsCOW_normal */
4106 /* splice us in between source and next-after-source. */
4107 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4108 SV_COW_NEXT_SV_SET(sstr, dstr);
4109 SvPV_set(dstr, SvPVX_mutable(sstr));
4113 /* SvIsCOW_shared_hash */
4114 DEBUG_C(PerlIO_printf(Perl_debug_log,
4115 "Copy on write: Sharing hash\n"));
4117 assert (SvTYPE(dstr) >= SVt_PV);
4119 HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)))));
4121 SvLEN_set(dstr, len);
4122 SvCUR_set(dstr, cur);
4123 SvREADONLY_on(dstr);
4127 { /* Passes the swipe test. */
4128 SvPV_set(dstr, SvPVX_mutable(sstr));
4129 SvLEN_set(dstr, SvLEN(sstr));
4130 SvCUR_set(dstr, SvCUR(sstr));
4133 (void)SvOK_off(sstr); /* NOTE: nukes most SvFLAGS on sstr */
4134 SvPV_set(sstr, NULL);
4140 if (sflags & SVp_NOK) {
4141 SvNV_set(dstr, SvNVX(sstr));
4143 if (sflags & SVp_IOK) {
4144 SvIV_set(dstr, SvIVX(sstr));
4145 /* Must do this otherwise some other overloaded use of 0x80000000
4146 gets confused. I guess SVpbm_VALID */
4147 if (sflags & SVf_IVisUV)
4150 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_NOK|SVp_NOK|SVf_UTF8);
4152 const MAGIC * const smg = SvVSTRING_mg(sstr);
4154 sv_magic(dstr, NULL, PERL_MAGIC_vstring,
4155 smg->mg_ptr, smg->mg_len);
4156 SvRMAGICAL_on(dstr);
4160 else if (sflags & (SVp_IOK|SVp_NOK)) {
4161 (void)SvOK_off(dstr);
4162 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_IVisUV|SVf_NOK|SVp_NOK);
4163 if (sflags & SVp_IOK) {
4164 /* XXXX Do we want to set IsUV for IV(ROK)? Be extra safe... */
4165 SvIV_set(dstr, SvIVX(sstr));
4167 if (sflags & SVp_NOK) {
4168 SvNV_set(dstr, SvNVX(sstr));
4172 if (isGV_with_GP(sstr)) {
4173 /* This stringification rule for globs is spread in 3 places.
4174 This feels bad. FIXME. */
4175 const U32 wasfake = sflags & SVf_FAKE;
4177 /* FAKE globs can get coerced, so need to turn this off
4178 temporarily if it is on. */
4180 gv_efullname3(dstr, MUTABLE_GV(sstr), "*");
4181 SvFLAGS(sstr) |= wasfake;
4184 (void)SvOK_off(dstr);
4186 if (SvTAINTED(sstr))
4191 =for apidoc sv_setsv_mg
4193 Like C<sv_setsv>, but also handles 'set' magic.
4199 Perl_sv_setsv_mg(pTHX_ SV *const dstr, register SV *const sstr)
4201 PERL_ARGS_ASSERT_SV_SETSV_MG;
4203 sv_setsv(dstr,sstr);
4207 #ifdef PERL_OLD_COPY_ON_WRITE
4209 Perl_sv_setsv_cow(pTHX_ SV *dstr, SV *sstr)
4211 STRLEN cur = SvCUR(sstr);
4212 STRLEN len = SvLEN(sstr);
4213 register char *new_pv;
4215 PERL_ARGS_ASSERT_SV_SETSV_COW;
4218 PerlIO_printf(Perl_debug_log, "Fast copy on write: %p -> %p\n",
4219 (void*)sstr, (void*)dstr);
4226 if (SvTHINKFIRST(dstr))
4227 sv_force_normal_flags(dstr, SV_COW_DROP_PV);
4228 else if (SvPVX_const(dstr))
4229 Safefree(SvPVX_const(dstr));
4233 SvUPGRADE(dstr, SVt_PVIV);
4235 assert (SvPOK(sstr));
4236 assert (SvPOKp(sstr));
4237 assert (!SvIOK(sstr));
4238 assert (!SvIOKp(sstr));
4239 assert (!SvNOK(sstr));
4240 assert (!SvNOKp(sstr));
4242 if (SvIsCOW(sstr)) {
4244 if (SvLEN(sstr) == 0) {
4245 /* source is a COW shared hash key. */
4246 DEBUG_C(PerlIO_printf(Perl_debug_log,
4247 "Fast copy on write: Sharing hash\n"));
4248 new_pv = HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr))));
4251 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4253 assert ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS);
4254 SvUPGRADE(sstr, SVt_PVIV);
4255 SvREADONLY_on(sstr);
4257 DEBUG_C(PerlIO_printf(Perl_debug_log,
4258 "Fast copy on write: Converting sstr to COW\n"));
4259 SV_COW_NEXT_SV_SET(dstr, sstr);
4261 SV_COW_NEXT_SV_SET(sstr, dstr);
4262 new_pv = SvPVX_mutable(sstr);
4265 SvPV_set(dstr, new_pv);
4266 SvFLAGS(dstr) = (SVt_PVIV|SVf_POK|SVp_POK|SVf_FAKE|SVf_READONLY);
4269 SvLEN_set(dstr, len);
4270 SvCUR_set(dstr, cur);
4279 =for apidoc sv_setpvn
4281 Copies a string into an SV. The C<len> parameter indicates the number of
4282 bytes to be copied. If the C<ptr> argument is NULL the SV will become
4283 undefined. Does not handle 'set' magic. See C<sv_setpvn_mg>.
4289 Perl_sv_setpvn(pTHX_ register SV *const sv, register const char *const ptr, register const STRLEN len)
4292 register char *dptr;
4294 PERL_ARGS_ASSERT_SV_SETPVN;
4296 SV_CHECK_THINKFIRST_COW_DROP(sv);
4302 /* len is STRLEN which is unsigned, need to copy to signed */
4305 Perl_croak(aTHX_ "panic: sv_setpvn called with negative strlen");
4307 SvUPGRADE(sv, SVt_PV);
4309 dptr = SvGROW(sv, len + 1);
4310 Move(ptr,dptr,len,char);
4313 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4318 =for apidoc sv_setpvn_mg
4320 Like C<sv_setpvn>, but also handles 'set' magic.
4326 Perl_sv_setpvn_mg(pTHX_ register SV *const sv, register const char *const ptr, register const STRLEN len)
4328 PERL_ARGS_ASSERT_SV_SETPVN_MG;
4330 sv_setpvn(sv,ptr,len);
4335 =for apidoc sv_setpv
4337 Copies a string into an SV. The string must be null-terminated. Does not
4338 handle 'set' magic. See C<sv_setpv_mg>.
4344 Perl_sv_setpv(pTHX_ register SV *const sv, register const char *const ptr)
4347 register STRLEN len;
4349 PERL_ARGS_ASSERT_SV_SETPV;
4351 SV_CHECK_THINKFIRST_COW_DROP(sv);
4357 SvUPGRADE(sv, SVt_PV);
4359 SvGROW(sv, len + 1);
4360 Move(ptr,SvPVX(sv),len+1,char);
4362 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4367 =for apidoc sv_setpv_mg
4369 Like C<sv_setpv>, but also handles 'set' magic.
4375 Perl_sv_setpv_mg(pTHX_ register SV *const sv, register const char *const ptr)
4377 PERL_ARGS_ASSERT_SV_SETPV_MG;
4384 =for apidoc sv_usepvn_flags
4386 Tells an SV to use C<ptr> to find its string value. Normally the
4387 string is stored inside the SV but sv_usepvn allows the SV to use an
4388 outside string. The C<ptr> should point to memory that was allocated
4389 by C<malloc>. The string length, C<len>, must be supplied. By default
4390 this function will realloc (i.e. move) the memory pointed to by C<ptr>,
4391 so that pointer should not be freed or used by the programmer after
4392 giving it to sv_usepvn, and neither should any pointers from "behind"
4393 that pointer (e.g. ptr + 1) be used.
4395 If C<flags> & SV_SMAGIC is true, will call SvSETMAGIC. If C<flags> &
4396 SV_HAS_TRAILING_NUL is true, then C<ptr[len]> must be NUL, and the realloc
4397 will be skipped. (i.e. the buffer is actually at least 1 byte longer than
4398 C<len>, and already meets the requirements for storing in C<SvPVX>)
4404 Perl_sv_usepvn_flags(pTHX_ SV *const sv, char *ptr, const STRLEN len, const U32 flags)
4409 PERL_ARGS_ASSERT_SV_USEPVN_FLAGS;
4411 SV_CHECK_THINKFIRST_COW_DROP(sv);
4412 SvUPGRADE(sv, SVt_PV);
4415 if (flags & SV_SMAGIC)
4419 if (SvPVX_const(sv))
4423 if (flags & SV_HAS_TRAILING_NUL)
4424 assert(ptr[len] == '\0');
4427 allocate = (flags & SV_HAS_TRAILING_NUL)
4429 #ifdef Perl_safesysmalloc_size
4432 PERL_STRLEN_ROUNDUP(len + 1);
4434 if (flags & SV_HAS_TRAILING_NUL) {
4435 /* It's long enough - do nothing.
4436 Specfically Perl_newCONSTSUB is relying on this. */
4439 /* Force a move to shake out bugs in callers. */
4440 char *new_ptr = (char*)safemalloc(allocate);
4441 Copy(ptr, new_ptr, len, char);
4442 PoisonFree(ptr,len,char);
4446 ptr = (char*) saferealloc (ptr, allocate);
4449 #ifdef Perl_safesysmalloc_size
4450 SvLEN_set(sv, Perl_safesysmalloc_size(ptr));
4452 SvLEN_set(sv, allocate);
4456 if (!(flags & SV_HAS_TRAILING_NUL)) {
4459 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4461 if (flags & SV_SMAGIC)
4465 #ifdef PERL_OLD_COPY_ON_WRITE
4466 /* Need to do this *after* making the SV normal, as we need the buffer
4467 pointer to remain valid until after we've copied it. If we let go too early,
4468 another thread could invalidate it by unsharing last of the same hash key
4469 (which it can do by means other than releasing copy-on-write Svs)
4470 or by changing the other copy-on-write SVs in the loop. */
4472 S_sv_release_COW(pTHX_ register SV *sv, const char *pvx, SV *after)
4474 PERL_ARGS_ASSERT_SV_RELEASE_COW;
4476 { /* this SV was SvIsCOW_normal(sv) */
4477 /* we need to find the SV pointing to us. */
4478 SV *current = SV_COW_NEXT_SV(after);
4480 if (current == sv) {
4481 /* The SV we point to points back to us (there were only two of us
4483 Hence other SV is no longer copy on write either. */
4485 SvREADONLY_off(after);
4487 /* We need to follow the pointers around the loop. */
4489 while ((next = SV_COW_NEXT_SV(current)) != sv) {
4492 /* don't loop forever if the structure is bust, and we have
4493 a pointer into a closed loop. */
4494 assert (current != after);
4495 assert (SvPVX_const(current) == pvx);
4497 /* Make the SV before us point to the SV after us. */
4498 SV_COW_NEXT_SV_SET(current, after);
4504 =for apidoc sv_force_normal_flags
4506 Undo various types of fakery on an SV: if the PV is a shared string, make
4507 a private copy; if we're a ref, stop refing; if we're a glob, downgrade to
4508 an xpvmg; if we're a copy-on-write scalar, this is the on-write time when
4509 we do the copy, and is also used locally. If C<SV_COW_DROP_PV> is set
4510 then a copy-on-write scalar drops its PV buffer (if any) and becomes
4511 SvPOK_off rather than making a copy. (Used where this scalar is about to be
4512 set to some other value.) In addition, the C<flags> parameter gets passed to
4513 C<sv_unref_flags()> when unrefing. C<sv_force_normal> calls this function
4514 with flags set to 0.
4520 Perl_sv_force_normal_flags(pTHX_ register SV *const sv, const U32 flags)
4524 PERL_ARGS_ASSERT_SV_FORCE_NORMAL_FLAGS;
4526 #ifdef PERL_OLD_COPY_ON_WRITE
4527 if (SvREADONLY(sv)) {
4529 const char * const pvx = SvPVX_const(sv);
4530 const STRLEN len = SvLEN(sv);
4531 const STRLEN cur = SvCUR(sv);
4532 /* next COW sv in the loop. If len is 0 then this is a shared-hash
4533 key scalar, so we mustn't attempt to call SV_COW_NEXT_SV(), as
4534 we'll fail an assertion. */
4535 SV * const next = len ? SV_COW_NEXT_SV(sv) : 0;
4538 PerlIO_printf(Perl_debug_log,
4539 "Copy on write: Force normal %ld\n",
4545 /* This SV doesn't own the buffer, so need to Newx() a new one: */
4548 if (flags & SV_COW_DROP_PV) {
4549 /* OK, so we don't need to copy our buffer. */
4552 SvGROW(sv, cur + 1);
4553 Move(pvx,SvPVX(sv),cur,char);
4558 sv_release_COW(sv, pvx, next);
4560 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4566 else if (IN_PERL_RUNTIME)
4567 Perl_croak(aTHX_ "%s", PL_no_modify);
4570 if (SvREADONLY(sv)) {
4572 const char * const pvx = SvPVX_const(sv);
4573 const STRLEN len = SvCUR(sv);
4578 SvGROW(sv, len + 1);
4579 Move(pvx,SvPVX(sv),len,char);
4581 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4583 else if (IN_PERL_RUNTIME)
4584 Perl_croak(aTHX_ "%s", PL_no_modify);
4588 sv_unref_flags(sv, flags);
4589 else if (SvFAKE(sv) && SvTYPE(sv) == SVt_PVGV)
4596 Efficient removal of characters from the beginning of the string buffer.
4597 SvPOK(sv) must be true and the C<ptr> must be a pointer to somewhere inside
4598 the string buffer. The C<ptr> becomes the first character of the adjusted
4599 string. Uses the "OOK hack".
4600 Beware: after this function returns, C<ptr> and SvPVX_const(sv) may no longer
4601 refer to the same chunk of data.
4607 Perl_sv_chop(pTHX_ register SV *const sv, register const char *const ptr)
4613 const U8 *real_start;
4617 PERL_ARGS_ASSERT_SV_CHOP;
4619 if (!ptr || !SvPOKp(sv))
4621 delta = ptr - SvPVX_const(sv);
4623 /* Nothing to do. */
4626 /* SvPVX(sv) may move in SV_CHECK_THINKFIRST(sv), but after this line,
4627 nothing uses the value of ptr any more. */
4628 max_delta = SvLEN(sv) ? SvLEN(sv) : SvCUR(sv);
4629 if (ptr <= SvPVX_const(sv))
4630 Perl_croak(aTHX_ "panic: sv_chop ptr=%p, start=%p, end=%p",
4631 ptr, SvPVX_const(sv), SvPVX_const(sv) + max_delta);
4632 SV_CHECK_THINKFIRST(sv);
4633 if (delta > max_delta)
4634 Perl_croak(aTHX_ "panic: sv_chop ptr=%p (was %p), start=%p, end=%p",
4635 SvPVX_const(sv) + delta, ptr, SvPVX_const(sv),
4636 SvPVX_const(sv) + max_delta);
4639 if (!SvLEN(sv)) { /* make copy of shared string */
4640 const char *pvx = SvPVX_const(sv);
4641 const STRLEN len = SvCUR(sv);
4642 SvGROW(sv, len + 1);
4643 Move(pvx,SvPVX(sv),len,char);
4646 SvFLAGS(sv) |= SVf_OOK;
4649 SvOOK_offset(sv, old_delta);
4651 SvLEN_set(sv, SvLEN(sv) - delta);
4652 SvCUR_set(sv, SvCUR(sv) - delta);
4653 SvPV_set(sv, SvPVX(sv) + delta);
4655 p = (U8 *)SvPVX_const(sv);
4660 real_start = p - delta;
4664 if (delta < 0x100) {
4668 p -= sizeof(STRLEN);
4669 Copy((U8*)&delta, p, sizeof(STRLEN), U8);
4673 /* Fill the preceding buffer with sentinals to verify that no-one is
4675 while (p > real_start) {
4683 =for apidoc sv_catpvn
4685 Concatenates the string onto the end of the string which is in the SV. The
4686 C<len> indicates number of bytes to copy. If the SV has the UTF-8
4687 status set, then the bytes appended should be valid UTF-8.
4688 Handles 'get' magic, but not 'set' magic. See C<sv_catpvn_mg>.
4690 =for apidoc sv_catpvn_flags
4692 Concatenates the string onto the end of the string which is in the SV. The
4693 C<len> indicates number of bytes to copy. If the SV has the UTF-8
4694 status set, then the bytes appended should be valid UTF-8.
4695 If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<dsv> if
4696 appropriate, else not. C<sv_catpvn> and C<sv_catpvn_nomg> are implemented
4697 in terms of this function.
4703 Perl_sv_catpvn_flags(pTHX_ register SV *const dsv, register const char *sstr, register const STRLEN slen, const I32 flags)
4707 const char * const dstr = SvPV_force_flags(dsv, dlen, flags);
4709 PERL_ARGS_ASSERT_SV_CATPVN_FLAGS;
4711 SvGROW(dsv, dlen + slen + 1);
4713 sstr = SvPVX_const(dsv);
4714 Move(sstr, SvPVX(dsv) + dlen, slen, char);
4715 SvCUR_set(dsv, SvCUR(dsv) + slen);
4717 (void)SvPOK_only_UTF8(dsv); /* validate pointer */
4719 if (flags & SV_SMAGIC)
4724 =for apidoc sv_catsv
4726 Concatenates the string from SV C<ssv> onto the end of the string in
4727 SV C<dsv>. Modifies C<dsv> but not C<ssv>. Handles 'get' magic, but
4728 not 'set' magic. See C<sv_catsv_mg>.
4730 =for apidoc sv_catsv_flags
4732 Concatenates the string from SV C<ssv> onto the end of the string in
4733 SV C<dsv>. Modifies C<dsv> but not C<ssv>. If C<flags> has C<SV_GMAGIC>
4734 bit set, will C<mg_get> on the SVs if appropriate, else not. C<sv_catsv>
4735 and C<sv_catsv_nomg> are implemented in terms of this function.
4740 Perl_sv_catsv_flags(pTHX_ SV *const dsv, register SV *const ssv, const I32 flags)
4744 PERL_ARGS_ASSERT_SV_CATSV_FLAGS;
4748 const char *spv = SvPV_const(ssv, slen);
4750 /* sutf8 and dutf8 were type bool, but under USE_ITHREADS,
4751 gcc version 2.95.2 20000220 (Debian GNU/Linux) for
4752 Linux xxx 2.2.17 on sparc64 with gcc -O2, we erroneously
4753 get dutf8 = 0x20000000, (i.e. SVf_UTF8) even though
4754 dsv->sv_flags doesn't have that bit set.
4755 Andy Dougherty 12 Oct 2001
4757 const I32 sutf8 = DO_UTF8(ssv);
4760 if (SvGMAGICAL(dsv) && (flags & SV_GMAGIC))
4762 dutf8 = DO_UTF8(dsv);
4764 if (dutf8 != sutf8) {
4766 /* Not modifying source SV, so taking a temporary copy. */
4767 SV* const csv = newSVpvn_flags(spv, slen, SVs_TEMP);
4769 sv_utf8_upgrade(csv);
4770 spv = SvPV_const(csv, slen);
4773 /* Leave enough space for the cat that's about to happen */
4774 sv_utf8_upgrade_flags_grow(dsv, 0, slen);
4776 sv_catpvn_nomg(dsv, spv, slen);
4779 if (flags & SV_SMAGIC)
4784 =for apidoc sv_catpv
4786 Concatenates the string onto the end of the string which is in the SV.
4787 If the SV has the UTF-8 status set, then the bytes appended should be
4788 valid UTF-8. Handles 'get' magic, but not 'set' magic. See C<sv_catpv_mg>.
4793 Perl_sv_catpv(pTHX_ register SV *const sv, register const char *ptr)
4796 register STRLEN len;
4800 PERL_ARGS_ASSERT_SV_CATPV;
4804 junk = SvPV_force(sv, tlen);
4806 SvGROW(sv, tlen + len + 1);
4808 ptr = SvPVX_const(sv);
4809 Move(ptr,SvPVX(sv)+tlen,len+1,char);
4810 SvCUR_set(sv, SvCUR(sv) + len);
4811 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4816 =for apidoc sv_catpv_mg
4818 Like C<sv_catpv>, but also handles 'set' magic.
4824 Perl_sv_catpv_mg(pTHX_ register SV *const sv, register const char *const ptr)
4826 PERL_ARGS_ASSERT_SV_CATPV_MG;
4835 Creates a new SV. A non-zero C<len> parameter indicates the number of
4836 bytes of preallocated string space the SV should have. An extra byte for a
4837 trailing NUL is also reserved. (SvPOK is not set for the SV even if string
4838 space is allocated.) The reference count for the new SV is set to 1.
4840 In 5.9.3, newSV() replaces the older NEWSV() API, and drops the first
4841 parameter, I<x>, a debug aid which allowed callers to identify themselves.
4842 This aid has been superseded by a new build option, PERL_MEM_LOG (see
4843 L<perlhack/PERL_MEM_LOG>). The older API is still there for use in XS
4844 modules supporting older perls.
4850 Perl_newSV(pTHX_ const STRLEN len)
4857 sv_upgrade(sv, SVt_PV);
4858 SvGROW(sv, len + 1);
4863 =for apidoc sv_magicext
4865 Adds magic to an SV, upgrading it if necessary. Applies the
4866 supplied vtable and returns a pointer to the magic added.
4868 Note that C<sv_magicext> will allow things that C<sv_magic> will not.
4869 In particular, you can add magic to SvREADONLY SVs, and add more than
4870 one instance of the same 'how'.
4872 If C<namlen> is greater than zero then a C<savepvn> I<copy> of C<name> is
4873 stored, if C<namlen> is zero then C<name> is stored as-is and - as another
4874 special case - if C<(name && namlen == HEf_SVKEY)> then C<name> is assumed
4875 to contain an C<SV*> and is stored as-is with its REFCNT incremented.
4877 (This is now used as a subroutine by C<sv_magic>.)
4882 Perl_sv_magicext(pTHX_ SV *const sv, SV *const obj, const int how,
4883 const MGVTBL *const vtable, const char *const name, const I32 namlen)
4888 PERL_ARGS_ASSERT_SV_MAGICEXT;
4890 SvUPGRADE(sv, SVt_PVMG);
4891 Newxz(mg, 1, MAGIC);
4892 mg->mg_moremagic = SvMAGIC(sv);
4893 SvMAGIC_set(sv, mg);
4895 /* Sometimes a magic contains a reference loop, where the sv and
4896 object refer to each other. To prevent a reference loop that
4897 would prevent such objects being freed, we look for such loops
4898 and if we find one we avoid incrementing the object refcount.
4900 Note we cannot do this to avoid self-tie loops as intervening RV must
4901 have its REFCNT incremented to keep it in existence.
4904 if (!obj || obj == sv ||
4905 how == PERL_MAGIC_arylen ||
4906 how == PERL_MAGIC_symtab ||
4907 (SvTYPE(obj) == SVt_PVGV &&
4908 (GvSV(obj) == sv || GvHV(obj) == (const HV *)sv
4909 || GvAV(obj) == (const AV *)sv || GvCV(obj) == (const CV *)sv
4910 || GvIOp(obj) == (const IO *)sv || GvFORM(obj) == (const CV *)sv)))
4915 mg->mg_obj = SvREFCNT_inc_simple(obj);
4916 mg->mg_flags |= MGf_REFCOUNTED;
4919 /* Normal self-ties simply pass a null object, and instead of
4920 using mg_obj directly, use the SvTIED_obj macro to produce a
4921 new RV as needed. For glob "self-ties", we are tieing the PVIO
4922 with an RV obj pointing to the glob containing the PVIO. In
4923 this case, to avoid a reference loop, we need to weaken the
4927 if (how == PERL_MAGIC_tiedscalar && SvTYPE(sv) == SVt_PVIO &&
4928 obj && SvROK(obj) && GvIO(SvRV(obj)) == (const IO *)sv)
4934 mg->mg_len = namlen;
4937 mg->mg_ptr = savepvn(name, namlen);
4938 else if (namlen == HEf_SVKEY) {
4939 /* Yes, this is casting away const. This is only for the case of
4940 HEf_SVKEY. I think we need to document this abberation of the
4941 constness of the API, rather than making name non-const, as
4942 that change propagating outwards a long way. */
4943 mg->mg_ptr = (char*)SvREFCNT_inc_simple_NN((SV *)name);
4945 mg->mg_ptr = (char *) name;
4947 mg->mg_virtual = (MGVTBL *) vtable;
4951 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
4956 =for apidoc sv_magic
4958 Adds magic to an SV. First upgrades C<sv> to type C<SVt_PVMG> if necessary,
4959 then adds a new magic item of type C<how> to the head of the magic list.
4961 See C<sv_magicext> (which C<sv_magic> now calls) for a description of the
4962 handling of the C<name> and C<namlen> arguments.
4964 You need to use C<sv_magicext> to add magic to SvREADONLY SVs and also
4965 to add more than one instance of the same 'how'.
4971 Perl_sv_magic(pTHX_ register SV *const sv, SV *const obj, const int how,
4972 const char *const name, const I32 namlen)
4975 const MGVTBL *vtable;
4978 PERL_ARGS_ASSERT_SV_MAGIC;
4980 #ifdef PERL_OLD_COPY_ON_WRITE
4982 sv_force_normal_flags(sv, 0);
4984 if (SvREADONLY(sv)) {
4986 /* its okay to attach magic to shared strings; the subsequent
4987 * upgrade to PVMG will unshare the string */
4988 !(SvFAKE(sv) && SvTYPE(sv) < SVt_PVMG)
4991 && how != PERL_MAGIC_regex_global
4992 && how != PERL_MAGIC_bm
4993 && how != PERL_MAGIC_fm
4994 && how != PERL_MAGIC_sv
4995 && how != PERL_MAGIC_backref
4998 Perl_croak(aTHX_ "%s", PL_no_modify);
5001 if (SvMAGICAL(sv) || (how == PERL_MAGIC_taint && SvTYPE(sv) >= SVt_PVMG)) {
5002 if (SvMAGIC(sv) && (mg = mg_find(sv, how))) {
5003 /* sv_magic() refuses to add a magic of the same 'how' as an
5006 if (how == PERL_MAGIC_taint) {
5008 /* Any scalar which already had taint magic on which someone
5009 (erroneously?) did SvIOK_on() or similar will now be
5010 incorrectly sporting public "OK" flags. */
5011 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
5019 vtable = &PL_vtbl_sv;
5021 case PERL_MAGIC_overload:
5022 vtable = &PL_vtbl_amagic;
5024 case PERL_MAGIC_overload_elem:
5025 vtable = &PL_vtbl_amagicelem;
5027 case PERL_MAGIC_overload_table:
5028 vtable = &PL_vtbl_ovrld;
5031 vtable = &PL_vtbl_bm;
5033 case PERL_MAGIC_regdata:
5034 vtable = &PL_vtbl_regdata;
5036 case PERL_MAGIC_regdatum:
5037 vtable = &PL_vtbl_regdatum;
5039 case PERL_MAGIC_env:
5040 vtable = &PL_vtbl_env;
5043 vtable = &PL_vtbl_fm;
5045 case PERL_MAGIC_envelem:
5046 vtable = &PL_vtbl_envelem;
5048 case PERL_MAGIC_regex_global:
5049 vtable = &PL_vtbl_mglob;
5051 case PERL_MAGIC_isa:
5052 vtable = &PL_vtbl_isa;
5054 case PERL_MAGIC_isaelem:
5055 vtable = &PL_vtbl_isaelem;
5057 case PERL_MAGIC_nkeys:
5058 vtable = &PL_vtbl_nkeys;
5060 case PERL_MAGIC_dbfile:
5063 case PERL_MAGIC_dbline:
5064 vtable = &PL_vtbl_dbline;
5066 #ifdef USE_LOCALE_COLLATE
5067 case PERL_MAGIC_collxfrm:
5068 vtable = &PL_vtbl_collxfrm;
5070 #endif /* USE_LOCALE_COLLATE */
5071 case PERL_MAGIC_tied:
5072 vtable = &PL_vtbl_pack;
5074 case PERL_MAGIC_tiedelem:
5075 case PERL_MAGIC_tiedscalar:
5076 vtable = &PL_vtbl_packelem;
5079 vtable = &PL_vtbl_regexp;
5081 case PERL_MAGIC_hints:
5082 /* As this vtable is all NULL, we can reuse it. */
5083 case PERL_MAGIC_sig:
5084 vtable = &PL_vtbl_sig;
5086 case PERL_MAGIC_sigelem:
5087 vtable = &PL_vtbl_sigelem;
5089 case PERL_MAGIC_taint:
5090 vtable = &PL_vtbl_taint;
5092 case PERL_MAGIC_uvar:
5093 vtable = &PL_vtbl_uvar;
5095 case PERL_MAGIC_vec:
5096 vtable = &PL_vtbl_vec;
5098 case PERL_MAGIC_arylen_p:
5099 case PERL_MAGIC_rhash:
5100 case PERL_MAGIC_symtab:
5101 case PERL_MAGIC_vstring:
5104 case PERL_MAGIC_utf8:
5105 vtable = &PL_vtbl_utf8;
5107 case PERL_MAGIC_substr:
5108 vtable = &PL_vtbl_substr;
5110 case PERL_MAGIC_defelem:
5111 vtable = &PL_vtbl_defelem;
5113 case PERL_MAGIC_arylen:
5114 vtable = &PL_vtbl_arylen;
5116 case PERL_MAGIC_pos:
5117 vtable = &PL_vtbl_pos;
5119 case PERL_MAGIC_backref:
5120 vtable = &PL_vtbl_backref;
5122 case PERL_MAGIC_hintselem:
5123 vtable = &PL_vtbl_hintselem;
5125 case PERL_MAGIC_ext:
5126 /* Reserved for use by extensions not perl internals. */
5127 /* Useful for attaching extension internal data to perl vars. */
5128 /* Note that multiple extensions may clash if magical scalars */
5129 /* etc holding private data from one are passed to another. */
5133 Perl_croak(aTHX_ "Don't know how to handle magic of type \\%o", how);
5136 /* Rest of work is done else where */
5137 mg = sv_magicext(sv,obj,how,vtable,name,namlen);
5140 case PERL_MAGIC_taint:
5143 case PERL_MAGIC_ext:
5144 case PERL_MAGIC_dbfile:
5151 =for apidoc sv_unmagic
5153 Removes all magic of type C<type> from an SV.
5159 Perl_sv_unmagic(pTHX_ SV *const sv, const int type)
5164 PERL_ARGS_ASSERT_SV_UNMAGIC;
5166 if (SvTYPE(sv) < SVt_PVMG || !SvMAGIC(sv))
5168 mgp = &(((XPVMG*) SvANY(sv))->xmg_u.xmg_magic);
5169 for (mg = *mgp; mg; mg = *mgp) {
5170 if (mg->mg_type == type) {
5171 const MGVTBL* const vtbl = mg->mg_virtual;
5172 *mgp = mg->mg_moremagic;
5173 if (vtbl && vtbl->svt_free)
5174 CALL_FPTR(vtbl->svt_free)(aTHX_ sv, mg);
5175 if (mg->mg_ptr && mg->mg_type != PERL_MAGIC_regex_global) {
5177 Safefree(mg->mg_ptr);
5178 else if (mg->mg_len == HEf_SVKEY)
5179 SvREFCNT_dec(MUTABLE_SV(mg->mg_ptr));
5180 else if (mg->mg_type == PERL_MAGIC_utf8)
5181 Safefree(mg->mg_ptr);
5183 if (mg->mg_flags & MGf_REFCOUNTED)
5184 SvREFCNT_dec(mg->mg_obj);
5188 mgp = &mg->mg_moremagic;
5192 SvFLAGS(sv) |= (SvFLAGS(sv) & (SVp_IOK|SVp_NOK|SVp_POK)) >> PRIVSHIFT;
5193 SvMAGIC_set(sv, NULL);
5200 =for apidoc sv_rvweaken
5202 Weaken a reference: set the C<SvWEAKREF> flag on this RV; give the
5203 referred-to SV C<PERL_MAGIC_backref> magic if it hasn't already; and
5204 push a back-reference to this RV onto the array of backreferences
5205 associated with that magic. If the RV is magical, set magic will be
5206 called after the RV is cleared.
5212 Perl_sv_rvweaken(pTHX_ SV *const sv)
5216 PERL_ARGS_ASSERT_SV_RVWEAKEN;
5218 if (!SvOK(sv)) /* let undefs pass */
5221 Perl_croak(aTHX_ "Can't weaken a nonreference");
5222 else if (SvWEAKREF(sv)) {
5223 if (ckWARN(WARN_MISC))
5224 Perl_warner(aTHX_ packWARN(WARN_MISC), "Reference is already weak");
5228 Perl_sv_add_backref(aTHX_ tsv, sv);
5234 /* Give tsv backref magic if it hasn't already got it, then push a
5235 * back-reference to sv onto the array associated with the backref magic.
5238 /* A discussion about the backreferences array and its refcount:
5240 * The AV holding the backreferences is pointed to either as the mg_obj of
5241 * PERL_MAGIC_backref, or in the specific case of a HV that has the hv_aux
5242 * structure, from the xhv_backreferences field. (A HV without hv_aux will
5243 * have the standard magic instead.) The array is created with a refcount
5244 * of 2. This means that if during global destruction the array gets
5245 * picked on first to have its refcount decremented by the random zapper,
5246 * it won't actually be freed, meaning it's still theere for when its
5247 * parent gets freed.
5248 * When the parent SV is freed, in the case of magic, the magic is freed,
5249 * Perl_magic_killbackrefs is called which decrements one refcount, then
5250 * mg_obj is freed which kills the second count.
5251 * In the vase of a HV being freed, one ref is removed by
5252 * Perl_hv_kill_backrefs, the other by Perl_sv_kill_backrefs, which it
5257 Perl_sv_add_backref(pTHX_ SV *const tsv, SV *const sv)
5262 PERL_ARGS_ASSERT_SV_ADD_BACKREF;
5264 if (SvTYPE(tsv) == SVt_PVHV) {
5265 AV **const avp = Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5269 /* There is no AV in the offical place - try a fixup. */
5270 MAGIC *const mg = mg_find(tsv, PERL_MAGIC_backref);
5273 /* Aha. They've got it stowed in magic. Bring it back. */
5274 av = MUTABLE_AV(mg->mg_obj);
5275 /* Stop mg_free decreasing the refernce count. */
5277 /* Stop mg_free even calling the destructor, given that
5278 there's no AV to free up. */
5280 sv_unmagic(tsv, PERL_MAGIC_backref);
5284 SvREFCNT_inc_simple_void(av); /* see discussion above */
5289 const MAGIC *const mg
5290 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
5292 av = MUTABLE_AV(mg->mg_obj);
5296 sv_magic(tsv, MUTABLE_SV(av), PERL_MAGIC_backref, NULL, 0);
5297 /* av now has a refcnt of 2; see discussion above */
5300 if (AvFILLp(av) >= AvMAX(av)) {
5301 av_extend(av, AvFILLp(av)+1);
5303 AvARRAY(av)[++AvFILLp(av)] = sv; /* av_push() */
5306 /* delete a back-reference to ourselves from the backref magic associated
5307 * with the SV we point to.
5311 S_sv_del_backref(pTHX_ SV *const tsv, SV *const sv)
5318 PERL_ARGS_ASSERT_SV_DEL_BACKREF;
5320 if (SvTYPE(tsv) == SVt_PVHV && SvOOK(tsv)) {
5321 av = *Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5322 /* We mustn't attempt to "fix up" the hash here by moving the
5323 backreference array back to the hv_aux structure, as that is stored
5324 in the main HvARRAY(), and hfreentries assumes that no-one
5325 reallocates HvARRAY() while it is running. */
5328 const MAGIC *const mg
5329 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
5331 av = MUTABLE_AV(mg->mg_obj);
5335 Perl_croak(aTHX_ "panic: del_backref");
5337 assert(!SvIS_FREED(av));
5340 /* We shouldn't be in here more than once, but for paranoia reasons lets
5342 for (i = AvFILLp(av); i >= 0; i--) {
5344 const SSize_t fill = AvFILLp(av);
5346 /* We weren't the last entry.
5347 An unordered list has this property that you can take the
5348 last element off the end to fill the hole, and it's still
5349 an unordered list :-)
5354 AvFILLp(av) = fill - 1;
5360 Perl_sv_kill_backrefs(pTHX_ SV *const sv, AV *const av)
5362 SV **svp = AvARRAY(av);
5364 PERL_ARGS_ASSERT_SV_KILL_BACKREFS;
5365 PERL_UNUSED_ARG(sv);
5367 assert(!svp || !SvIS_FREED(av));
5369 SV *const *const last = svp + AvFILLp(av);
5371 while (svp <= last) {
5373 SV *const referrer = *svp;
5374 if (SvWEAKREF(referrer)) {
5375 /* XXX Should we check that it hasn't changed? */
5376 SvRV_set(referrer, 0);
5378 SvWEAKREF_off(referrer);
5379 SvSETMAGIC(referrer);
5380 } else if (SvTYPE(referrer) == SVt_PVGV ||
5381 SvTYPE(referrer) == SVt_PVLV) {
5382 /* You lookin' at me? */
5383 assert(GvSTASH(referrer));
5384 assert(GvSTASH(referrer) == (const HV *)sv);
5385 GvSTASH(referrer) = 0;
5388 "panic: magic_killbackrefs (flags=%"UVxf")",
5389 (UV)SvFLAGS(referrer));
5397 SvREFCNT_dec(av); /* remove extra count added by sv_add_backref() */
5402 =for apidoc sv_insert
5404 Inserts a string at the specified offset/length within the SV. Similar to
5405 the Perl substr() function. Handles get magic.
5407 =for apidoc sv_insert_flags
5409 Same as C<sv_insert>, but the extra C<flags> are passed the C<SvPV_force_flags> that applies to C<bigstr>.
5415 Perl_sv_insert_flags(pTHX_ SV *const bigstr, const STRLEN offset, const STRLEN len, const char *const little, const STRLEN littlelen, const U32 flags)
5420 register char *midend;
5421 register char *bigend;
5425 PERL_ARGS_ASSERT_SV_INSERT_FLAGS;
5428 Perl_croak(aTHX_ "Can't modify non-existent substring");
5429 SvPV_force_flags(bigstr, curlen, flags);
5430 (void)SvPOK_only_UTF8(bigstr);
5431 if (offset + len > curlen) {
5432 SvGROW(bigstr, offset+len+1);
5433 Zero(SvPVX(bigstr)+curlen, offset+len-curlen, char);
5434 SvCUR_set(bigstr, offset+len);
5438 i = littlelen - len;
5439 if (i > 0) { /* string might grow */
5440 big = SvGROW(bigstr, SvCUR(bigstr) + i + 1);
5441 mid = big + offset + len;
5442 midend = bigend = big + SvCUR(bigstr);
5445 while (midend > mid) /* shove everything down */
5446 *--bigend = *--midend;
5447 Move(little,big+offset,littlelen,char);
5448 SvCUR_set(bigstr, SvCUR(bigstr) + i);
5453 Move(little,SvPVX(bigstr)+offset,len,char);
5458 big = SvPVX(bigstr);
5461 bigend = big + SvCUR(bigstr);
5463 if (midend > bigend)
5464 Perl_croak(aTHX_ "panic: sv_insert");
5466 if (mid - big > bigend - midend) { /* faster to shorten from end */
5468 Move(little, mid, littlelen,char);
5471 i = bigend - midend;
5473 Move(midend, mid, i,char);
5477 SvCUR_set(bigstr, mid - big);
5479 else if ((i = mid - big)) { /* faster from front */
5480 midend -= littlelen;
5482 Move(big, midend - i, i, char);
5483 sv_chop(bigstr,midend-i);
5485 Move(little, mid, littlelen,char);
5487 else if (littlelen) {
5488 midend -= littlelen;
5489 sv_chop(bigstr,midend);
5490 Move(little,midend,littlelen,char);
5493 sv_chop(bigstr,midend);
5499 =for apidoc sv_replace
5501 Make the first argument a copy of the second, then delete the original.
5502 The target SV physically takes over ownership of the body of the source SV
5503 and inherits its flags; however, the target keeps any magic it owns,
5504 and any magic in the source is discarded.
5505 Note that this is a rather specialist SV copying operation; most of the
5506 time you'll want to use C<sv_setsv> or one of its many macro front-ends.
5512 Perl_sv_replace(pTHX_ register SV *const sv, register SV *const nsv)
5515 const U32 refcnt = SvREFCNT(sv);
5517 PERL_ARGS_ASSERT_SV_REPLACE;
5519 SV_CHECK_THINKFIRST_COW_DROP(sv);
5520 if (SvREFCNT(nsv) != 1) {
5521 Perl_croak(aTHX_ "panic: reference miscount on nsv in sv_replace() (%"
5522 UVuf " != 1)", (UV) SvREFCNT(nsv));
5524 if (SvMAGICAL(sv)) {
5528 sv_upgrade(nsv, SVt_PVMG);
5529 SvMAGIC_set(nsv, SvMAGIC(sv));
5530 SvFLAGS(nsv) |= SvMAGICAL(sv);
5532 SvMAGIC_set(sv, NULL);
5536 assert(!SvREFCNT(sv));
5537 #ifdef DEBUG_LEAKING_SCALARS
5538 sv->sv_flags = nsv->sv_flags;
5539 sv->sv_any = nsv->sv_any;
5540 sv->sv_refcnt = nsv->sv_refcnt;
5541 sv->sv_u = nsv->sv_u;
5543 StructCopy(nsv,sv,SV);
5545 if(SvTYPE(sv) == SVt_IV) {
5547 = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
5551 #ifdef PERL_OLD_COPY_ON_WRITE
5552 if (SvIsCOW_normal(nsv)) {
5553 /* We need to follow the pointers around the loop to make the
5554 previous SV point to sv, rather than nsv. */
5557 while ((next = SV_COW_NEXT_SV(current)) != nsv) {
5560 assert(SvPVX_const(current) == SvPVX_const(nsv));
5562 /* Make the SV before us point to the SV after us. */
5564 PerlIO_printf(Perl_debug_log, "previous is\n");
5566 PerlIO_printf(Perl_debug_log,
5567 "move it from 0x%"UVxf" to 0x%"UVxf"\n",
5568 (UV) SV_COW_NEXT_SV(current), (UV) sv);
5570 SV_COW_NEXT_SV_SET(current, sv);
5573 SvREFCNT(sv) = refcnt;
5574 SvFLAGS(nsv) |= SVTYPEMASK; /* Mark as freed */
5580 =for apidoc sv_clear
5582 Clear an SV: call any destructors, free up any memory used by the body,
5583 and free the body itself. The SV's head is I<not> freed, although
5584 its type is set to all 1's so that it won't inadvertently be assumed
5585 to be live during global destruction etc.
5586 This function should only be called when REFCNT is zero. Most of the time
5587 you'll want to call C<sv_free()> (or its macro wrapper C<SvREFCNT_dec>)
5594 Perl_sv_clear(pTHX_ register SV *const sv)
5597 const U32 type = SvTYPE(sv);
5598 const struct body_details *const sv_type_details
5599 = bodies_by_type + type;
5602 PERL_ARGS_ASSERT_SV_CLEAR;
5603 assert(SvREFCNT(sv) == 0);
5604 assert(SvTYPE(sv) != SVTYPEMASK);
5606 if (type <= SVt_IV) {
5607 /* See the comment in sv.h about the collusion between this early
5608 return and the overloading of the NULL and IV slots in the size
5611 SV * const target = SvRV(sv);
5613 sv_del_backref(target, sv);
5615 SvREFCNT_dec(target);
5617 SvFLAGS(sv) &= SVf_BREAK;
5618 SvFLAGS(sv) |= SVTYPEMASK;
5623 if (PL_defstash && /* Still have a symbol table? */
5630 stash = SvSTASH(sv);
5631 destructor = StashHANDLER(stash,DESTROY);
5633 /* A constant subroutine can have no side effects, so
5634 don't bother calling it. */
5635 && !CvCONST(destructor)
5636 /* Don't bother calling an empty destructor */
5637 && (CvISXSUB(destructor)
5638 || CvSTART(destructor)->op_next->op_type != OP_LEAVESUB))
5640 SV* const tmpref = newRV(sv);
5641 SvREADONLY_on(tmpref); /* DESTROY() could be naughty */
5643 PUSHSTACKi(PERLSI_DESTROY);
5648 call_sv(MUTABLE_SV(destructor), G_DISCARD|G_EVAL|G_KEEPERR|G_VOID);
5654 if(SvREFCNT(tmpref) < 2) {
5655 /* tmpref is not kept alive! */
5657 SvRV_set(tmpref, NULL);
5660 SvREFCNT_dec(tmpref);
5662 } while (SvOBJECT(sv) && SvSTASH(sv) != stash);
5666 if (PL_in_clean_objs)
5667 Perl_croak(aTHX_ "DESTROY created new reference to dead object '%s'",
5669 /* DESTROY gave object new lease on life */
5675 SvREFCNT_dec(SvSTASH(sv)); /* possibly of changed persuasion */
5676 SvOBJECT_off(sv); /* Curse the object. */
5677 if (type != SVt_PVIO)
5678 --PL_sv_objcount; /* XXX Might want something more general */
5681 if (type >= SVt_PVMG) {
5682 if (type == SVt_PVMG && SvPAD_OUR(sv)) {
5683 SvREFCNT_dec(SvOURSTASH(sv));
5684 } else if (SvMAGIC(sv))
5686 if (type == SVt_PVMG && SvPAD_TYPED(sv))
5687 SvREFCNT_dec(SvSTASH(sv));
5690 /* case SVt_BIND: */
5693 IoIFP(sv) != PerlIO_stdin() &&
5694 IoIFP(sv) != PerlIO_stdout() &&
5695 IoIFP(sv) != PerlIO_stderr())
5697 io_close(MUTABLE_IO(sv), FALSE);
5699 if (IoDIRP(sv) && !(IoFLAGS(sv) & IOf_FAKE_DIRP))
5700 PerlDir_close(IoDIRP(sv));
5701 IoDIRP(sv) = (DIR*)NULL;
5702 Safefree(IoTOP_NAME(sv));
5703 Safefree(IoFMT_NAME(sv));
5704 Safefree(IoBOTTOM_NAME(sv));
5707 /* FIXME for plugins */
5708 pregfree2((REGEXP*) sv);
5712 cv_undef(MUTABLE_CV(sv));
5715 if (PL_last_swash_hv == (const HV *)sv) {
5716 PL_last_swash_hv = NULL;
5718 Perl_hv_kill_backrefs(aTHX_ MUTABLE_HV(sv));
5719 hv_undef(MUTABLE_HV(sv));
5722 if (PL_comppad == MUTABLE_AV(sv)) {
5726 av_undef(MUTABLE_AV(sv));
5729 if (LvTYPE(sv) == 'T') { /* for tie: return HE to pool */
5730 SvREFCNT_dec(HeKEY_sv((HE*)LvTARG(sv)));
5731 HeNEXT((HE*)LvTARG(sv)) = PL_hv_fetch_ent_mh;
5732 PL_hv_fetch_ent_mh = (HE*)LvTARG(sv);
5734 else if (LvTYPE(sv) != 't') /* unless tie: unrefcnted fake SV** */
5735 SvREFCNT_dec(LvTARG(sv));
5737 if (isGV_with_GP(sv)) {
5738 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
5739 && HvNAME_get(stash))
5740 mro_method_changed_in(stash);
5741 gp_free(MUTABLE_GV(sv));
5743 unshare_hek(GvNAME_HEK(sv));
5744 /* If we're in a stash, we don't own a reference to it. However it does
5745 have a back reference to us, which needs to be cleared. */
5746 if (!SvVALID(sv) && (stash = GvSTASH(sv)))
5747 sv_del_backref(MUTABLE_SV(stash), sv);
5749 /* FIXME. There are probably more unreferenced pointers to SVs in the
5750 interpreter struct that we should check and tidy in a similar
5752 if ((const GV *)sv == PL_last_in_gv)
5753 PL_last_in_gv = NULL;
5759 /* Don't bother with SvOOK_off(sv); as we're only going to free it. */
5762 SvOOK_offset(sv, offset);
5763 SvPV_set(sv, SvPVX_mutable(sv) - offset);
5764 /* Don't even bother with turning off the OOK flag. */
5767 SV * const target = SvRV(sv);
5769 sv_del_backref(target, sv);
5771 SvREFCNT_dec(target);
5773 #ifdef PERL_OLD_COPY_ON_WRITE
5774 else if (SvPVX_const(sv)) {
5777 PerlIO_printf(Perl_debug_log, "Copy on write: clear\n");
5781 sv_release_COW(sv, SvPVX_const(sv), SV_COW_NEXT_SV(sv));
5783 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
5787 } else if (SvLEN(sv)) {
5788 Safefree(SvPVX_const(sv));
5792 else if (SvPVX_const(sv) && SvLEN(sv))
5793 Safefree(SvPVX_mutable(sv));
5794 else if (SvPVX_const(sv) && SvREADONLY(sv) && SvFAKE(sv)) {
5795 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
5804 SvFLAGS(sv) &= SVf_BREAK;
5805 SvFLAGS(sv) |= SVTYPEMASK;
5807 if (sv_type_details->arena) {
5808 del_body(((char *)SvANY(sv) + sv_type_details->offset),
5809 &PL_body_roots[type]);
5811 else if (sv_type_details->body_size) {
5812 my_safefree(SvANY(sv));
5817 =for apidoc sv_newref
5819 Increment an SV's reference count. Use the C<SvREFCNT_inc()> wrapper
5826 Perl_sv_newref(pTHX_ SV *const sv)
5828 PERL_UNUSED_CONTEXT;
5837 Decrement an SV's reference count, and if it drops to zero, call
5838 C<sv_clear> to invoke destructors and free up any memory used by
5839 the body; finally, deallocate the SV's head itself.
5840 Normally called via a wrapper macro C<SvREFCNT_dec>.
5846 Perl_sv_free(pTHX_ SV *const sv)
5851 if (SvREFCNT(sv) == 0) {
5852 if (SvFLAGS(sv) & SVf_BREAK)
5853 /* this SV's refcnt has been artificially decremented to
5854 * trigger cleanup */
5856 if (PL_in_clean_all) /* All is fair */
5858 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
5859 /* make sure SvREFCNT(sv)==0 happens very seldom */
5860 SvREFCNT(sv) = (~(U32)0)/2;
5863 if (ckWARN_d(WARN_INTERNAL)) {
5864 #ifdef DEBUG_LEAKING_SCALARS_FORK_DUMP
5865 Perl_dump_sv_child(aTHX_ sv);
5867 #ifdef DEBUG_LEAKING_SCALARS
5870 #ifdef DEBUG_LEAKING_SCALARS_ABORT
5871 if (PL_warnhook == PERL_WARNHOOK_FATAL
5872 || ckDEAD(packWARN(WARN_INTERNAL))) {
5873 /* Don't let Perl_warner cause us to escape our fate: */
5877 /* This may not return: */
5878 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
5879 "Attempt to free unreferenced scalar: SV 0x%"UVxf
5880 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
5883 #ifdef DEBUG_LEAKING_SCALARS_ABORT
5888 if (--(SvREFCNT(sv)) > 0)
5890 Perl_sv_free2(aTHX_ sv);
5894 Perl_sv_free2(pTHX_ SV *const sv)
5898 PERL_ARGS_ASSERT_SV_FREE2;
5902 if (ckWARN_d(WARN_DEBUGGING))
5903 Perl_warner(aTHX_ packWARN(WARN_DEBUGGING),
5904 "Attempt to free temp prematurely: SV 0x%"UVxf
5905 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
5909 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
5910 /* make sure SvREFCNT(sv)==0 happens very seldom */
5911 SvREFCNT(sv) = (~(U32)0)/2;
5922 Returns the length of the string in the SV. Handles magic and type
5923 coercion. See also C<SvCUR>, which gives raw access to the xpv_cur slot.
5929 Perl_sv_len(pTHX_ register SV *const sv)
5937 len = mg_length(sv);
5939 (void)SvPV_const(sv, len);
5944 =for apidoc sv_len_utf8
5946 Returns the number of characters in the string in an SV, counting wide
5947 UTF-8 bytes as a single character. Handles magic and type coercion.
5953 * The length is cached in PERL_MAGIC_utf8, in the mg_len field. Also the
5954 * mg_ptr is used, by sv_pos_u2b() and sv_pos_b2u() - see the comments below.
5955 * (Note that the mg_len is not the length of the mg_ptr field.
5956 * This allows the cache to store the character length of the string without
5957 * needing to malloc() extra storage to attach to the mg_ptr.)
5962 Perl_sv_len_utf8(pTHX_ register SV *const sv)
5968 return mg_length(sv);
5972 const U8 *s = (U8*)SvPV_const(sv, len);
5976 MAGIC *mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_utf8) : NULL;
5978 if (mg && mg->mg_len != -1) {
5980 if (PL_utf8cache < 0) {
5981 const STRLEN real = Perl_utf8_length(aTHX_ s, s + len);
5983 /* Need to turn the assertions off otherwise we may
5984 recurse infinitely while printing error messages.
5986 SAVEI8(PL_utf8cache);
5988 Perl_croak(aTHX_ "panic: sv_len_utf8 cache %"UVuf
5989 " real %"UVuf" for %"SVf,
5990 (UV) ulen, (UV) real, SVfARG(sv));
5995 ulen = Perl_utf8_length(aTHX_ s, s + len);
5996 if (!SvREADONLY(sv)) {
5998 mg = sv_magicext(sv, 0, PERL_MAGIC_utf8,
5999 &PL_vtbl_utf8, 0, 0);
6007 return Perl_utf8_length(aTHX_ s, s + len);
6011 /* Walk forwards to find the byte corresponding to the passed in UTF-8
6014 S_sv_pos_u2b_forwards(const U8 *const start, const U8 *const send,
6017 const U8 *s = start;
6019 PERL_ARGS_ASSERT_SV_POS_U2B_FORWARDS;
6021 while (s < send && uoffset--)
6024 /* This is the existing behaviour. Possibly it should be a croak, as
6025 it's actually a bounds error */
6031 /* Given the length of the string in both bytes and UTF-8 characters, decide
6032 whether to walk forwards or backwards to find the byte corresponding to
6033 the passed in UTF-8 offset. */
6035 S_sv_pos_u2b_midway(const U8 *const start, const U8 *send,
6036 const STRLEN uoffset, const STRLEN uend)
6038 STRLEN backw = uend - uoffset;
6040 PERL_ARGS_ASSERT_SV_POS_U2B_MIDWAY;
6042 if (uoffset < 2 * backw) {
6043 /* The assumption is that going forwards is twice the speed of going
6044 forward (that's where the 2 * backw comes from).
6045 (The real figure of course depends on the UTF-8 data.) */
6046 return sv_pos_u2b_forwards(start, send, uoffset);
6051 while (UTF8_IS_CONTINUATION(*send))
6054 return send - start;
6057 /* For the string representation of the given scalar, find the byte
6058 corresponding to the passed in UTF-8 offset. uoffset0 and boffset0
6059 give another position in the string, *before* the sought offset, which
6060 (which is always true, as 0, 0 is a valid pair of positions), which should
6061 help reduce the amount of linear searching.
6062 If *mgp is non-NULL, it should point to the UTF-8 cache magic, which
6063 will be used to reduce the amount of linear searching. The cache will be
6064 created if necessary, and the found value offered to it for update. */
6066 S_sv_pos_u2b_cached(pTHX_ SV *const sv, MAGIC **const mgp, const U8 *const start,
6067 const U8 *const send, const STRLEN uoffset,
6068 STRLEN uoffset0, STRLEN boffset0)
6070 STRLEN boffset = 0; /* Actually always set, but let's keep gcc happy. */
6073 PERL_ARGS_ASSERT_SV_POS_U2B_CACHED;
6075 assert (uoffset >= uoffset0);
6077 if (SvMAGICAL(sv) && !SvREADONLY(sv) && PL_utf8cache
6078 && (*mgp || (*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
6079 if ((*mgp)->mg_ptr) {
6080 STRLEN *cache = (STRLEN *) (*mgp)->mg_ptr;
6081 if (cache[0] == uoffset) {
6082 /* An exact match. */
6085 if (cache[2] == uoffset) {
6086 /* An exact match. */
6090 if (cache[0] < uoffset) {
6091 /* The cache already knows part of the way. */
6092 if (cache[0] > uoffset0) {
6093 /* The cache knows more than the passed in pair */
6094 uoffset0 = cache[0];
6095 boffset0 = cache[1];
6097 if ((*mgp)->mg_len != -1) {
6098 /* And we know the end too. */
6100 + sv_pos_u2b_midway(start + boffset0, send,
6102 (*mgp)->mg_len - uoffset0);
6105 + sv_pos_u2b_forwards(start + boffset0,
6106 send, uoffset - uoffset0);
6109 else if (cache[2] < uoffset) {
6110 /* We're between the two cache entries. */
6111 if (cache[2] > uoffset0) {
6112 /* and the cache knows more than the passed in pair */
6113 uoffset0 = cache[2];
6114 boffset0 = cache[3];
6118 + sv_pos_u2b_midway(start + boffset0,
6121 cache[0] - uoffset0);
6124 + sv_pos_u2b_midway(start + boffset0,
6127 cache[2] - uoffset0);
6131 else if ((*mgp)->mg_len != -1) {
6132 /* If we can take advantage of a passed in offset, do so. */
6133 /* In fact, offset0 is either 0, or less than offset, so don't
6134 need to worry about the other possibility. */
6136 + sv_pos_u2b_midway(start + boffset0, send,
6138 (*mgp)->mg_len - uoffset0);
6143 if (!found || PL_utf8cache < 0) {
6144 const STRLEN real_boffset
6145 = boffset0 + sv_pos_u2b_forwards(start + boffset0,
6146 send, uoffset - uoffset0);
6148 if (found && PL_utf8cache < 0) {
6149 if (real_boffset != boffset) {
6150 /* Need to turn the assertions off otherwise we may recurse
6151 infinitely while printing error messages. */
6152 SAVEI8(PL_utf8cache);
6154 Perl_croak(aTHX_ "panic: sv_pos_u2b_cache cache %"UVuf
6155 " real %"UVuf" for %"SVf,
6156 (UV) boffset, (UV) real_boffset, SVfARG(sv));
6159 boffset = real_boffset;
6163 utf8_mg_pos_cache_update(sv, mgp, boffset, uoffset, send - start);
6169 =for apidoc sv_pos_u2b
6171 Converts the value pointed to by offsetp from a count of UTF-8 chars from
6172 the start of the string, to a count of the equivalent number of bytes; if
6173 lenp is non-zero, it does the same to lenp, but this time starting from
6174 the offset, rather than from the start of the string. Handles magic and
6181 * sv_pos_u2b() uses, like sv_pos_b2u(), the mg_ptr of the potential
6182 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6183 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
6188 Perl_sv_pos_u2b(pTHX_ register SV *const sv, I32 *const offsetp, I32 *const lenp)
6193 PERL_ARGS_ASSERT_SV_POS_U2B;
6198 start = (U8*)SvPV_const(sv, len);
6200 STRLEN uoffset = (STRLEN) *offsetp;
6201 const U8 * const send = start + len;
6203 const STRLEN boffset = sv_pos_u2b_cached(sv, &mg, start, send,
6206 *offsetp = (I32) boffset;
6209 /* Convert the relative offset to absolute. */
6210 const STRLEN uoffset2 = uoffset + (STRLEN) *lenp;
6211 const STRLEN boffset2
6212 = sv_pos_u2b_cached(sv, &mg, start, send, uoffset2,
6213 uoffset, boffset) - boffset;
6227 /* Create and update the UTF8 magic offset cache, with the proffered utf8/
6228 byte length pairing. The (byte) length of the total SV is passed in too,
6229 as blen, because for some (more esoteric) SVs, the call to SvPV_const()
6230 may not have updated SvCUR, so we can't rely on reading it directly.
6232 The proffered utf8/byte length pairing isn't used if the cache already has
6233 two pairs, and swapping either for the proffered pair would increase the
6234 RMS of the intervals between known byte offsets.
6236 The cache itself consists of 4 STRLEN values
6237 0: larger UTF-8 offset
6238 1: corresponding byte offset
6239 2: smaller UTF-8 offset
6240 3: corresponding byte offset
6242 Unused cache pairs have the value 0, 0.
6243 Keeping the cache "backwards" means that the invariant of
6244 cache[0] >= cache[2] is maintained even with empty slots, which means that
6245 the code that uses it doesn't need to worry if only 1 entry has actually
6246 been set to non-zero. It also makes the "position beyond the end of the
6247 cache" logic much simpler, as the first slot is always the one to start
6251 S_utf8_mg_pos_cache_update(pTHX_ SV *const sv, MAGIC **const mgp, const STRLEN byte,
6252 const STRLEN utf8, const STRLEN blen)
6256 PERL_ARGS_ASSERT_UTF8_MG_POS_CACHE_UPDATE;
6262 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, (MGVTBL*)&PL_vtbl_utf8, 0,
6264 (*mgp)->mg_len = -1;
6268 if (!(cache = (STRLEN *)(*mgp)->mg_ptr)) {
6269 Newxz(cache, PERL_MAGIC_UTF8_CACHESIZE * 2, STRLEN);
6270 (*mgp)->mg_ptr = (char *) cache;
6274 if (PL_utf8cache < 0) {
6275 const U8 *start = (const U8 *) SvPVX_const(sv);
6276 const STRLEN realutf8 = utf8_length(start, start + byte);
6278 if (realutf8 != utf8) {
6279 /* Need to turn the assertions off otherwise we may recurse
6280 infinitely while printing error messages. */
6281 SAVEI8(PL_utf8cache);
6283 Perl_croak(aTHX_ "panic: utf8_mg_pos_cache_update cache %"UVuf
6284 " real %"UVuf" for %"SVf, (UV) utf8, (UV) realutf8, SVfARG(sv));
6288 /* Cache is held with the later position first, to simplify the code
6289 that deals with unbounded ends. */
6291 ASSERT_UTF8_CACHE(cache);
6292 if (cache[1] == 0) {
6293 /* Cache is totally empty */
6296 } else if (cache[3] == 0) {
6297 if (byte > cache[1]) {
6298 /* New one is larger, so goes first. */
6299 cache[2] = cache[0];
6300 cache[3] = cache[1];
6308 #define THREEWAY_SQUARE(a,b,c,d) \
6309 ((float)((d) - (c))) * ((float)((d) - (c))) \
6310 + ((float)((c) - (b))) * ((float)((c) - (b))) \
6311 + ((float)((b) - (a))) * ((float)((b) - (a)))
6313 /* Cache has 2 slots in use, and we know three potential pairs.
6314 Keep the two that give the lowest RMS distance. Do the
6315 calcualation in bytes simply because we always know the byte
6316 length. squareroot has the same ordering as the positive value,
6317 so don't bother with the actual square root. */
6318 const float existing = THREEWAY_SQUARE(0, cache[3], cache[1], blen);
6319 if (byte > cache[1]) {
6320 /* New position is after the existing pair of pairs. */
6321 const float keep_earlier
6322 = THREEWAY_SQUARE(0, cache[3], byte, blen);
6323 const float keep_later
6324 = THREEWAY_SQUARE(0, cache[1], byte, blen);
6326 if (keep_later < keep_earlier) {
6327 if (keep_later < existing) {
6328 cache[2] = cache[0];
6329 cache[3] = cache[1];
6335 if (keep_earlier < existing) {
6341 else if (byte > cache[3]) {
6342 /* New position is between the existing pair of pairs. */
6343 const float keep_earlier
6344 = THREEWAY_SQUARE(0, cache[3], byte, blen);
6345 const float keep_later
6346 = THREEWAY_SQUARE(0, byte, cache[1], blen);
6348 if (keep_later < keep_earlier) {
6349 if (keep_later < existing) {
6355 if (keep_earlier < existing) {
6362 /* New position is before the existing pair of pairs. */
6363 const float keep_earlier
6364 = THREEWAY_SQUARE(0, byte, cache[3], blen);
6365 const float keep_later
6366 = THREEWAY_SQUARE(0, byte, cache[1], blen);
6368 if (keep_later < keep_earlier) {
6369 if (keep_later < existing) {
6375 if (keep_earlier < existing) {
6376 cache[0] = cache[2];
6377 cache[1] = cache[3];
6384 ASSERT_UTF8_CACHE(cache);
6387 /* We already know all of the way, now we may be able to walk back. The same
6388 assumption is made as in S_sv_pos_u2b_midway(), namely that walking
6389 backward is half the speed of walking forward. */
6391 S_sv_pos_b2u_midway(pTHX_ const U8 *const s, const U8 *const target,
6392 const U8 *end, STRLEN endu)
6394 const STRLEN forw = target - s;
6395 STRLEN backw = end - target;
6397 PERL_ARGS_ASSERT_SV_POS_B2U_MIDWAY;
6399 if (forw < 2 * backw) {
6400 return utf8_length(s, target);
6403 while (end > target) {
6405 while (UTF8_IS_CONTINUATION(*end)) {
6414 =for apidoc sv_pos_b2u
6416 Converts the value pointed to by offsetp from a count of bytes from the
6417 start of the string, to a count of the equivalent number of UTF-8 chars.
6418 Handles magic and type coercion.
6424 * sv_pos_b2u() uses, like sv_pos_u2b(), the mg_ptr of the potential
6425 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6430 Perl_sv_pos_b2u(pTHX_ register SV *const sv, I32 *const offsetp)
6433 const STRLEN byte = *offsetp;
6434 STRLEN len = 0; /* Actually always set, but let's keep gcc happy. */
6440 PERL_ARGS_ASSERT_SV_POS_B2U;
6445 s = (const U8*)SvPV_const(sv, blen);
6448 Perl_croak(aTHX_ "panic: sv_pos_b2u: bad byte offset");
6452 if (SvMAGICAL(sv) && !SvREADONLY(sv) && PL_utf8cache
6453 && (mg = mg_find(sv, PERL_MAGIC_utf8))) {
6455 STRLEN * const cache = (STRLEN *) mg->mg_ptr;
6456 if (cache[1] == byte) {
6457 /* An exact match. */
6458 *offsetp = cache[0];
6461 if (cache[3] == byte) {
6462 /* An exact match. */
6463 *offsetp = cache[2];
6467 if (cache[1] < byte) {
6468 /* We already know part of the way. */
6469 if (mg->mg_len != -1) {
6470 /* Actually, we know the end too. */
6472 + S_sv_pos_b2u_midway(aTHX_ s + cache[1], send,
6473 s + blen, mg->mg_len - cache[0]);
6475 len = cache[0] + utf8_length(s + cache[1], send);
6478 else if (cache[3] < byte) {
6479 /* We're between the two cached pairs, so we do the calculation
6480 offset by the byte/utf-8 positions for the earlier pair,
6481 then add the utf-8 characters from the string start to
6483 len = S_sv_pos_b2u_midway(aTHX_ s + cache[3], send,
6484 s + cache[1], cache[0] - cache[2])
6488 else { /* cache[3] > byte */
6489 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + cache[3],
6493 ASSERT_UTF8_CACHE(cache);
6495 } else if (mg->mg_len != -1) {
6496 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + blen, mg->mg_len);
6500 if (!found || PL_utf8cache < 0) {
6501 const STRLEN real_len = utf8_length(s, send);
6503 if (found && PL_utf8cache < 0) {
6504 if (len != real_len) {
6505 /* Need to turn the assertions off otherwise we may recurse
6506 infinitely while printing error messages. */
6507 SAVEI8(PL_utf8cache);
6509 Perl_croak(aTHX_ "panic: sv_pos_b2u cache %"UVuf
6510 " real %"UVuf" for %"SVf,
6511 (UV) len, (UV) real_len, SVfARG(sv));
6519 utf8_mg_pos_cache_update(sv, &mg, byte, len, blen);
6525 Returns a boolean indicating whether the strings in the two SVs are
6526 identical. Is UTF-8 and 'use bytes' aware, handles get magic, and will
6527 coerce its args to strings if necessary.
6533 Perl_sv_eq(pTHX_ register SV *sv1, register SV *sv2)
6542 SV* svrecode = NULL;
6549 /* if pv1 and pv2 are the same, second SvPV_const call may
6550 * invalidate pv1, so we may need to make a copy */
6551 if (sv1 == sv2 && (SvTHINKFIRST(sv1) || SvGMAGICAL(sv1))) {
6552 pv1 = SvPV_const(sv1, cur1);
6553 sv1 = newSVpvn_flags(pv1, cur1, SVs_TEMP | SvUTF8(sv2));
6555 pv1 = SvPV_const(sv1, cur1);
6563 pv2 = SvPV_const(sv2, cur2);
6565 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
6566 /* Differing utf8ness.
6567 * Do not UTF8size the comparands as a side-effect. */
6570 svrecode = newSVpvn(pv2, cur2);
6571 sv_recode_to_utf8(svrecode, PL_encoding);
6572 pv2 = SvPV_const(svrecode, cur2);
6575 svrecode = newSVpvn(pv1, cur1);
6576 sv_recode_to_utf8(svrecode, PL_encoding);
6577 pv1 = SvPV_const(svrecode, cur1);
6579 /* Now both are in UTF-8. */
6581 SvREFCNT_dec(svrecode);
6586 bool is_utf8 = TRUE;
6589 /* sv1 is the UTF-8 one,
6590 * if is equal it must be downgrade-able */
6591 char * const pv = (char*)bytes_from_utf8((const U8*)pv1,
6597 /* sv2 is the UTF-8 one,
6598 * if is equal it must be downgrade-able */
6599 char * const pv = (char *)bytes_from_utf8((const U8*)pv2,
6605 /* Downgrade not possible - cannot be eq */
6613 eq = (pv1 == pv2) || memEQ(pv1, pv2, cur1);
6615 SvREFCNT_dec(svrecode);
6625 Compares the strings in two SVs. Returns -1, 0, or 1 indicating whether the
6626 string in C<sv1> is less than, equal to, or greater than the string in
6627 C<sv2>. Is UTF-8 and 'use bytes' aware, handles get magic, and will
6628 coerce its args to strings if necessary. See also C<sv_cmp_locale>.
6634 Perl_sv_cmp(pTHX_ register SV *const sv1, register SV *const sv2)
6638 const char *pv1, *pv2;
6641 SV *svrecode = NULL;
6648 pv1 = SvPV_const(sv1, cur1);
6655 pv2 = SvPV_const(sv2, cur2);
6657 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
6658 /* Differing utf8ness.
6659 * Do not UTF8size the comparands as a side-effect. */
6662 svrecode = newSVpvn(pv2, cur2);
6663 sv_recode_to_utf8(svrecode, PL_encoding);
6664 pv2 = SvPV_const(svrecode, cur2);
6667 pv2 = tpv = (char*)bytes_to_utf8((const U8*)pv2, &cur2);
6672 svrecode = newSVpvn(pv1, cur1);
6673 sv_recode_to_utf8(svrecode, PL_encoding);
6674 pv1 = SvPV_const(svrecode, cur1);
6677 pv1 = tpv = (char*)bytes_to_utf8((const U8*)pv1, &cur1);
6683 cmp = cur2 ? -1 : 0;
6687 const I32 retval = memcmp((const void*)pv1, (const void*)pv2, cur1 < cur2 ? cur1 : cur2);
6690 cmp = retval < 0 ? -1 : 1;
6691 } else if (cur1 == cur2) {
6694 cmp = cur1 < cur2 ? -1 : 1;
6698 SvREFCNT_dec(svrecode);
6706 =for apidoc sv_cmp_locale
6708 Compares the strings in two SVs in a locale-aware manner. Is UTF-8 and
6709 'use bytes' aware, handles get magic, and will coerce its args to strings
6710 if necessary. See also C<sv_cmp>.
6716 Perl_sv_cmp_locale(pTHX_ register SV *const sv1, register SV *const sv2)
6719 #ifdef USE_LOCALE_COLLATE
6725 if (PL_collation_standard)
6729 pv1 = sv1 ? sv_collxfrm(sv1, &len1) : (char *) NULL;
6731 pv2 = sv2 ? sv_collxfrm(sv2, &len2) : (char *) NULL;
6733 if (!pv1 || !len1) {
6744 retval = memcmp((void*)pv1, (void*)pv2, len1 < len2 ? len1 : len2);
6747 return retval < 0 ? -1 : 1;
6750 * When the result of collation is equality, that doesn't mean
6751 * that there are no differences -- some locales exclude some
6752 * characters from consideration. So to avoid false equalities,
6753 * we use the raw string as a tiebreaker.
6759 #endif /* USE_LOCALE_COLLATE */
6761 return sv_cmp(sv1, sv2);
6765 #ifdef USE_LOCALE_COLLATE
6768 =for apidoc sv_collxfrm
6770 Add Collate Transform magic to an SV if it doesn't already have it.
6772 Any scalar variable may carry PERL_MAGIC_collxfrm magic that contains the
6773 scalar data of the variable, but transformed to such a format that a normal
6774 memory comparison can be used to compare the data according to the locale
6781 Perl_sv_collxfrm(pTHX_ SV *const sv, STRLEN *const nxp)
6786 PERL_ARGS_ASSERT_SV_COLLXFRM;
6788 mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_collxfrm) : (MAGIC *) NULL;
6789 if (!mg || !mg->mg_ptr || *(U32*)mg->mg_ptr != PL_collation_ix) {
6795 Safefree(mg->mg_ptr);
6796 s = SvPV_const(sv, len);
6797 if ((xf = mem_collxfrm(s, len, &xlen))) {
6799 #ifdef PERL_OLD_COPY_ON_WRITE
6801 sv_force_normal_flags(sv, 0);
6803 mg = sv_magicext(sv, 0, PERL_MAGIC_collxfrm, &PL_vtbl_collxfrm,
6817 if (mg && mg->mg_ptr) {
6819 return mg->mg_ptr + sizeof(PL_collation_ix);
6827 #endif /* USE_LOCALE_COLLATE */
6832 Get a line from the filehandle and store it into the SV, optionally
6833 appending to the currently-stored string.
6839 Perl_sv_gets(pTHX_ register SV *const sv, register PerlIO *const fp, I32 append)
6844 register STDCHAR rslast;
6845 register STDCHAR *bp;
6850 PERL_ARGS_ASSERT_SV_GETS;
6852 if (SvTHINKFIRST(sv))
6853 sv_force_normal_flags(sv, append ? 0 : SV_COW_DROP_PV);
6854 /* XXX. If you make this PVIV, then copy on write can copy scalars read
6856 However, perlbench says it's slower, because the existing swipe code
6857 is faster than copy on write.
6858 Swings and roundabouts. */
6859 SvUPGRADE(sv, SVt_PV);
6864 if (PerlIO_isutf8(fp)) {
6866 sv_utf8_upgrade_nomg(sv);
6867 sv_pos_u2b(sv,&append,0);
6869 } else if (SvUTF8(sv)) {
6870 SV * const tsv = newSV(0);
6871 sv_gets(tsv, fp, 0);
6872 sv_utf8_upgrade_nomg(tsv);
6873 SvCUR_set(sv,append);
6876 goto return_string_or_null;
6881 if (PerlIO_isutf8(fp))
6884 if (IN_PERL_COMPILETIME) {
6885 /* we always read code in line mode */
6889 else if (RsSNARF(PL_rs)) {
6890 /* If it is a regular disk file use size from stat() as estimate
6891 of amount we are going to read -- may result in mallocing
6892 more memory than we really need if the layers below reduce
6893 the size we read (e.g. CRLF or a gzip layer).
6896 if (!PerlLIO_fstat(PerlIO_fileno(fp), &st) && S_ISREG(st.st_mode)) {
6897 const Off_t offset = PerlIO_tell(fp);
6898 if (offset != (Off_t) -1 && st.st_size + append > offset) {
6899 (void) SvGROW(sv, (STRLEN)((st.st_size - offset) + append + 1));
6905 else if (RsRECORD(PL_rs)) {
6913 /* Grab the size of the record we're getting */
6914 recsize = SvUV(SvRV(PL_rs)); /* RsRECORD() guarantees > 0. */
6915 buffer = SvGROW(sv, (STRLEN)(recsize + append + 1)) + append;
6918 /* VMS wants read instead of fread, because fread doesn't respect */
6919 /* RMS record boundaries. This is not necessarily a good thing to be */
6920 /* doing, but we've got no other real choice - except avoid stdio
6921 as implementation - perhaps write a :vms layer ?
6923 fd = PerlIO_fileno(fp);
6924 if (fd == -1) { /* in-memory file from PerlIO::Scalar */
6925 bytesread = PerlIO_read(fp, buffer, recsize);
6928 bytesread = PerlLIO_read(fd, buffer, recsize);
6931 bytesread = PerlIO_read(fp, buffer, recsize);
6935 SvCUR_set(sv, bytesread + append);
6936 buffer[bytesread] = '\0';
6937 goto return_string_or_null;
6939 else if (RsPARA(PL_rs)) {
6945 /* Get $/ i.e. PL_rs into same encoding as stream wants */
6946 if (PerlIO_isutf8(fp)) {
6947 rsptr = SvPVutf8(PL_rs, rslen);
6950 if (SvUTF8(PL_rs)) {
6951 if (!sv_utf8_downgrade(PL_rs, TRUE)) {
6952 Perl_croak(aTHX_ "Wide character in $/");
6955 rsptr = SvPV_const(PL_rs, rslen);
6959 rslast = rslen ? rsptr[rslen - 1] : '\0';
6961 if (rspara) { /* have to do this both before and after */
6962 do { /* to make sure file boundaries work right */
6965 i = PerlIO_getc(fp);
6969 PerlIO_ungetc(fp,i);
6975 /* See if we know enough about I/O mechanism to cheat it ! */
6977 /* This used to be #ifdef test - it is made run-time test for ease
6978 of abstracting out stdio interface. One call should be cheap
6979 enough here - and may even be a macro allowing compile
6983 if (PerlIO_fast_gets(fp)) {
6986 * We're going to steal some values from the stdio struct
6987 * and put EVERYTHING in the innermost loop into registers.
6989 register STDCHAR *ptr;
6993 #if defined(VMS) && defined(PERLIO_IS_STDIO)
6994 /* An ungetc()d char is handled separately from the regular
6995 * buffer, so we getc() it back out and stuff it in the buffer.
6997 i = PerlIO_getc(fp);
6998 if (i == EOF) return 0;
6999 *(--((*fp)->_ptr)) = (unsigned char) i;
7003 /* Here is some breathtakingly efficient cheating */
7005 cnt = PerlIO_get_cnt(fp); /* get count into register */
7006 /* make sure we have the room */
7007 if ((I32)(SvLEN(sv) - append) <= cnt + 1) {
7008 /* Not room for all of it
7009 if we are looking for a separator and room for some
7011 if (rslen && cnt > 80 && (I32)SvLEN(sv) > append) {
7012 /* just process what we have room for */
7013 shortbuffered = cnt - SvLEN(sv) + append + 1;
7014 cnt -= shortbuffered;
7018 /* remember that cnt can be negative */
7019 SvGROW(sv, (STRLEN)(append + (cnt <= 0 ? 2 : (cnt + 1))));
7024 bp = (STDCHAR*)SvPVX_const(sv) + append; /* move these two too to registers */
7025 ptr = (STDCHAR*)PerlIO_get_ptr(fp);
7026 DEBUG_P(PerlIO_printf(Perl_debug_log,
7027 "Screamer: entering, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7028 DEBUG_P(PerlIO_printf(Perl_debug_log,
7029 "Screamer: entering: PerlIO * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7030 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7031 PTR2UV(PerlIO_has_base(fp) ? PerlIO_get_base(fp) : 0)));
7036 while (cnt > 0) { /* this | eat */
7038 if ((*bp++ = *ptr++) == rslast) /* really | dust */
7039 goto thats_all_folks; /* screams | sed :-) */
7043 Copy(ptr, bp, cnt, char); /* this | eat */
7044 bp += cnt; /* screams | dust */
7045 ptr += cnt; /* louder | sed :-) */
7050 if (shortbuffered) { /* oh well, must extend */
7051 cnt = shortbuffered;
7053 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
7055 SvGROW(sv, SvLEN(sv) + append + cnt + 2);
7056 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
7060 DEBUG_P(PerlIO_printf(Perl_debug_log,
7061 "Screamer: going to getc, ptr=%"UVuf", cnt=%ld\n",
7062 PTR2UV(ptr),(long)cnt));
7063 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* deregisterize cnt and ptr */
7065 DEBUG_P(PerlIO_printf(Perl_debug_log,
7066 "Screamer: pre: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7067 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7068 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7070 /* This used to call 'filbuf' in stdio form, but as that behaves like
7071 getc when cnt <= 0 we use PerlIO_getc here to avoid introducing
7072 another abstraction. */
7073 i = PerlIO_getc(fp); /* get more characters */
7075 DEBUG_P(PerlIO_printf(Perl_debug_log,
7076 "Screamer: post: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7077 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7078 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7080 cnt = PerlIO_get_cnt(fp);
7081 ptr = (STDCHAR*)PerlIO_get_ptr(fp); /* reregisterize cnt and ptr */
7082 DEBUG_P(PerlIO_printf(Perl_debug_log,
7083 "Screamer: after getc, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7085 if (i == EOF) /* all done for ever? */
7086 goto thats_really_all_folks;
7088 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
7090 SvGROW(sv, bpx + cnt + 2);
7091 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
7093 *bp++ = (STDCHAR)i; /* store character from PerlIO_getc */
7095 if (rslen && (STDCHAR)i == rslast) /* all done for now? */
7096 goto thats_all_folks;
7100 if ((rslen > 1 && (STRLEN)(bp - (STDCHAR*)SvPVX_const(sv)) < rslen) ||
7101 memNE((char*)bp - rslen, rsptr, rslen))
7102 goto screamer; /* go back to the fray */
7103 thats_really_all_folks:
7105 cnt += shortbuffered;
7106 DEBUG_P(PerlIO_printf(Perl_debug_log,
7107 "Screamer: quitting, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7108 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* put these back or we're in trouble */
7109 DEBUG_P(PerlIO_printf(Perl_debug_log,
7110 "Screamer: end: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7111 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7112 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7114 SvCUR_set(sv, bp - (STDCHAR*)SvPVX_const(sv)); /* set length */
7115 DEBUG_P(PerlIO_printf(Perl_debug_log,
7116 "Screamer: done, len=%ld, string=|%.*s|\n",
7117 (long)SvCUR(sv),(int)SvCUR(sv),SvPVX_const(sv)));
7121 /*The big, slow, and stupid way. */
7122 #ifdef USE_HEAP_INSTEAD_OF_STACK /* Even slower way. */
7123 STDCHAR *buf = NULL;
7124 Newx(buf, 8192, STDCHAR);
7132 register const STDCHAR * const bpe = buf + sizeof(buf);
7134 while ((i = PerlIO_getc(fp)) != EOF && (*bp++ = (STDCHAR)i) != rslast && bp < bpe)
7135 ; /* keep reading */
7139 cnt = PerlIO_read(fp,(char*)buf, sizeof(buf));
7140 /* Accomodate broken VAXC compiler, which applies U8 cast to
7141 * both args of ?: operator, causing EOF to change into 255
7144 i = (U8)buf[cnt - 1];
7150 cnt = 0; /* we do need to re-set the sv even when cnt <= 0 */
7152 sv_catpvn(sv, (char *) buf, cnt);
7154 sv_setpvn(sv, (char *) buf, cnt);
7156 if (i != EOF && /* joy */
7158 SvCUR(sv) < rslen ||
7159 memNE(SvPVX_const(sv) + SvCUR(sv) - rslen, rsptr, rslen)))
7163 * If we're reading from a TTY and we get a short read,
7164 * indicating that the user hit his EOF character, we need
7165 * to notice it now, because if we try to read from the TTY
7166 * again, the EOF condition will disappear.
7168 * The comparison of cnt to sizeof(buf) is an optimization
7169 * that prevents unnecessary calls to feof().
7173 if (!(cnt < (I32)sizeof(buf) && PerlIO_eof(fp)))
7177 #ifdef USE_HEAP_INSTEAD_OF_STACK
7182 if (rspara) { /* have to do this both before and after */
7183 while (i != EOF) { /* to make sure file boundaries work right */
7184 i = PerlIO_getc(fp);
7186 PerlIO_ungetc(fp,i);
7192 return_string_or_null:
7193 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7199 Auto-increment of the value in the SV, doing string to numeric conversion
7200 if necessary. Handles 'get' magic.
7206 Perl_sv_inc(pTHX_ register SV *const sv)
7215 if (SvTHINKFIRST(sv)) {
7217 sv_force_normal_flags(sv, 0);
7218 if (SvREADONLY(sv)) {
7219 if (IN_PERL_RUNTIME)
7220 Perl_croak(aTHX_ "%s", PL_no_modify);
7224 if (SvAMAGIC(sv) && AMG_CALLun(sv,inc))
7226 i = PTR2IV(SvRV(sv));
7231 flags = SvFLAGS(sv);
7232 if ((flags & (SVp_NOK|SVp_IOK)) == SVp_NOK) {
7233 /* It's (privately or publicly) a float, but not tested as an
7234 integer, so test it to see. */
7236 flags = SvFLAGS(sv);
7238 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
7239 /* It's publicly an integer, or privately an integer-not-float */
7240 #ifdef PERL_PRESERVE_IVUV
7244 if (SvUVX(sv) == UV_MAX)
7245 sv_setnv(sv, UV_MAX_P1);
7247 (void)SvIOK_only_UV(sv);
7248 SvUV_set(sv, SvUVX(sv) + 1);
7250 if (SvIVX(sv) == IV_MAX)
7251 sv_setuv(sv, (UV)IV_MAX + 1);
7253 (void)SvIOK_only(sv);
7254 SvIV_set(sv, SvIVX(sv) + 1);
7259 if (flags & SVp_NOK) {
7260 const NV was = SvNVX(sv);
7261 if (NV_OVERFLOWS_INTEGERS_AT &&
7262 was >= NV_OVERFLOWS_INTEGERS_AT && ckWARN(WARN_IMPRECISION)) {
7263 Perl_warner(aTHX_ packWARN(WARN_IMPRECISION),
7264 "Lost precision when incrementing %" NVff " by 1",
7267 (void)SvNOK_only(sv);
7268 SvNV_set(sv, was + 1.0);
7272 if (!(flags & SVp_POK) || !*SvPVX_const(sv)) {
7273 if ((flags & SVTYPEMASK) < SVt_PVIV)
7274 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV ? SVt_PVIV : SVt_IV));
7275 (void)SvIOK_only(sv);
7280 while (isALPHA(*d)) d++;
7281 while (isDIGIT(*d)) d++;
7283 #ifdef PERL_PRESERVE_IVUV
7284 /* Got to punt this as an integer if needs be, but we don't issue
7285 warnings. Probably ought to make the sv_iv_please() that does
7286 the conversion if possible, and silently. */
7287 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
7288 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
7289 /* Need to try really hard to see if it's an integer.
7290 9.22337203685478e+18 is an integer.
7291 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
7292 so $a="9.22337203685478e+18"; $a+0; $a++
7293 needs to be the same as $a="9.22337203685478e+18"; $a++
7300 /* sv_2iv *should* have made this an NV */
7301 if (flags & SVp_NOK) {
7302 (void)SvNOK_only(sv);
7303 SvNV_set(sv, SvNVX(sv) + 1.0);
7306 /* I don't think we can get here. Maybe I should assert this
7307 And if we do get here I suspect that sv_setnv will croak. NWC
7309 #if defined(USE_LONG_DOUBLE)
7310 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",
7311 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7313 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_inc punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
7314 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7317 #endif /* PERL_PRESERVE_IVUV */
7318 sv_setnv(sv,Atof(SvPVX_const(sv)) + 1.0);
7322 while (d >= SvPVX_const(sv)) {
7330 /* MKS: The original code here died if letters weren't consecutive.
7331 * at least it didn't have to worry about non-C locales. The
7332 * new code assumes that ('z'-'a')==('Z'-'A'), letters are
7333 * arranged in order (although not consecutively) and that only
7334 * [A-Za-z] are accepted by isALPHA in the C locale.
7336 if (*d != 'z' && *d != 'Z') {
7337 do { ++*d; } while (!isALPHA(*d));
7340 *(d--) -= 'z' - 'a';
7345 *(d--) -= 'z' - 'a' + 1;
7349 /* oh,oh, the number grew */
7350 SvGROW(sv, SvCUR(sv) + 2);
7351 SvCUR_set(sv, SvCUR(sv) + 1);
7352 for (d = SvPVX(sv) + SvCUR(sv); d > SvPVX_const(sv); d--)
7363 Auto-decrement of the value in the SV, doing string to numeric conversion
7364 if necessary. Handles 'get' magic.
7370 Perl_sv_dec(pTHX_ register SV *const sv)
7378 if (SvTHINKFIRST(sv)) {
7380 sv_force_normal_flags(sv, 0);
7381 if (SvREADONLY(sv)) {
7382 if (IN_PERL_RUNTIME)
7383 Perl_croak(aTHX_ "%s", PL_no_modify);
7387 if (SvAMAGIC(sv) && AMG_CALLun(sv,dec))
7389 i = PTR2IV(SvRV(sv));
7394 /* Unlike sv_inc we don't have to worry about string-never-numbers
7395 and keeping them magic. But we mustn't warn on punting */
7396 flags = SvFLAGS(sv);
7397 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
7398 /* It's publicly an integer, or privately an integer-not-float */
7399 #ifdef PERL_PRESERVE_IVUV
7403 if (SvUVX(sv) == 0) {
7404 (void)SvIOK_only(sv);
7408 (void)SvIOK_only_UV(sv);
7409 SvUV_set(sv, SvUVX(sv) - 1);
7412 if (SvIVX(sv) == IV_MIN) {
7413 sv_setnv(sv, (NV)IV_MIN);
7417 (void)SvIOK_only(sv);
7418 SvIV_set(sv, SvIVX(sv) - 1);
7423 if (flags & SVp_NOK) {
7426 const NV was = SvNVX(sv);
7427 if (NV_OVERFLOWS_INTEGERS_AT &&
7428 was <= -NV_OVERFLOWS_INTEGERS_AT && ckWARN(WARN_IMPRECISION)) {
7429 Perl_warner(aTHX_ packWARN(WARN_IMPRECISION),
7430 "Lost precision when decrementing %" NVff " by 1",
7433 (void)SvNOK_only(sv);
7434 SvNV_set(sv, was - 1.0);
7438 if (!(flags & SVp_POK)) {
7439 if ((flags & SVTYPEMASK) < SVt_PVIV)
7440 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV) ? SVt_PVIV : SVt_IV);
7442 (void)SvIOK_only(sv);
7445 #ifdef PERL_PRESERVE_IVUV
7447 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
7448 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
7449 /* Need to try really hard to see if it's an integer.
7450 9.22337203685478e+18 is an integer.
7451 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
7452 so $a="9.22337203685478e+18"; $a+0; $a--
7453 needs to be the same as $a="9.22337203685478e+18"; $a--
7460 /* sv_2iv *should* have made this an NV */
7461 if (flags & SVp_NOK) {
7462 (void)SvNOK_only(sv);
7463 SvNV_set(sv, SvNVX(sv) - 1.0);
7466 /* I don't think we can get here. Maybe I should assert this
7467 And if we do get here I suspect that sv_setnv will croak. NWC
7469 #if defined(USE_LONG_DOUBLE)
7470 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",
7471 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7473 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_dec punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
7474 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7478 #endif /* PERL_PRESERVE_IVUV */
7479 sv_setnv(sv,Atof(SvPVX_const(sv)) - 1.0); /* punt */
7483 =for apidoc sv_mortalcopy
7485 Creates a new SV which is a copy of the original SV (using C<sv_setsv>).
7486 The new SV is marked as mortal. It will be destroyed "soon", either by an
7487 explicit call to FREETMPS, or by an implicit call at places such as
7488 statement boundaries. See also C<sv_newmortal> and C<sv_2mortal>.
7493 /* Make a string that will exist for the duration of the expression
7494 * evaluation. Actually, it may have to last longer than that, but
7495 * hopefully we won't free it until it has been assigned to a
7496 * permanent location. */
7499 Perl_sv_mortalcopy(pTHX_ SV *const oldstr)
7505 sv_setsv(sv,oldstr);
7507 PL_tmps_stack[++PL_tmps_ix] = sv;
7513 =for apidoc sv_newmortal
7515 Creates a new null SV which is mortal. The reference count of the SV is
7516 set to 1. It will be destroyed "soon", either by an explicit call to
7517 FREETMPS, or by an implicit call at places such as statement boundaries.
7518 See also C<sv_mortalcopy> and C<sv_2mortal>.
7524 Perl_sv_newmortal(pTHX)
7530 SvFLAGS(sv) = SVs_TEMP;
7532 PL_tmps_stack[++PL_tmps_ix] = sv;
7538 =for apidoc newSVpvn_flags
7540 Creates a new SV and copies a string into it. The reference count for the
7541 SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
7542 string. You are responsible for ensuring that the source string is at least
7543 C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
7544 Currently the only flag bits accepted are C<SVf_UTF8> and C<SVs_TEMP>.
7545 If C<SVs_TEMP> is set, then C<sv2mortal()> is called on the result before
7546 returning. If C<SVf_UTF8> is set, then it will be set on the new SV.
7547 C<newSVpvn_utf8()> is a convenience wrapper for this function, defined as
7549 #define newSVpvn_utf8(s, len, u) \
7550 newSVpvn_flags((s), (len), (u) ? SVf_UTF8 : 0)
7556 Perl_newSVpvn_flags(pTHX_ const char *const s, const STRLEN len, const U32 flags)
7561 /* All the flags we don't support must be zero.
7562 And we're new code so I'm going to assert this from the start. */
7563 assert(!(flags & ~(SVf_UTF8|SVs_TEMP)));
7565 sv_setpvn(sv,s,len);
7566 SvFLAGS(sv) |= (flags & SVf_UTF8);
7567 return (flags & SVs_TEMP) ? sv_2mortal(sv) : sv;
7571 =for apidoc sv_2mortal
7573 Marks an existing SV as mortal. The SV will be destroyed "soon", either
7574 by an explicit call to FREETMPS, or by an implicit call at places such as
7575 statement boundaries. SvTEMP() is turned on which means that the SV's
7576 string buffer can be "stolen" if this SV is copied. See also C<sv_newmortal>
7577 and C<sv_mortalcopy>.
7583 Perl_sv_2mortal(pTHX_ register SV *const sv)
7588 if (SvREADONLY(sv) && SvIMMORTAL(sv))
7591 PL_tmps_stack[++PL_tmps_ix] = sv;
7599 Creates a new SV and copies a string into it. The reference count for the
7600 SV is set to 1. If C<len> is zero, Perl will compute the length using
7601 strlen(). For efficiency, consider using C<newSVpvn> instead.
7607 Perl_newSVpv(pTHX_ const char *const s, const STRLEN len)
7613 sv_setpvn(sv, s, len || s == NULL ? len : strlen(s));
7618 =for apidoc newSVpvn
7620 Creates a new SV and copies a string into it. The reference count for the
7621 SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
7622 string. You are responsible for ensuring that the source string is at least
7623 C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
7629 Perl_newSVpvn(pTHX_ const char *const s, const STRLEN len)
7635 sv_setpvn(sv,s,len);
7640 =for apidoc newSVhek
7642 Creates a new SV from the hash key structure. It will generate scalars that
7643 point to the shared string table where possible. Returns a new (undefined)
7644 SV if the hek is NULL.
7650 Perl_newSVhek(pTHX_ const HEK *const hek)
7660 if (HEK_LEN(hek) == HEf_SVKEY) {
7661 return newSVsv(*(SV**)HEK_KEY(hek));
7663 const int flags = HEK_FLAGS(hek);
7664 if (flags & HVhek_WASUTF8) {
7666 Andreas would like keys he put in as utf8 to come back as utf8
7668 STRLEN utf8_len = HEK_LEN(hek);
7669 const U8 *as_utf8 = bytes_to_utf8 ((U8*)HEK_KEY(hek), &utf8_len);
7670 SV * const sv = newSVpvn ((const char*)as_utf8, utf8_len);
7673 Safefree (as_utf8); /* bytes_to_utf8() allocates a new string */
7675 } else if (flags & (HVhek_REHASH|HVhek_UNSHARED)) {
7676 /* We don't have a pointer to the hv, so we have to replicate the
7677 flag into every HEK. This hv is using custom a hasing
7678 algorithm. Hence we can't return a shared string scalar, as
7679 that would contain the (wrong) hash value, and might get passed
7680 into an hv routine with a regular hash.
7681 Similarly, a hash that isn't using shared hash keys has to have
7682 the flag in every key so that we know not to try to call
7683 share_hek_kek on it. */
7685 SV * const sv = newSVpvn (HEK_KEY(hek), HEK_LEN(hek));
7690 /* This will be overwhelminly the most common case. */
7692 /* Inline most of newSVpvn_share(), because share_hek_hek() is far
7693 more efficient than sharepvn(). */
7697 sv_upgrade(sv, SVt_PV);
7698 SvPV_set(sv, (char *)HEK_KEY(share_hek_hek(hek)));
7699 SvCUR_set(sv, HEK_LEN(hek));
7712 =for apidoc newSVpvn_share
7714 Creates a new SV with its SvPVX_const pointing to a shared string in the string
7715 table. If the string does not already exist in the table, it is created
7716 first. Turns on READONLY and FAKE. If the C<hash> parameter is non-zero, that
7717 value is used; otherwise the hash is computed. The string's hash can be later
7718 be retrieved from the SV with the C<SvSHARED_HASH()> macro. The idea here is
7719 that as the string table is used for shared hash keys these strings will have
7720 SvPVX_const == HeKEY and hash lookup will avoid string compare.
7726 Perl_newSVpvn_share(pTHX_ const char *src, I32 len, U32 hash)
7730 bool is_utf8 = FALSE;
7731 const char *const orig_src = src;
7734 STRLEN tmplen = -len;
7736 /* See the note in hv.c:hv_fetch() --jhi */
7737 src = (char*)bytes_from_utf8((const U8*)src, &tmplen, &is_utf8);
7741 PERL_HASH(hash, src, len);
7743 /* The logic for this is inlined in S_mro_get_linear_isa_dfs(), so if it
7744 changes here, update it there too. */
7745 sv_upgrade(sv, SVt_PV);
7746 SvPV_set(sv, sharepvn(src, is_utf8?-len:len, hash));
7754 if (src != orig_src)
7760 #if defined(PERL_IMPLICIT_CONTEXT)
7762 /* pTHX_ magic can't cope with varargs, so this is a no-context
7763 * version of the main function, (which may itself be aliased to us).
7764 * Don't access this version directly.
7768 Perl_newSVpvf_nocontext(const char *const pat, ...)
7774 PERL_ARGS_ASSERT_NEWSVPVF_NOCONTEXT;
7776 va_start(args, pat);
7777 sv = vnewSVpvf(pat, &args);
7784 =for apidoc newSVpvf
7786 Creates a new SV and initializes it with the string formatted like
7793 Perl_newSVpvf(pTHX_ const char *const pat, ...)
7798 PERL_ARGS_ASSERT_NEWSVPVF;
7800 va_start(args, pat);
7801 sv = vnewSVpvf(pat, &args);
7806 /* backend for newSVpvf() and newSVpvf_nocontext() */
7809 Perl_vnewSVpvf(pTHX_ const char *const pat, va_list *const args)
7814 PERL_ARGS_ASSERT_VNEWSVPVF;
7817 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
7824 Creates a new SV and copies a floating point value into it.
7825 The reference count for the SV is set to 1.
7831 Perl_newSVnv(pTHX_ const NV n)
7844 Creates a new SV and copies an integer into it. The reference count for the
7851 Perl_newSViv(pTHX_ const IV i)
7864 Creates a new SV and copies an unsigned integer into it.
7865 The reference count for the SV is set to 1.
7871 Perl_newSVuv(pTHX_ const UV u)
7882 =for apidoc newSV_type
7884 Creates a new SV, of the type specified. The reference count for the new SV
7891 Perl_newSV_type(pTHX_ const svtype type)
7896 sv_upgrade(sv, type);
7901 =for apidoc newRV_noinc
7903 Creates an RV wrapper for an SV. The reference count for the original
7904 SV is B<not> incremented.
7910 Perl_newRV_noinc(pTHX_ SV *const tmpRef)
7913 register SV *sv = newSV_type(SVt_IV);
7915 PERL_ARGS_ASSERT_NEWRV_NOINC;
7918 SvRV_set(sv, tmpRef);
7923 /* newRV_inc is the official function name to use now.
7924 * newRV_inc is in fact #defined to newRV in sv.h
7928 Perl_newRV(pTHX_ SV *const sv)
7932 PERL_ARGS_ASSERT_NEWRV;
7934 return newRV_noinc(SvREFCNT_inc_simple_NN(sv));
7940 Creates a new SV which is an exact duplicate of the original SV.
7947 Perl_newSVsv(pTHX_ register SV *const old)
7954 if (SvTYPE(old) == SVTYPEMASK) {
7955 if (ckWARN_d(WARN_INTERNAL))
7956 Perl_warner(aTHX_ packWARN(WARN_INTERNAL), "semi-panic: attempt to dup freed string");
7960 /* SV_GMAGIC is the default for sv_setv()
7961 SV_NOSTEAL prevents TEMP buffers being, well, stolen, and saves games
7962 with SvTEMP_off and SvTEMP_on round a call to sv_setsv. */
7963 sv_setsv_flags(sv, old, SV_GMAGIC | SV_NOSTEAL);
7968 =for apidoc sv_reset
7970 Underlying implementation for the C<reset> Perl function.
7971 Note that the perl-level function is vaguely deprecated.
7977 Perl_sv_reset(pTHX_ register const char *s, HV *const stash)
7980 char todo[PERL_UCHAR_MAX+1];
7982 PERL_ARGS_ASSERT_SV_RESET;
7987 if (!*s) { /* reset ?? searches */
7988 MAGIC * const mg = mg_find((const SV *)stash, PERL_MAGIC_symtab);
7990 const U32 count = mg->mg_len / sizeof(PMOP**);
7991 PMOP **pmp = (PMOP**) mg->mg_ptr;
7992 PMOP *const *const end = pmp + count;
7996 SvREADONLY_off(PL_regex_pad[(*pmp)->op_pmoffset]);
7998 (*pmp)->op_pmflags &= ~PMf_USED;
8006 /* reset variables */
8008 if (!HvARRAY(stash))
8011 Zero(todo, 256, char);
8014 I32 i = (unsigned char)*s;
8018 max = (unsigned char)*s++;
8019 for ( ; i <= max; i++) {
8022 for (i = 0; i <= (I32) HvMAX(stash); i++) {
8024 for (entry = HvARRAY(stash)[i];
8026 entry = HeNEXT(entry))
8031 if (!todo[(U8)*HeKEY(entry)])
8033 gv = MUTABLE_GV(HeVAL(entry));
8036 if (SvTHINKFIRST(sv)) {
8037 if (!SvREADONLY(sv) && SvROK(sv))
8039 /* XXX Is this continue a bug? Why should THINKFIRST
8040 exempt us from resetting arrays and hashes? */
8044 if (SvTYPE(sv) >= SVt_PV) {
8046 if (SvPVX_const(sv) != NULL)
8054 if (GvHV(gv) && !HvNAME_get(GvHV(gv))) {
8056 Perl_die(aTHX_ "Can't reset %%ENV on this system");
8059 # if defined(USE_ENVIRON_ARRAY)
8062 # endif /* USE_ENVIRON_ARRAY */
8073 Using various gambits, try to get an IO from an SV: the IO slot if its a
8074 GV; or the recursive result if we're an RV; or the IO slot of the symbol
8075 named after the PV if we're a string.
8081 Perl_sv_2io(pTHX_ SV *const sv)
8086 PERL_ARGS_ASSERT_SV_2IO;
8088 switch (SvTYPE(sv)) {
8090 io = MUTABLE_IO(sv);
8093 if (isGV_with_GP(sv)) {
8094 gv = MUTABLE_GV(sv);
8097 Perl_croak(aTHX_ "Bad filehandle: %s", GvNAME(gv));
8103 Perl_croak(aTHX_ PL_no_usym, "filehandle");
8105 return sv_2io(SvRV(sv));
8106 gv = gv_fetchsv(sv, 0, SVt_PVIO);
8112 Perl_croak(aTHX_ "Bad filehandle: %"SVf, SVfARG(sv));
8121 Using various gambits, try to get a CV from an SV; in addition, try if
8122 possible to set C<*st> and C<*gvp> to the stash and GV associated with it.
8123 The flags in C<lref> are passed to sv_fetchsv.
8129 Perl_sv_2cv(pTHX_ SV *sv, HV **const st, GV **const gvp, const I32 lref)
8135 PERL_ARGS_ASSERT_SV_2CV;
8142 switch (SvTYPE(sv)) {
8146 return MUTABLE_CV(sv);
8153 if (isGV_with_GP(sv)) {
8154 gv = MUTABLE_GV(sv);
8163 SV * const *sp = &sv; /* Used in tryAMAGICunDEREF macro. */
8165 tryAMAGICunDEREF(to_cv);
8168 if (SvTYPE(sv) == SVt_PVCV) {
8169 cv = MUTABLE_CV(sv);
8174 else if(isGV_with_GP(sv))
8175 gv = MUTABLE_GV(sv);
8177 Perl_croak(aTHX_ "Not a subroutine reference");
8179 else if (isGV_with_GP(sv)) {
8181 gv = MUTABLE_GV(sv);
8184 gv = gv_fetchsv(sv, lref, SVt_PVCV); /* Calls get magic */
8190 /* Some flags to gv_fetchsv mean don't really create the GV */
8191 if (!isGV_with_GP(gv)) {
8197 if (lref && !GvCVu(gv)) {
8201 gv_efullname3(tmpsv, gv, NULL);
8202 /* XXX this is probably not what they think they're getting.
8203 * It has the same effect as "sub name;", i.e. just a forward
8205 newSUB(start_subparse(FALSE, 0),
8206 newSVOP(OP_CONST, 0, tmpsv),
8210 Perl_croak(aTHX_ "Unable to create sub named \"%"SVf"\"",
8211 SVfARG(SvOK(sv) ? sv : &PL_sv_no));
8220 Returns true if the SV has a true value by Perl's rules.
8221 Use the C<SvTRUE> macro instead, which may call C<sv_true()> or may
8222 instead use an in-line version.
8228 Perl_sv_true(pTHX_ register SV *const sv)
8233 register const XPV* const tXpv = (XPV*)SvANY(sv);
8235 (tXpv->xpv_cur > 1 ||
8236 (tXpv->xpv_cur && *sv->sv_u.svu_pv != '0')))
8243 return SvIVX(sv) != 0;
8246 return SvNVX(sv) != 0.0;
8248 return sv_2bool(sv);
8254 =for apidoc sv_pvn_force
8256 Get a sensible string out of the SV somehow.
8257 A private implementation of the C<SvPV_force> macro for compilers which
8258 can't cope with complex macro expressions. Always use the macro instead.
8260 =for apidoc sv_pvn_force_flags
8262 Get a sensible string out of the SV somehow.
8263 If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<sv> if
8264 appropriate, else not. C<sv_pvn_force> and C<sv_pvn_force_nomg> are
8265 implemented in terms of this function.
8266 You normally want to use the various wrapper macros instead: see
8267 C<SvPV_force> and C<SvPV_force_nomg>
8273 Perl_sv_pvn_force_flags(pTHX_ SV *const sv, STRLEN *const lp, const I32 flags)
8277 PERL_ARGS_ASSERT_SV_PVN_FORCE_FLAGS;
8279 if (SvTHINKFIRST(sv) && !SvROK(sv))
8280 sv_force_normal_flags(sv, 0);
8290 if (SvREADONLY(sv) && !(flags & SV_MUTABLE_RETURN)) {
8291 const char * const ref = sv_reftype(sv,0);
8293 Perl_croak(aTHX_ "Can't coerce readonly %s to string in %s",
8294 ref, OP_NAME(PL_op));
8296 Perl_croak(aTHX_ "Can't coerce readonly %s to string", ref);
8298 if ((SvTYPE(sv) > SVt_PVLV && SvTYPE(sv) != SVt_PVFM)
8299 || isGV_with_GP(sv))
8300 Perl_croak(aTHX_ "Can't coerce %s to string in %s", sv_reftype(sv,0),
8302 s = sv_2pv_flags(sv, &len, flags);
8306 if (s != SvPVX_const(sv)) { /* Almost, but not quite, sv_setpvn() */
8309 SvUPGRADE(sv, SVt_PV); /* Never FALSE */
8310 SvGROW(sv, len + 1);
8311 Move(s,SvPVX(sv),len,char);
8313 SvPVX(sv)[len] = '\0';
8316 SvPOK_on(sv); /* validate pointer */
8318 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
8319 PTR2UV(sv),SvPVX_const(sv)));
8322 return SvPVX_mutable(sv);
8326 =for apidoc sv_pvbyten_force
8328 The backend for the C<SvPVbytex_force> macro. Always use the macro instead.
8334 Perl_sv_pvbyten_force(pTHX_ SV *const sv, STRLEN *const lp)
8336 PERL_ARGS_ASSERT_SV_PVBYTEN_FORCE;
8338 sv_pvn_force(sv,lp);
8339 sv_utf8_downgrade(sv,0);
8345 =for apidoc sv_pvutf8n_force
8347 The backend for the C<SvPVutf8x_force> macro. Always use the macro instead.
8353 Perl_sv_pvutf8n_force(pTHX_ SV *const sv, STRLEN *const lp)
8355 PERL_ARGS_ASSERT_SV_PVUTF8N_FORCE;
8357 sv_pvn_force(sv,lp);
8358 sv_utf8_upgrade(sv);
8364 =for apidoc sv_reftype
8366 Returns a string describing what the SV is a reference to.
8372 Perl_sv_reftype(pTHX_ const SV *const sv, const int ob)
8374 PERL_ARGS_ASSERT_SV_REFTYPE;
8376 /* The fact that I don't need to downcast to char * everywhere, only in ?:
8377 inside return suggests a const propagation bug in g++. */
8378 if (ob && SvOBJECT(sv)) {
8379 char * const name = HvNAME_get(SvSTASH(sv));
8380 return name ? name : (char *) "__ANON__";
8383 switch (SvTYPE(sv)) {
8398 case SVt_PVLV: return (char *) (SvROK(sv) ? "REF"
8399 /* tied lvalues should appear to be
8400 * scalars for backwards compatitbility */
8401 : (LvTYPE(sv) == 't' || LvTYPE(sv) == 'T')
8402 ? "SCALAR" : "LVALUE");
8403 case SVt_PVAV: return "ARRAY";
8404 case SVt_PVHV: return "HASH";
8405 case SVt_PVCV: return "CODE";
8406 case SVt_PVGV: return (char *) (isGV_with_GP(sv)
8407 ? "GLOB" : "SCALAR");
8408 case SVt_PVFM: return "FORMAT";
8409 case SVt_PVIO: return "IO";
8410 case SVt_BIND: return "BIND";
8411 case SVt_REGEXP: return "REGEXP";
8412 default: return "UNKNOWN";
8418 =for apidoc sv_isobject
8420 Returns a boolean indicating whether the SV is an RV pointing to a blessed
8421 object. If the SV is not an RV, or if the object is not blessed, then this
8428 Perl_sv_isobject(pTHX_ SV *sv)
8444 Returns a boolean indicating whether the SV is blessed into the specified
8445 class. This does not check for subtypes; use C<sv_derived_from> to verify
8446 an inheritance relationship.
8452 Perl_sv_isa(pTHX_ SV *sv, const char *const name)
8456 PERL_ARGS_ASSERT_SV_ISA;
8466 hvname = HvNAME_get(SvSTASH(sv));
8470 return strEQ(hvname, name);
8476 Creates a new SV for the RV, C<rv>, to point to. If C<rv> is not an RV then
8477 it will be upgraded to one. If C<classname> is non-null then the new SV will
8478 be blessed in the specified package. The new SV is returned and its
8479 reference count is 1.
8485 Perl_newSVrv(pTHX_ SV *const rv, const char *const classname)
8490 PERL_ARGS_ASSERT_NEWSVRV;
8494 SV_CHECK_THINKFIRST_COW_DROP(rv);
8495 (void)SvAMAGIC_off(rv);
8497 if (SvTYPE(rv) >= SVt_PVMG) {
8498 const U32 refcnt = SvREFCNT(rv);
8502 SvREFCNT(rv) = refcnt;
8504 sv_upgrade(rv, SVt_IV);
8505 } else if (SvROK(rv)) {
8506 SvREFCNT_dec(SvRV(rv));
8508 prepare_SV_for_RV(rv);
8516 HV* const stash = gv_stashpv(classname, GV_ADD);
8517 (void)sv_bless(rv, stash);
8523 =for apidoc sv_setref_pv
8525 Copies a pointer into a new SV, optionally blessing the SV. The C<rv>
8526 argument will be upgraded to an RV. That RV will be modified to point to
8527 the new SV. If the C<pv> argument is NULL then C<PL_sv_undef> will be placed
8528 into the SV. The C<classname> argument indicates the package for the
8529 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
8530 will have a reference count of 1, and the RV will be returned.
8532 Do not use with other Perl types such as HV, AV, SV, CV, because those
8533 objects will become corrupted by the pointer copy process.
8535 Note that C<sv_setref_pvn> copies the string while this copies the pointer.
8541 Perl_sv_setref_pv(pTHX_ SV *const rv, const char *const classname, void *const pv)
8545 PERL_ARGS_ASSERT_SV_SETREF_PV;
8548 sv_setsv(rv, &PL_sv_undef);
8552 sv_setiv(newSVrv(rv,classname), PTR2IV(pv));
8557 =for apidoc sv_setref_iv
8559 Copies an integer into a new SV, optionally blessing the SV. The C<rv>
8560 argument will be upgraded to an RV. That RV will be modified to point to
8561 the new SV. The C<classname> argument indicates the package for the
8562 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
8563 will have a reference count of 1, and the RV will be returned.
8569 Perl_sv_setref_iv(pTHX_ SV *const rv, const char *const classname, const IV iv)
8571 PERL_ARGS_ASSERT_SV_SETREF_IV;
8573 sv_setiv(newSVrv(rv,classname), iv);
8578 =for apidoc sv_setref_uv
8580 Copies an unsigned integer into a new SV, optionally blessing the SV. The C<rv>
8581 argument will be upgraded to an RV. That RV will be modified to point to
8582 the new SV. The C<classname> argument indicates the package for the
8583 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
8584 will have a reference count of 1, and the RV will be returned.
8590 Perl_sv_setref_uv(pTHX_ SV *const rv, const char *const classname, const UV uv)
8592 PERL_ARGS_ASSERT_SV_SETREF_UV;
8594 sv_setuv(newSVrv(rv,classname), uv);
8599 =for apidoc sv_setref_nv
8601 Copies a double into a new SV, optionally blessing the SV. The C<rv>
8602 argument will be upgraded to an RV. That RV will be modified to point to
8603 the new SV. The C<classname> argument indicates the package for the
8604 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
8605 will have a reference count of 1, and the RV will be returned.
8611 Perl_sv_setref_nv(pTHX_ SV *const rv, const char *const classname, const NV nv)
8613 PERL_ARGS_ASSERT_SV_SETREF_NV;
8615 sv_setnv(newSVrv(rv,classname), nv);
8620 =for apidoc sv_setref_pvn
8622 Copies a string into a new SV, optionally blessing the SV. The length of the
8623 string must be specified with C<n>. The C<rv> argument will be upgraded to
8624 an RV. That RV will be modified to point to the new SV. The C<classname>
8625 argument indicates the package for the blessing. Set C<classname> to
8626 C<NULL> to avoid the blessing. The new SV will have a reference count
8627 of 1, and the RV will be returned.
8629 Note that C<sv_setref_pv> copies the pointer while this copies the string.
8635 Perl_sv_setref_pvn(pTHX_ SV *const rv, const char *const classname,
8636 const char *const pv, const STRLEN n)
8638 PERL_ARGS_ASSERT_SV_SETREF_PVN;
8640 sv_setpvn(newSVrv(rv,classname), pv, n);
8645 =for apidoc sv_bless
8647 Blesses an SV into a specified package. The SV must be an RV. The package
8648 must be designated by its stash (see C<gv_stashpv()>). The reference count
8649 of the SV is unaffected.
8655 Perl_sv_bless(pTHX_ SV *const sv, HV *const stash)
8660 PERL_ARGS_ASSERT_SV_BLESS;
8663 Perl_croak(aTHX_ "Can't bless non-reference value");
8665 if (SvFLAGS(tmpRef) & (SVs_OBJECT|SVf_READONLY)) {
8666 if (SvIsCOW(tmpRef))
8667 sv_force_normal_flags(tmpRef, 0);
8668 if (SvREADONLY(tmpRef))
8669 Perl_croak(aTHX_ "%s", PL_no_modify);
8670 if (SvOBJECT(tmpRef)) {
8671 if (SvTYPE(tmpRef) != SVt_PVIO)
8673 SvREFCNT_dec(SvSTASH(tmpRef));
8676 SvOBJECT_on(tmpRef);
8677 if (SvTYPE(tmpRef) != SVt_PVIO)
8679 SvUPGRADE(tmpRef, SVt_PVMG);
8680 SvSTASH_set(tmpRef, MUTABLE_HV(SvREFCNT_inc_simple(stash)));
8685 (void)SvAMAGIC_off(sv);
8687 if(SvSMAGICAL(tmpRef))
8688 if(mg_find(tmpRef, PERL_MAGIC_ext) || mg_find(tmpRef, PERL_MAGIC_uvar))
8696 /* Downgrades a PVGV to a PVMG.
8700 S_sv_unglob(pTHX_ SV *const sv)
8705 SV * const temp = sv_newmortal();
8707 PERL_ARGS_ASSERT_SV_UNGLOB;
8709 assert(SvTYPE(sv) == SVt_PVGV);
8711 gv_efullname3(temp, MUTABLE_GV(sv), "*");
8714 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
8715 && HvNAME_get(stash))
8716 mro_method_changed_in(stash);
8717 gp_free(MUTABLE_GV(sv));
8720 sv_del_backref(MUTABLE_SV(GvSTASH(sv)), sv);
8724 if (GvNAME_HEK(sv)) {
8725 unshare_hek(GvNAME_HEK(sv));
8727 isGV_with_GP_off(sv);
8729 /* need to keep SvANY(sv) in the right arena */
8730 xpvmg = new_XPVMG();
8731 StructCopy(SvANY(sv), xpvmg, XPVMG);
8732 del_XPVGV(SvANY(sv));
8735 SvFLAGS(sv) &= ~SVTYPEMASK;
8736 SvFLAGS(sv) |= SVt_PVMG;
8738 /* Intentionally not calling any local SET magic, as this isn't so much a
8739 set operation as merely an internal storage change. */
8740 sv_setsv_flags(sv, temp, 0);
8744 =for apidoc sv_unref_flags
8746 Unsets the RV status of the SV, and decrements the reference count of
8747 whatever was being referenced by the RV. This can almost be thought of
8748 as a reversal of C<newSVrv>. The C<cflags> argument can contain
8749 C<SV_IMMEDIATE_UNREF> to force the reference count to be decremented
8750 (otherwise the decrementing is conditional on the reference count being
8751 different from one or the reference being a readonly SV).
8758 Perl_sv_unref_flags(pTHX_ SV *const ref, const U32 flags)
8760 SV* const target = SvRV(ref);
8762 PERL_ARGS_ASSERT_SV_UNREF_FLAGS;
8764 if (SvWEAKREF(ref)) {
8765 sv_del_backref(target, ref);
8767 SvRV_set(ref, NULL);
8770 SvRV_set(ref, NULL);
8772 /* You can't have a || SvREADONLY(target) here, as $a = $$a, where $a was
8773 assigned to as BEGIN {$a = \"Foo"} will fail. */
8774 if (SvREFCNT(target) != 1 || (flags & SV_IMMEDIATE_UNREF))
8775 SvREFCNT_dec(target);
8776 else /* XXX Hack, but hard to make $a=$a->[1] work otherwise */
8777 sv_2mortal(target); /* Schedule for freeing later */
8781 =for apidoc sv_untaint
8783 Untaint an SV. Use C<SvTAINTED_off> instead.
8788 Perl_sv_untaint(pTHX_ SV *const sv)
8790 PERL_ARGS_ASSERT_SV_UNTAINT;
8792 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
8793 MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
8800 =for apidoc sv_tainted
8802 Test an SV for taintedness. Use C<SvTAINTED> instead.
8807 Perl_sv_tainted(pTHX_ SV *const sv)
8809 PERL_ARGS_ASSERT_SV_TAINTED;
8811 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
8812 const MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
8813 if (mg && (mg->mg_len & 1) )
8820 =for apidoc sv_setpviv
8822 Copies an integer into the given SV, also updating its string value.
8823 Does not handle 'set' magic. See C<sv_setpviv_mg>.
8829 Perl_sv_setpviv(pTHX_ SV *const sv, const IV iv)
8831 char buf[TYPE_CHARS(UV)];
8833 char * const ptr = uiv_2buf(buf, iv, 0, 0, &ebuf);
8835 PERL_ARGS_ASSERT_SV_SETPVIV;
8837 sv_setpvn(sv, ptr, ebuf - ptr);
8841 =for apidoc sv_setpviv_mg
8843 Like C<sv_setpviv>, but also handles 'set' magic.
8849 Perl_sv_setpviv_mg(pTHX_ SV *const sv, const IV iv)
8851 PERL_ARGS_ASSERT_SV_SETPVIV_MG;
8857 #if defined(PERL_IMPLICIT_CONTEXT)
8859 /* pTHX_ magic can't cope with varargs, so this is a no-context
8860 * version of the main function, (which may itself be aliased to us).
8861 * Don't access this version directly.
8865 Perl_sv_setpvf_nocontext(SV *const sv, const char *const pat, ...)
8870 PERL_ARGS_ASSERT_SV_SETPVF_NOCONTEXT;
8872 va_start(args, pat);
8873 sv_vsetpvf(sv, pat, &args);
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_mg_nocontext(SV *const sv, const char *const pat, ...)
8888 PERL_ARGS_ASSERT_SV_SETPVF_MG_NOCONTEXT;
8890 va_start(args, pat);
8891 sv_vsetpvf_mg(sv, pat, &args);
8897 =for apidoc sv_setpvf
8899 Works like C<sv_catpvf> but copies the text into the SV instead of
8900 appending it. Does not handle 'set' magic. See C<sv_setpvf_mg>.
8906 Perl_sv_setpvf(pTHX_ SV *const sv, const char *const pat, ...)
8910 PERL_ARGS_ASSERT_SV_SETPVF;
8912 va_start(args, pat);
8913 sv_vsetpvf(sv, pat, &args);
8918 =for apidoc sv_vsetpvf
8920 Works like C<sv_vcatpvf> but copies the text into the SV instead of
8921 appending it. Does not handle 'set' magic. See C<sv_vsetpvf_mg>.
8923 Usually used via its frontend C<sv_setpvf>.
8929 Perl_sv_vsetpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
8931 PERL_ARGS_ASSERT_SV_VSETPVF;
8933 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
8937 =for apidoc sv_setpvf_mg
8939 Like C<sv_setpvf>, but also handles 'set' magic.
8945 Perl_sv_setpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
8949 PERL_ARGS_ASSERT_SV_SETPVF_MG;
8951 va_start(args, pat);
8952 sv_vsetpvf_mg(sv, pat, &args);
8957 =for apidoc sv_vsetpvf_mg
8959 Like C<sv_vsetpvf>, but also handles 'set' magic.
8961 Usually used via its frontend C<sv_setpvf_mg>.
8967 Perl_sv_vsetpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
8969 PERL_ARGS_ASSERT_SV_VSETPVF_MG;
8971 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
8975 #if defined(PERL_IMPLICIT_CONTEXT)
8977 /* pTHX_ magic can't cope with varargs, so this is a no-context
8978 * version of the main function, (which may itself be aliased to us).
8979 * Don't access this version directly.
8983 Perl_sv_catpvf_nocontext(SV *const sv, const char *const pat, ...)
8988 PERL_ARGS_ASSERT_SV_CATPVF_NOCONTEXT;
8990 va_start(args, pat);
8991 sv_vcatpvf(sv, pat, &args);
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_mg_nocontext(SV *const sv, const char *const pat, ...)
9006 PERL_ARGS_ASSERT_SV_CATPVF_MG_NOCONTEXT;
9008 va_start(args, pat);
9009 sv_vcatpvf_mg(sv, pat, &args);
9015 =for apidoc sv_catpvf
9017 Processes its arguments like C<sprintf> and appends the formatted
9018 output to an SV. If the appended data contains "wide" characters
9019 (including, but not limited to, SVs with a UTF-8 PV formatted with %s,
9020 and characters >255 formatted with %c), the original SV might get
9021 upgraded to UTF-8. Handles 'get' magic, but not 'set' magic. See
9022 C<sv_catpvf_mg>. If the original SV was UTF-8, the pattern should be
9023 valid UTF-8; if the original SV was bytes, the pattern should be too.
9028 Perl_sv_catpvf(pTHX_ SV *const sv, const char *const pat, ...)
9032 PERL_ARGS_ASSERT_SV_CATPVF;
9034 va_start(args, pat);
9035 sv_vcatpvf(sv, pat, &args);
9040 =for apidoc sv_vcatpvf
9042 Processes its arguments like C<vsprintf> and appends the formatted output
9043 to an SV. Does not handle 'set' magic. See C<sv_vcatpvf_mg>.
9045 Usually used via its frontend C<sv_catpvf>.
9051 Perl_sv_vcatpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9053 PERL_ARGS_ASSERT_SV_VCATPVF;
9055 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9059 =for apidoc sv_catpvf_mg
9061 Like C<sv_catpvf>, but also handles 'set' magic.
9067 Perl_sv_catpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
9071 PERL_ARGS_ASSERT_SV_CATPVF_MG;
9073 va_start(args, pat);
9074 sv_vcatpvf_mg(sv, pat, &args);
9079 =for apidoc sv_vcatpvf_mg
9081 Like C<sv_vcatpvf>, but also handles 'set' magic.
9083 Usually used via its frontend C<sv_catpvf_mg>.
9089 Perl_sv_vcatpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9091 PERL_ARGS_ASSERT_SV_VCATPVF_MG;
9093 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9098 =for apidoc sv_vsetpvfn
9100 Works like C<sv_vcatpvfn> but copies the text into the SV instead of
9103 Usually used via one of its frontends C<sv_vsetpvf> and C<sv_vsetpvf_mg>.
9109 Perl_sv_vsetpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
9110 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
9112 PERL_ARGS_ASSERT_SV_VSETPVFN;
9115 sv_vcatpvfn(sv, pat, patlen, args, svargs, svmax, maybe_tainted);
9119 S_expect_number(pTHX_ char **const pattern)
9124 PERL_ARGS_ASSERT_EXPECT_NUMBER;
9126 switch (**pattern) {
9127 case '1': case '2': case '3':
9128 case '4': case '5': case '6':
9129 case '7': case '8': case '9':
9130 var = *(*pattern)++ - '0';
9131 while (isDIGIT(**pattern)) {
9132 const I32 tmp = var * 10 + (*(*pattern)++ - '0');
9134 Perl_croak(aTHX_ "Integer overflow in format string for %s", (PL_op ? OP_NAME(PL_op) : "sv_vcatpvfn"));
9142 S_F0convert(NV nv, char *const endbuf, STRLEN *const len)
9144 const int neg = nv < 0;
9147 PERL_ARGS_ASSERT_F0CONVERT;
9155 if (uv & 1 && uv == nv)
9156 uv--; /* Round to even */
9158 const unsigned dig = uv % 10;
9171 =for apidoc sv_vcatpvfn
9173 Processes its arguments like C<vsprintf> and appends the formatted output
9174 to an SV. Uses an array of SVs if the C style variable argument list is
9175 missing (NULL). When running with taint checks enabled, indicates via
9176 C<maybe_tainted> if results are untrustworthy (often due to the use of
9179 Usually used via one of its frontends C<sv_vcatpvf> and C<sv_vcatpvf_mg>.
9185 #define VECTORIZE_ARGS vecsv = va_arg(*args, SV*);\
9186 vecstr = (U8*)SvPV_const(vecsv,veclen);\
9187 vec_utf8 = DO_UTF8(vecsv);
9189 /* XXX maybe_tainted is never assigned to, so the doc above is lying. */
9192 Perl_sv_vcatpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
9193 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
9201 static const char nullstr[] = "(null)";
9203 bool has_utf8 = DO_UTF8(sv); /* has the result utf8? */
9204 const bool pat_utf8 = has_utf8; /* the pattern is in utf8? */
9206 /* Times 4: a decimal digit takes more than 3 binary digits.
9207 * NV_DIG: mantissa takes than many decimal digits.
9208 * Plus 32: Playing safe. */
9209 char ebuf[IV_DIG * 4 + NV_DIG + 32];
9210 /* large enough for "%#.#f" --chip */
9211 /* what about long double NVs? --jhi */
9213 PERL_ARGS_ASSERT_SV_VCATPVFN;
9214 PERL_UNUSED_ARG(maybe_tainted);
9216 /* no matter what, this is a string now */
9217 (void)SvPV_force(sv, origlen);
9219 /* special-case "", "%s", and "%-p" (SVf - see below) */
9222 if (patlen == 2 && pat[0] == '%' && pat[1] == 's') {
9224 const char * const s = va_arg(*args, char*);
9225 sv_catpv(sv, s ? s : nullstr);
9227 else if (svix < svmax) {
9228 sv_catsv(sv, *svargs);
9232 if (args && patlen == 3 && pat[0] == '%' &&
9233 pat[1] == '-' && pat[2] == 'p') {
9234 argsv = MUTABLE_SV(va_arg(*args, void*));
9235 sv_catsv(sv, argsv);
9239 #ifndef USE_LONG_DOUBLE
9240 /* special-case "%.<number>[gf]" */
9241 if ( !args && patlen <= 5 && pat[0] == '%' && pat[1] == '.'
9242 && (pat[patlen-1] == 'g' || pat[patlen-1] == 'f') ) {
9243 unsigned digits = 0;
9247 while (*pp >= '0' && *pp <= '9')
9248 digits = 10 * digits + (*pp++ - '0');
9249 if (pp - pat == (int)patlen - 1) {
9257 /* Add check for digits != 0 because it seems that some
9258 gconverts are buggy in this case, and we don't yet have
9259 a Configure test for this. */
9260 if (digits && digits < sizeof(ebuf) - NV_DIG - 10) {
9261 /* 0, point, slack */
9262 Gconvert(nv, (int)digits, 0, ebuf);
9264 if (*ebuf) /* May return an empty string for digits==0 */
9267 } else if (!digits) {
9270 if ((p = F0convert(nv, ebuf + sizeof ebuf, &l))) {
9271 sv_catpvn(sv, p, l);
9277 #endif /* !USE_LONG_DOUBLE */
9279 if (!args && svix < svmax && DO_UTF8(*svargs))
9282 patend = (char*)pat + patlen;
9283 for (p = (char*)pat; p < patend; p = q) {
9286 bool vectorize = FALSE;
9287 bool vectorarg = FALSE;
9288 bool vec_utf8 = FALSE;
9294 bool has_precis = FALSE;
9296 const I32 osvix = svix;
9297 bool is_utf8 = FALSE; /* is this item utf8? */
9298 #ifdef HAS_LDBL_SPRINTF_BUG
9299 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
9300 with sfio - Allen <allens@cpan.org> */
9301 bool fix_ldbl_sprintf_bug = FALSE;
9305 U8 utf8buf[UTF8_MAXBYTES+1];
9306 STRLEN esignlen = 0;
9308 const char *eptr = NULL;
9309 const char *fmtstart;
9312 const U8 *vecstr = NULL;
9319 /* we need a long double target in case HAS_LONG_DOUBLE but
9322 #if defined(HAS_LONG_DOUBLE) && LONG_DOUBLESIZE > DOUBLESIZE
9330 const char *dotstr = ".";
9331 STRLEN dotstrlen = 1;
9332 I32 efix = 0; /* explicit format parameter index */
9333 I32 ewix = 0; /* explicit width index */
9334 I32 epix = 0; /* explicit precision index */
9335 I32 evix = 0; /* explicit vector index */
9336 bool asterisk = FALSE;
9338 /* echo everything up to the next format specification */
9339 for (q = p; q < patend && *q != '%'; ++q) ;
9341 if (has_utf8 && !pat_utf8)
9342 sv_catpvn_utf8_upgrade(sv, p, q - p, nsv);
9344 sv_catpvn(sv, p, q - p);
9353 We allow format specification elements in this order:
9354 \d+\$ explicit format parameter index
9356 v|\*(\d+\$)?v vector with optional (optionally specified) arg
9357 0 flag (as above): repeated to allow "v02"
9358 \d+|\*(\d+\$)? width using optional (optionally specified) arg
9359 \.(\d*|\*(\d+\$)?) precision using optional (optionally specified) arg
9361 [%bcdefginopsuxDFOUX] format (mandatory)
9366 As of perl5.9.3, printf format checking is on by default.
9367 Internally, perl uses %p formats to provide an escape to
9368 some extended formatting. This block deals with those
9369 extensions: if it does not match, (char*)q is reset and
9370 the normal format processing code is used.
9372 Currently defined extensions are:
9373 %p include pointer address (standard)
9374 %-p (SVf) include an SV (previously %_)
9375 %-<num>p include an SV with precision <num>
9376 %<num>p reserved for future extensions
9378 Robin Barker 2005-07-14
9380 %1p (VDf) removed. RMB 2007-10-19
9387 n = expect_number(&q);
9394 argsv = MUTABLE_SV(va_arg(*args, void*));
9395 eptr = SvPV_const(argsv, elen);
9401 if (ckWARN_d(WARN_INTERNAL))
9402 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
9403 "internal %%<num>p might conflict with future printf extensions");
9409 if ( (width = expect_number(&q)) ) {
9424 if (plus == '+' && *q == ' ') /* '+' over ' ' */
9453 if ( (ewix = expect_number(&q)) )
9462 if ((vectorarg = asterisk)) {
9475 width = expect_number(&q);
9481 vecsv = va_arg(*args, SV*);
9483 vecsv = (evix > 0 && evix <= svmax)
9484 ? svargs[evix-1] : &PL_sv_undef;
9486 vecsv = svix < svmax ? svargs[svix++] : &PL_sv_undef;
9488 dotstr = SvPV_const(vecsv, dotstrlen);
9489 /* Keep the DO_UTF8 test *after* the SvPV call, else things go
9490 bad with tied or overloaded values that return UTF8. */
9493 else if (has_utf8) {
9494 vecsv = sv_mortalcopy(vecsv);
9495 sv_utf8_upgrade(vecsv);
9496 dotstr = SvPV_const(vecsv, dotstrlen);
9503 else if (efix ? (efix > 0 && efix <= svmax) : svix < svmax) {
9504 vecsv = svargs[efix ? efix-1 : svix++];
9505 vecstr = (U8*)SvPV_const(vecsv,veclen);
9506 vec_utf8 = DO_UTF8(vecsv);
9508 /* if this is a version object, we need to convert
9509 * back into v-string notation and then let the
9510 * vectorize happen normally
9512 if (sv_derived_from(vecsv, "version")) {
9513 char *version = savesvpv(vecsv);
9514 if ( hv_exists(MUTABLE_HV(SvRV(vecsv)), "alpha", 5 ) ) {
9515 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
9516 "vector argument not supported with alpha versions");
9519 vecsv = sv_newmortal();
9520 scan_vstring(version, version + veclen, vecsv);
9521 vecstr = (U8*)SvPV_const(vecsv, veclen);
9522 vec_utf8 = DO_UTF8(vecsv);
9534 i = va_arg(*args, int);
9536 i = (ewix ? ewix <= svmax : svix < svmax) ?
9537 SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
9539 width = (i < 0) ? -i : i;
9549 if ( ((epix = expect_number(&q))) && (*q++ != '$') )
9551 /* XXX: todo, support specified precision parameter */
9555 i = va_arg(*args, int);
9557 i = (ewix ? ewix <= svmax : svix < svmax)
9558 ? SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
9560 has_precis = !(i < 0);
9565 precis = precis * 10 + (*q++ - '0');
9574 case 'I': /* Ix, I32x, and I64x */
9576 if (q[1] == '6' && q[2] == '4') {
9582 if (q[1] == '3' && q[2] == '2') {
9592 #if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
9603 #if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
9604 if (*(q + 1) == 'l') { /* lld, llf */
9630 if (!vectorize && !args) {
9632 const I32 i = efix-1;
9633 argsv = (i >= 0 && i < svmax) ? svargs[i] : &PL_sv_undef;
9635 argsv = (svix >= 0 && svix < svmax)
9636 ? svargs[svix++] : &PL_sv_undef;
9647 uv = (args) ? va_arg(*args, int) : SvIV(argsv);
9649 (!UNI_IS_INVARIANT(uv) && SvUTF8(sv)))
9651 eptr = (char*)utf8buf;
9652 elen = uvchr_to_utf8((U8*)eptr, uv) - utf8buf;
9666 eptr = va_arg(*args, char*);
9668 elen = strlen(eptr);
9670 eptr = (char *)nullstr;
9671 elen = sizeof nullstr - 1;
9675 eptr = SvPV_const(argsv, elen);
9676 if (DO_UTF8(argsv)) {
9677 I32 old_precis = precis;
9678 if (has_precis && precis < elen) {
9680 sv_pos_u2b(argsv, &p, 0); /* sticks at end */
9683 if (width) { /* fudge width (can't fudge elen) */
9684 if (has_precis && precis < elen)
9685 width += precis - old_precis;
9687 width += elen - sv_len_utf8(argsv);
9694 if (has_precis && elen > precis)
9701 if (alt || vectorize)
9703 uv = PTR2UV(args ? va_arg(*args, void*) : argsv);
9724 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
9733 esignbuf[esignlen++] = plus;
9737 case 'h': iv = (short)va_arg(*args, int); break;
9738 case 'l': iv = va_arg(*args, long); break;
9739 case 'V': iv = va_arg(*args, IV); break;
9740 default: iv = va_arg(*args, int); break;
9743 iv = va_arg(*args, Quad_t); break;
9750 IV tiv = SvIV(argsv); /* work around GCC bug #13488 */
9752 case 'h': iv = (short)tiv; break;
9753 case 'l': iv = (long)tiv; break;
9755 default: iv = tiv; break;
9758 iv = (Quad_t)tiv; break;
9764 if ( !vectorize ) /* we already set uv above */
9769 esignbuf[esignlen++] = plus;
9773 esignbuf[esignlen++] = '-';
9817 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
9828 case 'h': uv = (unsigned short)va_arg(*args, unsigned); break;
9829 case 'l': uv = va_arg(*args, unsigned long); break;
9830 case 'V': uv = va_arg(*args, UV); break;
9831 default: uv = va_arg(*args, unsigned); break;
9834 uv = va_arg(*args, Uquad_t); break;
9841 UV tuv = SvUV(argsv); /* work around GCC bug #13488 */
9843 case 'h': uv = (unsigned short)tuv; break;
9844 case 'l': uv = (unsigned long)tuv; break;
9846 default: uv = tuv; break;
9849 uv = (Uquad_t)tuv; break;
9858 char *ptr = ebuf + sizeof ebuf;
9859 bool tempalt = uv ? alt : FALSE; /* Vectors can't change alt */
9865 p = (char *)((c == 'X') ? PL_hexdigit + 16 : PL_hexdigit);
9871 esignbuf[esignlen++] = '0';
9872 esignbuf[esignlen++] = c; /* 'x' or 'X' */
9880 if (alt && *ptr != '0')
9889 esignbuf[esignlen++] = '0';
9890 esignbuf[esignlen++] = c;
9893 default: /* it had better be ten or less */
9897 } while (uv /= base);
9900 elen = (ebuf + sizeof ebuf) - ptr;
9904 zeros = precis - elen;
9905 else if (precis == 0 && elen == 1 && *eptr == '0'
9906 && !(base == 8 && alt)) /* "%#.0o" prints "0" */
9909 /* a precision nullifies the 0 flag. */
9916 /* FLOATING POINT */
9919 c = 'f'; /* maybe %F isn't supported here */
9927 /* This is evil, but floating point is even more evil */
9929 /* for SV-style calling, we can only get NV
9930 for C-style calling, we assume %f is double;
9931 for simplicity we allow any of %Lf, %llf, %qf for long double
9935 #if defined(USE_LONG_DOUBLE)
9939 /* [perl #20339] - we should accept and ignore %lf rather than die */
9943 #if defined(USE_LONG_DOUBLE)
9944 intsize = args ? 0 : 'q';
9948 #if defined(HAS_LONG_DOUBLE)
9957 /* now we need (long double) if intsize == 'q', else (double) */
9959 #if LONG_DOUBLESIZE > DOUBLESIZE
9961 va_arg(*args, long double) :
9962 va_arg(*args, double)
9964 va_arg(*args, double)
9969 /* nv * 0 will be NaN for NaN, +Inf and -Inf, and 0 for anything
9970 else. frexp() has some unspecified behaviour for those three */
9971 if (c != 'e' && c != 'E' && (nv * 0) == 0) {
9973 /* FIXME: if HAS_LONG_DOUBLE but not USE_LONG_DOUBLE this
9974 will cast our (long double) to (double) */
9975 (void)Perl_frexp(nv, &i);
9976 if (i == PERL_INT_MIN)
9977 Perl_die(aTHX_ "panic: frexp");
9979 need = BIT_DIGITS(i);
9981 need += has_precis ? precis : 6; /* known default */
9986 #ifdef HAS_LDBL_SPRINTF_BUG
9987 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
9988 with sfio - Allen <allens@cpan.org> */
9991 # define MY_DBL_MAX DBL_MAX
9992 # else /* XXX guessing! HUGE_VAL may be defined as infinity, so not using */
9993 # if DOUBLESIZE >= 8
9994 # define MY_DBL_MAX 1.7976931348623157E+308L
9996 # define MY_DBL_MAX 3.40282347E+38L
10000 # ifdef HAS_LDBL_SPRINTF_BUG_LESS1 /* only between -1L & 1L - Allen */
10001 # define MY_DBL_MAX_BUG 1L
10003 # define MY_DBL_MAX_BUG MY_DBL_MAX
10007 # define MY_DBL_MIN DBL_MIN
10008 # else /* XXX guessing! -Allen */
10009 # if DOUBLESIZE >= 8
10010 # define MY_DBL_MIN 2.2250738585072014E-308L
10012 # define MY_DBL_MIN 1.17549435E-38L
10016 if ((intsize == 'q') && (c == 'f') &&
10017 ((nv < MY_DBL_MAX_BUG) && (nv > -MY_DBL_MAX_BUG)) &&
10018 (need < DBL_DIG)) {
10019 /* it's going to be short enough that
10020 * long double precision is not needed */
10022 if ((nv <= 0L) && (nv >= -0L))
10023 fix_ldbl_sprintf_bug = TRUE; /* 0 is 0 - easiest */
10025 /* would use Perl_fp_class as a double-check but not
10026 * functional on IRIX - see perl.h comments */
10028 if ((nv >= MY_DBL_MIN) || (nv <= -MY_DBL_MIN)) {
10029 /* It's within the range that a double can represent */
10030 #if defined(DBL_MAX) && !defined(DBL_MIN)
10031 if ((nv >= ((long double)1/DBL_MAX)) ||
10032 (nv <= (-(long double)1/DBL_MAX)))
10034 fix_ldbl_sprintf_bug = TRUE;
10037 if (fix_ldbl_sprintf_bug == TRUE) {
10047 # undef MY_DBL_MAX_BUG
10050 #endif /* HAS_LDBL_SPRINTF_BUG */
10052 need += 20; /* fudge factor */
10053 if (PL_efloatsize < need) {
10054 Safefree(PL_efloatbuf);
10055 PL_efloatsize = need + 20; /* more fudge */
10056 Newx(PL_efloatbuf, PL_efloatsize, char);
10057 PL_efloatbuf[0] = '\0';
10060 if ( !(width || left || plus || alt) && fill != '0'
10061 && has_precis && intsize != 'q' ) { /* Shortcuts */
10062 /* See earlier comment about buggy Gconvert when digits,
10064 if ( c == 'g' && precis) {
10065 Gconvert((NV)nv, (int)precis, 0, PL_efloatbuf);
10066 /* May return an empty string for digits==0 */
10067 if (*PL_efloatbuf) {
10068 elen = strlen(PL_efloatbuf);
10069 goto float_converted;
10071 } else if ( c == 'f' && !precis) {
10072 if ((eptr = F0convert(nv, ebuf + sizeof ebuf, &elen)))
10077 char *ptr = ebuf + sizeof ebuf;
10080 /* FIXME: what to do if HAS_LONG_DOUBLE but not PERL_PRIfldbl? */
10081 #if defined(HAS_LONG_DOUBLE) && defined(PERL_PRIfldbl)
10082 if (intsize == 'q') {
10083 /* Copy the one or more characters in a long double
10084 * format before the 'base' ([efgEFG]) character to
10085 * the format string. */
10086 static char const prifldbl[] = PERL_PRIfldbl;
10087 char const *p = prifldbl + sizeof(prifldbl) - 3;
10088 while (p >= prifldbl) { *--ptr = *p--; }
10093 do { *--ptr = '0' + (base % 10); } while (base /= 10);
10098 do { *--ptr = '0' + (base % 10); } while (base /= 10);
10110 /* No taint. Otherwise we are in the strange situation
10111 * where printf() taints but print($float) doesn't.
10113 #if defined(HAS_LONG_DOUBLE)
10114 elen = ((intsize == 'q')
10115 ? my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, nv)
10116 : my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, (double)nv));
10118 elen = my_sprintf(PL_efloatbuf, ptr, nv);
10122 eptr = PL_efloatbuf;
10130 i = SvCUR(sv) - origlen;
10133 case 'h': *(va_arg(*args, short*)) = i; break;
10134 default: *(va_arg(*args, int*)) = i; break;
10135 case 'l': *(va_arg(*args, long*)) = i; break;
10136 case 'V': *(va_arg(*args, IV*)) = i; break;
10139 *(va_arg(*args, Quad_t*)) = i; break;
10146 sv_setuv_mg(argsv, (UV)i);
10147 continue; /* not "break" */
10154 && (PL_op->op_type == OP_PRTF || PL_op->op_type == OP_SPRINTF)
10155 && ckWARN(WARN_PRINTF))
10157 SV * const msg = sv_newmortal();
10158 Perl_sv_setpvf(aTHX_ msg, "Invalid conversion in %sprintf: ",
10159 (PL_op->op_type == OP_PRTF) ? "" : "s");
10160 if (fmtstart < patend) {
10161 const char * const fmtend = q < patend ? q : patend;
10163 sv_catpvs(msg, "\"%");
10164 for (f = fmtstart; f < fmtend; f++) {
10166 sv_catpvn(msg, f, 1);
10168 Perl_sv_catpvf(aTHX_ msg,
10169 "\\%03"UVof, (UV)*f & 0xFF);
10172 sv_catpvs(msg, "\"");
10174 sv_catpvs(msg, "end of string");
10176 Perl_warner(aTHX_ packWARN(WARN_PRINTF), "%"SVf, SVfARG(msg)); /* yes, this is reentrant */
10179 /* output mangled stuff ... */
10185 /* ... right here, because formatting flags should not apply */
10186 SvGROW(sv, SvCUR(sv) + elen + 1);
10188 Copy(eptr, p, elen, char);
10191 SvCUR_set(sv, p - SvPVX_const(sv));
10193 continue; /* not "break" */
10196 if (is_utf8 != has_utf8) {
10199 sv_utf8_upgrade(sv);
10202 const STRLEN old_elen = elen;
10203 SV * const nsv = newSVpvn_flags(eptr, elen, SVs_TEMP);
10204 sv_utf8_upgrade(nsv);
10205 eptr = SvPVX_const(nsv);
10208 if (width) { /* fudge width (can't fudge elen) */
10209 width += elen - old_elen;
10215 have = esignlen + zeros + elen;
10217 Perl_croak_nocontext("%s", PL_memory_wrap);
10219 need = (have > width ? have : width);
10222 if (need >= (((STRLEN)~0) - SvCUR(sv) - dotstrlen - 1))
10223 Perl_croak_nocontext("%s", PL_memory_wrap);
10224 SvGROW(sv, SvCUR(sv) + need + dotstrlen + 1);
10226 if (esignlen && fill == '0') {
10228 for (i = 0; i < (int)esignlen; i++)
10229 *p++ = esignbuf[i];
10231 if (gap && !left) {
10232 memset(p, fill, gap);
10235 if (esignlen && fill != '0') {
10237 for (i = 0; i < (int)esignlen; i++)
10238 *p++ = esignbuf[i];
10242 for (i = zeros; i; i--)
10246 Copy(eptr, p, elen, char);
10250 memset(p, ' ', gap);
10255 Copy(dotstr, p, dotstrlen, char);
10259 vectorize = FALSE; /* done iterating over vecstr */
10266 SvCUR_set(sv, p - SvPVX_const(sv));
10274 /* =========================================================================
10276 =head1 Cloning an interpreter
10278 All the macros and functions in this section are for the private use of
10279 the main function, perl_clone().
10281 The foo_dup() functions make an exact copy of an existing foo thingy.
10282 During the course of a cloning, a hash table is used to map old addresses
10283 to new addresses. The table is created and manipulated with the
10284 ptr_table_* functions.
10288 * =========================================================================*/
10291 #if defined(USE_ITHREADS)
10293 /* XXX Remove this so it doesn't have to go thru the macro and return for nothing */
10294 #ifndef GpREFCNT_inc
10295 # define GpREFCNT_inc(gp) ((gp) ? (++(gp)->gp_refcnt, (gp)) : (GP*)NULL)
10299 /* Certain cases in Perl_ss_dup have been merged, by relying on the fact
10300 that currently av_dup, gv_dup and hv_dup are the same as sv_dup.
10301 If this changes, please unmerge ss_dup.
10302 Likewise, sv_dup_inc_multiple() relies on this fact. */
10303 #define sv_dup_inc(s,t) SvREFCNT_inc(sv_dup(s,t))
10304 #define sv_dup_inc_NN(s,t) SvREFCNT_inc_NN(sv_dup(s,t))
10305 #define av_dup(s,t) MUTABLE_AV(sv_dup((const SV *)s,t))
10306 #define av_dup_inc(s,t) MUTABLE_AV(SvREFCNT_inc(sv_dup((const SV *)s,t)))
10307 #define hv_dup(s,t) MUTABLE_HV(sv_dup((const SV *)s,t))
10308 #define hv_dup_inc(s,t) MUTABLE_HV(SvREFCNT_inc(sv_dup((const SV *)s,t)))
10309 #define cv_dup(s,t) MUTABLE_CV(sv_dup((const SV *)s,t))
10310 #define cv_dup_inc(s,t) MUTABLE_CV(SvREFCNT_inc(sv_dup((const SV *)s,t)))
10311 #define io_dup(s,t) MUTABLE_IO(sv_dup((const SV *)s,t))
10312 #define io_dup_inc(s,t) MUTABLE_IO(SvREFCNT_inc(sv_dup((const SV *)s,t)))
10313 #define gv_dup(s,t) MUTABLE_GV(sv_dup((const SV *)s,t))
10314 #define gv_dup_inc(s,t) MUTABLE_GV(SvREFCNT_inc(sv_dup((const SV *)s,t)))
10315 #define SAVEPV(p) ((p) ? savepv(p) : NULL)
10316 #define SAVEPVN(p,n) ((p) ? savepvn(p,n) : NULL)
10318 /* clone a parser */
10321 Perl_parser_dup(pTHX_ const yy_parser *const proto, CLONE_PARAMS *const param)
10325 PERL_ARGS_ASSERT_PARSER_DUP;
10330 /* look for it in the table first */
10331 parser = (yy_parser *)ptr_table_fetch(PL_ptr_table, proto);
10335 /* create anew and remember what it is */
10336 Newxz(parser, 1, yy_parser);
10337 ptr_table_store(PL_ptr_table, proto, parser);
10339 parser->yyerrstatus = 0;
10340 parser->yychar = YYEMPTY; /* Cause a token to be read. */
10342 /* XXX these not yet duped */
10343 parser->old_parser = NULL;
10344 parser->stack = NULL;
10346 parser->stack_size = 0;
10347 /* XXX parser->stack->state = 0; */
10349 /* XXX eventually, just Copy() most of the parser struct ? */
10351 parser->lex_brackets = proto->lex_brackets;
10352 parser->lex_casemods = proto->lex_casemods;
10353 parser->lex_brackstack = savepvn(proto->lex_brackstack,
10354 (proto->lex_brackets < 120 ? 120 : proto->lex_brackets));
10355 parser->lex_casestack = savepvn(proto->lex_casestack,
10356 (proto->lex_casemods < 12 ? 12 : proto->lex_casemods));
10357 parser->lex_defer = proto->lex_defer;
10358 parser->lex_dojoin = proto->lex_dojoin;
10359 parser->lex_expect = proto->lex_expect;
10360 parser->lex_formbrack = proto->lex_formbrack;
10361 parser->lex_inpat = proto->lex_inpat;
10362 parser->lex_inwhat = proto->lex_inwhat;
10363 parser->lex_op = proto->lex_op;
10364 parser->lex_repl = sv_dup_inc(proto->lex_repl, param);
10365 parser->lex_starts = proto->lex_starts;
10366 parser->lex_stuff = sv_dup_inc(proto->lex_stuff, param);
10367 parser->multi_close = proto->multi_close;
10368 parser->multi_open = proto->multi_open;
10369 parser->multi_start = proto->multi_start;
10370 parser->multi_end = proto->multi_end;
10371 parser->pending_ident = proto->pending_ident;
10372 parser->preambled = proto->preambled;
10373 parser->sublex_info = proto->sublex_info; /* XXX not quite right */
10374 parser->linestr = sv_dup_inc(proto->linestr, param);
10375 parser->expect = proto->expect;
10376 parser->copline = proto->copline;
10377 parser->last_lop_op = proto->last_lop_op;
10378 parser->lex_state = proto->lex_state;
10379 parser->rsfp = fp_dup(proto->rsfp, '<', param);
10380 /* rsfp_filters entries have fake IoDIRP() */
10381 parser->rsfp_filters= av_dup_inc(proto->rsfp_filters, param);
10382 parser->in_my = proto->in_my;
10383 parser->in_my_stash = hv_dup(proto->in_my_stash, param);
10384 parser->error_count = proto->error_count;
10387 parser->linestr = sv_dup_inc(proto->linestr, param);
10390 char * const ols = SvPVX(proto->linestr);
10391 char * const ls = SvPVX(parser->linestr);
10393 parser->bufptr = ls + (proto->bufptr >= ols ?
10394 proto->bufptr - ols : 0);
10395 parser->oldbufptr = ls + (proto->oldbufptr >= ols ?
10396 proto->oldbufptr - ols : 0);
10397 parser->oldoldbufptr= ls + (proto->oldoldbufptr >= ols ?
10398 proto->oldoldbufptr - ols : 0);
10399 parser->linestart = ls + (proto->linestart >= ols ?
10400 proto->linestart - ols : 0);
10401 parser->last_uni = ls + (proto->last_uni >= ols ?
10402 proto->last_uni - ols : 0);
10403 parser->last_lop = ls + (proto->last_lop >= ols ?
10404 proto->last_lop - ols : 0);
10406 parser->bufend = ls + SvCUR(parser->linestr);
10409 Copy(proto->tokenbuf, parser->tokenbuf, 256, char);
10413 parser->endwhite = proto->endwhite;
10414 parser->faketokens = proto->faketokens;
10415 parser->lasttoke = proto->lasttoke;
10416 parser->nextwhite = proto->nextwhite;
10417 parser->realtokenstart = proto->realtokenstart;
10418 parser->skipwhite = proto->skipwhite;
10419 parser->thisclose = proto->thisclose;
10420 parser->thismad = proto->thismad;
10421 parser->thisopen = proto->thisopen;
10422 parser->thisstuff = proto->thisstuff;
10423 parser->thistoken = proto->thistoken;
10424 parser->thiswhite = proto->thiswhite;
10426 Copy(proto->nexttoke, parser->nexttoke, 5, NEXTTOKE);
10427 parser->curforce = proto->curforce;
10429 Copy(proto->nextval, parser->nextval, 5, YYSTYPE);
10430 Copy(proto->nexttype, parser->nexttype, 5, I32);
10431 parser->nexttoke = proto->nexttoke;
10437 /* duplicate a file handle */
10440 Perl_fp_dup(pTHX_ PerlIO *const fp, const char type, CLONE_PARAMS *const param)
10444 PERL_ARGS_ASSERT_FP_DUP;
10445 PERL_UNUSED_ARG(type);
10448 return (PerlIO*)NULL;
10450 /* look for it in the table first */
10451 ret = (PerlIO*)ptr_table_fetch(PL_ptr_table, fp);
10455 /* create anew and remember what it is */
10456 ret = PerlIO_fdupopen(aTHX_ fp, param, PERLIO_DUP_CLONE);
10457 ptr_table_store(PL_ptr_table, fp, ret);
10461 /* duplicate a directory handle */
10464 Perl_dirp_dup(pTHX_ DIR *const dp)
10466 PERL_UNUSED_CONTEXT;
10473 /* duplicate a typeglob */
10476 Perl_gp_dup(pTHX_ GP *const gp, CLONE_PARAMS *const param)
10480 PERL_ARGS_ASSERT_GP_DUP;
10484 /* look for it in the table first */
10485 ret = (GP*)ptr_table_fetch(PL_ptr_table, gp);
10489 /* create anew and remember what it is */
10491 ptr_table_store(PL_ptr_table, gp, ret);
10494 /* ret->gp_refcnt must be 0 before any other dups are called. We're relying
10495 on Newxz() to do this for us. */
10496 ret->gp_sv = sv_dup_inc(gp->gp_sv, param);
10497 ret->gp_io = io_dup_inc(gp->gp_io, param);
10498 ret->gp_form = cv_dup_inc(gp->gp_form, param);
10499 ret->gp_av = av_dup_inc(gp->gp_av, param);
10500 ret->gp_hv = hv_dup_inc(gp->gp_hv, param);
10501 ret->gp_egv = gv_dup(gp->gp_egv, param);/* GvEGV is not refcounted */
10502 ret->gp_cv = cv_dup_inc(gp->gp_cv, param);
10503 ret->gp_cvgen = gp->gp_cvgen;
10504 ret->gp_line = gp->gp_line;
10505 ret->gp_file_hek = hek_dup(gp->gp_file_hek, param);
10509 /* duplicate a chain of magic */
10512 Perl_mg_dup(pTHX_ MAGIC *mg, CLONE_PARAMS *const param)
10514 MAGIC *mgret = NULL;
10515 MAGIC **mgprev_p = &mgret;
10517 PERL_ARGS_ASSERT_MG_DUP;
10519 for (; mg; mg = mg->mg_moremagic) {
10521 Newx(nmg, 1, MAGIC);
10523 mgprev_p = &(nmg->mg_moremagic);
10525 /* There was a comment "XXX copy dynamic vtable?" but as we don't have
10526 dynamic vtables, I'm not sure why Sarathy wrote it. The comment dates
10527 from the original commit adding Perl_mg_dup() - revision 4538.
10528 Similarly there is the annotation "XXX random ptr?" next to the
10529 assignment to nmg->mg_ptr. */
10532 /* FIXME for plugins
10533 if (nmg->mg_type == PERL_MAGIC_qr) {
10534 nmg->mg_obj = MUTABLE_SV(CALLREGDUPE((REGEXP*)nmg->mg_obj, param));
10538 if(nmg->mg_type == PERL_MAGIC_backref) {
10539 /* The backref AV has its reference count deliberately bumped by
10542 = SvREFCNT_inc(av_dup_inc((const AV *) nmg->mg_obj, param));
10545 nmg->mg_obj = (nmg->mg_flags & MGf_REFCOUNTED)
10546 ? sv_dup_inc(nmg->mg_obj, param)
10547 : sv_dup(nmg->mg_obj, param);
10550 if (nmg->mg_ptr && nmg->mg_type != PERL_MAGIC_regex_global) {
10551 if (nmg->mg_len > 0) {
10552 nmg->mg_ptr = SAVEPVN(nmg->mg_ptr, nmg->mg_len);
10553 if (nmg->mg_type == PERL_MAGIC_overload_table &&
10554 AMT_AMAGIC((AMT*)nmg->mg_ptr))
10556 AMT * const namtp = (AMT*)nmg->mg_ptr;
10557 sv_dup_inc_multiple((SV**)(namtp->table),
10558 (SV**)(namtp->table), NofAMmeth, param);
10561 else if (nmg->mg_len == HEf_SVKEY)
10562 nmg->mg_ptr = (char*)sv_dup_inc((const SV *)nmg->mg_ptr, param);
10564 if ((nmg->mg_flags & MGf_DUP) && nmg->mg_virtual && nmg->mg_virtual->svt_dup) {
10565 CALL_FPTR(nmg->mg_virtual->svt_dup)(aTHX_ nmg, param);
10571 #endif /* USE_ITHREADS */
10573 /* create a new pointer-mapping table */
10576 Perl_ptr_table_new(pTHX)
10579 PERL_UNUSED_CONTEXT;
10581 Newx(tbl, 1, PTR_TBL_t);
10582 tbl->tbl_max = 511;
10583 tbl->tbl_items = 0;
10584 Newxz(tbl->tbl_ary, tbl->tbl_max + 1, PTR_TBL_ENT_t*);
10588 #define PTR_TABLE_HASH(ptr) \
10589 ((PTR2UV(ptr) >> 3) ^ (PTR2UV(ptr) >> (3 + 7)) ^ (PTR2UV(ptr) >> (3 + 17)))
10592 we use the PTE_SVSLOT 'reservation' made above, both here (in the
10593 following define) and at call to new_body_inline made below in
10594 Perl_ptr_table_store()
10597 #define del_pte(p) del_body_type(p, PTE_SVSLOT)
10599 /* map an existing pointer using a table */
10601 STATIC PTR_TBL_ENT_t *
10602 S_ptr_table_find(PTR_TBL_t *const tbl, const void *const sv)
10604 PTR_TBL_ENT_t *tblent;
10605 const UV hash = PTR_TABLE_HASH(sv);
10607 PERL_ARGS_ASSERT_PTR_TABLE_FIND;
10609 tblent = tbl->tbl_ary[hash & tbl->tbl_max];
10610 for (; tblent; tblent = tblent->next) {
10611 if (tblent->oldval == sv)
10618 Perl_ptr_table_fetch(pTHX_ PTR_TBL_t *const tbl, const void *const sv)
10620 PTR_TBL_ENT_t const *const tblent = ptr_table_find(tbl, sv);
10622 PERL_ARGS_ASSERT_PTR_TABLE_FETCH;
10623 PERL_UNUSED_CONTEXT;
10625 return tblent ? tblent->newval : NULL;
10628 /* add a new entry to a pointer-mapping table */
10631 Perl_ptr_table_store(pTHX_ PTR_TBL_t *const tbl, const void *const oldsv, void *const newsv)
10633 PTR_TBL_ENT_t *tblent = ptr_table_find(tbl, oldsv);
10635 PERL_ARGS_ASSERT_PTR_TABLE_STORE;
10636 PERL_UNUSED_CONTEXT;
10639 tblent->newval = newsv;
10641 const UV entry = PTR_TABLE_HASH(oldsv) & tbl->tbl_max;
10643 new_body_inline(tblent, PTE_SVSLOT);
10645 tblent->oldval = oldsv;
10646 tblent->newval = newsv;
10647 tblent->next = tbl->tbl_ary[entry];
10648 tbl->tbl_ary[entry] = tblent;
10650 if (tblent->next && tbl->tbl_items > tbl->tbl_max)
10651 ptr_table_split(tbl);
10655 /* double the hash bucket size of an existing ptr table */
10658 Perl_ptr_table_split(pTHX_ PTR_TBL_t *const tbl)
10660 PTR_TBL_ENT_t **ary = tbl->tbl_ary;
10661 const UV oldsize = tbl->tbl_max + 1;
10662 UV newsize = oldsize * 2;
10665 PERL_ARGS_ASSERT_PTR_TABLE_SPLIT;
10666 PERL_UNUSED_CONTEXT;
10668 Renew(ary, newsize, PTR_TBL_ENT_t*);
10669 Zero(&ary[oldsize], newsize-oldsize, PTR_TBL_ENT_t*);
10670 tbl->tbl_max = --newsize;
10671 tbl->tbl_ary = ary;
10672 for (i=0; i < oldsize; i++, ary++) {
10673 PTR_TBL_ENT_t **curentp, **entp, *ent;
10676 curentp = ary + oldsize;
10677 for (entp = ary, ent = *ary; ent; ent = *entp) {
10678 if ((newsize & PTR_TABLE_HASH(ent->oldval)) != i) {
10680 ent->next = *curentp;
10690 /* remove all the entries from a ptr table */
10693 Perl_ptr_table_clear(pTHX_ PTR_TBL_t *const tbl)
10695 if (tbl && tbl->tbl_items) {
10696 register PTR_TBL_ENT_t * const * const array = tbl->tbl_ary;
10697 UV riter = tbl->tbl_max;
10700 PTR_TBL_ENT_t *entry = array[riter];
10703 PTR_TBL_ENT_t * const oentry = entry;
10704 entry = entry->next;
10709 tbl->tbl_items = 0;
10713 /* clear and free a ptr table */
10716 Perl_ptr_table_free(pTHX_ PTR_TBL_t *const tbl)
10721 ptr_table_clear(tbl);
10722 Safefree(tbl->tbl_ary);
10726 #if defined(USE_ITHREADS)
10729 Perl_rvpv_dup(pTHX_ SV *const dstr, const SV *const sstr, CLONE_PARAMS *const param)
10731 PERL_ARGS_ASSERT_RVPV_DUP;
10734 SvRV_set(dstr, SvWEAKREF(sstr)
10735 ? sv_dup(SvRV_const(sstr), param)
10736 : sv_dup_inc(SvRV_const(sstr), param));
10739 else if (SvPVX_const(sstr)) {
10740 /* Has something there */
10742 /* Normal PV - clone whole allocated space */
10743 SvPV_set(dstr, SAVEPVN(SvPVX_const(sstr), SvLEN(sstr)-1));
10744 if (SvREADONLY(sstr) && SvFAKE(sstr)) {
10745 /* Not that normal - actually sstr is copy on write.
10746 But we are a true, independant SV, so: */
10747 SvREADONLY_off(dstr);
10752 /* Special case - not normally malloced for some reason */
10753 if (isGV_with_GP(sstr)) {
10754 /* Don't need to do anything here. */
10756 else if ((SvREADONLY(sstr) && SvFAKE(sstr))) {
10757 /* A "shared" PV - clone it as "shared" PV */
10759 HEK_KEY(hek_dup(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)),
10763 /* Some other special case - random pointer */
10764 SvPV_set(dstr, (char *) SvPVX_const(sstr));
10769 /* Copy the NULL */
10770 SvPV_set(dstr, NULL);
10774 /* duplicate a list of SVs. source and dest may point to the same memory. */
10776 S_sv_dup_inc_multiple(pTHX_ SV *const *source, SV **dest,
10777 SSize_t items, CLONE_PARAMS *const param)
10779 PERL_ARGS_ASSERT_SV_DUP_INC_MULTIPLE;
10781 while (items-- > 0) {
10782 *dest++ = sv_dup_inc(*source++, param);
10788 /* duplicate an SV of any type (including AV, HV etc) */
10791 Perl_sv_dup(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
10796 PERL_ARGS_ASSERT_SV_DUP;
10800 if (SvTYPE(sstr) == SVTYPEMASK) {
10801 #ifdef DEBUG_LEAKING_SCALARS_ABORT
10806 /* look for it in the table first */
10807 dstr = MUTABLE_SV(ptr_table_fetch(PL_ptr_table, sstr));
10811 if(param->flags & CLONEf_JOIN_IN) {
10812 /** We are joining here so we don't want do clone
10813 something that is bad **/
10814 if (SvTYPE(sstr) == SVt_PVHV) {
10815 const HEK * const hvname = HvNAME_HEK(sstr);
10817 /** don't clone stashes if they already exist **/
10818 return MUTABLE_SV(gv_stashpvn(HEK_KEY(hvname), HEK_LEN(hvname), 0));
10822 /* create anew and remember what it is */
10825 #ifdef DEBUG_LEAKING_SCALARS
10826 dstr->sv_debug_optype = sstr->sv_debug_optype;
10827 dstr->sv_debug_line = sstr->sv_debug_line;
10828 dstr->sv_debug_inpad = sstr->sv_debug_inpad;
10829 dstr->sv_debug_cloned = 1;
10830 dstr->sv_debug_file = savepv(sstr->sv_debug_file);
10833 ptr_table_store(PL_ptr_table, sstr, dstr);
10836 SvFLAGS(dstr) = SvFLAGS(sstr);
10837 SvFLAGS(dstr) &= ~SVf_OOK; /* don't propagate OOK hack */
10838 SvREFCNT(dstr) = 0; /* must be before any other dups! */
10841 if (SvANY(sstr) && PL_watch_pvx && SvPVX_const(sstr) == PL_watch_pvx)
10842 PerlIO_printf(Perl_debug_log, "watch at %p hit, found string \"%s\"\n",
10843 (void*)PL_watch_pvx, SvPVX_const(sstr));
10846 /* don't clone objects whose class has asked us not to */
10847 if (SvOBJECT(sstr) && ! (SvFLAGS(SvSTASH(sstr)) & SVphv_CLONEABLE)) {
10852 switch (SvTYPE(sstr)) {
10854 SvANY(dstr) = NULL;
10857 SvANY(dstr) = (XPVIV*)((char*)&(dstr->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
10859 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
10861 SvIV_set(dstr, SvIVX(sstr));
10865 SvANY(dstr) = new_XNV();
10866 SvNV_set(dstr, SvNVX(sstr));
10868 /* case SVt_BIND: */
10871 /* These are all the types that need complex bodies allocating. */
10873 const svtype sv_type = SvTYPE(sstr);
10874 const struct body_details *const sv_type_details
10875 = bodies_by_type + sv_type;
10879 Perl_croak(aTHX_ "Bizarre SvTYPE [%" IVdf "]", (IV)SvTYPE(sstr));
10894 assert(sv_type_details->body_size);
10895 if (sv_type_details->arena) {
10896 new_body_inline(new_body, sv_type);
10898 = (void*)((char*)new_body - sv_type_details->offset);
10900 new_body = new_NOARENA(sv_type_details);
10904 SvANY(dstr) = new_body;
10907 Copy(((char*)SvANY(sstr)) + sv_type_details->offset,
10908 ((char*)SvANY(dstr)) + sv_type_details->offset,
10909 sv_type_details->copy, char);
10911 Copy(((char*)SvANY(sstr)),
10912 ((char*)SvANY(dstr)),
10913 sv_type_details->body_size + sv_type_details->offset, char);
10916 if (sv_type != SVt_PVAV && sv_type != SVt_PVHV
10917 && !isGV_with_GP(dstr))
10918 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
10920 /* The Copy above means that all the source (unduplicated) pointers
10921 are now in the destination. We can check the flags and the
10922 pointers in either, but it's possible that there's less cache
10923 missing by always going for the destination.
10924 FIXME - instrument and check that assumption */
10925 if (sv_type >= SVt_PVMG) {
10926 if ((sv_type == SVt_PVMG) && SvPAD_OUR(dstr)) {
10927 SvOURSTASH_set(dstr, hv_dup_inc(SvOURSTASH(dstr), param));
10928 } else if (SvMAGIC(dstr))
10929 SvMAGIC_set(dstr, mg_dup(SvMAGIC(dstr), param));
10931 SvSTASH_set(dstr, hv_dup_inc(SvSTASH(dstr), param));
10934 /* The cast silences a GCC warning about unhandled types. */
10935 switch ((int)sv_type) {
10945 /* FIXME for plugins */
10946 re_dup_guts((REGEXP*) sstr, (REGEXP*) dstr, param);
10949 /* XXX LvTARGOFF sometimes holds PMOP* when DEBUGGING */
10950 if (LvTYPE(dstr) == 't') /* for tie: unrefcnted fake (SV**) */
10951 LvTARG(dstr) = dstr;
10952 else if (LvTYPE(dstr) == 'T') /* for tie: fake HE */
10953 LvTARG(dstr) = MUTABLE_SV(he_dup((HE*)LvTARG(dstr), 0, param));
10955 LvTARG(dstr) = sv_dup_inc(LvTARG(dstr), param);
10957 if(isGV_with_GP(sstr)) {
10958 GvNAME_HEK(dstr) = hek_dup(GvNAME_HEK(dstr), param);
10959 /* Don't call sv_add_backref here as it's going to be
10960 created as part of the magic cloning of the symbol
10962 /* Danger Will Robinson - GvGP(dstr) isn't initialised
10963 at the point of this comment. */
10964 GvSTASH(dstr) = hv_dup(GvSTASH(dstr), param);
10965 GvGP(dstr) = gp_dup(GvGP(sstr), param);
10966 (void)GpREFCNT_inc(GvGP(dstr));
10968 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
10971 IoIFP(dstr) = fp_dup(IoIFP(dstr), IoTYPE(dstr), param);
10972 if (IoOFP(dstr) == IoIFP(sstr))
10973 IoOFP(dstr) = IoIFP(dstr);
10975 IoOFP(dstr) = fp_dup(IoOFP(dstr), IoTYPE(dstr), param);
10976 /* PL_parser->rsfp_filters entries have fake IoDIRP() */
10977 if(IoFLAGS(dstr) & IOf_FAKE_DIRP) {
10978 /* I have no idea why fake dirp (rsfps)
10979 should be treated differently but otherwise
10980 we end up with leaks -- sky*/
10981 IoTOP_GV(dstr) = gv_dup_inc(IoTOP_GV(dstr), param);
10982 IoFMT_GV(dstr) = gv_dup_inc(IoFMT_GV(dstr), param);
10983 IoBOTTOM_GV(dstr) = gv_dup_inc(IoBOTTOM_GV(dstr), param);
10985 IoTOP_GV(dstr) = gv_dup(IoTOP_GV(dstr), param);
10986 IoFMT_GV(dstr) = gv_dup(IoFMT_GV(dstr), param);
10987 IoBOTTOM_GV(dstr) = gv_dup(IoBOTTOM_GV(dstr), param);
10988 if (IoDIRP(dstr)) {
10989 IoDIRP(dstr) = dirp_dup(IoDIRP(dstr));
10992 /* IoDIRP(dstr) is already a copy of IoDIRP(sstr) */
10995 IoTOP_NAME(dstr) = SAVEPV(IoTOP_NAME(dstr));
10996 IoFMT_NAME(dstr) = SAVEPV(IoFMT_NAME(dstr));
10997 IoBOTTOM_NAME(dstr) = SAVEPV(IoBOTTOM_NAME(dstr));
11000 /* avoid cloning an empty array */
11001 if (AvARRAY((const AV *)sstr) && AvFILLp((const AV *)sstr) >= 0) {
11002 SV **dst_ary, **src_ary;
11003 SSize_t items = AvFILLp((const AV *)sstr) + 1;
11005 src_ary = AvARRAY((const AV *)sstr);
11006 Newxz(dst_ary, AvMAX((const AV *)sstr)+1, SV*);
11007 ptr_table_store(PL_ptr_table, src_ary, dst_ary);
11008 AvARRAY(MUTABLE_AV(dstr)) = dst_ary;
11009 AvALLOC((const AV *)dstr) = dst_ary;
11010 if (AvREAL((const AV *)sstr)) {
11011 dst_ary = sv_dup_inc_multiple(src_ary, dst_ary, items,
11015 while (items-- > 0)
11016 *dst_ary++ = sv_dup(*src_ary++, param);
11018 items = AvMAX((const AV *)sstr) - AvFILLp((const AV *)sstr);
11019 while (items-- > 0) {
11020 *dst_ary++ = &PL_sv_undef;
11024 AvARRAY(MUTABLE_AV(dstr)) = NULL;
11025 AvALLOC((const AV *)dstr) = (SV**)NULL;
11026 AvMAX( (const AV *)dstr) = -1;
11027 AvFILLp((const AV *)dstr) = -1;
11031 if (HvARRAY((const HV *)sstr)) {
11033 const bool sharekeys = !!HvSHAREKEYS(sstr);
11034 XPVHV * const dxhv = (XPVHV*)SvANY(dstr);
11035 XPVHV * const sxhv = (XPVHV*)SvANY(sstr);
11037 Newx(darray, PERL_HV_ARRAY_ALLOC_BYTES(dxhv->xhv_max+1)
11038 + (SvOOK(sstr) ? sizeof(struct xpvhv_aux) : 0),
11040 HvARRAY(dstr) = (HE**)darray;
11041 while (i <= sxhv->xhv_max) {
11042 const HE * const source = HvARRAY(sstr)[i];
11043 HvARRAY(dstr)[i] = source
11044 ? he_dup(source, sharekeys, param) : 0;
11049 const struct xpvhv_aux * const saux = HvAUX(sstr);
11050 struct xpvhv_aux * const daux = HvAUX(dstr);
11051 /* This flag isn't copied. */
11052 /* SvOOK_on(hv) attacks the IV flags. */
11053 SvFLAGS(dstr) |= SVf_OOK;
11055 hvname = saux->xhv_name;
11056 daux->xhv_name = hek_dup(hvname, param);
11058 daux->xhv_riter = saux->xhv_riter;
11059 daux->xhv_eiter = saux->xhv_eiter
11060 ? he_dup(saux->xhv_eiter,
11061 (bool)!!HvSHAREKEYS(sstr), param) : 0;
11062 /* backref array needs refcnt=2; see sv_add_backref */
11063 daux->xhv_backreferences =
11064 saux->xhv_backreferences
11065 ? MUTABLE_AV(SvREFCNT_inc(
11066 sv_dup_inc((const SV *)saux->xhv_backreferences, param)))
11069 daux->xhv_mro_meta = saux->xhv_mro_meta
11070 ? mro_meta_dup(saux->xhv_mro_meta, param)
11073 /* Record stashes for possible cloning in Perl_clone(). */
11075 av_push(param->stashes, dstr);
11079 HvARRAY(MUTABLE_HV(dstr)) = NULL;
11082 if (!(param->flags & CLONEf_COPY_STACKS)) {
11086 /* NOTE: not refcounted */
11087 CvSTASH(dstr) = hv_dup(CvSTASH(dstr), param);
11089 if (!CvISXSUB(dstr))
11090 CvROOT(dstr) = OpREFCNT_inc(CvROOT(dstr));
11092 if (CvCONST(dstr) && CvISXSUB(dstr)) {
11093 CvXSUBANY(dstr).any_ptr =
11094 sv_dup_inc((const SV *)CvXSUBANY(dstr).any_ptr, param);
11096 /* don't dup if copying back - CvGV isn't refcounted, so the
11097 * duped GV may never be freed. A bit of a hack! DAPM */
11098 CvGV(dstr) = (param->flags & CLONEf_JOIN_IN) ?
11099 NULL : gv_dup(CvGV(dstr), param) ;
11100 PAD_DUP(CvPADLIST(dstr), CvPADLIST(sstr), param);
11102 CvWEAKOUTSIDE(sstr)
11103 ? cv_dup( CvOUTSIDE(dstr), param)
11104 : cv_dup_inc(CvOUTSIDE(dstr), param);
11105 if (!CvISXSUB(dstr))
11106 CvFILE(dstr) = SAVEPV(CvFILE(dstr));
11112 if (SvOBJECT(dstr) && SvTYPE(dstr) != SVt_PVIO)
11118 /* duplicate a context */
11121 Perl_cx_dup(pTHX_ PERL_CONTEXT *cxs, I32 ix, I32 max, CLONE_PARAMS* param)
11123 PERL_CONTEXT *ncxs;
11125 PERL_ARGS_ASSERT_CX_DUP;
11128 return (PERL_CONTEXT*)NULL;
11130 /* look for it in the table first */
11131 ncxs = (PERL_CONTEXT*)ptr_table_fetch(PL_ptr_table, cxs);
11135 /* create anew and remember what it is */
11136 Newx(ncxs, max + 1, PERL_CONTEXT);
11137 ptr_table_store(PL_ptr_table, cxs, ncxs);
11138 Copy(cxs, ncxs, max + 1, PERL_CONTEXT);
11141 PERL_CONTEXT * const ncx = &ncxs[ix];
11142 if (CxTYPE(ncx) == CXt_SUBST) {
11143 Perl_croak(aTHX_ "Cloning substitution context is unimplemented");
11146 switch (CxTYPE(ncx)) {
11148 ncx->blk_sub.cv = (ncx->blk_sub.olddepth == 0
11149 ? cv_dup_inc(ncx->blk_sub.cv, param)
11150 : cv_dup(ncx->blk_sub.cv,param));
11151 ncx->blk_sub.argarray = (CxHASARGS(ncx)
11152 ? av_dup_inc(ncx->blk_sub.argarray,
11155 ncx->blk_sub.savearray = av_dup_inc(ncx->blk_sub.savearray,
11157 ncx->blk_sub.oldcomppad = (PAD*)ptr_table_fetch(PL_ptr_table,
11158 ncx->blk_sub.oldcomppad);
11161 ncx->blk_eval.old_namesv = sv_dup_inc(ncx->blk_eval.old_namesv,
11163 ncx->blk_eval.cur_text = sv_dup(ncx->blk_eval.cur_text, param);
11165 case CXt_LOOP_LAZYSV:
11166 ncx->blk_loop.state_u.lazysv.end
11167 = sv_dup_inc(ncx->blk_loop.state_u.lazysv.end, param);
11168 /* We are taking advantage of av_dup_inc and sv_dup_inc
11169 actually being the same function, and order equivalance of
11171 We can assert the later [but only at run time :-(] */
11172 assert ((void *) &ncx->blk_loop.state_u.ary.ary ==
11173 (void *) &ncx->blk_loop.state_u.lazysv.cur);
11175 ncx->blk_loop.state_u.ary.ary
11176 = av_dup_inc(ncx->blk_loop.state_u.ary.ary, param);
11177 case CXt_LOOP_LAZYIV:
11178 case CXt_LOOP_PLAIN:
11179 if (CxPADLOOP(ncx)) {
11180 ncx->blk_loop.oldcomppad
11181 = (PAD*)ptr_table_fetch(PL_ptr_table,
11182 ncx->blk_loop.oldcomppad);
11184 ncx->blk_loop.oldcomppad
11185 = (PAD*)gv_dup((const GV *)ncx->blk_loop.oldcomppad,
11190 ncx->blk_format.cv = cv_dup(ncx->blk_format.cv, param);
11191 ncx->blk_format.gv = gv_dup(ncx->blk_format.gv, param);
11192 ncx->blk_format.dfoutgv = gv_dup_inc(ncx->blk_format.dfoutgv,
11205 /* duplicate a stack info structure */
11208 Perl_si_dup(pTHX_ PERL_SI *si, CLONE_PARAMS* param)
11212 PERL_ARGS_ASSERT_SI_DUP;
11215 return (PERL_SI*)NULL;
11217 /* look for it in the table first */
11218 nsi = (PERL_SI*)ptr_table_fetch(PL_ptr_table, si);
11222 /* create anew and remember what it is */
11223 Newxz(nsi, 1, PERL_SI);
11224 ptr_table_store(PL_ptr_table, si, nsi);
11226 nsi->si_stack = av_dup_inc(si->si_stack, param);
11227 nsi->si_cxix = si->si_cxix;
11228 nsi->si_cxmax = si->si_cxmax;
11229 nsi->si_cxstack = cx_dup(si->si_cxstack, si->si_cxix, si->si_cxmax, param);
11230 nsi->si_type = si->si_type;
11231 nsi->si_prev = si_dup(si->si_prev, param);
11232 nsi->si_next = si_dup(si->si_next, param);
11233 nsi->si_markoff = si->si_markoff;
11238 #define POPINT(ss,ix) ((ss)[--(ix)].any_i32)
11239 #define TOPINT(ss,ix) ((ss)[ix].any_i32)
11240 #define POPLONG(ss,ix) ((ss)[--(ix)].any_long)
11241 #define TOPLONG(ss,ix) ((ss)[ix].any_long)
11242 #define POPIV(ss,ix) ((ss)[--(ix)].any_iv)
11243 #define TOPIV(ss,ix) ((ss)[ix].any_iv)
11244 #define POPBOOL(ss,ix) ((ss)[--(ix)].any_bool)
11245 #define TOPBOOL(ss,ix) ((ss)[ix].any_bool)
11246 #define POPPTR(ss,ix) ((ss)[--(ix)].any_ptr)
11247 #define TOPPTR(ss,ix) ((ss)[ix].any_ptr)
11248 #define POPDPTR(ss,ix) ((ss)[--(ix)].any_dptr)
11249 #define TOPDPTR(ss,ix) ((ss)[ix].any_dptr)
11250 #define POPDXPTR(ss,ix) ((ss)[--(ix)].any_dxptr)
11251 #define TOPDXPTR(ss,ix) ((ss)[ix].any_dxptr)
11254 #define pv_dup_inc(p) SAVEPV(p)
11255 #define pv_dup(p) SAVEPV(p)
11256 #define svp_dup_inc(p,pp) any_dup(p,pp)
11258 /* map any object to the new equivent - either something in the
11259 * ptr table, or something in the interpreter structure
11263 Perl_any_dup(pTHX_ void *v, const PerlInterpreter *proto_perl)
11267 PERL_ARGS_ASSERT_ANY_DUP;
11270 return (void*)NULL;
11272 /* look for it in the table first */
11273 ret = ptr_table_fetch(PL_ptr_table, v);
11277 /* see if it is part of the interpreter structure */
11278 if (v >= (void*)proto_perl && v < (void*)(proto_perl+1))
11279 ret = (void*)(((char*)aTHX) + (((char*)v) - (char*)proto_perl));
11287 /* duplicate the save stack */
11290 Perl_ss_dup(pTHX_ PerlInterpreter *proto_perl, CLONE_PARAMS* param)
11293 ANY * const ss = proto_perl->Isavestack;
11294 const I32 max = proto_perl->Isavestack_max;
11295 I32 ix = proto_perl->Isavestack_ix;
11308 void (*dptr) (void*);
11309 void (*dxptr) (pTHX_ void*);
11311 PERL_ARGS_ASSERT_SS_DUP;
11313 Newxz(nss, max, ANY);
11316 const I32 type = POPINT(ss,ix);
11317 TOPINT(nss,ix) = type;
11319 case SAVEt_HELEM: /* hash element */
11320 sv = (const SV *)POPPTR(ss,ix);
11321 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11323 case SAVEt_ITEM: /* normal string */
11324 case SAVEt_SV: /* scalar reference */
11325 sv = (const SV *)POPPTR(ss,ix);
11326 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11329 case SAVEt_MORTALIZESV:
11330 sv = (const SV *)POPPTR(ss,ix);
11331 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11333 case SAVEt_SHARED_PVREF: /* char* in shared space */
11334 c = (char*)POPPTR(ss,ix);
11335 TOPPTR(nss,ix) = savesharedpv(c);
11336 ptr = POPPTR(ss,ix);
11337 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11339 case SAVEt_GENERIC_SVREF: /* generic sv */
11340 case SAVEt_SVREF: /* scalar reference */
11341 sv = (const SV *)POPPTR(ss,ix);
11342 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11343 ptr = POPPTR(ss,ix);
11344 TOPPTR(nss,ix) = svp_dup_inc((SV**)ptr, proto_perl);/* XXXXX */
11346 case SAVEt_HV: /* hash reference */
11347 case SAVEt_AV: /* array reference */
11348 sv = (const SV *) POPPTR(ss,ix);
11349 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11351 case SAVEt_COMPPAD:
11353 sv = (const SV *) POPPTR(ss,ix);
11354 TOPPTR(nss,ix) = sv_dup(sv, param);
11356 case SAVEt_INT: /* int reference */
11357 ptr = POPPTR(ss,ix);
11358 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11359 intval = (int)POPINT(ss,ix);
11360 TOPINT(nss,ix) = intval;
11362 case SAVEt_LONG: /* long reference */
11363 ptr = POPPTR(ss,ix);
11364 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11366 case SAVEt_CLEARSV:
11367 longval = (long)POPLONG(ss,ix);
11368 TOPLONG(nss,ix) = longval;
11370 case SAVEt_I32: /* I32 reference */
11371 case SAVEt_I16: /* I16 reference */
11372 case SAVEt_I8: /* I8 reference */
11373 case SAVEt_COP_ARYBASE: /* call CopARYBASE_set */
11374 ptr = POPPTR(ss,ix);
11375 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11377 TOPINT(nss,ix) = i;
11379 case SAVEt_IV: /* IV reference */
11380 ptr = POPPTR(ss,ix);
11381 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11383 TOPIV(nss,ix) = iv;
11385 case SAVEt_HPTR: /* HV* reference */
11386 case SAVEt_APTR: /* AV* reference */
11387 case SAVEt_SPTR: /* SV* reference */
11388 ptr = POPPTR(ss,ix);
11389 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11390 sv = (const SV *)POPPTR(ss,ix);
11391 TOPPTR(nss,ix) = sv_dup(sv, param);
11393 case SAVEt_VPTR: /* random* reference */
11394 ptr = POPPTR(ss,ix);
11395 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11396 ptr = POPPTR(ss,ix);
11397 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11399 case SAVEt_GENERIC_PVREF: /* generic char* */
11400 case SAVEt_PPTR: /* char* reference */
11401 ptr = POPPTR(ss,ix);
11402 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11403 c = (char*)POPPTR(ss,ix);
11404 TOPPTR(nss,ix) = pv_dup(c);
11406 case SAVEt_GP: /* scalar reference */
11407 gp = (GP*)POPPTR(ss,ix);
11408 TOPPTR(nss,ix) = gp = gp_dup(gp, param);
11409 (void)GpREFCNT_inc(gp);
11410 gv = (const GV *)POPPTR(ss,ix);
11411 TOPPTR(nss,ix) = gv_dup_inc(gv, param);
11414 ptr = POPPTR(ss,ix);
11415 if (ptr && (((OP*)ptr)->op_private & OPpREFCOUNTED)) {
11416 /* these are assumed to be refcounted properly */
11418 switch (((OP*)ptr)->op_type) {
11420 case OP_LEAVESUBLV:
11424 case OP_LEAVEWRITE:
11425 TOPPTR(nss,ix) = ptr;
11428 (void) OpREFCNT_inc(o);
11432 TOPPTR(nss,ix) = NULL;
11437 TOPPTR(nss,ix) = NULL;
11440 hv = (const HV *)POPPTR(ss,ix);
11441 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
11443 TOPINT(nss,ix) = i;
11446 c = (char*)POPPTR(ss,ix);
11447 TOPPTR(nss,ix) = pv_dup_inc(c);
11449 case SAVEt_STACK_POS: /* Position on Perl stack */
11451 TOPINT(nss,ix) = i;
11453 case SAVEt_DESTRUCTOR:
11454 ptr = POPPTR(ss,ix);
11455 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
11456 dptr = POPDPTR(ss,ix);
11457 TOPDPTR(nss,ix) = DPTR2FPTR(void (*)(void*),
11458 any_dup(FPTR2DPTR(void *, dptr),
11461 case SAVEt_DESTRUCTOR_X:
11462 ptr = POPPTR(ss,ix);
11463 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
11464 dxptr = POPDXPTR(ss,ix);
11465 TOPDXPTR(nss,ix) = DPTR2FPTR(void (*)(pTHX_ void*),
11466 any_dup(FPTR2DPTR(void *, dxptr),
11469 case SAVEt_REGCONTEXT:
11472 TOPINT(nss,ix) = i;
11475 case SAVEt_AELEM: /* array element */
11476 sv = (const SV *)POPPTR(ss,ix);
11477 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11479 TOPINT(nss,ix) = i;
11480 av = (const AV *)POPPTR(ss,ix);
11481 TOPPTR(nss,ix) = av_dup_inc(av, param);
11484 ptr = POPPTR(ss,ix);
11485 TOPPTR(nss,ix) = ptr;
11488 ptr = POPPTR(ss,ix);
11491 ((struct refcounted_he *)ptr)->refcounted_he_refcnt++;
11492 HINTS_REFCNT_UNLOCK;
11494 TOPPTR(nss,ix) = ptr;
11496 TOPINT(nss,ix) = i;
11497 if (i & HINT_LOCALIZE_HH) {
11498 hv = (const HV *)POPPTR(ss,ix);
11499 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
11502 case SAVEt_PADSV_AND_MORTALIZE:
11503 longval = (long)POPLONG(ss,ix);
11504 TOPLONG(nss,ix) = longval;
11505 ptr = POPPTR(ss,ix);
11506 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11507 sv = (const SV *)POPPTR(ss,ix);
11508 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11511 ptr = POPPTR(ss,ix);
11512 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11513 longval = (long)POPBOOL(ss,ix);
11514 TOPBOOL(nss,ix) = (bool)longval;
11516 case SAVEt_SET_SVFLAGS:
11518 TOPINT(nss,ix) = i;
11520 TOPINT(nss,ix) = i;
11521 sv = (const SV *)POPPTR(ss,ix);
11522 TOPPTR(nss,ix) = sv_dup(sv, param);
11524 case SAVEt_RE_STATE:
11526 const struct re_save_state *const old_state
11527 = (struct re_save_state *)
11528 (ss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
11529 struct re_save_state *const new_state
11530 = (struct re_save_state *)
11531 (nss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
11533 Copy(old_state, new_state, 1, struct re_save_state);
11534 ix -= SAVESTACK_ALLOC_FOR_RE_SAVE_STATE;
11536 new_state->re_state_bostr
11537 = pv_dup(old_state->re_state_bostr);
11538 new_state->re_state_reginput
11539 = pv_dup(old_state->re_state_reginput);
11540 new_state->re_state_regeol
11541 = pv_dup(old_state->re_state_regeol);
11542 new_state->re_state_regoffs
11543 = (regexp_paren_pair*)
11544 any_dup(old_state->re_state_regoffs, proto_perl);
11545 new_state->re_state_reglastparen
11546 = (U32*) any_dup(old_state->re_state_reglastparen,
11548 new_state->re_state_reglastcloseparen
11549 = (U32*)any_dup(old_state->re_state_reglastcloseparen,
11551 /* XXX This just has to be broken. The old save_re_context
11552 code did SAVEGENERICPV(PL_reg_start_tmp);
11553 PL_reg_start_tmp is char **.
11554 Look above to what the dup code does for
11555 SAVEt_GENERIC_PVREF
11556 It can never have worked.
11557 So this is merely a faithful copy of the exiting bug: */
11558 new_state->re_state_reg_start_tmp
11559 = (char **) pv_dup((char *)
11560 old_state->re_state_reg_start_tmp);
11561 /* I assume that it only ever "worked" because no-one called
11562 (pseudo)fork while the regexp engine had re-entered itself.
11564 #ifdef PERL_OLD_COPY_ON_WRITE
11565 new_state->re_state_nrs
11566 = sv_dup(old_state->re_state_nrs, param);
11568 new_state->re_state_reg_magic
11569 = (MAGIC*) any_dup(old_state->re_state_reg_magic,
11571 new_state->re_state_reg_oldcurpm
11572 = (PMOP*) any_dup(old_state->re_state_reg_oldcurpm,
11574 new_state->re_state_reg_curpm
11575 = (PMOP*) any_dup(old_state->re_state_reg_curpm,
11577 new_state->re_state_reg_oldsaved
11578 = pv_dup(old_state->re_state_reg_oldsaved);
11579 new_state->re_state_reg_poscache
11580 = pv_dup(old_state->re_state_reg_poscache);
11581 new_state->re_state_reg_starttry
11582 = pv_dup(old_state->re_state_reg_starttry);
11585 case SAVEt_COMPILE_WARNINGS:
11586 ptr = POPPTR(ss,ix);
11587 TOPPTR(nss,ix) = DUP_WARNINGS((STRLEN*)ptr);
11590 ptr = POPPTR(ss,ix);
11591 TOPPTR(nss,ix) = parser_dup((const yy_parser*)ptr, param);
11595 "panic: ss_dup inconsistency (%"IVdf")", (IV) type);
11603 /* if sv is a stash, call $class->CLONE_SKIP(), and set the SVphv_CLONEABLE
11604 * flag to the result. This is done for each stash before cloning starts,
11605 * so we know which stashes want their objects cloned */
11608 do_mark_cloneable_stash(pTHX_ SV *const sv)
11610 const HEK * const hvname = HvNAME_HEK((const HV *)sv);
11612 GV* const cloner = gv_fetchmethod_autoload(MUTABLE_HV(sv), "CLONE_SKIP", 0);
11613 SvFLAGS(sv) |= SVphv_CLONEABLE; /* clone objects by default */
11614 if (cloner && GvCV(cloner)) {
11621 mXPUSHs(newSVhek(hvname));
11623 call_sv(MUTABLE_SV(GvCV(cloner)), G_SCALAR);
11630 SvFLAGS(sv) &= ~SVphv_CLONEABLE;
11638 =for apidoc perl_clone
11640 Create and return a new interpreter by cloning the current one.
11642 perl_clone takes these flags as parameters:
11644 CLONEf_COPY_STACKS - is used to, well, copy the stacks also,
11645 without it we only clone the data and zero the stacks,
11646 with it we copy the stacks and the new perl interpreter is
11647 ready to run at the exact same point as the previous one.
11648 The pseudo-fork code uses COPY_STACKS while the
11649 threads->create doesn't.
11651 CLONEf_KEEP_PTR_TABLE
11652 perl_clone keeps a ptr_table with the pointer of the old
11653 variable as a key and the new variable as a value,
11654 this allows it to check if something has been cloned and not
11655 clone it again but rather just use the value and increase the
11656 refcount. If KEEP_PTR_TABLE is not set then perl_clone will kill
11657 the ptr_table using the function
11658 C<ptr_table_free(PL_ptr_table); PL_ptr_table = NULL;>,
11659 reason to keep it around is if you want to dup some of your own
11660 variable who are outside the graph perl scans, example of this
11661 code is in threads.xs create
11664 This is a win32 thing, it is ignored on unix, it tells perls
11665 win32host code (which is c++) to clone itself, this is needed on
11666 win32 if you want to run two threads at the same time,
11667 if you just want to do some stuff in a separate perl interpreter
11668 and then throw it away and return to the original one,
11669 you don't need to do anything.
11674 /* XXX the above needs expanding by someone who actually understands it ! */
11675 EXTERN_C PerlInterpreter *
11676 perl_clone_host(PerlInterpreter* proto_perl, UV flags);
11679 perl_clone(PerlInterpreter *proto_perl, UV flags)
11682 #ifdef PERL_IMPLICIT_SYS
11684 PERL_ARGS_ASSERT_PERL_CLONE;
11686 /* perlhost.h so we need to call into it
11687 to clone the host, CPerlHost should have a c interface, sky */
11689 if (flags & CLONEf_CLONE_HOST) {
11690 return perl_clone_host(proto_perl,flags);
11692 return perl_clone_using(proto_perl, flags,
11694 proto_perl->IMemShared,
11695 proto_perl->IMemParse,
11697 proto_perl->IStdIO,
11701 proto_perl->IProc);
11705 perl_clone_using(PerlInterpreter *proto_perl, UV flags,
11706 struct IPerlMem* ipM, struct IPerlMem* ipMS,
11707 struct IPerlMem* ipMP, struct IPerlEnv* ipE,
11708 struct IPerlStdIO* ipStd, struct IPerlLIO* ipLIO,
11709 struct IPerlDir* ipD, struct IPerlSock* ipS,
11710 struct IPerlProc* ipP)
11712 /* XXX many of the string copies here can be optimized if they're
11713 * constants; they need to be allocated as common memory and just
11714 * their pointers copied. */
11717 CLONE_PARAMS clone_params;
11718 CLONE_PARAMS* const param = &clone_params;
11720 PerlInterpreter * const my_perl = (PerlInterpreter*)(*ipM->pMalloc)(ipM, sizeof(PerlInterpreter));
11722 PERL_ARGS_ASSERT_PERL_CLONE_USING;
11724 /* for each stash, determine whether its objects should be cloned */
11725 S_visit(proto_perl, do_mark_cloneable_stash, SVt_PVHV, SVTYPEMASK);
11726 PERL_SET_THX(my_perl);
11729 PoisonNew(my_perl, 1, PerlInterpreter);
11735 PL_savestack_ix = 0;
11736 PL_savestack_max = -1;
11737 PL_sig_pending = 0;
11739 Zero(&PL_debug_pad, 1, struct perl_debug_pad);
11740 # else /* !DEBUGGING */
11741 Zero(my_perl, 1, PerlInterpreter);
11742 # endif /* DEBUGGING */
11744 /* host pointers */
11746 PL_MemShared = ipMS;
11747 PL_MemParse = ipMP;
11754 #else /* !PERL_IMPLICIT_SYS */
11756 CLONE_PARAMS clone_params;
11757 CLONE_PARAMS* param = &clone_params;
11758 PerlInterpreter * const my_perl = (PerlInterpreter*)PerlMem_malloc(sizeof(PerlInterpreter));
11760 PERL_ARGS_ASSERT_PERL_CLONE;
11762 /* for each stash, determine whether its objects should be cloned */
11763 S_visit(proto_perl, do_mark_cloneable_stash, SVt_PVHV, SVTYPEMASK);
11764 PERL_SET_THX(my_perl);
11767 PoisonNew(my_perl, 1, PerlInterpreter);
11773 PL_savestack_ix = 0;
11774 PL_savestack_max = -1;
11775 PL_sig_pending = 0;
11777 Zero(&PL_debug_pad, 1, struct perl_debug_pad);
11778 # else /* !DEBUGGING */
11779 Zero(my_perl, 1, PerlInterpreter);
11780 # endif /* DEBUGGING */
11781 #endif /* PERL_IMPLICIT_SYS */
11782 param->flags = flags;
11783 param->proto_perl = proto_perl;
11785 INIT_TRACK_MEMPOOL(my_perl->Imemory_debug_header, my_perl);
11787 PL_body_arenas = NULL;
11788 Zero(&PL_body_roots, 1, PL_body_roots);
11790 PL_nice_chunk = NULL;
11791 PL_nice_chunk_size = 0;
11793 PL_sv_objcount = 0;
11795 PL_sv_arenaroot = NULL;
11797 PL_debug = proto_perl->Idebug;
11799 PL_hash_seed = proto_perl->Ihash_seed;
11800 PL_rehash_seed = proto_perl->Irehash_seed;
11802 #ifdef USE_REENTRANT_API
11803 /* XXX: things like -Dm will segfault here in perlio, but doing
11804 * PERL_SET_CONTEXT(proto_perl);
11805 * breaks too many other things
11807 Perl_reentrant_init(aTHX);
11810 /* create SV map for pointer relocation */
11811 PL_ptr_table = ptr_table_new();
11813 /* initialize these special pointers as early as possible */
11814 SvANY(&PL_sv_undef) = NULL;
11815 SvREFCNT(&PL_sv_undef) = (~(U32)0)/2;
11816 SvFLAGS(&PL_sv_undef) = SVf_READONLY|SVt_NULL;
11817 ptr_table_store(PL_ptr_table, &proto_perl->Isv_undef, &PL_sv_undef);
11819 SvANY(&PL_sv_no) = new_XPVNV();
11820 SvREFCNT(&PL_sv_no) = (~(U32)0)/2;
11821 SvFLAGS(&PL_sv_no) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
11822 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
11823 SvPV_set(&PL_sv_no, savepvn(PL_No, 0));
11824 SvCUR_set(&PL_sv_no, 0);
11825 SvLEN_set(&PL_sv_no, 1);
11826 SvIV_set(&PL_sv_no, 0);
11827 SvNV_set(&PL_sv_no, 0);
11828 ptr_table_store(PL_ptr_table, &proto_perl->Isv_no, &PL_sv_no);
11830 SvANY(&PL_sv_yes) = new_XPVNV();
11831 SvREFCNT(&PL_sv_yes) = (~(U32)0)/2;
11832 SvFLAGS(&PL_sv_yes) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
11833 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
11834 SvPV_set(&PL_sv_yes, savepvn(PL_Yes, 1));
11835 SvCUR_set(&PL_sv_yes, 1);
11836 SvLEN_set(&PL_sv_yes, 2);
11837 SvIV_set(&PL_sv_yes, 1);
11838 SvNV_set(&PL_sv_yes, 1);
11839 ptr_table_store(PL_ptr_table, &proto_perl->Isv_yes, &PL_sv_yes);
11841 /* create (a non-shared!) shared string table */
11842 PL_strtab = newHV();
11843 HvSHAREKEYS_off(PL_strtab);
11844 hv_ksplit(PL_strtab, HvTOTALKEYS(proto_perl->Istrtab));
11845 ptr_table_store(PL_ptr_table, proto_perl->Istrtab, PL_strtab);
11847 PL_compiling = proto_perl->Icompiling;
11849 /* These two PVs will be free'd special way so must set them same way op.c does */
11850 PL_compiling.cop_stashpv = savesharedpv(PL_compiling.cop_stashpv);
11851 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_stashpv, PL_compiling.cop_stashpv);
11853 PL_compiling.cop_file = savesharedpv(PL_compiling.cop_file);
11854 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_file, PL_compiling.cop_file);
11856 ptr_table_store(PL_ptr_table, &proto_perl->Icompiling, &PL_compiling);
11857 PL_compiling.cop_warnings = DUP_WARNINGS(PL_compiling.cop_warnings);
11858 if (PL_compiling.cop_hints_hash) {
11860 PL_compiling.cop_hints_hash->refcounted_he_refcnt++;
11861 HINTS_REFCNT_UNLOCK;
11863 PL_curcop = (COP*)any_dup(proto_perl->Icurcop, proto_perl);
11864 #ifdef PERL_DEBUG_READONLY_OPS
11869 /* pseudo environmental stuff */
11870 PL_origargc = proto_perl->Iorigargc;
11871 PL_origargv = proto_perl->Iorigargv;
11873 param->stashes = newAV(); /* Setup array of objects to call clone on */
11875 /* Set tainting stuff before PerlIO_debug can possibly get called */
11876 PL_tainting = proto_perl->Itainting;
11877 PL_taint_warn = proto_perl->Itaint_warn;
11879 #ifdef PERLIO_LAYERS
11880 /* Clone PerlIO tables as soon as we can handle general xx_dup() */
11881 PerlIO_clone(aTHX_ proto_perl, param);
11884 PL_envgv = gv_dup(proto_perl->Ienvgv, param);
11885 PL_incgv = gv_dup(proto_perl->Iincgv, param);
11886 PL_hintgv = gv_dup(proto_perl->Ihintgv, param);
11887 PL_origfilename = SAVEPV(proto_perl->Iorigfilename);
11888 PL_diehook = sv_dup_inc(proto_perl->Idiehook, param);
11889 PL_warnhook = sv_dup_inc(proto_perl->Iwarnhook, param);
11892 PL_minus_c = proto_perl->Iminus_c;
11893 PL_patchlevel = sv_dup_inc(proto_perl->Ipatchlevel, param);
11894 PL_localpatches = proto_perl->Ilocalpatches;
11895 PL_splitstr = proto_perl->Isplitstr;
11896 PL_minus_n = proto_perl->Iminus_n;
11897 PL_minus_p = proto_perl->Iminus_p;
11898 PL_minus_l = proto_perl->Iminus_l;
11899 PL_minus_a = proto_perl->Iminus_a;
11900 PL_minus_E = proto_perl->Iminus_E;
11901 PL_minus_F = proto_perl->Iminus_F;
11902 PL_doswitches = proto_perl->Idoswitches;
11903 PL_dowarn = proto_perl->Idowarn;
11904 PL_doextract = proto_perl->Idoextract;
11905 PL_sawampersand = proto_perl->Isawampersand;
11906 PL_unsafe = proto_perl->Iunsafe;
11907 PL_inplace = SAVEPV(proto_perl->Iinplace);
11908 PL_e_script = sv_dup_inc(proto_perl->Ie_script, param);
11909 PL_perldb = proto_perl->Iperldb;
11910 PL_perl_destruct_level = proto_perl->Iperl_destruct_level;
11911 PL_exit_flags = proto_perl->Iexit_flags;
11913 /* magical thingies */
11914 /* XXX time(&PL_basetime) when asked for? */
11915 PL_basetime = proto_perl->Ibasetime;
11916 PL_formfeed = sv_dup(proto_perl->Iformfeed, param);
11918 PL_maxsysfd = proto_perl->Imaxsysfd;
11919 PL_statusvalue = proto_perl->Istatusvalue;
11921 PL_statusvalue_vms = proto_perl->Istatusvalue_vms;
11923 PL_statusvalue_posix = proto_perl->Istatusvalue_posix;
11925 PL_encoding = sv_dup(proto_perl->Iencoding, param);
11927 sv_setpvs(PERL_DEBUG_PAD(0), ""); /* For regex debugging. */
11928 sv_setpvs(PERL_DEBUG_PAD(1), ""); /* ext/re needs these */
11929 sv_setpvs(PERL_DEBUG_PAD(2), ""); /* even without DEBUGGING. */
11932 /* RE engine related */
11933 Zero(&PL_reg_state, 1, struct re_save_state);
11934 PL_reginterp_cnt = 0;
11935 PL_regmatch_slab = NULL;
11937 /* Clone the regex array */
11938 /* ORANGE FIXME for plugins, probably in the SV dup code.
11939 newSViv(PTR2IV(CALLREGDUPE(
11940 INT2PTR(REGEXP *, SvIVX(regex)), param))))
11942 PL_regex_padav = av_dup_inc(proto_perl->Iregex_padav, param);
11943 PL_regex_pad = AvARRAY(PL_regex_padav);
11945 /* shortcuts to various I/O objects */
11946 PL_ofsgv = gv_dup(proto_perl->Iofsgv, param);
11947 PL_stdingv = gv_dup(proto_perl->Istdingv, param);
11948 PL_stderrgv = gv_dup(proto_perl->Istderrgv, param);
11949 PL_defgv = gv_dup(proto_perl->Idefgv, param);
11950 PL_argvgv = gv_dup(proto_perl->Iargvgv, param);
11951 PL_argvoutgv = gv_dup(proto_perl->Iargvoutgv, param);
11952 PL_argvout_stack = av_dup_inc(proto_perl->Iargvout_stack, param);
11954 /* shortcuts to regexp stuff */
11955 PL_replgv = gv_dup(proto_perl->Ireplgv, param);
11957 /* shortcuts to misc objects */
11958 PL_errgv = gv_dup(proto_perl->Ierrgv, param);
11960 /* shortcuts to debugging objects */
11961 PL_DBgv = gv_dup(proto_perl->IDBgv, param);
11962 PL_DBline = gv_dup(proto_perl->IDBline, param);
11963 PL_DBsub = gv_dup(proto_perl->IDBsub, param);
11964 PL_DBsingle = sv_dup(proto_perl->IDBsingle, param);
11965 PL_DBtrace = sv_dup(proto_perl->IDBtrace, param);
11966 PL_DBsignal = sv_dup(proto_perl->IDBsignal, param);
11967 PL_dbargs = av_dup(proto_perl->Idbargs, param);
11969 /* symbol tables */
11970 PL_defstash = hv_dup_inc(proto_perl->Idefstash, param);
11971 PL_curstash = hv_dup(proto_perl->Icurstash, param);
11972 PL_debstash = hv_dup(proto_perl->Idebstash, param);
11973 PL_globalstash = hv_dup(proto_perl->Iglobalstash, param);
11974 PL_curstname = sv_dup_inc(proto_perl->Icurstname, param);
11976 PL_beginav = av_dup_inc(proto_perl->Ibeginav, param);
11977 PL_beginav_save = av_dup_inc(proto_perl->Ibeginav_save, param);
11978 PL_checkav_save = av_dup_inc(proto_perl->Icheckav_save, param);
11979 PL_unitcheckav = av_dup_inc(proto_perl->Iunitcheckav, param);
11980 PL_unitcheckav_save = av_dup_inc(proto_perl->Iunitcheckav_save, param);
11981 PL_endav = av_dup_inc(proto_perl->Iendav, param);
11982 PL_checkav = av_dup_inc(proto_perl->Icheckav, param);
11983 PL_initav = av_dup_inc(proto_perl->Iinitav, param);
11985 PL_sub_generation = proto_perl->Isub_generation;
11986 PL_isarev = hv_dup_inc(proto_perl->Iisarev, param);
11988 /* funky return mechanisms */
11989 PL_forkprocess = proto_perl->Iforkprocess;
11991 /* subprocess state */
11992 PL_fdpid = av_dup_inc(proto_perl->Ifdpid, param);
11994 /* internal state */
11995 PL_maxo = proto_perl->Imaxo;
11996 if (proto_perl->Iop_mask)
11997 PL_op_mask = SAVEPVN(proto_perl->Iop_mask, PL_maxo);
12000 /* PL_asserting = proto_perl->Iasserting; */
12002 /* current interpreter roots */
12003 PL_main_cv = cv_dup_inc(proto_perl->Imain_cv, param);
12005 PL_main_root = OpREFCNT_inc(proto_perl->Imain_root);
12007 PL_main_start = proto_perl->Imain_start;
12008 PL_eval_root = proto_perl->Ieval_root;
12009 PL_eval_start = proto_perl->Ieval_start;
12011 /* runtime control stuff */
12012 PL_curcopdb = (COP*)any_dup(proto_perl->Icurcopdb, proto_perl);
12014 PL_filemode = proto_perl->Ifilemode;
12015 PL_lastfd = proto_perl->Ilastfd;
12016 PL_oldname = proto_perl->Ioldname; /* XXX not quite right */
12019 PL_gensym = proto_perl->Igensym;
12020 PL_preambleav = av_dup_inc(proto_perl->Ipreambleav, param);
12021 PL_laststatval = proto_perl->Ilaststatval;
12022 PL_laststype = proto_perl->Ilaststype;
12025 PL_ors_sv = sv_dup_inc(proto_perl->Iors_sv, param);
12027 /* interpreter atexit processing */
12028 PL_exitlistlen = proto_perl->Iexitlistlen;
12029 if (PL_exitlistlen) {
12030 Newx(PL_exitlist, PL_exitlistlen, PerlExitListEntry);
12031 Copy(proto_perl->Iexitlist, PL_exitlist, PL_exitlistlen, PerlExitListEntry);
12034 PL_exitlist = (PerlExitListEntry*)NULL;
12036 PL_my_cxt_size = proto_perl->Imy_cxt_size;
12037 if (PL_my_cxt_size) {
12038 Newx(PL_my_cxt_list, PL_my_cxt_size, void *);
12039 Copy(proto_perl->Imy_cxt_list, PL_my_cxt_list, PL_my_cxt_size, void *);
12040 #ifdef PERL_GLOBAL_STRUCT_PRIVATE
12041 Newx(PL_my_cxt_keys, PL_my_cxt_size, const char *);
12042 Copy(proto_perl->Imy_cxt_keys, PL_my_cxt_keys, PL_my_cxt_size, char *);
12046 PL_my_cxt_list = (void**)NULL;
12047 #ifdef PERL_GLOBAL_STRUCT_PRIVATE
12048 PL_my_cxt_keys = (const char**)NULL;
12051 PL_modglobal = hv_dup_inc(proto_perl->Imodglobal, param);
12052 PL_custom_op_names = hv_dup_inc(proto_perl->Icustom_op_names,param);
12053 PL_custom_op_descs = hv_dup_inc(proto_perl->Icustom_op_descs,param);
12055 PL_profiledata = NULL;
12057 PL_compcv = cv_dup(proto_perl->Icompcv, param);
12059 PAD_CLONE_VARS(proto_perl, param);
12061 #ifdef HAVE_INTERP_INTERN
12062 sys_intern_dup(&proto_perl->Isys_intern, &PL_sys_intern);
12065 /* more statics moved here */
12066 PL_generation = proto_perl->Igeneration;
12067 PL_DBcv = cv_dup(proto_perl->IDBcv, param);
12069 PL_in_clean_objs = proto_perl->Iin_clean_objs;
12070 PL_in_clean_all = proto_perl->Iin_clean_all;
12072 PL_uid = proto_perl->Iuid;
12073 PL_euid = proto_perl->Ieuid;
12074 PL_gid = proto_perl->Igid;
12075 PL_egid = proto_perl->Iegid;
12076 PL_nomemok = proto_perl->Inomemok;
12077 PL_an = proto_perl->Ian;
12078 PL_evalseq = proto_perl->Ievalseq;
12079 PL_origenviron = proto_perl->Iorigenviron; /* XXX not quite right */
12080 PL_origalen = proto_perl->Iorigalen;
12081 #ifdef PERL_USES_PL_PIDSTATUS
12082 PL_pidstatus = newHV(); /* XXX flag for cloning? */
12084 PL_osname = SAVEPV(proto_perl->Iosname);
12085 PL_sighandlerp = proto_perl->Isighandlerp;
12087 PL_runops = proto_perl->Irunops;
12089 PL_parser = parser_dup(proto_perl->Iparser, param);
12091 PL_subline = proto_perl->Isubline;
12092 PL_subname = sv_dup_inc(proto_perl->Isubname, param);
12095 PL_cryptseen = proto_perl->Icryptseen;
12098 PL_hints = proto_perl->Ihints;
12100 PL_amagic_generation = proto_perl->Iamagic_generation;
12102 #ifdef USE_LOCALE_COLLATE
12103 PL_collation_ix = proto_perl->Icollation_ix;
12104 PL_collation_name = SAVEPV(proto_perl->Icollation_name);
12105 PL_collation_standard = proto_perl->Icollation_standard;
12106 PL_collxfrm_base = proto_perl->Icollxfrm_base;
12107 PL_collxfrm_mult = proto_perl->Icollxfrm_mult;
12108 #endif /* USE_LOCALE_COLLATE */
12110 #ifdef USE_LOCALE_NUMERIC
12111 PL_numeric_name = SAVEPV(proto_perl->Inumeric_name);
12112 PL_numeric_standard = proto_perl->Inumeric_standard;
12113 PL_numeric_local = proto_perl->Inumeric_local;
12114 PL_numeric_radix_sv = sv_dup_inc(proto_perl->Inumeric_radix_sv, param);
12115 #endif /* !USE_LOCALE_NUMERIC */
12117 /* utf8 character classes */
12118 PL_utf8_alnum = sv_dup_inc(proto_perl->Iutf8_alnum, param);
12119 PL_utf8_alnumc = sv_dup_inc(proto_perl->Iutf8_alnumc, param);
12120 PL_utf8_ascii = sv_dup_inc(proto_perl->Iutf8_ascii, param);
12121 PL_utf8_alpha = sv_dup_inc(proto_perl->Iutf8_alpha, param);
12122 PL_utf8_space = sv_dup_inc(proto_perl->Iutf8_space, param);
12123 PL_utf8_cntrl = sv_dup_inc(proto_perl->Iutf8_cntrl, param);
12124 PL_utf8_graph = sv_dup_inc(proto_perl->Iutf8_graph, param);
12125 PL_utf8_digit = sv_dup_inc(proto_perl->Iutf8_digit, param);
12126 PL_utf8_upper = sv_dup_inc(proto_perl->Iutf8_upper, param);
12127 PL_utf8_lower = sv_dup_inc(proto_perl->Iutf8_lower, param);
12128 PL_utf8_print = sv_dup_inc(proto_perl->Iutf8_print, param);
12129 PL_utf8_punct = sv_dup_inc(proto_perl->Iutf8_punct, param);
12130 PL_utf8_xdigit = sv_dup_inc(proto_perl->Iutf8_xdigit, param);
12131 PL_utf8_mark = sv_dup_inc(proto_perl->Iutf8_mark, param);
12132 PL_utf8_toupper = sv_dup_inc(proto_perl->Iutf8_toupper, param);
12133 PL_utf8_totitle = sv_dup_inc(proto_perl->Iutf8_totitle, param);
12134 PL_utf8_tolower = sv_dup_inc(proto_perl->Iutf8_tolower, param);
12135 PL_utf8_tofold = sv_dup_inc(proto_perl->Iutf8_tofold, param);
12136 PL_utf8_idstart = sv_dup_inc(proto_perl->Iutf8_idstart, param);
12137 PL_utf8_idcont = sv_dup_inc(proto_perl->Iutf8_idcont, param);
12139 /* Did the locale setup indicate UTF-8? */
12140 PL_utf8locale = proto_perl->Iutf8locale;
12141 /* Unicode features (see perlrun/-C) */
12142 PL_unicode = proto_perl->Iunicode;
12144 /* Pre-5.8 signals control */
12145 PL_signals = proto_perl->Isignals;
12147 /* times() ticks per second */
12148 PL_clocktick = proto_perl->Iclocktick;
12150 /* Recursion stopper for PerlIO_find_layer */
12151 PL_in_load_module = proto_perl->Iin_load_module;
12153 /* sort() routine */
12154 PL_sort_RealCmp = proto_perl->Isort_RealCmp;
12156 /* Not really needed/useful since the reenrant_retint is "volatile",
12157 * but do it for consistency's sake. */
12158 PL_reentrant_retint = proto_perl->Ireentrant_retint;
12160 /* Hooks to shared SVs and locks. */
12161 PL_sharehook = proto_perl->Isharehook;
12162 PL_lockhook = proto_perl->Ilockhook;
12163 PL_unlockhook = proto_perl->Iunlockhook;
12164 PL_threadhook = proto_perl->Ithreadhook;
12165 PL_destroyhook = proto_perl->Idestroyhook;
12167 #ifdef THREADS_HAVE_PIDS
12168 PL_ppid = proto_perl->Ippid;
12172 PL_last_swash_hv = NULL; /* reinits on demand */
12173 PL_last_swash_klen = 0;
12174 PL_last_swash_key[0]= '\0';
12175 PL_last_swash_tmps = (U8*)NULL;
12176 PL_last_swash_slen = 0;
12178 PL_glob_index = proto_perl->Iglob_index;
12179 PL_srand_called = proto_perl->Isrand_called;
12181 if (proto_perl->Ipsig_pend) {
12182 Newxz(PL_psig_pend, SIG_SIZE, int);
12185 PL_psig_pend = (int*)NULL;
12188 if (proto_perl->Ipsig_name) {
12189 Newx(PL_psig_name, 2 * SIG_SIZE, SV*);
12190 sv_dup_inc_multiple(proto_perl->Ipsig_name, PL_psig_name, 2 * SIG_SIZE,
12192 PL_psig_ptr = PL_psig_name + SIG_SIZE;
12195 PL_psig_ptr = (SV**)NULL;
12196 PL_psig_name = (SV**)NULL;
12199 /* intrpvar.h stuff */
12201 if (flags & CLONEf_COPY_STACKS) {
12202 /* next allocation will be PL_tmps_stack[PL_tmps_ix+1] */
12203 PL_tmps_ix = proto_perl->Itmps_ix;
12204 PL_tmps_max = proto_perl->Itmps_max;
12205 PL_tmps_floor = proto_perl->Itmps_floor;
12206 Newx(PL_tmps_stack, PL_tmps_max, SV*);
12207 sv_dup_inc_multiple(proto_perl->Itmps_stack, PL_tmps_stack, PL_tmps_ix,
12210 /* next PUSHMARK() sets *(PL_markstack_ptr+1) */
12211 i = proto_perl->Imarkstack_max - proto_perl->Imarkstack;
12212 Newxz(PL_markstack, i, I32);
12213 PL_markstack_max = PL_markstack + (proto_perl->Imarkstack_max
12214 - proto_perl->Imarkstack);
12215 PL_markstack_ptr = PL_markstack + (proto_perl->Imarkstack_ptr
12216 - proto_perl->Imarkstack);
12217 Copy(proto_perl->Imarkstack, PL_markstack,
12218 PL_markstack_ptr - PL_markstack + 1, I32);
12220 /* next push_scope()/ENTER sets PL_scopestack[PL_scopestack_ix]
12221 * NOTE: unlike the others! */
12222 PL_scopestack_ix = proto_perl->Iscopestack_ix;
12223 PL_scopestack_max = proto_perl->Iscopestack_max;
12224 Newxz(PL_scopestack, PL_scopestack_max, I32);
12225 Copy(proto_perl->Iscopestack, PL_scopestack, PL_scopestack_ix, I32);
12227 /* NOTE: si_dup() looks at PL_markstack */
12228 PL_curstackinfo = si_dup(proto_perl->Icurstackinfo, param);
12230 /* PL_curstack = PL_curstackinfo->si_stack; */
12231 PL_curstack = av_dup(proto_perl->Icurstack, param);
12232 PL_mainstack = av_dup(proto_perl->Imainstack, param);
12234 /* next PUSHs() etc. set *(PL_stack_sp+1) */
12235 PL_stack_base = AvARRAY(PL_curstack);
12236 PL_stack_sp = PL_stack_base + (proto_perl->Istack_sp
12237 - proto_perl->Istack_base);
12238 PL_stack_max = PL_stack_base + AvMAX(PL_curstack);
12240 /* next SSPUSHFOO() sets PL_savestack[PL_savestack_ix]
12241 * NOTE: unlike the others! */
12242 PL_savestack_ix = proto_perl->Isavestack_ix;
12243 PL_savestack_max = proto_perl->Isavestack_max;
12244 /*Newxz(PL_savestack, PL_savestack_max, ANY);*/
12245 PL_savestack = ss_dup(proto_perl, param);
12249 ENTER; /* perl_destruct() wants to LEAVE; */
12251 /* although we're not duplicating the tmps stack, we should still
12252 * add entries for any SVs on the tmps stack that got cloned by a
12253 * non-refcount means (eg a temp in @_); otherwise they will be
12256 for (i = 0; i<= proto_perl->Itmps_ix; i++) {
12257 SV * const nsv = MUTABLE_SV(ptr_table_fetch(PL_ptr_table,
12258 proto_perl->Itmps_stack[i]));
12259 if (nsv && !SvREFCNT(nsv)) {
12261 PL_tmps_stack[++PL_tmps_ix] = SvREFCNT_inc_simple(nsv);
12266 PL_start_env = proto_perl->Istart_env; /* XXXXXX */
12267 PL_top_env = &PL_start_env;
12269 PL_op = proto_perl->Iop;
12272 PL_Xpv = (XPV*)NULL;
12273 my_perl->Ina = proto_perl->Ina;
12275 PL_statbuf = proto_perl->Istatbuf;
12276 PL_statcache = proto_perl->Istatcache;
12277 PL_statgv = gv_dup(proto_perl->Istatgv, param);
12278 PL_statname = sv_dup_inc(proto_perl->Istatname, param);
12280 PL_timesbuf = proto_perl->Itimesbuf;
12283 PL_tainted = proto_perl->Itainted;
12284 PL_curpm = proto_perl->Icurpm; /* XXX No PMOP ref count */
12285 PL_rs = sv_dup_inc(proto_perl->Irs, param);
12286 PL_last_in_gv = gv_dup(proto_perl->Ilast_in_gv, param);
12287 PL_defoutgv = gv_dup_inc(proto_perl->Idefoutgv, param);
12288 PL_chopset = proto_perl->Ichopset; /* XXX never deallocated */
12289 PL_toptarget = sv_dup_inc(proto_perl->Itoptarget, param);
12290 PL_bodytarget = sv_dup_inc(proto_perl->Ibodytarget, param);
12291 PL_formtarget = sv_dup(proto_perl->Iformtarget, param);
12293 PL_restartop = proto_perl->Irestartop;
12294 PL_in_eval = proto_perl->Iin_eval;
12295 PL_delaymagic = proto_perl->Idelaymagic;
12296 PL_dirty = proto_perl->Idirty;
12297 PL_localizing = proto_perl->Ilocalizing;
12299 PL_errors = sv_dup_inc(proto_perl->Ierrors, param);
12300 PL_hv_fetch_ent_mh = NULL;
12301 PL_modcount = proto_perl->Imodcount;
12302 PL_lastgotoprobe = NULL;
12303 PL_dumpindent = proto_perl->Idumpindent;
12305 PL_sortcop = (OP*)any_dup(proto_perl->Isortcop, proto_perl);
12306 PL_sortstash = hv_dup(proto_perl->Isortstash, param);
12307 PL_firstgv = gv_dup(proto_perl->Ifirstgv, param);
12308 PL_secondgv = gv_dup(proto_perl->Isecondgv, param);
12309 PL_efloatbuf = NULL; /* reinits on demand */
12310 PL_efloatsize = 0; /* reinits on demand */
12314 PL_screamfirst = NULL;
12315 PL_screamnext = NULL;
12316 PL_maxscream = -1; /* reinits on demand */
12317 PL_lastscream = NULL;
12320 PL_regdummy = proto_perl->Iregdummy;
12321 PL_colorset = 0; /* reinits PL_colors[] */
12322 /*PL_colors[6] = {0,0,0,0,0,0};*/
12326 /* Pluggable optimizer */
12327 PL_peepp = proto_perl->Ipeepp;
12329 PL_stashcache = newHV();
12331 PL_watchaddr = (char **) ptr_table_fetch(PL_ptr_table,
12332 proto_perl->Iwatchaddr);
12333 PL_watchok = PL_watchaddr ? * PL_watchaddr : NULL;
12334 if (PL_debug && PL_watchaddr) {
12335 PerlIO_printf(Perl_debug_log,
12336 "WATCHING: %"UVxf" cloned as %"UVxf" with value %"UVxf"\n",
12337 PTR2UV(proto_perl->Iwatchaddr), PTR2UV(PL_watchaddr),
12338 PTR2UV(PL_watchok));
12341 PL_registered_mros = hv_dup_inc(proto_perl->Iregistered_mros, param);
12343 if (!(flags & CLONEf_KEEP_PTR_TABLE)) {
12344 ptr_table_free(PL_ptr_table);
12345 PL_ptr_table = NULL;
12348 /* Call the ->CLONE method, if it exists, for each of the stashes
12349 identified by sv_dup() above.
12351 while(av_len(param->stashes) != -1) {
12352 HV* const stash = MUTABLE_HV(av_shift(param->stashes));
12353 GV* const cloner = gv_fetchmethod_autoload(stash, "CLONE", 0);
12354 if (cloner && GvCV(cloner)) {
12359 mXPUSHs(newSVhek(HvNAME_HEK(stash)));
12361 call_sv(MUTABLE_SV(GvCV(cloner)), G_DISCARD);
12367 SvREFCNT_dec(param->stashes);
12369 /* orphaned? eg threads->new inside BEGIN or use */
12370 if (PL_compcv && ! SvREFCNT(PL_compcv)) {
12371 SvREFCNT_inc_simple_void(PL_compcv);
12372 SAVEFREESV(PL_compcv);
12378 #endif /* USE_ITHREADS */
12381 =head1 Unicode Support
12383 =for apidoc sv_recode_to_utf8
12385 The encoding is assumed to be an Encode object, on entry the PV
12386 of the sv is assumed to be octets in that encoding, and the sv
12387 will be converted into Unicode (and UTF-8).
12389 If the sv already is UTF-8 (or if it is not POK), or if the encoding
12390 is not a reference, nothing is done to the sv. If the encoding is not
12391 an C<Encode::XS> Encoding object, bad things will happen.
12392 (See F<lib/encoding.pm> and L<Encode>).
12394 The PV of the sv is returned.
12399 Perl_sv_recode_to_utf8(pTHX_ SV *sv, SV *encoding)
12403 PERL_ARGS_ASSERT_SV_RECODE_TO_UTF8;
12405 if (SvPOK(sv) && !SvUTF8(sv) && !IN_BYTES && SvROK(encoding)) {
12419 Passing sv_yes is wrong - it needs to be or'ed set of constants
12420 for Encode::XS, while UTf-8 decode (currently) assumes a true value means
12421 remove converted chars from source.
12423 Both will default the value - let them.
12425 XPUSHs(&PL_sv_yes);
12428 call_method("decode", G_SCALAR);
12432 s = SvPV_const(uni, len);
12433 if (s != SvPVX_const(sv)) {
12434 SvGROW(sv, len + 1);
12435 Move(s, SvPVX(sv), len + 1, char);
12436 SvCUR_set(sv, len);
12443 return SvPOKp(sv) ? SvPVX(sv) : NULL;
12447 =for apidoc sv_cat_decode
12449 The encoding is assumed to be an Encode object, the PV of the ssv is
12450 assumed to be octets in that encoding and decoding the input starts
12451 from the position which (PV + *offset) pointed to. The dsv will be
12452 concatenated the decoded UTF-8 string from ssv. Decoding will terminate
12453 when the string tstr appears in decoding output or the input ends on
12454 the PV of the ssv. The value which the offset points will be modified
12455 to the last input position on the ssv.
12457 Returns TRUE if the terminator was found, else returns FALSE.
12462 Perl_sv_cat_decode(pTHX_ SV *dsv, SV *encoding,
12463 SV *ssv, int *offset, char *tstr, int tlen)
12468 PERL_ARGS_ASSERT_SV_CAT_DECODE;
12470 if (SvPOK(ssv) && SvPOK(dsv) && SvROK(encoding) && offset) {
12481 offsv = newSViv(*offset);
12483 mXPUSHp(tstr, tlen);
12485 call_method("cat_decode", G_SCALAR);
12487 ret = SvTRUE(TOPs);
12488 *offset = SvIV(offsv);
12494 Perl_croak(aTHX_ "Invalid argument to sv_cat_decode");
12499 /* ---------------------------------------------------------------------
12501 * support functions for report_uninit()
12504 /* the maxiumum size of array or hash where we will scan looking
12505 * for the undefined element that triggered the warning */
12507 #define FUV_MAX_SEARCH_SIZE 1000
12509 /* Look for an entry in the hash whose value has the same SV as val;
12510 * If so, return a mortal copy of the key. */
12513 S_find_hash_subscript(pTHX_ const HV *const hv, const SV *const val)
12516 register HE **array;
12519 PERL_ARGS_ASSERT_FIND_HASH_SUBSCRIPT;
12521 if (!hv || SvMAGICAL(hv) || !HvARRAY(hv) ||
12522 (HvTOTALKEYS(hv) > FUV_MAX_SEARCH_SIZE))
12525 array = HvARRAY(hv);
12527 for (i=HvMAX(hv); i>0; i--) {
12528 register HE *entry;
12529 for (entry = array[i]; entry; entry = HeNEXT(entry)) {
12530 if (HeVAL(entry) != val)
12532 if ( HeVAL(entry) == &PL_sv_undef ||
12533 HeVAL(entry) == &PL_sv_placeholder)
12537 if (HeKLEN(entry) == HEf_SVKEY)
12538 return sv_mortalcopy(HeKEY_sv(entry));
12539 return sv_2mortal(newSVhek(HeKEY_hek(entry)));
12545 /* Look for an entry in the array whose value has the same SV as val;
12546 * If so, return the index, otherwise return -1. */
12549 S_find_array_subscript(pTHX_ const AV *const av, const SV *const val)
12553 PERL_ARGS_ASSERT_FIND_ARRAY_SUBSCRIPT;
12555 if (!av || SvMAGICAL(av) || !AvARRAY(av) ||
12556 (AvFILLp(av) > FUV_MAX_SEARCH_SIZE))
12559 if (val != &PL_sv_undef) {
12560 SV ** const svp = AvARRAY(av);
12563 for (i=AvFILLp(av); i>=0; i--)
12570 /* S_varname(): return the name of a variable, optionally with a subscript.
12571 * If gv is non-zero, use the name of that global, along with gvtype (one
12572 * of "$", "@", "%"); otherwise use the name of the lexical at pad offset
12573 * targ. Depending on the value of the subscript_type flag, return:
12576 #define FUV_SUBSCRIPT_NONE 1 /* "@foo" */
12577 #define FUV_SUBSCRIPT_ARRAY 2 /* "$foo[aindex]" */
12578 #define FUV_SUBSCRIPT_HASH 3 /* "$foo{keyname}" */
12579 #define FUV_SUBSCRIPT_WITHIN 4 /* "within @foo" */
12582 S_varname(pTHX_ const GV *const gv, const char gvtype, PADOFFSET targ,
12583 const SV *const keyname, I32 aindex, int subscript_type)
12586 SV * const name = sv_newmortal();
12589 buffer[0] = gvtype;
12592 /* as gv_fullname4(), but add literal '^' for $^FOO names */
12594 gv_fullname4(name, gv, buffer, 0);
12596 if ((unsigned int)SvPVX(name)[1] <= 26) {
12598 buffer[1] = SvPVX(name)[1] + 'A' - 1;
12600 /* Swap the 1 unprintable control character for the 2 byte pretty
12601 version - ie substr($name, 1, 1) = $buffer; */
12602 sv_insert(name, 1, 1, buffer, 2);
12606 CV * const cv = find_runcv(NULL);
12610 if (!cv || !CvPADLIST(cv))
12612 av = MUTABLE_AV((*av_fetch(CvPADLIST(cv), 0, FALSE)));
12613 sv = *av_fetch(av, targ, FALSE);
12614 sv_setpvn(name, SvPV_nolen_const(sv), SvCUR(sv));
12617 if (subscript_type == FUV_SUBSCRIPT_HASH) {
12618 SV * const sv = newSV(0);
12619 *SvPVX(name) = '$';
12620 Perl_sv_catpvf(aTHX_ name, "{%s}",
12621 pv_display(sv,SvPVX_const(keyname), SvCUR(keyname), 0, 32));
12624 else if (subscript_type == FUV_SUBSCRIPT_ARRAY) {
12625 *SvPVX(name) = '$';
12626 Perl_sv_catpvf(aTHX_ name, "[%"IVdf"]", (IV)aindex);
12628 else if (subscript_type == FUV_SUBSCRIPT_WITHIN) {
12629 /* We know that name has no magic, so can use 0 instead of SV_GMAGIC */
12630 Perl_sv_insert_flags(aTHX_ name, 0, 0, STR_WITH_LEN("within "), 0);
12638 =for apidoc find_uninit_var
12640 Find the name of the undefined variable (if any) that caused the operator o
12641 to issue a "Use of uninitialized value" warning.
12642 If match is true, only return a name if it's value matches uninit_sv.
12643 So roughly speaking, if a unary operator (such as OP_COS) generates a
12644 warning, then following the direct child of the op may yield an
12645 OP_PADSV or OP_GV that gives the name of the undefined variable. On the
12646 other hand, with OP_ADD there are two branches to follow, so we only print
12647 the variable name if we get an exact match.
12649 The name is returned as a mortal SV.
12651 Assumes that PL_op is the op that originally triggered the error, and that
12652 PL_comppad/PL_curpad points to the currently executing pad.
12658 S_find_uninit_var(pTHX_ const OP *const obase, const SV *const uninit_sv,
12664 const OP *o, *o2, *kid;
12666 if (!obase || (match && (!uninit_sv || uninit_sv == &PL_sv_undef ||
12667 uninit_sv == &PL_sv_placeholder)))
12670 switch (obase->op_type) {
12677 const bool pad = (obase->op_type == OP_PADAV || obase->op_type == OP_PADHV);
12678 const bool hash = (obase->op_type == OP_PADHV || obase->op_type == OP_RV2HV);
12681 int subscript_type = FUV_SUBSCRIPT_WITHIN;
12683 if (pad) { /* @lex, %lex */
12684 sv = PAD_SVl(obase->op_targ);
12688 if (cUNOPx(obase)->op_first->op_type == OP_GV) {
12689 /* @global, %global */
12690 gv = cGVOPx_gv(cUNOPx(obase)->op_first);
12693 sv = hash ? MUTABLE_SV(GvHV(gv)): MUTABLE_SV(GvAV(gv));
12695 else /* @{expr}, %{expr} */
12696 return find_uninit_var(cUNOPx(obase)->op_first,
12700 /* attempt to find a match within the aggregate */
12702 keysv = find_hash_subscript((const HV*)sv, uninit_sv);
12704 subscript_type = FUV_SUBSCRIPT_HASH;
12707 index = find_array_subscript((const AV *)sv, uninit_sv);
12709 subscript_type = FUV_SUBSCRIPT_ARRAY;
12712 if (match && subscript_type == FUV_SUBSCRIPT_WITHIN)
12715 return varname(gv, hash ? '%' : '@', obase->op_targ,
12716 keysv, index, subscript_type);
12720 if (match && PAD_SVl(obase->op_targ) != uninit_sv)
12722 return varname(NULL, '$', obase->op_targ,
12723 NULL, 0, FUV_SUBSCRIPT_NONE);
12726 gv = cGVOPx_gv(obase);
12727 if (!gv || (match && GvSV(gv) != uninit_sv))
12729 return varname(gv, '$', 0, NULL, 0, FUV_SUBSCRIPT_NONE);
12732 if (obase->op_flags & OPf_SPECIAL) { /* lexical array */
12735 AV *av = MUTABLE_AV(PAD_SV(obase->op_targ));
12736 if (!av || SvRMAGICAL(av))
12738 svp = av_fetch(av, (I32)obase->op_private, FALSE);
12739 if (!svp || *svp != uninit_sv)
12742 return varname(NULL, '$', obase->op_targ,
12743 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
12746 gv = cGVOPx_gv(obase);
12751 AV *const av = GvAV(gv);
12752 if (!av || SvRMAGICAL(av))
12754 svp = av_fetch(av, (I32)obase->op_private, FALSE);
12755 if (!svp || *svp != uninit_sv)
12758 return varname(gv, '$', 0,
12759 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
12764 o = cUNOPx(obase)->op_first;
12765 if (!o || o->op_type != OP_NULL ||
12766 ! (o->op_targ == OP_AELEM || o->op_targ == OP_HELEM))
12768 return find_uninit_var(cBINOPo->op_last, uninit_sv, match);
12772 if (PL_op == obase)
12773 /* $a[uninit_expr] or $h{uninit_expr} */
12774 return find_uninit_var(cBINOPx(obase)->op_last, uninit_sv, match);
12777 o = cBINOPx(obase)->op_first;
12778 kid = cBINOPx(obase)->op_last;
12780 /* get the av or hv, and optionally the gv */
12782 if (o->op_type == OP_PADAV || o->op_type == OP_PADHV) {
12783 sv = PAD_SV(o->op_targ);
12785 else if ((o->op_type == OP_RV2AV || o->op_type == OP_RV2HV)
12786 && cUNOPo->op_first->op_type == OP_GV)
12788 gv = cGVOPx_gv(cUNOPo->op_first);
12792 == OP_RV2HV ? MUTABLE_SV(GvHV(gv)) : MUTABLE_SV(GvAV(gv));
12797 if (kid && kid->op_type == OP_CONST && SvOK(cSVOPx_sv(kid))) {
12798 /* index is constant */
12802 if (obase->op_type == OP_HELEM) {
12803 HE* he = hv_fetch_ent(MUTABLE_HV(sv), cSVOPx_sv(kid), 0, 0);
12804 if (!he || HeVAL(he) != uninit_sv)
12808 SV * const * const svp = av_fetch(MUTABLE_AV(sv), SvIV(cSVOPx_sv(kid)), FALSE);
12809 if (!svp || *svp != uninit_sv)
12813 if (obase->op_type == OP_HELEM)
12814 return varname(gv, '%', o->op_targ,
12815 cSVOPx_sv(kid), 0, FUV_SUBSCRIPT_HASH);
12817 return varname(gv, '@', o->op_targ, NULL,
12818 SvIV(cSVOPx_sv(kid)), FUV_SUBSCRIPT_ARRAY);
12821 /* index is an expression;
12822 * attempt to find a match within the aggregate */
12823 if (obase->op_type == OP_HELEM) {
12824 SV * const keysv = find_hash_subscript((const HV*)sv, uninit_sv);
12826 return varname(gv, '%', o->op_targ,
12827 keysv, 0, FUV_SUBSCRIPT_HASH);
12831 = find_array_subscript((const AV *)sv, uninit_sv);
12833 return varname(gv, '@', o->op_targ,
12834 NULL, index, FUV_SUBSCRIPT_ARRAY);
12839 (o->op_type == OP_PADAV || o->op_type == OP_RV2AV)
12841 o->op_targ, NULL, 0, FUV_SUBSCRIPT_WITHIN);
12846 /* only examine RHS */
12847 return find_uninit_var(cBINOPx(obase)->op_first, uninit_sv, match);
12850 o = cUNOPx(obase)->op_first;
12851 if (o->op_type == OP_PUSHMARK)
12854 if (!o->op_sibling) {
12855 /* one-arg version of open is highly magical */
12857 if (o->op_type == OP_GV) { /* open FOO; */
12859 if (match && GvSV(gv) != uninit_sv)
12861 return varname(gv, '$', 0,
12862 NULL, 0, FUV_SUBSCRIPT_NONE);
12864 /* other possibilities not handled are:
12865 * open $x; or open my $x; should return '${*$x}'
12866 * open expr; should return '$'.expr ideally
12872 /* ops where $_ may be an implicit arg */
12876 if ( !(obase->op_flags & OPf_STACKED)) {
12877 if (uninit_sv == ((obase->op_private & OPpTARGET_MY)
12878 ? PAD_SVl(obase->op_targ)
12881 sv = sv_newmortal();
12882 sv_setpvs(sv, "$_");
12891 match = 1; /* print etc can return undef on defined args */
12892 /* skip filehandle as it can't produce 'undef' warning */
12893 o = cUNOPx(obase)->op_first;
12894 if ((obase->op_flags & OPf_STACKED) && o->op_type == OP_PUSHMARK)
12895 o = o->op_sibling->op_sibling;
12899 case OP_ENTEREVAL: /* could be eval $undef or $x='$undef'; eval $x */
12901 case OP_CUSTOM: /* XS or custom code could trigger random warnings */
12903 /* the following ops are capable of returning PL_sv_undef even for
12904 * defined arg(s) */
12923 case OP_GETPEERNAME:
12971 case OP_SMARTMATCH:
12980 /* XXX tmp hack: these two may call an XS sub, and currently
12981 XS subs don't have a SUB entry on the context stack, so CV and
12982 pad determination goes wrong, and BAD things happen. So, just
12983 don't try to determine the value under those circumstances.
12984 Need a better fix at dome point. DAPM 11/2007 */
12989 /* def-ness of rval pos() is independent of the def-ness of its arg */
12990 if ( !(obase->op_flags & OPf_MOD))
12995 if (SvROK(PL_rs) && uninit_sv == SvRV(PL_rs))
12996 return newSVpvs_flags("${$/}", SVs_TEMP);
13001 if (!(obase->op_flags & OPf_KIDS))
13003 o = cUNOPx(obase)->op_first;
13009 /* if all except one arg are constant, or have no side-effects,
13010 * or are optimized away, then it's unambiguous */
13012 for (kid=o; kid; kid = kid->op_sibling) {
13014 const OPCODE type = kid->op_type;
13015 if ( (type == OP_CONST && SvOK(cSVOPx_sv(kid)))
13016 || (type == OP_NULL && ! (kid->op_flags & OPf_KIDS))
13017 || (type == OP_PUSHMARK)
13021 if (o2) { /* more than one found */
13028 return find_uninit_var(o2, uninit_sv, match);
13030 /* scan all args */
13032 sv = find_uninit_var(o, uninit_sv, 1);
13044 =for apidoc report_uninit
13046 Print appropriate "Use of uninitialized variable" warning
13052 Perl_report_uninit(pTHX_ const SV *uninit_sv)
13056 SV* varname = NULL;
13058 varname = find_uninit_var(PL_op, uninit_sv,0);
13060 sv_insert(varname, 0, 0, " ", 1);
13062 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
13063 varname ? SvPV_nolen_const(varname) : "",
13064 " in ", OP_DESC(PL_op));
13067 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
13073 * c-indentation-style: bsd
13074 * c-basic-offset: 4
13075 * indent-tabs-mode: t
13078 * ex: set ts=8 sts=4 sw=4 noet: