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. */
1863 S_glob_2pv(pTHX_ GV * const gv, STRLEN * const len)
1865 const U32 wasfake = SvFLAGS(gv) & SVf_FAKE;
1866 SV *const buffer = sv_newmortal();
1868 PERL_ARGS_ASSERT_GLOB_2PV;
1870 /* FAKE globs can get coerced, so need to turn this off temporarily if it
1873 gv_efullname3(buffer, gv, "*");
1874 SvFLAGS(gv) |= wasfake;
1876 assert(SvPOK(buffer));
1878 *len = SvCUR(buffer);
1880 return SvPVX(buffer);
1883 /* Actually, ISO C leaves conversion of UV to IV undefined, but
1884 until proven guilty, assume that things are not that bad... */
1889 As 64 bit platforms often have an NV that doesn't preserve all bits of
1890 an IV (an assumption perl has been based on to date) it becomes necessary
1891 to remove the assumption that the NV always carries enough precision to
1892 recreate the IV whenever needed, and that the NV is the canonical form.
1893 Instead, IV/UV and NV need to be given equal rights. So as to not lose
1894 precision as a side effect of conversion (which would lead to insanity
1895 and the dragon(s) in t/op/numconvert.t getting very angry) the intent is
1896 1) to distinguish between IV/UV/NV slots that have cached a valid
1897 conversion where precision was lost and IV/UV/NV slots that have a
1898 valid conversion which has lost no precision
1899 2) to ensure that if a numeric conversion to one form is requested that
1900 would lose precision, the precise conversion (or differently
1901 imprecise conversion) is also performed and cached, to prevent
1902 requests for different numeric formats on the same SV causing
1903 lossy conversion chains. (lossless conversion chains are perfectly
1908 SvIOKp is true if the IV slot contains a valid value
1909 SvIOK is true only if the IV value is accurate (UV if SvIOK_UV true)
1910 SvNOKp is true if the NV slot contains a valid value
1911 SvNOK is true only if the NV value is accurate
1914 while converting from PV to NV, check to see if converting that NV to an
1915 IV(or UV) would lose accuracy over a direct conversion from PV to
1916 IV(or UV). If it would, cache both conversions, return NV, but mark
1917 SV as IOK NOKp (ie not NOK).
1919 While converting from PV to IV, check to see if converting that IV to an
1920 NV would lose accuracy over a direct conversion from PV to NV. If it
1921 would, cache both conversions, flag similarly.
1923 Before, the SV value "3.2" could become NV=3.2 IV=3 NOK, IOK quite
1924 correctly because if IV & NV were set NV *always* overruled.
1925 Now, "3.2" will become NV=3.2 IV=3 NOK, IOKp, because the flag's meaning
1926 changes - now IV and NV together means that the two are interchangeable:
1927 SvIVX == (IV) SvNVX && SvNVX == (NV) SvIVX;
1929 The benefit of this is that operations such as pp_add know that if
1930 SvIOK is true for both left and right operands, then integer addition
1931 can be used instead of floating point (for cases where the result won't
1932 overflow). Before, floating point was always used, which could lead to
1933 loss of precision compared with integer addition.
1935 * making IV and NV equal status should make maths accurate on 64 bit
1937 * may speed up maths somewhat if pp_add and friends start to use
1938 integers when possible instead of fp. (Hopefully the overhead in
1939 looking for SvIOK and checking for overflow will not outweigh the
1940 fp to integer speedup)
1941 * will slow down integer operations (callers of SvIV) on "inaccurate"
1942 values, as the change from SvIOK to SvIOKp will cause a call into
1943 sv_2iv each time rather than a macro access direct to the IV slot
1944 * should speed up number->string conversion on integers as IV is
1945 favoured when IV and NV are equally accurate
1947 ####################################################################
1948 You had better be using SvIOK_notUV if you want an IV for arithmetic:
1949 SvIOK is true if (IV or UV), so you might be getting (IV)SvUV.
1950 On the other hand, SvUOK is true iff UV.
1951 ####################################################################
1953 Your mileage will vary depending your CPU's relative fp to integer
1957 #ifndef NV_PRESERVES_UV
1958 # define IS_NUMBER_UNDERFLOW_IV 1
1959 # define IS_NUMBER_UNDERFLOW_UV 2
1960 # define IS_NUMBER_IV_AND_UV 2
1961 # define IS_NUMBER_OVERFLOW_IV 4
1962 # define IS_NUMBER_OVERFLOW_UV 5
1964 /* sv_2iuv_non_preserve(): private routine for use by sv_2iv() and sv_2uv() */
1966 /* For sv_2nv these three cases are "SvNOK and don't bother casting" */
1968 S_sv_2iuv_non_preserve(pTHX_ register SV *const sv
1976 PERL_ARGS_ASSERT_SV_2IUV_NON_PRESERVE;
1978 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));
1979 if (SvNVX(sv) < (NV)IV_MIN) {
1980 (void)SvIOKp_on(sv);
1982 SvIV_set(sv, IV_MIN);
1983 return IS_NUMBER_UNDERFLOW_IV;
1985 if (SvNVX(sv) > (NV)UV_MAX) {
1986 (void)SvIOKp_on(sv);
1989 SvUV_set(sv, UV_MAX);
1990 return IS_NUMBER_OVERFLOW_UV;
1992 (void)SvIOKp_on(sv);
1994 /* Can't use strtol etc to convert this string. (See truth table in
1996 if (SvNVX(sv) <= (UV)IV_MAX) {
1997 SvIV_set(sv, I_V(SvNVX(sv)));
1998 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
1999 SvIOK_on(sv); /* Integer is precise. NOK, IOK */
2001 /* Integer is imprecise. NOK, IOKp */
2003 return SvNVX(sv) < 0 ? IS_NUMBER_UNDERFLOW_UV : IS_NUMBER_IV_AND_UV;
2006 SvUV_set(sv, U_V(SvNVX(sv)));
2007 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
2008 if (SvUVX(sv) == UV_MAX) {
2009 /* As we know that NVs don't preserve UVs, UV_MAX cannot
2010 possibly be preserved by NV. Hence, it must be overflow.
2012 return IS_NUMBER_OVERFLOW_UV;
2014 SvIOK_on(sv); /* Integer is precise. NOK, UOK */
2016 /* Integer is imprecise. NOK, IOKp */
2018 return IS_NUMBER_OVERFLOW_IV;
2020 #endif /* !NV_PRESERVES_UV*/
2023 S_sv_2iuv_common(pTHX_ SV *const sv)
2027 PERL_ARGS_ASSERT_SV_2IUV_COMMON;
2030 /* erm. not sure. *should* never get NOKp (without NOK) from sv_2nv
2031 * without also getting a cached IV/UV from it at the same time
2032 * (ie PV->NV conversion should detect loss of accuracy and cache
2033 * IV or UV at same time to avoid this. */
2034 /* IV-over-UV optimisation - choose to cache IV if possible */
2036 if (SvTYPE(sv) == SVt_NV)
2037 sv_upgrade(sv, SVt_PVNV);
2039 (void)SvIOKp_on(sv); /* Must do this first, to clear any SvOOK */
2040 /* < not <= as for NV doesn't preserve UV, ((NV)IV_MAX+1) will almost
2041 certainly cast into the IV range at IV_MAX, whereas the correct
2042 answer is the UV IV_MAX +1. Hence < ensures that dodgy boundary
2044 #if defined(NAN_COMPARE_BROKEN) && defined(Perl_isnan)
2045 if (Perl_isnan(SvNVX(sv))) {
2051 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2052 SvIV_set(sv, I_V(SvNVX(sv)));
2053 if (SvNVX(sv) == (NV) SvIVX(sv)
2054 #ifndef NV_PRESERVES_UV
2055 && (((UV)1 << NV_PRESERVES_UV_BITS) >
2056 (UV)(SvIVX(sv) > 0 ? SvIVX(sv) : -SvIVX(sv)))
2057 /* Don't flag it as "accurately an integer" if the number
2058 came from a (by definition imprecise) NV operation, and
2059 we're outside the range of NV integer precision */
2063 SvIOK_on(sv); /* Can this go wrong with rounding? NWC */
2065 /* scalar has trailing garbage, eg "42a" */
2067 DEBUG_c(PerlIO_printf(Perl_debug_log,
2068 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (precise)\n",
2074 /* IV not precise. No need to convert from PV, as NV
2075 conversion would already have cached IV if it detected
2076 that PV->IV would be better than PV->NV->IV
2077 flags already correct - don't set public IOK. */
2078 DEBUG_c(PerlIO_printf(Perl_debug_log,
2079 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (imprecise)\n",
2084 /* Can the above go wrong if SvIVX == IV_MIN and SvNVX < IV_MIN,
2085 but the cast (NV)IV_MIN rounds to a the value less (more
2086 negative) than IV_MIN which happens to be equal to SvNVX ??
2087 Analogous to 0xFFFFFFFFFFFFFFFF rounding up to NV (2**64) and
2088 NV rounding back to 0xFFFFFFFFFFFFFFFF, so UVX == UV(NVX) and
2089 (NV)UVX == NVX are both true, but the values differ. :-(
2090 Hopefully for 2s complement IV_MIN is something like
2091 0x8000000000000000 which will be exact. NWC */
2094 SvUV_set(sv, U_V(SvNVX(sv)));
2096 (SvNVX(sv) == (NV) SvUVX(sv))
2097 #ifndef NV_PRESERVES_UV
2098 /* Make sure it's not 0xFFFFFFFFFFFFFFFF */
2099 /*&& (SvUVX(sv) != UV_MAX) irrelevant with code below */
2100 && (((UV)1 << NV_PRESERVES_UV_BITS) > SvUVX(sv))
2101 /* Don't flag it as "accurately an integer" if the number
2102 came from a (by definition imprecise) NV operation, and
2103 we're outside the range of NV integer precision */
2109 DEBUG_c(PerlIO_printf(Perl_debug_log,
2110 "0x%"UVxf" 2iv(%"UVuf" => %"IVdf") (as unsigned)\n",
2116 else if (SvPOKp(sv) && SvLEN(sv)) {
2118 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2119 /* We want to avoid a possible problem when we cache an IV/ a UV which
2120 may be later translated to an NV, and the resulting NV is not
2121 the same as the direct translation of the initial string
2122 (eg 123.456 can shortcut to the IV 123 with atol(), but we must
2123 be careful to ensure that the value with the .456 is around if the
2124 NV value is requested in the future).
2126 This means that if we cache such an IV/a UV, we need to cache the
2127 NV as well. Moreover, we trade speed for space, and do not
2128 cache the NV if we are sure it's not needed.
2131 /* SVt_PVNV is one higher than SVt_PVIV, hence this order */
2132 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2133 == IS_NUMBER_IN_UV) {
2134 /* It's definitely an integer, only upgrade to PVIV */
2135 if (SvTYPE(sv) < SVt_PVIV)
2136 sv_upgrade(sv, SVt_PVIV);
2138 } else if (SvTYPE(sv) < SVt_PVNV)
2139 sv_upgrade(sv, SVt_PVNV);
2141 /* If NVs preserve UVs then we only use the UV value if we know that
2142 we aren't going to call atof() below. If NVs don't preserve UVs
2143 then the value returned may have more precision than atof() will
2144 return, even though value isn't perfectly accurate. */
2145 if ((numtype & (IS_NUMBER_IN_UV
2146 #ifdef NV_PRESERVES_UV
2149 )) == IS_NUMBER_IN_UV) {
2150 /* This won't turn off the public IOK flag if it was set above */
2151 (void)SvIOKp_on(sv);
2153 if (!(numtype & IS_NUMBER_NEG)) {
2155 if (value <= (UV)IV_MAX) {
2156 SvIV_set(sv, (IV)value);
2158 /* it didn't overflow, and it was positive. */
2159 SvUV_set(sv, value);
2163 /* 2s complement assumption */
2164 if (value <= (UV)IV_MIN) {
2165 SvIV_set(sv, -(IV)value);
2167 /* Too negative for an IV. This is a double upgrade, but
2168 I'm assuming it will be rare. */
2169 if (SvTYPE(sv) < SVt_PVNV)
2170 sv_upgrade(sv, SVt_PVNV);
2174 SvNV_set(sv, -(NV)value);
2175 SvIV_set(sv, IV_MIN);
2179 /* For !NV_PRESERVES_UV and IS_NUMBER_IN_UV and IS_NUMBER_NOT_INT we
2180 will be in the previous block to set the IV slot, and the next
2181 block to set the NV slot. So no else here. */
2183 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2184 != IS_NUMBER_IN_UV) {
2185 /* It wasn't an (integer that doesn't overflow the UV). */
2186 SvNV_set(sv, Atof(SvPVX_const(sv)));
2188 if (! numtype && ckWARN(WARN_NUMERIC))
2191 #if defined(USE_LONG_DOUBLE)
2192 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%" PERL_PRIgldbl ")\n",
2193 PTR2UV(sv), SvNVX(sv)));
2195 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"NVgf")\n",
2196 PTR2UV(sv), SvNVX(sv)));
2199 #ifdef NV_PRESERVES_UV
2200 (void)SvIOKp_on(sv);
2202 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2203 SvIV_set(sv, I_V(SvNVX(sv)));
2204 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
2207 NOOP; /* Integer is imprecise. NOK, IOKp */
2209 /* UV will not work better than IV */
2211 if (SvNVX(sv) > (NV)UV_MAX) {
2213 /* Integer is inaccurate. NOK, IOKp, is UV */
2214 SvUV_set(sv, UV_MAX);
2216 SvUV_set(sv, U_V(SvNVX(sv)));
2217 /* 0xFFFFFFFFFFFFFFFF not an issue in here, NVs
2218 NV preservse UV so can do correct comparison. */
2219 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
2222 NOOP; /* Integer is imprecise. NOK, IOKp, is UV */
2227 #else /* NV_PRESERVES_UV */
2228 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2229 == (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT)) {
2230 /* The IV/UV slot will have been set from value returned by
2231 grok_number above. The NV slot has just been set using
2234 assert (SvIOKp(sv));
2236 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2237 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2238 /* Small enough to preserve all bits. */
2239 (void)SvIOKp_on(sv);
2241 SvIV_set(sv, I_V(SvNVX(sv)));
2242 if ((NV)(SvIVX(sv)) == SvNVX(sv))
2244 /* Assumption: first non-preserved integer is < IV_MAX,
2245 this NV is in the preserved range, therefore: */
2246 if (!(U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))
2248 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);
2252 0 0 already failed to read UV.
2253 0 1 already failed to read UV.
2254 1 0 you won't get here in this case. IV/UV
2255 slot set, public IOK, Atof() unneeded.
2256 1 1 already read UV.
2257 so there's no point in sv_2iuv_non_preserve() attempting
2258 to use atol, strtol, strtoul etc. */
2260 sv_2iuv_non_preserve (sv, numtype);
2262 sv_2iuv_non_preserve (sv);
2266 #endif /* NV_PRESERVES_UV */
2267 /* It might be more code efficient to go through the entire logic above
2268 and conditionally set with SvIOKp_on() rather than SvIOK(), but it
2269 gets complex and potentially buggy, so more programmer efficient
2270 to do it this way, by turning off the public flags: */
2272 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK);
2276 if (isGV_with_GP(sv))
2277 return glob_2number(MUTABLE_GV(sv));
2279 if (!(SvFLAGS(sv) & SVs_PADTMP)) {
2280 if (!PL_localizing && ckWARN(WARN_UNINITIALIZED))
2283 if (SvTYPE(sv) < SVt_IV)
2284 /* Typically the caller expects that sv_any is not NULL now. */
2285 sv_upgrade(sv, SVt_IV);
2286 /* Return 0 from the caller. */
2293 =for apidoc sv_2iv_flags
2295 Return the integer value of an SV, doing any necessary string
2296 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2297 Normally used via the C<SvIV(sv)> and C<SvIVx(sv)> macros.
2303 Perl_sv_2iv_flags(pTHX_ register SV *const sv, const I32 flags)
2308 if (SvGMAGICAL(sv) || (SvTYPE(sv) == SVt_PVGV && SvVALID(sv))) {
2309 /* FBMs use the same flag bit as SVf_IVisUV, so must let them
2310 cache IVs just in case. In practice it seems that they never
2311 actually anywhere accessible by user Perl code, let alone get used
2312 in anything other than a string context. */
2313 if (flags & SV_GMAGIC)
2318 return I_V(SvNVX(sv));
2320 if (SvPOKp(sv) && SvLEN(sv)) {
2323 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2325 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2326 == IS_NUMBER_IN_UV) {
2327 /* It's definitely an integer */
2328 if (numtype & IS_NUMBER_NEG) {
2329 if (value < (UV)IV_MIN)
2332 if (value < (UV)IV_MAX)
2337 if (ckWARN(WARN_NUMERIC))
2340 return I_V(Atof(SvPVX_const(sv)));
2345 assert(SvTYPE(sv) >= SVt_PVMG);
2346 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
2347 } else if (SvTHINKFIRST(sv)) {
2351 SV * const tmpstr=AMG_CALLun(sv,numer);
2352 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2353 return SvIV(tmpstr);
2356 return PTR2IV(SvRV(sv));
2359 sv_force_normal_flags(sv, 0);
2361 if (SvREADONLY(sv) && !SvOK(sv)) {
2362 if (ckWARN(WARN_UNINITIALIZED))
2368 if (S_sv_2iuv_common(aTHX_ sv))
2371 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"IVdf")\n",
2372 PTR2UV(sv),SvIVX(sv)));
2373 return SvIsUV(sv) ? (IV)SvUVX(sv) : SvIVX(sv);
2377 =for apidoc sv_2uv_flags
2379 Return the unsigned integer value of an SV, doing any necessary string
2380 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2381 Normally used via the C<SvUV(sv)> and C<SvUVx(sv)> macros.
2387 Perl_sv_2uv_flags(pTHX_ register SV *const sv, const I32 flags)
2392 if (SvGMAGICAL(sv) || (SvTYPE(sv) == SVt_PVGV && SvVALID(sv))) {
2393 /* FBMs use the same flag bit as SVf_IVisUV, so must let them
2394 cache IVs just in case. */
2395 if (flags & SV_GMAGIC)
2400 return U_V(SvNVX(sv));
2401 if (SvPOKp(sv) && SvLEN(sv)) {
2404 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2406 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2407 == IS_NUMBER_IN_UV) {
2408 /* It's definitely an integer */
2409 if (!(numtype & IS_NUMBER_NEG))
2413 if (ckWARN(WARN_NUMERIC))
2416 return U_V(Atof(SvPVX_const(sv)));
2421 assert(SvTYPE(sv) >= SVt_PVMG);
2422 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
2423 } else if (SvTHINKFIRST(sv)) {
2427 SV *const tmpstr = AMG_CALLun(sv,numer);
2428 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2429 return SvUV(tmpstr);
2432 return PTR2UV(SvRV(sv));
2435 sv_force_normal_flags(sv, 0);
2437 if (SvREADONLY(sv) && !SvOK(sv)) {
2438 if (ckWARN(WARN_UNINITIALIZED))
2444 if (S_sv_2iuv_common(aTHX_ sv))
2448 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2uv(%"UVuf")\n",
2449 PTR2UV(sv),SvUVX(sv)));
2450 return SvIsUV(sv) ? SvUVX(sv) : (UV)SvIVX(sv);
2456 Return the num value of an SV, doing any necessary string or integer
2457 conversion, magic etc. Normally used via the C<SvNV(sv)> and C<SvNVx(sv)>
2464 Perl_sv_2nv(pTHX_ register SV *const sv)
2469 if (SvGMAGICAL(sv) || (SvTYPE(sv) == SVt_PVGV && SvVALID(sv))) {
2470 /* FBMs use the same flag bit as SVf_IVisUV, so must let them
2471 cache IVs just in case. */
2475 if ((SvPOKp(sv) && SvLEN(sv)) && !SvIOKp(sv)) {
2476 if (!SvIOKp(sv) && ckWARN(WARN_NUMERIC) &&
2477 !grok_number(SvPVX_const(sv), SvCUR(sv), NULL))
2479 return Atof(SvPVX_const(sv));
2483 return (NV)SvUVX(sv);
2485 return (NV)SvIVX(sv);
2490 assert(SvTYPE(sv) >= SVt_PVMG);
2491 /* This falls through to the report_uninit near the end of the
2493 } else if (SvTHINKFIRST(sv)) {
2497 SV *const tmpstr = AMG_CALLun(sv,numer);
2498 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2499 return SvNV(tmpstr);
2502 return PTR2NV(SvRV(sv));
2505 sv_force_normal_flags(sv, 0);
2507 if (SvREADONLY(sv) && !SvOK(sv)) {
2508 if (ckWARN(WARN_UNINITIALIZED))
2513 if (SvTYPE(sv) < SVt_NV) {
2514 /* The logic to use SVt_PVNV if necessary is in sv_upgrade. */
2515 sv_upgrade(sv, SVt_NV);
2516 #ifdef USE_LONG_DOUBLE
2518 STORE_NUMERIC_LOCAL_SET_STANDARD();
2519 PerlIO_printf(Perl_debug_log,
2520 "0x%"UVxf" num(%" PERL_PRIgldbl ")\n",
2521 PTR2UV(sv), SvNVX(sv));
2522 RESTORE_NUMERIC_LOCAL();
2526 STORE_NUMERIC_LOCAL_SET_STANDARD();
2527 PerlIO_printf(Perl_debug_log, "0x%"UVxf" num(%"NVgf")\n",
2528 PTR2UV(sv), SvNVX(sv));
2529 RESTORE_NUMERIC_LOCAL();
2533 else if (SvTYPE(sv) < SVt_PVNV)
2534 sv_upgrade(sv, SVt_PVNV);
2539 SvNV_set(sv, SvIsUV(sv) ? (NV)SvUVX(sv) : (NV)SvIVX(sv));
2540 #ifdef NV_PRESERVES_UV
2546 /* Only set the public NV OK flag if this NV preserves the IV */
2547 /* Check it's not 0xFFFFFFFFFFFFFFFF */
2549 SvIsUV(sv) ? ((SvUVX(sv) != UV_MAX)&&(SvUVX(sv) == U_V(SvNVX(sv))))
2550 : (SvIVX(sv) == I_V(SvNVX(sv))))
2556 else if (SvPOKp(sv) && SvLEN(sv)) {
2558 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2559 if (!SvIOKp(sv) && !numtype && ckWARN(WARN_NUMERIC))
2561 #ifdef NV_PRESERVES_UV
2562 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2563 == IS_NUMBER_IN_UV) {
2564 /* It's definitely an integer */
2565 SvNV_set(sv, (numtype & IS_NUMBER_NEG) ? -(NV)value : (NV)value);
2567 SvNV_set(sv, Atof(SvPVX_const(sv)));
2573 SvNV_set(sv, Atof(SvPVX_const(sv)));
2574 /* Only set the public NV OK flag if this NV preserves the value in
2575 the PV at least as well as an IV/UV would.
2576 Not sure how to do this 100% reliably. */
2577 /* if that shift count is out of range then Configure's test is
2578 wonky. We shouldn't be in here with NV_PRESERVES_UV_BITS ==
2580 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2581 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2582 SvNOK_on(sv); /* Definitely small enough to preserve all bits */
2583 } else if (!(numtype & IS_NUMBER_IN_UV)) {
2584 /* Can't use strtol etc to convert this string, so don't try.
2585 sv_2iv and sv_2uv will use the NV to convert, not the PV. */
2588 /* value has been set. It may not be precise. */
2589 if ((numtype & IS_NUMBER_NEG) && (value > (UV)IV_MIN)) {
2590 /* 2s complement assumption for (UV)IV_MIN */
2591 SvNOK_on(sv); /* Integer is too negative. */
2596 if (numtype & IS_NUMBER_NEG) {
2597 SvIV_set(sv, -(IV)value);
2598 } else if (value <= (UV)IV_MAX) {
2599 SvIV_set(sv, (IV)value);
2601 SvUV_set(sv, value);
2605 if (numtype & IS_NUMBER_NOT_INT) {
2606 /* I believe that even if the original PV had decimals,
2607 they are lost beyond the limit of the FP precision.
2608 However, neither is canonical, so both only get p
2609 flags. NWC, 2000/11/25 */
2610 /* Both already have p flags, so do nothing */
2612 const NV nv = SvNVX(sv);
2613 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2614 if (SvIVX(sv) == I_V(nv)) {
2617 /* It had no "." so it must be integer. */
2621 /* between IV_MAX and NV(UV_MAX).
2622 Could be slightly > UV_MAX */
2624 if (numtype & IS_NUMBER_NOT_INT) {
2625 /* UV and NV both imprecise. */
2627 const UV nv_as_uv = U_V(nv);
2629 if (value == nv_as_uv && SvUVX(sv) != UV_MAX) {
2638 /* It might be more code efficient to go through the entire logic above
2639 and conditionally set with SvNOKp_on() rather than SvNOK(), but it
2640 gets complex and potentially buggy, so more programmer efficient
2641 to do it this way, by turning off the public flags: */
2643 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK);
2644 #endif /* NV_PRESERVES_UV */
2647 if (isGV_with_GP(sv)) {
2648 glob_2number(MUTABLE_GV(sv));
2652 if (!PL_localizing && !(SvFLAGS(sv) & SVs_PADTMP) && ckWARN(WARN_UNINITIALIZED))
2654 assert (SvTYPE(sv) >= SVt_NV);
2655 /* Typically the caller expects that sv_any is not NULL now. */
2656 /* XXX Ilya implies that this is a bug in callers that assume this
2657 and ideally should be fixed. */
2660 #if defined(USE_LONG_DOUBLE)
2662 STORE_NUMERIC_LOCAL_SET_STANDARD();
2663 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2nv(%" PERL_PRIgldbl ")\n",
2664 PTR2UV(sv), SvNVX(sv));
2665 RESTORE_NUMERIC_LOCAL();
2669 STORE_NUMERIC_LOCAL_SET_STANDARD();
2670 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 1nv(%"NVgf")\n",
2671 PTR2UV(sv), SvNVX(sv));
2672 RESTORE_NUMERIC_LOCAL();
2681 Return an SV with the numeric value of the source SV, doing any necessary
2682 reference or overload conversion. You must use the C<SvNUM(sv)> macro to
2683 access this function.
2689 Perl_sv_2num(pTHX_ register SV *const sv)
2691 PERL_ARGS_ASSERT_SV_2NUM;
2696 SV * const tmpsv = AMG_CALLun(sv,numer);
2697 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
2698 return sv_2num(tmpsv);
2700 return sv_2mortal(newSVuv(PTR2UV(SvRV(sv))));
2703 /* uiv_2buf(): private routine for use by sv_2pv_flags(): print an IV or
2704 * UV as a string towards the end of buf, and return pointers to start and
2707 * We assume that buf is at least TYPE_CHARS(UV) long.
2711 S_uiv_2buf(char *const buf, const IV iv, UV uv, const int is_uv, char **const peob)
2713 char *ptr = buf + TYPE_CHARS(UV);
2714 char * const ebuf = ptr;
2717 PERL_ARGS_ASSERT_UIV_2BUF;
2729 *--ptr = '0' + (char)(uv % 10);
2738 =for apidoc sv_2pv_flags
2740 Returns a pointer to the string value of an SV, and sets *lp to its length.
2741 If flags includes SV_GMAGIC, does an mg_get() first. Coerces sv to a string
2743 Normally invoked via the C<SvPV_flags> macro. C<sv_2pv()> and C<sv_2pv_nomg>
2744 usually end up here too.
2750 Perl_sv_2pv_flags(pTHX_ register SV *const sv, STRLEN *const lp, const I32 flags)
2760 if (SvGMAGICAL(sv)) {
2761 if (flags & SV_GMAGIC)
2766 if (flags & SV_MUTABLE_RETURN)
2767 return SvPVX_mutable(sv);
2768 if (flags & SV_CONST_RETURN)
2769 return (char *)SvPVX_const(sv);
2772 if (SvIOKp(sv) || SvNOKp(sv)) {
2773 char tbuf[64]; /* Must fit sprintf/Gconvert of longest IV/NV */
2778 ? my_snprintf(tbuf, sizeof(tbuf), "%"UVuf, (UV)SvUVX(sv))
2779 : my_snprintf(tbuf, sizeof(tbuf), "%"IVdf, (IV)SvIVX(sv));
2781 Gconvert(SvNVX(sv), NV_DIG, 0, tbuf);
2788 #ifdef FIXNEGATIVEZERO
2789 if (len == 2 && tbuf[0] == '-' && tbuf[1] == '0') {
2795 SvUPGRADE(sv, SVt_PV);
2798 s = SvGROW_mutable(sv, len + 1);
2801 return (char*)memcpy(s, tbuf, len + 1);
2807 assert(SvTYPE(sv) >= SVt_PVMG);
2808 /* This falls through to the report_uninit near the end of the
2810 } else if (SvTHINKFIRST(sv)) {
2814 SV *const tmpstr = AMG_CALLun(sv,string);
2815 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2817 /* char *pv = lp ? SvPV(tmpstr, *lp) : SvPV_nolen(tmpstr);
2821 if ((SvFLAGS(tmpstr) & (SVf_POK)) == SVf_POK) {
2822 if (flags & SV_CONST_RETURN) {
2823 pv = (char *) SvPVX_const(tmpstr);
2825 pv = (flags & SV_MUTABLE_RETURN)
2826 ? SvPVX_mutable(tmpstr) : SvPVX(tmpstr);
2829 *lp = SvCUR(tmpstr);
2831 pv = sv_2pv_flags(tmpstr, lp, flags);
2844 SV *const referent = SvRV(sv);
2848 retval = buffer = savepvn("NULLREF", len);
2849 } else if (SvTYPE(referent) == SVt_REGEXP) {
2850 REGEXP * const re = (REGEXP *)MUTABLE_PTR(referent);
2855 /* If the regex is UTF-8 we want the containing scalar to
2856 have an UTF-8 flag too */
2862 if ((seen_evals = RX_SEEN_EVALS(re)))
2863 PL_reginterp_cnt += seen_evals;
2866 *lp = RX_WRAPLEN(re);
2868 return RX_WRAPPED(re);
2870 const char *const typestr = sv_reftype(referent, 0);
2871 const STRLEN typelen = strlen(typestr);
2872 UV addr = PTR2UV(referent);
2873 const char *stashname = NULL;
2874 STRLEN stashnamelen = 0; /* hush, gcc */
2875 const char *buffer_end;
2877 if (SvOBJECT(referent)) {
2878 const HEK *const name = HvNAME_HEK(SvSTASH(referent));
2881 stashname = HEK_KEY(name);
2882 stashnamelen = HEK_LEN(name);
2884 if (HEK_UTF8(name)) {
2890 stashname = "__ANON__";
2893 len = stashnamelen + 1 /* = */ + typelen + 3 /* (0x */
2894 + 2 * sizeof(UV) + 2 /* )\0 */;
2896 len = typelen + 3 /* (0x */
2897 + 2 * sizeof(UV) + 2 /* )\0 */;
2900 Newx(buffer, len, char);
2901 buffer_end = retval = buffer + len;
2903 /* Working backwards */
2907 *--retval = PL_hexdigit[addr & 15];
2908 } while (addr >>= 4);
2914 memcpy(retval, typestr, typelen);
2918 retval -= stashnamelen;
2919 memcpy(retval, stashname, stashnamelen);
2921 /* retval may not neccesarily have reached the start of the
2923 assert (retval >= buffer);
2925 len = buffer_end - retval - 1; /* -1 for that \0 */
2933 if (SvREADONLY(sv) && !SvOK(sv)) {
2936 if (flags & SV_UNDEF_RETURNS_NULL)
2938 if (ckWARN(WARN_UNINITIALIZED))
2943 if (SvIOK(sv) || ((SvIOKp(sv) && !SvNOKp(sv)))) {
2944 /* I'm assuming that if both IV and NV are equally valid then
2945 converting the IV is going to be more efficient */
2946 const U32 isUIOK = SvIsUV(sv);
2947 char buf[TYPE_CHARS(UV)];
2951 if (SvTYPE(sv) < SVt_PVIV)
2952 sv_upgrade(sv, SVt_PVIV);
2953 ptr = uiv_2buf(buf, SvIVX(sv), SvUVX(sv), isUIOK, &ebuf);
2955 /* inlined from sv_setpvn */
2956 s = SvGROW_mutable(sv, len + 1);
2957 Move(ptr, s, len, char);
2961 else if (SvNOKp(sv)) {
2963 if (SvTYPE(sv) < SVt_PVNV)
2964 sv_upgrade(sv, SVt_PVNV);
2965 /* The +20 is pure guesswork. Configure test needed. --jhi */
2966 s = SvGROW_mutable(sv, NV_DIG + 20);
2967 /* some Xenix systems wipe out errno here */
2969 if (SvNVX(sv) == 0.0)
2970 my_strlcpy(s, "0", SvLEN(sv));
2974 Gconvert(SvNVX(sv), NV_DIG, 0, s);
2977 #ifdef FIXNEGATIVEZERO
2978 if (*s == '-' && s[1] == '0' && !s[2]) {
2990 if (isGV_with_GP(sv))
2991 return glob_2pv(MUTABLE_GV(sv), lp);
2995 if (flags & SV_UNDEF_RETURNS_NULL)
2997 if (!PL_localizing && !(SvFLAGS(sv) & SVs_PADTMP) && ckWARN(WARN_UNINITIALIZED))
2999 if (SvTYPE(sv) < SVt_PV)
3000 /* Typically the caller expects that sv_any is not NULL now. */
3001 sv_upgrade(sv, SVt_PV);
3005 const STRLEN len = s - SvPVX_const(sv);
3011 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
3012 PTR2UV(sv),SvPVX_const(sv)));
3013 if (flags & SV_CONST_RETURN)
3014 return (char *)SvPVX_const(sv);
3015 if (flags & SV_MUTABLE_RETURN)
3016 return SvPVX_mutable(sv);
3021 =for apidoc sv_copypv
3023 Copies a stringified representation of the source SV into the
3024 destination SV. Automatically performs any necessary mg_get and
3025 coercion of numeric values into strings. Guaranteed to preserve
3026 UTF8 flag even from overloaded objects. Similar in nature to
3027 sv_2pv[_flags] but operates directly on an SV instead of just the
3028 string. Mostly uses sv_2pv_flags to do its work, except when that
3029 would lose the UTF-8'ness of the PV.
3035 Perl_sv_copypv(pTHX_ SV *const dsv, register SV *const ssv)
3038 const char * const s = SvPV_const(ssv,len);
3040 PERL_ARGS_ASSERT_SV_COPYPV;
3042 sv_setpvn(dsv,s,len);
3050 =for apidoc sv_2pvbyte
3052 Return a pointer to the byte-encoded representation of the SV, and set *lp
3053 to its length. May cause the SV to be downgraded from UTF-8 as a
3056 Usually accessed via the C<SvPVbyte> macro.
3062 Perl_sv_2pvbyte(pTHX_ register SV *const sv, STRLEN *const lp)
3064 PERL_ARGS_ASSERT_SV_2PVBYTE;
3066 sv_utf8_downgrade(sv,0);
3067 return lp ? SvPV(sv,*lp) : SvPV_nolen(sv);
3071 =for apidoc sv_2pvutf8
3073 Return a pointer to the UTF-8-encoded representation of the SV, and set *lp
3074 to its length. May cause the SV to be upgraded to UTF-8 as a side-effect.
3076 Usually accessed via the C<SvPVutf8> macro.
3082 Perl_sv_2pvutf8(pTHX_ register SV *const sv, STRLEN *const lp)
3084 PERL_ARGS_ASSERT_SV_2PVUTF8;
3086 sv_utf8_upgrade(sv);
3087 return lp ? SvPV(sv,*lp) : SvPV_nolen(sv);
3092 =for apidoc sv_2bool
3094 This function is only called on magical items, and is only used by
3095 sv_true() or its macro equivalent.
3101 Perl_sv_2bool(pTHX_ register SV *const sv)
3105 PERL_ARGS_ASSERT_SV_2BOOL;
3113 SV * const tmpsv = AMG_CALLun(sv,bool_);
3114 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
3115 return (bool)SvTRUE(tmpsv);
3117 return SvRV(sv) != 0;
3120 register XPV* const Xpvtmp = (XPV*)SvANY(sv);
3122 (*sv->sv_u.svu_pv > '0' ||
3123 Xpvtmp->xpv_cur > 1 ||
3124 (Xpvtmp->xpv_cur && *sv->sv_u.svu_pv != '0')))
3131 return SvIVX(sv) != 0;
3134 return SvNVX(sv) != 0.0;
3136 if (isGV_with_GP(sv))
3146 =for apidoc sv_utf8_upgrade
3148 Converts the PV of an SV to its UTF-8-encoded form.
3149 Forces the SV to string form if it is not already.
3150 Will C<mg_get> on C<sv> if appropriate.
3151 Always sets the SvUTF8 flag to avoid future validity checks even
3152 if the whole string is the same in UTF-8 as not.
3153 Returns the number of bytes in the converted string
3155 This is not as a general purpose byte encoding to Unicode interface:
3156 use the Encode extension for that.
3158 =for apidoc sv_utf8_upgrade_nomg
3160 Like sv_utf8_upgrade, but doesn't do magic on C<sv>
3162 =for apidoc sv_utf8_upgrade_flags
3164 Converts the PV of an SV to its UTF-8-encoded form.
3165 Forces the SV to string form if it is not already.
3166 Always sets the SvUTF8 flag to avoid future validity checks even
3167 if all the bytes are invariant in UTF-8. If C<flags> has C<SV_GMAGIC> bit set,
3168 will C<mg_get> on C<sv> if appropriate, else not.
3169 Returns the number of bytes in the converted string
3170 C<sv_utf8_upgrade> and
3171 C<sv_utf8_upgrade_nomg> are implemented in terms of this function.
3173 This is not as a general purpose byte encoding to Unicode interface:
3174 use the Encode extension for that.
3178 The grow version is currently not externally documented. It adds a parameter,
3179 extra, which is the number of unused bytes the string of 'sv' is guaranteed to
3180 have free after it upon return. This allows the caller to reserve extra space
3181 that it intends to fill, to avoid extra grows.
3183 Also externally undocumented for the moment is the flag SV_FORCE_UTF8_UPGRADE,
3184 which can be used to tell this function to not first check to see if there are
3185 any characters that are different in UTF-8 (variant characters) which would
3186 force it to allocate a new string to sv, but to assume there are. Typically
3187 this flag is used by a routine that has already parsed the string to find that
3188 there are such characters, and passes this information on so that the work
3189 doesn't have to be repeated.
3191 (One might think that the calling routine could pass in the position of the
3192 first such variant, so it wouldn't have to be found again. But that is not the
3193 case, because typically when the caller is likely to use this flag, it won't be
3194 calling this routine unless it finds something that won't fit into a byte.
3195 Otherwise it tries to not upgrade and just use bytes. But some things that
3196 do fit into a byte are variants in utf8, and the caller may not have been
3197 keeping track of these.)
3199 If the routine itself changes the string, it adds a trailing NUL. Such a NUL
3200 isn't guaranteed due to having other routines do the work in some input cases,
3201 or if the input is already flagged as being in utf8.
3203 The speed of this could perhaps be improved for many cases if someone wanted to
3204 write a fast function that counts the number of variant characters in a string,
3205 especially if it could return the position of the first one.
3210 Perl_sv_utf8_upgrade_flags_grow(pTHX_ register SV *const sv, const I32 flags, STRLEN extra)
3214 PERL_ARGS_ASSERT_SV_UTF8_UPGRADE_FLAGS_GROW;
3216 if (sv == &PL_sv_undef)
3220 if (SvREADONLY(sv) && (SvPOKp(sv) || SvIOKp(sv) || SvNOKp(sv))) {
3221 (void) sv_2pv_flags(sv,&len, flags);
3223 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3227 (void) SvPV_force(sv,len);
3232 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3237 sv_force_normal_flags(sv, 0);
3240 if (PL_encoding && !(flags & SV_UTF8_NO_ENCODING)) {
3241 sv_recode_to_utf8(sv, PL_encoding);
3242 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3246 if (SvCUR(sv) > 0) { /* Assume Latin-1/EBCDIC */
3247 /* This function could be much more efficient if we
3248 * had a FLAG in SVs to signal if there are any variant
3249 * chars in the PV. Given that there isn't such a flag
3250 * make the loop as fast as possible (although there are certainly ways
3251 * to speed this up, eg. through vectorization) */
3252 U8 * s = (U8 *) SvPVX_const(sv);
3253 U8 * e = (U8 *) SvEND(sv);
3255 STRLEN two_byte_count = 0;
3257 if (flags & SV_FORCE_UTF8_UPGRADE) goto must_be_utf8;
3259 /* See if really will need to convert to utf8. We mustn't rely on our
3260 * incoming SV being well formed and having a trailing '\0', as certain
3261 * code in pp_formline can send us partially built SVs. */
3265 if (NATIVE_IS_INVARIANT(ch)) continue;
3267 t--; /* t already incremented; re-point to first variant */
3272 /* utf8 conversion not needed because all are invariants. Mark as
3273 * UTF-8 even if no variant - saves scanning loop */
3279 /* Here, the string should be converted to utf8, either because of an
3280 * input flag (two_byte_count = 0), or because a character that
3281 * requires 2 bytes was found (two_byte_count = 1). t points either to
3282 * the beginning of the string (if we didn't examine anything), or to
3283 * the first variant. In either case, everything from s to t - 1 will
3284 * occupy only 1 byte each on output.
3286 * There are two main ways to convert. One is to create a new string
3287 * and go through the input starting from the beginning, appending each
3288 * converted value onto the new string as we go along. It's probably
3289 * best to allocate enough space in the string for the worst possible
3290 * case rather than possibly running out of space and having to
3291 * reallocate and then copy what we've done so far. Since everything
3292 * from s to t - 1 is invariant, the destination can be initialized
3293 * with these using a fast memory copy
3295 * The other way is to figure out exactly how big the string should be
3296 * by parsing the entire input. Then you don't have to make it big
3297 * enough to handle the worst possible case, and more importantly, if
3298 * the string you already have is large enough, you don't have to
3299 * allocate a new string, you can copy the last character in the input
3300 * string to the final position(s) that will be occupied by the
3301 * converted string and go backwards, stopping at t, since everything
3302 * before that is invariant.
3304 * There are advantages and disadvantages to each method.
3306 * In the first method, we can allocate a new string, do the memory
3307 * copy from the s to t - 1, and then proceed through the rest of the
3308 * string byte-by-byte.
3310 * In the second method, we proceed through the rest of the input
3311 * string just calculating how big the converted string will be. Then
3312 * there are two cases:
3313 * 1) if the string has enough extra space to handle the converted
3314 * value. We go backwards through the string, converting until we
3315 * get to the position we are at now, and then stop. If this
3316 * position is far enough along in the string, this method is
3317 * faster than the other method. If the memory copy were the same
3318 * speed as the byte-by-byte loop, that position would be about
3319 * half-way, as at the half-way mark, parsing to the end and back
3320 * is one complete string's parse, the same amount as starting
3321 * over and going all the way through. Actually, it would be
3322 * somewhat less than half-way, as it's faster to just count bytes
3323 * than to also copy, and we don't have the overhead of allocating
3324 * a new string, changing the scalar to use it, and freeing the
3325 * existing one. But if the memory copy is fast, the break-even
3326 * point is somewhere after half way. The counting loop could be
3327 * sped up by vectorization, etc, to move the break-even point
3328 * further towards the beginning.
3329 * 2) if the string doesn't have enough space to handle the converted
3330 * value. A new string will have to be allocated, and one might
3331 * as well, given that, start from the beginning doing the first
3332 * method. We've spent extra time parsing the string and in
3333 * exchange all we've gotten is that we know precisely how big to
3334 * make the new one. Perl is more optimized for time than space,
3335 * so this case is a loser.
3336 * So what I've decided to do is not use the 2nd method unless it is
3337 * guaranteed that a new string won't have to be allocated, assuming
3338 * the worst case. I also decided not to put any more conditions on it
3339 * than this, for now. It seems likely that, since the worst case is
3340 * twice as big as the unknown portion of the string (plus 1), we won't
3341 * be guaranteed enough space, causing us to go to the first method,
3342 * unless the string is short, or the first variant character is near
3343 * the end of it. In either of these cases, it seems best to use the
3344 * 2nd method. The only circumstance I can think of where this would
3345 * be really slower is if the string had once had much more data in it
3346 * than it does now, but there is still a substantial amount in it */
3349 STRLEN invariant_head = t - s;
3350 STRLEN size = invariant_head + (e - t) * 2 + 1 + extra;
3351 if (SvLEN(sv) < size) {
3353 /* Here, have decided to allocate a new string */
3358 Newx(dst, size, U8);
3360 /* If no known invariants at the beginning of the input string,
3361 * set so starts from there. Otherwise, can use memory copy to
3362 * get up to where we are now, and then start from here */
3364 if (invariant_head <= 0) {
3367 Copy(s, dst, invariant_head, char);
3368 d = dst + invariant_head;
3372 const UV uv = NATIVE8_TO_UNI(*t++);
3373 if (UNI_IS_INVARIANT(uv))
3374 *d++ = (U8)UNI_TO_NATIVE(uv);
3376 *d++ = (U8)UTF8_EIGHT_BIT_HI(uv);
3377 *d++ = (U8)UTF8_EIGHT_BIT_LO(uv);
3381 SvPV_free(sv); /* No longer using pre-existing string */
3382 SvPV_set(sv, (char*)dst);
3383 SvCUR_set(sv, d - dst);
3384 SvLEN_set(sv, size);
3387 /* Here, have decided to get the exact size of the string.
3388 * Currently this happens only when we know that there is
3389 * guaranteed enough space to fit the converted string, so
3390 * don't have to worry about growing. If two_byte_count is 0,
3391 * then t points to the first byte of the string which hasn't
3392 * been examined yet. Otherwise two_byte_count is 1, and t
3393 * points to the first byte in the string that will expand to
3394 * two. Depending on this, start examining at t or 1 after t.
3397 U8 *d = t + two_byte_count;
3400 /* Count up the remaining bytes that expand to two */
3403 const U8 chr = *d++;
3404 if (! NATIVE_IS_INVARIANT(chr)) two_byte_count++;
3407 /* The string will expand by just the number of bytes that
3408 * occupy two positions. But we are one afterwards because of
3409 * the increment just above. This is the place to put the
3410 * trailing NUL, and to set the length before we decrement */
3412 d += two_byte_count;
3413 SvCUR_set(sv, d - s);
3417 /* Having decremented d, it points to the position to put the
3418 * very last byte of the expanded string. Go backwards through
3419 * the string, copying and expanding as we go, stopping when we
3420 * get to the part that is invariant the rest of the way down */
3424 const U8 ch = NATIVE8_TO_UNI(*e--);
3425 if (UNI_IS_INVARIANT(ch)) {
3426 *d-- = UNI_TO_NATIVE(ch);
3428 *d-- = (U8)UTF8_EIGHT_BIT_LO(ch);
3429 *d-- = (U8)UTF8_EIGHT_BIT_HI(ch);
3436 /* Mark as UTF-8 even if no variant - saves scanning loop */
3442 =for apidoc sv_utf8_downgrade
3444 Attempts to convert the PV of an SV from characters to bytes.
3445 If the PV contains a character that cannot fit
3446 in a byte, this conversion will fail;
3447 in this case, either returns false or, if C<fail_ok> is not
3450 This is not as a general purpose Unicode to byte encoding interface:
3451 use the Encode extension for that.
3457 Perl_sv_utf8_downgrade(pTHX_ register SV *const sv, const bool fail_ok)
3461 PERL_ARGS_ASSERT_SV_UTF8_DOWNGRADE;
3463 if (SvPOKp(sv) && SvUTF8(sv)) {
3469 sv_force_normal_flags(sv, 0);
3471 s = (U8 *) SvPV(sv, len);
3472 if (!utf8_to_bytes(s, &len)) {
3477 Perl_croak(aTHX_ "Wide character in %s",
3480 Perl_croak(aTHX_ "Wide character");
3491 =for apidoc sv_utf8_encode
3493 Converts the PV of an SV to UTF-8, but then turns the C<SvUTF8>
3494 flag off so that it looks like octets again.
3500 Perl_sv_utf8_encode(pTHX_ register SV *const sv)
3502 PERL_ARGS_ASSERT_SV_UTF8_ENCODE;
3505 sv_force_normal_flags(sv, 0);
3507 if (SvREADONLY(sv)) {
3508 Perl_croak(aTHX_ "%s", PL_no_modify);
3510 (void) sv_utf8_upgrade(sv);
3515 =for apidoc sv_utf8_decode
3517 If the PV of the SV is an octet sequence in UTF-8
3518 and contains a multiple-byte character, the C<SvUTF8> flag is turned on
3519 so that it looks like a character. If the PV contains only single-byte
3520 characters, the C<SvUTF8> flag stays being off.
3521 Scans PV for validity and returns false if the PV is invalid UTF-8.
3527 Perl_sv_utf8_decode(pTHX_ register SV *const sv)
3529 PERL_ARGS_ASSERT_SV_UTF8_DECODE;
3535 /* The octets may have got themselves encoded - get them back as
3538 if (!sv_utf8_downgrade(sv, TRUE))
3541 /* it is actually just a matter of turning the utf8 flag on, but
3542 * we want to make sure everything inside is valid utf8 first.
3544 c = (const U8 *) SvPVX_const(sv);
3545 if (!is_utf8_string(c, SvCUR(sv)+1))
3547 e = (const U8 *) SvEND(sv);
3550 if (!UTF8_IS_INVARIANT(ch)) {
3560 =for apidoc sv_setsv
3562 Copies the contents of the source SV C<ssv> into the destination SV
3563 C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3564 function if the source SV needs to be reused. Does not handle 'set' magic.
3565 Loosely speaking, it performs a copy-by-value, obliterating any previous
3566 content of the destination.
3568 You probably want to use one of the assortment of wrappers, such as
3569 C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3570 C<SvSetMagicSV_nosteal>.
3572 =for apidoc sv_setsv_flags
3574 Copies the contents of the source SV C<ssv> into the destination SV
3575 C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3576 function if the source SV needs to be reused. Does not handle 'set' magic.
3577 Loosely speaking, it performs a copy-by-value, obliterating any previous
3578 content of the destination.
3579 If the C<flags> parameter has the C<SV_GMAGIC> bit set, will C<mg_get> on
3580 C<ssv> if appropriate, else not. If the C<flags> parameter has the
3581 C<NOSTEAL> bit set then the buffers of temps will not be stolen. <sv_setsv>
3582 and C<sv_setsv_nomg> are implemented in terms of this function.
3584 You probably want to use one of the assortment of wrappers, such as
3585 C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3586 C<SvSetMagicSV_nosteal>.
3588 This is the primary function for copying scalars, and most other
3589 copy-ish functions and macros use this underneath.
3595 S_glob_assign_glob(pTHX_ SV *const dstr, SV *const sstr, const int dtype)
3597 I32 mro_changes = 0; /* 1 = method, 2 = isa */
3599 PERL_ARGS_ASSERT_GLOB_ASSIGN_GLOB;
3601 if (dtype != SVt_PVGV) {
3602 const char * const name = GvNAME(sstr);
3603 const STRLEN len = GvNAMELEN(sstr);
3605 if (dtype >= SVt_PV) {
3611 SvUPGRADE(dstr, SVt_PVGV);
3612 (void)SvOK_off(dstr);
3613 /* FIXME - why are we doing this, then turning it off and on again
3615 isGV_with_GP_on(dstr);
3617 GvSTASH(dstr) = GvSTASH(sstr);
3619 Perl_sv_add_backref(aTHX_ MUTABLE_SV(GvSTASH(dstr)), dstr);
3620 gv_name_set(MUTABLE_GV(dstr), name, len, GV_ADD);
3621 SvFAKE_on(dstr); /* can coerce to non-glob */
3624 #ifdef GV_UNIQUE_CHECK
3625 if (GvUNIQUE((const GV *)dstr)) {
3626 Perl_croak(aTHX_ "%s", PL_no_modify);
3630 if(GvGP(MUTABLE_GV(sstr))) {
3631 /* If source has method cache entry, clear it */
3633 SvREFCNT_dec(GvCV(sstr));
3637 /* If source has a real method, then a method is
3639 else if(GvCV((const GV *)sstr)) {
3644 /* If dest already had a real method, that's a change as well */
3645 if(!mro_changes && GvGP(MUTABLE_GV(dstr)) && GvCVu((const GV *)dstr)) {
3649 if(strEQ(GvNAME((const GV *)dstr),"ISA"))
3652 gp_free(MUTABLE_GV(dstr));
3653 isGV_with_GP_off(dstr);
3654 (void)SvOK_off(dstr);
3655 isGV_with_GP_on(dstr);
3656 GvINTRO_off(dstr); /* one-shot flag */
3657 GvGP(dstr) = gp_ref(GvGP(sstr));
3658 if (SvTAINTED(sstr))
3660 if (GvIMPORTED(dstr) != GVf_IMPORTED
3661 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3663 GvIMPORTED_on(dstr);
3666 if(mro_changes == 2) mro_isa_changed_in(GvSTASH(dstr));
3667 else if(mro_changes) mro_method_changed_in(GvSTASH(dstr));
3672 S_glob_assign_ref(pTHX_ SV *const dstr, SV *const sstr)
3674 SV * const sref = SvREFCNT_inc(SvRV(sstr));
3676 const int intro = GvINTRO(dstr);
3679 const U32 stype = SvTYPE(sref);
3681 PERL_ARGS_ASSERT_GLOB_ASSIGN_REF;
3683 #ifdef GV_UNIQUE_CHECK
3684 if (GvUNIQUE((const GV *)dstr)) {
3685 Perl_croak(aTHX_ "%s", PL_no_modify);
3690 GvINTRO_off(dstr); /* one-shot flag */
3691 GvLINE(dstr) = CopLINE(PL_curcop);
3692 GvEGV(dstr) = MUTABLE_GV(dstr);
3697 location = (SV **) &GvCV(dstr);
3698 import_flag = GVf_IMPORTED_CV;
3701 location = (SV **) &GvHV(dstr);
3702 import_flag = GVf_IMPORTED_HV;
3705 location = (SV **) &GvAV(dstr);
3706 import_flag = GVf_IMPORTED_AV;
3709 location = (SV **) &GvIOp(dstr);
3712 location = (SV **) &GvFORM(dstr);
3714 location = &GvSV(dstr);
3715 import_flag = GVf_IMPORTED_SV;
3718 if (stype == SVt_PVCV) {
3719 /*if (GvCVGEN(dstr) && (GvCV(dstr) != (const CV *)sref || GvCVGEN(dstr))) {*/
3720 if (GvCVGEN(dstr)) {
3721 SvREFCNT_dec(GvCV(dstr));
3723 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3726 SAVEGENERICSV(*location);
3730 if (stype == SVt_PVCV && (*location != sref || GvCVGEN(dstr))) {
3731 CV* const cv = MUTABLE_CV(*location);
3733 if (!GvCVGEN((const GV *)dstr) &&
3734 (CvROOT(cv) || CvXSUB(cv)))
3736 /* Redefining a sub - warning is mandatory if
3737 it was a const and its value changed. */
3738 if (CvCONST(cv) && CvCONST((const CV *)sref)
3740 == cv_const_sv((const CV *)sref)) {
3742 /* They are 2 constant subroutines generated from
3743 the same constant. This probably means that
3744 they are really the "same" proxy subroutine
3745 instantiated in 2 places. Most likely this is
3746 when a constant is exported twice. Don't warn.
3749 else if (ckWARN(WARN_REDEFINE)
3751 && (!CvCONST((const CV *)sref)
3752 || sv_cmp(cv_const_sv(cv),
3753 cv_const_sv((const CV *)
3755 Perl_warner(aTHX_ packWARN(WARN_REDEFINE),
3758 ? "Constant subroutine %s::%s redefined"
3759 : "Subroutine %s::%s redefined"),
3760 HvNAME_get(GvSTASH((const GV *)dstr)),
3761 GvENAME(MUTABLE_GV(dstr)));
3765 cv_ckproto_len(cv, (const GV *)dstr,
3766 SvPOK(sref) ? SvPVX_const(sref) : NULL,
3767 SvPOK(sref) ? SvCUR(sref) : 0);
3769 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3770 GvASSUMECV_on(dstr);
3771 if(GvSTASH(dstr)) mro_method_changed_in(GvSTASH(dstr)); /* sub foo { 1 } sub bar { 2 } *bar = \&foo */
3774 if (import_flag && !(GvFLAGS(dstr) & import_flag)
3775 && CopSTASH_ne(PL_curcop, GvSTASH(dstr))) {
3776 GvFLAGS(dstr) |= import_flag;
3781 if (SvTAINTED(sstr))
3787 Perl_sv_setsv_flags(pTHX_ SV *dstr, register SV* sstr, const I32 flags)
3790 register U32 sflags;
3792 register svtype stype;
3794 PERL_ARGS_ASSERT_SV_SETSV_FLAGS;
3799 if (SvIS_FREED(dstr)) {
3800 Perl_croak(aTHX_ "panic: attempt to copy value %" SVf
3801 " to a freed scalar %p", SVfARG(sstr), (void *)dstr);
3803 SV_CHECK_THINKFIRST_COW_DROP(dstr);
3805 sstr = &PL_sv_undef;
3806 if (SvIS_FREED(sstr)) {
3807 Perl_croak(aTHX_ "panic: attempt to copy freed scalar %p to %p",
3808 (void*)sstr, (void*)dstr);
3810 stype = SvTYPE(sstr);
3811 dtype = SvTYPE(dstr);
3813 (void)SvAMAGIC_off(dstr);
3816 /* need to nuke the magic */
3820 /* There's a lot of redundancy below but we're going for speed here */
3825 if (dtype != SVt_PVGV) {
3826 (void)SvOK_off(dstr);
3834 sv_upgrade(dstr, SVt_IV);
3838 sv_upgrade(dstr, SVt_PVIV);
3841 goto end_of_first_switch;
3843 (void)SvIOK_only(dstr);
3844 SvIV_set(dstr, SvIVX(sstr));
3847 /* SvTAINTED can only be true if the SV has taint magic, which in
3848 turn means that the SV type is PVMG (or greater). This is the
3849 case statement for SVt_IV, so this cannot be true (whatever gcov
3851 assert(!SvTAINTED(sstr));
3856 if (dtype < SVt_PV && dtype != SVt_IV)
3857 sv_upgrade(dstr, SVt_IV);
3865 sv_upgrade(dstr, SVt_NV);
3869 sv_upgrade(dstr, SVt_PVNV);
3872 goto end_of_first_switch;
3874 SvNV_set(dstr, SvNVX(sstr));
3875 (void)SvNOK_only(dstr);
3876 /* SvTAINTED can only be true if the SV has taint magic, which in
3877 turn means that the SV type is PVMG (or greater). This is the
3878 case statement for SVt_NV, so this cannot be true (whatever gcov
3880 assert(!SvTAINTED(sstr));
3886 #ifdef PERL_OLD_COPY_ON_WRITE
3887 if ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS) {
3888 if (dtype < SVt_PVIV)
3889 sv_upgrade(dstr, SVt_PVIV);
3897 sv_upgrade(dstr, SVt_PV);
3900 if (dtype < SVt_PVIV)
3901 sv_upgrade(dstr, SVt_PVIV);
3904 if (dtype < SVt_PVNV)
3905 sv_upgrade(dstr, SVt_PVNV);
3909 const char * const type = sv_reftype(sstr,0);
3911 Perl_croak(aTHX_ "Bizarre copy of %s in %s", type, OP_NAME(PL_op));
3913 Perl_croak(aTHX_ "Bizarre copy of %s", type);
3917 /* case SVt_BIND: */
3920 if (isGV_with_GP(sstr) && dtype <= SVt_PVGV) {
3921 glob_assign_glob(dstr, sstr, dtype);
3924 /* SvVALID means that this PVGV is playing at being an FBM. */
3928 if (SvGMAGICAL(sstr) && (flags & SV_GMAGIC)) {
3930 if (SvTYPE(sstr) != stype) {
3931 stype = SvTYPE(sstr);
3932 if (isGV_with_GP(sstr) && stype == SVt_PVGV && dtype <= SVt_PVGV) {
3933 glob_assign_glob(dstr, sstr, dtype);
3938 if (stype == SVt_PVLV)
3939 SvUPGRADE(dstr, SVt_PVNV);
3941 SvUPGRADE(dstr, (svtype)stype);
3943 end_of_first_switch:
3945 /* dstr may have been upgraded. */
3946 dtype = SvTYPE(dstr);
3947 sflags = SvFLAGS(sstr);
3949 if (dtype == SVt_PVCV || dtype == SVt_PVFM) {
3950 /* Assigning to a subroutine sets the prototype. */
3953 const char *const ptr = SvPV_const(sstr, len);
3955 SvGROW(dstr, len + 1);
3956 Copy(ptr, SvPVX(dstr), len + 1, char);
3957 SvCUR_set(dstr, len);
3959 SvFLAGS(dstr) |= sflags & SVf_UTF8;
3963 } else if (dtype == SVt_PVAV || dtype == SVt_PVHV) {
3964 const char * const type = sv_reftype(dstr,0);
3966 Perl_croak(aTHX_ "Cannot copy to %s in %s", type, OP_NAME(PL_op));
3968 Perl_croak(aTHX_ "Cannot copy to %s", type);
3969 } else if (sflags & SVf_ROK) {
3970 if (isGV_with_GP(dstr) && dtype == SVt_PVGV
3971 && SvTYPE(SvRV(sstr)) == SVt_PVGV && isGV_with_GP(SvRV(sstr))) {
3974 if (GvIMPORTED(dstr) != GVf_IMPORTED
3975 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3977 GvIMPORTED_on(dstr);
3982 glob_assign_glob(dstr, sstr, dtype);
3986 if (dtype >= SVt_PV) {
3987 if (dtype == SVt_PVGV && isGV_with_GP(dstr)) {
3988 glob_assign_ref(dstr, sstr);
3991 if (SvPVX_const(dstr)) {
3997 (void)SvOK_off(dstr);
3998 SvRV_set(dstr, SvREFCNT_inc(SvRV(sstr)));
3999 SvFLAGS(dstr) |= sflags & SVf_ROK;
4000 assert(!(sflags & SVp_NOK));
4001 assert(!(sflags & SVp_IOK));
4002 assert(!(sflags & SVf_NOK));
4003 assert(!(sflags & SVf_IOK));
4005 else if (dtype == SVt_PVGV && isGV_with_GP(dstr)) {
4006 if (!(sflags & SVf_OK)) {
4007 if (ckWARN(WARN_MISC))
4008 Perl_warner(aTHX_ packWARN(WARN_MISC),
4009 "Undefined value assigned to typeglob");
4012 GV *gv = gv_fetchsv(sstr, GV_ADD, SVt_PVGV);
4013 if (dstr != (const SV *)gv) {
4015 gp_free(MUTABLE_GV(dstr));
4016 GvGP(dstr) = gp_ref(GvGP(gv));
4020 else if (sflags & SVp_POK) {
4024 * Check to see if we can just swipe the string. If so, it's a
4025 * possible small lose on short strings, but a big win on long ones.
4026 * It might even be a win on short strings if SvPVX_const(dstr)
4027 * has to be allocated and SvPVX_const(sstr) has to be freed.
4028 * Likewise if we can set up COW rather than doing an actual copy, we
4029 * drop to the else clause, as the swipe code and the COW setup code
4030 * have much in common.
4033 /* Whichever path we take through the next code, we want this true,
4034 and doing it now facilitates the COW check. */
4035 (void)SvPOK_only(dstr);
4038 /* If we're already COW then this clause is not true, and if COW
4039 is allowed then we drop down to the else and make dest COW
4040 with us. If caller hasn't said that we're allowed to COW
4041 shared hash keys then we don't do the COW setup, even if the
4042 source scalar is a shared hash key scalar. */
4043 (((flags & SV_COW_SHARED_HASH_KEYS)
4044 ? (sflags & (SVf_FAKE|SVf_READONLY)) != (SVf_FAKE|SVf_READONLY)
4045 : 1 /* If making a COW copy is forbidden then the behaviour we
4046 desire is as if the source SV isn't actually already
4047 COW, even if it is. So we act as if the source flags
4048 are not COW, rather than actually testing them. */
4050 #ifndef PERL_OLD_COPY_ON_WRITE
4051 /* The change that added SV_COW_SHARED_HASH_KEYS makes the logic
4052 when PERL_OLD_COPY_ON_WRITE is defined a little wrong.
4053 Conceptually PERL_OLD_COPY_ON_WRITE being defined should
4054 override SV_COW_SHARED_HASH_KEYS, because it means "always COW"
4055 but in turn, it's somewhat dead code, never expected to go
4056 live, but more kept as a placeholder on how to do it better
4057 in a newer implementation. */
4058 /* If we are COW and dstr is a suitable target then we drop down
4059 into the else and make dest a COW of us. */
4060 || (SvFLAGS(dstr) & CAN_COW_MASK) != CAN_COW_FLAGS
4065 (sflags & SVs_TEMP) && /* slated for free anyway? */
4066 !(sflags & SVf_OOK) && /* and not involved in OOK hack? */
4067 (!(flags & SV_NOSTEAL)) &&
4068 /* and we're allowed to steal temps */
4069 SvREFCNT(sstr) == 1 && /* and no other references to it? */
4070 SvLEN(sstr) && /* and really is a string */
4071 /* and won't be needed again, potentially */
4072 !(PL_op && PL_op->op_type == OP_AASSIGN))
4073 #ifdef PERL_OLD_COPY_ON_WRITE
4074 && ((flags & SV_COW_SHARED_HASH_KEYS)
4075 ? (!((sflags & CAN_COW_MASK) == CAN_COW_FLAGS
4076 && (SvFLAGS(dstr) & CAN_COW_MASK) == CAN_COW_FLAGS
4077 && SvTYPE(sstr) >= SVt_PVIV))
4081 /* Failed the swipe test, and it's not a shared hash key either.
4082 Have to copy the string. */
4083 STRLEN len = SvCUR(sstr);
4084 SvGROW(dstr, len + 1); /* inlined from sv_setpvn */
4085 Move(SvPVX_const(sstr),SvPVX(dstr),len,char);
4086 SvCUR_set(dstr, len);
4087 *SvEND(dstr) = '\0';
4089 /* If PERL_OLD_COPY_ON_WRITE is not defined, then isSwipe will always
4091 /* Either it's a shared hash key, or it's suitable for
4092 copy-on-write or we can swipe the string. */
4094 PerlIO_printf(Perl_debug_log, "Copy on write: sstr --> dstr\n");
4098 #ifdef PERL_OLD_COPY_ON_WRITE
4100 /* I believe I should acquire a global SV mutex if
4101 it's a COW sv (not a shared hash key) to stop
4102 it going un copy-on-write.
4103 If the source SV has gone un copy on write between up there
4104 and down here, then (assert() that) it is of the correct
4105 form to make it copy on write again */
4106 if ((sflags & (SVf_FAKE | SVf_READONLY))
4107 != (SVf_FAKE | SVf_READONLY)) {
4108 SvREADONLY_on(sstr);
4110 /* Make the source SV into a loop of 1.
4111 (about to become 2) */
4112 SV_COW_NEXT_SV_SET(sstr, sstr);
4116 /* Initial code is common. */
4117 if (SvPVX_const(dstr)) { /* we know that dtype >= SVt_PV */
4122 /* making another shared SV. */
4123 STRLEN cur = SvCUR(sstr);
4124 STRLEN len = SvLEN(sstr);
4125 #ifdef PERL_OLD_COPY_ON_WRITE
4127 assert (SvTYPE(dstr) >= SVt_PVIV);
4128 /* SvIsCOW_normal */
4129 /* splice us in between source and next-after-source. */
4130 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4131 SV_COW_NEXT_SV_SET(sstr, dstr);
4132 SvPV_set(dstr, SvPVX_mutable(sstr));
4136 /* SvIsCOW_shared_hash */
4137 DEBUG_C(PerlIO_printf(Perl_debug_log,
4138 "Copy on write: Sharing hash\n"));
4140 assert (SvTYPE(dstr) >= SVt_PV);
4142 HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)))));
4144 SvLEN_set(dstr, len);
4145 SvCUR_set(dstr, cur);
4146 SvREADONLY_on(dstr);
4148 /* Relesase a global SV mutex. */
4151 { /* Passes the swipe test. */
4152 SvPV_set(dstr, SvPVX_mutable(sstr));
4153 SvLEN_set(dstr, SvLEN(sstr));
4154 SvCUR_set(dstr, SvCUR(sstr));
4157 (void)SvOK_off(sstr); /* NOTE: nukes most SvFLAGS on sstr */
4158 SvPV_set(sstr, NULL);
4164 if (sflags & SVp_NOK) {
4165 SvNV_set(dstr, SvNVX(sstr));
4167 if (sflags & SVp_IOK) {
4168 SvIV_set(dstr, SvIVX(sstr));
4169 /* Must do this otherwise some other overloaded use of 0x80000000
4170 gets confused. I guess SVpbm_VALID */
4171 if (sflags & SVf_IVisUV)
4174 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_NOK|SVp_NOK|SVf_UTF8);
4176 const MAGIC * const smg = SvVSTRING_mg(sstr);
4178 sv_magic(dstr, NULL, PERL_MAGIC_vstring,
4179 smg->mg_ptr, smg->mg_len);
4180 SvRMAGICAL_on(dstr);
4184 else if (sflags & (SVp_IOK|SVp_NOK)) {
4185 (void)SvOK_off(dstr);
4186 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_IVisUV|SVf_NOK|SVp_NOK);
4187 if (sflags & SVp_IOK) {
4188 /* XXXX Do we want to set IsUV for IV(ROK)? Be extra safe... */
4189 SvIV_set(dstr, SvIVX(sstr));
4191 if (sflags & SVp_NOK) {
4192 SvNV_set(dstr, SvNVX(sstr));
4196 if (isGV_with_GP(sstr)) {
4197 /* This stringification rule for globs is spread in 3 places.
4198 This feels bad. FIXME. */
4199 const U32 wasfake = sflags & SVf_FAKE;
4201 /* FAKE globs can get coerced, so need to turn this off
4202 temporarily if it is on. */
4204 gv_efullname3(dstr, MUTABLE_GV(sstr), "*");
4205 SvFLAGS(sstr) |= wasfake;
4208 (void)SvOK_off(dstr);
4210 if (SvTAINTED(sstr))
4215 =for apidoc sv_setsv_mg
4217 Like C<sv_setsv>, but also handles 'set' magic.
4223 Perl_sv_setsv_mg(pTHX_ SV *const dstr, register SV *const sstr)
4225 PERL_ARGS_ASSERT_SV_SETSV_MG;
4227 sv_setsv(dstr,sstr);
4231 #ifdef PERL_OLD_COPY_ON_WRITE
4233 Perl_sv_setsv_cow(pTHX_ SV *dstr, SV *sstr)
4235 STRLEN cur = SvCUR(sstr);
4236 STRLEN len = SvLEN(sstr);
4237 register char *new_pv;
4239 PERL_ARGS_ASSERT_SV_SETSV_COW;
4242 PerlIO_printf(Perl_debug_log, "Fast copy on write: %p -> %p\n",
4243 (void*)sstr, (void*)dstr);
4250 if (SvTHINKFIRST(dstr))
4251 sv_force_normal_flags(dstr, SV_COW_DROP_PV);
4252 else if (SvPVX_const(dstr))
4253 Safefree(SvPVX_const(dstr));
4257 SvUPGRADE(dstr, SVt_PVIV);
4259 assert (SvPOK(sstr));
4260 assert (SvPOKp(sstr));
4261 assert (!SvIOK(sstr));
4262 assert (!SvIOKp(sstr));
4263 assert (!SvNOK(sstr));
4264 assert (!SvNOKp(sstr));
4266 if (SvIsCOW(sstr)) {
4268 if (SvLEN(sstr) == 0) {
4269 /* source is a COW shared hash key. */
4270 DEBUG_C(PerlIO_printf(Perl_debug_log,
4271 "Fast copy on write: Sharing hash\n"));
4272 new_pv = HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr))));
4275 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4277 assert ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS);
4278 SvUPGRADE(sstr, SVt_PVIV);
4279 SvREADONLY_on(sstr);
4281 DEBUG_C(PerlIO_printf(Perl_debug_log,
4282 "Fast copy on write: Converting sstr to COW\n"));
4283 SV_COW_NEXT_SV_SET(dstr, sstr);
4285 SV_COW_NEXT_SV_SET(sstr, dstr);
4286 new_pv = SvPVX_mutable(sstr);
4289 SvPV_set(dstr, new_pv);
4290 SvFLAGS(dstr) = (SVt_PVIV|SVf_POK|SVp_POK|SVf_FAKE|SVf_READONLY);
4293 SvLEN_set(dstr, len);
4294 SvCUR_set(dstr, cur);
4303 =for apidoc sv_setpvn
4305 Copies a string into an SV. The C<len> parameter indicates the number of
4306 bytes to be copied. If the C<ptr> argument is NULL the SV will become
4307 undefined. Does not handle 'set' magic. See C<sv_setpvn_mg>.
4313 Perl_sv_setpvn(pTHX_ register SV *const sv, register const char *const ptr, register const STRLEN len)
4316 register char *dptr;
4318 PERL_ARGS_ASSERT_SV_SETPVN;
4320 SV_CHECK_THINKFIRST_COW_DROP(sv);
4326 /* len is STRLEN which is unsigned, need to copy to signed */
4329 Perl_croak(aTHX_ "panic: sv_setpvn called with negative strlen");
4331 SvUPGRADE(sv, SVt_PV);
4333 dptr = SvGROW(sv, len + 1);
4334 Move(ptr,dptr,len,char);
4337 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4342 =for apidoc sv_setpvn_mg
4344 Like C<sv_setpvn>, but also handles 'set' magic.
4350 Perl_sv_setpvn_mg(pTHX_ register SV *const sv, register const char *const ptr, register const STRLEN len)
4352 PERL_ARGS_ASSERT_SV_SETPVN_MG;
4354 sv_setpvn(sv,ptr,len);
4359 =for apidoc sv_setpv
4361 Copies a string into an SV. The string must be null-terminated. Does not
4362 handle 'set' magic. See C<sv_setpv_mg>.
4368 Perl_sv_setpv(pTHX_ register SV *const sv, register const char *const ptr)
4371 register STRLEN len;
4373 PERL_ARGS_ASSERT_SV_SETPV;
4375 SV_CHECK_THINKFIRST_COW_DROP(sv);
4381 SvUPGRADE(sv, SVt_PV);
4383 SvGROW(sv, len + 1);
4384 Move(ptr,SvPVX(sv),len+1,char);
4386 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4391 =for apidoc sv_setpv_mg
4393 Like C<sv_setpv>, but also handles 'set' magic.
4399 Perl_sv_setpv_mg(pTHX_ register SV *const sv, register const char *const ptr)
4401 PERL_ARGS_ASSERT_SV_SETPV_MG;
4408 =for apidoc sv_usepvn_flags
4410 Tells an SV to use C<ptr> to find its string value. Normally the
4411 string is stored inside the SV but sv_usepvn allows the SV to use an
4412 outside string. The C<ptr> should point to memory that was allocated
4413 by C<malloc>. The string length, C<len>, must be supplied. By default
4414 this function will realloc (i.e. move) the memory pointed to by C<ptr>,
4415 so that pointer should not be freed or used by the programmer after
4416 giving it to sv_usepvn, and neither should any pointers from "behind"
4417 that pointer (e.g. ptr + 1) be used.
4419 If C<flags> & SV_SMAGIC is true, will call SvSETMAGIC. If C<flags> &
4420 SV_HAS_TRAILING_NUL is true, then C<ptr[len]> must be NUL, and the realloc
4421 will be skipped. (i.e. the buffer is actually at least 1 byte longer than
4422 C<len>, and already meets the requirements for storing in C<SvPVX>)
4428 Perl_sv_usepvn_flags(pTHX_ SV *const sv, char *ptr, const STRLEN len, const U32 flags)
4433 PERL_ARGS_ASSERT_SV_USEPVN_FLAGS;
4435 SV_CHECK_THINKFIRST_COW_DROP(sv);
4436 SvUPGRADE(sv, SVt_PV);
4439 if (flags & SV_SMAGIC)
4443 if (SvPVX_const(sv))
4447 if (flags & SV_HAS_TRAILING_NUL)
4448 assert(ptr[len] == '\0');
4451 allocate = (flags & SV_HAS_TRAILING_NUL)
4453 #ifdef Perl_safesysmalloc_size
4456 PERL_STRLEN_ROUNDUP(len + 1);
4458 if (flags & SV_HAS_TRAILING_NUL) {
4459 /* It's long enough - do nothing.
4460 Specfically Perl_newCONSTSUB is relying on this. */
4463 /* Force a move to shake out bugs in callers. */
4464 char *new_ptr = (char*)safemalloc(allocate);
4465 Copy(ptr, new_ptr, len, char);
4466 PoisonFree(ptr,len,char);
4470 ptr = (char*) saferealloc (ptr, allocate);
4473 #ifdef Perl_safesysmalloc_size
4474 SvLEN_set(sv, Perl_safesysmalloc_size(ptr));
4476 SvLEN_set(sv, allocate);
4480 if (!(flags & SV_HAS_TRAILING_NUL)) {
4483 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4485 if (flags & SV_SMAGIC)
4489 #ifdef PERL_OLD_COPY_ON_WRITE
4490 /* Need to do this *after* making the SV normal, as we need the buffer
4491 pointer to remain valid until after we've copied it. If we let go too early,
4492 another thread could invalidate it by unsharing last of the same hash key
4493 (which it can do by means other than releasing copy-on-write Svs)
4494 or by changing the other copy-on-write SVs in the loop. */
4496 S_sv_release_COW(pTHX_ register SV *sv, const char *pvx, SV *after)
4498 PERL_ARGS_ASSERT_SV_RELEASE_COW;
4500 { /* this SV was SvIsCOW_normal(sv) */
4501 /* we need to find the SV pointing to us. */
4502 SV *current = SV_COW_NEXT_SV(after);
4504 if (current == sv) {
4505 /* The SV we point to points back to us (there were only two of us
4507 Hence other SV is no longer copy on write either. */
4509 SvREADONLY_off(after);
4511 /* We need to follow the pointers around the loop. */
4513 while ((next = SV_COW_NEXT_SV(current)) != sv) {
4516 /* don't loop forever if the structure is bust, and we have
4517 a pointer into a closed loop. */
4518 assert (current != after);
4519 assert (SvPVX_const(current) == pvx);
4521 /* Make the SV before us point to the SV after us. */
4522 SV_COW_NEXT_SV_SET(current, after);
4528 =for apidoc sv_force_normal_flags
4530 Undo various types of fakery on an SV: if the PV is a shared string, make
4531 a private copy; if we're a ref, stop refing; if we're a glob, downgrade to
4532 an xpvmg; if we're a copy-on-write scalar, this is the on-write time when
4533 we do the copy, and is also used locally. If C<SV_COW_DROP_PV> is set
4534 then a copy-on-write scalar drops its PV buffer (if any) and becomes
4535 SvPOK_off rather than making a copy. (Used where this scalar is about to be
4536 set to some other value.) In addition, the C<flags> parameter gets passed to
4537 C<sv_unref_flags()> when unrefing. C<sv_force_normal> calls this function
4538 with flags set to 0.
4544 Perl_sv_force_normal_flags(pTHX_ register SV *const sv, const U32 flags)
4548 PERL_ARGS_ASSERT_SV_FORCE_NORMAL_FLAGS;
4550 #ifdef PERL_OLD_COPY_ON_WRITE
4551 if (SvREADONLY(sv)) {
4552 /* At this point I believe I should acquire a global SV mutex. */
4554 const char * const pvx = SvPVX_const(sv);
4555 const STRLEN len = SvLEN(sv);
4556 const STRLEN cur = SvCUR(sv);
4557 /* next COW sv in the loop. If len is 0 then this is a shared-hash
4558 key scalar, so we mustn't attempt to call SV_COW_NEXT_SV(), as
4559 we'll fail an assertion. */
4560 SV * const next = len ? SV_COW_NEXT_SV(sv) : 0;
4563 PerlIO_printf(Perl_debug_log,
4564 "Copy on write: Force normal %ld\n",
4570 /* This SV doesn't own the buffer, so need to Newx() a new one: */
4573 if (flags & SV_COW_DROP_PV) {
4574 /* OK, so we don't need to copy our buffer. */
4577 SvGROW(sv, cur + 1);
4578 Move(pvx,SvPVX(sv),cur,char);
4583 sv_release_COW(sv, pvx, next);
4585 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4591 else if (IN_PERL_RUNTIME)
4592 Perl_croak(aTHX_ "%s", PL_no_modify);
4593 /* At this point I believe that I can drop the global SV mutex. */
4596 if (SvREADONLY(sv)) {
4598 const char * const pvx = SvPVX_const(sv);
4599 const STRLEN len = SvCUR(sv);
4604 SvGROW(sv, len + 1);
4605 Move(pvx,SvPVX(sv),len,char);
4607 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4609 else if (IN_PERL_RUNTIME)
4610 Perl_croak(aTHX_ "%s", PL_no_modify);
4614 sv_unref_flags(sv, flags);
4615 else if (SvFAKE(sv) && SvTYPE(sv) == SVt_PVGV)
4622 Efficient removal of characters from the beginning of the string buffer.
4623 SvPOK(sv) must be true and the C<ptr> must be a pointer to somewhere inside
4624 the string buffer. The C<ptr> becomes the first character of the adjusted
4625 string. Uses the "OOK hack".
4626 Beware: after this function returns, C<ptr> and SvPVX_const(sv) may no longer
4627 refer to the same chunk of data.
4633 Perl_sv_chop(pTHX_ register SV *const sv, register const char *const ptr)
4639 const U8 *real_start;
4643 PERL_ARGS_ASSERT_SV_CHOP;
4645 if (!ptr || !SvPOKp(sv))
4647 delta = ptr - SvPVX_const(sv);
4649 /* Nothing to do. */
4652 /* SvPVX(sv) may move in SV_CHECK_THINKFIRST(sv), but after this line,
4653 nothing uses the value of ptr any more. */
4654 max_delta = SvLEN(sv) ? SvLEN(sv) : SvCUR(sv);
4655 if (ptr <= SvPVX_const(sv))
4656 Perl_croak(aTHX_ "panic: sv_chop ptr=%p, start=%p, end=%p",
4657 ptr, SvPVX_const(sv), SvPVX_const(sv) + max_delta);
4658 SV_CHECK_THINKFIRST(sv);
4659 if (delta > max_delta)
4660 Perl_croak(aTHX_ "panic: sv_chop ptr=%p (was %p), start=%p, end=%p",
4661 SvPVX_const(sv) + delta, ptr, SvPVX_const(sv),
4662 SvPVX_const(sv) + max_delta);
4665 if (!SvLEN(sv)) { /* make copy of shared string */
4666 const char *pvx = SvPVX_const(sv);
4667 const STRLEN len = SvCUR(sv);
4668 SvGROW(sv, len + 1);
4669 Move(pvx,SvPVX(sv),len,char);
4672 SvFLAGS(sv) |= SVf_OOK;
4675 SvOOK_offset(sv, old_delta);
4677 SvLEN_set(sv, SvLEN(sv) - delta);
4678 SvCUR_set(sv, SvCUR(sv) - delta);
4679 SvPV_set(sv, SvPVX(sv) + delta);
4681 p = (U8 *)SvPVX_const(sv);
4686 real_start = p - delta;
4690 if (delta < 0x100) {
4694 p -= sizeof(STRLEN);
4695 Copy((U8*)&delta, p, sizeof(STRLEN), U8);
4699 /* Fill the preceding buffer with sentinals to verify that no-one is
4701 while (p > real_start) {
4709 =for apidoc sv_catpvn
4711 Concatenates the string onto the end of the string which is in the SV. The
4712 C<len> indicates number of bytes to copy. If the SV has the UTF-8
4713 status set, then the bytes appended should be valid UTF-8.
4714 Handles 'get' magic, but not 'set' magic. See C<sv_catpvn_mg>.
4716 =for apidoc sv_catpvn_flags
4718 Concatenates the string onto the end of the string which is in the SV. The
4719 C<len> indicates number of bytes to copy. If the SV has the UTF-8
4720 status set, then the bytes appended should be valid UTF-8.
4721 If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<dsv> if
4722 appropriate, else not. C<sv_catpvn> and C<sv_catpvn_nomg> are implemented
4723 in terms of this function.
4729 Perl_sv_catpvn_flags(pTHX_ register SV *const dsv, register const char *sstr, register const STRLEN slen, const I32 flags)
4733 const char * const dstr = SvPV_force_flags(dsv, dlen, flags);
4735 PERL_ARGS_ASSERT_SV_CATPVN_FLAGS;
4737 SvGROW(dsv, dlen + slen + 1);
4739 sstr = SvPVX_const(dsv);
4740 Move(sstr, SvPVX(dsv) + dlen, slen, char);
4741 SvCUR_set(dsv, SvCUR(dsv) + slen);
4743 (void)SvPOK_only_UTF8(dsv); /* validate pointer */
4745 if (flags & SV_SMAGIC)
4750 =for apidoc sv_catsv
4752 Concatenates the string from SV C<ssv> onto the end of the string in
4753 SV C<dsv>. Modifies C<dsv> but not C<ssv>. Handles 'get' magic, but
4754 not 'set' magic. See C<sv_catsv_mg>.
4756 =for apidoc sv_catsv_flags
4758 Concatenates the string from SV C<ssv> onto the end of the string in
4759 SV C<dsv>. Modifies C<dsv> but not C<ssv>. If C<flags> has C<SV_GMAGIC>
4760 bit set, will C<mg_get> on the SVs if appropriate, else not. C<sv_catsv>
4761 and C<sv_catsv_nomg> are implemented in terms of this function.
4766 Perl_sv_catsv_flags(pTHX_ SV *const dsv, register SV *const ssv, const I32 flags)
4770 PERL_ARGS_ASSERT_SV_CATSV_FLAGS;
4774 const char *spv = SvPV_const(ssv, slen);
4776 /* sutf8 and dutf8 were type bool, but under USE_ITHREADS,
4777 gcc version 2.95.2 20000220 (Debian GNU/Linux) for
4778 Linux xxx 2.2.17 on sparc64 with gcc -O2, we erroneously
4779 get dutf8 = 0x20000000, (i.e. SVf_UTF8) even though
4780 dsv->sv_flags doesn't have that bit set.
4781 Andy Dougherty 12 Oct 2001
4783 const I32 sutf8 = DO_UTF8(ssv);
4786 if (SvGMAGICAL(dsv) && (flags & SV_GMAGIC))
4788 dutf8 = DO_UTF8(dsv);
4790 if (dutf8 != sutf8) {
4792 /* Not modifying source SV, so taking a temporary copy. */
4793 SV* const csv = newSVpvn_flags(spv, slen, SVs_TEMP);
4795 sv_utf8_upgrade(csv);
4796 spv = SvPV_const(csv, slen);
4799 /* Leave enough space for the cat that's about to happen */
4800 sv_utf8_upgrade_flags_grow(dsv, 0, slen);
4802 sv_catpvn_nomg(dsv, spv, slen);
4805 if (flags & SV_SMAGIC)
4810 =for apidoc sv_catpv
4812 Concatenates the string onto the end of the string which is in the SV.
4813 If the SV has the UTF-8 status set, then the bytes appended should be
4814 valid UTF-8. Handles 'get' magic, but not 'set' magic. See C<sv_catpv_mg>.
4819 Perl_sv_catpv(pTHX_ register SV *const sv, register const char *ptr)
4822 register STRLEN len;
4826 PERL_ARGS_ASSERT_SV_CATPV;
4830 junk = SvPV_force(sv, tlen);
4832 SvGROW(sv, tlen + len + 1);
4834 ptr = SvPVX_const(sv);
4835 Move(ptr,SvPVX(sv)+tlen,len+1,char);
4836 SvCUR_set(sv, SvCUR(sv) + len);
4837 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4842 =for apidoc sv_catpv_mg
4844 Like C<sv_catpv>, but also handles 'set' magic.
4850 Perl_sv_catpv_mg(pTHX_ register SV *const sv, register const char *const ptr)
4852 PERL_ARGS_ASSERT_SV_CATPV_MG;
4861 Creates a new SV. A non-zero C<len> parameter indicates the number of
4862 bytes of preallocated string space the SV should have. An extra byte for a
4863 trailing NUL is also reserved. (SvPOK is not set for the SV even if string
4864 space is allocated.) The reference count for the new SV is set to 1.
4866 In 5.9.3, newSV() replaces the older NEWSV() API, and drops the first
4867 parameter, I<x>, a debug aid which allowed callers to identify themselves.
4868 This aid has been superseded by a new build option, PERL_MEM_LOG (see
4869 L<perlhack/PERL_MEM_LOG>). The older API is still there for use in XS
4870 modules supporting older perls.
4876 Perl_newSV(pTHX_ const STRLEN len)
4883 sv_upgrade(sv, SVt_PV);
4884 SvGROW(sv, len + 1);
4889 =for apidoc sv_magicext
4891 Adds magic to an SV, upgrading it if necessary. Applies the
4892 supplied vtable and returns a pointer to the magic added.
4894 Note that C<sv_magicext> will allow things that C<sv_magic> will not.
4895 In particular, you can add magic to SvREADONLY SVs, and add more than
4896 one instance of the same 'how'.
4898 If C<namlen> is greater than zero then a C<savepvn> I<copy> of C<name> is
4899 stored, if C<namlen> is zero then C<name> is stored as-is and - as another
4900 special case - if C<(name && namlen == HEf_SVKEY)> then C<name> is assumed
4901 to contain an C<SV*> and is stored as-is with its REFCNT incremented.
4903 (This is now used as a subroutine by C<sv_magic>.)
4908 Perl_sv_magicext(pTHX_ SV *const sv, SV *const obj, const int how,
4909 const MGVTBL *const vtable, const char *const name, const I32 namlen)
4914 PERL_ARGS_ASSERT_SV_MAGICEXT;
4916 SvUPGRADE(sv, SVt_PVMG);
4917 Newxz(mg, 1, MAGIC);
4918 mg->mg_moremagic = SvMAGIC(sv);
4919 SvMAGIC_set(sv, mg);
4921 /* Sometimes a magic contains a reference loop, where the sv and
4922 object refer to each other. To prevent a reference loop that
4923 would prevent such objects being freed, we look for such loops
4924 and if we find one we avoid incrementing the object refcount.
4926 Note we cannot do this to avoid self-tie loops as intervening RV must
4927 have its REFCNT incremented to keep it in existence.
4930 if (!obj || obj == sv ||
4931 how == PERL_MAGIC_arylen ||
4932 how == PERL_MAGIC_symtab ||
4933 (SvTYPE(obj) == SVt_PVGV &&
4934 (GvSV(obj) == sv || GvHV(obj) == (const HV *)sv
4935 || GvAV(obj) == (const AV *)sv || GvCV(obj) == (const CV *)sv
4936 || GvIOp(obj) == (const IO *)sv || GvFORM(obj) == (const CV *)sv)))
4941 mg->mg_obj = SvREFCNT_inc_simple(obj);
4942 mg->mg_flags |= MGf_REFCOUNTED;
4945 /* Normal self-ties simply pass a null object, and instead of
4946 using mg_obj directly, use the SvTIED_obj macro to produce a
4947 new RV as needed. For glob "self-ties", we are tieing the PVIO
4948 with an RV obj pointing to the glob containing the PVIO. In
4949 this case, to avoid a reference loop, we need to weaken the
4953 if (how == PERL_MAGIC_tiedscalar && SvTYPE(sv) == SVt_PVIO &&
4954 obj && SvROK(obj) && GvIO(SvRV(obj)) == (const IO *)sv)
4960 mg->mg_len = namlen;
4963 mg->mg_ptr = savepvn(name, namlen);
4964 else if (namlen == HEf_SVKEY) {
4965 /* Yes, this is casting away const. This is only for the case of
4966 HEf_SVKEY. I think we need to document this abberation of the
4967 constness of the API, rather than making name non-const, as
4968 that change propagating outwards a long way. */
4969 mg->mg_ptr = (char*)SvREFCNT_inc_simple_NN((SV *)name);
4971 mg->mg_ptr = (char *) name;
4973 mg->mg_virtual = (MGVTBL *) vtable;
4977 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
4982 =for apidoc sv_magic
4984 Adds magic to an SV. First upgrades C<sv> to type C<SVt_PVMG> if necessary,
4985 then adds a new magic item of type C<how> to the head of the magic list.
4987 See C<sv_magicext> (which C<sv_magic> now calls) for a description of the
4988 handling of the C<name> and C<namlen> arguments.
4990 You need to use C<sv_magicext> to add magic to SvREADONLY SVs and also
4991 to add more than one instance of the same 'how'.
4997 Perl_sv_magic(pTHX_ register SV *const sv, SV *const obj, const int how,
4998 const char *const name, const I32 namlen)
5001 const MGVTBL *vtable;
5004 PERL_ARGS_ASSERT_SV_MAGIC;
5006 #ifdef PERL_OLD_COPY_ON_WRITE
5008 sv_force_normal_flags(sv, 0);
5010 if (SvREADONLY(sv)) {
5012 /* its okay to attach magic to shared strings; the subsequent
5013 * upgrade to PVMG will unshare the string */
5014 !(SvFAKE(sv) && SvTYPE(sv) < SVt_PVMG)
5017 && how != PERL_MAGIC_regex_global
5018 && how != PERL_MAGIC_bm
5019 && how != PERL_MAGIC_fm
5020 && how != PERL_MAGIC_sv
5021 && how != PERL_MAGIC_backref
5024 Perl_croak(aTHX_ "%s", PL_no_modify);
5027 if (SvMAGICAL(sv) || (how == PERL_MAGIC_taint && SvTYPE(sv) >= SVt_PVMG)) {
5028 if (SvMAGIC(sv) && (mg = mg_find(sv, how))) {
5029 /* sv_magic() refuses to add a magic of the same 'how' as an
5032 if (how == PERL_MAGIC_taint) {
5034 /* Any scalar which already had taint magic on which someone
5035 (erroneously?) did SvIOK_on() or similar will now be
5036 incorrectly sporting public "OK" flags. */
5037 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
5045 vtable = &PL_vtbl_sv;
5047 case PERL_MAGIC_overload:
5048 vtable = &PL_vtbl_amagic;
5050 case PERL_MAGIC_overload_elem:
5051 vtable = &PL_vtbl_amagicelem;
5053 case PERL_MAGIC_overload_table:
5054 vtable = &PL_vtbl_ovrld;
5057 vtable = &PL_vtbl_bm;
5059 case PERL_MAGIC_regdata:
5060 vtable = &PL_vtbl_regdata;
5062 case PERL_MAGIC_regdatum:
5063 vtable = &PL_vtbl_regdatum;
5065 case PERL_MAGIC_env:
5066 vtable = &PL_vtbl_env;
5069 vtable = &PL_vtbl_fm;
5071 case PERL_MAGIC_envelem:
5072 vtable = &PL_vtbl_envelem;
5074 case PERL_MAGIC_regex_global:
5075 vtable = &PL_vtbl_mglob;
5077 case PERL_MAGIC_isa:
5078 vtable = &PL_vtbl_isa;
5080 case PERL_MAGIC_isaelem:
5081 vtable = &PL_vtbl_isaelem;
5083 case PERL_MAGIC_nkeys:
5084 vtable = &PL_vtbl_nkeys;
5086 case PERL_MAGIC_dbfile:
5089 case PERL_MAGIC_dbline:
5090 vtable = &PL_vtbl_dbline;
5092 #ifdef USE_LOCALE_COLLATE
5093 case PERL_MAGIC_collxfrm:
5094 vtable = &PL_vtbl_collxfrm;
5096 #endif /* USE_LOCALE_COLLATE */
5097 case PERL_MAGIC_tied:
5098 vtable = &PL_vtbl_pack;
5100 case PERL_MAGIC_tiedelem:
5101 case PERL_MAGIC_tiedscalar:
5102 vtable = &PL_vtbl_packelem;
5105 vtable = &PL_vtbl_regexp;
5107 case PERL_MAGIC_hints:
5108 /* As this vtable is all NULL, we can reuse it. */
5109 case PERL_MAGIC_sig:
5110 vtable = &PL_vtbl_sig;
5112 case PERL_MAGIC_sigelem:
5113 vtable = &PL_vtbl_sigelem;
5115 case PERL_MAGIC_taint:
5116 vtable = &PL_vtbl_taint;
5118 case PERL_MAGIC_uvar:
5119 vtable = &PL_vtbl_uvar;
5121 case PERL_MAGIC_vec:
5122 vtable = &PL_vtbl_vec;
5124 case PERL_MAGIC_arylen_p:
5125 case PERL_MAGIC_rhash:
5126 case PERL_MAGIC_symtab:
5127 case PERL_MAGIC_vstring:
5130 case PERL_MAGIC_utf8:
5131 vtable = &PL_vtbl_utf8;
5133 case PERL_MAGIC_substr:
5134 vtable = &PL_vtbl_substr;
5136 case PERL_MAGIC_defelem:
5137 vtable = &PL_vtbl_defelem;
5139 case PERL_MAGIC_arylen:
5140 vtable = &PL_vtbl_arylen;
5142 case PERL_MAGIC_pos:
5143 vtable = &PL_vtbl_pos;
5145 case PERL_MAGIC_backref:
5146 vtable = &PL_vtbl_backref;
5148 case PERL_MAGIC_hintselem:
5149 vtable = &PL_vtbl_hintselem;
5151 case PERL_MAGIC_ext:
5152 /* Reserved for use by extensions not perl internals. */
5153 /* Useful for attaching extension internal data to perl vars. */
5154 /* Note that multiple extensions may clash if magical scalars */
5155 /* etc holding private data from one are passed to another. */
5159 Perl_croak(aTHX_ "Don't know how to handle magic of type \\%o", how);
5162 /* Rest of work is done else where */
5163 mg = sv_magicext(sv,obj,how,vtable,name,namlen);
5166 case PERL_MAGIC_taint:
5169 case PERL_MAGIC_ext:
5170 case PERL_MAGIC_dbfile:
5177 =for apidoc sv_unmagic
5179 Removes all magic of type C<type> from an SV.
5185 Perl_sv_unmagic(pTHX_ SV *const sv, const int type)
5190 PERL_ARGS_ASSERT_SV_UNMAGIC;
5192 if (SvTYPE(sv) < SVt_PVMG || !SvMAGIC(sv))
5194 mgp = &(((XPVMG*) SvANY(sv))->xmg_u.xmg_magic);
5195 for (mg = *mgp; mg; mg = *mgp) {
5196 if (mg->mg_type == type) {
5197 const MGVTBL* const vtbl = mg->mg_virtual;
5198 *mgp = mg->mg_moremagic;
5199 if (vtbl && vtbl->svt_free)
5200 CALL_FPTR(vtbl->svt_free)(aTHX_ sv, mg);
5201 if (mg->mg_ptr && mg->mg_type != PERL_MAGIC_regex_global) {
5203 Safefree(mg->mg_ptr);
5204 else if (mg->mg_len == HEf_SVKEY)
5205 SvREFCNT_dec(MUTABLE_SV(mg->mg_ptr));
5206 else if (mg->mg_type == PERL_MAGIC_utf8)
5207 Safefree(mg->mg_ptr);
5209 if (mg->mg_flags & MGf_REFCOUNTED)
5210 SvREFCNT_dec(mg->mg_obj);
5214 mgp = &mg->mg_moremagic;
5218 SvFLAGS(sv) |= (SvFLAGS(sv) & (SVp_IOK|SVp_NOK|SVp_POK)) >> PRIVSHIFT;
5219 SvMAGIC_set(sv, NULL);
5226 =for apidoc sv_rvweaken
5228 Weaken a reference: set the C<SvWEAKREF> flag on this RV; give the
5229 referred-to SV C<PERL_MAGIC_backref> magic if it hasn't already; and
5230 push a back-reference to this RV onto the array of backreferences
5231 associated with that magic. If the RV is magical, set magic will be
5232 called after the RV is cleared.
5238 Perl_sv_rvweaken(pTHX_ SV *const sv)
5242 PERL_ARGS_ASSERT_SV_RVWEAKEN;
5244 if (!SvOK(sv)) /* let undefs pass */
5247 Perl_croak(aTHX_ "Can't weaken a nonreference");
5248 else if (SvWEAKREF(sv)) {
5249 if (ckWARN(WARN_MISC))
5250 Perl_warner(aTHX_ packWARN(WARN_MISC), "Reference is already weak");
5254 Perl_sv_add_backref(aTHX_ tsv, sv);
5260 /* Give tsv backref magic if it hasn't already got it, then push a
5261 * back-reference to sv onto the array associated with the backref magic.
5264 /* A discussion about the backreferences array and its refcount:
5266 * The AV holding the backreferences is pointed to either as the mg_obj of
5267 * PERL_MAGIC_backref, or in the specific case of a HV that has the hv_aux
5268 * structure, from the xhv_backreferences field. (A HV without hv_aux will
5269 * have the standard magic instead.) The array is created with a refcount
5270 * of 2. This means that if during global destruction the array gets
5271 * picked on first to have its refcount decremented by the random zapper,
5272 * it won't actually be freed, meaning it's still theere for when its
5273 * parent gets freed.
5274 * When the parent SV is freed, in the case of magic, the magic is freed,
5275 * Perl_magic_killbackrefs is called which decrements one refcount, then
5276 * mg_obj is freed which kills the second count.
5277 * In the vase of a HV being freed, one ref is removed by
5278 * Perl_hv_kill_backrefs, the other by Perl_sv_kill_backrefs, which it
5283 Perl_sv_add_backref(pTHX_ SV *const tsv, SV *const sv)
5288 PERL_ARGS_ASSERT_SV_ADD_BACKREF;
5290 if (SvTYPE(tsv) == SVt_PVHV) {
5291 AV **const avp = Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5295 /* There is no AV in the offical place - try a fixup. */
5296 MAGIC *const mg = mg_find(tsv, PERL_MAGIC_backref);
5299 /* Aha. They've got it stowed in magic. Bring it back. */
5300 av = MUTABLE_AV(mg->mg_obj);
5301 /* Stop mg_free decreasing the refernce count. */
5303 /* Stop mg_free even calling the destructor, given that
5304 there's no AV to free up. */
5306 sv_unmagic(tsv, PERL_MAGIC_backref);
5310 SvREFCNT_inc_simple_void(av); /* see discussion above */
5315 const MAGIC *const mg
5316 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
5318 av = MUTABLE_AV(mg->mg_obj);
5322 sv_magic(tsv, MUTABLE_SV(av), PERL_MAGIC_backref, NULL, 0);
5323 /* av now has a refcnt of 2; see discussion above */
5326 if (AvFILLp(av) >= AvMAX(av)) {
5327 av_extend(av, AvFILLp(av)+1);
5329 AvARRAY(av)[++AvFILLp(av)] = sv; /* av_push() */
5332 /* delete a back-reference to ourselves from the backref magic associated
5333 * with the SV we point to.
5337 S_sv_del_backref(pTHX_ SV *const tsv, SV *const sv)
5344 PERL_ARGS_ASSERT_SV_DEL_BACKREF;
5346 if (SvTYPE(tsv) == SVt_PVHV && SvOOK(tsv)) {
5347 av = *Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5348 /* We mustn't attempt to "fix up" the hash here by moving the
5349 backreference array back to the hv_aux structure, as that is stored
5350 in the main HvARRAY(), and hfreentries assumes that no-one
5351 reallocates HvARRAY() while it is running. */
5354 const MAGIC *const mg
5355 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
5357 av = MUTABLE_AV(mg->mg_obj);
5361 Perl_croak(aTHX_ "panic: del_backref");
5363 assert(!SvIS_FREED(av));
5366 /* We shouldn't be in here more than once, but for paranoia reasons lets
5368 for (i = AvFILLp(av); i >= 0; i--) {
5370 const SSize_t fill = AvFILLp(av);
5372 /* We weren't the last entry.
5373 An unordered list has this property that you can take the
5374 last element off the end to fill the hole, and it's still
5375 an unordered list :-)
5380 AvFILLp(av) = fill - 1;
5386 Perl_sv_kill_backrefs(pTHX_ SV *const sv, AV *const av)
5388 SV **svp = AvARRAY(av);
5390 PERL_ARGS_ASSERT_SV_KILL_BACKREFS;
5391 PERL_UNUSED_ARG(sv);
5393 assert(!svp || !SvIS_FREED(av));
5395 SV *const *const last = svp + AvFILLp(av);
5397 while (svp <= last) {
5399 SV *const referrer = *svp;
5400 if (SvWEAKREF(referrer)) {
5401 /* XXX Should we check that it hasn't changed? */
5402 SvRV_set(referrer, 0);
5404 SvWEAKREF_off(referrer);
5405 SvSETMAGIC(referrer);
5406 } else if (SvTYPE(referrer) == SVt_PVGV ||
5407 SvTYPE(referrer) == SVt_PVLV) {
5408 /* You lookin' at me? */
5409 assert(GvSTASH(referrer));
5410 assert(GvSTASH(referrer) == (const HV *)sv);
5411 GvSTASH(referrer) = 0;
5414 "panic: magic_killbackrefs (flags=%"UVxf")",
5415 (UV)SvFLAGS(referrer));
5423 SvREFCNT_dec(av); /* remove extra count added by sv_add_backref() */
5428 =for apidoc sv_insert
5430 Inserts a string at the specified offset/length within the SV. Similar to
5431 the Perl substr() function. Handles get magic.
5433 =for apidoc sv_insert_flags
5435 Same as C<sv_insert>, but the extra C<flags> are passed the C<SvPV_force_flags> that applies to C<bigstr>.
5441 Perl_sv_insert_flags(pTHX_ SV *const bigstr, const STRLEN offset, const STRLEN len, const char *const little, const STRLEN littlelen, const U32 flags)
5446 register char *midend;
5447 register char *bigend;
5451 PERL_ARGS_ASSERT_SV_INSERT_FLAGS;
5454 Perl_croak(aTHX_ "Can't modify non-existent substring");
5455 SvPV_force_flags(bigstr, curlen, flags);
5456 (void)SvPOK_only_UTF8(bigstr);
5457 if (offset + len > curlen) {
5458 SvGROW(bigstr, offset+len+1);
5459 Zero(SvPVX(bigstr)+curlen, offset+len-curlen, char);
5460 SvCUR_set(bigstr, offset+len);
5464 i = littlelen - len;
5465 if (i > 0) { /* string might grow */
5466 big = SvGROW(bigstr, SvCUR(bigstr) + i + 1);
5467 mid = big + offset + len;
5468 midend = bigend = big + SvCUR(bigstr);
5471 while (midend > mid) /* shove everything down */
5472 *--bigend = *--midend;
5473 Move(little,big+offset,littlelen,char);
5474 SvCUR_set(bigstr, SvCUR(bigstr) + i);
5479 Move(little,SvPVX(bigstr)+offset,len,char);
5484 big = SvPVX(bigstr);
5487 bigend = big + SvCUR(bigstr);
5489 if (midend > bigend)
5490 Perl_croak(aTHX_ "panic: sv_insert");
5492 if (mid - big > bigend - midend) { /* faster to shorten from end */
5494 Move(little, mid, littlelen,char);
5497 i = bigend - midend;
5499 Move(midend, mid, i,char);
5503 SvCUR_set(bigstr, mid - big);
5505 else if ((i = mid - big)) { /* faster from front */
5506 midend -= littlelen;
5508 Move(big, midend - i, i, char);
5509 sv_chop(bigstr,midend-i);
5511 Move(little, mid, littlelen,char);
5513 else if (littlelen) {
5514 midend -= littlelen;
5515 sv_chop(bigstr,midend);
5516 Move(little,midend,littlelen,char);
5519 sv_chop(bigstr,midend);
5525 =for apidoc sv_replace
5527 Make the first argument a copy of the second, then delete the original.
5528 The target SV physically takes over ownership of the body of the source SV
5529 and inherits its flags; however, the target keeps any magic it owns,
5530 and any magic in the source is discarded.
5531 Note that this is a rather specialist SV copying operation; most of the
5532 time you'll want to use C<sv_setsv> or one of its many macro front-ends.
5538 Perl_sv_replace(pTHX_ register SV *const sv, register SV *const nsv)
5541 const U32 refcnt = SvREFCNT(sv);
5543 PERL_ARGS_ASSERT_SV_REPLACE;
5545 SV_CHECK_THINKFIRST_COW_DROP(sv);
5546 if (SvREFCNT(nsv) != 1) {
5547 Perl_croak(aTHX_ "panic: reference miscount on nsv in sv_replace() (%"
5548 UVuf " != 1)", (UV) SvREFCNT(nsv));
5550 if (SvMAGICAL(sv)) {
5554 sv_upgrade(nsv, SVt_PVMG);
5555 SvMAGIC_set(nsv, SvMAGIC(sv));
5556 SvFLAGS(nsv) |= SvMAGICAL(sv);
5558 SvMAGIC_set(sv, NULL);
5562 assert(!SvREFCNT(sv));
5563 #ifdef DEBUG_LEAKING_SCALARS
5564 sv->sv_flags = nsv->sv_flags;
5565 sv->sv_any = nsv->sv_any;
5566 sv->sv_refcnt = nsv->sv_refcnt;
5567 sv->sv_u = nsv->sv_u;
5569 StructCopy(nsv,sv,SV);
5571 if(SvTYPE(sv) == SVt_IV) {
5573 = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
5577 #ifdef PERL_OLD_COPY_ON_WRITE
5578 if (SvIsCOW_normal(nsv)) {
5579 /* We need to follow the pointers around the loop to make the
5580 previous SV point to sv, rather than nsv. */
5583 while ((next = SV_COW_NEXT_SV(current)) != nsv) {
5586 assert(SvPVX_const(current) == SvPVX_const(nsv));
5588 /* Make the SV before us point to the SV after us. */
5590 PerlIO_printf(Perl_debug_log, "previous is\n");
5592 PerlIO_printf(Perl_debug_log,
5593 "move it from 0x%"UVxf" to 0x%"UVxf"\n",
5594 (UV) SV_COW_NEXT_SV(current), (UV) sv);
5596 SV_COW_NEXT_SV_SET(current, sv);
5599 SvREFCNT(sv) = refcnt;
5600 SvFLAGS(nsv) |= SVTYPEMASK; /* Mark as freed */
5606 =for apidoc sv_clear
5608 Clear an SV: call any destructors, free up any memory used by the body,
5609 and free the body itself. The SV's head is I<not> freed, although
5610 its type is set to all 1's so that it won't inadvertently be assumed
5611 to be live during global destruction etc.
5612 This function should only be called when REFCNT is zero. Most of the time
5613 you'll want to call C<sv_free()> (or its macro wrapper C<SvREFCNT_dec>)
5620 Perl_sv_clear(pTHX_ register SV *const sv)
5623 const U32 type = SvTYPE(sv);
5624 const struct body_details *const sv_type_details
5625 = bodies_by_type + type;
5628 PERL_ARGS_ASSERT_SV_CLEAR;
5629 assert(SvREFCNT(sv) == 0);
5630 assert(SvTYPE(sv) != SVTYPEMASK);
5632 if (type <= SVt_IV) {
5633 /* See the comment in sv.h about the collusion between this early
5634 return and the overloading of the NULL and IV slots in the size
5637 SV * const target = SvRV(sv);
5639 sv_del_backref(target, sv);
5641 SvREFCNT_dec(target);
5643 SvFLAGS(sv) &= SVf_BREAK;
5644 SvFLAGS(sv) |= SVTYPEMASK;
5649 if (PL_defstash && /* Still have a symbol table? */
5656 stash = SvSTASH(sv);
5657 destructor = StashHANDLER(stash,DESTROY);
5659 /* Don't bother calling an empty destructor */
5660 && (CvISXSUB(destructor)
5661 || CvSTART(destructor)->op_next->op_type != OP_LEAVESUB))
5663 SV* const tmpref = newRV(sv);
5664 SvREADONLY_on(tmpref); /* DESTROY() could be naughty */
5666 PUSHSTACKi(PERLSI_DESTROY);
5671 call_sv(MUTABLE_SV(destructor), G_DISCARD|G_EVAL|G_KEEPERR|G_VOID);
5677 if(SvREFCNT(tmpref) < 2) {
5678 /* tmpref is not kept alive! */
5680 SvRV_set(tmpref, NULL);
5683 SvREFCNT_dec(tmpref);
5685 } while (SvOBJECT(sv) && SvSTASH(sv) != stash);
5689 if (PL_in_clean_objs)
5690 Perl_croak(aTHX_ "DESTROY created new reference to dead object '%s'",
5692 /* DESTROY gave object new lease on life */
5698 SvREFCNT_dec(SvSTASH(sv)); /* possibly of changed persuasion */
5699 SvOBJECT_off(sv); /* Curse the object. */
5700 if (type != SVt_PVIO)
5701 --PL_sv_objcount; /* XXX Might want something more general */
5704 if (type >= SVt_PVMG) {
5705 if (type == SVt_PVMG && SvPAD_OUR(sv)) {
5706 SvREFCNT_dec(SvOURSTASH(sv));
5707 } else if (SvMAGIC(sv))
5709 if (type == SVt_PVMG && SvPAD_TYPED(sv))
5710 SvREFCNT_dec(SvSTASH(sv));
5713 /* case SVt_BIND: */
5716 IoIFP(sv) != PerlIO_stdin() &&
5717 IoIFP(sv) != PerlIO_stdout() &&
5718 IoIFP(sv) != PerlIO_stderr())
5720 io_close(MUTABLE_IO(sv), FALSE);
5722 if (IoDIRP(sv) && !(IoFLAGS(sv) & IOf_FAKE_DIRP))
5723 PerlDir_close(IoDIRP(sv));
5724 IoDIRP(sv) = (DIR*)NULL;
5725 Safefree(IoTOP_NAME(sv));
5726 Safefree(IoFMT_NAME(sv));
5727 Safefree(IoBOTTOM_NAME(sv));
5730 /* FIXME for plugins */
5731 pregfree2((REGEXP*) sv);
5735 cv_undef(MUTABLE_CV(sv));
5738 if (PL_last_swash_hv == (const HV *)sv) {
5739 PL_last_swash_hv = NULL;
5741 Perl_hv_kill_backrefs(aTHX_ MUTABLE_HV(sv));
5742 hv_undef(MUTABLE_HV(sv));
5745 if (PL_comppad == MUTABLE_AV(sv)) {
5749 av_undef(MUTABLE_AV(sv));
5752 if (LvTYPE(sv) == 'T') { /* for tie: return HE to pool */
5753 SvREFCNT_dec(HeKEY_sv((HE*)LvTARG(sv)));
5754 HeNEXT((HE*)LvTARG(sv)) = PL_hv_fetch_ent_mh;
5755 PL_hv_fetch_ent_mh = (HE*)LvTARG(sv);
5757 else if (LvTYPE(sv) != 't') /* unless tie: unrefcnted fake SV** */
5758 SvREFCNT_dec(LvTARG(sv));
5760 if (isGV_with_GP(sv)) {
5761 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
5762 && HvNAME_get(stash))
5763 mro_method_changed_in(stash);
5764 gp_free(MUTABLE_GV(sv));
5766 unshare_hek(GvNAME_HEK(sv));
5767 /* If we're in a stash, we don't own a reference to it. However it does
5768 have a back reference to us, which needs to be cleared. */
5769 if (!SvVALID(sv) && (stash = GvSTASH(sv)))
5770 sv_del_backref(MUTABLE_SV(stash), sv);
5772 /* FIXME. There are probably more unreferenced pointers to SVs in the
5773 interpreter struct that we should check and tidy in a similar
5775 if ((const GV *)sv == PL_last_in_gv)
5776 PL_last_in_gv = NULL;
5782 /* Don't bother with SvOOK_off(sv); as we're only going to free it. */
5785 SvOOK_offset(sv, offset);
5786 SvPV_set(sv, SvPVX_mutable(sv) - offset);
5787 /* Don't even bother with turning off the OOK flag. */
5790 SV * const target = SvRV(sv);
5792 sv_del_backref(target, sv);
5794 SvREFCNT_dec(target);
5796 #ifdef PERL_OLD_COPY_ON_WRITE
5797 else if (SvPVX_const(sv)) {
5799 /* I believe I need to grab the global SV mutex here and
5800 then recheck the COW status. */
5802 PerlIO_printf(Perl_debug_log, "Copy on write: clear\n");
5806 sv_release_COW(sv, SvPVX_const(sv), SV_COW_NEXT_SV(sv));
5808 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
5811 /* And drop it here. */
5813 } else if (SvLEN(sv)) {
5814 Safefree(SvPVX_const(sv));
5818 else if (SvPVX_const(sv) && SvLEN(sv))
5819 Safefree(SvPVX_mutable(sv));
5820 else if (SvPVX_const(sv) && SvREADONLY(sv) && SvFAKE(sv)) {
5821 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
5830 SvFLAGS(sv) &= SVf_BREAK;
5831 SvFLAGS(sv) |= SVTYPEMASK;
5833 if (sv_type_details->arena) {
5834 del_body(((char *)SvANY(sv) + sv_type_details->offset),
5835 &PL_body_roots[type]);
5837 else if (sv_type_details->body_size) {
5838 my_safefree(SvANY(sv));
5843 =for apidoc sv_newref
5845 Increment an SV's reference count. Use the C<SvREFCNT_inc()> wrapper
5852 Perl_sv_newref(pTHX_ SV *const sv)
5854 PERL_UNUSED_CONTEXT;
5863 Decrement an SV's reference count, and if it drops to zero, call
5864 C<sv_clear> to invoke destructors and free up any memory used by
5865 the body; finally, deallocate the SV's head itself.
5866 Normally called via a wrapper macro C<SvREFCNT_dec>.
5872 Perl_sv_free(pTHX_ SV *const sv)
5877 if (SvREFCNT(sv) == 0) {
5878 if (SvFLAGS(sv) & SVf_BREAK)
5879 /* this SV's refcnt has been artificially decremented to
5880 * trigger cleanup */
5882 if (PL_in_clean_all) /* All is fair */
5884 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
5885 /* make sure SvREFCNT(sv)==0 happens very seldom */
5886 SvREFCNT(sv) = (~(U32)0)/2;
5889 if (ckWARN_d(WARN_INTERNAL)) {
5890 #ifdef DEBUG_LEAKING_SCALARS_FORK_DUMP
5891 Perl_dump_sv_child(aTHX_ sv);
5893 #ifdef DEBUG_LEAKING_SCALARS
5896 #ifdef DEBUG_LEAKING_SCALARS_ABORT
5897 if (PL_warnhook == PERL_WARNHOOK_FATAL
5898 || ckDEAD(packWARN(WARN_INTERNAL))) {
5899 /* Don't let Perl_warner cause us to escape our fate: */
5903 /* This may not return: */
5904 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
5905 "Attempt to free unreferenced scalar: SV 0x%"UVxf
5906 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
5909 #ifdef DEBUG_LEAKING_SCALARS_ABORT
5914 if (--(SvREFCNT(sv)) > 0)
5916 Perl_sv_free2(aTHX_ sv);
5920 Perl_sv_free2(pTHX_ SV *const sv)
5924 PERL_ARGS_ASSERT_SV_FREE2;
5928 if (ckWARN_d(WARN_DEBUGGING))
5929 Perl_warner(aTHX_ packWARN(WARN_DEBUGGING),
5930 "Attempt to free temp prematurely: SV 0x%"UVxf
5931 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
5935 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
5936 /* make sure SvREFCNT(sv)==0 happens very seldom */
5937 SvREFCNT(sv) = (~(U32)0)/2;
5948 Returns the length of the string in the SV. Handles magic and type
5949 coercion. See also C<SvCUR>, which gives raw access to the xpv_cur slot.
5955 Perl_sv_len(pTHX_ register SV *const sv)
5963 len = mg_length(sv);
5965 (void)SvPV_const(sv, len);
5970 =for apidoc sv_len_utf8
5972 Returns the number of characters in the string in an SV, counting wide
5973 UTF-8 bytes as a single character. Handles magic and type coercion.
5979 * The length is cached in PERL_MAGIC_utf8, in the mg_len field. Also the
5980 * mg_ptr is used, by sv_pos_u2b() and sv_pos_b2u() - see the comments below.
5981 * (Note that the mg_len is not the length of the mg_ptr field.
5982 * This allows the cache to store the character length of the string without
5983 * needing to malloc() extra storage to attach to the mg_ptr.)
5988 Perl_sv_len_utf8(pTHX_ register SV *const sv)
5994 return mg_length(sv);
5998 const U8 *s = (U8*)SvPV_const(sv, len);
6002 MAGIC *mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_utf8) : NULL;
6004 if (mg && mg->mg_len != -1) {
6006 if (PL_utf8cache < 0) {
6007 const STRLEN real = Perl_utf8_length(aTHX_ s, s + len);
6009 /* Need to turn the assertions off otherwise we may
6010 recurse infinitely while printing error messages.
6012 SAVEI8(PL_utf8cache);
6014 Perl_croak(aTHX_ "panic: sv_len_utf8 cache %"UVuf
6015 " real %"UVuf" for %"SVf,
6016 (UV) ulen, (UV) real, SVfARG(sv));
6021 ulen = Perl_utf8_length(aTHX_ s, s + len);
6022 if (!SvREADONLY(sv)) {
6024 mg = sv_magicext(sv, 0, PERL_MAGIC_utf8,
6025 &PL_vtbl_utf8, 0, 0);
6033 return Perl_utf8_length(aTHX_ s, s + len);
6037 /* Walk forwards to find the byte corresponding to the passed in UTF-8
6040 S_sv_pos_u2b_forwards(const U8 *const start, const U8 *const send,
6043 const U8 *s = start;
6045 PERL_ARGS_ASSERT_SV_POS_U2B_FORWARDS;
6047 while (s < send && uoffset--)
6050 /* This is the existing behaviour. Possibly it should be a croak, as
6051 it's actually a bounds error */
6057 /* Given the length of the string in both bytes and UTF-8 characters, decide
6058 whether to walk forwards or backwards to find the byte corresponding to
6059 the passed in UTF-8 offset. */
6061 S_sv_pos_u2b_midway(const U8 *const start, const U8 *send,
6062 const STRLEN uoffset, const STRLEN uend)
6064 STRLEN backw = uend - uoffset;
6066 PERL_ARGS_ASSERT_SV_POS_U2B_MIDWAY;
6068 if (uoffset < 2 * backw) {
6069 /* The assumption is that going forwards is twice the speed of going
6070 forward (that's where the 2 * backw comes from).
6071 (The real figure of course depends on the UTF-8 data.) */
6072 return sv_pos_u2b_forwards(start, send, uoffset);
6077 while (UTF8_IS_CONTINUATION(*send))
6080 return send - start;
6083 /* For the string representation of the given scalar, find the byte
6084 corresponding to the passed in UTF-8 offset. uoffset0 and boffset0
6085 give another position in the string, *before* the sought offset, which
6086 (which is always true, as 0, 0 is a valid pair of positions), which should
6087 help reduce the amount of linear searching.
6088 If *mgp is non-NULL, it should point to the UTF-8 cache magic, which
6089 will be used to reduce the amount of linear searching. The cache will be
6090 created if necessary, and the found value offered to it for update. */
6092 S_sv_pos_u2b_cached(pTHX_ SV *const sv, MAGIC **const mgp, const U8 *const start,
6093 const U8 *const send, const STRLEN uoffset,
6094 STRLEN uoffset0, STRLEN boffset0)
6096 STRLEN boffset = 0; /* Actually always set, but let's keep gcc happy. */
6099 PERL_ARGS_ASSERT_SV_POS_U2B_CACHED;
6101 assert (uoffset >= uoffset0);
6103 if (SvMAGICAL(sv) && !SvREADONLY(sv) && PL_utf8cache
6104 && (*mgp || (*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
6105 if ((*mgp)->mg_ptr) {
6106 STRLEN *cache = (STRLEN *) (*mgp)->mg_ptr;
6107 if (cache[0] == uoffset) {
6108 /* An exact match. */
6111 if (cache[2] == uoffset) {
6112 /* An exact match. */
6116 if (cache[0] < uoffset) {
6117 /* The cache already knows part of the way. */
6118 if (cache[0] > uoffset0) {
6119 /* The cache knows more than the passed in pair */
6120 uoffset0 = cache[0];
6121 boffset0 = cache[1];
6123 if ((*mgp)->mg_len != -1) {
6124 /* And we know the end too. */
6126 + sv_pos_u2b_midway(start + boffset0, send,
6128 (*mgp)->mg_len - uoffset0);
6131 + sv_pos_u2b_forwards(start + boffset0,
6132 send, uoffset - uoffset0);
6135 else if (cache[2] < uoffset) {
6136 /* We're between the two cache entries. */
6137 if (cache[2] > uoffset0) {
6138 /* and the cache knows more than the passed in pair */
6139 uoffset0 = cache[2];
6140 boffset0 = cache[3];
6144 + sv_pos_u2b_midway(start + boffset0,
6147 cache[0] - uoffset0);
6150 + sv_pos_u2b_midway(start + boffset0,
6153 cache[2] - uoffset0);
6157 else if ((*mgp)->mg_len != -1) {
6158 /* If we can take advantage of a passed in offset, do so. */
6159 /* In fact, offset0 is either 0, or less than offset, so don't
6160 need to worry about the other possibility. */
6162 + sv_pos_u2b_midway(start + boffset0, send,
6164 (*mgp)->mg_len - uoffset0);
6169 if (!found || PL_utf8cache < 0) {
6170 const STRLEN real_boffset
6171 = boffset0 + sv_pos_u2b_forwards(start + boffset0,
6172 send, uoffset - uoffset0);
6174 if (found && PL_utf8cache < 0) {
6175 if (real_boffset != boffset) {
6176 /* Need to turn the assertions off otherwise we may recurse
6177 infinitely while printing error messages. */
6178 SAVEI8(PL_utf8cache);
6180 Perl_croak(aTHX_ "panic: sv_pos_u2b_cache cache %"UVuf
6181 " real %"UVuf" for %"SVf,
6182 (UV) boffset, (UV) real_boffset, SVfARG(sv));
6185 boffset = real_boffset;
6189 utf8_mg_pos_cache_update(sv, mgp, boffset, uoffset, send - start);
6195 =for apidoc sv_pos_u2b
6197 Converts the value pointed to by offsetp from a count of UTF-8 chars from
6198 the start of the string, to a count of the equivalent number of bytes; if
6199 lenp is non-zero, it does the same to lenp, but this time starting from
6200 the offset, rather than from the start of the string. Handles magic and
6207 * sv_pos_u2b() uses, like sv_pos_b2u(), the mg_ptr of the potential
6208 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6209 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
6214 Perl_sv_pos_u2b(pTHX_ register SV *const sv, I32 *const offsetp, I32 *const lenp)
6219 PERL_ARGS_ASSERT_SV_POS_U2B;
6224 start = (U8*)SvPV_const(sv, len);
6226 STRLEN uoffset = (STRLEN) *offsetp;
6227 const U8 * const send = start + len;
6229 const STRLEN boffset = sv_pos_u2b_cached(sv, &mg, start, send,
6232 *offsetp = (I32) boffset;
6235 /* Convert the relative offset to absolute. */
6236 const STRLEN uoffset2 = uoffset + (STRLEN) *lenp;
6237 const STRLEN boffset2
6238 = sv_pos_u2b_cached(sv, &mg, start, send, uoffset2,
6239 uoffset, boffset) - boffset;
6253 /* Create and update the UTF8 magic offset cache, with the proffered utf8/
6254 byte length pairing. The (byte) length of the total SV is passed in too,
6255 as blen, because for some (more esoteric) SVs, the call to SvPV_const()
6256 may not have updated SvCUR, so we can't rely on reading it directly.
6258 The proffered utf8/byte length pairing isn't used if the cache already has
6259 two pairs, and swapping either for the proffered pair would increase the
6260 RMS of the intervals between known byte offsets.
6262 The cache itself consists of 4 STRLEN values
6263 0: larger UTF-8 offset
6264 1: corresponding byte offset
6265 2: smaller UTF-8 offset
6266 3: corresponding byte offset
6268 Unused cache pairs have the value 0, 0.
6269 Keeping the cache "backwards" means that the invariant of
6270 cache[0] >= cache[2] is maintained even with empty slots, which means that
6271 the code that uses it doesn't need to worry if only 1 entry has actually
6272 been set to non-zero. It also makes the "position beyond the end of the
6273 cache" logic much simpler, as the first slot is always the one to start
6277 S_utf8_mg_pos_cache_update(pTHX_ SV *const sv, MAGIC **const mgp, const STRLEN byte,
6278 const STRLEN utf8, const STRLEN blen)
6282 PERL_ARGS_ASSERT_UTF8_MG_POS_CACHE_UPDATE;
6288 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, (MGVTBL*)&PL_vtbl_utf8, 0,
6290 (*mgp)->mg_len = -1;
6294 if (!(cache = (STRLEN *)(*mgp)->mg_ptr)) {
6295 Newxz(cache, PERL_MAGIC_UTF8_CACHESIZE * 2, STRLEN);
6296 (*mgp)->mg_ptr = (char *) cache;
6300 if (PL_utf8cache < 0) {
6301 const U8 *start = (const U8 *) SvPVX_const(sv);
6302 const STRLEN realutf8 = utf8_length(start, start + byte);
6304 if (realutf8 != utf8) {
6305 /* Need to turn the assertions off otherwise we may recurse
6306 infinitely while printing error messages. */
6307 SAVEI8(PL_utf8cache);
6309 Perl_croak(aTHX_ "panic: utf8_mg_pos_cache_update cache %"UVuf
6310 " real %"UVuf" for %"SVf, (UV) utf8, (UV) realutf8, SVfARG(sv));
6314 /* Cache is held with the later position first, to simplify the code
6315 that deals with unbounded ends. */
6317 ASSERT_UTF8_CACHE(cache);
6318 if (cache[1] == 0) {
6319 /* Cache is totally empty */
6322 } else if (cache[3] == 0) {
6323 if (byte > cache[1]) {
6324 /* New one is larger, so goes first. */
6325 cache[2] = cache[0];
6326 cache[3] = cache[1];
6334 #define THREEWAY_SQUARE(a,b,c,d) \
6335 ((float)((d) - (c))) * ((float)((d) - (c))) \
6336 + ((float)((c) - (b))) * ((float)((c) - (b))) \
6337 + ((float)((b) - (a))) * ((float)((b) - (a)))
6339 /* Cache has 2 slots in use, and we know three potential pairs.
6340 Keep the two that give the lowest RMS distance. Do the
6341 calcualation in bytes simply because we always know the byte
6342 length. squareroot has the same ordering as the positive value,
6343 so don't bother with the actual square root. */
6344 const float existing = THREEWAY_SQUARE(0, cache[3], cache[1], blen);
6345 if (byte > cache[1]) {
6346 /* New position is after the existing pair of pairs. */
6347 const float keep_earlier
6348 = THREEWAY_SQUARE(0, cache[3], byte, blen);
6349 const float keep_later
6350 = THREEWAY_SQUARE(0, cache[1], byte, blen);
6352 if (keep_later < keep_earlier) {
6353 if (keep_later < existing) {
6354 cache[2] = cache[0];
6355 cache[3] = cache[1];
6361 if (keep_earlier < existing) {
6367 else if (byte > cache[3]) {
6368 /* New position is between the existing pair of pairs. */
6369 const float keep_earlier
6370 = THREEWAY_SQUARE(0, cache[3], byte, blen);
6371 const float keep_later
6372 = THREEWAY_SQUARE(0, byte, cache[1], blen);
6374 if (keep_later < keep_earlier) {
6375 if (keep_later < existing) {
6381 if (keep_earlier < existing) {
6388 /* New position is before the existing pair of pairs. */
6389 const float keep_earlier
6390 = THREEWAY_SQUARE(0, byte, cache[3], blen);
6391 const float keep_later
6392 = THREEWAY_SQUARE(0, byte, cache[1], blen);
6394 if (keep_later < keep_earlier) {
6395 if (keep_later < existing) {
6401 if (keep_earlier < existing) {
6402 cache[0] = cache[2];
6403 cache[1] = cache[3];
6410 ASSERT_UTF8_CACHE(cache);
6413 /* We already know all of the way, now we may be able to walk back. The same
6414 assumption is made as in S_sv_pos_u2b_midway(), namely that walking
6415 backward is half the speed of walking forward. */
6417 S_sv_pos_b2u_midway(pTHX_ const U8 *const s, const U8 *const target,
6418 const U8 *end, STRLEN endu)
6420 const STRLEN forw = target - s;
6421 STRLEN backw = end - target;
6423 PERL_ARGS_ASSERT_SV_POS_B2U_MIDWAY;
6425 if (forw < 2 * backw) {
6426 return utf8_length(s, target);
6429 while (end > target) {
6431 while (UTF8_IS_CONTINUATION(*end)) {
6440 =for apidoc sv_pos_b2u
6442 Converts the value pointed to by offsetp from a count of bytes from the
6443 start of the string, to a count of the equivalent number of UTF-8 chars.
6444 Handles magic and type coercion.
6450 * sv_pos_b2u() uses, like sv_pos_u2b(), the mg_ptr of the potential
6451 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6456 Perl_sv_pos_b2u(pTHX_ register SV *const sv, I32 *const offsetp)
6459 const STRLEN byte = *offsetp;
6460 STRLEN len = 0; /* Actually always set, but let's keep gcc happy. */
6466 PERL_ARGS_ASSERT_SV_POS_B2U;
6471 s = (const U8*)SvPV_const(sv, blen);
6474 Perl_croak(aTHX_ "panic: sv_pos_b2u: bad byte offset");
6478 if (SvMAGICAL(sv) && !SvREADONLY(sv) && PL_utf8cache
6479 && (mg = mg_find(sv, PERL_MAGIC_utf8))) {
6481 STRLEN * const cache = (STRLEN *) mg->mg_ptr;
6482 if (cache[1] == byte) {
6483 /* An exact match. */
6484 *offsetp = cache[0];
6487 if (cache[3] == byte) {
6488 /* An exact match. */
6489 *offsetp = cache[2];
6493 if (cache[1] < byte) {
6494 /* We already know part of the way. */
6495 if (mg->mg_len != -1) {
6496 /* Actually, we know the end too. */
6498 + S_sv_pos_b2u_midway(aTHX_ s + cache[1], send,
6499 s + blen, mg->mg_len - cache[0]);
6501 len = cache[0] + utf8_length(s + cache[1], send);
6504 else if (cache[3] < byte) {
6505 /* We're between the two cached pairs, so we do the calculation
6506 offset by the byte/utf-8 positions for the earlier pair,
6507 then add the utf-8 characters from the string start to
6509 len = S_sv_pos_b2u_midway(aTHX_ s + cache[3], send,
6510 s + cache[1], cache[0] - cache[2])
6514 else { /* cache[3] > byte */
6515 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + cache[3],
6519 ASSERT_UTF8_CACHE(cache);
6521 } else if (mg->mg_len != -1) {
6522 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + blen, mg->mg_len);
6526 if (!found || PL_utf8cache < 0) {
6527 const STRLEN real_len = utf8_length(s, send);
6529 if (found && PL_utf8cache < 0) {
6530 if (len != real_len) {
6531 /* Need to turn the assertions off otherwise we may recurse
6532 infinitely while printing error messages. */
6533 SAVEI8(PL_utf8cache);
6535 Perl_croak(aTHX_ "panic: sv_pos_b2u cache %"UVuf
6536 " real %"UVuf" for %"SVf,
6537 (UV) len, (UV) real_len, SVfARG(sv));
6545 utf8_mg_pos_cache_update(sv, &mg, byte, len, blen);
6551 Returns a boolean indicating whether the strings in the two SVs are
6552 identical. Is UTF-8 and 'use bytes' aware, handles get magic, and will
6553 coerce its args to strings if necessary.
6559 Perl_sv_eq(pTHX_ register SV *sv1, register SV *sv2)
6568 SV* svrecode = NULL;
6575 /* if pv1 and pv2 are the same, second SvPV_const call may
6576 * invalidate pv1, so we may need to make a copy */
6577 if (sv1 == sv2 && (SvTHINKFIRST(sv1) || SvGMAGICAL(sv1))) {
6578 pv1 = SvPV_const(sv1, cur1);
6579 sv1 = newSVpvn_flags(pv1, cur1, SVs_TEMP | SvUTF8(sv2));
6581 pv1 = SvPV_const(sv1, cur1);
6589 pv2 = SvPV_const(sv2, cur2);
6591 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
6592 /* Differing utf8ness.
6593 * Do not UTF8size the comparands as a side-effect. */
6596 svrecode = newSVpvn(pv2, cur2);
6597 sv_recode_to_utf8(svrecode, PL_encoding);
6598 pv2 = SvPV_const(svrecode, cur2);
6601 svrecode = newSVpvn(pv1, cur1);
6602 sv_recode_to_utf8(svrecode, PL_encoding);
6603 pv1 = SvPV_const(svrecode, cur1);
6605 /* Now both are in UTF-8. */
6607 SvREFCNT_dec(svrecode);
6612 bool is_utf8 = TRUE;
6615 /* sv1 is the UTF-8 one,
6616 * if is equal it must be downgrade-able */
6617 char * const pv = (char*)bytes_from_utf8((const U8*)pv1,
6623 /* sv2 is the UTF-8 one,
6624 * if is equal it must be downgrade-able */
6625 char * const pv = (char *)bytes_from_utf8((const U8*)pv2,
6631 /* Downgrade not possible - cannot be eq */
6639 eq = (pv1 == pv2) || memEQ(pv1, pv2, cur1);
6641 SvREFCNT_dec(svrecode);
6651 Compares the strings in two SVs. Returns -1, 0, or 1 indicating whether the
6652 string in C<sv1> is less than, equal to, or greater than the string in
6653 C<sv2>. Is UTF-8 and 'use bytes' aware, handles get magic, and will
6654 coerce its args to strings if necessary. See also C<sv_cmp_locale>.
6660 Perl_sv_cmp(pTHX_ register SV *const sv1, register SV *const sv2)
6664 const char *pv1, *pv2;
6667 SV *svrecode = NULL;
6674 pv1 = SvPV_const(sv1, cur1);
6681 pv2 = SvPV_const(sv2, cur2);
6683 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
6684 /* Differing utf8ness.
6685 * Do not UTF8size the comparands as a side-effect. */
6688 svrecode = newSVpvn(pv2, cur2);
6689 sv_recode_to_utf8(svrecode, PL_encoding);
6690 pv2 = SvPV_const(svrecode, cur2);
6693 pv2 = tpv = (char*)bytes_to_utf8((const U8*)pv2, &cur2);
6698 svrecode = newSVpvn(pv1, cur1);
6699 sv_recode_to_utf8(svrecode, PL_encoding);
6700 pv1 = SvPV_const(svrecode, cur1);
6703 pv1 = tpv = (char*)bytes_to_utf8((const U8*)pv1, &cur1);
6709 cmp = cur2 ? -1 : 0;
6713 const I32 retval = memcmp((const void*)pv1, (const void*)pv2, cur1 < cur2 ? cur1 : cur2);
6716 cmp = retval < 0 ? -1 : 1;
6717 } else if (cur1 == cur2) {
6720 cmp = cur1 < cur2 ? -1 : 1;
6724 SvREFCNT_dec(svrecode);
6732 =for apidoc sv_cmp_locale
6734 Compares the strings in two SVs in a locale-aware manner. Is UTF-8 and
6735 'use bytes' aware, handles get magic, and will coerce its args to strings
6736 if necessary. See also C<sv_cmp>.
6742 Perl_sv_cmp_locale(pTHX_ register SV *const sv1, register SV *const sv2)
6745 #ifdef USE_LOCALE_COLLATE
6751 if (PL_collation_standard)
6755 pv1 = sv1 ? sv_collxfrm(sv1, &len1) : (char *) NULL;
6757 pv2 = sv2 ? sv_collxfrm(sv2, &len2) : (char *) NULL;
6759 if (!pv1 || !len1) {
6770 retval = memcmp((void*)pv1, (void*)pv2, len1 < len2 ? len1 : len2);
6773 return retval < 0 ? -1 : 1;
6776 * When the result of collation is equality, that doesn't mean
6777 * that there are no differences -- some locales exclude some
6778 * characters from consideration. So to avoid false equalities,
6779 * we use the raw string as a tiebreaker.
6785 #endif /* USE_LOCALE_COLLATE */
6787 return sv_cmp(sv1, sv2);
6791 #ifdef USE_LOCALE_COLLATE
6794 =for apidoc sv_collxfrm
6796 Add Collate Transform magic to an SV if it doesn't already have it.
6798 Any scalar variable may carry PERL_MAGIC_collxfrm magic that contains the
6799 scalar data of the variable, but transformed to such a format that a normal
6800 memory comparison can be used to compare the data according to the locale
6807 Perl_sv_collxfrm(pTHX_ SV *const sv, STRLEN *const nxp)
6812 PERL_ARGS_ASSERT_SV_COLLXFRM;
6814 mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_collxfrm) : (MAGIC *) NULL;
6815 if (!mg || !mg->mg_ptr || *(U32*)mg->mg_ptr != PL_collation_ix) {
6821 Safefree(mg->mg_ptr);
6822 s = SvPV_const(sv, len);
6823 if ((xf = mem_collxfrm(s, len, &xlen))) {
6825 #ifdef PERL_OLD_COPY_ON_WRITE
6827 sv_force_normal_flags(sv, 0);
6829 mg = sv_magicext(sv, 0, PERL_MAGIC_collxfrm, &PL_vtbl_collxfrm,
6843 if (mg && mg->mg_ptr) {
6845 return mg->mg_ptr + sizeof(PL_collation_ix);
6853 #endif /* USE_LOCALE_COLLATE */
6858 Get a line from the filehandle and store it into the SV, optionally
6859 appending to the currently-stored string.
6865 Perl_sv_gets(pTHX_ register SV *const sv, register PerlIO *const fp, I32 append)
6870 register STDCHAR rslast;
6871 register STDCHAR *bp;
6876 PERL_ARGS_ASSERT_SV_GETS;
6878 if (SvTHINKFIRST(sv))
6879 sv_force_normal_flags(sv, append ? 0 : SV_COW_DROP_PV);
6880 /* XXX. If you make this PVIV, then copy on write can copy scalars read
6882 However, perlbench says it's slower, because the existing swipe code
6883 is faster than copy on write.
6884 Swings and roundabouts. */
6885 SvUPGRADE(sv, SVt_PV);
6890 if (PerlIO_isutf8(fp)) {
6892 sv_utf8_upgrade_nomg(sv);
6893 sv_pos_u2b(sv,&append,0);
6895 } else if (SvUTF8(sv)) {
6896 SV * const tsv = newSV(0);
6897 sv_gets(tsv, fp, 0);
6898 sv_utf8_upgrade_nomg(tsv);
6899 SvCUR_set(sv,append);
6902 goto return_string_or_null;
6907 if (PerlIO_isutf8(fp))
6910 if (IN_PERL_COMPILETIME) {
6911 /* we always read code in line mode */
6915 else if (RsSNARF(PL_rs)) {
6916 /* If it is a regular disk file use size from stat() as estimate
6917 of amount we are going to read -- may result in mallocing
6918 more memory than we really need if the layers below reduce
6919 the size we read (e.g. CRLF or a gzip layer).
6922 if (!PerlLIO_fstat(PerlIO_fileno(fp), &st) && S_ISREG(st.st_mode)) {
6923 const Off_t offset = PerlIO_tell(fp);
6924 if (offset != (Off_t) -1 && st.st_size + append > offset) {
6925 (void) SvGROW(sv, (STRLEN)((st.st_size - offset) + append + 1));
6931 else if (RsRECORD(PL_rs)) {
6939 /* Grab the size of the record we're getting */
6940 recsize = SvUV(SvRV(PL_rs)); /* RsRECORD() guarantees > 0. */
6941 buffer = SvGROW(sv, (STRLEN)(recsize + append + 1)) + append;
6944 /* VMS wants read instead of fread, because fread doesn't respect */
6945 /* RMS record boundaries. This is not necessarily a good thing to be */
6946 /* doing, but we've got no other real choice - except avoid stdio
6947 as implementation - perhaps write a :vms layer ?
6949 fd = PerlIO_fileno(fp);
6950 if (fd == -1) { /* in-memory file from PerlIO::Scalar */
6951 bytesread = PerlIO_read(fp, buffer, recsize);
6954 bytesread = PerlLIO_read(fd, buffer, recsize);
6957 bytesread = PerlIO_read(fp, buffer, recsize);
6961 SvCUR_set(sv, bytesread + append);
6962 buffer[bytesread] = '\0';
6963 goto return_string_or_null;
6965 else if (RsPARA(PL_rs)) {
6971 /* Get $/ i.e. PL_rs into same encoding as stream wants */
6972 if (PerlIO_isutf8(fp)) {
6973 rsptr = SvPVutf8(PL_rs, rslen);
6976 if (SvUTF8(PL_rs)) {
6977 if (!sv_utf8_downgrade(PL_rs, TRUE)) {
6978 Perl_croak(aTHX_ "Wide character in $/");
6981 rsptr = SvPV_const(PL_rs, rslen);
6985 rslast = rslen ? rsptr[rslen - 1] : '\0';
6987 if (rspara) { /* have to do this both before and after */
6988 do { /* to make sure file boundaries work right */
6991 i = PerlIO_getc(fp);
6995 PerlIO_ungetc(fp,i);
7001 /* See if we know enough about I/O mechanism to cheat it ! */
7003 /* This used to be #ifdef test - it is made run-time test for ease
7004 of abstracting out stdio interface. One call should be cheap
7005 enough here - and may even be a macro allowing compile
7009 if (PerlIO_fast_gets(fp)) {
7012 * We're going to steal some values from the stdio struct
7013 * and put EVERYTHING in the innermost loop into registers.
7015 register STDCHAR *ptr;
7019 #if defined(VMS) && defined(PERLIO_IS_STDIO)
7020 /* An ungetc()d char is handled separately from the regular
7021 * buffer, so we getc() it back out and stuff it in the buffer.
7023 i = PerlIO_getc(fp);
7024 if (i == EOF) return 0;
7025 *(--((*fp)->_ptr)) = (unsigned char) i;
7029 /* Here is some breathtakingly efficient cheating */
7031 cnt = PerlIO_get_cnt(fp); /* get count into register */
7032 /* make sure we have the room */
7033 if ((I32)(SvLEN(sv) - append) <= cnt + 1) {
7034 /* Not room for all of it
7035 if we are looking for a separator and room for some
7037 if (rslen && cnt > 80 && (I32)SvLEN(sv) > append) {
7038 /* just process what we have room for */
7039 shortbuffered = cnt - SvLEN(sv) + append + 1;
7040 cnt -= shortbuffered;
7044 /* remember that cnt can be negative */
7045 SvGROW(sv, (STRLEN)(append + (cnt <= 0 ? 2 : (cnt + 1))));
7050 bp = (STDCHAR*)SvPVX_const(sv) + append; /* move these two too to registers */
7051 ptr = (STDCHAR*)PerlIO_get_ptr(fp);
7052 DEBUG_P(PerlIO_printf(Perl_debug_log,
7053 "Screamer: entering, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7054 DEBUG_P(PerlIO_printf(Perl_debug_log,
7055 "Screamer: entering: PerlIO * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7056 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7057 PTR2UV(PerlIO_has_base(fp) ? PerlIO_get_base(fp) : 0)));
7062 while (cnt > 0) { /* this | eat */
7064 if ((*bp++ = *ptr++) == rslast) /* really | dust */
7065 goto thats_all_folks; /* screams | sed :-) */
7069 Copy(ptr, bp, cnt, char); /* this | eat */
7070 bp += cnt; /* screams | dust */
7071 ptr += cnt; /* louder | sed :-) */
7076 if (shortbuffered) { /* oh well, must extend */
7077 cnt = shortbuffered;
7079 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
7081 SvGROW(sv, SvLEN(sv) + append + cnt + 2);
7082 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
7086 DEBUG_P(PerlIO_printf(Perl_debug_log,
7087 "Screamer: going to getc, ptr=%"UVuf", cnt=%ld\n",
7088 PTR2UV(ptr),(long)cnt));
7089 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* deregisterize cnt and ptr */
7091 DEBUG_P(PerlIO_printf(Perl_debug_log,
7092 "Screamer: pre: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7093 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7094 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7096 /* This used to call 'filbuf' in stdio form, but as that behaves like
7097 getc when cnt <= 0 we use PerlIO_getc here to avoid introducing
7098 another abstraction. */
7099 i = PerlIO_getc(fp); /* get more characters */
7101 DEBUG_P(PerlIO_printf(Perl_debug_log,
7102 "Screamer: post: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7103 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7104 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7106 cnt = PerlIO_get_cnt(fp);
7107 ptr = (STDCHAR*)PerlIO_get_ptr(fp); /* reregisterize cnt and ptr */
7108 DEBUG_P(PerlIO_printf(Perl_debug_log,
7109 "Screamer: after getc, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7111 if (i == EOF) /* all done for ever? */
7112 goto thats_really_all_folks;
7114 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
7116 SvGROW(sv, bpx + cnt + 2);
7117 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
7119 *bp++ = (STDCHAR)i; /* store character from PerlIO_getc */
7121 if (rslen && (STDCHAR)i == rslast) /* all done for now? */
7122 goto thats_all_folks;
7126 if ((rslen > 1 && (STRLEN)(bp - (STDCHAR*)SvPVX_const(sv)) < rslen) ||
7127 memNE((char*)bp - rslen, rsptr, rslen))
7128 goto screamer; /* go back to the fray */
7129 thats_really_all_folks:
7131 cnt += shortbuffered;
7132 DEBUG_P(PerlIO_printf(Perl_debug_log,
7133 "Screamer: quitting, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7134 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* put these back or we're in trouble */
7135 DEBUG_P(PerlIO_printf(Perl_debug_log,
7136 "Screamer: end: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7137 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7138 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7140 SvCUR_set(sv, bp - (STDCHAR*)SvPVX_const(sv)); /* set length */
7141 DEBUG_P(PerlIO_printf(Perl_debug_log,
7142 "Screamer: done, len=%ld, string=|%.*s|\n",
7143 (long)SvCUR(sv),(int)SvCUR(sv),SvPVX_const(sv)));
7147 /*The big, slow, and stupid way. */
7148 #ifdef USE_HEAP_INSTEAD_OF_STACK /* Even slower way. */
7149 STDCHAR *buf = NULL;
7150 Newx(buf, 8192, STDCHAR);
7158 register const STDCHAR * const bpe = buf + sizeof(buf);
7160 while ((i = PerlIO_getc(fp)) != EOF && (*bp++ = (STDCHAR)i) != rslast && bp < bpe)
7161 ; /* keep reading */
7165 cnt = PerlIO_read(fp,(char*)buf, sizeof(buf));
7166 /* Accomodate broken VAXC compiler, which applies U8 cast to
7167 * both args of ?: operator, causing EOF to change into 255
7170 i = (U8)buf[cnt - 1];
7176 cnt = 0; /* we do need to re-set the sv even when cnt <= 0 */
7178 sv_catpvn(sv, (char *) buf, cnt);
7180 sv_setpvn(sv, (char *) buf, cnt);
7182 if (i != EOF && /* joy */
7184 SvCUR(sv) < rslen ||
7185 memNE(SvPVX_const(sv) + SvCUR(sv) - rslen, rsptr, rslen)))
7189 * If we're reading from a TTY and we get a short read,
7190 * indicating that the user hit his EOF character, we need
7191 * to notice it now, because if we try to read from the TTY
7192 * again, the EOF condition will disappear.
7194 * The comparison of cnt to sizeof(buf) is an optimization
7195 * that prevents unnecessary calls to feof().
7199 if (!(cnt < (I32)sizeof(buf) && PerlIO_eof(fp)))
7203 #ifdef USE_HEAP_INSTEAD_OF_STACK
7208 if (rspara) { /* have to do this both before and after */
7209 while (i != EOF) { /* to make sure file boundaries work right */
7210 i = PerlIO_getc(fp);
7212 PerlIO_ungetc(fp,i);
7218 return_string_or_null:
7219 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7225 Auto-increment of the value in the SV, doing string to numeric conversion
7226 if necessary. Handles 'get' magic.
7232 Perl_sv_inc(pTHX_ register SV *const sv)
7241 if (SvTHINKFIRST(sv)) {
7243 sv_force_normal_flags(sv, 0);
7244 if (SvREADONLY(sv)) {
7245 if (IN_PERL_RUNTIME)
7246 Perl_croak(aTHX_ "%s", PL_no_modify);
7250 if (SvAMAGIC(sv) && AMG_CALLun(sv,inc))
7252 i = PTR2IV(SvRV(sv));
7257 flags = SvFLAGS(sv);
7258 if ((flags & (SVp_NOK|SVp_IOK)) == SVp_NOK) {
7259 /* It's (privately or publicly) a float, but not tested as an
7260 integer, so test it to see. */
7262 flags = SvFLAGS(sv);
7264 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
7265 /* It's publicly an integer, or privately an integer-not-float */
7266 #ifdef PERL_PRESERVE_IVUV
7270 if (SvUVX(sv) == UV_MAX)
7271 sv_setnv(sv, UV_MAX_P1);
7273 (void)SvIOK_only_UV(sv);
7274 SvUV_set(sv, SvUVX(sv) + 1);
7276 if (SvIVX(sv) == IV_MAX)
7277 sv_setuv(sv, (UV)IV_MAX + 1);
7279 (void)SvIOK_only(sv);
7280 SvIV_set(sv, SvIVX(sv) + 1);
7285 if (flags & SVp_NOK) {
7286 const NV was = SvNVX(sv);
7287 if (NV_OVERFLOWS_INTEGERS_AT &&
7288 was >= NV_OVERFLOWS_INTEGERS_AT && ckWARN(WARN_IMPRECISION)) {
7289 Perl_warner(aTHX_ packWARN(WARN_IMPRECISION),
7290 "Lost precision when incrementing %" NVff " by 1",
7293 (void)SvNOK_only(sv);
7294 SvNV_set(sv, was + 1.0);
7298 if (!(flags & SVp_POK) || !*SvPVX_const(sv)) {
7299 if ((flags & SVTYPEMASK) < SVt_PVIV)
7300 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV ? SVt_PVIV : SVt_IV));
7301 (void)SvIOK_only(sv);
7306 while (isALPHA(*d)) d++;
7307 while (isDIGIT(*d)) d++;
7309 #ifdef PERL_PRESERVE_IVUV
7310 /* Got to punt this as an integer if needs be, but we don't issue
7311 warnings. Probably ought to make the sv_iv_please() that does
7312 the conversion if possible, and silently. */
7313 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
7314 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
7315 /* Need to try really hard to see if it's an integer.
7316 9.22337203685478e+18 is an integer.
7317 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
7318 so $a="9.22337203685478e+18"; $a+0; $a++
7319 needs to be the same as $a="9.22337203685478e+18"; $a++
7326 /* sv_2iv *should* have made this an NV */
7327 if (flags & SVp_NOK) {
7328 (void)SvNOK_only(sv);
7329 SvNV_set(sv, SvNVX(sv) + 1.0);
7332 /* I don't think we can get here. Maybe I should assert this
7333 And if we do get here I suspect that sv_setnv will croak. NWC
7335 #if defined(USE_LONG_DOUBLE)
7336 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",
7337 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7339 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_inc punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
7340 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7343 #endif /* PERL_PRESERVE_IVUV */
7344 sv_setnv(sv,Atof(SvPVX_const(sv)) + 1.0);
7348 while (d >= SvPVX_const(sv)) {
7356 /* MKS: The original code here died if letters weren't consecutive.
7357 * at least it didn't have to worry about non-C locales. The
7358 * new code assumes that ('z'-'a')==('Z'-'A'), letters are
7359 * arranged in order (although not consecutively) and that only
7360 * [A-Za-z] are accepted by isALPHA in the C locale.
7362 if (*d != 'z' && *d != 'Z') {
7363 do { ++*d; } while (!isALPHA(*d));
7366 *(d--) -= 'z' - 'a';
7371 *(d--) -= 'z' - 'a' + 1;
7375 /* oh,oh, the number grew */
7376 SvGROW(sv, SvCUR(sv) + 2);
7377 SvCUR_set(sv, SvCUR(sv) + 1);
7378 for (d = SvPVX(sv) + SvCUR(sv); d > SvPVX_const(sv); d--)
7389 Auto-decrement of the value in the SV, doing string to numeric conversion
7390 if necessary. Handles 'get' magic.
7396 Perl_sv_dec(pTHX_ register SV *const sv)
7404 if (SvTHINKFIRST(sv)) {
7406 sv_force_normal_flags(sv, 0);
7407 if (SvREADONLY(sv)) {
7408 if (IN_PERL_RUNTIME)
7409 Perl_croak(aTHX_ "%s", PL_no_modify);
7413 if (SvAMAGIC(sv) && AMG_CALLun(sv,dec))
7415 i = PTR2IV(SvRV(sv));
7420 /* Unlike sv_inc we don't have to worry about string-never-numbers
7421 and keeping them magic. But we mustn't warn on punting */
7422 flags = SvFLAGS(sv);
7423 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
7424 /* It's publicly an integer, or privately an integer-not-float */
7425 #ifdef PERL_PRESERVE_IVUV
7429 if (SvUVX(sv) == 0) {
7430 (void)SvIOK_only(sv);
7434 (void)SvIOK_only_UV(sv);
7435 SvUV_set(sv, SvUVX(sv) - 1);
7438 if (SvIVX(sv) == IV_MIN) {
7439 sv_setnv(sv, (NV)IV_MIN);
7443 (void)SvIOK_only(sv);
7444 SvIV_set(sv, SvIVX(sv) - 1);
7449 if (flags & SVp_NOK) {
7452 const NV was = SvNVX(sv);
7453 if (NV_OVERFLOWS_INTEGERS_AT &&
7454 was <= -NV_OVERFLOWS_INTEGERS_AT && ckWARN(WARN_IMPRECISION)) {
7455 Perl_warner(aTHX_ packWARN(WARN_IMPRECISION),
7456 "Lost precision when decrementing %" NVff " by 1",
7459 (void)SvNOK_only(sv);
7460 SvNV_set(sv, was - 1.0);
7464 if (!(flags & SVp_POK)) {
7465 if ((flags & SVTYPEMASK) < SVt_PVIV)
7466 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV) ? SVt_PVIV : SVt_IV);
7468 (void)SvIOK_only(sv);
7471 #ifdef PERL_PRESERVE_IVUV
7473 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
7474 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
7475 /* Need to try really hard to see if it's an integer.
7476 9.22337203685478e+18 is an integer.
7477 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
7478 so $a="9.22337203685478e+18"; $a+0; $a--
7479 needs to be the same as $a="9.22337203685478e+18"; $a--
7486 /* sv_2iv *should* have made this an NV */
7487 if (flags & SVp_NOK) {
7488 (void)SvNOK_only(sv);
7489 SvNV_set(sv, SvNVX(sv) - 1.0);
7492 /* I don't think we can get here. Maybe I should assert this
7493 And if we do get here I suspect that sv_setnv will croak. NWC
7495 #if defined(USE_LONG_DOUBLE)
7496 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",
7497 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7499 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_dec punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
7500 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7504 #endif /* PERL_PRESERVE_IVUV */
7505 sv_setnv(sv,Atof(SvPVX_const(sv)) - 1.0); /* punt */
7509 =for apidoc sv_mortalcopy
7511 Creates a new SV which is a copy of the original SV (using C<sv_setsv>).
7512 The new SV is marked as mortal. It will be destroyed "soon", either by an
7513 explicit call to FREETMPS, or by an implicit call at places such as
7514 statement boundaries. See also C<sv_newmortal> and C<sv_2mortal>.
7519 /* Make a string that will exist for the duration of the expression
7520 * evaluation. Actually, it may have to last longer than that, but
7521 * hopefully we won't free it until it has been assigned to a
7522 * permanent location. */
7525 Perl_sv_mortalcopy(pTHX_ SV *const oldstr)
7531 sv_setsv(sv,oldstr);
7533 PL_tmps_stack[++PL_tmps_ix] = sv;
7539 =for apidoc sv_newmortal
7541 Creates a new null SV which is mortal. The reference count of the SV is
7542 set to 1. It will be destroyed "soon", either by an explicit call to
7543 FREETMPS, or by an implicit call at places such as statement boundaries.
7544 See also C<sv_mortalcopy> and C<sv_2mortal>.
7550 Perl_sv_newmortal(pTHX)
7556 SvFLAGS(sv) = SVs_TEMP;
7558 PL_tmps_stack[++PL_tmps_ix] = sv;
7564 =for apidoc newSVpvn_flags
7566 Creates a new SV and copies a string into it. The reference count for the
7567 SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
7568 string. You are responsible for ensuring that the source string is at least
7569 C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
7570 Currently the only flag bits accepted are C<SVf_UTF8> and C<SVs_TEMP>.
7571 If C<SVs_TEMP> is set, then C<sv2mortal()> is called on the result before
7572 returning. If C<SVf_UTF8> is set, then it will be set on the new SV.
7573 C<newSVpvn_utf8()> is a convenience wrapper for this function, defined as
7575 #define newSVpvn_utf8(s, len, u) \
7576 newSVpvn_flags((s), (len), (u) ? SVf_UTF8 : 0)
7582 Perl_newSVpvn_flags(pTHX_ const char *const s, const STRLEN len, const U32 flags)
7587 /* All the flags we don't support must be zero.
7588 And we're new code so I'm going to assert this from the start. */
7589 assert(!(flags & ~(SVf_UTF8|SVs_TEMP)));
7591 sv_setpvn(sv,s,len);
7592 SvFLAGS(sv) |= (flags & SVf_UTF8);
7593 return (flags & SVs_TEMP) ? sv_2mortal(sv) : sv;
7597 =for apidoc sv_2mortal
7599 Marks an existing SV as mortal. The SV will be destroyed "soon", either
7600 by an explicit call to FREETMPS, or by an implicit call at places such as
7601 statement boundaries. SvTEMP() is turned on which means that the SV's
7602 string buffer can be "stolen" if this SV is copied. See also C<sv_newmortal>
7603 and C<sv_mortalcopy>.
7609 Perl_sv_2mortal(pTHX_ register SV *const sv)
7614 if (SvREADONLY(sv) && SvIMMORTAL(sv))
7617 PL_tmps_stack[++PL_tmps_ix] = sv;
7625 Creates a new SV and copies a string into it. The reference count for the
7626 SV is set to 1. If C<len> is zero, Perl will compute the length using
7627 strlen(). For efficiency, consider using C<newSVpvn> instead.
7633 Perl_newSVpv(pTHX_ const char *const s, const STRLEN len)
7639 sv_setpvn(sv, s, len || s == NULL ? len : strlen(s));
7644 =for apidoc newSVpvn
7646 Creates a new SV and copies a string into it. The reference count for the
7647 SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
7648 string. You are responsible for ensuring that the source string is at least
7649 C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
7655 Perl_newSVpvn(pTHX_ const char *const s, const STRLEN len)
7661 sv_setpvn(sv,s,len);
7666 =for apidoc newSVhek
7668 Creates a new SV from the hash key structure. It will generate scalars that
7669 point to the shared string table where possible. Returns a new (undefined)
7670 SV if the hek is NULL.
7676 Perl_newSVhek(pTHX_ const HEK *const hek)
7686 if (HEK_LEN(hek) == HEf_SVKEY) {
7687 return newSVsv(*(SV**)HEK_KEY(hek));
7689 const int flags = HEK_FLAGS(hek);
7690 if (flags & HVhek_WASUTF8) {
7692 Andreas would like keys he put in as utf8 to come back as utf8
7694 STRLEN utf8_len = HEK_LEN(hek);
7695 const U8 *as_utf8 = bytes_to_utf8 ((U8*)HEK_KEY(hek), &utf8_len);
7696 SV * const sv = newSVpvn ((const char*)as_utf8, utf8_len);
7699 Safefree (as_utf8); /* bytes_to_utf8() allocates a new string */
7701 } else if (flags & (HVhek_REHASH|HVhek_UNSHARED)) {
7702 /* We don't have a pointer to the hv, so we have to replicate the
7703 flag into every HEK. This hv is using custom a hasing
7704 algorithm. Hence we can't return a shared string scalar, as
7705 that would contain the (wrong) hash value, and might get passed
7706 into an hv routine with a regular hash.
7707 Similarly, a hash that isn't using shared hash keys has to have
7708 the flag in every key so that we know not to try to call
7709 share_hek_kek on it. */
7711 SV * const sv = newSVpvn (HEK_KEY(hek), HEK_LEN(hek));
7716 /* This will be overwhelminly the most common case. */
7718 /* Inline most of newSVpvn_share(), because share_hek_hek() is far
7719 more efficient than sharepvn(). */
7723 sv_upgrade(sv, SVt_PV);
7724 SvPV_set(sv, (char *)HEK_KEY(share_hek_hek(hek)));
7725 SvCUR_set(sv, HEK_LEN(hek));
7738 =for apidoc newSVpvn_share
7740 Creates a new SV with its SvPVX_const pointing to a shared string in the string
7741 table. If the string does not already exist in the table, it is created
7742 first. Turns on READONLY and FAKE. If the C<hash> parameter is non-zero, that
7743 value is used; otherwise the hash is computed. The string's hash can be later
7744 be retrieved from the SV with the C<SvSHARED_HASH()> macro. The idea here is
7745 that as the string table is used for shared hash keys these strings will have
7746 SvPVX_const == HeKEY and hash lookup will avoid string compare.
7752 Perl_newSVpvn_share(pTHX_ const char *src, I32 len, U32 hash)
7756 bool is_utf8 = FALSE;
7757 const char *const orig_src = src;
7760 STRLEN tmplen = -len;
7762 /* See the note in hv.c:hv_fetch() --jhi */
7763 src = (char*)bytes_from_utf8((const U8*)src, &tmplen, &is_utf8);
7767 PERL_HASH(hash, src, len);
7769 /* The logic for this is inlined in S_mro_get_linear_isa_dfs(), so if it
7770 changes here, update it there too. */
7771 sv_upgrade(sv, SVt_PV);
7772 SvPV_set(sv, sharepvn(src, is_utf8?-len:len, hash));
7780 if (src != orig_src)
7786 #if defined(PERL_IMPLICIT_CONTEXT)
7788 /* pTHX_ magic can't cope with varargs, so this is a no-context
7789 * version of the main function, (which may itself be aliased to us).
7790 * Don't access this version directly.
7794 Perl_newSVpvf_nocontext(const char *const pat, ...)
7800 PERL_ARGS_ASSERT_NEWSVPVF_NOCONTEXT;
7802 va_start(args, pat);
7803 sv = vnewSVpvf(pat, &args);
7810 =for apidoc newSVpvf
7812 Creates a new SV and initializes it with the string formatted like
7819 Perl_newSVpvf(pTHX_ const char *const pat, ...)
7824 PERL_ARGS_ASSERT_NEWSVPVF;
7826 va_start(args, pat);
7827 sv = vnewSVpvf(pat, &args);
7832 /* backend for newSVpvf() and newSVpvf_nocontext() */
7835 Perl_vnewSVpvf(pTHX_ const char *const pat, va_list *const args)
7840 PERL_ARGS_ASSERT_VNEWSVPVF;
7843 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
7850 Creates a new SV and copies a floating point value into it.
7851 The reference count for the SV is set to 1.
7857 Perl_newSVnv(pTHX_ const NV n)
7870 Creates a new SV and copies an integer into it. The reference count for the
7877 Perl_newSViv(pTHX_ const IV i)
7890 Creates a new SV and copies an unsigned integer into it.
7891 The reference count for the SV is set to 1.
7897 Perl_newSVuv(pTHX_ const UV u)
7908 =for apidoc newSV_type
7910 Creates a new SV, of the type specified. The reference count for the new SV
7917 Perl_newSV_type(pTHX_ const svtype type)
7922 sv_upgrade(sv, type);
7927 =for apidoc newRV_noinc
7929 Creates an RV wrapper for an SV. The reference count for the original
7930 SV is B<not> incremented.
7936 Perl_newRV_noinc(pTHX_ SV *const tmpRef)
7939 register SV *sv = newSV_type(SVt_IV);
7941 PERL_ARGS_ASSERT_NEWRV_NOINC;
7944 SvRV_set(sv, tmpRef);
7949 /* newRV_inc is the official function name to use now.
7950 * newRV_inc is in fact #defined to newRV in sv.h
7954 Perl_newRV(pTHX_ SV *const sv)
7958 PERL_ARGS_ASSERT_NEWRV;
7960 return newRV_noinc(SvREFCNT_inc_simple_NN(sv));
7966 Creates a new SV which is an exact duplicate of the original SV.
7973 Perl_newSVsv(pTHX_ register SV *const old)
7980 if (SvTYPE(old) == SVTYPEMASK) {
7981 if (ckWARN_d(WARN_INTERNAL))
7982 Perl_warner(aTHX_ packWARN(WARN_INTERNAL), "semi-panic: attempt to dup freed string");
7986 /* SV_GMAGIC is the default for sv_setv()
7987 SV_NOSTEAL prevents TEMP buffers being, well, stolen, and saves games
7988 with SvTEMP_off and SvTEMP_on round a call to sv_setsv. */
7989 sv_setsv_flags(sv, old, SV_GMAGIC | SV_NOSTEAL);
7994 =for apidoc sv_reset
7996 Underlying implementation for the C<reset> Perl function.
7997 Note that the perl-level function is vaguely deprecated.
8003 Perl_sv_reset(pTHX_ register const char *s, HV *const stash)
8006 char todo[PERL_UCHAR_MAX+1];
8008 PERL_ARGS_ASSERT_SV_RESET;
8013 if (!*s) { /* reset ?? searches */
8014 MAGIC * const mg = mg_find((const SV *)stash, PERL_MAGIC_symtab);
8016 const U32 count = mg->mg_len / sizeof(PMOP**);
8017 PMOP **pmp = (PMOP**) mg->mg_ptr;
8018 PMOP *const *const end = pmp + count;
8022 SvREADONLY_off(PL_regex_pad[(*pmp)->op_pmoffset]);
8024 (*pmp)->op_pmflags &= ~PMf_USED;
8032 /* reset variables */
8034 if (!HvARRAY(stash))
8037 Zero(todo, 256, char);
8040 I32 i = (unsigned char)*s;
8044 max = (unsigned char)*s++;
8045 for ( ; i <= max; i++) {
8048 for (i = 0; i <= (I32) HvMAX(stash); i++) {
8050 for (entry = HvARRAY(stash)[i];
8052 entry = HeNEXT(entry))
8057 if (!todo[(U8)*HeKEY(entry)])
8059 gv = MUTABLE_GV(HeVAL(entry));
8062 if (SvTHINKFIRST(sv)) {
8063 if (!SvREADONLY(sv) && SvROK(sv))
8065 /* XXX Is this continue a bug? Why should THINKFIRST
8066 exempt us from resetting arrays and hashes? */
8070 if (SvTYPE(sv) >= SVt_PV) {
8072 if (SvPVX_const(sv) != NULL)
8080 if (GvHV(gv) && !HvNAME_get(GvHV(gv))) {
8082 Perl_die(aTHX_ "Can't reset %%ENV on this system");
8085 # if defined(USE_ENVIRON_ARRAY)
8088 # endif /* USE_ENVIRON_ARRAY */
8099 Using various gambits, try to get an IO from an SV: the IO slot if its a
8100 GV; or the recursive result if we're an RV; or the IO slot of the symbol
8101 named after the PV if we're a string.
8107 Perl_sv_2io(pTHX_ SV *const sv)
8112 PERL_ARGS_ASSERT_SV_2IO;
8114 switch (SvTYPE(sv)) {
8116 io = MUTABLE_IO(sv);
8119 if (isGV_with_GP(sv)) {
8120 gv = MUTABLE_GV(sv);
8123 Perl_croak(aTHX_ "Bad filehandle: %s", GvNAME(gv));
8129 Perl_croak(aTHX_ PL_no_usym, "filehandle");
8131 return sv_2io(SvRV(sv));
8132 gv = gv_fetchsv(sv, 0, SVt_PVIO);
8138 Perl_croak(aTHX_ "Bad filehandle: %"SVf, SVfARG(sv));
8147 Using various gambits, try to get a CV from an SV; in addition, try if
8148 possible to set C<*st> and C<*gvp> to the stash and GV associated with it.
8149 The flags in C<lref> are passed to sv_fetchsv.
8155 Perl_sv_2cv(pTHX_ SV *sv, HV **const st, GV **const gvp, const I32 lref)
8161 PERL_ARGS_ASSERT_SV_2CV;
8168 switch (SvTYPE(sv)) {
8172 return MUTABLE_CV(sv);
8179 if (isGV_with_GP(sv)) {
8180 gv = MUTABLE_GV(sv);
8189 SV * const *sp = &sv; /* Used in tryAMAGICunDEREF macro. */
8191 tryAMAGICunDEREF(to_cv);
8194 if (SvTYPE(sv) == SVt_PVCV) {
8195 cv = MUTABLE_CV(sv);
8200 else if(isGV_with_GP(sv))
8201 gv = MUTABLE_GV(sv);
8203 Perl_croak(aTHX_ "Not a subroutine reference");
8205 else if (isGV_with_GP(sv)) {
8207 gv = MUTABLE_GV(sv);
8210 gv = gv_fetchsv(sv, lref, SVt_PVCV); /* Calls get magic */
8216 /* Some flags to gv_fetchsv mean don't really create the GV */
8217 if (!isGV_with_GP(gv)) {
8223 if (lref && !GvCVu(gv)) {
8227 gv_efullname3(tmpsv, gv, NULL);
8228 /* XXX this is probably not what they think they're getting.
8229 * It has the same effect as "sub name;", i.e. just a forward
8231 newSUB(start_subparse(FALSE, 0),
8232 newSVOP(OP_CONST, 0, tmpsv),
8236 Perl_croak(aTHX_ "Unable to create sub named \"%"SVf"\"",
8237 SVfARG(SvOK(sv) ? sv : &PL_sv_no));
8246 Returns true if the SV has a true value by Perl's rules.
8247 Use the C<SvTRUE> macro instead, which may call C<sv_true()> or may
8248 instead use an in-line version.
8254 Perl_sv_true(pTHX_ register SV *const sv)
8259 register const XPV* const tXpv = (XPV*)SvANY(sv);
8261 (tXpv->xpv_cur > 1 ||
8262 (tXpv->xpv_cur && *sv->sv_u.svu_pv != '0')))
8269 return SvIVX(sv) != 0;
8272 return SvNVX(sv) != 0.0;
8274 return sv_2bool(sv);
8280 =for apidoc sv_pvn_force
8282 Get a sensible string out of the SV somehow.
8283 A private implementation of the C<SvPV_force> macro for compilers which
8284 can't cope with complex macro expressions. Always use the macro instead.
8286 =for apidoc sv_pvn_force_flags
8288 Get a sensible string out of the SV somehow.
8289 If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<sv> if
8290 appropriate, else not. C<sv_pvn_force> and C<sv_pvn_force_nomg> are
8291 implemented in terms of this function.
8292 You normally want to use the various wrapper macros instead: see
8293 C<SvPV_force> and C<SvPV_force_nomg>
8299 Perl_sv_pvn_force_flags(pTHX_ SV *const sv, STRLEN *const lp, const I32 flags)
8303 PERL_ARGS_ASSERT_SV_PVN_FORCE_FLAGS;
8305 if (SvTHINKFIRST(sv) && !SvROK(sv))
8306 sv_force_normal_flags(sv, 0);
8316 if (SvREADONLY(sv) && !(flags & SV_MUTABLE_RETURN)) {
8317 const char * const ref = sv_reftype(sv,0);
8319 Perl_croak(aTHX_ "Can't coerce readonly %s to string in %s",
8320 ref, OP_NAME(PL_op));
8322 Perl_croak(aTHX_ "Can't coerce readonly %s to string", ref);
8324 if ((SvTYPE(sv) > SVt_PVLV && SvTYPE(sv) != SVt_PVFM)
8325 || isGV_with_GP(sv))
8326 Perl_croak(aTHX_ "Can't coerce %s to string in %s", sv_reftype(sv,0),
8328 s = sv_2pv_flags(sv, &len, flags);
8332 if (s != SvPVX_const(sv)) { /* Almost, but not quite, sv_setpvn() */
8335 SvUPGRADE(sv, SVt_PV); /* Never FALSE */
8336 SvGROW(sv, len + 1);
8337 Move(s,SvPVX(sv),len,char);
8339 SvPVX(sv)[len] = '\0';
8342 SvPOK_on(sv); /* validate pointer */
8344 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
8345 PTR2UV(sv),SvPVX_const(sv)));
8348 return SvPVX_mutable(sv);
8352 =for apidoc sv_pvbyten_force
8354 The backend for the C<SvPVbytex_force> macro. Always use the macro instead.
8360 Perl_sv_pvbyten_force(pTHX_ SV *const sv, STRLEN *const lp)
8362 PERL_ARGS_ASSERT_SV_PVBYTEN_FORCE;
8364 sv_pvn_force(sv,lp);
8365 sv_utf8_downgrade(sv,0);
8371 =for apidoc sv_pvutf8n_force
8373 The backend for the C<SvPVutf8x_force> macro. Always use the macro instead.
8379 Perl_sv_pvutf8n_force(pTHX_ SV *const sv, STRLEN *const lp)
8381 PERL_ARGS_ASSERT_SV_PVUTF8N_FORCE;
8383 sv_pvn_force(sv,lp);
8384 sv_utf8_upgrade(sv);
8390 =for apidoc sv_reftype
8392 Returns a string describing what the SV is a reference to.
8398 Perl_sv_reftype(pTHX_ const SV *const sv, const int ob)
8400 PERL_ARGS_ASSERT_SV_REFTYPE;
8402 /* The fact that I don't need to downcast to char * everywhere, only in ?:
8403 inside return suggests a const propagation bug in g++. */
8404 if (ob && SvOBJECT(sv)) {
8405 char * const name = HvNAME_get(SvSTASH(sv));
8406 return name ? name : (char *) "__ANON__";
8409 switch (SvTYPE(sv)) {
8424 case SVt_PVLV: return (char *) (SvROK(sv) ? "REF"
8425 /* tied lvalues should appear to be
8426 * scalars for backwards compatitbility */
8427 : (LvTYPE(sv) == 't' || LvTYPE(sv) == 'T')
8428 ? "SCALAR" : "LVALUE");
8429 case SVt_PVAV: return "ARRAY";
8430 case SVt_PVHV: return "HASH";
8431 case SVt_PVCV: return "CODE";
8432 case SVt_PVGV: return (char *) (isGV_with_GP(sv)
8433 ? "GLOB" : "SCALAR");
8434 case SVt_PVFM: return "FORMAT";
8435 case SVt_PVIO: return "IO";
8436 case SVt_BIND: return "BIND";
8437 case SVt_REGEXP: return "REGEXP";
8438 default: return "UNKNOWN";
8444 =for apidoc sv_isobject
8446 Returns a boolean indicating whether the SV is an RV pointing to a blessed
8447 object. If the SV is not an RV, or if the object is not blessed, then this
8454 Perl_sv_isobject(pTHX_ SV *sv)
8470 Returns a boolean indicating whether the SV is blessed into the specified
8471 class. This does not check for subtypes; use C<sv_derived_from> to verify
8472 an inheritance relationship.
8478 Perl_sv_isa(pTHX_ SV *sv, const char *const name)
8482 PERL_ARGS_ASSERT_SV_ISA;
8492 hvname = HvNAME_get(SvSTASH(sv));
8496 return strEQ(hvname, name);
8502 Creates a new SV for the RV, C<rv>, to point to. If C<rv> is not an RV then
8503 it will be upgraded to one. If C<classname> is non-null then the new SV will
8504 be blessed in the specified package. The new SV is returned and its
8505 reference count is 1.
8511 Perl_newSVrv(pTHX_ SV *const rv, const char *const classname)
8516 PERL_ARGS_ASSERT_NEWSVRV;
8520 SV_CHECK_THINKFIRST_COW_DROP(rv);
8521 (void)SvAMAGIC_off(rv);
8523 if (SvTYPE(rv) >= SVt_PVMG) {
8524 const U32 refcnt = SvREFCNT(rv);
8528 SvREFCNT(rv) = refcnt;
8530 sv_upgrade(rv, SVt_IV);
8531 } else if (SvROK(rv)) {
8532 SvREFCNT_dec(SvRV(rv));
8534 prepare_SV_for_RV(rv);
8542 HV* const stash = gv_stashpv(classname, GV_ADD);
8543 (void)sv_bless(rv, stash);
8549 =for apidoc sv_setref_pv
8551 Copies a pointer into a new SV, optionally blessing the SV. The C<rv>
8552 argument will be upgraded to an RV. That RV will be modified to point to
8553 the new SV. If the C<pv> argument is NULL then C<PL_sv_undef> will be placed
8554 into the SV. The C<classname> argument indicates the package for the
8555 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
8556 will have a reference count of 1, and the RV will be returned.
8558 Do not use with other Perl types such as HV, AV, SV, CV, because those
8559 objects will become corrupted by the pointer copy process.
8561 Note that C<sv_setref_pvn> copies the string while this copies the pointer.
8567 Perl_sv_setref_pv(pTHX_ SV *const rv, const char *const classname, void *const pv)
8571 PERL_ARGS_ASSERT_SV_SETREF_PV;
8574 sv_setsv(rv, &PL_sv_undef);
8578 sv_setiv(newSVrv(rv,classname), PTR2IV(pv));
8583 =for apidoc sv_setref_iv
8585 Copies an integer into a new SV, optionally blessing the SV. The C<rv>
8586 argument will be upgraded to an RV. That RV will be modified to point to
8587 the new SV. The C<classname> argument indicates the package for the
8588 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
8589 will have a reference count of 1, and the RV will be returned.
8595 Perl_sv_setref_iv(pTHX_ SV *const rv, const char *const classname, const IV iv)
8597 PERL_ARGS_ASSERT_SV_SETREF_IV;
8599 sv_setiv(newSVrv(rv,classname), iv);
8604 =for apidoc sv_setref_uv
8606 Copies an unsigned integer into a new SV, optionally blessing the SV. The C<rv>
8607 argument will be upgraded to an RV. That RV will be modified to point to
8608 the new SV. The C<classname> argument indicates the package for the
8609 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
8610 will have a reference count of 1, and the RV will be returned.
8616 Perl_sv_setref_uv(pTHX_ SV *const rv, const char *const classname, const UV uv)
8618 PERL_ARGS_ASSERT_SV_SETREF_UV;
8620 sv_setuv(newSVrv(rv,classname), uv);
8625 =for apidoc sv_setref_nv
8627 Copies a double into a new SV, optionally blessing the SV. The C<rv>
8628 argument will be upgraded to an RV. That RV will be modified to point to
8629 the new SV. The C<classname> argument indicates the package for the
8630 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
8631 will have a reference count of 1, and the RV will be returned.
8637 Perl_sv_setref_nv(pTHX_ SV *const rv, const char *const classname, const NV nv)
8639 PERL_ARGS_ASSERT_SV_SETREF_NV;
8641 sv_setnv(newSVrv(rv,classname), nv);
8646 =for apidoc sv_setref_pvn
8648 Copies a string into a new SV, optionally blessing the SV. The length of the
8649 string must be specified with C<n>. The C<rv> argument will be upgraded to
8650 an RV. That RV will be modified to point to the new SV. The C<classname>
8651 argument indicates the package for the blessing. Set C<classname> to
8652 C<NULL> to avoid the blessing. The new SV will have a reference count
8653 of 1, and the RV will be returned.
8655 Note that C<sv_setref_pv> copies the pointer while this copies the string.
8661 Perl_sv_setref_pvn(pTHX_ SV *const rv, const char *const classname,
8662 const char *const pv, const STRLEN n)
8664 PERL_ARGS_ASSERT_SV_SETREF_PVN;
8666 sv_setpvn(newSVrv(rv,classname), pv, n);
8671 =for apidoc sv_bless
8673 Blesses an SV into a specified package. The SV must be an RV. The package
8674 must be designated by its stash (see C<gv_stashpv()>). The reference count
8675 of the SV is unaffected.
8681 Perl_sv_bless(pTHX_ SV *const sv, HV *const stash)
8686 PERL_ARGS_ASSERT_SV_BLESS;
8689 Perl_croak(aTHX_ "Can't bless non-reference value");
8691 if (SvFLAGS(tmpRef) & (SVs_OBJECT|SVf_READONLY)) {
8692 if (SvIsCOW(tmpRef))
8693 sv_force_normal_flags(tmpRef, 0);
8694 if (SvREADONLY(tmpRef))
8695 Perl_croak(aTHX_ "%s", PL_no_modify);
8696 if (SvOBJECT(tmpRef)) {
8697 if (SvTYPE(tmpRef) != SVt_PVIO)
8699 SvREFCNT_dec(SvSTASH(tmpRef));
8702 SvOBJECT_on(tmpRef);
8703 if (SvTYPE(tmpRef) != SVt_PVIO)
8705 SvUPGRADE(tmpRef, SVt_PVMG);
8706 SvSTASH_set(tmpRef, MUTABLE_HV(SvREFCNT_inc_simple(stash)));
8711 (void)SvAMAGIC_off(sv);
8713 if(SvSMAGICAL(tmpRef))
8714 if(mg_find(tmpRef, PERL_MAGIC_ext) || mg_find(tmpRef, PERL_MAGIC_uvar))
8722 /* Downgrades a PVGV to a PVMG.
8726 S_sv_unglob(pTHX_ SV *const sv)
8731 SV * const temp = sv_newmortal();
8733 PERL_ARGS_ASSERT_SV_UNGLOB;
8735 assert(SvTYPE(sv) == SVt_PVGV);
8737 gv_efullname3(temp, MUTABLE_GV(sv), "*");
8740 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
8741 && HvNAME_get(stash))
8742 mro_method_changed_in(stash);
8743 gp_free(MUTABLE_GV(sv));
8746 sv_del_backref(MUTABLE_SV(GvSTASH(sv)), sv);
8750 if (GvNAME_HEK(sv)) {
8751 unshare_hek(GvNAME_HEK(sv));
8753 isGV_with_GP_off(sv);
8755 /* need to keep SvANY(sv) in the right arena */
8756 xpvmg = new_XPVMG();
8757 StructCopy(SvANY(sv), xpvmg, XPVMG);
8758 del_XPVGV(SvANY(sv));
8761 SvFLAGS(sv) &= ~SVTYPEMASK;
8762 SvFLAGS(sv) |= SVt_PVMG;
8764 /* Intentionally not calling any local SET magic, as this isn't so much a
8765 set operation as merely an internal storage change. */
8766 sv_setsv_flags(sv, temp, 0);
8770 =for apidoc sv_unref_flags
8772 Unsets the RV status of the SV, and decrements the reference count of
8773 whatever was being referenced by the RV. This can almost be thought of
8774 as a reversal of C<newSVrv>. The C<cflags> argument can contain
8775 C<SV_IMMEDIATE_UNREF> to force the reference count to be decremented
8776 (otherwise the decrementing is conditional on the reference count being
8777 different from one or the reference being a readonly SV).
8784 Perl_sv_unref_flags(pTHX_ SV *const ref, const U32 flags)
8786 SV* const target = SvRV(ref);
8788 PERL_ARGS_ASSERT_SV_UNREF_FLAGS;
8790 if (SvWEAKREF(ref)) {
8791 sv_del_backref(target, ref);
8793 SvRV_set(ref, NULL);
8796 SvRV_set(ref, NULL);
8798 /* You can't have a || SvREADONLY(target) here, as $a = $$a, where $a was
8799 assigned to as BEGIN {$a = \"Foo"} will fail. */
8800 if (SvREFCNT(target) != 1 || (flags & SV_IMMEDIATE_UNREF))
8801 SvREFCNT_dec(target);
8802 else /* XXX Hack, but hard to make $a=$a->[1] work otherwise */
8803 sv_2mortal(target); /* Schedule for freeing later */
8807 =for apidoc sv_untaint
8809 Untaint an SV. Use C<SvTAINTED_off> instead.
8814 Perl_sv_untaint(pTHX_ SV *const sv)
8816 PERL_ARGS_ASSERT_SV_UNTAINT;
8818 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
8819 MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
8826 =for apidoc sv_tainted
8828 Test an SV for taintedness. Use C<SvTAINTED> instead.
8833 Perl_sv_tainted(pTHX_ SV *const sv)
8835 PERL_ARGS_ASSERT_SV_TAINTED;
8837 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
8838 const MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
8839 if (mg && (mg->mg_len & 1) )
8846 =for apidoc sv_setpviv
8848 Copies an integer into the given SV, also updating its string value.
8849 Does not handle 'set' magic. See C<sv_setpviv_mg>.
8855 Perl_sv_setpviv(pTHX_ SV *const sv, const IV iv)
8857 char buf[TYPE_CHARS(UV)];
8859 char * const ptr = uiv_2buf(buf, iv, 0, 0, &ebuf);
8861 PERL_ARGS_ASSERT_SV_SETPVIV;
8863 sv_setpvn(sv, ptr, ebuf - ptr);
8867 =for apidoc sv_setpviv_mg
8869 Like C<sv_setpviv>, but also handles 'set' magic.
8875 Perl_sv_setpviv_mg(pTHX_ SV *const sv, const IV iv)
8877 PERL_ARGS_ASSERT_SV_SETPVIV_MG;
8883 #if defined(PERL_IMPLICIT_CONTEXT)
8885 /* pTHX_ magic can't cope with varargs, so this is a no-context
8886 * version of the main function, (which may itself be aliased to us).
8887 * Don't access this version directly.
8891 Perl_sv_setpvf_nocontext(SV *const sv, const char *const pat, ...)
8896 PERL_ARGS_ASSERT_SV_SETPVF_NOCONTEXT;
8898 va_start(args, pat);
8899 sv_vsetpvf(sv, pat, &args);
8903 /* pTHX_ magic can't cope with varargs, so this is a no-context
8904 * version of the main function, (which may itself be aliased to us).
8905 * Don't access this version directly.
8909 Perl_sv_setpvf_mg_nocontext(SV *const sv, const char *const pat, ...)
8914 PERL_ARGS_ASSERT_SV_SETPVF_MG_NOCONTEXT;
8916 va_start(args, pat);
8917 sv_vsetpvf_mg(sv, pat, &args);
8923 =for apidoc sv_setpvf
8925 Works like C<sv_catpvf> but copies the text into the SV instead of
8926 appending it. Does not handle 'set' magic. See C<sv_setpvf_mg>.
8932 Perl_sv_setpvf(pTHX_ SV *const sv, const char *const pat, ...)
8936 PERL_ARGS_ASSERT_SV_SETPVF;
8938 va_start(args, pat);
8939 sv_vsetpvf(sv, pat, &args);
8944 =for apidoc sv_vsetpvf
8946 Works like C<sv_vcatpvf> but copies the text into the SV instead of
8947 appending it. Does not handle 'set' magic. See C<sv_vsetpvf_mg>.
8949 Usually used via its frontend C<sv_setpvf>.
8955 Perl_sv_vsetpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
8957 PERL_ARGS_ASSERT_SV_VSETPVF;
8959 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
8963 =for apidoc sv_setpvf_mg
8965 Like C<sv_setpvf>, but also handles 'set' magic.
8971 Perl_sv_setpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
8975 PERL_ARGS_ASSERT_SV_SETPVF_MG;
8977 va_start(args, pat);
8978 sv_vsetpvf_mg(sv, pat, &args);
8983 =for apidoc sv_vsetpvf_mg
8985 Like C<sv_vsetpvf>, but also handles 'set' magic.
8987 Usually used via its frontend C<sv_setpvf_mg>.
8993 Perl_sv_vsetpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
8995 PERL_ARGS_ASSERT_SV_VSETPVF_MG;
8997 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9001 #if defined(PERL_IMPLICIT_CONTEXT)
9003 /* pTHX_ magic can't cope with varargs, so this is a no-context
9004 * version of the main function, (which may itself be aliased to us).
9005 * Don't access this version directly.
9009 Perl_sv_catpvf_nocontext(SV *const sv, const char *const pat, ...)
9014 PERL_ARGS_ASSERT_SV_CATPVF_NOCONTEXT;
9016 va_start(args, pat);
9017 sv_vcatpvf(sv, pat, &args);
9021 /* pTHX_ magic can't cope with varargs, so this is a no-context
9022 * version of the main function, (which may itself be aliased to us).
9023 * Don't access this version directly.
9027 Perl_sv_catpvf_mg_nocontext(SV *const sv, const char *const pat, ...)
9032 PERL_ARGS_ASSERT_SV_CATPVF_MG_NOCONTEXT;
9034 va_start(args, pat);
9035 sv_vcatpvf_mg(sv, pat, &args);
9041 =for apidoc sv_catpvf
9043 Processes its arguments like C<sprintf> and appends the formatted
9044 output to an SV. If the appended data contains "wide" characters
9045 (including, but not limited to, SVs with a UTF-8 PV formatted with %s,
9046 and characters >255 formatted with %c), the original SV might get
9047 upgraded to UTF-8. Handles 'get' magic, but not 'set' magic. See
9048 C<sv_catpvf_mg>. If the original SV was UTF-8, the pattern should be
9049 valid UTF-8; if the original SV was bytes, the pattern should be too.
9054 Perl_sv_catpvf(pTHX_ SV *const sv, const char *const pat, ...)
9058 PERL_ARGS_ASSERT_SV_CATPVF;
9060 va_start(args, pat);
9061 sv_vcatpvf(sv, pat, &args);
9066 =for apidoc sv_vcatpvf
9068 Processes its arguments like C<vsprintf> and appends the formatted output
9069 to an SV. Does not handle 'set' magic. See C<sv_vcatpvf_mg>.
9071 Usually used via its frontend C<sv_catpvf>.
9077 Perl_sv_vcatpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9079 PERL_ARGS_ASSERT_SV_VCATPVF;
9081 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9085 =for apidoc sv_catpvf_mg
9087 Like C<sv_catpvf>, but also handles 'set' magic.
9093 Perl_sv_catpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
9097 PERL_ARGS_ASSERT_SV_CATPVF_MG;
9099 va_start(args, pat);
9100 sv_vcatpvf_mg(sv, pat, &args);
9105 =for apidoc sv_vcatpvf_mg
9107 Like C<sv_vcatpvf>, but also handles 'set' magic.
9109 Usually used via its frontend C<sv_catpvf_mg>.
9115 Perl_sv_vcatpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9117 PERL_ARGS_ASSERT_SV_VCATPVF_MG;
9119 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9124 =for apidoc sv_vsetpvfn
9126 Works like C<sv_vcatpvfn> but copies the text into the SV instead of
9129 Usually used via one of its frontends C<sv_vsetpvf> and C<sv_vsetpvf_mg>.
9135 Perl_sv_vsetpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
9136 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
9138 PERL_ARGS_ASSERT_SV_VSETPVFN;
9141 sv_vcatpvfn(sv, pat, patlen, args, svargs, svmax, maybe_tainted);
9145 S_expect_number(pTHX_ char **const pattern)
9150 PERL_ARGS_ASSERT_EXPECT_NUMBER;
9152 switch (**pattern) {
9153 case '1': case '2': case '3':
9154 case '4': case '5': case '6':
9155 case '7': case '8': case '9':
9156 var = *(*pattern)++ - '0';
9157 while (isDIGIT(**pattern)) {
9158 const I32 tmp = var * 10 + (*(*pattern)++ - '0');
9160 Perl_croak(aTHX_ "Integer overflow in format string for %s", (PL_op ? OP_NAME(PL_op) : "sv_vcatpvfn"));
9168 S_F0convert(NV nv, char *const endbuf, STRLEN *const len)
9170 const int neg = nv < 0;
9173 PERL_ARGS_ASSERT_F0CONVERT;
9181 if (uv & 1 && uv == nv)
9182 uv--; /* Round to even */
9184 const unsigned dig = uv % 10;
9197 =for apidoc sv_vcatpvfn
9199 Processes its arguments like C<vsprintf> and appends the formatted output
9200 to an SV. Uses an array of SVs if the C style variable argument list is
9201 missing (NULL). When running with taint checks enabled, indicates via
9202 C<maybe_tainted> if results are untrustworthy (often due to the use of
9205 Usually used via one of its frontends C<sv_vcatpvf> and C<sv_vcatpvf_mg>.
9211 #define VECTORIZE_ARGS vecsv = va_arg(*args, SV*);\
9212 vecstr = (U8*)SvPV_const(vecsv,veclen);\
9213 vec_utf8 = DO_UTF8(vecsv);
9215 /* XXX maybe_tainted is never assigned to, so the doc above is lying. */
9218 Perl_sv_vcatpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
9219 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
9227 static const char nullstr[] = "(null)";
9229 bool has_utf8 = DO_UTF8(sv); /* has the result utf8? */
9230 const bool pat_utf8 = has_utf8; /* the pattern is in utf8? */
9232 /* Times 4: a decimal digit takes more than 3 binary digits.
9233 * NV_DIG: mantissa takes than many decimal digits.
9234 * Plus 32: Playing safe. */
9235 char ebuf[IV_DIG * 4 + NV_DIG + 32];
9236 /* large enough for "%#.#f" --chip */
9237 /* what about long double NVs? --jhi */
9239 PERL_ARGS_ASSERT_SV_VCATPVFN;
9240 PERL_UNUSED_ARG(maybe_tainted);
9242 /* no matter what, this is a string now */
9243 (void)SvPV_force(sv, origlen);
9245 /* special-case "", "%s", and "%-p" (SVf - see below) */
9248 if (patlen == 2 && pat[0] == '%' && pat[1] == 's') {
9250 const char * const s = va_arg(*args, char*);
9251 sv_catpv(sv, s ? s : nullstr);
9253 else if (svix < svmax) {
9254 sv_catsv(sv, *svargs);
9258 if (args && patlen == 3 && pat[0] == '%' &&
9259 pat[1] == '-' && pat[2] == 'p') {
9260 argsv = MUTABLE_SV(va_arg(*args, void*));
9261 sv_catsv(sv, argsv);
9265 #ifndef USE_LONG_DOUBLE
9266 /* special-case "%.<number>[gf]" */
9267 if ( !args && patlen <= 5 && pat[0] == '%' && pat[1] == '.'
9268 && (pat[patlen-1] == 'g' || pat[patlen-1] == 'f') ) {
9269 unsigned digits = 0;
9273 while (*pp >= '0' && *pp <= '9')
9274 digits = 10 * digits + (*pp++ - '0');
9275 if (pp - pat == (int)patlen - 1) {
9283 /* Add check for digits != 0 because it seems that some
9284 gconverts are buggy in this case, and we don't yet have
9285 a Configure test for this. */
9286 if (digits && digits < sizeof(ebuf) - NV_DIG - 10) {
9287 /* 0, point, slack */
9288 Gconvert(nv, (int)digits, 0, ebuf);
9290 if (*ebuf) /* May return an empty string for digits==0 */
9293 } else if (!digits) {
9296 if ((p = F0convert(nv, ebuf + sizeof ebuf, &l))) {
9297 sv_catpvn(sv, p, l);
9303 #endif /* !USE_LONG_DOUBLE */
9305 if (!args && svix < svmax && DO_UTF8(*svargs))
9308 patend = (char*)pat + patlen;
9309 for (p = (char*)pat; p < patend; p = q) {
9312 bool vectorize = FALSE;
9313 bool vectorarg = FALSE;
9314 bool vec_utf8 = FALSE;
9320 bool has_precis = FALSE;
9322 const I32 osvix = svix;
9323 bool is_utf8 = FALSE; /* is this item utf8? */
9324 #ifdef HAS_LDBL_SPRINTF_BUG
9325 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
9326 with sfio - Allen <allens@cpan.org> */
9327 bool fix_ldbl_sprintf_bug = FALSE;
9331 U8 utf8buf[UTF8_MAXBYTES+1];
9332 STRLEN esignlen = 0;
9334 const char *eptr = NULL;
9335 const char *fmtstart;
9338 const U8 *vecstr = NULL;
9345 /* we need a long double target in case HAS_LONG_DOUBLE but
9348 #if defined(HAS_LONG_DOUBLE) && LONG_DOUBLESIZE > DOUBLESIZE
9356 const char *dotstr = ".";
9357 STRLEN dotstrlen = 1;
9358 I32 efix = 0; /* explicit format parameter index */
9359 I32 ewix = 0; /* explicit width index */
9360 I32 epix = 0; /* explicit precision index */
9361 I32 evix = 0; /* explicit vector index */
9362 bool asterisk = FALSE;
9364 /* echo everything up to the next format specification */
9365 for (q = p; q < patend && *q != '%'; ++q) ;
9367 if (has_utf8 && !pat_utf8)
9368 sv_catpvn_utf8_upgrade(sv, p, q - p, nsv);
9370 sv_catpvn(sv, p, q - p);
9379 We allow format specification elements in this order:
9380 \d+\$ explicit format parameter index
9382 v|\*(\d+\$)?v vector with optional (optionally specified) arg
9383 0 flag (as above): repeated to allow "v02"
9384 \d+|\*(\d+\$)? width using optional (optionally specified) arg
9385 \.(\d*|\*(\d+\$)?) precision using optional (optionally specified) arg
9387 [%bcdefginopsuxDFOUX] format (mandatory)
9392 As of perl5.9.3, printf format checking is on by default.
9393 Internally, perl uses %p formats to provide an escape to
9394 some extended formatting. This block deals with those
9395 extensions: if it does not match, (char*)q is reset and
9396 the normal format processing code is used.
9398 Currently defined extensions are:
9399 %p include pointer address (standard)
9400 %-p (SVf) include an SV (previously %_)
9401 %-<num>p include an SV with precision <num>
9402 %<num>p reserved for future extensions
9404 Robin Barker 2005-07-14
9406 %1p (VDf) removed. RMB 2007-10-19
9413 n = expect_number(&q);
9420 argsv = MUTABLE_SV(va_arg(*args, void*));
9421 eptr = SvPV_const(argsv, elen);
9427 if (ckWARN_d(WARN_INTERNAL))
9428 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
9429 "internal %%<num>p might conflict with future printf extensions");
9435 if ( (width = expect_number(&q)) ) {
9450 if (plus == '+' && *q == ' ') /* '+' over ' ' */
9479 if ( (ewix = expect_number(&q)) )
9488 if ((vectorarg = asterisk)) {
9501 width = expect_number(&q);
9507 vecsv = va_arg(*args, SV*);
9509 vecsv = (evix > 0 && evix <= svmax)
9510 ? svargs[evix-1] : &PL_sv_undef;
9512 vecsv = svix < svmax ? svargs[svix++] : &PL_sv_undef;
9514 dotstr = SvPV_const(vecsv, dotstrlen);
9515 /* Keep the DO_UTF8 test *after* the SvPV call, else things go
9516 bad with tied or overloaded values that return UTF8. */
9519 else if (has_utf8) {
9520 vecsv = sv_mortalcopy(vecsv);
9521 sv_utf8_upgrade(vecsv);
9522 dotstr = SvPV_const(vecsv, dotstrlen);
9529 else if (efix ? (efix > 0 && efix <= svmax) : svix < svmax) {
9530 vecsv = svargs[efix ? efix-1 : svix++];
9531 vecstr = (U8*)SvPV_const(vecsv,veclen);
9532 vec_utf8 = DO_UTF8(vecsv);
9534 /* if this is a version object, we need to convert
9535 * back into v-string notation and then let the
9536 * vectorize happen normally
9538 if (sv_derived_from(vecsv, "version")) {
9539 char *version = savesvpv(vecsv);
9540 if ( hv_exists(MUTABLE_HV(SvRV(vecsv)), "alpha", 5 ) ) {
9541 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
9542 "vector argument not supported with alpha versions");
9545 vecsv = sv_newmortal();
9546 scan_vstring(version, version + veclen, vecsv);
9547 vecstr = (U8*)SvPV_const(vecsv, veclen);
9548 vec_utf8 = DO_UTF8(vecsv);
9560 i = va_arg(*args, int);
9562 i = (ewix ? ewix <= svmax : svix < svmax) ?
9563 SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
9565 width = (i < 0) ? -i : i;
9575 if ( ((epix = expect_number(&q))) && (*q++ != '$') )
9577 /* XXX: todo, support specified precision parameter */
9581 i = va_arg(*args, int);
9583 i = (ewix ? ewix <= svmax : svix < svmax)
9584 ? SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
9586 has_precis = !(i < 0);
9591 precis = precis * 10 + (*q++ - '0');
9600 case 'I': /* Ix, I32x, and I64x */
9602 if (q[1] == '6' && q[2] == '4') {
9608 if (q[1] == '3' && q[2] == '2') {
9618 #if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
9629 #if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
9630 if (*(q + 1) == 'l') { /* lld, llf */
9656 if (!vectorize && !args) {
9658 const I32 i = efix-1;
9659 argsv = (i >= 0 && i < svmax) ? svargs[i] : &PL_sv_undef;
9661 argsv = (svix >= 0 && svix < svmax)
9662 ? svargs[svix++] : &PL_sv_undef;
9673 uv = (args) ? va_arg(*args, int) : SvIV(argsv);
9675 (!UNI_IS_INVARIANT(uv) && SvUTF8(sv)))
9677 eptr = (char*)utf8buf;
9678 elen = uvchr_to_utf8((U8*)eptr, uv) - utf8buf;
9692 eptr = va_arg(*args, char*);
9694 #ifdef MACOS_TRADITIONAL
9695 /* On MacOS, %#s format is used for Pascal strings */
9700 elen = strlen(eptr);
9702 eptr = (char *)nullstr;
9703 elen = sizeof nullstr - 1;
9707 eptr = SvPV_const(argsv, elen);
9708 if (DO_UTF8(argsv)) {
9709 I32 old_precis = precis;
9710 if (has_precis && precis < elen) {
9712 sv_pos_u2b(argsv, &p, 0); /* sticks at end */
9715 if (width) { /* fudge width (can't fudge elen) */
9716 if (has_precis && precis < elen)
9717 width += precis - old_precis;
9719 width += elen - sv_len_utf8(argsv);
9726 if (has_precis && elen > precis)
9733 if (alt || vectorize)
9735 uv = PTR2UV(args ? va_arg(*args, void*) : argsv);
9756 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
9765 esignbuf[esignlen++] = plus;
9769 case 'h': iv = (short)va_arg(*args, int); break;
9770 case 'l': iv = va_arg(*args, long); break;
9771 case 'V': iv = va_arg(*args, IV); break;
9772 default: iv = va_arg(*args, int); break;
9775 iv = va_arg(*args, Quad_t); break;
9782 IV tiv = SvIV(argsv); /* work around GCC bug #13488 */
9784 case 'h': iv = (short)tiv; break;
9785 case 'l': iv = (long)tiv; break;
9787 default: iv = tiv; break;
9790 iv = (Quad_t)tiv; break;
9796 if ( !vectorize ) /* we already set uv above */
9801 esignbuf[esignlen++] = plus;
9805 esignbuf[esignlen++] = '-';
9849 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
9860 case 'h': uv = (unsigned short)va_arg(*args, unsigned); break;
9861 case 'l': uv = va_arg(*args, unsigned long); break;
9862 case 'V': uv = va_arg(*args, UV); break;
9863 default: uv = va_arg(*args, unsigned); break;
9866 uv = va_arg(*args, Uquad_t); break;
9873 UV tuv = SvUV(argsv); /* work around GCC bug #13488 */
9875 case 'h': uv = (unsigned short)tuv; break;
9876 case 'l': uv = (unsigned long)tuv; break;
9878 default: uv = tuv; break;
9881 uv = (Uquad_t)tuv; break;
9890 char *ptr = ebuf + sizeof ebuf;
9891 bool tempalt = uv ? alt : FALSE; /* Vectors can't change alt */
9897 p = (char *)((c == 'X') ? PL_hexdigit + 16 : PL_hexdigit);
9903 esignbuf[esignlen++] = '0';
9904 esignbuf[esignlen++] = c; /* 'x' or 'X' */
9912 if (alt && *ptr != '0')
9921 esignbuf[esignlen++] = '0';
9922 esignbuf[esignlen++] = c;
9925 default: /* it had better be ten or less */
9929 } while (uv /= base);
9932 elen = (ebuf + sizeof ebuf) - ptr;
9936 zeros = precis - elen;
9937 else if (precis == 0 && elen == 1 && *eptr == '0'
9938 && !(base == 8 && alt)) /* "%#.0o" prints "0" */
9941 /* a precision nullifies the 0 flag. */
9948 /* FLOATING POINT */
9951 c = 'f'; /* maybe %F isn't supported here */
9959 /* This is evil, but floating point is even more evil */
9961 /* for SV-style calling, we can only get NV
9962 for C-style calling, we assume %f is double;
9963 for simplicity we allow any of %Lf, %llf, %qf for long double
9967 #if defined(USE_LONG_DOUBLE)
9971 /* [perl #20339] - we should accept and ignore %lf rather than die */
9975 #if defined(USE_LONG_DOUBLE)
9976 intsize = args ? 0 : 'q';
9980 #if defined(HAS_LONG_DOUBLE)
9989 /* now we need (long double) if intsize == 'q', else (double) */
9991 #if LONG_DOUBLESIZE > DOUBLESIZE
9993 va_arg(*args, long double) :
9994 va_arg(*args, double)
9996 va_arg(*args, double)
10001 /* nv * 0 will be NaN for NaN, +Inf and -Inf, and 0 for anything
10002 else. frexp() has some unspecified behaviour for those three */
10003 if (c != 'e' && c != 'E' && (nv * 0) == 0) {
10005 /* FIXME: if HAS_LONG_DOUBLE but not USE_LONG_DOUBLE this
10006 will cast our (long double) to (double) */
10007 (void)Perl_frexp(nv, &i);
10008 if (i == PERL_INT_MIN)
10009 Perl_die(aTHX_ "panic: frexp");
10011 need = BIT_DIGITS(i);
10013 need += has_precis ? precis : 6; /* known default */
10018 #ifdef HAS_LDBL_SPRINTF_BUG
10019 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
10020 with sfio - Allen <allens@cpan.org> */
10023 # define MY_DBL_MAX DBL_MAX
10024 # else /* XXX guessing! HUGE_VAL may be defined as infinity, so not using */
10025 # if DOUBLESIZE >= 8
10026 # define MY_DBL_MAX 1.7976931348623157E+308L
10028 # define MY_DBL_MAX 3.40282347E+38L
10032 # ifdef HAS_LDBL_SPRINTF_BUG_LESS1 /* only between -1L & 1L - Allen */
10033 # define MY_DBL_MAX_BUG 1L
10035 # define MY_DBL_MAX_BUG MY_DBL_MAX
10039 # define MY_DBL_MIN DBL_MIN
10040 # else /* XXX guessing! -Allen */
10041 # if DOUBLESIZE >= 8
10042 # define MY_DBL_MIN 2.2250738585072014E-308L
10044 # define MY_DBL_MIN 1.17549435E-38L
10048 if ((intsize == 'q') && (c == 'f') &&
10049 ((nv < MY_DBL_MAX_BUG) && (nv > -MY_DBL_MAX_BUG)) &&
10050 (need < DBL_DIG)) {
10051 /* it's going to be short enough that
10052 * long double precision is not needed */
10054 if ((nv <= 0L) && (nv >= -0L))
10055 fix_ldbl_sprintf_bug = TRUE; /* 0 is 0 - easiest */
10057 /* would use Perl_fp_class as a double-check but not
10058 * functional on IRIX - see perl.h comments */
10060 if ((nv >= MY_DBL_MIN) || (nv <= -MY_DBL_MIN)) {
10061 /* It's within the range that a double can represent */
10062 #if defined(DBL_MAX) && !defined(DBL_MIN)
10063 if ((nv >= ((long double)1/DBL_MAX)) ||
10064 (nv <= (-(long double)1/DBL_MAX)))
10066 fix_ldbl_sprintf_bug = TRUE;
10069 if (fix_ldbl_sprintf_bug == TRUE) {
10079 # undef MY_DBL_MAX_BUG
10082 #endif /* HAS_LDBL_SPRINTF_BUG */
10084 need += 20; /* fudge factor */
10085 if (PL_efloatsize < need) {
10086 Safefree(PL_efloatbuf);
10087 PL_efloatsize = need + 20; /* more fudge */
10088 Newx(PL_efloatbuf, PL_efloatsize, char);
10089 PL_efloatbuf[0] = '\0';
10092 if ( !(width || left || plus || alt) && fill != '0'
10093 && has_precis && intsize != 'q' ) { /* Shortcuts */
10094 /* See earlier comment about buggy Gconvert when digits,
10096 if ( c == 'g' && precis) {
10097 Gconvert((NV)nv, (int)precis, 0, PL_efloatbuf);
10098 /* May return an empty string for digits==0 */
10099 if (*PL_efloatbuf) {
10100 elen = strlen(PL_efloatbuf);
10101 goto float_converted;
10103 } else if ( c == 'f' && !precis) {
10104 if ((eptr = F0convert(nv, ebuf + sizeof ebuf, &elen)))
10109 char *ptr = ebuf + sizeof ebuf;
10112 /* FIXME: what to do if HAS_LONG_DOUBLE but not PERL_PRIfldbl? */
10113 #if defined(HAS_LONG_DOUBLE) && defined(PERL_PRIfldbl)
10114 if (intsize == 'q') {
10115 /* Copy the one or more characters in a long double
10116 * format before the 'base' ([efgEFG]) character to
10117 * the format string. */
10118 static char const prifldbl[] = PERL_PRIfldbl;
10119 char const *p = prifldbl + sizeof(prifldbl) - 3;
10120 while (p >= prifldbl) { *--ptr = *p--; }
10125 do { *--ptr = '0' + (base % 10); } while (base /= 10);
10130 do { *--ptr = '0' + (base % 10); } while (base /= 10);
10142 /* No taint. Otherwise we are in the strange situation
10143 * where printf() taints but print($float) doesn't.
10145 #if defined(HAS_LONG_DOUBLE)
10146 elen = ((intsize == 'q')
10147 ? my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, nv)
10148 : my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, (double)nv));
10150 elen = my_sprintf(PL_efloatbuf, ptr, nv);
10154 eptr = PL_efloatbuf;
10162 i = SvCUR(sv) - origlen;
10165 case 'h': *(va_arg(*args, short*)) = i; break;
10166 default: *(va_arg(*args, int*)) = i; break;
10167 case 'l': *(va_arg(*args, long*)) = i; break;
10168 case 'V': *(va_arg(*args, IV*)) = i; break;
10171 *(va_arg(*args, Quad_t*)) = i; break;
10178 sv_setuv_mg(argsv, (UV)i);
10179 continue; /* not "break" */
10186 && (PL_op->op_type == OP_PRTF || PL_op->op_type == OP_SPRINTF)
10187 && ckWARN(WARN_PRINTF))
10189 SV * const msg = sv_newmortal();
10190 Perl_sv_setpvf(aTHX_ msg, "Invalid conversion in %sprintf: ",
10191 (PL_op->op_type == OP_PRTF) ? "" : "s");
10192 if (fmtstart < patend) {
10193 const char * const fmtend = q < patend ? q : patend;
10195 sv_catpvs(msg, "\"%");
10196 for (f = fmtstart; f < fmtend; f++) {
10198 sv_catpvn(msg, f, 1);
10200 Perl_sv_catpvf(aTHX_ msg,
10201 "\\%03"UVof, (UV)*f & 0xFF);
10204 sv_catpvs(msg, "\"");
10206 sv_catpvs(msg, "end of string");
10208 Perl_warner(aTHX_ packWARN(WARN_PRINTF), "%"SVf, SVfARG(msg)); /* yes, this is reentrant */
10211 /* output mangled stuff ... */
10217 /* ... right here, because formatting flags should not apply */
10218 SvGROW(sv, SvCUR(sv) + elen + 1);
10220 Copy(eptr, p, elen, char);
10223 SvCUR_set(sv, p - SvPVX_const(sv));
10225 continue; /* not "break" */
10228 if (is_utf8 != has_utf8) {
10231 sv_utf8_upgrade(sv);
10234 const STRLEN old_elen = elen;
10235 SV * const nsv = newSVpvn_flags(eptr, elen, SVs_TEMP);
10236 sv_utf8_upgrade(nsv);
10237 eptr = SvPVX_const(nsv);
10240 if (width) { /* fudge width (can't fudge elen) */
10241 width += elen - old_elen;
10247 have = esignlen + zeros + elen;
10249 Perl_croak_nocontext("%s", PL_memory_wrap);
10251 need = (have > width ? have : width);
10254 if (need >= (((STRLEN)~0) - SvCUR(sv) - dotstrlen - 1))
10255 Perl_croak_nocontext("%s", PL_memory_wrap);
10256 SvGROW(sv, SvCUR(sv) + need + dotstrlen + 1);
10258 if (esignlen && fill == '0') {
10260 for (i = 0; i < (int)esignlen; i++)
10261 *p++ = esignbuf[i];
10263 if (gap && !left) {
10264 memset(p, fill, gap);
10267 if (esignlen && fill != '0') {
10269 for (i = 0; i < (int)esignlen; i++)
10270 *p++ = esignbuf[i];
10274 for (i = zeros; i; i--)
10278 Copy(eptr, p, elen, char);
10282 memset(p, ' ', gap);
10287 Copy(dotstr, p, dotstrlen, char);
10291 vectorize = FALSE; /* done iterating over vecstr */
10298 SvCUR_set(sv, p - SvPVX_const(sv));
10306 /* =========================================================================
10308 =head1 Cloning an interpreter
10310 All the macros and functions in this section are for the private use of
10311 the main function, perl_clone().
10313 The foo_dup() functions make an exact copy of an existing foo thingy.
10314 During the course of a cloning, a hash table is used to map old addresses
10315 to new addresses. The table is created and manipulated with the
10316 ptr_table_* functions.
10320 ============================================================================*/
10323 #if defined(USE_ITHREADS)
10325 /* XXX Remove this so it doesn't have to go thru the macro and return for nothing */
10326 #ifndef GpREFCNT_inc
10327 # define GpREFCNT_inc(gp) ((gp) ? (++(gp)->gp_refcnt, (gp)) : (GP*)NULL)
10331 /* Certain cases in Perl_ss_dup have been merged, by relying on the fact
10332 that currently av_dup, gv_dup and hv_dup are the same as sv_dup.
10333 If this changes, please unmerge ss_dup. */
10334 #define sv_dup_inc(s,t) SvREFCNT_inc(sv_dup(s,t))
10335 #define sv_dup_inc_NN(s,t) SvREFCNT_inc_NN(sv_dup(s,t))
10336 #define av_dup(s,t) MUTABLE_AV(sv_dup((const SV *)s,t))
10337 #define av_dup_inc(s,t) MUTABLE_AV(SvREFCNT_inc(sv_dup((const SV *)s,t)))
10338 #define hv_dup(s,t) MUTABLE_HV(sv_dup((const SV *)s,t))
10339 #define hv_dup_inc(s,t) MUTABLE_HV(SvREFCNT_inc(sv_dup((const SV *)s,t)))
10340 #define cv_dup(s,t) MUTABLE_CV(sv_dup((const SV *)s,t))
10341 #define cv_dup_inc(s,t) MUTABLE_CV(SvREFCNT_inc(sv_dup((const SV *)s,t)))
10342 #define io_dup(s,t) MUTABLE_IO(sv_dup((const SV *)s,t))
10343 #define io_dup_inc(s,t) MUTABLE_IO(SvREFCNT_inc(sv_dup((const SV *)s,t)))
10344 #define gv_dup(s,t) MUTABLE_GV(sv_dup((const SV *)s,t))
10345 #define gv_dup_inc(s,t) MUTABLE_GV(SvREFCNT_inc(sv_dup((const SV *)s,t)))
10346 #define SAVEPV(p) ((p) ? savepv(p) : NULL)
10347 #define SAVEPVN(p,n) ((p) ? savepvn(p,n) : NULL)
10349 /* clone a parser */
10352 Perl_parser_dup(pTHX_ const yy_parser *const proto, CLONE_PARAMS *const param)
10356 PERL_ARGS_ASSERT_PARSER_DUP;
10361 /* look for it in the table first */
10362 parser = (yy_parser *)ptr_table_fetch(PL_ptr_table, proto);
10366 /* create anew and remember what it is */
10367 Newxz(parser, 1, yy_parser);
10368 ptr_table_store(PL_ptr_table, proto, parser);
10370 parser->yyerrstatus = 0;
10371 parser->yychar = YYEMPTY; /* Cause a token to be read. */
10373 /* XXX these not yet duped */
10374 parser->old_parser = NULL;
10375 parser->stack = NULL;
10377 parser->stack_size = 0;
10378 /* XXX parser->stack->state = 0; */
10380 /* XXX eventually, just Copy() most of the parser struct ? */
10382 parser->lex_brackets = proto->lex_brackets;
10383 parser->lex_casemods = proto->lex_casemods;
10384 parser->lex_brackstack = savepvn(proto->lex_brackstack,
10385 (proto->lex_brackets < 120 ? 120 : proto->lex_brackets));
10386 parser->lex_casestack = savepvn(proto->lex_casestack,
10387 (proto->lex_casemods < 12 ? 12 : proto->lex_casemods));
10388 parser->lex_defer = proto->lex_defer;
10389 parser->lex_dojoin = proto->lex_dojoin;
10390 parser->lex_expect = proto->lex_expect;
10391 parser->lex_formbrack = proto->lex_formbrack;
10392 parser->lex_inpat = proto->lex_inpat;
10393 parser->lex_inwhat = proto->lex_inwhat;
10394 parser->lex_op = proto->lex_op;
10395 parser->lex_repl = sv_dup_inc(proto->lex_repl, param);
10396 parser->lex_starts = proto->lex_starts;
10397 parser->lex_stuff = sv_dup_inc(proto->lex_stuff, param);
10398 parser->multi_close = proto->multi_close;
10399 parser->multi_open = proto->multi_open;
10400 parser->multi_start = proto->multi_start;
10401 parser->multi_end = proto->multi_end;
10402 parser->pending_ident = proto->pending_ident;
10403 parser->preambled = proto->preambled;
10404 parser->sublex_info = proto->sublex_info; /* XXX not quite right */
10405 parser->linestr = sv_dup_inc(proto->linestr, param);
10406 parser->expect = proto->expect;
10407 parser->copline = proto->copline;
10408 parser->last_lop_op = proto->last_lop_op;
10409 parser->lex_state = proto->lex_state;
10410 parser->rsfp = fp_dup(proto->rsfp, '<', param);
10411 /* rsfp_filters entries have fake IoDIRP() */
10412 parser->rsfp_filters= av_dup_inc(proto->rsfp_filters, param);
10413 parser->in_my = proto->in_my;
10414 parser->in_my_stash = hv_dup(proto->in_my_stash, param);
10415 parser->error_count = proto->error_count;
10418 parser->linestr = sv_dup_inc(proto->linestr, param);
10421 char * const ols = SvPVX(proto->linestr);
10422 char * const ls = SvPVX(parser->linestr);
10424 parser->bufptr = ls + (proto->bufptr >= ols ?
10425 proto->bufptr - ols : 0);
10426 parser->oldbufptr = ls + (proto->oldbufptr >= ols ?
10427 proto->oldbufptr - ols : 0);
10428 parser->oldoldbufptr= ls + (proto->oldoldbufptr >= ols ?
10429 proto->oldoldbufptr - ols : 0);
10430 parser->linestart = ls + (proto->linestart >= ols ?
10431 proto->linestart - ols : 0);
10432 parser->last_uni = ls + (proto->last_uni >= ols ?
10433 proto->last_uni - ols : 0);
10434 parser->last_lop = ls + (proto->last_lop >= ols ?
10435 proto->last_lop - ols : 0);
10437 parser->bufend = ls + SvCUR(parser->linestr);
10440 Copy(proto->tokenbuf, parser->tokenbuf, 256, char);
10444 parser->endwhite = proto->endwhite;
10445 parser->faketokens = proto->faketokens;
10446 parser->lasttoke = proto->lasttoke;
10447 parser->nextwhite = proto->nextwhite;
10448 parser->realtokenstart = proto->realtokenstart;
10449 parser->skipwhite = proto->skipwhite;
10450 parser->thisclose = proto->thisclose;
10451 parser->thismad = proto->thismad;
10452 parser->thisopen = proto->thisopen;
10453 parser->thisstuff = proto->thisstuff;
10454 parser->thistoken = proto->thistoken;
10455 parser->thiswhite = proto->thiswhite;
10457 Copy(proto->nexttoke, parser->nexttoke, 5, NEXTTOKE);
10458 parser->curforce = proto->curforce;
10460 Copy(proto->nextval, parser->nextval, 5, YYSTYPE);
10461 Copy(proto->nexttype, parser->nexttype, 5, I32);
10462 parser->nexttoke = proto->nexttoke;
10468 /* duplicate a file handle */
10471 Perl_fp_dup(pTHX_ PerlIO *const fp, const char type, CLONE_PARAMS *const param)
10475 PERL_ARGS_ASSERT_FP_DUP;
10476 PERL_UNUSED_ARG(type);
10479 return (PerlIO*)NULL;
10481 /* look for it in the table first */
10482 ret = (PerlIO*)ptr_table_fetch(PL_ptr_table, fp);
10486 /* create anew and remember what it is */
10487 ret = PerlIO_fdupopen(aTHX_ fp, param, PERLIO_DUP_CLONE);
10488 ptr_table_store(PL_ptr_table, fp, ret);
10492 /* duplicate a directory handle */
10495 Perl_dirp_dup(pTHX_ DIR *const dp)
10497 PERL_UNUSED_CONTEXT;
10504 /* duplicate a typeglob */
10507 Perl_gp_dup(pTHX_ GP *const gp, CLONE_PARAMS *const param)
10511 PERL_ARGS_ASSERT_GP_DUP;
10515 /* look for it in the table first */
10516 ret = (GP*)ptr_table_fetch(PL_ptr_table, gp);
10520 /* create anew and remember what it is */
10522 ptr_table_store(PL_ptr_table, gp, ret);
10525 ret->gp_refcnt = 0; /* must be before any other dups! */
10526 ret->gp_sv = sv_dup_inc(gp->gp_sv, param);
10527 ret->gp_io = io_dup_inc(gp->gp_io, param);
10528 ret->gp_form = cv_dup_inc(gp->gp_form, param);
10529 ret->gp_av = av_dup_inc(gp->gp_av, param);
10530 ret->gp_hv = hv_dup_inc(gp->gp_hv, param);
10531 ret->gp_egv = gv_dup(gp->gp_egv, param);/* GvEGV is not refcounted */
10532 ret->gp_cv = cv_dup_inc(gp->gp_cv, param);
10533 ret->gp_cvgen = gp->gp_cvgen;
10534 ret->gp_line = gp->gp_line;
10535 ret->gp_file_hek = hek_dup(gp->gp_file_hek, param);
10539 /* duplicate a chain of magic */
10542 Perl_mg_dup(pTHX_ MAGIC *mg, CLONE_PARAMS *const param)
10544 MAGIC *mgprev = (MAGIC*)NULL;
10547 PERL_ARGS_ASSERT_MG_DUP;
10550 return (MAGIC*)NULL;
10551 /* look for it in the table first */
10552 mgret = (MAGIC*)ptr_table_fetch(PL_ptr_table, mg);
10556 for (; mg; mg = mg->mg_moremagic) {
10558 Newxz(nmg, 1, MAGIC);
10560 mgprev->mg_moremagic = nmg;
10563 nmg->mg_virtual = mg->mg_virtual; /* XXX copy dynamic vtable? */
10564 nmg->mg_private = mg->mg_private;
10565 nmg->mg_type = mg->mg_type;
10566 nmg->mg_flags = mg->mg_flags;
10567 /* FIXME for plugins
10568 if (mg->mg_type == PERL_MAGIC_qr) {
10569 nmg->mg_obj = MUTABLE_SV(CALLREGDUPE((REGEXP*)mg->mg_obj, param));
10573 if(mg->mg_type == PERL_MAGIC_backref) {
10574 /* The backref AV has its reference count deliberately bumped by
10577 = SvREFCNT_inc(av_dup_inc((const AV *) mg->mg_obj, param));
10580 nmg->mg_obj = (mg->mg_flags & MGf_REFCOUNTED)
10581 ? sv_dup_inc(mg->mg_obj, param)
10582 : sv_dup(mg->mg_obj, param);
10584 nmg->mg_len = mg->mg_len;
10585 nmg->mg_ptr = mg->mg_ptr; /* XXX random ptr? */
10586 if (mg->mg_ptr && mg->mg_type != PERL_MAGIC_regex_global) {
10587 if (mg->mg_len > 0) {
10588 nmg->mg_ptr = SAVEPVN(mg->mg_ptr, mg->mg_len);
10589 if (mg->mg_type == PERL_MAGIC_overload_table &&
10590 AMT_AMAGIC((AMT*)mg->mg_ptr))
10592 const AMT * const amtp = (AMT*)mg->mg_ptr;
10593 AMT * const namtp = (AMT*)nmg->mg_ptr;
10595 for (i = 1; i < NofAMmeth; i++) {
10596 namtp->table[i] = cv_dup_inc(amtp->table[i], param);
10600 else if (mg->mg_len == HEf_SVKEY)
10601 nmg->mg_ptr = (char*)sv_dup_inc((const SV *)mg->mg_ptr, param);
10603 if ((mg->mg_flags & MGf_DUP) && mg->mg_virtual && mg->mg_virtual->svt_dup) {
10604 CALL_FPTR(nmg->mg_virtual->svt_dup)(aTHX_ nmg, param);
10611 #endif /* USE_ITHREADS */
10613 /* create a new pointer-mapping table */
10616 Perl_ptr_table_new(pTHX)
10619 PERL_UNUSED_CONTEXT;
10621 Newxz(tbl, 1, PTR_TBL_t);
10622 tbl->tbl_max = 511;
10623 tbl->tbl_items = 0;
10624 Newxz(tbl->tbl_ary, tbl->tbl_max + 1, PTR_TBL_ENT_t*);
10628 #define PTR_TABLE_HASH(ptr) \
10629 ((PTR2UV(ptr) >> 3) ^ (PTR2UV(ptr) >> (3 + 7)) ^ (PTR2UV(ptr) >> (3 + 17)))
10632 we use the PTE_SVSLOT 'reservation' made above, both here (in the
10633 following define) and at call to new_body_inline made below in
10634 Perl_ptr_table_store()
10637 #define del_pte(p) del_body_type(p, PTE_SVSLOT)
10639 /* map an existing pointer using a table */
10641 STATIC PTR_TBL_ENT_t *
10642 S_ptr_table_find(PTR_TBL_t *const tbl, const void *const sv)
10644 PTR_TBL_ENT_t *tblent;
10645 const UV hash = PTR_TABLE_HASH(sv);
10647 PERL_ARGS_ASSERT_PTR_TABLE_FIND;
10649 tblent = tbl->tbl_ary[hash & tbl->tbl_max];
10650 for (; tblent; tblent = tblent->next) {
10651 if (tblent->oldval == sv)
10658 Perl_ptr_table_fetch(pTHX_ PTR_TBL_t *const tbl, const void *const sv)
10660 PTR_TBL_ENT_t const *const tblent = ptr_table_find(tbl, sv);
10662 PERL_ARGS_ASSERT_PTR_TABLE_FETCH;
10663 PERL_UNUSED_CONTEXT;
10665 return tblent ? tblent->newval : NULL;
10668 /* add a new entry to a pointer-mapping table */
10671 Perl_ptr_table_store(pTHX_ PTR_TBL_t *const tbl, const void *const oldsv, void *const newsv)
10673 PTR_TBL_ENT_t *tblent = ptr_table_find(tbl, oldsv);
10675 PERL_ARGS_ASSERT_PTR_TABLE_STORE;
10676 PERL_UNUSED_CONTEXT;
10679 tblent->newval = newsv;
10681 const UV entry = PTR_TABLE_HASH(oldsv) & tbl->tbl_max;
10683 new_body_inline(tblent, PTE_SVSLOT);
10685 tblent->oldval = oldsv;
10686 tblent->newval = newsv;
10687 tblent->next = tbl->tbl_ary[entry];
10688 tbl->tbl_ary[entry] = tblent;
10690 if (tblent->next && tbl->tbl_items > tbl->tbl_max)
10691 ptr_table_split(tbl);
10695 /* double the hash bucket size of an existing ptr table */
10698 Perl_ptr_table_split(pTHX_ PTR_TBL_t *const tbl)
10700 PTR_TBL_ENT_t **ary = tbl->tbl_ary;
10701 const UV oldsize = tbl->tbl_max + 1;
10702 UV newsize = oldsize * 2;
10705 PERL_ARGS_ASSERT_PTR_TABLE_SPLIT;
10706 PERL_UNUSED_CONTEXT;
10708 Renew(ary, newsize, PTR_TBL_ENT_t*);
10709 Zero(&ary[oldsize], newsize-oldsize, PTR_TBL_ENT_t*);
10710 tbl->tbl_max = --newsize;
10711 tbl->tbl_ary = ary;
10712 for (i=0; i < oldsize; i++, ary++) {
10713 PTR_TBL_ENT_t **curentp, **entp, *ent;
10716 curentp = ary + oldsize;
10717 for (entp = ary, ent = *ary; ent; ent = *entp) {
10718 if ((newsize & PTR_TABLE_HASH(ent->oldval)) != i) {
10720 ent->next = *curentp;
10730 /* remove all the entries from a ptr table */
10733 Perl_ptr_table_clear(pTHX_ PTR_TBL_t *const tbl)
10735 if (tbl && tbl->tbl_items) {
10736 register PTR_TBL_ENT_t * const * const array = tbl->tbl_ary;
10737 UV riter = tbl->tbl_max;
10740 PTR_TBL_ENT_t *entry = array[riter];
10743 PTR_TBL_ENT_t * const oentry = entry;
10744 entry = entry->next;
10749 tbl->tbl_items = 0;
10753 /* clear and free a ptr table */
10756 Perl_ptr_table_free(pTHX_ PTR_TBL_t *const tbl)
10761 ptr_table_clear(tbl);
10762 Safefree(tbl->tbl_ary);
10766 #if defined(USE_ITHREADS)
10769 Perl_rvpv_dup(pTHX_ SV *const dstr, const SV *const sstr, CLONE_PARAMS *const param)
10771 PERL_ARGS_ASSERT_RVPV_DUP;
10774 SvRV_set(dstr, SvWEAKREF(sstr)
10775 ? sv_dup(SvRV_const(sstr), param)
10776 : sv_dup_inc(SvRV_const(sstr), param));
10779 else if (SvPVX_const(sstr)) {
10780 /* Has something there */
10782 /* Normal PV - clone whole allocated space */
10783 SvPV_set(dstr, SAVEPVN(SvPVX_const(sstr), SvLEN(sstr)-1));
10784 if (SvREADONLY(sstr) && SvFAKE(sstr)) {
10785 /* Not that normal - actually sstr is copy on write.
10786 But we are a true, independant SV, so: */
10787 SvREADONLY_off(dstr);
10792 /* Special case - not normally malloced for some reason */
10793 if (isGV_with_GP(sstr)) {
10794 /* Don't need to do anything here. */
10796 else if ((SvREADONLY(sstr) && SvFAKE(sstr))) {
10797 /* A "shared" PV - clone it as "shared" PV */
10799 HEK_KEY(hek_dup(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)),
10803 /* Some other special case - random pointer */
10804 SvPV_set(dstr, (char *) SvPVX_const(sstr));
10809 /* Copy the NULL */
10810 SvPV_set(dstr, NULL);
10814 /* duplicate an SV of any type (including AV, HV etc) */
10817 Perl_sv_dup(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
10822 PERL_ARGS_ASSERT_SV_DUP;
10826 if (SvTYPE(sstr) == SVTYPEMASK) {
10827 #ifdef DEBUG_LEAKING_SCALARS_ABORT
10832 /* look for it in the table first */
10833 dstr = MUTABLE_SV(ptr_table_fetch(PL_ptr_table, sstr));
10837 if(param->flags & CLONEf_JOIN_IN) {
10838 /** We are joining here so we don't want do clone
10839 something that is bad **/
10840 if (SvTYPE(sstr) == SVt_PVHV) {
10841 const HEK * const hvname = HvNAME_HEK(sstr);
10843 /** don't clone stashes if they already exist **/
10844 return MUTABLE_SV(gv_stashpvn(HEK_KEY(hvname), HEK_LEN(hvname), 0));
10848 /* create anew and remember what it is */
10851 #ifdef DEBUG_LEAKING_SCALARS
10852 dstr->sv_debug_optype = sstr->sv_debug_optype;
10853 dstr->sv_debug_line = sstr->sv_debug_line;
10854 dstr->sv_debug_inpad = sstr->sv_debug_inpad;
10855 dstr->sv_debug_cloned = 1;
10856 dstr->sv_debug_file = savepv(sstr->sv_debug_file);
10859 ptr_table_store(PL_ptr_table, sstr, dstr);
10862 SvFLAGS(dstr) = SvFLAGS(sstr);
10863 SvFLAGS(dstr) &= ~SVf_OOK; /* don't propagate OOK hack */
10864 SvREFCNT(dstr) = 0; /* must be before any other dups! */
10867 if (SvANY(sstr) && PL_watch_pvx && SvPVX_const(sstr) == PL_watch_pvx)
10868 PerlIO_printf(Perl_debug_log, "watch at %p hit, found string \"%s\"\n",
10869 (void*)PL_watch_pvx, SvPVX_const(sstr));
10872 /* don't clone objects whose class has asked us not to */
10873 if (SvOBJECT(sstr) && ! (SvFLAGS(SvSTASH(sstr)) & SVphv_CLONEABLE)) {
10878 switch (SvTYPE(sstr)) {
10880 SvANY(dstr) = NULL;
10883 SvANY(dstr) = (XPVIV*)((char*)&(dstr->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
10885 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
10887 SvIV_set(dstr, SvIVX(sstr));
10891 SvANY(dstr) = new_XNV();
10892 SvNV_set(dstr, SvNVX(sstr));
10894 /* case SVt_BIND: */
10897 /* These are all the types that need complex bodies allocating. */
10899 const svtype sv_type = SvTYPE(sstr);
10900 const struct body_details *const sv_type_details
10901 = bodies_by_type + sv_type;
10905 Perl_croak(aTHX_ "Bizarre SvTYPE [%" IVdf "]", (IV)SvTYPE(sstr));
10909 if (GvUNIQUE((const GV *)sstr)) {
10910 NOOP; /* Do sharing here, and fall through */
10923 assert(sv_type_details->body_size);
10924 if (sv_type_details->arena) {
10925 new_body_inline(new_body, sv_type);
10927 = (void*)((char*)new_body - sv_type_details->offset);
10929 new_body = new_NOARENA(sv_type_details);
10933 SvANY(dstr) = new_body;
10936 Copy(((char*)SvANY(sstr)) + sv_type_details->offset,
10937 ((char*)SvANY(dstr)) + sv_type_details->offset,
10938 sv_type_details->copy, char);
10940 Copy(((char*)SvANY(sstr)),
10941 ((char*)SvANY(dstr)),
10942 sv_type_details->body_size + sv_type_details->offset, char);
10945 if (sv_type != SVt_PVAV && sv_type != SVt_PVHV
10946 && !isGV_with_GP(dstr))
10947 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
10949 /* The Copy above means that all the source (unduplicated) pointers
10950 are now in the destination. We can check the flags and the
10951 pointers in either, but it's possible that there's less cache
10952 missing by always going for the destination.
10953 FIXME - instrument and check that assumption */
10954 if (sv_type >= SVt_PVMG) {
10955 if ((sv_type == SVt_PVMG) && SvPAD_OUR(dstr)) {
10956 SvOURSTASH_set(dstr, hv_dup_inc(SvOURSTASH(dstr), param));
10957 } else if (SvMAGIC(dstr))
10958 SvMAGIC_set(dstr, mg_dup(SvMAGIC(dstr), param));
10960 SvSTASH_set(dstr, hv_dup_inc(SvSTASH(dstr), param));
10963 /* The cast silences a GCC warning about unhandled types. */
10964 switch ((int)sv_type) {
10974 /* FIXME for plugins */
10975 re_dup_guts((REGEXP*) sstr, (REGEXP*) dstr, param);
10978 /* XXX LvTARGOFF sometimes holds PMOP* when DEBUGGING */
10979 if (LvTYPE(dstr) == 't') /* for tie: unrefcnted fake (SV**) */
10980 LvTARG(dstr) = dstr;
10981 else if (LvTYPE(dstr) == 'T') /* for tie: fake HE */
10982 LvTARG(dstr) = MUTABLE_SV(he_dup((HE*)LvTARG(dstr), 0, param));
10984 LvTARG(dstr) = sv_dup_inc(LvTARG(dstr), param);
10986 if(isGV_with_GP(sstr)) {
10987 if (GvNAME_HEK(dstr))
10988 GvNAME_HEK(dstr) = hek_dup(GvNAME_HEK(dstr), param);
10989 /* Don't call sv_add_backref here as it's going to be
10990 created as part of the magic cloning of the symbol
10992 /* Danger Will Robinson - GvGP(dstr) isn't initialised
10993 at the point of this comment. */
10994 GvSTASH(dstr) = hv_dup(GvSTASH(dstr), param);
10995 GvGP(dstr) = gp_dup(GvGP(sstr), param);
10996 (void)GpREFCNT_inc(GvGP(dstr));
10998 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
11001 IoIFP(dstr) = fp_dup(IoIFP(dstr), IoTYPE(dstr), param);
11002 if (IoOFP(dstr) == IoIFP(sstr))
11003 IoOFP(dstr) = IoIFP(dstr);
11005 IoOFP(dstr) = fp_dup(IoOFP(dstr), IoTYPE(dstr), param);
11006 /* PL_parser->rsfp_filters entries have fake IoDIRP() */
11007 if(IoFLAGS(dstr) & IOf_FAKE_DIRP) {
11008 /* I have no idea why fake dirp (rsfps)
11009 should be treated differently but otherwise
11010 we end up with leaks -- sky*/
11011 IoTOP_GV(dstr) = gv_dup_inc(IoTOP_GV(dstr), param);
11012 IoFMT_GV(dstr) = gv_dup_inc(IoFMT_GV(dstr), param);
11013 IoBOTTOM_GV(dstr) = gv_dup_inc(IoBOTTOM_GV(dstr), param);
11015 IoTOP_GV(dstr) = gv_dup(IoTOP_GV(dstr), param);
11016 IoFMT_GV(dstr) = gv_dup(IoFMT_GV(dstr), param);
11017 IoBOTTOM_GV(dstr) = gv_dup(IoBOTTOM_GV(dstr), param);
11018 if (IoDIRP(dstr)) {
11019 IoDIRP(dstr) = dirp_dup(IoDIRP(dstr));
11022 /* IoDIRP(dstr) is already a copy of IoDIRP(sstr) */
11025 IoTOP_NAME(dstr) = SAVEPV(IoTOP_NAME(dstr));
11026 IoFMT_NAME(dstr) = SAVEPV(IoFMT_NAME(dstr));
11027 IoBOTTOM_NAME(dstr) = SAVEPV(IoBOTTOM_NAME(dstr));
11030 /* avoid cloning an empty array */
11031 if (AvARRAY((const AV *)sstr) && AvFILLp((const AV *)sstr) >= 0) {
11032 SV **dst_ary, **src_ary;
11033 SSize_t items = AvFILLp((const AV *)sstr) + 1;
11035 src_ary = AvARRAY((const AV *)sstr);
11036 Newxz(dst_ary, AvMAX((const AV *)sstr)+1, SV*);
11037 ptr_table_store(PL_ptr_table, src_ary, dst_ary);
11038 AvARRAY(MUTABLE_AV(dstr)) = dst_ary;
11039 AvALLOC((const AV *)dstr) = dst_ary;
11040 if (AvREAL((const AV *)sstr)) {
11041 while (items-- > 0)
11042 *dst_ary++ = sv_dup_inc(*src_ary++, param);
11045 while (items-- > 0)
11046 *dst_ary++ = sv_dup(*src_ary++, param);
11048 items = AvMAX((const AV *)sstr) - AvFILLp((const AV *)sstr);
11049 while (items-- > 0) {
11050 *dst_ary++ = &PL_sv_undef;
11054 AvARRAY(MUTABLE_AV(dstr)) = NULL;
11055 AvALLOC((const AV *)dstr) = (SV**)NULL;
11056 AvMAX( (const AV *)dstr) = -1;
11057 AvFILLp((const AV *)dstr) = -1;
11061 if (HvARRAY((const HV *)sstr)) {
11063 const bool sharekeys = !!HvSHAREKEYS(sstr);
11064 XPVHV * const dxhv = (XPVHV*)SvANY(dstr);
11065 XPVHV * const sxhv = (XPVHV*)SvANY(sstr);
11067 Newx(darray, PERL_HV_ARRAY_ALLOC_BYTES(dxhv->xhv_max+1)
11068 + (SvOOK(sstr) ? sizeof(struct xpvhv_aux) : 0),
11070 HvARRAY(dstr) = (HE**)darray;
11071 while (i <= sxhv->xhv_max) {
11072 const HE * const source = HvARRAY(sstr)[i];
11073 HvARRAY(dstr)[i] = source
11074 ? he_dup(source, sharekeys, param) : 0;
11079 const struct xpvhv_aux * const saux = HvAUX(sstr);
11080 struct xpvhv_aux * const daux = HvAUX(dstr);
11081 /* This flag isn't copied. */
11082 /* SvOOK_on(hv) attacks the IV flags. */
11083 SvFLAGS(dstr) |= SVf_OOK;
11085 hvname = saux->xhv_name;
11086 daux->xhv_name = hvname ? hek_dup(hvname, param) : hvname;
11088 daux->xhv_riter = saux->xhv_riter;
11089 daux->xhv_eiter = saux->xhv_eiter
11090 ? he_dup(saux->xhv_eiter,
11091 (bool)!!HvSHAREKEYS(sstr), param) : 0;
11092 /* backref array needs refcnt=2; see sv_add_backref */
11093 daux->xhv_backreferences =
11094 saux->xhv_backreferences
11095 ? MUTABLE_AV(SvREFCNT_inc(
11096 sv_dup_inc((const SV *)saux->xhv_backreferences, param)))
11099 daux->xhv_mro_meta = saux->xhv_mro_meta
11100 ? mro_meta_dup(saux->xhv_mro_meta, param)
11103 /* Record stashes for possible cloning in Perl_clone(). */
11105 av_push(param->stashes, dstr);
11109 HvARRAY(MUTABLE_HV(dstr)) = NULL;
11112 if (!(param->flags & CLONEf_COPY_STACKS)) {
11116 /* NOTE: not refcounted */
11117 CvSTASH(dstr) = hv_dup(CvSTASH(dstr), param);
11119 if (!CvISXSUB(dstr))
11120 CvROOT(dstr) = OpREFCNT_inc(CvROOT(dstr));
11122 if (CvCONST(dstr) && CvISXSUB(dstr)) {
11123 CvXSUBANY(dstr).any_ptr = GvUNIQUE(CvGV(dstr)) ?
11124 SvREFCNT_inc(CvXSUBANY(dstr).any_ptr) :
11125 sv_dup_inc((const SV *)CvXSUBANY(dstr).any_ptr, param);
11127 /* don't dup if copying back - CvGV isn't refcounted, so the
11128 * duped GV may never be freed. A bit of a hack! DAPM */
11129 CvGV(dstr) = (param->flags & CLONEf_JOIN_IN) ?
11130 NULL : gv_dup(CvGV(dstr), param) ;
11131 PAD_DUP(CvPADLIST(dstr), CvPADLIST(sstr), param);
11133 CvWEAKOUTSIDE(sstr)
11134 ? cv_dup( CvOUTSIDE(dstr), param)
11135 : cv_dup_inc(CvOUTSIDE(dstr), param);
11136 if (!CvISXSUB(dstr))
11137 CvFILE(dstr) = SAVEPV(CvFILE(dstr));
11143 if (SvOBJECT(dstr) && SvTYPE(dstr) != SVt_PVIO)
11149 /* duplicate a context */
11152 Perl_cx_dup(pTHX_ PERL_CONTEXT *cxs, I32 ix, I32 max, CLONE_PARAMS* param)
11154 PERL_CONTEXT *ncxs;
11156 PERL_ARGS_ASSERT_CX_DUP;
11159 return (PERL_CONTEXT*)NULL;
11161 /* look for it in the table first */
11162 ncxs = (PERL_CONTEXT*)ptr_table_fetch(PL_ptr_table, cxs);
11166 /* create anew and remember what it is */
11167 Newx(ncxs, max + 1, PERL_CONTEXT);
11168 ptr_table_store(PL_ptr_table, cxs, ncxs);
11169 Copy(cxs, ncxs, max + 1, PERL_CONTEXT);
11172 PERL_CONTEXT * const ncx = &ncxs[ix];
11173 if (CxTYPE(ncx) == CXt_SUBST) {
11174 Perl_croak(aTHX_ "Cloning substitution context is unimplemented");
11177 switch (CxTYPE(ncx)) {
11179 ncx->blk_sub.cv = (ncx->blk_sub.olddepth == 0
11180 ? cv_dup_inc(ncx->blk_sub.cv, param)
11181 : cv_dup(ncx->blk_sub.cv,param));
11182 ncx->blk_sub.argarray = (CxHASARGS(ncx)
11183 ? av_dup_inc(ncx->blk_sub.argarray,
11186 ncx->blk_sub.savearray = av_dup_inc(ncx->blk_sub.savearray,
11188 ncx->blk_sub.oldcomppad = (PAD*)ptr_table_fetch(PL_ptr_table,
11189 ncx->blk_sub.oldcomppad);
11192 ncx->blk_eval.old_namesv = sv_dup_inc(ncx->blk_eval.old_namesv,
11194 ncx->blk_eval.cur_text = sv_dup(ncx->blk_eval.cur_text, param);
11196 case CXt_LOOP_LAZYSV:
11197 ncx->blk_loop.state_u.lazysv.end
11198 = sv_dup_inc(ncx->blk_loop.state_u.lazysv.end, param);
11199 /* We are taking advantage of av_dup_inc and sv_dup_inc
11200 actually being the same function, and order equivalance of
11202 We can assert the later [but only at run time :-(] */
11203 assert ((void *) &ncx->blk_loop.state_u.ary.ary ==
11204 (void *) &ncx->blk_loop.state_u.lazysv.cur);
11206 ncx->blk_loop.state_u.ary.ary
11207 = av_dup_inc(ncx->blk_loop.state_u.ary.ary, param);
11208 case CXt_LOOP_LAZYIV:
11209 case CXt_LOOP_PLAIN:
11210 if (CxPADLOOP(ncx)) {
11211 ncx->blk_loop.oldcomppad
11212 = (PAD*)ptr_table_fetch(PL_ptr_table,
11213 ncx->blk_loop.oldcomppad);
11215 ncx->blk_loop.oldcomppad
11216 = (PAD*)gv_dup((const GV *)ncx->blk_loop.oldcomppad,
11221 ncx->blk_format.cv = cv_dup(ncx->blk_format.cv, param);
11222 ncx->blk_format.gv = gv_dup(ncx->blk_format.gv, param);
11223 ncx->blk_format.dfoutgv = gv_dup_inc(ncx->blk_format.dfoutgv,
11236 /* duplicate a stack info structure */
11239 Perl_si_dup(pTHX_ PERL_SI *si, CLONE_PARAMS* param)
11243 PERL_ARGS_ASSERT_SI_DUP;
11246 return (PERL_SI*)NULL;
11248 /* look for it in the table first */
11249 nsi = (PERL_SI*)ptr_table_fetch(PL_ptr_table, si);
11253 /* create anew and remember what it is */
11254 Newxz(nsi, 1, PERL_SI);
11255 ptr_table_store(PL_ptr_table, si, nsi);
11257 nsi->si_stack = av_dup_inc(si->si_stack, param);
11258 nsi->si_cxix = si->si_cxix;
11259 nsi->si_cxmax = si->si_cxmax;
11260 nsi->si_cxstack = cx_dup(si->si_cxstack, si->si_cxix, si->si_cxmax, param);
11261 nsi->si_type = si->si_type;
11262 nsi->si_prev = si_dup(si->si_prev, param);
11263 nsi->si_next = si_dup(si->si_next, param);
11264 nsi->si_markoff = si->si_markoff;
11269 #define POPINT(ss,ix) ((ss)[--(ix)].any_i32)
11270 #define TOPINT(ss,ix) ((ss)[ix].any_i32)
11271 #define POPLONG(ss,ix) ((ss)[--(ix)].any_long)
11272 #define TOPLONG(ss,ix) ((ss)[ix].any_long)
11273 #define POPIV(ss,ix) ((ss)[--(ix)].any_iv)
11274 #define TOPIV(ss,ix) ((ss)[ix].any_iv)
11275 #define POPBOOL(ss,ix) ((ss)[--(ix)].any_bool)
11276 #define TOPBOOL(ss,ix) ((ss)[ix].any_bool)
11277 #define POPPTR(ss,ix) ((ss)[--(ix)].any_ptr)
11278 #define TOPPTR(ss,ix) ((ss)[ix].any_ptr)
11279 #define POPDPTR(ss,ix) ((ss)[--(ix)].any_dptr)
11280 #define TOPDPTR(ss,ix) ((ss)[ix].any_dptr)
11281 #define POPDXPTR(ss,ix) ((ss)[--(ix)].any_dxptr)
11282 #define TOPDXPTR(ss,ix) ((ss)[ix].any_dxptr)
11285 #define pv_dup_inc(p) SAVEPV(p)
11286 #define pv_dup(p) SAVEPV(p)
11287 #define svp_dup_inc(p,pp) any_dup(p,pp)
11289 /* map any object to the new equivent - either something in the
11290 * ptr table, or something in the interpreter structure
11294 Perl_any_dup(pTHX_ void *v, const PerlInterpreter *proto_perl)
11298 PERL_ARGS_ASSERT_ANY_DUP;
11301 return (void*)NULL;
11303 /* look for it in the table first */
11304 ret = ptr_table_fetch(PL_ptr_table, v);
11308 /* see if it is part of the interpreter structure */
11309 if (v >= (void*)proto_perl && v < (void*)(proto_perl+1))
11310 ret = (void*)(((char*)aTHX) + (((char*)v) - (char*)proto_perl));
11318 /* duplicate the save stack */
11321 Perl_ss_dup(pTHX_ PerlInterpreter *proto_perl, CLONE_PARAMS* param)
11324 ANY * const ss = proto_perl->Isavestack;
11325 const I32 max = proto_perl->Isavestack_max;
11326 I32 ix = proto_perl->Isavestack_ix;
11339 void (*dptr) (void*);
11340 void (*dxptr) (pTHX_ void*);
11342 PERL_ARGS_ASSERT_SS_DUP;
11344 Newxz(nss, max, ANY);
11347 const I32 type = POPINT(ss,ix);
11348 TOPINT(nss,ix) = type;
11350 case SAVEt_HELEM: /* hash element */
11351 sv = (const SV *)POPPTR(ss,ix);
11352 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11354 case SAVEt_ITEM: /* normal string */
11355 case SAVEt_SV: /* scalar reference */
11356 sv = (const SV *)POPPTR(ss,ix);
11357 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11360 case SAVEt_MORTALIZESV:
11361 sv = (const SV *)POPPTR(ss,ix);
11362 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11364 case SAVEt_SHARED_PVREF: /* char* in shared space */
11365 c = (char*)POPPTR(ss,ix);
11366 TOPPTR(nss,ix) = savesharedpv(c);
11367 ptr = POPPTR(ss,ix);
11368 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11370 case SAVEt_GENERIC_SVREF: /* generic sv */
11371 case SAVEt_SVREF: /* scalar reference */
11372 sv = (const SV *)POPPTR(ss,ix);
11373 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11374 ptr = POPPTR(ss,ix);
11375 TOPPTR(nss,ix) = svp_dup_inc((SV**)ptr, proto_perl);/* XXXXX */
11377 case SAVEt_HV: /* hash reference */
11378 case SAVEt_AV: /* array reference */
11379 sv = (const SV *) POPPTR(ss,ix);
11380 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11382 case SAVEt_COMPPAD:
11384 sv = (const SV *) POPPTR(ss,ix);
11385 TOPPTR(nss,ix) = sv_dup(sv, param);
11387 case SAVEt_INT: /* int reference */
11388 ptr = POPPTR(ss,ix);
11389 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11390 intval = (int)POPINT(ss,ix);
11391 TOPINT(nss,ix) = intval;
11393 case SAVEt_LONG: /* long reference */
11394 ptr = POPPTR(ss,ix);
11395 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11397 case SAVEt_CLEARSV:
11398 longval = (long)POPLONG(ss,ix);
11399 TOPLONG(nss,ix) = longval;
11401 case SAVEt_I32: /* I32 reference */
11402 case SAVEt_I16: /* I16 reference */
11403 case SAVEt_I8: /* I8 reference */
11404 case SAVEt_COP_ARYBASE: /* call CopARYBASE_set */
11405 ptr = POPPTR(ss,ix);
11406 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11408 TOPINT(nss,ix) = i;
11410 case SAVEt_IV: /* IV reference */
11411 ptr = POPPTR(ss,ix);
11412 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11414 TOPIV(nss,ix) = iv;
11416 case SAVEt_HPTR: /* HV* reference */
11417 case SAVEt_APTR: /* AV* reference */
11418 case SAVEt_SPTR: /* SV* reference */
11419 ptr = POPPTR(ss,ix);
11420 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11421 sv = (const SV *)POPPTR(ss,ix);
11422 TOPPTR(nss,ix) = sv_dup(sv, param);
11424 case SAVEt_VPTR: /* random* reference */
11425 ptr = POPPTR(ss,ix);
11426 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11427 ptr = POPPTR(ss,ix);
11428 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11430 case SAVEt_GENERIC_PVREF: /* generic char* */
11431 case SAVEt_PPTR: /* char* reference */
11432 ptr = POPPTR(ss,ix);
11433 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11434 c = (char*)POPPTR(ss,ix);
11435 TOPPTR(nss,ix) = pv_dup(c);
11437 case SAVEt_GP: /* scalar reference */
11438 gp = (GP*)POPPTR(ss,ix);
11439 TOPPTR(nss,ix) = gp = gp_dup(gp, param);
11440 (void)GpREFCNT_inc(gp);
11441 gv = (const GV *)POPPTR(ss,ix);
11442 TOPPTR(nss,ix) = gv_dup_inc(gv, param);
11445 ptr = POPPTR(ss,ix);
11446 if (ptr && (((OP*)ptr)->op_private & OPpREFCOUNTED)) {
11447 /* these are assumed to be refcounted properly */
11449 switch (((OP*)ptr)->op_type) {
11451 case OP_LEAVESUBLV:
11455 case OP_LEAVEWRITE:
11456 TOPPTR(nss,ix) = ptr;
11459 (void) OpREFCNT_inc(o);
11463 TOPPTR(nss,ix) = NULL;
11468 TOPPTR(nss,ix) = NULL;
11471 hv = (const HV *)POPPTR(ss,ix);
11472 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
11474 TOPINT(nss,ix) = i;
11477 c = (char*)POPPTR(ss,ix);
11478 TOPPTR(nss,ix) = pv_dup_inc(c);
11480 case SAVEt_STACK_POS: /* Position on Perl stack */
11482 TOPINT(nss,ix) = i;
11484 case SAVEt_DESTRUCTOR:
11485 ptr = POPPTR(ss,ix);
11486 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
11487 dptr = POPDPTR(ss,ix);
11488 TOPDPTR(nss,ix) = DPTR2FPTR(void (*)(void*),
11489 any_dup(FPTR2DPTR(void *, dptr),
11492 case SAVEt_DESTRUCTOR_X:
11493 ptr = POPPTR(ss,ix);
11494 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
11495 dxptr = POPDXPTR(ss,ix);
11496 TOPDXPTR(nss,ix) = DPTR2FPTR(void (*)(pTHX_ void*),
11497 any_dup(FPTR2DPTR(void *, dxptr),
11500 case SAVEt_REGCONTEXT:
11503 TOPINT(nss,ix) = i;
11506 case SAVEt_AELEM: /* array element */
11507 sv = (const SV *)POPPTR(ss,ix);
11508 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11510 TOPINT(nss,ix) = i;
11511 av = (const AV *)POPPTR(ss,ix);
11512 TOPPTR(nss,ix) = av_dup_inc(av, param);
11515 ptr = POPPTR(ss,ix);
11516 TOPPTR(nss,ix) = ptr;
11519 ptr = POPPTR(ss,ix);
11522 ((struct refcounted_he *)ptr)->refcounted_he_refcnt++;
11523 HINTS_REFCNT_UNLOCK;
11525 TOPPTR(nss,ix) = ptr;
11527 TOPINT(nss,ix) = i;
11528 if (i & HINT_LOCALIZE_HH) {
11529 hv = (const HV *)POPPTR(ss,ix);
11530 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
11533 case SAVEt_PADSV_AND_MORTALIZE:
11534 longval = (long)POPLONG(ss,ix);
11535 TOPLONG(nss,ix) = longval;
11536 ptr = POPPTR(ss,ix);
11537 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11538 sv = (const SV *)POPPTR(ss,ix);
11539 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11542 ptr = POPPTR(ss,ix);
11543 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11544 longval = (long)POPBOOL(ss,ix);
11545 TOPBOOL(nss,ix) = (bool)longval;
11547 case SAVEt_SET_SVFLAGS:
11549 TOPINT(nss,ix) = i;
11551 TOPINT(nss,ix) = i;
11552 sv = (const SV *)POPPTR(ss,ix);
11553 TOPPTR(nss,ix) = sv_dup(sv, param);
11555 case SAVEt_RE_STATE:
11557 const struct re_save_state *const old_state
11558 = (struct re_save_state *)
11559 (ss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
11560 struct re_save_state *const new_state
11561 = (struct re_save_state *)
11562 (nss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
11564 Copy(old_state, new_state, 1, struct re_save_state);
11565 ix -= SAVESTACK_ALLOC_FOR_RE_SAVE_STATE;
11567 new_state->re_state_bostr
11568 = pv_dup(old_state->re_state_bostr);
11569 new_state->re_state_reginput
11570 = pv_dup(old_state->re_state_reginput);
11571 new_state->re_state_regeol
11572 = pv_dup(old_state->re_state_regeol);
11573 new_state->re_state_regoffs
11574 = (regexp_paren_pair*)
11575 any_dup(old_state->re_state_regoffs, proto_perl);
11576 new_state->re_state_reglastparen
11577 = (U32*) any_dup(old_state->re_state_reglastparen,
11579 new_state->re_state_reglastcloseparen
11580 = (U32*)any_dup(old_state->re_state_reglastcloseparen,
11582 /* XXX This just has to be broken. The old save_re_context
11583 code did SAVEGENERICPV(PL_reg_start_tmp);
11584 PL_reg_start_tmp is char **.
11585 Look above to what the dup code does for
11586 SAVEt_GENERIC_PVREF
11587 It can never have worked.
11588 So this is merely a faithful copy of the exiting bug: */
11589 new_state->re_state_reg_start_tmp
11590 = (char **) pv_dup((char *)
11591 old_state->re_state_reg_start_tmp);
11592 /* I assume that it only ever "worked" because no-one called
11593 (pseudo)fork while the regexp engine had re-entered itself.
11595 #ifdef PERL_OLD_COPY_ON_WRITE
11596 new_state->re_state_nrs
11597 = sv_dup(old_state->re_state_nrs, param);
11599 new_state->re_state_reg_magic
11600 = (MAGIC*) any_dup(old_state->re_state_reg_magic,
11602 new_state->re_state_reg_oldcurpm
11603 = (PMOP*) any_dup(old_state->re_state_reg_oldcurpm,
11605 new_state->re_state_reg_curpm
11606 = (PMOP*) any_dup(old_state->re_state_reg_curpm,
11608 new_state->re_state_reg_oldsaved
11609 = pv_dup(old_state->re_state_reg_oldsaved);
11610 new_state->re_state_reg_poscache
11611 = pv_dup(old_state->re_state_reg_poscache);
11612 new_state->re_state_reg_starttry
11613 = pv_dup(old_state->re_state_reg_starttry);
11616 case SAVEt_COMPILE_WARNINGS:
11617 ptr = POPPTR(ss,ix);
11618 TOPPTR(nss,ix) = DUP_WARNINGS((STRLEN*)ptr);
11621 ptr = POPPTR(ss,ix);
11622 TOPPTR(nss,ix) = parser_dup((const yy_parser*)ptr, param);
11626 "panic: ss_dup inconsistency (%"IVdf")", (IV) type);
11634 /* if sv is a stash, call $class->CLONE_SKIP(), and set the SVphv_CLONEABLE
11635 * flag to the result. This is done for each stash before cloning starts,
11636 * so we know which stashes want their objects cloned */
11639 do_mark_cloneable_stash(pTHX_ SV *const sv)
11641 const HEK * const hvname = HvNAME_HEK((const HV *)sv);
11643 GV* const cloner = gv_fetchmethod_autoload(MUTABLE_HV(sv), "CLONE_SKIP", 0);
11644 SvFLAGS(sv) |= SVphv_CLONEABLE; /* clone objects by default */
11645 if (cloner && GvCV(cloner)) {
11652 mXPUSHs(newSVhek(hvname));
11654 call_sv(MUTABLE_SV(GvCV(cloner)), G_SCALAR);
11661 SvFLAGS(sv) &= ~SVphv_CLONEABLE;
11669 =for apidoc perl_clone
11671 Create and return a new interpreter by cloning the current one.
11673 perl_clone takes these flags as parameters:
11675 CLONEf_COPY_STACKS - is used to, well, copy the stacks also,
11676 without it we only clone the data and zero the stacks,
11677 with it we copy the stacks and the new perl interpreter is
11678 ready to run at the exact same point as the previous one.
11679 The pseudo-fork code uses COPY_STACKS while the
11680 threads->create doesn't.
11682 CLONEf_KEEP_PTR_TABLE
11683 perl_clone keeps a ptr_table with the pointer of the old
11684 variable as a key and the new variable as a value,
11685 this allows it to check if something has been cloned and not
11686 clone it again but rather just use the value and increase the
11687 refcount. If KEEP_PTR_TABLE is not set then perl_clone will kill
11688 the ptr_table using the function
11689 C<ptr_table_free(PL_ptr_table); PL_ptr_table = NULL;>,
11690 reason to keep it around is if you want to dup some of your own
11691 variable who are outside the graph perl scans, example of this
11692 code is in threads.xs create
11695 This is a win32 thing, it is ignored on unix, it tells perls
11696 win32host code (which is c++) to clone itself, this is needed on
11697 win32 if you want to run two threads at the same time,
11698 if you just want to do some stuff in a separate perl interpreter
11699 and then throw it away and return to the original one,
11700 you don't need to do anything.
11705 /* XXX the above needs expanding by someone who actually understands it ! */
11706 EXTERN_C PerlInterpreter *
11707 perl_clone_host(PerlInterpreter* proto_perl, UV flags);
11710 perl_clone(PerlInterpreter *proto_perl, UV flags)
11713 #ifdef PERL_IMPLICIT_SYS
11715 PERL_ARGS_ASSERT_PERL_CLONE;
11717 /* perlhost.h so we need to call into it
11718 to clone the host, CPerlHost should have a c interface, sky */
11720 if (flags & CLONEf_CLONE_HOST) {
11721 return perl_clone_host(proto_perl,flags);
11723 return perl_clone_using(proto_perl, flags,
11725 proto_perl->IMemShared,
11726 proto_perl->IMemParse,
11728 proto_perl->IStdIO,
11732 proto_perl->IProc);
11736 perl_clone_using(PerlInterpreter *proto_perl, UV flags,
11737 struct IPerlMem* ipM, struct IPerlMem* ipMS,
11738 struct IPerlMem* ipMP, struct IPerlEnv* ipE,
11739 struct IPerlStdIO* ipStd, struct IPerlLIO* ipLIO,
11740 struct IPerlDir* ipD, struct IPerlSock* ipS,
11741 struct IPerlProc* ipP)
11743 /* XXX many of the string copies here can be optimized if they're
11744 * constants; they need to be allocated as common memory and just
11745 * their pointers copied. */
11748 CLONE_PARAMS clone_params;
11749 CLONE_PARAMS* const param = &clone_params;
11751 PerlInterpreter * const my_perl = (PerlInterpreter*)(*ipM->pMalloc)(ipM, sizeof(PerlInterpreter));
11753 PERL_ARGS_ASSERT_PERL_CLONE_USING;
11755 /* for each stash, determine whether its objects should be cloned */
11756 S_visit(proto_perl, do_mark_cloneable_stash, SVt_PVHV, SVTYPEMASK);
11757 PERL_SET_THX(my_perl);
11760 PoisonNew(my_perl, 1, PerlInterpreter);
11766 PL_savestack_ix = 0;
11767 PL_savestack_max = -1;
11768 PL_sig_pending = 0;
11770 Zero(&PL_debug_pad, 1, struct perl_debug_pad);
11771 # else /* !DEBUGGING */
11772 Zero(my_perl, 1, PerlInterpreter);
11773 # endif /* DEBUGGING */
11775 /* host pointers */
11777 PL_MemShared = ipMS;
11778 PL_MemParse = ipMP;
11785 #else /* !PERL_IMPLICIT_SYS */
11787 CLONE_PARAMS clone_params;
11788 CLONE_PARAMS* param = &clone_params;
11789 PerlInterpreter * const my_perl = (PerlInterpreter*)PerlMem_malloc(sizeof(PerlInterpreter));
11791 PERL_ARGS_ASSERT_PERL_CLONE;
11793 /* for each stash, determine whether its objects should be cloned */
11794 S_visit(proto_perl, do_mark_cloneable_stash, SVt_PVHV, SVTYPEMASK);
11795 PERL_SET_THX(my_perl);
11798 PoisonNew(my_perl, 1, PerlInterpreter);
11804 PL_savestack_ix = 0;
11805 PL_savestack_max = -1;
11806 PL_sig_pending = 0;
11808 Zero(&PL_debug_pad, 1, struct perl_debug_pad);
11809 # else /* !DEBUGGING */
11810 Zero(my_perl, 1, PerlInterpreter);
11811 # endif /* DEBUGGING */
11812 #endif /* PERL_IMPLICIT_SYS */
11813 param->flags = flags;
11814 param->proto_perl = proto_perl;
11816 INIT_TRACK_MEMPOOL(my_perl->Imemory_debug_header, my_perl);
11818 PL_body_arenas = NULL;
11819 Zero(&PL_body_roots, 1, PL_body_roots);
11821 PL_nice_chunk = NULL;
11822 PL_nice_chunk_size = 0;
11824 PL_sv_objcount = 0;
11826 PL_sv_arenaroot = NULL;
11828 PL_debug = proto_perl->Idebug;
11830 PL_hash_seed = proto_perl->Ihash_seed;
11831 PL_rehash_seed = proto_perl->Irehash_seed;
11833 #ifdef USE_REENTRANT_API
11834 /* XXX: things like -Dm will segfault here in perlio, but doing
11835 * PERL_SET_CONTEXT(proto_perl);
11836 * breaks too many other things
11838 Perl_reentrant_init(aTHX);
11841 /* create SV map for pointer relocation */
11842 PL_ptr_table = ptr_table_new();
11844 /* initialize these special pointers as early as possible */
11845 SvANY(&PL_sv_undef) = NULL;
11846 SvREFCNT(&PL_sv_undef) = (~(U32)0)/2;
11847 SvFLAGS(&PL_sv_undef) = SVf_READONLY|SVt_NULL;
11848 ptr_table_store(PL_ptr_table, &proto_perl->Isv_undef, &PL_sv_undef);
11850 SvANY(&PL_sv_no) = new_XPVNV();
11851 SvREFCNT(&PL_sv_no) = (~(U32)0)/2;
11852 SvFLAGS(&PL_sv_no) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
11853 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
11854 SvPV_set(&PL_sv_no, savepvn(PL_No, 0));
11855 SvCUR_set(&PL_sv_no, 0);
11856 SvLEN_set(&PL_sv_no, 1);
11857 SvIV_set(&PL_sv_no, 0);
11858 SvNV_set(&PL_sv_no, 0);
11859 ptr_table_store(PL_ptr_table, &proto_perl->Isv_no, &PL_sv_no);
11861 SvANY(&PL_sv_yes) = new_XPVNV();
11862 SvREFCNT(&PL_sv_yes) = (~(U32)0)/2;
11863 SvFLAGS(&PL_sv_yes) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
11864 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
11865 SvPV_set(&PL_sv_yes, savepvn(PL_Yes, 1));
11866 SvCUR_set(&PL_sv_yes, 1);
11867 SvLEN_set(&PL_sv_yes, 2);
11868 SvIV_set(&PL_sv_yes, 1);
11869 SvNV_set(&PL_sv_yes, 1);
11870 ptr_table_store(PL_ptr_table, &proto_perl->Isv_yes, &PL_sv_yes);
11872 /* create (a non-shared!) shared string table */
11873 PL_strtab = newHV();
11874 HvSHAREKEYS_off(PL_strtab);
11875 hv_ksplit(PL_strtab, HvTOTALKEYS(proto_perl->Istrtab));
11876 ptr_table_store(PL_ptr_table, proto_perl->Istrtab, PL_strtab);
11878 PL_compiling = proto_perl->Icompiling;
11880 /* These two PVs will be free'd special way so must set them same way op.c does */
11881 PL_compiling.cop_stashpv = savesharedpv(PL_compiling.cop_stashpv);
11882 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_stashpv, PL_compiling.cop_stashpv);
11884 PL_compiling.cop_file = savesharedpv(PL_compiling.cop_file);
11885 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_file, PL_compiling.cop_file);
11887 ptr_table_store(PL_ptr_table, &proto_perl->Icompiling, &PL_compiling);
11888 PL_compiling.cop_warnings = DUP_WARNINGS(PL_compiling.cop_warnings);
11889 if (PL_compiling.cop_hints_hash) {
11891 PL_compiling.cop_hints_hash->refcounted_he_refcnt++;
11892 HINTS_REFCNT_UNLOCK;
11894 PL_curcop = (COP*)any_dup(proto_perl->Icurcop, proto_perl);
11895 #ifdef PERL_DEBUG_READONLY_OPS
11900 /* pseudo environmental stuff */
11901 PL_origargc = proto_perl->Iorigargc;
11902 PL_origargv = proto_perl->Iorigargv;
11904 param->stashes = newAV(); /* Setup array of objects to call clone on */
11906 /* Set tainting stuff before PerlIO_debug can possibly get called */
11907 PL_tainting = proto_perl->Itainting;
11908 PL_taint_warn = proto_perl->Itaint_warn;
11910 #ifdef PERLIO_LAYERS
11911 /* Clone PerlIO tables as soon as we can handle general xx_dup() */
11912 PerlIO_clone(aTHX_ proto_perl, param);
11915 PL_envgv = gv_dup(proto_perl->Ienvgv, param);
11916 PL_incgv = gv_dup(proto_perl->Iincgv, param);
11917 PL_hintgv = gv_dup(proto_perl->Ihintgv, param);
11918 PL_origfilename = SAVEPV(proto_perl->Iorigfilename);
11919 PL_diehook = sv_dup_inc(proto_perl->Idiehook, param);
11920 PL_warnhook = sv_dup_inc(proto_perl->Iwarnhook, param);
11923 PL_minus_c = proto_perl->Iminus_c;
11924 PL_patchlevel = sv_dup_inc(proto_perl->Ipatchlevel, param);
11925 PL_localpatches = proto_perl->Ilocalpatches;
11926 PL_splitstr = proto_perl->Isplitstr;
11927 PL_minus_n = proto_perl->Iminus_n;
11928 PL_minus_p = proto_perl->Iminus_p;
11929 PL_minus_l = proto_perl->Iminus_l;
11930 PL_minus_a = proto_perl->Iminus_a;
11931 PL_minus_E = proto_perl->Iminus_E;
11932 PL_minus_F = proto_perl->Iminus_F;
11933 PL_doswitches = proto_perl->Idoswitches;
11934 PL_dowarn = proto_perl->Idowarn;
11935 PL_doextract = proto_perl->Idoextract;
11936 PL_sawampersand = proto_perl->Isawampersand;
11937 PL_unsafe = proto_perl->Iunsafe;
11938 PL_inplace = SAVEPV(proto_perl->Iinplace);
11939 PL_e_script = sv_dup_inc(proto_perl->Ie_script, param);
11940 PL_perldb = proto_perl->Iperldb;
11941 PL_perl_destruct_level = proto_perl->Iperl_destruct_level;
11942 PL_exit_flags = proto_perl->Iexit_flags;
11944 /* magical thingies */
11945 /* XXX time(&PL_basetime) when asked for? */
11946 PL_basetime = proto_perl->Ibasetime;
11947 PL_formfeed = sv_dup(proto_perl->Iformfeed, param);
11949 PL_maxsysfd = proto_perl->Imaxsysfd;
11950 PL_statusvalue = proto_perl->Istatusvalue;
11952 PL_statusvalue_vms = proto_perl->Istatusvalue_vms;
11954 PL_statusvalue_posix = proto_perl->Istatusvalue_posix;
11956 PL_encoding = sv_dup(proto_perl->Iencoding, param);
11958 sv_setpvs(PERL_DEBUG_PAD(0), ""); /* For regex debugging. */
11959 sv_setpvs(PERL_DEBUG_PAD(1), ""); /* ext/re needs these */
11960 sv_setpvs(PERL_DEBUG_PAD(2), ""); /* even without DEBUGGING. */
11963 /* RE engine related */
11964 Zero(&PL_reg_state, 1, struct re_save_state);
11965 PL_reginterp_cnt = 0;
11966 PL_regmatch_slab = NULL;
11968 /* Clone the regex array */
11969 /* ORANGE FIXME for plugins, probably in the SV dup code.
11970 newSViv(PTR2IV(CALLREGDUPE(
11971 INT2PTR(REGEXP *, SvIVX(regex)), param))))
11973 PL_regex_padav = av_dup_inc(proto_perl->Iregex_padav, param);
11974 PL_regex_pad = AvARRAY(PL_regex_padav);
11976 /* shortcuts to various I/O objects */
11977 PL_ofsgv = gv_dup(proto_perl->Iofsgv, param);
11978 PL_stdingv = gv_dup(proto_perl->Istdingv, param);
11979 PL_stderrgv = gv_dup(proto_perl->Istderrgv, param);
11980 PL_defgv = gv_dup(proto_perl->Idefgv, param);
11981 PL_argvgv = gv_dup(proto_perl->Iargvgv, param);
11982 PL_argvoutgv = gv_dup(proto_perl->Iargvoutgv, param);
11983 PL_argvout_stack = av_dup_inc(proto_perl->Iargvout_stack, param);
11985 /* shortcuts to regexp stuff */
11986 PL_replgv = gv_dup(proto_perl->Ireplgv, param);
11988 /* shortcuts to misc objects */
11989 PL_errgv = gv_dup(proto_perl->Ierrgv, param);
11991 /* shortcuts to debugging objects */
11992 PL_DBgv = gv_dup(proto_perl->IDBgv, param);
11993 PL_DBline = gv_dup(proto_perl->IDBline, param);
11994 PL_DBsub = gv_dup(proto_perl->IDBsub, param);
11995 PL_DBsingle = sv_dup(proto_perl->IDBsingle, param);
11996 PL_DBtrace = sv_dup(proto_perl->IDBtrace, param);
11997 PL_DBsignal = sv_dup(proto_perl->IDBsignal, param);
11998 PL_dbargs = av_dup(proto_perl->Idbargs, param);
12000 /* symbol tables */
12001 PL_defstash = hv_dup_inc(proto_perl->Idefstash, param);
12002 PL_curstash = hv_dup(proto_perl->Icurstash, param);
12003 PL_debstash = hv_dup(proto_perl->Idebstash, param);
12004 PL_globalstash = hv_dup(proto_perl->Iglobalstash, param);
12005 PL_curstname = sv_dup_inc(proto_perl->Icurstname, param);
12007 PL_beginav = av_dup_inc(proto_perl->Ibeginav, param);
12008 PL_beginav_save = av_dup_inc(proto_perl->Ibeginav_save, param);
12009 PL_checkav_save = av_dup_inc(proto_perl->Icheckav_save, param);
12010 PL_unitcheckav = av_dup_inc(proto_perl->Iunitcheckav, param);
12011 PL_unitcheckav_save = av_dup_inc(proto_perl->Iunitcheckav_save, param);
12012 PL_endav = av_dup_inc(proto_perl->Iendav, param);
12013 PL_checkav = av_dup_inc(proto_perl->Icheckav, param);
12014 PL_initav = av_dup_inc(proto_perl->Iinitav, param);
12016 PL_sub_generation = proto_perl->Isub_generation;
12017 PL_isarev = hv_dup_inc(proto_perl->Iisarev, param);
12019 /* funky return mechanisms */
12020 PL_forkprocess = proto_perl->Iforkprocess;
12022 /* subprocess state */
12023 PL_fdpid = av_dup_inc(proto_perl->Ifdpid, param);
12025 /* internal state */
12026 PL_maxo = proto_perl->Imaxo;
12027 if (proto_perl->Iop_mask)
12028 PL_op_mask = SAVEPVN(proto_perl->Iop_mask, PL_maxo);
12031 /* PL_asserting = proto_perl->Iasserting; */
12033 /* current interpreter roots */
12034 PL_main_cv = cv_dup_inc(proto_perl->Imain_cv, param);
12036 PL_main_root = OpREFCNT_inc(proto_perl->Imain_root);
12038 PL_main_start = proto_perl->Imain_start;
12039 PL_eval_root = proto_perl->Ieval_root;
12040 PL_eval_start = proto_perl->Ieval_start;
12042 /* runtime control stuff */
12043 PL_curcopdb = (COP*)any_dup(proto_perl->Icurcopdb, proto_perl);
12045 PL_filemode = proto_perl->Ifilemode;
12046 PL_lastfd = proto_perl->Ilastfd;
12047 PL_oldname = proto_perl->Ioldname; /* XXX not quite right */
12050 PL_gensym = proto_perl->Igensym;
12051 PL_preambleav = av_dup_inc(proto_perl->Ipreambleav, param);
12052 PL_laststatval = proto_perl->Ilaststatval;
12053 PL_laststype = proto_perl->Ilaststype;
12056 PL_ors_sv = sv_dup_inc(proto_perl->Iors_sv, param);
12058 /* interpreter atexit processing */
12059 PL_exitlistlen = proto_perl->Iexitlistlen;
12060 if (PL_exitlistlen) {
12061 Newx(PL_exitlist, PL_exitlistlen, PerlExitListEntry);
12062 Copy(proto_perl->Iexitlist, PL_exitlist, PL_exitlistlen, PerlExitListEntry);
12065 PL_exitlist = (PerlExitListEntry*)NULL;
12067 PL_my_cxt_size = proto_perl->Imy_cxt_size;
12068 if (PL_my_cxt_size) {
12069 Newx(PL_my_cxt_list, PL_my_cxt_size, void *);
12070 Copy(proto_perl->Imy_cxt_list, PL_my_cxt_list, PL_my_cxt_size, void *);
12071 #ifdef PERL_GLOBAL_STRUCT_PRIVATE
12072 Newx(PL_my_cxt_keys, PL_my_cxt_size, const char *);
12073 Copy(proto_perl->Imy_cxt_keys, PL_my_cxt_keys, PL_my_cxt_size, char *);
12077 PL_my_cxt_list = (void**)NULL;
12078 #ifdef PERL_GLOBAL_STRUCT_PRIVATE
12079 PL_my_cxt_keys = (const char**)NULL;
12082 PL_modglobal = hv_dup_inc(proto_perl->Imodglobal, param);
12083 PL_custom_op_names = hv_dup_inc(proto_perl->Icustom_op_names,param);
12084 PL_custom_op_descs = hv_dup_inc(proto_perl->Icustom_op_descs,param);
12086 PL_profiledata = NULL;
12088 PL_compcv = cv_dup(proto_perl->Icompcv, param);
12090 PAD_CLONE_VARS(proto_perl, param);
12092 #ifdef HAVE_INTERP_INTERN
12093 sys_intern_dup(&proto_perl->Isys_intern, &PL_sys_intern);
12096 /* more statics moved here */
12097 PL_generation = proto_perl->Igeneration;
12098 PL_DBcv = cv_dup(proto_perl->IDBcv, param);
12100 PL_in_clean_objs = proto_perl->Iin_clean_objs;
12101 PL_in_clean_all = proto_perl->Iin_clean_all;
12103 PL_uid = proto_perl->Iuid;
12104 PL_euid = proto_perl->Ieuid;
12105 PL_gid = proto_perl->Igid;
12106 PL_egid = proto_perl->Iegid;
12107 PL_nomemok = proto_perl->Inomemok;
12108 PL_an = proto_perl->Ian;
12109 PL_evalseq = proto_perl->Ievalseq;
12110 PL_origenviron = proto_perl->Iorigenviron; /* XXX not quite right */
12111 PL_origalen = proto_perl->Iorigalen;
12112 #ifdef PERL_USES_PL_PIDSTATUS
12113 PL_pidstatus = newHV(); /* XXX flag for cloning? */
12115 PL_osname = SAVEPV(proto_perl->Iosname);
12116 PL_sighandlerp = proto_perl->Isighandlerp;
12118 PL_runops = proto_perl->Irunops;
12120 PL_parser = parser_dup(proto_perl->Iparser, param);
12122 PL_subline = proto_perl->Isubline;
12123 PL_subname = sv_dup_inc(proto_perl->Isubname, param);
12126 PL_cryptseen = proto_perl->Icryptseen;
12129 PL_hints = proto_perl->Ihints;
12131 PL_amagic_generation = proto_perl->Iamagic_generation;
12133 #ifdef USE_LOCALE_COLLATE
12134 PL_collation_ix = proto_perl->Icollation_ix;
12135 PL_collation_name = SAVEPV(proto_perl->Icollation_name);
12136 PL_collation_standard = proto_perl->Icollation_standard;
12137 PL_collxfrm_base = proto_perl->Icollxfrm_base;
12138 PL_collxfrm_mult = proto_perl->Icollxfrm_mult;
12139 #endif /* USE_LOCALE_COLLATE */
12141 #ifdef USE_LOCALE_NUMERIC
12142 PL_numeric_name = SAVEPV(proto_perl->Inumeric_name);
12143 PL_numeric_standard = proto_perl->Inumeric_standard;
12144 PL_numeric_local = proto_perl->Inumeric_local;
12145 PL_numeric_radix_sv = sv_dup_inc(proto_perl->Inumeric_radix_sv, param);
12146 #endif /* !USE_LOCALE_NUMERIC */
12148 /* utf8 character classes */
12149 PL_utf8_alnum = sv_dup_inc(proto_perl->Iutf8_alnum, param);
12150 PL_utf8_alnumc = sv_dup_inc(proto_perl->Iutf8_alnumc, param);
12151 PL_utf8_ascii = sv_dup_inc(proto_perl->Iutf8_ascii, param);
12152 PL_utf8_alpha = sv_dup_inc(proto_perl->Iutf8_alpha, param);
12153 PL_utf8_space = sv_dup_inc(proto_perl->Iutf8_space, param);
12154 PL_utf8_cntrl = sv_dup_inc(proto_perl->Iutf8_cntrl, param);
12155 PL_utf8_graph = sv_dup_inc(proto_perl->Iutf8_graph, param);
12156 PL_utf8_digit = sv_dup_inc(proto_perl->Iutf8_digit, param);
12157 PL_utf8_upper = sv_dup_inc(proto_perl->Iutf8_upper, param);
12158 PL_utf8_lower = sv_dup_inc(proto_perl->Iutf8_lower, param);
12159 PL_utf8_print = sv_dup_inc(proto_perl->Iutf8_print, param);
12160 PL_utf8_punct = sv_dup_inc(proto_perl->Iutf8_punct, param);
12161 PL_utf8_xdigit = sv_dup_inc(proto_perl->Iutf8_xdigit, param);
12162 PL_utf8_mark = sv_dup_inc(proto_perl->Iutf8_mark, param);
12163 PL_utf8_toupper = sv_dup_inc(proto_perl->Iutf8_toupper, param);
12164 PL_utf8_totitle = sv_dup_inc(proto_perl->Iutf8_totitle, param);
12165 PL_utf8_tolower = sv_dup_inc(proto_perl->Iutf8_tolower, param);
12166 PL_utf8_tofold = sv_dup_inc(proto_perl->Iutf8_tofold, param);
12167 PL_utf8_idstart = sv_dup_inc(proto_perl->Iutf8_idstart, param);
12168 PL_utf8_idcont = sv_dup_inc(proto_perl->Iutf8_idcont, param);
12170 /* Did the locale setup indicate UTF-8? */
12171 PL_utf8locale = proto_perl->Iutf8locale;
12172 /* Unicode features (see perlrun/-C) */
12173 PL_unicode = proto_perl->Iunicode;
12175 /* Pre-5.8 signals control */
12176 PL_signals = proto_perl->Isignals;
12178 /* times() ticks per second */
12179 PL_clocktick = proto_perl->Iclocktick;
12181 /* Recursion stopper for PerlIO_find_layer */
12182 PL_in_load_module = proto_perl->Iin_load_module;
12184 /* sort() routine */
12185 PL_sort_RealCmp = proto_perl->Isort_RealCmp;
12187 /* Not really needed/useful since the reenrant_retint is "volatile",
12188 * but do it for consistency's sake. */
12189 PL_reentrant_retint = proto_perl->Ireentrant_retint;
12191 /* Hooks to shared SVs and locks. */
12192 PL_sharehook = proto_perl->Isharehook;
12193 PL_lockhook = proto_perl->Ilockhook;
12194 PL_unlockhook = proto_perl->Iunlockhook;
12195 PL_threadhook = proto_perl->Ithreadhook;
12196 PL_destroyhook = proto_perl->Idestroyhook;
12198 #ifdef THREADS_HAVE_PIDS
12199 PL_ppid = proto_perl->Ippid;
12203 PL_last_swash_hv = NULL; /* reinits on demand */
12204 PL_last_swash_klen = 0;
12205 PL_last_swash_key[0]= '\0';
12206 PL_last_swash_tmps = (U8*)NULL;
12207 PL_last_swash_slen = 0;
12209 PL_glob_index = proto_perl->Iglob_index;
12210 PL_srand_called = proto_perl->Isrand_called;
12211 PL_bitcount = NULL; /* reinits on demand */
12213 if (proto_perl->Ipsig_pend) {
12214 Newxz(PL_psig_pend, SIG_SIZE, int);
12217 PL_psig_pend = (int*)NULL;
12220 if (proto_perl->Ipsig_ptr) {
12221 Newxz(PL_psig_ptr, SIG_SIZE, SV*);
12222 Newxz(PL_psig_name, SIG_SIZE, SV*);
12223 for (i = 1; i < SIG_SIZE; i++) {
12224 PL_psig_ptr[i] = sv_dup_inc(proto_perl->Ipsig_ptr[i], param);
12225 PL_psig_name[i] = sv_dup_inc(proto_perl->Ipsig_name[i], param);
12229 PL_psig_ptr = (SV**)NULL;
12230 PL_psig_name = (SV**)NULL;
12233 /* intrpvar.h stuff */
12235 if (flags & CLONEf_COPY_STACKS) {
12236 /* next allocation will be PL_tmps_stack[PL_tmps_ix+1] */
12237 PL_tmps_ix = proto_perl->Itmps_ix;
12238 PL_tmps_max = proto_perl->Itmps_max;
12239 PL_tmps_floor = proto_perl->Itmps_floor;
12240 Newxz(PL_tmps_stack, PL_tmps_max, SV*);
12242 while (i <= PL_tmps_ix) {
12243 PL_tmps_stack[i] = sv_dup_inc(proto_perl->Itmps_stack[i], param);
12247 /* next PUSHMARK() sets *(PL_markstack_ptr+1) */
12248 i = proto_perl->Imarkstack_max - proto_perl->Imarkstack;
12249 Newxz(PL_markstack, i, I32);
12250 PL_markstack_max = PL_markstack + (proto_perl->Imarkstack_max
12251 - proto_perl->Imarkstack);
12252 PL_markstack_ptr = PL_markstack + (proto_perl->Imarkstack_ptr
12253 - proto_perl->Imarkstack);
12254 Copy(proto_perl->Imarkstack, PL_markstack,
12255 PL_markstack_ptr - PL_markstack + 1, I32);
12257 /* next push_scope()/ENTER sets PL_scopestack[PL_scopestack_ix]
12258 * NOTE: unlike the others! */
12259 PL_scopestack_ix = proto_perl->Iscopestack_ix;
12260 PL_scopestack_max = proto_perl->Iscopestack_max;
12261 Newxz(PL_scopestack, PL_scopestack_max, I32);
12262 Copy(proto_perl->Iscopestack, PL_scopestack, PL_scopestack_ix, I32);
12264 /* NOTE: si_dup() looks at PL_markstack */
12265 PL_curstackinfo = si_dup(proto_perl->Icurstackinfo, param);
12267 /* PL_curstack = PL_curstackinfo->si_stack; */
12268 PL_curstack = av_dup(proto_perl->Icurstack, param);
12269 PL_mainstack = av_dup(proto_perl->Imainstack, param);
12271 /* next PUSHs() etc. set *(PL_stack_sp+1) */
12272 PL_stack_base = AvARRAY(PL_curstack);
12273 PL_stack_sp = PL_stack_base + (proto_perl->Istack_sp
12274 - proto_perl->Istack_base);
12275 PL_stack_max = PL_stack_base + AvMAX(PL_curstack);
12277 /* next SSPUSHFOO() sets PL_savestack[PL_savestack_ix]
12278 * NOTE: unlike the others! */
12279 PL_savestack_ix = proto_perl->Isavestack_ix;
12280 PL_savestack_max = proto_perl->Isavestack_max;
12281 /*Newxz(PL_savestack, PL_savestack_max, ANY);*/
12282 PL_savestack = ss_dup(proto_perl, param);
12286 ENTER; /* perl_destruct() wants to LEAVE; */
12288 /* although we're not duplicating the tmps stack, we should still
12289 * add entries for any SVs on the tmps stack that got cloned by a
12290 * non-refcount means (eg a temp in @_); otherwise they will be
12293 for (i = 0; i<= proto_perl->Itmps_ix; i++) {
12294 SV * const nsv = MUTABLE_SV(ptr_table_fetch(PL_ptr_table,
12295 proto_perl->Itmps_stack[i]));
12296 if (nsv && !SvREFCNT(nsv)) {
12298 PL_tmps_stack[++PL_tmps_ix] = SvREFCNT_inc_simple(nsv);
12303 PL_start_env = proto_perl->Istart_env; /* XXXXXX */
12304 PL_top_env = &PL_start_env;
12306 PL_op = proto_perl->Iop;
12309 PL_Xpv = (XPV*)NULL;
12310 my_perl->Ina = proto_perl->Ina;
12312 PL_statbuf = proto_perl->Istatbuf;
12313 PL_statcache = proto_perl->Istatcache;
12314 PL_statgv = gv_dup(proto_perl->Istatgv, param);
12315 PL_statname = sv_dup_inc(proto_perl->Istatname, param);
12317 PL_timesbuf = proto_perl->Itimesbuf;
12320 PL_tainted = proto_perl->Itainted;
12321 PL_curpm = proto_perl->Icurpm; /* XXX No PMOP ref count */
12322 PL_rs = sv_dup_inc(proto_perl->Irs, param);
12323 PL_last_in_gv = gv_dup(proto_perl->Ilast_in_gv, param);
12324 PL_defoutgv = gv_dup_inc(proto_perl->Idefoutgv, param);
12325 PL_chopset = proto_perl->Ichopset; /* XXX never deallocated */
12326 PL_toptarget = sv_dup_inc(proto_perl->Itoptarget, param);
12327 PL_bodytarget = sv_dup_inc(proto_perl->Ibodytarget, param);
12328 PL_formtarget = sv_dup(proto_perl->Iformtarget, param);
12330 PL_restartop = proto_perl->Irestartop;
12331 PL_in_eval = proto_perl->Iin_eval;
12332 PL_delaymagic = proto_perl->Idelaymagic;
12333 PL_dirty = proto_perl->Idirty;
12334 PL_localizing = proto_perl->Ilocalizing;
12336 PL_errors = sv_dup_inc(proto_perl->Ierrors, param);
12337 PL_hv_fetch_ent_mh = NULL;
12338 PL_modcount = proto_perl->Imodcount;
12339 PL_lastgotoprobe = NULL;
12340 PL_dumpindent = proto_perl->Idumpindent;
12342 PL_sortcop = (OP*)any_dup(proto_perl->Isortcop, proto_perl);
12343 PL_sortstash = hv_dup(proto_perl->Isortstash, param);
12344 PL_firstgv = gv_dup(proto_perl->Ifirstgv, param);
12345 PL_secondgv = gv_dup(proto_perl->Isecondgv, param);
12346 PL_efloatbuf = NULL; /* reinits on demand */
12347 PL_efloatsize = 0; /* reinits on demand */
12351 PL_screamfirst = NULL;
12352 PL_screamnext = NULL;
12353 PL_maxscream = -1; /* reinits on demand */
12354 PL_lastscream = NULL;
12357 PL_regdummy = proto_perl->Iregdummy;
12358 PL_colorset = 0; /* reinits PL_colors[] */
12359 /*PL_colors[6] = {0,0,0,0,0,0};*/
12363 /* Pluggable optimizer */
12364 PL_peepp = proto_perl->Ipeepp;
12366 PL_stashcache = newHV();
12368 PL_watchaddr = (char **) ptr_table_fetch(PL_ptr_table,
12369 proto_perl->Iwatchaddr);
12370 PL_watchok = PL_watchaddr ? * PL_watchaddr : NULL;
12371 if (PL_debug && PL_watchaddr) {
12372 PerlIO_printf(Perl_debug_log,
12373 "WATCHING: %"UVxf" cloned as %"UVxf" with value %"UVxf"\n",
12374 PTR2UV(proto_perl->Iwatchaddr), PTR2UV(PL_watchaddr),
12375 PTR2UV(PL_watchok));
12378 PL_registered_mros = hv_dup_inc(proto_perl->Iregistered_mros, param);
12380 if (!(flags & CLONEf_KEEP_PTR_TABLE)) {
12381 ptr_table_free(PL_ptr_table);
12382 PL_ptr_table = NULL;
12385 /* Call the ->CLONE method, if it exists, for each of the stashes
12386 identified by sv_dup() above.
12388 while(av_len(param->stashes) != -1) {
12389 HV* const stash = MUTABLE_HV(av_shift(param->stashes));
12390 GV* const cloner = gv_fetchmethod_autoload(stash, "CLONE", 0);
12391 if (cloner && GvCV(cloner)) {
12396 mXPUSHs(newSVhek(HvNAME_HEK(stash)));
12398 call_sv(MUTABLE_SV(GvCV(cloner)), G_DISCARD);
12404 SvREFCNT_dec(param->stashes);
12406 /* orphaned? eg threads->new inside BEGIN or use */
12407 if (PL_compcv && ! SvREFCNT(PL_compcv)) {
12408 SvREFCNT_inc_simple_void(PL_compcv);
12409 SAVEFREESV(PL_compcv);
12415 #endif /* USE_ITHREADS */
12418 =head1 Unicode Support
12420 =for apidoc sv_recode_to_utf8
12422 The encoding is assumed to be an Encode object, on entry the PV
12423 of the sv is assumed to be octets in that encoding, and the sv
12424 will be converted into Unicode (and UTF-8).
12426 If the sv already is UTF-8 (or if it is not POK), or if the encoding
12427 is not a reference, nothing is done to the sv. If the encoding is not
12428 an C<Encode::XS> Encoding object, bad things will happen.
12429 (See F<lib/encoding.pm> and L<Encode>).
12431 The PV of the sv is returned.
12436 Perl_sv_recode_to_utf8(pTHX_ SV *sv, SV *encoding)
12440 PERL_ARGS_ASSERT_SV_RECODE_TO_UTF8;
12442 if (SvPOK(sv) && !SvUTF8(sv) && !IN_BYTES && SvROK(encoding)) {
12456 Passing sv_yes is wrong - it needs to be or'ed set of constants
12457 for Encode::XS, while UTf-8 decode (currently) assumes a true value means
12458 remove converted chars from source.
12460 Both will default the value - let them.
12462 XPUSHs(&PL_sv_yes);
12465 call_method("decode", G_SCALAR);
12469 s = SvPV_const(uni, len);
12470 if (s != SvPVX_const(sv)) {
12471 SvGROW(sv, len + 1);
12472 Move(s, SvPVX(sv), len + 1, char);
12473 SvCUR_set(sv, len);
12480 return SvPOKp(sv) ? SvPVX(sv) : NULL;
12484 =for apidoc sv_cat_decode
12486 The encoding is assumed to be an Encode object, the PV of the ssv is
12487 assumed to be octets in that encoding and decoding the input starts
12488 from the position which (PV + *offset) pointed to. The dsv will be
12489 concatenated the decoded UTF-8 string from ssv. Decoding will terminate
12490 when the string tstr appears in decoding output or the input ends on
12491 the PV of the ssv. The value which the offset points will be modified
12492 to the last input position on the ssv.
12494 Returns TRUE if the terminator was found, else returns FALSE.
12499 Perl_sv_cat_decode(pTHX_ SV *dsv, SV *encoding,
12500 SV *ssv, int *offset, char *tstr, int tlen)
12505 PERL_ARGS_ASSERT_SV_CAT_DECODE;
12507 if (SvPOK(ssv) && SvPOK(dsv) && SvROK(encoding) && offset) {
12518 offsv = newSViv(*offset);
12520 mXPUSHp(tstr, tlen);
12522 call_method("cat_decode", G_SCALAR);
12524 ret = SvTRUE(TOPs);
12525 *offset = SvIV(offsv);
12531 Perl_croak(aTHX_ "Invalid argument to sv_cat_decode");
12536 /* ---------------------------------------------------------------------
12538 * support functions for report_uninit()
12541 /* the maxiumum size of array or hash where we will scan looking
12542 * for the undefined element that triggered the warning */
12544 #define FUV_MAX_SEARCH_SIZE 1000
12546 /* Look for an entry in the hash whose value has the same SV as val;
12547 * If so, return a mortal copy of the key. */
12550 S_find_hash_subscript(pTHX_ const HV *const hv, const SV *const val)
12553 register HE **array;
12556 PERL_ARGS_ASSERT_FIND_HASH_SUBSCRIPT;
12558 if (!hv || SvMAGICAL(hv) || !HvARRAY(hv) ||
12559 (HvTOTALKEYS(hv) > FUV_MAX_SEARCH_SIZE))
12562 array = HvARRAY(hv);
12564 for (i=HvMAX(hv); i>0; i--) {
12565 register HE *entry;
12566 for (entry = array[i]; entry; entry = HeNEXT(entry)) {
12567 if (HeVAL(entry) != val)
12569 if ( HeVAL(entry) == &PL_sv_undef ||
12570 HeVAL(entry) == &PL_sv_placeholder)
12574 if (HeKLEN(entry) == HEf_SVKEY)
12575 return sv_mortalcopy(HeKEY_sv(entry));
12576 return sv_2mortal(newSVhek(HeKEY_hek(entry)));
12582 /* Look for an entry in the array whose value has the same SV as val;
12583 * If so, return the index, otherwise return -1. */
12586 S_find_array_subscript(pTHX_ const AV *const av, const SV *const val)
12590 PERL_ARGS_ASSERT_FIND_ARRAY_SUBSCRIPT;
12592 if (!av || SvMAGICAL(av) || !AvARRAY(av) ||
12593 (AvFILLp(av) > FUV_MAX_SEARCH_SIZE))
12596 if (val != &PL_sv_undef) {
12597 SV ** const svp = AvARRAY(av);
12600 for (i=AvFILLp(av); i>=0; i--)
12607 /* S_varname(): return the name of a variable, optionally with a subscript.
12608 * If gv is non-zero, use the name of that global, along with gvtype (one
12609 * of "$", "@", "%"); otherwise use the name of the lexical at pad offset
12610 * targ. Depending on the value of the subscript_type flag, return:
12613 #define FUV_SUBSCRIPT_NONE 1 /* "@foo" */
12614 #define FUV_SUBSCRIPT_ARRAY 2 /* "$foo[aindex]" */
12615 #define FUV_SUBSCRIPT_HASH 3 /* "$foo{keyname}" */
12616 #define FUV_SUBSCRIPT_WITHIN 4 /* "within @foo" */
12619 S_varname(pTHX_ const GV *const gv, const char gvtype, PADOFFSET targ,
12620 const SV *const keyname, I32 aindex, int subscript_type)
12623 SV * const name = sv_newmortal();
12626 buffer[0] = gvtype;
12629 /* as gv_fullname4(), but add literal '^' for $^FOO names */
12631 gv_fullname4(name, gv, buffer, 0);
12633 if ((unsigned int)SvPVX(name)[1] <= 26) {
12635 buffer[1] = SvPVX(name)[1] + 'A' - 1;
12637 /* Swap the 1 unprintable control character for the 2 byte pretty
12638 version - ie substr($name, 1, 1) = $buffer; */
12639 sv_insert(name, 1, 1, buffer, 2);
12643 CV * const cv = find_runcv(NULL);
12647 if (!cv || !CvPADLIST(cv))
12649 av = MUTABLE_AV((*av_fetch(CvPADLIST(cv), 0, FALSE)));
12650 sv = *av_fetch(av, targ, FALSE);
12651 sv_setpvn(name, SvPV_nolen_const(sv), SvCUR(sv));
12654 if (subscript_type == FUV_SUBSCRIPT_HASH) {
12655 SV * const sv = newSV(0);
12656 *SvPVX(name) = '$';
12657 Perl_sv_catpvf(aTHX_ name, "{%s}",
12658 pv_display(sv,SvPVX_const(keyname), SvCUR(keyname), 0, 32));
12661 else if (subscript_type == FUV_SUBSCRIPT_ARRAY) {
12662 *SvPVX(name) = '$';
12663 Perl_sv_catpvf(aTHX_ name, "[%"IVdf"]", (IV)aindex);
12665 else if (subscript_type == FUV_SUBSCRIPT_WITHIN) {
12666 /* We know that name has no magic, so can use 0 instead of SV_GMAGIC */
12667 Perl_sv_insert_flags(aTHX_ name, 0, 0, STR_WITH_LEN("within "), 0);
12675 =for apidoc find_uninit_var
12677 Find the name of the undefined variable (if any) that caused the operator o
12678 to issue a "Use of uninitialized value" warning.
12679 If match is true, only return a name if it's value matches uninit_sv.
12680 So roughly speaking, if a unary operator (such as OP_COS) generates a
12681 warning, then following the direct child of the op may yield an
12682 OP_PADSV or OP_GV that gives the name of the undefined variable. On the
12683 other hand, with OP_ADD there are two branches to follow, so we only print
12684 the variable name if we get an exact match.
12686 The name is returned as a mortal SV.
12688 Assumes that PL_op is the op that originally triggered the error, and that
12689 PL_comppad/PL_curpad points to the currently executing pad.
12695 S_find_uninit_var(pTHX_ const OP *const obase, const SV *const uninit_sv,
12701 const OP *o, *o2, *kid;
12703 if (!obase || (match && (!uninit_sv || uninit_sv == &PL_sv_undef ||
12704 uninit_sv == &PL_sv_placeholder)))
12707 switch (obase->op_type) {
12714 const bool pad = (obase->op_type == OP_PADAV || obase->op_type == OP_PADHV);
12715 const bool hash = (obase->op_type == OP_PADHV || obase->op_type == OP_RV2HV);
12718 int subscript_type = FUV_SUBSCRIPT_WITHIN;
12720 if (pad) { /* @lex, %lex */
12721 sv = PAD_SVl(obase->op_targ);
12725 if (cUNOPx(obase)->op_first->op_type == OP_GV) {
12726 /* @global, %global */
12727 gv = cGVOPx_gv(cUNOPx(obase)->op_first);
12730 sv = hash ? MUTABLE_SV(GvHV(gv)): MUTABLE_SV(GvAV(gv));
12732 else /* @{expr}, %{expr} */
12733 return find_uninit_var(cUNOPx(obase)->op_first,
12737 /* attempt to find a match within the aggregate */
12739 keysv = find_hash_subscript((const HV*)sv, uninit_sv);
12741 subscript_type = FUV_SUBSCRIPT_HASH;
12744 index = find_array_subscript((const AV *)sv, uninit_sv);
12746 subscript_type = FUV_SUBSCRIPT_ARRAY;
12749 if (match && subscript_type == FUV_SUBSCRIPT_WITHIN)
12752 return varname(gv, hash ? '%' : '@', obase->op_targ,
12753 keysv, index, subscript_type);
12757 if (match && PAD_SVl(obase->op_targ) != uninit_sv)
12759 return varname(NULL, '$', obase->op_targ,
12760 NULL, 0, FUV_SUBSCRIPT_NONE);
12763 gv = cGVOPx_gv(obase);
12764 if (!gv || (match && GvSV(gv) != uninit_sv))
12766 return varname(gv, '$', 0, NULL, 0, FUV_SUBSCRIPT_NONE);
12769 if (obase->op_flags & OPf_SPECIAL) { /* lexical array */
12772 AV *av = MUTABLE_AV(PAD_SV(obase->op_targ));
12773 if (!av || SvRMAGICAL(av))
12775 svp = av_fetch(av, (I32)obase->op_private, FALSE);
12776 if (!svp || *svp != uninit_sv)
12779 return varname(NULL, '$', obase->op_targ,
12780 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
12783 gv = cGVOPx_gv(obase);
12788 AV *const av = GvAV(gv);
12789 if (!av || SvRMAGICAL(av))
12791 svp = av_fetch(av, (I32)obase->op_private, FALSE);
12792 if (!svp || *svp != uninit_sv)
12795 return varname(gv, '$', 0,
12796 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
12801 o = cUNOPx(obase)->op_first;
12802 if (!o || o->op_type != OP_NULL ||
12803 ! (o->op_targ == OP_AELEM || o->op_targ == OP_HELEM))
12805 return find_uninit_var(cBINOPo->op_last, uninit_sv, match);
12809 if (PL_op == obase)
12810 /* $a[uninit_expr] or $h{uninit_expr} */
12811 return find_uninit_var(cBINOPx(obase)->op_last, uninit_sv, match);
12814 o = cBINOPx(obase)->op_first;
12815 kid = cBINOPx(obase)->op_last;
12817 /* get the av or hv, and optionally the gv */
12819 if (o->op_type == OP_PADAV || o->op_type == OP_PADHV) {
12820 sv = PAD_SV(o->op_targ);
12822 else if ((o->op_type == OP_RV2AV || o->op_type == OP_RV2HV)
12823 && cUNOPo->op_first->op_type == OP_GV)
12825 gv = cGVOPx_gv(cUNOPo->op_first);
12829 == OP_RV2HV ? MUTABLE_SV(GvHV(gv)) : MUTABLE_SV(GvAV(gv));
12834 if (kid && kid->op_type == OP_CONST && SvOK(cSVOPx_sv(kid))) {
12835 /* index is constant */
12839 if (obase->op_type == OP_HELEM) {
12840 HE* he = hv_fetch_ent(MUTABLE_HV(sv), cSVOPx_sv(kid), 0, 0);
12841 if (!he || HeVAL(he) != uninit_sv)
12845 SV * const * const svp = av_fetch(MUTABLE_AV(sv), SvIV(cSVOPx_sv(kid)), FALSE);
12846 if (!svp || *svp != uninit_sv)
12850 if (obase->op_type == OP_HELEM)
12851 return varname(gv, '%', o->op_targ,
12852 cSVOPx_sv(kid), 0, FUV_SUBSCRIPT_HASH);
12854 return varname(gv, '@', o->op_targ, NULL,
12855 SvIV(cSVOPx_sv(kid)), FUV_SUBSCRIPT_ARRAY);
12858 /* index is an expression;
12859 * attempt to find a match within the aggregate */
12860 if (obase->op_type == OP_HELEM) {
12861 SV * const keysv = find_hash_subscript((const HV*)sv, uninit_sv);
12863 return varname(gv, '%', o->op_targ,
12864 keysv, 0, FUV_SUBSCRIPT_HASH);
12868 = find_array_subscript((const AV *)sv, uninit_sv);
12870 return varname(gv, '@', o->op_targ,
12871 NULL, index, FUV_SUBSCRIPT_ARRAY);
12876 (o->op_type == OP_PADAV || o->op_type == OP_RV2AV)
12878 o->op_targ, NULL, 0, FUV_SUBSCRIPT_WITHIN);
12883 /* only examine RHS */
12884 return find_uninit_var(cBINOPx(obase)->op_first, uninit_sv, match);
12887 o = cUNOPx(obase)->op_first;
12888 if (o->op_type == OP_PUSHMARK)
12891 if (!o->op_sibling) {
12892 /* one-arg version of open is highly magical */
12894 if (o->op_type == OP_GV) { /* open FOO; */
12896 if (match && GvSV(gv) != uninit_sv)
12898 return varname(gv, '$', 0,
12899 NULL, 0, FUV_SUBSCRIPT_NONE);
12901 /* other possibilities not handled are:
12902 * open $x; or open my $x; should return '${*$x}'
12903 * open expr; should return '$'.expr ideally
12909 /* ops where $_ may be an implicit arg */
12913 if ( !(obase->op_flags & OPf_STACKED)) {
12914 if (uninit_sv == ((obase->op_private & OPpTARGET_MY)
12915 ? PAD_SVl(obase->op_targ)
12918 sv = sv_newmortal();
12919 sv_setpvs(sv, "$_");
12928 match = 1; /* print etc can return undef on defined args */
12929 /* skip filehandle as it can't produce 'undef' warning */
12930 o = cUNOPx(obase)->op_first;
12931 if ((obase->op_flags & OPf_STACKED) && o->op_type == OP_PUSHMARK)
12932 o = o->op_sibling->op_sibling;
12936 case OP_ENTEREVAL: /* could be eval $undef or $x='$undef'; eval $x */
12938 case OP_CUSTOM: /* XS or custom code could trigger random warnings */
12940 /* the following ops are capable of returning PL_sv_undef even for
12941 * defined arg(s) */
12960 case OP_GETPEERNAME:
13008 case OP_SMARTMATCH:
13017 /* XXX tmp hack: these two may call an XS sub, and currently
13018 XS subs don't have a SUB entry on the context stack, so CV and
13019 pad determination goes wrong, and BAD things happen. So, just
13020 don't try to determine the value under those circumstances.
13021 Need a better fix at dome point. DAPM 11/2007 */
13026 /* def-ness of rval pos() is independent of the def-ness of its arg */
13027 if ( !(obase->op_flags & OPf_MOD))
13032 if (SvROK(PL_rs) && uninit_sv == SvRV(PL_rs))
13033 return newSVpvs_flags("${$/}", SVs_TEMP);
13038 if (!(obase->op_flags & OPf_KIDS))
13040 o = cUNOPx(obase)->op_first;
13046 /* if all except one arg are constant, or have no side-effects,
13047 * or are optimized away, then it's unambiguous */
13049 for (kid=o; kid; kid = kid->op_sibling) {
13051 const OPCODE type = kid->op_type;
13052 if ( (type == OP_CONST && SvOK(cSVOPx_sv(kid)))
13053 || (type == OP_NULL && ! (kid->op_flags & OPf_KIDS))
13054 || (type == OP_PUSHMARK)
13058 if (o2) { /* more than one found */
13065 return find_uninit_var(o2, uninit_sv, match);
13067 /* scan all args */
13069 sv = find_uninit_var(o, uninit_sv, 1);
13081 =for apidoc report_uninit
13083 Print appropriate "Use of uninitialized variable" warning
13089 Perl_report_uninit(pTHX_ const SV *uninit_sv)
13093 SV* varname = NULL;
13095 varname = find_uninit_var(PL_op, uninit_sv,0);
13097 sv_insert(varname, 0, 0, " ", 1);
13099 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
13100 varname ? SvPV_nolen_const(varname) : "",
13101 " in ", OP_DESC(PL_op));
13104 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
13110 * c-indentation-style: bsd
13111 * c-basic-offset: 4
13112 * indent-tabs-mode: t
13115 * ex: set ts=8 sts=4 sw=4 noet: