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 if(GvGP(MUTABLE_GV(sstr))) {
3625 /* If source has method cache entry, clear it */
3627 SvREFCNT_dec(GvCV(sstr));
3631 /* If source has a real method, then a method is
3633 else if(GvCV((const GV *)sstr)) {
3638 /* If dest already had a real method, that's a change as well */
3639 if(!mro_changes && GvGP(MUTABLE_GV(dstr)) && GvCVu((const GV *)dstr)) {
3643 if(strEQ(GvNAME((const GV *)dstr),"ISA"))
3646 gp_free(MUTABLE_GV(dstr));
3647 isGV_with_GP_off(dstr);
3648 (void)SvOK_off(dstr);
3649 isGV_with_GP_on(dstr);
3650 GvINTRO_off(dstr); /* one-shot flag */
3651 GvGP(dstr) = gp_ref(GvGP(sstr));
3652 if (SvTAINTED(sstr))
3654 if (GvIMPORTED(dstr) != GVf_IMPORTED
3655 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3657 GvIMPORTED_on(dstr);
3660 if(mro_changes == 2) mro_isa_changed_in(GvSTASH(dstr));
3661 else if(mro_changes) mro_method_changed_in(GvSTASH(dstr));
3666 S_glob_assign_ref(pTHX_ SV *const dstr, SV *const sstr)
3668 SV * const sref = SvREFCNT_inc(SvRV(sstr));
3670 const int intro = GvINTRO(dstr);
3673 const U32 stype = SvTYPE(sref);
3675 PERL_ARGS_ASSERT_GLOB_ASSIGN_REF;
3678 GvINTRO_off(dstr); /* one-shot flag */
3679 GvLINE(dstr) = CopLINE(PL_curcop);
3680 GvEGV(dstr) = MUTABLE_GV(dstr);
3685 location = (SV **) &GvCV(dstr);
3686 import_flag = GVf_IMPORTED_CV;
3689 location = (SV **) &GvHV(dstr);
3690 import_flag = GVf_IMPORTED_HV;
3693 location = (SV **) &GvAV(dstr);
3694 import_flag = GVf_IMPORTED_AV;
3697 location = (SV **) &GvIOp(dstr);
3700 location = (SV **) &GvFORM(dstr);
3702 location = &GvSV(dstr);
3703 import_flag = GVf_IMPORTED_SV;
3706 if (stype == SVt_PVCV) {
3707 /*if (GvCVGEN(dstr) && (GvCV(dstr) != (const CV *)sref || GvCVGEN(dstr))) {*/
3708 if (GvCVGEN(dstr)) {
3709 SvREFCNT_dec(GvCV(dstr));
3711 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3714 SAVEGENERICSV(*location);
3718 if (stype == SVt_PVCV && (*location != sref || GvCVGEN(dstr))) {
3719 CV* const cv = MUTABLE_CV(*location);
3721 if (!GvCVGEN((const GV *)dstr) &&
3722 (CvROOT(cv) || CvXSUB(cv)))
3724 /* Redefining a sub - warning is mandatory if
3725 it was a const and its value changed. */
3726 if (CvCONST(cv) && CvCONST((const CV *)sref)
3728 == cv_const_sv((const CV *)sref)) {
3730 /* They are 2 constant subroutines generated from
3731 the same constant. This probably means that
3732 they are really the "same" proxy subroutine
3733 instantiated in 2 places. Most likely this is
3734 when a constant is exported twice. Don't warn.
3737 else if (ckWARN(WARN_REDEFINE)
3739 && (!CvCONST((const CV *)sref)
3740 || sv_cmp(cv_const_sv(cv),
3741 cv_const_sv((const CV *)
3743 Perl_warner(aTHX_ packWARN(WARN_REDEFINE),
3746 ? "Constant subroutine %s::%s redefined"
3747 : "Subroutine %s::%s redefined"),
3748 HvNAME_get(GvSTASH((const GV *)dstr)),
3749 GvENAME(MUTABLE_GV(dstr)));
3753 cv_ckproto_len(cv, (const GV *)dstr,
3754 SvPOK(sref) ? SvPVX_const(sref) : NULL,
3755 SvPOK(sref) ? SvCUR(sref) : 0);
3757 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3758 GvASSUMECV_on(dstr);
3759 if(GvSTASH(dstr)) mro_method_changed_in(GvSTASH(dstr)); /* sub foo { 1 } sub bar { 2 } *bar = \&foo */
3762 if (import_flag && !(GvFLAGS(dstr) & import_flag)
3763 && CopSTASH_ne(PL_curcop, GvSTASH(dstr))) {
3764 GvFLAGS(dstr) |= import_flag;
3769 if (SvTAINTED(sstr))
3775 Perl_sv_setsv_flags(pTHX_ SV *dstr, register SV* sstr, const I32 flags)
3778 register U32 sflags;
3780 register svtype stype;
3782 PERL_ARGS_ASSERT_SV_SETSV_FLAGS;
3787 if (SvIS_FREED(dstr)) {
3788 Perl_croak(aTHX_ "panic: attempt to copy value %" SVf
3789 " to a freed scalar %p", SVfARG(sstr), (void *)dstr);
3791 SV_CHECK_THINKFIRST_COW_DROP(dstr);
3793 sstr = &PL_sv_undef;
3794 if (SvIS_FREED(sstr)) {
3795 Perl_croak(aTHX_ "panic: attempt to copy freed scalar %p to %p",
3796 (void*)sstr, (void*)dstr);
3798 stype = SvTYPE(sstr);
3799 dtype = SvTYPE(dstr);
3801 (void)SvAMAGIC_off(dstr);
3804 /* need to nuke the magic */
3808 /* There's a lot of redundancy below but we're going for speed here */
3813 if (dtype != SVt_PVGV) {
3814 (void)SvOK_off(dstr);
3822 sv_upgrade(dstr, SVt_IV);
3826 sv_upgrade(dstr, SVt_PVIV);
3829 goto end_of_first_switch;
3831 (void)SvIOK_only(dstr);
3832 SvIV_set(dstr, SvIVX(sstr));
3835 /* SvTAINTED can only be true if the SV has taint magic, which in
3836 turn means that the SV type is PVMG (or greater). This is the
3837 case statement for SVt_IV, so this cannot be true (whatever gcov
3839 assert(!SvTAINTED(sstr));
3844 if (dtype < SVt_PV && dtype != SVt_IV)
3845 sv_upgrade(dstr, SVt_IV);
3853 sv_upgrade(dstr, SVt_NV);
3857 sv_upgrade(dstr, SVt_PVNV);
3860 goto end_of_first_switch;
3862 SvNV_set(dstr, SvNVX(sstr));
3863 (void)SvNOK_only(dstr);
3864 /* SvTAINTED can only be true if the SV has taint magic, which in
3865 turn means that the SV type is PVMG (or greater). This is the
3866 case statement for SVt_NV, so this cannot be true (whatever gcov
3868 assert(!SvTAINTED(sstr));
3874 #ifdef PERL_OLD_COPY_ON_WRITE
3875 if ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS) {
3876 if (dtype < SVt_PVIV)
3877 sv_upgrade(dstr, SVt_PVIV);
3885 sv_upgrade(dstr, SVt_PV);
3888 if (dtype < SVt_PVIV)
3889 sv_upgrade(dstr, SVt_PVIV);
3892 if (dtype < SVt_PVNV)
3893 sv_upgrade(dstr, SVt_PVNV);
3897 const char * const type = sv_reftype(sstr,0);
3899 Perl_croak(aTHX_ "Bizarre copy of %s in %s", type, OP_NAME(PL_op));
3901 Perl_croak(aTHX_ "Bizarre copy of %s", type);
3905 /* case SVt_BIND: */
3908 if (isGV_with_GP(sstr) && dtype <= SVt_PVGV) {
3909 glob_assign_glob(dstr, sstr, dtype);
3912 /* SvVALID means that this PVGV is playing at being an FBM. */
3916 if (SvGMAGICAL(sstr) && (flags & SV_GMAGIC)) {
3918 if (SvTYPE(sstr) != stype) {
3919 stype = SvTYPE(sstr);
3920 if (isGV_with_GP(sstr) && stype == SVt_PVGV && dtype <= SVt_PVGV) {
3921 glob_assign_glob(dstr, sstr, dtype);
3926 if (stype == SVt_PVLV)
3927 SvUPGRADE(dstr, SVt_PVNV);
3929 SvUPGRADE(dstr, (svtype)stype);
3931 end_of_first_switch:
3933 /* dstr may have been upgraded. */
3934 dtype = SvTYPE(dstr);
3935 sflags = SvFLAGS(sstr);
3937 if (dtype == SVt_PVCV || dtype == SVt_PVFM) {
3938 /* Assigning to a subroutine sets the prototype. */
3941 const char *const ptr = SvPV_const(sstr, len);
3943 SvGROW(dstr, len + 1);
3944 Copy(ptr, SvPVX(dstr), len + 1, char);
3945 SvCUR_set(dstr, len);
3947 SvFLAGS(dstr) |= sflags & SVf_UTF8;
3951 } else if (dtype == SVt_PVAV || dtype == SVt_PVHV) {
3952 const char * const type = sv_reftype(dstr,0);
3954 Perl_croak(aTHX_ "Cannot copy to %s in %s", type, OP_NAME(PL_op));
3956 Perl_croak(aTHX_ "Cannot copy to %s", type);
3957 } else if (sflags & SVf_ROK) {
3958 if (isGV_with_GP(dstr) && dtype == SVt_PVGV
3959 && SvTYPE(SvRV(sstr)) == SVt_PVGV && isGV_with_GP(SvRV(sstr))) {
3962 if (GvIMPORTED(dstr) != GVf_IMPORTED
3963 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3965 GvIMPORTED_on(dstr);
3970 glob_assign_glob(dstr, sstr, dtype);
3974 if (dtype >= SVt_PV) {
3975 if (dtype == SVt_PVGV && isGV_with_GP(dstr)) {
3976 glob_assign_ref(dstr, sstr);
3979 if (SvPVX_const(dstr)) {
3985 (void)SvOK_off(dstr);
3986 SvRV_set(dstr, SvREFCNT_inc(SvRV(sstr)));
3987 SvFLAGS(dstr) |= sflags & SVf_ROK;
3988 assert(!(sflags & SVp_NOK));
3989 assert(!(sflags & SVp_IOK));
3990 assert(!(sflags & SVf_NOK));
3991 assert(!(sflags & SVf_IOK));
3993 else if (dtype == SVt_PVGV && isGV_with_GP(dstr)) {
3994 if (!(sflags & SVf_OK)) {
3995 if (ckWARN(WARN_MISC))
3996 Perl_warner(aTHX_ packWARN(WARN_MISC),
3997 "Undefined value assigned to typeglob");
4000 GV *gv = gv_fetchsv(sstr, GV_ADD, SVt_PVGV);
4001 if (dstr != (const SV *)gv) {
4003 gp_free(MUTABLE_GV(dstr));
4004 GvGP(dstr) = gp_ref(GvGP(gv));
4008 else if (sflags & SVp_POK) {
4012 * Check to see if we can just swipe the string. If so, it's a
4013 * possible small lose on short strings, but a big win on long ones.
4014 * It might even be a win on short strings if SvPVX_const(dstr)
4015 * has to be allocated and SvPVX_const(sstr) has to be freed.
4016 * Likewise if we can set up COW rather than doing an actual copy, we
4017 * drop to the else clause, as the swipe code and the COW setup code
4018 * have much in common.
4021 /* Whichever path we take through the next code, we want this true,
4022 and doing it now facilitates the COW check. */
4023 (void)SvPOK_only(dstr);
4026 /* If we're already COW then this clause is not true, and if COW
4027 is allowed then we drop down to the else and make dest COW
4028 with us. If caller hasn't said that we're allowed to COW
4029 shared hash keys then we don't do the COW setup, even if the
4030 source scalar is a shared hash key scalar. */
4031 (((flags & SV_COW_SHARED_HASH_KEYS)
4032 ? (sflags & (SVf_FAKE|SVf_READONLY)) != (SVf_FAKE|SVf_READONLY)
4033 : 1 /* If making a COW copy is forbidden then the behaviour we
4034 desire is as if the source SV isn't actually already
4035 COW, even if it is. So we act as if the source flags
4036 are not COW, rather than actually testing them. */
4038 #ifndef PERL_OLD_COPY_ON_WRITE
4039 /* The change that added SV_COW_SHARED_HASH_KEYS makes the logic
4040 when PERL_OLD_COPY_ON_WRITE is defined a little wrong.
4041 Conceptually PERL_OLD_COPY_ON_WRITE being defined should
4042 override SV_COW_SHARED_HASH_KEYS, because it means "always COW"
4043 but in turn, it's somewhat dead code, never expected to go
4044 live, but more kept as a placeholder on how to do it better
4045 in a newer implementation. */
4046 /* If we are COW and dstr is a suitable target then we drop down
4047 into the else and make dest a COW of us. */
4048 || (SvFLAGS(dstr) & CAN_COW_MASK) != CAN_COW_FLAGS
4053 (sflags & SVs_TEMP) && /* slated for free anyway? */
4054 !(sflags & SVf_OOK) && /* and not involved in OOK hack? */
4055 (!(flags & SV_NOSTEAL)) &&
4056 /* and we're allowed to steal temps */
4057 SvREFCNT(sstr) == 1 && /* and no other references to it? */
4058 SvLEN(sstr) && /* and really is a string */
4059 /* and won't be needed again, potentially */
4060 !(PL_op && PL_op->op_type == OP_AASSIGN))
4061 #ifdef PERL_OLD_COPY_ON_WRITE
4062 && ((flags & SV_COW_SHARED_HASH_KEYS)
4063 ? (!((sflags & CAN_COW_MASK) == CAN_COW_FLAGS
4064 && (SvFLAGS(dstr) & CAN_COW_MASK) == CAN_COW_FLAGS
4065 && SvTYPE(sstr) >= SVt_PVIV && SvTYPE(sstr) != SVt_PVFM))
4069 /* Failed the swipe test, and it's not a shared hash key either.
4070 Have to copy the string. */
4071 STRLEN len = SvCUR(sstr);
4072 SvGROW(dstr, len + 1); /* inlined from sv_setpvn */
4073 Move(SvPVX_const(sstr),SvPVX(dstr),len,char);
4074 SvCUR_set(dstr, len);
4075 *SvEND(dstr) = '\0';
4077 /* If PERL_OLD_COPY_ON_WRITE is not defined, then isSwipe will always
4079 /* Either it's a shared hash key, or it's suitable for
4080 copy-on-write or we can swipe the string. */
4082 PerlIO_printf(Perl_debug_log, "Copy on write: sstr --> dstr\n");
4086 #ifdef PERL_OLD_COPY_ON_WRITE
4088 /* I believe I should acquire a global SV mutex if
4089 it's a COW sv (not a shared hash key) to stop
4090 it going un copy-on-write.
4091 If the source SV has gone un copy on write between up there
4092 and down here, then (assert() that) it is of the correct
4093 form to make it copy on write again */
4094 if ((sflags & (SVf_FAKE | SVf_READONLY))
4095 != (SVf_FAKE | SVf_READONLY)) {
4096 SvREADONLY_on(sstr);
4098 /* Make the source SV into a loop of 1.
4099 (about to become 2) */
4100 SV_COW_NEXT_SV_SET(sstr, sstr);
4104 /* Initial code is common. */
4105 if (SvPVX_const(dstr)) { /* we know that dtype >= SVt_PV */
4110 /* making another shared SV. */
4111 STRLEN cur = SvCUR(sstr);
4112 STRLEN len = SvLEN(sstr);
4113 #ifdef PERL_OLD_COPY_ON_WRITE
4115 assert (SvTYPE(dstr) >= SVt_PVIV);
4116 /* SvIsCOW_normal */
4117 /* splice us in between source and next-after-source. */
4118 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4119 SV_COW_NEXT_SV_SET(sstr, dstr);
4120 SvPV_set(dstr, SvPVX_mutable(sstr));
4124 /* SvIsCOW_shared_hash */
4125 DEBUG_C(PerlIO_printf(Perl_debug_log,
4126 "Copy on write: Sharing hash\n"));
4128 assert (SvTYPE(dstr) >= SVt_PV);
4130 HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)))));
4132 SvLEN_set(dstr, len);
4133 SvCUR_set(dstr, cur);
4134 SvREADONLY_on(dstr);
4136 /* Relesase a global SV mutex. */
4139 { /* Passes the swipe test. */
4140 SvPV_set(dstr, SvPVX_mutable(sstr));
4141 SvLEN_set(dstr, SvLEN(sstr));
4142 SvCUR_set(dstr, SvCUR(sstr));
4145 (void)SvOK_off(sstr); /* NOTE: nukes most SvFLAGS on sstr */
4146 SvPV_set(sstr, NULL);
4152 if (sflags & SVp_NOK) {
4153 SvNV_set(dstr, SvNVX(sstr));
4155 if (sflags & SVp_IOK) {
4156 SvIV_set(dstr, SvIVX(sstr));
4157 /* Must do this otherwise some other overloaded use of 0x80000000
4158 gets confused. I guess SVpbm_VALID */
4159 if (sflags & SVf_IVisUV)
4162 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_NOK|SVp_NOK|SVf_UTF8);
4164 const MAGIC * const smg = SvVSTRING_mg(sstr);
4166 sv_magic(dstr, NULL, PERL_MAGIC_vstring,
4167 smg->mg_ptr, smg->mg_len);
4168 SvRMAGICAL_on(dstr);
4172 else if (sflags & (SVp_IOK|SVp_NOK)) {
4173 (void)SvOK_off(dstr);
4174 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_IVisUV|SVf_NOK|SVp_NOK);
4175 if (sflags & SVp_IOK) {
4176 /* XXXX Do we want to set IsUV for IV(ROK)? Be extra safe... */
4177 SvIV_set(dstr, SvIVX(sstr));
4179 if (sflags & SVp_NOK) {
4180 SvNV_set(dstr, SvNVX(sstr));
4184 if (isGV_with_GP(sstr)) {
4185 /* This stringification rule for globs is spread in 3 places.
4186 This feels bad. FIXME. */
4187 const U32 wasfake = sflags & SVf_FAKE;
4189 /* FAKE globs can get coerced, so need to turn this off
4190 temporarily if it is on. */
4192 gv_efullname3(dstr, MUTABLE_GV(sstr), "*");
4193 SvFLAGS(sstr) |= wasfake;
4196 (void)SvOK_off(dstr);
4198 if (SvTAINTED(sstr))
4203 =for apidoc sv_setsv_mg
4205 Like C<sv_setsv>, but also handles 'set' magic.
4211 Perl_sv_setsv_mg(pTHX_ SV *const dstr, register SV *const sstr)
4213 PERL_ARGS_ASSERT_SV_SETSV_MG;
4215 sv_setsv(dstr,sstr);
4219 #ifdef PERL_OLD_COPY_ON_WRITE
4221 Perl_sv_setsv_cow(pTHX_ SV *dstr, SV *sstr)
4223 STRLEN cur = SvCUR(sstr);
4224 STRLEN len = SvLEN(sstr);
4225 register char *new_pv;
4227 PERL_ARGS_ASSERT_SV_SETSV_COW;
4230 PerlIO_printf(Perl_debug_log, "Fast copy on write: %p -> %p\n",
4231 (void*)sstr, (void*)dstr);
4238 if (SvTHINKFIRST(dstr))
4239 sv_force_normal_flags(dstr, SV_COW_DROP_PV);
4240 else if (SvPVX_const(dstr))
4241 Safefree(SvPVX_const(dstr));
4245 SvUPGRADE(dstr, SVt_PVIV);
4247 assert (SvPOK(sstr));
4248 assert (SvPOKp(sstr));
4249 assert (!SvIOK(sstr));
4250 assert (!SvIOKp(sstr));
4251 assert (!SvNOK(sstr));
4252 assert (!SvNOKp(sstr));
4254 if (SvIsCOW(sstr)) {
4256 if (SvLEN(sstr) == 0) {
4257 /* source is a COW shared hash key. */
4258 DEBUG_C(PerlIO_printf(Perl_debug_log,
4259 "Fast copy on write: Sharing hash\n"));
4260 new_pv = HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr))));
4263 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4265 assert ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS);
4266 SvUPGRADE(sstr, SVt_PVIV);
4267 SvREADONLY_on(sstr);
4269 DEBUG_C(PerlIO_printf(Perl_debug_log,
4270 "Fast copy on write: Converting sstr to COW\n"));
4271 SV_COW_NEXT_SV_SET(dstr, sstr);
4273 SV_COW_NEXT_SV_SET(sstr, dstr);
4274 new_pv = SvPVX_mutable(sstr);
4277 SvPV_set(dstr, new_pv);
4278 SvFLAGS(dstr) = (SVt_PVIV|SVf_POK|SVp_POK|SVf_FAKE|SVf_READONLY);
4281 SvLEN_set(dstr, len);
4282 SvCUR_set(dstr, cur);
4291 =for apidoc sv_setpvn
4293 Copies a string into an SV. The C<len> parameter indicates the number of
4294 bytes to be copied. If the C<ptr> argument is NULL the SV will become
4295 undefined. Does not handle 'set' magic. See C<sv_setpvn_mg>.
4301 Perl_sv_setpvn(pTHX_ register SV *const sv, register const char *const ptr, register const STRLEN len)
4304 register char *dptr;
4306 PERL_ARGS_ASSERT_SV_SETPVN;
4308 SV_CHECK_THINKFIRST_COW_DROP(sv);
4314 /* len is STRLEN which is unsigned, need to copy to signed */
4317 Perl_croak(aTHX_ "panic: sv_setpvn called with negative strlen");
4319 SvUPGRADE(sv, SVt_PV);
4321 dptr = SvGROW(sv, len + 1);
4322 Move(ptr,dptr,len,char);
4325 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4330 =for apidoc sv_setpvn_mg
4332 Like C<sv_setpvn>, but also handles 'set' magic.
4338 Perl_sv_setpvn_mg(pTHX_ register SV *const sv, register const char *const ptr, register const STRLEN len)
4340 PERL_ARGS_ASSERT_SV_SETPVN_MG;
4342 sv_setpvn(sv,ptr,len);
4347 =for apidoc sv_setpv
4349 Copies a string into an SV. The string must be null-terminated. Does not
4350 handle 'set' magic. See C<sv_setpv_mg>.
4356 Perl_sv_setpv(pTHX_ register SV *const sv, register const char *const ptr)
4359 register STRLEN len;
4361 PERL_ARGS_ASSERT_SV_SETPV;
4363 SV_CHECK_THINKFIRST_COW_DROP(sv);
4369 SvUPGRADE(sv, SVt_PV);
4371 SvGROW(sv, len + 1);
4372 Move(ptr,SvPVX(sv),len+1,char);
4374 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4379 =for apidoc sv_setpv_mg
4381 Like C<sv_setpv>, but also handles 'set' magic.
4387 Perl_sv_setpv_mg(pTHX_ register SV *const sv, register const char *const ptr)
4389 PERL_ARGS_ASSERT_SV_SETPV_MG;
4396 =for apidoc sv_usepvn_flags
4398 Tells an SV to use C<ptr> to find its string value. Normally the
4399 string is stored inside the SV but sv_usepvn allows the SV to use an
4400 outside string. The C<ptr> should point to memory that was allocated
4401 by C<malloc>. The string length, C<len>, must be supplied. By default
4402 this function will realloc (i.e. move) the memory pointed to by C<ptr>,
4403 so that pointer should not be freed or used by the programmer after
4404 giving it to sv_usepvn, and neither should any pointers from "behind"
4405 that pointer (e.g. ptr + 1) be used.
4407 If C<flags> & SV_SMAGIC is true, will call SvSETMAGIC. If C<flags> &
4408 SV_HAS_TRAILING_NUL is true, then C<ptr[len]> must be NUL, and the realloc
4409 will be skipped. (i.e. the buffer is actually at least 1 byte longer than
4410 C<len>, and already meets the requirements for storing in C<SvPVX>)
4416 Perl_sv_usepvn_flags(pTHX_ SV *const sv, char *ptr, const STRLEN len, const U32 flags)
4421 PERL_ARGS_ASSERT_SV_USEPVN_FLAGS;
4423 SV_CHECK_THINKFIRST_COW_DROP(sv);
4424 SvUPGRADE(sv, SVt_PV);
4427 if (flags & SV_SMAGIC)
4431 if (SvPVX_const(sv))
4435 if (flags & SV_HAS_TRAILING_NUL)
4436 assert(ptr[len] == '\0');
4439 allocate = (flags & SV_HAS_TRAILING_NUL)
4441 #ifdef Perl_safesysmalloc_size
4444 PERL_STRLEN_ROUNDUP(len + 1);
4446 if (flags & SV_HAS_TRAILING_NUL) {
4447 /* It's long enough - do nothing.
4448 Specfically Perl_newCONSTSUB is relying on this. */
4451 /* Force a move to shake out bugs in callers. */
4452 char *new_ptr = (char*)safemalloc(allocate);
4453 Copy(ptr, new_ptr, len, char);
4454 PoisonFree(ptr,len,char);
4458 ptr = (char*) saferealloc (ptr, allocate);
4461 #ifdef Perl_safesysmalloc_size
4462 SvLEN_set(sv, Perl_safesysmalloc_size(ptr));
4464 SvLEN_set(sv, allocate);
4468 if (!(flags & SV_HAS_TRAILING_NUL)) {
4471 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4473 if (flags & SV_SMAGIC)
4477 #ifdef PERL_OLD_COPY_ON_WRITE
4478 /* Need to do this *after* making the SV normal, as we need the buffer
4479 pointer to remain valid until after we've copied it. If we let go too early,
4480 another thread could invalidate it by unsharing last of the same hash key
4481 (which it can do by means other than releasing copy-on-write Svs)
4482 or by changing the other copy-on-write SVs in the loop. */
4484 S_sv_release_COW(pTHX_ register SV *sv, const char *pvx, SV *after)
4486 PERL_ARGS_ASSERT_SV_RELEASE_COW;
4488 { /* this SV was SvIsCOW_normal(sv) */
4489 /* we need to find the SV pointing to us. */
4490 SV *current = SV_COW_NEXT_SV(after);
4492 if (current == sv) {
4493 /* The SV we point to points back to us (there were only two of us
4495 Hence other SV is no longer copy on write either. */
4497 SvREADONLY_off(after);
4499 /* We need to follow the pointers around the loop. */
4501 while ((next = SV_COW_NEXT_SV(current)) != sv) {
4504 /* don't loop forever if the structure is bust, and we have
4505 a pointer into a closed loop. */
4506 assert (current != after);
4507 assert (SvPVX_const(current) == pvx);
4509 /* Make the SV before us point to the SV after us. */
4510 SV_COW_NEXT_SV_SET(current, after);
4516 =for apidoc sv_force_normal_flags
4518 Undo various types of fakery on an SV: if the PV is a shared string, make
4519 a private copy; if we're a ref, stop refing; if we're a glob, downgrade to
4520 an xpvmg; if we're a copy-on-write scalar, this is the on-write time when
4521 we do the copy, and is also used locally. If C<SV_COW_DROP_PV> is set
4522 then a copy-on-write scalar drops its PV buffer (if any) and becomes
4523 SvPOK_off rather than making a copy. (Used where this scalar is about to be
4524 set to some other value.) In addition, the C<flags> parameter gets passed to
4525 C<sv_unref_flags()> when unrefing. C<sv_force_normal> calls this function
4526 with flags set to 0.
4532 Perl_sv_force_normal_flags(pTHX_ register SV *const sv, const U32 flags)
4536 PERL_ARGS_ASSERT_SV_FORCE_NORMAL_FLAGS;
4538 #ifdef PERL_OLD_COPY_ON_WRITE
4539 if (SvREADONLY(sv)) {
4540 /* At this point I believe I should acquire a global SV mutex. */
4542 const char * const pvx = SvPVX_const(sv);
4543 const STRLEN len = SvLEN(sv);
4544 const STRLEN cur = SvCUR(sv);
4545 /* next COW sv in the loop. If len is 0 then this is a shared-hash
4546 key scalar, so we mustn't attempt to call SV_COW_NEXT_SV(), as
4547 we'll fail an assertion. */
4548 SV * const next = len ? SV_COW_NEXT_SV(sv) : 0;
4551 PerlIO_printf(Perl_debug_log,
4552 "Copy on write: Force normal %ld\n",
4558 /* This SV doesn't own the buffer, so need to Newx() a new one: */
4561 if (flags & SV_COW_DROP_PV) {
4562 /* OK, so we don't need to copy our buffer. */
4565 SvGROW(sv, cur + 1);
4566 Move(pvx,SvPVX(sv),cur,char);
4571 sv_release_COW(sv, pvx, next);
4573 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4579 else if (IN_PERL_RUNTIME)
4580 Perl_croak(aTHX_ "%s", PL_no_modify);
4581 /* At this point I believe that I can drop the global SV mutex. */
4584 if (SvREADONLY(sv)) {
4586 const char * const pvx = SvPVX_const(sv);
4587 const STRLEN len = SvCUR(sv);
4592 SvGROW(sv, len + 1);
4593 Move(pvx,SvPVX(sv),len,char);
4595 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4597 else if (IN_PERL_RUNTIME)
4598 Perl_croak(aTHX_ "%s", PL_no_modify);
4602 sv_unref_flags(sv, flags);
4603 else if (SvFAKE(sv) && SvTYPE(sv) == SVt_PVGV)
4610 Efficient removal of characters from the beginning of the string buffer.
4611 SvPOK(sv) must be true and the C<ptr> must be a pointer to somewhere inside
4612 the string buffer. The C<ptr> becomes the first character of the adjusted
4613 string. Uses the "OOK hack".
4614 Beware: after this function returns, C<ptr> and SvPVX_const(sv) may no longer
4615 refer to the same chunk of data.
4621 Perl_sv_chop(pTHX_ register SV *const sv, register const char *const ptr)
4627 const U8 *real_start;
4631 PERL_ARGS_ASSERT_SV_CHOP;
4633 if (!ptr || !SvPOKp(sv))
4635 delta = ptr - SvPVX_const(sv);
4637 /* Nothing to do. */
4640 /* SvPVX(sv) may move in SV_CHECK_THINKFIRST(sv), but after this line,
4641 nothing uses the value of ptr any more. */
4642 max_delta = SvLEN(sv) ? SvLEN(sv) : SvCUR(sv);
4643 if (ptr <= SvPVX_const(sv))
4644 Perl_croak(aTHX_ "panic: sv_chop ptr=%p, start=%p, end=%p",
4645 ptr, SvPVX_const(sv), SvPVX_const(sv) + max_delta);
4646 SV_CHECK_THINKFIRST(sv);
4647 if (delta > max_delta)
4648 Perl_croak(aTHX_ "panic: sv_chop ptr=%p (was %p), start=%p, end=%p",
4649 SvPVX_const(sv) + delta, ptr, SvPVX_const(sv),
4650 SvPVX_const(sv) + max_delta);
4653 if (!SvLEN(sv)) { /* make copy of shared string */
4654 const char *pvx = SvPVX_const(sv);
4655 const STRLEN len = SvCUR(sv);
4656 SvGROW(sv, len + 1);
4657 Move(pvx,SvPVX(sv),len,char);
4660 SvFLAGS(sv) |= SVf_OOK;
4663 SvOOK_offset(sv, old_delta);
4665 SvLEN_set(sv, SvLEN(sv) - delta);
4666 SvCUR_set(sv, SvCUR(sv) - delta);
4667 SvPV_set(sv, SvPVX(sv) + delta);
4669 p = (U8 *)SvPVX_const(sv);
4674 real_start = p - delta;
4678 if (delta < 0x100) {
4682 p -= sizeof(STRLEN);
4683 Copy((U8*)&delta, p, sizeof(STRLEN), U8);
4687 /* Fill the preceding buffer with sentinals to verify that no-one is
4689 while (p > real_start) {
4697 =for apidoc sv_catpvn
4699 Concatenates the string onto the end of the string which is in the SV. The
4700 C<len> indicates number of bytes to copy. If the SV has the UTF-8
4701 status set, then the bytes appended should be valid UTF-8.
4702 Handles 'get' magic, but not 'set' magic. See C<sv_catpvn_mg>.
4704 =for apidoc sv_catpvn_flags
4706 Concatenates the string onto the end of the string which is in the SV. The
4707 C<len> indicates number of bytes to copy. If the SV has the UTF-8
4708 status set, then the bytes appended should be valid UTF-8.
4709 If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<dsv> if
4710 appropriate, else not. C<sv_catpvn> and C<sv_catpvn_nomg> are implemented
4711 in terms of this function.
4717 Perl_sv_catpvn_flags(pTHX_ register SV *const dsv, register const char *sstr, register const STRLEN slen, const I32 flags)
4721 const char * const dstr = SvPV_force_flags(dsv, dlen, flags);
4723 PERL_ARGS_ASSERT_SV_CATPVN_FLAGS;
4725 SvGROW(dsv, dlen + slen + 1);
4727 sstr = SvPVX_const(dsv);
4728 Move(sstr, SvPVX(dsv) + dlen, slen, char);
4729 SvCUR_set(dsv, SvCUR(dsv) + slen);
4731 (void)SvPOK_only_UTF8(dsv); /* validate pointer */
4733 if (flags & SV_SMAGIC)
4738 =for apidoc sv_catsv
4740 Concatenates the string from SV C<ssv> onto the end of the string in
4741 SV C<dsv>. Modifies C<dsv> but not C<ssv>. Handles 'get' magic, but
4742 not 'set' magic. See C<sv_catsv_mg>.
4744 =for apidoc sv_catsv_flags
4746 Concatenates the string from SV C<ssv> onto the end of the string in
4747 SV C<dsv>. Modifies C<dsv> but not C<ssv>. If C<flags> has C<SV_GMAGIC>
4748 bit set, will C<mg_get> on the SVs if appropriate, else not. C<sv_catsv>
4749 and C<sv_catsv_nomg> are implemented in terms of this function.
4754 Perl_sv_catsv_flags(pTHX_ SV *const dsv, register SV *const ssv, const I32 flags)
4758 PERL_ARGS_ASSERT_SV_CATSV_FLAGS;
4762 const char *spv = SvPV_const(ssv, slen);
4764 /* sutf8 and dutf8 were type bool, but under USE_ITHREADS,
4765 gcc version 2.95.2 20000220 (Debian GNU/Linux) for
4766 Linux xxx 2.2.17 on sparc64 with gcc -O2, we erroneously
4767 get dutf8 = 0x20000000, (i.e. SVf_UTF8) even though
4768 dsv->sv_flags doesn't have that bit set.
4769 Andy Dougherty 12 Oct 2001
4771 const I32 sutf8 = DO_UTF8(ssv);
4774 if (SvGMAGICAL(dsv) && (flags & SV_GMAGIC))
4776 dutf8 = DO_UTF8(dsv);
4778 if (dutf8 != sutf8) {
4780 /* Not modifying source SV, so taking a temporary copy. */
4781 SV* const csv = newSVpvn_flags(spv, slen, SVs_TEMP);
4783 sv_utf8_upgrade(csv);
4784 spv = SvPV_const(csv, slen);
4787 /* Leave enough space for the cat that's about to happen */
4788 sv_utf8_upgrade_flags_grow(dsv, 0, slen);
4790 sv_catpvn_nomg(dsv, spv, slen);
4793 if (flags & SV_SMAGIC)
4798 =for apidoc sv_catpv
4800 Concatenates the string onto the end of the string which is in the SV.
4801 If the SV has the UTF-8 status set, then the bytes appended should be
4802 valid UTF-8. Handles 'get' magic, but not 'set' magic. See C<sv_catpv_mg>.
4807 Perl_sv_catpv(pTHX_ register SV *const sv, register const char *ptr)
4810 register STRLEN len;
4814 PERL_ARGS_ASSERT_SV_CATPV;
4818 junk = SvPV_force(sv, tlen);
4820 SvGROW(sv, tlen + len + 1);
4822 ptr = SvPVX_const(sv);
4823 Move(ptr,SvPVX(sv)+tlen,len+1,char);
4824 SvCUR_set(sv, SvCUR(sv) + len);
4825 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4830 =for apidoc sv_catpv_mg
4832 Like C<sv_catpv>, but also handles 'set' magic.
4838 Perl_sv_catpv_mg(pTHX_ register SV *const sv, register const char *const ptr)
4840 PERL_ARGS_ASSERT_SV_CATPV_MG;
4849 Creates a new SV. A non-zero C<len> parameter indicates the number of
4850 bytes of preallocated string space the SV should have. An extra byte for a
4851 trailing NUL is also reserved. (SvPOK is not set for the SV even if string
4852 space is allocated.) The reference count for the new SV is set to 1.
4854 In 5.9.3, newSV() replaces the older NEWSV() API, and drops the first
4855 parameter, I<x>, a debug aid which allowed callers to identify themselves.
4856 This aid has been superseded by a new build option, PERL_MEM_LOG (see
4857 L<perlhack/PERL_MEM_LOG>). The older API is still there for use in XS
4858 modules supporting older perls.
4864 Perl_newSV(pTHX_ const STRLEN len)
4871 sv_upgrade(sv, SVt_PV);
4872 SvGROW(sv, len + 1);
4877 =for apidoc sv_magicext
4879 Adds magic to an SV, upgrading it if necessary. Applies the
4880 supplied vtable and returns a pointer to the magic added.
4882 Note that C<sv_magicext> will allow things that C<sv_magic> will not.
4883 In particular, you can add magic to SvREADONLY SVs, and add more than
4884 one instance of the same 'how'.
4886 If C<namlen> is greater than zero then a C<savepvn> I<copy> of C<name> is
4887 stored, if C<namlen> is zero then C<name> is stored as-is and - as another
4888 special case - if C<(name && namlen == HEf_SVKEY)> then C<name> is assumed
4889 to contain an C<SV*> and is stored as-is with its REFCNT incremented.
4891 (This is now used as a subroutine by C<sv_magic>.)
4896 Perl_sv_magicext(pTHX_ SV *const sv, SV *const obj, const int how,
4897 const MGVTBL *const vtable, const char *const name, const I32 namlen)
4902 PERL_ARGS_ASSERT_SV_MAGICEXT;
4904 SvUPGRADE(sv, SVt_PVMG);
4905 Newxz(mg, 1, MAGIC);
4906 mg->mg_moremagic = SvMAGIC(sv);
4907 SvMAGIC_set(sv, mg);
4909 /* Sometimes a magic contains a reference loop, where the sv and
4910 object refer to each other. To prevent a reference loop that
4911 would prevent such objects being freed, we look for such loops
4912 and if we find one we avoid incrementing the object refcount.
4914 Note we cannot do this to avoid self-tie loops as intervening RV must
4915 have its REFCNT incremented to keep it in existence.
4918 if (!obj || obj == sv ||
4919 how == PERL_MAGIC_arylen ||
4920 how == PERL_MAGIC_symtab ||
4921 (SvTYPE(obj) == SVt_PVGV &&
4922 (GvSV(obj) == sv || GvHV(obj) == (const HV *)sv
4923 || GvAV(obj) == (const AV *)sv || GvCV(obj) == (const CV *)sv
4924 || GvIOp(obj) == (const IO *)sv || GvFORM(obj) == (const CV *)sv)))
4929 mg->mg_obj = SvREFCNT_inc_simple(obj);
4930 mg->mg_flags |= MGf_REFCOUNTED;
4933 /* Normal self-ties simply pass a null object, and instead of
4934 using mg_obj directly, use the SvTIED_obj macro to produce a
4935 new RV as needed. For glob "self-ties", we are tieing the PVIO
4936 with an RV obj pointing to the glob containing the PVIO. In
4937 this case, to avoid a reference loop, we need to weaken the
4941 if (how == PERL_MAGIC_tiedscalar && SvTYPE(sv) == SVt_PVIO &&
4942 obj && SvROK(obj) && GvIO(SvRV(obj)) == (const IO *)sv)
4948 mg->mg_len = namlen;
4951 mg->mg_ptr = savepvn(name, namlen);
4952 else if (namlen == HEf_SVKEY) {
4953 /* Yes, this is casting away const. This is only for the case of
4954 HEf_SVKEY. I think we need to document this abberation of the
4955 constness of the API, rather than making name non-const, as
4956 that change propagating outwards a long way. */
4957 mg->mg_ptr = (char*)SvREFCNT_inc_simple_NN((SV *)name);
4959 mg->mg_ptr = (char *) name;
4961 mg->mg_virtual = (MGVTBL *) vtable;
4965 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
4970 =for apidoc sv_magic
4972 Adds magic to an SV. First upgrades C<sv> to type C<SVt_PVMG> if necessary,
4973 then adds a new magic item of type C<how> to the head of the magic list.
4975 See C<sv_magicext> (which C<sv_magic> now calls) for a description of the
4976 handling of the C<name> and C<namlen> arguments.
4978 You need to use C<sv_magicext> to add magic to SvREADONLY SVs and also
4979 to add more than one instance of the same 'how'.
4985 Perl_sv_magic(pTHX_ register SV *const sv, SV *const obj, const int how,
4986 const char *const name, const I32 namlen)
4989 const MGVTBL *vtable;
4992 PERL_ARGS_ASSERT_SV_MAGIC;
4994 #ifdef PERL_OLD_COPY_ON_WRITE
4996 sv_force_normal_flags(sv, 0);
4998 if (SvREADONLY(sv)) {
5000 /* its okay to attach magic to shared strings; the subsequent
5001 * upgrade to PVMG will unshare the string */
5002 !(SvFAKE(sv) && SvTYPE(sv) < SVt_PVMG)
5005 && how != PERL_MAGIC_regex_global
5006 && how != PERL_MAGIC_bm
5007 && how != PERL_MAGIC_fm
5008 && how != PERL_MAGIC_sv
5009 && how != PERL_MAGIC_backref
5012 Perl_croak(aTHX_ "%s", PL_no_modify);
5015 if (SvMAGICAL(sv) || (how == PERL_MAGIC_taint && SvTYPE(sv) >= SVt_PVMG)) {
5016 if (SvMAGIC(sv) && (mg = mg_find(sv, how))) {
5017 /* sv_magic() refuses to add a magic of the same 'how' as an
5020 if (how == PERL_MAGIC_taint) {
5022 /* Any scalar which already had taint magic on which someone
5023 (erroneously?) did SvIOK_on() or similar will now be
5024 incorrectly sporting public "OK" flags. */
5025 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
5033 vtable = &PL_vtbl_sv;
5035 case PERL_MAGIC_overload:
5036 vtable = &PL_vtbl_amagic;
5038 case PERL_MAGIC_overload_elem:
5039 vtable = &PL_vtbl_amagicelem;
5041 case PERL_MAGIC_overload_table:
5042 vtable = &PL_vtbl_ovrld;
5045 vtable = &PL_vtbl_bm;
5047 case PERL_MAGIC_regdata:
5048 vtable = &PL_vtbl_regdata;
5050 case PERL_MAGIC_regdatum:
5051 vtable = &PL_vtbl_regdatum;
5053 case PERL_MAGIC_env:
5054 vtable = &PL_vtbl_env;
5057 vtable = &PL_vtbl_fm;
5059 case PERL_MAGIC_envelem:
5060 vtable = &PL_vtbl_envelem;
5062 case PERL_MAGIC_regex_global:
5063 vtable = &PL_vtbl_mglob;
5065 case PERL_MAGIC_isa:
5066 vtable = &PL_vtbl_isa;
5068 case PERL_MAGIC_isaelem:
5069 vtable = &PL_vtbl_isaelem;
5071 case PERL_MAGIC_nkeys:
5072 vtable = &PL_vtbl_nkeys;
5074 case PERL_MAGIC_dbfile:
5077 case PERL_MAGIC_dbline:
5078 vtable = &PL_vtbl_dbline;
5080 #ifdef USE_LOCALE_COLLATE
5081 case PERL_MAGIC_collxfrm:
5082 vtable = &PL_vtbl_collxfrm;
5084 #endif /* USE_LOCALE_COLLATE */
5085 case PERL_MAGIC_tied:
5086 vtable = &PL_vtbl_pack;
5088 case PERL_MAGIC_tiedelem:
5089 case PERL_MAGIC_tiedscalar:
5090 vtable = &PL_vtbl_packelem;
5093 vtable = &PL_vtbl_regexp;
5095 case PERL_MAGIC_hints:
5096 /* As this vtable is all NULL, we can reuse it. */
5097 case PERL_MAGIC_sig:
5098 vtable = &PL_vtbl_sig;
5100 case PERL_MAGIC_sigelem:
5101 vtable = &PL_vtbl_sigelem;
5103 case PERL_MAGIC_taint:
5104 vtable = &PL_vtbl_taint;
5106 case PERL_MAGIC_uvar:
5107 vtable = &PL_vtbl_uvar;
5109 case PERL_MAGIC_vec:
5110 vtable = &PL_vtbl_vec;
5112 case PERL_MAGIC_arylen_p:
5113 case PERL_MAGIC_rhash:
5114 case PERL_MAGIC_symtab:
5115 case PERL_MAGIC_vstring:
5118 case PERL_MAGIC_utf8:
5119 vtable = &PL_vtbl_utf8;
5121 case PERL_MAGIC_substr:
5122 vtable = &PL_vtbl_substr;
5124 case PERL_MAGIC_defelem:
5125 vtable = &PL_vtbl_defelem;
5127 case PERL_MAGIC_arylen:
5128 vtable = &PL_vtbl_arylen;
5130 case PERL_MAGIC_pos:
5131 vtable = &PL_vtbl_pos;
5133 case PERL_MAGIC_backref:
5134 vtable = &PL_vtbl_backref;
5136 case PERL_MAGIC_hintselem:
5137 vtable = &PL_vtbl_hintselem;
5139 case PERL_MAGIC_ext:
5140 /* Reserved for use by extensions not perl internals. */
5141 /* Useful for attaching extension internal data to perl vars. */
5142 /* Note that multiple extensions may clash if magical scalars */
5143 /* etc holding private data from one are passed to another. */
5147 Perl_croak(aTHX_ "Don't know how to handle magic of type \\%o", how);
5150 /* Rest of work is done else where */
5151 mg = sv_magicext(sv,obj,how,vtable,name,namlen);
5154 case PERL_MAGIC_taint:
5157 case PERL_MAGIC_ext:
5158 case PERL_MAGIC_dbfile:
5165 =for apidoc sv_unmagic
5167 Removes all magic of type C<type> from an SV.
5173 Perl_sv_unmagic(pTHX_ SV *const sv, const int type)
5178 PERL_ARGS_ASSERT_SV_UNMAGIC;
5180 if (SvTYPE(sv) < SVt_PVMG || !SvMAGIC(sv))
5182 mgp = &(((XPVMG*) SvANY(sv))->xmg_u.xmg_magic);
5183 for (mg = *mgp; mg; mg = *mgp) {
5184 if (mg->mg_type == type) {
5185 const MGVTBL* const vtbl = mg->mg_virtual;
5186 *mgp = mg->mg_moremagic;
5187 if (vtbl && vtbl->svt_free)
5188 CALL_FPTR(vtbl->svt_free)(aTHX_ sv, mg);
5189 if (mg->mg_ptr && mg->mg_type != PERL_MAGIC_regex_global) {
5191 Safefree(mg->mg_ptr);
5192 else if (mg->mg_len == HEf_SVKEY)
5193 SvREFCNT_dec(MUTABLE_SV(mg->mg_ptr));
5194 else if (mg->mg_type == PERL_MAGIC_utf8)
5195 Safefree(mg->mg_ptr);
5197 if (mg->mg_flags & MGf_REFCOUNTED)
5198 SvREFCNT_dec(mg->mg_obj);
5202 mgp = &mg->mg_moremagic;
5206 SvFLAGS(sv) |= (SvFLAGS(sv) & (SVp_IOK|SVp_NOK|SVp_POK)) >> PRIVSHIFT;
5207 SvMAGIC_set(sv, NULL);
5214 =for apidoc sv_rvweaken
5216 Weaken a reference: set the C<SvWEAKREF> flag on this RV; give the
5217 referred-to SV C<PERL_MAGIC_backref> magic if it hasn't already; and
5218 push a back-reference to this RV onto the array of backreferences
5219 associated with that magic. If the RV is magical, set magic will be
5220 called after the RV is cleared.
5226 Perl_sv_rvweaken(pTHX_ SV *const sv)
5230 PERL_ARGS_ASSERT_SV_RVWEAKEN;
5232 if (!SvOK(sv)) /* let undefs pass */
5235 Perl_croak(aTHX_ "Can't weaken a nonreference");
5236 else if (SvWEAKREF(sv)) {
5237 if (ckWARN(WARN_MISC))
5238 Perl_warner(aTHX_ packWARN(WARN_MISC), "Reference is already weak");
5242 Perl_sv_add_backref(aTHX_ tsv, sv);
5248 /* Give tsv backref magic if it hasn't already got it, then push a
5249 * back-reference to sv onto the array associated with the backref magic.
5252 /* A discussion about the backreferences array and its refcount:
5254 * The AV holding the backreferences is pointed to either as the mg_obj of
5255 * PERL_MAGIC_backref, or in the specific case of a HV that has the hv_aux
5256 * structure, from the xhv_backreferences field. (A HV without hv_aux will
5257 * have the standard magic instead.) The array is created with a refcount
5258 * of 2. This means that if during global destruction the array gets
5259 * picked on first to have its refcount decremented by the random zapper,
5260 * it won't actually be freed, meaning it's still theere for when its
5261 * parent gets freed.
5262 * When the parent SV is freed, in the case of magic, the magic is freed,
5263 * Perl_magic_killbackrefs is called which decrements one refcount, then
5264 * mg_obj is freed which kills the second count.
5265 * In the vase of a HV being freed, one ref is removed by
5266 * Perl_hv_kill_backrefs, the other by Perl_sv_kill_backrefs, which it
5271 Perl_sv_add_backref(pTHX_ SV *const tsv, SV *const sv)
5276 PERL_ARGS_ASSERT_SV_ADD_BACKREF;
5278 if (SvTYPE(tsv) == SVt_PVHV) {
5279 AV **const avp = Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5283 /* There is no AV in the offical place - try a fixup. */
5284 MAGIC *const mg = mg_find(tsv, PERL_MAGIC_backref);
5287 /* Aha. They've got it stowed in magic. Bring it back. */
5288 av = MUTABLE_AV(mg->mg_obj);
5289 /* Stop mg_free decreasing the refernce count. */
5291 /* Stop mg_free even calling the destructor, given that
5292 there's no AV to free up. */
5294 sv_unmagic(tsv, PERL_MAGIC_backref);
5298 SvREFCNT_inc_simple_void(av); /* see discussion above */
5303 const MAGIC *const mg
5304 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
5306 av = MUTABLE_AV(mg->mg_obj);
5310 sv_magic(tsv, MUTABLE_SV(av), PERL_MAGIC_backref, NULL, 0);
5311 /* av now has a refcnt of 2; see discussion above */
5314 if (AvFILLp(av) >= AvMAX(av)) {
5315 av_extend(av, AvFILLp(av)+1);
5317 AvARRAY(av)[++AvFILLp(av)] = sv; /* av_push() */
5320 /* delete a back-reference to ourselves from the backref magic associated
5321 * with the SV we point to.
5325 S_sv_del_backref(pTHX_ SV *const tsv, SV *const sv)
5332 PERL_ARGS_ASSERT_SV_DEL_BACKREF;
5334 if (SvTYPE(tsv) == SVt_PVHV && SvOOK(tsv)) {
5335 av = *Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5336 /* We mustn't attempt to "fix up" the hash here by moving the
5337 backreference array back to the hv_aux structure, as that is stored
5338 in the main HvARRAY(), and hfreentries assumes that no-one
5339 reallocates HvARRAY() while it is running. */
5342 const MAGIC *const mg
5343 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
5345 av = MUTABLE_AV(mg->mg_obj);
5349 Perl_croak(aTHX_ "panic: del_backref");
5351 assert(!SvIS_FREED(av));
5354 /* We shouldn't be in here more than once, but for paranoia reasons lets
5356 for (i = AvFILLp(av); i >= 0; i--) {
5358 const SSize_t fill = AvFILLp(av);
5360 /* We weren't the last entry.
5361 An unordered list has this property that you can take the
5362 last element off the end to fill the hole, and it's still
5363 an unordered list :-)
5368 AvFILLp(av) = fill - 1;
5374 Perl_sv_kill_backrefs(pTHX_ SV *const sv, AV *const av)
5376 SV **svp = AvARRAY(av);
5378 PERL_ARGS_ASSERT_SV_KILL_BACKREFS;
5379 PERL_UNUSED_ARG(sv);
5381 assert(!svp || !SvIS_FREED(av));
5383 SV *const *const last = svp + AvFILLp(av);
5385 while (svp <= last) {
5387 SV *const referrer = *svp;
5388 if (SvWEAKREF(referrer)) {
5389 /* XXX Should we check that it hasn't changed? */
5390 SvRV_set(referrer, 0);
5392 SvWEAKREF_off(referrer);
5393 SvSETMAGIC(referrer);
5394 } else if (SvTYPE(referrer) == SVt_PVGV ||
5395 SvTYPE(referrer) == SVt_PVLV) {
5396 /* You lookin' at me? */
5397 assert(GvSTASH(referrer));
5398 assert(GvSTASH(referrer) == (const HV *)sv);
5399 GvSTASH(referrer) = 0;
5402 "panic: magic_killbackrefs (flags=%"UVxf")",
5403 (UV)SvFLAGS(referrer));
5411 SvREFCNT_dec(av); /* remove extra count added by sv_add_backref() */
5416 =for apidoc sv_insert
5418 Inserts a string at the specified offset/length within the SV. Similar to
5419 the Perl substr() function. Handles get magic.
5421 =for apidoc sv_insert_flags
5423 Same as C<sv_insert>, but the extra C<flags> are passed the C<SvPV_force_flags> that applies to C<bigstr>.
5429 Perl_sv_insert_flags(pTHX_ SV *const bigstr, const STRLEN offset, const STRLEN len, const char *const little, const STRLEN littlelen, const U32 flags)
5434 register char *midend;
5435 register char *bigend;
5439 PERL_ARGS_ASSERT_SV_INSERT_FLAGS;
5442 Perl_croak(aTHX_ "Can't modify non-existent substring");
5443 SvPV_force_flags(bigstr, curlen, flags);
5444 (void)SvPOK_only_UTF8(bigstr);
5445 if (offset + len > curlen) {
5446 SvGROW(bigstr, offset+len+1);
5447 Zero(SvPVX(bigstr)+curlen, offset+len-curlen, char);
5448 SvCUR_set(bigstr, offset+len);
5452 i = littlelen - len;
5453 if (i > 0) { /* string might grow */
5454 big = SvGROW(bigstr, SvCUR(bigstr) + i + 1);
5455 mid = big + offset + len;
5456 midend = bigend = big + SvCUR(bigstr);
5459 while (midend > mid) /* shove everything down */
5460 *--bigend = *--midend;
5461 Move(little,big+offset,littlelen,char);
5462 SvCUR_set(bigstr, SvCUR(bigstr) + i);
5467 Move(little,SvPVX(bigstr)+offset,len,char);
5472 big = SvPVX(bigstr);
5475 bigend = big + SvCUR(bigstr);
5477 if (midend > bigend)
5478 Perl_croak(aTHX_ "panic: sv_insert");
5480 if (mid - big > bigend - midend) { /* faster to shorten from end */
5482 Move(little, mid, littlelen,char);
5485 i = bigend - midend;
5487 Move(midend, mid, i,char);
5491 SvCUR_set(bigstr, mid - big);
5493 else if ((i = mid - big)) { /* faster from front */
5494 midend -= littlelen;
5496 Move(big, midend - i, i, char);
5497 sv_chop(bigstr,midend-i);
5499 Move(little, mid, littlelen,char);
5501 else if (littlelen) {
5502 midend -= littlelen;
5503 sv_chop(bigstr,midend);
5504 Move(little,midend,littlelen,char);
5507 sv_chop(bigstr,midend);
5513 =for apidoc sv_replace
5515 Make the first argument a copy of the second, then delete the original.
5516 The target SV physically takes over ownership of the body of the source SV
5517 and inherits its flags; however, the target keeps any magic it owns,
5518 and any magic in the source is discarded.
5519 Note that this is a rather specialist SV copying operation; most of the
5520 time you'll want to use C<sv_setsv> or one of its many macro front-ends.
5526 Perl_sv_replace(pTHX_ register SV *const sv, register SV *const nsv)
5529 const U32 refcnt = SvREFCNT(sv);
5531 PERL_ARGS_ASSERT_SV_REPLACE;
5533 SV_CHECK_THINKFIRST_COW_DROP(sv);
5534 if (SvREFCNT(nsv) != 1) {
5535 Perl_croak(aTHX_ "panic: reference miscount on nsv in sv_replace() (%"
5536 UVuf " != 1)", (UV) SvREFCNT(nsv));
5538 if (SvMAGICAL(sv)) {
5542 sv_upgrade(nsv, SVt_PVMG);
5543 SvMAGIC_set(nsv, SvMAGIC(sv));
5544 SvFLAGS(nsv) |= SvMAGICAL(sv);
5546 SvMAGIC_set(sv, NULL);
5550 assert(!SvREFCNT(sv));
5551 #ifdef DEBUG_LEAKING_SCALARS
5552 sv->sv_flags = nsv->sv_flags;
5553 sv->sv_any = nsv->sv_any;
5554 sv->sv_refcnt = nsv->sv_refcnt;
5555 sv->sv_u = nsv->sv_u;
5557 StructCopy(nsv,sv,SV);
5559 if(SvTYPE(sv) == SVt_IV) {
5561 = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
5565 #ifdef PERL_OLD_COPY_ON_WRITE
5566 if (SvIsCOW_normal(nsv)) {
5567 /* We need to follow the pointers around the loop to make the
5568 previous SV point to sv, rather than nsv. */
5571 while ((next = SV_COW_NEXT_SV(current)) != nsv) {
5574 assert(SvPVX_const(current) == SvPVX_const(nsv));
5576 /* Make the SV before us point to the SV after us. */
5578 PerlIO_printf(Perl_debug_log, "previous is\n");
5580 PerlIO_printf(Perl_debug_log,
5581 "move it from 0x%"UVxf" to 0x%"UVxf"\n",
5582 (UV) SV_COW_NEXT_SV(current), (UV) sv);
5584 SV_COW_NEXT_SV_SET(current, sv);
5587 SvREFCNT(sv) = refcnt;
5588 SvFLAGS(nsv) |= SVTYPEMASK; /* Mark as freed */
5594 =for apidoc sv_clear
5596 Clear an SV: call any destructors, free up any memory used by the body,
5597 and free the body itself. The SV's head is I<not> freed, although
5598 its type is set to all 1's so that it won't inadvertently be assumed
5599 to be live during global destruction etc.
5600 This function should only be called when REFCNT is zero. Most of the time
5601 you'll want to call C<sv_free()> (or its macro wrapper C<SvREFCNT_dec>)
5608 Perl_sv_clear(pTHX_ register SV *const sv)
5611 const U32 type = SvTYPE(sv);
5612 const struct body_details *const sv_type_details
5613 = bodies_by_type + type;
5616 PERL_ARGS_ASSERT_SV_CLEAR;
5617 assert(SvREFCNT(sv) == 0);
5618 assert(SvTYPE(sv) != SVTYPEMASK);
5620 if (type <= SVt_IV) {
5621 /* See the comment in sv.h about the collusion between this early
5622 return and the overloading of the NULL and IV slots in the size
5625 SV * const target = SvRV(sv);
5627 sv_del_backref(target, sv);
5629 SvREFCNT_dec(target);
5631 SvFLAGS(sv) &= SVf_BREAK;
5632 SvFLAGS(sv) |= SVTYPEMASK;
5637 if (PL_defstash && /* Still have a symbol table? */
5644 stash = SvSTASH(sv);
5645 destructor = StashHANDLER(stash,DESTROY);
5647 /* A constant subroutine can have no side effects, so
5648 don't bother calling it. */
5649 && !CvCONST(destructor)
5650 /* Don't bother calling an empty destructor */
5651 && (CvISXSUB(destructor)
5652 || CvSTART(destructor)->op_next->op_type != OP_LEAVESUB))
5654 SV* const tmpref = newRV(sv);
5655 SvREADONLY_on(tmpref); /* DESTROY() could be naughty */
5657 PUSHSTACKi(PERLSI_DESTROY);
5662 call_sv(MUTABLE_SV(destructor), G_DISCARD|G_EVAL|G_KEEPERR|G_VOID);
5668 if(SvREFCNT(tmpref) < 2) {
5669 /* tmpref is not kept alive! */
5671 SvRV_set(tmpref, NULL);
5674 SvREFCNT_dec(tmpref);
5676 } while (SvOBJECT(sv) && SvSTASH(sv) != stash);
5680 if (PL_in_clean_objs)
5681 Perl_croak(aTHX_ "DESTROY created new reference to dead object '%s'",
5683 /* DESTROY gave object new lease on life */
5689 SvREFCNT_dec(SvSTASH(sv)); /* possibly of changed persuasion */
5690 SvOBJECT_off(sv); /* Curse the object. */
5691 if (type != SVt_PVIO)
5692 --PL_sv_objcount; /* XXX Might want something more general */
5695 if (type >= SVt_PVMG) {
5696 if (type == SVt_PVMG && SvPAD_OUR(sv)) {
5697 SvREFCNT_dec(SvOURSTASH(sv));
5698 } else if (SvMAGIC(sv))
5700 if (type == SVt_PVMG && SvPAD_TYPED(sv))
5701 SvREFCNT_dec(SvSTASH(sv));
5704 /* case SVt_BIND: */
5707 IoIFP(sv) != PerlIO_stdin() &&
5708 IoIFP(sv) != PerlIO_stdout() &&
5709 IoIFP(sv) != PerlIO_stderr())
5711 io_close(MUTABLE_IO(sv), FALSE);
5713 if (IoDIRP(sv) && !(IoFLAGS(sv) & IOf_FAKE_DIRP))
5714 PerlDir_close(IoDIRP(sv));
5715 IoDIRP(sv) = (DIR*)NULL;
5716 Safefree(IoTOP_NAME(sv));
5717 Safefree(IoFMT_NAME(sv));
5718 Safefree(IoBOTTOM_NAME(sv));
5721 /* FIXME for plugins */
5722 pregfree2((REGEXP*) sv);
5726 cv_undef(MUTABLE_CV(sv));
5729 if (PL_last_swash_hv == (const HV *)sv) {
5730 PL_last_swash_hv = NULL;
5732 Perl_hv_kill_backrefs(aTHX_ MUTABLE_HV(sv));
5733 hv_undef(MUTABLE_HV(sv));
5736 if (PL_comppad == MUTABLE_AV(sv)) {
5740 av_undef(MUTABLE_AV(sv));
5743 if (LvTYPE(sv) == 'T') { /* for tie: return HE to pool */
5744 SvREFCNT_dec(HeKEY_sv((HE*)LvTARG(sv)));
5745 HeNEXT((HE*)LvTARG(sv)) = PL_hv_fetch_ent_mh;
5746 PL_hv_fetch_ent_mh = (HE*)LvTARG(sv);
5748 else if (LvTYPE(sv) != 't') /* unless tie: unrefcnted fake SV** */
5749 SvREFCNT_dec(LvTARG(sv));
5751 if (isGV_with_GP(sv)) {
5752 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
5753 && HvNAME_get(stash))
5754 mro_method_changed_in(stash);
5755 gp_free(MUTABLE_GV(sv));
5757 unshare_hek(GvNAME_HEK(sv));
5758 /* If we're in a stash, we don't own a reference to it. However it does
5759 have a back reference to us, which needs to be cleared. */
5760 if (!SvVALID(sv) && (stash = GvSTASH(sv)))
5761 sv_del_backref(MUTABLE_SV(stash), sv);
5763 /* FIXME. There are probably more unreferenced pointers to SVs in the
5764 interpreter struct that we should check and tidy in a similar
5766 if ((const GV *)sv == PL_last_in_gv)
5767 PL_last_in_gv = NULL;
5773 /* Don't bother with SvOOK_off(sv); as we're only going to free it. */
5776 SvOOK_offset(sv, offset);
5777 SvPV_set(sv, SvPVX_mutable(sv) - offset);
5778 /* Don't even bother with turning off the OOK flag. */
5781 SV * const target = SvRV(sv);
5783 sv_del_backref(target, sv);
5785 SvREFCNT_dec(target);
5787 #ifdef PERL_OLD_COPY_ON_WRITE
5788 else if (SvPVX_const(sv)) {
5790 /* I believe I need to grab the global SV mutex here and
5791 then recheck the COW status. */
5793 PerlIO_printf(Perl_debug_log, "Copy on write: clear\n");
5797 sv_release_COW(sv, SvPVX_const(sv), SV_COW_NEXT_SV(sv));
5799 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
5802 /* And drop it here. */
5804 } else if (SvLEN(sv)) {
5805 Safefree(SvPVX_const(sv));
5809 else if (SvPVX_const(sv) && SvLEN(sv))
5810 Safefree(SvPVX_mutable(sv));
5811 else if (SvPVX_const(sv) && SvREADONLY(sv) && SvFAKE(sv)) {
5812 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
5821 SvFLAGS(sv) &= SVf_BREAK;
5822 SvFLAGS(sv) |= SVTYPEMASK;
5824 if (sv_type_details->arena) {
5825 del_body(((char *)SvANY(sv) + sv_type_details->offset),
5826 &PL_body_roots[type]);
5828 else if (sv_type_details->body_size) {
5829 my_safefree(SvANY(sv));
5834 =for apidoc sv_newref
5836 Increment an SV's reference count. Use the C<SvREFCNT_inc()> wrapper
5843 Perl_sv_newref(pTHX_ SV *const sv)
5845 PERL_UNUSED_CONTEXT;
5854 Decrement an SV's reference count, and if it drops to zero, call
5855 C<sv_clear> to invoke destructors and free up any memory used by
5856 the body; finally, deallocate the SV's head itself.
5857 Normally called via a wrapper macro C<SvREFCNT_dec>.
5863 Perl_sv_free(pTHX_ SV *const sv)
5868 if (SvREFCNT(sv) == 0) {
5869 if (SvFLAGS(sv) & SVf_BREAK)
5870 /* this SV's refcnt has been artificially decremented to
5871 * trigger cleanup */
5873 if (PL_in_clean_all) /* All is fair */
5875 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
5876 /* make sure SvREFCNT(sv)==0 happens very seldom */
5877 SvREFCNT(sv) = (~(U32)0)/2;
5880 if (ckWARN_d(WARN_INTERNAL)) {
5881 #ifdef DEBUG_LEAKING_SCALARS_FORK_DUMP
5882 Perl_dump_sv_child(aTHX_ sv);
5884 #ifdef DEBUG_LEAKING_SCALARS
5887 #ifdef DEBUG_LEAKING_SCALARS_ABORT
5888 if (PL_warnhook == PERL_WARNHOOK_FATAL
5889 || ckDEAD(packWARN(WARN_INTERNAL))) {
5890 /* Don't let Perl_warner cause us to escape our fate: */
5894 /* This may not return: */
5895 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
5896 "Attempt to free unreferenced scalar: SV 0x%"UVxf
5897 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
5900 #ifdef DEBUG_LEAKING_SCALARS_ABORT
5905 if (--(SvREFCNT(sv)) > 0)
5907 Perl_sv_free2(aTHX_ sv);
5911 Perl_sv_free2(pTHX_ SV *const sv)
5915 PERL_ARGS_ASSERT_SV_FREE2;
5919 if (ckWARN_d(WARN_DEBUGGING))
5920 Perl_warner(aTHX_ packWARN(WARN_DEBUGGING),
5921 "Attempt to free temp prematurely: SV 0x%"UVxf
5922 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
5926 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
5927 /* make sure SvREFCNT(sv)==0 happens very seldom */
5928 SvREFCNT(sv) = (~(U32)0)/2;
5939 Returns the length of the string in the SV. Handles magic and type
5940 coercion. See also C<SvCUR>, which gives raw access to the xpv_cur slot.
5946 Perl_sv_len(pTHX_ register SV *const sv)
5954 len = mg_length(sv);
5956 (void)SvPV_const(sv, len);
5961 =for apidoc sv_len_utf8
5963 Returns the number of characters in the string in an SV, counting wide
5964 UTF-8 bytes as a single character. Handles magic and type coercion.
5970 * The length is cached in PERL_MAGIC_utf8, in the mg_len field. Also the
5971 * mg_ptr is used, by sv_pos_u2b() and sv_pos_b2u() - see the comments below.
5972 * (Note that the mg_len is not the length of the mg_ptr field.
5973 * This allows the cache to store the character length of the string without
5974 * needing to malloc() extra storage to attach to the mg_ptr.)
5979 Perl_sv_len_utf8(pTHX_ register SV *const sv)
5985 return mg_length(sv);
5989 const U8 *s = (U8*)SvPV_const(sv, len);
5993 MAGIC *mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_utf8) : NULL;
5995 if (mg && mg->mg_len != -1) {
5997 if (PL_utf8cache < 0) {
5998 const STRLEN real = Perl_utf8_length(aTHX_ s, s + len);
6000 /* Need to turn the assertions off otherwise we may
6001 recurse infinitely while printing error messages.
6003 SAVEI8(PL_utf8cache);
6005 Perl_croak(aTHX_ "panic: sv_len_utf8 cache %"UVuf
6006 " real %"UVuf" for %"SVf,
6007 (UV) ulen, (UV) real, SVfARG(sv));
6012 ulen = Perl_utf8_length(aTHX_ s, s + len);
6013 if (!SvREADONLY(sv)) {
6015 mg = sv_magicext(sv, 0, PERL_MAGIC_utf8,
6016 &PL_vtbl_utf8, 0, 0);
6024 return Perl_utf8_length(aTHX_ s, s + len);
6028 /* Walk forwards to find the byte corresponding to the passed in UTF-8
6031 S_sv_pos_u2b_forwards(const U8 *const start, const U8 *const send,
6034 const U8 *s = start;
6036 PERL_ARGS_ASSERT_SV_POS_U2B_FORWARDS;
6038 while (s < send && uoffset--)
6041 /* This is the existing behaviour. Possibly it should be a croak, as
6042 it's actually a bounds error */
6048 /* Given the length of the string in both bytes and UTF-8 characters, decide
6049 whether to walk forwards or backwards to find the byte corresponding to
6050 the passed in UTF-8 offset. */
6052 S_sv_pos_u2b_midway(const U8 *const start, const U8 *send,
6053 const STRLEN uoffset, const STRLEN uend)
6055 STRLEN backw = uend - uoffset;
6057 PERL_ARGS_ASSERT_SV_POS_U2B_MIDWAY;
6059 if (uoffset < 2 * backw) {
6060 /* The assumption is that going forwards is twice the speed of going
6061 forward (that's where the 2 * backw comes from).
6062 (The real figure of course depends on the UTF-8 data.) */
6063 return sv_pos_u2b_forwards(start, send, uoffset);
6068 while (UTF8_IS_CONTINUATION(*send))
6071 return send - start;
6074 /* For the string representation of the given scalar, find the byte
6075 corresponding to the passed in UTF-8 offset. uoffset0 and boffset0
6076 give another position in the string, *before* the sought offset, which
6077 (which is always true, as 0, 0 is a valid pair of positions), which should
6078 help reduce the amount of linear searching.
6079 If *mgp is non-NULL, it should point to the UTF-8 cache magic, which
6080 will be used to reduce the amount of linear searching. The cache will be
6081 created if necessary, and the found value offered to it for update. */
6083 S_sv_pos_u2b_cached(pTHX_ SV *const sv, MAGIC **const mgp, const U8 *const start,
6084 const U8 *const send, const STRLEN uoffset,
6085 STRLEN uoffset0, STRLEN boffset0)
6087 STRLEN boffset = 0; /* Actually always set, but let's keep gcc happy. */
6090 PERL_ARGS_ASSERT_SV_POS_U2B_CACHED;
6092 assert (uoffset >= uoffset0);
6094 if (SvMAGICAL(sv) && !SvREADONLY(sv) && PL_utf8cache
6095 && (*mgp || (*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
6096 if ((*mgp)->mg_ptr) {
6097 STRLEN *cache = (STRLEN *) (*mgp)->mg_ptr;
6098 if (cache[0] == uoffset) {
6099 /* An exact match. */
6102 if (cache[2] == uoffset) {
6103 /* An exact match. */
6107 if (cache[0] < uoffset) {
6108 /* The cache already knows part of the way. */
6109 if (cache[0] > uoffset0) {
6110 /* The cache knows more than the passed in pair */
6111 uoffset0 = cache[0];
6112 boffset0 = cache[1];
6114 if ((*mgp)->mg_len != -1) {
6115 /* And we know the end too. */
6117 + sv_pos_u2b_midway(start + boffset0, send,
6119 (*mgp)->mg_len - uoffset0);
6122 + sv_pos_u2b_forwards(start + boffset0,
6123 send, uoffset - uoffset0);
6126 else if (cache[2] < uoffset) {
6127 /* We're between the two cache entries. */
6128 if (cache[2] > uoffset0) {
6129 /* and the cache knows more than the passed in pair */
6130 uoffset0 = cache[2];
6131 boffset0 = cache[3];
6135 + sv_pos_u2b_midway(start + boffset0,
6138 cache[0] - uoffset0);
6141 + sv_pos_u2b_midway(start + boffset0,
6144 cache[2] - uoffset0);
6148 else if ((*mgp)->mg_len != -1) {
6149 /* If we can take advantage of a passed in offset, do so. */
6150 /* In fact, offset0 is either 0, or less than offset, so don't
6151 need to worry about the other possibility. */
6153 + sv_pos_u2b_midway(start + boffset0, send,
6155 (*mgp)->mg_len - uoffset0);
6160 if (!found || PL_utf8cache < 0) {
6161 const STRLEN real_boffset
6162 = boffset0 + sv_pos_u2b_forwards(start + boffset0,
6163 send, uoffset - uoffset0);
6165 if (found && PL_utf8cache < 0) {
6166 if (real_boffset != boffset) {
6167 /* Need to turn the assertions off otherwise we may recurse
6168 infinitely while printing error messages. */
6169 SAVEI8(PL_utf8cache);
6171 Perl_croak(aTHX_ "panic: sv_pos_u2b_cache cache %"UVuf
6172 " real %"UVuf" for %"SVf,
6173 (UV) boffset, (UV) real_boffset, SVfARG(sv));
6176 boffset = real_boffset;
6180 utf8_mg_pos_cache_update(sv, mgp, boffset, uoffset, send - start);
6186 =for apidoc sv_pos_u2b
6188 Converts the value pointed to by offsetp from a count of UTF-8 chars from
6189 the start of the string, to a count of the equivalent number of bytes; if
6190 lenp is non-zero, it does the same to lenp, but this time starting from
6191 the offset, rather than from the start of the string. Handles magic and
6198 * sv_pos_u2b() uses, like sv_pos_b2u(), the mg_ptr of the potential
6199 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6200 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
6205 Perl_sv_pos_u2b(pTHX_ register SV *const sv, I32 *const offsetp, I32 *const lenp)
6210 PERL_ARGS_ASSERT_SV_POS_U2B;
6215 start = (U8*)SvPV_const(sv, len);
6217 STRLEN uoffset = (STRLEN) *offsetp;
6218 const U8 * const send = start + len;
6220 const STRLEN boffset = sv_pos_u2b_cached(sv, &mg, start, send,
6223 *offsetp = (I32) boffset;
6226 /* Convert the relative offset to absolute. */
6227 const STRLEN uoffset2 = uoffset + (STRLEN) *lenp;
6228 const STRLEN boffset2
6229 = sv_pos_u2b_cached(sv, &mg, start, send, uoffset2,
6230 uoffset, boffset) - boffset;
6244 /* Create and update the UTF8 magic offset cache, with the proffered utf8/
6245 byte length pairing. The (byte) length of the total SV is passed in too,
6246 as blen, because for some (more esoteric) SVs, the call to SvPV_const()
6247 may not have updated SvCUR, so we can't rely on reading it directly.
6249 The proffered utf8/byte length pairing isn't used if the cache already has
6250 two pairs, and swapping either for the proffered pair would increase the
6251 RMS of the intervals between known byte offsets.
6253 The cache itself consists of 4 STRLEN values
6254 0: larger UTF-8 offset
6255 1: corresponding byte offset
6256 2: smaller UTF-8 offset
6257 3: corresponding byte offset
6259 Unused cache pairs have the value 0, 0.
6260 Keeping the cache "backwards" means that the invariant of
6261 cache[0] >= cache[2] is maintained even with empty slots, which means that
6262 the code that uses it doesn't need to worry if only 1 entry has actually
6263 been set to non-zero. It also makes the "position beyond the end of the
6264 cache" logic much simpler, as the first slot is always the one to start
6268 S_utf8_mg_pos_cache_update(pTHX_ SV *const sv, MAGIC **const mgp, const STRLEN byte,
6269 const STRLEN utf8, const STRLEN blen)
6273 PERL_ARGS_ASSERT_UTF8_MG_POS_CACHE_UPDATE;
6279 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, (MGVTBL*)&PL_vtbl_utf8, 0,
6281 (*mgp)->mg_len = -1;
6285 if (!(cache = (STRLEN *)(*mgp)->mg_ptr)) {
6286 Newxz(cache, PERL_MAGIC_UTF8_CACHESIZE * 2, STRLEN);
6287 (*mgp)->mg_ptr = (char *) cache;
6291 if (PL_utf8cache < 0) {
6292 const U8 *start = (const U8 *) SvPVX_const(sv);
6293 const STRLEN realutf8 = utf8_length(start, start + byte);
6295 if (realutf8 != utf8) {
6296 /* Need to turn the assertions off otherwise we may recurse
6297 infinitely while printing error messages. */
6298 SAVEI8(PL_utf8cache);
6300 Perl_croak(aTHX_ "panic: utf8_mg_pos_cache_update cache %"UVuf
6301 " real %"UVuf" for %"SVf, (UV) utf8, (UV) realutf8, SVfARG(sv));
6305 /* Cache is held with the later position first, to simplify the code
6306 that deals with unbounded ends. */
6308 ASSERT_UTF8_CACHE(cache);
6309 if (cache[1] == 0) {
6310 /* Cache is totally empty */
6313 } else if (cache[3] == 0) {
6314 if (byte > cache[1]) {
6315 /* New one is larger, so goes first. */
6316 cache[2] = cache[0];
6317 cache[3] = cache[1];
6325 #define THREEWAY_SQUARE(a,b,c,d) \
6326 ((float)((d) - (c))) * ((float)((d) - (c))) \
6327 + ((float)((c) - (b))) * ((float)((c) - (b))) \
6328 + ((float)((b) - (a))) * ((float)((b) - (a)))
6330 /* Cache has 2 slots in use, and we know three potential pairs.
6331 Keep the two that give the lowest RMS distance. Do the
6332 calcualation in bytes simply because we always know the byte
6333 length. squareroot has the same ordering as the positive value,
6334 so don't bother with the actual square root. */
6335 const float existing = THREEWAY_SQUARE(0, cache[3], cache[1], blen);
6336 if (byte > cache[1]) {
6337 /* New position is after the existing pair of pairs. */
6338 const float keep_earlier
6339 = THREEWAY_SQUARE(0, cache[3], byte, blen);
6340 const float keep_later
6341 = THREEWAY_SQUARE(0, cache[1], byte, blen);
6343 if (keep_later < keep_earlier) {
6344 if (keep_later < existing) {
6345 cache[2] = cache[0];
6346 cache[3] = cache[1];
6352 if (keep_earlier < existing) {
6358 else if (byte > cache[3]) {
6359 /* New position is between the existing pair of pairs. */
6360 const float keep_earlier
6361 = THREEWAY_SQUARE(0, cache[3], byte, blen);
6362 const float keep_later
6363 = THREEWAY_SQUARE(0, byte, cache[1], blen);
6365 if (keep_later < keep_earlier) {
6366 if (keep_later < existing) {
6372 if (keep_earlier < existing) {
6379 /* New position is before the existing pair of pairs. */
6380 const float keep_earlier
6381 = THREEWAY_SQUARE(0, byte, cache[3], blen);
6382 const float keep_later
6383 = THREEWAY_SQUARE(0, byte, cache[1], blen);
6385 if (keep_later < keep_earlier) {
6386 if (keep_later < existing) {
6392 if (keep_earlier < existing) {
6393 cache[0] = cache[2];
6394 cache[1] = cache[3];
6401 ASSERT_UTF8_CACHE(cache);
6404 /* We already know all of the way, now we may be able to walk back. The same
6405 assumption is made as in S_sv_pos_u2b_midway(), namely that walking
6406 backward is half the speed of walking forward. */
6408 S_sv_pos_b2u_midway(pTHX_ const U8 *const s, const U8 *const target,
6409 const U8 *end, STRLEN endu)
6411 const STRLEN forw = target - s;
6412 STRLEN backw = end - target;
6414 PERL_ARGS_ASSERT_SV_POS_B2U_MIDWAY;
6416 if (forw < 2 * backw) {
6417 return utf8_length(s, target);
6420 while (end > target) {
6422 while (UTF8_IS_CONTINUATION(*end)) {
6431 =for apidoc sv_pos_b2u
6433 Converts the value pointed to by offsetp from a count of bytes from the
6434 start of the string, to a count of the equivalent number of UTF-8 chars.
6435 Handles magic and type coercion.
6441 * sv_pos_b2u() uses, like sv_pos_u2b(), the mg_ptr of the potential
6442 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6447 Perl_sv_pos_b2u(pTHX_ register SV *const sv, I32 *const offsetp)
6450 const STRLEN byte = *offsetp;
6451 STRLEN len = 0; /* Actually always set, but let's keep gcc happy. */
6457 PERL_ARGS_ASSERT_SV_POS_B2U;
6462 s = (const U8*)SvPV_const(sv, blen);
6465 Perl_croak(aTHX_ "panic: sv_pos_b2u: bad byte offset");
6469 if (SvMAGICAL(sv) && !SvREADONLY(sv) && PL_utf8cache
6470 && (mg = mg_find(sv, PERL_MAGIC_utf8))) {
6472 STRLEN * const cache = (STRLEN *) mg->mg_ptr;
6473 if (cache[1] == byte) {
6474 /* An exact match. */
6475 *offsetp = cache[0];
6478 if (cache[3] == byte) {
6479 /* An exact match. */
6480 *offsetp = cache[2];
6484 if (cache[1] < byte) {
6485 /* We already know part of the way. */
6486 if (mg->mg_len != -1) {
6487 /* Actually, we know the end too. */
6489 + S_sv_pos_b2u_midway(aTHX_ s + cache[1], send,
6490 s + blen, mg->mg_len - cache[0]);
6492 len = cache[0] + utf8_length(s + cache[1], send);
6495 else if (cache[3] < byte) {
6496 /* We're between the two cached pairs, so we do the calculation
6497 offset by the byte/utf-8 positions for the earlier pair,
6498 then add the utf-8 characters from the string start to
6500 len = S_sv_pos_b2u_midway(aTHX_ s + cache[3], send,
6501 s + cache[1], cache[0] - cache[2])
6505 else { /* cache[3] > byte */
6506 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + cache[3],
6510 ASSERT_UTF8_CACHE(cache);
6512 } else if (mg->mg_len != -1) {
6513 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + blen, mg->mg_len);
6517 if (!found || PL_utf8cache < 0) {
6518 const STRLEN real_len = utf8_length(s, send);
6520 if (found && PL_utf8cache < 0) {
6521 if (len != real_len) {
6522 /* Need to turn the assertions off otherwise we may recurse
6523 infinitely while printing error messages. */
6524 SAVEI8(PL_utf8cache);
6526 Perl_croak(aTHX_ "panic: sv_pos_b2u cache %"UVuf
6527 " real %"UVuf" for %"SVf,
6528 (UV) len, (UV) real_len, SVfARG(sv));
6536 utf8_mg_pos_cache_update(sv, &mg, byte, len, blen);
6542 Returns a boolean indicating whether the strings in the two SVs are
6543 identical. Is UTF-8 and 'use bytes' aware, handles get magic, and will
6544 coerce its args to strings if necessary.
6550 Perl_sv_eq(pTHX_ register SV *sv1, register SV *sv2)
6559 SV* svrecode = NULL;
6566 /* if pv1 and pv2 are the same, second SvPV_const call may
6567 * invalidate pv1, so we may need to make a copy */
6568 if (sv1 == sv2 && (SvTHINKFIRST(sv1) || SvGMAGICAL(sv1))) {
6569 pv1 = SvPV_const(sv1, cur1);
6570 sv1 = newSVpvn_flags(pv1, cur1, SVs_TEMP | SvUTF8(sv2));
6572 pv1 = SvPV_const(sv1, cur1);
6580 pv2 = SvPV_const(sv2, cur2);
6582 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
6583 /* Differing utf8ness.
6584 * Do not UTF8size the comparands as a side-effect. */
6587 svrecode = newSVpvn(pv2, cur2);
6588 sv_recode_to_utf8(svrecode, PL_encoding);
6589 pv2 = SvPV_const(svrecode, cur2);
6592 svrecode = newSVpvn(pv1, cur1);
6593 sv_recode_to_utf8(svrecode, PL_encoding);
6594 pv1 = SvPV_const(svrecode, cur1);
6596 /* Now both are in UTF-8. */
6598 SvREFCNT_dec(svrecode);
6603 bool is_utf8 = TRUE;
6606 /* sv1 is the UTF-8 one,
6607 * if is equal it must be downgrade-able */
6608 char * const pv = (char*)bytes_from_utf8((const U8*)pv1,
6614 /* sv2 is the UTF-8 one,
6615 * if is equal it must be downgrade-able */
6616 char * const pv = (char *)bytes_from_utf8((const U8*)pv2,
6622 /* Downgrade not possible - cannot be eq */
6630 eq = (pv1 == pv2) || memEQ(pv1, pv2, cur1);
6632 SvREFCNT_dec(svrecode);
6642 Compares the strings in two SVs. Returns -1, 0, or 1 indicating whether the
6643 string in C<sv1> is less than, equal to, or greater than the string in
6644 C<sv2>. Is UTF-8 and 'use bytes' aware, handles get magic, and will
6645 coerce its args to strings if necessary. See also C<sv_cmp_locale>.
6651 Perl_sv_cmp(pTHX_ register SV *const sv1, register SV *const sv2)
6655 const char *pv1, *pv2;
6658 SV *svrecode = NULL;
6665 pv1 = SvPV_const(sv1, cur1);
6672 pv2 = SvPV_const(sv2, cur2);
6674 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
6675 /* Differing utf8ness.
6676 * Do not UTF8size the comparands as a side-effect. */
6679 svrecode = newSVpvn(pv2, cur2);
6680 sv_recode_to_utf8(svrecode, PL_encoding);
6681 pv2 = SvPV_const(svrecode, cur2);
6684 pv2 = tpv = (char*)bytes_to_utf8((const U8*)pv2, &cur2);
6689 svrecode = newSVpvn(pv1, cur1);
6690 sv_recode_to_utf8(svrecode, PL_encoding);
6691 pv1 = SvPV_const(svrecode, cur1);
6694 pv1 = tpv = (char*)bytes_to_utf8((const U8*)pv1, &cur1);
6700 cmp = cur2 ? -1 : 0;
6704 const I32 retval = memcmp((const void*)pv1, (const void*)pv2, cur1 < cur2 ? cur1 : cur2);
6707 cmp = retval < 0 ? -1 : 1;
6708 } else if (cur1 == cur2) {
6711 cmp = cur1 < cur2 ? -1 : 1;
6715 SvREFCNT_dec(svrecode);
6723 =for apidoc sv_cmp_locale
6725 Compares the strings in two SVs in a locale-aware manner. Is UTF-8 and
6726 'use bytes' aware, handles get magic, and will coerce its args to strings
6727 if necessary. See also C<sv_cmp>.
6733 Perl_sv_cmp_locale(pTHX_ register SV *const sv1, register SV *const sv2)
6736 #ifdef USE_LOCALE_COLLATE
6742 if (PL_collation_standard)
6746 pv1 = sv1 ? sv_collxfrm(sv1, &len1) : (char *) NULL;
6748 pv2 = sv2 ? sv_collxfrm(sv2, &len2) : (char *) NULL;
6750 if (!pv1 || !len1) {
6761 retval = memcmp((void*)pv1, (void*)pv2, len1 < len2 ? len1 : len2);
6764 return retval < 0 ? -1 : 1;
6767 * When the result of collation is equality, that doesn't mean
6768 * that there are no differences -- some locales exclude some
6769 * characters from consideration. So to avoid false equalities,
6770 * we use the raw string as a tiebreaker.
6776 #endif /* USE_LOCALE_COLLATE */
6778 return sv_cmp(sv1, sv2);
6782 #ifdef USE_LOCALE_COLLATE
6785 =for apidoc sv_collxfrm
6787 Add Collate Transform magic to an SV if it doesn't already have it.
6789 Any scalar variable may carry PERL_MAGIC_collxfrm magic that contains the
6790 scalar data of the variable, but transformed to such a format that a normal
6791 memory comparison can be used to compare the data according to the locale
6798 Perl_sv_collxfrm(pTHX_ SV *const sv, STRLEN *const nxp)
6803 PERL_ARGS_ASSERT_SV_COLLXFRM;
6805 mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_collxfrm) : (MAGIC *) NULL;
6806 if (!mg || !mg->mg_ptr || *(U32*)mg->mg_ptr != PL_collation_ix) {
6812 Safefree(mg->mg_ptr);
6813 s = SvPV_const(sv, len);
6814 if ((xf = mem_collxfrm(s, len, &xlen))) {
6816 #ifdef PERL_OLD_COPY_ON_WRITE
6818 sv_force_normal_flags(sv, 0);
6820 mg = sv_magicext(sv, 0, PERL_MAGIC_collxfrm, &PL_vtbl_collxfrm,
6834 if (mg && mg->mg_ptr) {
6836 return mg->mg_ptr + sizeof(PL_collation_ix);
6844 #endif /* USE_LOCALE_COLLATE */
6849 Get a line from the filehandle and store it into the SV, optionally
6850 appending to the currently-stored string.
6856 Perl_sv_gets(pTHX_ register SV *const sv, register PerlIO *const fp, I32 append)
6861 register STDCHAR rslast;
6862 register STDCHAR *bp;
6867 PERL_ARGS_ASSERT_SV_GETS;
6869 if (SvTHINKFIRST(sv))
6870 sv_force_normal_flags(sv, append ? 0 : SV_COW_DROP_PV);
6871 /* XXX. If you make this PVIV, then copy on write can copy scalars read
6873 However, perlbench says it's slower, because the existing swipe code
6874 is faster than copy on write.
6875 Swings and roundabouts. */
6876 SvUPGRADE(sv, SVt_PV);
6881 if (PerlIO_isutf8(fp)) {
6883 sv_utf8_upgrade_nomg(sv);
6884 sv_pos_u2b(sv,&append,0);
6886 } else if (SvUTF8(sv)) {
6887 SV * const tsv = newSV(0);
6888 sv_gets(tsv, fp, 0);
6889 sv_utf8_upgrade_nomg(tsv);
6890 SvCUR_set(sv,append);
6893 goto return_string_or_null;
6898 if (PerlIO_isutf8(fp))
6901 if (IN_PERL_COMPILETIME) {
6902 /* we always read code in line mode */
6906 else if (RsSNARF(PL_rs)) {
6907 /* If it is a regular disk file use size from stat() as estimate
6908 of amount we are going to read -- may result in mallocing
6909 more memory than we really need if the layers below reduce
6910 the size we read (e.g. CRLF or a gzip layer).
6913 if (!PerlLIO_fstat(PerlIO_fileno(fp), &st) && S_ISREG(st.st_mode)) {
6914 const Off_t offset = PerlIO_tell(fp);
6915 if (offset != (Off_t) -1 && st.st_size + append > offset) {
6916 (void) SvGROW(sv, (STRLEN)((st.st_size - offset) + append + 1));
6922 else if (RsRECORD(PL_rs)) {
6930 /* Grab the size of the record we're getting */
6931 recsize = SvUV(SvRV(PL_rs)); /* RsRECORD() guarantees > 0. */
6932 buffer = SvGROW(sv, (STRLEN)(recsize + append + 1)) + append;
6935 /* VMS wants read instead of fread, because fread doesn't respect */
6936 /* RMS record boundaries. This is not necessarily a good thing to be */
6937 /* doing, but we've got no other real choice - except avoid stdio
6938 as implementation - perhaps write a :vms layer ?
6940 fd = PerlIO_fileno(fp);
6941 if (fd == -1) { /* in-memory file from PerlIO::Scalar */
6942 bytesread = PerlIO_read(fp, buffer, recsize);
6945 bytesread = PerlLIO_read(fd, buffer, recsize);
6948 bytesread = PerlIO_read(fp, buffer, recsize);
6952 SvCUR_set(sv, bytesread + append);
6953 buffer[bytesread] = '\0';
6954 goto return_string_or_null;
6956 else if (RsPARA(PL_rs)) {
6962 /* Get $/ i.e. PL_rs into same encoding as stream wants */
6963 if (PerlIO_isutf8(fp)) {
6964 rsptr = SvPVutf8(PL_rs, rslen);
6967 if (SvUTF8(PL_rs)) {
6968 if (!sv_utf8_downgrade(PL_rs, TRUE)) {
6969 Perl_croak(aTHX_ "Wide character in $/");
6972 rsptr = SvPV_const(PL_rs, rslen);
6976 rslast = rslen ? rsptr[rslen - 1] : '\0';
6978 if (rspara) { /* have to do this both before and after */
6979 do { /* to make sure file boundaries work right */
6982 i = PerlIO_getc(fp);
6986 PerlIO_ungetc(fp,i);
6992 /* See if we know enough about I/O mechanism to cheat it ! */
6994 /* This used to be #ifdef test - it is made run-time test for ease
6995 of abstracting out stdio interface. One call should be cheap
6996 enough here - and may even be a macro allowing compile
7000 if (PerlIO_fast_gets(fp)) {
7003 * We're going to steal some values from the stdio struct
7004 * and put EVERYTHING in the innermost loop into registers.
7006 register STDCHAR *ptr;
7010 #if defined(VMS) && defined(PERLIO_IS_STDIO)
7011 /* An ungetc()d char is handled separately from the regular
7012 * buffer, so we getc() it back out and stuff it in the buffer.
7014 i = PerlIO_getc(fp);
7015 if (i == EOF) return 0;
7016 *(--((*fp)->_ptr)) = (unsigned char) i;
7020 /* Here is some breathtakingly efficient cheating */
7022 cnt = PerlIO_get_cnt(fp); /* get count into register */
7023 /* make sure we have the room */
7024 if ((I32)(SvLEN(sv) - append) <= cnt + 1) {
7025 /* Not room for all of it
7026 if we are looking for a separator and room for some
7028 if (rslen && cnt > 80 && (I32)SvLEN(sv) > append) {
7029 /* just process what we have room for */
7030 shortbuffered = cnt - SvLEN(sv) + append + 1;
7031 cnt -= shortbuffered;
7035 /* remember that cnt can be negative */
7036 SvGROW(sv, (STRLEN)(append + (cnt <= 0 ? 2 : (cnt + 1))));
7041 bp = (STDCHAR*)SvPVX_const(sv) + append; /* move these two too to registers */
7042 ptr = (STDCHAR*)PerlIO_get_ptr(fp);
7043 DEBUG_P(PerlIO_printf(Perl_debug_log,
7044 "Screamer: entering, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7045 DEBUG_P(PerlIO_printf(Perl_debug_log,
7046 "Screamer: entering: PerlIO * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7047 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7048 PTR2UV(PerlIO_has_base(fp) ? PerlIO_get_base(fp) : 0)));
7053 while (cnt > 0) { /* this | eat */
7055 if ((*bp++ = *ptr++) == rslast) /* really | dust */
7056 goto thats_all_folks; /* screams | sed :-) */
7060 Copy(ptr, bp, cnt, char); /* this | eat */
7061 bp += cnt; /* screams | dust */
7062 ptr += cnt; /* louder | sed :-) */
7067 if (shortbuffered) { /* oh well, must extend */
7068 cnt = shortbuffered;
7070 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
7072 SvGROW(sv, SvLEN(sv) + append + cnt + 2);
7073 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
7077 DEBUG_P(PerlIO_printf(Perl_debug_log,
7078 "Screamer: going to getc, ptr=%"UVuf", cnt=%ld\n",
7079 PTR2UV(ptr),(long)cnt));
7080 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* deregisterize cnt and ptr */
7082 DEBUG_P(PerlIO_printf(Perl_debug_log,
7083 "Screamer: pre: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7084 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7085 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7087 /* This used to call 'filbuf' in stdio form, but as that behaves like
7088 getc when cnt <= 0 we use PerlIO_getc here to avoid introducing
7089 another abstraction. */
7090 i = PerlIO_getc(fp); /* get more characters */
7092 DEBUG_P(PerlIO_printf(Perl_debug_log,
7093 "Screamer: post: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7094 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7095 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7097 cnt = PerlIO_get_cnt(fp);
7098 ptr = (STDCHAR*)PerlIO_get_ptr(fp); /* reregisterize cnt and ptr */
7099 DEBUG_P(PerlIO_printf(Perl_debug_log,
7100 "Screamer: after getc, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7102 if (i == EOF) /* all done for ever? */
7103 goto thats_really_all_folks;
7105 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
7107 SvGROW(sv, bpx + cnt + 2);
7108 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
7110 *bp++ = (STDCHAR)i; /* store character from PerlIO_getc */
7112 if (rslen && (STDCHAR)i == rslast) /* all done for now? */
7113 goto thats_all_folks;
7117 if ((rslen > 1 && (STRLEN)(bp - (STDCHAR*)SvPVX_const(sv)) < rslen) ||
7118 memNE((char*)bp - rslen, rsptr, rslen))
7119 goto screamer; /* go back to the fray */
7120 thats_really_all_folks:
7122 cnt += shortbuffered;
7123 DEBUG_P(PerlIO_printf(Perl_debug_log,
7124 "Screamer: quitting, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7125 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* put these back or we're in trouble */
7126 DEBUG_P(PerlIO_printf(Perl_debug_log,
7127 "Screamer: end: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7128 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7129 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7131 SvCUR_set(sv, bp - (STDCHAR*)SvPVX_const(sv)); /* set length */
7132 DEBUG_P(PerlIO_printf(Perl_debug_log,
7133 "Screamer: done, len=%ld, string=|%.*s|\n",
7134 (long)SvCUR(sv),(int)SvCUR(sv),SvPVX_const(sv)));
7138 /*The big, slow, and stupid way. */
7139 #ifdef USE_HEAP_INSTEAD_OF_STACK /* Even slower way. */
7140 STDCHAR *buf = NULL;
7141 Newx(buf, 8192, STDCHAR);
7149 register const STDCHAR * const bpe = buf + sizeof(buf);
7151 while ((i = PerlIO_getc(fp)) != EOF && (*bp++ = (STDCHAR)i) != rslast && bp < bpe)
7152 ; /* keep reading */
7156 cnt = PerlIO_read(fp,(char*)buf, sizeof(buf));
7157 /* Accomodate broken VAXC compiler, which applies U8 cast to
7158 * both args of ?: operator, causing EOF to change into 255
7161 i = (U8)buf[cnt - 1];
7167 cnt = 0; /* we do need to re-set the sv even when cnt <= 0 */
7169 sv_catpvn(sv, (char *) buf, cnt);
7171 sv_setpvn(sv, (char *) buf, cnt);
7173 if (i != EOF && /* joy */
7175 SvCUR(sv) < rslen ||
7176 memNE(SvPVX_const(sv) + SvCUR(sv) - rslen, rsptr, rslen)))
7180 * If we're reading from a TTY and we get a short read,
7181 * indicating that the user hit his EOF character, we need
7182 * to notice it now, because if we try to read from the TTY
7183 * again, the EOF condition will disappear.
7185 * The comparison of cnt to sizeof(buf) is an optimization
7186 * that prevents unnecessary calls to feof().
7190 if (!(cnt < (I32)sizeof(buf) && PerlIO_eof(fp)))
7194 #ifdef USE_HEAP_INSTEAD_OF_STACK
7199 if (rspara) { /* have to do this both before and after */
7200 while (i != EOF) { /* to make sure file boundaries work right */
7201 i = PerlIO_getc(fp);
7203 PerlIO_ungetc(fp,i);
7209 return_string_or_null:
7210 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7216 Auto-increment of the value in the SV, doing string to numeric conversion
7217 if necessary. Handles 'get' magic.
7223 Perl_sv_inc(pTHX_ register SV *const sv)
7232 if (SvTHINKFIRST(sv)) {
7234 sv_force_normal_flags(sv, 0);
7235 if (SvREADONLY(sv)) {
7236 if (IN_PERL_RUNTIME)
7237 Perl_croak(aTHX_ "%s", PL_no_modify);
7241 if (SvAMAGIC(sv) && AMG_CALLun(sv,inc))
7243 i = PTR2IV(SvRV(sv));
7248 flags = SvFLAGS(sv);
7249 if ((flags & (SVp_NOK|SVp_IOK)) == SVp_NOK) {
7250 /* It's (privately or publicly) a float, but not tested as an
7251 integer, so test it to see. */
7253 flags = SvFLAGS(sv);
7255 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
7256 /* It's publicly an integer, or privately an integer-not-float */
7257 #ifdef PERL_PRESERVE_IVUV
7261 if (SvUVX(sv) == UV_MAX)
7262 sv_setnv(sv, UV_MAX_P1);
7264 (void)SvIOK_only_UV(sv);
7265 SvUV_set(sv, SvUVX(sv) + 1);
7267 if (SvIVX(sv) == IV_MAX)
7268 sv_setuv(sv, (UV)IV_MAX + 1);
7270 (void)SvIOK_only(sv);
7271 SvIV_set(sv, SvIVX(sv) + 1);
7276 if (flags & SVp_NOK) {
7277 const NV was = SvNVX(sv);
7278 if (NV_OVERFLOWS_INTEGERS_AT &&
7279 was >= NV_OVERFLOWS_INTEGERS_AT && ckWARN(WARN_IMPRECISION)) {
7280 Perl_warner(aTHX_ packWARN(WARN_IMPRECISION),
7281 "Lost precision when incrementing %" NVff " by 1",
7284 (void)SvNOK_only(sv);
7285 SvNV_set(sv, was + 1.0);
7289 if (!(flags & SVp_POK) || !*SvPVX_const(sv)) {
7290 if ((flags & SVTYPEMASK) < SVt_PVIV)
7291 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV ? SVt_PVIV : SVt_IV));
7292 (void)SvIOK_only(sv);
7297 while (isALPHA(*d)) d++;
7298 while (isDIGIT(*d)) d++;
7300 #ifdef PERL_PRESERVE_IVUV
7301 /* Got to punt this as an integer if needs be, but we don't issue
7302 warnings. Probably ought to make the sv_iv_please() that does
7303 the conversion if possible, and silently. */
7304 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
7305 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
7306 /* Need to try really hard to see if it's an integer.
7307 9.22337203685478e+18 is an integer.
7308 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
7309 so $a="9.22337203685478e+18"; $a+0; $a++
7310 needs to be the same as $a="9.22337203685478e+18"; $a++
7317 /* sv_2iv *should* have made this an NV */
7318 if (flags & SVp_NOK) {
7319 (void)SvNOK_only(sv);
7320 SvNV_set(sv, SvNVX(sv) + 1.0);
7323 /* I don't think we can get here. Maybe I should assert this
7324 And if we do get here I suspect that sv_setnv will croak. NWC
7326 #if defined(USE_LONG_DOUBLE)
7327 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",
7328 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7330 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_inc punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
7331 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7334 #endif /* PERL_PRESERVE_IVUV */
7335 sv_setnv(sv,Atof(SvPVX_const(sv)) + 1.0);
7339 while (d >= SvPVX_const(sv)) {
7347 /* MKS: The original code here died if letters weren't consecutive.
7348 * at least it didn't have to worry about non-C locales. The
7349 * new code assumes that ('z'-'a')==('Z'-'A'), letters are
7350 * arranged in order (although not consecutively) and that only
7351 * [A-Za-z] are accepted by isALPHA in the C locale.
7353 if (*d != 'z' && *d != 'Z') {
7354 do { ++*d; } while (!isALPHA(*d));
7357 *(d--) -= 'z' - 'a';
7362 *(d--) -= 'z' - 'a' + 1;
7366 /* oh,oh, the number grew */
7367 SvGROW(sv, SvCUR(sv) + 2);
7368 SvCUR_set(sv, SvCUR(sv) + 1);
7369 for (d = SvPVX(sv) + SvCUR(sv); d > SvPVX_const(sv); d--)
7380 Auto-decrement of the value in the SV, doing string to numeric conversion
7381 if necessary. Handles 'get' magic.
7387 Perl_sv_dec(pTHX_ register SV *const sv)
7395 if (SvTHINKFIRST(sv)) {
7397 sv_force_normal_flags(sv, 0);
7398 if (SvREADONLY(sv)) {
7399 if (IN_PERL_RUNTIME)
7400 Perl_croak(aTHX_ "%s", PL_no_modify);
7404 if (SvAMAGIC(sv) && AMG_CALLun(sv,dec))
7406 i = PTR2IV(SvRV(sv));
7411 /* Unlike sv_inc we don't have to worry about string-never-numbers
7412 and keeping them magic. But we mustn't warn on punting */
7413 flags = SvFLAGS(sv);
7414 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
7415 /* It's publicly an integer, or privately an integer-not-float */
7416 #ifdef PERL_PRESERVE_IVUV
7420 if (SvUVX(sv) == 0) {
7421 (void)SvIOK_only(sv);
7425 (void)SvIOK_only_UV(sv);
7426 SvUV_set(sv, SvUVX(sv) - 1);
7429 if (SvIVX(sv) == IV_MIN) {
7430 sv_setnv(sv, (NV)IV_MIN);
7434 (void)SvIOK_only(sv);
7435 SvIV_set(sv, SvIVX(sv) - 1);
7440 if (flags & SVp_NOK) {
7443 const NV was = SvNVX(sv);
7444 if (NV_OVERFLOWS_INTEGERS_AT &&
7445 was <= -NV_OVERFLOWS_INTEGERS_AT && ckWARN(WARN_IMPRECISION)) {
7446 Perl_warner(aTHX_ packWARN(WARN_IMPRECISION),
7447 "Lost precision when decrementing %" NVff " by 1",
7450 (void)SvNOK_only(sv);
7451 SvNV_set(sv, was - 1.0);
7455 if (!(flags & SVp_POK)) {
7456 if ((flags & SVTYPEMASK) < SVt_PVIV)
7457 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV) ? SVt_PVIV : SVt_IV);
7459 (void)SvIOK_only(sv);
7462 #ifdef PERL_PRESERVE_IVUV
7464 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
7465 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
7466 /* Need to try really hard to see if it's an integer.
7467 9.22337203685478e+18 is an integer.
7468 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
7469 so $a="9.22337203685478e+18"; $a+0; $a--
7470 needs to be the same as $a="9.22337203685478e+18"; $a--
7477 /* sv_2iv *should* have made this an NV */
7478 if (flags & SVp_NOK) {
7479 (void)SvNOK_only(sv);
7480 SvNV_set(sv, SvNVX(sv) - 1.0);
7483 /* I don't think we can get here. Maybe I should assert this
7484 And if we do get here I suspect that sv_setnv will croak. NWC
7486 #if defined(USE_LONG_DOUBLE)
7487 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",
7488 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7490 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_dec punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
7491 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7495 #endif /* PERL_PRESERVE_IVUV */
7496 sv_setnv(sv,Atof(SvPVX_const(sv)) - 1.0); /* punt */
7500 =for apidoc sv_mortalcopy
7502 Creates a new SV which is a copy of the original SV (using C<sv_setsv>).
7503 The new SV is marked as mortal. It will be destroyed "soon", either by an
7504 explicit call to FREETMPS, or by an implicit call at places such as
7505 statement boundaries. See also C<sv_newmortal> and C<sv_2mortal>.
7510 /* Make a string that will exist for the duration of the expression
7511 * evaluation. Actually, it may have to last longer than that, but
7512 * hopefully we won't free it until it has been assigned to a
7513 * permanent location. */
7516 Perl_sv_mortalcopy(pTHX_ SV *const oldstr)
7522 sv_setsv(sv,oldstr);
7524 PL_tmps_stack[++PL_tmps_ix] = sv;
7530 =for apidoc sv_newmortal
7532 Creates a new null SV which is mortal. The reference count of the SV is
7533 set to 1. It will be destroyed "soon", either by an explicit call to
7534 FREETMPS, or by an implicit call at places such as statement boundaries.
7535 See also C<sv_mortalcopy> and C<sv_2mortal>.
7541 Perl_sv_newmortal(pTHX)
7547 SvFLAGS(sv) = SVs_TEMP;
7549 PL_tmps_stack[++PL_tmps_ix] = sv;
7555 =for apidoc newSVpvn_flags
7557 Creates a new SV and copies a string into it. The reference count for the
7558 SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
7559 string. You are responsible for ensuring that the source string is at least
7560 C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
7561 Currently the only flag bits accepted are C<SVf_UTF8> and C<SVs_TEMP>.
7562 If C<SVs_TEMP> is set, then C<sv2mortal()> is called on the result before
7563 returning. If C<SVf_UTF8> is set, then it will be set on the new SV.
7564 C<newSVpvn_utf8()> is a convenience wrapper for this function, defined as
7566 #define newSVpvn_utf8(s, len, u) \
7567 newSVpvn_flags((s), (len), (u) ? SVf_UTF8 : 0)
7573 Perl_newSVpvn_flags(pTHX_ const char *const s, const STRLEN len, const U32 flags)
7578 /* All the flags we don't support must be zero.
7579 And we're new code so I'm going to assert this from the start. */
7580 assert(!(flags & ~(SVf_UTF8|SVs_TEMP)));
7582 sv_setpvn(sv,s,len);
7583 SvFLAGS(sv) |= (flags & SVf_UTF8);
7584 return (flags & SVs_TEMP) ? sv_2mortal(sv) : sv;
7588 =for apidoc sv_2mortal
7590 Marks an existing SV as mortal. The SV will be destroyed "soon", either
7591 by an explicit call to FREETMPS, or by an implicit call at places such as
7592 statement boundaries. SvTEMP() is turned on which means that the SV's
7593 string buffer can be "stolen" if this SV is copied. See also C<sv_newmortal>
7594 and C<sv_mortalcopy>.
7600 Perl_sv_2mortal(pTHX_ register SV *const sv)
7605 if (SvREADONLY(sv) && SvIMMORTAL(sv))
7608 PL_tmps_stack[++PL_tmps_ix] = sv;
7616 Creates a new SV and copies a string into it. The reference count for the
7617 SV is set to 1. If C<len> is zero, Perl will compute the length using
7618 strlen(). For efficiency, consider using C<newSVpvn> instead.
7624 Perl_newSVpv(pTHX_ const char *const s, const STRLEN len)
7630 sv_setpvn(sv, s, len || s == NULL ? len : strlen(s));
7635 =for apidoc newSVpvn
7637 Creates a new SV and copies a string into it. The reference count for the
7638 SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
7639 string. You are responsible for ensuring that the source string is at least
7640 C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
7646 Perl_newSVpvn(pTHX_ const char *const s, const STRLEN len)
7652 sv_setpvn(sv,s,len);
7657 =for apidoc newSVhek
7659 Creates a new SV from the hash key structure. It will generate scalars that
7660 point to the shared string table where possible. Returns a new (undefined)
7661 SV if the hek is NULL.
7667 Perl_newSVhek(pTHX_ const HEK *const hek)
7677 if (HEK_LEN(hek) == HEf_SVKEY) {
7678 return newSVsv(*(SV**)HEK_KEY(hek));
7680 const int flags = HEK_FLAGS(hek);
7681 if (flags & HVhek_WASUTF8) {
7683 Andreas would like keys he put in as utf8 to come back as utf8
7685 STRLEN utf8_len = HEK_LEN(hek);
7686 const U8 *as_utf8 = bytes_to_utf8 ((U8*)HEK_KEY(hek), &utf8_len);
7687 SV * const sv = newSVpvn ((const char*)as_utf8, utf8_len);
7690 Safefree (as_utf8); /* bytes_to_utf8() allocates a new string */
7692 } else if (flags & (HVhek_REHASH|HVhek_UNSHARED)) {
7693 /* We don't have a pointer to the hv, so we have to replicate the
7694 flag into every HEK. This hv is using custom a hasing
7695 algorithm. Hence we can't return a shared string scalar, as
7696 that would contain the (wrong) hash value, and might get passed
7697 into an hv routine with a regular hash.
7698 Similarly, a hash that isn't using shared hash keys has to have
7699 the flag in every key so that we know not to try to call
7700 share_hek_kek on it. */
7702 SV * const sv = newSVpvn (HEK_KEY(hek), HEK_LEN(hek));
7707 /* This will be overwhelminly the most common case. */
7709 /* Inline most of newSVpvn_share(), because share_hek_hek() is far
7710 more efficient than sharepvn(). */
7714 sv_upgrade(sv, SVt_PV);
7715 SvPV_set(sv, (char *)HEK_KEY(share_hek_hek(hek)));
7716 SvCUR_set(sv, HEK_LEN(hek));
7729 =for apidoc newSVpvn_share
7731 Creates a new SV with its SvPVX_const pointing to a shared string in the string
7732 table. If the string does not already exist in the table, it is created
7733 first. Turns on READONLY and FAKE. If the C<hash> parameter is non-zero, that
7734 value is used; otherwise the hash is computed. The string's hash can be later
7735 be retrieved from the SV with the C<SvSHARED_HASH()> macro. The idea here is
7736 that as the string table is used for shared hash keys these strings will have
7737 SvPVX_const == HeKEY and hash lookup will avoid string compare.
7743 Perl_newSVpvn_share(pTHX_ const char *src, I32 len, U32 hash)
7747 bool is_utf8 = FALSE;
7748 const char *const orig_src = src;
7751 STRLEN tmplen = -len;
7753 /* See the note in hv.c:hv_fetch() --jhi */
7754 src = (char*)bytes_from_utf8((const U8*)src, &tmplen, &is_utf8);
7758 PERL_HASH(hash, src, len);
7760 /* The logic for this is inlined in S_mro_get_linear_isa_dfs(), so if it
7761 changes here, update it there too. */
7762 sv_upgrade(sv, SVt_PV);
7763 SvPV_set(sv, sharepvn(src, is_utf8?-len:len, hash));
7771 if (src != orig_src)
7777 #if defined(PERL_IMPLICIT_CONTEXT)
7779 /* pTHX_ magic can't cope with varargs, so this is a no-context
7780 * version of the main function, (which may itself be aliased to us).
7781 * Don't access this version directly.
7785 Perl_newSVpvf_nocontext(const char *const pat, ...)
7791 PERL_ARGS_ASSERT_NEWSVPVF_NOCONTEXT;
7793 va_start(args, pat);
7794 sv = vnewSVpvf(pat, &args);
7801 =for apidoc newSVpvf
7803 Creates a new SV and initializes it with the string formatted like
7810 Perl_newSVpvf(pTHX_ const char *const pat, ...)
7815 PERL_ARGS_ASSERT_NEWSVPVF;
7817 va_start(args, pat);
7818 sv = vnewSVpvf(pat, &args);
7823 /* backend for newSVpvf() and newSVpvf_nocontext() */
7826 Perl_vnewSVpvf(pTHX_ const char *const pat, va_list *const args)
7831 PERL_ARGS_ASSERT_VNEWSVPVF;
7834 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
7841 Creates a new SV and copies a floating point value into it.
7842 The reference count for the SV is set to 1.
7848 Perl_newSVnv(pTHX_ const NV n)
7861 Creates a new SV and copies an integer into it. The reference count for the
7868 Perl_newSViv(pTHX_ const IV i)
7881 Creates a new SV and copies an unsigned integer into it.
7882 The reference count for the SV is set to 1.
7888 Perl_newSVuv(pTHX_ const UV u)
7899 =for apidoc newSV_type
7901 Creates a new SV, of the type specified. The reference count for the new SV
7908 Perl_newSV_type(pTHX_ const svtype type)
7913 sv_upgrade(sv, type);
7918 =for apidoc newRV_noinc
7920 Creates an RV wrapper for an SV. The reference count for the original
7921 SV is B<not> incremented.
7927 Perl_newRV_noinc(pTHX_ SV *const tmpRef)
7930 register SV *sv = newSV_type(SVt_IV);
7932 PERL_ARGS_ASSERT_NEWRV_NOINC;
7935 SvRV_set(sv, tmpRef);
7940 /* newRV_inc is the official function name to use now.
7941 * newRV_inc is in fact #defined to newRV in sv.h
7945 Perl_newRV(pTHX_ SV *const sv)
7949 PERL_ARGS_ASSERT_NEWRV;
7951 return newRV_noinc(SvREFCNT_inc_simple_NN(sv));
7957 Creates a new SV which is an exact duplicate of the original SV.
7964 Perl_newSVsv(pTHX_ register SV *const old)
7971 if (SvTYPE(old) == SVTYPEMASK) {
7972 if (ckWARN_d(WARN_INTERNAL))
7973 Perl_warner(aTHX_ packWARN(WARN_INTERNAL), "semi-panic: attempt to dup freed string");
7977 /* SV_GMAGIC is the default for sv_setv()
7978 SV_NOSTEAL prevents TEMP buffers being, well, stolen, and saves games
7979 with SvTEMP_off and SvTEMP_on round a call to sv_setsv. */
7980 sv_setsv_flags(sv, old, SV_GMAGIC | SV_NOSTEAL);
7985 =for apidoc sv_reset
7987 Underlying implementation for the C<reset> Perl function.
7988 Note that the perl-level function is vaguely deprecated.
7994 Perl_sv_reset(pTHX_ register const char *s, HV *const stash)
7997 char todo[PERL_UCHAR_MAX+1];
7999 PERL_ARGS_ASSERT_SV_RESET;
8004 if (!*s) { /* reset ?? searches */
8005 MAGIC * const mg = mg_find((const SV *)stash, PERL_MAGIC_symtab);
8007 const U32 count = mg->mg_len / sizeof(PMOP**);
8008 PMOP **pmp = (PMOP**) mg->mg_ptr;
8009 PMOP *const *const end = pmp + count;
8013 SvREADONLY_off(PL_regex_pad[(*pmp)->op_pmoffset]);
8015 (*pmp)->op_pmflags &= ~PMf_USED;
8023 /* reset variables */
8025 if (!HvARRAY(stash))
8028 Zero(todo, 256, char);
8031 I32 i = (unsigned char)*s;
8035 max = (unsigned char)*s++;
8036 for ( ; i <= max; i++) {
8039 for (i = 0; i <= (I32) HvMAX(stash); i++) {
8041 for (entry = HvARRAY(stash)[i];
8043 entry = HeNEXT(entry))
8048 if (!todo[(U8)*HeKEY(entry)])
8050 gv = MUTABLE_GV(HeVAL(entry));
8053 if (SvTHINKFIRST(sv)) {
8054 if (!SvREADONLY(sv) && SvROK(sv))
8056 /* XXX Is this continue a bug? Why should THINKFIRST
8057 exempt us from resetting arrays and hashes? */
8061 if (SvTYPE(sv) >= SVt_PV) {
8063 if (SvPVX_const(sv) != NULL)
8071 if (GvHV(gv) && !HvNAME_get(GvHV(gv))) {
8073 Perl_die(aTHX_ "Can't reset %%ENV on this system");
8076 # if defined(USE_ENVIRON_ARRAY)
8079 # endif /* USE_ENVIRON_ARRAY */
8090 Using various gambits, try to get an IO from an SV: the IO slot if its a
8091 GV; or the recursive result if we're an RV; or the IO slot of the symbol
8092 named after the PV if we're a string.
8098 Perl_sv_2io(pTHX_ SV *const sv)
8103 PERL_ARGS_ASSERT_SV_2IO;
8105 switch (SvTYPE(sv)) {
8107 io = MUTABLE_IO(sv);
8110 if (isGV_with_GP(sv)) {
8111 gv = MUTABLE_GV(sv);
8114 Perl_croak(aTHX_ "Bad filehandle: %s", GvNAME(gv));
8120 Perl_croak(aTHX_ PL_no_usym, "filehandle");
8122 return sv_2io(SvRV(sv));
8123 gv = gv_fetchsv(sv, 0, SVt_PVIO);
8129 Perl_croak(aTHX_ "Bad filehandle: %"SVf, SVfARG(sv));
8138 Using various gambits, try to get a CV from an SV; in addition, try if
8139 possible to set C<*st> and C<*gvp> to the stash and GV associated with it.
8140 The flags in C<lref> are passed to sv_fetchsv.
8146 Perl_sv_2cv(pTHX_ SV *sv, HV **const st, GV **const gvp, const I32 lref)
8152 PERL_ARGS_ASSERT_SV_2CV;
8159 switch (SvTYPE(sv)) {
8163 return MUTABLE_CV(sv);
8170 if (isGV_with_GP(sv)) {
8171 gv = MUTABLE_GV(sv);
8180 SV * const *sp = &sv; /* Used in tryAMAGICunDEREF macro. */
8182 tryAMAGICunDEREF(to_cv);
8185 if (SvTYPE(sv) == SVt_PVCV) {
8186 cv = MUTABLE_CV(sv);
8191 else if(isGV_with_GP(sv))
8192 gv = MUTABLE_GV(sv);
8194 Perl_croak(aTHX_ "Not a subroutine reference");
8196 else if (isGV_with_GP(sv)) {
8198 gv = MUTABLE_GV(sv);
8201 gv = gv_fetchsv(sv, lref, SVt_PVCV); /* Calls get magic */
8207 /* Some flags to gv_fetchsv mean don't really create the GV */
8208 if (!isGV_with_GP(gv)) {
8214 if (lref && !GvCVu(gv)) {
8218 gv_efullname3(tmpsv, gv, NULL);
8219 /* XXX this is probably not what they think they're getting.
8220 * It has the same effect as "sub name;", i.e. just a forward
8222 newSUB(start_subparse(FALSE, 0),
8223 newSVOP(OP_CONST, 0, tmpsv),
8227 Perl_croak(aTHX_ "Unable to create sub named \"%"SVf"\"",
8228 SVfARG(SvOK(sv) ? sv : &PL_sv_no));
8237 Returns true if the SV has a true value by Perl's rules.
8238 Use the C<SvTRUE> macro instead, which may call C<sv_true()> or may
8239 instead use an in-line version.
8245 Perl_sv_true(pTHX_ register SV *const sv)
8250 register const XPV* const tXpv = (XPV*)SvANY(sv);
8252 (tXpv->xpv_cur > 1 ||
8253 (tXpv->xpv_cur && *sv->sv_u.svu_pv != '0')))
8260 return SvIVX(sv) != 0;
8263 return SvNVX(sv) != 0.0;
8265 return sv_2bool(sv);
8271 =for apidoc sv_pvn_force
8273 Get a sensible string out of the SV somehow.
8274 A private implementation of the C<SvPV_force> macro for compilers which
8275 can't cope with complex macro expressions. Always use the macro instead.
8277 =for apidoc sv_pvn_force_flags
8279 Get a sensible string out of the SV somehow.
8280 If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<sv> if
8281 appropriate, else not. C<sv_pvn_force> and C<sv_pvn_force_nomg> are
8282 implemented in terms of this function.
8283 You normally want to use the various wrapper macros instead: see
8284 C<SvPV_force> and C<SvPV_force_nomg>
8290 Perl_sv_pvn_force_flags(pTHX_ SV *const sv, STRLEN *const lp, const I32 flags)
8294 PERL_ARGS_ASSERT_SV_PVN_FORCE_FLAGS;
8296 if (SvTHINKFIRST(sv) && !SvROK(sv))
8297 sv_force_normal_flags(sv, 0);
8307 if (SvREADONLY(sv) && !(flags & SV_MUTABLE_RETURN)) {
8308 const char * const ref = sv_reftype(sv,0);
8310 Perl_croak(aTHX_ "Can't coerce readonly %s to string in %s",
8311 ref, OP_NAME(PL_op));
8313 Perl_croak(aTHX_ "Can't coerce readonly %s to string", ref);
8315 if ((SvTYPE(sv) > SVt_PVLV && SvTYPE(sv) != SVt_PVFM)
8316 || isGV_with_GP(sv))
8317 Perl_croak(aTHX_ "Can't coerce %s to string in %s", sv_reftype(sv,0),
8319 s = sv_2pv_flags(sv, &len, flags);
8323 if (s != SvPVX_const(sv)) { /* Almost, but not quite, sv_setpvn() */
8326 SvUPGRADE(sv, SVt_PV); /* Never FALSE */
8327 SvGROW(sv, len + 1);
8328 Move(s,SvPVX(sv),len,char);
8330 SvPVX(sv)[len] = '\0';
8333 SvPOK_on(sv); /* validate pointer */
8335 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
8336 PTR2UV(sv),SvPVX_const(sv)));
8339 return SvPVX_mutable(sv);
8343 =for apidoc sv_pvbyten_force
8345 The backend for the C<SvPVbytex_force> macro. Always use the macro instead.
8351 Perl_sv_pvbyten_force(pTHX_ SV *const sv, STRLEN *const lp)
8353 PERL_ARGS_ASSERT_SV_PVBYTEN_FORCE;
8355 sv_pvn_force(sv,lp);
8356 sv_utf8_downgrade(sv,0);
8362 =for apidoc sv_pvutf8n_force
8364 The backend for the C<SvPVutf8x_force> macro. Always use the macro instead.
8370 Perl_sv_pvutf8n_force(pTHX_ SV *const sv, STRLEN *const lp)
8372 PERL_ARGS_ASSERT_SV_PVUTF8N_FORCE;
8374 sv_pvn_force(sv,lp);
8375 sv_utf8_upgrade(sv);
8381 =for apidoc sv_reftype
8383 Returns a string describing what the SV is a reference to.
8389 Perl_sv_reftype(pTHX_ const SV *const sv, const int ob)
8391 PERL_ARGS_ASSERT_SV_REFTYPE;
8393 /* The fact that I don't need to downcast to char * everywhere, only in ?:
8394 inside return suggests a const propagation bug in g++. */
8395 if (ob && SvOBJECT(sv)) {
8396 char * const name = HvNAME_get(SvSTASH(sv));
8397 return name ? name : (char *) "__ANON__";
8400 switch (SvTYPE(sv)) {
8415 case SVt_PVLV: return (char *) (SvROK(sv) ? "REF"
8416 /* tied lvalues should appear to be
8417 * scalars for backwards compatitbility */
8418 : (LvTYPE(sv) == 't' || LvTYPE(sv) == 'T')
8419 ? "SCALAR" : "LVALUE");
8420 case SVt_PVAV: return "ARRAY";
8421 case SVt_PVHV: return "HASH";
8422 case SVt_PVCV: return "CODE";
8423 case SVt_PVGV: return (char *) (isGV_with_GP(sv)
8424 ? "GLOB" : "SCALAR");
8425 case SVt_PVFM: return "FORMAT";
8426 case SVt_PVIO: return "IO";
8427 case SVt_BIND: return "BIND";
8428 case SVt_REGEXP: return "REGEXP";
8429 default: return "UNKNOWN";
8435 =for apidoc sv_isobject
8437 Returns a boolean indicating whether the SV is an RV pointing to a blessed
8438 object. If the SV is not an RV, or if the object is not blessed, then this
8445 Perl_sv_isobject(pTHX_ SV *sv)
8461 Returns a boolean indicating whether the SV is blessed into the specified
8462 class. This does not check for subtypes; use C<sv_derived_from> to verify
8463 an inheritance relationship.
8469 Perl_sv_isa(pTHX_ SV *sv, const char *const name)
8473 PERL_ARGS_ASSERT_SV_ISA;
8483 hvname = HvNAME_get(SvSTASH(sv));
8487 return strEQ(hvname, name);
8493 Creates a new SV for the RV, C<rv>, to point to. If C<rv> is not an RV then
8494 it will be upgraded to one. If C<classname> is non-null then the new SV will
8495 be blessed in the specified package. The new SV is returned and its
8496 reference count is 1.
8502 Perl_newSVrv(pTHX_ SV *const rv, const char *const classname)
8507 PERL_ARGS_ASSERT_NEWSVRV;
8511 SV_CHECK_THINKFIRST_COW_DROP(rv);
8512 (void)SvAMAGIC_off(rv);
8514 if (SvTYPE(rv) >= SVt_PVMG) {
8515 const U32 refcnt = SvREFCNT(rv);
8519 SvREFCNT(rv) = refcnt;
8521 sv_upgrade(rv, SVt_IV);
8522 } else if (SvROK(rv)) {
8523 SvREFCNT_dec(SvRV(rv));
8525 prepare_SV_for_RV(rv);
8533 HV* const stash = gv_stashpv(classname, GV_ADD);
8534 (void)sv_bless(rv, stash);
8540 =for apidoc sv_setref_pv
8542 Copies a pointer into a new SV, optionally blessing the SV. The C<rv>
8543 argument will be upgraded to an RV. That RV will be modified to point to
8544 the new SV. If the C<pv> argument is NULL then C<PL_sv_undef> will be placed
8545 into the SV. The C<classname> argument indicates the package for the
8546 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
8547 will have a reference count of 1, and the RV will be returned.
8549 Do not use with other Perl types such as HV, AV, SV, CV, because those
8550 objects will become corrupted by the pointer copy process.
8552 Note that C<sv_setref_pvn> copies the string while this copies the pointer.
8558 Perl_sv_setref_pv(pTHX_ SV *const rv, const char *const classname, void *const pv)
8562 PERL_ARGS_ASSERT_SV_SETREF_PV;
8565 sv_setsv(rv, &PL_sv_undef);
8569 sv_setiv(newSVrv(rv,classname), PTR2IV(pv));
8574 =for apidoc sv_setref_iv
8576 Copies an integer into a new SV, optionally blessing the SV. The C<rv>
8577 argument will be upgraded to an RV. That RV will be modified to point to
8578 the new SV. The C<classname> argument indicates the package for the
8579 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
8580 will have a reference count of 1, and the RV will be returned.
8586 Perl_sv_setref_iv(pTHX_ SV *const rv, const char *const classname, const IV iv)
8588 PERL_ARGS_ASSERT_SV_SETREF_IV;
8590 sv_setiv(newSVrv(rv,classname), iv);
8595 =for apidoc sv_setref_uv
8597 Copies an unsigned integer into a new SV, optionally blessing the SV. The C<rv>
8598 argument will be upgraded to an RV. That RV will be modified to point to
8599 the new SV. The C<classname> argument indicates the package for the
8600 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
8601 will have a reference count of 1, and the RV will be returned.
8607 Perl_sv_setref_uv(pTHX_ SV *const rv, const char *const classname, const UV uv)
8609 PERL_ARGS_ASSERT_SV_SETREF_UV;
8611 sv_setuv(newSVrv(rv,classname), uv);
8616 =for apidoc sv_setref_nv
8618 Copies a double into a new SV, optionally blessing the SV. The C<rv>
8619 argument will be upgraded to an RV. That RV will be modified to point to
8620 the new SV. The C<classname> argument indicates the package for the
8621 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
8622 will have a reference count of 1, and the RV will be returned.
8628 Perl_sv_setref_nv(pTHX_ SV *const rv, const char *const classname, const NV nv)
8630 PERL_ARGS_ASSERT_SV_SETREF_NV;
8632 sv_setnv(newSVrv(rv,classname), nv);
8637 =for apidoc sv_setref_pvn
8639 Copies a string into a new SV, optionally blessing the SV. The length of the
8640 string must be specified with C<n>. The C<rv> argument will be upgraded to
8641 an RV. That RV will be modified to point to the new SV. The C<classname>
8642 argument indicates the package for the blessing. Set C<classname> to
8643 C<NULL> to avoid the blessing. The new SV will have a reference count
8644 of 1, and the RV will be returned.
8646 Note that C<sv_setref_pv> copies the pointer while this copies the string.
8652 Perl_sv_setref_pvn(pTHX_ SV *const rv, const char *const classname,
8653 const char *const pv, const STRLEN n)
8655 PERL_ARGS_ASSERT_SV_SETREF_PVN;
8657 sv_setpvn(newSVrv(rv,classname), pv, n);
8662 =for apidoc sv_bless
8664 Blesses an SV into a specified package. The SV must be an RV. The package
8665 must be designated by its stash (see C<gv_stashpv()>). The reference count
8666 of the SV is unaffected.
8672 Perl_sv_bless(pTHX_ SV *const sv, HV *const stash)
8677 PERL_ARGS_ASSERT_SV_BLESS;
8680 Perl_croak(aTHX_ "Can't bless non-reference value");
8682 if (SvFLAGS(tmpRef) & (SVs_OBJECT|SVf_READONLY)) {
8683 if (SvIsCOW(tmpRef))
8684 sv_force_normal_flags(tmpRef, 0);
8685 if (SvREADONLY(tmpRef))
8686 Perl_croak(aTHX_ "%s", PL_no_modify);
8687 if (SvOBJECT(tmpRef)) {
8688 if (SvTYPE(tmpRef) != SVt_PVIO)
8690 SvREFCNT_dec(SvSTASH(tmpRef));
8693 SvOBJECT_on(tmpRef);
8694 if (SvTYPE(tmpRef) != SVt_PVIO)
8696 SvUPGRADE(tmpRef, SVt_PVMG);
8697 SvSTASH_set(tmpRef, MUTABLE_HV(SvREFCNT_inc_simple(stash)));
8702 (void)SvAMAGIC_off(sv);
8704 if(SvSMAGICAL(tmpRef))
8705 if(mg_find(tmpRef, PERL_MAGIC_ext) || mg_find(tmpRef, PERL_MAGIC_uvar))
8713 /* Downgrades a PVGV to a PVMG.
8717 S_sv_unglob(pTHX_ SV *const sv)
8722 SV * const temp = sv_newmortal();
8724 PERL_ARGS_ASSERT_SV_UNGLOB;
8726 assert(SvTYPE(sv) == SVt_PVGV);
8728 gv_efullname3(temp, MUTABLE_GV(sv), "*");
8731 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
8732 && HvNAME_get(stash))
8733 mro_method_changed_in(stash);
8734 gp_free(MUTABLE_GV(sv));
8737 sv_del_backref(MUTABLE_SV(GvSTASH(sv)), sv);
8741 if (GvNAME_HEK(sv)) {
8742 unshare_hek(GvNAME_HEK(sv));
8744 isGV_with_GP_off(sv);
8746 /* need to keep SvANY(sv) in the right arena */
8747 xpvmg = new_XPVMG();
8748 StructCopy(SvANY(sv), xpvmg, XPVMG);
8749 del_XPVGV(SvANY(sv));
8752 SvFLAGS(sv) &= ~SVTYPEMASK;
8753 SvFLAGS(sv) |= SVt_PVMG;
8755 /* Intentionally not calling any local SET magic, as this isn't so much a
8756 set operation as merely an internal storage change. */
8757 sv_setsv_flags(sv, temp, 0);
8761 =for apidoc sv_unref_flags
8763 Unsets the RV status of the SV, and decrements the reference count of
8764 whatever was being referenced by the RV. This can almost be thought of
8765 as a reversal of C<newSVrv>. The C<cflags> argument can contain
8766 C<SV_IMMEDIATE_UNREF> to force the reference count to be decremented
8767 (otherwise the decrementing is conditional on the reference count being
8768 different from one or the reference being a readonly SV).
8775 Perl_sv_unref_flags(pTHX_ SV *const ref, const U32 flags)
8777 SV* const target = SvRV(ref);
8779 PERL_ARGS_ASSERT_SV_UNREF_FLAGS;
8781 if (SvWEAKREF(ref)) {
8782 sv_del_backref(target, ref);
8784 SvRV_set(ref, NULL);
8787 SvRV_set(ref, NULL);
8789 /* You can't have a || SvREADONLY(target) here, as $a = $$a, where $a was
8790 assigned to as BEGIN {$a = \"Foo"} will fail. */
8791 if (SvREFCNT(target) != 1 || (flags & SV_IMMEDIATE_UNREF))
8792 SvREFCNT_dec(target);
8793 else /* XXX Hack, but hard to make $a=$a->[1] work otherwise */
8794 sv_2mortal(target); /* Schedule for freeing later */
8798 =for apidoc sv_untaint
8800 Untaint an SV. Use C<SvTAINTED_off> instead.
8805 Perl_sv_untaint(pTHX_ SV *const sv)
8807 PERL_ARGS_ASSERT_SV_UNTAINT;
8809 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
8810 MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
8817 =for apidoc sv_tainted
8819 Test an SV for taintedness. Use C<SvTAINTED> instead.
8824 Perl_sv_tainted(pTHX_ SV *const sv)
8826 PERL_ARGS_ASSERT_SV_TAINTED;
8828 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
8829 const MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
8830 if (mg && (mg->mg_len & 1) )
8837 =for apidoc sv_setpviv
8839 Copies an integer into the given SV, also updating its string value.
8840 Does not handle 'set' magic. See C<sv_setpviv_mg>.
8846 Perl_sv_setpviv(pTHX_ SV *const sv, const IV iv)
8848 char buf[TYPE_CHARS(UV)];
8850 char * const ptr = uiv_2buf(buf, iv, 0, 0, &ebuf);
8852 PERL_ARGS_ASSERT_SV_SETPVIV;
8854 sv_setpvn(sv, ptr, ebuf - ptr);
8858 =for apidoc sv_setpviv_mg
8860 Like C<sv_setpviv>, but also handles 'set' magic.
8866 Perl_sv_setpviv_mg(pTHX_ SV *const sv, const IV iv)
8868 PERL_ARGS_ASSERT_SV_SETPVIV_MG;
8874 #if defined(PERL_IMPLICIT_CONTEXT)
8876 /* pTHX_ magic can't cope with varargs, so this is a no-context
8877 * version of the main function, (which may itself be aliased to us).
8878 * Don't access this version directly.
8882 Perl_sv_setpvf_nocontext(SV *const sv, const char *const pat, ...)
8887 PERL_ARGS_ASSERT_SV_SETPVF_NOCONTEXT;
8889 va_start(args, pat);
8890 sv_vsetpvf(sv, pat, &args);
8894 /* pTHX_ magic can't cope with varargs, so this is a no-context
8895 * version of the main function, (which may itself be aliased to us).
8896 * Don't access this version directly.
8900 Perl_sv_setpvf_mg_nocontext(SV *const sv, const char *const pat, ...)
8905 PERL_ARGS_ASSERT_SV_SETPVF_MG_NOCONTEXT;
8907 va_start(args, pat);
8908 sv_vsetpvf_mg(sv, pat, &args);
8914 =for apidoc sv_setpvf
8916 Works like C<sv_catpvf> but copies the text into the SV instead of
8917 appending it. Does not handle 'set' magic. See C<sv_setpvf_mg>.
8923 Perl_sv_setpvf(pTHX_ SV *const sv, const char *const pat, ...)
8927 PERL_ARGS_ASSERT_SV_SETPVF;
8929 va_start(args, pat);
8930 sv_vsetpvf(sv, pat, &args);
8935 =for apidoc sv_vsetpvf
8937 Works like C<sv_vcatpvf> but copies the text into the SV instead of
8938 appending it. Does not handle 'set' magic. See C<sv_vsetpvf_mg>.
8940 Usually used via its frontend C<sv_setpvf>.
8946 Perl_sv_vsetpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
8948 PERL_ARGS_ASSERT_SV_VSETPVF;
8950 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
8954 =for apidoc sv_setpvf_mg
8956 Like C<sv_setpvf>, but also handles 'set' magic.
8962 Perl_sv_setpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
8966 PERL_ARGS_ASSERT_SV_SETPVF_MG;
8968 va_start(args, pat);
8969 sv_vsetpvf_mg(sv, pat, &args);
8974 =for apidoc sv_vsetpvf_mg
8976 Like C<sv_vsetpvf>, but also handles 'set' magic.
8978 Usually used via its frontend C<sv_setpvf_mg>.
8984 Perl_sv_vsetpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
8986 PERL_ARGS_ASSERT_SV_VSETPVF_MG;
8988 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
8992 #if defined(PERL_IMPLICIT_CONTEXT)
8994 /* pTHX_ magic can't cope with varargs, so this is a no-context
8995 * version of the main function, (which may itself be aliased to us).
8996 * Don't access this version directly.
9000 Perl_sv_catpvf_nocontext(SV *const sv, const char *const pat, ...)
9005 PERL_ARGS_ASSERT_SV_CATPVF_NOCONTEXT;
9007 va_start(args, pat);
9008 sv_vcatpvf(sv, pat, &args);
9012 /* pTHX_ magic can't cope with varargs, so this is a no-context
9013 * version of the main function, (which may itself be aliased to us).
9014 * Don't access this version directly.
9018 Perl_sv_catpvf_mg_nocontext(SV *const sv, const char *const pat, ...)
9023 PERL_ARGS_ASSERT_SV_CATPVF_MG_NOCONTEXT;
9025 va_start(args, pat);
9026 sv_vcatpvf_mg(sv, pat, &args);
9032 =for apidoc sv_catpvf
9034 Processes its arguments like C<sprintf> and appends the formatted
9035 output to an SV. If the appended data contains "wide" characters
9036 (including, but not limited to, SVs with a UTF-8 PV formatted with %s,
9037 and characters >255 formatted with %c), the original SV might get
9038 upgraded to UTF-8. Handles 'get' magic, but not 'set' magic. See
9039 C<sv_catpvf_mg>. If the original SV was UTF-8, the pattern should be
9040 valid UTF-8; if the original SV was bytes, the pattern should be too.
9045 Perl_sv_catpvf(pTHX_ SV *const sv, const char *const pat, ...)
9049 PERL_ARGS_ASSERT_SV_CATPVF;
9051 va_start(args, pat);
9052 sv_vcatpvf(sv, pat, &args);
9057 =for apidoc sv_vcatpvf
9059 Processes its arguments like C<vsprintf> and appends the formatted output
9060 to an SV. Does not handle 'set' magic. See C<sv_vcatpvf_mg>.
9062 Usually used via its frontend C<sv_catpvf>.
9068 Perl_sv_vcatpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9070 PERL_ARGS_ASSERT_SV_VCATPVF;
9072 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9076 =for apidoc sv_catpvf_mg
9078 Like C<sv_catpvf>, but also handles 'set' magic.
9084 Perl_sv_catpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
9088 PERL_ARGS_ASSERT_SV_CATPVF_MG;
9090 va_start(args, pat);
9091 sv_vcatpvf_mg(sv, pat, &args);
9096 =for apidoc sv_vcatpvf_mg
9098 Like C<sv_vcatpvf>, but also handles 'set' magic.
9100 Usually used via its frontend C<sv_catpvf_mg>.
9106 Perl_sv_vcatpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9108 PERL_ARGS_ASSERT_SV_VCATPVF_MG;
9110 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9115 =for apidoc sv_vsetpvfn
9117 Works like C<sv_vcatpvfn> but copies the text into the SV instead of
9120 Usually used via one of its frontends C<sv_vsetpvf> and C<sv_vsetpvf_mg>.
9126 Perl_sv_vsetpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
9127 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
9129 PERL_ARGS_ASSERT_SV_VSETPVFN;
9132 sv_vcatpvfn(sv, pat, patlen, args, svargs, svmax, maybe_tainted);
9136 S_expect_number(pTHX_ char **const pattern)
9141 PERL_ARGS_ASSERT_EXPECT_NUMBER;
9143 switch (**pattern) {
9144 case '1': case '2': case '3':
9145 case '4': case '5': case '6':
9146 case '7': case '8': case '9':
9147 var = *(*pattern)++ - '0';
9148 while (isDIGIT(**pattern)) {
9149 const I32 tmp = var * 10 + (*(*pattern)++ - '0');
9151 Perl_croak(aTHX_ "Integer overflow in format string for %s", (PL_op ? OP_NAME(PL_op) : "sv_vcatpvfn"));
9159 S_F0convert(NV nv, char *const endbuf, STRLEN *const len)
9161 const int neg = nv < 0;
9164 PERL_ARGS_ASSERT_F0CONVERT;
9172 if (uv & 1 && uv == nv)
9173 uv--; /* Round to even */
9175 const unsigned dig = uv % 10;
9188 =for apidoc sv_vcatpvfn
9190 Processes its arguments like C<vsprintf> and appends the formatted output
9191 to an SV. Uses an array of SVs if the C style variable argument list is
9192 missing (NULL). When running with taint checks enabled, indicates via
9193 C<maybe_tainted> if results are untrustworthy (often due to the use of
9196 Usually used via one of its frontends C<sv_vcatpvf> and C<sv_vcatpvf_mg>.
9202 #define VECTORIZE_ARGS vecsv = va_arg(*args, SV*);\
9203 vecstr = (U8*)SvPV_const(vecsv,veclen);\
9204 vec_utf8 = DO_UTF8(vecsv);
9206 /* XXX maybe_tainted is never assigned to, so the doc above is lying. */
9209 Perl_sv_vcatpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
9210 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
9218 static const char nullstr[] = "(null)";
9220 bool has_utf8 = DO_UTF8(sv); /* has the result utf8? */
9221 const bool pat_utf8 = has_utf8; /* the pattern is in utf8? */
9223 /* Times 4: a decimal digit takes more than 3 binary digits.
9224 * NV_DIG: mantissa takes than many decimal digits.
9225 * Plus 32: Playing safe. */
9226 char ebuf[IV_DIG * 4 + NV_DIG + 32];
9227 /* large enough for "%#.#f" --chip */
9228 /* what about long double NVs? --jhi */
9230 PERL_ARGS_ASSERT_SV_VCATPVFN;
9231 PERL_UNUSED_ARG(maybe_tainted);
9233 /* no matter what, this is a string now */
9234 (void)SvPV_force(sv, origlen);
9236 /* special-case "", "%s", and "%-p" (SVf - see below) */
9239 if (patlen == 2 && pat[0] == '%' && pat[1] == 's') {
9241 const char * const s = va_arg(*args, char*);
9242 sv_catpv(sv, s ? s : nullstr);
9244 else if (svix < svmax) {
9245 sv_catsv(sv, *svargs);
9249 if (args && patlen == 3 && pat[0] == '%' &&
9250 pat[1] == '-' && pat[2] == 'p') {
9251 argsv = MUTABLE_SV(va_arg(*args, void*));
9252 sv_catsv(sv, argsv);
9256 #ifndef USE_LONG_DOUBLE
9257 /* special-case "%.<number>[gf]" */
9258 if ( !args && patlen <= 5 && pat[0] == '%' && pat[1] == '.'
9259 && (pat[patlen-1] == 'g' || pat[patlen-1] == 'f') ) {
9260 unsigned digits = 0;
9264 while (*pp >= '0' && *pp <= '9')
9265 digits = 10 * digits + (*pp++ - '0');
9266 if (pp - pat == (int)patlen - 1) {
9274 /* Add check for digits != 0 because it seems that some
9275 gconverts are buggy in this case, and we don't yet have
9276 a Configure test for this. */
9277 if (digits && digits < sizeof(ebuf) - NV_DIG - 10) {
9278 /* 0, point, slack */
9279 Gconvert(nv, (int)digits, 0, ebuf);
9281 if (*ebuf) /* May return an empty string for digits==0 */
9284 } else if (!digits) {
9287 if ((p = F0convert(nv, ebuf + sizeof ebuf, &l))) {
9288 sv_catpvn(sv, p, l);
9294 #endif /* !USE_LONG_DOUBLE */
9296 if (!args && svix < svmax && DO_UTF8(*svargs))
9299 patend = (char*)pat + patlen;
9300 for (p = (char*)pat; p < patend; p = q) {
9303 bool vectorize = FALSE;
9304 bool vectorarg = FALSE;
9305 bool vec_utf8 = FALSE;
9311 bool has_precis = FALSE;
9313 const I32 osvix = svix;
9314 bool is_utf8 = FALSE; /* is this item utf8? */
9315 #ifdef HAS_LDBL_SPRINTF_BUG
9316 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
9317 with sfio - Allen <allens@cpan.org> */
9318 bool fix_ldbl_sprintf_bug = FALSE;
9322 U8 utf8buf[UTF8_MAXBYTES+1];
9323 STRLEN esignlen = 0;
9325 const char *eptr = NULL;
9326 const char *fmtstart;
9329 const U8 *vecstr = NULL;
9336 /* we need a long double target in case HAS_LONG_DOUBLE but
9339 #if defined(HAS_LONG_DOUBLE) && LONG_DOUBLESIZE > DOUBLESIZE
9347 const char *dotstr = ".";
9348 STRLEN dotstrlen = 1;
9349 I32 efix = 0; /* explicit format parameter index */
9350 I32 ewix = 0; /* explicit width index */
9351 I32 epix = 0; /* explicit precision index */
9352 I32 evix = 0; /* explicit vector index */
9353 bool asterisk = FALSE;
9355 /* echo everything up to the next format specification */
9356 for (q = p; q < patend && *q != '%'; ++q) ;
9358 if (has_utf8 && !pat_utf8)
9359 sv_catpvn_utf8_upgrade(sv, p, q - p, nsv);
9361 sv_catpvn(sv, p, q - p);
9370 We allow format specification elements in this order:
9371 \d+\$ explicit format parameter index
9373 v|\*(\d+\$)?v vector with optional (optionally specified) arg
9374 0 flag (as above): repeated to allow "v02"
9375 \d+|\*(\d+\$)? width using optional (optionally specified) arg
9376 \.(\d*|\*(\d+\$)?) precision using optional (optionally specified) arg
9378 [%bcdefginopsuxDFOUX] format (mandatory)
9383 As of perl5.9.3, printf format checking is on by default.
9384 Internally, perl uses %p formats to provide an escape to
9385 some extended formatting. This block deals with those
9386 extensions: if it does not match, (char*)q is reset and
9387 the normal format processing code is used.
9389 Currently defined extensions are:
9390 %p include pointer address (standard)
9391 %-p (SVf) include an SV (previously %_)
9392 %-<num>p include an SV with precision <num>
9393 %<num>p reserved for future extensions
9395 Robin Barker 2005-07-14
9397 %1p (VDf) removed. RMB 2007-10-19
9404 n = expect_number(&q);
9411 argsv = MUTABLE_SV(va_arg(*args, void*));
9412 eptr = SvPV_const(argsv, elen);
9418 if (ckWARN_d(WARN_INTERNAL))
9419 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
9420 "internal %%<num>p might conflict with future printf extensions");
9426 if ( (width = expect_number(&q)) ) {
9441 if (plus == '+' && *q == ' ') /* '+' over ' ' */
9470 if ( (ewix = expect_number(&q)) )
9479 if ((vectorarg = asterisk)) {
9492 width = expect_number(&q);
9498 vecsv = va_arg(*args, SV*);
9500 vecsv = (evix > 0 && evix <= svmax)
9501 ? svargs[evix-1] : &PL_sv_undef;
9503 vecsv = svix < svmax ? svargs[svix++] : &PL_sv_undef;
9505 dotstr = SvPV_const(vecsv, dotstrlen);
9506 /* Keep the DO_UTF8 test *after* the SvPV call, else things go
9507 bad with tied or overloaded values that return UTF8. */
9510 else if (has_utf8) {
9511 vecsv = sv_mortalcopy(vecsv);
9512 sv_utf8_upgrade(vecsv);
9513 dotstr = SvPV_const(vecsv, dotstrlen);
9520 else if (efix ? (efix > 0 && efix <= svmax) : svix < svmax) {
9521 vecsv = svargs[efix ? efix-1 : svix++];
9522 vecstr = (U8*)SvPV_const(vecsv,veclen);
9523 vec_utf8 = DO_UTF8(vecsv);
9525 /* if this is a version object, we need to convert
9526 * back into v-string notation and then let the
9527 * vectorize happen normally
9529 if (sv_derived_from(vecsv, "version")) {
9530 char *version = savesvpv(vecsv);
9531 if ( hv_exists(MUTABLE_HV(SvRV(vecsv)), "alpha", 5 ) ) {
9532 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
9533 "vector argument not supported with alpha versions");
9536 vecsv = sv_newmortal();
9537 scan_vstring(version, version + veclen, vecsv);
9538 vecstr = (U8*)SvPV_const(vecsv, veclen);
9539 vec_utf8 = DO_UTF8(vecsv);
9551 i = va_arg(*args, int);
9553 i = (ewix ? ewix <= svmax : svix < svmax) ?
9554 SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
9556 width = (i < 0) ? -i : i;
9566 if ( ((epix = expect_number(&q))) && (*q++ != '$') )
9568 /* XXX: todo, support specified precision parameter */
9572 i = va_arg(*args, int);
9574 i = (ewix ? ewix <= svmax : svix < svmax)
9575 ? SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
9577 has_precis = !(i < 0);
9582 precis = precis * 10 + (*q++ - '0');
9591 case 'I': /* Ix, I32x, and I64x */
9593 if (q[1] == '6' && q[2] == '4') {
9599 if (q[1] == '3' && q[2] == '2') {
9609 #if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
9620 #if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
9621 if (*(q + 1) == 'l') { /* lld, llf */
9647 if (!vectorize && !args) {
9649 const I32 i = efix-1;
9650 argsv = (i >= 0 && i < svmax) ? svargs[i] : &PL_sv_undef;
9652 argsv = (svix >= 0 && svix < svmax)
9653 ? svargs[svix++] : &PL_sv_undef;
9664 uv = (args) ? va_arg(*args, int) : SvIV(argsv);
9666 (!UNI_IS_INVARIANT(uv) && SvUTF8(sv)))
9668 eptr = (char*)utf8buf;
9669 elen = uvchr_to_utf8((U8*)eptr, uv) - utf8buf;
9683 eptr = va_arg(*args, char*);
9685 #ifdef MACOS_TRADITIONAL
9686 /* On MacOS, %#s format is used for Pascal strings */
9691 elen = strlen(eptr);
9693 eptr = (char *)nullstr;
9694 elen = sizeof nullstr - 1;
9698 eptr = SvPV_const(argsv, elen);
9699 if (DO_UTF8(argsv)) {
9700 I32 old_precis = precis;
9701 if (has_precis && precis < elen) {
9703 sv_pos_u2b(argsv, &p, 0); /* sticks at end */
9706 if (width) { /* fudge width (can't fudge elen) */
9707 if (has_precis && precis < elen)
9708 width += precis - old_precis;
9710 width += elen - sv_len_utf8(argsv);
9717 if (has_precis && elen > precis)
9724 if (alt || vectorize)
9726 uv = PTR2UV(args ? va_arg(*args, void*) : argsv);
9747 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
9756 esignbuf[esignlen++] = plus;
9760 case 'h': iv = (short)va_arg(*args, int); break;
9761 case 'l': iv = va_arg(*args, long); break;
9762 case 'V': iv = va_arg(*args, IV); break;
9763 default: iv = va_arg(*args, int); break;
9766 iv = va_arg(*args, Quad_t); break;
9773 IV tiv = SvIV(argsv); /* work around GCC bug #13488 */
9775 case 'h': iv = (short)tiv; break;
9776 case 'l': iv = (long)tiv; break;
9778 default: iv = tiv; break;
9781 iv = (Quad_t)tiv; break;
9787 if ( !vectorize ) /* we already set uv above */
9792 esignbuf[esignlen++] = plus;
9796 esignbuf[esignlen++] = '-';
9840 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
9851 case 'h': uv = (unsigned short)va_arg(*args, unsigned); break;
9852 case 'l': uv = va_arg(*args, unsigned long); break;
9853 case 'V': uv = va_arg(*args, UV); break;
9854 default: uv = va_arg(*args, unsigned); break;
9857 uv = va_arg(*args, Uquad_t); break;
9864 UV tuv = SvUV(argsv); /* work around GCC bug #13488 */
9866 case 'h': uv = (unsigned short)tuv; break;
9867 case 'l': uv = (unsigned long)tuv; break;
9869 default: uv = tuv; break;
9872 uv = (Uquad_t)tuv; break;
9881 char *ptr = ebuf + sizeof ebuf;
9882 bool tempalt = uv ? alt : FALSE; /* Vectors can't change alt */
9888 p = (char *)((c == 'X') ? PL_hexdigit + 16 : PL_hexdigit);
9894 esignbuf[esignlen++] = '0';
9895 esignbuf[esignlen++] = c; /* 'x' or 'X' */
9903 if (alt && *ptr != '0')
9912 esignbuf[esignlen++] = '0';
9913 esignbuf[esignlen++] = c;
9916 default: /* it had better be ten or less */
9920 } while (uv /= base);
9923 elen = (ebuf + sizeof ebuf) - ptr;
9927 zeros = precis - elen;
9928 else if (precis == 0 && elen == 1 && *eptr == '0'
9929 && !(base == 8 && alt)) /* "%#.0o" prints "0" */
9932 /* a precision nullifies the 0 flag. */
9939 /* FLOATING POINT */
9942 c = 'f'; /* maybe %F isn't supported here */
9950 /* This is evil, but floating point is even more evil */
9952 /* for SV-style calling, we can only get NV
9953 for C-style calling, we assume %f is double;
9954 for simplicity we allow any of %Lf, %llf, %qf for long double
9958 #if defined(USE_LONG_DOUBLE)
9962 /* [perl #20339] - we should accept and ignore %lf rather than die */
9966 #if defined(USE_LONG_DOUBLE)
9967 intsize = args ? 0 : 'q';
9971 #if defined(HAS_LONG_DOUBLE)
9980 /* now we need (long double) if intsize == 'q', else (double) */
9982 #if LONG_DOUBLESIZE > DOUBLESIZE
9984 va_arg(*args, long double) :
9985 va_arg(*args, double)
9987 va_arg(*args, double)
9992 /* nv * 0 will be NaN for NaN, +Inf and -Inf, and 0 for anything
9993 else. frexp() has some unspecified behaviour for those three */
9994 if (c != 'e' && c != 'E' && (nv * 0) == 0) {
9996 /* FIXME: if HAS_LONG_DOUBLE but not USE_LONG_DOUBLE this
9997 will cast our (long double) to (double) */
9998 (void)Perl_frexp(nv, &i);
9999 if (i == PERL_INT_MIN)
10000 Perl_die(aTHX_ "panic: frexp");
10002 need = BIT_DIGITS(i);
10004 need += has_precis ? precis : 6; /* known default */
10009 #ifdef HAS_LDBL_SPRINTF_BUG
10010 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
10011 with sfio - Allen <allens@cpan.org> */
10014 # define MY_DBL_MAX DBL_MAX
10015 # else /* XXX guessing! HUGE_VAL may be defined as infinity, so not using */
10016 # if DOUBLESIZE >= 8
10017 # define MY_DBL_MAX 1.7976931348623157E+308L
10019 # define MY_DBL_MAX 3.40282347E+38L
10023 # ifdef HAS_LDBL_SPRINTF_BUG_LESS1 /* only between -1L & 1L - Allen */
10024 # define MY_DBL_MAX_BUG 1L
10026 # define MY_DBL_MAX_BUG MY_DBL_MAX
10030 # define MY_DBL_MIN DBL_MIN
10031 # else /* XXX guessing! -Allen */
10032 # if DOUBLESIZE >= 8
10033 # define MY_DBL_MIN 2.2250738585072014E-308L
10035 # define MY_DBL_MIN 1.17549435E-38L
10039 if ((intsize == 'q') && (c == 'f') &&
10040 ((nv < MY_DBL_MAX_BUG) && (nv > -MY_DBL_MAX_BUG)) &&
10041 (need < DBL_DIG)) {
10042 /* it's going to be short enough that
10043 * long double precision is not needed */
10045 if ((nv <= 0L) && (nv >= -0L))
10046 fix_ldbl_sprintf_bug = TRUE; /* 0 is 0 - easiest */
10048 /* would use Perl_fp_class as a double-check but not
10049 * functional on IRIX - see perl.h comments */
10051 if ((nv >= MY_DBL_MIN) || (nv <= -MY_DBL_MIN)) {
10052 /* It's within the range that a double can represent */
10053 #if defined(DBL_MAX) && !defined(DBL_MIN)
10054 if ((nv >= ((long double)1/DBL_MAX)) ||
10055 (nv <= (-(long double)1/DBL_MAX)))
10057 fix_ldbl_sprintf_bug = TRUE;
10060 if (fix_ldbl_sprintf_bug == TRUE) {
10070 # undef MY_DBL_MAX_BUG
10073 #endif /* HAS_LDBL_SPRINTF_BUG */
10075 need += 20; /* fudge factor */
10076 if (PL_efloatsize < need) {
10077 Safefree(PL_efloatbuf);
10078 PL_efloatsize = need + 20; /* more fudge */
10079 Newx(PL_efloatbuf, PL_efloatsize, char);
10080 PL_efloatbuf[0] = '\0';
10083 if ( !(width || left || plus || alt) && fill != '0'
10084 && has_precis && intsize != 'q' ) { /* Shortcuts */
10085 /* See earlier comment about buggy Gconvert when digits,
10087 if ( c == 'g' && precis) {
10088 Gconvert((NV)nv, (int)precis, 0, PL_efloatbuf);
10089 /* May return an empty string for digits==0 */
10090 if (*PL_efloatbuf) {
10091 elen = strlen(PL_efloatbuf);
10092 goto float_converted;
10094 } else if ( c == 'f' && !precis) {
10095 if ((eptr = F0convert(nv, ebuf + sizeof ebuf, &elen)))
10100 char *ptr = ebuf + sizeof ebuf;
10103 /* FIXME: what to do if HAS_LONG_DOUBLE but not PERL_PRIfldbl? */
10104 #if defined(HAS_LONG_DOUBLE) && defined(PERL_PRIfldbl)
10105 if (intsize == 'q') {
10106 /* Copy the one or more characters in a long double
10107 * format before the 'base' ([efgEFG]) character to
10108 * the format string. */
10109 static char const prifldbl[] = PERL_PRIfldbl;
10110 char const *p = prifldbl + sizeof(prifldbl) - 3;
10111 while (p >= prifldbl) { *--ptr = *p--; }
10116 do { *--ptr = '0' + (base % 10); } while (base /= 10);
10121 do { *--ptr = '0' + (base % 10); } while (base /= 10);
10133 /* No taint. Otherwise we are in the strange situation
10134 * where printf() taints but print($float) doesn't.
10136 #if defined(HAS_LONG_DOUBLE)
10137 elen = ((intsize == 'q')
10138 ? my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, nv)
10139 : my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, (double)nv));
10141 elen = my_sprintf(PL_efloatbuf, ptr, nv);
10145 eptr = PL_efloatbuf;
10153 i = SvCUR(sv) - origlen;
10156 case 'h': *(va_arg(*args, short*)) = i; break;
10157 default: *(va_arg(*args, int*)) = i; break;
10158 case 'l': *(va_arg(*args, long*)) = i; break;
10159 case 'V': *(va_arg(*args, IV*)) = i; break;
10162 *(va_arg(*args, Quad_t*)) = i; break;
10169 sv_setuv_mg(argsv, (UV)i);
10170 continue; /* not "break" */
10177 && (PL_op->op_type == OP_PRTF || PL_op->op_type == OP_SPRINTF)
10178 && ckWARN(WARN_PRINTF))
10180 SV * const msg = sv_newmortal();
10181 Perl_sv_setpvf(aTHX_ msg, "Invalid conversion in %sprintf: ",
10182 (PL_op->op_type == OP_PRTF) ? "" : "s");
10183 if (fmtstart < patend) {
10184 const char * const fmtend = q < patend ? q : patend;
10186 sv_catpvs(msg, "\"%");
10187 for (f = fmtstart; f < fmtend; f++) {
10189 sv_catpvn(msg, f, 1);
10191 Perl_sv_catpvf(aTHX_ msg,
10192 "\\%03"UVof, (UV)*f & 0xFF);
10195 sv_catpvs(msg, "\"");
10197 sv_catpvs(msg, "end of string");
10199 Perl_warner(aTHX_ packWARN(WARN_PRINTF), "%"SVf, SVfARG(msg)); /* yes, this is reentrant */
10202 /* output mangled stuff ... */
10208 /* ... right here, because formatting flags should not apply */
10209 SvGROW(sv, SvCUR(sv) + elen + 1);
10211 Copy(eptr, p, elen, char);
10214 SvCUR_set(sv, p - SvPVX_const(sv));
10216 continue; /* not "break" */
10219 if (is_utf8 != has_utf8) {
10222 sv_utf8_upgrade(sv);
10225 const STRLEN old_elen = elen;
10226 SV * const nsv = newSVpvn_flags(eptr, elen, SVs_TEMP);
10227 sv_utf8_upgrade(nsv);
10228 eptr = SvPVX_const(nsv);
10231 if (width) { /* fudge width (can't fudge elen) */
10232 width += elen - old_elen;
10238 have = esignlen + zeros + elen;
10240 Perl_croak_nocontext("%s", PL_memory_wrap);
10242 need = (have > width ? have : width);
10245 if (need >= (((STRLEN)~0) - SvCUR(sv) - dotstrlen - 1))
10246 Perl_croak_nocontext("%s", PL_memory_wrap);
10247 SvGROW(sv, SvCUR(sv) + need + dotstrlen + 1);
10249 if (esignlen && fill == '0') {
10251 for (i = 0; i < (int)esignlen; i++)
10252 *p++ = esignbuf[i];
10254 if (gap && !left) {
10255 memset(p, fill, gap);
10258 if (esignlen && fill != '0') {
10260 for (i = 0; i < (int)esignlen; i++)
10261 *p++ = esignbuf[i];
10265 for (i = zeros; i; i--)
10269 Copy(eptr, p, elen, char);
10273 memset(p, ' ', gap);
10278 Copy(dotstr, p, dotstrlen, char);
10282 vectorize = FALSE; /* done iterating over vecstr */
10289 SvCUR_set(sv, p - SvPVX_const(sv));
10297 /* =========================================================================
10299 =head1 Cloning an interpreter
10301 All the macros and functions in this section are for the private use of
10302 the main function, perl_clone().
10304 The foo_dup() functions make an exact copy of an existing foo thingy.
10305 During the course of a cloning, a hash table is used to map old addresses
10306 to new addresses. The table is created and manipulated with the
10307 ptr_table_* functions.
10311 * =========================================================================*/
10314 #if defined(USE_ITHREADS)
10316 /* XXX Remove this so it doesn't have to go thru the macro and return for nothing */
10317 #ifndef GpREFCNT_inc
10318 # define GpREFCNT_inc(gp) ((gp) ? (++(gp)->gp_refcnt, (gp)) : (GP*)NULL)
10322 /* Certain cases in Perl_ss_dup have been merged, by relying on the fact
10323 that currently av_dup, gv_dup and hv_dup are the same as sv_dup.
10324 If this changes, please unmerge ss_dup. */
10325 #define sv_dup_inc(s,t) SvREFCNT_inc(sv_dup(s,t))
10326 #define sv_dup_inc_NN(s,t) SvREFCNT_inc_NN(sv_dup(s,t))
10327 #define av_dup(s,t) MUTABLE_AV(sv_dup((const SV *)s,t))
10328 #define av_dup_inc(s,t) MUTABLE_AV(SvREFCNT_inc(sv_dup((const SV *)s,t)))
10329 #define hv_dup(s,t) MUTABLE_HV(sv_dup((const SV *)s,t))
10330 #define hv_dup_inc(s,t) MUTABLE_HV(SvREFCNT_inc(sv_dup((const SV *)s,t)))
10331 #define cv_dup(s,t) MUTABLE_CV(sv_dup((const SV *)s,t))
10332 #define cv_dup_inc(s,t) MUTABLE_CV(SvREFCNT_inc(sv_dup((const SV *)s,t)))
10333 #define io_dup(s,t) MUTABLE_IO(sv_dup((const SV *)s,t))
10334 #define io_dup_inc(s,t) MUTABLE_IO(SvREFCNT_inc(sv_dup((const SV *)s,t)))
10335 #define gv_dup(s,t) MUTABLE_GV(sv_dup((const SV *)s,t))
10336 #define gv_dup_inc(s,t) MUTABLE_GV(SvREFCNT_inc(sv_dup((const SV *)s,t)))
10337 #define SAVEPV(p) ((p) ? savepv(p) : NULL)
10338 #define SAVEPVN(p,n) ((p) ? savepvn(p,n) : NULL)
10340 /* clone a parser */
10343 Perl_parser_dup(pTHX_ const yy_parser *const proto, CLONE_PARAMS *const param)
10347 PERL_ARGS_ASSERT_PARSER_DUP;
10352 /* look for it in the table first */
10353 parser = (yy_parser *)ptr_table_fetch(PL_ptr_table, proto);
10357 /* create anew and remember what it is */
10358 Newxz(parser, 1, yy_parser);
10359 ptr_table_store(PL_ptr_table, proto, parser);
10361 parser->yyerrstatus = 0;
10362 parser->yychar = YYEMPTY; /* Cause a token to be read. */
10364 /* XXX these not yet duped */
10365 parser->old_parser = NULL;
10366 parser->stack = NULL;
10368 parser->stack_size = 0;
10369 /* XXX parser->stack->state = 0; */
10371 /* XXX eventually, just Copy() most of the parser struct ? */
10373 parser->lex_brackets = proto->lex_brackets;
10374 parser->lex_casemods = proto->lex_casemods;
10375 parser->lex_brackstack = savepvn(proto->lex_brackstack,
10376 (proto->lex_brackets < 120 ? 120 : proto->lex_brackets));
10377 parser->lex_casestack = savepvn(proto->lex_casestack,
10378 (proto->lex_casemods < 12 ? 12 : proto->lex_casemods));
10379 parser->lex_defer = proto->lex_defer;
10380 parser->lex_dojoin = proto->lex_dojoin;
10381 parser->lex_expect = proto->lex_expect;
10382 parser->lex_formbrack = proto->lex_formbrack;
10383 parser->lex_inpat = proto->lex_inpat;
10384 parser->lex_inwhat = proto->lex_inwhat;
10385 parser->lex_op = proto->lex_op;
10386 parser->lex_repl = sv_dup_inc(proto->lex_repl, param);
10387 parser->lex_starts = proto->lex_starts;
10388 parser->lex_stuff = sv_dup_inc(proto->lex_stuff, param);
10389 parser->multi_close = proto->multi_close;
10390 parser->multi_open = proto->multi_open;
10391 parser->multi_start = proto->multi_start;
10392 parser->multi_end = proto->multi_end;
10393 parser->pending_ident = proto->pending_ident;
10394 parser->preambled = proto->preambled;
10395 parser->sublex_info = proto->sublex_info; /* XXX not quite right */
10396 parser->linestr = sv_dup_inc(proto->linestr, param);
10397 parser->expect = proto->expect;
10398 parser->copline = proto->copline;
10399 parser->last_lop_op = proto->last_lop_op;
10400 parser->lex_state = proto->lex_state;
10401 parser->rsfp = fp_dup(proto->rsfp, '<', param);
10402 /* rsfp_filters entries have fake IoDIRP() */
10403 parser->rsfp_filters= av_dup_inc(proto->rsfp_filters, param);
10404 parser->in_my = proto->in_my;
10405 parser->in_my_stash = hv_dup(proto->in_my_stash, param);
10406 parser->error_count = proto->error_count;
10409 parser->linestr = sv_dup_inc(proto->linestr, param);
10412 char * const ols = SvPVX(proto->linestr);
10413 char * const ls = SvPVX(parser->linestr);
10415 parser->bufptr = ls + (proto->bufptr >= ols ?
10416 proto->bufptr - ols : 0);
10417 parser->oldbufptr = ls + (proto->oldbufptr >= ols ?
10418 proto->oldbufptr - ols : 0);
10419 parser->oldoldbufptr= ls + (proto->oldoldbufptr >= ols ?
10420 proto->oldoldbufptr - ols : 0);
10421 parser->linestart = ls + (proto->linestart >= ols ?
10422 proto->linestart - ols : 0);
10423 parser->last_uni = ls + (proto->last_uni >= ols ?
10424 proto->last_uni - ols : 0);
10425 parser->last_lop = ls + (proto->last_lop >= ols ?
10426 proto->last_lop - ols : 0);
10428 parser->bufend = ls + SvCUR(parser->linestr);
10431 Copy(proto->tokenbuf, parser->tokenbuf, 256, char);
10435 parser->endwhite = proto->endwhite;
10436 parser->faketokens = proto->faketokens;
10437 parser->lasttoke = proto->lasttoke;
10438 parser->nextwhite = proto->nextwhite;
10439 parser->realtokenstart = proto->realtokenstart;
10440 parser->skipwhite = proto->skipwhite;
10441 parser->thisclose = proto->thisclose;
10442 parser->thismad = proto->thismad;
10443 parser->thisopen = proto->thisopen;
10444 parser->thisstuff = proto->thisstuff;
10445 parser->thistoken = proto->thistoken;
10446 parser->thiswhite = proto->thiswhite;
10448 Copy(proto->nexttoke, parser->nexttoke, 5, NEXTTOKE);
10449 parser->curforce = proto->curforce;
10451 Copy(proto->nextval, parser->nextval, 5, YYSTYPE);
10452 Copy(proto->nexttype, parser->nexttype, 5, I32);
10453 parser->nexttoke = proto->nexttoke;
10459 /* duplicate a file handle */
10462 Perl_fp_dup(pTHX_ PerlIO *const fp, const char type, CLONE_PARAMS *const param)
10466 PERL_ARGS_ASSERT_FP_DUP;
10467 PERL_UNUSED_ARG(type);
10470 return (PerlIO*)NULL;
10472 /* look for it in the table first */
10473 ret = (PerlIO*)ptr_table_fetch(PL_ptr_table, fp);
10477 /* create anew and remember what it is */
10478 ret = PerlIO_fdupopen(aTHX_ fp, param, PERLIO_DUP_CLONE);
10479 ptr_table_store(PL_ptr_table, fp, ret);
10483 /* duplicate a directory handle */
10486 Perl_dirp_dup(pTHX_ DIR *const dp)
10488 PERL_UNUSED_CONTEXT;
10495 /* duplicate a typeglob */
10498 Perl_gp_dup(pTHX_ GP *const gp, CLONE_PARAMS *const param)
10502 PERL_ARGS_ASSERT_GP_DUP;
10506 /* look for it in the table first */
10507 ret = (GP*)ptr_table_fetch(PL_ptr_table, gp);
10511 /* create anew and remember what it is */
10513 ptr_table_store(PL_ptr_table, gp, ret);
10516 ret->gp_refcnt = 0; /* must be before any other dups! */
10517 ret->gp_sv = sv_dup_inc(gp->gp_sv, param);
10518 ret->gp_io = io_dup_inc(gp->gp_io, param);
10519 ret->gp_form = cv_dup_inc(gp->gp_form, param);
10520 ret->gp_av = av_dup_inc(gp->gp_av, param);
10521 ret->gp_hv = hv_dup_inc(gp->gp_hv, param);
10522 ret->gp_egv = gv_dup(gp->gp_egv, param);/* GvEGV is not refcounted */
10523 ret->gp_cv = cv_dup_inc(gp->gp_cv, param);
10524 ret->gp_cvgen = gp->gp_cvgen;
10525 ret->gp_line = gp->gp_line;
10526 ret->gp_file_hek = hek_dup(gp->gp_file_hek, param);
10530 /* duplicate a chain of magic */
10533 Perl_mg_dup(pTHX_ MAGIC *mg, CLONE_PARAMS *const param)
10535 MAGIC *mgprev = (MAGIC*)NULL;
10538 PERL_ARGS_ASSERT_MG_DUP;
10541 return (MAGIC*)NULL;
10542 /* look for it in the table first */
10543 mgret = (MAGIC*)ptr_table_fetch(PL_ptr_table, mg);
10547 for (; mg; mg = mg->mg_moremagic) {
10549 Newxz(nmg, 1, MAGIC);
10551 mgprev->mg_moremagic = nmg;
10554 nmg->mg_virtual = mg->mg_virtual; /* XXX copy dynamic vtable? */
10555 nmg->mg_private = mg->mg_private;
10556 nmg->mg_type = mg->mg_type;
10557 nmg->mg_flags = mg->mg_flags;
10558 /* FIXME for plugins
10559 if (mg->mg_type == PERL_MAGIC_qr) {
10560 nmg->mg_obj = MUTABLE_SV(CALLREGDUPE((REGEXP*)mg->mg_obj, param));
10564 if(mg->mg_type == PERL_MAGIC_backref) {
10565 /* The backref AV has its reference count deliberately bumped by
10568 = SvREFCNT_inc(av_dup_inc((const AV *) mg->mg_obj, param));
10571 nmg->mg_obj = (mg->mg_flags & MGf_REFCOUNTED)
10572 ? sv_dup_inc(mg->mg_obj, param)
10573 : sv_dup(mg->mg_obj, param);
10575 nmg->mg_len = mg->mg_len;
10576 nmg->mg_ptr = mg->mg_ptr; /* XXX random ptr? */
10577 if (mg->mg_ptr && mg->mg_type != PERL_MAGIC_regex_global) {
10578 if (mg->mg_len > 0) {
10579 nmg->mg_ptr = SAVEPVN(mg->mg_ptr, mg->mg_len);
10580 if (mg->mg_type == PERL_MAGIC_overload_table &&
10581 AMT_AMAGIC((AMT*)mg->mg_ptr))
10583 const AMT * const amtp = (AMT*)mg->mg_ptr;
10584 AMT * const namtp = (AMT*)nmg->mg_ptr;
10586 for (i = 1; i < NofAMmeth; i++) {
10587 namtp->table[i] = cv_dup_inc(amtp->table[i], param);
10591 else if (mg->mg_len == HEf_SVKEY)
10592 nmg->mg_ptr = (char*)sv_dup_inc((const SV *)mg->mg_ptr, param);
10594 if ((mg->mg_flags & MGf_DUP) && mg->mg_virtual && mg->mg_virtual->svt_dup) {
10595 CALL_FPTR(nmg->mg_virtual->svt_dup)(aTHX_ nmg, param);
10602 #endif /* USE_ITHREADS */
10604 /* create a new pointer-mapping table */
10607 Perl_ptr_table_new(pTHX)
10610 PERL_UNUSED_CONTEXT;
10612 Newxz(tbl, 1, PTR_TBL_t);
10613 tbl->tbl_max = 511;
10614 tbl->tbl_items = 0;
10615 Newxz(tbl->tbl_ary, tbl->tbl_max + 1, PTR_TBL_ENT_t*);
10619 #define PTR_TABLE_HASH(ptr) \
10620 ((PTR2UV(ptr) >> 3) ^ (PTR2UV(ptr) >> (3 + 7)) ^ (PTR2UV(ptr) >> (3 + 17)))
10623 we use the PTE_SVSLOT 'reservation' made above, both here (in the
10624 following define) and at call to new_body_inline made below in
10625 Perl_ptr_table_store()
10628 #define del_pte(p) del_body_type(p, PTE_SVSLOT)
10630 /* map an existing pointer using a table */
10632 STATIC PTR_TBL_ENT_t *
10633 S_ptr_table_find(PTR_TBL_t *const tbl, const void *const sv)
10635 PTR_TBL_ENT_t *tblent;
10636 const UV hash = PTR_TABLE_HASH(sv);
10638 PERL_ARGS_ASSERT_PTR_TABLE_FIND;
10640 tblent = tbl->tbl_ary[hash & tbl->tbl_max];
10641 for (; tblent; tblent = tblent->next) {
10642 if (tblent->oldval == sv)
10649 Perl_ptr_table_fetch(pTHX_ PTR_TBL_t *const tbl, const void *const sv)
10651 PTR_TBL_ENT_t const *const tblent = ptr_table_find(tbl, sv);
10653 PERL_ARGS_ASSERT_PTR_TABLE_FETCH;
10654 PERL_UNUSED_CONTEXT;
10656 return tblent ? tblent->newval : NULL;
10659 /* add a new entry to a pointer-mapping table */
10662 Perl_ptr_table_store(pTHX_ PTR_TBL_t *const tbl, const void *const oldsv, void *const newsv)
10664 PTR_TBL_ENT_t *tblent = ptr_table_find(tbl, oldsv);
10666 PERL_ARGS_ASSERT_PTR_TABLE_STORE;
10667 PERL_UNUSED_CONTEXT;
10670 tblent->newval = newsv;
10672 const UV entry = PTR_TABLE_HASH(oldsv) & tbl->tbl_max;
10674 new_body_inline(tblent, PTE_SVSLOT);
10676 tblent->oldval = oldsv;
10677 tblent->newval = newsv;
10678 tblent->next = tbl->tbl_ary[entry];
10679 tbl->tbl_ary[entry] = tblent;
10681 if (tblent->next && tbl->tbl_items > tbl->tbl_max)
10682 ptr_table_split(tbl);
10686 /* double the hash bucket size of an existing ptr table */
10689 Perl_ptr_table_split(pTHX_ PTR_TBL_t *const tbl)
10691 PTR_TBL_ENT_t **ary = tbl->tbl_ary;
10692 const UV oldsize = tbl->tbl_max + 1;
10693 UV newsize = oldsize * 2;
10696 PERL_ARGS_ASSERT_PTR_TABLE_SPLIT;
10697 PERL_UNUSED_CONTEXT;
10699 Renew(ary, newsize, PTR_TBL_ENT_t*);
10700 Zero(&ary[oldsize], newsize-oldsize, PTR_TBL_ENT_t*);
10701 tbl->tbl_max = --newsize;
10702 tbl->tbl_ary = ary;
10703 for (i=0; i < oldsize; i++, ary++) {
10704 PTR_TBL_ENT_t **curentp, **entp, *ent;
10707 curentp = ary + oldsize;
10708 for (entp = ary, ent = *ary; ent; ent = *entp) {
10709 if ((newsize & PTR_TABLE_HASH(ent->oldval)) != i) {
10711 ent->next = *curentp;
10721 /* remove all the entries from a ptr table */
10724 Perl_ptr_table_clear(pTHX_ PTR_TBL_t *const tbl)
10726 if (tbl && tbl->tbl_items) {
10727 register PTR_TBL_ENT_t * const * const array = tbl->tbl_ary;
10728 UV riter = tbl->tbl_max;
10731 PTR_TBL_ENT_t *entry = array[riter];
10734 PTR_TBL_ENT_t * const oentry = entry;
10735 entry = entry->next;
10740 tbl->tbl_items = 0;
10744 /* clear and free a ptr table */
10747 Perl_ptr_table_free(pTHX_ PTR_TBL_t *const tbl)
10752 ptr_table_clear(tbl);
10753 Safefree(tbl->tbl_ary);
10757 #if defined(USE_ITHREADS)
10760 Perl_rvpv_dup(pTHX_ SV *const dstr, const SV *const sstr, CLONE_PARAMS *const param)
10762 PERL_ARGS_ASSERT_RVPV_DUP;
10765 SvRV_set(dstr, SvWEAKREF(sstr)
10766 ? sv_dup(SvRV_const(sstr), param)
10767 : sv_dup_inc(SvRV_const(sstr), param));
10770 else if (SvPVX_const(sstr)) {
10771 /* Has something there */
10773 /* Normal PV - clone whole allocated space */
10774 SvPV_set(dstr, SAVEPVN(SvPVX_const(sstr), SvLEN(sstr)-1));
10775 if (SvREADONLY(sstr) && SvFAKE(sstr)) {
10776 /* Not that normal - actually sstr is copy on write.
10777 But we are a true, independant SV, so: */
10778 SvREADONLY_off(dstr);
10783 /* Special case - not normally malloced for some reason */
10784 if (isGV_with_GP(sstr)) {
10785 /* Don't need to do anything here. */
10787 else if ((SvREADONLY(sstr) && SvFAKE(sstr))) {
10788 /* A "shared" PV - clone it as "shared" PV */
10790 HEK_KEY(hek_dup(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)),
10794 /* Some other special case - random pointer */
10795 SvPV_set(dstr, (char *) SvPVX_const(sstr));
10800 /* Copy the NULL */
10801 SvPV_set(dstr, NULL);
10805 /* duplicate an SV of any type (including AV, HV etc) */
10808 Perl_sv_dup(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
10813 PERL_ARGS_ASSERT_SV_DUP;
10817 if (SvTYPE(sstr) == SVTYPEMASK) {
10818 #ifdef DEBUG_LEAKING_SCALARS_ABORT
10823 /* look for it in the table first */
10824 dstr = MUTABLE_SV(ptr_table_fetch(PL_ptr_table, sstr));
10828 if(param->flags & CLONEf_JOIN_IN) {
10829 /** We are joining here so we don't want do clone
10830 something that is bad **/
10831 if (SvTYPE(sstr) == SVt_PVHV) {
10832 const HEK * const hvname = HvNAME_HEK(sstr);
10834 /** don't clone stashes if they already exist **/
10835 return MUTABLE_SV(gv_stashpvn(HEK_KEY(hvname), HEK_LEN(hvname), 0));
10839 /* create anew and remember what it is */
10842 #ifdef DEBUG_LEAKING_SCALARS
10843 dstr->sv_debug_optype = sstr->sv_debug_optype;
10844 dstr->sv_debug_line = sstr->sv_debug_line;
10845 dstr->sv_debug_inpad = sstr->sv_debug_inpad;
10846 dstr->sv_debug_cloned = 1;
10847 dstr->sv_debug_file = savepv(sstr->sv_debug_file);
10850 ptr_table_store(PL_ptr_table, sstr, dstr);
10853 SvFLAGS(dstr) = SvFLAGS(sstr);
10854 SvFLAGS(dstr) &= ~SVf_OOK; /* don't propagate OOK hack */
10855 SvREFCNT(dstr) = 0; /* must be before any other dups! */
10858 if (SvANY(sstr) && PL_watch_pvx && SvPVX_const(sstr) == PL_watch_pvx)
10859 PerlIO_printf(Perl_debug_log, "watch at %p hit, found string \"%s\"\n",
10860 (void*)PL_watch_pvx, SvPVX_const(sstr));
10863 /* don't clone objects whose class has asked us not to */
10864 if (SvOBJECT(sstr) && ! (SvFLAGS(SvSTASH(sstr)) & SVphv_CLONEABLE)) {
10869 switch (SvTYPE(sstr)) {
10871 SvANY(dstr) = NULL;
10874 SvANY(dstr) = (XPVIV*)((char*)&(dstr->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
10876 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
10878 SvIV_set(dstr, SvIVX(sstr));
10882 SvANY(dstr) = new_XNV();
10883 SvNV_set(dstr, SvNVX(sstr));
10885 /* case SVt_BIND: */
10888 /* These are all the types that need complex bodies allocating. */
10890 const svtype sv_type = SvTYPE(sstr);
10891 const struct body_details *const sv_type_details
10892 = bodies_by_type + sv_type;
10896 Perl_croak(aTHX_ "Bizarre SvTYPE [%" IVdf "]", (IV)SvTYPE(sstr));
10911 assert(sv_type_details->body_size);
10912 if (sv_type_details->arena) {
10913 new_body_inline(new_body, sv_type);
10915 = (void*)((char*)new_body - sv_type_details->offset);
10917 new_body = new_NOARENA(sv_type_details);
10921 SvANY(dstr) = new_body;
10924 Copy(((char*)SvANY(sstr)) + sv_type_details->offset,
10925 ((char*)SvANY(dstr)) + sv_type_details->offset,
10926 sv_type_details->copy, char);
10928 Copy(((char*)SvANY(sstr)),
10929 ((char*)SvANY(dstr)),
10930 sv_type_details->body_size + sv_type_details->offset, char);
10933 if (sv_type != SVt_PVAV && sv_type != SVt_PVHV
10934 && !isGV_with_GP(dstr))
10935 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
10937 /* The Copy above means that all the source (unduplicated) pointers
10938 are now in the destination. We can check the flags and the
10939 pointers in either, but it's possible that there's less cache
10940 missing by always going for the destination.
10941 FIXME - instrument and check that assumption */
10942 if (sv_type >= SVt_PVMG) {
10943 if ((sv_type == SVt_PVMG) && SvPAD_OUR(dstr)) {
10944 SvOURSTASH_set(dstr, hv_dup_inc(SvOURSTASH(dstr), param));
10945 } else if (SvMAGIC(dstr))
10946 SvMAGIC_set(dstr, mg_dup(SvMAGIC(dstr), param));
10948 SvSTASH_set(dstr, hv_dup_inc(SvSTASH(dstr), param));
10951 /* The cast silences a GCC warning about unhandled types. */
10952 switch ((int)sv_type) {
10962 /* FIXME for plugins */
10963 re_dup_guts((REGEXP*) sstr, (REGEXP*) dstr, param);
10966 /* XXX LvTARGOFF sometimes holds PMOP* when DEBUGGING */
10967 if (LvTYPE(dstr) == 't') /* for tie: unrefcnted fake (SV**) */
10968 LvTARG(dstr) = dstr;
10969 else if (LvTYPE(dstr) == 'T') /* for tie: fake HE */
10970 LvTARG(dstr) = MUTABLE_SV(he_dup((HE*)LvTARG(dstr), 0, param));
10972 LvTARG(dstr) = sv_dup_inc(LvTARG(dstr), param);
10974 if(isGV_with_GP(sstr)) {
10975 if (GvNAME_HEK(dstr))
10976 GvNAME_HEK(dstr) = hek_dup(GvNAME_HEK(dstr), param);
10977 /* Don't call sv_add_backref here as it's going to be
10978 created as part of the magic cloning of the symbol
10980 /* Danger Will Robinson - GvGP(dstr) isn't initialised
10981 at the point of this comment. */
10982 GvSTASH(dstr) = hv_dup(GvSTASH(dstr), param);
10983 GvGP(dstr) = gp_dup(GvGP(sstr), param);
10984 (void)GpREFCNT_inc(GvGP(dstr));
10986 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
10989 IoIFP(dstr) = fp_dup(IoIFP(dstr), IoTYPE(dstr), param);
10990 if (IoOFP(dstr) == IoIFP(sstr))
10991 IoOFP(dstr) = IoIFP(dstr);
10993 IoOFP(dstr) = fp_dup(IoOFP(dstr), IoTYPE(dstr), param);
10994 /* PL_parser->rsfp_filters entries have fake IoDIRP() */
10995 if(IoFLAGS(dstr) & IOf_FAKE_DIRP) {
10996 /* I have no idea why fake dirp (rsfps)
10997 should be treated differently but otherwise
10998 we end up with leaks -- sky*/
10999 IoTOP_GV(dstr) = gv_dup_inc(IoTOP_GV(dstr), param);
11000 IoFMT_GV(dstr) = gv_dup_inc(IoFMT_GV(dstr), param);
11001 IoBOTTOM_GV(dstr) = gv_dup_inc(IoBOTTOM_GV(dstr), param);
11003 IoTOP_GV(dstr) = gv_dup(IoTOP_GV(dstr), param);
11004 IoFMT_GV(dstr) = gv_dup(IoFMT_GV(dstr), param);
11005 IoBOTTOM_GV(dstr) = gv_dup(IoBOTTOM_GV(dstr), param);
11006 if (IoDIRP(dstr)) {
11007 IoDIRP(dstr) = dirp_dup(IoDIRP(dstr));
11010 /* IoDIRP(dstr) is already a copy of IoDIRP(sstr) */
11013 IoTOP_NAME(dstr) = SAVEPV(IoTOP_NAME(dstr));
11014 IoFMT_NAME(dstr) = SAVEPV(IoFMT_NAME(dstr));
11015 IoBOTTOM_NAME(dstr) = SAVEPV(IoBOTTOM_NAME(dstr));
11018 /* avoid cloning an empty array */
11019 if (AvARRAY((const AV *)sstr) && AvFILLp((const AV *)sstr) >= 0) {
11020 SV **dst_ary, **src_ary;
11021 SSize_t items = AvFILLp((const AV *)sstr) + 1;
11023 src_ary = AvARRAY((const AV *)sstr);
11024 Newxz(dst_ary, AvMAX((const AV *)sstr)+1, SV*);
11025 ptr_table_store(PL_ptr_table, src_ary, dst_ary);
11026 AvARRAY(MUTABLE_AV(dstr)) = dst_ary;
11027 AvALLOC((const AV *)dstr) = dst_ary;
11028 if (AvREAL((const AV *)sstr)) {
11029 while (items-- > 0)
11030 *dst_ary++ = sv_dup_inc(*src_ary++, param);
11033 while (items-- > 0)
11034 *dst_ary++ = sv_dup(*src_ary++, param);
11036 items = AvMAX((const AV *)sstr) - AvFILLp((const AV *)sstr);
11037 while (items-- > 0) {
11038 *dst_ary++ = &PL_sv_undef;
11042 AvARRAY(MUTABLE_AV(dstr)) = NULL;
11043 AvALLOC((const AV *)dstr) = (SV**)NULL;
11044 AvMAX( (const AV *)dstr) = -1;
11045 AvFILLp((const AV *)dstr) = -1;
11049 if (HvARRAY((const HV *)sstr)) {
11051 const bool sharekeys = !!HvSHAREKEYS(sstr);
11052 XPVHV * const dxhv = (XPVHV*)SvANY(dstr);
11053 XPVHV * const sxhv = (XPVHV*)SvANY(sstr);
11055 Newx(darray, PERL_HV_ARRAY_ALLOC_BYTES(dxhv->xhv_max+1)
11056 + (SvOOK(sstr) ? sizeof(struct xpvhv_aux) : 0),
11058 HvARRAY(dstr) = (HE**)darray;
11059 while (i <= sxhv->xhv_max) {
11060 const HE * const source = HvARRAY(sstr)[i];
11061 HvARRAY(dstr)[i] = source
11062 ? he_dup(source, sharekeys, param) : 0;
11067 const struct xpvhv_aux * const saux = HvAUX(sstr);
11068 struct xpvhv_aux * const daux = HvAUX(dstr);
11069 /* This flag isn't copied. */
11070 /* SvOOK_on(hv) attacks the IV flags. */
11071 SvFLAGS(dstr) |= SVf_OOK;
11073 hvname = saux->xhv_name;
11074 daux->xhv_name = hvname ? hek_dup(hvname, param) : hvname;
11076 daux->xhv_riter = saux->xhv_riter;
11077 daux->xhv_eiter = saux->xhv_eiter
11078 ? he_dup(saux->xhv_eiter,
11079 (bool)!!HvSHAREKEYS(sstr), param) : 0;
11080 /* backref array needs refcnt=2; see sv_add_backref */
11081 daux->xhv_backreferences =
11082 saux->xhv_backreferences
11083 ? MUTABLE_AV(SvREFCNT_inc(
11084 sv_dup_inc((const SV *)saux->xhv_backreferences, param)))
11087 daux->xhv_mro_meta = saux->xhv_mro_meta
11088 ? mro_meta_dup(saux->xhv_mro_meta, param)
11091 /* Record stashes for possible cloning in Perl_clone(). */
11093 av_push(param->stashes, dstr);
11097 HvARRAY(MUTABLE_HV(dstr)) = NULL;
11100 if (!(param->flags & CLONEf_COPY_STACKS)) {
11104 /* NOTE: not refcounted */
11105 CvSTASH(dstr) = hv_dup(CvSTASH(dstr), param);
11107 if (!CvISXSUB(dstr))
11108 CvROOT(dstr) = OpREFCNT_inc(CvROOT(dstr));
11110 if (CvCONST(dstr) && CvISXSUB(dstr)) {
11111 CvXSUBANY(dstr).any_ptr =
11112 sv_dup_inc((const SV *)CvXSUBANY(dstr).any_ptr, param);
11114 /* don't dup if copying back - CvGV isn't refcounted, so the
11115 * duped GV may never be freed. A bit of a hack! DAPM */
11116 CvGV(dstr) = (param->flags & CLONEf_JOIN_IN) ?
11117 NULL : gv_dup(CvGV(dstr), param) ;
11118 PAD_DUP(CvPADLIST(dstr), CvPADLIST(sstr), param);
11120 CvWEAKOUTSIDE(sstr)
11121 ? cv_dup( CvOUTSIDE(dstr), param)
11122 : cv_dup_inc(CvOUTSIDE(dstr), param);
11123 if (!CvISXSUB(dstr))
11124 CvFILE(dstr) = SAVEPV(CvFILE(dstr));
11130 if (SvOBJECT(dstr) && SvTYPE(dstr) != SVt_PVIO)
11136 /* duplicate a context */
11139 Perl_cx_dup(pTHX_ PERL_CONTEXT *cxs, I32 ix, I32 max, CLONE_PARAMS* param)
11141 PERL_CONTEXT *ncxs;
11143 PERL_ARGS_ASSERT_CX_DUP;
11146 return (PERL_CONTEXT*)NULL;
11148 /* look for it in the table first */
11149 ncxs = (PERL_CONTEXT*)ptr_table_fetch(PL_ptr_table, cxs);
11153 /* create anew and remember what it is */
11154 Newx(ncxs, max + 1, PERL_CONTEXT);
11155 ptr_table_store(PL_ptr_table, cxs, ncxs);
11156 Copy(cxs, ncxs, max + 1, PERL_CONTEXT);
11159 PERL_CONTEXT * const ncx = &ncxs[ix];
11160 if (CxTYPE(ncx) == CXt_SUBST) {
11161 Perl_croak(aTHX_ "Cloning substitution context is unimplemented");
11164 switch (CxTYPE(ncx)) {
11166 ncx->blk_sub.cv = (ncx->blk_sub.olddepth == 0
11167 ? cv_dup_inc(ncx->blk_sub.cv, param)
11168 : cv_dup(ncx->blk_sub.cv,param));
11169 ncx->blk_sub.argarray = (CxHASARGS(ncx)
11170 ? av_dup_inc(ncx->blk_sub.argarray,
11173 ncx->blk_sub.savearray = av_dup_inc(ncx->blk_sub.savearray,
11175 ncx->blk_sub.oldcomppad = (PAD*)ptr_table_fetch(PL_ptr_table,
11176 ncx->blk_sub.oldcomppad);
11179 ncx->blk_eval.old_namesv = sv_dup_inc(ncx->blk_eval.old_namesv,
11181 ncx->blk_eval.cur_text = sv_dup(ncx->blk_eval.cur_text, param);
11183 case CXt_LOOP_LAZYSV:
11184 ncx->blk_loop.state_u.lazysv.end
11185 = sv_dup_inc(ncx->blk_loop.state_u.lazysv.end, param);
11186 /* We are taking advantage of av_dup_inc and sv_dup_inc
11187 actually being the same function, and order equivalance of
11189 We can assert the later [but only at run time :-(] */
11190 assert ((void *) &ncx->blk_loop.state_u.ary.ary ==
11191 (void *) &ncx->blk_loop.state_u.lazysv.cur);
11193 ncx->blk_loop.state_u.ary.ary
11194 = av_dup_inc(ncx->blk_loop.state_u.ary.ary, param);
11195 case CXt_LOOP_LAZYIV:
11196 case CXt_LOOP_PLAIN:
11197 if (CxPADLOOP(ncx)) {
11198 ncx->blk_loop.oldcomppad
11199 = (PAD*)ptr_table_fetch(PL_ptr_table,
11200 ncx->blk_loop.oldcomppad);
11202 ncx->blk_loop.oldcomppad
11203 = (PAD*)gv_dup((const GV *)ncx->blk_loop.oldcomppad,
11208 ncx->blk_format.cv = cv_dup(ncx->blk_format.cv, param);
11209 ncx->blk_format.gv = gv_dup(ncx->blk_format.gv, param);
11210 ncx->blk_format.dfoutgv = gv_dup_inc(ncx->blk_format.dfoutgv,
11223 /* duplicate a stack info structure */
11226 Perl_si_dup(pTHX_ PERL_SI *si, CLONE_PARAMS* param)
11230 PERL_ARGS_ASSERT_SI_DUP;
11233 return (PERL_SI*)NULL;
11235 /* look for it in the table first */
11236 nsi = (PERL_SI*)ptr_table_fetch(PL_ptr_table, si);
11240 /* create anew and remember what it is */
11241 Newxz(nsi, 1, PERL_SI);
11242 ptr_table_store(PL_ptr_table, si, nsi);
11244 nsi->si_stack = av_dup_inc(si->si_stack, param);
11245 nsi->si_cxix = si->si_cxix;
11246 nsi->si_cxmax = si->si_cxmax;
11247 nsi->si_cxstack = cx_dup(si->si_cxstack, si->si_cxix, si->si_cxmax, param);
11248 nsi->si_type = si->si_type;
11249 nsi->si_prev = si_dup(si->si_prev, param);
11250 nsi->si_next = si_dup(si->si_next, param);
11251 nsi->si_markoff = si->si_markoff;
11256 #define POPINT(ss,ix) ((ss)[--(ix)].any_i32)
11257 #define TOPINT(ss,ix) ((ss)[ix].any_i32)
11258 #define POPLONG(ss,ix) ((ss)[--(ix)].any_long)
11259 #define TOPLONG(ss,ix) ((ss)[ix].any_long)
11260 #define POPIV(ss,ix) ((ss)[--(ix)].any_iv)
11261 #define TOPIV(ss,ix) ((ss)[ix].any_iv)
11262 #define POPBOOL(ss,ix) ((ss)[--(ix)].any_bool)
11263 #define TOPBOOL(ss,ix) ((ss)[ix].any_bool)
11264 #define POPPTR(ss,ix) ((ss)[--(ix)].any_ptr)
11265 #define TOPPTR(ss,ix) ((ss)[ix].any_ptr)
11266 #define POPDPTR(ss,ix) ((ss)[--(ix)].any_dptr)
11267 #define TOPDPTR(ss,ix) ((ss)[ix].any_dptr)
11268 #define POPDXPTR(ss,ix) ((ss)[--(ix)].any_dxptr)
11269 #define TOPDXPTR(ss,ix) ((ss)[ix].any_dxptr)
11272 #define pv_dup_inc(p) SAVEPV(p)
11273 #define pv_dup(p) SAVEPV(p)
11274 #define svp_dup_inc(p,pp) any_dup(p,pp)
11276 /* map any object to the new equivent - either something in the
11277 * ptr table, or something in the interpreter structure
11281 Perl_any_dup(pTHX_ void *v, const PerlInterpreter *proto_perl)
11285 PERL_ARGS_ASSERT_ANY_DUP;
11288 return (void*)NULL;
11290 /* look for it in the table first */
11291 ret = ptr_table_fetch(PL_ptr_table, v);
11295 /* see if it is part of the interpreter structure */
11296 if (v >= (void*)proto_perl && v < (void*)(proto_perl+1))
11297 ret = (void*)(((char*)aTHX) + (((char*)v) - (char*)proto_perl));
11305 /* duplicate the save stack */
11308 Perl_ss_dup(pTHX_ PerlInterpreter *proto_perl, CLONE_PARAMS* param)
11311 ANY * const ss = proto_perl->Isavestack;
11312 const I32 max = proto_perl->Isavestack_max;
11313 I32 ix = proto_perl->Isavestack_ix;
11326 void (*dptr) (void*);
11327 void (*dxptr) (pTHX_ void*);
11329 PERL_ARGS_ASSERT_SS_DUP;
11331 Newxz(nss, max, ANY);
11334 const I32 type = POPINT(ss,ix);
11335 TOPINT(nss,ix) = type;
11337 case SAVEt_HELEM: /* hash element */
11338 sv = (const SV *)POPPTR(ss,ix);
11339 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11341 case SAVEt_ITEM: /* normal string */
11342 case SAVEt_SV: /* scalar reference */
11343 sv = (const SV *)POPPTR(ss,ix);
11344 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11347 case SAVEt_MORTALIZESV:
11348 sv = (const SV *)POPPTR(ss,ix);
11349 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11351 case SAVEt_SHARED_PVREF: /* char* in shared space */
11352 c = (char*)POPPTR(ss,ix);
11353 TOPPTR(nss,ix) = savesharedpv(c);
11354 ptr = POPPTR(ss,ix);
11355 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11357 case SAVEt_GENERIC_SVREF: /* generic sv */
11358 case SAVEt_SVREF: /* scalar reference */
11359 sv = (const SV *)POPPTR(ss,ix);
11360 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11361 ptr = POPPTR(ss,ix);
11362 TOPPTR(nss,ix) = svp_dup_inc((SV**)ptr, proto_perl);/* XXXXX */
11364 case SAVEt_HV: /* hash reference */
11365 case SAVEt_AV: /* array reference */
11366 sv = (const SV *) POPPTR(ss,ix);
11367 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11369 case SAVEt_COMPPAD:
11371 sv = (const SV *) POPPTR(ss,ix);
11372 TOPPTR(nss,ix) = sv_dup(sv, param);
11374 case SAVEt_INT: /* int reference */
11375 ptr = POPPTR(ss,ix);
11376 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11377 intval = (int)POPINT(ss,ix);
11378 TOPINT(nss,ix) = intval;
11380 case SAVEt_LONG: /* long reference */
11381 ptr = POPPTR(ss,ix);
11382 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11384 case SAVEt_CLEARSV:
11385 longval = (long)POPLONG(ss,ix);
11386 TOPLONG(nss,ix) = longval;
11388 case SAVEt_I32: /* I32 reference */
11389 case SAVEt_I16: /* I16 reference */
11390 case SAVEt_I8: /* I8 reference */
11391 case SAVEt_COP_ARYBASE: /* call CopARYBASE_set */
11392 ptr = POPPTR(ss,ix);
11393 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11395 TOPINT(nss,ix) = i;
11397 case SAVEt_IV: /* IV reference */
11398 ptr = POPPTR(ss,ix);
11399 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11401 TOPIV(nss,ix) = iv;
11403 case SAVEt_HPTR: /* HV* reference */
11404 case SAVEt_APTR: /* AV* reference */
11405 case SAVEt_SPTR: /* SV* reference */
11406 ptr = POPPTR(ss,ix);
11407 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11408 sv = (const SV *)POPPTR(ss,ix);
11409 TOPPTR(nss,ix) = sv_dup(sv, param);
11411 case SAVEt_VPTR: /* random* reference */
11412 ptr = POPPTR(ss,ix);
11413 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11414 ptr = POPPTR(ss,ix);
11415 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11417 case SAVEt_GENERIC_PVREF: /* generic char* */
11418 case SAVEt_PPTR: /* char* reference */
11419 ptr = POPPTR(ss,ix);
11420 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11421 c = (char*)POPPTR(ss,ix);
11422 TOPPTR(nss,ix) = pv_dup(c);
11424 case SAVEt_GP: /* scalar reference */
11425 gp = (GP*)POPPTR(ss,ix);
11426 TOPPTR(nss,ix) = gp = gp_dup(gp, param);
11427 (void)GpREFCNT_inc(gp);
11428 gv = (const GV *)POPPTR(ss,ix);
11429 TOPPTR(nss,ix) = gv_dup_inc(gv, param);
11432 ptr = POPPTR(ss,ix);
11433 if (ptr && (((OP*)ptr)->op_private & OPpREFCOUNTED)) {
11434 /* these are assumed to be refcounted properly */
11436 switch (((OP*)ptr)->op_type) {
11438 case OP_LEAVESUBLV:
11442 case OP_LEAVEWRITE:
11443 TOPPTR(nss,ix) = ptr;
11446 (void) OpREFCNT_inc(o);
11450 TOPPTR(nss,ix) = NULL;
11455 TOPPTR(nss,ix) = NULL;
11458 hv = (const HV *)POPPTR(ss,ix);
11459 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
11461 TOPINT(nss,ix) = i;
11464 c = (char*)POPPTR(ss,ix);
11465 TOPPTR(nss,ix) = pv_dup_inc(c);
11467 case SAVEt_STACK_POS: /* Position on Perl stack */
11469 TOPINT(nss,ix) = i;
11471 case SAVEt_DESTRUCTOR:
11472 ptr = POPPTR(ss,ix);
11473 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
11474 dptr = POPDPTR(ss,ix);
11475 TOPDPTR(nss,ix) = DPTR2FPTR(void (*)(void*),
11476 any_dup(FPTR2DPTR(void *, dptr),
11479 case SAVEt_DESTRUCTOR_X:
11480 ptr = POPPTR(ss,ix);
11481 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
11482 dxptr = POPDXPTR(ss,ix);
11483 TOPDXPTR(nss,ix) = DPTR2FPTR(void (*)(pTHX_ void*),
11484 any_dup(FPTR2DPTR(void *, dxptr),
11487 case SAVEt_REGCONTEXT:
11490 TOPINT(nss,ix) = i;
11493 case SAVEt_AELEM: /* array element */
11494 sv = (const SV *)POPPTR(ss,ix);
11495 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11497 TOPINT(nss,ix) = i;
11498 av = (const AV *)POPPTR(ss,ix);
11499 TOPPTR(nss,ix) = av_dup_inc(av, param);
11502 ptr = POPPTR(ss,ix);
11503 TOPPTR(nss,ix) = ptr;
11506 ptr = POPPTR(ss,ix);
11509 ((struct refcounted_he *)ptr)->refcounted_he_refcnt++;
11510 HINTS_REFCNT_UNLOCK;
11512 TOPPTR(nss,ix) = ptr;
11514 TOPINT(nss,ix) = i;
11515 if (i & HINT_LOCALIZE_HH) {
11516 hv = (const HV *)POPPTR(ss,ix);
11517 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
11520 case SAVEt_PADSV_AND_MORTALIZE:
11521 longval = (long)POPLONG(ss,ix);
11522 TOPLONG(nss,ix) = longval;
11523 ptr = POPPTR(ss,ix);
11524 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11525 sv = (const SV *)POPPTR(ss,ix);
11526 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11529 ptr = POPPTR(ss,ix);
11530 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11531 longval = (long)POPBOOL(ss,ix);
11532 TOPBOOL(nss,ix) = (bool)longval;
11534 case SAVEt_SET_SVFLAGS:
11536 TOPINT(nss,ix) = i;
11538 TOPINT(nss,ix) = i;
11539 sv = (const SV *)POPPTR(ss,ix);
11540 TOPPTR(nss,ix) = sv_dup(sv, param);
11542 case SAVEt_RE_STATE:
11544 const struct re_save_state *const old_state
11545 = (struct re_save_state *)
11546 (ss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
11547 struct re_save_state *const new_state
11548 = (struct re_save_state *)
11549 (nss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
11551 Copy(old_state, new_state, 1, struct re_save_state);
11552 ix -= SAVESTACK_ALLOC_FOR_RE_SAVE_STATE;
11554 new_state->re_state_bostr
11555 = pv_dup(old_state->re_state_bostr);
11556 new_state->re_state_reginput
11557 = pv_dup(old_state->re_state_reginput);
11558 new_state->re_state_regeol
11559 = pv_dup(old_state->re_state_regeol);
11560 new_state->re_state_regoffs
11561 = (regexp_paren_pair*)
11562 any_dup(old_state->re_state_regoffs, proto_perl);
11563 new_state->re_state_reglastparen
11564 = (U32*) any_dup(old_state->re_state_reglastparen,
11566 new_state->re_state_reglastcloseparen
11567 = (U32*)any_dup(old_state->re_state_reglastcloseparen,
11569 /* XXX This just has to be broken. The old save_re_context
11570 code did SAVEGENERICPV(PL_reg_start_tmp);
11571 PL_reg_start_tmp is char **.
11572 Look above to what the dup code does for
11573 SAVEt_GENERIC_PVREF
11574 It can never have worked.
11575 So this is merely a faithful copy of the exiting bug: */
11576 new_state->re_state_reg_start_tmp
11577 = (char **) pv_dup((char *)
11578 old_state->re_state_reg_start_tmp);
11579 /* I assume that it only ever "worked" because no-one called
11580 (pseudo)fork while the regexp engine had re-entered itself.
11582 #ifdef PERL_OLD_COPY_ON_WRITE
11583 new_state->re_state_nrs
11584 = sv_dup(old_state->re_state_nrs, param);
11586 new_state->re_state_reg_magic
11587 = (MAGIC*) any_dup(old_state->re_state_reg_magic,
11589 new_state->re_state_reg_oldcurpm
11590 = (PMOP*) any_dup(old_state->re_state_reg_oldcurpm,
11592 new_state->re_state_reg_curpm
11593 = (PMOP*) any_dup(old_state->re_state_reg_curpm,
11595 new_state->re_state_reg_oldsaved
11596 = pv_dup(old_state->re_state_reg_oldsaved);
11597 new_state->re_state_reg_poscache
11598 = pv_dup(old_state->re_state_reg_poscache);
11599 new_state->re_state_reg_starttry
11600 = pv_dup(old_state->re_state_reg_starttry);
11603 case SAVEt_COMPILE_WARNINGS:
11604 ptr = POPPTR(ss,ix);
11605 TOPPTR(nss,ix) = DUP_WARNINGS((STRLEN*)ptr);
11608 ptr = POPPTR(ss,ix);
11609 TOPPTR(nss,ix) = parser_dup((const yy_parser*)ptr, param);
11613 "panic: ss_dup inconsistency (%"IVdf")", (IV) type);
11621 /* if sv is a stash, call $class->CLONE_SKIP(), and set the SVphv_CLONEABLE
11622 * flag to the result. This is done for each stash before cloning starts,
11623 * so we know which stashes want their objects cloned */
11626 do_mark_cloneable_stash(pTHX_ SV *const sv)
11628 const HEK * const hvname = HvNAME_HEK((const HV *)sv);
11630 GV* const cloner = gv_fetchmethod_autoload(MUTABLE_HV(sv), "CLONE_SKIP", 0);
11631 SvFLAGS(sv) |= SVphv_CLONEABLE; /* clone objects by default */
11632 if (cloner && GvCV(cloner)) {
11639 mXPUSHs(newSVhek(hvname));
11641 call_sv(MUTABLE_SV(GvCV(cloner)), G_SCALAR);
11648 SvFLAGS(sv) &= ~SVphv_CLONEABLE;
11656 =for apidoc perl_clone
11658 Create and return a new interpreter by cloning the current one.
11660 perl_clone takes these flags as parameters:
11662 CLONEf_COPY_STACKS - is used to, well, copy the stacks also,
11663 without it we only clone the data and zero the stacks,
11664 with it we copy the stacks and the new perl interpreter is
11665 ready to run at the exact same point as the previous one.
11666 The pseudo-fork code uses COPY_STACKS while the
11667 threads->create doesn't.
11669 CLONEf_KEEP_PTR_TABLE
11670 perl_clone keeps a ptr_table with the pointer of the old
11671 variable as a key and the new variable as a value,
11672 this allows it to check if something has been cloned and not
11673 clone it again but rather just use the value and increase the
11674 refcount. If KEEP_PTR_TABLE is not set then perl_clone will kill
11675 the ptr_table using the function
11676 C<ptr_table_free(PL_ptr_table); PL_ptr_table = NULL;>,
11677 reason to keep it around is if you want to dup some of your own
11678 variable who are outside the graph perl scans, example of this
11679 code is in threads.xs create
11682 This is a win32 thing, it is ignored on unix, it tells perls
11683 win32host code (which is c++) to clone itself, this is needed on
11684 win32 if you want to run two threads at the same time,
11685 if you just want to do some stuff in a separate perl interpreter
11686 and then throw it away and return to the original one,
11687 you don't need to do anything.
11692 /* XXX the above needs expanding by someone who actually understands it ! */
11693 EXTERN_C PerlInterpreter *
11694 perl_clone_host(PerlInterpreter* proto_perl, UV flags);
11697 perl_clone(PerlInterpreter *proto_perl, UV flags)
11700 #ifdef PERL_IMPLICIT_SYS
11702 PERL_ARGS_ASSERT_PERL_CLONE;
11704 /* perlhost.h so we need to call into it
11705 to clone the host, CPerlHost should have a c interface, sky */
11707 if (flags & CLONEf_CLONE_HOST) {
11708 return perl_clone_host(proto_perl,flags);
11710 return perl_clone_using(proto_perl, flags,
11712 proto_perl->IMemShared,
11713 proto_perl->IMemParse,
11715 proto_perl->IStdIO,
11719 proto_perl->IProc);
11723 perl_clone_using(PerlInterpreter *proto_perl, UV flags,
11724 struct IPerlMem* ipM, struct IPerlMem* ipMS,
11725 struct IPerlMem* ipMP, struct IPerlEnv* ipE,
11726 struct IPerlStdIO* ipStd, struct IPerlLIO* ipLIO,
11727 struct IPerlDir* ipD, struct IPerlSock* ipS,
11728 struct IPerlProc* ipP)
11730 /* XXX many of the string copies here can be optimized if they're
11731 * constants; they need to be allocated as common memory and just
11732 * their pointers copied. */
11735 CLONE_PARAMS clone_params;
11736 CLONE_PARAMS* const param = &clone_params;
11738 PerlInterpreter * const my_perl = (PerlInterpreter*)(*ipM->pMalloc)(ipM, sizeof(PerlInterpreter));
11740 PERL_ARGS_ASSERT_PERL_CLONE_USING;
11742 /* for each stash, determine whether its objects should be cloned */
11743 S_visit(proto_perl, do_mark_cloneable_stash, SVt_PVHV, SVTYPEMASK);
11744 PERL_SET_THX(my_perl);
11747 PoisonNew(my_perl, 1, PerlInterpreter);
11753 PL_savestack_ix = 0;
11754 PL_savestack_max = -1;
11755 PL_sig_pending = 0;
11757 Zero(&PL_debug_pad, 1, struct perl_debug_pad);
11758 # else /* !DEBUGGING */
11759 Zero(my_perl, 1, PerlInterpreter);
11760 # endif /* DEBUGGING */
11762 /* host pointers */
11764 PL_MemShared = ipMS;
11765 PL_MemParse = ipMP;
11772 #else /* !PERL_IMPLICIT_SYS */
11774 CLONE_PARAMS clone_params;
11775 CLONE_PARAMS* param = &clone_params;
11776 PerlInterpreter * const my_perl = (PerlInterpreter*)PerlMem_malloc(sizeof(PerlInterpreter));
11778 PERL_ARGS_ASSERT_PERL_CLONE;
11780 /* for each stash, determine whether its objects should be cloned */
11781 S_visit(proto_perl, do_mark_cloneable_stash, SVt_PVHV, SVTYPEMASK);
11782 PERL_SET_THX(my_perl);
11785 PoisonNew(my_perl, 1, PerlInterpreter);
11791 PL_savestack_ix = 0;
11792 PL_savestack_max = -1;
11793 PL_sig_pending = 0;
11795 Zero(&PL_debug_pad, 1, struct perl_debug_pad);
11796 # else /* !DEBUGGING */
11797 Zero(my_perl, 1, PerlInterpreter);
11798 # endif /* DEBUGGING */
11799 #endif /* PERL_IMPLICIT_SYS */
11800 param->flags = flags;
11801 param->proto_perl = proto_perl;
11803 INIT_TRACK_MEMPOOL(my_perl->Imemory_debug_header, my_perl);
11805 PL_body_arenas = NULL;
11806 Zero(&PL_body_roots, 1, PL_body_roots);
11808 PL_nice_chunk = NULL;
11809 PL_nice_chunk_size = 0;
11811 PL_sv_objcount = 0;
11813 PL_sv_arenaroot = NULL;
11815 PL_debug = proto_perl->Idebug;
11817 PL_hash_seed = proto_perl->Ihash_seed;
11818 PL_rehash_seed = proto_perl->Irehash_seed;
11820 #ifdef USE_REENTRANT_API
11821 /* XXX: things like -Dm will segfault here in perlio, but doing
11822 * PERL_SET_CONTEXT(proto_perl);
11823 * breaks too many other things
11825 Perl_reentrant_init(aTHX);
11828 /* create SV map for pointer relocation */
11829 PL_ptr_table = ptr_table_new();
11831 /* initialize these special pointers as early as possible */
11832 SvANY(&PL_sv_undef) = NULL;
11833 SvREFCNT(&PL_sv_undef) = (~(U32)0)/2;
11834 SvFLAGS(&PL_sv_undef) = SVf_READONLY|SVt_NULL;
11835 ptr_table_store(PL_ptr_table, &proto_perl->Isv_undef, &PL_sv_undef);
11837 SvANY(&PL_sv_no) = new_XPVNV();
11838 SvREFCNT(&PL_sv_no) = (~(U32)0)/2;
11839 SvFLAGS(&PL_sv_no) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
11840 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
11841 SvPV_set(&PL_sv_no, savepvn(PL_No, 0));
11842 SvCUR_set(&PL_sv_no, 0);
11843 SvLEN_set(&PL_sv_no, 1);
11844 SvIV_set(&PL_sv_no, 0);
11845 SvNV_set(&PL_sv_no, 0);
11846 ptr_table_store(PL_ptr_table, &proto_perl->Isv_no, &PL_sv_no);
11848 SvANY(&PL_sv_yes) = new_XPVNV();
11849 SvREFCNT(&PL_sv_yes) = (~(U32)0)/2;
11850 SvFLAGS(&PL_sv_yes) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
11851 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
11852 SvPV_set(&PL_sv_yes, savepvn(PL_Yes, 1));
11853 SvCUR_set(&PL_sv_yes, 1);
11854 SvLEN_set(&PL_sv_yes, 2);
11855 SvIV_set(&PL_sv_yes, 1);
11856 SvNV_set(&PL_sv_yes, 1);
11857 ptr_table_store(PL_ptr_table, &proto_perl->Isv_yes, &PL_sv_yes);
11859 /* create (a non-shared!) shared string table */
11860 PL_strtab = newHV();
11861 HvSHAREKEYS_off(PL_strtab);
11862 hv_ksplit(PL_strtab, HvTOTALKEYS(proto_perl->Istrtab));
11863 ptr_table_store(PL_ptr_table, proto_perl->Istrtab, PL_strtab);
11865 PL_compiling = proto_perl->Icompiling;
11867 /* These two PVs will be free'd special way so must set them same way op.c does */
11868 PL_compiling.cop_stashpv = savesharedpv(PL_compiling.cop_stashpv);
11869 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_stashpv, PL_compiling.cop_stashpv);
11871 PL_compiling.cop_file = savesharedpv(PL_compiling.cop_file);
11872 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_file, PL_compiling.cop_file);
11874 ptr_table_store(PL_ptr_table, &proto_perl->Icompiling, &PL_compiling);
11875 PL_compiling.cop_warnings = DUP_WARNINGS(PL_compiling.cop_warnings);
11876 if (PL_compiling.cop_hints_hash) {
11878 PL_compiling.cop_hints_hash->refcounted_he_refcnt++;
11879 HINTS_REFCNT_UNLOCK;
11881 PL_curcop = (COP*)any_dup(proto_perl->Icurcop, proto_perl);
11882 #ifdef PERL_DEBUG_READONLY_OPS
11887 /* pseudo environmental stuff */
11888 PL_origargc = proto_perl->Iorigargc;
11889 PL_origargv = proto_perl->Iorigargv;
11891 param->stashes = newAV(); /* Setup array of objects to call clone on */
11893 /* Set tainting stuff before PerlIO_debug can possibly get called */
11894 PL_tainting = proto_perl->Itainting;
11895 PL_taint_warn = proto_perl->Itaint_warn;
11897 #ifdef PERLIO_LAYERS
11898 /* Clone PerlIO tables as soon as we can handle general xx_dup() */
11899 PerlIO_clone(aTHX_ proto_perl, param);
11902 PL_envgv = gv_dup(proto_perl->Ienvgv, param);
11903 PL_incgv = gv_dup(proto_perl->Iincgv, param);
11904 PL_hintgv = gv_dup(proto_perl->Ihintgv, param);
11905 PL_origfilename = SAVEPV(proto_perl->Iorigfilename);
11906 PL_diehook = sv_dup_inc(proto_perl->Idiehook, param);
11907 PL_warnhook = sv_dup_inc(proto_perl->Iwarnhook, param);
11910 PL_minus_c = proto_perl->Iminus_c;
11911 PL_patchlevel = sv_dup_inc(proto_perl->Ipatchlevel, param);
11912 PL_localpatches = proto_perl->Ilocalpatches;
11913 PL_splitstr = proto_perl->Isplitstr;
11914 PL_minus_n = proto_perl->Iminus_n;
11915 PL_minus_p = proto_perl->Iminus_p;
11916 PL_minus_l = proto_perl->Iminus_l;
11917 PL_minus_a = proto_perl->Iminus_a;
11918 PL_minus_E = proto_perl->Iminus_E;
11919 PL_minus_F = proto_perl->Iminus_F;
11920 PL_doswitches = proto_perl->Idoswitches;
11921 PL_dowarn = proto_perl->Idowarn;
11922 PL_doextract = proto_perl->Idoextract;
11923 PL_sawampersand = proto_perl->Isawampersand;
11924 PL_unsafe = proto_perl->Iunsafe;
11925 PL_inplace = SAVEPV(proto_perl->Iinplace);
11926 PL_e_script = sv_dup_inc(proto_perl->Ie_script, param);
11927 PL_perldb = proto_perl->Iperldb;
11928 PL_perl_destruct_level = proto_perl->Iperl_destruct_level;
11929 PL_exit_flags = proto_perl->Iexit_flags;
11931 /* magical thingies */
11932 /* XXX time(&PL_basetime) when asked for? */
11933 PL_basetime = proto_perl->Ibasetime;
11934 PL_formfeed = sv_dup(proto_perl->Iformfeed, param);
11936 PL_maxsysfd = proto_perl->Imaxsysfd;
11937 PL_statusvalue = proto_perl->Istatusvalue;
11939 PL_statusvalue_vms = proto_perl->Istatusvalue_vms;
11941 PL_statusvalue_posix = proto_perl->Istatusvalue_posix;
11943 PL_encoding = sv_dup(proto_perl->Iencoding, param);
11945 sv_setpvs(PERL_DEBUG_PAD(0), ""); /* For regex debugging. */
11946 sv_setpvs(PERL_DEBUG_PAD(1), ""); /* ext/re needs these */
11947 sv_setpvs(PERL_DEBUG_PAD(2), ""); /* even without DEBUGGING. */
11950 /* RE engine related */
11951 Zero(&PL_reg_state, 1, struct re_save_state);
11952 PL_reginterp_cnt = 0;
11953 PL_regmatch_slab = NULL;
11955 /* Clone the regex array */
11956 /* ORANGE FIXME for plugins, probably in the SV dup code.
11957 newSViv(PTR2IV(CALLREGDUPE(
11958 INT2PTR(REGEXP *, SvIVX(regex)), param))))
11960 PL_regex_padav = av_dup_inc(proto_perl->Iregex_padav, param);
11961 PL_regex_pad = AvARRAY(PL_regex_padav);
11963 /* shortcuts to various I/O objects */
11964 PL_ofsgv = gv_dup(proto_perl->Iofsgv, param);
11965 PL_stdingv = gv_dup(proto_perl->Istdingv, param);
11966 PL_stderrgv = gv_dup(proto_perl->Istderrgv, param);
11967 PL_defgv = gv_dup(proto_perl->Idefgv, param);
11968 PL_argvgv = gv_dup(proto_perl->Iargvgv, param);
11969 PL_argvoutgv = gv_dup(proto_perl->Iargvoutgv, param);
11970 PL_argvout_stack = av_dup_inc(proto_perl->Iargvout_stack, param);
11972 /* shortcuts to regexp stuff */
11973 PL_replgv = gv_dup(proto_perl->Ireplgv, param);
11975 /* shortcuts to misc objects */
11976 PL_errgv = gv_dup(proto_perl->Ierrgv, param);
11978 /* shortcuts to debugging objects */
11979 PL_DBgv = gv_dup(proto_perl->IDBgv, param);
11980 PL_DBline = gv_dup(proto_perl->IDBline, param);
11981 PL_DBsub = gv_dup(proto_perl->IDBsub, param);
11982 PL_DBsingle = sv_dup(proto_perl->IDBsingle, param);
11983 PL_DBtrace = sv_dup(proto_perl->IDBtrace, param);
11984 PL_DBsignal = sv_dup(proto_perl->IDBsignal, param);
11985 PL_dbargs = av_dup(proto_perl->Idbargs, param);
11987 /* symbol tables */
11988 PL_defstash = hv_dup_inc(proto_perl->Idefstash, param);
11989 PL_curstash = hv_dup(proto_perl->Icurstash, param);
11990 PL_debstash = hv_dup(proto_perl->Idebstash, param);
11991 PL_globalstash = hv_dup(proto_perl->Iglobalstash, param);
11992 PL_curstname = sv_dup_inc(proto_perl->Icurstname, param);
11994 PL_beginav = av_dup_inc(proto_perl->Ibeginav, param);
11995 PL_beginav_save = av_dup_inc(proto_perl->Ibeginav_save, param);
11996 PL_checkav_save = av_dup_inc(proto_perl->Icheckav_save, param);
11997 PL_unitcheckav = av_dup_inc(proto_perl->Iunitcheckav, param);
11998 PL_unitcheckav_save = av_dup_inc(proto_perl->Iunitcheckav_save, param);
11999 PL_endav = av_dup_inc(proto_perl->Iendav, param);
12000 PL_checkav = av_dup_inc(proto_perl->Icheckav, param);
12001 PL_initav = av_dup_inc(proto_perl->Iinitav, param);
12003 PL_sub_generation = proto_perl->Isub_generation;
12004 PL_isarev = hv_dup_inc(proto_perl->Iisarev, param);
12006 /* funky return mechanisms */
12007 PL_forkprocess = proto_perl->Iforkprocess;
12009 /* subprocess state */
12010 PL_fdpid = av_dup_inc(proto_perl->Ifdpid, param);
12012 /* internal state */
12013 PL_maxo = proto_perl->Imaxo;
12014 if (proto_perl->Iop_mask)
12015 PL_op_mask = SAVEPVN(proto_perl->Iop_mask, PL_maxo);
12018 /* PL_asserting = proto_perl->Iasserting; */
12020 /* current interpreter roots */
12021 PL_main_cv = cv_dup_inc(proto_perl->Imain_cv, param);
12023 PL_main_root = OpREFCNT_inc(proto_perl->Imain_root);
12025 PL_main_start = proto_perl->Imain_start;
12026 PL_eval_root = proto_perl->Ieval_root;
12027 PL_eval_start = proto_perl->Ieval_start;
12029 /* runtime control stuff */
12030 PL_curcopdb = (COP*)any_dup(proto_perl->Icurcopdb, proto_perl);
12032 PL_filemode = proto_perl->Ifilemode;
12033 PL_lastfd = proto_perl->Ilastfd;
12034 PL_oldname = proto_perl->Ioldname; /* XXX not quite right */
12037 PL_gensym = proto_perl->Igensym;
12038 PL_preambleav = av_dup_inc(proto_perl->Ipreambleav, param);
12039 PL_laststatval = proto_perl->Ilaststatval;
12040 PL_laststype = proto_perl->Ilaststype;
12043 PL_ors_sv = sv_dup_inc(proto_perl->Iors_sv, param);
12045 /* interpreter atexit processing */
12046 PL_exitlistlen = proto_perl->Iexitlistlen;
12047 if (PL_exitlistlen) {
12048 Newx(PL_exitlist, PL_exitlistlen, PerlExitListEntry);
12049 Copy(proto_perl->Iexitlist, PL_exitlist, PL_exitlistlen, PerlExitListEntry);
12052 PL_exitlist = (PerlExitListEntry*)NULL;
12054 PL_my_cxt_size = proto_perl->Imy_cxt_size;
12055 if (PL_my_cxt_size) {
12056 Newx(PL_my_cxt_list, PL_my_cxt_size, void *);
12057 Copy(proto_perl->Imy_cxt_list, PL_my_cxt_list, PL_my_cxt_size, void *);
12058 #ifdef PERL_GLOBAL_STRUCT_PRIVATE
12059 Newx(PL_my_cxt_keys, PL_my_cxt_size, const char *);
12060 Copy(proto_perl->Imy_cxt_keys, PL_my_cxt_keys, PL_my_cxt_size, char *);
12064 PL_my_cxt_list = (void**)NULL;
12065 #ifdef PERL_GLOBAL_STRUCT_PRIVATE
12066 PL_my_cxt_keys = (const char**)NULL;
12069 PL_modglobal = hv_dup_inc(proto_perl->Imodglobal, param);
12070 PL_custom_op_names = hv_dup_inc(proto_perl->Icustom_op_names,param);
12071 PL_custom_op_descs = hv_dup_inc(proto_perl->Icustom_op_descs,param);
12073 PL_profiledata = NULL;
12075 PL_compcv = cv_dup(proto_perl->Icompcv, param);
12077 PAD_CLONE_VARS(proto_perl, param);
12079 #ifdef HAVE_INTERP_INTERN
12080 sys_intern_dup(&proto_perl->Isys_intern, &PL_sys_intern);
12083 /* more statics moved here */
12084 PL_generation = proto_perl->Igeneration;
12085 PL_DBcv = cv_dup(proto_perl->IDBcv, param);
12087 PL_in_clean_objs = proto_perl->Iin_clean_objs;
12088 PL_in_clean_all = proto_perl->Iin_clean_all;
12090 PL_uid = proto_perl->Iuid;
12091 PL_euid = proto_perl->Ieuid;
12092 PL_gid = proto_perl->Igid;
12093 PL_egid = proto_perl->Iegid;
12094 PL_nomemok = proto_perl->Inomemok;
12095 PL_an = proto_perl->Ian;
12096 PL_evalseq = proto_perl->Ievalseq;
12097 PL_origenviron = proto_perl->Iorigenviron; /* XXX not quite right */
12098 PL_origalen = proto_perl->Iorigalen;
12099 #ifdef PERL_USES_PL_PIDSTATUS
12100 PL_pidstatus = newHV(); /* XXX flag for cloning? */
12102 PL_osname = SAVEPV(proto_perl->Iosname);
12103 PL_sighandlerp = proto_perl->Isighandlerp;
12105 PL_runops = proto_perl->Irunops;
12107 PL_parser = parser_dup(proto_perl->Iparser, param);
12109 PL_subline = proto_perl->Isubline;
12110 PL_subname = sv_dup_inc(proto_perl->Isubname, param);
12113 PL_cryptseen = proto_perl->Icryptseen;
12116 PL_hints = proto_perl->Ihints;
12118 PL_amagic_generation = proto_perl->Iamagic_generation;
12120 #ifdef USE_LOCALE_COLLATE
12121 PL_collation_ix = proto_perl->Icollation_ix;
12122 PL_collation_name = SAVEPV(proto_perl->Icollation_name);
12123 PL_collation_standard = proto_perl->Icollation_standard;
12124 PL_collxfrm_base = proto_perl->Icollxfrm_base;
12125 PL_collxfrm_mult = proto_perl->Icollxfrm_mult;
12126 #endif /* USE_LOCALE_COLLATE */
12128 #ifdef USE_LOCALE_NUMERIC
12129 PL_numeric_name = SAVEPV(proto_perl->Inumeric_name);
12130 PL_numeric_standard = proto_perl->Inumeric_standard;
12131 PL_numeric_local = proto_perl->Inumeric_local;
12132 PL_numeric_radix_sv = sv_dup_inc(proto_perl->Inumeric_radix_sv, param);
12133 #endif /* !USE_LOCALE_NUMERIC */
12135 /* utf8 character classes */
12136 PL_utf8_alnum = sv_dup_inc(proto_perl->Iutf8_alnum, param);
12137 PL_utf8_alnumc = sv_dup_inc(proto_perl->Iutf8_alnumc, param);
12138 PL_utf8_ascii = sv_dup_inc(proto_perl->Iutf8_ascii, param);
12139 PL_utf8_alpha = sv_dup_inc(proto_perl->Iutf8_alpha, param);
12140 PL_utf8_space = sv_dup_inc(proto_perl->Iutf8_space, param);
12141 PL_utf8_cntrl = sv_dup_inc(proto_perl->Iutf8_cntrl, param);
12142 PL_utf8_graph = sv_dup_inc(proto_perl->Iutf8_graph, param);
12143 PL_utf8_digit = sv_dup_inc(proto_perl->Iutf8_digit, param);
12144 PL_utf8_upper = sv_dup_inc(proto_perl->Iutf8_upper, param);
12145 PL_utf8_lower = sv_dup_inc(proto_perl->Iutf8_lower, param);
12146 PL_utf8_print = sv_dup_inc(proto_perl->Iutf8_print, param);
12147 PL_utf8_punct = sv_dup_inc(proto_perl->Iutf8_punct, param);
12148 PL_utf8_xdigit = sv_dup_inc(proto_perl->Iutf8_xdigit, param);
12149 PL_utf8_mark = sv_dup_inc(proto_perl->Iutf8_mark, param);
12150 PL_utf8_toupper = sv_dup_inc(proto_perl->Iutf8_toupper, param);
12151 PL_utf8_totitle = sv_dup_inc(proto_perl->Iutf8_totitle, param);
12152 PL_utf8_tolower = sv_dup_inc(proto_perl->Iutf8_tolower, param);
12153 PL_utf8_tofold = sv_dup_inc(proto_perl->Iutf8_tofold, param);
12154 PL_utf8_idstart = sv_dup_inc(proto_perl->Iutf8_idstart, param);
12155 PL_utf8_idcont = sv_dup_inc(proto_perl->Iutf8_idcont, param);
12157 /* Did the locale setup indicate UTF-8? */
12158 PL_utf8locale = proto_perl->Iutf8locale;
12159 /* Unicode features (see perlrun/-C) */
12160 PL_unicode = proto_perl->Iunicode;
12162 /* Pre-5.8 signals control */
12163 PL_signals = proto_perl->Isignals;
12165 /* times() ticks per second */
12166 PL_clocktick = proto_perl->Iclocktick;
12168 /* Recursion stopper for PerlIO_find_layer */
12169 PL_in_load_module = proto_perl->Iin_load_module;
12171 /* sort() routine */
12172 PL_sort_RealCmp = proto_perl->Isort_RealCmp;
12174 /* Not really needed/useful since the reenrant_retint is "volatile",
12175 * but do it for consistency's sake. */
12176 PL_reentrant_retint = proto_perl->Ireentrant_retint;
12178 /* Hooks to shared SVs and locks. */
12179 PL_sharehook = proto_perl->Isharehook;
12180 PL_lockhook = proto_perl->Ilockhook;
12181 PL_unlockhook = proto_perl->Iunlockhook;
12182 PL_threadhook = proto_perl->Ithreadhook;
12183 PL_destroyhook = proto_perl->Idestroyhook;
12185 #ifdef THREADS_HAVE_PIDS
12186 PL_ppid = proto_perl->Ippid;
12190 PL_last_swash_hv = NULL; /* reinits on demand */
12191 PL_last_swash_klen = 0;
12192 PL_last_swash_key[0]= '\0';
12193 PL_last_swash_tmps = (U8*)NULL;
12194 PL_last_swash_slen = 0;
12196 PL_glob_index = proto_perl->Iglob_index;
12197 PL_srand_called = proto_perl->Isrand_called;
12198 PL_bitcount = NULL; /* reinits on demand */
12200 if (proto_perl->Ipsig_pend) {
12201 Newxz(PL_psig_pend, SIG_SIZE, int);
12204 PL_psig_pend = (int*)NULL;
12207 if (proto_perl->Ipsig_ptr) {
12208 Newxz(PL_psig_ptr, SIG_SIZE, SV*);
12209 Newxz(PL_psig_name, SIG_SIZE, SV*);
12210 for (i = 1; i < SIG_SIZE; i++) {
12211 PL_psig_ptr[i] = sv_dup_inc(proto_perl->Ipsig_ptr[i], param);
12212 PL_psig_name[i] = sv_dup_inc(proto_perl->Ipsig_name[i], param);
12216 PL_psig_ptr = (SV**)NULL;
12217 PL_psig_name = (SV**)NULL;
12220 /* intrpvar.h stuff */
12222 if (flags & CLONEf_COPY_STACKS) {
12223 /* next allocation will be PL_tmps_stack[PL_tmps_ix+1] */
12224 PL_tmps_ix = proto_perl->Itmps_ix;
12225 PL_tmps_max = proto_perl->Itmps_max;
12226 PL_tmps_floor = proto_perl->Itmps_floor;
12227 Newxz(PL_tmps_stack, PL_tmps_max, SV*);
12229 while (i <= PL_tmps_ix) {
12230 PL_tmps_stack[i] = sv_dup_inc(proto_perl->Itmps_stack[i], param);
12234 /* next PUSHMARK() sets *(PL_markstack_ptr+1) */
12235 i = proto_perl->Imarkstack_max - proto_perl->Imarkstack;
12236 Newxz(PL_markstack, i, I32);
12237 PL_markstack_max = PL_markstack + (proto_perl->Imarkstack_max
12238 - proto_perl->Imarkstack);
12239 PL_markstack_ptr = PL_markstack + (proto_perl->Imarkstack_ptr
12240 - proto_perl->Imarkstack);
12241 Copy(proto_perl->Imarkstack, PL_markstack,
12242 PL_markstack_ptr - PL_markstack + 1, I32);
12244 /* next push_scope()/ENTER sets PL_scopestack[PL_scopestack_ix]
12245 * NOTE: unlike the others! */
12246 PL_scopestack_ix = proto_perl->Iscopestack_ix;
12247 PL_scopestack_max = proto_perl->Iscopestack_max;
12248 Newxz(PL_scopestack, PL_scopestack_max, I32);
12249 Copy(proto_perl->Iscopestack, PL_scopestack, PL_scopestack_ix, I32);
12251 /* NOTE: si_dup() looks at PL_markstack */
12252 PL_curstackinfo = si_dup(proto_perl->Icurstackinfo, param);
12254 /* PL_curstack = PL_curstackinfo->si_stack; */
12255 PL_curstack = av_dup(proto_perl->Icurstack, param);
12256 PL_mainstack = av_dup(proto_perl->Imainstack, param);
12258 /* next PUSHs() etc. set *(PL_stack_sp+1) */
12259 PL_stack_base = AvARRAY(PL_curstack);
12260 PL_stack_sp = PL_stack_base + (proto_perl->Istack_sp
12261 - proto_perl->Istack_base);
12262 PL_stack_max = PL_stack_base + AvMAX(PL_curstack);
12264 /* next SSPUSHFOO() sets PL_savestack[PL_savestack_ix]
12265 * NOTE: unlike the others! */
12266 PL_savestack_ix = proto_perl->Isavestack_ix;
12267 PL_savestack_max = proto_perl->Isavestack_max;
12268 /*Newxz(PL_savestack, PL_savestack_max, ANY);*/
12269 PL_savestack = ss_dup(proto_perl, param);
12273 ENTER; /* perl_destruct() wants to LEAVE; */
12275 /* although we're not duplicating the tmps stack, we should still
12276 * add entries for any SVs on the tmps stack that got cloned by a
12277 * non-refcount means (eg a temp in @_); otherwise they will be
12280 for (i = 0; i<= proto_perl->Itmps_ix; i++) {
12281 SV * const nsv = MUTABLE_SV(ptr_table_fetch(PL_ptr_table,
12282 proto_perl->Itmps_stack[i]));
12283 if (nsv && !SvREFCNT(nsv)) {
12285 PL_tmps_stack[++PL_tmps_ix] = SvREFCNT_inc_simple(nsv);
12290 PL_start_env = proto_perl->Istart_env; /* XXXXXX */
12291 PL_top_env = &PL_start_env;
12293 PL_op = proto_perl->Iop;
12296 PL_Xpv = (XPV*)NULL;
12297 my_perl->Ina = proto_perl->Ina;
12299 PL_statbuf = proto_perl->Istatbuf;
12300 PL_statcache = proto_perl->Istatcache;
12301 PL_statgv = gv_dup(proto_perl->Istatgv, param);
12302 PL_statname = sv_dup_inc(proto_perl->Istatname, param);
12304 PL_timesbuf = proto_perl->Itimesbuf;
12307 PL_tainted = proto_perl->Itainted;
12308 PL_curpm = proto_perl->Icurpm; /* XXX No PMOP ref count */
12309 PL_rs = sv_dup_inc(proto_perl->Irs, param);
12310 PL_last_in_gv = gv_dup(proto_perl->Ilast_in_gv, param);
12311 PL_defoutgv = gv_dup_inc(proto_perl->Idefoutgv, param);
12312 PL_chopset = proto_perl->Ichopset; /* XXX never deallocated */
12313 PL_toptarget = sv_dup_inc(proto_perl->Itoptarget, param);
12314 PL_bodytarget = sv_dup_inc(proto_perl->Ibodytarget, param);
12315 PL_formtarget = sv_dup(proto_perl->Iformtarget, param);
12317 PL_restartop = proto_perl->Irestartop;
12318 PL_in_eval = proto_perl->Iin_eval;
12319 PL_delaymagic = proto_perl->Idelaymagic;
12320 PL_dirty = proto_perl->Idirty;
12321 PL_localizing = proto_perl->Ilocalizing;
12323 PL_errors = sv_dup_inc(proto_perl->Ierrors, param);
12324 PL_hv_fetch_ent_mh = NULL;
12325 PL_modcount = proto_perl->Imodcount;
12326 PL_lastgotoprobe = NULL;
12327 PL_dumpindent = proto_perl->Idumpindent;
12329 PL_sortcop = (OP*)any_dup(proto_perl->Isortcop, proto_perl);
12330 PL_sortstash = hv_dup(proto_perl->Isortstash, param);
12331 PL_firstgv = gv_dup(proto_perl->Ifirstgv, param);
12332 PL_secondgv = gv_dup(proto_perl->Isecondgv, param);
12333 PL_efloatbuf = NULL; /* reinits on demand */
12334 PL_efloatsize = 0; /* reinits on demand */
12338 PL_screamfirst = NULL;
12339 PL_screamnext = NULL;
12340 PL_maxscream = -1; /* reinits on demand */
12341 PL_lastscream = NULL;
12344 PL_regdummy = proto_perl->Iregdummy;
12345 PL_colorset = 0; /* reinits PL_colors[] */
12346 /*PL_colors[6] = {0,0,0,0,0,0};*/
12350 /* Pluggable optimizer */
12351 PL_peepp = proto_perl->Ipeepp;
12353 PL_stashcache = newHV();
12355 PL_watchaddr = (char **) ptr_table_fetch(PL_ptr_table,
12356 proto_perl->Iwatchaddr);
12357 PL_watchok = PL_watchaddr ? * PL_watchaddr : NULL;
12358 if (PL_debug && PL_watchaddr) {
12359 PerlIO_printf(Perl_debug_log,
12360 "WATCHING: %"UVxf" cloned as %"UVxf" with value %"UVxf"\n",
12361 PTR2UV(proto_perl->Iwatchaddr), PTR2UV(PL_watchaddr),
12362 PTR2UV(PL_watchok));
12365 PL_registered_mros = hv_dup_inc(proto_perl->Iregistered_mros, param);
12367 if (!(flags & CLONEf_KEEP_PTR_TABLE)) {
12368 ptr_table_free(PL_ptr_table);
12369 PL_ptr_table = NULL;
12372 /* Call the ->CLONE method, if it exists, for each of the stashes
12373 identified by sv_dup() above.
12375 while(av_len(param->stashes) != -1) {
12376 HV* const stash = MUTABLE_HV(av_shift(param->stashes));
12377 GV* const cloner = gv_fetchmethod_autoload(stash, "CLONE", 0);
12378 if (cloner && GvCV(cloner)) {
12383 mXPUSHs(newSVhek(HvNAME_HEK(stash)));
12385 call_sv(MUTABLE_SV(GvCV(cloner)), G_DISCARD);
12391 SvREFCNT_dec(param->stashes);
12393 /* orphaned? eg threads->new inside BEGIN or use */
12394 if (PL_compcv && ! SvREFCNT(PL_compcv)) {
12395 SvREFCNT_inc_simple_void(PL_compcv);
12396 SAVEFREESV(PL_compcv);
12402 #endif /* USE_ITHREADS */
12405 =head1 Unicode Support
12407 =for apidoc sv_recode_to_utf8
12409 The encoding is assumed to be an Encode object, on entry the PV
12410 of the sv is assumed to be octets in that encoding, and the sv
12411 will be converted into Unicode (and UTF-8).
12413 If the sv already is UTF-8 (or if it is not POK), or if the encoding
12414 is not a reference, nothing is done to the sv. If the encoding is not
12415 an C<Encode::XS> Encoding object, bad things will happen.
12416 (See F<lib/encoding.pm> and L<Encode>).
12418 The PV of the sv is returned.
12423 Perl_sv_recode_to_utf8(pTHX_ SV *sv, SV *encoding)
12427 PERL_ARGS_ASSERT_SV_RECODE_TO_UTF8;
12429 if (SvPOK(sv) && !SvUTF8(sv) && !IN_BYTES && SvROK(encoding)) {
12443 Passing sv_yes is wrong - it needs to be or'ed set of constants
12444 for Encode::XS, while UTf-8 decode (currently) assumes a true value means
12445 remove converted chars from source.
12447 Both will default the value - let them.
12449 XPUSHs(&PL_sv_yes);
12452 call_method("decode", G_SCALAR);
12456 s = SvPV_const(uni, len);
12457 if (s != SvPVX_const(sv)) {
12458 SvGROW(sv, len + 1);
12459 Move(s, SvPVX(sv), len + 1, char);
12460 SvCUR_set(sv, len);
12467 return SvPOKp(sv) ? SvPVX(sv) : NULL;
12471 =for apidoc sv_cat_decode
12473 The encoding is assumed to be an Encode object, the PV of the ssv is
12474 assumed to be octets in that encoding and decoding the input starts
12475 from the position which (PV + *offset) pointed to. The dsv will be
12476 concatenated the decoded UTF-8 string from ssv. Decoding will terminate
12477 when the string tstr appears in decoding output or the input ends on
12478 the PV of the ssv. The value which the offset points will be modified
12479 to the last input position on the ssv.
12481 Returns TRUE if the terminator was found, else returns FALSE.
12486 Perl_sv_cat_decode(pTHX_ SV *dsv, SV *encoding,
12487 SV *ssv, int *offset, char *tstr, int tlen)
12492 PERL_ARGS_ASSERT_SV_CAT_DECODE;
12494 if (SvPOK(ssv) && SvPOK(dsv) && SvROK(encoding) && offset) {
12505 offsv = newSViv(*offset);
12507 mXPUSHp(tstr, tlen);
12509 call_method("cat_decode", G_SCALAR);
12511 ret = SvTRUE(TOPs);
12512 *offset = SvIV(offsv);
12518 Perl_croak(aTHX_ "Invalid argument to sv_cat_decode");
12523 /* ---------------------------------------------------------------------
12525 * support functions for report_uninit()
12528 /* the maxiumum size of array or hash where we will scan looking
12529 * for the undefined element that triggered the warning */
12531 #define FUV_MAX_SEARCH_SIZE 1000
12533 /* Look for an entry in the hash whose value has the same SV as val;
12534 * If so, return a mortal copy of the key. */
12537 S_find_hash_subscript(pTHX_ const HV *const hv, const SV *const val)
12540 register HE **array;
12543 PERL_ARGS_ASSERT_FIND_HASH_SUBSCRIPT;
12545 if (!hv || SvMAGICAL(hv) || !HvARRAY(hv) ||
12546 (HvTOTALKEYS(hv) > FUV_MAX_SEARCH_SIZE))
12549 array = HvARRAY(hv);
12551 for (i=HvMAX(hv); i>0; i--) {
12552 register HE *entry;
12553 for (entry = array[i]; entry; entry = HeNEXT(entry)) {
12554 if (HeVAL(entry) != val)
12556 if ( HeVAL(entry) == &PL_sv_undef ||
12557 HeVAL(entry) == &PL_sv_placeholder)
12561 if (HeKLEN(entry) == HEf_SVKEY)
12562 return sv_mortalcopy(HeKEY_sv(entry));
12563 return sv_2mortal(newSVhek(HeKEY_hek(entry)));
12569 /* Look for an entry in the array whose value has the same SV as val;
12570 * If so, return the index, otherwise return -1. */
12573 S_find_array_subscript(pTHX_ const AV *const av, const SV *const val)
12577 PERL_ARGS_ASSERT_FIND_ARRAY_SUBSCRIPT;
12579 if (!av || SvMAGICAL(av) || !AvARRAY(av) ||
12580 (AvFILLp(av) > FUV_MAX_SEARCH_SIZE))
12583 if (val != &PL_sv_undef) {
12584 SV ** const svp = AvARRAY(av);
12587 for (i=AvFILLp(av); i>=0; i--)
12594 /* S_varname(): return the name of a variable, optionally with a subscript.
12595 * If gv is non-zero, use the name of that global, along with gvtype (one
12596 * of "$", "@", "%"); otherwise use the name of the lexical at pad offset
12597 * targ. Depending on the value of the subscript_type flag, return:
12600 #define FUV_SUBSCRIPT_NONE 1 /* "@foo" */
12601 #define FUV_SUBSCRIPT_ARRAY 2 /* "$foo[aindex]" */
12602 #define FUV_SUBSCRIPT_HASH 3 /* "$foo{keyname}" */
12603 #define FUV_SUBSCRIPT_WITHIN 4 /* "within @foo" */
12606 S_varname(pTHX_ const GV *const gv, const char gvtype, PADOFFSET targ,
12607 const SV *const keyname, I32 aindex, int subscript_type)
12610 SV * const name = sv_newmortal();
12613 buffer[0] = gvtype;
12616 /* as gv_fullname4(), but add literal '^' for $^FOO names */
12618 gv_fullname4(name, gv, buffer, 0);
12620 if ((unsigned int)SvPVX(name)[1] <= 26) {
12622 buffer[1] = SvPVX(name)[1] + 'A' - 1;
12624 /* Swap the 1 unprintable control character for the 2 byte pretty
12625 version - ie substr($name, 1, 1) = $buffer; */
12626 sv_insert(name, 1, 1, buffer, 2);
12630 CV * const cv = find_runcv(NULL);
12634 if (!cv || !CvPADLIST(cv))
12636 av = MUTABLE_AV((*av_fetch(CvPADLIST(cv), 0, FALSE)));
12637 sv = *av_fetch(av, targ, FALSE);
12638 sv_setpvn(name, SvPV_nolen_const(sv), SvCUR(sv));
12641 if (subscript_type == FUV_SUBSCRIPT_HASH) {
12642 SV * const sv = newSV(0);
12643 *SvPVX(name) = '$';
12644 Perl_sv_catpvf(aTHX_ name, "{%s}",
12645 pv_display(sv,SvPVX_const(keyname), SvCUR(keyname), 0, 32));
12648 else if (subscript_type == FUV_SUBSCRIPT_ARRAY) {
12649 *SvPVX(name) = '$';
12650 Perl_sv_catpvf(aTHX_ name, "[%"IVdf"]", (IV)aindex);
12652 else if (subscript_type == FUV_SUBSCRIPT_WITHIN) {
12653 /* We know that name has no magic, so can use 0 instead of SV_GMAGIC */
12654 Perl_sv_insert_flags(aTHX_ name, 0, 0, STR_WITH_LEN("within "), 0);
12662 =for apidoc find_uninit_var
12664 Find the name of the undefined variable (if any) that caused the operator o
12665 to issue a "Use of uninitialized value" warning.
12666 If match is true, only return a name if it's value matches uninit_sv.
12667 So roughly speaking, if a unary operator (such as OP_COS) generates a
12668 warning, then following the direct child of the op may yield an
12669 OP_PADSV or OP_GV that gives the name of the undefined variable. On the
12670 other hand, with OP_ADD there are two branches to follow, so we only print
12671 the variable name if we get an exact match.
12673 The name is returned as a mortal SV.
12675 Assumes that PL_op is the op that originally triggered the error, and that
12676 PL_comppad/PL_curpad points to the currently executing pad.
12682 S_find_uninit_var(pTHX_ const OP *const obase, const SV *const uninit_sv,
12688 const OP *o, *o2, *kid;
12690 if (!obase || (match && (!uninit_sv || uninit_sv == &PL_sv_undef ||
12691 uninit_sv == &PL_sv_placeholder)))
12694 switch (obase->op_type) {
12701 const bool pad = (obase->op_type == OP_PADAV || obase->op_type == OP_PADHV);
12702 const bool hash = (obase->op_type == OP_PADHV || obase->op_type == OP_RV2HV);
12705 int subscript_type = FUV_SUBSCRIPT_WITHIN;
12707 if (pad) { /* @lex, %lex */
12708 sv = PAD_SVl(obase->op_targ);
12712 if (cUNOPx(obase)->op_first->op_type == OP_GV) {
12713 /* @global, %global */
12714 gv = cGVOPx_gv(cUNOPx(obase)->op_first);
12717 sv = hash ? MUTABLE_SV(GvHV(gv)): MUTABLE_SV(GvAV(gv));
12719 else /* @{expr}, %{expr} */
12720 return find_uninit_var(cUNOPx(obase)->op_first,
12724 /* attempt to find a match within the aggregate */
12726 keysv = find_hash_subscript((const HV*)sv, uninit_sv);
12728 subscript_type = FUV_SUBSCRIPT_HASH;
12731 index = find_array_subscript((const AV *)sv, uninit_sv);
12733 subscript_type = FUV_SUBSCRIPT_ARRAY;
12736 if (match && subscript_type == FUV_SUBSCRIPT_WITHIN)
12739 return varname(gv, hash ? '%' : '@', obase->op_targ,
12740 keysv, index, subscript_type);
12744 if (match && PAD_SVl(obase->op_targ) != uninit_sv)
12746 return varname(NULL, '$', obase->op_targ,
12747 NULL, 0, FUV_SUBSCRIPT_NONE);
12750 gv = cGVOPx_gv(obase);
12751 if (!gv || (match && GvSV(gv) != uninit_sv))
12753 return varname(gv, '$', 0, NULL, 0, FUV_SUBSCRIPT_NONE);
12756 if (obase->op_flags & OPf_SPECIAL) { /* lexical array */
12759 AV *av = MUTABLE_AV(PAD_SV(obase->op_targ));
12760 if (!av || SvRMAGICAL(av))
12762 svp = av_fetch(av, (I32)obase->op_private, FALSE);
12763 if (!svp || *svp != uninit_sv)
12766 return varname(NULL, '$', obase->op_targ,
12767 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
12770 gv = cGVOPx_gv(obase);
12775 AV *const av = GvAV(gv);
12776 if (!av || SvRMAGICAL(av))
12778 svp = av_fetch(av, (I32)obase->op_private, FALSE);
12779 if (!svp || *svp != uninit_sv)
12782 return varname(gv, '$', 0,
12783 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
12788 o = cUNOPx(obase)->op_first;
12789 if (!o || o->op_type != OP_NULL ||
12790 ! (o->op_targ == OP_AELEM || o->op_targ == OP_HELEM))
12792 return find_uninit_var(cBINOPo->op_last, uninit_sv, match);
12796 if (PL_op == obase)
12797 /* $a[uninit_expr] or $h{uninit_expr} */
12798 return find_uninit_var(cBINOPx(obase)->op_last, uninit_sv, match);
12801 o = cBINOPx(obase)->op_first;
12802 kid = cBINOPx(obase)->op_last;
12804 /* get the av or hv, and optionally the gv */
12806 if (o->op_type == OP_PADAV || o->op_type == OP_PADHV) {
12807 sv = PAD_SV(o->op_targ);
12809 else if ((o->op_type == OP_RV2AV || o->op_type == OP_RV2HV)
12810 && cUNOPo->op_first->op_type == OP_GV)
12812 gv = cGVOPx_gv(cUNOPo->op_first);
12816 == OP_RV2HV ? MUTABLE_SV(GvHV(gv)) : MUTABLE_SV(GvAV(gv));
12821 if (kid && kid->op_type == OP_CONST && SvOK(cSVOPx_sv(kid))) {
12822 /* index is constant */
12826 if (obase->op_type == OP_HELEM) {
12827 HE* he = hv_fetch_ent(MUTABLE_HV(sv), cSVOPx_sv(kid), 0, 0);
12828 if (!he || HeVAL(he) != uninit_sv)
12832 SV * const * const svp = av_fetch(MUTABLE_AV(sv), SvIV(cSVOPx_sv(kid)), FALSE);
12833 if (!svp || *svp != uninit_sv)
12837 if (obase->op_type == OP_HELEM)
12838 return varname(gv, '%', o->op_targ,
12839 cSVOPx_sv(kid), 0, FUV_SUBSCRIPT_HASH);
12841 return varname(gv, '@', o->op_targ, NULL,
12842 SvIV(cSVOPx_sv(kid)), FUV_SUBSCRIPT_ARRAY);
12845 /* index is an expression;
12846 * attempt to find a match within the aggregate */
12847 if (obase->op_type == OP_HELEM) {
12848 SV * const keysv = find_hash_subscript((const HV*)sv, uninit_sv);
12850 return varname(gv, '%', o->op_targ,
12851 keysv, 0, FUV_SUBSCRIPT_HASH);
12855 = find_array_subscript((const AV *)sv, uninit_sv);
12857 return varname(gv, '@', o->op_targ,
12858 NULL, index, FUV_SUBSCRIPT_ARRAY);
12863 (o->op_type == OP_PADAV || o->op_type == OP_RV2AV)
12865 o->op_targ, NULL, 0, FUV_SUBSCRIPT_WITHIN);
12870 /* only examine RHS */
12871 return find_uninit_var(cBINOPx(obase)->op_first, uninit_sv, match);
12874 o = cUNOPx(obase)->op_first;
12875 if (o->op_type == OP_PUSHMARK)
12878 if (!o->op_sibling) {
12879 /* one-arg version of open is highly magical */
12881 if (o->op_type == OP_GV) { /* open FOO; */
12883 if (match && GvSV(gv) != uninit_sv)
12885 return varname(gv, '$', 0,
12886 NULL, 0, FUV_SUBSCRIPT_NONE);
12888 /* other possibilities not handled are:
12889 * open $x; or open my $x; should return '${*$x}'
12890 * open expr; should return '$'.expr ideally
12896 /* ops where $_ may be an implicit arg */
12900 if ( !(obase->op_flags & OPf_STACKED)) {
12901 if (uninit_sv == ((obase->op_private & OPpTARGET_MY)
12902 ? PAD_SVl(obase->op_targ)
12905 sv = sv_newmortal();
12906 sv_setpvs(sv, "$_");
12915 match = 1; /* print etc can return undef on defined args */
12916 /* skip filehandle as it can't produce 'undef' warning */
12917 o = cUNOPx(obase)->op_first;
12918 if ((obase->op_flags & OPf_STACKED) && o->op_type == OP_PUSHMARK)
12919 o = o->op_sibling->op_sibling;
12923 case OP_ENTEREVAL: /* could be eval $undef or $x='$undef'; eval $x */
12925 case OP_CUSTOM: /* XS or custom code could trigger random warnings */
12927 /* the following ops are capable of returning PL_sv_undef even for
12928 * defined arg(s) */
12947 case OP_GETPEERNAME:
12995 case OP_SMARTMATCH:
13004 /* XXX tmp hack: these two may call an XS sub, and currently
13005 XS subs don't have a SUB entry on the context stack, so CV and
13006 pad determination goes wrong, and BAD things happen. So, just
13007 don't try to determine the value under those circumstances.
13008 Need a better fix at dome point. DAPM 11/2007 */
13013 /* def-ness of rval pos() is independent of the def-ness of its arg */
13014 if ( !(obase->op_flags & OPf_MOD))
13019 if (SvROK(PL_rs) && uninit_sv == SvRV(PL_rs))
13020 return newSVpvs_flags("${$/}", SVs_TEMP);
13025 if (!(obase->op_flags & OPf_KIDS))
13027 o = cUNOPx(obase)->op_first;
13033 /* if all except one arg are constant, or have no side-effects,
13034 * or are optimized away, then it's unambiguous */
13036 for (kid=o; kid; kid = kid->op_sibling) {
13038 const OPCODE type = kid->op_type;
13039 if ( (type == OP_CONST && SvOK(cSVOPx_sv(kid)))
13040 || (type == OP_NULL && ! (kid->op_flags & OPf_KIDS))
13041 || (type == OP_PUSHMARK)
13045 if (o2) { /* more than one found */
13052 return find_uninit_var(o2, uninit_sv, match);
13054 /* scan all args */
13056 sv = find_uninit_var(o, uninit_sv, 1);
13068 =for apidoc report_uninit
13070 Print appropriate "Use of uninitialized variable" warning
13076 Perl_report_uninit(pTHX_ const SV *uninit_sv)
13080 SV* varname = NULL;
13082 varname = find_uninit_var(PL_op, uninit_sv,0);
13084 sv_insert(varname, 0, 0, " ", 1);
13086 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
13087 varname ? SvPV_nolen_const(varname) : "",
13088 " in ", OP_DESC(PL_op));
13091 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
13097 * c-indentation-style: bsd
13098 * c-basic-offset: 4
13099 * indent-tabs-mode: t
13102 * ex: set ts=8 sts=4 sw=4 noet: