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
36 # if __STDC_VERSION__ >= 199901L && !defined(VMS)
47 /* Missing proto on LynxOS */
48 char *gconvert(double, int, int, char *);
51 #ifdef PERL_UTF8_CACHE_ASSERT
52 /* if adding more checks watch out for the following tests:
53 * t/op/index.t t/op/length.t t/op/pat.t t/op/substr.t
54 * lib/utf8.t lib/Unicode/Collate/t/index.t
57 # define ASSERT_UTF8_CACHE(cache) \
58 STMT_START { if (cache) { assert((cache)[0] <= (cache)[1]); \
59 assert((cache)[2] <= (cache)[3]); \
60 assert((cache)[3] <= (cache)[1]);} \
63 # define ASSERT_UTF8_CACHE(cache) NOOP
66 #ifdef PERL_OLD_COPY_ON_WRITE
67 #define SV_COW_NEXT_SV(sv) INT2PTR(SV *,SvUVX(sv))
68 #define SV_COW_NEXT_SV_SET(current,next) SvUV_set(current, PTR2UV(next))
69 /* This is a pessimistic view. Scalar must be purely a read-write PV to copy-
73 /* ============================================================================
75 =head1 Allocation and deallocation of SVs.
77 An SV (or AV, HV, etc.) is allocated in two parts: the head (struct
78 sv, av, hv...) contains type and reference count information, and for
79 many types, a pointer to the body (struct xrv, xpv, xpviv...), which
80 contains fields specific to each type. Some types store all they need
81 in the head, so don't have a body.
83 In all but the most memory-paranoid configurations (ex: PURIFY), heads
84 and bodies are allocated out of arenas, which by default are
85 approximately 4K chunks of memory parcelled up into N heads or bodies.
86 Sv-bodies are allocated by their sv-type, guaranteeing size
87 consistency needed to allocate safely from arrays.
89 For SV-heads, the first slot in each arena is reserved, and holds a
90 link to the next arena, some flags, and a note of the number of slots.
91 Snaked through each arena chain is a linked list of free items; when
92 this becomes empty, an extra arena is allocated and divided up into N
93 items which are threaded into the free list.
95 SV-bodies are similar, but they use arena-sets by default, which
96 separate the link and info from the arena itself, and reclaim the 1st
97 slot in the arena. SV-bodies are further described later.
99 The following global variables are associated with arenas:
101 PL_sv_arenaroot pointer to list of SV arenas
102 PL_sv_root pointer to list of free SV structures
104 PL_body_arenas head of linked-list of body arenas
105 PL_body_roots[] array of pointers to list of free bodies of svtype
106 arrays are indexed by the svtype needed
108 A few special SV heads are not allocated from an arena, but are
109 instead directly created in the interpreter structure, eg PL_sv_undef.
110 The size of arenas can be changed from the default by setting
111 PERL_ARENA_SIZE appropriately at compile time.
113 The SV arena serves the secondary purpose of allowing still-live SVs
114 to be located and destroyed during final cleanup.
116 At the lowest level, the macros new_SV() and del_SV() grab and free
117 an SV head. (If debugging with -DD, del_SV() calls the function S_del_sv()
118 to return the SV to the free list with error checking.) new_SV() calls
119 more_sv() / sv_add_arena() to add an extra arena if the free list is empty.
120 SVs in the free list have their SvTYPE field set to all ones.
122 At the time of very final cleanup, sv_free_arenas() is called from
123 perl_destruct() to physically free all the arenas allocated since the
124 start of the interpreter.
126 The function visit() scans the SV arenas list, and calls a specified
127 function for each SV it finds which is still live - ie which has an SvTYPE
128 other than all 1's, and a non-zero SvREFCNT. visit() is used by the
129 following functions (specified as [function that calls visit()] / [function
130 called by visit() for each SV]):
132 sv_report_used() / do_report_used()
133 dump all remaining SVs (debugging aid)
135 sv_clean_objs() / do_clean_objs(),do_clean_named_objs(),
136 do_clean_named_io_objs()
137 Attempt to free all objects pointed to by RVs,
138 and try to do the same for all objects indirectly
139 referenced by typeglobs too. Called once from
140 perl_destruct(), prior to calling sv_clean_all()
143 sv_clean_all() / do_clean_all()
144 SvREFCNT_dec(sv) each remaining SV, possibly
145 triggering an sv_free(). It also sets the
146 SVf_BREAK flag on the SV to indicate that the
147 refcnt has been artificially lowered, and thus
148 stopping sv_free() from giving spurious warnings
149 about SVs which unexpectedly have a refcnt
150 of zero. called repeatedly from perl_destruct()
151 until there are no SVs left.
153 =head2 Arena allocator API Summary
155 Private API to rest of sv.c
159 new_XPVNV(), del_XPVGV(),
164 sv_report_used(), sv_clean_objs(), sv_clean_all(), sv_free_arenas()
168 * ========================================================================= */
171 * "A time to plant, and a time to uproot what was planted..."
175 # define MEM_LOG_NEW_SV(sv, file, line, func) \
176 Perl_mem_log_new_sv(sv, file, line, func)
177 # define MEM_LOG_DEL_SV(sv, file, line, func) \
178 Perl_mem_log_del_sv(sv, file, line, func)
180 # define MEM_LOG_NEW_SV(sv, file, line, func) NOOP
181 # define MEM_LOG_DEL_SV(sv, file, line, func) NOOP
184 #ifdef DEBUG_LEAKING_SCALARS
185 # define FREE_SV_DEBUG_FILE(sv) Safefree((sv)->sv_debug_file)
186 # define DEBUG_SV_SERIAL(sv) \
187 DEBUG_m(PerlIO_printf(Perl_debug_log, "0x%"UVxf": (%05ld) del_SV\n", \
188 PTR2UV(sv), (long)(sv)->sv_debug_serial))
190 # define FREE_SV_DEBUG_FILE(sv)
191 # define DEBUG_SV_SERIAL(sv) NOOP
195 # define SvARENA_CHAIN(sv) ((sv)->sv_u.svu_rv)
196 # define SvARENA_CHAIN_SET(sv,val) (sv)->sv_u.svu_rv = MUTABLE_SV((val))
197 /* Whilst I'd love to do this, it seems that things like to check on
199 # define POSION_SV_HEAD(sv) PoisonNew(sv, 1, struct STRUCT_SV)
201 # define POSION_SV_HEAD(sv) PoisonNew(&SvANY(sv), 1, void *), \
202 PoisonNew(&SvREFCNT(sv), 1, U32)
204 # define SvARENA_CHAIN(sv) SvANY(sv)
205 # define SvARENA_CHAIN_SET(sv,val) SvANY(sv) = (void *)(val)
206 # define POSION_SV_HEAD(sv)
209 /* Mark an SV head as unused, and add to free list.
211 * If SVf_BREAK is set, skip adding it to the free list, as this SV had
212 * its refcount artificially decremented during global destruction, so
213 * there may be dangling pointers to it. The last thing we want in that
214 * case is for it to be reused. */
216 #define plant_SV(p) \
218 const U32 old_flags = SvFLAGS(p); \
219 MEM_LOG_DEL_SV(p, __FILE__, __LINE__, FUNCTION__); \
220 DEBUG_SV_SERIAL(p); \
221 FREE_SV_DEBUG_FILE(p); \
223 SvFLAGS(p) = SVTYPEMASK; \
224 if (!(old_flags & SVf_BREAK)) { \
225 SvARENA_CHAIN_SET(p, PL_sv_root); \
231 #define uproot_SV(p) \
234 PL_sv_root = MUTABLE_SV(SvARENA_CHAIN(p)); \
239 /* make some more SVs by adding another arena */
246 char *chunk; /* must use New here to match call to */
247 Newx(chunk,PERL_ARENA_SIZE,char); /* Safefree() in sv_free_arenas() */
248 sv_add_arena(chunk, PERL_ARENA_SIZE, 0);
253 /* new_SV(): return a new, empty SV head */
255 #ifdef DEBUG_LEAKING_SCALARS
256 /* provide a real function for a debugger to play with */
258 S_new_SV(pTHX_ const char *file, int line, const char *func)
265 sv = S_more_sv(aTHX);
269 sv->sv_debug_optype = PL_op ? PL_op->op_type : 0;
270 sv->sv_debug_line = (U16) (PL_parser && PL_parser->copline != NOLINE
276 sv->sv_debug_inpad = 0;
277 sv->sv_debug_parent = NULL;
278 sv->sv_debug_file = PL_curcop ? savepv(CopFILE(PL_curcop)): NULL;
280 sv->sv_debug_serial = PL_sv_serial++;
282 MEM_LOG_NEW_SV(sv, file, line, func);
283 DEBUG_m(PerlIO_printf(Perl_debug_log, "0x%"UVxf": (%05ld) new_SV (from %s:%d [%s])\n",
284 PTR2UV(sv), (long)sv->sv_debug_serial, file, line, func));
288 # define new_SV(p) (p)=S_new_SV(aTHX_ __FILE__, __LINE__, FUNCTION__)
296 (p) = S_more_sv(aTHX); \
300 MEM_LOG_NEW_SV(p, __FILE__, __LINE__, FUNCTION__); \
305 /* del_SV(): return an empty SV head to the free list */
318 S_del_sv(pTHX_ SV *p)
322 PERL_ARGS_ASSERT_DEL_SV;
327 for (sva = PL_sv_arenaroot; sva; sva = MUTABLE_SV(SvANY(sva))) {
328 const SV * const sv = sva + 1;
329 const SV * const svend = &sva[SvREFCNT(sva)];
330 if (p >= sv && p < svend) {
336 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL),
337 "Attempt to free non-arena SV: 0x%"UVxf
338 pTHX__FORMAT, PTR2UV(p) pTHX__VALUE);
345 #else /* ! DEBUGGING */
347 #define del_SV(p) plant_SV(p)
349 #endif /* DEBUGGING */
353 =head1 SV Manipulation Functions
355 =for apidoc sv_add_arena
357 Given a chunk of memory, link it to the head of the list of arenas,
358 and split it into a list of free SVs.
364 S_sv_add_arena(pTHX_ char *const ptr, const U32 size, const U32 flags)
367 SV *const sva = MUTABLE_SV(ptr);
371 PERL_ARGS_ASSERT_SV_ADD_ARENA;
373 /* The first SV in an arena isn't an SV. */
374 SvANY(sva) = (void *) PL_sv_arenaroot; /* ptr to next arena */
375 SvREFCNT(sva) = size / sizeof(SV); /* number of SV slots */
376 SvFLAGS(sva) = flags; /* FAKE if not to be freed */
378 PL_sv_arenaroot = sva;
379 PL_sv_root = sva + 1;
381 svend = &sva[SvREFCNT(sva) - 1];
384 SvARENA_CHAIN_SET(sv, (sv + 1));
388 /* Must always set typemask because it's always checked in on cleanup
389 when the arenas are walked looking for objects. */
390 SvFLAGS(sv) = SVTYPEMASK;
393 SvARENA_CHAIN_SET(sv, 0);
397 SvFLAGS(sv) = SVTYPEMASK;
400 /* visit(): call the named function for each non-free SV in the arenas
401 * whose flags field matches the flags/mask args. */
404 S_visit(pTHX_ SVFUNC_t f, const U32 flags, const U32 mask)
410 PERL_ARGS_ASSERT_VISIT;
412 for (sva = PL_sv_arenaroot; sva; sva = MUTABLE_SV(SvANY(sva))) {
413 register const SV * const svend = &sva[SvREFCNT(sva)];
415 for (sv = sva + 1; sv < svend; ++sv) {
416 if (SvTYPE(sv) != (svtype)SVTYPEMASK
417 && (sv->sv_flags & mask) == flags
430 /* called by sv_report_used() for each live SV */
433 do_report_used(pTHX_ SV *const sv)
435 if (SvTYPE(sv) != (svtype)SVTYPEMASK) {
436 PerlIO_printf(Perl_debug_log, "****\n");
443 =for apidoc sv_report_used
445 Dump the contents of all SVs not yet freed (debugging aid).
451 Perl_sv_report_used(pTHX)
454 visit(do_report_used, 0, 0);
460 /* called by sv_clean_objs() for each live SV */
463 do_clean_objs(pTHX_ SV *const ref)
468 SV * const target = SvRV(ref);
469 if (SvOBJECT(target)) {
470 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning object ref:\n "), sv_dump(ref)));
471 if (SvWEAKREF(ref)) {
472 sv_del_backref(target, ref);
478 SvREFCNT_dec(target);
483 /* XXX Might want to check arrays, etc. */
487 /* clear any slots in a GV which hold objects - except IO;
488 * called by sv_clean_objs() for each live GV */
491 do_clean_named_objs(pTHX_ SV *const sv)
495 assert(SvTYPE(sv) == SVt_PVGV);
496 assert(isGV_with_GP(sv));
500 /* freeing GP entries may indirectly free the current GV;
501 * hold onto it while we mess with the GP slots */
504 if ( ((obj = GvSV(sv) )) && SvOBJECT(obj)) {
505 DEBUG_D((PerlIO_printf(Perl_debug_log,
506 "Cleaning named glob SV object:\n "), sv_dump(obj)));
510 if ( ((obj = MUTABLE_SV(GvAV(sv)) )) && SvOBJECT(obj)) {
511 DEBUG_D((PerlIO_printf(Perl_debug_log,
512 "Cleaning named glob AV object:\n "), sv_dump(obj)));
516 if ( ((obj = MUTABLE_SV(GvHV(sv)) )) && SvOBJECT(obj)) {
517 DEBUG_D((PerlIO_printf(Perl_debug_log,
518 "Cleaning named glob HV object:\n "), sv_dump(obj)));
522 if ( ((obj = MUTABLE_SV(GvCV(sv)) )) && SvOBJECT(obj)) {
523 DEBUG_D((PerlIO_printf(Perl_debug_log,
524 "Cleaning named glob CV object:\n "), sv_dump(obj)));
528 SvREFCNT_dec(sv); /* undo the inc above */
531 /* clear any IO slots in a GV which hold objects (except stderr, defout);
532 * called by sv_clean_objs() for each live GV */
535 do_clean_named_io_objs(pTHX_ SV *const sv)
539 assert(SvTYPE(sv) == SVt_PVGV);
540 assert(isGV_with_GP(sv));
541 if (!GvGP(sv) || sv == (SV*)PL_stderrgv || sv == (SV*)PL_defoutgv)
545 if ( ((obj = MUTABLE_SV(GvIO(sv)) )) && SvOBJECT(obj)) {
546 DEBUG_D((PerlIO_printf(Perl_debug_log,
547 "Cleaning named glob IO object:\n "), sv_dump(obj)));
551 SvREFCNT_dec(sv); /* undo the inc above */
554 /* Void wrapper to pass to visit() */
556 do_curse(pTHX_ SV * const sv) {
557 if ((PL_stderrgv && GvGP(PL_stderrgv) && (SV*)GvIO(PL_stderrgv) == sv)
558 || (PL_defoutgv && GvGP(PL_defoutgv) && (SV*)GvIO(PL_defoutgv) == sv))
564 =for apidoc sv_clean_objs
566 Attempt to destroy all objects not yet freed.
572 Perl_sv_clean_objs(pTHX)
576 PL_in_clean_objs = TRUE;
577 visit(do_clean_objs, SVf_ROK, SVf_ROK);
578 /* Some barnacles may yet remain, clinging to typeglobs.
579 * Run the non-IO destructors first: they may want to output
580 * error messages, close files etc */
581 visit(do_clean_named_objs, SVt_PVGV|SVpgv_GP, SVTYPEMASK|SVp_POK|SVpgv_GP);
582 visit(do_clean_named_io_objs, SVt_PVGV|SVpgv_GP, SVTYPEMASK|SVp_POK|SVpgv_GP);
583 /* And if there are some very tenacious barnacles clinging to arrays,
584 closures, or what have you.... */
585 visit(do_curse, SVs_OBJECT, SVs_OBJECT);
586 olddef = PL_defoutgv;
587 PL_defoutgv = NULL; /* disable skip of PL_defoutgv */
588 if (olddef && isGV_with_GP(olddef))
589 do_clean_named_io_objs(aTHX_ MUTABLE_SV(olddef));
590 olderr = PL_stderrgv;
591 PL_stderrgv = NULL; /* disable skip of PL_stderrgv */
592 if (olderr && isGV_with_GP(olderr))
593 do_clean_named_io_objs(aTHX_ MUTABLE_SV(olderr));
594 SvREFCNT_dec(olddef);
595 PL_in_clean_objs = FALSE;
598 /* called by sv_clean_all() for each live SV */
601 do_clean_all(pTHX_ SV *const sv)
604 if (sv == (const SV *) PL_fdpid || sv == (const SV *)PL_strtab) {
605 /* don't clean pid table and strtab */
608 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning loops: SV at 0x%"UVxf"\n", PTR2UV(sv)) ));
609 SvFLAGS(sv) |= SVf_BREAK;
614 =for apidoc sv_clean_all
616 Decrement the refcnt of each remaining SV, possibly triggering a
617 cleanup. This function may have to be called multiple times to free
618 SVs which are in complex self-referential hierarchies.
624 Perl_sv_clean_all(pTHX)
628 PL_in_clean_all = TRUE;
629 cleaned = visit(do_clean_all, 0,0);
634 ARENASETS: a meta-arena implementation which separates arena-info
635 into struct arena_set, which contains an array of struct
636 arena_descs, each holding info for a single arena. By separating
637 the meta-info from the arena, we recover the 1st slot, formerly
638 borrowed for list management. The arena_set is about the size of an
639 arena, avoiding the needless malloc overhead of a naive linked-list.
641 The cost is 1 arena-set malloc per ~320 arena-mallocs, + the unused
642 memory in the last arena-set (1/2 on average). In trade, we get
643 back the 1st slot in each arena (ie 1.7% of a CV-arena, less for
644 smaller types). The recovery of the wasted space allows use of
645 small arenas for large, rare body types, by changing array* fields
646 in body_details_by_type[] below.
649 char *arena; /* the raw storage, allocated aligned */
650 size_t size; /* its size ~4k typ */
651 svtype utype; /* bodytype stored in arena */
656 /* Get the maximum number of elements in set[] such that struct arena_set
657 will fit within PERL_ARENA_SIZE, which is probably just under 4K, and
658 therefore likely to be 1 aligned memory page. */
660 #define ARENAS_PER_SET ((PERL_ARENA_SIZE - sizeof(struct arena_set*) \
661 - 2 * sizeof(int)) / sizeof (struct arena_desc))
664 struct arena_set* next;
665 unsigned int set_size; /* ie ARENAS_PER_SET */
666 unsigned int curr; /* index of next available arena-desc */
667 struct arena_desc set[ARENAS_PER_SET];
671 =for apidoc sv_free_arenas
673 Deallocate the memory used by all arenas. Note that all the individual SV
674 heads and bodies within the arenas must already have been freed.
679 Perl_sv_free_arenas(pTHX)
686 /* Free arenas here, but be careful about fake ones. (We assume
687 contiguity of the fake ones with the corresponding real ones.) */
689 for (sva = PL_sv_arenaroot; sva; sva = svanext) {
690 svanext = MUTABLE_SV(SvANY(sva));
691 while (svanext && SvFAKE(svanext))
692 svanext = MUTABLE_SV(SvANY(svanext));
699 struct arena_set *aroot = (struct arena_set*) PL_body_arenas;
702 struct arena_set *current = aroot;
705 assert(aroot->set[i].arena);
706 Safefree(aroot->set[i].arena);
714 i = PERL_ARENA_ROOTS_SIZE;
716 PL_body_roots[i] = 0;
723 Here are mid-level routines that manage the allocation of bodies out
724 of the various arenas. There are 5 kinds of arenas:
726 1. SV-head arenas, which are discussed and handled above
727 2. regular body arenas
728 3. arenas for reduced-size bodies
731 Arena types 2 & 3 are chained by body-type off an array of
732 arena-root pointers, which is indexed by svtype. Some of the
733 larger/less used body types are malloced singly, since a large
734 unused block of them is wasteful. Also, several svtypes dont have
735 bodies; the data fits into the sv-head itself. The arena-root
736 pointer thus has a few unused root-pointers (which may be hijacked
737 later for arena types 4,5)
739 3 differs from 2 as an optimization; some body types have several
740 unused fields in the front of the structure (which are kept in-place
741 for consistency). These bodies can be allocated in smaller chunks,
742 because the leading fields arent accessed. Pointers to such bodies
743 are decremented to point at the unused 'ghost' memory, knowing that
744 the pointers are used with offsets to the real memory.
747 =head1 SV-Body Allocation
749 Allocation of SV-bodies is similar to SV-heads, differing as follows;
750 the allocation mechanism is used for many body types, so is somewhat
751 more complicated, it uses arena-sets, and has no need for still-live
754 At the outermost level, (new|del)_X*V macros return bodies of the
755 appropriate type. These macros call either (new|del)_body_type or
756 (new|del)_body_allocated macro pairs, depending on specifics of the
757 type. Most body types use the former pair, the latter pair is used to
758 allocate body types with "ghost fields".
760 "ghost fields" are fields that are unused in certain types, and
761 consequently don't need to actually exist. They are declared because
762 they're part of a "base type", which allows use of functions as
763 methods. The simplest examples are AVs and HVs, 2 aggregate types
764 which don't use the fields which support SCALAR semantics.
766 For these types, the arenas are carved up into appropriately sized
767 chunks, we thus avoid wasted memory for those unaccessed members.
768 When bodies are allocated, we adjust the pointer back in memory by the
769 size of the part not allocated, so it's as if we allocated the full
770 structure. (But things will all go boom if you write to the part that
771 is "not there", because you'll be overwriting the last members of the
772 preceding structure in memory.)
774 We calculate the correction using the STRUCT_OFFSET macro on the first
775 member present. If the allocated structure is smaller (no initial NV
776 actually allocated) then the net effect is to subtract the size of the NV
777 from the pointer, to return a new pointer as if an initial NV were actually
778 allocated. (We were using structures named *_allocated for this, but
779 this turned out to be a subtle bug, because a structure without an NV
780 could have a lower alignment constraint, but the compiler is allowed to
781 optimised accesses based on the alignment constraint of the actual pointer
782 to the full structure, for example, using a single 64 bit load instruction
783 because it "knows" that two adjacent 32 bit members will be 8-byte aligned.)
785 This is the same trick as was used for NV and IV bodies. Ironically it
786 doesn't need to be used for NV bodies any more, because NV is now at
787 the start of the structure. IV bodies don't need it either, because
788 they are no longer allocated.
790 In turn, the new_body_* allocators call S_new_body(), which invokes
791 new_body_inline macro, which takes a lock, and takes a body off the
792 linked list at PL_body_roots[sv_type], calling Perl_more_bodies() if
793 necessary to refresh an empty list. Then the lock is released, and
794 the body is returned.
796 Perl_more_bodies allocates a new arena, and carves it up into an array of N
797 bodies, which it strings into a linked list. It looks up arena-size
798 and body-size from the body_details table described below, thus
799 supporting the multiple body-types.
801 If PURIFY is defined, or PERL_ARENA_SIZE=0, arenas are not used, and
802 the (new|del)_X*V macros are mapped directly to malloc/free.
804 For each sv-type, struct body_details bodies_by_type[] carries
805 parameters which control these aspects of SV handling:
807 Arena_size determines whether arenas are used for this body type, and if
808 so, how big they are. PURIFY or PERL_ARENA_SIZE=0 set this field to
809 zero, forcing individual mallocs and frees.
811 Body_size determines how big a body is, and therefore how many fit into
812 each arena. Offset carries the body-pointer adjustment needed for
813 "ghost fields", and is used in *_allocated macros.
815 But its main purpose is to parameterize info needed in
816 Perl_sv_upgrade(). The info here dramatically simplifies the function
817 vs the implementation in 5.8.8, making it table-driven. All fields
818 are used for this, except for arena_size.
820 For the sv-types that have no bodies, arenas are not used, so those
821 PL_body_roots[sv_type] are unused, and can be overloaded. In
822 something of a special case, SVt_NULL is borrowed for HE arenas;
823 PL_body_roots[HE_SVSLOT=SVt_NULL] is filled by S_more_he, but the
824 bodies_by_type[SVt_NULL] slot is not used, as the table is not
829 struct body_details {
830 U8 body_size; /* Size to allocate */
831 U8 copy; /* Size of structure to copy (may be shorter) */
833 unsigned int type : 4; /* We have space for a sanity check. */
834 unsigned int cant_upgrade : 1; /* Cannot upgrade this type */
835 unsigned int zero_nv : 1; /* zero the NV when upgrading from this */
836 unsigned int arena : 1; /* Allocated from an arena */
837 size_t arena_size; /* Size of arena to allocate */
845 /* With -DPURFIY we allocate everything directly, and don't use arenas.
846 This seems a rather elegant way to simplify some of the code below. */
847 #define HASARENA FALSE
849 #define HASARENA TRUE
851 #define NOARENA FALSE
853 /* Size the arenas to exactly fit a given number of bodies. A count
854 of 0 fits the max number bodies into a PERL_ARENA_SIZE.block,
855 simplifying the default. If count > 0, the arena is sized to fit
856 only that many bodies, allowing arenas to be used for large, rare
857 bodies (XPVFM, XPVIO) without undue waste. The arena size is
858 limited by PERL_ARENA_SIZE, so we can safely oversize the
861 #define FIT_ARENA0(body_size) \
862 ((size_t)(PERL_ARENA_SIZE / body_size) * body_size)
863 #define FIT_ARENAn(count,body_size) \
864 ( count * body_size <= PERL_ARENA_SIZE) \
865 ? count * body_size \
866 : FIT_ARENA0 (body_size)
867 #define FIT_ARENA(count,body_size) \
869 ? FIT_ARENAn (count, body_size) \
870 : FIT_ARENA0 (body_size)
872 /* Calculate the length to copy. Specifically work out the length less any
873 final padding the compiler needed to add. See the comment in sv_upgrade
874 for why copying the padding proved to be a bug. */
876 #define copy_length(type, last_member) \
877 STRUCT_OFFSET(type, last_member) \
878 + sizeof (((type*)SvANY((const SV *)0))->last_member)
880 static const struct body_details bodies_by_type[] = {
881 /* HEs use this offset for their arena. */
882 { 0, 0, 0, SVt_NULL, FALSE, NONV, NOARENA, 0 },
884 /* The bind placeholder pretends to be an RV for now.
885 Also it's marked as "can't upgrade" to stop anyone using it before it's
887 { 0, 0, 0, SVt_BIND, TRUE, NONV, NOARENA, 0 },
889 /* IVs are in the head, so the allocation size is 0. */
891 sizeof(IV), /* This is used to copy out the IV body. */
892 STRUCT_OFFSET(XPVIV, xiv_iv), SVt_IV, FALSE, NONV,
893 NOARENA /* IVS don't need an arena */, 0
896 { sizeof(NV), sizeof(NV),
897 STRUCT_OFFSET(XPVNV, xnv_u),
898 SVt_NV, FALSE, HADNV, HASARENA, FIT_ARENA(0, sizeof(NV)) },
900 { sizeof(XPV) - STRUCT_OFFSET(XPV, xpv_cur),
901 copy_length(XPV, xpv_len) - STRUCT_OFFSET(XPV, xpv_cur),
902 + STRUCT_OFFSET(XPV, xpv_cur),
903 SVt_PV, FALSE, NONV, HASARENA,
904 FIT_ARENA(0, sizeof(XPV) - STRUCT_OFFSET(XPV, xpv_cur)) },
906 { sizeof(XPVIV) - STRUCT_OFFSET(XPV, xpv_cur),
907 copy_length(XPVIV, xiv_u) - STRUCT_OFFSET(XPV, xpv_cur),
908 + STRUCT_OFFSET(XPV, xpv_cur),
909 SVt_PVIV, FALSE, NONV, HASARENA,
910 FIT_ARENA(0, sizeof(XPVIV) - STRUCT_OFFSET(XPV, xpv_cur)) },
912 { sizeof(XPVNV) - STRUCT_OFFSET(XPV, xpv_cur),
913 copy_length(XPVNV, xnv_u) - STRUCT_OFFSET(XPV, xpv_cur),
914 + STRUCT_OFFSET(XPV, xpv_cur),
915 SVt_PVNV, FALSE, HADNV, HASARENA,
916 FIT_ARENA(0, sizeof(XPVNV) - STRUCT_OFFSET(XPV, xpv_cur)) },
918 { sizeof(XPVMG), copy_length(XPVMG, xnv_u), 0, SVt_PVMG, FALSE, HADNV,
919 HASARENA, FIT_ARENA(0, sizeof(XPVMG)) },
924 SVt_REGEXP, FALSE, NONV, HASARENA,
925 FIT_ARENA(0, sizeof(regexp))
928 { sizeof(XPVGV), sizeof(XPVGV), 0, SVt_PVGV, TRUE, HADNV,
929 HASARENA, FIT_ARENA(0, sizeof(XPVGV)) },
931 { sizeof(XPVLV), sizeof(XPVLV), 0, SVt_PVLV, TRUE, HADNV,
932 HASARENA, FIT_ARENA(0, sizeof(XPVLV)) },
935 copy_length(XPVAV, xav_alloc),
937 SVt_PVAV, TRUE, NONV, HASARENA,
938 FIT_ARENA(0, sizeof(XPVAV)) },
941 copy_length(XPVHV, xhv_max),
943 SVt_PVHV, TRUE, NONV, HASARENA,
944 FIT_ARENA(0, sizeof(XPVHV)) },
949 SVt_PVCV, TRUE, NONV, HASARENA,
950 FIT_ARENA(0, sizeof(XPVCV)) },
955 SVt_PVFM, TRUE, NONV, NOARENA,
956 FIT_ARENA(20, sizeof(XPVFM)) },
961 SVt_PVIO, TRUE, NONV, HASARENA,
962 FIT_ARENA(24, sizeof(XPVIO)) },
965 #define new_body_allocated(sv_type) \
966 (void *)((char *)S_new_body(aTHX_ sv_type) \
967 - bodies_by_type[sv_type].offset)
969 /* return a thing to the free list */
971 #define del_body(thing, root) \
973 void ** const thing_copy = (void **)thing; \
974 *thing_copy = *root; \
975 *root = (void*)thing_copy; \
980 #define new_XNV() safemalloc(sizeof(XPVNV))
981 #define new_XPVNV() safemalloc(sizeof(XPVNV))
982 #define new_XPVMG() safemalloc(sizeof(XPVMG))
984 #define del_XPVGV(p) safefree(p)
988 #define new_XNV() new_body_allocated(SVt_NV)
989 #define new_XPVNV() new_body_allocated(SVt_PVNV)
990 #define new_XPVMG() new_body_allocated(SVt_PVMG)
992 #define del_XPVGV(p) del_body(p + bodies_by_type[SVt_PVGV].offset, \
993 &PL_body_roots[SVt_PVGV])
997 /* no arena for you! */
999 #define new_NOARENA(details) \
1000 safemalloc((details)->body_size + (details)->offset)
1001 #define new_NOARENAZ(details) \
1002 safecalloc((details)->body_size + (details)->offset, 1)
1005 Perl_more_bodies (pTHX_ const svtype sv_type, const size_t body_size,
1006 const size_t arena_size)
1009 void ** const root = &PL_body_roots[sv_type];
1010 struct arena_desc *adesc;
1011 struct arena_set *aroot = (struct arena_set *) PL_body_arenas;
1015 const size_t good_arena_size = Perl_malloc_good_size(arena_size);
1016 #if defined(DEBUGGING) && !defined(PERL_GLOBAL_STRUCT_PRIVATE)
1017 static bool done_sanity_check;
1019 /* PERL_GLOBAL_STRUCT_PRIVATE cannot coexist with global
1020 * variables like done_sanity_check. */
1021 if (!done_sanity_check) {
1022 unsigned int i = SVt_LAST;
1024 done_sanity_check = TRUE;
1027 assert (bodies_by_type[i].type == i);
1033 /* may need new arena-set to hold new arena */
1034 if (!aroot || aroot->curr >= aroot->set_size) {
1035 struct arena_set *newroot;
1036 Newxz(newroot, 1, struct arena_set);
1037 newroot->set_size = ARENAS_PER_SET;
1038 newroot->next = aroot;
1040 PL_body_arenas = (void *) newroot;
1041 DEBUG_m(PerlIO_printf(Perl_debug_log, "new arenaset %p\n", (void*)aroot));
1044 /* ok, now have arena-set with at least 1 empty/available arena-desc */
1045 curr = aroot->curr++;
1046 adesc = &(aroot->set[curr]);
1047 assert(!adesc->arena);
1049 Newx(adesc->arena, good_arena_size, char);
1050 adesc->size = good_arena_size;
1051 adesc->utype = sv_type;
1052 DEBUG_m(PerlIO_printf(Perl_debug_log, "arena %d added: %p size %"UVuf"\n",
1053 curr, (void*)adesc->arena, (UV)good_arena_size));
1055 start = (char *) adesc->arena;
1057 /* Get the address of the byte after the end of the last body we can fit.
1058 Remember, this is integer division: */
1059 end = start + good_arena_size / body_size * body_size;
1061 /* computed count doesn't reflect the 1st slot reservation */
1062 #if defined(MYMALLOC) || defined(HAS_MALLOC_GOOD_SIZE)
1063 DEBUG_m(PerlIO_printf(Perl_debug_log,
1064 "arena %p end %p arena-size %d (from %d) type %d "
1066 (void*)start, (void*)end, (int)good_arena_size,
1067 (int)arena_size, sv_type, (int)body_size,
1068 (int)good_arena_size / (int)body_size));
1070 DEBUG_m(PerlIO_printf(Perl_debug_log,
1071 "arena %p end %p arena-size %d type %d size %d ct %d\n",
1072 (void*)start, (void*)end,
1073 (int)arena_size, sv_type, (int)body_size,
1074 (int)good_arena_size / (int)body_size));
1076 *root = (void *)start;
1079 /* Where the next body would start: */
1080 char * const next = start + body_size;
1083 /* This is the last body: */
1084 assert(next == end);
1086 *(void **)start = 0;
1090 *(void**) start = (void *)next;
1095 /* grab a new thing from the free list, allocating more if necessary.
1096 The inline version is used for speed in hot routines, and the
1097 function using it serves the rest (unless PURIFY).
1099 #define new_body_inline(xpv, sv_type) \
1101 void ** const r3wt = &PL_body_roots[sv_type]; \
1102 xpv = (PTR_TBL_ENT_t*) (*((void **)(r3wt)) \
1103 ? *((void **)(r3wt)) : Perl_more_bodies(aTHX_ sv_type, \
1104 bodies_by_type[sv_type].body_size,\
1105 bodies_by_type[sv_type].arena_size)); \
1106 *(r3wt) = *(void**)(xpv); \
1112 S_new_body(pTHX_ const svtype sv_type)
1116 new_body_inline(xpv, sv_type);
1122 static const struct body_details fake_rv =
1123 { 0, 0, 0, SVt_IV, FALSE, NONV, NOARENA, 0 };
1126 =for apidoc sv_upgrade
1128 Upgrade an SV to a more complex form. Generally adds a new body type to the
1129 SV, then copies across as much information as possible from the old body.
1130 It croaks if the SV is already in a more complex form than requested. You
1131 generally want to use the C<SvUPGRADE> macro wrapper, which checks the type
1132 before calling C<sv_upgrade>, and hence does not croak. See also
1139 Perl_sv_upgrade(pTHX_ register SV *const sv, svtype new_type)
1144 const svtype old_type = SvTYPE(sv);
1145 const struct body_details *new_type_details;
1146 const struct body_details *old_type_details
1147 = bodies_by_type + old_type;
1148 SV *referant = NULL;
1150 PERL_ARGS_ASSERT_SV_UPGRADE;
1152 if (old_type == new_type)
1155 /* This clause was purposefully added ahead of the early return above to
1156 the shared string hackery for (sort {$a <=> $b} keys %hash), with the
1157 inference by Nick I-S that it would fix other troublesome cases. See
1158 changes 7162, 7163 (f130fd4589cf5fbb24149cd4db4137c8326f49c1 and parent)
1160 Given that shared hash key scalars are no longer PVIV, but PV, there is
1161 no longer need to unshare so as to free up the IVX slot for its proper
1162 purpose. So it's safe to move the early return earlier. */
1164 if (new_type != SVt_PV && SvIsCOW(sv)) {
1165 sv_force_normal_flags(sv, 0);
1168 old_body = SvANY(sv);
1170 /* Copying structures onto other structures that have been neatly zeroed
1171 has a subtle gotcha. Consider XPVMG
1173 +------+------+------+------+------+-------+-------+
1174 | NV | CUR | LEN | IV | MAGIC | STASH |
1175 +------+------+------+------+------+-------+-------+
1176 0 4 8 12 16 20 24 28
1178 where NVs are aligned to 8 bytes, so that sizeof that structure is
1179 actually 32 bytes long, with 4 bytes of padding at the end:
1181 +------+------+------+------+------+-------+-------+------+
1182 | NV | CUR | LEN | IV | MAGIC | STASH | ??? |
1183 +------+------+------+------+------+-------+-------+------+
1184 0 4 8 12 16 20 24 28 32
1186 so what happens if you allocate memory for this structure:
1188 +------+------+------+------+------+-------+-------+------+------+...
1189 | NV | CUR | LEN | IV | MAGIC | STASH | GP | NAME |
1190 +------+------+------+------+------+-------+-------+------+------+...
1191 0 4 8 12 16 20 24 28 32 36
1193 zero it, then copy sizeof(XPVMG) bytes on top of it? Not quite what you
1194 expect, because you copy the area marked ??? onto GP. Now, ??? may have
1195 started out as zero once, but it's quite possible that it isn't. So now,
1196 rather than a nicely zeroed GP, you have it pointing somewhere random.
1199 (In fact, GP ends up pointing at a previous GP structure, because the
1200 principle cause of the padding in XPVMG getting garbage is a copy of
1201 sizeof(XPVMG) bytes from a XPVGV structure in sv_unglob. Right now
1202 this happens to be moot because XPVGV has been re-ordered, with GP
1203 no longer after STASH)
1205 So we are careful and work out the size of used parts of all the
1213 referant = SvRV(sv);
1214 old_type_details = &fake_rv;
1215 if (new_type == SVt_NV)
1216 new_type = SVt_PVNV;
1218 if (new_type < SVt_PVIV) {
1219 new_type = (new_type == SVt_NV)
1220 ? SVt_PVNV : SVt_PVIV;
1225 if (new_type < SVt_PVNV) {
1226 new_type = SVt_PVNV;
1230 assert(new_type > SVt_PV);
1231 assert(SVt_IV < SVt_PV);
1232 assert(SVt_NV < SVt_PV);
1239 /* Because the XPVMG of PL_mess_sv isn't allocated from the arena,
1240 there's no way that it can be safely upgraded, because perl.c
1241 expects to Safefree(SvANY(PL_mess_sv)) */
1242 assert(sv != PL_mess_sv);
1243 /* This flag bit is used to mean other things in other scalar types.
1244 Given that it only has meaning inside the pad, it shouldn't be set
1245 on anything that can get upgraded. */
1246 assert(!SvPAD_TYPED(sv));
1249 if (old_type_details->cant_upgrade)
1250 Perl_croak(aTHX_ "Can't upgrade %s (%" UVuf ") to %" UVuf,
1251 sv_reftype(sv, 0), (UV) old_type, (UV) new_type);
1254 if (old_type > new_type)
1255 Perl_croak(aTHX_ "sv_upgrade from type %d down to type %d",
1256 (int)old_type, (int)new_type);
1258 new_type_details = bodies_by_type + new_type;
1260 SvFLAGS(sv) &= ~SVTYPEMASK;
1261 SvFLAGS(sv) |= new_type;
1263 /* This can't happen, as SVt_NULL is <= all values of new_type, so one of
1264 the return statements above will have triggered. */
1265 assert (new_type != SVt_NULL);
1268 assert(old_type == SVt_NULL);
1269 SvANY(sv) = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
1273 assert(old_type == SVt_NULL);
1274 SvANY(sv) = new_XNV();
1279 assert(new_type_details->body_size);
1282 assert(new_type_details->arena);
1283 assert(new_type_details->arena_size);
1284 /* This points to the start of the allocated area. */
1285 new_body_inline(new_body, new_type);
1286 Zero(new_body, new_type_details->body_size, char);
1287 new_body = ((char *)new_body) - new_type_details->offset;
1289 /* We always allocated the full length item with PURIFY. To do this
1290 we fake things so that arena is false for all 16 types.. */
1291 new_body = new_NOARENAZ(new_type_details);
1293 SvANY(sv) = new_body;
1294 if (new_type == SVt_PVAV) {
1298 if (old_type_details->body_size) {
1301 /* It will have been zeroed when the new body was allocated.
1302 Lets not write to it, in case it confuses a write-back
1308 #ifndef NODEFAULT_SHAREKEYS
1309 HvSHAREKEYS_on(sv); /* key-sharing on by default */
1311 HvMAX(sv) = 7; /* (start with 8 buckets) */
1314 /* SVt_NULL isn't the only thing upgraded to AV or HV.
1315 The target created by newSVrv also is, and it can have magic.
1316 However, it never has SvPVX set.
1318 if (old_type == SVt_IV) {
1320 } else if (old_type >= SVt_PV) {
1321 assert(SvPVX_const(sv) == 0);
1324 if (old_type >= SVt_PVMG) {
1325 SvMAGIC_set(sv, ((XPVMG*)old_body)->xmg_u.xmg_magic);
1326 SvSTASH_set(sv, ((XPVMG*)old_body)->xmg_stash);
1328 sv->sv_u.svu_array = NULL; /* or svu_hash */
1334 /* This ensures that SvTHINKFIRST(sv) is true, and hence that
1335 sv_force_normal_flags(sv) is called. */
1338 /* XXX Is this still needed? Was it ever needed? Surely as there is
1339 no route from NV to PVIV, NOK can never be true */
1340 assert(!SvNOKp(sv));
1351 assert(new_type_details->body_size);
1352 /* We always allocated the full length item with PURIFY. To do this
1353 we fake things so that arena is false for all 16 types.. */
1354 if(new_type_details->arena) {
1355 /* This points to the start of the allocated area. */
1356 new_body_inline(new_body, new_type);
1357 Zero(new_body, new_type_details->body_size, char);
1358 new_body = ((char *)new_body) - new_type_details->offset;
1360 new_body = new_NOARENAZ(new_type_details);
1362 SvANY(sv) = new_body;
1364 if (old_type_details->copy) {
1365 /* There is now the potential for an upgrade from something without
1366 an offset (PVNV or PVMG) to something with one (PVCV, PVFM) */
1367 int offset = old_type_details->offset;
1368 int length = old_type_details->copy;
1370 if (new_type_details->offset > old_type_details->offset) {
1371 const int difference
1372 = new_type_details->offset - old_type_details->offset;
1373 offset += difference;
1374 length -= difference;
1376 assert (length >= 0);
1378 Copy((char *)old_body + offset, (char *)new_body + offset, length,
1382 #ifndef NV_ZERO_IS_ALLBITS_ZERO
1383 /* If NV 0.0 is stores as all bits 0 then Zero() already creates a
1384 * correct 0.0 for us. Otherwise, if the old body didn't have an
1385 * NV slot, but the new one does, then we need to initialise the
1386 * freshly created NV slot with whatever the correct bit pattern is
1388 if (old_type_details->zero_nv && !new_type_details->zero_nv
1389 && !isGV_with_GP(sv))
1393 if (new_type == SVt_PVIO) {
1394 IO * const io = MUTABLE_IO(sv);
1395 GV *iogv = gv_fetchpvs("IO::File::", GV_ADD, SVt_PVHV);
1398 /* Clear the stashcache because a new IO could overrule a package
1400 hv_clear(PL_stashcache);
1402 SvSTASH_set(io, MUTABLE_HV(SvREFCNT_inc(GvHV(iogv))));
1403 IoPAGE_LEN(sv) = 60;
1405 if (old_type < SVt_PV) {
1406 /* referant will be NULL unless the old type was SVt_IV emulating
1408 sv->sv_u.svu_rv = referant;
1412 Perl_croak(aTHX_ "panic: sv_upgrade to unknown type %lu",
1413 (unsigned long)new_type);
1416 if (old_type > SVt_IV) {
1420 /* Note that there is an assumption that all bodies of types that
1421 can be upgraded came from arenas. Only the more complex non-
1422 upgradable types are allowed to be directly malloc()ed. */
1423 assert(old_type_details->arena);
1424 del_body((void*)((char*)old_body + old_type_details->offset),
1425 &PL_body_roots[old_type]);
1431 =for apidoc sv_backoff
1433 Remove any string offset. You should normally use the C<SvOOK_off> macro
1440 Perl_sv_backoff(pTHX_ register SV *const sv)
1443 const char * const s = SvPVX_const(sv);
1445 PERL_ARGS_ASSERT_SV_BACKOFF;
1446 PERL_UNUSED_CONTEXT;
1449 assert(SvTYPE(sv) != SVt_PVHV);
1450 assert(SvTYPE(sv) != SVt_PVAV);
1452 SvOOK_offset(sv, delta);
1454 SvLEN_set(sv, SvLEN(sv) + delta);
1455 SvPV_set(sv, SvPVX(sv) - delta);
1456 Move(s, SvPVX(sv), SvCUR(sv)+1, char);
1457 SvFLAGS(sv) &= ~SVf_OOK;
1464 Expands the character buffer in the SV. If necessary, uses C<sv_unref> and
1465 upgrades the SV to C<SVt_PV>. Returns a pointer to the character buffer.
1466 Use the C<SvGROW> wrapper instead.
1472 Perl_sv_grow(pTHX_ register SV *const sv, register STRLEN newlen)
1476 PERL_ARGS_ASSERT_SV_GROW;
1478 if (PL_madskills && newlen >= 0x100000) {
1479 PerlIO_printf(Perl_debug_log,
1480 "Allocation too large: %"UVxf"\n", (UV)newlen);
1482 #ifdef HAS_64K_LIMIT
1483 if (newlen >= 0x10000) {
1484 PerlIO_printf(Perl_debug_log,
1485 "Allocation too large: %"UVxf"\n", (UV)newlen);
1488 #endif /* HAS_64K_LIMIT */
1491 if (SvTYPE(sv) < SVt_PV) {
1492 sv_upgrade(sv, SVt_PV);
1493 s = SvPVX_mutable(sv);
1495 else if (SvOOK(sv)) { /* pv is offset? */
1497 s = SvPVX_mutable(sv);
1498 if (newlen > SvLEN(sv))
1499 newlen += 10 * (newlen - SvCUR(sv)); /* avoid copy each time */
1500 #ifdef HAS_64K_LIMIT
1501 if (newlen >= 0x10000)
1506 s = SvPVX_mutable(sv);
1508 if (newlen > SvLEN(sv)) { /* need more room? */
1509 STRLEN minlen = SvCUR(sv);
1510 minlen += (minlen >> PERL_STRLEN_EXPAND_SHIFT) + 10;
1511 if (newlen < minlen)
1513 #ifndef Perl_safesysmalloc_size
1514 newlen = PERL_STRLEN_ROUNDUP(newlen);
1516 if (SvLEN(sv) && s) {
1517 s = (char*)saferealloc(s, newlen);
1520 s = (char*)safemalloc(newlen);
1521 if (SvPVX_const(sv) && SvCUR(sv)) {
1522 Move(SvPVX_const(sv), s, (newlen < SvCUR(sv)) ? newlen : SvCUR(sv), char);
1526 #ifdef Perl_safesysmalloc_size
1527 /* Do this here, do it once, do it right, and then we will never get
1528 called back into sv_grow() unless there really is some growing
1530 SvLEN_set(sv, Perl_safesysmalloc_size(s));
1532 SvLEN_set(sv, newlen);
1539 =for apidoc sv_setiv
1541 Copies an integer into the given SV, upgrading first if necessary.
1542 Does not handle 'set' magic. See also C<sv_setiv_mg>.
1548 Perl_sv_setiv(pTHX_ register SV *const sv, const IV i)
1552 PERL_ARGS_ASSERT_SV_SETIV;
1554 SV_CHECK_THINKFIRST_COW_DROP(sv);
1555 switch (SvTYPE(sv)) {
1558 sv_upgrade(sv, SVt_IV);
1561 sv_upgrade(sv, SVt_PVIV);
1565 if (!isGV_with_GP(sv))
1572 /* diag_listed_as: Can't coerce %s to %s in %s */
1573 Perl_croak(aTHX_ "Can't coerce %s to integer in %s", sv_reftype(sv,0),
1577 (void)SvIOK_only(sv); /* validate number */
1583 =for apidoc sv_setiv_mg
1585 Like C<sv_setiv>, but also handles 'set' magic.
1591 Perl_sv_setiv_mg(pTHX_ register SV *const sv, const IV i)
1593 PERL_ARGS_ASSERT_SV_SETIV_MG;
1600 =for apidoc sv_setuv
1602 Copies an unsigned integer into the given SV, upgrading first if necessary.
1603 Does not handle 'set' magic. See also C<sv_setuv_mg>.
1609 Perl_sv_setuv(pTHX_ register SV *const sv, const UV u)
1611 PERL_ARGS_ASSERT_SV_SETUV;
1613 /* With the if statement to ensure that integers are stored as IVs whenever
1615 u=1.49 s=0.52 cu=72.49 cs=10.64 scripts=270 tests=20865
1618 u=1.35 s=0.47 cu=73.45 cs=11.43 scripts=270 tests=20865
1620 If you wish to remove the following if statement, so that this routine
1621 (and its callers) always return UVs, please benchmark to see what the
1622 effect is. Modern CPUs may be different. Or may not :-)
1624 if (u <= (UV)IV_MAX) {
1625 sv_setiv(sv, (IV)u);
1634 =for apidoc sv_setuv_mg
1636 Like C<sv_setuv>, but also handles 'set' magic.
1642 Perl_sv_setuv_mg(pTHX_ register SV *const sv, const UV u)
1644 PERL_ARGS_ASSERT_SV_SETUV_MG;
1651 =for apidoc sv_setnv
1653 Copies a double into the given SV, upgrading first if necessary.
1654 Does not handle 'set' magic. See also C<sv_setnv_mg>.
1660 Perl_sv_setnv(pTHX_ register SV *const sv, const NV num)
1664 PERL_ARGS_ASSERT_SV_SETNV;
1666 SV_CHECK_THINKFIRST_COW_DROP(sv);
1667 switch (SvTYPE(sv)) {
1670 sv_upgrade(sv, SVt_NV);
1674 sv_upgrade(sv, SVt_PVNV);
1678 if (!isGV_with_GP(sv))
1685 /* diag_listed_as: Can't coerce %s to %s in %s */
1686 Perl_croak(aTHX_ "Can't coerce %s to number in %s", sv_reftype(sv,0),
1691 (void)SvNOK_only(sv); /* validate number */
1696 =for apidoc sv_setnv_mg
1698 Like C<sv_setnv>, but also handles 'set' magic.
1704 Perl_sv_setnv_mg(pTHX_ register SV *const sv, const NV num)
1706 PERL_ARGS_ASSERT_SV_SETNV_MG;
1712 /* Print an "isn't numeric" warning, using a cleaned-up,
1713 * printable version of the offending string
1717 S_not_a_number(pTHX_ SV *const sv)
1724 PERL_ARGS_ASSERT_NOT_A_NUMBER;
1727 dsv = newSVpvs_flags("", SVs_TEMP);
1728 pv = sv_uni_display(dsv, sv, 10, UNI_DISPLAY_ISPRINT);
1731 const char * const limit = tmpbuf + sizeof(tmpbuf) - 8;
1732 /* each *s can expand to 4 chars + "...\0",
1733 i.e. need room for 8 chars */
1735 const char *s = SvPVX_const(sv);
1736 const char * const end = s + SvCUR(sv);
1737 for ( ; s < end && d < limit; s++ ) {
1739 if (ch & 128 && !isPRINT_LC(ch)) {
1748 else if (ch == '\r') {
1752 else if (ch == '\f') {
1756 else if (ch == '\\') {
1760 else if (ch == '\0') {
1764 else if (isPRINT_LC(ch))
1781 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
1782 /* diag_listed_as: Argument "%s" isn't numeric%s */
1783 "Argument \"%s\" isn't numeric in %s", pv,
1786 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
1787 /* diag_listed_as: Argument "%s" isn't numeric%s */
1788 "Argument \"%s\" isn't numeric", pv);
1792 =for apidoc looks_like_number
1794 Test if the content of an SV looks like a number (or is a number).
1795 C<Inf> and C<Infinity> are treated as numbers (so will not issue a
1796 non-numeric warning), even if your atof() doesn't grok them. Get-magic is
1803 Perl_looks_like_number(pTHX_ SV *const sv)
1805 register const char *sbegin;
1808 PERL_ARGS_ASSERT_LOOKS_LIKE_NUMBER;
1810 if (SvPOK(sv) || SvPOKp(sv)) {
1811 sbegin = SvPV_nomg_const(sv, len);
1814 return SvFLAGS(sv) & (SVf_NOK|SVp_NOK|SVf_IOK|SVp_IOK);
1815 return grok_number(sbegin, len, NULL);
1819 S_glob_2number(pTHX_ GV * const gv)
1821 PERL_ARGS_ASSERT_GLOB_2NUMBER;
1823 /* We know that all GVs stringify to something that is not-a-number,
1824 so no need to test that. */
1825 if (ckWARN(WARN_NUMERIC))
1827 SV *const buffer = sv_newmortal();
1828 gv_efullname3(buffer, gv, "*");
1829 not_a_number(buffer);
1831 /* We just want something true to return, so that S_sv_2iuv_common
1832 can tail call us and return true. */
1836 /* Actually, ISO C leaves conversion of UV to IV undefined, but
1837 until proven guilty, assume that things are not that bad... */
1842 As 64 bit platforms often have an NV that doesn't preserve all bits of
1843 an IV (an assumption perl has been based on to date) it becomes necessary
1844 to remove the assumption that the NV always carries enough precision to
1845 recreate the IV whenever needed, and that the NV is the canonical form.
1846 Instead, IV/UV and NV need to be given equal rights. So as to not lose
1847 precision as a side effect of conversion (which would lead to insanity
1848 and the dragon(s) in t/op/numconvert.t getting very angry) the intent is
1849 1) to distinguish between IV/UV/NV slots that have cached a valid
1850 conversion where precision was lost and IV/UV/NV slots that have a
1851 valid conversion which has lost no precision
1852 2) to ensure that if a numeric conversion to one form is requested that
1853 would lose precision, the precise conversion (or differently
1854 imprecise conversion) is also performed and cached, to prevent
1855 requests for different numeric formats on the same SV causing
1856 lossy conversion chains. (lossless conversion chains are perfectly
1861 SvIOKp is true if the IV slot contains a valid value
1862 SvIOK is true only if the IV value is accurate (UV if SvIOK_UV true)
1863 SvNOKp is true if the NV slot contains a valid value
1864 SvNOK is true only if the NV value is accurate
1867 while converting from PV to NV, check to see if converting that NV to an
1868 IV(or UV) would lose accuracy over a direct conversion from PV to
1869 IV(or UV). If it would, cache both conversions, return NV, but mark
1870 SV as IOK NOKp (ie not NOK).
1872 While converting from PV to IV, check to see if converting that IV to an
1873 NV would lose accuracy over a direct conversion from PV to NV. If it
1874 would, cache both conversions, flag similarly.
1876 Before, the SV value "3.2" could become NV=3.2 IV=3 NOK, IOK quite
1877 correctly because if IV & NV were set NV *always* overruled.
1878 Now, "3.2" will become NV=3.2 IV=3 NOK, IOKp, because the flag's meaning
1879 changes - now IV and NV together means that the two are interchangeable:
1880 SvIVX == (IV) SvNVX && SvNVX == (NV) SvIVX;
1882 The benefit of this is that operations such as pp_add know that if
1883 SvIOK is true for both left and right operands, then integer addition
1884 can be used instead of floating point (for cases where the result won't
1885 overflow). Before, floating point was always used, which could lead to
1886 loss of precision compared with integer addition.
1888 * making IV and NV equal status should make maths accurate on 64 bit
1890 * may speed up maths somewhat if pp_add and friends start to use
1891 integers when possible instead of fp. (Hopefully the overhead in
1892 looking for SvIOK and checking for overflow will not outweigh the
1893 fp to integer speedup)
1894 * will slow down integer operations (callers of SvIV) on "inaccurate"
1895 values, as the change from SvIOK to SvIOKp will cause a call into
1896 sv_2iv each time rather than a macro access direct to the IV slot
1897 * should speed up number->string conversion on integers as IV is
1898 favoured when IV and NV are equally accurate
1900 ####################################################################
1901 You had better be using SvIOK_notUV if you want an IV for arithmetic:
1902 SvIOK is true if (IV or UV), so you might be getting (IV)SvUV.
1903 On the other hand, SvUOK is true iff UV.
1904 ####################################################################
1906 Your mileage will vary depending your CPU's relative fp to integer
1910 #ifndef NV_PRESERVES_UV
1911 # define IS_NUMBER_UNDERFLOW_IV 1
1912 # define IS_NUMBER_UNDERFLOW_UV 2
1913 # define IS_NUMBER_IV_AND_UV 2
1914 # define IS_NUMBER_OVERFLOW_IV 4
1915 # define IS_NUMBER_OVERFLOW_UV 5
1917 /* sv_2iuv_non_preserve(): private routine for use by sv_2iv() and sv_2uv() */
1919 /* For sv_2nv these three cases are "SvNOK and don't bother casting" */
1921 S_sv_2iuv_non_preserve(pTHX_ register SV *const sv
1929 PERL_ARGS_ASSERT_SV_2IUV_NON_PRESERVE;
1931 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));
1932 if (SvNVX(sv) < (NV)IV_MIN) {
1933 (void)SvIOKp_on(sv);
1935 SvIV_set(sv, IV_MIN);
1936 return IS_NUMBER_UNDERFLOW_IV;
1938 if (SvNVX(sv) > (NV)UV_MAX) {
1939 (void)SvIOKp_on(sv);
1942 SvUV_set(sv, UV_MAX);
1943 return IS_NUMBER_OVERFLOW_UV;
1945 (void)SvIOKp_on(sv);
1947 /* Can't use strtol etc to convert this string. (See truth table in
1949 if (SvNVX(sv) <= (UV)IV_MAX) {
1950 SvIV_set(sv, I_V(SvNVX(sv)));
1951 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
1952 SvIOK_on(sv); /* Integer is precise. NOK, IOK */
1954 /* Integer is imprecise. NOK, IOKp */
1956 return SvNVX(sv) < 0 ? IS_NUMBER_UNDERFLOW_UV : IS_NUMBER_IV_AND_UV;
1959 SvUV_set(sv, U_V(SvNVX(sv)));
1960 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
1961 if (SvUVX(sv) == UV_MAX) {
1962 /* As we know that NVs don't preserve UVs, UV_MAX cannot
1963 possibly be preserved by NV. Hence, it must be overflow.
1965 return IS_NUMBER_OVERFLOW_UV;
1967 SvIOK_on(sv); /* Integer is precise. NOK, UOK */
1969 /* Integer is imprecise. NOK, IOKp */
1971 return IS_NUMBER_OVERFLOW_IV;
1973 #endif /* !NV_PRESERVES_UV*/
1976 S_sv_2iuv_common(pTHX_ SV *const sv)
1980 PERL_ARGS_ASSERT_SV_2IUV_COMMON;
1983 /* erm. not sure. *should* never get NOKp (without NOK) from sv_2nv
1984 * without also getting a cached IV/UV from it at the same time
1985 * (ie PV->NV conversion should detect loss of accuracy and cache
1986 * IV or UV at same time to avoid this. */
1987 /* IV-over-UV optimisation - choose to cache IV if possible */
1989 if (SvTYPE(sv) == SVt_NV)
1990 sv_upgrade(sv, SVt_PVNV);
1992 (void)SvIOKp_on(sv); /* Must do this first, to clear any SvOOK */
1993 /* < not <= as for NV doesn't preserve UV, ((NV)IV_MAX+1) will almost
1994 certainly cast into the IV range at IV_MAX, whereas the correct
1995 answer is the UV IV_MAX +1. Hence < ensures that dodgy boundary
1997 #if defined(NAN_COMPARE_BROKEN) && defined(Perl_isnan)
1998 if (Perl_isnan(SvNVX(sv))) {
2004 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2005 SvIV_set(sv, I_V(SvNVX(sv)));
2006 if (SvNVX(sv) == (NV) SvIVX(sv)
2007 #ifndef NV_PRESERVES_UV
2008 && (((UV)1 << NV_PRESERVES_UV_BITS) >
2009 (UV)(SvIVX(sv) > 0 ? SvIVX(sv) : -SvIVX(sv)))
2010 /* Don't flag it as "accurately an integer" if the number
2011 came from a (by definition imprecise) NV operation, and
2012 we're outside the range of NV integer precision */
2016 SvIOK_on(sv); /* Can this go wrong with rounding? NWC */
2018 /* scalar has trailing garbage, eg "42a" */
2020 DEBUG_c(PerlIO_printf(Perl_debug_log,
2021 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (precise)\n",
2027 /* IV not precise. No need to convert from PV, as NV
2028 conversion would already have cached IV if it detected
2029 that PV->IV would be better than PV->NV->IV
2030 flags already correct - don't set public IOK. */
2031 DEBUG_c(PerlIO_printf(Perl_debug_log,
2032 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (imprecise)\n",
2037 /* Can the above go wrong if SvIVX == IV_MIN and SvNVX < IV_MIN,
2038 but the cast (NV)IV_MIN rounds to a the value less (more
2039 negative) than IV_MIN which happens to be equal to SvNVX ??
2040 Analogous to 0xFFFFFFFFFFFFFFFF rounding up to NV (2**64) and
2041 NV rounding back to 0xFFFFFFFFFFFFFFFF, so UVX == UV(NVX) and
2042 (NV)UVX == NVX are both true, but the values differ. :-(
2043 Hopefully for 2s complement IV_MIN is something like
2044 0x8000000000000000 which will be exact. NWC */
2047 SvUV_set(sv, U_V(SvNVX(sv)));
2049 (SvNVX(sv) == (NV) SvUVX(sv))
2050 #ifndef NV_PRESERVES_UV
2051 /* Make sure it's not 0xFFFFFFFFFFFFFFFF */
2052 /*&& (SvUVX(sv) != UV_MAX) irrelevant with code below */
2053 && (((UV)1 << NV_PRESERVES_UV_BITS) > SvUVX(sv))
2054 /* Don't flag it as "accurately an integer" if the number
2055 came from a (by definition imprecise) NV operation, and
2056 we're outside the range of NV integer precision */
2062 DEBUG_c(PerlIO_printf(Perl_debug_log,
2063 "0x%"UVxf" 2iv(%"UVuf" => %"IVdf") (as unsigned)\n",
2069 else if (SvPOKp(sv) && SvLEN(sv)) {
2071 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2072 /* We want to avoid a possible problem when we cache an IV/ a UV which
2073 may be later translated to an NV, and the resulting NV is not
2074 the same as the direct translation of the initial string
2075 (eg 123.456 can shortcut to the IV 123 with atol(), but we must
2076 be careful to ensure that the value with the .456 is around if the
2077 NV value is requested in the future).
2079 This means that if we cache such an IV/a UV, we need to cache the
2080 NV as well. Moreover, we trade speed for space, and do not
2081 cache the NV if we are sure it's not needed.
2084 /* SVt_PVNV is one higher than SVt_PVIV, hence this order */
2085 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2086 == IS_NUMBER_IN_UV) {
2087 /* It's definitely an integer, only upgrade to PVIV */
2088 if (SvTYPE(sv) < SVt_PVIV)
2089 sv_upgrade(sv, SVt_PVIV);
2091 } else if (SvTYPE(sv) < SVt_PVNV)
2092 sv_upgrade(sv, SVt_PVNV);
2094 /* If NVs preserve UVs then we only use the UV value if we know that
2095 we aren't going to call atof() below. If NVs don't preserve UVs
2096 then the value returned may have more precision than atof() will
2097 return, even though value isn't perfectly accurate. */
2098 if ((numtype & (IS_NUMBER_IN_UV
2099 #ifdef NV_PRESERVES_UV
2102 )) == IS_NUMBER_IN_UV) {
2103 /* This won't turn off the public IOK flag if it was set above */
2104 (void)SvIOKp_on(sv);
2106 if (!(numtype & IS_NUMBER_NEG)) {
2108 if (value <= (UV)IV_MAX) {
2109 SvIV_set(sv, (IV)value);
2111 /* it didn't overflow, and it was positive. */
2112 SvUV_set(sv, value);
2116 /* 2s complement assumption */
2117 if (value <= (UV)IV_MIN) {
2118 SvIV_set(sv, -(IV)value);
2120 /* Too negative for an IV. This is a double upgrade, but
2121 I'm assuming it will be rare. */
2122 if (SvTYPE(sv) < SVt_PVNV)
2123 sv_upgrade(sv, SVt_PVNV);
2127 SvNV_set(sv, -(NV)value);
2128 SvIV_set(sv, IV_MIN);
2132 /* For !NV_PRESERVES_UV and IS_NUMBER_IN_UV and IS_NUMBER_NOT_INT we
2133 will be in the previous block to set the IV slot, and the next
2134 block to set the NV slot. So no else here. */
2136 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2137 != IS_NUMBER_IN_UV) {
2138 /* It wasn't an (integer that doesn't overflow the UV). */
2139 SvNV_set(sv, Atof(SvPVX_const(sv)));
2141 if (! numtype && ckWARN(WARN_NUMERIC))
2144 #if defined(USE_LONG_DOUBLE)
2145 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%" PERL_PRIgldbl ")\n",
2146 PTR2UV(sv), SvNVX(sv)));
2148 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"NVgf")\n",
2149 PTR2UV(sv), SvNVX(sv)));
2152 #ifdef NV_PRESERVES_UV
2153 (void)SvIOKp_on(sv);
2155 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2156 SvIV_set(sv, I_V(SvNVX(sv)));
2157 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
2160 NOOP; /* Integer is imprecise. NOK, IOKp */
2162 /* UV will not work better than IV */
2164 if (SvNVX(sv) > (NV)UV_MAX) {
2166 /* Integer is inaccurate. NOK, IOKp, is UV */
2167 SvUV_set(sv, UV_MAX);
2169 SvUV_set(sv, U_V(SvNVX(sv)));
2170 /* 0xFFFFFFFFFFFFFFFF not an issue in here, NVs
2171 NV preservse UV so can do correct comparison. */
2172 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
2175 NOOP; /* Integer is imprecise. NOK, IOKp, is UV */
2180 #else /* NV_PRESERVES_UV */
2181 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2182 == (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT)) {
2183 /* The IV/UV slot will have been set from value returned by
2184 grok_number above. The NV slot has just been set using
2187 assert (SvIOKp(sv));
2189 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2190 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2191 /* Small enough to preserve all bits. */
2192 (void)SvIOKp_on(sv);
2194 SvIV_set(sv, I_V(SvNVX(sv)));
2195 if ((NV)(SvIVX(sv)) == SvNVX(sv))
2197 /* Assumption: first non-preserved integer is < IV_MAX,
2198 this NV is in the preserved range, therefore: */
2199 if (!(U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))
2201 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);
2205 0 0 already failed to read UV.
2206 0 1 already failed to read UV.
2207 1 0 you won't get here in this case. IV/UV
2208 slot set, public IOK, Atof() unneeded.
2209 1 1 already read UV.
2210 so there's no point in sv_2iuv_non_preserve() attempting
2211 to use atol, strtol, strtoul etc. */
2213 sv_2iuv_non_preserve (sv, numtype);
2215 sv_2iuv_non_preserve (sv);
2219 #endif /* NV_PRESERVES_UV */
2220 /* It might be more code efficient to go through the entire logic above
2221 and conditionally set with SvIOKp_on() rather than SvIOK(), but it
2222 gets complex and potentially buggy, so more programmer efficient
2223 to do it this way, by turning off the public flags: */
2225 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK);
2229 if (isGV_with_GP(sv))
2230 return glob_2number(MUTABLE_GV(sv));
2232 if (!SvPADTMP(sv)) {
2233 if (!PL_localizing && ckWARN(WARN_UNINITIALIZED))
2236 if (SvTYPE(sv) < SVt_IV)
2237 /* Typically the caller expects that sv_any is not NULL now. */
2238 sv_upgrade(sv, SVt_IV);
2239 /* Return 0 from the caller. */
2246 =for apidoc sv_2iv_flags
2248 Return the integer value of an SV, doing any necessary string
2249 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2250 Normally used via the C<SvIV(sv)> and C<SvIVx(sv)> macros.
2256 Perl_sv_2iv_flags(pTHX_ register SV *const sv, const I32 flags)
2263 if (SvGMAGICAL(sv) && (flags & SV_GMAGIC))
2269 if (flags & SV_SKIP_OVERLOAD)
2271 tmpstr = AMG_CALLunary(sv, numer_amg);
2272 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2273 return SvIV(tmpstr);
2276 return PTR2IV(SvRV(sv));
2280 /* FBMs use the space for SvIVX and SvNVX for other purposes, and use
2281 the same flag bit as SVf_IVisUV, so must not let them cache IVs.
2282 In practice they are extremely unlikely to actually get anywhere
2283 accessible by user Perl code - the only way that I'm aware of is when
2284 a constant subroutine which is used as the second argument to index.
2289 return I_V(SvNVX(sv));
2290 if (SvPOKp(sv) && SvLEN(sv)) {
2293 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2295 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2296 == IS_NUMBER_IN_UV) {
2297 /* It's definitely an integer */
2298 if (numtype & IS_NUMBER_NEG) {
2299 if (value < (UV)IV_MIN)
2302 if (value < (UV)IV_MAX)
2307 if (ckWARN(WARN_NUMERIC))
2310 return I_V(Atof(SvPVX_const(sv)));
2312 if (ckWARN(WARN_UNINITIALIZED))
2317 if (SvTHINKFIRST(sv)) {
2319 sv_force_normal_flags(sv, 0);
2321 if (SvREADONLY(sv) && !SvOK(sv)) {
2322 if (ckWARN(WARN_UNINITIALIZED))
2329 if (S_sv_2iuv_common(aTHX_ sv))
2333 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"IVdf")\n",
2334 PTR2UV(sv),SvIVX(sv)));
2335 return SvIsUV(sv) ? (IV)SvUVX(sv) : SvIVX(sv);
2339 =for apidoc sv_2uv_flags
2341 Return the unsigned integer value of an SV, doing any necessary string
2342 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2343 Normally used via the C<SvUV(sv)> and C<SvUVx(sv)> macros.
2349 Perl_sv_2uv_flags(pTHX_ register SV *const sv, const I32 flags)
2356 if (SvGMAGICAL(sv) && (flags & SV_GMAGIC))
2362 if (flags & SV_SKIP_OVERLOAD)
2364 tmpstr = AMG_CALLunary(sv, numer_amg);
2365 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2366 return SvUV(tmpstr);
2369 return PTR2UV(SvRV(sv));
2373 /* FBMs use the space for SvIVX and SvNVX for other purposes, and use
2374 the same flag bit as SVf_IVisUV, so must not let them cache IVs. */
2378 return U_V(SvNVX(sv));
2379 if (SvPOKp(sv) && SvLEN(sv)) {
2382 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2384 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2385 == IS_NUMBER_IN_UV) {
2386 /* It's definitely an integer */
2387 if (!(numtype & IS_NUMBER_NEG))
2391 if (ckWARN(WARN_NUMERIC))
2394 return U_V(Atof(SvPVX_const(sv)));
2396 if (ckWARN(WARN_UNINITIALIZED))
2401 if (SvTHINKFIRST(sv)) {
2403 sv_force_normal_flags(sv, 0);
2405 if (SvREADONLY(sv) && !SvOK(sv)) {
2406 if (ckWARN(WARN_UNINITIALIZED))
2413 if (S_sv_2iuv_common(aTHX_ sv))
2417 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2uv(%"UVuf")\n",
2418 PTR2UV(sv),SvUVX(sv)));
2419 return SvIsUV(sv) ? SvUVX(sv) : (UV)SvIVX(sv);
2423 =for apidoc sv_2nv_flags
2425 Return the num value of an SV, doing any necessary string or integer
2426 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2427 Normally used via the C<SvNV(sv)> and C<SvNVx(sv)> macros.
2433 Perl_sv_2nv_flags(pTHX_ register SV *const sv, const I32 flags)
2438 if (SvGMAGICAL(sv) || SvVALID(sv)) {
2439 /* FBMs use the space for SvIVX and SvNVX for other purposes, and use
2440 the same flag bit as SVf_IVisUV, so must not let them cache NVs. */
2441 if (flags & SV_GMAGIC)
2445 if ((SvPOKp(sv) && SvLEN(sv)) && !SvIOKp(sv)) {
2446 if (!SvIOKp(sv) && ckWARN(WARN_NUMERIC) &&
2447 !grok_number(SvPVX_const(sv), SvCUR(sv), NULL))
2449 return Atof(SvPVX_const(sv));
2453 return (NV)SvUVX(sv);
2455 return (NV)SvIVX(sv);
2460 assert(SvTYPE(sv) >= SVt_PVMG);
2461 /* This falls through to the report_uninit near the end of the
2463 } else if (SvTHINKFIRST(sv)) {
2468 if (flags & SV_SKIP_OVERLOAD)
2470 tmpstr = AMG_CALLunary(sv, numer_amg);
2471 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2472 return SvNV(tmpstr);
2475 return PTR2NV(SvRV(sv));
2478 sv_force_normal_flags(sv, 0);
2480 if (SvREADONLY(sv) && !SvOK(sv)) {
2481 if (ckWARN(WARN_UNINITIALIZED))
2486 if (SvTYPE(sv) < SVt_NV) {
2487 /* The logic to use SVt_PVNV if necessary is in sv_upgrade. */
2488 sv_upgrade(sv, SVt_NV);
2489 #ifdef USE_LONG_DOUBLE
2491 STORE_NUMERIC_LOCAL_SET_STANDARD();
2492 PerlIO_printf(Perl_debug_log,
2493 "0x%"UVxf" num(%" PERL_PRIgldbl ")\n",
2494 PTR2UV(sv), SvNVX(sv));
2495 RESTORE_NUMERIC_LOCAL();
2499 STORE_NUMERIC_LOCAL_SET_STANDARD();
2500 PerlIO_printf(Perl_debug_log, "0x%"UVxf" num(%"NVgf")\n",
2501 PTR2UV(sv), SvNVX(sv));
2502 RESTORE_NUMERIC_LOCAL();
2506 else if (SvTYPE(sv) < SVt_PVNV)
2507 sv_upgrade(sv, SVt_PVNV);
2512 SvNV_set(sv, SvIsUV(sv) ? (NV)SvUVX(sv) : (NV)SvIVX(sv));
2513 #ifdef NV_PRESERVES_UV
2519 /* Only set the public NV OK flag if this NV preserves the IV */
2520 /* Check it's not 0xFFFFFFFFFFFFFFFF */
2522 SvIsUV(sv) ? ((SvUVX(sv) != UV_MAX)&&(SvUVX(sv) == U_V(SvNVX(sv))))
2523 : (SvIVX(sv) == I_V(SvNVX(sv))))
2529 else if (SvPOKp(sv) && SvLEN(sv)) {
2531 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2532 if (!SvIOKp(sv) && !numtype && ckWARN(WARN_NUMERIC))
2534 #ifdef NV_PRESERVES_UV
2535 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2536 == IS_NUMBER_IN_UV) {
2537 /* It's definitely an integer */
2538 SvNV_set(sv, (numtype & IS_NUMBER_NEG) ? -(NV)value : (NV)value);
2540 SvNV_set(sv, Atof(SvPVX_const(sv)));
2546 SvNV_set(sv, Atof(SvPVX_const(sv)));
2547 /* Only set the public NV OK flag if this NV preserves the value in
2548 the PV at least as well as an IV/UV would.
2549 Not sure how to do this 100% reliably. */
2550 /* if that shift count is out of range then Configure's test is
2551 wonky. We shouldn't be in here with NV_PRESERVES_UV_BITS ==
2553 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2554 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2555 SvNOK_on(sv); /* Definitely small enough to preserve all bits */
2556 } else if (!(numtype & IS_NUMBER_IN_UV)) {
2557 /* Can't use strtol etc to convert this string, so don't try.
2558 sv_2iv and sv_2uv will use the NV to convert, not the PV. */
2561 /* value has been set. It may not be precise. */
2562 if ((numtype & IS_NUMBER_NEG) && (value > (UV)IV_MIN)) {
2563 /* 2s complement assumption for (UV)IV_MIN */
2564 SvNOK_on(sv); /* Integer is too negative. */
2569 if (numtype & IS_NUMBER_NEG) {
2570 SvIV_set(sv, -(IV)value);
2571 } else if (value <= (UV)IV_MAX) {
2572 SvIV_set(sv, (IV)value);
2574 SvUV_set(sv, value);
2578 if (numtype & IS_NUMBER_NOT_INT) {
2579 /* I believe that even if the original PV had decimals,
2580 they are lost beyond the limit of the FP precision.
2581 However, neither is canonical, so both only get p
2582 flags. NWC, 2000/11/25 */
2583 /* Both already have p flags, so do nothing */
2585 const NV nv = SvNVX(sv);
2586 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2587 if (SvIVX(sv) == I_V(nv)) {
2590 /* It had no "." so it must be integer. */
2594 /* between IV_MAX and NV(UV_MAX).
2595 Could be slightly > UV_MAX */
2597 if (numtype & IS_NUMBER_NOT_INT) {
2598 /* UV and NV both imprecise. */
2600 const UV nv_as_uv = U_V(nv);
2602 if (value == nv_as_uv && SvUVX(sv) != UV_MAX) {
2611 /* It might be more code efficient to go through the entire logic above
2612 and conditionally set with SvNOKp_on() rather than SvNOK(), but it
2613 gets complex and potentially buggy, so more programmer efficient
2614 to do it this way, by turning off the public flags: */
2616 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK);
2617 #endif /* NV_PRESERVES_UV */
2620 if (isGV_with_GP(sv)) {
2621 glob_2number(MUTABLE_GV(sv));
2625 if (!PL_localizing && !SvPADTMP(sv) && ckWARN(WARN_UNINITIALIZED))
2627 assert (SvTYPE(sv) >= SVt_NV);
2628 /* Typically the caller expects that sv_any is not NULL now. */
2629 /* XXX Ilya implies that this is a bug in callers that assume this
2630 and ideally should be fixed. */
2633 #if defined(USE_LONG_DOUBLE)
2635 STORE_NUMERIC_LOCAL_SET_STANDARD();
2636 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2nv(%" PERL_PRIgldbl ")\n",
2637 PTR2UV(sv), SvNVX(sv));
2638 RESTORE_NUMERIC_LOCAL();
2642 STORE_NUMERIC_LOCAL_SET_STANDARD();
2643 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 1nv(%"NVgf")\n",
2644 PTR2UV(sv), SvNVX(sv));
2645 RESTORE_NUMERIC_LOCAL();
2654 Return an SV with the numeric value of the source SV, doing any necessary
2655 reference or overload conversion. You must use the C<SvNUM(sv)> macro to
2656 access this function.
2662 Perl_sv_2num(pTHX_ register SV *const sv)
2664 PERL_ARGS_ASSERT_SV_2NUM;
2669 SV * const tmpsv = AMG_CALLunary(sv, numer_amg);
2670 TAINT_IF(tmpsv && SvTAINTED(tmpsv));
2671 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
2672 return sv_2num(tmpsv);
2674 return sv_2mortal(newSVuv(PTR2UV(SvRV(sv))));
2677 /* uiv_2buf(): private routine for use by sv_2pv_flags(): print an IV or
2678 * UV as a string towards the end of buf, and return pointers to start and
2681 * We assume that buf is at least TYPE_CHARS(UV) long.
2685 S_uiv_2buf(char *const buf, const IV iv, UV uv, const int is_uv, char **const peob)
2687 char *ptr = buf + TYPE_CHARS(UV);
2688 char * const ebuf = ptr;
2691 PERL_ARGS_ASSERT_UIV_2BUF;
2703 *--ptr = '0' + (char)(uv % 10);
2712 =for apidoc sv_2pv_flags
2714 Returns a pointer to the string value of an SV, and sets *lp to its length.
2715 If flags includes SV_GMAGIC, does an mg_get() first. Coerces sv to a
2716 string if necessary. Normally invoked via the C<SvPV_flags> macro.
2717 C<sv_2pv()> and C<sv_2pv_nomg> usually end up here too.
2723 Perl_sv_2pv_flags(pTHX_ register SV *const sv, STRLEN *const lp, const I32 flags)
2733 if (SvGMAGICAL(sv) && (flags & SV_GMAGIC))
2738 if (flags & SV_SKIP_OVERLOAD)
2740 tmpstr = AMG_CALLunary(sv, string_amg);
2741 TAINT_IF(tmpstr && SvTAINTED(tmpstr));
2742 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2744 /* char *pv = lp ? SvPV(tmpstr, *lp) : SvPV_nolen(tmpstr);
2748 if ((SvFLAGS(tmpstr) & (SVf_POK)) == SVf_POK) {
2749 if (flags & SV_CONST_RETURN) {
2750 pv = (char *) SvPVX_const(tmpstr);
2752 pv = (flags & SV_MUTABLE_RETURN)
2753 ? SvPVX_mutable(tmpstr) : SvPVX(tmpstr);
2756 *lp = SvCUR(tmpstr);
2758 pv = sv_2pv_flags(tmpstr, lp, flags);
2771 SV *const referent = SvRV(sv);
2775 retval = buffer = savepvn("NULLREF", len);
2776 } else if (SvTYPE(referent) == SVt_REGEXP &&
2777 (!(PL_curcop->cop_hints & HINT_NO_AMAGIC) ||
2778 amagic_is_enabled(string_amg))) {
2779 REGEXP * const re = (REGEXP *)MUTABLE_PTR(referent);
2783 /* If the regex is UTF-8 we want the containing scalar to
2784 have an UTF-8 flag too */
2791 *lp = RX_WRAPLEN(re);
2793 return RX_WRAPPED(re);
2795 const char *const typestr = sv_reftype(referent, 0);
2796 const STRLEN typelen = strlen(typestr);
2797 UV addr = PTR2UV(referent);
2798 const char *stashname = NULL;
2799 STRLEN stashnamelen = 0; /* hush, gcc */
2800 const char *buffer_end;
2802 if (SvOBJECT(referent)) {
2803 const HEK *const name = HvNAME_HEK(SvSTASH(referent));
2806 stashname = HEK_KEY(name);
2807 stashnamelen = HEK_LEN(name);
2809 if (HEK_UTF8(name)) {
2815 stashname = "__ANON__";
2818 len = stashnamelen + 1 /* = */ + typelen + 3 /* (0x */
2819 + 2 * sizeof(UV) + 2 /* )\0 */;
2821 len = typelen + 3 /* (0x */
2822 + 2 * sizeof(UV) + 2 /* )\0 */;
2825 Newx(buffer, len, char);
2826 buffer_end = retval = buffer + len;
2828 /* Working backwards */
2832 *--retval = PL_hexdigit[addr & 15];
2833 } while (addr >>= 4);
2839 memcpy(retval, typestr, typelen);
2843 retval -= stashnamelen;
2844 memcpy(retval, stashname, stashnamelen);
2846 /* retval may not necessarily have reached the start of the
2848 assert (retval >= buffer);
2850 len = buffer_end - retval - 1; /* -1 for that \0 */
2862 if (flags & SV_MUTABLE_RETURN)
2863 return SvPVX_mutable(sv);
2864 if (flags & SV_CONST_RETURN)
2865 return (char *)SvPVX_const(sv);
2870 /* I'm assuming that if both IV and NV are equally valid then
2871 converting the IV is going to be more efficient */
2872 const U32 isUIOK = SvIsUV(sv);
2873 char buf[TYPE_CHARS(UV)];
2877 if (SvTYPE(sv) < SVt_PVIV)
2878 sv_upgrade(sv, SVt_PVIV);
2879 ptr = uiv_2buf(buf, SvIVX(sv), SvUVX(sv), isUIOK, &ebuf);
2881 /* inlined from sv_setpvn */
2882 s = SvGROW_mutable(sv, len + 1);
2883 Move(ptr, s, len, char);
2887 else if (SvNOK(sv)) {
2888 if (SvTYPE(sv) < SVt_PVNV)
2889 sv_upgrade(sv, SVt_PVNV);
2890 if (SvNVX(sv) == 0.0) {
2891 s = SvGROW_mutable(sv, 2);
2896 /* The +20 is pure guesswork. Configure test needed. --jhi */
2897 s = SvGROW_mutable(sv, NV_DIG + 20);
2898 /* some Xenix systems wipe out errno here */
2899 Gconvert(SvNVX(sv), NV_DIG, 0, s);
2908 else if (isGV_with_GP(sv)) {
2909 GV *const gv = MUTABLE_GV(sv);
2910 SV *const buffer = sv_newmortal();
2912 gv_efullname3(buffer, gv, "*");
2914 assert(SvPOK(buffer));
2918 *lp = SvCUR(buffer);
2919 return SvPVX(buffer);
2924 if (flags & SV_UNDEF_RETURNS_NULL)
2926 if (!PL_localizing && !SvPADTMP(sv) && ckWARN(WARN_UNINITIALIZED))
2928 /* Typically the caller expects that sv_any is not NULL now. */
2929 if (!SvREADONLY(sv) && SvTYPE(sv) < SVt_PV)
2930 sv_upgrade(sv, SVt_PV);
2935 const STRLEN len = s - SvPVX_const(sv);
2941 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
2942 PTR2UV(sv),SvPVX_const(sv)));
2943 if (flags & SV_CONST_RETURN)
2944 return (char *)SvPVX_const(sv);
2945 if (flags & SV_MUTABLE_RETURN)
2946 return SvPVX_mutable(sv);
2951 =for apidoc sv_copypv
2953 Copies a stringified representation of the source SV into the
2954 destination SV. Automatically performs any necessary mg_get and
2955 coercion of numeric values into strings. Guaranteed to preserve
2956 UTF8 flag even from overloaded objects. Similar in nature to
2957 sv_2pv[_flags] but operates directly on an SV instead of just the
2958 string. Mostly uses sv_2pv_flags to do its work, except when that
2959 would lose the UTF-8'ness of the PV.
2961 =for apidoc sv_copypv_nomg
2963 Like sv_copypv, but doesn't invoke get magic first.
2965 =for apidoc sv_copypv_flags
2967 Implementation of sv_copypv and sv_copypv_nomg. Calls get magic iff flags
2974 Perl_sv_copypv(pTHX_ SV *const dsv, register SV *const ssv)
2976 PERL_ARGS_ASSERT_SV_COPYPV;
2978 sv_copypv_flags(dsv, ssv, 0);
2982 Perl_sv_copypv_flags(pTHX_ SV *const dsv, register SV *const ssv, const I32 flags)
2987 PERL_ARGS_ASSERT_SV_COPYPV_FLAGS;
2989 if ((flags & SV_GMAGIC) && SvGMAGICAL(ssv))
2991 s = SvPV_nomg_const(ssv,len);
2992 sv_setpvn(dsv,s,len);
3000 =for apidoc sv_2pvbyte
3002 Return a pointer to the byte-encoded representation of the SV, and set *lp
3003 to its length. May cause the SV to be downgraded from UTF-8 as a
3006 Usually accessed via the C<SvPVbyte> macro.
3012 Perl_sv_2pvbyte(pTHX_ register SV *sv, STRLEN *const lp)
3014 PERL_ARGS_ASSERT_SV_2PVBYTE;
3016 if (((SvREADONLY(sv) || SvFAKE(sv)) && !SvIsCOW(sv))
3017 || isGV_with_GP(sv) || SvROK(sv)) {
3018 SV *sv2 = sv_newmortal();
3022 else SvGETMAGIC(sv);
3023 sv_utf8_downgrade(sv,0);
3024 return lp ? SvPV_nomg(sv,*lp) : SvPV_nomg_nolen(sv);
3028 =for apidoc sv_2pvutf8
3030 Return a pointer to the UTF-8-encoded representation of the SV, and set *lp
3031 to its length. May cause the SV to be upgraded to UTF-8 as a side-effect.
3033 Usually accessed via the C<SvPVutf8> macro.
3039 Perl_sv_2pvutf8(pTHX_ register SV *sv, STRLEN *const lp)
3041 PERL_ARGS_ASSERT_SV_2PVUTF8;
3043 if (((SvREADONLY(sv) || SvFAKE(sv)) && !SvIsCOW(sv))
3044 || isGV_with_GP(sv) || SvROK(sv))
3045 sv = sv_mortalcopy(sv);
3048 sv_utf8_upgrade_nomg(sv);
3049 return lp ? SvPV_nomg(sv,*lp) : SvPV_nomg_nolen(sv);
3054 =for apidoc sv_2bool
3056 This macro is only used by sv_true() or its macro equivalent, and only if
3057 the latter's argument is neither SvPOK, SvIOK nor SvNOK.
3058 It calls sv_2bool_flags with the SV_GMAGIC flag.
3060 =for apidoc sv_2bool_flags
3062 This function is only used by sv_true() and friends, and only if
3063 the latter's argument is neither SvPOK, SvIOK nor SvNOK. If the flags
3064 contain SV_GMAGIC, then it does an mg_get() first.
3071 Perl_sv_2bool_flags(pTHX_ register SV *const sv, const I32 flags)
3075 PERL_ARGS_ASSERT_SV_2BOOL_FLAGS;
3077 if(flags & SV_GMAGIC) SvGETMAGIC(sv);
3083 SV * const tmpsv = AMG_CALLunary(sv, bool__amg);
3084 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
3085 return cBOOL(SvTRUE(tmpsv));
3087 return SvRV(sv) != 0;
3089 return SvTRUE_common(sv, isGV_with_GP(sv) ? 1 : 0);
3093 =for apidoc sv_utf8_upgrade
3095 Converts the PV of an SV to its UTF-8-encoded form.
3096 Forces the SV to string form if it is not already.
3097 Will C<mg_get> on C<sv> if appropriate.
3098 Always sets the SvUTF8 flag to avoid future validity checks even
3099 if the whole string is the same in UTF-8 as not.
3100 Returns the number of bytes in the converted string
3102 This is not as a general purpose byte encoding to Unicode interface:
3103 use the Encode extension for that.
3105 =for apidoc sv_utf8_upgrade_nomg
3107 Like sv_utf8_upgrade, but doesn't do magic on C<sv>.
3109 =for apidoc sv_utf8_upgrade_flags
3111 Converts the PV of an SV to its UTF-8-encoded form.
3112 Forces the SV to string form if it is not already.
3113 Always sets the SvUTF8 flag to avoid future validity checks even
3114 if all the bytes are invariant in UTF-8.
3115 If C<flags> has C<SV_GMAGIC> bit set,
3116 will C<mg_get> on C<sv> if appropriate, else not.
3117 Returns the number of bytes in the converted string
3118 C<sv_utf8_upgrade> and
3119 C<sv_utf8_upgrade_nomg> are implemented in terms of this function.
3121 This is not as a general purpose byte encoding to Unicode interface:
3122 use the Encode extension for that.
3126 The grow version is currently not externally documented. It adds a parameter,
3127 extra, which is the number of unused bytes the string of 'sv' is guaranteed to
3128 have free after it upon return. This allows the caller to reserve extra space
3129 that it intends to fill, to avoid extra grows.
3131 Also externally undocumented for the moment is the flag SV_FORCE_UTF8_UPGRADE,
3132 which can be used to tell this function to not first check to see if there are
3133 any characters that are different in UTF-8 (variant characters) which would
3134 force it to allocate a new string to sv, but to assume there are. Typically
3135 this flag is used by a routine that has already parsed the string to find that
3136 there are such characters, and passes this information on so that the work
3137 doesn't have to be repeated.
3139 (One might think that the calling routine could pass in the position of the
3140 first such variant, so it wouldn't have to be found again. But that is not the
3141 case, because typically when the caller is likely to use this flag, it won't be
3142 calling this routine unless it finds something that won't fit into a byte.
3143 Otherwise it tries to not upgrade and just use bytes. But some things that
3144 do fit into a byte are variants in utf8, and the caller may not have been
3145 keeping track of these.)
3147 If the routine itself changes the string, it adds a trailing NUL. Such a NUL
3148 isn't guaranteed due to having other routines do the work in some input cases,
3149 or if the input is already flagged as being in utf8.
3151 The speed of this could perhaps be improved for many cases if someone wanted to
3152 write a fast function that counts the number of variant characters in a string,
3153 especially if it could return the position of the first one.
3158 Perl_sv_utf8_upgrade_flags_grow(pTHX_ register SV *const sv, const I32 flags, STRLEN extra)
3162 PERL_ARGS_ASSERT_SV_UTF8_UPGRADE_FLAGS_GROW;
3164 if (sv == &PL_sv_undef)
3168 if (SvREADONLY(sv) && (SvPOKp(sv) || SvIOKp(sv) || SvNOKp(sv))) {
3169 (void) sv_2pv_flags(sv,&len, flags);
3171 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3175 (void) SvPV_force_flags(sv,len,flags & SV_GMAGIC);
3180 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3185 sv_force_normal_flags(sv, 0);
3188 if (PL_encoding && !(flags & SV_UTF8_NO_ENCODING)) {
3189 sv_recode_to_utf8(sv, PL_encoding);
3190 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3194 if (SvCUR(sv) == 0) {
3195 if (extra) SvGROW(sv, extra);
3196 } else { /* Assume Latin-1/EBCDIC */
3197 /* This function could be much more efficient if we
3198 * had a FLAG in SVs to signal if there are any variant
3199 * chars in the PV. Given that there isn't such a flag
3200 * make the loop as fast as possible (although there are certainly ways
3201 * to speed this up, eg. through vectorization) */
3202 U8 * s = (U8 *) SvPVX_const(sv);
3203 U8 * e = (U8 *) SvEND(sv);
3205 STRLEN two_byte_count = 0;
3207 if (flags & SV_FORCE_UTF8_UPGRADE) goto must_be_utf8;
3209 /* See if really will need to convert to utf8. We mustn't rely on our
3210 * incoming SV being well formed and having a trailing '\0', as certain
3211 * code in pp_formline can send us partially built SVs. */
3215 if (NATIVE_IS_INVARIANT(ch)) continue;
3217 t--; /* t already incremented; re-point to first variant */
3222 /* utf8 conversion not needed because all are invariants. Mark as
3223 * UTF-8 even if no variant - saves scanning loop */
3225 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3230 /* Here, the string should be converted to utf8, either because of an
3231 * input flag (two_byte_count = 0), or because a character that
3232 * requires 2 bytes was found (two_byte_count = 1). t points either to
3233 * the beginning of the string (if we didn't examine anything), or to
3234 * the first variant. In either case, everything from s to t - 1 will
3235 * occupy only 1 byte each on output.
3237 * There are two main ways to convert. One is to create a new string
3238 * and go through the input starting from the beginning, appending each
3239 * converted value onto the new string as we go along. It's probably
3240 * best to allocate enough space in the string for the worst possible
3241 * case rather than possibly running out of space and having to
3242 * reallocate and then copy what we've done so far. Since everything
3243 * from s to t - 1 is invariant, the destination can be initialized
3244 * with these using a fast memory copy
3246 * The other way is to figure out exactly how big the string should be
3247 * by parsing the entire input. Then you don't have to make it big
3248 * enough to handle the worst possible case, and more importantly, if
3249 * the string you already have is large enough, you don't have to
3250 * allocate a new string, you can copy the last character in the input
3251 * string to the final position(s) that will be occupied by the
3252 * converted string and go backwards, stopping at t, since everything
3253 * before that is invariant.
3255 * There are advantages and disadvantages to each method.
3257 * In the first method, we can allocate a new string, do the memory
3258 * copy from the s to t - 1, and then proceed through the rest of the
3259 * string byte-by-byte.
3261 * In the second method, we proceed through the rest of the input
3262 * string just calculating how big the converted string will be. Then
3263 * there are two cases:
3264 * 1) if the string has enough extra space to handle the converted
3265 * value. We go backwards through the string, converting until we
3266 * get to the position we are at now, and then stop. If this
3267 * position is far enough along in the string, this method is
3268 * faster than the other method. If the memory copy were the same
3269 * speed as the byte-by-byte loop, that position would be about
3270 * half-way, as at the half-way mark, parsing to the end and back
3271 * is one complete string's parse, the same amount as starting
3272 * over and going all the way through. Actually, it would be
3273 * somewhat less than half-way, as it's faster to just count bytes
3274 * than to also copy, and we don't have the overhead of allocating
3275 * a new string, changing the scalar to use it, and freeing the
3276 * existing one. But if the memory copy is fast, the break-even
3277 * point is somewhere after half way. The counting loop could be
3278 * sped up by vectorization, etc, to move the break-even point
3279 * further towards the beginning.
3280 * 2) if the string doesn't have enough space to handle the converted
3281 * value. A new string will have to be allocated, and one might
3282 * as well, given that, start from the beginning doing the first
3283 * method. We've spent extra time parsing the string and in
3284 * exchange all we've gotten is that we know precisely how big to
3285 * make the new one. Perl is more optimized for time than space,
3286 * so this case is a loser.
3287 * So what I've decided to do is not use the 2nd method unless it is
3288 * guaranteed that a new string won't have to be allocated, assuming
3289 * the worst case. I also decided not to put any more conditions on it
3290 * than this, for now. It seems likely that, since the worst case is
3291 * twice as big as the unknown portion of the string (plus 1), we won't
3292 * be guaranteed enough space, causing us to go to the first method,
3293 * unless the string is short, or the first variant character is near
3294 * the end of it. In either of these cases, it seems best to use the
3295 * 2nd method. The only circumstance I can think of where this would
3296 * be really slower is if the string had once had much more data in it
3297 * than it does now, but there is still a substantial amount in it */
3300 STRLEN invariant_head = t - s;
3301 STRLEN size = invariant_head + (e - t) * 2 + 1 + extra;
3302 if (SvLEN(sv) < size) {
3304 /* Here, have decided to allocate a new string */
3309 Newx(dst, size, U8);
3311 /* If no known invariants at the beginning of the input string,
3312 * set so starts from there. Otherwise, can use memory copy to
3313 * get up to where we are now, and then start from here */
3315 if (invariant_head <= 0) {
3318 Copy(s, dst, invariant_head, char);
3319 d = dst + invariant_head;
3323 const UV uv = NATIVE8_TO_UNI(*t++);
3324 if (UNI_IS_INVARIANT(uv))
3325 *d++ = (U8)UNI_TO_NATIVE(uv);
3327 *d++ = (U8)UTF8_EIGHT_BIT_HI(uv);
3328 *d++ = (U8)UTF8_EIGHT_BIT_LO(uv);
3332 SvPV_free(sv); /* No longer using pre-existing string */
3333 SvPV_set(sv, (char*)dst);
3334 SvCUR_set(sv, d - dst);
3335 SvLEN_set(sv, size);
3338 /* Here, have decided to get the exact size of the string.
3339 * Currently this happens only when we know that there is
3340 * guaranteed enough space to fit the converted string, so
3341 * don't have to worry about growing. If two_byte_count is 0,
3342 * then t points to the first byte of the string which hasn't
3343 * been examined yet. Otherwise two_byte_count is 1, and t
3344 * points to the first byte in the string that will expand to
3345 * two. Depending on this, start examining at t or 1 after t.
3348 U8 *d = t + two_byte_count;
3351 /* Count up the remaining bytes that expand to two */
3354 const U8 chr = *d++;
3355 if (! NATIVE_IS_INVARIANT(chr)) two_byte_count++;
3358 /* The string will expand by just the number of bytes that
3359 * occupy two positions. But we are one afterwards because of
3360 * the increment just above. This is the place to put the
3361 * trailing NUL, and to set the length before we decrement */
3363 d += two_byte_count;
3364 SvCUR_set(sv, d - s);
3368 /* Having decremented d, it points to the position to put the
3369 * very last byte of the expanded string. Go backwards through
3370 * the string, copying and expanding as we go, stopping when we
3371 * get to the part that is invariant the rest of the way down */
3375 const U8 ch = NATIVE8_TO_UNI(*e--);
3376 if (UNI_IS_INVARIANT(ch)) {
3377 *d-- = UNI_TO_NATIVE(ch);
3379 *d-- = (U8)UTF8_EIGHT_BIT_LO(ch);
3380 *d-- = (U8)UTF8_EIGHT_BIT_HI(ch);
3385 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
3386 /* Update pos. We do it at the end rather than during
3387 * the upgrade, to avoid slowing down the common case
3388 * (upgrade without pos) */
3389 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
3391 I32 pos = mg->mg_len;
3392 if (pos > 0 && (U32)pos > invariant_head) {
3393 U8 *d = (U8*) SvPVX(sv) + invariant_head;
3394 STRLEN n = (U32)pos - invariant_head;
3396 if (UTF8_IS_START(*d))
3401 mg->mg_len = d - (U8*)SvPVX(sv);
3404 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
3405 magic_setutf8(sv,mg); /* clear UTF8 cache */
3410 /* Mark as UTF-8 even if no variant - saves scanning loop */
3416 =for apidoc sv_utf8_downgrade
3418 Attempts to convert the PV of an SV from characters to bytes.
3419 If the PV contains a character that cannot fit
3420 in a byte, this conversion will fail;
3421 in this case, either returns false or, if C<fail_ok> is not
3424 This is not as a general purpose Unicode to byte encoding interface:
3425 use the Encode extension for that.
3431 Perl_sv_utf8_downgrade(pTHX_ register SV *const sv, const bool fail_ok)
3435 PERL_ARGS_ASSERT_SV_UTF8_DOWNGRADE;
3437 if (SvPOKp(sv) && SvUTF8(sv)) {
3441 int mg_flags = SV_GMAGIC;
3444 sv_force_normal_flags(sv, 0);
3446 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
3448 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
3450 I32 pos = mg->mg_len;
3452 sv_pos_b2u(sv, &pos);
3453 mg_flags = 0; /* sv_pos_b2u does get magic */
3457 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
3458 magic_setutf8(sv,mg); /* clear UTF8 cache */
3461 s = (U8 *) SvPV_flags(sv, len, mg_flags);
3463 if (!utf8_to_bytes(s, &len)) {
3468 Perl_croak(aTHX_ "Wide character in %s",
3471 Perl_croak(aTHX_ "Wide character");
3482 =for apidoc sv_utf8_encode
3484 Converts the PV of an SV to UTF-8, but then turns the C<SvUTF8>
3485 flag off so that it looks like octets again.
3491 Perl_sv_utf8_encode(pTHX_ register SV *const sv)
3493 PERL_ARGS_ASSERT_SV_UTF8_ENCODE;
3495 if (SvREADONLY(sv)) {
3496 sv_force_normal_flags(sv, 0);
3498 (void) sv_utf8_upgrade(sv);
3503 =for apidoc sv_utf8_decode
3505 If the PV of the SV is an octet sequence in UTF-8
3506 and contains a multiple-byte character, the C<SvUTF8> flag is turned on
3507 so that it looks like a character. If the PV contains only single-byte
3508 characters, the C<SvUTF8> flag stays off.
3509 Scans PV for validity and returns false if the PV is invalid UTF-8.
3515 Perl_sv_utf8_decode(pTHX_ register SV *const sv)
3517 PERL_ARGS_ASSERT_SV_UTF8_DECODE;
3520 const U8 *start, *c;
3523 /* The octets may have got themselves encoded - get them back as
3526 if (!sv_utf8_downgrade(sv, TRUE))
3529 /* it is actually just a matter of turning the utf8 flag on, but
3530 * we want to make sure everything inside is valid utf8 first.
3532 c = start = (const U8 *) SvPVX_const(sv);
3533 if (!is_utf8_string(c, SvCUR(sv)))
3535 e = (const U8 *) SvEND(sv);
3538 if (!UTF8_IS_INVARIANT(ch)) {
3543 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
3544 /* adjust pos to the start of a UTF8 char sequence */
3545 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
3547 I32 pos = mg->mg_len;
3549 for (c = start + pos; c > start; c--) {
3550 if (UTF8_IS_START(*c))
3553 mg->mg_len = c - start;
3556 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
3557 magic_setutf8(sv,mg); /* clear UTF8 cache */
3564 =for apidoc sv_setsv
3566 Copies the contents of the source SV C<ssv> into the destination SV
3567 C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3568 function if the source SV needs to be reused. Does not handle 'set' magic.
3569 Loosely speaking, it performs a copy-by-value, obliterating any previous
3570 content of the destination.
3572 You probably want to use one of the assortment of wrappers, such as
3573 C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3574 C<SvSetMagicSV_nosteal>.
3576 =for apidoc sv_setsv_flags
3578 Copies the contents of the source SV C<ssv> into the destination SV
3579 C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3580 function if the source SV needs to be reused. Does not handle 'set' magic.
3581 Loosely speaking, it performs a copy-by-value, obliterating any previous
3582 content of the destination.
3583 If the C<flags> parameter has the C<SV_GMAGIC> bit set, will C<mg_get> on
3584 C<ssv> if appropriate, else not. If the C<flags>
3585 parameter has the C<NOSTEAL> bit set then the
3586 buffers of temps will not be stolen. <sv_setsv>
3587 and C<sv_setsv_nomg> are implemented in terms of this function.
3589 You probably want to use one of the assortment of wrappers, such as
3590 C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3591 C<SvSetMagicSV_nosteal>.
3593 This is the primary function for copying scalars, and most other
3594 copy-ish functions and macros use this underneath.
3600 S_glob_assign_glob(pTHX_ SV *const dstr, SV *const sstr, const int dtype)
3602 I32 mro_changes = 0; /* 1 = method, 2 = isa, 3 = recursive isa */
3603 HV *old_stash = NULL;
3605 PERL_ARGS_ASSERT_GLOB_ASSIGN_GLOB;
3607 if (dtype != SVt_PVGV && !isGV_with_GP(dstr)) {
3608 const char * const name = GvNAME(sstr);
3609 const STRLEN len = GvNAMELEN(sstr);
3611 if (dtype >= SVt_PV) {
3617 SvUPGRADE(dstr, SVt_PVGV);
3618 (void)SvOK_off(dstr);
3619 /* We have to turn this on here, even though we turn it off
3620 below, as GvSTASH will fail an assertion otherwise. */
3621 isGV_with_GP_on(dstr);
3623 GvSTASH(dstr) = GvSTASH(sstr);
3625 Perl_sv_add_backref(aTHX_ MUTABLE_SV(GvSTASH(dstr)), dstr);
3626 gv_name_set(MUTABLE_GV(dstr), name, len,
3627 GV_ADD | (GvNAMEUTF8(sstr) ? SVf_UTF8 : 0 ));
3628 SvFAKE_on(dstr); /* can coerce to non-glob */
3631 if(GvGP(MUTABLE_GV(sstr))) {
3632 /* If source has method cache entry, clear it */
3634 SvREFCNT_dec(GvCV(sstr));
3635 GvCV_set(sstr, NULL);
3638 /* If source has a real method, then a method is
3641 GvCV((const GV *)sstr) && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3647 /* If dest already had a real method, that's a change as well */
3649 !mro_changes && GvGP(MUTABLE_GV(dstr)) && GvCVu((const GV *)dstr)
3650 && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3655 /* We don't need to check the name of the destination if it was not a
3656 glob to begin with. */
3657 if(dtype == SVt_PVGV) {
3658 const char * const name = GvNAME((const GV *)dstr);
3661 /* The stash may have been detached from the symbol table, so
3663 && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3667 const STRLEN len = GvNAMELEN(dstr);
3668 if ((len > 1 && name[len-2] == ':' && name[len-1] == ':')
3669 || (len == 1 && name[0] == ':')) {
3672 /* Set aside the old stash, so we can reset isa caches on
3674 if((old_stash = GvHV(dstr)))
3675 /* Make sure we do not lose it early. */
3676 SvREFCNT_inc_simple_void_NN(
3677 sv_2mortal((SV *)old_stash)
3683 gp_free(MUTABLE_GV(dstr));
3684 isGV_with_GP_off(dstr); /* SvOK_off does not like globs. */
3685 (void)SvOK_off(dstr);
3686 isGV_with_GP_on(dstr);
3687 GvINTRO_off(dstr); /* one-shot flag */
3688 GvGP_set(dstr, gp_ref(GvGP(sstr)));
3689 if (SvTAINTED(sstr))
3691 if (GvIMPORTED(dstr) != GVf_IMPORTED
3692 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3694 GvIMPORTED_on(dstr);
3697 if(mro_changes == 2) {
3698 if (GvAV((const GV *)sstr)) {
3700 SV * const sref = (SV *)GvAV((const GV *)dstr);
3701 if (SvSMAGICAL(sref) && (mg = mg_find(sref, PERL_MAGIC_isa))) {
3702 if (SvTYPE(mg->mg_obj) != SVt_PVAV) {
3703 AV * const ary = newAV();
3704 av_push(ary, mg->mg_obj); /* takes the refcount */
3705 mg->mg_obj = (SV *)ary;
3707 av_push((AV *)mg->mg_obj, SvREFCNT_inc_simple_NN(dstr));
3709 else sv_magic(sref, dstr, PERL_MAGIC_isa, NULL, 0);
3711 mro_isa_changed_in(GvSTASH(dstr));
3713 else if(mro_changes == 3) {
3714 HV * const stash = GvHV(dstr);
3715 if(old_stash ? (HV *)HvENAME_get(old_stash) : stash)
3721 else if(mro_changes) mro_method_changed_in(GvSTASH(dstr));
3726 S_glob_assign_ref(pTHX_ SV *const dstr, SV *const sstr)
3728 SV * const sref = SvREFCNT_inc(SvRV(sstr));
3730 const int intro = GvINTRO(dstr);
3733 const U32 stype = SvTYPE(sref);
3735 PERL_ARGS_ASSERT_GLOB_ASSIGN_REF;
3738 GvINTRO_off(dstr); /* one-shot flag */
3739 GvLINE(dstr) = CopLINE(PL_curcop);
3740 GvEGV(dstr) = MUTABLE_GV(dstr);
3745 location = (SV **) &(GvGP(dstr)->gp_cv); /* XXX bypassing GvCV_set */
3746 import_flag = GVf_IMPORTED_CV;
3749 location = (SV **) &GvHV(dstr);
3750 import_flag = GVf_IMPORTED_HV;
3753 location = (SV **) &GvAV(dstr);
3754 import_flag = GVf_IMPORTED_AV;
3757 location = (SV **) &GvIOp(dstr);
3760 location = (SV **) &GvFORM(dstr);
3763 location = &GvSV(dstr);
3764 import_flag = GVf_IMPORTED_SV;
3767 if (stype == SVt_PVCV) {
3768 /*if (GvCVGEN(dstr) && (GvCV(dstr) != (const CV *)sref || GvCVGEN(dstr))) {*/
3769 if (GvCVGEN(dstr)) {
3770 SvREFCNT_dec(GvCV(dstr));
3771 GvCV_set(dstr, NULL);
3772 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3775 SAVEGENERICSV(*location);
3779 if (stype == SVt_PVCV && (*location != sref || GvCVGEN(dstr))) {
3780 CV* const cv = MUTABLE_CV(*location);
3782 if (!GvCVGEN((const GV *)dstr) &&
3783 (CvROOT(cv) || CvXSUB(cv)) &&
3784 /* redundant check that avoids creating the extra SV
3785 most of the time: */
3786 (CvCONST(cv) || ckWARN(WARN_REDEFINE)))
3788 SV * const new_const_sv =
3789 CvCONST((const CV *)sref)
3790 ? cv_const_sv((const CV *)sref)
3792 report_redefined_cv(
3793 sv_2mortal(Perl_newSVpvf(aTHX_
3796 HvNAME_HEK(GvSTASH((const GV *)dstr))
3798 HEKfARG(GvENAME_HEK(MUTABLE_GV(dstr)))
3801 CvCONST((const CV *)sref) ? &new_const_sv : NULL
3805 cv_ckproto_len_flags(cv, (const GV *)dstr,
3806 SvPOK(sref) ? CvPROTO(sref) : NULL,
3807 SvPOK(sref) ? CvPROTOLEN(sref) : 0,
3808 SvPOK(sref) ? SvUTF8(sref) : 0);
3810 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3811 GvASSUMECV_on(dstr);
3812 if(GvSTASH(dstr)) mro_method_changed_in(GvSTASH(dstr)); /* sub foo { 1 } sub bar { 2 } *bar = \&foo */
3815 if (import_flag && !(GvFLAGS(dstr) & import_flag)
3816 && CopSTASH_ne(PL_curcop, GvSTASH(dstr))) {
3817 GvFLAGS(dstr) |= import_flag;
3819 if (stype == SVt_PVHV) {
3820 const char * const name = GvNAME((GV*)dstr);
3821 const STRLEN len = GvNAMELEN(dstr);
3824 (len > 1 && name[len-2] == ':' && name[len-1] == ':')
3825 || (len == 1 && name[0] == ':')
3827 && (!dref || HvENAME_get(dref))
3830 (HV *)sref, (HV *)dref,
3836 stype == SVt_PVAV && sref != dref
3837 && strEQ(GvNAME((GV*)dstr), "ISA")
3838 /* The stash may have been detached from the symbol table, so
3839 check its name before doing anything. */
3840 && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3843 MAGIC * const omg = dref && SvSMAGICAL(dref)
3844 ? mg_find(dref, PERL_MAGIC_isa)
3846 if (SvSMAGICAL(sref) && (mg = mg_find(sref, PERL_MAGIC_isa))) {
3847 if (SvTYPE(mg->mg_obj) != SVt_PVAV) {
3848 AV * const ary = newAV();
3849 av_push(ary, mg->mg_obj); /* takes the refcount */
3850 mg->mg_obj = (SV *)ary;
3853 if (SvTYPE(omg->mg_obj) == SVt_PVAV) {
3854 SV **svp = AvARRAY((AV *)omg->mg_obj);
3855 I32 items = AvFILLp((AV *)omg->mg_obj) + 1;
3859 SvREFCNT_inc_simple_NN(*svp++)
3865 SvREFCNT_inc_simple_NN(omg->mg_obj)
3869 av_push((AV *)mg->mg_obj,SvREFCNT_inc_simple_NN(dstr));
3874 sref, omg ? omg->mg_obj : dstr, PERL_MAGIC_isa, NULL, 0
3876 mg = mg_find(sref, PERL_MAGIC_isa);
3878 /* Since the *ISA assignment could have affected more than
3879 one stash, don't call mro_isa_changed_in directly, but let
3880 magic_clearisa do it for us, as it already has the logic for
3881 dealing with globs vs arrays of globs. */
3883 Perl_magic_clearisa(aTHX_ NULL, mg);
3888 if (SvTAINTED(sstr))
3894 Perl_sv_setsv_flags(pTHX_ SV *dstr, register SV* sstr, const I32 flags)
3897 register U32 sflags;
3899 register svtype stype;
3901 PERL_ARGS_ASSERT_SV_SETSV_FLAGS;
3906 if (SvIS_FREED(dstr)) {
3907 Perl_croak(aTHX_ "panic: attempt to copy value %" SVf
3908 " to a freed scalar %p", SVfARG(sstr), (void *)dstr);
3910 SV_CHECK_THINKFIRST_COW_DROP(dstr);
3912 sstr = &PL_sv_undef;
3913 if (SvIS_FREED(sstr)) {
3914 Perl_croak(aTHX_ "panic: attempt to copy freed scalar %p to %p",
3915 (void*)sstr, (void*)dstr);
3917 stype = SvTYPE(sstr);
3918 dtype = SvTYPE(dstr);
3920 /* There's a lot of redundancy below but we're going for speed here */
3925 if (dtype != SVt_PVGV && dtype != SVt_PVLV) {
3926 (void)SvOK_off(dstr);
3934 sv_upgrade(dstr, SVt_IV);
3938 sv_upgrade(dstr, SVt_PVIV);
3942 goto end_of_first_switch;
3944 (void)SvIOK_only(dstr);
3945 SvIV_set(dstr, SvIVX(sstr));
3948 /* SvTAINTED can only be true if the SV has taint magic, which in
3949 turn means that the SV type is PVMG (or greater). This is the
3950 case statement for SVt_IV, so this cannot be true (whatever gcov
3952 assert(!SvTAINTED(sstr));
3957 if (dtype < SVt_PV && dtype != SVt_IV)
3958 sv_upgrade(dstr, SVt_IV);
3966 sv_upgrade(dstr, SVt_NV);
3970 sv_upgrade(dstr, SVt_PVNV);
3974 goto end_of_first_switch;
3976 SvNV_set(dstr, SvNVX(sstr));
3977 (void)SvNOK_only(dstr);
3978 /* SvTAINTED can only be true if the SV has taint magic, which in
3979 turn means that the SV type is PVMG (or greater). This is the
3980 case statement for SVt_NV, so this cannot be true (whatever gcov
3982 assert(!SvTAINTED(sstr));
3988 #ifdef PERL_OLD_COPY_ON_WRITE
3989 if ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS) {
3990 if (dtype < SVt_PVIV)
3991 sv_upgrade(dstr, SVt_PVIV);
3998 sv_upgrade(dstr, SVt_PV);
4001 if (dtype < SVt_PVIV)
4002 sv_upgrade(dstr, SVt_PVIV);
4005 if (dtype < SVt_PVNV)
4006 sv_upgrade(dstr, SVt_PVNV);
4010 const char * const type = sv_reftype(sstr,0);
4012 /* diag_listed_as: Bizarre copy of %s */
4013 Perl_croak(aTHX_ "Bizarre copy of %s in %s", type, OP_DESC(PL_op));
4015 Perl_croak(aTHX_ "Bizarre copy of %s", type);
4020 if (dtype < SVt_REGEXP)
4021 sv_upgrade(dstr, SVt_REGEXP);
4024 /* case SVt_BIND: */
4028 if (SvGMAGICAL(sstr) && (flags & SV_GMAGIC)) {
4030 if (SvTYPE(sstr) != stype)
4031 stype = SvTYPE(sstr);
4033 if (isGV_with_GP(sstr) && dtype <= SVt_PVLV) {
4034 glob_assign_glob(dstr, sstr, dtype);
4037 if (stype == SVt_PVLV)
4038 SvUPGRADE(dstr, SVt_PVNV);
4040 SvUPGRADE(dstr, (svtype)stype);
4042 end_of_first_switch:
4044 /* dstr may have been upgraded. */
4045 dtype = SvTYPE(dstr);
4046 sflags = SvFLAGS(sstr);
4048 if (dtype == SVt_PVCV || dtype == SVt_PVFM) {
4049 /* Assigning to a subroutine sets the prototype. */
4052 const char *const ptr = SvPV_const(sstr, len);
4054 SvGROW(dstr, len + 1);
4055 Copy(ptr, SvPVX(dstr), len + 1, char);
4056 SvCUR_set(dstr, len);
4058 SvFLAGS(dstr) |= sflags & SVf_UTF8;
4059 CvAUTOLOAD_off(dstr);
4063 } else if (dtype == SVt_PVAV || dtype == SVt_PVHV) {
4064 const char * const type = sv_reftype(dstr,0);
4066 /* diag_listed_as: Cannot copy to %s */
4067 Perl_croak(aTHX_ "Cannot copy to %s in %s", type, OP_DESC(PL_op));
4069 Perl_croak(aTHX_ "Cannot copy to %s", type);
4070 } else if (sflags & SVf_ROK) {
4071 if (isGV_with_GP(dstr)
4072 && SvTYPE(SvRV(sstr)) == SVt_PVGV && isGV_with_GP(SvRV(sstr))) {
4075 if (GvIMPORTED(dstr) != GVf_IMPORTED
4076 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
4078 GvIMPORTED_on(dstr);
4083 glob_assign_glob(dstr, sstr, dtype);
4087 if (dtype >= SVt_PV) {
4088 if (isGV_with_GP(dstr)) {
4089 glob_assign_ref(dstr, sstr);
4092 if (SvPVX_const(dstr)) {
4098 (void)SvOK_off(dstr);
4099 SvRV_set(dstr, SvREFCNT_inc(SvRV(sstr)));
4100 SvFLAGS(dstr) |= sflags & SVf_ROK;
4101 assert(!(sflags & SVp_NOK));
4102 assert(!(sflags & SVp_IOK));
4103 assert(!(sflags & SVf_NOK));
4104 assert(!(sflags & SVf_IOK));
4106 else if (isGV_with_GP(dstr)) {
4107 if (!(sflags & SVf_OK)) {
4108 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
4109 "Undefined value assigned to typeglob");
4112 GV *gv = gv_fetchsv_nomg(sstr, GV_ADD, SVt_PVGV);
4113 if (dstr != (const SV *)gv) {
4114 const char * const name = GvNAME((const GV *)dstr);
4115 const STRLEN len = GvNAMELEN(dstr);
4116 HV *old_stash = NULL;
4117 bool reset_isa = FALSE;
4118 if ((len > 1 && name[len-2] == ':' && name[len-1] == ':')
4119 || (len == 1 && name[0] == ':')) {
4120 /* Set aside the old stash, so we can reset isa caches
4121 on its subclasses. */
4122 if((old_stash = GvHV(dstr))) {
4123 /* Make sure we do not lose it early. */
4124 SvREFCNT_inc_simple_void_NN(
4125 sv_2mortal((SV *)old_stash)
4132 gp_free(MUTABLE_GV(dstr));
4133 GvGP_set(dstr, gp_ref(GvGP(gv)));
4136 HV * const stash = GvHV(dstr);
4138 old_stash ? (HV *)HvENAME_get(old_stash) : stash
4148 else if (dtype == SVt_REGEXP && stype == SVt_REGEXP) {
4149 reg_temp_copy((REGEXP*)dstr, (REGEXP*)sstr);
4151 else if (sflags & SVp_POK) {
4155 * Check to see if we can just swipe the string. If so, it's a
4156 * possible small lose on short strings, but a big win on long ones.
4157 * It might even be a win on short strings if SvPVX_const(dstr)
4158 * has to be allocated and SvPVX_const(sstr) has to be freed.
4159 * Likewise if we can set up COW rather than doing an actual copy, we
4160 * drop to the else clause, as the swipe code and the COW setup code
4161 * have much in common.
4164 /* Whichever path we take through the next code, we want this true,
4165 and doing it now facilitates the COW check. */
4166 (void)SvPOK_only(dstr);
4169 /* If we're already COW then this clause is not true, and if COW
4170 is allowed then we drop down to the else and make dest COW
4171 with us. If caller hasn't said that we're allowed to COW
4172 shared hash keys then we don't do the COW setup, even if the
4173 source scalar is a shared hash key scalar. */
4174 (((flags & SV_COW_SHARED_HASH_KEYS)
4175 ? (sflags & (SVf_FAKE|SVf_READONLY)) != (SVf_FAKE|SVf_READONLY)
4176 : 1 /* If making a COW copy is forbidden then the behaviour we
4177 desire is as if the source SV isn't actually already
4178 COW, even if it is. So we act as if the source flags
4179 are not COW, rather than actually testing them. */
4181 #ifndef PERL_OLD_COPY_ON_WRITE
4182 /* The change that added SV_COW_SHARED_HASH_KEYS makes the logic
4183 when PERL_OLD_COPY_ON_WRITE is defined a little wrong.
4184 Conceptually PERL_OLD_COPY_ON_WRITE being defined should
4185 override SV_COW_SHARED_HASH_KEYS, because it means "always COW"
4186 but in turn, it's somewhat dead code, never expected to go
4187 live, but more kept as a placeholder on how to do it better
4188 in a newer implementation. */
4189 /* If we are COW and dstr is a suitable target then we drop down
4190 into the else and make dest a COW of us. */
4191 || (SvFLAGS(dstr) & CAN_COW_MASK) != CAN_COW_FLAGS
4196 (sflags & SVs_TEMP) && /* slated for free anyway? */
4197 !(sflags & SVf_OOK) && /* and not involved in OOK hack? */
4198 (!(flags & SV_NOSTEAL)) &&
4199 /* and we're allowed to steal temps */
4200 SvREFCNT(sstr) == 1 && /* and no other references to it? */
4201 SvLEN(sstr)) /* and really is a string */
4202 #ifdef PERL_OLD_COPY_ON_WRITE
4203 && ((flags & SV_COW_SHARED_HASH_KEYS)
4204 ? (!((sflags & CAN_COW_MASK) == CAN_COW_FLAGS
4205 && (SvFLAGS(dstr) & CAN_COW_MASK) == CAN_COW_FLAGS
4206 && SvTYPE(sstr) >= SVt_PVIV && SvTYPE(sstr) != SVt_PVFM))
4210 /* Failed the swipe test, and it's not a shared hash key either.
4211 Have to copy the string. */
4212 STRLEN len = SvCUR(sstr);
4213 SvGROW(dstr, len + 1); /* inlined from sv_setpvn */
4214 Move(SvPVX_const(sstr),SvPVX(dstr),len,char);
4215 SvCUR_set(dstr, len);
4216 *SvEND(dstr) = '\0';
4218 /* If PERL_OLD_COPY_ON_WRITE is not defined, then isSwipe will always
4220 /* Either it's a shared hash key, or it's suitable for
4221 copy-on-write or we can swipe the string. */
4223 PerlIO_printf(Perl_debug_log, "Copy on write: sstr --> dstr\n");
4227 #ifdef PERL_OLD_COPY_ON_WRITE
4229 if ((sflags & (SVf_FAKE | SVf_READONLY))
4230 != (SVf_FAKE | SVf_READONLY)) {
4231 SvREADONLY_on(sstr);
4233 /* Make the source SV into a loop of 1.
4234 (about to become 2) */
4235 SV_COW_NEXT_SV_SET(sstr, sstr);
4239 /* Initial code is common. */
4240 if (SvPVX_const(dstr)) { /* we know that dtype >= SVt_PV */
4245 /* making another shared SV. */
4246 STRLEN cur = SvCUR(sstr);
4247 STRLEN len = SvLEN(sstr);
4248 #ifdef PERL_OLD_COPY_ON_WRITE
4250 assert (SvTYPE(dstr) >= SVt_PVIV);
4251 /* SvIsCOW_normal */
4252 /* splice us in between source and next-after-source. */
4253 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4254 SV_COW_NEXT_SV_SET(sstr, dstr);
4255 SvPV_set(dstr, SvPVX_mutable(sstr));
4259 /* SvIsCOW_shared_hash */
4260 DEBUG_C(PerlIO_printf(Perl_debug_log,
4261 "Copy on write: Sharing hash\n"));
4263 assert (SvTYPE(dstr) >= SVt_PV);
4265 HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)))));
4267 SvLEN_set(dstr, len);
4268 SvCUR_set(dstr, cur);
4269 SvREADONLY_on(dstr);
4273 { /* Passes the swipe test. */
4274 SvPV_set(dstr, SvPVX_mutable(sstr));
4275 SvLEN_set(dstr, SvLEN(sstr));
4276 SvCUR_set(dstr, SvCUR(sstr));
4279 (void)SvOK_off(sstr); /* NOTE: nukes most SvFLAGS on sstr */
4280 SvPV_set(sstr, NULL);
4286 if (sflags & SVp_NOK) {
4287 SvNV_set(dstr, SvNVX(sstr));
4289 if (sflags & SVp_IOK) {
4290 SvIV_set(dstr, SvIVX(sstr));
4291 /* Must do this otherwise some other overloaded use of 0x80000000
4292 gets confused. I guess SVpbm_VALID */
4293 if (sflags & SVf_IVisUV)
4296 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_NOK|SVp_NOK|SVf_UTF8);
4298 const MAGIC * const smg = SvVSTRING_mg(sstr);
4300 sv_magic(dstr, NULL, PERL_MAGIC_vstring,
4301 smg->mg_ptr, smg->mg_len);
4302 SvRMAGICAL_on(dstr);
4306 else if (sflags & (SVp_IOK|SVp_NOK)) {
4307 (void)SvOK_off(dstr);
4308 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_IVisUV|SVf_NOK|SVp_NOK);
4309 if (sflags & SVp_IOK) {
4310 /* XXXX Do we want to set IsUV for IV(ROK)? Be extra safe... */
4311 SvIV_set(dstr, SvIVX(sstr));
4313 if (sflags & SVp_NOK) {
4314 SvNV_set(dstr, SvNVX(sstr));
4318 if (isGV_with_GP(sstr)) {
4319 gv_efullname3(dstr, MUTABLE_GV(sstr), "*");
4322 (void)SvOK_off(dstr);
4324 if (SvTAINTED(sstr))
4329 =for apidoc sv_setsv_mg
4331 Like C<sv_setsv>, but also handles 'set' magic.
4337 Perl_sv_setsv_mg(pTHX_ SV *const dstr, register SV *const sstr)
4339 PERL_ARGS_ASSERT_SV_SETSV_MG;
4341 sv_setsv(dstr,sstr);
4345 #ifdef PERL_OLD_COPY_ON_WRITE
4347 Perl_sv_setsv_cow(pTHX_ SV *dstr, SV *sstr)
4349 STRLEN cur = SvCUR(sstr);
4350 STRLEN len = SvLEN(sstr);
4351 register char *new_pv;
4353 PERL_ARGS_ASSERT_SV_SETSV_COW;
4356 PerlIO_printf(Perl_debug_log, "Fast copy on write: %p -> %p\n",
4357 (void*)sstr, (void*)dstr);
4364 if (SvTHINKFIRST(dstr))
4365 sv_force_normal_flags(dstr, SV_COW_DROP_PV);
4366 else if (SvPVX_const(dstr))
4367 Safefree(SvPVX_const(dstr));
4371 SvUPGRADE(dstr, SVt_PVIV);
4373 assert (SvPOK(sstr));
4374 assert (SvPOKp(sstr));
4375 assert (!SvIOK(sstr));
4376 assert (!SvIOKp(sstr));
4377 assert (!SvNOK(sstr));
4378 assert (!SvNOKp(sstr));
4380 if (SvIsCOW(sstr)) {
4382 if (SvLEN(sstr) == 0) {
4383 /* source is a COW shared hash key. */
4384 DEBUG_C(PerlIO_printf(Perl_debug_log,
4385 "Fast copy on write: Sharing hash\n"));
4386 new_pv = HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr))));
4389 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4391 assert ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS);
4392 SvUPGRADE(sstr, SVt_PVIV);
4393 SvREADONLY_on(sstr);
4395 DEBUG_C(PerlIO_printf(Perl_debug_log,
4396 "Fast copy on write: Converting sstr to COW\n"));
4397 SV_COW_NEXT_SV_SET(dstr, sstr);
4399 SV_COW_NEXT_SV_SET(sstr, dstr);
4400 new_pv = SvPVX_mutable(sstr);
4403 SvPV_set(dstr, new_pv);
4404 SvFLAGS(dstr) = (SVt_PVIV|SVf_POK|SVp_POK|SVf_FAKE|SVf_READONLY);
4407 SvLEN_set(dstr, len);
4408 SvCUR_set(dstr, cur);
4417 =for apidoc sv_setpvn
4419 Copies a string into an SV. The C<len> parameter indicates the number of
4420 bytes to be copied. If the C<ptr> argument is NULL the SV will become
4421 undefined. Does not handle 'set' magic. See C<sv_setpvn_mg>.
4427 Perl_sv_setpvn(pTHX_ register SV *const sv, register const char *const ptr, register const STRLEN len)
4430 register char *dptr;
4432 PERL_ARGS_ASSERT_SV_SETPVN;
4434 SV_CHECK_THINKFIRST_COW_DROP(sv);
4440 /* len is STRLEN which is unsigned, need to copy to signed */
4443 Perl_croak(aTHX_ "panic: sv_setpvn called with negative strlen %"
4446 SvUPGRADE(sv, SVt_PV);
4448 dptr = SvGROW(sv, len + 1);
4449 Move(ptr,dptr,len,char);
4452 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4454 if (SvTYPE(sv) == SVt_PVCV) CvAUTOLOAD_off(sv);
4458 =for apidoc sv_setpvn_mg
4460 Like C<sv_setpvn>, but also handles 'set' magic.
4466 Perl_sv_setpvn_mg(pTHX_ register SV *const sv, register const char *const ptr, register const STRLEN len)
4468 PERL_ARGS_ASSERT_SV_SETPVN_MG;
4470 sv_setpvn(sv,ptr,len);
4475 =for apidoc sv_setpv
4477 Copies a string into an SV. The string must be null-terminated. Does not
4478 handle 'set' magic. See C<sv_setpv_mg>.
4484 Perl_sv_setpv(pTHX_ register SV *const sv, register const char *const ptr)
4487 register STRLEN len;
4489 PERL_ARGS_ASSERT_SV_SETPV;
4491 SV_CHECK_THINKFIRST_COW_DROP(sv);
4497 SvUPGRADE(sv, SVt_PV);
4499 SvGROW(sv, len + 1);
4500 Move(ptr,SvPVX(sv),len+1,char);
4502 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4504 if (SvTYPE(sv) == SVt_PVCV) CvAUTOLOAD_off(sv);
4508 =for apidoc sv_setpv_mg
4510 Like C<sv_setpv>, but also handles 'set' magic.
4516 Perl_sv_setpv_mg(pTHX_ register SV *const sv, register const char *const ptr)
4518 PERL_ARGS_ASSERT_SV_SETPV_MG;
4525 Perl_sv_sethek(pTHX_ register SV *const sv, const HEK *const hek)
4529 PERL_ARGS_ASSERT_SV_SETHEK;
4535 if (HEK_LEN(hek) == HEf_SVKEY) {
4536 sv_setsv(sv, *(SV**)HEK_KEY(hek));
4539 const int flags = HEK_FLAGS(hek);
4540 if (flags & HVhek_WASUTF8) {
4541 STRLEN utf8_len = HEK_LEN(hek);
4542 char *as_utf8 = (char *)bytes_to_utf8((U8*)HEK_KEY(hek), &utf8_len);
4543 sv_usepvn_flags(sv, as_utf8, utf8_len, SV_HAS_TRAILING_NUL);
4546 } else if (flags & (HVhek_REHASH|HVhek_UNSHARED)) {
4547 sv_setpvn(sv, HEK_KEY(hek), HEK_LEN(hek));
4550 else SvUTF8_off(sv);
4554 SV_CHECK_THINKFIRST_COW_DROP(sv);
4555 SvUPGRADE(sv, SVt_PV);
4556 Safefree(SvPVX(sv));
4557 SvPV_set(sv,(char *)HEK_KEY(share_hek_hek(hek)));
4558 SvCUR_set(sv, HEK_LEN(hek));
4565 else SvUTF8_off(sv);
4573 =for apidoc sv_usepvn_flags
4575 Tells an SV to use C<ptr> to find its string value. Normally the
4576 string is stored inside the SV but sv_usepvn allows the SV to use an
4577 outside string. The C<ptr> should point to memory that was allocated
4578 by C<malloc>. It must be the start of a mallocked block
4579 of memory, and not a pointer to the middle of it. The
4580 string length, C<len>, must be supplied. By default
4581 this function will realloc (i.e. move) the memory pointed to by C<ptr>,
4582 so that pointer should not be freed or used by the programmer after
4583 giving it to sv_usepvn, and neither should any pointers from "behind"
4584 that pointer (e.g. ptr + 1) be used.
4586 If C<flags> & SV_SMAGIC is true, will call SvSETMAGIC. If C<flags> &
4587 SV_HAS_TRAILING_NUL is true, then C<ptr[len]> must be NUL, and the realloc
4588 will be skipped (i.e. the buffer is actually at least 1 byte longer than
4589 C<len>, and already meets the requirements for storing in C<SvPVX>).
4595 Perl_sv_usepvn_flags(pTHX_ SV *const sv, char *ptr, const STRLEN len, const U32 flags)
4600 PERL_ARGS_ASSERT_SV_USEPVN_FLAGS;
4602 SV_CHECK_THINKFIRST_COW_DROP(sv);
4603 SvUPGRADE(sv, SVt_PV);
4606 if (flags & SV_SMAGIC)
4610 if (SvPVX_const(sv))
4614 if (flags & SV_HAS_TRAILING_NUL)
4615 assert(ptr[len] == '\0');
4618 allocate = (flags & SV_HAS_TRAILING_NUL)
4620 #ifdef Perl_safesysmalloc_size
4623 PERL_STRLEN_ROUNDUP(len + 1);
4625 if (flags & SV_HAS_TRAILING_NUL) {
4626 /* It's long enough - do nothing.
4627 Specifically Perl_newCONSTSUB is relying on this. */
4630 /* Force a move to shake out bugs in callers. */
4631 char *new_ptr = (char*)safemalloc(allocate);
4632 Copy(ptr, new_ptr, len, char);
4633 PoisonFree(ptr,len,char);
4637 ptr = (char*) saferealloc (ptr, allocate);
4640 #ifdef Perl_safesysmalloc_size
4641 SvLEN_set(sv, Perl_safesysmalloc_size(ptr));
4643 SvLEN_set(sv, allocate);
4647 if (!(flags & SV_HAS_TRAILING_NUL)) {
4650 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4652 if (flags & SV_SMAGIC)
4656 #ifdef PERL_OLD_COPY_ON_WRITE
4657 /* Need to do this *after* making the SV normal, as we need the buffer
4658 pointer to remain valid until after we've copied it. If we let go too early,
4659 another thread could invalidate it by unsharing last of the same hash key
4660 (which it can do by means other than releasing copy-on-write Svs)
4661 or by changing the other copy-on-write SVs in the loop. */
4663 S_sv_release_COW(pTHX_ register SV *sv, const char *pvx, SV *after)
4665 PERL_ARGS_ASSERT_SV_RELEASE_COW;
4667 { /* this SV was SvIsCOW_normal(sv) */
4668 /* we need to find the SV pointing to us. */
4669 SV *current = SV_COW_NEXT_SV(after);
4671 if (current == sv) {
4672 /* The SV we point to points back to us (there were only two of us
4674 Hence other SV is no longer copy on write either. */
4676 SvREADONLY_off(after);
4678 /* We need to follow the pointers around the loop. */
4680 while ((next = SV_COW_NEXT_SV(current)) != sv) {
4683 /* don't loop forever if the structure is bust, and we have
4684 a pointer into a closed loop. */
4685 assert (current != after);
4686 assert (SvPVX_const(current) == pvx);
4688 /* Make the SV before us point to the SV after us. */
4689 SV_COW_NEXT_SV_SET(current, after);
4695 =for apidoc sv_force_normal_flags
4697 Undo various types of fakery on an SV, where fakery means
4698 "more than" a string: if the PV is a shared string, make
4699 a private copy; if we're a ref, stop refing; if we're a glob, downgrade to
4700 an xpvmg; if we're a copy-on-write scalar, this is the on-write time when
4701 we do the copy, and is also used locally; if this is a
4702 vstring, drop the vstring magic. If C<SV_COW_DROP_PV> is set
4703 then a copy-on-write scalar drops its PV buffer (if any) and becomes
4704 SvPOK_off rather than making a copy. (Used where this
4705 scalar is about to be set to some other value.) In addition,
4706 the C<flags> parameter gets passed to C<sv_unref_flags()>
4707 when unreffing. C<sv_force_normal> calls this function
4708 with flags set to 0.
4714 Perl_sv_force_normal_flags(pTHX_ register SV *const sv, const U32 flags)
4718 PERL_ARGS_ASSERT_SV_FORCE_NORMAL_FLAGS;
4720 #ifdef PERL_OLD_COPY_ON_WRITE
4721 if (SvREADONLY(sv)) {
4723 const char * const pvx = SvPVX_const(sv);
4724 const STRLEN len = SvLEN(sv);
4725 const STRLEN cur = SvCUR(sv);
4726 /* next COW sv in the loop. If len is 0 then this is a shared-hash
4727 key scalar, so we mustn't attempt to call SV_COW_NEXT_SV(), as
4728 we'll fail an assertion. */
4729 SV * const next = len ? SV_COW_NEXT_SV(sv) : 0;
4732 PerlIO_printf(Perl_debug_log,
4733 "Copy on write: Force normal %ld\n",
4739 /* This SV doesn't own the buffer, so need to Newx() a new one: */
4742 if (flags & SV_COW_DROP_PV) {
4743 /* OK, so we don't need to copy our buffer. */
4746 SvGROW(sv, cur + 1);
4747 Move(pvx,SvPVX(sv),cur,char);
4752 sv_release_COW(sv, pvx, next);
4754 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4760 else if (IN_PERL_RUNTIME)
4761 Perl_croak_no_modify(aTHX);
4764 if (SvREADONLY(sv)) {
4766 const char * const pvx = SvPVX_const(sv);
4767 const STRLEN len = SvCUR(sv);
4772 if (flags & SV_COW_DROP_PV) {
4773 /* OK, so we don't need to copy our buffer. */
4776 SvGROW(sv, len + 1);
4777 Move(pvx,SvPVX(sv),len,char);
4780 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4782 else if (IN_PERL_RUNTIME)
4783 Perl_croak_no_modify(aTHX);
4787 sv_unref_flags(sv, flags);
4788 else if (SvFAKE(sv) && isGV_with_GP(sv))
4789 sv_unglob(sv, flags);
4790 else if (SvFAKE(sv) && SvTYPE(sv) == SVt_REGEXP) {
4791 /* Need to downgrade the REGEXP to a simple(r) scalar. This is analogous
4792 to sv_unglob. We only need it here, so inline it. */
4793 const svtype new_type = SvMAGIC(sv) || SvSTASH(sv) ? SVt_PVMG : SVt_PV;
4794 SV *const temp = newSV_type(new_type);
4795 void *const temp_p = SvANY(sv);
4797 if (new_type == SVt_PVMG) {
4798 SvMAGIC_set(temp, SvMAGIC(sv));
4799 SvMAGIC_set(sv, NULL);
4800 SvSTASH_set(temp, SvSTASH(sv));
4801 SvSTASH_set(sv, NULL);
4803 SvCUR_set(temp, SvCUR(sv));
4804 /* Remember that SvPVX is in the head, not the body. */
4806 SvLEN_set(temp, SvLEN(sv));
4807 /* This signals "buffer is owned by someone else" in sv_clear,
4808 which is the least effort way to stop it freeing the buffer.
4810 SvLEN_set(sv, SvLEN(sv)+1);
4812 /* Their buffer is already owned by someone else. */
4813 SvPVX(sv) = savepvn(SvPVX(sv), SvCUR(sv));
4814 SvLEN_set(temp, SvCUR(sv)+1);
4817 /* Now swap the rest of the bodies. */
4819 SvFLAGS(sv) &= ~(SVf_FAKE|SVTYPEMASK);
4820 SvFLAGS(sv) |= new_type;
4821 SvANY(sv) = SvANY(temp);
4823 SvFLAGS(temp) &= ~(SVTYPEMASK);
4824 SvFLAGS(temp) |= SVt_REGEXP|SVf_FAKE;
4825 SvANY(temp) = temp_p;
4829 else if (SvVOK(sv)) sv_unmagic(sv, PERL_MAGIC_vstring);
4835 Efficient removal of characters from the beginning of the string buffer.
4836 SvPOK(sv), or at least SvPOKp(sv), must be true and the C<ptr> must be a
4837 pointer to somewhere inside the string buffer. The C<ptr> becomes the first
4838 character of the adjusted string. Uses the "OOK hack". On return, only
4839 SvPOK(sv) and SvPOKp(sv) among the OK flags will be true.
4841 Beware: after this function returns, C<ptr> and SvPVX_const(sv) may no longer
4842 refer to the same chunk of data.
4844 The unfortunate similarity of this function's name to that of Perl's C<chop>
4845 operator is strictly coincidental. This function works from the left;
4846 C<chop> works from the right.
4852 Perl_sv_chop(pTHX_ register SV *const sv, register const char *const ptr)
4863 PERL_ARGS_ASSERT_SV_CHOP;
4865 if (!ptr || !SvPOKp(sv))
4867 delta = ptr - SvPVX_const(sv);
4869 /* Nothing to do. */
4872 max_delta = SvLEN(sv) ? SvLEN(sv) : SvCUR(sv);
4873 if (delta > max_delta)
4874 Perl_croak(aTHX_ "panic: sv_chop ptr=%p, start=%p, end=%p",
4875 ptr, SvPVX_const(sv), SvPVX_const(sv) + max_delta);
4876 /* SvPVX(sv) may move in SV_CHECK_THINKFIRST(sv), so don't use ptr any more */
4877 SV_CHECK_THINKFIRST(sv);
4878 SvPOK_only_UTF8(sv);
4881 if (!SvLEN(sv)) { /* make copy of shared string */
4882 const char *pvx = SvPVX_const(sv);
4883 const STRLEN len = SvCUR(sv);
4884 SvGROW(sv, len + 1);
4885 Move(pvx,SvPVX(sv),len,char);
4891 SvOOK_offset(sv, old_delta);
4893 SvLEN_set(sv, SvLEN(sv) - delta);
4894 SvCUR_set(sv, SvCUR(sv) - delta);
4895 SvPV_set(sv, SvPVX(sv) + delta);
4897 p = (U8 *)SvPVX_const(sv);
4900 /* how many bytes were evacuated? we will fill them with sentinel
4901 bytes, except for the part holding the new offset of course. */
4904 evacn += (old_delta < 0x100 ? 1 : 1 + sizeof(STRLEN));
4906 assert(evacn <= delta + old_delta);
4912 if (delta < 0x100) {
4916 p -= sizeof(STRLEN);
4917 Copy((U8*)&delta, p, sizeof(STRLEN), U8);
4921 /* Fill the preceding buffer with sentinals to verify that no-one is
4931 =for apidoc sv_catpvn
4933 Concatenates the string onto the end of the string which is in the SV. The
4934 C<len> indicates number of bytes to copy. If the SV has the UTF-8
4935 status set, then the bytes appended should be valid UTF-8.
4936 Handles 'get' magic, but not 'set' magic. See C<sv_catpvn_mg>.
4938 =for apidoc sv_catpvn_flags
4940 Concatenates the string onto the end of the string which is in the SV. The
4941 C<len> indicates number of bytes to copy. If the SV has the UTF-8
4942 status set, then the bytes appended should be valid UTF-8.
4943 If C<flags> has the C<SV_SMAGIC> bit set, will
4944 C<mg_set> on C<dsv> afterwards if appropriate.
4945 C<sv_catpvn> and C<sv_catpvn_nomg> are implemented
4946 in terms of this function.
4952 Perl_sv_catpvn_flags(pTHX_ register SV *const dsv, register const char *sstr, register const STRLEN slen, const I32 flags)
4956 const char * const dstr = SvPV_force_flags(dsv, dlen, flags);
4958 PERL_ARGS_ASSERT_SV_CATPVN_FLAGS;
4959 assert((flags & (SV_CATBYTES|SV_CATUTF8)) != (SV_CATBYTES|SV_CATUTF8));
4961 if (!(flags & SV_CATBYTES) || !SvUTF8(dsv)) {
4962 if (flags & SV_CATUTF8 && !SvUTF8(dsv)) {
4963 sv_utf8_upgrade_flags_grow(dsv, 0, slen + 1);
4966 else SvGROW(dsv, dlen + slen + 1);
4968 sstr = SvPVX_const(dsv);
4969 Move(sstr, SvPVX(dsv) + dlen, slen, char);
4970 SvCUR_set(dsv, SvCUR(dsv) + slen);
4973 /* We inline bytes_to_utf8, to avoid an extra malloc. */
4974 const char * const send = sstr + slen;
4977 /* Something this code does not account for, which I think is
4978 impossible; it would require the same pv to be treated as
4979 bytes *and* utf8, which would indicate a bug elsewhere. */
4980 assert(sstr != dstr);
4982 SvGROW(dsv, dlen + slen * 2 + 1);
4983 d = (U8 *)SvPVX(dsv) + dlen;
4985 while (sstr < send) {
4986 const UV uv = NATIVE_TO_ASCII((U8)*sstr++);
4987 if (UNI_IS_INVARIANT(uv))
4988 *d++ = (U8)UTF_TO_NATIVE(uv);
4990 *d++ = (U8)UTF8_EIGHT_BIT_HI(uv);
4991 *d++ = (U8)UTF8_EIGHT_BIT_LO(uv);
4994 SvCUR_set(dsv, d-(const U8 *)SvPVX(dsv));
4997 (void)SvPOK_only_UTF8(dsv); /* validate pointer */
4999 if (flags & SV_SMAGIC)
5004 =for apidoc sv_catsv
5006 Concatenates the string from SV C<ssv> onto the end of the string in SV
5007 C<dsv>. If C<ssv> is null, does nothing; otherwise modifies only C<dsv>.
5008 Handles 'get' magic on both SVs, but no 'set' magic. See C<sv_catsv_mg> and
5011 =for apidoc sv_catsv_flags
5013 Concatenates the string from SV C<ssv> onto the end of the string in SV
5014 C<dsv>. If C<ssv> is null, does nothing; otherwise modifies only C<dsv>.
5015 If C<flags> include C<SV_GMAGIC> bit set, will call C<mg_get> on both SVs if
5016 appropriate. If C<flags> include C<SV_SMAGIC>, C<mg_set> will be called on
5017 the modified SV afterward, if appropriate. C<sv_catsv>, C<sv_catsv_nomg>,
5018 and C<sv_catsv_mg> are implemented in terms of this function.
5023 Perl_sv_catsv_flags(pTHX_ SV *const dsv, register SV *const ssv, const I32 flags)
5027 PERL_ARGS_ASSERT_SV_CATSV_FLAGS;
5031 const char *spv = SvPV_flags_const(ssv, slen, flags);
5033 if (flags & SV_GMAGIC)
5035 sv_catpvn_flags(dsv, spv, slen,
5036 DO_UTF8(ssv) ? SV_CATUTF8 : SV_CATBYTES);
5037 if (flags & SV_SMAGIC)
5044 =for apidoc sv_catpv
5046 Concatenates the string onto the end of the string which is in the SV.
5047 If the SV has the UTF-8 status set, then the bytes appended should be
5048 valid UTF-8. Handles 'get' magic, but not 'set' magic. See C<sv_catpv_mg>.
5053 Perl_sv_catpv(pTHX_ register SV *const sv, register const char *ptr)
5056 register STRLEN len;
5060 PERL_ARGS_ASSERT_SV_CATPV;
5064 junk = SvPV_force(sv, tlen);
5066 SvGROW(sv, tlen + len + 1);
5068 ptr = SvPVX_const(sv);
5069 Move(ptr,SvPVX(sv)+tlen,len+1,char);
5070 SvCUR_set(sv, SvCUR(sv) + len);
5071 (void)SvPOK_only_UTF8(sv); /* validate pointer */
5076 =for apidoc sv_catpv_flags
5078 Concatenates the string onto the end of the string which is in the SV.
5079 If the SV has the UTF-8 status set, then the bytes appended should
5080 be valid UTF-8. If C<flags> has the C<SV_SMAGIC> bit set, will C<mg_set>
5081 on the modified SV if appropriate.
5087 Perl_sv_catpv_flags(pTHX_ SV *dstr, const char *sstr, const I32 flags)
5089 PERL_ARGS_ASSERT_SV_CATPV_FLAGS;
5090 sv_catpvn_flags(dstr, sstr, strlen(sstr), flags);
5094 =for apidoc sv_catpv_mg
5096 Like C<sv_catpv>, but also handles 'set' magic.
5102 Perl_sv_catpv_mg(pTHX_ register SV *const sv, register const char *const ptr)
5104 PERL_ARGS_ASSERT_SV_CATPV_MG;
5113 Creates a new SV. A non-zero C<len> parameter indicates the number of
5114 bytes of preallocated string space the SV should have. An extra byte for a
5115 trailing NUL is also reserved. (SvPOK is not set for the SV even if string
5116 space is allocated.) The reference count for the new SV is set to 1.
5118 In 5.9.3, newSV() replaces the older NEWSV() API, and drops the first
5119 parameter, I<x>, a debug aid which allowed callers to identify themselves.
5120 This aid has been superseded by a new build option, PERL_MEM_LOG (see
5121 L<perlhacktips/PERL_MEM_LOG>). The older API is still there for use in XS
5122 modules supporting older perls.
5128 Perl_newSV(pTHX_ const STRLEN len)
5135 sv_upgrade(sv, SVt_PV);
5136 SvGROW(sv, len + 1);
5141 =for apidoc sv_magicext
5143 Adds magic to an SV, upgrading it if necessary. Applies the
5144 supplied vtable and returns a pointer to the magic added.
5146 Note that C<sv_magicext> will allow things that C<sv_magic> will not.
5147 In particular, you can add magic to SvREADONLY SVs, and add more than
5148 one instance of the same 'how'.
5150 If C<namlen> is greater than zero then a C<savepvn> I<copy> of C<name> is
5151 stored, if C<namlen> is zero then C<name> is stored as-is and - as another
5152 special case - if C<(name && namlen == HEf_SVKEY)> then C<name> is assumed
5153 to contain an C<SV*> and is stored as-is with its REFCNT incremented.
5155 (This is now used as a subroutine by C<sv_magic>.)
5160 Perl_sv_magicext(pTHX_ SV *const sv, SV *const obj, const int how,
5161 const MGVTBL *const vtable, const char *const name, const I32 namlen)
5166 PERL_ARGS_ASSERT_SV_MAGICEXT;
5168 SvUPGRADE(sv, SVt_PVMG);
5169 Newxz(mg, 1, MAGIC);
5170 mg->mg_moremagic = SvMAGIC(sv);
5171 SvMAGIC_set(sv, mg);
5173 /* Sometimes a magic contains a reference loop, where the sv and
5174 object refer to each other. To prevent a reference loop that
5175 would prevent such objects being freed, we look for such loops
5176 and if we find one we avoid incrementing the object refcount.
5178 Note we cannot do this to avoid self-tie loops as intervening RV must
5179 have its REFCNT incremented to keep it in existence.
5182 if (!obj || obj == sv ||
5183 how == PERL_MAGIC_arylen ||
5184 how == PERL_MAGIC_symtab ||
5185 (SvTYPE(obj) == SVt_PVGV &&
5186 (GvSV(obj) == sv || GvHV(obj) == (const HV *)sv
5187 || GvAV(obj) == (const AV *)sv || GvCV(obj) == (const CV *)sv
5188 || GvIOp(obj) == (const IO *)sv || GvFORM(obj) == (const CV *)sv)))
5193 mg->mg_obj = SvREFCNT_inc_simple(obj);
5194 mg->mg_flags |= MGf_REFCOUNTED;
5197 /* Normal self-ties simply pass a null object, and instead of
5198 using mg_obj directly, use the SvTIED_obj macro to produce a
5199 new RV as needed. For glob "self-ties", we are tieing the PVIO
5200 with an RV obj pointing to the glob containing the PVIO. In
5201 this case, to avoid a reference loop, we need to weaken the
5205 if (how == PERL_MAGIC_tiedscalar && SvTYPE(sv) == SVt_PVIO &&
5206 obj && SvROK(obj) && GvIO(SvRV(obj)) == (const IO *)sv)
5212 mg->mg_len = namlen;
5215 mg->mg_ptr = savepvn(name, namlen);
5216 else if (namlen == HEf_SVKEY) {
5217 /* Yes, this is casting away const. This is only for the case of
5218 HEf_SVKEY. I think we need to document this aberation of the
5219 constness of the API, rather than making name non-const, as
5220 that change propagating outwards a long way. */
5221 mg->mg_ptr = (char*)SvREFCNT_inc_simple_NN((SV *)name);
5223 mg->mg_ptr = (char *) name;
5225 mg->mg_virtual = (MGVTBL *) vtable;
5232 =for apidoc sv_magic
5234 Adds magic to an SV. First upgrades C<sv> to type C<SVt_PVMG> if
5235 necessary, then adds a new magic item of type C<how> to the head of the
5238 See C<sv_magicext> (which C<sv_magic> now calls) for a description of the
5239 handling of the C<name> and C<namlen> arguments.
5241 You need to use C<sv_magicext> to add magic to SvREADONLY SVs and also
5242 to add more than one instance of the same 'how'.
5248 Perl_sv_magic(pTHX_ register SV *const sv, SV *const obj, const int how,
5249 const char *const name, const I32 namlen)
5252 const MGVTBL *vtable;
5255 unsigned int vtable_index;
5257 PERL_ARGS_ASSERT_SV_MAGIC;
5259 if (how < 0 || (unsigned)how > C_ARRAY_LENGTH(PL_magic_data)
5260 || ((flags = PL_magic_data[how]),
5261 (vtable_index = flags & PERL_MAGIC_VTABLE_MASK)
5262 > magic_vtable_max))
5263 Perl_croak(aTHX_ "Don't know how to handle magic of type \\%o", how);
5265 /* PERL_MAGIC_ext is reserved for use by extensions not perl internals.
5266 Useful for attaching extension internal data to perl vars.
5267 Note that multiple extensions may clash if magical scalars
5268 etc holding private data from one are passed to another. */
5270 vtable = (vtable_index == magic_vtable_max)
5271 ? NULL : PL_magic_vtables + vtable_index;
5273 #ifdef PERL_OLD_COPY_ON_WRITE
5275 sv_force_normal_flags(sv, 0);
5277 if (SvREADONLY(sv)) {
5279 /* its okay to attach magic to shared strings */
5283 && !PERL_MAGIC_TYPE_READONLY_ACCEPTABLE(how)
5286 Perl_croak_no_modify(aTHX);
5289 if (SvMAGICAL(sv) || (how == PERL_MAGIC_taint && SvTYPE(sv) >= SVt_PVMG)) {
5290 if (SvMAGIC(sv) && (mg = mg_find(sv, how))) {
5291 /* sv_magic() refuses to add a magic of the same 'how' as an
5294 if (how == PERL_MAGIC_taint)
5300 /* Rest of work is done else where */
5301 mg = sv_magicext(sv,obj,how,vtable,name,namlen);
5304 case PERL_MAGIC_taint:
5307 case PERL_MAGIC_ext:
5308 case PERL_MAGIC_dbfile:
5315 S_sv_unmagicext_flags(pTHX_ SV *const sv, const int type, MGVTBL *vtbl, const U32 flags)
5322 if (SvTYPE(sv) < SVt_PVMG || !SvMAGIC(sv))
5324 mgp = &(((XPVMG*) SvANY(sv))->xmg_u.xmg_magic);
5325 for (mg = *mgp; mg; mg = *mgp) {
5326 const MGVTBL* const virt = mg->mg_virtual;
5327 if (mg->mg_type == type && (!flags || virt == vtbl)) {
5328 *mgp = mg->mg_moremagic;
5329 if (virt && virt->svt_free)
5330 virt->svt_free(aTHX_ sv, mg);
5331 if (mg->mg_ptr && mg->mg_type != PERL_MAGIC_regex_global) {
5333 Safefree(mg->mg_ptr);
5334 else if (mg->mg_len == HEf_SVKEY)
5335 SvREFCNT_dec(MUTABLE_SV(mg->mg_ptr));
5336 else if (mg->mg_type == PERL_MAGIC_utf8)
5337 Safefree(mg->mg_ptr);
5339 if (mg->mg_flags & MGf_REFCOUNTED)
5340 SvREFCNT_dec(mg->mg_obj);
5344 mgp = &mg->mg_moremagic;
5347 if (SvMAGICAL(sv)) /* if we're under save_magic, wait for restore_magic; */
5348 mg_magical(sv); /* else fix the flags now */
5352 SvFLAGS(sv) |= (SvFLAGS(sv) & (SVp_IOK|SVp_NOK|SVp_POK)) >> PRIVSHIFT;
5358 =for apidoc sv_unmagic
5360 Removes all magic of type C<type> from an SV.
5366 Perl_sv_unmagic(pTHX_ SV *const sv, const int type)
5368 PERL_ARGS_ASSERT_SV_UNMAGIC;
5369 return S_sv_unmagicext_flags(aTHX_ sv, type, NULL, 0);
5373 =for apidoc sv_unmagicext
5375 Removes all magic of type C<type> with the specified C<vtbl> from an SV.
5381 Perl_sv_unmagicext(pTHX_ SV *const sv, const int type, MGVTBL *vtbl)
5383 PERL_ARGS_ASSERT_SV_UNMAGICEXT;
5384 return S_sv_unmagicext_flags(aTHX_ sv, type, vtbl, 1);
5388 =for apidoc sv_rvweaken
5390 Weaken a reference: set the C<SvWEAKREF> flag on this RV; give the
5391 referred-to SV C<PERL_MAGIC_backref> magic if it hasn't already; and
5392 push a back-reference to this RV onto the array of backreferences
5393 associated with that magic. If the RV is magical, set magic will be
5394 called after the RV is cleared.
5400 Perl_sv_rvweaken(pTHX_ SV *const sv)
5404 PERL_ARGS_ASSERT_SV_RVWEAKEN;
5406 if (!SvOK(sv)) /* let undefs pass */
5409 Perl_croak(aTHX_ "Can't weaken a nonreference");
5410 else if (SvWEAKREF(sv)) {
5411 Perl_ck_warner(aTHX_ packWARN(WARN_MISC), "Reference is already weak");
5414 else if (SvREADONLY(sv)) croak_no_modify();
5416 Perl_sv_add_backref(aTHX_ tsv, sv);
5422 /* Give tsv backref magic if it hasn't already got it, then push a
5423 * back-reference to sv onto the array associated with the backref magic.
5425 * As an optimisation, if there's only one backref and it's not an AV,
5426 * store it directly in the HvAUX or mg_obj slot, avoiding the need to
5427 * allocate an AV. (Whether the slot holds an AV tells us whether this is
5431 /* A discussion about the backreferences array and its refcount:
5433 * The AV holding the backreferences is pointed to either as the mg_obj of
5434 * PERL_MAGIC_backref, or in the specific case of a HV, from the
5435 * xhv_backreferences field. The array is created with a refcount
5436 * of 2. This means that if during global destruction the array gets
5437 * picked on before its parent to have its refcount decremented by the
5438 * random zapper, it won't actually be freed, meaning it's still there for
5439 * when its parent gets freed.
5441 * When the parent SV is freed, the extra ref is killed by
5442 * Perl_sv_kill_backrefs. The other ref is killed, in the case of magic,
5443 * by mg_free() / MGf_REFCOUNTED, or for a hash, by Perl_hv_kill_backrefs.
5445 * When a single backref SV is stored directly, it is not reference
5450 Perl_sv_add_backref(pTHX_ SV *const tsv, SV *const sv)
5457 PERL_ARGS_ASSERT_SV_ADD_BACKREF;
5459 /* find slot to store array or singleton backref */
5461 if (SvTYPE(tsv) == SVt_PVHV) {
5462 svp = (SV**)Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5465 (SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL))))
5467 sv_magic(tsv, NULL, PERL_MAGIC_backref, NULL, 0);
5468 mg = mg_find(tsv, PERL_MAGIC_backref);
5470 svp = &(mg->mg_obj);
5473 /* create or retrieve the array */
5475 if ( (!*svp && SvTYPE(sv) == SVt_PVAV)
5476 || (*svp && SvTYPE(*svp) != SVt_PVAV)
5481 SvREFCNT_inc_simple_void(av);
5482 /* av now has a refcnt of 2; see discussion above */
5484 /* move single existing backref to the array */
5486 AvARRAY(av)[++AvFILLp(av)] = *svp; /* av_push() */
5490 mg->mg_flags |= MGf_REFCOUNTED;
5493 av = MUTABLE_AV(*svp);
5496 /* optimisation: store single backref directly in HvAUX or mg_obj */
5500 /* push new backref */
5501 assert(SvTYPE(av) == SVt_PVAV);
5502 if (AvFILLp(av) >= AvMAX(av)) {
5503 av_extend(av, AvFILLp(av)+1);
5505 AvARRAY(av)[++AvFILLp(av)] = sv; /* av_push() */
5508 /* delete a back-reference to ourselves from the backref magic associated
5509 * with the SV we point to.
5513 Perl_sv_del_backref(pTHX_ SV *const tsv, SV *const sv)
5518 PERL_ARGS_ASSERT_SV_DEL_BACKREF;
5520 if (SvTYPE(tsv) == SVt_PVHV) {
5522 svp = (SV**)Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5524 else if (SvIS_FREED(tsv) && PL_phase == PERL_PHASE_DESTRUCT) {
5525 /* It's possible for the the last (strong) reference to tsv to have
5526 become freed *before* the last thing holding a weak reference.
5527 If both survive longer than the backreferences array, then when
5528 the referent's reference count drops to 0 and it is freed, it's
5529 not able to chase the backreferences, so they aren't NULLed.
5531 For example, a CV holds a weak reference to its stash. If both the
5532 CV and the stash survive longer than the backreferences array,
5533 and the CV gets picked for the SvBREAK() treatment first,
5534 *and* it turns out that the stash is only being kept alive because
5535 of an our variable in the pad of the CV, then midway during CV
5536 destruction the stash gets freed, but CvSTASH() isn't set to NULL.
5537 It ends up pointing to the freed HV. Hence it's chased in here, and
5538 if this block wasn't here, it would hit the !svp panic just below.
5540 I don't believe that "better" destruction ordering is going to help
5541 here - during global destruction there's always going to be the
5542 chance that something goes out of order. We've tried to make it
5543 foolproof before, and it only resulted in evolutionary pressure on
5544 fools. Which made us look foolish for our hubris. :-(
5550 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
5551 svp = mg ? &(mg->mg_obj) : NULL;
5555 Perl_croak(aTHX_ "panic: del_backref, svp=0");
5557 /* It's possible that sv is being freed recursively part way through the
5558 freeing of tsv. If this happens, the backreferences array of tsv has
5559 already been freed, and so svp will be NULL. If this is the case,
5560 we should not panic. Instead, nothing needs doing, so return. */
5561 if (PL_phase == PERL_PHASE_DESTRUCT && SvREFCNT(tsv) == 0)
5563 Perl_croak(aTHX_ "panic: del_backref, *svp=%p phase=%s refcnt=%" UVuf,
5564 *svp, PL_phase_names[PL_phase], (UV)SvREFCNT(tsv));
5567 if (SvTYPE(*svp) == SVt_PVAV) {
5571 AV * const av = (AV*)*svp;
5573 assert(!SvIS_FREED(av));
5577 /* for an SV with N weak references to it, if all those
5578 * weak refs are deleted, then sv_del_backref will be called
5579 * N times and O(N^2) compares will be done within the backref
5580 * array. To ameliorate this potential slowness, we:
5581 * 1) make sure this code is as tight as possible;
5582 * 2) when looking for SV, look for it at both the head and tail of the
5583 * array first before searching the rest, since some create/destroy
5584 * patterns will cause the backrefs to be freed in order.
5591 SV **p = &svp[fill];
5592 SV *const topsv = *p;
5599 /* We weren't the last entry.
5600 An unordered list has this property that you
5601 can take the last element off the end to fill
5602 the hole, and it's still an unordered list :-)
5608 break; /* should only be one */
5615 AvFILLp(av) = fill-1;
5617 else if (SvIS_FREED(*svp) && PL_phase == PERL_PHASE_DESTRUCT) {
5618 /* freed AV; skip */
5621 /* optimisation: only a single backref, stored directly */
5623 Perl_croak(aTHX_ "panic: del_backref, *svp=%p, sv=%p", *svp, sv);
5630 Perl_sv_kill_backrefs(pTHX_ SV *const sv, AV *const av)
5636 PERL_ARGS_ASSERT_SV_KILL_BACKREFS;
5641 /* after multiple passes through Perl_sv_clean_all() for a thingy
5642 * that has badly leaked, the backref array may have gotten freed,
5643 * since we only protect it against 1 round of cleanup */
5644 if (SvIS_FREED(av)) {
5645 if (PL_in_clean_all) /* All is fair */
5648 "panic: magic_killbackrefs (freed backref AV/SV)");
5652 is_array = (SvTYPE(av) == SVt_PVAV);
5654 assert(!SvIS_FREED(av));
5657 last = svp + AvFILLp(av);
5660 /* optimisation: only a single backref, stored directly */
5666 while (svp <= last) {
5668 SV *const referrer = *svp;
5669 if (SvWEAKREF(referrer)) {
5670 /* XXX Should we check that it hasn't changed? */
5671 assert(SvROK(referrer));
5672 SvRV_set(referrer, 0);
5674 SvWEAKREF_off(referrer);
5675 SvSETMAGIC(referrer);
5676 } else if (SvTYPE(referrer) == SVt_PVGV ||
5677 SvTYPE(referrer) == SVt_PVLV) {
5678 assert(SvTYPE(sv) == SVt_PVHV); /* stash backref */
5679 /* You lookin' at me? */
5680 assert(GvSTASH(referrer));
5681 assert(GvSTASH(referrer) == (const HV *)sv);
5682 GvSTASH(referrer) = 0;
5683 } else if (SvTYPE(referrer) == SVt_PVCV ||
5684 SvTYPE(referrer) == SVt_PVFM) {
5685 if (SvTYPE(sv) == SVt_PVHV) { /* stash backref */
5686 /* You lookin' at me? */
5687 assert(CvSTASH(referrer));
5688 assert(CvSTASH(referrer) == (const HV *)sv);
5689 SvANY(MUTABLE_CV(referrer))->xcv_stash = 0;
5692 assert(SvTYPE(sv) == SVt_PVGV);
5693 /* You lookin' at me? */
5694 assert(CvGV(referrer));
5695 assert(CvGV(referrer) == (const GV *)sv);
5696 anonymise_cv_maybe(MUTABLE_GV(sv),
5697 MUTABLE_CV(referrer));
5702 "panic: magic_killbackrefs (flags=%"UVxf")",
5703 (UV)SvFLAGS(referrer));
5714 SvREFCNT_dec(av); /* remove extra count added by sv_add_backref() */
5720 =for apidoc sv_insert
5722 Inserts a string at the specified offset/length within the SV. Similar to
5723 the Perl substr() function. Handles get magic.
5725 =for apidoc sv_insert_flags
5727 Same as C<sv_insert>, but the extra C<flags> are passed to the
5728 C<SvPV_force_flags> that applies to C<bigstr>.
5734 Perl_sv_insert_flags(pTHX_ SV *const bigstr, const STRLEN offset, const STRLEN len, const char *const little, const STRLEN littlelen, const U32 flags)
5739 register char *midend;
5740 register char *bigend;
5741 register SSize_t i; /* better be sizeof(STRLEN) or bad things happen */
5744 PERL_ARGS_ASSERT_SV_INSERT_FLAGS;
5747 Perl_croak(aTHX_ "Can't modify nonexistent substring");
5748 SvPV_force_flags(bigstr, curlen, flags);
5749 (void)SvPOK_only_UTF8(bigstr);
5750 if (offset + len > curlen) {
5751 SvGROW(bigstr, offset+len+1);
5752 Zero(SvPVX(bigstr)+curlen, offset+len-curlen, char);
5753 SvCUR_set(bigstr, offset+len);
5757 i = littlelen - len;
5758 if (i > 0) { /* string might grow */
5759 big = SvGROW(bigstr, SvCUR(bigstr) + i + 1);
5760 mid = big + offset + len;
5761 midend = bigend = big + SvCUR(bigstr);
5764 while (midend > mid) /* shove everything down */
5765 *--bigend = *--midend;
5766 Move(little,big+offset,littlelen,char);
5767 SvCUR_set(bigstr, SvCUR(bigstr) + i);
5772 Move(little,SvPVX(bigstr)+offset,len,char);
5777 big = SvPVX(bigstr);
5780 bigend = big + SvCUR(bigstr);
5782 if (midend > bigend)
5783 Perl_croak(aTHX_ "panic: sv_insert, midend=%p, bigend=%p",
5786 if (mid - big > bigend - midend) { /* faster to shorten from end */
5788 Move(little, mid, littlelen,char);
5791 i = bigend - midend;
5793 Move(midend, mid, i,char);
5797 SvCUR_set(bigstr, mid - big);
5799 else if ((i = mid - big)) { /* faster from front */
5800 midend -= littlelen;
5802 Move(big, midend - i, i, char);
5803 sv_chop(bigstr,midend-i);
5805 Move(little, mid, littlelen,char);
5807 else if (littlelen) {
5808 midend -= littlelen;
5809 sv_chop(bigstr,midend);
5810 Move(little,midend,littlelen,char);
5813 sv_chop(bigstr,midend);
5819 =for apidoc sv_replace
5821 Make the first argument a copy of the second, then delete the original.
5822 The target SV physically takes over ownership of the body of the source SV
5823 and inherits its flags; however, the target keeps any magic it owns,
5824 and any magic in the source is discarded.
5825 Note that this is a rather specialist SV copying operation; most of the
5826 time you'll want to use C<sv_setsv> or one of its many macro front-ends.
5832 Perl_sv_replace(pTHX_ register SV *const sv, register SV *const nsv)
5835 const U32 refcnt = SvREFCNT(sv);
5837 PERL_ARGS_ASSERT_SV_REPLACE;
5839 SV_CHECK_THINKFIRST_COW_DROP(sv);
5840 if (SvREFCNT(nsv) != 1) {
5841 Perl_croak(aTHX_ "panic: reference miscount on nsv in sv_replace()"
5842 " (%" UVuf " != 1)", (UV) SvREFCNT(nsv));
5844 if (SvMAGICAL(sv)) {
5848 sv_upgrade(nsv, SVt_PVMG);
5849 SvMAGIC_set(nsv, SvMAGIC(sv));
5850 SvFLAGS(nsv) |= SvMAGICAL(sv);
5852 SvMAGIC_set(sv, NULL);
5856 assert(!SvREFCNT(sv));
5857 #ifdef DEBUG_LEAKING_SCALARS
5858 sv->sv_flags = nsv->sv_flags;
5859 sv->sv_any = nsv->sv_any;
5860 sv->sv_refcnt = nsv->sv_refcnt;
5861 sv->sv_u = nsv->sv_u;
5863 StructCopy(nsv,sv,SV);
5865 if(SvTYPE(sv) == SVt_IV) {
5867 = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
5871 #ifdef PERL_OLD_COPY_ON_WRITE
5872 if (SvIsCOW_normal(nsv)) {
5873 /* We need to follow the pointers around the loop to make the
5874 previous SV point to sv, rather than nsv. */
5877 while ((next = SV_COW_NEXT_SV(current)) != nsv) {
5880 assert(SvPVX_const(current) == SvPVX_const(nsv));
5882 /* Make the SV before us point to the SV after us. */
5884 PerlIO_printf(Perl_debug_log, "previous is\n");
5886 PerlIO_printf(Perl_debug_log,
5887 "move it from 0x%"UVxf" to 0x%"UVxf"\n",
5888 (UV) SV_COW_NEXT_SV(current), (UV) sv);
5890 SV_COW_NEXT_SV_SET(current, sv);
5893 SvREFCNT(sv) = refcnt;
5894 SvFLAGS(nsv) |= SVTYPEMASK; /* Mark as freed */
5899 /* We're about to free a GV which has a CV that refers back to us.
5900 * If that CV will outlive us, make it anonymous (i.e. fix up its CvGV
5904 S_anonymise_cv_maybe(pTHX_ GV *gv, CV* cv)
5909 PERL_ARGS_ASSERT_ANONYMISE_CV_MAYBE;
5912 assert(SvREFCNT(gv) == 0);
5913 assert(isGV(gv) && isGV_with_GP(gv));
5915 assert(!CvANON(cv));
5916 assert(CvGV(cv) == gv);
5918 /* will the CV shortly be freed by gp_free() ? */
5919 if (GvCV(gv) == cv && GvGP(gv)->gp_refcnt < 2 && SvREFCNT(cv) < 2) {
5920 SvANY(cv)->xcv_gv = NULL;
5924 /* if not, anonymise: */
5925 gvname = (GvSTASH(gv) && HvNAME(GvSTASH(gv)) && HvENAME(GvSTASH(gv)))
5926 ? newSVhek(HvENAME_HEK(GvSTASH(gv)))
5927 : newSVpvn_flags( "__ANON__", 8, 0 );
5928 sv_catpvs(gvname, "::__ANON__");
5929 anongv = gv_fetchsv(gvname, GV_ADDMULTI, SVt_PVCV);
5930 SvREFCNT_dec(gvname);
5934 SvANY(cv)->xcv_gv = MUTABLE_GV(SvREFCNT_inc(anongv));
5939 =for apidoc sv_clear
5941 Clear an SV: call any destructors, free up any memory used by the body,
5942 and free the body itself. The SV's head is I<not> freed, although
5943 its type is set to all 1's so that it won't inadvertently be assumed
5944 to be live during global destruction etc.
5945 This function should only be called when REFCNT is zero. Most of the time
5946 you'll want to call C<sv_free()> (or its macro wrapper C<SvREFCNT_dec>)
5953 Perl_sv_clear(pTHX_ SV *const orig_sv)
5958 const struct body_details *sv_type_details;
5961 register SV *sv = orig_sv;
5964 PERL_ARGS_ASSERT_SV_CLEAR;
5966 /* within this loop, sv is the SV currently being freed, and
5967 * iter_sv is the most recent AV or whatever that's being iterated
5968 * over to provide more SVs */
5974 assert(SvREFCNT(sv) == 0);
5975 assert(SvTYPE(sv) != (svtype)SVTYPEMASK);
5977 if (type <= SVt_IV) {
5978 /* See the comment in sv.h about the collusion between this
5979 * early return and the overloading of the NULL slots in the
5983 SvFLAGS(sv) &= SVf_BREAK;
5984 SvFLAGS(sv) |= SVTYPEMASK;
5988 assert(!SvOBJECT(sv) || type >= SVt_PVMG); /* objs are always >= MG */
5990 if (type >= SVt_PVMG) {
5992 if (!curse(sv, 1)) goto get_next_sv;
5993 type = SvTYPE(sv); /* destructor may have changed it */
5995 /* Free back-references before magic, in case the magic calls
5996 * Perl code that has weak references to sv. */
5997 if (type == SVt_PVHV) {
5998 Perl_hv_kill_backrefs(aTHX_ MUTABLE_HV(sv));
6002 else if (type == SVt_PVMG && SvPAD_OUR(sv)) {
6003 SvREFCNT_dec(SvOURSTASH(sv));
6004 } else if (SvMAGIC(sv)) {
6005 /* Free back-references before other types of magic. */
6006 sv_unmagic(sv, PERL_MAGIC_backref);
6010 if (type == SVt_PVMG && SvPAD_TYPED(sv))
6011 SvREFCNT_dec(SvSTASH(sv));
6014 /* case SVt_BIND: */
6017 IoIFP(sv) != PerlIO_stdin() &&
6018 IoIFP(sv) != PerlIO_stdout() &&
6019 IoIFP(sv) != PerlIO_stderr() &&
6020 !(IoFLAGS(sv) & IOf_FAKE_DIRP))
6022 io_close(MUTABLE_IO(sv), FALSE);
6024 if (IoDIRP(sv) && !(IoFLAGS(sv) & IOf_FAKE_DIRP))
6025 PerlDir_close(IoDIRP(sv));
6026 IoDIRP(sv) = (DIR*)NULL;
6027 Safefree(IoTOP_NAME(sv));
6028 Safefree(IoFMT_NAME(sv));
6029 Safefree(IoBOTTOM_NAME(sv));
6030 if ((const GV *)sv == PL_statgv)
6034 /* FIXME for plugins */
6035 pregfree2((REGEXP*) sv);
6039 cv_undef(MUTABLE_CV(sv));
6040 /* If we're in a stash, we don't own a reference to it.
6041 * However it does have a back reference to us, which needs to
6043 if ((stash = CvSTASH(sv)))
6044 sv_del_backref(MUTABLE_SV(stash), sv);
6047 if (PL_last_swash_hv == (const HV *)sv) {
6048 PL_last_swash_hv = NULL;
6050 if (HvTOTALKEYS((HV*)sv) > 0) {
6052 /* this statement should match the one at the beginning of
6053 * hv_undef_flags() */
6054 if ( PL_phase != PERL_PHASE_DESTRUCT
6055 && (name = HvNAME((HV*)sv)))
6058 (void)hv_delete(PL_stashcache, name,
6059 HvNAMEUTF8((HV*)sv) ? -HvNAMELEN_get((HV*)sv) : HvNAMELEN_get((HV*)sv), G_DISCARD);
6060 hv_name_set((HV*)sv, NULL, 0, 0);
6063 /* save old iter_sv in unused SvSTASH field */
6064 assert(!SvOBJECT(sv));
6065 SvSTASH(sv) = (HV*)iter_sv;
6068 /* save old hash_index in unused SvMAGIC field */
6069 assert(!SvMAGICAL(sv));
6070 assert(!SvMAGIC(sv));
6071 ((XPVMG*) SvANY(sv))->xmg_u.xmg_hash_index = hash_index;
6074 next_sv = Perl_hfree_next_entry(aTHX_ (HV*)sv, &hash_index);
6075 goto get_next_sv; /* process this new sv */
6077 /* free empty hash */
6078 Perl_hv_undef_flags(aTHX_ MUTABLE_HV(sv), HV_NAME_SETALL);
6079 assert(!HvARRAY((HV*)sv));
6083 AV* av = MUTABLE_AV(sv);
6084 if (PL_comppad == av) {
6088 if (AvREAL(av) && AvFILLp(av) > -1) {
6089 next_sv = AvARRAY(av)[AvFILLp(av)--];
6090 /* save old iter_sv in top-most slot of AV,
6091 * and pray that it doesn't get wiped in the meantime */
6092 AvARRAY(av)[AvMAX(av)] = iter_sv;
6094 goto get_next_sv; /* process this new sv */
6096 Safefree(AvALLOC(av));
6101 if (LvTYPE(sv) == 'T') { /* for tie: return HE to pool */
6102 SvREFCNT_dec(HeKEY_sv((HE*)LvTARG(sv)));
6103 HeNEXT((HE*)LvTARG(sv)) = PL_hv_fetch_ent_mh;
6104 PL_hv_fetch_ent_mh = (HE*)LvTARG(sv);
6106 else if (LvTYPE(sv) != 't') /* unless tie: unrefcnted fake SV** */
6107 SvREFCNT_dec(LvTARG(sv));
6109 if (isGV_with_GP(sv)) {
6110 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
6111 && HvENAME_get(stash))
6112 mro_method_changed_in(stash);
6113 gp_free(MUTABLE_GV(sv));
6115 unshare_hek(GvNAME_HEK(sv));
6116 /* If we're in a stash, we don't own a reference to it.
6117 * However it does have a back reference to us, which
6118 * needs to be cleared. */
6119 if (!SvVALID(sv) && (stash = GvSTASH(sv)))
6120 sv_del_backref(MUTABLE_SV(stash), sv);
6122 /* FIXME. There are probably more unreferenced pointers to SVs
6123 * in the interpreter struct that we should check and tidy in
6124 * a similar fashion to this: */
6125 /* See also S_sv_unglob, which does the same thing. */
6126 if ((const GV *)sv == PL_last_in_gv)
6127 PL_last_in_gv = NULL;
6128 else if ((const GV *)sv == PL_statgv)
6135 /* Don't bother with SvOOK_off(sv); as we're only going to
6139 SvOOK_offset(sv, offset);
6140 SvPV_set(sv, SvPVX_mutable(sv) - offset);
6141 /* Don't even bother with turning off the OOK flag. */
6146 SV * const target = SvRV(sv);
6148 sv_del_backref(target, sv);
6153 #ifdef PERL_OLD_COPY_ON_WRITE
6154 else if (SvPVX_const(sv)
6155 && !(SvTYPE(sv) == SVt_PVIO
6156 && !(IoFLAGS(sv) & IOf_FAKE_DIRP)))
6160 PerlIO_printf(Perl_debug_log, "Copy on write: clear\n");
6164 sv_release_COW(sv, SvPVX_const(sv), SV_COW_NEXT_SV(sv));
6166 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
6170 } else if (SvLEN(sv)) {
6171 Safefree(SvPVX_const(sv));
6175 else if (SvPVX_const(sv) && SvLEN(sv)
6176 && !(SvTYPE(sv) == SVt_PVIO
6177 && !(IoFLAGS(sv) & IOf_FAKE_DIRP)))
6178 Safefree(SvPVX_mutable(sv));
6179 else if (SvPVX_const(sv) && SvIsCOW(sv)) {
6180 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
6191 SvFLAGS(sv) &= SVf_BREAK;
6192 SvFLAGS(sv) |= SVTYPEMASK;
6194 sv_type_details = bodies_by_type + type;
6195 if (sv_type_details->arena) {
6196 del_body(((char *)SvANY(sv) + sv_type_details->offset),
6197 &PL_body_roots[type]);
6199 else if (sv_type_details->body_size) {
6200 safefree(SvANY(sv));
6204 /* caller is responsible for freeing the head of the original sv */
6205 if (sv != orig_sv && !SvREFCNT(sv))
6208 /* grab and free next sv, if any */
6216 else if (!iter_sv) {
6218 } else if (SvTYPE(iter_sv) == SVt_PVAV) {
6219 AV *const av = (AV*)iter_sv;
6220 if (AvFILLp(av) > -1) {
6221 sv = AvARRAY(av)[AvFILLp(av)--];
6223 else { /* no more elements of current AV to free */
6226 /* restore previous value, squirrelled away */
6227 iter_sv = AvARRAY(av)[AvMAX(av)];
6228 Safefree(AvALLOC(av));
6231 } else if (SvTYPE(iter_sv) == SVt_PVHV) {
6232 sv = Perl_hfree_next_entry(aTHX_ (HV*)iter_sv, &hash_index);
6233 if (!sv && !HvTOTALKEYS((HV *)iter_sv)) {
6234 /* no more elements of current HV to free */
6237 /* Restore previous values of iter_sv and hash_index,
6238 * squirrelled away */
6239 assert(!SvOBJECT(sv));
6240 iter_sv = (SV*)SvSTASH(sv);
6241 assert(!SvMAGICAL(sv));
6242 hash_index = ((XPVMG*) SvANY(sv))->xmg_u.xmg_hash_index;
6244 /* perl -DA does not like rubbish in SvMAGIC. */
6248 /* free any remaining detritus from the hash struct */
6249 Perl_hv_undef_flags(aTHX_ MUTABLE_HV(sv), HV_NAME_SETALL);
6250 assert(!HvARRAY((HV*)sv));
6255 /* unrolled SvREFCNT_dec and sv_free2 follows: */
6259 if (!SvREFCNT(sv)) {
6263 if (--(SvREFCNT(sv)))
6267 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEBUGGING),
6268 "Attempt to free temp prematurely: SV 0x%"UVxf
6269 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
6273 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
6274 /* make sure SvREFCNT(sv)==0 happens very seldom */
6275 SvREFCNT(sv) = (~(U32)0)/2;
6284 /* This routine curses the sv itself, not the object referenced by sv. So
6285 sv does not have to be ROK. */
6288 S_curse(pTHX_ SV * const sv, const bool check_refcnt) {
6291 PERL_ARGS_ASSERT_CURSE;
6292 assert(SvOBJECT(sv));
6294 if (PL_defstash && /* Still have a symbol table? */
6301 stash = SvSTASH(sv);
6302 destructor = StashHANDLER(stash,DESTROY);
6304 /* A constant subroutine can have no side effects, so
6305 don't bother calling it. */
6306 && !CvCONST(destructor)
6307 /* Don't bother calling an empty destructor or one that
6308 returns immediately. */
6309 && (CvISXSUB(destructor)
6310 || (CvSTART(destructor)
6311 && (CvSTART(destructor)->op_next->op_type
6313 && (CvSTART(destructor)->op_next->op_type
6315 || CvSTART(destructor)->op_next->op_next->op_type
6321 SV* const tmpref = newRV(sv);
6322 SvREADONLY_on(tmpref); /* DESTROY() could be naughty */
6324 PUSHSTACKi(PERLSI_DESTROY);
6329 call_sv(MUTABLE_SV(destructor),
6330 G_DISCARD|G_EVAL|G_KEEPERR|G_VOID);
6334 if(SvREFCNT(tmpref) < 2) {
6335 /* tmpref is not kept alive! */
6337 SvRV_set(tmpref, NULL);
6340 SvREFCNT_dec(tmpref);
6342 } while (SvOBJECT(sv) && SvSTASH(sv) != stash);
6345 if (check_refcnt && SvREFCNT(sv)) {
6346 if (PL_in_clean_objs)
6348 "DESTROY created new reference to dead object '%"HEKf"'",
6349 HEKfARG(HvNAME_HEK(stash)));
6350 /* DESTROY gave object new lease on life */
6356 SvREFCNT_dec(SvSTASH(sv)); /* possibly of changed persuasion */
6357 SvOBJECT_off(sv); /* Curse the object. */
6358 if (SvTYPE(sv) != SVt_PVIO)
6359 --PL_sv_objcount;/* XXX Might want something more general */
6365 =for apidoc sv_newref
6367 Increment an SV's reference count. Use the C<SvREFCNT_inc()> wrapper
6374 Perl_sv_newref(pTHX_ SV *const sv)
6376 PERL_UNUSED_CONTEXT;
6385 Decrement an SV's reference count, and if it drops to zero, call
6386 C<sv_clear> to invoke destructors and free up any memory used by
6387 the body; finally, deallocate the SV's head itself.
6388 Normally called via a wrapper macro C<SvREFCNT_dec>.
6394 Perl_sv_free(pTHX_ SV *const sv)
6399 if (SvREFCNT(sv) == 0) {
6400 if (SvFLAGS(sv) & SVf_BREAK)
6401 /* this SV's refcnt has been artificially decremented to
6402 * trigger cleanup */
6404 if (PL_in_clean_all) /* All is fair */
6406 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
6407 /* make sure SvREFCNT(sv)==0 happens very seldom */
6408 SvREFCNT(sv) = (~(U32)0)/2;
6411 if (ckWARN_d(WARN_INTERNAL)) {
6412 #ifdef DEBUG_LEAKING_SCALARS_FORK_DUMP
6413 Perl_dump_sv_child(aTHX_ sv);
6415 #ifdef DEBUG_LEAKING_SCALARS
6418 #ifdef DEBUG_LEAKING_SCALARS_ABORT
6419 if (PL_warnhook == PERL_WARNHOOK_FATAL
6420 || ckDEAD(packWARN(WARN_INTERNAL))) {
6421 /* Don't let Perl_warner cause us to escape our fate: */
6425 /* This may not return: */
6426 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
6427 "Attempt to free unreferenced scalar: SV 0x%"UVxf
6428 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
6431 #ifdef DEBUG_LEAKING_SCALARS_ABORT
6436 if (--(SvREFCNT(sv)) > 0)
6438 Perl_sv_free2(aTHX_ sv);
6442 Perl_sv_free2(pTHX_ SV *const sv)
6446 PERL_ARGS_ASSERT_SV_FREE2;
6450 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEBUGGING),
6451 "Attempt to free temp prematurely: SV 0x%"UVxf
6452 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
6456 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
6457 /* make sure SvREFCNT(sv)==0 happens very seldom */
6458 SvREFCNT(sv) = (~(U32)0)/2;
6469 Returns the length of the string in the SV. Handles magic and type
6470 coercion. See also C<SvCUR>, which gives raw access to the xpv_cur slot.
6476 Perl_sv_len(pTHX_ register SV *const sv)
6484 len = mg_length(sv);
6486 (void)SvPV_const(sv, len);
6491 =for apidoc sv_len_utf8
6493 Returns the number of characters in the string in an SV, counting wide
6494 UTF-8 bytes as a single character. Handles magic and type coercion.
6500 * The length is cached in PERL_MAGIC_utf8, in the mg_len field. Also the
6501 * mg_ptr is used, by sv_pos_u2b() and sv_pos_b2u() - see the comments below.
6502 * (Note that the mg_len is not the length of the mg_ptr field.
6503 * This allows the cache to store the character length of the string without
6504 * needing to malloc() extra storage to attach to the mg_ptr.)
6509 Perl_sv_len_utf8(pTHX_ register SV *const sv)
6515 return mg_length(sv);
6519 const U8 *s = (U8*)SvPV_const(sv, len);
6523 MAGIC *mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_utf8) : NULL;
6525 if (mg && (mg->mg_len != -1 || mg->mg_ptr)) {
6526 if (mg->mg_len != -1)
6529 /* We can use the offset cache for a headstart.
6530 The longer value is stored in the first pair. */
6531 STRLEN *cache = (STRLEN *) mg->mg_ptr;
6533 ulen = cache[0] + Perl_utf8_length(aTHX_ s + cache[1],
6537 if (PL_utf8cache < 0) {
6538 const STRLEN real = Perl_utf8_length(aTHX_ s, s + len);
6539 assert_uft8_cache_coherent("sv_len_utf8", ulen, real, sv);
6543 ulen = Perl_utf8_length(aTHX_ s, s + len);
6544 utf8_mg_len_cache_update(sv, &mg, ulen);
6548 return Perl_utf8_length(aTHX_ s, s + len);
6552 /* Walk forwards to find the byte corresponding to the passed in UTF-8
6555 S_sv_pos_u2b_forwards(const U8 *const start, const U8 *const send,
6556 STRLEN *const uoffset_p, bool *const at_end)
6558 const U8 *s = start;
6559 STRLEN uoffset = *uoffset_p;
6561 PERL_ARGS_ASSERT_SV_POS_U2B_FORWARDS;
6563 while (s < send && uoffset) {
6570 else if (s > send) {
6572 /* This is the existing behaviour. Possibly it should be a croak, as
6573 it's actually a bounds error */
6576 *uoffset_p -= uoffset;
6580 /* Given the length of the string in both bytes and UTF-8 characters, decide
6581 whether to walk forwards or backwards to find the byte corresponding to
6582 the passed in UTF-8 offset. */
6584 S_sv_pos_u2b_midway(const U8 *const start, const U8 *send,
6585 STRLEN uoffset, const STRLEN uend)
6587 STRLEN backw = uend - uoffset;
6589 PERL_ARGS_ASSERT_SV_POS_U2B_MIDWAY;
6591 if (uoffset < 2 * backw) {
6592 /* The assumption is that going forwards is twice the speed of going
6593 forward (that's where the 2 * backw comes from).
6594 (The real figure of course depends on the UTF-8 data.) */
6595 const U8 *s = start;
6597 while (s < send && uoffset--)
6607 while (UTF8_IS_CONTINUATION(*send))
6610 return send - start;
6613 /* For the string representation of the given scalar, find the byte
6614 corresponding to the passed in UTF-8 offset. uoffset0 and boffset0
6615 give another position in the string, *before* the sought offset, which
6616 (which is always true, as 0, 0 is a valid pair of positions), which should
6617 help reduce the amount of linear searching.
6618 If *mgp is non-NULL, it should point to the UTF-8 cache magic, which
6619 will be used to reduce the amount of linear searching. The cache will be
6620 created if necessary, and the found value offered to it for update. */
6622 S_sv_pos_u2b_cached(pTHX_ SV *const sv, MAGIC **const mgp, const U8 *const start,
6623 const U8 *const send, STRLEN uoffset,
6624 STRLEN uoffset0, STRLEN boffset0)
6626 STRLEN boffset = 0; /* Actually always set, but let's keep gcc happy. */
6628 bool at_end = FALSE;
6630 PERL_ARGS_ASSERT_SV_POS_U2B_CACHED;
6632 assert (uoffset >= uoffset0);
6639 && (*mgp || (SvTYPE(sv) >= SVt_PVMG &&
6640 (*mgp = mg_find(sv, PERL_MAGIC_utf8))))) {
6641 if ((*mgp)->mg_ptr) {
6642 STRLEN *cache = (STRLEN *) (*mgp)->mg_ptr;
6643 if (cache[0] == uoffset) {
6644 /* An exact match. */
6647 if (cache[2] == uoffset) {
6648 /* An exact match. */
6652 if (cache[0] < uoffset) {
6653 /* The cache already knows part of the way. */
6654 if (cache[0] > uoffset0) {
6655 /* The cache knows more than the passed in pair */
6656 uoffset0 = cache[0];
6657 boffset0 = cache[1];
6659 if ((*mgp)->mg_len != -1) {
6660 /* And we know the end too. */
6662 + sv_pos_u2b_midway(start + boffset0, send,
6664 (*mgp)->mg_len - uoffset0);
6666 uoffset -= uoffset0;
6668 + sv_pos_u2b_forwards(start + boffset0,
6669 send, &uoffset, &at_end);
6670 uoffset += uoffset0;
6673 else if (cache[2] < uoffset) {
6674 /* We're between the two cache entries. */
6675 if (cache[2] > uoffset0) {
6676 /* and the cache knows more than the passed in pair */
6677 uoffset0 = cache[2];
6678 boffset0 = cache[3];
6682 + sv_pos_u2b_midway(start + boffset0,
6685 cache[0] - uoffset0);
6688 + sv_pos_u2b_midway(start + boffset0,
6691 cache[2] - uoffset0);
6695 else if ((*mgp)->mg_len != -1) {
6696 /* If we can take advantage of a passed in offset, do so. */
6697 /* In fact, offset0 is either 0, or less than offset, so don't
6698 need to worry about the other possibility. */
6700 + sv_pos_u2b_midway(start + boffset0, send,
6702 (*mgp)->mg_len - uoffset0);
6707 if (!found || PL_utf8cache < 0) {
6708 STRLEN real_boffset;
6709 uoffset -= uoffset0;
6710 real_boffset = boffset0 + sv_pos_u2b_forwards(start + boffset0,
6711 send, &uoffset, &at_end);
6712 uoffset += uoffset0;
6714 if (found && PL_utf8cache < 0)
6715 assert_uft8_cache_coherent("sv_pos_u2b_cache", boffset,
6717 boffset = real_boffset;
6722 utf8_mg_len_cache_update(sv, mgp, uoffset);
6724 utf8_mg_pos_cache_update(sv, mgp, boffset, uoffset, send - start);
6731 =for apidoc sv_pos_u2b_flags
6733 Converts the value pointed to by offsetp from a count of UTF-8 chars from
6734 the start of the string, to a count of the equivalent number of bytes; if
6735 lenp is non-zero, it does the same to lenp, but this time starting from
6736 the offset, rather than from the start
6737 of the string. Handles type coercion.
6738 I<flags> is passed to C<SvPV_flags>, and usually should be
6739 C<SV_GMAGIC|SV_CONST_RETURN> to handle magic.
6745 * sv_pos_u2b_flags() uses, like sv_pos_b2u(), the mg_ptr of the potential
6746 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6747 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
6752 Perl_sv_pos_u2b_flags(pTHX_ SV *const sv, STRLEN uoffset, STRLEN *const lenp,
6759 PERL_ARGS_ASSERT_SV_POS_U2B_FLAGS;
6761 start = (U8*)SvPV_flags(sv, len, flags);
6763 const U8 * const send = start + len;
6765 boffset = sv_pos_u2b_cached(sv, &mg, start, send, uoffset, 0, 0);
6768 && *lenp /* don't bother doing work for 0, as its bytes equivalent
6769 is 0, and *lenp is already set to that. */) {
6770 /* Convert the relative offset to absolute. */
6771 const STRLEN uoffset2 = uoffset + *lenp;
6772 const STRLEN boffset2
6773 = sv_pos_u2b_cached(sv, &mg, start, send, uoffset2,
6774 uoffset, boffset) - boffset;
6788 =for apidoc sv_pos_u2b
6790 Converts the value pointed to by offsetp from a count of UTF-8 chars from
6791 the start of the string, to a count of the equivalent number of bytes; if
6792 lenp is non-zero, it does the same to lenp, but this time starting from
6793 the offset, rather than from the start of the string. Handles magic and
6796 Use C<sv_pos_u2b_flags> in preference, which correctly handles strings longer
6803 * sv_pos_u2b() uses, like sv_pos_b2u(), the mg_ptr of the potential
6804 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6805 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
6809 /* This function is subject to size and sign problems */
6812 Perl_sv_pos_u2b(pTHX_ register SV *const sv, I32 *const offsetp, I32 *const lenp)
6814 PERL_ARGS_ASSERT_SV_POS_U2B;
6817 STRLEN ulen = (STRLEN)*lenp;
6818 *offsetp = (I32)sv_pos_u2b_flags(sv, (STRLEN)*offsetp, &ulen,
6819 SV_GMAGIC|SV_CONST_RETURN);
6822 *offsetp = (I32)sv_pos_u2b_flags(sv, (STRLEN)*offsetp, NULL,
6823 SV_GMAGIC|SV_CONST_RETURN);
6828 S_utf8_mg_len_cache_update(pTHX_ SV *const sv, MAGIC **const mgp,
6831 PERL_ARGS_ASSERT_UTF8_MG_LEN_CACHE_UPDATE;
6835 if (!*mgp && (SvTYPE(sv) < SVt_PVMG ||
6836 !(*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
6837 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, &PL_vtbl_utf8, 0, 0);
6841 (*mgp)->mg_len = ulen;
6842 /* For now, treat "overflowed" as "still unknown". See RT #72924. */
6843 if (ulen != (STRLEN) (*mgp)->mg_len)
6844 (*mgp)->mg_len = -1;
6847 /* Create and update the UTF8 magic offset cache, with the proffered utf8/
6848 byte length pairing. The (byte) length of the total SV is passed in too,
6849 as blen, because for some (more esoteric) SVs, the call to SvPV_const()
6850 may not have updated SvCUR, so we can't rely on reading it directly.
6852 The proffered utf8/byte length pairing isn't used if the cache already has
6853 two pairs, and swapping either for the proffered pair would increase the
6854 RMS of the intervals between known byte offsets.
6856 The cache itself consists of 4 STRLEN values
6857 0: larger UTF-8 offset
6858 1: corresponding byte offset
6859 2: smaller UTF-8 offset
6860 3: corresponding byte offset
6862 Unused cache pairs have the value 0, 0.
6863 Keeping the cache "backwards" means that the invariant of
6864 cache[0] >= cache[2] is maintained even with empty slots, which means that
6865 the code that uses it doesn't need to worry if only 1 entry has actually
6866 been set to non-zero. It also makes the "position beyond the end of the
6867 cache" logic much simpler, as the first slot is always the one to start
6871 S_utf8_mg_pos_cache_update(pTHX_ SV *const sv, MAGIC **const mgp, const STRLEN byte,
6872 const STRLEN utf8, const STRLEN blen)
6876 PERL_ARGS_ASSERT_UTF8_MG_POS_CACHE_UPDATE;
6881 if (!*mgp && (SvTYPE(sv) < SVt_PVMG ||
6882 !(*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
6883 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, (MGVTBL*)&PL_vtbl_utf8, 0,
6885 (*mgp)->mg_len = -1;
6889 if (!(cache = (STRLEN *)(*mgp)->mg_ptr)) {
6890 Newxz(cache, PERL_MAGIC_UTF8_CACHESIZE * 2, STRLEN);
6891 (*mgp)->mg_ptr = (char *) cache;
6895 if (PL_utf8cache < 0 && SvPOKp(sv)) {
6896 /* SvPOKp() because it's possible that sv has string overloading, and
6897 therefore is a reference, hence SvPVX() is actually a pointer.
6898 This cures the (very real) symptoms of RT 69422, but I'm not actually
6899 sure whether we should even be caching the results of UTF-8
6900 operations on overloading, given that nothing stops overloading
6901 returning a different value every time it's called. */
6902 const U8 *start = (const U8 *) SvPVX_const(sv);
6903 const STRLEN realutf8 = utf8_length(start, start + byte);
6905 assert_uft8_cache_coherent("utf8_mg_pos_cache_update", utf8, realutf8,
6909 /* Cache is held with the later position first, to simplify the code
6910 that deals with unbounded ends. */
6912 ASSERT_UTF8_CACHE(cache);
6913 if (cache[1] == 0) {
6914 /* Cache is totally empty */
6917 } else if (cache[3] == 0) {
6918 if (byte > cache[1]) {
6919 /* New one is larger, so goes first. */
6920 cache[2] = cache[0];
6921 cache[3] = cache[1];
6929 #define THREEWAY_SQUARE(a,b,c,d) \
6930 ((float)((d) - (c))) * ((float)((d) - (c))) \
6931 + ((float)((c) - (b))) * ((float)((c) - (b))) \
6932 + ((float)((b) - (a))) * ((float)((b) - (a)))
6934 /* Cache has 2 slots in use, and we know three potential pairs.
6935 Keep the two that give the lowest RMS distance. Do the
6936 calculation in bytes simply because we always know the byte
6937 length. squareroot has the same ordering as the positive value,
6938 so don't bother with the actual square root. */
6939 const float existing = THREEWAY_SQUARE(0, cache[3], cache[1], blen);
6940 if (byte > cache[1]) {
6941 /* New position is after the existing pair of pairs. */
6942 const float keep_earlier
6943 = THREEWAY_SQUARE(0, cache[3], byte, blen);
6944 const float keep_later
6945 = THREEWAY_SQUARE(0, cache[1], byte, blen);
6947 if (keep_later < keep_earlier) {
6948 if (keep_later < existing) {
6949 cache[2] = cache[0];
6950 cache[3] = cache[1];
6956 if (keep_earlier < existing) {
6962 else if (byte > cache[3]) {
6963 /* New position is between the existing pair of pairs. */
6964 const float keep_earlier
6965 = THREEWAY_SQUARE(0, cache[3], byte, blen);
6966 const float keep_later
6967 = THREEWAY_SQUARE(0, byte, cache[1], blen);
6969 if (keep_later < keep_earlier) {
6970 if (keep_later < existing) {
6976 if (keep_earlier < existing) {
6983 /* New position is before the existing pair of pairs. */
6984 const float keep_earlier
6985 = THREEWAY_SQUARE(0, byte, cache[3], blen);
6986 const float keep_later
6987 = THREEWAY_SQUARE(0, byte, cache[1], blen);
6989 if (keep_later < keep_earlier) {
6990 if (keep_later < existing) {
6996 if (keep_earlier < existing) {
6997 cache[0] = cache[2];
6998 cache[1] = cache[3];
7005 ASSERT_UTF8_CACHE(cache);
7008 /* We already know all of the way, now we may be able to walk back. The same
7009 assumption is made as in S_sv_pos_u2b_midway(), namely that walking
7010 backward is half the speed of walking forward. */
7012 S_sv_pos_b2u_midway(pTHX_ const U8 *const s, const U8 *const target,
7013 const U8 *end, STRLEN endu)
7015 const STRLEN forw = target - s;
7016 STRLEN backw = end - target;
7018 PERL_ARGS_ASSERT_SV_POS_B2U_MIDWAY;
7020 if (forw < 2 * backw) {
7021 return utf8_length(s, target);
7024 while (end > target) {
7026 while (UTF8_IS_CONTINUATION(*end)) {
7035 =for apidoc sv_pos_b2u
7037 Converts the value pointed to by offsetp from a count of bytes from the
7038 start of the string, to a count of the equivalent number of UTF-8 chars.
7039 Handles magic and type coercion.
7045 * sv_pos_b2u() uses, like sv_pos_u2b(), the mg_ptr of the potential
7046 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
7051 Perl_sv_pos_b2u(pTHX_ register SV *const sv, I32 *const offsetp)
7054 const STRLEN byte = *offsetp;
7055 STRLEN len = 0; /* Actually always set, but let's keep gcc happy. */
7061 PERL_ARGS_ASSERT_SV_POS_B2U;
7066 s = (const U8*)SvPV_const(sv, blen);
7069 Perl_croak(aTHX_ "panic: sv_pos_b2u: bad byte offset, blen=%"UVuf
7070 ", byte=%"UVuf, (UV)blen, (UV)byte);
7076 && SvTYPE(sv) >= SVt_PVMG
7077 && (mg = mg_find(sv, PERL_MAGIC_utf8)))
7080 STRLEN * const cache = (STRLEN *) mg->mg_ptr;
7081 if (cache[1] == byte) {
7082 /* An exact match. */
7083 *offsetp = cache[0];
7086 if (cache[3] == byte) {
7087 /* An exact match. */
7088 *offsetp = cache[2];
7092 if (cache[1] < byte) {
7093 /* We already know part of the way. */
7094 if (mg->mg_len != -1) {
7095 /* Actually, we know the end too. */
7097 + S_sv_pos_b2u_midway(aTHX_ s + cache[1], send,
7098 s + blen, mg->mg_len - cache[0]);
7100 len = cache[0] + utf8_length(s + cache[1], send);
7103 else if (cache[3] < byte) {
7104 /* We're between the two cached pairs, so we do the calculation
7105 offset by the byte/utf-8 positions for the earlier pair,
7106 then add the utf-8 characters from the string start to
7108 len = S_sv_pos_b2u_midway(aTHX_ s + cache[3], send,
7109 s + cache[1], cache[0] - cache[2])
7113 else { /* cache[3] > byte */
7114 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + cache[3],
7118 ASSERT_UTF8_CACHE(cache);
7120 } else if (mg->mg_len != -1) {
7121 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + blen, mg->mg_len);
7125 if (!found || PL_utf8cache < 0) {
7126 const STRLEN real_len = utf8_length(s, send);
7128 if (found && PL_utf8cache < 0)
7129 assert_uft8_cache_coherent("sv_pos_b2u", len, real_len, sv);
7136 utf8_mg_len_cache_update(sv, &mg, len);
7138 utf8_mg_pos_cache_update(sv, &mg, byte, len, blen);
7143 S_assert_uft8_cache_coherent(pTHX_ const char *const func, STRLEN from_cache,
7144 STRLEN real, SV *const sv)
7146 PERL_ARGS_ASSERT_ASSERT_UFT8_CACHE_COHERENT;
7148 /* As this is debugging only code, save space by keeping this test here,
7149 rather than inlining it in all the callers. */
7150 if (from_cache == real)
7153 /* Need to turn the assertions off otherwise we may recurse infinitely
7154 while printing error messages. */
7155 SAVEI8(PL_utf8cache);
7157 Perl_croak(aTHX_ "panic: %s cache %"UVuf" real %"UVuf" for %"SVf,
7158 func, (UV) from_cache, (UV) real, SVfARG(sv));
7164 Returns a boolean indicating whether the strings in the two SVs are
7165 identical. Is UTF-8 and 'use bytes' aware, handles get magic, and will
7166 coerce its args to strings if necessary.
7168 =for apidoc sv_eq_flags
7170 Returns a boolean indicating whether the strings in the two SVs are
7171 identical. Is UTF-8 and 'use bytes' aware and coerces its args to strings
7172 if necessary. If the flags include SV_GMAGIC, it handles get-magic, too.
7178 Perl_sv_eq_flags(pTHX_ register SV *sv1, register SV *sv2, const U32 flags)
7186 SV* svrecode = NULL;
7193 /* if pv1 and pv2 are the same, second SvPV_const call may
7194 * invalidate pv1 (if we are handling magic), so we may need to
7196 if (sv1 == sv2 && flags & SV_GMAGIC
7197 && (SvTHINKFIRST(sv1) || SvGMAGICAL(sv1))) {
7198 pv1 = SvPV_const(sv1, cur1);
7199 sv1 = newSVpvn_flags(pv1, cur1, SVs_TEMP | SvUTF8(sv2));
7201 pv1 = SvPV_flags_const(sv1, cur1, flags);
7209 pv2 = SvPV_flags_const(sv2, cur2, flags);
7211 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
7212 /* Differing utf8ness.
7213 * Do not UTF8size the comparands as a side-effect. */
7216 svrecode = newSVpvn(pv2, cur2);
7217 sv_recode_to_utf8(svrecode, PL_encoding);
7218 pv2 = SvPV_const(svrecode, cur2);
7221 svrecode = newSVpvn(pv1, cur1);
7222 sv_recode_to_utf8(svrecode, PL_encoding);
7223 pv1 = SvPV_const(svrecode, cur1);
7225 /* Now both are in UTF-8. */
7227 SvREFCNT_dec(svrecode);
7233 /* sv1 is the UTF-8 one */
7234 return bytes_cmp_utf8((const U8*)pv2, cur2,
7235 (const U8*)pv1, cur1) == 0;
7238 /* sv2 is the UTF-8 one */
7239 return bytes_cmp_utf8((const U8*)pv1, cur1,
7240 (const U8*)pv2, cur2) == 0;
7246 eq = (pv1 == pv2) || memEQ(pv1, pv2, cur1);
7248 SvREFCNT_dec(svrecode);
7256 Compares the strings in two SVs. Returns -1, 0, or 1 indicating whether the
7257 string in C<sv1> is less than, equal to, or greater than the string in
7258 C<sv2>. Is UTF-8 and 'use bytes' aware, handles get magic, and will
7259 coerce its args to strings if necessary. See also C<sv_cmp_locale>.
7261 =for apidoc sv_cmp_flags
7263 Compares the strings in two SVs. Returns -1, 0, or 1 indicating whether the
7264 string in C<sv1> is less than, equal to, or greater than the string in
7265 C<sv2>. Is UTF-8 and 'use bytes' aware and will coerce its args to strings
7266 if necessary. If the flags include SV_GMAGIC, it handles get magic. See
7267 also C<sv_cmp_locale_flags>.
7273 Perl_sv_cmp(pTHX_ register SV *const sv1, register SV *const sv2)
7275 return sv_cmp_flags(sv1, sv2, SV_GMAGIC);
7279 Perl_sv_cmp_flags(pTHX_ register SV *const sv1, register SV *const sv2,
7284 const char *pv1, *pv2;
7287 SV *svrecode = NULL;
7294 pv1 = SvPV_flags_const(sv1, cur1, flags);
7301 pv2 = SvPV_flags_const(sv2, cur2, flags);
7303 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
7304 /* Differing utf8ness.
7305 * Do not UTF8size the comparands as a side-effect. */
7308 svrecode = newSVpvn(pv2, cur2);
7309 sv_recode_to_utf8(svrecode, PL_encoding);
7310 pv2 = SvPV_const(svrecode, cur2);
7313 const int retval = -bytes_cmp_utf8((const U8*)pv2, cur2,
7314 (const U8*)pv1, cur1);
7315 return retval ? retval < 0 ? -1 : +1 : 0;
7320 svrecode = newSVpvn(pv1, cur1);
7321 sv_recode_to_utf8(svrecode, PL_encoding);
7322 pv1 = SvPV_const(svrecode, cur1);
7325 const int retval = bytes_cmp_utf8((const U8*)pv1, cur1,
7326 (const U8*)pv2, cur2);
7327 return retval ? retval < 0 ? -1 : +1 : 0;
7333 cmp = cur2 ? -1 : 0;
7337 const I32 retval = memcmp((const void*)pv1, (const void*)pv2, cur1 < cur2 ? cur1 : cur2);
7340 cmp = retval < 0 ? -1 : 1;
7341 } else if (cur1 == cur2) {
7344 cmp = cur1 < cur2 ? -1 : 1;
7348 SvREFCNT_dec(svrecode);
7356 =for apidoc sv_cmp_locale
7358 Compares the strings in two SVs in a locale-aware manner. Is UTF-8 and
7359 'use bytes' aware, handles get magic, and will coerce its args to strings
7360 if necessary. See also C<sv_cmp>.
7362 =for apidoc sv_cmp_locale_flags
7364 Compares the strings in two SVs in a locale-aware manner. Is UTF-8 and
7365 'use bytes' aware and will coerce its args to strings if necessary. If the
7366 flags contain SV_GMAGIC, it handles get magic. See also C<sv_cmp_flags>.
7372 Perl_sv_cmp_locale(pTHX_ register SV *const sv1, register SV *const sv2)
7374 return sv_cmp_locale_flags(sv1, sv2, SV_GMAGIC);
7378 Perl_sv_cmp_locale_flags(pTHX_ register SV *const sv1, register SV *const sv2,
7382 #ifdef USE_LOCALE_COLLATE
7388 if (PL_collation_standard)
7392 pv1 = sv1 ? sv_collxfrm_flags(sv1, &len1, flags) : (char *) NULL;
7394 pv2 = sv2 ? sv_collxfrm_flags(sv2, &len2, flags) : (char *) NULL;
7396 if (!pv1 || !len1) {
7407 retval = memcmp((void*)pv1, (void*)pv2, len1 < len2 ? len1 : len2);
7410 return retval < 0 ? -1 : 1;
7413 * When the result of collation is equality, that doesn't mean
7414 * that there are no differences -- some locales exclude some
7415 * characters from consideration. So to avoid false equalities,
7416 * we use the raw string as a tiebreaker.
7422 #endif /* USE_LOCALE_COLLATE */
7424 return sv_cmp(sv1, sv2);
7428 #ifdef USE_LOCALE_COLLATE
7431 =for apidoc sv_collxfrm
7433 This calls C<sv_collxfrm_flags> with the SV_GMAGIC flag. See
7434 C<sv_collxfrm_flags>.
7436 =for apidoc sv_collxfrm_flags
7438 Add Collate Transform magic to an SV if it doesn't already have it. If the
7439 flags contain SV_GMAGIC, it handles get-magic.
7441 Any scalar variable may carry PERL_MAGIC_collxfrm magic that contains the
7442 scalar data of the variable, but transformed to such a format that a normal
7443 memory comparison can be used to compare the data according to the locale
7450 Perl_sv_collxfrm_flags(pTHX_ SV *const sv, STRLEN *const nxp, const I32 flags)
7455 PERL_ARGS_ASSERT_SV_COLLXFRM_FLAGS;
7457 mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_collxfrm) : (MAGIC *) NULL;
7458 if (!mg || !mg->mg_ptr || *(U32*)mg->mg_ptr != PL_collation_ix) {
7464 Safefree(mg->mg_ptr);
7465 s = SvPV_flags_const(sv, len, flags);
7466 if ((xf = mem_collxfrm(s, len, &xlen))) {
7468 #ifdef PERL_OLD_COPY_ON_WRITE
7470 sv_force_normal_flags(sv, 0);
7472 mg = sv_magicext(sv, 0, PERL_MAGIC_collxfrm, &PL_vtbl_collxfrm,
7486 if (mg && mg->mg_ptr) {
7488 return mg->mg_ptr + sizeof(PL_collation_ix);
7496 #endif /* USE_LOCALE_COLLATE */
7499 S_sv_gets_append_to_utf8(pTHX_ SV *const sv, PerlIO *const fp, I32 append)
7501 SV * const tsv = newSV(0);
7504 sv_gets(tsv, fp, 0);
7505 sv_utf8_upgrade_nomg(tsv);
7506 SvCUR_set(sv,append);
7509 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7513 S_sv_gets_read_record(pTHX_ SV *const sv, PerlIO *const fp, I32 append)
7516 const U32 recsize = SvUV(SvRV(PL_rs)); /* RsRECORD() guarantees > 0. */
7517 /* Grab the size of the record we're getting */
7518 char *const buffer = SvGROW(sv, (STRLEN)(recsize + append + 1)) + append;
7525 /* VMS wants read instead of fread, because fread doesn't respect */
7526 /* RMS record boundaries. This is not necessarily a good thing to be */
7527 /* doing, but we've got no other real choice - except avoid stdio
7528 as implementation - perhaps write a :vms layer ?
7530 fd = PerlIO_fileno(fp);
7532 bytesread = PerlLIO_read(fd, buffer, recsize);
7534 else /* in-memory file from PerlIO::Scalar */
7537 bytesread = PerlIO_read(fp, buffer, recsize);
7542 SvCUR_set(sv, bytesread + append);
7543 buffer[bytesread] = '\0';
7544 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7550 Get a line from the filehandle and store it into the SV, optionally
7551 appending to the currently-stored string. If C<append> is not 0, the
7552 line is appended to the SV instead of overwriting it. C<append> should
7553 be set to the byte offset that the appended string should start at
7554 in the SV (typically, C<SvCUR(sv)> is a suitable choice).
7560 Perl_sv_gets(pTHX_ register SV *const sv, register PerlIO *const fp, I32 append)
7565 register STDCHAR rslast;
7566 register STDCHAR *bp;
7571 PERL_ARGS_ASSERT_SV_GETS;
7573 if (SvTHINKFIRST(sv))
7574 sv_force_normal_flags(sv, append ? 0 : SV_COW_DROP_PV);
7575 /* XXX. If you make this PVIV, then copy on write can copy scalars read
7577 However, perlbench says it's slower, because the existing swipe code
7578 is faster than copy on write.
7579 Swings and roundabouts. */
7580 SvUPGRADE(sv, SVt_PV);
7583 if (PerlIO_isutf8(fp)) {
7585 sv_utf8_upgrade_nomg(sv);
7586 sv_pos_u2b(sv,&append,0);
7588 } else if (SvUTF8(sv)) {
7589 return S_sv_gets_append_to_utf8(aTHX_ sv, fp, append);
7597 if (PerlIO_isutf8(fp))
7600 if (IN_PERL_COMPILETIME) {
7601 /* we always read code in line mode */
7605 else if (RsSNARF(PL_rs)) {
7606 /* If it is a regular disk file use size from stat() as estimate
7607 of amount we are going to read -- may result in mallocing
7608 more memory than we really need if the layers below reduce
7609 the size we read (e.g. CRLF or a gzip layer).
7612 if (!PerlLIO_fstat(PerlIO_fileno(fp), &st) && S_ISREG(st.st_mode)) {
7613 const Off_t offset = PerlIO_tell(fp);
7614 if (offset != (Off_t) -1 && st.st_size + append > offset) {
7615 (void) SvGROW(sv, (STRLEN)((st.st_size - offset) + append + 1));
7621 else if (RsRECORD(PL_rs)) {
7622 return S_sv_gets_read_record(aTHX_ sv, fp, append);
7624 else if (RsPARA(PL_rs)) {
7630 /* Get $/ i.e. PL_rs into same encoding as stream wants */
7631 if (PerlIO_isutf8(fp)) {
7632 rsptr = SvPVutf8(PL_rs, rslen);
7635 if (SvUTF8(PL_rs)) {
7636 if (!sv_utf8_downgrade(PL_rs, TRUE)) {
7637 Perl_croak(aTHX_ "Wide character in $/");
7640 rsptr = SvPV_const(PL_rs, rslen);
7644 rslast = rslen ? rsptr[rslen - 1] : '\0';
7646 if (rspara) { /* have to do this both before and after */
7647 do { /* to make sure file boundaries work right */
7650 i = PerlIO_getc(fp);
7654 PerlIO_ungetc(fp,i);
7660 /* See if we know enough about I/O mechanism to cheat it ! */
7662 /* This used to be #ifdef test - it is made run-time test for ease
7663 of abstracting out stdio interface. One call should be cheap
7664 enough here - and may even be a macro allowing compile
7668 if (PerlIO_fast_gets(fp)) {
7671 * We're going to steal some values from the stdio struct
7672 * and put EVERYTHING in the innermost loop into registers.
7674 register STDCHAR *ptr;
7678 #if defined(VMS) && defined(PERLIO_IS_STDIO)
7679 /* An ungetc()d char is handled separately from the regular
7680 * buffer, so we getc() it back out and stuff it in the buffer.
7682 i = PerlIO_getc(fp);
7683 if (i == EOF) return 0;
7684 *(--((*fp)->_ptr)) = (unsigned char) i;
7688 /* Here is some breathtakingly efficient cheating */
7690 cnt = PerlIO_get_cnt(fp); /* get count into register */
7691 /* make sure we have the room */
7692 if ((I32)(SvLEN(sv) - append) <= cnt + 1) {
7693 /* Not room for all of it
7694 if we are looking for a separator and room for some
7696 if (rslen && cnt > 80 && (I32)SvLEN(sv) > append) {
7697 /* just process what we have room for */
7698 shortbuffered = cnt - SvLEN(sv) + append + 1;
7699 cnt -= shortbuffered;
7703 /* remember that cnt can be negative */
7704 SvGROW(sv, (STRLEN)(append + (cnt <= 0 ? 2 : (cnt + 1))));
7709 bp = (STDCHAR*)SvPVX_const(sv) + append; /* move these two too to registers */
7710 ptr = (STDCHAR*)PerlIO_get_ptr(fp);
7711 DEBUG_P(PerlIO_printf(Perl_debug_log,
7712 "Screamer: entering, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7713 DEBUG_P(PerlIO_printf(Perl_debug_log,
7714 "Screamer: entering: PerlIO * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7715 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7716 PTR2UV(PerlIO_has_base(fp) ? PerlIO_get_base(fp) : 0)));
7721 while (cnt > 0) { /* this | eat */
7723 if ((*bp++ = *ptr++) == rslast) /* really | dust */
7724 goto thats_all_folks; /* screams | sed :-) */
7728 Copy(ptr, bp, cnt, char); /* this | eat */
7729 bp += cnt; /* screams | dust */
7730 ptr += cnt; /* louder | sed :-) */
7732 assert (!shortbuffered);
7733 goto cannot_be_shortbuffered;
7737 if (shortbuffered) { /* oh well, must extend */
7738 cnt = shortbuffered;
7740 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
7742 SvGROW(sv, SvLEN(sv) + append + cnt + 2);
7743 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
7747 cannot_be_shortbuffered:
7748 DEBUG_P(PerlIO_printf(Perl_debug_log,
7749 "Screamer: going to getc, ptr=%"UVuf", cnt=%ld\n",
7750 PTR2UV(ptr),(long)cnt));
7751 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* deregisterize cnt and ptr */
7753 DEBUG_Pv(PerlIO_printf(Perl_debug_log,
7754 "Screamer: pre: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7755 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7756 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7758 /* This used to call 'filbuf' in stdio form, but as that behaves like
7759 getc when cnt <= 0 we use PerlIO_getc here to avoid introducing
7760 another abstraction. */
7761 i = PerlIO_getc(fp); /* get more characters */
7763 DEBUG_Pv(PerlIO_printf(Perl_debug_log,
7764 "Screamer: post: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7765 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7766 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7768 cnt = PerlIO_get_cnt(fp);
7769 ptr = (STDCHAR*)PerlIO_get_ptr(fp); /* reregisterize cnt and ptr */
7770 DEBUG_P(PerlIO_printf(Perl_debug_log,
7771 "Screamer: after getc, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7773 if (i == EOF) /* all done for ever? */
7774 goto thats_really_all_folks;
7776 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
7778 SvGROW(sv, bpx + cnt + 2);
7779 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
7781 *bp++ = (STDCHAR)i; /* store character from PerlIO_getc */
7783 if (rslen && (STDCHAR)i == rslast) /* all done for now? */
7784 goto thats_all_folks;
7788 if ((rslen > 1 && (STRLEN)(bp - (STDCHAR*)SvPVX_const(sv)) < rslen) ||
7789 memNE((char*)bp - rslen, rsptr, rslen))
7790 goto screamer; /* go back to the fray */
7791 thats_really_all_folks:
7793 cnt += shortbuffered;
7794 DEBUG_P(PerlIO_printf(Perl_debug_log,
7795 "Screamer: quitting, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7796 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* put these back or we're in trouble */
7797 DEBUG_P(PerlIO_printf(Perl_debug_log,
7798 "Screamer: end: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7799 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7800 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7802 SvCUR_set(sv, bp - (STDCHAR*)SvPVX_const(sv)); /* set length */
7803 DEBUG_P(PerlIO_printf(Perl_debug_log,
7804 "Screamer: done, len=%ld, string=|%.*s|\n",
7805 (long)SvCUR(sv),(int)SvCUR(sv),SvPVX_const(sv)));
7809 /*The big, slow, and stupid way. */
7810 #ifdef USE_HEAP_INSTEAD_OF_STACK /* Even slower way. */
7811 STDCHAR *buf = NULL;
7812 Newx(buf, 8192, STDCHAR);
7820 register const STDCHAR * const bpe = buf + sizeof(buf);
7822 while ((i = PerlIO_getc(fp)) != EOF && (*bp++ = (STDCHAR)i) != rslast && bp < bpe)
7823 ; /* keep reading */
7827 cnt = PerlIO_read(fp,(char*)buf, sizeof(buf));
7828 /* Accommodate broken VAXC compiler, which applies U8 cast to
7829 * both args of ?: operator, causing EOF to change into 255
7832 i = (U8)buf[cnt - 1];
7838 cnt = 0; /* we do need to re-set the sv even when cnt <= 0 */
7840 sv_catpvn_nomg(sv, (char *) buf, cnt);
7842 sv_setpvn(sv, (char *) buf, cnt); /* "nomg" is implied */
7844 if (i != EOF && /* joy */
7846 SvCUR(sv) < rslen ||
7847 memNE(SvPVX_const(sv) + SvCUR(sv) - rslen, rsptr, rslen)))
7851 * If we're reading from a TTY and we get a short read,
7852 * indicating that the user hit his EOF character, we need
7853 * to notice it now, because if we try to read from the TTY
7854 * again, the EOF condition will disappear.
7856 * The comparison of cnt to sizeof(buf) is an optimization
7857 * that prevents unnecessary calls to feof().
7861 if (!(cnt < (I32)sizeof(buf) && PerlIO_eof(fp)))
7865 #ifdef USE_HEAP_INSTEAD_OF_STACK
7870 if (rspara) { /* have to do this both before and after */
7871 while (i != EOF) { /* to make sure file boundaries work right */
7872 i = PerlIO_getc(fp);
7874 PerlIO_ungetc(fp,i);
7880 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7886 Auto-increment of the value in the SV, doing string to numeric conversion
7887 if necessary. Handles 'get' magic and operator overloading.
7893 Perl_sv_inc(pTHX_ register SV *const sv)
7902 =for apidoc sv_inc_nomg
7904 Auto-increment of the value in the SV, doing string to numeric conversion
7905 if necessary. Handles operator overloading. Skips handling 'get' magic.
7911 Perl_sv_inc_nomg(pTHX_ register SV *const sv)
7919 if (SvTHINKFIRST(sv)) {
7920 if (SvIsCOW(sv) || isGV_with_GP(sv))
7921 sv_force_normal_flags(sv, 0);
7922 if (SvREADONLY(sv)) {
7923 if (IN_PERL_RUNTIME)
7924 Perl_croak_no_modify(aTHX);
7928 if (SvAMAGIC(sv) && AMG_CALLunary(sv, inc_amg))
7930 i = PTR2IV(SvRV(sv));
7935 flags = SvFLAGS(sv);
7936 if ((flags & (SVp_NOK|SVp_IOK)) == SVp_NOK) {
7937 /* It's (privately or publicly) a float, but not tested as an
7938 integer, so test it to see. */
7940 flags = SvFLAGS(sv);
7942 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
7943 /* It's publicly an integer, or privately an integer-not-float */
7944 #ifdef PERL_PRESERVE_IVUV
7948 if (SvUVX(sv) == UV_MAX)
7949 sv_setnv(sv, UV_MAX_P1);
7951 (void)SvIOK_only_UV(sv);
7952 SvUV_set(sv, SvUVX(sv) + 1);
7954 if (SvIVX(sv) == IV_MAX)
7955 sv_setuv(sv, (UV)IV_MAX + 1);
7957 (void)SvIOK_only(sv);
7958 SvIV_set(sv, SvIVX(sv) + 1);
7963 if (flags & SVp_NOK) {
7964 const NV was = SvNVX(sv);
7965 if (NV_OVERFLOWS_INTEGERS_AT &&
7966 was >= NV_OVERFLOWS_INTEGERS_AT) {
7967 /* diag_listed_as: Lost precision when %s %f by 1 */
7968 Perl_ck_warner(aTHX_ packWARN(WARN_IMPRECISION),
7969 "Lost precision when incrementing %" NVff " by 1",
7972 (void)SvNOK_only(sv);
7973 SvNV_set(sv, was + 1.0);
7977 if (!(flags & SVp_POK) || !*SvPVX_const(sv)) {
7978 if ((flags & SVTYPEMASK) < SVt_PVIV)
7979 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV ? SVt_PVIV : SVt_IV));
7980 (void)SvIOK_only(sv);
7985 while (isALPHA(*d)) d++;
7986 while (isDIGIT(*d)) d++;
7987 if (d < SvEND(sv)) {
7988 #ifdef PERL_PRESERVE_IVUV
7989 /* Got to punt this as an integer if needs be, but we don't issue
7990 warnings. Probably ought to make the sv_iv_please() that does
7991 the conversion if possible, and silently. */
7992 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
7993 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
7994 /* Need to try really hard to see if it's an integer.
7995 9.22337203685478e+18 is an integer.
7996 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
7997 so $a="9.22337203685478e+18"; $a+0; $a++
7998 needs to be the same as $a="9.22337203685478e+18"; $a++
8005 /* sv_2iv *should* have made this an NV */
8006 if (flags & SVp_NOK) {
8007 (void)SvNOK_only(sv);
8008 SvNV_set(sv, SvNVX(sv) + 1.0);
8011 /* I don't think we can get here. Maybe I should assert this
8012 And if we do get here I suspect that sv_setnv will croak. NWC
8014 #if defined(USE_LONG_DOUBLE)
8015 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",
8016 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
8018 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_inc punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
8019 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
8022 #endif /* PERL_PRESERVE_IVUV */
8023 sv_setnv(sv,Atof(SvPVX_const(sv)) + 1.0);
8027 while (d >= SvPVX_const(sv)) {
8035 /* MKS: The original code here died if letters weren't consecutive.
8036 * at least it didn't have to worry about non-C locales. The
8037 * new code assumes that ('z'-'a')==('Z'-'A'), letters are
8038 * arranged in order (although not consecutively) and that only
8039 * [A-Za-z] are accepted by isALPHA in the C locale.
8041 if (*d != 'z' && *d != 'Z') {
8042 do { ++*d; } while (!isALPHA(*d));
8045 *(d--) -= 'z' - 'a';
8050 *(d--) -= 'z' - 'a' + 1;
8054 /* oh,oh, the number grew */
8055 SvGROW(sv, SvCUR(sv) + 2);
8056 SvCUR_set(sv, SvCUR(sv) + 1);
8057 for (d = SvPVX(sv) + SvCUR(sv); d > SvPVX_const(sv); d--)
8068 Auto-decrement of the value in the SV, doing string to numeric conversion
8069 if necessary. Handles 'get' magic and operator overloading.
8075 Perl_sv_dec(pTHX_ register SV *const sv)
8085 =for apidoc sv_dec_nomg
8087 Auto-decrement of the value in the SV, doing string to numeric conversion
8088 if necessary. Handles operator overloading. Skips handling 'get' magic.
8094 Perl_sv_dec_nomg(pTHX_ register SV *const sv)
8101 if (SvTHINKFIRST(sv)) {
8102 if (SvIsCOW(sv) || isGV_with_GP(sv))
8103 sv_force_normal_flags(sv, 0);
8104 if (SvREADONLY(sv)) {
8105 if (IN_PERL_RUNTIME)
8106 Perl_croak_no_modify(aTHX);
8110 if (SvAMAGIC(sv) && AMG_CALLunary(sv, dec_amg))
8112 i = PTR2IV(SvRV(sv));
8117 /* Unlike sv_inc we don't have to worry about string-never-numbers
8118 and keeping them magic. But we mustn't warn on punting */
8119 flags = SvFLAGS(sv);
8120 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
8121 /* It's publicly an integer, or privately an integer-not-float */
8122 #ifdef PERL_PRESERVE_IVUV
8126 if (SvUVX(sv) == 0) {
8127 (void)SvIOK_only(sv);
8131 (void)SvIOK_only_UV(sv);
8132 SvUV_set(sv, SvUVX(sv) - 1);
8135 if (SvIVX(sv) == IV_MIN) {
8136 sv_setnv(sv, (NV)IV_MIN);
8140 (void)SvIOK_only(sv);
8141 SvIV_set(sv, SvIVX(sv) - 1);
8146 if (flags & SVp_NOK) {
8149 const NV was = SvNVX(sv);
8150 if (NV_OVERFLOWS_INTEGERS_AT &&
8151 was <= -NV_OVERFLOWS_INTEGERS_AT) {
8152 /* diag_listed_as: Lost precision when %s %f by 1 */
8153 Perl_ck_warner(aTHX_ packWARN(WARN_IMPRECISION),
8154 "Lost precision when decrementing %" NVff " by 1",
8157 (void)SvNOK_only(sv);
8158 SvNV_set(sv, was - 1.0);
8162 if (!(flags & SVp_POK)) {
8163 if ((flags & SVTYPEMASK) < SVt_PVIV)
8164 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV) ? SVt_PVIV : SVt_IV);
8166 (void)SvIOK_only(sv);
8169 #ifdef PERL_PRESERVE_IVUV
8171 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
8172 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
8173 /* Need to try really hard to see if it's an integer.
8174 9.22337203685478e+18 is an integer.
8175 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
8176 so $a="9.22337203685478e+18"; $a+0; $a--
8177 needs to be the same as $a="9.22337203685478e+18"; $a--
8184 /* sv_2iv *should* have made this an NV */
8185 if (flags & SVp_NOK) {
8186 (void)SvNOK_only(sv);
8187 SvNV_set(sv, SvNVX(sv) - 1.0);
8190 /* I don't think we can get here. Maybe I should assert this
8191 And if we do get here I suspect that sv_setnv will croak. NWC
8193 #if defined(USE_LONG_DOUBLE)
8194 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",
8195 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
8197 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_dec punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
8198 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
8202 #endif /* PERL_PRESERVE_IVUV */
8203 sv_setnv(sv,Atof(SvPVX_const(sv)) - 1.0); /* punt */
8206 /* this define is used to eliminate a chunk of duplicated but shared logic
8207 * it has the suffix __SV_C to signal that it isnt API, and isnt meant to be
8208 * used anywhere but here - yves
8210 #define PUSH_EXTEND_MORTAL__SV_C(AnSv) \
8213 PL_tmps_stack[++PL_tmps_ix] = (AnSv); \
8217 =for apidoc sv_mortalcopy
8219 Creates a new SV which is a copy of the original SV (using C<sv_setsv>).
8220 The new SV is marked as mortal. It will be destroyed "soon", either by an
8221 explicit call to FREETMPS, or by an implicit call at places such as
8222 statement boundaries. See also C<sv_newmortal> and C<sv_2mortal>.
8227 /* Make a string that will exist for the duration of the expression
8228 * evaluation. Actually, it may have to last longer than that, but
8229 * hopefully we won't free it until it has been assigned to a
8230 * permanent location. */
8233 Perl_sv_mortalcopy(pTHX_ SV *const oldstr)
8239 sv_setsv(sv,oldstr);
8240 PUSH_EXTEND_MORTAL__SV_C(sv);
8246 =for apidoc sv_newmortal
8248 Creates a new null SV which is mortal. The reference count of the SV is
8249 set to 1. It will be destroyed "soon", either by an explicit call to
8250 FREETMPS, or by an implicit call at places such as statement boundaries.
8251 See also C<sv_mortalcopy> and C<sv_2mortal>.
8257 Perl_sv_newmortal(pTHX)
8263 SvFLAGS(sv) = SVs_TEMP;
8264 PUSH_EXTEND_MORTAL__SV_C(sv);
8270 =for apidoc newSVpvn_flags
8272 Creates a new SV and copies a string into it. The reference count for the
8273 SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
8274 string. You are responsible for ensuring that the source string is at least
8275 C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
8276 Currently the only flag bits accepted are C<SVf_UTF8> and C<SVs_TEMP>.
8277 If C<SVs_TEMP> is set, then C<sv_2mortal()> is called on the result before
8278 returning. If C<SVf_UTF8> is set, C<s>
8279 is considered to be in UTF-8 and the
8280 C<SVf_UTF8> flag will be set on the new SV.
8281 C<newSVpvn_utf8()> is a convenience wrapper for this function, defined as
8283 #define newSVpvn_utf8(s, len, u) \
8284 newSVpvn_flags((s), (len), (u) ? SVf_UTF8 : 0)
8290 Perl_newSVpvn_flags(pTHX_ const char *const s, const STRLEN len, const U32 flags)
8295 /* All the flags we don't support must be zero.
8296 And we're new code so I'm going to assert this from the start. */
8297 assert(!(flags & ~(SVf_UTF8|SVs_TEMP)));
8299 sv_setpvn(sv,s,len);
8301 /* This code used to a sv_2mortal(), however we now unroll the call to sv_2mortal()
8302 * and do what it does ourselves here.
8303 * Since we have asserted that flags can only have the SVf_UTF8 and/or SVs_TEMP flags
8304 * set above we can use it to enable the sv flags directly (bypassing SvTEMP_on), which
8305 * in turn means we dont need to mask out the SVf_UTF8 flag below, which means that we
8306 * eliminate quite a few steps than it looks - Yves (explaining patch by gfx)
8309 SvFLAGS(sv) |= flags;
8311 if(flags & SVs_TEMP){
8312 PUSH_EXTEND_MORTAL__SV_C(sv);
8319 =for apidoc sv_2mortal
8321 Marks an existing SV as mortal. The SV will be destroyed "soon", either
8322 by an explicit call to FREETMPS, or by an implicit call at places such as
8323 statement boundaries. SvTEMP() is turned on which means that the SV's
8324 string buffer can be "stolen" if this SV is copied. See also C<sv_newmortal>
8325 and C<sv_mortalcopy>.
8331 Perl_sv_2mortal(pTHX_ register SV *const sv)
8336 if (SvREADONLY(sv) && SvIMMORTAL(sv))
8338 PUSH_EXTEND_MORTAL__SV_C(sv);
8346 Creates a new SV and copies a string into it. The reference count for the
8347 SV is set to 1. If C<len> is zero, Perl will compute the length using
8348 strlen(). For efficiency, consider using C<newSVpvn> instead.
8354 Perl_newSVpv(pTHX_ const char *const s, const STRLEN len)
8360 sv_setpvn(sv, s, len || s == NULL ? len : strlen(s));
8365 =for apidoc newSVpvn
8367 Creates a new SV and copies a buffer into it, which may contain NUL characters
8368 (C<\0>) and other binary data. The reference count for the SV is set to 1.
8369 Note that if C<len> is zero, Perl will create a zero length (Perl) string. You
8370 are responsible for ensuring that the source buffer is at least
8371 C<len> bytes long. If the C<buffer> argument is NULL the new SV will be
8378 Perl_newSVpvn(pTHX_ const char *const buffer, const STRLEN len)
8384 sv_setpvn(sv,buffer,len);
8389 =for apidoc newSVhek
8391 Creates a new SV from the hash key structure. It will generate scalars that
8392 point to the shared string table where possible. Returns a new (undefined)
8393 SV if the hek is NULL.
8399 Perl_newSVhek(pTHX_ const HEK *const hek)
8409 if (HEK_LEN(hek) == HEf_SVKEY) {
8410 return newSVsv(*(SV**)HEK_KEY(hek));
8412 const int flags = HEK_FLAGS(hek);
8413 if (flags & HVhek_WASUTF8) {
8415 Andreas would like keys he put in as utf8 to come back as utf8
8417 STRLEN utf8_len = HEK_LEN(hek);
8418 SV * const sv = newSV_type(SVt_PV);
8419 char *as_utf8 = (char *)bytes_to_utf8 ((U8*)HEK_KEY(hek), &utf8_len);
8420 /* bytes_to_utf8() allocates a new string, which we can repurpose: */
8421 sv_usepvn_flags(sv, as_utf8, utf8_len, SV_HAS_TRAILING_NUL);
8424 } else if (flags & (HVhek_REHASH|HVhek_UNSHARED)) {
8425 /* We don't have a pointer to the hv, so we have to replicate the
8426 flag into every HEK. This hv is using custom a hasing
8427 algorithm. Hence we can't return a shared string scalar, as
8428 that would contain the (wrong) hash value, and might get passed
8429 into an hv routine with a regular hash.
8430 Similarly, a hash that isn't using shared hash keys has to have
8431 the flag in every key so that we know not to try to call
8432 share_hek_hek on it. */
8434 SV * const sv = newSVpvn (HEK_KEY(hek), HEK_LEN(hek));
8439 /* This will be overwhelminly the most common case. */
8441 /* Inline most of newSVpvn_share(), because share_hek_hek() is far
8442 more efficient than sharepvn(). */
8446 sv_upgrade(sv, SVt_PV);
8447 SvPV_set(sv, (char *)HEK_KEY(share_hek_hek(hek)));
8448 SvCUR_set(sv, HEK_LEN(hek));
8461 =for apidoc newSVpvn_share
8463 Creates a new SV with its SvPVX_const pointing to a shared string in the string
8464 table. If the string does not already exist in the table, it is
8465 created first. Turns on READONLY and FAKE. If the C<hash> parameter
8466 is non-zero, that value is used; otherwise the hash is computed.
8467 The string's hash can later be retrieved from the SV
8468 with the C<SvSHARED_HASH()> macro. The idea here is
8469 that as the string table is used for shared hash keys these strings will have
8470 SvPVX_const == HeKEY and hash lookup will avoid string compare.
8476 Perl_newSVpvn_share(pTHX_ const char *src, I32 len, U32 hash)
8480 bool is_utf8 = FALSE;
8481 const char *const orig_src = src;
8484 STRLEN tmplen = -len;
8486 /* See the note in hv.c:hv_fetch() --jhi */
8487 src = (char*)bytes_from_utf8((const U8*)src, &tmplen, &is_utf8);
8491 PERL_HASH(hash, src, len);
8493 /* The logic for this is inlined in S_mro_get_linear_isa_dfs(), so if it
8494 changes here, update it there too. */
8495 sv_upgrade(sv, SVt_PV);
8496 SvPV_set(sv, sharepvn(src, is_utf8?-len:len, hash));
8504 if (src != orig_src)
8510 =for apidoc newSVpv_share
8512 Like C<newSVpvn_share>, but takes a nul-terminated string instead of a
8519 Perl_newSVpv_share(pTHX_ const char *src, U32 hash)
8521 return newSVpvn_share(src, strlen(src), hash);
8524 #if defined(PERL_IMPLICIT_CONTEXT)
8526 /* pTHX_ magic can't cope with varargs, so this is a no-context
8527 * version of the main function, (which may itself be aliased to us).
8528 * Don't access this version directly.
8532 Perl_newSVpvf_nocontext(const char *const pat, ...)
8538 PERL_ARGS_ASSERT_NEWSVPVF_NOCONTEXT;
8540 va_start(args, pat);
8541 sv = vnewSVpvf(pat, &args);
8548 =for apidoc newSVpvf
8550 Creates a new SV and initializes it with the string formatted like
8557 Perl_newSVpvf(pTHX_ const char *const pat, ...)
8562 PERL_ARGS_ASSERT_NEWSVPVF;
8564 va_start(args, pat);
8565 sv = vnewSVpvf(pat, &args);
8570 /* backend for newSVpvf() and newSVpvf_nocontext() */
8573 Perl_vnewSVpvf(pTHX_ const char *const pat, va_list *const args)
8578 PERL_ARGS_ASSERT_VNEWSVPVF;
8581 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
8588 Creates a new SV and copies a floating point value into it.
8589 The reference count for the SV is set to 1.
8595 Perl_newSVnv(pTHX_ const NV n)
8608 Creates a new SV and copies an integer into it. The reference count for the
8615 Perl_newSViv(pTHX_ const IV i)
8628 Creates a new SV and copies an unsigned integer into it.
8629 The reference count for the SV is set to 1.
8635 Perl_newSVuv(pTHX_ const UV u)
8646 =for apidoc newSV_type
8648 Creates a new SV, of the type specified. The reference count for the new SV
8655 Perl_newSV_type(pTHX_ const svtype type)
8660 sv_upgrade(sv, type);
8665 =for apidoc newRV_noinc
8667 Creates an RV wrapper for an SV. The reference count for the original
8668 SV is B<not> incremented.
8674 Perl_newRV_noinc(pTHX_ SV *const tmpRef)
8677 register SV *sv = newSV_type(SVt_IV);
8679 PERL_ARGS_ASSERT_NEWRV_NOINC;
8682 SvRV_set(sv, tmpRef);
8687 /* newRV_inc is the official function name to use now.
8688 * newRV_inc is in fact #defined to newRV in sv.h
8692 Perl_newRV(pTHX_ SV *const sv)
8696 PERL_ARGS_ASSERT_NEWRV;
8698 return newRV_noinc(SvREFCNT_inc_simple_NN(sv));
8704 Creates a new SV which is an exact duplicate of the original SV.
8711 Perl_newSVsv(pTHX_ register SV *const old)
8718 if (SvTYPE(old) == (svtype)SVTYPEMASK) {
8719 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL), "semi-panic: attempt to dup freed string");
8723 /* SV_GMAGIC is the default for sv_setv()
8724 SV_NOSTEAL prevents TEMP buffers being, well, stolen, and saves games
8725 with SvTEMP_off and SvTEMP_on round a call to sv_setsv. */
8726 sv_setsv_flags(sv, old, SV_GMAGIC | SV_NOSTEAL);
8731 =for apidoc sv_reset
8733 Underlying implementation for the C<reset> Perl function.
8734 Note that the perl-level function is vaguely deprecated.
8740 Perl_sv_reset(pTHX_ register const char *s, HV *const stash)
8743 char todo[PERL_UCHAR_MAX+1];
8745 PERL_ARGS_ASSERT_SV_RESET;
8750 if (!*s) { /* reset ?? searches */
8751 MAGIC * const mg = mg_find((const SV *)stash, PERL_MAGIC_symtab);
8753 const U32 count = mg->mg_len / sizeof(PMOP**);
8754 PMOP **pmp = (PMOP**) mg->mg_ptr;
8755 PMOP *const *const end = pmp + count;
8759 SvREADONLY_off(PL_regex_pad[(*pmp)->op_pmoffset]);
8761 (*pmp)->op_pmflags &= ~PMf_USED;
8769 /* reset variables */
8771 if (!HvARRAY(stash))
8774 Zero(todo, 256, char);
8777 I32 i = (unsigned char)*s;
8781 max = (unsigned char)*s++;
8782 for ( ; i <= max; i++) {
8785 for (i = 0; i <= (I32) HvMAX(stash); i++) {
8787 for (entry = HvARRAY(stash)[i];
8789 entry = HeNEXT(entry))
8794 if (!todo[(U8)*HeKEY(entry)])
8796 gv = MUTABLE_GV(HeVAL(entry));
8799 if (SvTHINKFIRST(sv)) {
8800 if (!SvREADONLY(sv) && SvROK(sv))
8802 /* XXX Is this continue a bug? Why should THINKFIRST
8803 exempt us from resetting arrays and hashes? */
8807 if (SvTYPE(sv) >= SVt_PV) {
8809 if (SvPVX_const(sv) != NULL)
8817 if (GvHV(gv) && !HvNAME_get(GvHV(gv))) {
8819 Perl_die(aTHX_ "Can't reset %%ENV on this system");
8822 # if defined(USE_ENVIRON_ARRAY)
8825 # endif /* USE_ENVIRON_ARRAY */
8836 Using various gambits, try to get an IO from an SV: the IO slot if its a
8837 GV; or the recursive result if we're an RV; or the IO slot of the symbol
8838 named after the PV if we're a string.
8840 'Get' magic is ignored on the sv passed in, but will be called on
8841 C<SvRV(sv)> if sv is an RV.
8847 Perl_sv_2io(pTHX_ SV *const sv)
8852 PERL_ARGS_ASSERT_SV_2IO;
8854 switch (SvTYPE(sv)) {
8856 io = MUTABLE_IO(sv);
8860 if (isGV_with_GP(sv)) {
8861 gv = MUTABLE_GV(sv);
8864 Perl_croak(aTHX_ "Bad filehandle: %"HEKf,
8865 HEKfARG(GvNAME_HEK(gv)));
8871 Perl_croak(aTHX_ PL_no_usym, "filehandle");
8873 SvGETMAGIC(SvRV(sv));
8874 return sv_2io(SvRV(sv));
8876 gv = gv_fetchsv_nomg(sv, 0, SVt_PVIO);
8883 if (SvGMAGICAL(sv)) {
8884 newsv = sv_newmortal();
8885 sv_setsv_nomg(newsv, sv);
8887 Perl_croak(aTHX_ "Bad filehandle: %"SVf, SVfARG(newsv));
8897 Using various gambits, try to get a CV from an SV; in addition, try if
8898 possible to set C<*st> and C<*gvp> to the stash and GV associated with it.
8899 The flags in C<lref> are passed to gv_fetchsv.
8905 Perl_sv_2cv(pTHX_ SV *sv, HV **const st, GV **const gvp, const I32 lref)
8911 PERL_ARGS_ASSERT_SV_2CV;
8918 switch (SvTYPE(sv)) {
8922 return MUTABLE_CV(sv);
8932 sv = amagic_deref_call(sv, to_cv_amg);
8935 if (SvTYPE(sv) == SVt_PVCV) {
8936 cv = MUTABLE_CV(sv);
8941 else if(SvGETMAGIC(sv), isGV_with_GP(sv))
8942 gv = MUTABLE_GV(sv);
8944 Perl_croak(aTHX_ "Not a subroutine reference");
8946 else if (isGV_with_GP(sv)) {
8947 gv = MUTABLE_GV(sv);
8950 gv = gv_fetchsv_nomg(sv, lref, SVt_PVCV);
8957 /* Some flags to gv_fetchsv mean don't really create the GV */
8958 if (!isGV_with_GP(gv)) {
8963 if (lref & ~GV_ADDMG && !GvCVu(gv)) {
8964 /* XXX this is probably not what they think they're getting.
8965 * It has the same effect as "sub name;", i.e. just a forward
8976 Returns true if the SV has a true value by Perl's rules.
8977 Use the C<SvTRUE> macro instead, which may call C<sv_true()> or may
8978 instead use an in-line version.
8984 Perl_sv_true(pTHX_ register SV *const sv)
8989 register const XPV* const tXpv = (XPV*)SvANY(sv);
8991 (tXpv->xpv_cur > 1 ||
8992 (tXpv->xpv_cur && *sv->sv_u.svu_pv != '0')))
8999 return SvIVX(sv) != 0;
9002 return SvNVX(sv) != 0.0;
9004 return sv_2bool(sv);
9010 =for apidoc sv_pvn_force
9012 Get a sensible string out of the SV somehow.
9013 A private implementation of the C<SvPV_force> macro for compilers which
9014 can't cope with complex macro expressions. Always use the macro instead.
9016 =for apidoc sv_pvn_force_flags
9018 Get a sensible string out of the SV somehow.
9019 If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<sv> if
9020 appropriate, else not. C<sv_pvn_force> and C<sv_pvn_force_nomg> are
9021 implemented in terms of this function.
9022 You normally want to use the various wrapper macros instead: see
9023 C<SvPV_force> and C<SvPV_force_nomg>
9029 Perl_sv_pvn_force_flags(pTHX_ SV *const sv, STRLEN *const lp, const I32 flags)
9033 PERL_ARGS_ASSERT_SV_PVN_FORCE_FLAGS;
9035 if (flags & SV_GMAGIC) SvGETMAGIC(sv);
9036 if (SvTHINKFIRST(sv) && !SvROK(sv))
9037 sv_force_normal_flags(sv, 0);
9047 if (SvREADONLY(sv) && !(flags & SV_MUTABLE_RETURN)) {
9048 const char * const ref = sv_reftype(sv,0);
9050 Perl_croak(aTHX_ "Can't coerce readonly %s to string in %s",
9051 ref, OP_DESC(PL_op));
9053 Perl_croak(aTHX_ "Can't coerce readonly %s to string", ref);
9055 if ((SvTYPE(sv) > SVt_PVLV && SvTYPE(sv) != SVt_PVFM)
9056 || isGV_with_GP(sv))
9057 /* diag_listed_as: Can't coerce %s to %s in %s */
9058 Perl_croak(aTHX_ "Can't coerce %s to string in %s", sv_reftype(sv,0),
9060 s = sv_2pv_flags(sv, &len, flags &~ SV_GMAGIC);
9067 if (s != SvPVX_const(sv)) { /* Almost, but not quite, sv_setpvn() */
9070 SvUPGRADE(sv, SVt_PV); /* Never FALSE */
9071 SvGROW(sv, len + 1);
9072 Move(s,SvPVX(sv),len,char);
9074 SvPVX(sv)[len] = '\0';
9077 SvPOK_on(sv); /* validate pointer */
9079 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
9080 PTR2UV(sv),SvPVX_const(sv)));
9083 (void)SvPOK_only_UTF8(sv);
9084 return SvPVX_mutable(sv);
9088 =for apidoc sv_pvbyten_force
9090 The backend for the C<SvPVbytex_force> macro. Always use the macro
9097 Perl_sv_pvbyten_force(pTHX_ SV *const sv, STRLEN *const lp)
9099 PERL_ARGS_ASSERT_SV_PVBYTEN_FORCE;
9101 sv_pvn_force(sv,lp);
9102 sv_utf8_downgrade(sv,0);
9108 =for apidoc sv_pvutf8n_force
9110 The backend for the C<SvPVutf8x_force> macro. Always use the macro
9117 Perl_sv_pvutf8n_force(pTHX_ SV *const sv, STRLEN *const lp)
9119 PERL_ARGS_ASSERT_SV_PVUTF8N_FORCE;
9121 sv_pvn_force(sv,lp);
9122 sv_utf8_upgrade(sv);
9128 =for apidoc sv_reftype
9130 Returns a string describing what the SV is a reference to.
9136 Perl_sv_reftype(pTHX_ const SV *const sv, const int ob)
9138 PERL_ARGS_ASSERT_SV_REFTYPE;
9139 if (ob && SvOBJECT(sv)) {
9140 return SvPV_nolen_const(sv_ref(NULL, sv, ob));
9143 switch (SvTYPE(sv)) {
9158 case SVt_PVLV: return (char *) (SvROK(sv) ? "REF"
9159 /* tied lvalues should appear to be
9160 * scalars for backwards compatibility */
9161 : (LvTYPE(sv) == 't' || LvTYPE(sv) == 'T')
9162 ? "SCALAR" : "LVALUE");
9163 case SVt_PVAV: return "ARRAY";
9164 case SVt_PVHV: return "HASH";
9165 case SVt_PVCV: return "CODE";
9166 case SVt_PVGV: return (char *) (isGV_with_GP(sv)
9167 ? "GLOB" : "SCALAR");
9168 case SVt_PVFM: return "FORMAT";
9169 case SVt_PVIO: return "IO";
9170 case SVt_BIND: return "BIND";
9171 case SVt_REGEXP: return "REGEXP";
9172 default: return "UNKNOWN";
9180 Returns a SV describing what the SV passed in is a reference to.
9186 Perl_sv_ref(pTHX_ register SV *dst, const SV *const sv, const int ob)
9188 PERL_ARGS_ASSERT_SV_REF;
9191 dst = sv_newmortal();
9193 if (ob && SvOBJECT(sv)) {
9194 HvNAME_get(SvSTASH(sv))
9195 ? sv_sethek(dst, HvNAME_HEK(SvSTASH(sv)))
9196 : sv_setpvn(dst, "__ANON__", 8);
9199 const char * reftype = sv_reftype(sv, 0);
9200 sv_setpv(dst, reftype);
9206 =for apidoc sv_isobject
9208 Returns a boolean indicating whether the SV is an RV pointing to a blessed
9209 object. If the SV is not an RV, or if the object is not blessed, then this
9216 Perl_sv_isobject(pTHX_ SV *sv)
9232 Returns a boolean indicating whether the SV is blessed into the specified
9233 class. This does not check for subtypes; use C<sv_derived_from> to verify
9234 an inheritance relationship.
9240 Perl_sv_isa(pTHX_ SV *sv, const char *const name)
9244 PERL_ARGS_ASSERT_SV_ISA;
9254 hvname = HvNAME_get(SvSTASH(sv));
9258 return strEQ(hvname, name);
9264 Creates a new SV for the RV, C<rv>, to point to. If C<rv> is not an RV then
9265 it will be upgraded to one. If C<classname> is non-null then the new SV will
9266 be blessed in the specified package. The new SV is returned and its
9267 reference count is 1.
9273 Perl_newSVrv(pTHX_ SV *const rv, const char *const classname)
9278 PERL_ARGS_ASSERT_NEWSVRV;
9282 SV_CHECK_THINKFIRST_COW_DROP(rv);
9284 if (SvTYPE(rv) >= SVt_PVMG) {
9285 const U32 refcnt = SvREFCNT(rv);
9289 SvREFCNT(rv) = refcnt;
9291 sv_upgrade(rv, SVt_IV);
9292 } else if (SvROK(rv)) {
9293 SvREFCNT_dec(SvRV(rv));
9295 prepare_SV_for_RV(rv);
9303 HV* const stash = gv_stashpv(classname, GV_ADD);
9304 (void)sv_bless(rv, stash);
9310 =for apidoc sv_setref_pv
9312 Copies a pointer into a new SV, optionally blessing the SV. The C<rv>
9313 argument will be upgraded to an RV. That RV will be modified to point to
9314 the new SV. If the C<pv> argument is NULL then C<PL_sv_undef> will be placed
9315 into the SV. The C<classname> argument indicates the package for the
9316 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9317 will have a reference count of 1, and the RV will be returned.
9319 Do not use with other Perl types such as HV, AV, SV, CV, because those
9320 objects will become corrupted by the pointer copy process.
9322 Note that C<sv_setref_pvn> copies the string while this copies the pointer.
9328 Perl_sv_setref_pv(pTHX_ SV *const rv, const char *const classname, void *const pv)
9332 PERL_ARGS_ASSERT_SV_SETREF_PV;
9335 sv_setsv(rv, &PL_sv_undef);
9339 sv_setiv(newSVrv(rv,classname), PTR2IV(pv));
9344 =for apidoc sv_setref_iv
9346 Copies an integer into a new SV, optionally blessing the SV. The C<rv>
9347 argument will be upgraded to an RV. That RV will be modified to point to
9348 the new SV. The C<classname> argument indicates the package for the
9349 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9350 will have a reference count of 1, and the RV will be returned.
9356 Perl_sv_setref_iv(pTHX_ SV *const rv, const char *const classname, const IV iv)
9358 PERL_ARGS_ASSERT_SV_SETREF_IV;
9360 sv_setiv(newSVrv(rv,classname), iv);
9365 =for apidoc sv_setref_uv
9367 Copies an unsigned integer into a new SV, optionally blessing the SV. The C<rv>
9368 argument will be upgraded to an RV. That RV will be modified to point to
9369 the new SV. The C<classname> argument indicates the package for the
9370 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9371 will have a reference count of 1, and the RV will be returned.
9377 Perl_sv_setref_uv(pTHX_ SV *const rv, const char *const classname, const UV uv)
9379 PERL_ARGS_ASSERT_SV_SETREF_UV;
9381 sv_setuv(newSVrv(rv,classname), uv);
9386 =for apidoc sv_setref_nv
9388 Copies a double into a new SV, optionally blessing the SV. The C<rv>
9389 argument will be upgraded to an RV. That RV will be modified to point to
9390 the new SV. The C<classname> argument indicates the package for the
9391 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9392 will have a reference count of 1, and the RV will be returned.
9398 Perl_sv_setref_nv(pTHX_ SV *const rv, const char *const classname, const NV nv)
9400 PERL_ARGS_ASSERT_SV_SETREF_NV;
9402 sv_setnv(newSVrv(rv,classname), nv);
9407 =for apidoc sv_setref_pvn
9409 Copies a string into a new SV, optionally blessing the SV. The length of the
9410 string must be specified with C<n>. The C<rv> argument will be upgraded to
9411 an RV. That RV will be modified to point to the new SV. The C<classname>
9412 argument indicates the package for the blessing. Set C<classname> to
9413 C<NULL> to avoid the blessing. The new SV will have a reference count
9414 of 1, and the RV will be returned.
9416 Note that C<sv_setref_pv> copies the pointer while this copies the string.
9422 Perl_sv_setref_pvn(pTHX_ SV *const rv, const char *const classname,
9423 const char *const pv, const STRLEN n)
9425 PERL_ARGS_ASSERT_SV_SETREF_PVN;
9427 sv_setpvn(newSVrv(rv,classname), pv, n);
9432 =for apidoc sv_bless
9434 Blesses an SV into a specified package. The SV must be an RV. The package
9435 must be designated by its stash (see C<gv_stashpv()>). The reference count
9436 of the SV is unaffected.
9442 Perl_sv_bless(pTHX_ SV *const sv, HV *const stash)
9447 PERL_ARGS_ASSERT_SV_BLESS;
9450 Perl_croak(aTHX_ "Can't bless non-reference value");
9452 if (SvFLAGS(tmpRef) & (SVs_OBJECT|SVf_READONLY)) {
9453 if (SvIsCOW(tmpRef))
9454 sv_force_normal_flags(tmpRef, 0);
9455 if (SvREADONLY(tmpRef))
9456 Perl_croak_no_modify(aTHX);
9457 if (SvOBJECT(tmpRef)) {
9458 if (SvTYPE(tmpRef) != SVt_PVIO)
9460 SvREFCNT_dec(SvSTASH(tmpRef));
9463 SvOBJECT_on(tmpRef);
9464 if (SvTYPE(tmpRef) != SVt_PVIO)
9466 SvUPGRADE(tmpRef, SVt_PVMG);
9467 SvSTASH_set(tmpRef, MUTABLE_HV(SvREFCNT_inc_simple(stash)));
9469 if(SvSMAGICAL(tmpRef))
9470 if(mg_find(tmpRef, PERL_MAGIC_ext) || mg_find(tmpRef, PERL_MAGIC_uvar))
9478 /* Downgrades a PVGV to a PVMG. If it's actually a PVLV, we leave the type
9479 * as it is after unglobbing it.
9482 PERL_STATIC_INLINE void
9483 S_sv_unglob(pTHX_ SV *const sv, U32 flags)
9488 SV * const temp = flags & SV_COW_DROP_PV ? NULL : sv_newmortal();
9490 PERL_ARGS_ASSERT_SV_UNGLOB;
9492 assert(SvTYPE(sv) == SVt_PVGV || SvTYPE(sv) == SVt_PVLV);
9494 if (!(flags & SV_COW_DROP_PV))
9495 gv_efullname3(temp, MUTABLE_GV(sv), "*");
9498 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
9499 && HvNAME_get(stash))
9500 mro_method_changed_in(stash);
9501 gp_free(MUTABLE_GV(sv));
9504 sv_del_backref(MUTABLE_SV(GvSTASH(sv)), sv);
9508 if (GvNAME_HEK(sv)) {
9509 unshare_hek(GvNAME_HEK(sv));
9511 isGV_with_GP_off(sv);
9513 if(SvTYPE(sv) == SVt_PVGV) {
9514 /* need to keep SvANY(sv) in the right arena */
9515 xpvmg = new_XPVMG();
9516 StructCopy(SvANY(sv), xpvmg, XPVMG);
9517 del_XPVGV(SvANY(sv));
9520 SvFLAGS(sv) &= ~SVTYPEMASK;
9521 SvFLAGS(sv) |= SVt_PVMG;
9524 /* Intentionally not calling any local SET magic, as this isn't so much a
9525 set operation as merely an internal storage change. */
9526 if (flags & SV_COW_DROP_PV) SvOK_off(sv);
9527 else sv_setsv_flags(sv, temp, 0);
9529 if ((const GV *)sv == PL_last_in_gv)
9530 PL_last_in_gv = NULL;
9531 else if ((const GV *)sv == PL_statgv)
9536 =for apidoc sv_unref_flags
9538 Unsets the RV status of the SV, and decrements the reference count of
9539 whatever was being referenced by the RV. This can almost be thought of
9540 as a reversal of C<newSVrv>. The C<cflags> argument can contain
9541 C<SV_IMMEDIATE_UNREF> to force the reference count to be decremented
9542 (otherwise the decrementing is conditional on the reference count being
9543 different from one or the reference being a readonly SV).
9550 Perl_sv_unref_flags(pTHX_ SV *const ref, const U32 flags)
9552 SV* const target = SvRV(ref);
9554 PERL_ARGS_ASSERT_SV_UNREF_FLAGS;
9556 if (SvWEAKREF(ref)) {
9557 sv_del_backref(target, ref);
9559 SvRV_set(ref, NULL);
9562 SvRV_set(ref, NULL);
9564 /* You can't have a || SvREADONLY(target) here, as $a = $$a, where $a was
9565 assigned to as BEGIN {$a = \"Foo"} will fail. */
9566 if (SvREFCNT(target) != 1 || (flags & SV_IMMEDIATE_UNREF))
9567 SvREFCNT_dec(target);
9568 else /* XXX Hack, but hard to make $a=$a->[1] work otherwise */
9569 sv_2mortal(target); /* Schedule for freeing later */
9573 =for apidoc sv_untaint
9575 Untaint an SV. Use C<SvTAINTED_off> instead.
9581 Perl_sv_untaint(pTHX_ SV *const sv)
9583 PERL_ARGS_ASSERT_SV_UNTAINT;
9585 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
9586 MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
9593 =for apidoc sv_tainted
9595 Test an SV for taintedness. Use C<SvTAINTED> instead.
9601 Perl_sv_tainted(pTHX_ SV *const sv)
9603 PERL_ARGS_ASSERT_SV_TAINTED;
9605 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
9606 const MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
9607 if (mg && (mg->mg_len & 1) )
9614 =for apidoc sv_setpviv
9616 Copies an integer into the given SV, also updating its string value.
9617 Does not handle 'set' magic. See C<sv_setpviv_mg>.
9623 Perl_sv_setpviv(pTHX_ SV *const sv, const IV iv)
9625 char buf[TYPE_CHARS(UV)];
9627 char * const ptr = uiv_2buf(buf, iv, 0, 0, &ebuf);
9629 PERL_ARGS_ASSERT_SV_SETPVIV;
9631 sv_setpvn(sv, ptr, ebuf - ptr);
9635 =for apidoc sv_setpviv_mg
9637 Like C<sv_setpviv>, but also handles 'set' magic.
9643 Perl_sv_setpviv_mg(pTHX_ SV *const sv, const IV iv)
9645 PERL_ARGS_ASSERT_SV_SETPVIV_MG;
9651 #if defined(PERL_IMPLICIT_CONTEXT)
9653 /* pTHX_ magic can't cope with varargs, so this is a no-context
9654 * version of the main function, (which may itself be aliased to us).
9655 * Don't access this version directly.
9659 Perl_sv_setpvf_nocontext(SV *const sv, const char *const pat, ...)
9664 PERL_ARGS_ASSERT_SV_SETPVF_NOCONTEXT;
9666 va_start(args, pat);
9667 sv_vsetpvf(sv, pat, &args);
9671 /* pTHX_ magic can't cope with varargs, so this is a no-context
9672 * version of the main function, (which may itself be aliased to us).
9673 * Don't access this version directly.
9677 Perl_sv_setpvf_mg_nocontext(SV *const sv, const char *const pat, ...)
9682 PERL_ARGS_ASSERT_SV_SETPVF_MG_NOCONTEXT;
9684 va_start(args, pat);
9685 sv_vsetpvf_mg(sv, pat, &args);
9691 =for apidoc sv_setpvf
9693 Works like C<sv_catpvf> but copies the text into the SV instead of
9694 appending it. Does not handle 'set' magic. See C<sv_setpvf_mg>.
9700 Perl_sv_setpvf(pTHX_ SV *const sv, const char *const pat, ...)
9704 PERL_ARGS_ASSERT_SV_SETPVF;
9706 va_start(args, pat);
9707 sv_vsetpvf(sv, pat, &args);
9712 =for apidoc sv_vsetpvf
9714 Works like C<sv_vcatpvf> but copies the text into the SV instead of
9715 appending it. Does not handle 'set' magic. See C<sv_vsetpvf_mg>.
9717 Usually used via its frontend C<sv_setpvf>.
9723 Perl_sv_vsetpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9725 PERL_ARGS_ASSERT_SV_VSETPVF;
9727 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9731 =for apidoc sv_setpvf_mg
9733 Like C<sv_setpvf>, but also handles 'set' magic.
9739 Perl_sv_setpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
9743 PERL_ARGS_ASSERT_SV_SETPVF_MG;
9745 va_start(args, pat);
9746 sv_vsetpvf_mg(sv, pat, &args);
9751 =for apidoc sv_vsetpvf_mg
9753 Like C<sv_vsetpvf>, but also handles 'set' magic.
9755 Usually used via its frontend C<sv_setpvf_mg>.
9761 Perl_sv_vsetpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9763 PERL_ARGS_ASSERT_SV_VSETPVF_MG;
9765 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9769 #if defined(PERL_IMPLICIT_CONTEXT)
9771 /* pTHX_ magic can't cope with varargs, so this is a no-context
9772 * version of the main function, (which may itself be aliased to us).
9773 * Don't access this version directly.
9777 Perl_sv_catpvf_nocontext(SV *const sv, const char *const pat, ...)
9782 PERL_ARGS_ASSERT_SV_CATPVF_NOCONTEXT;
9784 va_start(args, pat);
9785 sv_vcatpvf(sv, pat, &args);
9789 /* pTHX_ magic can't cope with varargs, so this is a no-context
9790 * version of the main function, (which may itself be aliased to us).
9791 * Don't access this version directly.
9795 Perl_sv_catpvf_mg_nocontext(SV *const sv, const char *const pat, ...)
9800 PERL_ARGS_ASSERT_SV_CATPVF_MG_NOCONTEXT;
9802 va_start(args, pat);
9803 sv_vcatpvf_mg(sv, pat, &args);
9809 =for apidoc sv_catpvf
9811 Processes its arguments like C<sprintf> and appends the formatted
9812 output to an SV. If the appended data contains "wide" characters
9813 (including, but not limited to, SVs with a UTF-8 PV formatted with %s,
9814 and characters >255 formatted with %c), the original SV might get
9815 upgraded to UTF-8. Handles 'get' magic, but not 'set' magic. See
9816 C<sv_catpvf_mg>. If the original SV was UTF-8, the pattern should be
9817 valid UTF-8; if the original SV was bytes, the pattern should be too.
9822 Perl_sv_catpvf(pTHX_ SV *const sv, const char *const pat, ...)
9826 PERL_ARGS_ASSERT_SV_CATPVF;
9828 va_start(args, pat);
9829 sv_vcatpvf(sv, pat, &args);
9834 =for apidoc sv_vcatpvf
9836 Processes its arguments like C<vsprintf> and appends the formatted output
9837 to an SV. Does not handle 'set' magic. See C<sv_vcatpvf_mg>.
9839 Usually used via its frontend C<sv_catpvf>.
9845 Perl_sv_vcatpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9847 PERL_ARGS_ASSERT_SV_VCATPVF;
9849 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9853 =for apidoc sv_catpvf_mg
9855 Like C<sv_catpvf>, but also handles 'set' magic.
9861 Perl_sv_catpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
9865 PERL_ARGS_ASSERT_SV_CATPVF_MG;
9867 va_start(args, pat);
9868 sv_vcatpvf_mg(sv, pat, &args);
9873 =for apidoc sv_vcatpvf_mg
9875 Like C<sv_vcatpvf>, but also handles 'set' magic.
9877 Usually used via its frontend C<sv_catpvf_mg>.
9883 Perl_sv_vcatpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9885 PERL_ARGS_ASSERT_SV_VCATPVF_MG;
9887 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9892 =for apidoc sv_vsetpvfn
9894 Works like C<sv_vcatpvfn> but copies the text into the SV instead of
9897 Usually used via one of its frontends C<sv_vsetpvf> and C<sv_vsetpvf_mg>.
9903 Perl_sv_vsetpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
9904 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
9906 PERL_ARGS_ASSERT_SV_VSETPVFN;
9909 sv_vcatpvfn_flags(sv, pat, patlen, args, svargs, svmax, maybe_tainted, 0);
9914 * Warn of missing argument to sprintf, and then return a defined value
9915 * to avoid inappropriate "use of uninit" warnings [perl #71000].
9917 #define WARN_MISSING WARN_UNINITIALIZED /* Not sure we want a new category */
9919 S_vcatpvfn_missing_argument(pTHX) {
9920 if (ckWARN(WARN_MISSING)) {
9921 Perl_warner(aTHX_ packWARN(WARN_MISSING), "Missing argument in %s",
9922 PL_op ? OP_DESC(PL_op) : "sv_vcatpvfn()");
9929 S_expect_number(pTHX_ char **const pattern)
9934 PERL_ARGS_ASSERT_EXPECT_NUMBER;
9936 switch (**pattern) {
9937 case '1': case '2': case '3':
9938 case '4': case '5': case '6':
9939 case '7': case '8': case '9':
9940 var = *(*pattern)++ - '0';
9941 while (isDIGIT(**pattern)) {
9942 const I32 tmp = var * 10 + (*(*pattern)++ - '0');
9944 Perl_croak(aTHX_ "Integer overflow in format string for %s", (PL_op ? OP_DESC(PL_op) : "sv_vcatpvfn"));
9952 S_F0convert(NV nv, char *const endbuf, STRLEN *const len)
9954 const int neg = nv < 0;
9957 PERL_ARGS_ASSERT_F0CONVERT;
9965 if (uv & 1 && uv == nv)
9966 uv--; /* Round to even */
9968 const unsigned dig = uv % 10;
9981 =for apidoc sv_vcatpvfn
9983 =for apidoc sv_vcatpvfn_flags
9985 Processes its arguments like C<vsprintf> and appends the formatted output
9986 to an SV. Uses an array of SVs if the C style variable argument list is
9987 missing (NULL). When running with taint checks enabled, indicates via
9988 C<maybe_tainted> if results are untrustworthy (often due to the use of
9991 If called as C<sv_vcatpvfn> or flags include C<SV_GMAGIC>, calls get magic.
9993 Usually used via one of its frontends C<sv_vcatpvf> and C<sv_vcatpvf_mg>.
9998 #define VECTORIZE_ARGS vecsv = va_arg(*args, SV*);\
9999 vecstr = (U8*)SvPV_const(vecsv,veclen);\
10000 vec_utf8 = DO_UTF8(vecsv);
10002 /* XXX maybe_tainted is never assigned to, so the doc above is lying. */
10005 Perl_sv_vcatpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
10006 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
10008 PERL_ARGS_ASSERT_SV_VCATPVFN;
10010 sv_vcatpvfn_flags(sv, pat, patlen, args, svargs, svmax, maybe_tainted, SV_GMAGIC|SV_SMAGIC);
10014 Perl_sv_vcatpvfn_flags(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
10015 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted,
10021 const char *patend;
10024 static const char nullstr[] = "(null)";
10026 bool has_utf8 = DO_UTF8(sv); /* has the result utf8? */
10027 const bool pat_utf8 = has_utf8; /* the pattern is in utf8? */
10029 /* Times 4: a decimal digit takes more than 3 binary digits.
10030 * NV_DIG: mantissa takes than many decimal digits.
10031 * Plus 32: Playing safe. */
10032 char ebuf[IV_DIG * 4 + NV_DIG + 32];
10033 /* large enough for "%#.#f" --chip */
10034 /* what about long double NVs? --jhi */
10036 PERL_ARGS_ASSERT_SV_VCATPVFN_FLAGS;
10037 PERL_UNUSED_ARG(maybe_tainted);
10039 if (flags & SV_GMAGIC)
10042 /* no matter what, this is a string now */
10043 (void)SvPV_force_nomg(sv, origlen);
10045 /* special-case "", "%s", and "%-p" (SVf - see below) */
10048 if (patlen == 2 && pat[0] == '%' && pat[1] == 's') {
10050 const char * const s = va_arg(*args, char*);
10051 sv_catpv_nomg(sv, s ? s : nullstr);
10053 else if (svix < svmax) {
10054 /* we want get magic on the source but not the target. sv_catsv can't do that, though */
10055 SvGETMAGIC(*svargs);
10056 sv_catsv_nomg(sv, *svargs);
10059 S_vcatpvfn_missing_argument(aTHX);
10062 if (args && patlen == 3 && pat[0] == '%' &&
10063 pat[1] == '-' && pat[2] == 'p') {
10064 argsv = MUTABLE_SV(va_arg(*args, void*));
10065 sv_catsv_nomg(sv, argsv);
10069 #ifndef USE_LONG_DOUBLE
10070 /* special-case "%.<number>[gf]" */
10071 if ( !args && patlen <= 5 && pat[0] == '%' && pat[1] == '.'
10072 && (pat[patlen-1] == 'g' || pat[patlen-1] == 'f') ) {
10073 unsigned digits = 0;
10077 while (*pp >= '0' && *pp <= '9')
10078 digits = 10 * digits + (*pp++ - '0');
10079 if (pp - pat == (int)patlen - 1 && svix < svmax) {
10080 const NV nv = SvNV(*svargs);
10082 /* Add check for digits != 0 because it seems that some
10083 gconverts are buggy in this case, and we don't yet have
10084 a Configure test for this. */
10085 if (digits && digits < sizeof(ebuf) - NV_DIG - 10) {
10086 /* 0, point, slack */
10087 Gconvert(nv, (int)digits, 0, ebuf);
10088 sv_catpv_nomg(sv, ebuf);
10089 if (*ebuf) /* May return an empty string for digits==0 */
10092 } else if (!digits) {
10095 if ((p = F0convert(nv, ebuf + sizeof ebuf, &l))) {
10096 sv_catpvn_nomg(sv, p, l);
10102 #endif /* !USE_LONG_DOUBLE */
10104 if (!args && svix < svmax && DO_UTF8(*svargs))
10107 patend = (char*)pat + patlen;
10108 for (p = (char*)pat; p < patend; p = q) {
10111 bool vectorize = FALSE;
10112 bool vectorarg = FALSE;
10113 bool vec_utf8 = FALSE;
10119 bool has_precis = FALSE;
10121 const I32 osvix = svix;
10122 bool is_utf8 = FALSE; /* is this item utf8? */
10123 #ifdef HAS_LDBL_SPRINTF_BUG
10124 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
10125 with sfio - Allen <allens@cpan.org> */
10126 bool fix_ldbl_sprintf_bug = FALSE;
10130 U8 utf8buf[UTF8_MAXBYTES+1];
10131 STRLEN esignlen = 0;
10133 const char *eptr = NULL;
10134 const char *fmtstart;
10137 const U8 *vecstr = NULL;
10144 /* we need a long double target in case HAS_LONG_DOUBLE but
10145 not USE_LONG_DOUBLE
10147 #if defined(HAS_LONG_DOUBLE) && LONG_DOUBLESIZE > DOUBLESIZE
10155 const char *dotstr = ".";
10156 STRLEN dotstrlen = 1;
10157 I32 efix = 0; /* explicit format parameter index */
10158 I32 ewix = 0; /* explicit width index */
10159 I32 epix = 0; /* explicit precision index */
10160 I32 evix = 0; /* explicit vector index */
10161 bool asterisk = FALSE;
10163 /* echo everything up to the next format specification */
10164 for (q = p; q < patend && *q != '%'; ++q) ;
10166 if (has_utf8 && !pat_utf8)
10167 sv_catpvn_nomg_utf8_upgrade(sv, p, q - p, nsv);
10169 sv_catpvn_nomg(sv, p, q - p);
10178 We allow format specification elements in this order:
10179 \d+\$ explicit format parameter index
10181 v|\*(\d+\$)?v vector with optional (optionally specified) arg
10182 0 flag (as above): repeated to allow "v02"
10183 \d+|\*(\d+\$)? width using optional (optionally specified) arg
10184 \.(\d*|\*(\d+\$)?) precision using optional (optionally specified) arg
10186 [%bcdefginopsuxDFOUX] format (mandatory)
10191 As of perl5.9.3, printf format checking is on by default.
10192 Internally, perl uses %p formats to provide an escape to
10193 some extended formatting. This block deals with those
10194 extensions: if it does not match, (char*)q is reset and
10195 the normal format processing code is used.
10197 Currently defined extensions are:
10198 %p include pointer address (standard)
10199 %-p (SVf) include an SV (previously %_)
10200 %-<num>p include an SV with precision <num>
10202 %3p include a HEK with precision of 256
10203 %<num>p (where num != 2 or 3) reserved for future
10206 Robin Barker 2005-07-14 (but modified since)
10208 %1p (VDf) removed. RMB 2007-10-19
10215 n = expect_number(&q);
10217 if (sv) { /* SVf */
10222 argsv = MUTABLE_SV(va_arg(*args, void*));
10223 eptr = SvPV_const(argsv, elen);
10224 if (DO_UTF8(argsv))
10228 else if (n==2 || n==3) { /* HEKf */
10229 HEK * const hek = va_arg(*args, HEK *);
10230 eptr = HEK_KEY(hek);
10231 elen = HEK_LEN(hek);
10232 if (HEK_UTF8(hek)) is_utf8 = TRUE;
10233 if (n==3) precis = 256, has_precis = TRUE;
10237 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL),
10238 "internal %%<num>p might conflict with future printf extensions");
10244 if ( (width = expect_number(&q)) ) {
10259 if (plus == '+' && *q == ' ') /* '+' over ' ' */
10288 if ( (ewix = expect_number(&q)) )
10297 if ((vectorarg = asterisk)) {
10310 width = expect_number(&q);
10313 if (vectorize && vectorarg) {
10314 /* vectorizing, but not with the default "." */
10316 vecsv = va_arg(*args, SV*);
10318 vecsv = (evix > 0 && evix <= svmax)
10319 ? svargs[evix-1] : S_vcatpvfn_missing_argument(aTHX);
10321 vecsv = svix < svmax
10322 ? svargs[svix++] : S_vcatpvfn_missing_argument(aTHX);
10324 dotstr = SvPV_const(vecsv, dotstrlen);
10325 /* Keep the DO_UTF8 test *after* the SvPV call, else things go
10326 bad with tied or overloaded values that return UTF8. */
10327 if (DO_UTF8(vecsv))
10329 else if (has_utf8) {
10330 vecsv = sv_mortalcopy(vecsv);
10331 sv_utf8_upgrade(vecsv);
10332 dotstr = SvPV_const(vecsv, dotstrlen);
10339 i = va_arg(*args, int);
10341 i = (ewix ? ewix <= svmax : svix < svmax) ?
10342 SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
10344 width = (i < 0) ? -i : i;
10354 if ( ((epix = expect_number(&q))) && (*q++ != '$') )
10356 /* XXX: todo, support specified precision parameter */
10360 i = va_arg(*args, int);
10362 i = (ewix ? ewix <= svmax : svix < svmax)
10363 ? SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
10365 has_precis = !(i < 0);
10369 while (isDIGIT(*q))
10370 precis = precis * 10 + (*q++ - '0');
10379 else if (efix ? (efix > 0 && efix <= svmax) : svix < svmax) {
10380 vecsv = svargs[efix ? efix-1 : svix++];
10381 vecstr = (U8*)SvPV_const(vecsv,veclen);
10382 vec_utf8 = DO_UTF8(vecsv);
10384 /* if this is a version object, we need to convert
10385 * back into v-string notation and then let the
10386 * vectorize happen normally
10388 if (sv_isobject(vecsv) && sv_derived_from(vecsv, "version")) {
10389 char *version = savesvpv(vecsv);
10390 if ( hv_exists(MUTABLE_HV(SvRV(vecsv)), "alpha", 5 ) ) {
10391 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
10392 "vector argument not supported with alpha versions");
10395 vecsv = sv_newmortal();
10396 scan_vstring(version, version + veclen, vecsv);
10397 vecstr = (U8*)SvPV_const(vecsv, veclen);
10398 vec_utf8 = DO_UTF8(vecsv);
10412 case 'I': /* Ix, I32x, and I64x */
10413 # ifdef USE_64_BIT_INT
10414 if (q[1] == '6' && q[2] == '4') {
10420 if (q[1] == '3' && q[2] == '2') {
10424 # ifdef USE_64_BIT_INT
10430 #if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
10442 #if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
10443 if (*q == 'l') { /* lld, llf */
10452 if (*++q == 'h') { /* hhd, hhu */
10481 if (!vectorize && !args) {
10483 const I32 i = efix-1;
10484 argsv = (i >= 0 && i < svmax)
10485 ? svargs[i] : S_vcatpvfn_missing_argument(aTHX);
10487 argsv = (svix >= 0 && svix < svmax)
10488 ? svargs[svix++] : S_vcatpvfn_missing_argument(aTHX);
10492 switch (c = *q++) {
10499 uv = (args) ? va_arg(*args, int) : SvIV(argsv);
10501 (!UNI_IS_INVARIANT(uv) && SvUTF8(sv)))
10503 eptr = (char*)utf8buf;
10504 elen = uvchr_to_utf8((U8*)eptr, uv) - utf8buf;
10518 eptr = va_arg(*args, char*);
10520 elen = strlen(eptr);
10522 eptr = (char *)nullstr;
10523 elen = sizeof nullstr - 1;
10527 eptr = SvPV_const(argsv, elen);
10528 if (DO_UTF8(argsv)) {
10529 STRLEN old_precis = precis;
10530 if (has_precis && precis < elen) {
10531 STRLEN ulen = sv_len_utf8(argsv);
10532 I32 p = precis > ulen ? ulen : precis;
10533 sv_pos_u2b(argsv, &p, 0); /* sticks at end */
10536 if (width) { /* fudge width (can't fudge elen) */
10537 if (has_precis && precis < elen)
10538 width += precis - old_precis;
10540 width += elen - sv_len_utf8(argsv);
10547 if (has_precis && precis < elen)
10554 if (alt || vectorize)
10556 uv = PTR2UV(args ? va_arg(*args, void*) : argsv);
10577 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
10586 esignbuf[esignlen++] = plus;
10590 case 'c': iv = (char)va_arg(*args, int); break;
10591 case 'h': iv = (short)va_arg(*args, int); break;
10592 case 'l': iv = va_arg(*args, long); break;
10593 case 'V': iv = va_arg(*args, IV); break;
10594 case 'z': iv = va_arg(*args, SSize_t); break;
10595 case 't': iv = va_arg(*args, ptrdiff_t); break;
10596 default: iv = va_arg(*args, int); break;
10598 case 'j': iv = va_arg(*args, intmax_t); break;
10602 iv = va_arg(*args, Quad_t); break;
10609 IV tiv = SvIV(argsv); /* work around GCC bug #13488 */
10611 case 'c': iv = (char)tiv; break;
10612 case 'h': iv = (short)tiv; break;
10613 case 'l': iv = (long)tiv; break;
10615 default: iv = tiv; break;
10618 iv = (Quad_t)tiv; break;
10624 if ( !vectorize ) /* we already set uv above */
10629 esignbuf[esignlen++] = plus;
10633 esignbuf[esignlen++] = '-';
10677 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
10688 case 'c': uv = (unsigned char)va_arg(*args, unsigned); break;
10689 case 'h': uv = (unsigned short)va_arg(*args, unsigned); break;
10690 case 'l': uv = va_arg(*args, unsigned long); break;
10691 case 'V': uv = va_arg(*args, UV); break;
10692 case 'z': uv = va_arg(*args, Size_t); break;
10693 case 't': uv = va_arg(*args, ptrdiff_t); break; /* will sign extend, but there is no uptrdiff_t, so oh well */
10695 case 'j': uv = va_arg(*args, uintmax_t); break;
10697 default: uv = va_arg(*args, unsigned); break;
10700 uv = va_arg(*args, Uquad_t); break;
10707 UV tuv = SvUV(argsv); /* work around GCC bug #13488 */
10709 case 'c': uv = (unsigned char)tuv; break;
10710 case 'h': uv = (unsigned short)tuv; break;
10711 case 'l': uv = (unsigned long)tuv; break;
10713 default: uv = tuv; break;
10716 uv = (Uquad_t)tuv; break;
10725 char *ptr = ebuf + sizeof ebuf;
10726 bool tempalt = uv ? alt : FALSE; /* Vectors can't change alt */
10732 p = (char *)((c == 'X') ? PL_hexdigit + 16 : PL_hexdigit);
10736 } while (uv >>= 4);
10738 esignbuf[esignlen++] = '0';
10739 esignbuf[esignlen++] = c; /* 'x' or 'X' */
10745 *--ptr = '0' + dig;
10746 } while (uv >>= 3);
10747 if (alt && *ptr != '0')
10753 *--ptr = '0' + dig;
10754 } while (uv >>= 1);
10756 esignbuf[esignlen++] = '0';
10757 esignbuf[esignlen++] = c;
10760 default: /* it had better be ten or less */
10763 *--ptr = '0' + dig;
10764 } while (uv /= base);
10767 elen = (ebuf + sizeof ebuf) - ptr;
10771 zeros = precis - elen;
10772 else if (precis == 0 && elen == 1 && *eptr == '0'
10773 && !(base == 8 && alt)) /* "%#.0o" prints "0" */
10776 /* a precision nullifies the 0 flag. */
10783 /* FLOATING POINT */
10786 c = 'f'; /* maybe %F isn't supported here */
10788 case 'e': case 'E':
10790 case 'g': case 'G':
10794 /* This is evil, but floating point is even more evil */
10796 /* for SV-style calling, we can only get NV
10797 for C-style calling, we assume %f is double;
10798 for simplicity we allow any of %Lf, %llf, %qf for long double
10802 #if defined(USE_LONG_DOUBLE)
10806 /* [perl #20339] - we should accept and ignore %lf rather than die */
10810 #if defined(USE_LONG_DOUBLE)
10811 intsize = args ? 0 : 'q';
10815 #if defined(HAS_LONG_DOUBLE)
10828 /* now we need (long double) if intsize == 'q', else (double) */
10830 #if LONG_DOUBLESIZE > DOUBLESIZE
10832 va_arg(*args, long double) :
10833 va_arg(*args, double)
10835 va_arg(*args, double)
10840 /* nv * 0 will be NaN for NaN, +Inf and -Inf, and 0 for anything
10841 else. frexp() has some unspecified behaviour for those three */
10842 if (c != 'e' && c != 'E' && (nv * 0) == 0) {
10844 /* FIXME: if HAS_LONG_DOUBLE but not USE_LONG_DOUBLE this
10845 will cast our (long double) to (double) */
10846 (void)Perl_frexp(nv, &i);
10847 if (i == PERL_INT_MIN)
10848 Perl_die(aTHX_ "panic: frexp");
10850 need = BIT_DIGITS(i);
10852 need += has_precis ? precis : 6; /* known default */
10857 #ifdef HAS_LDBL_SPRINTF_BUG
10858 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
10859 with sfio - Allen <allens@cpan.org> */
10862 # define MY_DBL_MAX DBL_MAX
10863 # else /* XXX guessing! HUGE_VAL may be defined as infinity, so not using */
10864 # if DOUBLESIZE >= 8
10865 # define MY_DBL_MAX 1.7976931348623157E+308L
10867 # define MY_DBL_MAX 3.40282347E+38L
10871 # ifdef HAS_LDBL_SPRINTF_BUG_LESS1 /* only between -1L & 1L - Allen */
10872 # define MY_DBL_MAX_BUG 1L
10874 # define MY_DBL_MAX_BUG MY_DBL_MAX
10878 # define MY_DBL_MIN DBL_MIN
10879 # else /* XXX guessing! -Allen */
10880 # if DOUBLESIZE >= 8
10881 # define MY_DBL_MIN 2.2250738585072014E-308L
10883 # define MY_DBL_MIN 1.17549435E-38L
10887 if ((intsize == 'q') && (c == 'f') &&
10888 ((nv < MY_DBL_MAX_BUG) && (nv > -MY_DBL_MAX_BUG)) &&
10889 (need < DBL_DIG)) {
10890 /* it's going to be short enough that
10891 * long double precision is not needed */
10893 if ((nv <= 0L) && (nv >= -0L))
10894 fix_ldbl_sprintf_bug = TRUE; /* 0 is 0 - easiest */
10896 /* would use Perl_fp_class as a double-check but not
10897 * functional on IRIX - see perl.h comments */
10899 if ((nv >= MY_DBL_MIN) || (nv <= -MY_DBL_MIN)) {
10900 /* It's within the range that a double can represent */
10901 #if defined(DBL_MAX) && !defined(DBL_MIN)
10902 if ((nv >= ((long double)1/DBL_MAX)) ||
10903 (nv <= (-(long double)1/DBL_MAX)))
10905 fix_ldbl_sprintf_bug = TRUE;
10908 if (fix_ldbl_sprintf_bug == TRUE) {
10918 # undef MY_DBL_MAX_BUG
10921 #endif /* HAS_LDBL_SPRINTF_BUG */
10923 need += 20; /* fudge factor */
10924 if (PL_efloatsize < need) {
10925 Safefree(PL_efloatbuf);
10926 PL_efloatsize = need + 20; /* more fudge */
10927 Newx(PL_efloatbuf, PL_efloatsize, char);
10928 PL_efloatbuf[0] = '\0';
10931 if ( !(width || left || plus || alt) && fill != '0'
10932 && has_precis && intsize != 'q' ) { /* Shortcuts */
10933 /* See earlier comment about buggy Gconvert when digits,
10935 if ( c == 'g' && precis) {
10936 Gconvert((NV)nv, (int)precis, 0, PL_efloatbuf);
10937 /* May return an empty string for digits==0 */
10938 if (*PL_efloatbuf) {
10939 elen = strlen(PL_efloatbuf);
10940 goto float_converted;
10942 } else if ( c == 'f' && !precis) {
10943 if ((eptr = F0convert(nv, ebuf + sizeof ebuf, &elen)))
10948 char *ptr = ebuf + sizeof ebuf;
10951 /* FIXME: what to do if HAS_LONG_DOUBLE but not PERL_PRIfldbl? */
10952 #if defined(HAS_LONG_DOUBLE) && defined(PERL_PRIfldbl)
10953 if (intsize == 'q') {
10954 /* Copy the one or more characters in a long double
10955 * format before the 'base' ([efgEFG]) character to
10956 * the format string. */
10957 static char const prifldbl[] = PERL_PRIfldbl;
10958 char const *p = prifldbl + sizeof(prifldbl) - 3;
10959 while (p >= prifldbl) { *--ptr = *p--; }
10964 do { *--ptr = '0' + (base % 10); } while (base /= 10);
10969 do { *--ptr = '0' + (base % 10); } while (base /= 10);
10981 /* No taint. Otherwise we are in the strange situation
10982 * where printf() taints but print($float) doesn't.
10984 #if defined(HAS_LONG_DOUBLE)
10985 elen = ((intsize == 'q')
10986 ? my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, nv)
10987 : my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, (double)nv));
10989 elen = my_sprintf(PL_efloatbuf, ptr, nv);
10993 eptr = PL_efloatbuf;
11001 i = SvCUR(sv) - origlen;
11004 case 'c': *(va_arg(*args, char*)) = i; break;
11005 case 'h': *(va_arg(*args, short*)) = i; break;
11006 default: *(va_arg(*args, int*)) = i; break;
11007 case 'l': *(va_arg(*args, long*)) = i; break;
11008 case 'V': *(va_arg(*args, IV*)) = i; break;
11009 case 'z': *(va_arg(*args, SSize_t*)) = i; break;
11010 case 't': *(va_arg(*args, ptrdiff_t*)) = i; break;
11012 case 'j': *(va_arg(*args, intmax_t*)) = i; break;
11016 *(va_arg(*args, Quad_t*)) = i; break;
11023 sv_setuv_mg(argsv, has_utf8 ? (UV)sv_len_utf8(sv) : (UV)i);
11024 continue; /* not "break" */
11031 && (PL_op->op_type == OP_PRTF || PL_op->op_type == OP_SPRINTF)
11032 && ckWARN(WARN_PRINTF))
11034 SV * const msg = sv_newmortal();
11035 Perl_sv_setpvf(aTHX_ msg, "Invalid conversion in %sprintf: ",
11036 (PL_op->op_type == OP_PRTF) ? "" : "s");
11037 if (fmtstart < patend) {
11038 const char * const fmtend = q < patend ? q : patend;
11040 sv_catpvs(msg, "\"%");
11041 for (f = fmtstart; f < fmtend; f++) {
11043 sv_catpvn_nomg(msg, f, 1);
11045 Perl_sv_catpvf(aTHX_ msg,
11046 "\\%03"UVof, (UV)*f & 0xFF);
11049 sv_catpvs(msg, "\"");
11051 sv_catpvs(msg, "end of string");
11053 Perl_warner(aTHX_ packWARN(WARN_PRINTF), "%"SVf, SVfARG(msg)); /* yes, this is reentrant */
11056 /* output mangled stuff ... */
11062 /* ... right here, because formatting flags should not apply */
11063 SvGROW(sv, SvCUR(sv) + elen + 1);
11065 Copy(eptr, p, elen, char);
11068 SvCUR_set(sv, p - SvPVX_const(sv));
11070 continue; /* not "break" */
11073 if (is_utf8 != has_utf8) {
11076 sv_utf8_upgrade(sv);
11079 const STRLEN old_elen = elen;
11080 SV * const nsv = newSVpvn_flags(eptr, elen, SVs_TEMP);
11081 sv_utf8_upgrade(nsv);
11082 eptr = SvPVX_const(nsv);
11085 if (width) { /* fudge width (can't fudge elen) */
11086 width += elen - old_elen;
11092 have = esignlen + zeros + elen;
11094 Perl_croak_nocontext("%s", PL_memory_wrap);
11096 need = (have > width ? have : width);
11099 if (need >= (((STRLEN)~0) - SvCUR(sv) - dotstrlen - 1))
11100 Perl_croak_nocontext("%s", PL_memory_wrap);
11101 SvGROW(sv, SvCUR(sv) + need + dotstrlen + 1);
11103 if (esignlen && fill == '0') {
11105 for (i = 0; i < (int)esignlen; i++)
11106 *p++ = esignbuf[i];
11108 if (gap && !left) {
11109 memset(p, fill, gap);
11112 if (esignlen && fill != '0') {
11114 for (i = 0; i < (int)esignlen; i++)
11115 *p++ = esignbuf[i];
11119 for (i = zeros; i; i--)
11123 Copy(eptr, p, elen, char);
11127 memset(p, ' ', gap);
11132 Copy(dotstr, p, dotstrlen, char);
11136 vectorize = FALSE; /* done iterating over vecstr */
11143 SvCUR_set(sv, p - SvPVX_const(sv));
11152 /* =========================================================================
11154 =head1 Cloning an interpreter
11156 All the macros and functions in this section are for the private use of
11157 the main function, perl_clone().
11159 The foo_dup() functions make an exact copy of an existing foo thingy.
11160 During the course of a cloning, a hash table is used to map old addresses
11161 to new addresses. The table is created and manipulated with the
11162 ptr_table_* functions.
11166 * =========================================================================*/
11169 #if defined(USE_ITHREADS)
11171 /* XXX Remove this so it doesn't have to go thru the macro and return for nothing */
11172 #ifndef GpREFCNT_inc
11173 # define GpREFCNT_inc(gp) ((gp) ? (++(gp)->gp_refcnt, (gp)) : (GP*)NULL)
11177 /* Certain cases in Perl_ss_dup have been merged, by relying on the fact
11178 that currently av_dup, gv_dup and hv_dup are the same as sv_dup.
11179 If this changes, please unmerge ss_dup.
11180 Likewise, sv_dup_inc_multiple() relies on this fact. */
11181 #define sv_dup_inc_NN(s,t) SvREFCNT_inc_NN(sv_dup_inc(s,t))
11182 #define av_dup(s,t) MUTABLE_AV(sv_dup((const SV *)s,t))
11183 #define av_dup_inc(s,t) MUTABLE_AV(sv_dup_inc((const SV *)s,t))
11184 #define hv_dup(s,t) MUTABLE_HV(sv_dup((const SV *)s,t))
11185 #define hv_dup_inc(s,t) MUTABLE_HV(sv_dup_inc((const SV *)s,t))
11186 #define cv_dup(s,t) MUTABLE_CV(sv_dup((const SV *)s,t))
11187 #define cv_dup_inc(s,t) MUTABLE_CV(sv_dup_inc((const SV *)s,t))
11188 #define io_dup(s,t) MUTABLE_IO(sv_dup((const SV *)s,t))
11189 #define io_dup_inc(s,t) MUTABLE_IO(sv_dup_inc((const SV *)s,t))
11190 #define gv_dup(s,t) MUTABLE_GV(sv_dup((const SV *)s,t))
11191 #define gv_dup_inc(s,t) MUTABLE_GV(sv_dup_inc((const SV *)s,t))
11192 #define SAVEPV(p) ((p) ? savepv(p) : NULL)
11193 #define SAVEPVN(p,n) ((p) ? savepvn(p,n) : NULL)
11195 /* clone a parser */
11198 Perl_parser_dup(pTHX_ const yy_parser *const proto, CLONE_PARAMS *const param)
11202 PERL_ARGS_ASSERT_PARSER_DUP;
11207 /* look for it in the table first */
11208 parser = (yy_parser *)ptr_table_fetch(PL_ptr_table, proto);
11212 /* create anew and remember what it is */
11213 Newxz(parser, 1, yy_parser);
11214 ptr_table_store(PL_ptr_table, proto, parser);
11216 /* XXX these not yet duped */
11217 parser->old_parser = NULL;
11218 parser->stack = NULL;
11220 parser->stack_size = 0;
11221 /* XXX parser->stack->state = 0; */
11223 /* XXX eventually, just Copy() most of the parser struct ? */
11225 parser->lex_brackets = proto->lex_brackets;
11226 parser->lex_casemods = proto->lex_casemods;
11227 parser->lex_brackstack = savepvn(proto->lex_brackstack,
11228 (proto->lex_brackets < 120 ? 120 : proto->lex_brackets));
11229 parser->lex_casestack = savepvn(proto->lex_casestack,
11230 (proto->lex_casemods < 12 ? 12 : proto->lex_casemods));
11231 parser->lex_defer = proto->lex_defer;
11232 parser->lex_dojoin = proto->lex_dojoin;
11233 parser->lex_expect = proto->lex_expect;
11234 parser->lex_formbrack = proto->lex_formbrack;
11235 parser->lex_inpat = proto->lex_inpat;
11236 parser->lex_inwhat = proto->lex_inwhat;
11237 parser->lex_op = proto->lex_op;
11238 parser->lex_repl = sv_dup_inc(proto->lex_repl, param);
11239 parser->lex_starts = proto->lex_starts;
11240 parser->lex_stuff = sv_dup_inc(proto->lex_stuff, param);
11241 parser->multi_close = proto->multi_close;
11242 parser->multi_open = proto->multi_open;
11243 parser->multi_start = proto->multi_start;
11244 parser->multi_end = proto->multi_end;
11245 parser->pending_ident = proto->pending_ident;
11246 parser->preambled = proto->preambled;
11247 parser->sublex_info = proto->sublex_info; /* XXX not quite right */
11248 parser->linestr = sv_dup_inc(proto->linestr, param);
11249 parser->expect = proto->expect;
11250 parser->copline = proto->copline;
11251 parser->last_lop_op = proto->last_lop_op;
11252 parser->lex_state = proto->lex_state;
11253 parser->rsfp = fp_dup(proto->rsfp, '<', param);
11254 /* rsfp_filters entries have fake IoDIRP() */
11255 parser->rsfp_filters= av_dup_inc(proto->rsfp_filters, param);
11256 parser->in_my = proto->in_my;
11257 parser->in_my_stash = hv_dup(proto->in_my_stash, param);
11258 parser->error_count = proto->error_count;
11261 parser->linestr = sv_dup_inc(proto->linestr, param);
11264 char * const ols = SvPVX(proto->linestr);
11265 char * const ls = SvPVX(parser->linestr);
11267 parser->bufptr = ls + (proto->bufptr >= ols ?
11268 proto->bufptr - ols : 0);
11269 parser->oldbufptr = ls + (proto->oldbufptr >= ols ?
11270 proto->oldbufptr - ols : 0);
11271 parser->oldoldbufptr= ls + (proto->oldoldbufptr >= ols ?
11272 proto->oldoldbufptr - ols : 0);
11273 parser->linestart = ls + (proto->linestart >= ols ?
11274 proto->linestart - ols : 0);
11275 parser->last_uni = ls + (proto->last_uni >= ols ?
11276 proto->last_uni - ols : 0);
11277 parser->last_lop = ls + (proto->last_lop >= ols ?
11278 proto->last_lop - ols : 0);
11280 parser->bufend = ls + SvCUR(parser->linestr);
11283 Copy(proto->tokenbuf, parser->tokenbuf, 256, char);
11287 parser->endwhite = proto->endwhite;
11288 parser->faketokens = proto->faketokens;
11289 parser->lasttoke = proto->lasttoke;
11290 parser->nextwhite = proto->nextwhite;
11291 parser->realtokenstart = proto->realtokenstart;
11292 parser->skipwhite = proto->skipwhite;
11293 parser->thisclose = proto->thisclose;
11294 parser->thismad = proto->thismad;
11295 parser->thisopen = proto->thisopen;
11296 parser->thisstuff = proto->thisstuff;
11297 parser->thistoken = proto->thistoken;
11298 parser->thiswhite = proto->thiswhite;
11300 Copy(proto->nexttoke, parser->nexttoke, 5, NEXTTOKE);
11301 parser->curforce = proto->curforce;
11303 Copy(proto->nextval, parser->nextval, 5, YYSTYPE);
11304 Copy(proto->nexttype, parser->nexttype, 5, I32);
11305 parser->nexttoke = proto->nexttoke;
11308 /* XXX should clone saved_curcop here, but we aren't passed
11309 * proto_perl; so do it in perl_clone_using instead */
11315 /* duplicate a file handle */
11318 Perl_fp_dup(pTHX_ PerlIO *const fp, const char type, CLONE_PARAMS *const param)
11322 PERL_ARGS_ASSERT_FP_DUP;
11323 PERL_UNUSED_ARG(type);
11326 return (PerlIO*)NULL;
11328 /* look for it in the table first */
11329 ret = (PerlIO*)ptr_table_fetch(PL_ptr_table, fp);
11333 /* create anew and remember what it is */
11334 ret = PerlIO_fdupopen(aTHX_ fp, param, PERLIO_DUP_CLONE);
11335 ptr_table_store(PL_ptr_table, fp, ret);
11339 /* duplicate a directory handle */
11342 Perl_dirp_dup(pTHX_ DIR *const dp, CLONE_PARAMS *const param)
11348 register const Direntry_t *dirent;
11349 char smallbuf[256];
11355 PERL_UNUSED_CONTEXT;
11356 PERL_ARGS_ASSERT_DIRP_DUP;
11361 /* look for it in the table first */
11362 ret = (DIR*)ptr_table_fetch(PL_ptr_table, dp);
11368 PERL_UNUSED_ARG(param);
11372 /* open the current directory (so we can switch back) */
11373 if (!(pwd = PerlDir_open("."))) return (DIR *)NULL;
11375 /* chdir to our dir handle and open the present working directory */
11376 if (fchdir(my_dirfd(dp)) < 0 || !(ret = PerlDir_open("."))) {
11377 PerlDir_close(pwd);
11378 return (DIR *)NULL;
11380 /* Now we should have two dir handles pointing to the same dir. */
11382 /* Be nice to the calling code and chdir back to where we were. */
11383 fchdir(my_dirfd(pwd)); /* If this fails, then what? */
11385 /* We have no need of the pwd handle any more. */
11386 PerlDir_close(pwd);
11389 # define d_namlen(d) (d)->d_namlen
11391 # define d_namlen(d) strlen((d)->d_name)
11393 /* Iterate once through dp, to get the file name at the current posi-
11394 tion. Then step back. */
11395 pos = PerlDir_tell(dp);
11396 if ((dirent = PerlDir_read(dp))) {
11397 len = d_namlen(dirent);
11398 if (len <= sizeof smallbuf) name = smallbuf;
11399 else Newx(name, len, char);
11400 Move(dirent->d_name, name, len, char);
11402 PerlDir_seek(dp, pos);
11404 /* Iterate through the new dir handle, till we find a file with the
11406 if (!dirent) /* just before the end */
11408 pos = PerlDir_tell(ret);
11409 if (PerlDir_read(ret)) continue; /* not there yet */
11410 PerlDir_seek(ret, pos); /* step back */
11414 const long pos0 = PerlDir_tell(ret);
11416 pos = PerlDir_tell(ret);
11417 if ((dirent = PerlDir_read(ret))) {
11418 if (len == d_namlen(dirent)
11419 && memEQ(name, dirent->d_name, len)) {
11421 PerlDir_seek(ret, pos); /* step back */
11424 /* else we are not there yet; keep iterating */
11426 else { /* This is not meant to happen. The best we can do is
11427 reset the iterator to the beginning. */
11428 PerlDir_seek(ret, pos0);
11435 if (name && name != smallbuf)
11440 ret = win32_dirp_dup(dp, param);
11443 /* pop it in the pointer table */
11445 ptr_table_store(PL_ptr_table, dp, ret);
11450 /* duplicate a typeglob */
11453 Perl_gp_dup(pTHX_ GP *const gp, CLONE_PARAMS *const param)
11457 PERL_ARGS_ASSERT_GP_DUP;
11461 /* look for it in the table first */
11462 ret = (GP*)ptr_table_fetch(PL_ptr_table, gp);
11466 /* create anew and remember what it is */
11468 ptr_table_store(PL_ptr_table, gp, ret);
11471 /* ret->gp_refcnt must be 0 before any other dups are called. We're relying
11472 on Newxz() to do this for us. */
11473 ret->gp_sv = sv_dup_inc(gp->gp_sv, param);
11474 ret->gp_io = io_dup_inc(gp->gp_io, param);
11475 ret->gp_form = cv_dup_inc(gp->gp_form, param);
11476 ret->gp_av = av_dup_inc(gp->gp_av, param);
11477 ret->gp_hv = hv_dup_inc(gp->gp_hv, param);
11478 ret->gp_egv = gv_dup(gp->gp_egv, param);/* GvEGV is not refcounted */
11479 ret->gp_cv = cv_dup_inc(gp->gp_cv, param);
11480 ret->gp_cvgen = gp->gp_cvgen;
11481 ret->gp_line = gp->gp_line;
11482 ret->gp_file_hek = hek_dup(gp->gp_file_hek, param);
11486 /* duplicate a chain of magic */
11489 Perl_mg_dup(pTHX_ MAGIC *mg, CLONE_PARAMS *const param)
11491 MAGIC *mgret = NULL;
11492 MAGIC **mgprev_p = &mgret;
11494 PERL_ARGS_ASSERT_MG_DUP;
11496 for (; mg; mg = mg->mg_moremagic) {
11499 if ((param->flags & CLONEf_JOIN_IN)
11500 && mg->mg_type == PERL_MAGIC_backref)
11501 /* when joining, we let the individual SVs add themselves to
11502 * backref as needed. */
11505 Newx(nmg, 1, MAGIC);
11507 mgprev_p = &(nmg->mg_moremagic);
11509 /* There was a comment "XXX copy dynamic vtable?" but as we don't have
11510 dynamic vtables, I'm not sure why Sarathy wrote it. The comment dates
11511 from the original commit adding Perl_mg_dup() - revision 4538.
11512 Similarly there is the annotation "XXX random ptr?" next to the
11513 assignment to nmg->mg_ptr. */
11516 /* FIXME for plugins
11517 if (nmg->mg_type == PERL_MAGIC_qr) {
11518 nmg->mg_obj = MUTABLE_SV(CALLREGDUPE((REGEXP*)nmg->mg_obj, param));
11522 nmg->mg_obj = (nmg->mg_flags & MGf_REFCOUNTED)
11523 ? nmg->mg_type == PERL_MAGIC_backref
11524 /* The backref AV has its reference
11525 * count deliberately bumped by 1 */
11526 ? SvREFCNT_inc(av_dup_inc((const AV *)
11527 nmg->mg_obj, param))
11528 : sv_dup_inc(nmg->mg_obj, param)
11529 : sv_dup(nmg->mg_obj, param);
11531 if (nmg->mg_ptr && nmg->mg_type != PERL_MAGIC_regex_global) {
11532 if (nmg->mg_len > 0) {
11533 nmg->mg_ptr = SAVEPVN(nmg->mg_ptr, nmg->mg_len);
11534 if (nmg->mg_type == PERL_MAGIC_overload_table &&
11535 AMT_AMAGIC((AMT*)nmg->mg_ptr))
11537 AMT * const namtp = (AMT*)nmg->mg_ptr;
11538 sv_dup_inc_multiple((SV**)(namtp->table),
11539 (SV**)(namtp->table), NofAMmeth, param);
11542 else if (nmg->mg_len == HEf_SVKEY)
11543 nmg->mg_ptr = (char*)sv_dup_inc((const SV *)nmg->mg_ptr, param);
11545 if ((nmg->mg_flags & MGf_DUP) && nmg->mg_virtual && nmg->mg_virtual->svt_dup) {
11546 nmg->mg_virtual->svt_dup(aTHX_ nmg, param);
11552 #endif /* USE_ITHREADS */
11554 struct ptr_tbl_arena {
11555 struct ptr_tbl_arena *next;
11556 struct ptr_tbl_ent array[1023/3]; /* as ptr_tbl_ent has 3 pointers. */
11559 /* create a new pointer-mapping table */
11562 Perl_ptr_table_new(pTHX)
11565 PERL_UNUSED_CONTEXT;
11567 Newx(tbl, 1, PTR_TBL_t);
11568 tbl->tbl_max = 511;
11569 tbl->tbl_items = 0;
11570 tbl->tbl_arena = NULL;
11571 tbl->tbl_arena_next = NULL;
11572 tbl->tbl_arena_end = NULL;
11573 Newxz(tbl->tbl_ary, tbl->tbl_max + 1, PTR_TBL_ENT_t*);
11577 #define PTR_TABLE_HASH(ptr) \
11578 ((PTR2UV(ptr) >> 3) ^ (PTR2UV(ptr) >> (3 + 7)) ^ (PTR2UV(ptr) >> (3 + 17)))
11580 /* map an existing pointer using a table */
11582 STATIC PTR_TBL_ENT_t *
11583 S_ptr_table_find(PTR_TBL_t *const tbl, const void *const sv)
11585 PTR_TBL_ENT_t *tblent;
11586 const UV hash = PTR_TABLE_HASH(sv);
11588 PERL_ARGS_ASSERT_PTR_TABLE_FIND;
11590 tblent = tbl->tbl_ary[hash & tbl->tbl_max];
11591 for (; tblent; tblent = tblent->next) {
11592 if (tblent->oldval == sv)
11599 Perl_ptr_table_fetch(pTHX_ PTR_TBL_t *const tbl, const void *const sv)
11601 PTR_TBL_ENT_t const *const tblent = ptr_table_find(tbl, sv);
11603 PERL_ARGS_ASSERT_PTR_TABLE_FETCH;
11604 PERL_UNUSED_CONTEXT;
11606 return tblent ? tblent->newval : NULL;
11609 /* add a new entry to a pointer-mapping table */
11612 Perl_ptr_table_store(pTHX_ PTR_TBL_t *const tbl, const void *const oldsv, void *const newsv)
11614 PTR_TBL_ENT_t *tblent = ptr_table_find(tbl, oldsv);
11616 PERL_ARGS_ASSERT_PTR_TABLE_STORE;
11617 PERL_UNUSED_CONTEXT;
11620 tblent->newval = newsv;
11622 const UV entry = PTR_TABLE_HASH(oldsv) & tbl->tbl_max;
11624 if (tbl->tbl_arena_next == tbl->tbl_arena_end) {
11625 struct ptr_tbl_arena *new_arena;
11627 Newx(new_arena, 1, struct ptr_tbl_arena);
11628 new_arena->next = tbl->tbl_arena;
11629 tbl->tbl_arena = new_arena;
11630 tbl->tbl_arena_next = new_arena->array;
11631 tbl->tbl_arena_end = new_arena->array
11632 + sizeof(new_arena->array) / sizeof(new_arena->array[0]);
11635 tblent = tbl->tbl_arena_next++;
11637 tblent->oldval = oldsv;
11638 tblent->newval = newsv;
11639 tblent->next = tbl->tbl_ary[entry];
11640 tbl->tbl_ary[entry] = tblent;
11642 if (tblent->next && tbl->tbl_items > tbl->tbl_max)
11643 ptr_table_split(tbl);
11647 /* double the hash bucket size of an existing ptr table */
11650 Perl_ptr_table_split(pTHX_ PTR_TBL_t *const tbl)
11652 PTR_TBL_ENT_t **ary = tbl->tbl_ary;
11653 const UV oldsize = tbl->tbl_max + 1;
11654 UV newsize = oldsize * 2;
11657 PERL_ARGS_ASSERT_PTR_TABLE_SPLIT;
11658 PERL_UNUSED_CONTEXT;
11660 Renew(ary, newsize, PTR_TBL_ENT_t*);
11661 Zero(&ary[oldsize], newsize-oldsize, PTR_TBL_ENT_t*);
11662 tbl->tbl_max = --newsize;
11663 tbl->tbl_ary = ary;
11664 for (i=0; i < oldsize; i++, ary++) {
11665 PTR_TBL_ENT_t **entp = ary;
11666 PTR_TBL_ENT_t *ent = *ary;
11667 PTR_TBL_ENT_t **curentp;
11670 curentp = ary + oldsize;
11672 if ((newsize & PTR_TABLE_HASH(ent->oldval)) != i) {
11674 ent->next = *curentp;
11684 /* remove all the entries from a ptr table */
11685 /* Deprecated - will be removed post 5.14 */
11688 Perl_ptr_table_clear(pTHX_ PTR_TBL_t *const tbl)
11690 if (tbl && tbl->tbl_items) {
11691 struct ptr_tbl_arena *arena = tbl->tbl_arena;
11693 Zero(tbl->tbl_ary, tbl->tbl_max + 1, struct ptr_tbl_ent **);
11696 struct ptr_tbl_arena *next = arena->next;
11702 tbl->tbl_items = 0;
11703 tbl->tbl_arena = NULL;
11704 tbl->tbl_arena_next = NULL;
11705 tbl->tbl_arena_end = NULL;
11709 /* clear and free a ptr table */
11712 Perl_ptr_table_free(pTHX_ PTR_TBL_t *const tbl)
11714 struct ptr_tbl_arena *arena;
11720 arena = tbl->tbl_arena;
11723 struct ptr_tbl_arena *next = arena->next;
11729 Safefree(tbl->tbl_ary);
11733 #if defined(USE_ITHREADS)
11736 Perl_rvpv_dup(pTHX_ SV *const dstr, const SV *const sstr, CLONE_PARAMS *const param)
11738 PERL_ARGS_ASSERT_RVPV_DUP;
11741 if (SvWEAKREF(sstr)) {
11742 SvRV_set(dstr, sv_dup(SvRV_const(sstr), param));
11743 if (param->flags & CLONEf_JOIN_IN) {
11744 /* if joining, we add any back references individually rather
11745 * than copying the whole backref array */
11746 Perl_sv_add_backref(aTHX_ SvRV(dstr), dstr);
11750 SvRV_set(dstr, sv_dup_inc(SvRV_const(sstr), param));
11752 else if (SvPVX_const(sstr)) {
11753 /* Has something there */
11755 /* Normal PV - clone whole allocated space */
11756 SvPV_set(dstr, SAVEPVN(SvPVX_const(sstr), SvLEN(sstr)-1));
11757 if (SvREADONLY(sstr) && SvFAKE(sstr)) {
11758 /* Not that normal - actually sstr is copy on write.
11759 But we are a true, independent SV, so: */
11760 SvREADONLY_off(dstr);
11765 /* Special case - not normally malloced for some reason */
11766 if (isGV_with_GP(sstr)) {
11767 /* Don't need to do anything here. */
11769 else if ((SvREADONLY(sstr) && SvFAKE(sstr))) {
11770 /* A "shared" PV - clone it as "shared" PV */
11772 HEK_KEY(hek_dup(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)),
11776 /* Some other special case - random pointer */
11777 SvPV_set(dstr, (char *) SvPVX_const(sstr));
11782 /* Copy the NULL */
11783 SvPV_set(dstr, NULL);
11787 /* duplicate a list of SVs. source and dest may point to the same memory. */
11789 S_sv_dup_inc_multiple(pTHX_ SV *const *source, SV **dest,
11790 SSize_t items, CLONE_PARAMS *const param)
11792 PERL_ARGS_ASSERT_SV_DUP_INC_MULTIPLE;
11794 while (items-- > 0) {
11795 *dest++ = sv_dup_inc(*source++, param);
11801 /* duplicate an SV of any type (including AV, HV etc) */
11804 S_sv_dup_common(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
11809 PERL_ARGS_ASSERT_SV_DUP_COMMON;
11811 if (SvTYPE(sstr) == (svtype)SVTYPEMASK) {
11812 #ifdef DEBUG_LEAKING_SCALARS_ABORT
11817 /* look for it in the table first */
11818 dstr = MUTABLE_SV(ptr_table_fetch(PL_ptr_table, sstr));
11822 if(param->flags & CLONEf_JOIN_IN) {
11823 /** We are joining here so we don't want do clone
11824 something that is bad **/
11825 if (SvTYPE(sstr) == SVt_PVHV) {
11826 const HEK * const hvname = HvNAME_HEK(sstr);
11828 /** don't clone stashes if they already exist **/
11829 dstr = MUTABLE_SV(gv_stashpvn(HEK_KEY(hvname), HEK_LEN(hvname),
11830 HEK_UTF8(hvname) ? SVf_UTF8 : 0));
11831 ptr_table_store(PL_ptr_table, sstr, dstr);
11835 else if (SvTYPE(sstr) == SVt_PVGV && !SvFAKE(sstr)) {
11836 HV *stash = GvSTASH(sstr);
11837 const HEK * hvname;
11838 if (stash && (hvname = HvNAME_HEK(stash))) {
11839 /** don't clone GVs if they already exist **/
11841 stash = gv_stashpvn(HEK_KEY(hvname), HEK_LEN(hvname),
11842 HEK_UTF8(hvname) ? SVf_UTF8 : 0);
11844 stash, GvNAME(sstr),
11850 if (svp && *svp && SvTYPE(*svp) == SVt_PVGV) {
11851 ptr_table_store(PL_ptr_table, sstr, *svp);
11858 /* create anew and remember what it is */
11861 #ifdef DEBUG_LEAKING_SCALARS
11862 dstr->sv_debug_optype = sstr->sv_debug_optype;
11863 dstr->sv_debug_line = sstr->sv_debug_line;
11864 dstr->sv_debug_inpad = sstr->sv_debug_inpad;
11865 dstr->sv_debug_parent = (SV*)sstr;
11866 FREE_SV_DEBUG_FILE(dstr);
11867 dstr->sv_debug_file = savepv(sstr->sv_debug_file);
11870 ptr_table_store(PL_ptr_table, sstr, dstr);
11873 SvFLAGS(dstr) = SvFLAGS(sstr);
11874 SvFLAGS(dstr) &= ~SVf_OOK; /* don't propagate OOK hack */
11875 SvREFCNT(dstr) = 0; /* must be before any other dups! */
11878 if (SvANY(sstr) && PL_watch_pvx && SvPVX_const(sstr) == PL_watch_pvx)
11879 PerlIO_printf(Perl_debug_log, "watch at %p hit, found string \"%s\"\n",
11880 (void*)PL_watch_pvx, SvPVX_const(sstr));
11883 /* don't clone objects whose class has asked us not to */
11884 if (SvOBJECT(sstr) && ! (SvFLAGS(SvSTASH(sstr)) & SVphv_CLONEABLE)) {
11889 switch (SvTYPE(sstr)) {
11891 SvANY(dstr) = NULL;
11894 SvANY(dstr) = (XPVIV*)((char*)&(dstr->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
11896 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
11898 SvIV_set(dstr, SvIVX(sstr));
11902 SvANY(dstr) = new_XNV();
11903 SvNV_set(dstr, SvNVX(sstr));
11905 /* case SVt_BIND: */
11908 /* These are all the types that need complex bodies allocating. */
11910 const svtype sv_type = SvTYPE(sstr);
11911 const struct body_details *const sv_type_details
11912 = bodies_by_type + sv_type;
11916 Perl_croak(aTHX_ "Bizarre SvTYPE [%" IVdf "]", (IV)SvTYPE(sstr));
11931 assert(sv_type_details->body_size);
11932 if (sv_type_details->arena) {
11933 new_body_inline(new_body, sv_type);
11935 = (void*)((char*)new_body - sv_type_details->offset);
11937 new_body = new_NOARENA(sv_type_details);
11941 SvANY(dstr) = new_body;
11944 Copy(((char*)SvANY(sstr)) + sv_type_details->offset,
11945 ((char*)SvANY(dstr)) + sv_type_details->offset,
11946 sv_type_details->copy, char);
11948 Copy(((char*)SvANY(sstr)),
11949 ((char*)SvANY(dstr)),
11950 sv_type_details->body_size + sv_type_details->offset, char);
11953 if (sv_type != SVt_PVAV && sv_type != SVt_PVHV
11954 && !isGV_with_GP(dstr)
11955 && !(sv_type == SVt_PVIO && !(IoFLAGS(dstr) & IOf_FAKE_DIRP)))
11956 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
11958 /* The Copy above means that all the source (unduplicated) pointers
11959 are now in the destination. We can check the flags and the
11960 pointers in either, but it's possible that there's less cache
11961 missing by always going for the destination.
11962 FIXME - instrument and check that assumption */
11963 if (sv_type >= SVt_PVMG) {
11964 if ((sv_type == SVt_PVMG) && SvPAD_OUR(dstr)) {
11965 SvOURSTASH_set(dstr, hv_dup_inc(SvOURSTASH(dstr), param));
11966 } else if (SvMAGIC(dstr))
11967 SvMAGIC_set(dstr, mg_dup(SvMAGIC(dstr), param));
11969 SvSTASH_set(dstr, hv_dup_inc(SvSTASH(dstr), param));
11972 /* The cast silences a GCC warning about unhandled types. */
11973 switch ((int)sv_type) {
11983 /* FIXME for plugins */
11984 re_dup_guts((REGEXP*) sstr, (REGEXP*) dstr, param);
11987 /* XXX LvTARGOFF sometimes holds PMOP* when DEBUGGING */
11988 if (LvTYPE(dstr) == 't') /* for tie: unrefcnted fake (SV**) */
11989 LvTARG(dstr) = dstr;
11990 else if (LvTYPE(dstr) == 'T') /* for tie: fake HE */
11991 LvTARG(dstr) = MUTABLE_SV(he_dup((HE*)LvTARG(dstr), 0, param));
11993 LvTARG(dstr) = sv_dup_inc(LvTARG(dstr), param);
11995 /* non-GP case already handled above */
11996 if(isGV_with_GP(sstr)) {
11997 GvNAME_HEK(dstr) = hek_dup(GvNAME_HEK(dstr), param);
11998 /* Don't call sv_add_backref here as it's going to be
11999 created as part of the magic cloning of the symbol
12000 table--unless this is during a join and the stash
12001 is not actually being cloned. */
12002 /* Danger Will Robinson - GvGP(dstr) isn't initialised
12003 at the point of this comment. */
12004 GvSTASH(dstr) = hv_dup(GvSTASH(dstr), param);
12005 if (param->flags & CLONEf_JOIN_IN)
12006 Perl_sv_add_backref(aTHX_ MUTABLE_SV(GvSTASH(dstr)), dstr);
12007 GvGP_set(dstr, gp_dup(GvGP(sstr), param));
12008 (void)GpREFCNT_inc(GvGP(dstr));
12012 /* PL_parser->rsfp_filters entries have fake IoDIRP() */
12013 if(IoFLAGS(dstr) & IOf_FAKE_DIRP) {
12014 /* I have no idea why fake dirp (rsfps)
12015 should be treated differently but otherwise
12016 we end up with leaks -- sky*/
12017 IoTOP_GV(dstr) = gv_dup_inc(IoTOP_GV(dstr), param);
12018 IoFMT_GV(dstr) = gv_dup_inc(IoFMT_GV(dstr), param);
12019 IoBOTTOM_GV(dstr) = gv_dup_inc(IoBOTTOM_GV(dstr), param);
12021 IoTOP_GV(dstr) = gv_dup(IoTOP_GV(dstr), param);
12022 IoFMT_GV(dstr) = gv_dup(IoFMT_GV(dstr), param);
12023 IoBOTTOM_GV(dstr) = gv_dup(IoBOTTOM_GV(dstr), param);
12024 if (IoDIRP(dstr)) {
12025 IoDIRP(dstr) = dirp_dup(IoDIRP(dstr), param);
12028 /* IoDIRP(dstr) is already a copy of IoDIRP(sstr) */
12030 IoIFP(dstr) = fp_dup(IoIFP(sstr), IoTYPE(dstr), param);
12032 if (IoOFP(dstr) == IoIFP(sstr))
12033 IoOFP(dstr) = IoIFP(dstr);
12035 IoOFP(dstr) = fp_dup(IoOFP(dstr), IoTYPE(dstr), param);
12036 IoTOP_NAME(dstr) = SAVEPV(IoTOP_NAME(dstr));
12037 IoFMT_NAME(dstr) = SAVEPV(IoFMT_NAME(dstr));
12038 IoBOTTOM_NAME(dstr) = SAVEPV(IoBOTTOM_NAME(dstr));
12041 /* avoid cloning an empty array */
12042 if (AvARRAY((const AV *)sstr) && AvFILLp((const AV *)sstr) >= 0) {
12043 SV **dst_ary, **src_ary;
12044 SSize_t items = AvFILLp((const AV *)sstr) + 1;
12046 src_ary = AvARRAY((const AV *)sstr);
12047 Newxz(dst_ary, AvMAX((const AV *)sstr)+1, SV*);
12048 ptr_table_store(PL_ptr_table, src_ary, dst_ary);
12049 AvARRAY(MUTABLE_AV(dstr)) = dst_ary;
12050 AvALLOC((const AV *)dstr) = dst_ary;
12051 if (AvREAL((const AV *)sstr)) {
12052 dst_ary = sv_dup_inc_multiple(src_ary, dst_ary, items,
12056 while (items-- > 0)
12057 *dst_ary++ = sv_dup(*src_ary++, param);
12059 items = AvMAX((const AV *)sstr) - AvFILLp((const AV *)sstr);
12060 while (items-- > 0) {
12061 *dst_ary++ = &PL_sv_undef;
12065 AvARRAY(MUTABLE_AV(dstr)) = NULL;
12066 AvALLOC((const AV *)dstr) = (SV**)NULL;
12067 AvMAX( (const AV *)dstr) = -1;
12068 AvFILLp((const AV *)dstr) = -1;
12072 if (HvARRAY((const HV *)sstr)) {
12074 const bool sharekeys = !!HvSHAREKEYS(sstr);
12075 XPVHV * const dxhv = (XPVHV*)SvANY(dstr);
12076 XPVHV * const sxhv = (XPVHV*)SvANY(sstr);
12078 Newx(darray, PERL_HV_ARRAY_ALLOC_BYTES(dxhv->xhv_max+1)
12079 + (SvOOK(sstr) ? sizeof(struct xpvhv_aux) : 0),
12081 HvARRAY(dstr) = (HE**)darray;
12082 while (i <= sxhv->xhv_max) {
12083 const HE * const source = HvARRAY(sstr)[i];
12084 HvARRAY(dstr)[i] = source
12085 ? he_dup(source, sharekeys, param) : 0;
12089 const struct xpvhv_aux * const saux = HvAUX(sstr);
12090 struct xpvhv_aux * const daux = HvAUX(dstr);
12091 /* This flag isn't copied. */
12094 if (saux->xhv_name_count) {
12095 HEK ** const sname = saux->xhv_name_u.xhvnameu_names;
12097 = saux->xhv_name_count < 0
12098 ? -saux->xhv_name_count
12099 : saux->xhv_name_count;
12100 HEK **shekp = sname + count;
12102 Newx(daux->xhv_name_u.xhvnameu_names, count, HEK *);
12103 dhekp = daux->xhv_name_u.xhvnameu_names + count;
12104 while (shekp-- > sname) {
12106 *dhekp = hek_dup(*shekp, param);
12110 daux->xhv_name_u.xhvnameu_name
12111 = hek_dup(saux->xhv_name_u.xhvnameu_name,
12114 daux->xhv_name_count = saux->xhv_name_count;
12116 daux->xhv_riter = saux->xhv_riter;
12117 daux->xhv_eiter = saux->xhv_eiter
12118 ? he_dup(saux->xhv_eiter,
12119 cBOOL(HvSHAREKEYS(sstr)), param) : 0;
12120 /* backref array needs refcnt=2; see sv_add_backref */
12121 daux->xhv_backreferences =
12122 (param->flags & CLONEf_JOIN_IN)
12123 /* when joining, we let the individual GVs and
12124 * CVs add themselves to backref as
12125 * needed. This avoids pulling in stuff
12126 * that isn't required, and simplifies the
12127 * case where stashes aren't cloned back
12128 * if they already exist in the parent
12131 : saux->xhv_backreferences
12132 ? (SvTYPE(saux->xhv_backreferences) == SVt_PVAV)
12133 ? MUTABLE_AV(SvREFCNT_inc(
12134 sv_dup_inc((const SV *)
12135 saux->xhv_backreferences, param)))
12136 : MUTABLE_AV(sv_dup((const SV *)
12137 saux->xhv_backreferences, param))
12140 daux->xhv_mro_meta = saux->xhv_mro_meta
12141 ? mro_meta_dup(saux->xhv_mro_meta, param)
12144 /* Record stashes for possible cloning in Perl_clone(). */
12146 av_push(param->stashes, dstr);
12150 HvARRAY(MUTABLE_HV(dstr)) = NULL;
12153 if (!(param->flags & CLONEf_COPY_STACKS)) {
12158 /* NOTE: not refcounted */
12159 SvANY(MUTABLE_CV(dstr))->xcv_stash =
12160 hv_dup(CvSTASH(dstr), param);
12161 if ((param->flags & CLONEf_JOIN_IN) && CvSTASH(dstr))
12162 Perl_sv_add_backref(aTHX_ MUTABLE_SV(CvSTASH(dstr)), dstr);
12163 if (!CvISXSUB(dstr)) {
12165 CvROOT(dstr) = OpREFCNT_inc(CvROOT(dstr));
12167 CvSLABBED_off(dstr);
12168 } else if (CvCONST(dstr)) {
12169 CvXSUBANY(dstr).any_ptr =
12170 sv_dup_inc((const SV *)CvXSUBANY(dstr).any_ptr, param);
12172 assert(!CvSLABBED(dstr));
12173 if (CvDYNFILE(dstr)) CvFILE(dstr) = SAVEPV(CvFILE(dstr));
12174 /* don't dup if copying back - CvGV isn't refcounted, so the
12175 * duped GV may never be freed. A bit of a hack! DAPM */
12176 SvANY(MUTABLE_CV(dstr))->xcv_gv =
12178 ? gv_dup_inc(CvGV(sstr), param)
12179 : (param->flags & CLONEf_JOIN_IN)
12181 : gv_dup(CvGV(sstr), param);
12183 CvPADLIST(dstr) = padlist_dup(CvPADLIST(sstr), param);
12185 CvWEAKOUTSIDE(sstr)
12186 ? cv_dup( CvOUTSIDE(dstr), param)
12187 : cv_dup_inc(CvOUTSIDE(dstr), param);
12193 if (SvOBJECT(dstr) && SvTYPE(dstr) != SVt_PVIO)
12200 Perl_sv_dup_inc(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
12202 PERL_ARGS_ASSERT_SV_DUP_INC;
12203 return sstr ? SvREFCNT_inc(sv_dup_common(sstr, param)) : NULL;
12207 Perl_sv_dup(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
12209 SV *dstr = sstr ? sv_dup_common(sstr, param) : NULL;
12210 PERL_ARGS_ASSERT_SV_DUP;
12212 /* Track every SV that (at least initially) had a reference count of 0.
12213 We need to do this by holding an actual reference to it in this array.
12214 If we attempt to cheat, turn AvREAL_off(), and store only pointers
12215 (akin to the stashes hash, and the perl stack), we come unstuck if
12216 a weak reference (or other SV legitimately SvREFCNT() == 0 for this
12217 thread) is manipulated in a CLONE method, because CLONE runs before the
12218 unreferenced array is walked to find SVs still with SvREFCNT() == 0
12219 (and fix things up by giving each a reference via the temps stack).
12220 Instead, during CLONE, if the 0-referenced SV has SvREFCNT_inc() and
12221 then SvREFCNT_dec(), it will be cleaned up (and added to the free list)
12222 before the walk of unreferenced happens and a reference to that is SV
12223 added to the temps stack. At which point we have the same SV considered
12224 to be in use, and free to be re-used. Not good.
12226 if (dstr && !(param->flags & CLONEf_COPY_STACKS) && !SvREFCNT(dstr)) {
12227 assert(param->unreferenced);
12228 av_push(param->unreferenced, SvREFCNT_inc(dstr));
12234 /* duplicate a context */
12237 Perl_cx_dup(pTHX_ PERL_CONTEXT *cxs, I32 ix, I32 max, CLONE_PARAMS* param)
12239 PERL_CONTEXT *ncxs;
12241 PERL_ARGS_ASSERT_CX_DUP;
12244 return (PERL_CONTEXT*)NULL;
12246 /* look for it in the table first */
12247 ncxs = (PERL_CONTEXT*)ptr_table_fetch(PL_ptr_table, cxs);
12251 /* create anew and remember what it is */
12252 Newx(ncxs, max + 1, PERL_CONTEXT);
12253 ptr_table_store(PL_ptr_table, cxs, ncxs);
12254 Copy(cxs, ncxs, max + 1, PERL_CONTEXT);
12257 PERL_CONTEXT * const ncx = &ncxs[ix];
12258 if (CxTYPE(ncx) == CXt_SUBST) {
12259 Perl_croak(aTHX_ "Cloning substitution context is unimplemented");
12262 ncx->blk_oldcop = (COP*)any_dup(ncx->blk_oldcop, param->proto_perl);
12263 switch (CxTYPE(ncx)) {
12265 ncx->blk_sub.cv = (ncx->blk_sub.olddepth == 0
12266 ? cv_dup_inc(ncx->blk_sub.cv, param)
12267 : cv_dup(ncx->blk_sub.cv,param));
12268 ncx->blk_sub.argarray = (CxHASARGS(ncx)
12269 ? av_dup_inc(ncx->blk_sub.argarray,
12272 ncx->blk_sub.savearray = av_dup_inc(ncx->blk_sub.savearray,
12274 ncx->blk_sub.oldcomppad = (PAD*)ptr_table_fetch(PL_ptr_table,
12275 ncx->blk_sub.oldcomppad);
12278 ncx->blk_eval.old_namesv = sv_dup_inc(ncx->blk_eval.old_namesv,
12280 ncx->blk_eval.cur_text = sv_dup(ncx->blk_eval.cur_text, param);
12281 ncx->blk_eval.cv = cv_dup(ncx->blk_eval.cv, param);
12283 case CXt_LOOP_LAZYSV:
12284 ncx->blk_loop.state_u.lazysv.end
12285 = sv_dup_inc(ncx->blk_loop.state_u.lazysv.end, param);
12286 /* We are taking advantage of av_dup_inc and sv_dup_inc
12287 actually being the same function, and order equivalence of
12289 We can assert the later [but only at run time :-(] */
12290 assert ((void *) &ncx->blk_loop.state_u.ary.ary ==
12291 (void *) &ncx->blk_loop.state_u.lazysv.cur);
12293 ncx->blk_loop.state_u.ary.ary
12294 = av_dup_inc(ncx->blk_loop.state_u.ary.ary, param);
12295 case CXt_LOOP_LAZYIV:
12296 case CXt_LOOP_PLAIN:
12297 if (CxPADLOOP(ncx)) {
12298 ncx->blk_loop.itervar_u.oldcomppad
12299 = (PAD*)ptr_table_fetch(PL_ptr_table,
12300 ncx->blk_loop.itervar_u.oldcomppad);
12302 ncx->blk_loop.itervar_u.gv
12303 = gv_dup((const GV *)ncx->blk_loop.itervar_u.gv,
12308 ncx->blk_format.cv = cv_dup(ncx->blk_format.cv, param);
12309 ncx->blk_format.gv = gv_dup(ncx->blk_format.gv, param);
12310 ncx->blk_format.dfoutgv = gv_dup_inc(ncx->blk_format.dfoutgv,
12325 /* duplicate a stack info structure */
12328 Perl_si_dup(pTHX_ PERL_SI *si, CLONE_PARAMS* param)
12332 PERL_ARGS_ASSERT_SI_DUP;
12335 return (PERL_SI*)NULL;
12337 /* look for it in the table first */
12338 nsi = (PERL_SI*)ptr_table_fetch(PL_ptr_table, si);
12342 /* create anew and remember what it is */
12343 Newxz(nsi, 1, PERL_SI);
12344 ptr_table_store(PL_ptr_table, si, nsi);
12346 nsi->si_stack = av_dup_inc(si->si_stack, param);
12347 nsi->si_cxix = si->si_cxix;
12348 nsi->si_cxmax = si->si_cxmax;
12349 nsi->si_cxstack = cx_dup(si->si_cxstack, si->si_cxix, si->si_cxmax, param);
12350 nsi->si_type = si->si_type;
12351 nsi->si_prev = si_dup(si->si_prev, param);
12352 nsi->si_next = si_dup(si->si_next, param);
12353 nsi->si_markoff = si->si_markoff;
12358 #define POPINT(ss,ix) ((ss)[--(ix)].any_i32)
12359 #define TOPINT(ss,ix) ((ss)[ix].any_i32)
12360 #define POPLONG(ss,ix) ((ss)[--(ix)].any_long)
12361 #define TOPLONG(ss,ix) ((ss)[ix].any_long)
12362 #define POPIV(ss,ix) ((ss)[--(ix)].any_iv)
12363 #define TOPIV(ss,ix) ((ss)[ix].any_iv)
12364 #define POPUV(ss,ix) ((ss)[--(ix)].any_uv)
12365 #define TOPUV(ss,ix) ((ss)[ix].any_uv)
12366 #define POPBOOL(ss,ix) ((ss)[--(ix)].any_bool)
12367 #define TOPBOOL(ss,ix) ((ss)[ix].any_bool)
12368 #define POPPTR(ss,ix) ((ss)[--(ix)].any_ptr)
12369 #define TOPPTR(ss,ix) ((ss)[ix].any_ptr)
12370 #define POPDPTR(ss,ix) ((ss)[--(ix)].any_dptr)
12371 #define TOPDPTR(ss,ix) ((ss)[ix].any_dptr)
12372 #define POPDXPTR(ss,ix) ((ss)[--(ix)].any_dxptr)
12373 #define TOPDXPTR(ss,ix) ((ss)[ix].any_dxptr)
12376 #define pv_dup_inc(p) SAVEPV(p)
12377 #define pv_dup(p) SAVEPV(p)
12378 #define svp_dup_inc(p,pp) any_dup(p,pp)
12380 /* map any object to the new equivent - either something in the
12381 * ptr table, or something in the interpreter structure
12385 Perl_any_dup(pTHX_ void *v, const PerlInterpreter *proto_perl)
12389 PERL_ARGS_ASSERT_ANY_DUP;
12392 return (void*)NULL;
12394 /* look for it in the table first */
12395 ret = ptr_table_fetch(PL_ptr_table, v);
12399 /* see if it is part of the interpreter structure */
12400 if (v >= (void*)proto_perl && v < (void*)(proto_perl+1))
12401 ret = (void*)(((char*)aTHX) + (((char*)v) - (char*)proto_perl));
12409 /* duplicate the save stack */
12412 Perl_ss_dup(pTHX_ PerlInterpreter *proto_perl, CLONE_PARAMS* param)
12415 ANY * const ss = proto_perl->Isavestack;
12416 const I32 max = proto_perl->Isavestack_max;
12417 I32 ix = proto_perl->Isavestack_ix;
12430 void (*dptr) (void*);
12431 void (*dxptr) (pTHX_ void*);
12433 PERL_ARGS_ASSERT_SS_DUP;
12435 Newxz(nss, max, ANY);
12438 const UV uv = POPUV(ss,ix);
12439 const U8 type = (U8)uv & SAVE_MASK;
12441 TOPUV(nss,ix) = uv;
12443 case SAVEt_CLEARSV:
12445 case SAVEt_HELEM: /* hash element */
12446 sv = (const SV *)POPPTR(ss,ix);
12447 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12449 case SAVEt_ITEM: /* normal string */
12450 case SAVEt_GVSV: /* scalar slot in GV */
12451 case SAVEt_SV: /* scalar reference */
12452 sv = (const SV *)POPPTR(ss,ix);
12453 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12456 case SAVEt_MORTALIZESV:
12457 sv = (const SV *)POPPTR(ss,ix);
12458 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12460 case SAVEt_SHARED_PVREF: /* char* in shared space */
12461 c = (char*)POPPTR(ss,ix);
12462 TOPPTR(nss,ix) = savesharedpv(c);
12463 ptr = POPPTR(ss,ix);
12464 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12466 case SAVEt_GENERIC_SVREF: /* generic sv */
12467 case SAVEt_SVREF: /* scalar reference */
12468 sv = (const SV *)POPPTR(ss,ix);
12469 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12470 ptr = POPPTR(ss,ix);
12471 TOPPTR(nss,ix) = svp_dup_inc((SV**)ptr, proto_perl);/* XXXXX */
12473 case SAVEt_HV: /* hash reference */
12474 case SAVEt_AV: /* array reference */
12475 sv = (const SV *) POPPTR(ss,ix);
12476 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12478 case SAVEt_COMPPAD:
12480 sv = (const SV *) POPPTR(ss,ix);
12481 TOPPTR(nss,ix) = sv_dup(sv, param);
12483 case SAVEt_INT: /* int reference */
12484 ptr = POPPTR(ss,ix);
12485 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12486 intval = (int)POPINT(ss,ix);
12487 TOPINT(nss,ix) = intval;
12489 case SAVEt_LONG: /* long reference */
12490 ptr = POPPTR(ss,ix);
12491 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12492 longval = (long)POPLONG(ss,ix);
12493 TOPLONG(nss,ix) = longval;
12495 case SAVEt_I32: /* I32 reference */
12496 ptr = POPPTR(ss,ix);
12497 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12499 TOPINT(nss,ix) = i;
12501 case SAVEt_IV: /* IV reference */
12502 ptr = POPPTR(ss,ix);
12503 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12505 TOPIV(nss,ix) = iv;
12507 case SAVEt_HPTR: /* HV* reference */
12508 case SAVEt_APTR: /* AV* reference */
12509 case SAVEt_SPTR: /* SV* reference */
12510 ptr = POPPTR(ss,ix);
12511 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12512 sv = (const SV *)POPPTR(ss,ix);
12513 TOPPTR(nss,ix) = sv_dup(sv, param);
12515 case SAVEt_VPTR: /* random* reference */
12516 ptr = POPPTR(ss,ix);
12517 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12519 case SAVEt_INT_SMALL:
12520 case SAVEt_I32_SMALL:
12521 case SAVEt_I16: /* I16 reference */
12522 case SAVEt_I8: /* I8 reference */
12524 ptr = POPPTR(ss,ix);
12525 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12527 case SAVEt_GENERIC_PVREF: /* generic char* */
12528 case SAVEt_PPTR: /* char* reference */
12529 ptr = POPPTR(ss,ix);
12530 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12531 c = (char*)POPPTR(ss,ix);
12532 TOPPTR(nss,ix) = pv_dup(c);
12534 case SAVEt_GP: /* scalar reference */
12535 gp = (GP*)POPPTR(ss,ix);
12536 TOPPTR(nss,ix) = gp = gp_dup(gp, param);
12537 (void)GpREFCNT_inc(gp);
12538 gv = (const GV *)POPPTR(ss,ix);
12539 TOPPTR(nss,ix) = gv_dup_inc(gv, param);
12542 ptr = POPPTR(ss,ix);
12543 if (ptr && (((OP*)ptr)->op_private & OPpREFCOUNTED)) {
12544 /* these are assumed to be refcounted properly */
12546 switch (((OP*)ptr)->op_type) {
12548 case OP_LEAVESUBLV:
12552 case OP_LEAVEWRITE:
12553 TOPPTR(nss,ix) = ptr;
12556 (void) OpREFCNT_inc(o);
12560 TOPPTR(nss,ix) = NULL;
12565 TOPPTR(nss,ix) = NULL;
12567 case SAVEt_FREECOPHH:
12568 ptr = POPPTR(ss,ix);
12569 TOPPTR(nss,ix) = cophh_copy((COPHH *)ptr);
12572 hv = (const HV *)POPPTR(ss,ix);
12573 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
12575 TOPINT(nss,ix) = i;
12578 c = (char*)POPPTR(ss,ix);
12579 TOPPTR(nss,ix) = pv_dup_inc(c);
12581 case SAVEt_STACK_POS: /* Position on Perl stack */
12583 TOPINT(nss,ix) = i;
12585 case SAVEt_DESTRUCTOR:
12586 ptr = POPPTR(ss,ix);
12587 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
12588 dptr = POPDPTR(ss,ix);
12589 TOPDPTR(nss,ix) = DPTR2FPTR(void (*)(void*),
12590 any_dup(FPTR2DPTR(void *, dptr),
12593 case SAVEt_DESTRUCTOR_X:
12594 ptr = POPPTR(ss,ix);
12595 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
12596 dxptr = POPDXPTR(ss,ix);
12597 TOPDXPTR(nss,ix) = DPTR2FPTR(void (*)(pTHX_ void*),
12598 any_dup(FPTR2DPTR(void *, dxptr),
12601 case SAVEt_REGCONTEXT:
12603 ix -= uv >> SAVE_TIGHT_SHIFT;
12605 case SAVEt_AELEM: /* array element */
12606 sv = (const SV *)POPPTR(ss,ix);
12607 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12609 TOPINT(nss,ix) = i;
12610 av = (const AV *)POPPTR(ss,ix);
12611 TOPPTR(nss,ix) = av_dup_inc(av, param);
12614 ptr = POPPTR(ss,ix);
12615 TOPPTR(nss,ix) = ptr;
12618 ptr = POPPTR(ss,ix);
12619 ptr = cophh_copy((COPHH*)ptr);
12620 TOPPTR(nss,ix) = ptr;
12622 TOPINT(nss,ix) = i;
12623 if (i & HINT_LOCALIZE_HH) {
12624 hv = (const HV *)POPPTR(ss,ix);
12625 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
12628 case SAVEt_PADSV_AND_MORTALIZE:
12629 longval = (long)POPLONG(ss,ix);
12630 TOPLONG(nss,ix) = longval;
12631 ptr = POPPTR(ss,ix);
12632 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12633 sv = (const SV *)POPPTR(ss,ix);
12634 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12636 case SAVEt_SET_SVFLAGS:
12638 TOPINT(nss,ix) = i;
12640 TOPINT(nss,ix) = i;
12641 sv = (const SV *)POPPTR(ss,ix);
12642 TOPPTR(nss,ix) = sv_dup(sv, param);
12644 case SAVEt_RE_STATE:
12646 const struct re_save_state *const old_state
12647 = (struct re_save_state *)
12648 (ss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
12649 struct re_save_state *const new_state
12650 = (struct re_save_state *)
12651 (nss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
12653 Copy(old_state, new_state, 1, struct re_save_state);
12654 ix -= SAVESTACK_ALLOC_FOR_RE_SAVE_STATE;
12656 new_state->re_state_bostr
12657 = pv_dup(old_state->re_state_bostr);
12658 new_state->re_state_reginput
12659 = pv_dup(old_state->re_state_reginput);
12660 new_state->re_state_regeol
12661 = pv_dup(old_state->re_state_regeol);
12662 #ifdef PERL_OLD_COPY_ON_WRITE
12663 new_state->re_state_nrs
12664 = sv_dup(old_state->re_state_nrs, param);
12666 new_state->re_state_reg_magic
12667 = (MAGIC*) any_dup(old_state->re_state_reg_magic,
12669 new_state->re_state_reg_oldcurpm
12670 = (PMOP*) any_dup(old_state->re_state_reg_oldcurpm,
12672 new_state->re_state_reg_curpm
12673 = (PMOP*) any_dup(old_state->re_state_reg_curpm,
12675 new_state->re_state_reg_oldsaved
12676 = pv_dup(old_state->re_state_reg_oldsaved);
12677 new_state->re_state_reg_poscache
12678 = pv_dup(old_state->re_state_reg_poscache);
12679 new_state->re_state_reg_starttry
12680 = pv_dup(old_state->re_state_reg_starttry);
12683 case SAVEt_COMPILE_WARNINGS:
12684 ptr = POPPTR(ss,ix);
12685 TOPPTR(nss,ix) = DUP_WARNINGS((STRLEN*)ptr);
12688 ptr = POPPTR(ss,ix);
12689 TOPPTR(nss,ix) = parser_dup((const yy_parser*)ptr, param);
12693 "panic: ss_dup inconsistency (%"IVdf")", (IV) type);
12701 /* if sv is a stash, call $class->CLONE_SKIP(), and set the SVphv_CLONEABLE
12702 * flag to the result. This is done for each stash before cloning starts,
12703 * so we know which stashes want their objects cloned */
12706 do_mark_cloneable_stash(pTHX_ SV *const sv)
12708 const HEK * const hvname = HvNAME_HEK((const HV *)sv);
12710 GV* const cloner = gv_fetchmethod_autoload(MUTABLE_HV(sv), "CLONE_SKIP", 0);
12711 SvFLAGS(sv) |= SVphv_CLONEABLE; /* clone objects by default */
12712 if (cloner && GvCV(cloner)) {
12719 mXPUSHs(newSVhek(hvname));
12721 call_sv(MUTABLE_SV(GvCV(cloner)), G_SCALAR);
12728 SvFLAGS(sv) &= ~SVphv_CLONEABLE;
12736 =for apidoc perl_clone
12738 Create and return a new interpreter by cloning the current one.
12740 perl_clone takes these flags as parameters:
12742 CLONEf_COPY_STACKS - is used to, well, copy the stacks also,
12743 without it we only clone the data and zero the stacks,
12744 with it we copy the stacks and the new perl interpreter is
12745 ready to run at the exact same point as the previous one.
12746 The pseudo-fork code uses COPY_STACKS while the
12747 threads->create doesn't.
12749 CLONEf_KEEP_PTR_TABLE -
12750 perl_clone keeps a ptr_table with the pointer of the old
12751 variable as a key and the new variable as a value,
12752 this allows it to check if something has been cloned and not
12753 clone it again but rather just use the value and increase the
12754 refcount. If KEEP_PTR_TABLE is not set then perl_clone will kill
12755 the ptr_table using the function
12756 C<ptr_table_free(PL_ptr_table); PL_ptr_table = NULL;>,
12757 reason to keep it around is if you want to dup some of your own
12758 variable who are outside the graph perl scans, example of this
12759 code is in threads.xs create.
12761 CLONEf_CLONE_HOST -
12762 This is a win32 thing, it is ignored on unix, it tells perls
12763 win32host code (which is c++) to clone itself, this is needed on
12764 win32 if you want to run two threads at the same time,
12765 if you just want to do some stuff in a separate perl interpreter
12766 and then throw it away and return to the original one,
12767 you don't need to do anything.
12772 /* XXX the above needs expanding by someone who actually understands it ! */
12773 EXTERN_C PerlInterpreter *
12774 perl_clone_host(PerlInterpreter* proto_perl, UV flags);
12777 perl_clone(PerlInterpreter *proto_perl, UV flags)
12780 #ifdef PERL_IMPLICIT_SYS
12782 PERL_ARGS_ASSERT_PERL_CLONE;
12784 /* perlhost.h so we need to call into it
12785 to clone the host, CPerlHost should have a c interface, sky */
12787 if (flags & CLONEf_CLONE_HOST) {
12788 return perl_clone_host(proto_perl,flags);
12790 return perl_clone_using(proto_perl, flags,
12792 proto_perl->IMemShared,
12793 proto_perl->IMemParse,
12795 proto_perl->IStdIO,
12799 proto_perl->IProc);
12803 perl_clone_using(PerlInterpreter *proto_perl, UV flags,
12804 struct IPerlMem* ipM, struct IPerlMem* ipMS,
12805 struct IPerlMem* ipMP, struct IPerlEnv* ipE,
12806 struct IPerlStdIO* ipStd, struct IPerlLIO* ipLIO,
12807 struct IPerlDir* ipD, struct IPerlSock* ipS,
12808 struct IPerlProc* ipP)
12810 /* XXX many of the string copies here can be optimized if they're
12811 * constants; they need to be allocated as common memory and just
12812 * their pointers copied. */
12815 CLONE_PARAMS clone_params;
12816 CLONE_PARAMS* const param = &clone_params;
12818 PerlInterpreter * const my_perl = (PerlInterpreter*)(*ipM->pMalloc)(ipM, sizeof(PerlInterpreter));
12820 PERL_ARGS_ASSERT_PERL_CLONE_USING;
12821 #else /* !PERL_IMPLICIT_SYS */
12823 CLONE_PARAMS clone_params;
12824 CLONE_PARAMS* param = &clone_params;
12825 PerlInterpreter * const my_perl = (PerlInterpreter*)PerlMem_malloc(sizeof(PerlInterpreter));
12827 PERL_ARGS_ASSERT_PERL_CLONE;
12828 #endif /* PERL_IMPLICIT_SYS */
12830 /* for each stash, determine whether its objects should be cloned */
12831 S_visit(proto_perl, do_mark_cloneable_stash, SVt_PVHV, SVTYPEMASK);
12832 PERL_SET_THX(my_perl);
12835 PoisonNew(my_perl, 1, PerlInterpreter);
12838 PL_defstash = NULL; /* may be used by perl malloc() */
12841 PL_scopestack_name = 0;
12843 PL_savestack_ix = 0;
12844 PL_savestack_max = -1;
12845 PL_sig_pending = 0;
12847 Zero(&PL_debug_pad, 1, struct perl_debug_pad);
12848 # ifdef DEBUG_LEAKING_SCALARS
12849 PL_sv_serial = (((UV)my_perl >> 2) & 0xfff) * 1000000;
12851 #else /* !DEBUGGING */
12852 Zero(my_perl, 1, PerlInterpreter);
12853 #endif /* DEBUGGING */
12855 #ifdef PERL_IMPLICIT_SYS
12856 /* host pointers */
12858 PL_MemShared = ipMS;
12859 PL_MemParse = ipMP;
12866 #endif /* PERL_IMPLICIT_SYS */
12868 param->flags = flags;
12869 /* Nothing in the core code uses this, but we make it available to
12870 extensions (using mg_dup). */
12871 param->proto_perl = proto_perl;
12872 /* Likely nothing will use this, but it is initialised to be consistent
12873 with Perl_clone_params_new(). */
12874 param->new_perl = my_perl;
12875 param->unreferenced = NULL;
12877 INIT_TRACK_MEMPOOL(my_perl->Imemory_debug_header, my_perl);
12879 PL_body_arenas = NULL;
12880 Zero(&PL_body_roots, 1, PL_body_roots);
12883 PL_sv_objcount = 0;
12885 PL_sv_arenaroot = NULL;
12887 PL_debug = proto_perl->Idebug;
12889 PL_hash_seed = proto_perl->Ihash_seed;
12890 PL_rehash_seed = proto_perl->Irehash_seed;
12892 /* dbargs array probably holds garbage */
12895 PL_compiling = proto_perl->Icompiling;
12897 /* pseudo environmental stuff */
12898 PL_origargc = proto_perl->Iorigargc;
12899 PL_origargv = proto_perl->Iorigargv;
12901 /* Set tainting stuff before PerlIO_debug can possibly get called */
12902 PL_tainting = proto_perl->Itainting;
12903 PL_taint_warn = proto_perl->Itaint_warn;
12905 PL_minus_c = proto_perl->Iminus_c;
12907 PL_localpatches = proto_perl->Ilocalpatches;
12908 PL_splitstr = proto_perl->Isplitstr;
12909 PL_minus_n = proto_perl->Iminus_n;
12910 PL_minus_p = proto_perl->Iminus_p;
12911 PL_minus_l = proto_perl->Iminus_l;
12912 PL_minus_a = proto_perl->Iminus_a;
12913 PL_minus_E = proto_perl->Iminus_E;
12914 PL_minus_F = proto_perl->Iminus_F;
12915 PL_doswitches = proto_perl->Idoswitches;
12916 PL_dowarn = proto_perl->Idowarn;
12917 PL_sawampersand = proto_perl->Isawampersand;
12918 PL_unsafe = proto_perl->Iunsafe;
12919 PL_perldb = proto_perl->Iperldb;
12920 PL_perl_destruct_level = proto_perl->Iperl_destruct_level;
12921 PL_exit_flags = proto_perl->Iexit_flags;
12923 /* XXX time(&PL_basetime) when asked for? */
12924 PL_basetime = proto_perl->Ibasetime;
12926 PL_maxsysfd = proto_perl->Imaxsysfd;
12927 PL_statusvalue = proto_perl->Istatusvalue;
12929 PL_statusvalue_vms = proto_perl->Istatusvalue_vms;
12931 PL_statusvalue_posix = proto_perl->Istatusvalue_posix;
12934 /* RE engine related */
12935 Zero(&PL_reg_state, 1, struct re_save_state);
12936 PL_regmatch_slab = NULL;
12938 PL_sub_generation = proto_perl->Isub_generation;
12940 /* funky return mechanisms */
12941 PL_forkprocess = proto_perl->Iforkprocess;
12943 /* internal state */
12944 PL_maxo = proto_perl->Imaxo;
12946 PL_main_start = proto_perl->Imain_start;
12947 PL_eval_root = proto_perl->Ieval_root;
12948 PL_eval_start = proto_perl->Ieval_start;
12950 PL_filemode = proto_perl->Ifilemode;
12951 PL_lastfd = proto_perl->Ilastfd;
12952 PL_oldname = proto_perl->Ioldname; /* XXX not quite right */
12955 PL_gensym = proto_perl->Igensym;
12957 PL_laststatval = proto_perl->Ilaststatval;
12958 PL_laststype = proto_perl->Ilaststype;
12961 PL_profiledata = NULL;
12963 PL_generation = proto_perl->Igeneration;
12965 PL_in_clean_objs = proto_perl->Iin_clean_objs;
12966 PL_in_clean_all = proto_perl->Iin_clean_all;
12968 PL_delaymagic_uid = proto_perl->Idelaymagic_uid;
12969 PL_delaymagic_euid = proto_perl->Idelaymagic_euid;
12970 PL_delaymagic_gid = proto_perl->Idelaymagic_gid;
12971 PL_delaymagic_egid = proto_perl->Idelaymagic_egid;
12972 PL_nomemok = proto_perl->Inomemok;
12973 PL_an = proto_perl->Ian;
12974 PL_evalseq = proto_perl->Ievalseq;
12975 PL_origenviron = proto_perl->Iorigenviron; /* XXX not quite right */
12976 PL_origalen = proto_perl->Iorigalen;
12978 PL_sighandlerp = proto_perl->Isighandlerp;
12980 PL_runops = proto_perl->Irunops;
12982 PL_subline = proto_perl->Isubline;
12985 PL_cryptseen = proto_perl->Icryptseen;
12988 PL_hints = proto_perl->Ihints;
12990 #ifdef USE_LOCALE_COLLATE
12991 PL_collation_ix = proto_perl->Icollation_ix;
12992 PL_collation_standard = proto_perl->Icollation_standard;
12993 PL_collxfrm_base = proto_perl->Icollxfrm_base;
12994 PL_collxfrm_mult = proto_perl->Icollxfrm_mult;
12995 #endif /* USE_LOCALE_COLLATE */
12997 #ifdef USE_LOCALE_NUMERIC
12998 PL_numeric_standard = proto_perl->Inumeric_standard;
12999 PL_numeric_local = proto_perl->Inumeric_local;
13000 #endif /* !USE_LOCALE_NUMERIC */
13002 /* Did the locale setup indicate UTF-8? */
13003 PL_utf8locale = proto_perl->Iutf8locale;
13004 /* Unicode features (see perlrun/-C) */
13005 PL_unicode = proto_perl->Iunicode;
13007 /* Pre-5.8 signals control */
13008 PL_signals = proto_perl->Isignals;
13010 /* times() ticks per second */
13011 PL_clocktick = proto_perl->Iclocktick;
13013 /* Recursion stopper for PerlIO_find_layer */
13014 PL_in_load_module = proto_perl->Iin_load_module;
13016 /* sort() routine */
13017 PL_sort_RealCmp = proto_perl->Isort_RealCmp;
13019 /* Not really needed/useful since the reenrant_retint is "volatile",
13020 * but do it for consistency's sake. */
13021 PL_reentrant_retint = proto_perl->Ireentrant_retint;
13023 /* Hooks to shared SVs and locks. */
13024 PL_sharehook = proto_perl->Isharehook;
13025 PL_lockhook = proto_perl->Ilockhook;
13026 PL_unlockhook = proto_perl->Iunlockhook;
13027 PL_threadhook = proto_perl->Ithreadhook;
13028 PL_destroyhook = proto_perl->Idestroyhook;
13029 PL_signalhook = proto_perl->Isignalhook;
13031 PL_globhook = proto_perl->Iglobhook;
13034 PL_last_swash_hv = NULL; /* reinits on demand */
13035 PL_last_swash_klen = 0;
13036 PL_last_swash_key[0]= '\0';
13037 PL_last_swash_tmps = (U8*)NULL;
13038 PL_last_swash_slen = 0;
13040 PL_glob_index = proto_perl->Iglob_index;
13041 PL_srand_called = proto_perl->Isrand_called;
13043 if (flags & CLONEf_COPY_STACKS) {
13044 /* next allocation will be PL_tmps_stack[PL_tmps_ix+1] */
13045 PL_tmps_ix = proto_perl->Itmps_ix;
13046 PL_tmps_max = proto_perl->Itmps_max;
13047 PL_tmps_floor = proto_perl->Itmps_floor;
13049 /* next push_scope()/ENTER sets PL_scopestack[PL_scopestack_ix]
13050 * NOTE: unlike the others! */
13051 PL_scopestack_ix = proto_perl->Iscopestack_ix;
13052 PL_scopestack_max = proto_perl->Iscopestack_max;
13054 /* next SSPUSHFOO() sets PL_savestack[PL_savestack_ix]
13055 * NOTE: unlike the others! */
13056 PL_savestack_ix = proto_perl->Isavestack_ix;
13057 PL_savestack_max = proto_perl->Isavestack_max;
13060 PL_start_env = proto_perl->Istart_env; /* XXXXXX */
13061 PL_top_env = &PL_start_env;
13063 PL_op = proto_perl->Iop;
13066 PL_Xpv = (XPV*)NULL;
13067 my_perl->Ina = proto_perl->Ina;
13069 PL_statbuf = proto_perl->Istatbuf;
13070 PL_statcache = proto_perl->Istatcache;
13073 PL_timesbuf = proto_perl->Itimesbuf;
13076 PL_tainted = proto_perl->Itainted;
13077 PL_curpm = proto_perl->Icurpm; /* XXX No PMOP ref count */
13079 PL_chopset = proto_perl->Ichopset; /* XXX never deallocated */
13081 PL_restartjmpenv = proto_perl->Irestartjmpenv;
13082 PL_restartop = proto_perl->Irestartop;
13083 PL_in_eval = proto_perl->Iin_eval;
13084 PL_delaymagic = proto_perl->Idelaymagic;
13085 PL_phase = proto_perl->Iphase;
13086 PL_localizing = proto_perl->Ilocalizing;
13088 PL_hv_fetch_ent_mh = NULL;
13089 PL_modcount = proto_perl->Imodcount;
13090 PL_lastgotoprobe = NULL;
13091 PL_dumpindent = proto_perl->Idumpindent;
13093 PL_efloatbuf = NULL; /* reinits on demand */
13094 PL_efloatsize = 0; /* reinits on demand */
13098 PL_regdummy = proto_perl->Iregdummy;
13099 PL_colorset = 0; /* reinits PL_colors[] */
13100 /*PL_colors[6] = {0,0,0,0,0,0};*/
13102 /* Pluggable optimizer */
13103 PL_peepp = proto_perl->Ipeepp;
13104 PL_rpeepp = proto_perl->Irpeepp;
13105 /* op_free() hook */
13106 PL_opfreehook = proto_perl->Iopfreehook;
13108 #ifdef USE_REENTRANT_API
13109 /* XXX: things like -Dm will segfault here in perlio, but doing
13110 * PERL_SET_CONTEXT(proto_perl);
13111 * breaks too many other things
13113 Perl_reentrant_init(aTHX);
13116 /* create SV map for pointer relocation */
13117 PL_ptr_table = ptr_table_new();
13119 /* initialize these special pointers as early as possible */
13121 ptr_table_store(PL_ptr_table, &proto_perl->Isv_undef, &PL_sv_undef);
13122 ptr_table_store(PL_ptr_table, &proto_perl->Isv_no, &PL_sv_no);
13123 ptr_table_store(PL_ptr_table, &proto_perl->Isv_yes, &PL_sv_yes);
13125 /* create (a non-shared!) shared string table */
13126 PL_strtab = newHV();
13127 HvSHAREKEYS_off(PL_strtab);
13128 hv_ksplit(PL_strtab, HvTOTALKEYS(proto_perl->Istrtab));
13129 ptr_table_store(PL_ptr_table, proto_perl->Istrtab, PL_strtab);
13131 /* This PV will be free'd special way so must set it same way op.c does */
13132 PL_compiling.cop_file = savesharedpv(PL_compiling.cop_file);
13133 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_file, PL_compiling.cop_file);
13135 ptr_table_store(PL_ptr_table, &proto_perl->Icompiling, &PL_compiling);
13136 PL_compiling.cop_warnings = DUP_WARNINGS(PL_compiling.cop_warnings);
13137 CopHINTHASH_set(&PL_compiling, cophh_copy(CopHINTHASH_get(&PL_compiling)));
13138 PL_curcop = (COP*)any_dup(proto_perl->Icurcop, proto_perl);
13140 param->stashes = newAV(); /* Setup array of objects to call clone on */
13141 /* This makes no difference to the implementation, as it always pushes
13142 and shifts pointers to other SVs without changing their reference
13143 count, with the array becoming empty before it is freed. However, it
13144 makes it conceptually clear what is going on, and will avoid some
13145 work inside av.c, filling slots between AvFILL() and AvMAX() with
13146 &PL_sv_undef, and SvREFCNT_dec()ing those. */
13147 AvREAL_off(param->stashes);
13149 if (!(flags & CLONEf_COPY_STACKS)) {
13150 param->unreferenced = newAV();
13153 #ifdef PERLIO_LAYERS
13154 /* Clone PerlIO tables as soon as we can handle general xx_dup() */
13155 PerlIO_clone(aTHX_ proto_perl, param);
13158 PL_envgv = gv_dup(proto_perl->Ienvgv, param);
13159 PL_incgv = gv_dup(proto_perl->Iincgv, param);
13160 PL_hintgv = gv_dup(proto_perl->Ihintgv, param);
13161 PL_origfilename = SAVEPV(proto_perl->Iorigfilename);
13162 PL_diehook = sv_dup_inc(proto_perl->Idiehook, param);
13163 PL_warnhook = sv_dup_inc(proto_perl->Iwarnhook, param);
13166 PL_patchlevel = sv_dup_inc(proto_perl->Ipatchlevel, param);
13167 PL_apiversion = sv_dup_inc(proto_perl->Iapiversion, param);
13168 PL_inplace = SAVEPV(proto_perl->Iinplace);
13169 PL_e_script = sv_dup_inc(proto_perl->Ie_script, param);
13171 /* magical thingies */
13172 PL_formfeed = sv_dup(proto_perl->Iformfeed, param);
13174 PL_encoding = sv_dup(proto_perl->Iencoding, param);
13176 sv_setpvs(PERL_DEBUG_PAD(0), ""); /* For regex debugging. */
13177 sv_setpvs(PERL_DEBUG_PAD(1), ""); /* ext/re needs these */
13178 sv_setpvs(PERL_DEBUG_PAD(2), ""); /* even without DEBUGGING. */
13181 /* Clone the regex array */
13182 /* ORANGE FIXME for plugins, probably in the SV dup code.
13183 newSViv(PTR2IV(CALLREGDUPE(
13184 INT2PTR(REGEXP *, SvIVX(regex)), param))))
13186 PL_regex_padav = av_dup_inc(proto_perl->Iregex_padav, param);
13187 PL_regex_pad = AvARRAY(PL_regex_padav);
13189 PL_stashpadmax = proto_perl->Istashpadmax;
13190 PL_stashpadix = proto_perl->Istashpadix ;
13191 Newx(PL_stashpad, PL_stashpadmax, HV *);
13194 for (; o < PL_stashpadmax; ++o)
13195 PL_stashpad[o] = hv_dup(proto_perl->Istashpad[o], param);
13198 /* shortcuts to various I/O objects */
13199 PL_ofsgv = gv_dup_inc(proto_perl->Iofsgv, param);
13200 PL_stdingv = gv_dup(proto_perl->Istdingv, param);
13201 PL_stderrgv = gv_dup(proto_perl->Istderrgv, param);
13202 PL_defgv = gv_dup(proto_perl->Idefgv, param);
13203 PL_argvgv = gv_dup(proto_perl->Iargvgv, param);
13204 PL_argvoutgv = gv_dup(proto_perl->Iargvoutgv, param);
13205 PL_argvout_stack = av_dup_inc(proto_perl->Iargvout_stack, param);
13207 /* shortcuts to regexp stuff */
13208 PL_replgv = gv_dup(proto_perl->Ireplgv, param);
13210 /* shortcuts to misc objects */
13211 PL_errgv = gv_dup(proto_perl->Ierrgv, param);
13213 /* shortcuts to debugging objects */
13214 PL_DBgv = gv_dup(proto_perl->IDBgv, param);
13215 PL_DBline = gv_dup(proto_perl->IDBline, param);
13216 PL_DBsub = gv_dup(proto_perl->IDBsub, param);
13217 PL_DBsingle = sv_dup(proto_perl->IDBsingle, param);
13218 PL_DBtrace = sv_dup(proto_perl->IDBtrace, param);
13219 PL_DBsignal = sv_dup(proto_perl->IDBsignal, param);
13221 /* symbol tables */
13222 PL_defstash = hv_dup_inc(proto_perl->Idefstash, param);
13223 PL_curstash = hv_dup_inc(proto_perl->Icurstash, param);
13224 PL_debstash = hv_dup(proto_perl->Idebstash, param);
13225 PL_globalstash = hv_dup(proto_perl->Iglobalstash, param);
13226 PL_curstname = sv_dup_inc(proto_perl->Icurstname, param);
13228 PL_beginav = av_dup_inc(proto_perl->Ibeginav, param);
13229 PL_beginav_save = av_dup_inc(proto_perl->Ibeginav_save, param);
13230 PL_checkav_save = av_dup_inc(proto_perl->Icheckav_save, param);
13231 PL_unitcheckav = av_dup_inc(proto_perl->Iunitcheckav, param);
13232 PL_unitcheckav_save = av_dup_inc(proto_perl->Iunitcheckav_save, param);
13233 PL_endav = av_dup_inc(proto_perl->Iendav, param);
13234 PL_checkav = av_dup_inc(proto_perl->Icheckav, param);
13235 PL_initav = av_dup_inc(proto_perl->Iinitav, param);
13237 PL_isarev = hv_dup_inc(proto_perl->Iisarev, param);
13239 /* subprocess state */
13240 PL_fdpid = av_dup_inc(proto_perl->Ifdpid, param);
13242 if (proto_perl->Iop_mask)
13243 PL_op_mask = SAVEPVN(proto_perl->Iop_mask, PL_maxo);
13246 /* PL_asserting = proto_perl->Iasserting; */
13248 /* current interpreter roots */
13249 PL_main_cv = cv_dup_inc(proto_perl->Imain_cv, param);
13251 PL_main_root = OpREFCNT_inc(proto_perl->Imain_root);
13254 /* runtime control stuff */
13255 PL_curcopdb = (COP*)any_dup(proto_perl->Icurcopdb, proto_perl);
13257 PL_preambleav = av_dup_inc(proto_perl->Ipreambleav, param);
13259 PL_ors_sv = sv_dup_inc(proto_perl->Iors_sv, param);
13261 /* interpreter atexit processing */
13262 PL_exitlistlen = proto_perl->Iexitlistlen;
13263 if (PL_exitlistlen) {
13264 Newx(PL_exitlist, PL_exitlistlen, PerlExitListEntry);
13265 Copy(proto_perl->Iexitlist, PL_exitlist, PL_exitlistlen, PerlExitListEntry);
13268 PL_exitlist = (PerlExitListEntry*)NULL;
13270 PL_my_cxt_size = proto_perl->Imy_cxt_size;
13271 if (PL_my_cxt_size) {
13272 Newx(PL_my_cxt_list, PL_my_cxt_size, void *);
13273 Copy(proto_perl->Imy_cxt_list, PL_my_cxt_list, PL_my_cxt_size, void *);
13274 #ifdef PERL_GLOBAL_STRUCT_PRIVATE
13275 Newx(PL_my_cxt_keys, PL_my_cxt_size, const char *);
13276 Copy(proto_perl->Imy_cxt_keys, PL_my_cxt_keys, PL_my_cxt_size, char *);
13280 PL_my_cxt_list = (void**)NULL;
13281 #ifdef PERL_GLOBAL_STRUCT_PRIVATE
13282 PL_my_cxt_keys = (const char**)NULL;
13285 PL_modglobal = hv_dup_inc(proto_perl->Imodglobal, param);
13286 PL_custom_op_names = hv_dup_inc(proto_perl->Icustom_op_names,param);
13287 PL_custom_op_descs = hv_dup_inc(proto_perl->Icustom_op_descs,param);
13288 PL_custom_ops = hv_dup_inc(proto_perl->Icustom_ops, param);
13290 PL_compcv = cv_dup(proto_perl->Icompcv, param);
13292 PAD_CLONE_VARS(proto_perl, param);
13294 #ifdef HAVE_INTERP_INTERN
13295 sys_intern_dup(&proto_perl->Isys_intern, &PL_sys_intern);
13298 PL_DBcv = cv_dup(proto_perl->IDBcv, param);
13300 #ifdef PERL_USES_PL_PIDSTATUS
13301 PL_pidstatus = newHV(); /* XXX flag for cloning? */
13303 PL_osname = SAVEPV(proto_perl->Iosname);
13304 PL_parser = parser_dup(proto_perl->Iparser, param);
13306 /* XXX this only works if the saved cop has already been cloned */
13307 if (proto_perl->Iparser) {
13308 PL_parser->saved_curcop = (COP*)any_dup(
13309 proto_perl->Iparser->saved_curcop,
13313 PL_subname = sv_dup_inc(proto_perl->Isubname, param);
13315 #ifdef USE_LOCALE_COLLATE
13316 PL_collation_name = SAVEPV(proto_perl->Icollation_name);
13317 #endif /* USE_LOCALE_COLLATE */
13319 #ifdef USE_LOCALE_NUMERIC
13320 PL_numeric_name = SAVEPV(proto_perl->Inumeric_name);
13321 PL_numeric_radix_sv = sv_dup_inc(proto_perl->Inumeric_radix_sv, param);
13322 #endif /* !USE_LOCALE_NUMERIC */
13324 /* Unicode inversion lists */
13325 PL_ASCII = sv_dup_inc(proto_perl->IASCII, param);
13326 PL_Latin1 = sv_dup_inc(proto_perl->ILatin1, param);
13328 PL_PerlSpace = sv_dup_inc(proto_perl->IPerlSpace, param);
13329 PL_XPerlSpace = sv_dup_inc(proto_perl->IXPerlSpace, param);
13331 PL_L1PosixAlnum = sv_dup_inc(proto_perl->IL1PosixAlnum, param);
13332 PL_PosixAlnum = sv_dup_inc(proto_perl->IPosixAlnum, param);
13334 PL_L1PosixAlpha = sv_dup_inc(proto_perl->IL1PosixAlpha, param);
13335 PL_PosixAlpha = sv_dup_inc(proto_perl->IPosixAlpha, param);
13337 PL_PosixBlank = sv_dup_inc(proto_perl->IPosixBlank, param);
13338 PL_XPosixBlank = sv_dup_inc(proto_perl->IXPosixBlank, param);
13340 PL_L1Cased = sv_dup_inc(proto_perl->IL1Cased, param);
13342 PL_PosixCntrl = sv_dup_inc(proto_perl->IPosixCntrl, param);
13343 PL_XPosixCntrl = sv_dup_inc(proto_perl->IXPosixCntrl, param);
13345 PL_PosixDigit = sv_dup_inc(proto_perl->IPosixDigit, param);
13347 PL_L1PosixGraph = sv_dup_inc(proto_perl->IL1PosixGraph, param);
13348 PL_PosixGraph = sv_dup_inc(proto_perl->IPosixGraph, param);
13350 PL_L1PosixLower = sv_dup_inc(proto_perl->IL1PosixLower, param);
13351 PL_PosixLower = sv_dup_inc(proto_perl->IPosixLower, param);
13353 PL_L1PosixPrint = sv_dup_inc(proto_perl->IL1PosixPrint, param);
13354 PL_PosixPrint = sv_dup_inc(proto_perl->IPosixPrint, param);
13356 PL_L1PosixPunct = sv_dup_inc(proto_perl->IL1PosixPunct, param);
13357 PL_PosixPunct = sv_dup_inc(proto_perl->IPosixPunct, param);
13359 PL_PosixSpace = sv_dup_inc(proto_perl->IPosixSpace, param);
13360 PL_XPosixSpace = sv_dup_inc(proto_perl->IXPosixSpace, param);
13362 PL_L1PosixUpper = sv_dup_inc(proto_perl->IL1PosixUpper, param);
13363 PL_PosixUpper = sv_dup_inc(proto_perl->IPosixUpper, param);
13365 PL_L1PosixWord = sv_dup_inc(proto_perl->IL1PosixWord, param);
13366 PL_PosixWord = sv_dup_inc(proto_perl->IPosixWord, param);
13368 PL_PosixXDigit = sv_dup_inc(proto_perl->IPosixXDigit, param);
13369 PL_XPosixXDigit = sv_dup_inc(proto_perl->IXPosixXDigit, param);
13371 PL_VertSpace = sv_dup_inc(proto_perl->IVertSpace, param);
13373 PL_NonL1NonFinalFold = sv_dup_inc(proto_perl->INonL1NonFinalFold, param);
13375 /* utf8 character class swashes */
13376 PL_utf8_alnum = sv_dup_inc(proto_perl->Iutf8_alnum, param);
13377 PL_utf8_alpha = sv_dup_inc(proto_perl->Iutf8_alpha, param);
13378 PL_utf8_blank = sv_dup_inc(proto_perl->Iutf8_blank, param);
13379 PL_utf8_space = sv_dup_inc(proto_perl->Iutf8_space, param);
13380 PL_utf8_graph = sv_dup_inc(proto_perl->Iutf8_graph, param);
13381 PL_utf8_digit = sv_dup_inc(proto_perl->Iutf8_digit, param);
13382 PL_utf8_upper = sv_dup_inc(proto_perl->Iutf8_upper, param);
13383 PL_utf8_lower = sv_dup_inc(proto_perl->Iutf8_lower, param);
13384 PL_utf8_print = sv_dup_inc(proto_perl->Iutf8_print, param);
13385 PL_utf8_punct = sv_dup_inc(proto_perl->Iutf8_punct, param);
13386 PL_utf8_xdigit = sv_dup_inc(proto_perl->Iutf8_xdigit, param);
13387 PL_utf8_mark = sv_dup_inc(proto_perl->Iutf8_mark, param);
13388 PL_utf8_X_begin = sv_dup_inc(proto_perl->Iutf8_X_begin, param);
13389 PL_utf8_X_extend = sv_dup_inc(proto_perl->Iutf8_X_extend, param);
13390 PL_utf8_X_prepend = sv_dup_inc(proto_perl->Iutf8_X_prepend, param);
13391 PL_utf8_X_non_hangul = sv_dup_inc(proto_perl->Iutf8_X_non_hangul, param);
13392 PL_utf8_X_L = sv_dup_inc(proto_perl->Iutf8_X_L, param);
13393 PL_utf8_X_LV = sv_dup_inc(proto_perl->Iutf8_X_LV, param);
13394 PL_utf8_X_LVT = sv_dup_inc(proto_perl->Iutf8_X_LVT, param);
13395 PL_utf8_X_T = sv_dup_inc(proto_perl->Iutf8_X_T, param);
13396 PL_utf8_X_V = sv_dup_inc(proto_perl->Iutf8_X_V, param);
13397 PL_utf8_X_LV_LVT_V = sv_dup_inc(proto_perl->Iutf8_X_LV_LVT_V, param);
13398 PL_utf8_toupper = sv_dup_inc(proto_perl->Iutf8_toupper, param);
13399 PL_utf8_totitle = sv_dup_inc(proto_perl->Iutf8_totitle, param);
13400 PL_utf8_tolower = sv_dup_inc(proto_perl->Iutf8_tolower, param);
13401 PL_utf8_tofold = sv_dup_inc(proto_perl->Iutf8_tofold, param);
13402 PL_utf8_idstart = sv_dup_inc(proto_perl->Iutf8_idstart, param);
13403 PL_utf8_xidstart = sv_dup_inc(proto_perl->Iutf8_xidstart, param);
13404 PL_utf8_perl_idstart = sv_dup_inc(proto_perl->Iutf8_perl_idstart, param);
13405 PL_utf8_idcont = sv_dup_inc(proto_perl->Iutf8_idcont, param);
13406 PL_utf8_xidcont = sv_dup_inc(proto_perl->Iutf8_xidcont, param);
13407 PL_utf8_foldable = sv_dup_inc(proto_perl->Iutf8_foldable, param);
13408 PL_utf8_quotemeta = sv_dup_inc(proto_perl->Iutf8_quotemeta, param);
13409 PL_ASCII = sv_dup_inc(proto_perl->IASCII, param);
13410 PL_AboveLatin1 = sv_dup_inc(proto_perl->IAboveLatin1, param);
13411 PL_Latin1 = sv_dup_inc(proto_perl->ILatin1, param);
13414 if (proto_perl->Ipsig_pend) {
13415 Newxz(PL_psig_pend, SIG_SIZE, int);
13418 PL_psig_pend = (int*)NULL;
13421 if (proto_perl->Ipsig_name) {
13422 Newx(PL_psig_name, 2 * SIG_SIZE, SV*);
13423 sv_dup_inc_multiple(proto_perl->Ipsig_name, PL_psig_name, 2 * SIG_SIZE,
13425 PL_psig_ptr = PL_psig_name + SIG_SIZE;
13428 PL_psig_ptr = (SV**)NULL;
13429 PL_psig_name = (SV**)NULL;
13432 if (flags & CLONEf_COPY_STACKS) {
13433 Newx(PL_tmps_stack, PL_tmps_max, SV*);
13434 sv_dup_inc_multiple(proto_perl->Itmps_stack, PL_tmps_stack,
13435 PL_tmps_ix+1, param);
13437 /* next PUSHMARK() sets *(PL_markstack_ptr+1) */
13438 i = proto_perl->Imarkstack_max - proto_perl->Imarkstack;
13439 Newxz(PL_markstack, i, I32);
13440 PL_markstack_max = PL_markstack + (proto_perl->Imarkstack_max
13441 - proto_perl->Imarkstack);
13442 PL_markstack_ptr = PL_markstack + (proto_perl->Imarkstack_ptr
13443 - proto_perl->Imarkstack);
13444 Copy(proto_perl->Imarkstack, PL_markstack,
13445 PL_markstack_ptr - PL_markstack + 1, I32);
13447 /* next push_scope()/ENTER sets PL_scopestack[PL_scopestack_ix]
13448 * NOTE: unlike the others! */
13449 Newxz(PL_scopestack, PL_scopestack_max, I32);
13450 Copy(proto_perl->Iscopestack, PL_scopestack, PL_scopestack_ix, I32);
13453 Newxz(PL_scopestack_name, PL_scopestack_max, const char *);
13454 Copy(proto_perl->Iscopestack_name, PL_scopestack_name, PL_scopestack_ix, const char *);
13456 /* NOTE: si_dup() looks at PL_markstack */
13457 PL_curstackinfo = si_dup(proto_perl->Icurstackinfo, param);
13459 /* PL_curstack = PL_curstackinfo->si_stack; */
13460 PL_curstack = av_dup(proto_perl->Icurstack, param);
13461 PL_mainstack = av_dup(proto_perl->Imainstack, param);
13463 /* next PUSHs() etc. set *(PL_stack_sp+1) */
13464 PL_stack_base = AvARRAY(PL_curstack);
13465 PL_stack_sp = PL_stack_base + (proto_perl->Istack_sp
13466 - proto_perl->Istack_base);
13467 PL_stack_max = PL_stack_base + AvMAX(PL_curstack);
13469 /*Newxz(PL_savestack, PL_savestack_max, ANY);*/
13470 PL_savestack = ss_dup(proto_perl, param);
13474 ENTER; /* perl_destruct() wants to LEAVE; */
13477 PL_statgv = gv_dup(proto_perl->Istatgv, param);
13478 PL_statname = sv_dup_inc(proto_perl->Istatname, param);
13480 PL_rs = sv_dup_inc(proto_perl->Irs, param);
13481 PL_last_in_gv = gv_dup(proto_perl->Ilast_in_gv, param);
13482 PL_defoutgv = gv_dup_inc(proto_perl->Idefoutgv, param);
13483 PL_toptarget = sv_dup_inc(proto_perl->Itoptarget, param);
13484 PL_bodytarget = sv_dup_inc(proto_perl->Ibodytarget, param);
13485 PL_formtarget = sv_dup(proto_perl->Iformtarget, param);
13487 PL_errors = sv_dup_inc(proto_perl->Ierrors, param);
13489 PL_sortcop = (OP*)any_dup(proto_perl->Isortcop, proto_perl);
13490 PL_sortstash = hv_dup(proto_perl->Isortstash, param);
13491 PL_firstgv = gv_dup(proto_perl->Ifirstgv, param);
13492 PL_secondgv = gv_dup(proto_perl->Isecondgv, param);
13494 PL_stashcache = newHV();
13496 PL_watchaddr = (char **) ptr_table_fetch(PL_ptr_table,
13497 proto_perl->Iwatchaddr);
13498 PL_watchok = PL_watchaddr ? * PL_watchaddr : NULL;
13499 if (PL_debug && PL_watchaddr) {
13500 PerlIO_printf(Perl_debug_log,
13501 "WATCHING: %"UVxf" cloned as %"UVxf" with value %"UVxf"\n",
13502 PTR2UV(proto_perl->Iwatchaddr), PTR2UV(PL_watchaddr),
13503 PTR2UV(PL_watchok));
13506 PL_registered_mros = hv_dup_inc(proto_perl->Iregistered_mros, param);
13507 PL_blockhooks = av_dup_inc(proto_perl->Iblockhooks, param);
13508 PL_utf8_foldclosures = hv_dup_inc(proto_perl->Iutf8_foldclosures, param);
13510 /* Call the ->CLONE method, if it exists, for each of the stashes
13511 identified by sv_dup() above.
13513 while(av_len(param->stashes) != -1) {
13514 HV* const stash = MUTABLE_HV(av_shift(param->stashes));
13515 GV* const cloner = gv_fetchmethod_autoload(stash, "CLONE", 0);
13516 if (cloner && GvCV(cloner)) {
13521 mXPUSHs(newSVhek(HvNAME_HEK(stash)));
13523 call_sv(MUTABLE_SV(GvCV(cloner)), G_DISCARD);
13529 if (!(flags & CLONEf_KEEP_PTR_TABLE)) {
13530 ptr_table_free(PL_ptr_table);
13531 PL_ptr_table = NULL;
13534 if (!(flags & CLONEf_COPY_STACKS)) {
13535 unreferenced_to_tmp_stack(param->unreferenced);
13538 SvREFCNT_dec(param->stashes);
13540 /* orphaned? eg threads->new inside BEGIN or use */
13541 if (PL_compcv && ! SvREFCNT(PL_compcv)) {
13542 SvREFCNT_inc_simple_void(PL_compcv);
13543 SAVEFREESV(PL_compcv);
13550 S_unreferenced_to_tmp_stack(pTHX_ AV *const unreferenced)
13552 PERL_ARGS_ASSERT_UNREFERENCED_TO_TMP_STACK;
13554 if (AvFILLp(unreferenced) > -1) {
13555 SV **svp = AvARRAY(unreferenced);
13556 SV **const last = svp + AvFILLp(unreferenced);
13560 if (SvREFCNT(*svp) == 1)
13562 } while (++svp <= last);
13564 EXTEND_MORTAL(count);
13565 svp = AvARRAY(unreferenced);
13568 if (SvREFCNT(*svp) == 1) {
13569 /* Our reference is the only one to this SV. This means that
13570 in this thread, the scalar effectively has a 0 reference.
13571 That doesn't work (cleanup never happens), so donate our
13572 reference to it onto the save stack. */
13573 PL_tmps_stack[++PL_tmps_ix] = *svp;
13575 /* As an optimisation, because we are already walking the
13576 entire array, instead of above doing either
13577 SvREFCNT_inc(*svp) or *svp = &PL_sv_undef, we can instead
13578 release our reference to the scalar, so that at the end of
13579 the array owns zero references to the scalars it happens to
13580 point to. We are effectively converting the array from
13581 AvREAL() on to AvREAL() off. This saves the av_clear()
13582 (triggered by the SvREFCNT_dec(unreferenced) below) from
13583 walking the array a second time. */
13584 SvREFCNT_dec(*svp);
13587 } while (++svp <= last);
13588 AvREAL_off(unreferenced);
13590 SvREFCNT_dec(unreferenced);
13594 Perl_clone_params_del(CLONE_PARAMS *param)
13596 /* This seemingly funky ordering keeps the build with PERL_GLOBAL_STRUCT
13598 PerlInterpreter *const to = param->new_perl;
13600 PerlInterpreter *const was = PERL_GET_THX;
13602 PERL_ARGS_ASSERT_CLONE_PARAMS_DEL;
13608 SvREFCNT_dec(param->stashes);
13609 if (param->unreferenced)
13610 unreferenced_to_tmp_stack(param->unreferenced);
13620 Perl_clone_params_new(PerlInterpreter *const from, PerlInterpreter *const to)
13623 /* Need to play this game, as newAV() can call safesysmalloc(), and that
13624 does a dTHX; to get the context from thread local storage.
13625 FIXME - under PERL_CORE Newx(), Safefree() and friends should expand to
13626 a version that passes in my_perl. */
13627 PerlInterpreter *const was = PERL_GET_THX;
13628 CLONE_PARAMS *param;
13630 PERL_ARGS_ASSERT_CLONE_PARAMS_NEW;
13636 /* Given that we've set the context, we can do this unshared. */
13637 Newx(param, 1, CLONE_PARAMS);
13640 param->proto_perl = from;
13641 param->new_perl = to;
13642 param->stashes = (AV *)Perl_newSV_type(to, SVt_PVAV);
13643 AvREAL_off(param->stashes);
13644 param->unreferenced = (AV *)Perl_newSV_type(to, SVt_PVAV);
13652 #endif /* USE_ITHREADS */
13655 Perl_init_constants(pTHX)
13657 SvREFCNT(&PL_sv_undef) = (~(U32)0)/2;
13658 SvFLAGS(&PL_sv_undef) = SVf_READONLY|SVt_NULL;
13659 SvANY(&PL_sv_undef) = NULL;
13661 SvANY(&PL_sv_no) = new_XPVNV();
13662 SvREFCNT(&PL_sv_no) = (~(U32)0)/2;
13663 SvFLAGS(&PL_sv_no) = SVt_PVNV|SVf_READONLY
13664 |SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
13667 SvANY(&PL_sv_yes) = new_XPVNV();
13668 SvREFCNT(&PL_sv_yes) = (~(U32)0)/2;
13669 SvFLAGS(&PL_sv_yes) = SVt_PVNV|SVf_READONLY
13670 |SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
13673 SvPV_set(&PL_sv_no, (char*)PL_No);
13674 SvCUR_set(&PL_sv_no, 0);
13675 SvLEN_set(&PL_sv_no, 0);
13676 SvIV_set(&PL_sv_no, 0);
13677 SvNV_set(&PL_sv_no, 0);
13679 SvPV_set(&PL_sv_yes, (char*)PL_Yes);
13680 SvCUR_set(&PL_sv_yes, 1);
13681 SvLEN_set(&PL_sv_yes, 0);
13682 SvIV_set(&PL_sv_yes, 1);
13683 SvNV_set(&PL_sv_yes, 1);
13687 =head1 Unicode Support
13689 =for apidoc sv_recode_to_utf8
13691 The encoding is assumed to be an Encode object, on entry the PV
13692 of the sv is assumed to be octets in that encoding, and the sv
13693 will be converted into Unicode (and UTF-8).
13695 If the sv already is UTF-8 (or if it is not POK), or if the encoding
13696 is not a reference, nothing is done to the sv. If the encoding is not
13697 an C<Encode::XS> Encoding object, bad things will happen.
13698 (See F<lib/encoding.pm> and L<Encode>.)
13700 The PV of the sv is returned.
13705 Perl_sv_recode_to_utf8(pTHX_ SV *sv, SV *encoding)
13709 PERL_ARGS_ASSERT_SV_RECODE_TO_UTF8;
13711 if (SvPOK(sv) && !SvUTF8(sv) && !IN_BYTES && SvROK(encoding)) {
13725 Passing sv_yes is wrong - it needs to be or'ed set of constants
13726 for Encode::XS, while UTf-8 decode (currently) assumes a true value means
13727 remove converted chars from source.
13729 Both will default the value - let them.
13731 XPUSHs(&PL_sv_yes);
13734 call_method("decode", G_SCALAR);
13738 s = SvPV_const(uni, len);
13739 if (s != SvPVX_const(sv)) {
13740 SvGROW(sv, len + 1);
13741 Move(s, SvPVX(sv), len + 1, char);
13742 SvCUR_set(sv, len);
13746 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
13747 /* clear pos and any utf8 cache */
13748 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
13751 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
13752 magic_setutf8(sv,mg); /* clear UTF8 cache */
13757 return SvPOKp(sv) ? SvPVX(sv) : NULL;
13761 =for apidoc sv_cat_decode
13763 The encoding is assumed to be an Encode object, the PV of the ssv is
13764 assumed to be octets in that encoding and decoding the input starts
13765 from the position which (PV + *offset) pointed to. The dsv will be
13766 concatenated the decoded UTF-8 string from ssv. Decoding will terminate
13767 when the string tstr appears in decoding output or the input ends on
13768 the PV of the ssv. The value which the offset points will be modified
13769 to the last input position on the ssv.
13771 Returns TRUE if the terminator was found, else returns FALSE.
13776 Perl_sv_cat_decode(pTHX_ SV *dsv, SV *encoding,
13777 SV *ssv, int *offset, char *tstr, int tlen)
13782 PERL_ARGS_ASSERT_SV_CAT_DECODE;
13784 if (SvPOK(ssv) && SvPOK(dsv) && SvROK(encoding) && offset) {
13795 offsv = newSViv(*offset);
13797 mXPUSHp(tstr, tlen);
13799 call_method("cat_decode", G_SCALAR);
13801 ret = SvTRUE(TOPs);
13802 *offset = SvIV(offsv);
13808 Perl_croak(aTHX_ "Invalid argument to sv_cat_decode");
13813 /* ---------------------------------------------------------------------
13815 * support functions for report_uninit()
13818 /* the maxiumum size of array or hash where we will scan looking
13819 * for the undefined element that triggered the warning */
13821 #define FUV_MAX_SEARCH_SIZE 1000
13823 /* Look for an entry in the hash whose value has the same SV as val;
13824 * If so, return a mortal copy of the key. */
13827 S_find_hash_subscript(pTHX_ const HV *const hv, const SV *const val)
13830 register HE **array;
13833 PERL_ARGS_ASSERT_FIND_HASH_SUBSCRIPT;
13835 if (!hv || SvMAGICAL(hv) || !HvARRAY(hv) ||
13836 (HvTOTALKEYS(hv) > FUV_MAX_SEARCH_SIZE))
13839 array = HvARRAY(hv);
13841 for (i=HvMAX(hv); i>0; i--) {
13842 register HE *entry;
13843 for (entry = array[i]; entry; entry = HeNEXT(entry)) {
13844 if (HeVAL(entry) != val)
13846 if ( HeVAL(entry) == &PL_sv_undef ||
13847 HeVAL(entry) == &PL_sv_placeholder)
13851 if (HeKLEN(entry) == HEf_SVKEY)
13852 return sv_mortalcopy(HeKEY_sv(entry));
13853 return sv_2mortal(newSVhek(HeKEY_hek(entry)));
13859 /* Look for an entry in the array whose value has the same SV as val;
13860 * If so, return the index, otherwise return -1. */
13863 S_find_array_subscript(pTHX_ const AV *const av, const SV *const val)
13867 PERL_ARGS_ASSERT_FIND_ARRAY_SUBSCRIPT;
13869 if (!av || SvMAGICAL(av) || !AvARRAY(av) ||
13870 (AvFILLp(av) > FUV_MAX_SEARCH_SIZE))
13873 if (val != &PL_sv_undef) {
13874 SV ** const svp = AvARRAY(av);
13877 for (i=AvFILLp(av); i>=0; i--)
13884 /* varname(): return the name of a variable, optionally with a subscript.
13885 * If gv is non-zero, use the name of that global, along with gvtype (one
13886 * of "$", "@", "%"); otherwise use the name of the lexical at pad offset
13887 * targ. Depending on the value of the subscript_type flag, return:
13890 #define FUV_SUBSCRIPT_NONE 1 /* "@foo" */
13891 #define FUV_SUBSCRIPT_ARRAY 2 /* "$foo[aindex]" */
13892 #define FUV_SUBSCRIPT_HASH 3 /* "$foo{keyname}" */
13893 #define FUV_SUBSCRIPT_WITHIN 4 /* "within @foo" */
13896 Perl_varname(pTHX_ const GV *const gv, const char gvtype, PADOFFSET targ,
13897 const SV *const keyname, I32 aindex, int subscript_type)
13900 SV * const name = sv_newmortal();
13901 if (gv && isGV(gv)) {
13903 buffer[0] = gvtype;
13906 /* as gv_fullname4(), but add literal '^' for $^FOO names */
13908 gv_fullname4(name, gv, buffer, 0);
13910 if ((unsigned int)SvPVX(name)[1] <= 26) {
13912 buffer[1] = SvPVX(name)[1] + 'A' - 1;
13914 /* Swap the 1 unprintable control character for the 2 byte pretty
13915 version - ie substr($name, 1, 1) = $buffer; */
13916 sv_insert(name, 1, 1, buffer, 2);
13920 CV * const cv = gv ? ((CV *)gv) : find_runcv(NULL);
13924 assert(!cv || SvTYPE(cv) == SVt_PVCV);
13926 if (!cv || !CvPADLIST(cv))
13928 av = MUTABLE_AV((*av_fetch(CvPADLIST(cv), 0, FALSE)));
13929 sv = *av_fetch(av, targ, FALSE);
13930 sv_setsv(name, sv);
13933 if (subscript_type == FUV_SUBSCRIPT_HASH) {
13934 SV * const sv = newSV(0);
13935 *SvPVX(name) = '$';
13936 Perl_sv_catpvf(aTHX_ name, "{%s}",
13937 pv_pretty(sv, SvPVX_const(keyname), SvCUR(keyname), 32, NULL, NULL,
13938 PERL_PV_PRETTY_DUMP | PERL_PV_ESCAPE_UNI_DETECT ));
13941 else if (subscript_type == FUV_SUBSCRIPT_ARRAY) {
13942 *SvPVX(name) = '$';
13943 Perl_sv_catpvf(aTHX_ name, "[%"IVdf"]", (IV)aindex);
13945 else if (subscript_type == FUV_SUBSCRIPT_WITHIN) {
13946 /* We know that name has no magic, so can use 0 instead of SV_GMAGIC */
13947 Perl_sv_insert_flags(aTHX_ name, 0, 0, STR_WITH_LEN("within "), 0);
13955 =for apidoc find_uninit_var
13957 Find the name of the undefined variable (if any) that caused the operator
13958 to issue a "Use of uninitialized value" warning.
13959 If match is true, only return a name if its value matches uninit_sv.
13960 So roughly speaking, if a unary operator (such as OP_COS) generates a
13961 warning, then following the direct child of the op may yield an
13962 OP_PADSV or OP_GV that gives the name of the undefined variable. On the
13963 other hand, with OP_ADD there are two branches to follow, so we only print
13964 the variable name if we get an exact match.
13966 The name is returned as a mortal SV.
13968 Assumes that PL_op is the op that originally triggered the error, and that
13969 PL_comppad/PL_curpad points to the currently executing pad.
13975 S_find_uninit_var(pTHX_ const OP *const obase, const SV *const uninit_sv,
13981 const OP *o, *o2, *kid;
13983 if (!obase || (match && (!uninit_sv || uninit_sv == &PL_sv_undef ||
13984 uninit_sv == &PL_sv_placeholder)))
13987 switch (obase->op_type) {
13994 const bool pad = (obase->op_type == OP_PADAV || obase->op_type == OP_PADHV);
13995 const bool hash = (obase->op_type == OP_PADHV || obase->op_type == OP_RV2HV);
13998 int subscript_type = FUV_SUBSCRIPT_WITHIN;
14000 if (pad) { /* @lex, %lex */
14001 sv = PAD_SVl(obase->op_targ);
14005 if (cUNOPx(obase)->op_first->op_type == OP_GV) {
14006 /* @global, %global */
14007 gv = cGVOPx_gv(cUNOPx(obase)->op_first);
14010 sv = hash ? MUTABLE_SV(GvHV(gv)): MUTABLE_SV(GvAV(gv));
14012 else if (obase == PL_op) /* @{expr}, %{expr} */
14013 return find_uninit_var(cUNOPx(obase)->op_first,
14015 else /* @{expr}, %{expr} as a sub-expression */
14019 /* attempt to find a match within the aggregate */
14021 keysv = find_hash_subscript((const HV*)sv, uninit_sv);
14023 subscript_type = FUV_SUBSCRIPT_HASH;
14026 index = find_array_subscript((const AV *)sv, uninit_sv);
14028 subscript_type = FUV_SUBSCRIPT_ARRAY;
14031 if (match && subscript_type == FUV_SUBSCRIPT_WITHIN)
14034 return varname(gv, hash ? '%' : '@', obase->op_targ,
14035 keysv, index, subscript_type);
14039 if (cUNOPx(obase)->op_first->op_type == OP_GV) {
14041 gv = cGVOPx_gv(cUNOPx(obase)->op_first);
14042 if (!gv || !GvSTASH(gv))
14044 if (match && (GvSV(gv) != uninit_sv))
14046 return varname(gv, '$', 0, NULL, 0, FUV_SUBSCRIPT_NONE);
14049 return find_uninit_var(cUNOPx(obase)->op_first, uninit_sv, 1);
14052 if (match && PAD_SVl(obase->op_targ) != uninit_sv)
14054 return varname(NULL, '$', obase->op_targ,
14055 NULL, 0, FUV_SUBSCRIPT_NONE);
14058 gv = cGVOPx_gv(obase);
14059 if (!gv || (match && GvSV(gv) != uninit_sv) || !GvSTASH(gv))
14061 return varname(gv, '$', 0, NULL, 0, FUV_SUBSCRIPT_NONE);
14063 case OP_AELEMFAST_LEX:
14066 AV *av = MUTABLE_AV(PAD_SV(obase->op_targ));
14067 if (!av || SvRMAGICAL(av))
14069 svp = av_fetch(av, (I32)obase->op_private, FALSE);
14070 if (!svp || *svp != uninit_sv)
14073 return varname(NULL, '$', obase->op_targ,
14074 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
14077 gv = cGVOPx_gv(obase);
14082 AV *const av = GvAV(gv);
14083 if (!av || SvRMAGICAL(av))
14085 svp = av_fetch(av, (I32)obase->op_private, FALSE);
14086 if (!svp || *svp != uninit_sv)
14089 return varname(gv, '$', 0,
14090 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
14095 o = cUNOPx(obase)->op_first;
14096 if (!o || o->op_type != OP_NULL ||
14097 ! (o->op_targ == OP_AELEM || o->op_targ == OP_HELEM))
14099 return find_uninit_var(cBINOPo->op_last, uninit_sv, match);
14104 bool negate = FALSE;
14106 if (PL_op == obase)
14107 /* $a[uninit_expr] or $h{uninit_expr} */
14108 return find_uninit_var(cBINOPx(obase)->op_last, uninit_sv, match);
14111 o = cBINOPx(obase)->op_first;
14112 kid = cBINOPx(obase)->op_last;
14114 /* get the av or hv, and optionally the gv */
14116 if (o->op_type == OP_PADAV || o->op_type == OP_PADHV) {
14117 sv = PAD_SV(o->op_targ);
14119 else if ((o->op_type == OP_RV2AV || o->op_type == OP_RV2HV)
14120 && cUNOPo->op_first->op_type == OP_GV)
14122 gv = cGVOPx_gv(cUNOPo->op_first);
14126 == OP_RV2HV ? MUTABLE_SV(GvHV(gv)) : MUTABLE_SV(GvAV(gv));
14131 if (kid && kid->op_type == OP_NEGATE) {
14133 kid = cUNOPx(kid)->op_first;
14136 if (kid && kid->op_type == OP_CONST && SvOK(cSVOPx_sv(kid))) {
14137 /* index is constant */
14140 kidsv = sv_2mortal(newSVpvs("-"));
14141 sv_catsv(kidsv, cSVOPx_sv(kid));
14144 kidsv = cSVOPx_sv(kid);
14148 if (obase->op_type == OP_HELEM) {
14149 HE* he = hv_fetch_ent(MUTABLE_HV(sv), kidsv, 0, 0);
14150 if (!he || HeVAL(he) != uninit_sv)
14154 SV * const * const svp = av_fetch(MUTABLE_AV(sv),
14155 negate ? - SvIV(cSVOPx_sv(kid)) : SvIV(cSVOPx_sv(kid)),
14157 if (!svp || *svp != uninit_sv)
14161 if (obase->op_type == OP_HELEM)
14162 return varname(gv, '%', o->op_targ,
14163 kidsv, 0, FUV_SUBSCRIPT_HASH);
14165 return varname(gv, '@', o->op_targ, NULL,
14166 negate ? - SvIV(cSVOPx_sv(kid)) : SvIV(cSVOPx_sv(kid)),
14167 FUV_SUBSCRIPT_ARRAY);
14170 /* index is an expression;
14171 * attempt to find a match within the aggregate */
14172 if (obase->op_type == OP_HELEM) {
14173 SV * const keysv = find_hash_subscript((const HV*)sv, uninit_sv);
14175 return varname(gv, '%', o->op_targ,
14176 keysv, 0, FUV_SUBSCRIPT_HASH);
14180 = find_array_subscript((const AV *)sv, uninit_sv);
14182 return varname(gv, '@', o->op_targ,
14183 NULL, index, FUV_SUBSCRIPT_ARRAY);
14188 (o->op_type == OP_PADAV || o->op_type == OP_RV2AV)
14190 o->op_targ, NULL, 0, FUV_SUBSCRIPT_WITHIN);
14196 /* only examine RHS */
14197 return find_uninit_var(cBINOPx(obase)->op_first, uninit_sv, match);
14200 o = cUNOPx(obase)->op_first;
14201 if (o->op_type == OP_PUSHMARK)
14204 if (!o->op_sibling) {
14205 /* one-arg version of open is highly magical */
14207 if (o->op_type == OP_GV) { /* open FOO; */
14209 if (match && GvSV(gv) != uninit_sv)
14211 return varname(gv, '$', 0,
14212 NULL, 0, FUV_SUBSCRIPT_NONE);
14214 /* other possibilities not handled are:
14215 * open $x; or open my $x; should return '${*$x}'
14216 * open expr; should return '$'.expr ideally
14222 /* ops where $_ may be an implicit arg */
14227 if ( !(obase->op_flags & OPf_STACKED)) {
14228 if (uninit_sv == ((obase->op_private & OPpTARGET_MY)
14229 ? PAD_SVl(obase->op_targ)
14232 sv = sv_newmortal();
14233 sv_setpvs(sv, "$_");
14242 match = 1; /* print etc can return undef on defined args */
14243 /* skip filehandle as it can't produce 'undef' warning */
14244 o = cUNOPx(obase)->op_first;
14245 if ((obase->op_flags & OPf_STACKED) && o->op_type == OP_PUSHMARK)
14246 o = o->op_sibling->op_sibling;
14250 case OP_ENTEREVAL: /* could be eval $undef or $x='$undef'; eval $x */
14251 case OP_CUSTOM: /* XS or custom code could trigger random warnings */
14253 /* the following ops are capable of returning PL_sv_undef even for
14254 * defined arg(s) */
14273 case OP_GETPEERNAME:
14321 case OP_SMARTMATCH:
14330 /* XXX tmp hack: these two may call an XS sub, and currently
14331 XS subs don't have a SUB entry on the context stack, so CV and
14332 pad determination goes wrong, and BAD things happen. So, just
14333 don't try to determine the value under those circumstances.
14334 Need a better fix at dome point. DAPM 11/2007 */
14340 GV * const gv = gv_fetchpvs(".", GV_NOTQUAL, SVt_PV);
14341 if (gv && GvSV(gv) == uninit_sv)
14342 return newSVpvs_flags("$.", SVs_TEMP);
14347 /* def-ness of rval pos() is independent of the def-ness of its arg */
14348 if ( !(obase->op_flags & OPf_MOD))
14353 if (SvROK(PL_rs) && uninit_sv == SvRV(PL_rs))
14354 return newSVpvs_flags("${$/}", SVs_TEMP);
14359 if (!(obase->op_flags & OPf_KIDS))
14361 o = cUNOPx(obase)->op_first;
14367 /* This loop checks all the kid ops, skipping any that cannot pos-
14368 * sibly be responsible for the uninitialized value; i.e., defined
14369 * constants and ops that return nothing. If there is only one op
14370 * left that is not skipped, then we *know* it is responsible for
14371 * the uninitialized value. If there is more than one op left, we
14372 * have to look for an exact match in the while() loop below.
14375 for (kid=o; kid; kid = kid->op_sibling) {
14377 const OPCODE type = kid->op_type;
14378 if ( (type == OP_CONST && SvOK(cSVOPx_sv(kid)))
14379 || (type == OP_NULL && ! (kid->op_flags & OPf_KIDS))
14380 || (type == OP_PUSHMARK)
14384 if (o2) { /* more than one found */
14391 return find_uninit_var(o2, uninit_sv, match);
14393 /* scan all args */
14395 sv = find_uninit_var(o, uninit_sv, 1);
14407 =for apidoc report_uninit
14409 Print appropriate "Use of uninitialized variable" warning.
14415 Perl_report_uninit(pTHX_ const SV *uninit_sv)
14419 SV* varname = NULL;
14420 if (uninit_sv && PL_curpad) {
14421 varname = find_uninit_var(PL_op, uninit_sv,0);
14423 sv_insert(varname, 0, 0, " ", 1);
14425 /* diag_listed_as: Use of uninitialized value%s */
14426 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit_sv,
14427 SVfARG(varname ? varname : &PL_sv_no),
14428 " in ", OP_DESC(PL_op));
14431 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
14437 * c-indentation-style: bsd
14438 * c-basic-offset: 4
14439 * indent-tabs-mode: nil
14442 * ex: set ts=8 sts=4 sw=4 et: