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) != 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) != 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 /* 8 bytes on most ILP32 with IEEE doubles */
897 { sizeof(NV), sizeof(NV),
898 STRUCT_OFFSET(XPVNV, xnv_u),
899 SVt_NV, FALSE, HADNV, HASARENA, FIT_ARENA(0, sizeof(NV)) },
901 /* 8 bytes on most ILP32 with IEEE doubles */
902 { sizeof(XPV) - STRUCT_OFFSET(XPV, xpv_cur),
903 copy_length(XPV, xpv_len) - STRUCT_OFFSET(XPV, xpv_cur),
904 + STRUCT_OFFSET(XPV, xpv_cur),
905 SVt_PV, FALSE, NONV, HASARENA,
906 FIT_ARENA(0, sizeof(XPV) - STRUCT_OFFSET(XPV, xpv_cur)) },
909 { sizeof(XPVIV) - STRUCT_OFFSET(XPV, xpv_cur),
910 copy_length(XPVIV, xiv_u) - STRUCT_OFFSET(XPV, xpv_cur),
911 + STRUCT_OFFSET(XPV, xpv_cur),
912 SVt_PVIV, FALSE, NONV, HASARENA,
913 FIT_ARENA(0, sizeof(XPVIV) - STRUCT_OFFSET(XPV, xpv_cur)) },
916 { sizeof(XPVNV) - STRUCT_OFFSET(XPV, xpv_cur),
917 copy_length(XPVNV, xnv_u) - STRUCT_OFFSET(XPV, xpv_cur),
918 + STRUCT_OFFSET(XPV, xpv_cur),
919 SVt_PVNV, FALSE, HADNV, HASARENA,
920 FIT_ARENA(0, sizeof(XPVNV) - STRUCT_OFFSET(XPV, xpv_cur)) },
923 { sizeof(XPVMG), copy_length(XPVMG, xnv_u), 0, SVt_PVMG, FALSE, HADNV,
924 HASARENA, FIT_ARENA(0, sizeof(XPVMG)) },
930 SVt_REGEXP, FALSE, NONV, HASARENA,
931 FIT_ARENA(0, sizeof(regexp))
935 { sizeof(XPVGV), sizeof(XPVGV), 0, SVt_PVGV, TRUE, HADNV,
936 HASARENA, FIT_ARENA(0, sizeof(XPVGV)) },
939 { sizeof(XPVLV), sizeof(XPVLV), 0, SVt_PVLV, TRUE, HADNV,
940 HASARENA, FIT_ARENA(0, sizeof(XPVLV)) },
943 copy_length(XPVAV, xav_alloc),
945 SVt_PVAV, TRUE, NONV, HASARENA,
946 FIT_ARENA(0, sizeof(XPVAV)) },
949 copy_length(XPVHV, xhv_max),
951 SVt_PVHV, TRUE, NONV, HASARENA,
952 FIT_ARENA(0, sizeof(XPVHV)) },
958 SVt_PVCV, TRUE, NONV, HASARENA,
959 FIT_ARENA(0, sizeof(XPVCV)) },
964 SVt_PVFM, TRUE, NONV, NOARENA,
965 FIT_ARENA(20, sizeof(XPVFM)) },
967 /* XPVIO is 84 bytes, fits 48x */
971 SVt_PVIO, TRUE, NONV, HASARENA,
972 FIT_ARENA(24, sizeof(XPVIO)) },
975 #define new_body_allocated(sv_type) \
976 (void *)((char *)S_new_body(aTHX_ sv_type) \
977 - bodies_by_type[sv_type].offset)
979 /* return a thing to the free list */
981 #define del_body(thing, root) \
983 void ** const thing_copy = (void **)thing; \
984 *thing_copy = *root; \
985 *root = (void*)thing_copy; \
990 #define new_XNV() safemalloc(sizeof(XPVNV))
991 #define new_XPVNV() safemalloc(sizeof(XPVNV))
992 #define new_XPVMG() safemalloc(sizeof(XPVMG))
994 #define del_XPVGV(p) safefree(p)
998 #define new_XNV() new_body_allocated(SVt_NV)
999 #define new_XPVNV() new_body_allocated(SVt_PVNV)
1000 #define new_XPVMG() new_body_allocated(SVt_PVMG)
1002 #define del_XPVGV(p) del_body(p + bodies_by_type[SVt_PVGV].offset, \
1003 &PL_body_roots[SVt_PVGV])
1007 /* no arena for you! */
1009 #define new_NOARENA(details) \
1010 safemalloc((details)->body_size + (details)->offset)
1011 #define new_NOARENAZ(details) \
1012 safecalloc((details)->body_size + (details)->offset, 1)
1015 Perl_more_bodies (pTHX_ const svtype sv_type, const size_t body_size,
1016 const size_t arena_size)
1019 void ** const root = &PL_body_roots[sv_type];
1020 struct arena_desc *adesc;
1021 struct arena_set *aroot = (struct arena_set *) PL_body_arenas;
1025 const size_t good_arena_size = Perl_malloc_good_size(arena_size);
1026 #if defined(DEBUGGING) && !defined(PERL_GLOBAL_STRUCT_PRIVATE)
1027 static bool done_sanity_check;
1029 /* PERL_GLOBAL_STRUCT_PRIVATE cannot coexist with global
1030 * variables like done_sanity_check. */
1031 if (!done_sanity_check) {
1032 unsigned int i = SVt_LAST;
1034 done_sanity_check = TRUE;
1037 assert (bodies_by_type[i].type == i);
1043 /* may need new arena-set to hold new arena */
1044 if (!aroot || aroot->curr >= aroot->set_size) {
1045 struct arena_set *newroot;
1046 Newxz(newroot, 1, struct arena_set);
1047 newroot->set_size = ARENAS_PER_SET;
1048 newroot->next = aroot;
1050 PL_body_arenas = (void *) newroot;
1051 DEBUG_m(PerlIO_printf(Perl_debug_log, "new arenaset %p\n", (void*)aroot));
1054 /* ok, now have arena-set with at least 1 empty/available arena-desc */
1055 curr = aroot->curr++;
1056 adesc = &(aroot->set[curr]);
1057 assert(!adesc->arena);
1059 Newx(adesc->arena, good_arena_size, char);
1060 adesc->size = good_arena_size;
1061 adesc->utype = sv_type;
1062 DEBUG_m(PerlIO_printf(Perl_debug_log, "arena %d added: %p size %"UVuf"\n",
1063 curr, (void*)adesc->arena, (UV)good_arena_size));
1065 start = (char *) adesc->arena;
1067 /* Get the address of the byte after the end of the last body we can fit.
1068 Remember, this is integer division: */
1069 end = start + good_arena_size / body_size * body_size;
1071 /* computed count doesn't reflect the 1st slot reservation */
1072 #if defined(MYMALLOC) || defined(HAS_MALLOC_GOOD_SIZE)
1073 DEBUG_m(PerlIO_printf(Perl_debug_log,
1074 "arena %p end %p arena-size %d (from %d) type %d "
1076 (void*)start, (void*)end, (int)good_arena_size,
1077 (int)arena_size, sv_type, (int)body_size,
1078 (int)good_arena_size / (int)body_size));
1080 DEBUG_m(PerlIO_printf(Perl_debug_log,
1081 "arena %p end %p arena-size %d type %d size %d ct %d\n",
1082 (void*)start, (void*)end,
1083 (int)arena_size, sv_type, (int)body_size,
1084 (int)good_arena_size / (int)body_size));
1086 *root = (void *)start;
1089 /* Where the next body would start: */
1090 char * const next = start + body_size;
1093 /* This is the last body: */
1094 assert(next == end);
1096 *(void **)start = 0;
1100 *(void**) start = (void *)next;
1105 /* grab a new thing from the free list, allocating more if necessary.
1106 The inline version is used for speed in hot routines, and the
1107 function using it serves the rest (unless PURIFY).
1109 #define new_body_inline(xpv, sv_type) \
1111 void ** const r3wt = &PL_body_roots[sv_type]; \
1112 xpv = (PTR_TBL_ENT_t*) (*((void **)(r3wt)) \
1113 ? *((void **)(r3wt)) : Perl_more_bodies(aTHX_ sv_type, \
1114 bodies_by_type[sv_type].body_size,\
1115 bodies_by_type[sv_type].arena_size)); \
1116 *(r3wt) = *(void**)(xpv); \
1122 S_new_body(pTHX_ const svtype sv_type)
1126 new_body_inline(xpv, sv_type);
1132 static const struct body_details fake_rv =
1133 { 0, 0, 0, SVt_IV, FALSE, NONV, NOARENA, 0 };
1136 =for apidoc sv_upgrade
1138 Upgrade an SV to a more complex form. Generally adds a new body type to the
1139 SV, then copies across as much information as possible from the old body.
1140 You generally want to use the C<SvUPGRADE> macro wrapper. See also C<svtype>.
1146 Perl_sv_upgrade(pTHX_ register SV *const sv, svtype new_type)
1151 const svtype old_type = SvTYPE(sv);
1152 const struct body_details *new_type_details;
1153 const struct body_details *old_type_details
1154 = bodies_by_type + old_type;
1155 SV *referant = NULL;
1157 PERL_ARGS_ASSERT_SV_UPGRADE;
1159 if (old_type == new_type)
1162 /* This clause was purposefully added ahead of the early return above to
1163 the shared string hackery for (sort {$a <=> $b} keys %hash), with the
1164 inference by Nick I-S that it would fix other troublesome cases. See
1165 changes 7162, 7163 (f130fd4589cf5fbb24149cd4db4137c8326f49c1 and parent)
1167 Given that shared hash key scalars are no longer PVIV, but PV, there is
1168 no longer need to unshare so as to free up the IVX slot for its proper
1169 purpose. So it's safe to move the early return earlier. */
1171 if (new_type != SVt_PV && SvIsCOW(sv)) {
1172 sv_force_normal_flags(sv, 0);
1175 old_body = SvANY(sv);
1177 /* Copying structures onto other structures that have been neatly zeroed
1178 has a subtle gotcha. Consider XPVMG
1180 +------+------+------+------+------+-------+-------+
1181 | NV | CUR | LEN | IV | MAGIC | STASH |
1182 +------+------+------+------+------+-------+-------+
1183 0 4 8 12 16 20 24 28
1185 where NVs are aligned to 8 bytes, so that sizeof that structure is
1186 actually 32 bytes long, with 4 bytes of padding at the end:
1188 +------+------+------+------+------+-------+-------+------+
1189 | NV | CUR | LEN | IV | MAGIC | STASH | ??? |
1190 +------+------+------+------+------+-------+-------+------+
1191 0 4 8 12 16 20 24 28 32
1193 so what happens if you allocate memory for this structure:
1195 +------+------+------+------+------+-------+-------+------+------+...
1196 | NV | CUR | LEN | IV | MAGIC | STASH | GP | NAME |
1197 +------+------+------+------+------+-------+-------+------+------+...
1198 0 4 8 12 16 20 24 28 32 36
1200 zero it, then copy sizeof(XPVMG) bytes on top of it? Not quite what you
1201 expect, because you copy the area marked ??? onto GP. Now, ??? may have
1202 started out as zero once, but it's quite possible that it isn't. So now,
1203 rather than a nicely zeroed GP, you have it pointing somewhere random.
1206 (In fact, GP ends up pointing at a previous GP structure, because the
1207 principle cause of the padding in XPVMG getting garbage is a copy of
1208 sizeof(XPVMG) bytes from a XPVGV structure in sv_unglob. Right now
1209 this happens to be moot because XPVGV has been re-ordered, with GP
1210 no longer after STASH)
1212 So we are careful and work out the size of used parts of all the
1220 referant = SvRV(sv);
1221 old_type_details = &fake_rv;
1222 if (new_type == SVt_NV)
1223 new_type = SVt_PVNV;
1225 if (new_type < SVt_PVIV) {
1226 new_type = (new_type == SVt_NV)
1227 ? SVt_PVNV : SVt_PVIV;
1232 if (new_type < SVt_PVNV) {
1233 new_type = SVt_PVNV;
1237 assert(new_type > SVt_PV);
1238 assert(SVt_IV < SVt_PV);
1239 assert(SVt_NV < SVt_PV);
1246 /* Because the XPVMG of PL_mess_sv isn't allocated from the arena,
1247 there's no way that it can be safely upgraded, because perl.c
1248 expects to Safefree(SvANY(PL_mess_sv)) */
1249 assert(sv != PL_mess_sv);
1250 /* This flag bit is used to mean other things in other scalar types.
1251 Given that it only has meaning inside the pad, it shouldn't be set
1252 on anything that can get upgraded. */
1253 assert(!SvPAD_TYPED(sv));
1256 if (old_type_details->cant_upgrade)
1257 Perl_croak(aTHX_ "Can't upgrade %s (%" UVuf ") to %" UVuf,
1258 sv_reftype(sv, 0), (UV) old_type, (UV) new_type);
1261 if (old_type > new_type)
1262 Perl_croak(aTHX_ "sv_upgrade from type %d down to type %d",
1263 (int)old_type, (int)new_type);
1265 new_type_details = bodies_by_type + new_type;
1267 SvFLAGS(sv) &= ~SVTYPEMASK;
1268 SvFLAGS(sv) |= new_type;
1270 /* This can't happen, as SVt_NULL is <= all values of new_type, so one of
1271 the return statements above will have triggered. */
1272 assert (new_type != SVt_NULL);
1275 assert(old_type == SVt_NULL);
1276 SvANY(sv) = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
1280 assert(old_type == SVt_NULL);
1281 SvANY(sv) = new_XNV();
1286 assert(new_type_details->body_size);
1289 assert(new_type_details->arena);
1290 assert(new_type_details->arena_size);
1291 /* This points to the start of the allocated area. */
1292 new_body_inline(new_body, new_type);
1293 Zero(new_body, new_type_details->body_size, char);
1294 new_body = ((char *)new_body) - new_type_details->offset;
1296 /* We always allocated the full length item with PURIFY. To do this
1297 we fake things so that arena is false for all 16 types.. */
1298 new_body = new_NOARENAZ(new_type_details);
1300 SvANY(sv) = new_body;
1301 if (new_type == SVt_PVAV) {
1305 if (old_type_details->body_size) {
1308 /* It will have been zeroed when the new body was allocated.
1309 Lets not write to it, in case it confuses a write-back
1315 #ifndef NODEFAULT_SHAREKEYS
1316 HvSHAREKEYS_on(sv); /* key-sharing on by default */
1318 HvMAX(sv) = 7; /* (start with 8 buckets) */
1321 /* SVt_NULL isn't the only thing upgraded to AV or HV.
1322 The target created by newSVrv also is, and it can have magic.
1323 However, it never has SvPVX set.
1325 if (old_type == SVt_IV) {
1327 } else if (old_type >= SVt_PV) {
1328 assert(SvPVX_const(sv) == 0);
1331 if (old_type >= SVt_PVMG) {
1332 SvMAGIC_set(sv, ((XPVMG*)old_body)->xmg_u.xmg_magic);
1333 SvSTASH_set(sv, ((XPVMG*)old_body)->xmg_stash);
1335 sv->sv_u.svu_array = NULL; /* or svu_hash */
1341 /* This ensures that SvTHINKFIRST(sv) is true, and hence that
1342 sv_force_normal_flags(sv) is called. */
1345 /* XXX Is this still needed? Was it ever needed? Surely as there is
1346 no route from NV to PVIV, NOK can never be true */
1347 assert(!SvNOKp(sv));
1358 assert(new_type_details->body_size);
1359 /* We always allocated the full length item with PURIFY. To do this
1360 we fake things so that arena is false for all 16 types.. */
1361 if(new_type_details->arena) {
1362 /* This points to the start of the allocated area. */
1363 new_body_inline(new_body, new_type);
1364 Zero(new_body, new_type_details->body_size, char);
1365 new_body = ((char *)new_body) - new_type_details->offset;
1367 new_body = new_NOARENAZ(new_type_details);
1369 SvANY(sv) = new_body;
1371 if (old_type_details->copy) {
1372 /* There is now the potential for an upgrade from something without
1373 an offset (PVNV or PVMG) to something with one (PVCV, PVFM) */
1374 int offset = old_type_details->offset;
1375 int length = old_type_details->copy;
1377 if (new_type_details->offset > old_type_details->offset) {
1378 const int difference
1379 = new_type_details->offset - old_type_details->offset;
1380 offset += difference;
1381 length -= difference;
1383 assert (length >= 0);
1385 Copy((char *)old_body + offset, (char *)new_body + offset, length,
1389 #ifndef NV_ZERO_IS_ALLBITS_ZERO
1390 /* If NV 0.0 is stores as all bits 0 then Zero() already creates a
1391 * correct 0.0 for us. Otherwise, if the old body didn't have an
1392 * NV slot, but the new one does, then we need to initialise the
1393 * freshly created NV slot with whatever the correct bit pattern is
1395 if (old_type_details->zero_nv && !new_type_details->zero_nv
1396 && !isGV_with_GP(sv))
1400 if (new_type == SVt_PVIO) {
1401 IO * const io = MUTABLE_IO(sv);
1402 GV *iogv = gv_fetchpvs("IO::File::", GV_ADD, SVt_PVHV);
1405 /* Clear the stashcache because a new IO could overrule a package
1407 hv_clear(PL_stashcache);
1409 SvSTASH_set(io, MUTABLE_HV(SvREFCNT_inc(GvHV(iogv))));
1410 IoPAGE_LEN(sv) = 60;
1412 if (old_type < SVt_PV) {
1413 /* referant will be NULL unless the old type was SVt_IV emulating
1415 sv->sv_u.svu_rv = referant;
1419 Perl_croak(aTHX_ "panic: sv_upgrade to unknown type %lu",
1420 (unsigned long)new_type);
1423 if (old_type > SVt_IV) {
1427 /* Note that there is an assumption that all bodies of types that
1428 can be upgraded came from arenas. Only the more complex non-
1429 upgradable types are allowed to be directly malloc()ed. */
1430 assert(old_type_details->arena);
1431 del_body((void*)((char*)old_body + old_type_details->offset),
1432 &PL_body_roots[old_type]);
1438 =for apidoc sv_backoff
1440 Remove any string offset. You should normally use the C<SvOOK_off> macro
1447 Perl_sv_backoff(pTHX_ register SV *const sv)
1450 const char * const s = SvPVX_const(sv);
1452 PERL_ARGS_ASSERT_SV_BACKOFF;
1453 PERL_UNUSED_CONTEXT;
1456 assert(SvTYPE(sv) != SVt_PVHV);
1457 assert(SvTYPE(sv) != SVt_PVAV);
1459 SvOOK_offset(sv, delta);
1461 SvLEN_set(sv, SvLEN(sv) + delta);
1462 SvPV_set(sv, SvPVX(sv) - delta);
1463 Move(s, SvPVX(sv), SvCUR(sv)+1, char);
1464 SvFLAGS(sv) &= ~SVf_OOK;
1471 Expands the character buffer in the SV. If necessary, uses C<sv_unref> and
1472 upgrades the SV to C<SVt_PV>. Returns a pointer to the character buffer.
1473 Use the C<SvGROW> wrapper instead.
1479 Perl_sv_grow(pTHX_ register SV *const sv, register STRLEN newlen)
1483 PERL_ARGS_ASSERT_SV_GROW;
1485 if (PL_madskills && newlen >= 0x100000) {
1486 PerlIO_printf(Perl_debug_log,
1487 "Allocation too large: %"UVxf"\n", (UV)newlen);
1489 #ifdef HAS_64K_LIMIT
1490 if (newlen >= 0x10000) {
1491 PerlIO_printf(Perl_debug_log,
1492 "Allocation too large: %"UVxf"\n", (UV)newlen);
1495 #endif /* HAS_64K_LIMIT */
1498 if (SvTYPE(sv) < SVt_PV) {
1499 sv_upgrade(sv, SVt_PV);
1500 s = SvPVX_mutable(sv);
1502 else if (SvOOK(sv)) { /* pv is offset? */
1504 s = SvPVX_mutable(sv);
1505 if (newlen > SvLEN(sv))
1506 newlen += 10 * (newlen - SvCUR(sv)); /* avoid copy each time */
1507 #ifdef HAS_64K_LIMIT
1508 if (newlen >= 0x10000)
1513 s = SvPVX_mutable(sv);
1515 if (newlen > SvLEN(sv)) { /* need more room? */
1516 STRLEN minlen = SvCUR(sv);
1517 minlen += (minlen >> PERL_STRLEN_EXPAND_SHIFT) + 10;
1518 if (newlen < minlen)
1520 #ifndef Perl_safesysmalloc_size
1521 newlen = PERL_STRLEN_ROUNDUP(newlen);
1523 if (SvLEN(sv) && s) {
1524 s = (char*)saferealloc(s, newlen);
1527 s = (char*)safemalloc(newlen);
1528 if (SvPVX_const(sv) && SvCUR(sv)) {
1529 Move(SvPVX_const(sv), s, (newlen < SvCUR(sv)) ? newlen : SvCUR(sv), char);
1533 #ifdef Perl_safesysmalloc_size
1534 /* Do this here, do it once, do it right, and then we will never get
1535 called back into sv_grow() unless there really is some growing
1537 SvLEN_set(sv, Perl_safesysmalloc_size(s));
1539 SvLEN_set(sv, newlen);
1546 =for apidoc sv_setiv
1548 Copies an integer into the given SV, upgrading first if necessary.
1549 Does not handle 'set' magic. See also C<sv_setiv_mg>.
1555 Perl_sv_setiv(pTHX_ register SV *const sv, const IV i)
1559 PERL_ARGS_ASSERT_SV_SETIV;
1561 SV_CHECK_THINKFIRST_COW_DROP(sv);
1562 switch (SvTYPE(sv)) {
1565 sv_upgrade(sv, SVt_IV);
1568 sv_upgrade(sv, SVt_PVIV);
1572 if (!isGV_with_GP(sv))
1579 /* diag_listed_as: Can't coerce %s to %s in %s */
1580 Perl_croak(aTHX_ "Can't coerce %s to integer in %s", sv_reftype(sv,0),
1584 (void)SvIOK_only(sv); /* validate number */
1590 =for apidoc sv_setiv_mg
1592 Like C<sv_setiv>, but also handles 'set' magic.
1598 Perl_sv_setiv_mg(pTHX_ register SV *const sv, const IV i)
1600 PERL_ARGS_ASSERT_SV_SETIV_MG;
1607 =for apidoc sv_setuv
1609 Copies an unsigned integer into the given SV, upgrading first if necessary.
1610 Does not handle 'set' magic. See also C<sv_setuv_mg>.
1616 Perl_sv_setuv(pTHX_ register SV *const sv, const UV u)
1618 PERL_ARGS_ASSERT_SV_SETUV;
1620 /* With these two if statements:
1621 u=1.49 s=0.52 cu=72.49 cs=10.64 scripts=270 tests=20865
1624 u=1.35 s=0.47 cu=73.45 cs=11.43 scripts=270 tests=20865
1626 If you wish to remove them, please benchmark to see what the effect is
1628 if (u <= (UV)IV_MAX) {
1629 sv_setiv(sv, (IV)u);
1638 =for apidoc sv_setuv_mg
1640 Like C<sv_setuv>, but also handles 'set' magic.
1646 Perl_sv_setuv_mg(pTHX_ register SV *const sv, const UV u)
1648 PERL_ARGS_ASSERT_SV_SETUV_MG;
1655 =for apidoc sv_setnv
1657 Copies a double into the given SV, upgrading first if necessary.
1658 Does not handle 'set' magic. See also C<sv_setnv_mg>.
1664 Perl_sv_setnv(pTHX_ register SV *const sv, const NV num)
1668 PERL_ARGS_ASSERT_SV_SETNV;
1670 SV_CHECK_THINKFIRST_COW_DROP(sv);
1671 switch (SvTYPE(sv)) {
1674 sv_upgrade(sv, SVt_NV);
1678 sv_upgrade(sv, SVt_PVNV);
1682 if (!isGV_with_GP(sv))
1689 /* diag_listed_as: Can't coerce %s to %s in %s */
1690 Perl_croak(aTHX_ "Can't coerce %s to number in %s", sv_reftype(sv,0),
1695 (void)SvNOK_only(sv); /* validate number */
1700 =for apidoc sv_setnv_mg
1702 Like C<sv_setnv>, but also handles 'set' magic.
1708 Perl_sv_setnv_mg(pTHX_ register SV *const sv, const NV num)
1710 PERL_ARGS_ASSERT_SV_SETNV_MG;
1716 /* Print an "isn't numeric" warning, using a cleaned-up,
1717 * printable version of the offending string
1721 S_not_a_number(pTHX_ SV *const sv)
1728 PERL_ARGS_ASSERT_NOT_A_NUMBER;
1731 dsv = newSVpvs_flags("", SVs_TEMP);
1732 pv = sv_uni_display(dsv, sv, 10, 0);
1735 const char * const limit = tmpbuf + sizeof(tmpbuf) - 8;
1736 /* each *s can expand to 4 chars + "...\0",
1737 i.e. need room for 8 chars */
1739 const char *s = SvPVX_const(sv);
1740 const char * const end = s + SvCUR(sv);
1741 for ( ; s < end && d < limit; s++ ) {
1743 if (ch & 128 && !isPRINT_LC(ch)) {
1752 else if (ch == '\r') {
1756 else if (ch == '\f') {
1760 else if (ch == '\\') {
1764 else if (ch == '\0') {
1768 else if (isPRINT_LC(ch))
1785 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
1786 "Argument \"%s\" isn't numeric in %s", pv,
1789 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
1790 "Argument \"%s\" isn't numeric", pv);
1794 =for apidoc looks_like_number
1796 Test if the content of an SV looks like a number (or is a number).
1797 C<Inf> and C<Infinity> are treated as numbers (so will not issue a
1798 non-numeric warning), even if your atof() doesn't grok them.
1804 Perl_looks_like_number(pTHX_ SV *const sv)
1806 register const char *sbegin;
1809 PERL_ARGS_ASSERT_LOOKS_LIKE_NUMBER;
1812 sbegin = SvPVX_const(sv);
1815 else if (SvPOKp(sv))
1816 sbegin = SvPV_const(sv, len);
1818 return SvFLAGS(sv) & (SVf_NOK|SVp_NOK|SVf_IOK|SVp_IOK);
1819 return grok_number(sbegin, len, NULL);
1823 S_glob_2number(pTHX_ GV * const gv)
1825 const U32 wasfake = SvFLAGS(gv) & SVf_FAKE;
1826 SV *const buffer = sv_newmortal();
1828 PERL_ARGS_ASSERT_GLOB_2NUMBER;
1830 /* FAKE globs can get coerced, so need to turn this off temporarily if it
1833 gv_efullname3(buffer, gv, "*");
1834 SvFLAGS(gv) |= wasfake;
1836 /* We know that all GVs stringify to something that is not-a-number,
1837 so no need to test that. */
1838 if (ckWARN(WARN_NUMERIC))
1839 not_a_number(buffer);
1840 /* We just want something true to return, so that S_sv_2iuv_common
1841 can tail call us and return true. */
1845 /* Actually, ISO C leaves conversion of UV to IV undefined, but
1846 until proven guilty, assume that things are not that bad... */
1851 As 64 bit platforms often have an NV that doesn't preserve all bits of
1852 an IV (an assumption perl has been based on to date) it becomes necessary
1853 to remove the assumption that the NV always carries enough precision to
1854 recreate the IV whenever needed, and that the NV is the canonical form.
1855 Instead, IV/UV and NV need to be given equal rights. So as to not lose
1856 precision as a side effect of conversion (which would lead to insanity
1857 and the dragon(s) in t/op/numconvert.t getting very angry) the intent is
1858 1) to distinguish between IV/UV/NV slots that have cached a valid
1859 conversion where precision was lost and IV/UV/NV slots that have a
1860 valid conversion which has lost no precision
1861 2) to ensure that if a numeric conversion to one form is requested that
1862 would lose precision, the precise conversion (or differently
1863 imprecise conversion) is also performed and cached, to prevent
1864 requests for different numeric formats on the same SV causing
1865 lossy conversion chains. (lossless conversion chains are perfectly
1870 SvIOKp is true if the IV slot contains a valid value
1871 SvIOK is true only if the IV value is accurate (UV if SvIOK_UV true)
1872 SvNOKp is true if the NV slot contains a valid value
1873 SvNOK is true only if the NV value is accurate
1876 while converting from PV to NV, check to see if converting that NV to an
1877 IV(or UV) would lose accuracy over a direct conversion from PV to
1878 IV(or UV). If it would, cache both conversions, return NV, but mark
1879 SV as IOK NOKp (ie not NOK).
1881 While converting from PV to IV, check to see if converting that IV to an
1882 NV would lose accuracy over a direct conversion from PV to NV. If it
1883 would, cache both conversions, flag similarly.
1885 Before, the SV value "3.2" could become NV=3.2 IV=3 NOK, IOK quite
1886 correctly because if IV & NV were set NV *always* overruled.
1887 Now, "3.2" will become NV=3.2 IV=3 NOK, IOKp, because the flag's meaning
1888 changes - now IV and NV together means that the two are interchangeable:
1889 SvIVX == (IV) SvNVX && SvNVX == (NV) SvIVX;
1891 The benefit of this is that operations such as pp_add know that if
1892 SvIOK is true for both left and right operands, then integer addition
1893 can be used instead of floating point (for cases where the result won't
1894 overflow). Before, floating point was always used, which could lead to
1895 loss of precision compared with integer addition.
1897 * making IV and NV equal status should make maths accurate on 64 bit
1899 * may speed up maths somewhat if pp_add and friends start to use
1900 integers when possible instead of fp. (Hopefully the overhead in
1901 looking for SvIOK and checking for overflow will not outweigh the
1902 fp to integer speedup)
1903 * will slow down integer operations (callers of SvIV) on "inaccurate"
1904 values, as the change from SvIOK to SvIOKp will cause a call into
1905 sv_2iv each time rather than a macro access direct to the IV slot
1906 * should speed up number->string conversion on integers as IV is
1907 favoured when IV and NV are equally accurate
1909 ####################################################################
1910 You had better be using SvIOK_notUV if you want an IV for arithmetic:
1911 SvIOK is true if (IV or UV), so you might be getting (IV)SvUV.
1912 On the other hand, SvUOK is true iff UV.
1913 ####################################################################
1915 Your mileage will vary depending your CPU's relative fp to integer
1919 #ifndef NV_PRESERVES_UV
1920 # define IS_NUMBER_UNDERFLOW_IV 1
1921 # define IS_NUMBER_UNDERFLOW_UV 2
1922 # define IS_NUMBER_IV_AND_UV 2
1923 # define IS_NUMBER_OVERFLOW_IV 4
1924 # define IS_NUMBER_OVERFLOW_UV 5
1926 /* sv_2iuv_non_preserve(): private routine for use by sv_2iv() and sv_2uv() */
1928 /* For sv_2nv these three cases are "SvNOK and don't bother casting" */
1930 S_sv_2iuv_non_preserve(pTHX_ register SV *const sv
1938 PERL_ARGS_ASSERT_SV_2IUV_NON_PRESERVE;
1940 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));
1941 if (SvNVX(sv) < (NV)IV_MIN) {
1942 (void)SvIOKp_on(sv);
1944 SvIV_set(sv, IV_MIN);
1945 return IS_NUMBER_UNDERFLOW_IV;
1947 if (SvNVX(sv) > (NV)UV_MAX) {
1948 (void)SvIOKp_on(sv);
1951 SvUV_set(sv, UV_MAX);
1952 return IS_NUMBER_OVERFLOW_UV;
1954 (void)SvIOKp_on(sv);
1956 /* Can't use strtol etc to convert this string. (See truth table in
1958 if (SvNVX(sv) <= (UV)IV_MAX) {
1959 SvIV_set(sv, I_V(SvNVX(sv)));
1960 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
1961 SvIOK_on(sv); /* Integer is precise. NOK, IOK */
1963 /* Integer is imprecise. NOK, IOKp */
1965 return SvNVX(sv) < 0 ? IS_NUMBER_UNDERFLOW_UV : IS_NUMBER_IV_AND_UV;
1968 SvUV_set(sv, U_V(SvNVX(sv)));
1969 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
1970 if (SvUVX(sv) == UV_MAX) {
1971 /* As we know that NVs don't preserve UVs, UV_MAX cannot
1972 possibly be preserved by NV. Hence, it must be overflow.
1974 return IS_NUMBER_OVERFLOW_UV;
1976 SvIOK_on(sv); /* Integer is precise. NOK, UOK */
1978 /* Integer is imprecise. NOK, IOKp */
1980 return IS_NUMBER_OVERFLOW_IV;
1982 #endif /* !NV_PRESERVES_UV*/
1985 S_sv_2iuv_common(pTHX_ SV *const sv)
1989 PERL_ARGS_ASSERT_SV_2IUV_COMMON;
1992 /* erm. not sure. *should* never get NOKp (without NOK) from sv_2nv
1993 * without also getting a cached IV/UV from it at the same time
1994 * (ie PV->NV conversion should detect loss of accuracy and cache
1995 * IV or UV at same time to avoid this. */
1996 /* IV-over-UV optimisation - choose to cache IV if possible */
1998 if (SvTYPE(sv) == SVt_NV)
1999 sv_upgrade(sv, SVt_PVNV);
2001 (void)SvIOKp_on(sv); /* Must do this first, to clear any SvOOK */
2002 /* < not <= as for NV doesn't preserve UV, ((NV)IV_MAX+1) will almost
2003 certainly cast into the IV range at IV_MAX, whereas the correct
2004 answer is the UV IV_MAX +1. Hence < ensures that dodgy boundary
2006 #if defined(NAN_COMPARE_BROKEN) && defined(Perl_isnan)
2007 if (Perl_isnan(SvNVX(sv))) {
2013 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2014 SvIV_set(sv, I_V(SvNVX(sv)));
2015 if (SvNVX(sv) == (NV) SvIVX(sv)
2016 #ifndef NV_PRESERVES_UV
2017 && (((UV)1 << NV_PRESERVES_UV_BITS) >
2018 (UV)(SvIVX(sv) > 0 ? SvIVX(sv) : -SvIVX(sv)))
2019 /* Don't flag it as "accurately an integer" if the number
2020 came from a (by definition imprecise) NV operation, and
2021 we're outside the range of NV integer precision */
2025 SvIOK_on(sv); /* Can this go wrong with rounding? NWC */
2027 /* scalar has trailing garbage, eg "42a" */
2029 DEBUG_c(PerlIO_printf(Perl_debug_log,
2030 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (precise)\n",
2036 /* IV not precise. No need to convert from PV, as NV
2037 conversion would already have cached IV if it detected
2038 that PV->IV would be better than PV->NV->IV
2039 flags already correct - don't set public IOK. */
2040 DEBUG_c(PerlIO_printf(Perl_debug_log,
2041 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (imprecise)\n",
2046 /* Can the above go wrong if SvIVX == IV_MIN and SvNVX < IV_MIN,
2047 but the cast (NV)IV_MIN rounds to a the value less (more
2048 negative) than IV_MIN which happens to be equal to SvNVX ??
2049 Analogous to 0xFFFFFFFFFFFFFFFF rounding up to NV (2**64) and
2050 NV rounding back to 0xFFFFFFFFFFFFFFFF, so UVX == UV(NVX) and
2051 (NV)UVX == NVX are both true, but the values differ. :-(
2052 Hopefully for 2s complement IV_MIN is something like
2053 0x8000000000000000 which will be exact. NWC */
2056 SvUV_set(sv, U_V(SvNVX(sv)));
2058 (SvNVX(sv) == (NV) SvUVX(sv))
2059 #ifndef NV_PRESERVES_UV
2060 /* Make sure it's not 0xFFFFFFFFFFFFFFFF */
2061 /*&& (SvUVX(sv) != UV_MAX) irrelevant with code below */
2062 && (((UV)1 << NV_PRESERVES_UV_BITS) > SvUVX(sv))
2063 /* Don't flag it as "accurately an integer" if the number
2064 came from a (by definition imprecise) NV operation, and
2065 we're outside the range of NV integer precision */
2071 DEBUG_c(PerlIO_printf(Perl_debug_log,
2072 "0x%"UVxf" 2iv(%"UVuf" => %"IVdf") (as unsigned)\n",
2078 else if (SvPOKp(sv) && SvLEN(sv)) {
2080 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2081 /* We want to avoid a possible problem when we cache an IV/ a UV which
2082 may be later translated to an NV, and the resulting NV is not
2083 the same as the direct translation of the initial string
2084 (eg 123.456 can shortcut to the IV 123 with atol(), but we must
2085 be careful to ensure that the value with the .456 is around if the
2086 NV value is requested in the future).
2088 This means that if we cache such an IV/a UV, we need to cache the
2089 NV as well. Moreover, we trade speed for space, and do not
2090 cache the NV if we are sure it's not needed.
2093 /* SVt_PVNV is one higher than SVt_PVIV, hence this order */
2094 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2095 == IS_NUMBER_IN_UV) {
2096 /* It's definitely an integer, only upgrade to PVIV */
2097 if (SvTYPE(sv) < SVt_PVIV)
2098 sv_upgrade(sv, SVt_PVIV);
2100 } else if (SvTYPE(sv) < SVt_PVNV)
2101 sv_upgrade(sv, SVt_PVNV);
2103 /* If NVs preserve UVs then we only use the UV value if we know that
2104 we aren't going to call atof() below. If NVs don't preserve UVs
2105 then the value returned may have more precision than atof() will
2106 return, even though value isn't perfectly accurate. */
2107 if ((numtype & (IS_NUMBER_IN_UV
2108 #ifdef NV_PRESERVES_UV
2111 )) == IS_NUMBER_IN_UV) {
2112 /* This won't turn off the public IOK flag if it was set above */
2113 (void)SvIOKp_on(sv);
2115 if (!(numtype & IS_NUMBER_NEG)) {
2117 if (value <= (UV)IV_MAX) {
2118 SvIV_set(sv, (IV)value);
2120 /* it didn't overflow, and it was positive. */
2121 SvUV_set(sv, value);
2125 /* 2s complement assumption */
2126 if (value <= (UV)IV_MIN) {
2127 SvIV_set(sv, -(IV)value);
2129 /* Too negative for an IV. This is a double upgrade, but
2130 I'm assuming it will be rare. */
2131 if (SvTYPE(sv) < SVt_PVNV)
2132 sv_upgrade(sv, SVt_PVNV);
2136 SvNV_set(sv, -(NV)value);
2137 SvIV_set(sv, IV_MIN);
2141 /* For !NV_PRESERVES_UV and IS_NUMBER_IN_UV and IS_NUMBER_NOT_INT we
2142 will be in the previous block to set the IV slot, and the next
2143 block to set the NV slot. So no else here. */
2145 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2146 != IS_NUMBER_IN_UV) {
2147 /* It wasn't an (integer that doesn't overflow the UV). */
2148 SvNV_set(sv, Atof(SvPVX_const(sv)));
2150 if (! numtype && ckWARN(WARN_NUMERIC))
2153 #if defined(USE_LONG_DOUBLE)
2154 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%" PERL_PRIgldbl ")\n",
2155 PTR2UV(sv), SvNVX(sv)));
2157 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"NVgf")\n",
2158 PTR2UV(sv), SvNVX(sv)));
2161 #ifdef NV_PRESERVES_UV
2162 (void)SvIOKp_on(sv);
2164 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2165 SvIV_set(sv, I_V(SvNVX(sv)));
2166 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
2169 NOOP; /* Integer is imprecise. NOK, IOKp */
2171 /* UV will not work better than IV */
2173 if (SvNVX(sv) > (NV)UV_MAX) {
2175 /* Integer is inaccurate. NOK, IOKp, is UV */
2176 SvUV_set(sv, UV_MAX);
2178 SvUV_set(sv, U_V(SvNVX(sv)));
2179 /* 0xFFFFFFFFFFFFFFFF not an issue in here, NVs
2180 NV preservse UV so can do correct comparison. */
2181 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
2184 NOOP; /* Integer is imprecise. NOK, IOKp, is UV */
2189 #else /* NV_PRESERVES_UV */
2190 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2191 == (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT)) {
2192 /* The IV/UV slot will have been set from value returned by
2193 grok_number above. The NV slot has just been set using
2196 assert (SvIOKp(sv));
2198 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2199 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2200 /* Small enough to preserve all bits. */
2201 (void)SvIOKp_on(sv);
2203 SvIV_set(sv, I_V(SvNVX(sv)));
2204 if ((NV)(SvIVX(sv)) == SvNVX(sv))
2206 /* Assumption: first non-preserved integer is < IV_MAX,
2207 this NV is in the preserved range, therefore: */
2208 if (!(U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))
2210 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);
2214 0 0 already failed to read UV.
2215 0 1 already failed to read UV.
2216 1 0 you won't get here in this case. IV/UV
2217 slot set, public IOK, Atof() unneeded.
2218 1 1 already read UV.
2219 so there's no point in sv_2iuv_non_preserve() attempting
2220 to use atol, strtol, strtoul etc. */
2222 sv_2iuv_non_preserve (sv, numtype);
2224 sv_2iuv_non_preserve (sv);
2228 #endif /* NV_PRESERVES_UV */
2229 /* It might be more code efficient to go through the entire logic above
2230 and conditionally set with SvIOKp_on() rather than SvIOK(), but it
2231 gets complex and potentially buggy, so more programmer efficient
2232 to do it this way, by turning off the public flags: */
2234 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK);
2238 if (isGV_with_GP(sv))
2239 return glob_2number(MUTABLE_GV(sv));
2241 if (!(SvFLAGS(sv) & SVs_PADTMP)) {
2242 if (!PL_localizing && ckWARN(WARN_UNINITIALIZED))
2245 if (SvTYPE(sv) < SVt_IV)
2246 /* Typically the caller expects that sv_any is not NULL now. */
2247 sv_upgrade(sv, SVt_IV);
2248 /* Return 0 from the caller. */
2255 =for apidoc sv_2iv_flags
2257 Return the integer value of an SV, doing any necessary string
2258 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2259 Normally used via the C<SvIV(sv)> and C<SvIVx(sv)> macros.
2265 Perl_sv_2iv_flags(pTHX_ register SV *const sv, const I32 flags)
2270 if (SvGMAGICAL(sv) || (SvTYPE(sv) == SVt_PVGV && SvVALID(sv))) {
2271 /* FBMs use the space for SvIVX and SvNVX for other purposes, and use
2272 the same flag bit as SVf_IVisUV, so must not let them cache IVs.
2273 In practice they are extremely unlikely to actually get anywhere
2274 accessible by user Perl code - the only way that I'm aware of is when
2275 a constant subroutine which is used as the second argument to index.
2277 if (flags & SV_GMAGIC)
2282 return I_V(SvNVX(sv));
2284 if (SvPOKp(sv) && SvLEN(sv)) {
2287 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2289 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2290 == IS_NUMBER_IN_UV) {
2291 /* It's definitely an integer */
2292 if (numtype & IS_NUMBER_NEG) {
2293 if (value < (UV)IV_MIN)
2296 if (value < (UV)IV_MAX)
2301 if (ckWARN(WARN_NUMERIC))
2304 return I_V(Atof(SvPVX_const(sv)));
2309 assert(SvTYPE(sv) >= SVt_PVMG);
2310 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
2311 } else if (SvTHINKFIRST(sv)) {
2316 if (flags & SV_SKIP_OVERLOAD)
2318 tmpstr = AMG_CALLunary(sv, numer_amg);
2319 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2320 return SvIV(tmpstr);
2323 return PTR2IV(SvRV(sv));
2326 sv_force_normal_flags(sv, 0);
2328 if (SvREADONLY(sv) && !SvOK(sv)) {
2329 if (ckWARN(WARN_UNINITIALIZED))
2335 if (S_sv_2iuv_common(aTHX_ sv))
2338 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"IVdf")\n",
2339 PTR2UV(sv),SvIVX(sv)));
2340 return SvIsUV(sv) ? (IV)SvUVX(sv) : SvIVX(sv);
2344 =for apidoc sv_2uv_flags
2346 Return the unsigned integer value of an SV, doing any necessary string
2347 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2348 Normally used via the C<SvUV(sv)> and C<SvUVx(sv)> macros.
2354 Perl_sv_2uv_flags(pTHX_ register SV *const sv, const I32 flags)
2359 if (SvGMAGICAL(sv) || (SvTYPE(sv) == SVt_PVGV && SvVALID(sv))) {
2360 /* FBMs use the space for SvIVX and SvNVX for other purposes, and use
2361 the same flag bit as SVf_IVisUV, so must not let them cache IVs. */
2362 if (flags & SV_GMAGIC)
2367 return U_V(SvNVX(sv));
2368 if (SvPOKp(sv) && SvLEN(sv)) {
2371 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2373 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2374 == IS_NUMBER_IN_UV) {
2375 /* It's definitely an integer */
2376 if (!(numtype & IS_NUMBER_NEG))
2380 if (ckWARN(WARN_NUMERIC))
2383 return U_V(Atof(SvPVX_const(sv)));
2388 assert(SvTYPE(sv) >= SVt_PVMG);
2389 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
2390 } else if (SvTHINKFIRST(sv)) {
2395 if (flags & SV_SKIP_OVERLOAD)
2397 tmpstr = AMG_CALLunary(sv, numer_amg);
2398 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2399 return SvUV(tmpstr);
2402 return PTR2UV(SvRV(sv));
2405 sv_force_normal_flags(sv, 0);
2407 if (SvREADONLY(sv) && !SvOK(sv)) {
2408 if (ckWARN(WARN_UNINITIALIZED))
2414 if (S_sv_2iuv_common(aTHX_ sv))
2418 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2uv(%"UVuf")\n",
2419 PTR2UV(sv),SvUVX(sv)));
2420 return SvIsUV(sv) ? SvUVX(sv) : (UV)SvIVX(sv);
2424 =for apidoc sv_2nv_flags
2426 Return the num value of an SV, doing any necessary string or integer
2427 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2428 Normally used via the C<SvNV(sv)> and C<SvNVx(sv)> macros.
2434 Perl_sv_2nv_flags(pTHX_ register SV *const sv, const I32 flags)
2439 if (SvGMAGICAL(sv) || (SvTYPE(sv) == SVt_PVGV && SvVALID(sv))) {
2440 /* FBMs use the space for SvIVX and SvNVX for other purposes, and use
2441 the same flag bit as SVf_IVisUV, so must not let them cache NVs. */
2442 if (flags & SV_GMAGIC)
2446 if ((SvPOKp(sv) && SvLEN(sv)) && !SvIOKp(sv)) {
2447 if (!SvIOKp(sv) && ckWARN(WARN_NUMERIC) &&
2448 !grok_number(SvPVX_const(sv), SvCUR(sv), NULL))
2450 return Atof(SvPVX_const(sv));
2454 return (NV)SvUVX(sv);
2456 return (NV)SvIVX(sv);
2461 assert(SvTYPE(sv) >= SVt_PVMG);
2462 /* This falls through to the report_uninit near the end of the
2464 } else if (SvTHINKFIRST(sv)) {
2469 if (flags & SV_SKIP_OVERLOAD)
2471 tmpstr = AMG_CALLunary(sv, numer_amg);
2472 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2473 return SvNV(tmpstr);
2476 return PTR2NV(SvRV(sv));
2479 sv_force_normal_flags(sv, 0);
2481 if (SvREADONLY(sv) && !SvOK(sv)) {
2482 if (ckWARN(WARN_UNINITIALIZED))
2487 if (SvTYPE(sv) < SVt_NV) {
2488 /* The logic to use SVt_PVNV if necessary is in sv_upgrade. */
2489 sv_upgrade(sv, SVt_NV);
2490 #ifdef USE_LONG_DOUBLE
2492 STORE_NUMERIC_LOCAL_SET_STANDARD();
2493 PerlIO_printf(Perl_debug_log,
2494 "0x%"UVxf" num(%" PERL_PRIgldbl ")\n",
2495 PTR2UV(sv), SvNVX(sv));
2496 RESTORE_NUMERIC_LOCAL();
2500 STORE_NUMERIC_LOCAL_SET_STANDARD();
2501 PerlIO_printf(Perl_debug_log, "0x%"UVxf" num(%"NVgf")\n",
2502 PTR2UV(sv), SvNVX(sv));
2503 RESTORE_NUMERIC_LOCAL();
2507 else if (SvTYPE(sv) < SVt_PVNV)
2508 sv_upgrade(sv, SVt_PVNV);
2513 SvNV_set(sv, SvIsUV(sv) ? (NV)SvUVX(sv) : (NV)SvIVX(sv));
2514 #ifdef NV_PRESERVES_UV
2520 /* Only set the public NV OK flag if this NV preserves the IV */
2521 /* Check it's not 0xFFFFFFFFFFFFFFFF */
2523 SvIsUV(sv) ? ((SvUVX(sv) != UV_MAX)&&(SvUVX(sv) == U_V(SvNVX(sv))))
2524 : (SvIVX(sv) == I_V(SvNVX(sv))))
2530 else if (SvPOKp(sv) && SvLEN(sv)) {
2532 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2533 if (!SvIOKp(sv) && !numtype && ckWARN(WARN_NUMERIC))
2535 #ifdef NV_PRESERVES_UV
2536 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2537 == IS_NUMBER_IN_UV) {
2538 /* It's definitely an integer */
2539 SvNV_set(sv, (numtype & IS_NUMBER_NEG) ? -(NV)value : (NV)value);
2541 SvNV_set(sv, Atof(SvPVX_const(sv)));
2547 SvNV_set(sv, Atof(SvPVX_const(sv)));
2548 /* Only set the public NV OK flag if this NV preserves the value in
2549 the PV at least as well as an IV/UV would.
2550 Not sure how to do this 100% reliably. */
2551 /* if that shift count is out of range then Configure's test is
2552 wonky. We shouldn't be in here with NV_PRESERVES_UV_BITS ==
2554 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2555 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2556 SvNOK_on(sv); /* Definitely small enough to preserve all bits */
2557 } else if (!(numtype & IS_NUMBER_IN_UV)) {
2558 /* Can't use strtol etc to convert this string, so don't try.
2559 sv_2iv and sv_2uv will use the NV to convert, not the PV. */
2562 /* value has been set. It may not be precise. */
2563 if ((numtype & IS_NUMBER_NEG) && (value > (UV)IV_MIN)) {
2564 /* 2s complement assumption for (UV)IV_MIN */
2565 SvNOK_on(sv); /* Integer is too negative. */
2570 if (numtype & IS_NUMBER_NEG) {
2571 SvIV_set(sv, -(IV)value);
2572 } else if (value <= (UV)IV_MAX) {
2573 SvIV_set(sv, (IV)value);
2575 SvUV_set(sv, value);
2579 if (numtype & IS_NUMBER_NOT_INT) {
2580 /* I believe that even if the original PV had decimals,
2581 they are lost beyond the limit of the FP precision.
2582 However, neither is canonical, so both only get p
2583 flags. NWC, 2000/11/25 */
2584 /* Both already have p flags, so do nothing */
2586 const NV nv = SvNVX(sv);
2587 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2588 if (SvIVX(sv) == I_V(nv)) {
2591 /* It had no "." so it must be integer. */
2595 /* between IV_MAX and NV(UV_MAX).
2596 Could be slightly > UV_MAX */
2598 if (numtype & IS_NUMBER_NOT_INT) {
2599 /* UV and NV both imprecise. */
2601 const UV nv_as_uv = U_V(nv);
2603 if (value == nv_as_uv && SvUVX(sv) != UV_MAX) {
2612 /* It might be more code efficient to go through the entire logic above
2613 and conditionally set with SvNOKp_on() rather than SvNOK(), but it
2614 gets complex and potentially buggy, so more programmer efficient
2615 to do it this way, by turning off the public flags: */
2617 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK);
2618 #endif /* NV_PRESERVES_UV */
2621 if (isGV_with_GP(sv)) {
2622 glob_2number(MUTABLE_GV(sv));
2626 if (!PL_localizing && !(SvFLAGS(sv) & SVs_PADTMP) && ckWARN(WARN_UNINITIALIZED))
2628 assert (SvTYPE(sv) >= SVt_NV);
2629 /* Typically the caller expects that sv_any is not NULL now. */
2630 /* XXX Ilya implies that this is a bug in callers that assume this
2631 and ideally should be fixed. */
2634 #if defined(USE_LONG_DOUBLE)
2636 STORE_NUMERIC_LOCAL_SET_STANDARD();
2637 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2nv(%" PERL_PRIgldbl ")\n",
2638 PTR2UV(sv), SvNVX(sv));
2639 RESTORE_NUMERIC_LOCAL();
2643 STORE_NUMERIC_LOCAL_SET_STANDARD();
2644 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 1nv(%"NVgf")\n",
2645 PTR2UV(sv), SvNVX(sv));
2646 RESTORE_NUMERIC_LOCAL();
2655 Return an SV with the numeric value of the source SV, doing any necessary
2656 reference or overload conversion. You must use the C<SvNUM(sv)> macro to
2657 access this function.
2663 Perl_sv_2num(pTHX_ register SV *const sv)
2665 PERL_ARGS_ASSERT_SV_2NUM;
2670 SV * const tmpsv = AMG_CALLunary(sv, numer_amg);
2671 TAINT_IF(tmpsv && SvTAINTED(tmpsv));
2672 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
2673 return sv_2num(tmpsv);
2675 return sv_2mortal(newSVuv(PTR2UV(SvRV(sv))));
2678 /* uiv_2buf(): private routine for use by sv_2pv_flags(): print an IV or
2679 * UV as a string towards the end of buf, and return pointers to start and
2682 * We assume that buf is at least TYPE_CHARS(UV) long.
2686 S_uiv_2buf(char *const buf, const IV iv, UV uv, const int is_uv, char **const peob)
2688 char *ptr = buf + TYPE_CHARS(UV);
2689 char * const ebuf = ptr;
2692 PERL_ARGS_ASSERT_UIV_2BUF;
2704 *--ptr = '0' + (char)(uv % 10);
2713 =for apidoc sv_2pv_flags
2715 Returns a pointer to the string value of an SV, and sets *lp to its length.
2716 If flags includes SV_GMAGIC, does an mg_get() first. Coerces sv to a string
2718 Normally invoked via the C<SvPV_flags> macro. C<sv_2pv()> and C<sv_2pv_nomg>
2719 usually end up here too.
2725 Perl_sv_2pv_flags(pTHX_ register SV *const sv, STRLEN *const lp, const I32 flags)
2735 if (SvGMAGICAL(sv)) {
2736 if (flags & SV_GMAGIC)
2741 if (flags & SV_MUTABLE_RETURN)
2742 return SvPVX_mutable(sv);
2743 if (flags & SV_CONST_RETURN)
2744 return (char *)SvPVX_const(sv);
2747 if (SvIOKp(sv) || SvNOKp(sv)) {
2748 char tbuf[64]; /* Must fit sprintf/Gconvert of longest IV/NV */
2753 ? my_snprintf(tbuf, sizeof(tbuf), "%"UVuf, (UV)SvUVX(sv))
2754 : my_snprintf(tbuf, sizeof(tbuf), "%"IVdf, (IV)SvIVX(sv));
2755 } else if(SvNVX(sv) == 0.0) {
2760 Gconvert(SvNVX(sv), NV_DIG, 0, tbuf);
2767 SvUPGRADE(sv, SVt_PV);
2770 s = SvGROW_mutable(sv, len + 1);
2773 return (char*)memcpy(s, tbuf, len + 1);
2779 assert(SvTYPE(sv) >= SVt_PVMG);
2780 /* This falls through to the report_uninit near the end of the
2782 } else if (SvTHINKFIRST(sv)) {
2787 if (flags & SV_SKIP_OVERLOAD)
2789 tmpstr = AMG_CALLunary(sv, string_amg);
2790 TAINT_IF(tmpstr && SvTAINTED(tmpstr));
2791 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2793 /* char *pv = lp ? SvPV(tmpstr, *lp) : SvPV_nolen(tmpstr);
2797 if ((SvFLAGS(tmpstr) & (SVf_POK)) == SVf_POK) {
2798 if (flags & SV_CONST_RETURN) {
2799 pv = (char *) SvPVX_const(tmpstr);
2801 pv = (flags & SV_MUTABLE_RETURN)
2802 ? SvPVX_mutable(tmpstr) : SvPVX(tmpstr);
2805 *lp = SvCUR(tmpstr);
2807 pv = sv_2pv_flags(tmpstr, lp, flags);
2820 SV *const referent = SvRV(sv);
2824 retval = buffer = savepvn("NULLREF", len);
2825 } else if (SvTYPE(referent) == SVt_REGEXP) {
2826 REGEXP * const re = (REGEXP *)MUTABLE_PTR(referent);
2831 /* If the regex is UTF-8 we want the containing scalar to
2832 have an UTF-8 flag too */
2838 if ((seen_evals = RX_SEEN_EVALS(re)))
2839 PL_reginterp_cnt += seen_evals;
2842 *lp = RX_WRAPLEN(re);
2844 return RX_WRAPPED(re);
2846 const char *const typestr = sv_reftype(referent, 0);
2847 const STRLEN typelen = strlen(typestr);
2848 UV addr = PTR2UV(referent);
2849 const char *stashname = NULL;
2850 STRLEN stashnamelen = 0; /* hush, gcc */
2851 const char *buffer_end;
2853 if (SvOBJECT(referent)) {
2854 const HEK *const name = HvNAME_HEK(SvSTASH(referent));
2857 stashname = HEK_KEY(name);
2858 stashnamelen = HEK_LEN(name);
2860 if (HEK_UTF8(name)) {
2866 stashname = "__ANON__";
2869 len = stashnamelen + 1 /* = */ + typelen + 3 /* (0x */
2870 + 2 * sizeof(UV) + 2 /* )\0 */;
2872 len = typelen + 3 /* (0x */
2873 + 2 * sizeof(UV) + 2 /* )\0 */;
2876 Newx(buffer, len, char);
2877 buffer_end = retval = buffer + len;
2879 /* Working backwards */
2883 *--retval = PL_hexdigit[addr & 15];
2884 } while (addr >>= 4);
2890 memcpy(retval, typestr, typelen);
2894 retval -= stashnamelen;
2895 memcpy(retval, stashname, stashnamelen);
2897 /* retval may not necessarily have reached the start of the
2899 assert (retval >= buffer);
2901 len = buffer_end - retval - 1; /* -1 for that \0 */
2909 if (SvREADONLY(sv) && !SvOK(sv)) {
2912 if (flags & SV_UNDEF_RETURNS_NULL)
2914 if (ckWARN(WARN_UNINITIALIZED))
2919 if (SvIOK(sv) || ((SvIOKp(sv) && !SvNOKp(sv)))) {
2920 /* I'm assuming that if both IV and NV are equally valid then
2921 converting the IV is going to be more efficient */
2922 const U32 isUIOK = SvIsUV(sv);
2923 char buf[TYPE_CHARS(UV)];
2927 if (SvTYPE(sv) < SVt_PVIV)
2928 sv_upgrade(sv, SVt_PVIV);
2929 ptr = uiv_2buf(buf, SvIVX(sv), SvUVX(sv), isUIOK, &ebuf);
2931 /* inlined from sv_setpvn */
2932 s = SvGROW_mutable(sv, len + 1);
2933 Move(ptr, s, len, char);
2937 else if (SvNOKp(sv)) {
2938 if (SvTYPE(sv) < SVt_PVNV)
2939 sv_upgrade(sv, SVt_PVNV);
2940 if (SvNVX(sv) == 0.0) {
2941 s = SvGROW_mutable(sv, 2);
2946 /* The +20 is pure guesswork. Configure test needed. --jhi */
2947 s = SvGROW_mutable(sv, NV_DIG + 20);
2948 /* some Xenix systems wipe out errno here */
2949 Gconvert(SvNVX(sv), NV_DIG, 0, s);
2959 if (isGV_with_GP(sv)) {
2960 GV *const gv = MUTABLE_GV(sv);
2961 const U32 wasfake = SvFLAGS(gv) & SVf_FAKE;
2962 SV *const buffer = sv_newmortal();
2964 /* FAKE globs can get coerced, so need to turn this off temporarily
2967 gv_efullname3(buffer, gv, "*");
2968 SvFLAGS(gv) |= wasfake;
2970 if (SvPOK(buffer)) {
2972 *lp = SvCUR(buffer);
2974 return SvPVX(buffer);
2985 if (flags & SV_UNDEF_RETURNS_NULL)
2987 if (!PL_localizing && !(SvFLAGS(sv) & SVs_PADTMP) && ckWARN(WARN_UNINITIALIZED))
2989 if (SvTYPE(sv) < SVt_PV)
2990 /* Typically the caller expects that sv_any is not NULL now. */
2991 sv_upgrade(sv, SVt_PV);
2995 const STRLEN len = s - SvPVX_const(sv);
3001 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
3002 PTR2UV(sv),SvPVX_const(sv)));
3003 if (flags & SV_CONST_RETURN)
3004 return (char *)SvPVX_const(sv);
3005 if (flags & SV_MUTABLE_RETURN)
3006 return SvPVX_mutable(sv);
3011 =for apidoc sv_copypv
3013 Copies a stringified representation of the source SV into the
3014 destination SV. Automatically performs any necessary mg_get and
3015 coercion of numeric values into strings. Guaranteed to preserve
3016 UTF8 flag even from overloaded objects. Similar in nature to
3017 sv_2pv[_flags] but operates directly on an SV instead of just the
3018 string. Mostly uses sv_2pv_flags to do its work, except when that
3019 would lose the UTF-8'ness of the PV.
3025 Perl_sv_copypv(pTHX_ SV *const dsv, register SV *const ssv)
3028 const char * const s = SvPV_const(ssv,len);
3030 PERL_ARGS_ASSERT_SV_COPYPV;
3032 sv_setpvn(dsv,s,len);
3040 =for apidoc sv_2pvbyte
3042 Return a pointer to the byte-encoded representation of the SV, and set *lp
3043 to its length. May cause the SV to be downgraded from UTF-8 as a
3046 Usually accessed via the C<SvPVbyte> macro.
3052 Perl_sv_2pvbyte(pTHX_ register SV *const sv, STRLEN *const lp)
3054 PERL_ARGS_ASSERT_SV_2PVBYTE;
3057 sv_utf8_downgrade(sv,0);
3058 return lp ? SvPV_nomg(sv,*lp) : SvPV_nomg_nolen(sv);
3062 =for apidoc sv_2pvutf8
3064 Return a pointer to the UTF-8-encoded representation of the SV, and set *lp
3065 to its length. May cause the SV to be upgraded to UTF-8 as a side-effect.
3067 Usually accessed via the C<SvPVutf8> macro.
3073 Perl_sv_2pvutf8(pTHX_ register SV *const sv, STRLEN *const lp)
3075 PERL_ARGS_ASSERT_SV_2PVUTF8;
3077 sv_utf8_upgrade(sv);
3078 return lp ? SvPV(sv,*lp) : SvPV_nolen(sv);
3083 =for apidoc sv_2bool
3085 This macro is only used by sv_true() or its macro equivalent, and only if
3086 the latter's argument is neither SvPOK, SvIOK nor SvNOK.
3087 It calls sv_2bool_flags with the SV_GMAGIC flag.
3089 =for apidoc sv_2bool_flags
3091 This function is only used by sv_true() and friends, and only if
3092 the latter's argument is neither SvPOK, SvIOK nor SvNOK. If the flags
3093 contain SV_GMAGIC, then it does an mg_get() first.
3100 Perl_sv_2bool_flags(pTHX_ register SV *const sv, const I32 flags)
3104 PERL_ARGS_ASSERT_SV_2BOOL_FLAGS;
3106 if(flags & SV_GMAGIC) SvGETMAGIC(sv);
3112 SV * const tmpsv = AMG_CALLunary(sv, bool__amg);
3113 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
3114 return cBOOL(SvTRUE(tmpsv));
3116 return SvRV(sv) != 0;
3119 register XPV* const Xpvtmp = (XPV*)SvANY(sv);
3121 (*sv->sv_u.svu_pv > '0' ||
3122 Xpvtmp->xpv_cur > 1 ||
3123 (Xpvtmp->xpv_cur && *sv->sv_u.svu_pv != '0')))
3130 return SvIVX(sv) != 0;
3133 return SvNVX(sv) != 0.0;
3135 if (isGV_with_GP(sv))
3145 =for apidoc sv_utf8_upgrade
3147 Converts the PV of an SV to its UTF-8-encoded form.
3148 Forces the SV to string form if it is not already.
3149 Will C<mg_get> on C<sv> if appropriate.
3150 Always sets the SvUTF8 flag to avoid future validity checks even
3151 if the whole string is the same in UTF-8 as not.
3152 Returns the number of bytes in the converted string
3154 This is not as a general purpose byte encoding to Unicode interface:
3155 use the Encode extension for that.
3157 =for apidoc sv_utf8_upgrade_nomg
3159 Like sv_utf8_upgrade, but doesn't do magic on C<sv>
3161 =for apidoc sv_utf8_upgrade_flags
3163 Converts the PV of an SV to its UTF-8-encoded form.
3164 Forces the SV to string form if it is not already.
3165 Always sets the SvUTF8 flag to avoid future validity checks even
3166 if all the bytes are invariant in UTF-8. If C<flags> has C<SV_GMAGIC> bit set,
3167 will C<mg_get> on C<sv> if appropriate, else not.
3168 Returns the number of bytes in the converted string
3169 C<sv_utf8_upgrade> and
3170 C<sv_utf8_upgrade_nomg> are implemented in terms of this function.
3172 This is not as a general purpose byte encoding to Unicode interface:
3173 use the Encode extension for that.
3177 The grow version is currently not externally documented. It adds a parameter,
3178 extra, which is the number of unused bytes the string of 'sv' is guaranteed to
3179 have free after it upon return. This allows the caller to reserve extra space
3180 that it intends to fill, to avoid extra grows.
3182 Also externally undocumented for the moment is the flag SV_FORCE_UTF8_UPGRADE,
3183 which can be used to tell this function to not first check to see if there are
3184 any characters that are different in UTF-8 (variant characters) which would
3185 force it to allocate a new string to sv, but to assume there are. Typically
3186 this flag is used by a routine that has already parsed the string to find that
3187 there are such characters, and passes this information on so that the work
3188 doesn't have to be repeated.
3190 (One might think that the calling routine could pass in the position of the
3191 first such variant, so it wouldn't have to be found again. But that is not the
3192 case, because typically when the caller is likely to use this flag, it won't be
3193 calling this routine unless it finds something that won't fit into a byte.
3194 Otherwise it tries to not upgrade and just use bytes. But some things that
3195 do fit into a byte are variants in utf8, and the caller may not have been
3196 keeping track of these.)
3198 If the routine itself changes the string, it adds a trailing NUL. Such a NUL
3199 isn't guaranteed due to having other routines do the work in some input cases,
3200 or if the input is already flagged as being in utf8.
3202 The speed of this could perhaps be improved for many cases if someone wanted to
3203 write a fast function that counts the number of variant characters in a string,
3204 especially if it could return the position of the first one.
3209 Perl_sv_utf8_upgrade_flags_grow(pTHX_ register SV *const sv, const I32 flags, STRLEN extra)
3213 PERL_ARGS_ASSERT_SV_UTF8_UPGRADE_FLAGS_GROW;
3215 if (sv == &PL_sv_undef)
3219 if (SvREADONLY(sv) && (SvPOKp(sv) || SvIOKp(sv) || SvNOKp(sv))) {
3220 (void) sv_2pv_flags(sv,&len, flags);
3222 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3226 (void) SvPV_force_flags(sv,len,flags & SV_GMAGIC);
3231 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3236 sv_force_normal_flags(sv, 0);
3239 if (PL_encoding && !(flags & SV_UTF8_NO_ENCODING)) {
3240 sv_recode_to_utf8(sv, PL_encoding);
3241 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3245 if (SvCUR(sv) == 0) {
3246 if (extra) SvGROW(sv, extra);
3247 } else { /* Assume Latin-1/EBCDIC */
3248 /* This function could be much more efficient if we
3249 * had a FLAG in SVs to signal if there are any variant
3250 * chars in the PV. Given that there isn't such a flag
3251 * make the loop as fast as possible (although there are certainly ways
3252 * to speed this up, eg. through vectorization) */
3253 U8 * s = (U8 *) SvPVX_const(sv);
3254 U8 * e = (U8 *) SvEND(sv);
3256 STRLEN two_byte_count = 0;
3258 if (flags & SV_FORCE_UTF8_UPGRADE) goto must_be_utf8;
3260 /* See if really will need to convert to utf8. We mustn't rely on our
3261 * incoming SV being well formed and having a trailing '\0', as certain
3262 * code in pp_formline can send us partially built SVs. */
3266 if (NATIVE_IS_INVARIANT(ch)) continue;
3268 t--; /* t already incremented; re-point to first variant */
3273 /* utf8 conversion not needed because all are invariants. Mark as
3274 * UTF-8 even if no variant - saves scanning loop */
3280 /* Here, the string should be converted to utf8, either because of an
3281 * input flag (two_byte_count = 0), or because a character that
3282 * requires 2 bytes was found (two_byte_count = 1). t points either to
3283 * the beginning of the string (if we didn't examine anything), or to
3284 * the first variant. In either case, everything from s to t - 1 will
3285 * occupy only 1 byte each on output.
3287 * There are two main ways to convert. One is to create a new string
3288 * and go through the input starting from the beginning, appending each
3289 * converted value onto the new string as we go along. It's probably
3290 * best to allocate enough space in the string for the worst possible
3291 * case rather than possibly running out of space and having to
3292 * reallocate and then copy what we've done so far. Since everything
3293 * from s to t - 1 is invariant, the destination can be initialized
3294 * with these using a fast memory copy
3296 * The other way is to figure out exactly how big the string should be
3297 * by parsing the entire input. Then you don't have to make it big
3298 * enough to handle the worst possible case, and more importantly, if
3299 * the string you already have is large enough, you don't have to
3300 * allocate a new string, you can copy the last character in the input
3301 * string to the final position(s) that will be occupied by the
3302 * converted string and go backwards, stopping at t, since everything
3303 * before that is invariant.
3305 * There are advantages and disadvantages to each method.
3307 * In the first method, we can allocate a new string, do the memory
3308 * copy from the s to t - 1, and then proceed through the rest of the
3309 * string byte-by-byte.
3311 * In the second method, we proceed through the rest of the input
3312 * string just calculating how big the converted string will be. Then
3313 * there are two cases:
3314 * 1) if the string has enough extra space to handle the converted
3315 * value. We go backwards through the string, converting until we
3316 * get to the position we are at now, and then stop. If this
3317 * position is far enough along in the string, this method is
3318 * faster than the other method. If the memory copy were the same
3319 * speed as the byte-by-byte loop, that position would be about
3320 * half-way, as at the half-way mark, parsing to the end and back
3321 * is one complete string's parse, the same amount as starting
3322 * over and going all the way through. Actually, it would be
3323 * somewhat less than half-way, as it's faster to just count bytes
3324 * than to also copy, and we don't have the overhead of allocating
3325 * a new string, changing the scalar to use it, and freeing the
3326 * existing one. But if the memory copy is fast, the break-even
3327 * point is somewhere after half way. The counting loop could be
3328 * sped up by vectorization, etc, to move the break-even point
3329 * further towards the beginning.
3330 * 2) if the string doesn't have enough space to handle the converted
3331 * value. A new string will have to be allocated, and one might
3332 * as well, given that, start from the beginning doing the first
3333 * method. We've spent extra time parsing the string and in
3334 * exchange all we've gotten is that we know precisely how big to
3335 * make the new one. Perl is more optimized for time than space,
3336 * so this case is a loser.
3337 * So what I've decided to do is not use the 2nd method unless it is
3338 * guaranteed that a new string won't have to be allocated, assuming
3339 * the worst case. I also decided not to put any more conditions on it
3340 * than this, for now. It seems likely that, since the worst case is
3341 * twice as big as the unknown portion of the string (plus 1), we won't
3342 * be guaranteed enough space, causing us to go to the first method,
3343 * unless the string is short, or the first variant character is near
3344 * the end of it. In either of these cases, it seems best to use the
3345 * 2nd method. The only circumstance I can think of where this would
3346 * be really slower is if the string had once had much more data in it
3347 * than it does now, but there is still a substantial amount in it */
3350 STRLEN invariant_head = t - s;
3351 STRLEN size = invariant_head + (e - t) * 2 + 1 + extra;
3352 if (SvLEN(sv) < size) {
3354 /* Here, have decided to allocate a new string */
3359 Newx(dst, size, U8);
3361 /* If no known invariants at the beginning of the input string,
3362 * set so starts from there. Otherwise, can use memory copy to
3363 * get up to where we are now, and then start from here */
3365 if (invariant_head <= 0) {
3368 Copy(s, dst, invariant_head, char);
3369 d = dst + invariant_head;
3373 const UV uv = NATIVE8_TO_UNI(*t++);
3374 if (UNI_IS_INVARIANT(uv))
3375 *d++ = (U8)UNI_TO_NATIVE(uv);
3377 *d++ = (U8)UTF8_EIGHT_BIT_HI(uv);
3378 *d++ = (U8)UTF8_EIGHT_BIT_LO(uv);
3382 SvPV_free(sv); /* No longer using pre-existing string */
3383 SvPV_set(sv, (char*)dst);
3384 SvCUR_set(sv, d - dst);
3385 SvLEN_set(sv, size);
3388 /* Here, have decided to get the exact size of the string.
3389 * Currently this happens only when we know that there is
3390 * guaranteed enough space to fit the converted string, so
3391 * don't have to worry about growing. If two_byte_count is 0,
3392 * then t points to the first byte of the string which hasn't
3393 * been examined yet. Otherwise two_byte_count is 1, and t
3394 * points to the first byte in the string that will expand to
3395 * two. Depending on this, start examining at t or 1 after t.
3398 U8 *d = t + two_byte_count;
3401 /* Count up the remaining bytes that expand to two */
3404 const U8 chr = *d++;
3405 if (! NATIVE_IS_INVARIANT(chr)) two_byte_count++;
3408 /* The string will expand by just the number of bytes that
3409 * occupy two positions. But we are one afterwards because of
3410 * the increment just above. This is the place to put the
3411 * trailing NUL, and to set the length before we decrement */
3413 d += two_byte_count;
3414 SvCUR_set(sv, d - s);
3418 /* Having decremented d, it points to the position to put the
3419 * very last byte of the expanded string. Go backwards through
3420 * the string, copying and expanding as we go, stopping when we
3421 * get to the part that is invariant the rest of the way down */
3425 const U8 ch = NATIVE8_TO_UNI(*e--);
3426 if (UNI_IS_INVARIANT(ch)) {
3427 *d-- = UNI_TO_NATIVE(ch);
3429 *d-- = (U8)UTF8_EIGHT_BIT_LO(ch);
3430 *d-- = (U8)UTF8_EIGHT_BIT_HI(ch);
3435 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
3436 /* Update pos. We do it at the end rather than during
3437 * the upgrade, to avoid slowing down the common case
3438 * (upgrade without pos) */
3439 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
3441 I32 pos = mg->mg_len;
3442 if (pos > 0 && (U32)pos > invariant_head) {
3443 U8 *d = (U8*) SvPVX(sv) + invariant_head;
3444 STRLEN n = (U32)pos - invariant_head;
3446 if (UTF8_IS_START(*d))
3451 mg->mg_len = d - (U8*)SvPVX(sv);
3454 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
3455 magic_setutf8(sv,mg); /* clear UTF8 cache */
3460 /* Mark as UTF-8 even if no variant - saves scanning loop */
3466 =for apidoc sv_utf8_downgrade
3468 Attempts to convert the PV of an SV from characters to bytes.
3469 If the PV contains a character that cannot fit
3470 in a byte, this conversion will fail;
3471 in this case, either returns false or, if C<fail_ok> is not
3474 This is not as a general purpose Unicode to byte encoding interface:
3475 use the Encode extension for that.
3481 Perl_sv_utf8_downgrade(pTHX_ register SV *const sv, const bool fail_ok)
3485 PERL_ARGS_ASSERT_SV_UTF8_DOWNGRADE;
3487 if (SvPOKp(sv) && SvUTF8(sv)) {
3491 int mg_flags = SV_GMAGIC;
3494 sv_force_normal_flags(sv, 0);
3496 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
3498 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
3500 I32 pos = mg->mg_len;
3502 sv_pos_b2u(sv, &pos);
3503 mg_flags = 0; /* sv_pos_b2u does get magic */
3507 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
3508 magic_setutf8(sv,mg); /* clear UTF8 cache */
3511 s = (U8 *) SvPV_flags(sv, len, mg_flags);
3513 if (!utf8_to_bytes(s, &len)) {
3518 Perl_croak(aTHX_ "Wide character in %s",
3521 Perl_croak(aTHX_ "Wide character");
3532 =for apidoc sv_utf8_encode
3534 Converts the PV of an SV to UTF-8, but then turns the C<SvUTF8>
3535 flag off so that it looks like octets again.
3541 Perl_sv_utf8_encode(pTHX_ register SV *const sv)
3543 PERL_ARGS_ASSERT_SV_UTF8_ENCODE;
3546 sv_force_normal_flags(sv, 0);
3548 if (SvREADONLY(sv)) {
3549 Perl_croak_no_modify(aTHX);
3551 (void) sv_utf8_upgrade(sv);
3556 =for apidoc sv_utf8_decode
3558 If the PV of the SV is an octet sequence in UTF-8
3559 and contains a multiple-byte character, the C<SvUTF8> flag is turned on
3560 so that it looks like a character. If the PV contains only single-byte
3561 characters, the C<SvUTF8> flag stays off.
3562 Scans PV for validity and returns false if the PV is invalid UTF-8.
3568 Perl_sv_utf8_decode(pTHX_ register SV *const sv)
3570 PERL_ARGS_ASSERT_SV_UTF8_DECODE;
3573 const U8 *start, *c;
3576 /* The octets may have got themselves encoded - get them back as
3579 if (!sv_utf8_downgrade(sv, TRUE))
3582 /* it is actually just a matter of turning the utf8 flag on, but
3583 * we want to make sure everything inside is valid utf8 first.
3585 c = start = (const U8 *) SvPVX_const(sv);
3586 if (!is_utf8_string(c, SvCUR(sv)+1))
3588 e = (const U8 *) SvEND(sv);
3591 if (!UTF8_IS_INVARIANT(ch)) {
3596 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
3597 /* adjust pos to the start of a UTF8 char sequence */
3598 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
3600 I32 pos = mg->mg_len;
3602 for (c = start + pos; c > start; c--) {
3603 if (UTF8_IS_START(*c))
3606 mg->mg_len = c - start;
3609 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
3610 magic_setutf8(sv,mg); /* clear UTF8 cache */
3617 =for apidoc sv_setsv
3619 Copies the contents of the source SV C<ssv> into the destination SV
3620 C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3621 function if the source SV needs to be reused. Does not handle 'set' magic.
3622 Loosely speaking, it performs a copy-by-value, obliterating any previous
3623 content of the destination.
3625 You probably want to use one of the assortment of wrappers, such as
3626 C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3627 C<SvSetMagicSV_nosteal>.
3629 =for apidoc sv_setsv_flags
3631 Copies the contents of the source SV C<ssv> into the destination SV
3632 C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3633 function if the source SV needs to be reused. Does not handle 'set' magic.
3634 Loosely speaking, it performs a copy-by-value, obliterating any previous
3635 content of the destination.
3636 If the C<flags> parameter has the C<SV_GMAGIC> bit set, will C<mg_get> on
3637 C<ssv> if appropriate, else not. If the C<flags> parameter has the
3638 C<NOSTEAL> bit set then the buffers of temps will not be stolen. <sv_setsv>
3639 and C<sv_setsv_nomg> are implemented in terms of this function.
3641 You probably want to use one of the assortment of wrappers, such as
3642 C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3643 C<SvSetMagicSV_nosteal>.
3645 This is the primary function for copying scalars, and most other
3646 copy-ish functions and macros use this underneath.
3652 S_glob_assign_glob(pTHX_ SV *const dstr, SV *const sstr, const int dtype)
3654 I32 mro_changes = 0; /* 1 = method, 2 = isa, 3 = recursive isa */
3655 HV *old_stash = NULL;
3657 PERL_ARGS_ASSERT_GLOB_ASSIGN_GLOB;
3659 if (dtype != SVt_PVGV && !isGV_with_GP(dstr)) {
3660 const char * const name = GvNAME(sstr);
3661 const STRLEN len = GvNAMELEN(sstr);
3663 if (dtype >= SVt_PV) {
3669 SvUPGRADE(dstr, SVt_PVGV);
3670 (void)SvOK_off(dstr);
3671 /* FIXME - why are we doing this, then turning it off and on again
3673 isGV_with_GP_on(dstr);
3675 GvSTASH(dstr) = GvSTASH(sstr);
3677 Perl_sv_add_backref(aTHX_ MUTABLE_SV(GvSTASH(dstr)), dstr);
3678 gv_name_set(MUTABLE_GV(dstr), name, len, GV_ADD);
3679 SvFAKE_on(dstr); /* can coerce to non-glob */
3682 if(GvGP(MUTABLE_GV(sstr))) {
3683 /* If source has method cache entry, clear it */
3685 SvREFCNT_dec(GvCV(sstr));
3686 GvCV_set(sstr, NULL);
3689 /* If source has a real method, then a method is
3692 GvCV((const GV *)sstr) && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3698 /* If dest already had a real method, that's a change as well */
3700 !mro_changes && GvGP(MUTABLE_GV(dstr)) && GvCVu((const GV *)dstr)
3701 && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3706 /* We don’t need to check the name of the destination if it was not a
3707 glob to begin with. */
3708 if(dtype == SVt_PVGV) {
3709 const char * const name = GvNAME((const GV *)dstr);
3712 /* The stash may have been detached from the symbol table, so
3714 && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3715 && GvAV((const GV *)sstr)
3719 const STRLEN len = GvNAMELEN(dstr);
3720 if ((len > 1 && name[len-2] == ':' && name[len-1] == ':')
3721 || (len == 1 && name[0] == ':')) {
3724 /* Set aside the old stash, so we can reset isa caches on
3726 if((old_stash = GvHV(dstr)))
3727 /* Make sure we do not lose it early. */
3728 SvREFCNT_inc_simple_void_NN(
3729 sv_2mortal((SV *)old_stash)
3735 gp_free(MUTABLE_GV(dstr));
3736 isGV_with_GP_off(dstr);
3737 (void)SvOK_off(dstr);
3738 isGV_with_GP_on(dstr);
3739 GvINTRO_off(dstr); /* one-shot flag */
3740 GvGP_set(dstr, gp_ref(GvGP(sstr)));
3741 if (SvTAINTED(sstr))
3743 if (GvIMPORTED(dstr) != GVf_IMPORTED
3744 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3746 GvIMPORTED_on(dstr);
3749 if(mro_changes == 2) {
3751 SV * const sref = (SV *)GvAV((const GV *)dstr);
3752 if (SvSMAGICAL(sref) && (mg = mg_find(sref, PERL_MAGIC_isa))) {
3753 if (SvTYPE(mg->mg_obj) != SVt_PVAV) {
3754 AV * const ary = newAV();
3755 av_push(ary, mg->mg_obj); /* takes the refcount */
3756 mg->mg_obj = (SV *)ary;
3758 av_push((AV *)mg->mg_obj, SvREFCNT_inc_simple_NN(dstr));
3760 else sv_magic(sref, dstr, PERL_MAGIC_isa, NULL, 0);
3761 mro_isa_changed_in(GvSTASH(dstr));
3763 else if(mro_changes == 3) {
3764 HV * const stash = GvHV(dstr);
3765 if(old_stash ? (HV *)HvENAME_get(old_stash) : stash)
3771 else if(mro_changes) mro_method_changed_in(GvSTASH(dstr));
3776 S_glob_assign_ref(pTHX_ SV *const dstr, SV *const sstr)
3778 SV * const sref = SvREFCNT_inc(SvRV(sstr));
3780 const int intro = GvINTRO(dstr);
3783 const U32 stype = SvTYPE(sref);
3785 PERL_ARGS_ASSERT_GLOB_ASSIGN_REF;
3788 GvINTRO_off(dstr); /* one-shot flag */
3789 GvLINE(dstr) = CopLINE(PL_curcop);
3790 GvEGV(dstr) = MUTABLE_GV(dstr);
3795 location = (SV **) &(GvGP(dstr)->gp_cv); /* XXX bypassing GvCV_set */
3796 import_flag = GVf_IMPORTED_CV;
3799 location = (SV **) &GvHV(dstr);
3800 import_flag = GVf_IMPORTED_HV;
3803 location = (SV **) &GvAV(dstr);
3804 import_flag = GVf_IMPORTED_AV;
3807 location = (SV **) &GvIOp(dstr);
3810 location = (SV **) &GvFORM(dstr);
3813 location = &GvSV(dstr);
3814 import_flag = GVf_IMPORTED_SV;
3817 if (stype == SVt_PVCV) {
3818 /*if (GvCVGEN(dstr) && (GvCV(dstr) != (const CV *)sref || GvCVGEN(dstr))) {*/
3819 if (GvCVGEN(dstr)) {
3820 SvREFCNT_dec(GvCV(dstr));
3821 GvCV_set(dstr, NULL);
3822 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3825 SAVEGENERICSV(*location);
3829 if (stype == SVt_PVCV && (*location != sref || GvCVGEN(dstr))) {
3830 CV* const cv = MUTABLE_CV(*location);
3832 if (!GvCVGEN((const GV *)dstr) &&
3833 (CvROOT(cv) || CvXSUB(cv)))
3835 /* Redefining a sub - warning is mandatory if
3836 it was a const and its value changed. */
3837 if (CvCONST(cv) && CvCONST((const CV *)sref)
3839 == cv_const_sv((const CV *)sref)) {
3841 /* They are 2 constant subroutines generated from
3842 the same constant. This probably means that
3843 they are really the "same" proxy subroutine
3844 instantiated in 2 places. Most likely this is
3845 when a constant is exported twice. Don't warn.
3848 else if (ckWARN(WARN_REDEFINE)
3850 && (!CvCONST((const CV *)sref)
3851 || sv_cmp(cv_const_sv(cv),
3852 cv_const_sv((const CV *)
3854 Perl_warner(aTHX_ packWARN(WARN_REDEFINE),
3857 ? "Constant subroutine %s::%s redefined"
3858 : "Subroutine %s::%s redefined"),
3859 HvNAME_get(GvSTASH((const GV *)dstr)),
3860 GvENAME(MUTABLE_GV(dstr)));
3864 cv_ckproto_len(cv, (const GV *)dstr,
3865 SvPOK(sref) ? SvPVX_const(sref) : NULL,
3866 SvPOK(sref) ? SvCUR(sref) : 0);
3868 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3869 GvASSUMECV_on(dstr);
3870 if(GvSTASH(dstr)) mro_method_changed_in(GvSTASH(dstr)); /* sub foo { 1 } sub bar { 2 } *bar = \&foo */
3873 if (import_flag && !(GvFLAGS(dstr) & import_flag)
3874 && CopSTASH_ne(PL_curcop, GvSTASH(dstr))) {
3875 GvFLAGS(dstr) |= import_flag;
3877 if (stype == SVt_PVHV) {
3878 const char * const name = GvNAME((GV*)dstr);
3879 const STRLEN len = GvNAMELEN(dstr);
3882 (len > 1 && name[len-2] == ':' && name[len-1] == ':')
3883 || (len == 1 && name[0] == ':')
3885 && (!dref || HvENAME_get(dref))
3888 (HV *)sref, (HV *)dref,
3894 stype == SVt_PVAV && sref != dref
3895 && strEQ(GvNAME((GV*)dstr), "ISA")
3896 /* The stash may have been detached from the symbol table, so
3897 check its name before doing anything. */
3898 && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3901 MAGIC * const omg = dref && SvSMAGICAL(dref)
3902 ? mg_find(dref, PERL_MAGIC_isa)
3904 if (SvSMAGICAL(sref) && (mg = mg_find(sref, PERL_MAGIC_isa))) {
3905 if (SvTYPE(mg->mg_obj) != SVt_PVAV) {
3906 AV * const ary = newAV();
3907 av_push(ary, mg->mg_obj); /* takes the refcount */
3908 mg->mg_obj = (SV *)ary;
3911 if (SvTYPE(omg->mg_obj) == SVt_PVAV) {
3912 SV **svp = AvARRAY((AV *)omg->mg_obj);
3913 I32 items = AvFILLp((AV *)omg->mg_obj) + 1;
3917 SvREFCNT_inc_simple_NN(*svp++)
3923 SvREFCNT_inc_simple_NN(omg->mg_obj)
3927 av_push((AV *)mg->mg_obj,SvREFCNT_inc_simple_NN(dstr));
3932 sref, omg ? omg->mg_obj : dstr, PERL_MAGIC_isa, NULL, 0
3934 mg = mg_find(sref, PERL_MAGIC_isa);
3936 /* Since the *ISA assignment could have affected more than
3937 one stash, don’t call mro_isa_changed_in directly, but let
3938 magic_clearisa do it for us, as it already has the logic for
3939 dealing with globs vs arrays of globs. */
3941 Perl_magic_clearisa(aTHX_ NULL, mg);
3946 if (SvTAINTED(sstr))
3952 Perl_sv_setsv_flags(pTHX_ SV *dstr, register SV* sstr, const I32 flags)
3955 register U32 sflags;
3957 register svtype stype;
3959 PERL_ARGS_ASSERT_SV_SETSV_FLAGS;
3964 if (SvIS_FREED(dstr)) {
3965 Perl_croak(aTHX_ "panic: attempt to copy value %" SVf
3966 " to a freed scalar %p", SVfARG(sstr), (void *)dstr);
3968 SV_CHECK_THINKFIRST_COW_DROP(dstr);
3970 sstr = &PL_sv_undef;
3971 if (SvIS_FREED(sstr)) {
3972 Perl_croak(aTHX_ "panic: attempt to copy freed scalar %p to %p",
3973 (void*)sstr, (void*)dstr);
3975 stype = SvTYPE(sstr);
3976 dtype = SvTYPE(dstr);
3978 (void)SvAMAGIC_off(dstr);
3981 /* need to nuke the magic */
3985 /* There's a lot of redundancy below but we're going for speed here */
3990 if (dtype != SVt_PVGV && dtype != SVt_PVLV) {
3991 (void)SvOK_off(dstr);
3999 sv_upgrade(dstr, SVt_IV);
4003 sv_upgrade(dstr, SVt_PVIV);
4007 goto end_of_first_switch;
4009 (void)SvIOK_only(dstr);
4010 SvIV_set(dstr, SvIVX(sstr));
4013 /* SvTAINTED can only be true if the SV has taint magic, which in
4014 turn means that the SV type is PVMG (or greater). This is the
4015 case statement for SVt_IV, so this cannot be true (whatever gcov
4017 assert(!SvTAINTED(sstr));
4022 if (dtype < SVt_PV && dtype != SVt_IV)
4023 sv_upgrade(dstr, SVt_IV);
4031 sv_upgrade(dstr, SVt_NV);
4035 sv_upgrade(dstr, SVt_PVNV);
4039 goto end_of_first_switch;
4041 SvNV_set(dstr, SvNVX(sstr));
4042 (void)SvNOK_only(dstr);
4043 /* SvTAINTED can only be true if the SV has taint magic, which in
4044 turn means that the SV type is PVMG (or greater). This is the
4045 case statement for SVt_NV, so this cannot be true (whatever gcov
4047 assert(!SvTAINTED(sstr));
4053 #ifdef PERL_OLD_COPY_ON_WRITE
4054 if ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS) {
4055 if (dtype < SVt_PVIV)
4056 sv_upgrade(dstr, SVt_PVIV);
4063 sv_upgrade(dstr, SVt_PV);
4066 if (dtype < SVt_PVIV)
4067 sv_upgrade(dstr, SVt_PVIV);
4070 if (dtype < SVt_PVNV)
4071 sv_upgrade(dstr, SVt_PVNV);
4075 const char * const type = sv_reftype(sstr,0);
4077 Perl_croak(aTHX_ "Bizarre copy of %s in %s", type, OP_DESC(PL_op));
4079 Perl_croak(aTHX_ "Bizarre copy of %s", type);
4084 if (dtype < SVt_REGEXP)
4085 sv_upgrade(dstr, SVt_REGEXP);
4088 /* case SVt_BIND: */
4091 /* SvVALID means that this PVGV is playing at being an FBM. */
4094 if (SvGMAGICAL(sstr) && (flags & SV_GMAGIC)) {
4096 if (SvTYPE(sstr) != stype)
4097 stype = SvTYPE(sstr);
4099 if (isGV_with_GP(sstr) && dtype <= SVt_PVLV) {
4100 glob_assign_glob(dstr, sstr, dtype);
4103 if (stype == SVt_PVLV)
4104 SvUPGRADE(dstr, SVt_PVNV);
4106 SvUPGRADE(dstr, (svtype)stype);
4108 end_of_first_switch:
4110 /* dstr may have been upgraded. */
4111 dtype = SvTYPE(dstr);
4112 sflags = SvFLAGS(sstr);
4114 if (dtype == SVt_PVCV || dtype == SVt_PVFM) {
4115 /* Assigning to a subroutine sets the prototype. */
4118 const char *const ptr = SvPV_const(sstr, len);
4120 SvGROW(dstr, len + 1);
4121 Copy(ptr, SvPVX(dstr), len + 1, char);
4122 SvCUR_set(dstr, len);
4124 SvFLAGS(dstr) |= sflags & SVf_UTF8;
4128 } else if (dtype == SVt_PVAV || dtype == SVt_PVHV) {
4129 const char * const type = sv_reftype(dstr,0);
4131 Perl_croak(aTHX_ "Cannot copy to %s in %s", type, OP_DESC(PL_op));
4133 Perl_croak(aTHX_ "Cannot copy to %s", type);
4134 } else if (sflags & SVf_ROK) {
4135 if (isGV_with_GP(dstr)
4136 && SvTYPE(SvRV(sstr)) == SVt_PVGV && isGV_with_GP(SvRV(sstr))) {
4139 if (GvIMPORTED(dstr) != GVf_IMPORTED
4140 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
4142 GvIMPORTED_on(dstr);
4147 glob_assign_glob(dstr, sstr, dtype);
4151 if (dtype >= SVt_PV) {
4152 if (isGV_with_GP(dstr)) {
4153 glob_assign_ref(dstr, sstr);
4156 if (SvPVX_const(dstr)) {
4162 (void)SvOK_off(dstr);
4163 SvRV_set(dstr, SvREFCNT_inc(SvRV(sstr)));
4164 SvFLAGS(dstr) |= sflags & SVf_ROK;
4165 assert(!(sflags & SVp_NOK));
4166 assert(!(sflags & SVp_IOK));
4167 assert(!(sflags & SVf_NOK));
4168 assert(!(sflags & SVf_IOK));
4170 else if (isGV_with_GP(dstr)) {
4171 if (!(sflags & SVf_OK)) {
4172 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
4173 "Undefined value assigned to typeglob");
4176 GV *gv = gv_fetchsv(sstr, GV_ADD, SVt_PVGV);
4177 if (dstr != (const SV *)gv) {
4178 const char * const name = GvNAME((const GV *)dstr);
4179 const STRLEN len = GvNAMELEN(dstr);
4180 HV *old_stash = NULL;
4181 bool reset_isa = FALSE;
4182 if ((len > 1 && name[len-2] == ':' && name[len-1] == ':')
4183 || (len == 1 && name[0] == ':')) {
4184 /* Set aside the old stash, so we can reset isa caches
4185 on its subclasses. */
4186 if((old_stash = GvHV(dstr))) {
4187 /* Make sure we do not lose it early. */
4188 SvREFCNT_inc_simple_void_NN(
4189 sv_2mortal((SV *)old_stash)
4196 gp_free(MUTABLE_GV(dstr));
4197 GvGP_set(dstr, gp_ref(GvGP(gv)));
4200 HV * const stash = GvHV(dstr);
4202 old_stash ? (HV *)HvENAME_get(old_stash) : stash
4212 else if (dtype == SVt_REGEXP && stype == SVt_REGEXP) {
4213 reg_temp_copy((REGEXP*)dstr, (REGEXP*)sstr);
4215 else if (sflags & SVp_POK) {
4219 * Check to see if we can just swipe the string. If so, it's a
4220 * possible small lose on short strings, but a big win on long ones.
4221 * It might even be a win on short strings if SvPVX_const(dstr)
4222 * has to be allocated and SvPVX_const(sstr) has to be freed.
4223 * Likewise if we can set up COW rather than doing an actual copy, we
4224 * drop to the else clause, as the swipe code and the COW setup code
4225 * have much in common.
4228 /* Whichever path we take through the next code, we want this true,
4229 and doing it now facilitates the COW check. */
4230 (void)SvPOK_only(dstr);
4233 /* If we're already COW then this clause is not true, and if COW
4234 is allowed then we drop down to the else and make dest COW
4235 with us. If caller hasn't said that we're allowed to COW
4236 shared hash keys then we don't do the COW setup, even if the
4237 source scalar is a shared hash key scalar. */
4238 (((flags & SV_COW_SHARED_HASH_KEYS)
4239 ? (sflags & (SVf_FAKE|SVf_READONLY)) != (SVf_FAKE|SVf_READONLY)
4240 : 1 /* If making a COW copy is forbidden then the behaviour we
4241 desire is as if the source SV isn't actually already
4242 COW, even if it is. So we act as if the source flags
4243 are not COW, rather than actually testing them. */
4245 #ifndef PERL_OLD_COPY_ON_WRITE
4246 /* The change that added SV_COW_SHARED_HASH_KEYS makes the logic
4247 when PERL_OLD_COPY_ON_WRITE is defined a little wrong.
4248 Conceptually PERL_OLD_COPY_ON_WRITE being defined should
4249 override SV_COW_SHARED_HASH_KEYS, because it means "always COW"
4250 but in turn, it's somewhat dead code, never expected to go
4251 live, but more kept as a placeholder on how to do it better
4252 in a newer implementation. */
4253 /* If we are COW and dstr is a suitable target then we drop down
4254 into the else and make dest a COW of us. */
4255 || (SvFLAGS(dstr) & CAN_COW_MASK) != CAN_COW_FLAGS
4260 (sflags & SVs_TEMP) && /* slated for free anyway? */
4261 !(sflags & SVf_OOK) && /* and not involved in OOK hack? */
4262 (!(flags & SV_NOSTEAL)) &&
4263 /* and we're allowed to steal temps */
4264 SvREFCNT(sstr) == 1 && /* and no other references to it? */
4265 SvLEN(sstr)) /* and really is a string */
4266 #ifdef PERL_OLD_COPY_ON_WRITE
4267 && ((flags & SV_COW_SHARED_HASH_KEYS)
4268 ? (!((sflags & CAN_COW_MASK) == CAN_COW_FLAGS
4269 && (SvFLAGS(dstr) & CAN_COW_MASK) == CAN_COW_FLAGS
4270 && SvTYPE(sstr) >= SVt_PVIV && SvTYPE(sstr) != SVt_PVFM))
4274 /* Failed the swipe test, and it's not a shared hash key either.
4275 Have to copy the string. */
4276 STRLEN len = SvCUR(sstr);
4277 SvGROW(dstr, len + 1); /* inlined from sv_setpvn */
4278 Move(SvPVX_const(sstr),SvPVX(dstr),len,char);
4279 SvCUR_set(dstr, len);
4280 *SvEND(dstr) = '\0';
4282 /* If PERL_OLD_COPY_ON_WRITE is not defined, then isSwipe will always
4284 /* Either it's a shared hash key, or it's suitable for
4285 copy-on-write or we can swipe the string. */
4287 PerlIO_printf(Perl_debug_log, "Copy on write: sstr --> dstr\n");
4291 #ifdef PERL_OLD_COPY_ON_WRITE
4293 if ((sflags & (SVf_FAKE | SVf_READONLY))
4294 != (SVf_FAKE | SVf_READONLY)) {
4295 SvREADONLY_on(sstr);
4297 /* Make the source SV into a loop of 1.
4298 (about to become 2) */
4299 SV_COW_NEXT_SV_SET(sstr, sstr);
4303 /* Initial code is common. */
4304 if (SvPVX_const(dstr)) { /* we know that dtype >= SVt_PV */
4309 /* making another shared SV. */
4310 STRLEN cur = SvCUR(sstr);
4311 STRLEN len = SvLEN(sstr);
4312 #ifdef PERL_OLD_COPY_ON_WRITE
4314 assert (SvTYPE(dstr) >= SVt_PVIV);
4315 /* SvIsCOW_normal */
4316 /* splice us in between source and next-after-source. */
4317 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4318 SV_COW_NEXT_SV_SET(sstr, dstr);
4319 SvPV_set(dstr, SvPVX_mutable(sstr));
4323 /* SvIsCOW_shared_hash */
4324 DEBUG_C(PerlIO_printf(Perl_debug_log,
4325 "Copy on write: Sharing hash\n"));
4327 assert (SvTYPE(dstr) >= SVt_PV);
4329 HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)))));
4331 SvLEN_set(dstr, len);
4332 SvCUR_set(dstr, cur);
4333 SvREADONLY_on(dstr);
4337 { /* Passes the swipe test. */
4338 SvPV_set(dstr, SvPVX_mutable(sstr));
4339 SvLEN_set(dstr, SvLEN(sstr));
4340 SvCUR_set(dstr, SvCUR(sstr));
4343 (void)SvOK_off(sstr); /* NOTE: nukes most SvFLAGS on sstr */
4344 SvPV_set(sstr, NULL);
4350 if (sflags & SVp_NOK) {
4351 SvNV_set(dstr, SvNVX(sstr));
4353 if (sflags & SVp_IOK) {
4354 SvIV_set(dstr, SvIVX(sstr));
4355 /* Must do this otherwise some other overloaded use of 0x80000000
4356 gets confused. I guess SVpbm_VALID */
4357 if (sflags & SVf_IVisUV)
4360 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_NOK|SVp_NOK|SVf_UTF8);
4362 const MAGIC * const smg = SvVSTRING_mg(sstr);
4364 sv_magic(dstr, NULL, PERL_MAGIC_vstring,
4365 smg->mg_ptr, smg->mg_len);
4366 SvRMAGICAL_on(dstr);
4370 else if (sflags & (SVp_IOK|SVp_NOK)) {
4371 (void)SvOK_off(dstr);
4372 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_IVisUV|SVf_NOK|SVp_NOK);
4373 if (sflags & SVp_IOK) {
4374 /* XXXX Do we want to set IsUV for IV(ROK)? Be extra safe... */
4375 SvIV_set(dstr, SvIVX(sstr));
4377 if (sflags & SVp_NOK) {
4378 SvNV_set(dstr, SvNVX(sstr));
4382 if (isGV_with_GP(sstr)) {
4383 /* This stringification rule for globs is spread in 3 places.
4384 This feels bad. FIXME. */
4385 const U32 wasfake = sflags & SVf_FAKE;
4387 /* FAKE globs can get coerced, so need to turn this off
4388 temporarily if it is on. */
4390 gv_efullname3(dstr, MUTABLE_GV(sstr), "*");
4391 SvFLAGS(sstr) |= wasfake;
4394 (void)SvOK_off(dstr);
4396 if (SvTAINTED(sstr))
4401 =for apidoc sv_setsv_mg
4403 Like C<sv_setsv>, but also handles 'set' magic.
4409 Perl_sv_setsv_mg(pTHX_ SV *const dstr, register SV *const sstr)
4411 PERL_ARGS_ASSERT_SV_SETSV_MG;
4413 sv_setsv(dstr,sstr);
4417 #ifdef PERL_OLD_COPY_ON_WRITE
4419 Perl_sv_setsv_cow(pTHX_ SV *dstr, SV *sstr)
4421 STRLEN cur = SvCUR(sstr);
4422 STRLEN len = SvLEN(sstr);
4423 register char *new_pv;
4425 PERL_ARGS_ASSERT_SV_SETSV_COW;
4428 PerlIO_printf(Perl_debug_log, "Fast copy on write: %p -> %p\n",
4429 (void*)sstr, (void*)dstr);
4436 if (SvTHINKFIRST(dstr))
4437 sv_force_normal_flags(dstr, SV_COW_DROP_PV);
4438 else if (SvPVX_const(dstr))
4439 Safefree(SvPVX_const(dstr));
4443 SvUPGRADE(dstr, SVt_PVIV);
4445 assert (SvPOK(sstr));
4446 assert (SvPOKp(sstr));
4447 assert (!SvIOK(sstr));
4448 assert (!SvIOKp(sstr));
4449 assert (!SvNOK(sstr));
4450 assert (!SvNOKp(sstr));
4452 if (SvIsCOW(sstr)) {
4454 if (SvLEN(sstr) == 0) {
4455 /* source is a COW shared hash key. */
4456 DEBUG_C(PerlIO_printf(Perl_debug_log,
4457 "Fast copy on write: Sharing hash\n"));
4458 new_pv = HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr))));
4461 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4463 assert ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS);
4464 SvUPGRADE(sstr, SVt_PVIV);
4465 SvREADONLY_on(sstr);
4467 DEBUG_C(PerlIO_printf(Perl_debug_log,
4468 "Fast copy on write: Converting sstr to COW\n"));
4469 SV_COW_NEXT_SV_SET(dstr, sstr);
4471 SV_COW_NEXT_SV_SET(sstr, dstr);
4472 new_pv = SvPVX_mutable(sstr);
4475 SvPV_set(dstr, new_pv);
4476 SvFLAGS(dstr) = (SVt_PVIV|SVf_POK|SVp_POK|SVf_FAKE|SVf_READONLY);
4479 SvLEN_set(dstr, len);
4480 SvCUR_set(dstr, cur);
4489 =for apidoc sv_setpvn
4491 Copies a string into an SV. The C<len> parameter indicates the number of
4492 bytes to be copied. If the C<ptr> argument is NULL the SV will become
4493 undefined. Does not handle 'set' magic. See C<sv_setpvn_mg>.
4499 Perl_sv_setpvn(pTHX_ register SV *const sv, register const char *const ptr, register const STRLEN len)
4502 register char *dptr;
4504 PERL_ARGS_ASSERT_SV_SETPVN;
4506 SV_CHECK_THINKFIRST_COW_DROP(sv);
4512 /* len is STRLEN which is unsigned, need to copy to signed */
4515 Perl_croak(aTHX_ "panic: sv_setpvn called with negative strlen");
4517 SvUPGRADE(sv, SVt_PV);
4519 dptr = SvGROW(sv, len + 1);
4520 Move(ptr,dptr,len,char);
4523 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4528 =for apidoc sv_setpvn_mg
4530 Like C<sv_setpvn>, but also handles 'set' magic.
4536 Perl_sv_setpvn_mg(pTHX_ register SV *const sv, register const char *const ptr, register const STRLEN len)
4538 PERL_ARGS_ASSERT_SV_SETPVN_MG;
4540 sv_setpvn(sv,ptr,len);
4545 =for apidoc sv_setpv
4547 Copies a string into an SV. The string must be null-terminated. Does not
4548 handle 'set' magic. See C<sv_setpv_mg>.
4554 Perl_sv_setpv(pTHX_ register SV *const sv, register const char *const ptr)
4557 register STRLEN len;
4559 PERL_ARGS_ASSERT_SV_SETPV;
4561 SV_CHECK_THINKFIRST_COW_DROP(sv);
4567 SvUPGRADE(sv, SVt_PV);
4569 SvGROW(sv, len + 1);
4570 Move(ptr,SvPVX(sv),len+1,char);
4572 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4577 =for apidoc sv_setpv_mg
4579 Like C<sv_setpv>, but also handles 'set' magic.
4585 Perl_sv_setpv_mg(pTHX_ register SV *const sv, register const char *const ptr)
4587 PERL_ARGS_ASSERT_SV_SETPV_MG;
4594 =for apidoc sv_usepvn_flags
4596 Tells an SV to use C<ptr> to find its string value. Normally the
4597 string is stored inside the SV but sv_usepvn allows the SV to use an
4598 outside string. The C<ptr> should point to memory that was allocated
4599 by C<malloc>. The string length, C<len>, must be supplied. By default
4600 this function will realloc (i.e. move) the memory pointed to by C<ptr>,
4601 so that pointer should not be freed or used by the programmer after
4602 giving it to sv_usepvn, and neither should any pointers from "behind"
4603 that pointer (e.g. ptr + 1) be used.
4605 If C<flags> & SV_SMAGIC is true, will call SvSETMAGIC. If C<flags> &
4606 SV_HAS_TRAILING_NUL is true, then C<ptr[len]> must be NUL, and the realloc
4607 will be skipped. (i.e. the buffer is actually at least 1 byte longer than
4608 C<len>, and already meets the requirements for storing in C<SvPVX>)
4614 Perl_sv_usepvn_flags(pTHX_ SV *const sv, char *ptr, const STRLEN len, const U32 flags)
4619 PERL_ARGS_ASSERT_SV_USEPVN_FLAGS;
4621 SV_CHECK_THINKFIRST_COW_DROP(sv);
4622 SvUPGRADE(sv, SVt_PV);
4625 if (flags & SV_SMAGIC)
4629 if (SvPVX_const(sv))
4633 if (flags & SV_HAS_TRAILING_NUL)
4634 assert(ptr[len] == '\0');
4637 allocate = (flags & SV_HAS_TRAILING_NUL)
4639 #ifdef Perl_safesysmalloc_size
4642 PERL_STRLEN_ROUNDUP(len + 1);
4644 if (flags & SV_HAS_TRAILING_NUL) {
4645 /* It's long enough - do nothing.
4646 Specifically Perl_newCONSTSUB is relying on this. */
4649 /* Force a move to shake out bugs in callers. */
4650 char *new_ptr = (char*)safemalloc(allocate);
4651 Copy(ptr, new_ptr, len, char);
4652 PoisonFree(ptr,len,char);
4656 ptr = (char*) saferealloc (ptr, allocate);
4659 #ifdef Perl_safesysmalloc_size
4660 SvLEN_set(sv, Perl_safesysmalloc_size(ptr));
4662 SvLEN_set(sv, allocate);
4666 if (!(flags & SV_HAS_TRAILING_NUL)) {
4669 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4671 if (flags & SV_SMAGIC)
4675 #ifdef PERL_OLD_COPY_ON_WRITE
4676 /* Need to do this *after* making the SV normal, as we need the buffer
4677 pointer to remain valid until after we've copied it. If we let go too early,
4678 another thread could invalidate it by unsharing last of the same hash key
4679 (which it can do by means other than releasing copy-on-write Svs)
4680 or by changing the other copy-on-write SVs in the loop. */
4682 S_sv_release_COW(pTHX_ register SV *sv, const char *pvx, SV *after)
4684 PERL_ARGS_ASSERT_SV_RELEASE_COW;
4686 { /* this SV was SvIsCOW_normal(sv) */
4687 /* we need to find the SV pointing to us. */
4688 SV *current = SV_COW_NEXT_SV(after);
4690 if (current == sv) {
4691 /* The SV we point to points back to us (there were only two of us
4693 Hence other SV is no longer copy on write either. */
4695 SvREADONLY_off(after);
4697 /* We need to follow the pointers around the loop. */
4699 while ((next = SV_COW_NEXT_SV(current)) != sv) {
4702 /* don't loop forever if the structure is bust, and we have
4703 a pointer into a closed loop. */
4704 assert (current != after);
4705 assert (SvPVX_const(current) == pvx);
4707 /* Make the SV before us point to the SV after us. */
4708 SV_COW_NEXT_SV_SET(current, after);
4714 =for apidoc sv_force_normal_flags
4716 Undo various types of fakery on an SV: if the PV is a shared string, make
4717 a private copy; if we're a ref, stop refing; if we're a glob, downgrade to
4718 an xpvmg; if we're a copy-on-write scalar, this is the on-write time when
4719 we do the copy, and is also used locally. If C<SV_COW_DROP_PV> is set
4720 then a copy-on-write scalar drops its PV buffer (if any) and becomes
4721 SvPOK_off rather than making a copy. (Used where this scalar is about to be
4722 set to some other value.) In addition, the C<flags> parameter gets passed to
4723 C<sv_unref_flags()> when unreffing. C<sv_force_normal> calls this function
4724 with flags set to 0.
4730 Perl_sv_force_normal_flags(pTHX_ register SV *const sv, const U32 flags)
4734 PERL_ARGS_ASSERT_SV_FORCE_NORMAL_FLAGS;
4736 #ifdef PERL_OLD_COPY_ON_WRITE
4737 if (SvREADONLY(sv)) {
4739 const char * const pvx = SvPVX_const(sv);
4740 const STRLEN len = SvLEN(sv);
4741 const STRLEN cur = SvCUR(sv);
4742 /* next COW sv in the loop. If len is 0 then this is a shared-hash
4743 key scalar, so we mustn't attempt to call SV_COW_NEXT_SV(), as
4744 we'll fail an assertion. */
4745 SV * const next = len ? SV_COW_NEXT_SV(sv) : 0;
4748 PerlIO_printf(Perl_debug_log,
4749 "Copy on write: Force normal %ld\n",
4755 /* This SV doesn't own the buffer, so need to Newx() a new one: */
4758 if (flags & SV_COW_DROP_PV) {
4759 /* OK, so we don't need to copy our buffer. */
4762 SvGROW(sv, cur + 1);
4763 Move(pvx,SvPVX(sv),cur,char);
4768 sv_release_COW(sv, pvx, next);
4770 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4776 else if (IN_PERL_RUNTIME)
4777 Perl_croak_no_modify(aTHX);
4780 if (SvREADONLY(sv)) {
4782 const char * const pvx = SvPVX_const(sv);
4783 const STRLEN len = SvCUR(sv);
4788 SvGROW(sv, len + 1);
4789 Move(pvx,SvPVX(sv),len,char);
4791 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4793 else if (IN_PERL_RUNTIME)
4794 Perl_croak_no_modify(aTHX);
4798 sv_unref_flags(sv, flags);
4799 else if (SvFAKE(sv) && isGV_with_GP(sv))
4801 else if (SvFAKE(sv) && SvTYPE(sv) == SVt_REGEXP) {
4802 /* Need to downgrade the REGEXP to a simple(r) scalar. This is analogous
4803 to sv_unglob. We only need it here, so inline it. */
4804 const svtype new_type = SvMAGIC(sv) || SvSTASH(sv) ? SVt_PVMG : SVt_PV;
4805 SV *const temp = newSV_type(new_type);
4806 void *const temp_p = SvANY(sv);
4808 if (new_type == SVt_PVMG) {
4809 SvMAGIC_set(temp, SvMAGIC(sv));
4810 SvMAGIC_set(sv, NULL);
4811 SvSTASH_set(temp, SvSTASH(sv));
4812 SvSTASH_set(sv, NULL);
4814 SvCUR_set(temp, SvCUR(sv));
4815 /* Remember that SvPVX is in the head, not the body. */
4817 SvLEN_set(temp, SvLEN(sv));
4818 /* This signals "buffer is owned by someone else" in sv_clear,
4819 which is the least effort way to stop it freeing the buffer.
4821 SvLEN_set(sv, SvLEN(sv)+1);
4823 /* Their buffer is already owned by someone else. */
4824 SvPVX(sv) = savepvn(SvPVX(sv), SvCUR(sv));
4825 SvLEN_set(temp, SvCUR(sv)+1);
4828 /* Now swap the rest of the bodies. */
4830 SvFLAGS(sv) &= ~(SVf_FAKE|SVTYPEMASK);
4831 SvFLAGS(sv) |= new_type;
4832 SvANY(sv) = SvANY(temp);
4834 SvFLAGS(temp) &= ~(SVTYPEMASK);
4835 SvFLAGS(temp) |= SVt_REGEXP|SVf_FAKE;
4836 SvANY(temp) = temp_p;
4845 Efficient removal of characters from the beginning of the string buffer.
4846 SvPOK(sv) must be true and the C<ptr> must be a pointer to somewhere inside
4847 the string buffer. The C<ptr> becomes the first character of the adjusted
4848 string. Uses the "OOK hack".
4849 Beware: after this function returns, C<ptr> and SvPVX_const(sv) may no longer
4850 refer to the same chunk of data.
4856 Perl_sv_chop(pTHX_ register SV *const sv, register const char *const ptr)
4862 const U8 *real_start;
4866 PERL_ARGS_ASSERT_SV_CHOP;
4868 if (!ptr || !SvPOKp(sv))
4870 delta = ptr - SvPVX_const(sv);
4872 /* Nothing to do. */
4875 /* SvPVX(sv) may move in SV_CHECK_THINKFIRST(sv), but after this line,
4876 nothing uses the value of ptr any more. */
4877 max_delta = SvLEN(sv) ? SvLEN(sv) : SvCUR(sv);
4878 if (ptr <= SvPVX_const(sv))
4879 Perl_croak(aTHX_ "panic: sv_chop ptr=%p, start=%p, end=%p",
4880 ptr, SvPVX_const(sv), SvPVX_const(sv) + max_delta);
4881 SV_CHECK_THINKFIRST(sv);
4882 if (delta > max_delta)
4883 Perl_croak(aTHX_ "panic: sv_chop ptr=%p (was %p), start=%p, end=%p",
4884 SvPVX_const(sv) + delta, ptr, SvPVX_const(sv),
4885 SvPVX_const(sv) + max_delta);
4888 if (!SvLEN(sv)) { /* make copy of shared string */
4889 const char *pvx = SvPVX_const(sv);
4890 const STRLEN len = SvCUR(sv);
4891 SvGROW(sv, len + 1);
4892 Move(pvx,SvPVX(sv),len,char);
4895 SvFLAGS(sv) |= SVf_OOK;
4898 SvOOK_offset(sv, old_delta);
4900 SvLEN_set(sv, SvLEN(sv) - delta);
4901 SvCUR_set(sv, SvCUR(sv) - delta);
4902 SvPV_set(sv, SvPVX(sv) + delta);
4904 p = (U8 *)SvPVX_const(sv);
4909 real_start = p - delta;
4913 if (delta < 0x100) {
4917 p -= sizeof(STRLEN);
4918 Copy((U8*)&delta, p, sizeof(STRLEN), U8);
4922 /* Fill the preceding buffer with sentinals to verify that no-one is
4924 while (p > real_start) {
4932 =for apidoc sv_catpvn
4934 Concatenates the string onto the end of the string which is in the SV. The
4935 C<len> indicates number of bytes to copy. If the SV has the UTF-8
4936 status set, then the bytes appended should be valid UTF-8.
4937 Handles 'get' magic, but not 'set' magic. See C<sv_catpvn_mg>.
4939 =for apidoc sv_catpvn_flags
4941 Concatenates the string onto the end of the string which is in the SV. The
4942 C<len> indicates number of bytes to copy. If the SV has the UTF-8
4943 status set, then the bytes appended should be valid UTF-8.
4944 If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<dsv> if
4945 appropriate, else not. 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;
4960 SvGROW(dsv, dlen + slen + 1);
4962 sstr = SvPVX_const(dsv);
4963 Move(sstr, SvPVX(dsv) + dlen, slen, char);
4964 SvCUR_set(dsv, SvCUR(dsv) + slen);
4966 (void)SvPOK_only_UTF8(dsv); /* validate pointer */
4968 if (flags & SV_SMAGIC)
4973 =for apidoc sv_catsv
4975 Concatenates the string from SV C<ssv> onto the end of the string in
4976 SV C<dsv>. Modifies C<dsv> but not C<ssv>. Handles 'get' magic, but
4977 not 'set' magic. See C<sv_catsv_mg>.
4979 =for apidoc sv_catsv_flags
4981 Concatenates the string from SV C<ssv> onto the end of the string in
4982 SV C<dsv>. Modifies C<dsv> but not C<ssv>. If C<flags> has C<SV_GMAGIC>
4983 bit set, will C<mg_get> on the SVs if appropriate, else not. C<sv_catsv>
4984 and C<sv_catsv_nomg> are implemented in terms of this function.
4989 Perl_sv_catsv_flags(pTHX_ SV *const dsv, register SV *const ssv, const I32 flags)
4993 PERL_ARGS_ASSERT_SV_CATSV_FLAGS;
4997 const char *spv = SvPV_flags_const(ssv, slen, flags);
4999 /* sutf8 and dutf8 were type bool, but under USE_ITHREADS,
5000 gcc version 2.95.2 20000220 (Debian GNU/Linux) for
5001 Linux xxx 2.2.17 on sparc64 with gcc -O2, we erroneously
5002 get dutf8 = 0x20000000, (i.e. SVf_UTF8) even though
5003 dsv->sv_flags doesn't have that bit set.
5004 Andy Dougherty 12 Oct 2001
5006 const I32 sutf8 = DO_UTF8(ssv);
5009 if (SvGMAGICAL(dsv) && (flags & SV_GMAGIC))
5011 dutf8 = DO_UTF8(dsv);
5013 if (dutf8 != sutf8) {
5015 /* Not modifying source SV, so taking a temporary copy. */
5016 SV* const csv = newSVpvn_flags(spv, slen, SVs_TEMP);
5018 sv_utf8_upgrade(csv);
5019 spv = SvPV_const(csv, slen);
5022 /* Leave enough space for the cat that's about to happen */
5023 sv_utf8_upgrade_flags_grow(dsv, 0, slen);
5025 sv_catpvn_nomg(dsv, spv, slen);
5028 if (flags & SV_SMAGIC)
5033 =for apidoc sv_catpv
5035 Concatenates the string onto the end of the string which is in the SV.
5036 If the SV has the UTF-8 status set, then the bytes appended should be
5037 valid UTF-8. Handles 'get' magic, but not 'set' magic. See C<sv_catpv_mg>.
5042 Perl_sv_catpv(pTHX_ register SV *const sv, register const char *ptr)
5045 register STRLEN len;
5049 PERL_ARGS_ASSERT_SV_CATPV;
5053 junk = SvPV_force(sv, tlen);
5055 SvGROW(sv, tlen + len + 1);
5057 ptr = SvPVX_const(sv);
5058 Move(ptr,SvPVX(sv)+tlen,len+1,char);
5059 SvCUR_set(sv, SvCUR(sv) + len);
5060 (void)SvPOK_only_UTF8(sv); /* validate pointer */
5065 =for apidoc sv_catpv_flags
5067 Concatenates the string onto the end of the string which is in the SV.
5068 If the SV has the UTF-8 status set, then the bytes appended should
5069 be valid UTF-8. If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get>
5070 on the SVs if appropriate, else not.
5076 Perl_sv_catpv_flags(pTHX_ SV *dstr, const char *sstr, const I32 flags)
5078 PERL_ARGS_ASSERT_SV_CATPV_FLAGS;
5079 sv_catpvn_flags(dstr, sstr, strlen(sstr), flags);
5083 =for apidoc sv_catpv_mg
5085 Like C<sv_catpv>, but also handles 'set' magic.
5091 Perl_sv_catpv_mg(pTHX_ register SV *const sv, register const char *const ptr)
5093 PERL_ARGS_ASSERT_SV_CATPV_MG;
5102 Creates a new SV. A non-zero C<len> parameter indicates the number of
5103 bytes of preallocated string space the SV should have. An extra byte for a
5104 trailing NUL is also reserved. (SvPOK is not set for the SV even if string
5105 space is allocated.) The reference count for the new SV is set to 1.
5107 In 5.9.3, newSV() replaces the older NEWSV() API, and drops the first
5108 parameter, I<x>, a debug aid which allowed callers to identify themselves.
5109 This aid has been superseded by a new build option, PERL_MEM_LOG (see
5110 L<perlhack/PERL_MEM_LOG>). The older API is still there for use in XS
5111 modules supporting older perls.
5117 Perl_newSV(pTHX_ const STRLEN len)
5124 sv_upgrade(sv, SVt_PV);
5125 SvGROW(sv, len + 1);
5130 =for apidoc sv_magicext
5132 Adds magic to an SV, upgrading it if necessary. Applies the
5133 supplied vtable and returns a pointer to the magic added.
5135 Note that C<sv_magicext> will allow things that C<sv_magic> will not.
5136 In particular, you can add magic to SvREADONLY SVs, and add more than
5137 one instance of the same 'how'.
5139 If C<namlen> is greater than zero then a C<savepvn> I<copy> of C<name> is
5140 stored, if C<namlen> is zero then C<name> is stored as-is and - as another
5141 special case - if C<(name && namlen == HEf_SVKEY)> then C<name> is assumed
5142 to contain an C<SV*> and is stored as-is with its REFCNT incremented.
5144 (This is now used as a subroutine by C<sv_magic>.)
5149 Perl_sv_magicext(pTHX_ SV *const sv, SV *const obj, const int how,
5150 const MGVTBL *const vtable, const char *const name, const I32 namlen)
5155 PERL_ARGS_ASSERT_SV_MAGICEXT;
5157 SvUPGRADE(sv, SVt_PVMG);
5158 Newxz(mg, 1, MAGIC);
5159 mg->mg_moremagic = SvMAGIC(sv);
5160 SvMAGIC_set(sv, mg);
5162 /* Sometimes a magic contains a reference loop, where the sv and
5163 object refer to each other. To prevent a reference loop that
5164 would prevent such objects being freed, we look for such loops
5165 and if we find one we avoid incrementing the object refcount.
5167 Note we cannot do this to avoid self-tie loops as intervening RV must
5168 have its REFCNT incremented to keep it in existence.
5171 if (!obj || obj == sv ||
5172 how == PERL_MAGIC_arylen ||
5173 how == PERL_MAGIC_symtab ||
5174 (SvTYPE(obj) == SVt_PVGV &&
5175 (GvSV(obj) == sv || GvHV(obj) == (const HV *)sv
5176 || GvAV(obj) == (const AV *)sv || GvCV(obj) == (const CV *)sv
5177 || GvIOp(obj) == (const IO *)sv || GvFORM(obj) == (const CV *)sv)))
5182 mg->mg_obj = SvREFCNT_inc_simple(obj);
5183 mg->mg_flags |= MGf_REFCOUNTED;
5186 /* Normal self-ties simply pass a null object, and instead of
5187 using mg_obj directly, use the SvTIED_obj macro to produce a
5188 new RV as needed. For glob "self-ties", we are tieing the PVIO
5189 with an RV obj pointing to the glob containing the PVIO. In
5190 this case, to avoid a reference loop, we need to weaken the
5194 if (how == PERL_MAGIC_tiedscalar && SvTYPE(sv) == SVt_PVIO &&
5195 obj && SvROK(obj) && GvIO(SvRV(obj)) == (const IO *)sv)
5201 mg->mg_len = namlen;
5204 mg->mg_ptr = savepvn(name, namlen);
5205 else if (namlen == HEf_SVKEY) {
5206 /* Yes, this is casting away const. This is only for the case of
5207 HEf_SVKEY. I think we need to document this aberation of the
5208 constness of the API, rather than making name non-const, as
5209 that change propagating outwards a long way. */
5210 mg->mg_ptr = (char*)SvREFCNT_inc_simple_NN((SV *)name);
5212 mg->mg_ptr = (char *) name;
5214 mg->mg_virtual = (MGVTBL *) vtable;
5218 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
5223 =for apidoc sv_magic
5225 Adds magic to an SV. First upgrades C<sv> to type C<SVt_PVMG> if necessary,
5226 then adds a new magic item of type C<how> to the head of the magic list.
5228 See C<sv_magicext> (which C<sv_magic> now calls) for a description of the
5229 handling of the C<name> and C<namlen> arguments.
5231 You need to use C<sv_magicext> to add magic to SvREADONLY SVs and also
5232 to add more than one instance of the same 'how'.
5238 Perl_sv_magic(pTHX_ register SV *const sv, SV *const obj, const int how,
5239 const char *const name, const I32 namlen)
5242 const MGVTBL *vtable;
5245 PERL_ARGS_ASSERT_SV_MAGIC;
5247 #ifdef PERL_OLD_COPY_ON_WRITE
5249 sv_force_normal_flags(sv, 0);
5251 if (SvREADONLY(sv)) {
5253 /* its okay to attach magic to shared strings; the subsequent
5254 * upgrade to PVMG will unshare the string */
5255 !(SvFAKE(sv) && SvTYPE(sv) < SVt_PVMG)
5258 && how != PERL_MAGIC_regex_global
5259 && how != PERL_MAGIC_bm
5260 && how != PERL_MAGIC_fm
5261 && how != PERL_MAGIC_sv
5262 && how != PERL_MAGIC_backref
5265 Perl_croak_no_modify(aTHX);
5268 if (SvMAGICAL(sv) || (how == PERL_MAGIC_taint && SvTYPE(sv) >= SVt_PVMG)) {
5269 if (SvMAGIC(sv) && (mg = mg_find(sv, how))) {
5270 /* sv_magic() refuses to add a magic of the same 'how' as an
5273 if (how == PERL_MAGIC_taint) {
5275 /* Any scalar which already had taint magic on which someone
5276 (erroneously?) did SvIOK_on() or similar will now be
5277 incorrectly sporting public "OK" flags. */
5278 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
5286 vtable = &PL_vtbl_sv;
5288 case PERL_MAGIC_overload:
5289 vtable = &PL_vtbl_amagic;
5291 case PERL_MAGIC_overload_elem:
5292 vtable = &PL_vtbl_amagicelem;
5294 case PERL_MAGIC_overload_table:
5295 vtable = &PL_vtbl_ovrld;
5298 vtable = &PL_vtbl_bm;
5300 case PERL_MAGIC_regdata:
5301 vtable = &PL_vtbl_regdata;
5303 case PERL_MAGIC_regdatum:
5304 vtable = &PL_vtbl_regdatum;
5306 case PERL_MAGIC_env:
5307 vtable = &PL_vtbl_env;
5310 vtable = &PL_vtbl_fm;
5312 case PERL_MAGIC_envelem:
5313 vtable = &PL_vtbl_envelem;
5315 case PERL_MAGIC_regex_global:
5316 vtable = &PL_vtbl_mglob;
5318 case PERL_MAGIC_isa:
5319 vtable = &PL_vtbl_isa;
5321 case PERL_MAGIC_isaelem:
5322 vtable = &PL_vtbl_isaelem;
5324 case PERL_MAGIC_nkeys:
5325 vtable = &PL_vtbl_nkeys;
5327 case PERL_MAGIC_dbfile:
5330 case PERL_MAGIC_dbline:
5331 vtable = &PL_vtbl_dbline;
5333 #ifdef USE_LOCALE_COLLATE
5334 case PERL_MAGIC_collxfrm:
5335 vtable = &PL_vtbl_collxfrm;
5337 #endif /* USE_LOCALE_COLLATE */
5338 case PERL_MAGIC_tied:
5339 vtable = &PL_vtbl_pack;
5341 case PERL_MAGIC_tiedelem:
5342 case PERL_MAGIC_tiedscalar:
5343 vtable = &PL_vtbl_packelem;
5346 vtable = &PL_vtbl_regexp;
5348 case PERL_MAGIC_sig:
5349 vtable = &PL_vtbl_sig;
5351 case PERL_MAGIC_sigelem:
5352 vtable = &PL_vtbl_sigelem;
5354 case PERL_MAGIC_taint:
5355 vtable = &PL_vtbl_taint;
5357 case PERL_MAGIC_uvar:
5358 vtable = &PL_vtbl_uvar;
5360 case PERL_MAGIC_vec:
5361 vtable = &PL_vtbl_vec;
5363 case PERL_MAGIC_arylen_p:
5364 case PERL_MAGIC_rhash:
5365 case PERL_MAGIC_symtab:
5366 case PERL_MAGIC_vstring:
5367 case PERL_MAGIC_checkcall:
5370 case PERL_MAGIC_utf8:
5371 vtable = &PL_vtbl_utf8;
5373 case PERL_MAGIC_substr:
5374 vtable = &PL_vtbl_substr;
5376 case PERL_MAGIC_defelem:
5377 vtable = &PL_vtbl_defelem;
5379 case PERL_MAGIC_arylen:
5380 vtable = &PL_vtbl_arylen;
5382 case PERL_MAGIC_pos:
5383 vtable = &PL_vtbl_pos;
5385 case PERL_MAGIC_backref:
5386 vtable = &PL_vtbl_backref;
5388 case PERL_MAGIC_hintselem:
5389 vtable = &PL_vtbl_hintselem;
5391 case PERL_MAGIC_hints:
5392 vtable = &PL_vtbl_hints;
5394 case PERL_MAGIC_ext:
5395 /* Reserved for use by extensions not perl internals. */
5396 /* Useful for attaching extension internal data to perl vars. */
5397 /* Note that multiple extensions may clash if magical scalars */
5398 /* etc holding private data from one are passed to another. */
5402 Perl_croak(aTHX_ "Don't know how to handle magic of type \\%o", how);
5405 /* Rest of work is done else where */
5406 mg = sv_magicext(sv,obj,how,vtable,name,namlen);
5409 case PERL_MAGIC_taint:
5412 case PERL_MAGIC_ext:
5413 case PERL_MAGIC_dbfile:
5420 S_sv_unmagicext_flags(pTHX_ SV *const sv, const int type, MGVTBL *vtbl, const U32 flags)
5427 if (SvTYPE(sv) < SVt_PVMG || !SvMAGIC(sv))
5429 mgp = &(((XPVMG*) SvANY(sv))->xmg_u.xmg_magic);
5430 for (mg = *mgp; mg; mg = *mgp) {
5431 const MGVTBL* const virt = mg->mg_virtual;
5432 if (mg->mg_type == type && (!flags || virt == vtbl)) {
5433 *mgp = mg->mg_moremagic;
5434 if (virt && virt->svt_free)
5435 virt->svt_free(aTHX_ sv, mg);
5436 if (mg->mg_ptr && mg->mg_type != PERL_MAGIC_regex_global) {
5438 Safefree(mg->mg_ptr);
5439 else if (mg->mg_len == HEf_SVKEY)
5440 SvREFCNT_dec(MUTABLE_SV(mg->mg_ptr));
5441 else if (mg->mg_type == PERL_MAGIC_utf8)
5442 Safefree(mg->mg_ptr);
5444 if (mg->mg_flags & MGf_REFCOUNTED)
5445 SvREFCNT_dec(mg->mg_obj);
5449 mgp = &mg->mg_moremagic;
5452 if (SvMAGICAL(sv)) /* if we're under save_magic, wait for restore_magic; */
5453 mg_magical(sv); /* else fix the flags now */
5457 SvFLAGS(sv) |= (SvFLAGS(sv) & (SVp_IOK|SVp_NOK|SVp_POK)) >> PRIVSHIFT;
5463 =for apidoc sv_unmagic
5465 Removes all magic of type C<type> from an SV.
5471 Perl_sv_unmagic(pTHX_ SV *const sv, const int type)
5473 PERL_ARGS_ASSERT_SV_UNMAGIC;
5474 return S_sv_unmagicext_flags(aTHX_ sv, type, NULL, 0);
5478 =for apidoc sv_unmagicext
5480 Removes all magic of type C<type> with the specified C<vtbl> from an SV.
5486 Perl_sv_unmagicext(pTHX_ SV *const sv, const int type, MGVTBL *vtbl)
5488 PERL_ARGS_ASSERT_SV_UNMAGICEXT;
5489 return S_sv_unmagicext_flags(aTHX_ sv, type, vtbl, 1);
5493 =for apidoc sv_rvweaken
5495 Weaken a reference: set the C<SvWEAKREF> flag on this RV; give the
5496 referred-to SV C<PERL_MAGIC_backref> magic if it hasn't already; and
5497 push a back-reference to this RV onto the array of backreferences
5498 associated with that magic. If the RV is magical, set magic will be
5499 called after the RV is cleared.
5505 Perl_sv_rvweaken(pTHX_ SV *const sv)
5509 PERL_ARGS_ASSERT_SV_RVWEAKEN;
5511 if (!SvOK(sv)) /* let undefs pass */
5514 Perl_croak(aTHX_ "Can't weaken a nonreference");
5515 else if (SvWEAKREF(sv)) {
5516 Perl_ck_warner(aTHX_ packWARN(WARN_MISC), "Reference is already weak");
5520 Perl_sv_add_backref(aTHX_ tsv, sv);
5526 /* Give tsv backref magic if it hasn't already got it, then push a
5527 * back-reference to sv onto the array associated with the backref magic.
5529 * As an optimisation, if there's only one backref and it's not an AV,
5530 * store it directly in the HvAUX or mg_obj slot, avoiding the need to
5531 * allocate an AV. (Whether the slot holds an AV tells us whether this is
5535 /* A discussion about the backreferences array and its refcount:
5537 * The AV holding the backreferences is pointed to either as the mg_obj of
5538 * PERL_MAGIC_backref, or in the specific case of a HV, from the
5539 * xhv_backreferences field. The array is created with a refcount
5540 * of 2. This means that if during global destruction the array gets
5541 * picked on before its parent to have its refcount decremented by the
5542 * random zapper, it won't actually be freed, meaning it's still there for
5543 * when its parent gets freed.
5545 * When the parent SV is freed, the extra ref is killed by
5546 * Perl_sv_kill_backrefs. The other ref is killed, in the case of magic,
5547 * by mg_free() / MGf_REFCOUNTED, or for a hash, by Perl_hv_kill_backrefs.
5549 * When a single backref SV is stored directly, it is not reference
5554 Perl_sv_add_backref(pTHX_ SV *const tsv, SV *const sv)
5561 PERL_ARGS_ASSERT_SV_ADD_BACKREF;
5563 /* find slot to store array or singleton backref */
5565 if (SvTYPE(tsv) == SVt_PVHV) {
5566 svp = (SV**)Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5569 (SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL))))
5571 sv_magic(tsv, NULL, PERL_MAGIC_backref, NULL, 0);
5572 mg = mg_find(tsv, PERL_MAGIC_backref);
5574 svp = &(mg->mg_obj);
5577 /* create or retrieve the array */
5579 if ( (!*svp && SvTYPE(sv) == SVt_PVAV)
5580 || (*svp && SvTYPE(*svp) != SVt_PVAV)
5585 SvREFCNT_inc_simple_void(av);
5586 /* av now has a refcnt of 2; see discussion above */
5588 /* move single existing backref to the array */
5590 AvARRAY(av)[++AvFILLp(av)] = *svp; /* av_push() */
5594 mg->mg_flags |= MGf_REFCOUNTED;
5597 av = MUTABLE_AV(*svp);
5600 /* optimisation: store single backref directly in HvAUX or mg_obj */
5604 /* push new backref */
5605 assert(SvTYPE(av) == SVt_PVAV);
5606 if (AvFILLp(av) >= AvMAX(av)) {
5607 av_extend(av, AvFILLp(av)+1);
5609 AvARRAY(av)[++AvFILLp(av)] = sv; /* av_push() */
5612 /* delete a back-reference to ourselves from the backref magic associated
5613 * with the SV we point to.
5617 Perl_sv_del_backref(pTHX_ SV *const tsv, SV *const sv)
5622 PERL_ARGS_ASSERT_SV_DEL_BACKREF;
5624 if (SvTYPE(tsv) == SVt_PVHV) {
5626 svp = (SV**)Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5630 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
5631 svp = mg ? &(mg->mg_obj) : NULL;
5635 Perl_croak(aTHX_ "panic: del_backref");
5637 if (SvTYPE(*svp) == SVt_PVAV) {
5641 AV * const av = (AV*)*svp;
5643 assert(!SvIS_FREED(av));
5647 /* for an SV with N weak references to it, if all those
5648 * weak refs are deleted, then sv_del_backref will be called
5649 * N times and O(N^2) compares will be done within the backref
5650 * array. To ameliorate this potential slowness, we:
5651 * 1) make sure this code is as tight as possible;
5652 * 2) when looking for SV, look for it at both the head and tail of the
5653 * array first before searching the rest, since some create/destroy
5654 * patterns will cause the backrefs to be freed in order.
5661 SV **p = &svp[fill];
5662 SV *const topsv = *p;
5669 /* We weren't the last entry.
5670 An unordered list has this property that you
5671 can take the last element off the end to fill
5672 the hole, and it's still an unordered list :-)
5678 break; /* should only be one */
5685 AvFILLp(av) = fill-1;
5688 /* optimisation: only a single backref, stored directly */
5690 Perl_croak(aTHX_ "panic: del_backref");
5697 Perl_sv_kill_backrefs(pTHX_ SV *const sv, AV *const av)
5703 PERL_ARGS_ASSERT_SV_KILL_BACKREFS;
5708 /* after multiple passes through Perl_sv_clean_all() for a thinngy
5709 * that has badly leaked, the backref array may have gotten freed,
5710 * since we only protect it against 1 round of cleanup */
5711 if (SvIS_FREED(av)) {
5712 if (PL_in_clean_all) /* All is fair */
5715 "panic: magic_killbackrefs (freed backref AV/SV)");
5719 is_array = (SvTYPE(av) == SVt_PVAV);
5721 assert(!SvIS_FREED(av));
5724 last = svp + AvFILLp(av);
5727 /* optimisation: only a single backref, stored directly */
5733 while (svp <= last) {
5735 SV *const referrer = *svp;
5736 if (SvWEAKREF(referrer)) {
5737 /* XXX Should we check that it hasn't changed? */
5738 assert(SvROK(referrer));
5739 SvRV_set(referrer, 0);
5741 SvWEAKREF_off(referrer);
5742 SvSETMAGIC(referrer);
5743 } else if (SvTYPE(referrer) == SVt_PVGV ||
5744 SvTYPE(referrer) == SVt_PVLV) {
5745 assert(SvTYPE(sv) == SVt_PVHV); /* stash backref */
5746 /* You lookin' at me? */
5747 assert(GvSTASH(referrer));
5748 assert(GvSTASH(referrer) == (const HV *)sv);
5749 GvSTASH(referrer) = 0;
5750 } else if (SvTYPE(referrer) == SVt_PVCV ||
5751 SvTYPE(referrer) == SVt_PVFM) {
5752 if (SvTYPE(sv) == SVt_PVHV) { /* stash backref */
5753 /* You lookin' at me? */
5754 assert(CvSTASH(referrer));
5755 assert(CvSTASH(referrer) == (const HV *)sv);
5756 SvANY(MUTABLE_CV(referrer))->xcv_stash = 0;
5759 assert(SvTYPE(sv) == SVt_PVGV);
5760 /* You lookin' at me? */
5761 assert(CvGV(referrer));
5762 assert(CvGV(referrer) == (const GV *)sv);
5763 anonymise_cv_maybe(MUTABLE_GV(sv),
5764 MUTABLE_CV(referrer));
5769 "panic: magic_killbackrefs (flags=%"UVxf")",
5770 (UV)SvFLAGS(referrer));
5781 SvREFCNT_dec(av); /* remove extra count added by sv_add_backref() */
5787 =for apidoc sv_insert
5789 Inserts a string at the specified offset/length within the SV. Similar to
5790 the Perl substr() function. Handles get magic.
5792 =for apidoc sv_insert_flags
5794 Same as C<sv_insert>, but the extra C<flags> are passed the C<SvPV_force_flags> that applies to C<bigstr>.
5800 Perl_sv_insert_flags(pTHX_ SV *const bigstr, const STRLEN offset, const STRLEN len, const char *const little, const STRLEN littlelen, const U32 flags)
5805 register char *midend;
5806 register char *bigend;
5810 PERL_ARGS_ASSERT_SV_INSERT_FLAGS;
5813 Perl_croak(aTHX_ "Can't modify non-existent substring");
5814 SvPV_force_flags(bigstr, curlen, flags);
5815 (void)SvPOK_only_UTF8(bigstr);
5816 if (offset + len > curlen) {
5817 SvGROW(bigstr, offset+len+1);
5818 Zero(SvPVX(bigstr)+curlen, offset+len-curlen, char);
5819 SvCUR_set(bigstr, offset+len);
5823 i = littlelen - len;
5824 if (i > 0) { /* string might grow */
5825 big = SvGROW(bigstr, SvCUR(bigstr) + i + 1);
5826 mid = big + offset + len;
5827 midend = bigend = big + SvCUR(bigstr);
5830 while (midend > mid) /* shove everything down */
5831 *--bigend = *--midend;
5832 Move(little,big+offset,littlelen,char);
5833 SvCUR_set(bigstr, SvCUR(bigstr) + i);
5838 Move(little,SvPVX(bigstr)+offset,len,char);
5843 big = SvPVX(bigstr);
5846 bigend = big + SvCUR(bigstr);
5848 if (midend > bigend)
5849 Perl_croak(aTHX_ "panic: sv_insert");
5851 if (mid - big > bigend - midend) { /* faster to shorten from end */
5853 Move(little, mid, littlelen,char);
5856 i = bigend - midend;
5858 Move(midend, mid, i,char);
5862 SvCUR_set(bigstr, mid - big);
5864 else if ((i = mid - big)) { /* faster from front */
5865 midend -= littlelen;
5867 Move(big, midend - i, i, char);
5868 sv_chop(bigstr,midend-i);
5870 Move(little, mid, littlelen,char);
5872 else if (littlelen) {
5873 midend -= littlelen;
5874 sv_chop(bigstr,midend);
5875 Move(little,midend,littlelen,char);
5878 sv_chop(bigstr,midend);
5884 =for apidoc sv_replace
5886 Make the first argument a copy of the second, then delete the original.
5887 The target SV physically takes over ownership of the body of the source SV
5888 and inherits its flags; however, the target keeps any magic it owns,
5889 and any magic in the source is discarded.
5890 Note that this is a rather specialist SV copying operation; most of the
5891 time you'll want to use C<sv_setsv> or one of its many macro front-ends.
5897 Perl_sv_replace(pTHX_ register SV *const sv, register SV *const nsv)
5900 const U32 refcnt = SvREFCNT(sv);
5902 PERL_ARGS_ASSERT_SV_REPLACE;
5904 SV_CHECK_THINKFIRST_COW_DROP(sv);
5905 if (SvREFCNT(nsv) != 1) {
5906 Perl_croak(aTHX_ "panic: reference miscount on nsv in sv_replace()"
5907 " (%" UVuf " != 1)", (UV) SvREFCNT(nsv));
5909 if (SvMAGICAL(sv)) {
5913 sv_upgrade(nsv, SVt_PVMG);
5914 SvMAGIC_set(nsv, SvMAGIC(sv));
5915 SvFLAGS(nsv) |= SvMAGICAL(sv);
5917 SvMAGIC_set(sv, NULL);
5921 assert(!SvREFCNT(sv));
5922 #ifdef DEBUG_LEAKING_SCALARS
5923 sv->sv_flags = nsv->sv_flags;
5924 sv->sv_any = nsv->sv_any;
5925 sv->sv_refcnt = nsv->sv_refcnt;
5926 sv->sv_u = nsv->sv_u;
5928 StructCopy(nsv,sv,SV);
5930 if(SvTYPE(sv) == SVt_IV) {
5932 = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
5936 #ifdef PERL_OLD_COPY_ON_WRITE
5937 if (SvIsCOW_normal(nsv)) {
5938 /* We need to follow the pointers around the loop to make the
5939 previous SV point to sv, rather than nsv. */
5942 while ((next = SV_COW_NEXT_SV(current)) != nsv) {
5945 assert(SvPVX_const(current) == SvPVX_const(nsv));
5947 /* Make the SV before us point to the SV after us. */
5949 PerlIO_printf(Perl_debug_log, "previous is\n");
5951 PerlIO_printf(Perl_debug_log,
5952 "move it from 0x%"UVxf" to 0x%"UVxf"\n",
5953 (UV) SV_COW_NEXT_SV(current), (UV) sv);
5955 SV_COW_NEXT_SV_SET(current, sv);
5958 SvREFCNT(sv) = refcnt;
5959 SvFLAGS(nsv) |= SVTYPEMASK; /* Mark as freed */
5964 /* We're about to free a GV which has a CV that refers back to us.
5965 * If that CV will outlive us, make it anonymous (i.e. fix up its CvGV
5969 S_anonymise_cv_maybe(pTHX_ GV *gv, CV* cv)
5975 PERL_ARGS_ASSERT_ANONYMISE_CV_MAYBE;
5978 assert(SvREFCNT(gv) == 0);
5979 assert(isGV(gv) && isGV_with_GP(gv));
5981 assert(!CvANON(cv));
5982 assert(CvGV(cv) == gv);
5984 /* will the CV shortly be freed by gp_free() ? */
5985 if (GvCV(gv) == cv && GvGP(gv)->gp_refcnt < 2 && SvREFCNT(cv) < 2) {
5986 SvANY(cv)->xcv_gv = NULL;
5990 /* if not, anonymise: */
5991 stash = GvSTASH(gv) && HvNAME(GvSTASH(gv))
5992 ? HvENAME(GvSTASH(gv)) : NULL;
5993 gvname = Perl_newSVpvf(aTHX_ "%s::__ANON__",
5994 stash ? stash : "__ANON__");
5995 anongv = gv_fetchsv(gvname, GV_ADDMULTI, SVt_PVCV);
5996 SvREFCNT_dec(gvname);
6000 SvANY(cv)->xcv_gv = MUTABLE_GV(SvREFCNT_inc(anongv));
6005 =for apidoc sv_clear
6007 Clear an SV: call any destructors, free up any memory used by the body,
6008 and free the body itself. The SV's head is I<not> freed, although
6009 its type is set to all 1's so that it won't inadvertently be assumed
6010 to be live during global destruction etc.
6011 This function should only be called when REFCNT is zero. Most of the time
6012 you'll want to call C<sv_free()> (or its macro wrapper C<SvREFCNT_dec>)
6019 Perl_sv_clear(pTHX_ SV *const orig_sv)
6024 const struct body_details *sv_type_details;
6027 register SV *sv = orig_sv;
6030 PERL_ARGS_ASSERT_SV_CLEAR;
6032 /* within this loop, sv is the SV currently being freed, and
6033 * iter_sv is the most recent AV or whatever that's being iterated
6034 * over to provide more SVs */
6040 assert(SvREFCNT(sv) == 0);
6041 assert(SvTYPE(sv) != SVTYPEMASK);
6043 if (type <= SVt_IV) {
6044 /* See the comment in sv.h about the collusion between this
6045 * early return and the overloading of the NULL slots in the
6049 SvFLAGS(sv) &= SVf_BREAK;
6050 SvFLAGS(sv) |= SVTYPEMASK;
6054 assert(!SvOBJECT(sv) || type >= SVt_PVMG); /* objs are always >= MG */
6056 if (type >= SVt_PVMG) {
6058 if (!curse(sv, 1)) goto get_next_sv;
6059 type = SvTYPE(sv); /* destructor may have changed it */
6061 /* Free back-references before magic, in case the magic calls
6062 * Perl code that has weak references to sv. */
6063 if (type == SVt_PVHV) {
6064 Perl_hv_kill_backrefs(aTHX_ MUTABLE_HV(sv));
6068 else if (type == SVt_PVMG && SvPAD_OUR(sv)) {
6069 SvREFCNT_dec(SvOURSTASH(sv));
6070 } else if (SvMAGIC(sv)) {
6071 /* Free back-references before other types of magic. */
6072 sv_unmagic(sv, PERL_MAGIC_backref);
6075 if (type == SVt_PVMG && SvPAD_TYPED(sv))
6076 SvREFCNT_dec(SvSTASH(sv));
6079 /* case SVt_BIND: */
6082 IoIFP(sv) != PerlIO_stdin() &&
6083 IoIFP(sv) != PerlIO_stdout() &&
6084 IoIFP(sv) != PerlIO_stderr() &&
6085 !(IoFLAGS(sv) & IOf_FAKE_DIRP))
6087 io_close(MUTABLE_IO(sv), FALSE);
6089 if (IoDIRP(sv) && !(IoFLAGS(sv) & IOf_FAKE_DIRP))
6090 PerlDir_close(IoDIRP(sv));
6091 IoDIRP(sv) = (DIR*)NULL;
6092 Safefree(IoTOP_NAME(sv));
6093 Safefree(IoFMT_NAME(sv));
6094 Safefree(IoBOTTOM_NAME(sv));
6097 /* FIXME for plugins */
6098 pregfree2((REGEXP*) sv);
6102 cv_undef(MUTABLE_CV(sv));
6103 /* If we're in a stash, we don't own a reference to it.
6104 * However it does have a back reference to us, which needs to
6106 if ((stash = CvSTASH(sv)))
6107 sv_del_backref(MUTABLE_SV(stash), sv);
6110 if (PL_last_swash_hv == (const HV *)sv) {
6111 PL_last_swash_hv = NULL;
6113 if (HvTOTALKEYS((HV*)sv) > 0) {
6115 /* this statement should match the one at the beginning of
6116 * hv_undef_flags() */
6117 if ( PL_phase != PERL_PHASE_DESTRUCT
6118 && (name = HvNAME((HV*)sv)))
6121 (void)hv_delete(PL_stashcache, name,
6122 HvNAMELEN_get((HV*)sv), G_DISCARD);
6123 hv_name_set((HV*)sv, NULL, 0, 0);
6126 /* save old iter_sv in unused SvSTASH field */
6127 assert(!SvOBJECT(sv));
6128 SvSTASH(sv) = (HV*)iter_sv;
6131 /* XXX ideally we should save the old value of hash_index
6132 * too, but I can't think of any place to hide it. The
6133 * effect of not saving it is that for freeing hashes of
6134 * hashes, we become quadratic in scanning the HvARRAY of
6135 * the top hash looking for new entries to free; but
6136 * hopefully this will be dwarfed by the freeing of all
6137 * the nested hashes. */
6139 next_sv = Perl_hfree_next_entry(aTHX_ (HV*)sv, &hash_index);
6140 goto get_next_sv; /* process this new sv */
6142 /* free empty hash */
6143 Perl_hv_undef_flags(aTHX_ MUTABLE_HV(sv), HV_NAME_SETALL);
6144 assert(!HvARRAY((HV*)sv));
6148 AV* av = MUTABLE_AV(sv);
6149 if (PL_comppad == av) {
6153 if (AvREAL(av) && AvFILLp(av) > -1) {
6154 next_sv = AvARRAY(av)[AvFILLp(av)--];
6155 /* save old iter_sv in top-most slot of AV,
6156 * and pray that it doesn't get wiped in the meantime */
6157 AvARRAY(av)[AvMAX(av)] = iter_sv;
6159 goto get_next_sv; /* process this new sv */
6161 Safefree(AvALLOC(av));
6166 if (LvTYPE(sv) == 'T') { /* for tie: return HE to pool */
6167 SvREFCNT_dec(HeKEY_sv((HE*)LvTARG(sv)));
6168 HeNEXT((HE*)LvTARG(sv)) = PL_hv_fetch_ent_mh;
6169 PL_hv_fetch_ent_mh = (HE*)LvTARG(sv);
6171 else if (LvTYPE(sv) != 't') /* unless tie: unrefcnted fake SV** */
6172 SvREFCNT_dec(LvTARG(sv));
6174 if (isGV_with_GP(sv)) {
6175 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
6176 && HvENAME_get(stash))
6177 mro_method_changed_in(stash);
6178 gp_free(MUTABLE_GV(sv));
6180 unshare_hek(GvNAME_HEK(sv));
6181 /* If we're in a stash, we don't own a reference to it.
6182 * However it does have a back reference to us, which
6183 * needs to be cleared. */
6184 if (!SvVALID(sv) && (stash = GvSTASH(sv)))
6185 sv_del_backref(MUTABLE_SV(stash), sv);
6187 /* FIXME. There are probably more unreferenced pointers to SVs
6188 * in the interpreter struct that we should check and tidy in
6189 * a similar fashion to this: */
6190 if ((const GV *)sv == PL_last_in_gv)
6191 PL_last_in_gv = NULL;
6197 /* Don't bother with SvOOK_off(sv); as we're only going to
6201 SvOOK_offset(sv, offset);
6202 SvPV_set(sv, SvPVX_mutable(sv) - offset);
6203 /* Don't even bother with turning off the OOK flag. */
6208 SV * const target = SvRV(sv);
6210 sv_del_backref(target, sv);
6215 #ifdef PERL_OLD_COPY_ON_WRITE
6216 else if (SvPVX_const(sv)
6217 && !(SvTYPE(sv) == SVt_PVIO
6218 && !(IoFLAGS(sv) & IOf_FAKE_DIRP)))
6222 PerlIO_printf(Perl_debug_log, "Copy on write: clear\n");
6226 sv_release_COW(sv, SvPVX_const(sv), SV_COW_NEXT_SV(sv));
6228 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
6232 } else if (SvLEN(sv)) {
6233 Safefree(SvPVX_const(sv));
6237 else if (SvPVX_const(sv) && SvLEN(sv)
6238 && !(SvTYPE(sv) == SVt_PVIO
6239 && !(IoFLAGS(sv) & IOf_FAKE_DIRP)))
6240 Safefree(SvPVX_mutable(sv));
6241 else if (SvPVX_const(sv) && SvREADONLY(sv) && SvFAKE(sv)) {
6242 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
6253 SvFLAGS(sv) &= SVf_BREAK;
6254 SvFLAGS(sv) |= SVTYPEMASK;
6256 sv_type_details = bodies_by_type + type;
6257 if (sv_type_details->arena) {
6258 del_body(((char *)SvANY(sv) + sv_type_details->offset),
6259 &PL_body_roots[type]);
6261 else if (sv_type_details->body_size) {
6262 safefree(SvANY(sv));
6266 /* caller is responsible for freeing the head of the original sv */
6267 if (sv != orig_sv && !SvREFCNT(sv))
6270 /* grab and free next sv, if any */
6278 else if (!iter_sv) {
6280 } else if (SvTYPE(iter_sv) == SVt_PVAV) {
6281 AV *const av = (AV*)iter_sv;
6282 if (AvFILLp(av) > -1) {
6283 sv = AvARRAY(av)[AvFILLp(av)--];
6285 else { /* no more elements of current AV to free */
6288 /* restore previous value, squirrelled away */
6289 iter_sv = AvARRAY(av)[AvMAX(av)];
6290 Safefree(AvALLOC(av));
6293 } else if (SvTYPE(iter_sv) == SVt_PVHV) {
6294 sv = Perl_hfree_next_entry(aTHX_ (HV*)iter_sv, &hash_index);
6295 if (!sv) { /* no more elements of current HV to free */
6298 /* Restore previous value of iter_sv, squirrelled away */
6299 assert(!SvOBJECT(sv));
6300 iter_sv = (SV*)SvSTASH(sv);
6302 /* ideally we should restore the old hash_index here,
6303 * but we don't currently save the old value */
6306 /* free any remaining detritus from the hash struct */
6307 Perl_hv_undef_flags(aTHX_ MUTABLE_HV(sv), HV_NAME_SETALL);
6308 assert(!HvARRAY((HV*)sv));
6313 /* unrolled SvREFCNT_dec and sv_free2 follows: */
6317 if (!SvREFCNT(sv)) {
6321 if (--(SvREFCNT(sv)))
6325 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEBUGGING),
6326 "Attempt to free temp prematurely: SV 0x%"UVxf
6327 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
6331 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
6332 /* make sure SvREFCNT(sv)==0 happens very seldom */
6333 SvREFCNT(sv) = (~(U32)0)/2;
6342 /* This routine curses the sv itself, not the object referenced by sv. So
6343 sv does not have to be ROK. */
6346 S_curse(pTHX_ SV * const sv, const bool check_refcnt) {
6349 PERL_ARGS_ASSERT_CURSE;
6350 assert(SvOBJECT(sv));
6352 if (PL_defstash && /* Still have a symbol table? */
6359 stash = SvSTASH(sv);
6360 destructor = StashHANDLER(stash,DESTROY);
6362 /* A constant subroutine can have no side effects, so
6363 don't bother calling it. */
6364 && !CvCONST(destructor)
6365 /* Don't bother calling an empty destructor */
6366 && (CvISXSUB(destructor)
6367 || (CvSTART(destructor)
6368 && (CvSTART(destructor)->op_next->op_type
6371 SV* const tmpref = newRV(sv);
6372 SvREADONLY_on(tmpref); /* DESTROY() could be naughty */
6374 PUSHSTACKi(PERLSI_DESTROY);
6379 call_sv(MUTABLE_SV(destructor),
6380 G_DISCARD|G_EVAL|G_KEEPERR|G_VOID);
6384 if(SvREFCNT(tmpref) < 2) {
6385 /* tmpref is not kept alive! */
6387 SvRV_set(tmpref, NULL);
6390 SvREFCNT_dec(tmpref);
6392 } while (SvOBJECT(sv) && SvSTASH(sv) != stash);
6395 if (check_refcnt && SvREFCNT(sv)) {
6396 if (PL_in_clean_objs)
6398 "DESTROY created new reference to dead object '%s'",
6400 /* DESTROY gave object new lease on life */
6406 SvREFCNT_dec(SvSTASH(sv)); /* possibly of changed persuasion */
6407 SvOBJECT_off(sv); /* Curse the object. */
6408 if (SvTYPE(sv) != SVt_PVIO)
6409 --PL_sv_objcount;/* XXX Might want something more general */
6415 =for apidoc sv_newref
6417 Increment an SV's reference count. Use the C<SvREFCNT_inc()> wrapper
6424 Perl_sv_newref(pTHX_ SV *const sv)
6426 PERL_UNUSED_CONTEXT;
6435 Decrement an SV's reference count, and if it drops to zero, call
6436 C<sv_clear> to invoke destructors and free up any memory used by
6437 the body; finally, deallocate the SV's head itself.
6438 Normally called via a wrapper macro C<SvREFCNT_dec>.
6444 Perl_sv_free(pTHX_ SV *const sv)
6449 if (SvREFCNT(sv) == 0) {
6450 if (SvFLAGS(sv) & SVf_BREAK)
6451 /* this SV's refcnt has been artificially decremented to
6452 * trigger cleanup */
6454 if (PL_in_clean_all) /* All is fair */
6456 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
6457 /* make sure SvREFCNT(sv)==0 happens very seldom */
6458 SvREFCNT(sv) = (~(U32)0)/2;
6461 if (ckWARN_d(WARN_INTERNAL)) {
6462 #ifdef DEBUG_LEAKING_SCALARS_FORK_DUMP
6463 Perl_dump_sv_child(aTHX_ sv);
6465 #ifdef DEBUG_LEAKING_SCALARS
6468 #ifdef DEBUG_LEAKING_SCALARS_ABORT
6469 if (PL_warnhook == PERL_WARNHOOK_FATAL
6470 || ckDEAD(packWARN(WARN_INTERNAL))) {
6471 /* Don't let Perl_warner cause us to escape our fate: */
6475 /* This may not return: */
6476 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
6477 "Attempt to free unreferenced scalar: SV 0x%"UVxf
6478 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
6481 #ifdef DEBUG_LEAKING_SCALARS_ABORT
6486 if (--(SvREFCNT(sv)) > 0)
6488 Perl_sv_free2(aTHX_ sv);
6492 Perl_sv_free2(pTHX_ SV *const sv)
6496 PERL_ARGS_ASSERT_SV_FREE2;
6500 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEBUGGING),
6501 "Attempt to free temp prematurely: SV 0x%"UVxf
6502 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
6506 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
6507 /* make sure SvREFCNT(sv)==0 happens very seldom */
6508 SvREFCNT(sv) = (~(U32)0)/2;
6519 Returns the length of the string in the SV. Handles magic and type
6520 coercion. See also C<SvCUR>, which gives raw access to the xpv_cur slot.
6526 Perl_sv_len(pTHX_ register SV *const sv)
6534 len = mg_length(sv);
6536 (void)SvPV_const(sv, len);
6541 =for apidoc sv_len_utf8
6543 Returns the number of characters in the string in an SV, counting wide
6544 UTF-8 bytes as a single character. Handles magic and type coercion.
6550 * The length is cached in PERL_MAGIC_utf8, in the mg_len field. Also the
6551 * mg_ptr is used, by sv_pos_u2b() and sv_pos_b2u() - see the comments below.
6552 * (Note that the mg_len is not the length of the mg_ptr field.
6553 * This allows the cache to store the character length of the string without
6554 * needing to malloc() extra storage to attach to the mg_ptr.)
6559 Perl_sv_len_utf8(pTHX_ register SV *const sv)
6565 return mg_length(sv);
6569 const U8 *s = (U8*)SvPV_const(sv, len);
6573 MAGIC *mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_utf8) : NULL;
6575 if (mg && (mg->mg_len != -1 || mg->mg_ptr)) {
6576 if (mg->mg_len != -1)
6579 /* We can use the offset cache for a headstart.
6580 The longer value is stored in the first pair. */
6581 STRLEN *cache = (STRLEN *) mg->mg_ptr;
6583 ulen = cache[0] + Perl_utf8_length(aTHX_ s + cache[1],
6587 if (PL_utf8cache < 0) {
6588 const STRLEN real = Perl_utf8_length(aTHX_ s, s + len);
6589 assert_uft8_cache_coherent("sv_len_utf8", ulen, real, sv);
6593 ulen = Perl_utf8_length(aTHX_ s, s + len);
6594 utf8_mg_len_cache_update(sv, &mg, ulen);
6598 return Perl_utf8_length(aTHX_ s, s + len);
6602 /* Walk forwards to find the byte corresponding to the passed in UTF-8
6605 S_sv_pos_u2b_forwards(const U8 *const start, const U8 *const send,
6606 STRLEN *const uoffset_p, bool *const at_end)
6608 const U8 *s = start;
6609 STRLEN uoffset = *uoffset_p;
6611 PERL_ARGS_ASSERT_SV_POS_U2B_FORWARDS;
6613 while (s < send && uoffset) {
6620 else if (s > send) {
6622 /* This is the existing behaviour. Possibly it should be a croak, as
6623 it's actually a bounds error */
6626 *uoffset_p -= uoffset;
6630 /* Given the length of the string in both bytes and UTF-8 characters, decide
6631 whether to walk forwards or backwards to find the byte corresponding to
6632 the passed in UTF-8 offset. */
6634 S_sv_pos_u2b_midway(const U8 *const start, const U8 *send,
6635 STRLEN uoffset, const STRLEN uend)
6637 STRLEN backw = uend - uoffset;
6639 PERL_ARGS_ASSERT_SV_POS_U2B_MIDWAY;
6641 if (uoffset < 2 * backw) {
6642 /* The assumption is that going forwards is twice the speed of going
6643 forward (that's where the 2 * backw comes from).
6644 (The real figure of course depends on the UTF-8 data.) */
6645 const U8 *s = start;
6647 while (s < send && uoffset--)
6657 while (UTF8_IS_CONTINUATION(*send))
6660 return send - start;
6663 /* For the string representation of the given scalar, find the byte
6664 corresponding to the passed in UTF-8 offset. uoffset0 and boffset0
6665 give another position in the string, *before* the sought offset, which
6666 (which is always true, as 0, 0 is a valid pair of positions), which should
6667 help reduce the amount of linear searching.
6668 If *mgp is non-NULL, it should point to the UTF-8 cache magic, which
6669 will be used to reduce the amount of linear searching. The cache will be
6670 created if necessary, and the found value offered to it for update. */
6672 S_sv_pos_u2b_cached(pTHX_ SV *const sv, MAGIC **const mgp, const U8 *const start,
6673 const U8 *const send, STRLEN uoffset,
6674 STRLEN uoffset0, STRLEN boffset0)
6676 STRLEN boffset = 0; /* Actually always set, but let's keep gcc happy. */
6678 bool at_end = FALSE;
6680 PERL_ARGS_ASSERT_SV_POS_U2B_CACHED;
6682 assert (uoffset >= uoffset0);
6689 && (*mgp || (SvTYPE(sv) >= SVt_PVMG &&
6690 (*mgp = mg_find(sv, PERL_MAGIC_utf8))))) {
6691 if ((*mgp)->mg_ptr) {
6692 STRLEN *cache = (STRLEN *) (*mgp)->mg_ptr;
6693 if (cache[0] == uoffset) {
6694 /* An exact match. */
6697 if (cache[2] == uoffset) {
6698 /* An exact match. */
6702 if (cache[0] < uoffset) {
6703 /* The cache already knows part of the way. */
6704 if (cache[0] > uoffset0) {
6705 /* The cache knows more than the passed in pair */
6706 uoffset0 = cache[0];
6707 boffset0 = cache[1];
6709 if ((*mgp)->mg_len != -1) {
6710 /* And we know the end too. */
6712 + sv_pos_u2b_midway(start + boffset0, send,
6714 (*mgp)->mg_len - uoffset0);
6716 uoffset -= uoffset0;
6718 + sv_pos_u2b_forwards(start + boffset0,
6719 send, &uoffset, &at_end);
6720 uoffset += uoffset0;
6723 else if (cache[2] < uoffset) {
6724 /* We're between the two cache entries. */
6725 if (cache[2] > uoffset0) {
6726 /* and the cache knows more than the passed in pair */
6727 uoffset0 = cache[2];
6728 boffset0 = cache[3];
6732 + sv_pos_u2b_midway(start + boffset0,
6735 cache[0] - uoffset0);
6738 + sv_pos_u2b_midway(start + boffset0,
6741 cache[2] - uoffset0);
6745 else if ((*mgp)->mg_len != -1) {
6746 /* If we can take advantage of a passed in offset, do so. */
6747 /* In fact, offset0 is either 0, or less than offset, so don't
6748 need to worry about the other possibility. */
6750 + sv_pos_u2b_midway(start + boffset0, send,
6752 (*mgp)->mg_len - uoffset0);
6757 if (!found || PL_utf8cache < 0) {
6758 STRLEN real_boffset;
6759 uoffset -= uoffset0;
6760 real_boffset = boffset0 + sv_pos_u2b_forwards(start + boffset0,
6761 send, &uoffset, &at_end);
6762 uoffset += uoffset0;
6764 if (found && PL_utf8cache < 0)
6765 assert_uft8_cache_coherent("sv_pos_u2b_cache", boffset,
6767 boffset = real_boffset;
6772 utf8_mg_len_cache_update(sv, mgp, uoffset);
6774 utf8_mg_pos_cache_update(sv, mgp, boffset, uoffset, send - start);
6781 =for apidoc sv_pos_u2b_flags
6783 Converts the value pointed to by offsetp from a count of UTF-8 chars from
6784 the start of the string, to a count of the equivalent number of bytes; if
6785 lenp is non-zero, it does the same to lenp, but this time starting from
6786 the offset, rather than from the start of the string. Handles type coercion.
6787 I<flags> is passed to C<SvPV_flags>, and usually should be
6788 C<SV_GMAGIC|SV_CONST_RETURN> to handle magic.
6794 * sv_pos_u2b_flags() uses, like sv_pos_b2u(), the mg_ptr of the potential
6795 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6796 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
6801 Perl_sv_pos_u2b_flags(pTHX_ SV *const sv, STRLEN uoffset, STRLEN *const lenp,
6808 PERL_ARGS_ASSERT_SV_POS_U2B_FLAGS;
6810 start = (U8*)SvPV_flags(sv, len, flags);
6812 const U8 * const send = start + len;
6814 boffset = sv_pos_u2b_cached(sv, &mg, start, send, uoffset, 0, 0);
6817 && *lenp /* don't bother doing work for 0, as its bytes equivalent
6818 is 0, and *lenp is already set to that. */) {
6819 /* Convert the relative offset to absolute. */
6820 const STRLEN uoffset2 = uoffset + *lenp;
6821 const STRLEN boffset2
6822 = sv_pos_u2b_cached(sv, &mg, start, send, uoffset2,
6823 uoffset, boffset) - boffset;
6837 =for apidoc sv_pos_u2b
6839 Converts the value pointed to by offsetp from a count of UTF-8 chars from
6840 the start of the string, to a count of the equivalent number of bytes; if
6841 lenp is non-zero, it does the same to lenp, but this time starting from
6842 the offset, rather than from the start of the string. Handles magic and
6845 Use C<sv_pos_u2b_flags> in preference, which correctly handles strings longer
6852 * sv_pos_u2b() uses, like sv_pos_b2u(), the mg_ptr of the potential
6853 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6854 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
6858 /* This function is subject to size and sign problems */
6861 Perl_sv_pos_u2b(pTHX_ register SV *const sv, I32 *const offsetp, I32 *const lenp)
6863 PERL_ARGS_ASSERT_SV_POS_U2B;
6866 STRLEN ulen = (STRLEN)*lenp;
6867 *offsetp = (I32)sv_pos_u2b_flags(sv, (STRLEN)*offsetp, &ulen,
6868 SV_GMAGIC|SV_CONST_RETURN);
6871 *offsetp = (I32)sv_pos_u2b_flags(sv, (STRLEN)*offsetp, NULL,
6872 SV_GMAGIC|SV_CONST_RETURN);
6877 S_utf8_mg_len_cache_update(pTHX_ SV *const sv, MAGIC **const mgp,
6880 PERL_ARGS_ASSERT_UTF8_MG_LEN_CACHE_UPDATE;
6884 if (!*mgp && (SvTYPE(sv) < SVt_PVMG ||
6885 !(*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
6886 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, &PL_vtbl_utf8, 0, 0);
6890 (*mgp)->mg_len = ulen;
6891 /* For now, treat "overflowed" as "still unknown". See RT #72924. */
6892 if (ulen != (STRLEN) (*mgp)->mg_len)
6893 (*mgp)->mg_len = -1;
6896 /* Create and update the UTF8 magic offset cache, with the proffered utf8/
6897 byte length pairing. The (byte) length of the total SV is passed in too,
6898 as blen, because for some (more esoteric) SVs, the call to SvPV_const()
6899 may not have updated SvCUR, so we can't rely on reading it directly.
6901 The proffered utf8/byte length pairing isn't used if the cache already has
6902 two pairs, and swapping either for the proffered pair would increase the
6903 RMS of the intervals between known byte offsets.
6905 The cache itself consists of 4 STRLEN values
6906 0: larger UTF-8 offset
6907 1: corresponding byte offset
6908 2: smaller UTF-8 offset
6909 3: corresponding byte offset
6911 Unused cache pairs have the value 0, 0.
6912 Keeping the cache "backwards" means that the invariant of
6913 cache[0] >= cache[2] is maintained even with empty slots, which means that
6914 the code that uses it doesn't need to worry if only 1 entry has actually
6915 been set to non-zero. It also makes the "position beyond the end of the
6916 cache" logic much simpler, as the first slot is always the one to start
6920 S_utf8_mg_pos_cache_update(pTHX_ SV *const sv, MAGIC **const mgp, const STRLEN byte,
6921 const STRLEN utf8, const STRLEN blen)
6925 PERL_ARGS_ASSERT_UTF8_MG_POS_CACHE_UPDATE;
6930 if (!*mgp && (SvTYPE(sv) < SVt_PVMG ||
6931 !(*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
6932 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, (MGVTBL*)&PL_vtbl_utf8, 0,
6934 (*mgp)->mg_len = -1;
6938 if (!(cache = (STRLEN *)(*mgp)->mg_ptr)) {
6939 Newxz(cache, PERL_MAGIC_UTF8_CACHESIZE * 2, STRLEN);
6940 (*mgp)->mg_ptr = (char *) cache;
6944 if (PL_utf8cache < 0 && SvPOKp(sv)) {
6945 /* SvPOKp() because it's possible that sv has string overloading, and
6946 therefore is a reference, hence SvPVX() is actually a pointer.
6947 This cures the (very real) symptoms of RT 69422, but I'm not actually
6948 sure whether we should even be caching the results of UTF-8
6949 operations on overloading, given that nothing stops overloading
6950 returning a different value every time it's called. */
6951 const U8 *start = (const U8 *) SvPVX_const(sv);
6952 const STRLEN realutf8 = utf8_length(start, start + byte);
6954 assert_uft8_cache_coherent("utf8_mg_pos_cache_update", utf8, realutf8,
6958 /* Cache is held with the later position first, to simplify the code
6959 that deals with unbounded ends. */
6961 ASSERT_UTF8_CACHE(cache);
6962 if (cache[1] == 0) {
6963 /* Cache is totally empty */
6966 } else if (cache[3] == 0) {
6967 if (byte > cache[1]) {
6968 /* New one is larger, so goes first. */
6969 cache[2] = cache[0];
6970 cache[3] = cache[1];
6978 #define THREEWAY_SQUARE(a,b,c,d) \
6979 ((float)((d) - (c))) * ((float)((d) - (c))) \
6980 + ((float)((c) - (b))) * ((float)((c) - (b))) \
6981 + ((float)((b) - (a))) * ((float)((b) - (a)))
6983 /* Cache has 2 slots in use, and we know three potential pairs.
6984 Keep the two that give the lowest RMS distance. Do the
6985 calculation in bytes simply because we always know the byte
6986 length. squareroot has the same ordering as the positive value,
6987 so don't bother with the actual square root. */
6988 const float existing = THREEWAY_SQUARE(0, cache[3], cache[1], blen);
6989 if (byte > cache[1]) {
6990 /* New position is after the existing pair of pairs. */
6991 const float keep_earlier
6992 = THREEWAY_SQUARE(0, cache[3], byte, blen);
6993 const float keep_later
6994 = THREEWAY_SQUARE(0, cache[1], byte, blen);
6996 if (keep_later < keep_earlier) {
6997 if (keep_later < existing) {
6998 cache[2] = cache[0];
6999 cache[3] = cache[1];
7005 if (keep_earlier < existing) {
7011 else if (byte > cache[3]) {
7012 /* New position is between the existing pair of pairs. */
7013 const float keep_earlier
7014 = THREEWAY_SQUARE(0, cache[3], byte, blen);
7015 const float keep_later
7016 = THREEWAY_SQUARE(0, byte, cache[1], blen);
7018 if (keep_later < keep_earlier) {
7019 if (keep_later < existing) {
7025 if (keep_earlier < existing) {
7032 /* New position is before the existing pair of pairs. */
7033 const float keep_earlier
7034 = THREEWAY_SQUARE(0, byte, cache[3], blen);
7035 const float keep_later
7036 = THREEWAY_SQUARE(0, byte, cache[1], blen);
7038 if (keep_later < keep_earlier) {
7039 if (keep_later < existing) {
7045 if (keep_earlier < existing) {
7046 cache[0] = cache[2];
7047 cache[1] = cache[3];
7054 ASSERT_UTF8_CACHE(cache);
7057 /* We already know all of the way, now we may be able to walk back. The same
7058 assumption is made as in S_sv_pos_u2b_midway(), namely that walking
7059 backward is half the speed of walking forward. */
7061 S_sv_pos_b2u_midway(pTHX_ const U8 *const s, const U8 *const target,
7062 const U8 *end, STRLEN endu)
7064 const STRLEN forw = target - s;
7065 STRLEN backw = end - target;
7067 PERL_ARGS_ASSERT_SV_POS_B2U_MIDWAY;
7069 if (forw < 2 * backw) {
7070 return utf8_length(s, target);
7073 while (end > target) {
7075 while (UTF8_IS_CONTINUATION(*end)) {
7084 =for apidoc sv_pos_b2u
7086 Converts the value pointed to by offsetp from a count of bytes from the
7087 start of the string, to a count of the equivalent number of UTF-8 chars.
7088 Handles magic and type coercion.
7094 * sv_pos_b2u() uses, like sv_pos_u2b(), the mg_ptr of the potential
7095 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
7100 Perl_sv_pos_b2u(pTHX_ register SV *const sv, I32 *const offsetp)
7103 const STRLEN byte = *offsetp;
7104 STRLEN len = 0; /* Actually always set, but let's keep gcc happy. */
7110 PERL_ARGS_ASSERT_SV_POS_B2U;
7115 s = (const U8*)SvPV_const(sv, blen);
7118 Perl_croak(aTHX_ "panic: sv_pos_b2u: bad byte offset");
7124 && SvTYPE(sv) >= SVt_PVMG
7125 && (mg = mg_find(sv, PERL_MAGIC_utf8)))
7128 STRLEN * const cache = (STRLEN *) mg->mg_ptr;
7129 if (cache[1] == byte) {
7130 /* An exact match. */
7131 *offsetp = cache[0];
7134 if (cache[3] == byte) {
7135 /* An exact match. */
7136 *offsetp = cache[2];
7140 if (cache[1] < byte) {
7141 /* We already know part of the way. */
7142 if (mg->mg_len != -1) {
7143 /* Actually, we know the end too. */
7145 + S_sv_pos_b2u_midway(aTHX_ s + cache[1], send,
7146 s + blen, mg->mg_len - cache[0]);
7148 len = cache[0] + utf8_length(s + cache[1], send);
7151 else if (cache[3] < byte) {
7152 /* We're between the two cached pairs, so we do the calculation
7153 offset by the byte/utf-8 positions for the earlier pair,
7154 then add the utf-8 characters from the string start to
7156 len = S_sv_pos_b2u_midway(aTHX_ s + cache[3], send,
7157 s + cache[1], cache[0] - cache[2])
7161 else { /* cache[3] > byte */
7162 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + cache[3],
7166 ASSERT_UTF8_CACHE(cache);
7168 } else if (mg->mg_len != -1) {
7169 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + blen, mg->mg_len);
7173 if (!found || PL_utf8cache < 0) {
7174 const STRLEN real_len = utf8_length(s, send);
7176 if (found && PL_utf8cache < 0)
7177 assert_uft8_cache_coherent("sv_pos_b2u", len, real_len, sv);
7184 utf8_mg_len_cache_update(sv, &mg, len);
7186 utf8_mg_pos_cache_update(sv, &mg, byte, len, blen);
7191 S_assert_uft8_cache_coherent(pTHX_ const char *const func, STRLEN from_cache,
7192 STRLEN real, SV *const sv)
7194 PERL_ARGS_ASSERT_ASSERT_UFT8_CACHE_COHERENT;
7196 /* As this is debugging only code, save space by keeping this test here,
7197 rather than inlining it in all the callers. */
7198 if (from_cache == real)
7201 /* Need to turn the assertions off otherwise we may recurse infinitely
7202 while printing error messages. */
7203 SAVEI8(PL_utf8cache);
7205 Perl_croak(aTHX_ "panic: %s cache %"UVuf" real %"UVuf" for %"SVf,
7206 func, (UV) from_cache, (UV) real, SVfARG(sv));
7212 Returns a boolean indicating whether the strings in the two SVs are
7213 identical. Is UTF-8 and 'use bytes' aware, handles get magic, and will
7214 coerce its args to strings if necessary.
7216 =for apidoc sv_eq_flags
7218 Returns a boolean indicating whether the strings in the two SVs are
7219 identical. Is UTF-8 and 'use bytes' aware and coerces its args to strings
7220 if necessary. If the flags include SV_GMAGIC, it handles get-magic, too.
7226 Perl_sv_eq_flags(pTHX_ register SV *sv1, register SV *sv2, const U32 flags)
7235 SV* svrecode = NULL;
7242 /* if pv1 and pv2 are the same, second SvPV_const call may
7243 * invalidate pv1 (if we are handling magic), so we may need to
7245 if (sv1 == sv2 && flags & SV_GMAGIC
7246 && (SvTHINKFIRST(sv1) || SvGMAGICAL(sv1))) {
7247 pv1 = SvPV_const(sv1, cur1);
7248 sv1 = newSVpvn_flags(pv1, cur1, SVs_TEMP | SvUTF8(sv2));
7250 pv1 = SvPV_flags_const(sv1, cur1, flags);
7258 pv2 = SvPV_flags_const(sv2, cur2, flags);
7260 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
7261 /* Differing utf8ness.
7262 * Do not UTF8size the comparands as a side-effect. */
7265 svrecode = newSVpvn(pv2, cur2);
7266 sv_recode_to_utf8(svrecode, PL_encoding);
7267 pv2 = SvPV_const(svrecode, cur2);
7270 svrecode = newSVpvn(pv1, cur1);
7271 sv_recode_to_utf8(svrecode, PL_encoding);
7272 pv1 = SvPV_const(svrecode, cur1);
7274 /* Now both are in UTF-8. */
7276 SvREFCNT_dec(svrecode);
7282 /* sv1 is the UTF-8 one */
7283 return bytes_cmp_utf8((const U8*)pv2, cur2,
7284 (const U8*)pv1, cur1) == 0;
7287 /* sv2 is the UTF-8 one */
7288 return bytes_cmp_utf8((const U8*)pv1, cur1,
7289 (const U8*)pv2, cur2) == 0;
7295 eq = (pv1 == pv2) || memEQ(pv1, pv2, cur1);
7297 SvREFCNT_dec(svrecode);
7307 Compares the strings in two SVs. Returns -1, 0, or 1 indicating whether the
7308 string in C<sv1> is less than, equal to, or greater than the string in
7309 C<sv2>. Is UTF-8 and 'use bytes' aware, handles get magic, and will
7310 coerce its args to strings if necessary. See also C<sv_cmp_locale>.
7312 =for apidoc sv_cmp_flags
7314 Compares the strings in two SVs. Returns -1, 0, or 1 indicating whether the
7315 string in C<sv1> is less than, equal to, or greater than the string in
7316 C<sv2>. Is UTF-8 and 'use bytes' aware and will coerce its args to strings
7317 if necessary. If the flags include SV_GMAGIC, it handles get magic. See
7318 also C<sv_cmp_locale_flags>.
7324 Perl_sv_cmp(pTHX_ register SV *const sv1, register SV *const sv2)
7326 return sv_cmp_flags(sv1, sv2, SV_GMAGIC);
7330 Perl_sv_cmp_flags(pTHX_ register SV *const sv1, register SV *const sv2,
7335 const char *pv1, *pv2;
7338 SV *svrecode = NULL;
7345 pv1 = SvPV_flags_const(sv1, cur1, flags);
7352 pv2 = SvPV_flags_const(sv2, cur2, flags);
7354 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
7355 /* Differing utf8ness.
7356 * Do not UTF8size the comparands as a side-effect. */
7359 svrecode = newSVpvn(pv2, cur2);
7360 sv_recode_to_utf8(svrecode, PL_encoding);
7361 pv2 = SvPV_const(svrecode, cur2);
7364 const int retval = -bytes_cmp_utf8((const U8*)pv2, cur2,
7365 (const U8*)pv1, cur1);
7366 return retval ? retval < 0 ? -1 : +1 : 0;
7371 svrecode = newSVpvn(pv1, cur1);
7372 sv_recode_to_utf8(svrecode, PL_encoding);
7373 pv1 = SvPV_const(svrecode, cur1);
7376 const int retval = bytes_cmp_utf8((const U8*)pv1, cur1,
7377 (const U8*)pv2, cur2);
7378 return retval ? retval < 0 ? -1 : +1 : 0;
7384 cmp = cur2 ? -1 : 0;
7388 const I32 retval = memcmp((const void*)pv1, (const void*)pv2, cur1 < cur2 ? cur1 : cur2);
7391 cmp = retval < 0 ? -1 : 1;
7392 } else if (cur1 == cur2) {
7395 cmp = cur1 < cur2 ? -1 : 1;
7399 SvREFCNT_dec(svrecode);
7407 =for apidoc sv_cmp_locale
7409 Compares the strings in two SVs in a locale-aware manner. Is UTF-8 and
7410 'use bytes' aware, handles get magic, and will coerce its args to strings
7411 if necessary. See also C<sv_cmp>.
7413 =for apidoc sv_cmp_locale_flags
7415 Compares the strings in two SVs in a locale-aware manner. Is UTF-8 and
7416 'use bytes' aware and will coerce its args to strings if necessary. If the
7417 flags contain SV_GMAGIC, it handles get magic. See also C<sv_cmp_flags>.
7423 Perl_sv_cmp_locale(pTHX_ register SV *const sv1, register SV *const sv2)
7425 return sv_cmp_locale_flags(sv1, sv2, SV_GMAGIC);
7429 Perl_sv_cmp_locale_flags(pTHX_ register SV *const sv1, register SV *const sv2,
7433 #ifdef USE_LOCALE_COLLATE
7439 if (PL_collation_standard)
7443 pv1 = sv1 ? sv_collxfrm_flags(sv1, &len1, flags) : (char *) NULL;
7445 pv2 = sv2 ? sv_collxfrm_flags(sv2, &len2, flags) : (char *) NULL;
7447 if (!pv1 || !len1) {
7458 retval = memcmp((void*)pv1, (void*)pv2, len1 < len2 ? len1 : len2);
7461 return retval < 0 ? -1 : 1;
7464 * When the result of collation is equality, that doesn't mean
7465 * that there are no differences -- some locales exclude some
7466 * characters from consideration. So to avoid false equalities,
7467 * we use the raw string as a tiebreaker.
7473 #endif /* USE_LOCALE_COLLATE */
7475 return sv_cmp(sv1, sv2);
7479 #ifdef USE_LOCALE_COLLATE
7482 =for apidoc sv_collxfrm
7484 This calls C<sv_collxfrm_flags> with the SV_GMAGIC flag. See
7485 C<sv_collxfrm_flags>.
7487 =for apidoc sv_collxfrm_flags
7489 Add Collate Transform magic to an SV if it doesn't already have it. If the
7490 flags contain SV_GMAGIC, it handles get-magic.
7492 Any scalar variable may carry PERL_MAGIC_collxfrm magic that contains the
7493 scalar data of the variable, but transformed to such a format that a normal
7494 memory comparison can be used to compare the data according to the locale
7501 Perl_sv_collxfrm_flags(pTHX_ SV *const sv, STRLEN *const nxp, const I32 flags)
7506 PERL_ARGS_ASSERT_SV_COLLXFRM_FLAGS;
7508 mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_collxfrm) : (MAGIC *) NULL;
7509 if (!mg || !mg->mg_ptr || *(U32*)mg->mg_ptr != PL_collation_ix) {
7515 Safefree(mg->mg_ptr);
7516 s = SvPV_flags_const(sv, len, flags);
7517 if ((xf = mem_collxfrm(s, len, &xlen))) {
7519 #ifdef PERL_OLD_COPY_ON_WRITE
7521 sv_force_normal_flags(sv, 0);
7523 mg = sv_magicext(sv, 0, PERL_MAGIC_collxfrm, &PL_vtbl_collxfrm,
7537 if (mg && mg->mg_ptr) {
7539 return mg->mg_ptr + sizeof(PL_collation_ix);
7547 #endif /* USE_LOCALE_COLLATE */
7550 S_sv_gets_append_to_utf8(pTHX_ SV *const sv, PerlIO *const fp, I32 append)
7552 SV * const tsv = newSV(0);
7555 sv_gets(tsv, fp, 0);
7556 sv_utf8_upgrade_nomg(tsv);
7557 SvCUR_set(sv,append);
7560 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7564 S_sv_gets_read_record(pTHX_ SV *const sv, PerlIO *const fp, I32 append)
7567 const U32 recsize = SvUV(SvRV(PL_rs)); /* RsRECORD() guarantees > 0. */
7568 /* Grab the size of the record we're getting */
7569 char *const buffer = SvGROW(sv, (STRLEN)(recsize + append + 1)) + append;
7576 /* VMS wants read instead of fread, because fread doesn't respect */
7577 /* RMS record boundaries. This is not necessarily a good thing to be */
7578 /* doing, but we've got no other real choice - except avoid stdio
7579 as implementation - perhaps write a :vms layer ?
7581 fd = PerlIO_fileno(fp);
7583 bytesread = PerlLIO_read(fd, buffer, recsize);
7585 else /* in-memory file from PerlIO::Scalar */
7588 bytesread = PerlIO_read(fp, buffer, recsize);
7593 SvCUR_set(sv, bytesread + append);
7594 buffer[bytesread] = '\0';
7595 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7601 Get a line from the filehandle and store it into the SV, optionally
7602 appending to the currently-stored string.
7608 Perl_sv_gets(pTHX_ register SV *const sv, register PerlIO *const fp, I32 append)
7613 register STDCHAR rslast;
7614 register STDCHAR *bp;
7619 PERL_ARGS_ASSERT_SV_GETS;
7621 if (SvTHINKFIRST(sv))
7622 sv_force_normal_flags(sv, append ? 0 : SV_COW_DROP_PV);
7623 /* XXX. If you make this PVIV, then copy on write can copy scalars read
7625 However, perlbench says it's slower, because the existing swipe code
7626 is faster than copy on write.
7627 Swings and roundabouts. */
7628 SvUPGRADE(sv, SVt_PV);
7633 if (PerlIO_isutf8(fp)) {
7635 sv_utf8_upgrade_nomg(sv);
7636 sv_pos_u2b(sv,&append,0);
7638 } else if (SvUTF8(sv)) {
7639 return S_sv_gets_append_to_utf8(aTHX_ sv, fp, append);
7647 if (PerlIO_isutf8(fp))
7650 if (IN_PERL_COMPILETIME) {
7651 /* we always read code in line mode */
7655 else if (RsSNARF(PL_rs)) {
7656 /* If it is a regular disk file use size from stat() as estimate
7657 of amount we are going to read -- may result in mallocing
7658 more memory than we really need if the layers below reduce
7659 the size we read (e.g. CRLF or a gzip layer).
7662 if (!PerlLIO_fstat(PerlIO_fileno(fp), &st) && S_ISREG(st.st_mode)) {
7663 const Off_t offset = PerlIO_tell(fp);
7664 if (offset != (Off_t) -1 && st.st_size + append > offset) {
7665 (void) SvGROW(sv, (STRLEN)((st.st_size - offset) + append + 1));
7671 else if (RsRECORD(PL_rs)) {
7672 return S_sv_gets_read_record(aTHX_ sv, fp, append);
7674 else if (RsPARA(PL_rs)) {
7680 /* Get $/ i.e. PL_rs into same encoding as stream wants */
7681 if (PerlIO_isutf8(fp)) {
7682 rsptr = SvPVutf8(PL_rs, rslen);
7685 if (SvUTF8(PL_rs)) {
7686 if (!sv_utf8_downgrade(PL_rs, TRUE)) {
7687 Perl_croak(aTHX_ "Wide character in $/");
7690 rsptr = SvPV_const(PL_rs, rslen);
7694 rslast = rslen ? rsptr[rslen - 1] : '\0';
7696 if (rspara) { /* have to do this both before and after */
7697 do { /* to make sure file boundaries work right */
7700 i = PerlIO_getc(fp);
7704 PerlIO_ungetc(fp,i);
7710 /* See if we know enough about I/O mechanism to cheat it ! */
7712 /* This used to be #ifdef test - it is made run-time test for ease
7713 of abstracting out stdio interface. One call should be cheap
7714 enough here - and may even be a macro allowing compile
7718 if (PerlIO_fast_gets(fp)) {
7721 * We're going to steal some values from the stdio struct
7722 * and put EVERYTHING in the innermost loop into registers.
7724 register STDCHAR *ptr;
7728 #if defined(VMS) && defined(PERLIO_IS_STDIO)
7729 /* An ungetc()d char is handled separately from the regular
7730 * buffer, so we getc() it back out and stuff it in the buffer.
7732 i = PerlIO_getc(fp);
7733 if (i == EOF) return 0;
7734 *(--((*fp)->_ptr)) = (unsigned char) i;
7738 /* Here is some breathtakingly efficient cheating */
7740 cnt = PerlIO_get_cnt(fp); /* get count into register */
7741 /* make sure we have the room */
7742 if ((I32)(SvLEN(sv) - append) <= cnt + 1) {
7743 /* Not room for all of it
7744 if we are looking for a separator and room for some
7746 if (rslen && cnt > 80 && (I32)SvLEN(sv) > append) {
7747 /* just process what we have room for */
7748 shortbuffered = cnt - SvLEN(sv) + append + 1;
7749 cnt -= shortbuffered;
7753 /* remember that cnt can be negative */
7754 SvGROW(sv, (STRLEN)(append + (cnt <= 0 ? 2 : (cnt + 1))));
7759 bp = (STDCHAR*)SvPVX_const(sv) + append; /* move these two too to registers */
7760 ptr = (STDCHAR*)PerlIO_get_ptr(fp);
7761 DEBUG_P(PerlIO_printf(Perl_debug_log,
7762 "Screamer: entering, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7763 DEBUG_P(PerlIO_printf(Perl_debug_log,
7764 "Screamer: entering: PerlIO * 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)));
7771 while (cnt > 0) { /* this | eat */
7773 if ((*bp++ = *ptr++) == rslast) /* really | dust */
7774 goto thats_all_folks; /* screams | sed :-) */
7778 Copy(ptr, bp, cnt, char); /* this | eat */
7779 bp += cnt; /* screams | dust */
7780 ptr += cnt; /* louder | sed :-) */
7782 assert (!shortbuffered);
7783 goto cannot_be_shortbuffered;
7787 if (shortbuffered) { /* oh well, must extend */
7788 cnt = shortbuffered;
7790 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
7792 SvGROW(sv, SvLEN(sv) + append + cnt + 2);
7793 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
7797 cannot_be_shortbuffered:
7798 DEBUG_P(PerlIO_printf(Perl_debug_log,
7799 "Screamer: going to getc, ptr=%"UVuf", cnt=%ld\n",
7800 PTR2UV(ptr),(long)cnt));
7801 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* deregisterize cnt and ptr */
7803 DEBUG_Pv(PerlIO_printf(Perl_debug_log,
7804 "Screamer: pre: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7805 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7806 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7808 /* This used to call 'filbuf' in stdio form, but as that behaves like
7809 getc when cnt <= 0 we use PerlIO_getc here to avoid introducing
7810 another abstraction. */
7811 i = PerlIO_getc(fp); /* get more characters */
7813 DEBUG_Pv(PerlIO_printf(Perl_debug_log,
7814 "Screamer: post: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7815 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7816 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7818 cnt = PerlIO_get_cnt(fp);
7819 ptr = (STDCHAR*)PerlIO_get_ptr(fp); /* reregisterize cnt and ptr */
7820 DEBUG_P(PerlIO_printf(Perl_debug_log,
7821 "Screamer: after getc, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7823 if (i == EOF) /* all done for ever? */
7824 goto thats_really_all_folks;
7826 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
7828 SvGROW(sv, bpx + cnt + 2);
7829 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
7831 *bp++ = (STDCHAR)i; /* store character from PerlIO_getc */
7833 if (rslen && (STDCHAR)i == rslast) /* all done for now? */
7834 goto thats_all_folks;
7838 if ((rslen > 1 && (STRLEN)(bp - (STDCHAR*)SvPVX_const(sv)) < rslen) ||
7839 memNE((char*)bp - rslen, rsptr, rslen))
7840 goto screamer; /* go back to the fray */
7841 thats_really_all_folks:
7843 cnt += shortbuffered;
7844 DEBUG_P(PerlIO_printf(Perl_debug_log,
7845 "Screamer: quitting, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7846 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* put these back or we're in trouble */
7847 DEBUG_P(PerlIO_printf(Perl_debug_log,
7848 "Screamer: end: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7849 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7850 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7852 SvCUR_set(sv, bp - (STDCHAR*)SvPVX_const(sv)); /* set length */
7853 DEBUG_P(PerlIO_printf(Perl_debug_log,
7854 "Screamer: done, len=%ld, string=|%.*s|\n",
7855 (long)SvCUR(sv),(int)SvCUR(sv),SvPVX_const(sv)));
7859 /*The big, slow, and stupid way. */
7860 #ifdef USE_HEAP_INSTEAD_OF_STACK /* Even slower way. */
7861 STDCHAR *buf = NULL;
7862 Newx(buf, 8192, STDCHAR);
7870 register const STDCHAR * const bpe = buf + sizeof(buf);
7872 while ((i = PerlIO_getc(fp)) != EOF && (*bp++ = (STDCHAR)i) != rslast && bp < bpe)
7873 ; /* keep reading */
7877 cnt = PerlIO_read(fp,(char*)buf, sizeof(buf));
7878 /* Accommodate broken VAXC compiler, which applies U8 cast to
7879 * both args of ?: operator, causing EOF to change into 255
7882 i = (U8)buf[cnt - 1];
7888 cnt = 0; /* we do need to re-set the sv even when cnt <= 0 */
7890 sv_catpvn(sv, (char *) buf, cnt);
7892 sv_setpvn(sv, (char *) buf, cnt);
7894 if (i != EOF && /* joy */
7896 SvCUR(sv) < rslen ||
7897 memNE(SvPVX_const(sv) + SvCUR(sv) - rslen, rsptr, rslen)))
7901 * If we're reading from a TTY and we get a short read,
7902 * indicating that the user hit his EOF character, we need
7903 * to notice it now, because if we try to read from the TTY
7904 * again, the EOF condition will disappear.
7906 * The comparison of cnt to sizeof(buf) is an optimization
7907 * that prevents unnecessary calls to feof().
7911 if (!(cnt < (I32)sizeof(buf) && PerlIO_eof(fp)))
7915 #ifdef USE_HEAP_INSTEAD_OF_STACK
7920 if (rspara) { /* have to do this both before and after */
7921 while (i != EOF) { /* to make sure file boundaries work right */
7922 i = PerlIO_getc(fp);
7924 PerlIO_ungetc(fp,i);
7930 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7936 Auto-increment of the value in the SV, doing string to numeric conversion
7937 if necessary. Handles 'get' magic and operator overloading.
7943 Perl_sv_inc(pTHX_ register SV *const sv)
7952 =for apidoc sv_inc_nomg
7954 Auto-increment of the value in the SV, doing string to numeric conversion
7955 if necessary. Handles operator overloading. Skips handling 'get' magic.
7961 Perl_sv_inc_nomg(pTHX_ register SV *const sv)
7969 if (SvTHINKFIRST(sv)) {
7971 sv_force_normal_flags(sv, 0);
7972 if (SvREADONLY(sv)) {
7973 if (IN_PERL_RUNTIME)
7974 Perl_croak_no_modify(aTHX);
7978 if (SvAMAGIC(sv) && AMG_CALLunary(sv, inc_amg))
7980 i = PTR2IV(SvRV(sv));
7985 flags = SvFLAGS(sv);
7986 if ((flags & (SVp_NOK|SVp_IOK)) == SVp_NOK) {
7987 /* It's (privately or publicly) a float, but not tested as an
7988 integer, so test it to see. */
7990 flags = SvFLAGS(sv);
7992 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
7993 /* It's publicly an integer, or privately an integer-not-float */
7994 #ifdef PERL_PRESERVE_IVUV
7998 if (SvUVX(sv) == UV_MAX)
7999 sv_setnv(sv, UV_MAX_P1);
8001 (void)SvIOK_only_UV(sv);
8002 SvUV_set(sv, SvUVX(sv) + 1);
8004 if (SvIVX(sv) == IV_MAX)
8005 sv_setuv(sv, (UV)IV_MAX + 1);
8007 (void)SvIOK_only(sv);
8008 SvIV_set(sv, SvIVX(sv) + 1);
8013 if (flags & SVp_NOK) {
8014 const NV was = SvNVX(sv);
8015 if (NV_OVERFLOWS_INTEGERS_AT &&
8016 was >= NV_OVERFLOWS_INTEGERS_AT) {
8017 Perl_ck_warner(aTHX_ packWARN(WARN_IMPRECISION),
8018 "Lost precision when incrementing %" NVff " by 1",
8021 (void)SvNOK_only(sv);
8022 SvNV_set(sv, was + 1.0);
8026 if (!(flags & SVp_POK) || !*SvPVX_const(sv)) {
8027 if ((flags & SVTYPEMASK) < SVt_PVIV)
8028 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV ? SVt_PVIV : SVt_IV));
8029 (void)SvIOK_only(sv);
8034 while (isALPHA(*d)) d++;
8035 while (isDIGIT(*d)) d++;
8036 if (d < SvEND(sv)) {
8037 #ifdef PERL_PRESERVE_IVUV
8038 /* Got to punt this as an integer if needs be, but we don't issue
8039 warnings. Probably ought to make the sv_iv_please() that does
8040 the conversion if possible, and silently. */
8041 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
8042 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
8043 /* Need to try really hard to see if it's an integer.
8044 9.22337203685478e+18 is an integer.
8045 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
8046 so $a="9.22337203685478e+18"; $a+0; $a++
8047 needs to be the same as $a="9.22337203685478e+18"; $a++
8054 /* sv_2iv *should* have made this an NV */
8055 if (flags & SVp_NOK) {
8056 (void)SvNOK_only(sv);
8057 SvNV_set(sv, SvNVX(sv) + 1.0);
8060 /* I don't think we can get here. Maybe I should assert this
8061 And if we do get here I suspect that sv_setnv will croak. NWC
8063 #if defined(USE_LONG_DOUBLE)
8064 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",
8065 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
8067 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_inc punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
8068 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
8071 #endif /* PERL_PRESERVE_IVUV */
8072 sv_setnv(sv,Atof(SvPVX_const(sv)) + 1.0);
8076 while (d >= SvPVX_const(sv)) {
8084 /* MKS: The original code here died if letters weren't consecutive.
8085 * at least it didn't have to worry about non-C locales. The
8086 * new code assumes that ('z'-'a')==('Z'-'A'), letters are
8087 * arranged in order (although not consecutively) and that only
8088 * [A-Za-z] are accepted by isALPHA in the C locale.
8090 if (*d != 'z' && *d != 'Z') {
8091 do { ++*d; } while (!isALPHA(*d));
8094 *(d--) -= 'z' - 'a';
8099 *(d--) -= 'z' - 'a' + 1;
8103 /* oh,oh, the number grew */
8104 SvGROW(sv, SvCUR(sv) + 2);
8105 SvCUR_set(sv, SvCUR(sv) + 1);
8106 for (d = SvPVX(sv) + SvCUR(sv); d > SvPVX_const(sv); d--)
8117 Auto-decrement of the value in the SV, doing string to numeric conversion
8118 if necessary. Handles 'get' magic and operator overloading.
8124 Perl_sv_dec(pTHX_ register SV *const sv)
8134 =for apidoc sv_dec_nomg
8136 Auto-decrement of the value in the SV, doing string to numeric conversion
8137 if necessary. Handles operator overloading. Skips handling 'get' magic.
8143 Perl_sv_dec_nomg(pTHX_ register SV *const sv)
8150 if (SvTHINKFIRST(sv)) {
8152 sv_force_normal_flags(sv, 0);
8153 if (SvREADONLY(sv)) {
8154 if (IN_PERL_RUNTIME)
8155 Perl_croak_no_modify(aTHX);
8159 if (SvAMAGIC(sv) && AMG_CALLunary(sv, dec_amg))
8161 i = PTR2IV(SvRV(sv));
8166 /* Unlike sv_inc we don't have to worry about string-never-numbers
8167 and keeping them magic. But we mustn't warn on punting */
8168 flags = SvFLAGS(sv);
8169 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
8170 /* It's publicly an integer, or privately an integer-not-float */
8171 #ifdef PERL_PRESERVE_IVUV
8175 if (SvUVX(sv) == 0) {
8176 (void)SvIOK_only(sv);
8180 (void)SvIOK_only_UV(sv);
8181 SvUV_set(sv, SvUVX(sv) - 1);
8184 if (SvIVX(sv) == IV_MIN) {
8185 sv_setnv(sv, (NV)IV_MIN);
8189 (void)SvIOK_only(sv);
8190 SvIV_set(sv, SvIVX(sv) - 1);
8195 if (flags & SVp_NOK) {
8198 const NV was = SvNVX(sv);
8199 if (NV_OVERFLOWS_INTEGERS_AT &&
8200 was <= -NV_OVERFLOWS_INTEGERS_AT) {
8201 Perl_ck_warner(aTHX_ packWARN(WARN_IMPRECISION),
8202 "Lost precision when decrementing %" NVff " by 1",
8205 (void)SvNOK_only(sv);
8206 SvNV_set(sv, was - 1.0);
8210 if (!(flags & SVp_POK)) {
8211 if ((flags & SVTYPEMASK) < SVt_PVIV)
8212 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV) ? SVt_PVIV : SVt_IV);
8214 (void)SvIOK_only(sv);
8217 #ifdef PERL_PRESERVE_IVUV
8219 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
8220 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
8221 /* Need to try really hard to see if it's an integer.
8222 9.22337203685478e+18 is an integer.
8223 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
8224 so $a="9.22337203685478e+18"; $a+0; $a--
8225 needs to be the same as $a="9.22337203685478e+18"; $a--
8232 /* sv_2iv *should* have made this an NV */
8233 if (flags & SVp_NOK) {
8234 (void)SvNOK_only(sv);
8235 SvNV_set(sv, SvNVX(sv) - 1.0);
8238 /* I don't think we can get here. Maybe I should assert this
8239 And if we do get here I suspect that sv_setnv will croak. NWC
8241 #if defined(USE_LONG_DOUBLE)
8242 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",
8243 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
8245 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_dec punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
8246 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
8250 #endif /* PERL_PRESERVE_IVUV */
8251 sv_setnv(sv,Atof(SvPVX_const(sv)) - 1.0); /* punt */
8254 /* this define is used to eliminate a chunk of duplicated but shared logic
8255 * it has the suffix __SV_C to signal that it isnt API, and isnt meant to be
8256 * used anywhere but here - yves
8258 #define PUSH_EXTEND_MORTAL__SV_C(AnSv) \
8261 PL_tmps_stack[++PL_tmps_ix] = (AnSv); \
8265 =for apidoc sv_mortalcopy
8267 Creates a new SV which is a copy of the original SV (using C<sv_setsv>).
8268 The new SV is marked as mortal. It will be destroyed "soon", either by an
8269 explicit call to FREETMPS, or by an implicit call at places such as
8270 statement boundaries. See also C<sv_newmortal> and C<sv_2mortal>.
8275 /* Make a string that will exist for the duration of the expression
8276 * evaluation. Actually, it may have to last longer than that, but
8277 * hopefully we won't free it until it has been assigned to a
8278 * permanent location. */
8281 Perl_sv_mortalcopy(pTHX_ SV *const oldstr)
8287 sv_setsv(sv,oldstr);
8288 PUSH_EXTEND_MORTAL__SV_C(sv);
8294 =for apidoc sv_newmortal
8296 Creates a new null SV which is mortal. The reference count of the SV is
8297 set to 1. It will be destroyed "soon", either by an explicit call to
8298 FREETMPS, or by an implicit call at places such as statement boundaries.
8299 See also C<sv_mortalcopy> and C<sv_2mortal>.
8305 Perl_sv_newmortal(pTHX)
8311 SvFLAGS(sv) = SVs_TEMP;
8312 PUSH_EXTEND_MORTAL__SV_C(sv);
8318 =for apidoc newSVpvn_flags
8320 Creates a new SV and copies a string into it. The reference count for the
8321 SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
8322 string. You are responsible for ensuring that the source string is at least
8323 C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
8324 Currently the only flag bits accepted are C<SVf_UTF8> and C<SVs_TEMP>.
8325 If C<SVs_TEMP> is set, then C<sv_2mortal()> is called on the result before
8326 returning. If C<SVf_UTF8> is set, C<s> is considered to be in UTF-8 and the
8327 C<SVf_UTF8> flag will be set on the new SV.
8328 C<newSVpvn_utf8()> is a convenience wrapper for this function, defined as
8330 #define newSVpvn_utf8(s, len, u) \
8331 newSVpvn_flags((s), (len), (u) ? SVf_UTF8 : 0)
8337 Perl_newSVpvn_flags(pTHX_ const char *const s, const STRLEN len, const U32 flags)
8342 /* All the flags we don't support must be zero.
8343 And we're new code so I'm going to assert this from the start. */
8344 assert(!(flags & ~(SVf_UTF8|SVs_TEMP)));
8346 sv_setpvn(sv,s,len);
8348 /* This code used to a sv_2mortal(), however we now unroll the call to sv_2mortal()
8349 * and do what it does ourselves here.
8350 * Since we have asserted that flags can only have the SVf_UTF8 and/or SVs_TEMP flags
8351 * set above we can use it to enable the sv flags directly (bypassing SvTEMP_on), which
8352 * in turn means we dont need to mask out the SVf_UTF8 flag below, which means that we
8353 * eliminate quite a few steps than it looks - Yves (explaining patch by gfx)
8356 SvFLAGS(sv) |= flags;
8358 if(flags & SVs_TEMP){
8359 PUSH_EXTEND_MORTAL__SV_C(sv);
8366 =for apidoc sv_2mortal
8368 Marks an existing SV as mortal. The SV will be destroyed "soon", either
8369 by an explicit call to FREETMPS, or by an implicit call at places such as
8370 statement boundaries. SvTEMP() is turned on which means that the SV's
8371 string buffer can be "stolen" if this SV is copied. See also C<sv_newmortal>
8372 and C<sv_mortalcopy>.
8378 Perl_sv_2mortal(pTHX_ register SV *const sv)
8383 if (SvREADONLY(sv) && SvIMMORTAL(sv))
8385 PUSH_EXTEND_MORTAL__SV_C(sv);
8393 Creates a new SV and copies a string into it. The reference count for the
8394 SV is set to 1. If C<len> is zero, Perl will compute the length using
8395 strlen(). For efficiency, consider using C<newSVpvn> instead.
8401 Perl_newSVpv(pTHX_ const char *const s, const STRLEN len)
8407 sv_setpvn(sv, s, len || s == NULL ? len : strlen(s));
8412 =for apidoc newSVpvn
8414 Creates a new SV and copies a string into it. The reference count for the
8415 SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
8416 string. You are responsible for ensuring that the source string is at least
8417 C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
8423 Perl_newSVpvn(pTHX_ const char *const s, const STRLEN len)
8429 sv_setpvn(sv,s,len);
8434 =for apidoc newSVhek
8436 Creates a new SV from the hash key structure. It will generate scalars that
8437 point to the shared string table where possible. Returns a new (undefined)
8438 SV if the hek is NULL.
8444 Perl_newSVhek(pTHX_ const HEK *const hek)
8454 if (HEK_LEN(hek) == HEf_SVKEY) {
8455 return newSVsv(*(SV**)HEK_KEY(hek));
8457 const int flags = HEK_FLAGS(hek);
8458 if (flags & HVhek_WASUTF8) {
8460 Andreas would like keys he put in as utf8 to come back as utf8
8462 STRLEN utf8_len = HEK_LEN(hek);
8463 SV * const sv = newSV_type(SVt_PV);
8464 char *as_utf8 = (char *)bytes_to_utf8 ((U8*)HEK_KEY(hek), &utf8_len);
8465 /* bytes_to_utf8() allocates a new string, which we can repurpose: */
8466 sv_usepvn_flags(sv, as_utf8, utf8_len, SV_HAS_TRAILING_NUL);
8469 } else if (flags & (HVhek_REHASH|HVhek_UNSHARED)) {
8470 /* We don't have a pointer to the hv, so we have to replicate the
8471 flag into every HEK. This hv is using custom a hasing
8472 algorithm. Hence we can't return a shared string scalar, as
8473 that would contain the (wrong) hash value, and might get passed
8474 into an hv routine with a regular hash.
8475 Similarly, a hash that isn't using shared hash keys has to have
8476 the flag in every key so that we know not to try to call
8477 share_hek_kek on it. */
8479 SV * const sv = newSVpvn (HEK_KEY(hek), HEK_LEN(hek));
8484 /* This will be overwhelminly the most common case. */
8486 /* Inline most of newSVpvn_share(), because share_hek_hek() is far
8487 more efficient than sharepvn(). */
8491 sv_upgrade(sv, SVt_PV);
8492 SvPV_set(sv, (char *)HEK_KEY(share_hek_hek(hek)));
8493 SvCUR_set(sv, HEK_LEN(hek));
8506 =for apidoc newSVpvn_share
8508 Creates a new SV with its SvPVX_const pointing to a shared string in the string
8509 table. If the string does not already exist in the table, it is created
8510 first. Turns on READONLY and FAKE. If the C<hash> parameter is non-zero, that
8511 value is used; otherwise the hash is computed. The string's hash can be later
8512 be retrieved from the SV with the C<SvSHARED_HASH()> macro. The idea here is
8513 that as the string table is used for shared hash keys these strings will have
8514 SvPVX_const == HeKEY and hash lookup will avoid string compare.
8520 Perl_newSVpvn_share(pTHX_ const char *src, I32 len, U32 hash)
8524 bool is_utf8 = FALSE;
8525 const char *const orig_src = src;
8528 STRLEN tmplen = -len;
8530 /* See the note in hv.c:hv_fetch() --jhi */
8531 src = (char*)bytes_from_utf8((const U8*)src, &tmplen, &is_utf8);
8535 PERL_HASH(hash, src, len);
8537 /* The logic for this is inlined in S_mro_get_linear_isa_dfs(), so if it
8538 changes here, update it there too. */
8539 sv_upgrade(sv, SVt_PV);
8540 SvPV_set(sv, sharepvn(src, is_utf8?-len:len, hash));
8548 if (src != orig_src)
8554 =for apidoc newSVpv_share
8556 Like C<newSVpvn_share>, but takes a nul-terminated string instead of a
8563 Perl_newSVpv_share(pTHX_ const char *src, U32 hash)
8565 return newSVpvn_share(src, strlen(src), hash);
8568 #if defined(PERL_IMPLICIT_CONTEXT)
8570 /* pTHX_ magic can't cope with varargs, so this is a no-context
8571 * version of the main function, (which may itself be aliased to us).
8572 * Don't access this version directly.
8576 Perl_newSVpvf_nocontext(const char *const pat, ...)
8582 PERL_ARGS_ASSERT_NEWSVPVF_NOCONTEXT;
8584 va_start(args, pat);
8585 sv = vnewSVpvf(pat, &args);
8592 =for apidoc newSVpvf
8594 Creates a new SV and initializes it with the string formatted like
8601 Perl_newSVpvf(pTHX_ const char *const pat, ...)
8606 PERL_ARGS_ASSERT_NEWSVPVF;
8608 va_start(args, pat);
8609 sv = vnewSVpvf(pat, &args);
8614 /* backend for newSVpvf() and newSVpvf_nocontext() */
8617 Perl_vnewSVpvf(pTHX_ const char *const pat, va_list *const args)
8622 PERL_ARGS_ASSERT_VNEWSVPVF;
8625 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
8632 Creates a new SV and copies a floating point value into it.
8633 The reference count for the SV is set to 1.
8639 Perl_newSVnv(pTHX_ const NV n)
8652 Creates a new SV and copies an integer into it. The reference count for the
8659 Perl_newSViv(pTHX_ const IV i)
8672 Creates a new SV and copies an unsigned integer into it.
8673 The reference count for the SV is set to 1.
8679 Perl_newSVuv(pTHX_ const UV u)
8690 =for apidoc newSV_type
8692 Creates a new SV, of the type specified. The reference count for the new SV
8699 Perl_newSV_type(pTHX_ const svtype type)
8704 sv_upgrade(sv, type);
8709 =for apidoc newRV_noinc
8711 Creates an RV wrapper for an SV. The reference count for the original
8712 SV is B<not> incremented.
8718 Perl_newRV_noinc(pTHX_ SV *const tmpRef)
8721 register SV *sv = newSV_type(SVt_IV);
8723 PERL_ARGS_ASSERT_NEWRV_NOINC;
8726 SvRV_set(sv, tmpRef);
8731 /* newRV_inc is the official function name to use now.
8732 * newRV_inc is in fact #defined to newRV in sv.h
8736 Perl_newRV(pTHX_ SV *const sv)
8740 PERL_ARGS_ASSERT_NEWRV;
8742 return newRV_noinc(SvREFCNT_inc_simple_NN(sv));
8748 Creates a new SV which is an exact duplicate of the original SV.
8755 Perl_newSVsv(pTHX_ register SV *const old)
8762 if (SvTYPE(old) == SVTYPEMASK) {
8763 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL), "semi-panic: attempt to dup freed string");
8767 /* SV_GMAGIC is the default for sv_setv()
8768 SV_NOSTEAL prevents TEMP buffers being, well, stolen, and saves games
8769 with SvTEMP_off and SvTEMP_on round a call to sv_setsv. */
8770 sv_setsv_flags(sv, old, SV_GMAGIC | SV_NOSTEAL);
8775 =for apidoc sv_reset
8777 Underlying implementation for the C<reset> Perl function.
8778 Note that the perl-level function is vaguely deprecated.
8784 Perl_sv_reset(pTHX_ register const char *s, HV *const stash)
8787 char todo[PERL_UCHAR_MAX+1];
8789 PERL_ARGS_ASSERT_SV_RESET;
8794 if (!*s) { /* reset ?? searches */
8795 MAGIC * const mg = mg_find((const SV *)stash, PERL_MAGIC_symtab);
8797 const U32 count = mg->mg_len / sizeof(PMOP**);
8798 PMOP **pmp = (PMOP**) mg->mg_ptr;
8799 PMOP *const *const end = pmp + count;
8803 SvREADONLY_off(PL_regex_pad[(*pmp)->op_pmoffset]);
8805 (*pmp)->op_pmflags &= ~PMf_USED;
8813 /* reset variables */
8815 if (!HvARRAY(stash))
8818 Zero(todo, 256, char);
8821 I32 i = (unsigned char)*s;
8825 max = (unsigned char)*s++;
8826 for ( ; i <= max; i++) {
8829 for (i = 0; i <= (I32) HvMAX(stash); i++) {
8831 for (entry = HvARRAY(stash)[i];
8833 entry = HeNEXT(entry))
8838 if (!todo[(U8)*HeKEY(entry)])
8840 gv = MUTABLE_GV(HeVAL(entry));
8843 if (SvTHINKFIRST(sv)) {
8844 if (!SvREADONLY(sv) && SvROK(sv))
8846 /* XXX Is this continue a bug? Why should THINKFIRST
8847 exempt us from resetting arrays and hashes? */
8851 if (SvTYPE(sv) >= SVt_PV) {
8853 if (SvPVX_const(sv) != NULL)
8861 if (GvHV(gv) && !HvNAME_get(GvHV(gv))) {
8863 Perl_die(aTHX_ "Can't reset %%ENV on this system");
8866 # if defined(USE_ENVIRON_ARRAY)
8869 # endif /* USE_ENVIRON_ARRAY */
8880 Using various gambits, try to get an IO from an SV: the IO slot if its a
8881 GV; or the recursive result if we're an RV; or the IO slot of the symbol
8882 named after the PV if we're a string.
8888 Perl_sv_2io(pTHX_ SV *const sv)
8893 PERL_ARGS_ASSERT_SV_2IO;
8895 switch (SvTYPE(sv)) {
8897 io = MUTABLE_IO(sv);
8901 if (isGV_with_GP(sv)) {
8902 gv = MUTABLE_GV(sv);
8905 Perl_croak(aTHX_ "Bad filehandle: %s", GvNAME(gv));
8911 Perl_croak(aTHX_ PL_no_usym, "filehandle");
8913 return sv_2io(SvRV(sv));
8914 gv = gv_fetchsv(sv, 0, SVt_PVIO);
8920 Perl_croak(aTHX_ "Bad filehandle: %"SVf, SVfARG(sv));
8929 Using various gambits, try to get a CV from an SV; in addition, try if
8930 possible to set C<*st> and C<*gvp> to the stash and GV associated with it.
8931 The flags in C<lref> are passed to gv_fetchsv.
8937 Perl_sv_2cv(pTHX_ SV *sv, HV **const st, GV **const gvp, const I32 lref)
8943 PERL_ARGS_ASSERT_SV_2CV;
8950 switch (SvTYPE(sv)) {
8954 return MUTABLE_CV(sv);
8961 if (isGV_with_GP(sv)) {
8962 gv = MUTABLE_GV(sv);
8973 sv = amagic_deref_call(sv, to_cv_amg);
8974 /* At this point I'd like to do SPAGAIN, but really I need to
8975 force it upon my callers. Hmmm. This is a mess... */
8978 if (SvTYPE(sv) == SVt_PVCV) {
8979 cv = MUTABLE_CV(sv);
8984 else if(isGV_with_GP(sv))
8985 gv = MUTABLE_GV(sv);
8987 Perl_croak(aTHX_ "Not a subroutine reference");
8989 else if (isGV_with_GP(sv)) {
8991 gv = MUTABLE_GV(sv);
8994 gv = gv_fetchsv(sv, lref, SVt_PVCV); /* Calls get magic */
9000 /* Some flags to gv_fetchsv mean don't really create the GV */
9001 if (!isGV_with_GP(gv)) {
9007 if (lref && !GvCVu(gv)) {
9011 gv_efullname3(tmpsv, gv, NULL);
9012 /* XXX this is probably not what they think they're getting.
9013 * It has the same effect as "sub name;", i.e. just a forward
9015 newSUB(start_subparse(FALSE, 0),
9016 newSVOP(OP_CONST, 0, tmpsv),
9020 Perl_croak(aTHX_ "Unable to create sub named \"%"SVf"\"",
9021 SVfARG(SvOK(sv) ? sv : &PL_sv_no));
9030 Returns true if the SV has a true value by Perl's rules.
9031 Use the C<SvTRUE> macro instead, which may call C<sv_true()> or may
9032 instead use an in-line version.
9038 Perl_sv_true(pTHX_ register SV *const sv)
9043 register const XPV* const tXpv = (XPV*)SvANY(sv);
9045 (tXpv->xpv_cur > 1 ||
9046 (tXpv->xpv_cur && *sv->sv_u.svu_pv != '0')))
9053 return SvIVX(sv) != 0;
9056 return SvNVX(sv) != 0.0;
9058 return sv_2bool(sv);
9064 =for apidoc sv_pvn_force
9066 Get a sensible string out of the SV somehow.
9067 A private implementation of the C<SvPV_force> macro for compilers which
9068 can't cope with complex macro expressions. Always use the macro instead.
9070 =for apidoc sv_pvn_force_flags
9072 Get a sensible string out of the SV somehow.
9073 If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<sv> if
9074 appropriate, else not. C<sv_pvn_force> and C<sv_pvn_force_nomg> are
9075 implemented in terms of this function.
9076 You normally want to use the various wrapper macros instead: see
9077 C<SvPV_force> and C<SvPV_force_nomg>
9083 Perl_sv_pvn_force_flags(pTHX_ SV *const sv, STRLEN *const lp, const I32 flags)
9087 PERL_ARGS_ASSERT_SV_PVN_FORCE_FLAGS;
9089 if (SvTHINKFIRST(sv) && !SvROK(sv))
9090 sv_force_normal_flags(sv, 0);
9100 if (SvREADONLY(sv) && !(flags & SV_MUTABLE_RETURN)) {
9101 const char * const ref = sv_reftype(sv,0);
9103 Perl_croak(aTHX_ "Can't coerce readonly %s to string in %s",
9104 ref, OP_DESC(PL_op));
9106 Perl_croak(aTHX_ "Can't coerce readonly %s to string", ref);
9108 if ((SvTYPE(sv) > SVt_PVLV && SvTYPE(sv) != SVt_PVFM)
9109 || isGV_with_GP(sv))
9110 /* diag_listed_as: Can't coerce %s to %s in %s */
9111 Perl_croak(aTHX_ "Can't coerce %s to string in %s", sv_reftype(sv,0),
9113 s = sv_2pv_flags(sv, &len, flags);
9117 if (s != SvPVX_const(sv)) { /* Almost, but not quite, sv_setpvn() */
9120 SvUPGRADE(sv, SVt_PV); /* Never FALSE */
9121 SvGROW(sv, len + 1);
9122 Move(s,SvPVX(sv),len,char);
9124 SvPVX(sv)[len] = '\0';
9127 SvPOK_on(sv); /* validate pointer */
9129 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
9130 PTR2UV(sv),SvPVX_const(sv)));
9133 return SvPVX_mutable(sv);
9137 =for apidoc sv_pvbyten_force
9139 The backend for the C<SvPVbytex_force> macro. Always use the macro instead.
9145 Perl_sv_pvbyten_force(pTHX_ SV *const sv, STRLEN *const lp)
9147 PERL_ARGS_ASSERT_SV_PVBYTEN_FORCE;
9149 sv_pvn_force(sv,lp);
9150 sv_utf8_downgrade(sv,0);
9156 =for apidoc sv_pvutf8n_force
9158 The backend for the C<SvPVutf8x_force> macro. Always use the macro instead.
9164 Perl_sv_pvutf8n_force(pTHX_ SV *const sv, STRLEN *const lp)
9166 PERL_ARGS_ASSERT_SV_PVUTF8N_FORCE;
9168 sv_pvn_force(sv,lp);
9169 sv_utf8_upgrade(sv);
9175 =for apidoc sv_reftype
9177 Returns a string describing what the SV is a reference to.
9183 Perl_sv_reftype(pTHX_ const SV *const sv, const int ob)
9185 PERL_ARGS_ASSERT_SV_REFTYPE;
9187 /* The fact that I don't need to downcast to char * everywhere, only in ?:
9188 inside return suggests a const propagation bug in g++. */
9189 if (ob && SvOBJECT(sv)) {
9190 char * const name = HvNAME_get(SvSTASH(sv));
9191 return name ? name : (char *) "__ANON__";
9194 switch (SvTYPE(sv)) {
9209 case SVt_PVLV: return (char *) (SvROK(sv) ? "REF"
9210 /* tied lvalues should appear to be
9211 * scalars for backwards compatibility */
9212 : (LvTYPE(sv) == 't' || LvTYPE(sv) == 'T')
9213 ? "SCALAR" : "LVALUE");
9214 case SVt_PVAV: return "ARRAY";
9215 case SVt_PVHV: return "HASH";
9216 case SVt_PVCV: return "CODE";
9217 case SVt_PVGV: return (char *) (isGV_with_GP(sv)
9218 ? "GLOB" : "SCALAR");
9219 case SVt_PVFM: return "FORMAT";
9220 case SVt_PVIO: return "IO";
9221 case SVt_BIND: return "BIND";
9222 case SVt_REGEXP: return "REGEXP";
9223 default: return "UNKNOWN";
9229 =for apidoc sv_isobject
9231 Returns a boolean indicating whether the SV is an RV pointing to a blessed
9232 object. If the SV is not an RV, or if the object is not blessed, then this
9239 Perl_sv_isobject(pTHX_ SV *sv)
9255 Returns a boolean indicating whether the SV is blessed into the specified
9256 class. This does not check for subtypes; use C<sv_derived_from> to verify
9257 an inheritance relationship.
9263 Perl_sv_isa(pTHX_ SV *sv, const char *const name)
9267 PERL_ARGS_ASSERT_SV_ISA;
9277 hvname = HvNAME_get(SvSTASH(sv));
9281 return strEQ(hvname, name);
9287 Creates a new SV for the RV, C<rv>, to point to. If C<rv> is not an RV then
9288 it will be upgraded to one. If C<classname> is non-null then the new SV will
9289 be blessed in the specified package. The new SV is returned and its
9290 reference count is 1.
9296 Perl_newSVrv(pTHX_ SV *const rv, const char *const classname)
9301 PERL_ARGS_ASSERT_NEWSVRV;
9305 SV_CHECK_THINKFIRST_COW_DROP(rv);
9306 (void)SvAMAGIC_off(rv);
9308 if (SvTYPE(rv) >= SVt_PVMG) {
9309 const U32 refcnt = SvREFCNT(rv);
9313 SvREFCNT(rv) = refcnt;
9315 sv_upgrade(rv, SVt_IV);
9316 } else if (SvROK(rv)) {
9317 SvREFCNT_dec(SvRV(rv));
9319 prepare_SV_for_RV(rv);
9327 HV* const stash = gv_stashpv(classname, GV_ADD);
9328 (void)sv_bless(rv, stash);
9334 =for apidoc sv_setref_pv
9336 Copies a pointer into a new SV, optionally blessing the SV. The C<rv>
9337 argument will be upgraded to an RV. That RV will be modified to point to
9338 the new SV. If the C<pv> argument is NULL then C<PL_sv_undef> will be placed
9339 into the SV. The C<classname> argument indicates the package for the
9340 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9341 will have a reference count of 1, and the RV will be returned.
9343 Do not use with other Perl types such as HV, AV, SV, CV, because those
9344 objects will become corrupted by the pointer copy process.
9346 Note that C<sv_setref_pvn> copies the string while this copies the pointer.
9352 Perl_sv_setref_pv(pTHX_ SV *const rv, const char *const classname, void *const pv)
9356 PERL_ARGS_ASSERT_SV_SETREF_PV;
9359 sv_setsv(rv, &PL_sv_undef);
9363 sv_setiv(newSVrv(rv,classname), PTR2IV(pv));
9368 =for apidoc sv_setref_iv
9370 Copies an integer into a new SV, optionally blessing the SV. The C<rv>
9371 argument will be upgraded to an RV. That RV will be modified to point to
9372 the new SV. The C<classname> argument indicates the package for the
9373 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9374 will have a reference count of 1, and the RV will be returned.
9380 Perl_sv_setref_iv(pTHX_ SV *const rv, const char *const classname, const IV iv)
9382 PERL_ARGS_ASSERT_SV_SETREF_IV;
9384 sv_setiv(newSVrv(rv,classname), iv);
9389 =for apidoc sv_setref_uv
9391 Copies an unsigned integer into a new SV, optionally blessing the SV. The C<rv>
9392 argument will be upgraded to an RV. That RV will be modified to point to
9393 the new SV. The C<classname> argument indicates the package for the
9394 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9395 will have a reference count of 1, and the RV will be returned.
9401 Perl_sv_setref_uv(pTHX_ SV *const rv, const char *const classname, const UV uv)
9403 PERL_ARGS_ASSERT_SV_SETREF_UV;
9405 sv_setuv(newSVrv(rv,classname), uv);
9410 =for apidoc sv_setref_nv
9412 Copies a double into a new SV, optionally blessing the SV. The C<rv>
9413 argument will be upgraded to an RV. That RV will be modified to point to
9414 the new SV. The C<classname> argument indicates the package for the
9415 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9416 will have a reference count of 1, and the RV will be returned.
9422 Perl_sv_setref_nv(pTHX_ SV *const rv, const char *const classname, const NV nv)
9424 PERL_ARGS_ASSERT_SV_SETREF_NV;
9426 sv_setnv(newSVrv(rv,classname), nv);
9431 =for apidoc sv_setref_pvn
9433 Copies a string into a new SV, optionally blessing the SV. The length of the
9434 string must be specified with C<n>. The C<rv> argument will be upgraded to
9435 an RV. That RV will be modified to point to the new SV. The C<classname>
9436 argument indicates the package for the blessing. Set C<classname> to
9437 C<NULL> to avoid the blessing. The new SV will have a reference count
9438 of 1, and the RV will be returned.
9440 Note that C<sv_setref_pv> copies the pointer while this copies the string.
9446 Perl_sv_setref_pvn(pTHX_ SV *const rv, const char *const classname,
9447 const char *const pv, const STRLEN n)
9449 PERL_ARGS_ASSERT_SV_SETREF_PVN;
9451 sv_setpvn(newSVrv(rv,classname), pv, n);
9456 =for apidoc sv_bless
9458 Blesses an SV into a specified package. The SV must be an RV. The package
9459 must be designated by its stash (see C<gv_stashpv()>). The reference count
9460 of the SV is unaffected.
9466 Perl_sv_bless(pTHX_ SV *const sv, HV *const stash)
9471 PERL_ARGS_ASSERT_SV_BLESS;
9474 Perl_croak(aTHX_ "Can't bless non-reference value");
9476 if (SvFLAGS(tmpRef) & (SVs_OBJECT|SVf_READONLY)) {
9477 if (SvIsCOW(tmpRef))
9478 sv_force_normal_flags(tmpRef, 0);
9479 if (SvREADONLY(tmpRef))
9480 Perl_croak_no_modify(aTHX);
9481 if (SvOBJECT(tmpRef)) {
9482 if (SvTYPE(tmpRef) != SVt_PVIO)
9484 SvREFCNT_dec(SvSTASH(tmpRef));
9487 SvOBJECT_on(tmpRef);
9488 if (SvTYPE(tmpRef) != SVt_PVIO)
9490 SvUPGRADE(tmpRef, SVt_PVMG);
9491 SvSTASH_set(tmpRef, MUTABLE_HV(SvREFCNT_inc_simple(stash)));
9496 (void)SvAMAGIC_off(sv);
9498 if(SvSMAGICAL(tmpRef))
9499 if(mg_find(tmpRef, PERL_MAGIC_ext) || mg_find(tmpRef, PERL_MAGIC_uvar))
9507 /* Downgrades a PVGV to a PVMG. If it’s actually a PVLV, we leave the type
9508 * as it is after unglobbing it.
9512 S_sv_unglob(pTHX_ SV *const sv)
9517 SV * const temp = sv_newmortal();
9519 PERL_ARGS_ASSERT_SV_UNGLOB;
9521 assert(SvTYPE(sv) == SVt_PVGV || SvTYPE(sv) == SVt_PVLV);
9523 gv_efullname3(temp, MUTABLE_GV(sv), "*");
9526 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
9527 && HvNAME_get(stash))
9528 mro_method_changed_in(stash);
9529 gp_free(MUTABLE_GV(sv));
9532 sv_del_backref(MUTABLE_SV(GvSTASH(sv)), sv);
9536 if (GvNAME_HEK(sv)) {
9537 unshare_hek(GvNAME_HEK(sv));
9539 isGV_with_GP_off(sv);
9541 if(SvTYPE(sv) == SVt_PVGV) {
9542 /* need to keep SvANY(sv) in the right arena */
9543 xpvmg = new_XPVMG();
9544 StructCopy(SvANY(sv), xpvmg, XPVMG);
9545 del_XPVGV(SvANY(sv));
9548 SvFLAGS(sv) &= ~SVTYPEMASK;
9549 SvFLAGS(sv) |= SVt_PVMG;
9552 /* Intentionally not calling any local SET magic, as this isn't so much a
9553 set operation as merely an internal storage change. */
9554 sv_setsv_flags(sv, temp, 0);
9558 =for apidoc sv_unref_flags
9560 Unsets the RV status of the SV, and decrements the reference count of
9561 whatever was being referenced by the RV. This can almost be thought of
9562 as a reversal of C<newSVrv>. The C<cflags> argument can contain
9563 C<SV_IMMEDIATE_UNREF> to force the reference count to be decremented
9564 (otherwise the decrementing is conditional on the reference count being
9565 different from one or the reference being a readonly SV).
9572 Perl_sv_unref_flags(pTHX_ SV *const ref, const U32 flags)
9574 SV* const target = SvRV(ref);
9576 PERL_ARGS_ASSERT_SV_UNREF_FLAGS;
9578 if (SvWEAKREF(ref)) {
9579 sv_del_backref(target, ref);
9581 SvRV_set(ref, NULL);
9584 SvRV_set(ref, NULL);
9586 /* You can't have a || SvREADONLY(target) here, as $a = $$a, where $a was
9587 assigned to as BEGIN {$a = \"Foo"} will fail. */
9588 if (SvREFCNT(target) != 1 || (flags & SV_IMMEDIATE_UNREF))
9589 SvREFCNT_dec(target);
9590 else /* XXX Hack, but hard to make $a=$a->[1] work otherwise */
9591 sv_2mortal(target); /* Schedule for freeing later */
9595 =for apidoc sv_untaint
9597 Untaint an SV. Use C<SvTAINTED_off> instead.
9602 Perl_sv_untaint(pTHX_ SV *const sv)
9604 PERL_ARGS_ASSERT_SV_UNTAINT;
9606 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
9607 MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
9614 =for apidoc sv_tainted
9616 Test an SV for taintedness. Use C<SvTAINTED> instead.
9621 Perl_sv_tainted(pTHX_ SV *const sv)
9623 PERL_ARGS_ASSERT_SV_TAINTED;
9625 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
9626 const MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
9627 if (mg && (mg->mg_len & 1) )
9634 =for apidoc sv_setpviv
9636 Copies an integer into the given SV, also updating its string value.
9637 Does not handle 'set' magic. See C<sv_setpviv_mg>.
9643 Perl_sv_setpviv(pTHX_ SV *const sv, const IV iv)
9645 char buf[TYPE_CHARS(UV)];
9647 char * const ptr = uiv_2buf(buf, iv, 0, 0, &ebuf);
9649 PERL_ARGS_ASSERT_SV_SETPVIV;
9651 sv_setpvn(sv, ptr, ebuf - ptr);
9655 =for apidoc sv_setpviv_mg
9657 Like C<sv_setpviv>, but also handles 'set' magic.
9663 Perl_sv_setpviv_mg(pTHX_ SV *const sv, const IV iv)
9665 PERL_ARGS_ASSERT_SV_SETPVIV_MG;
9671 #if defined(PERL_IMPLICIT_CONTEXT)
9673 /* pTHX_ magic can't cope with varargs, so this is a no-context
9674 * version of the main function, (which may itself be aliased to us).
9675 * Don't access this version directly.
9679 Perl_sv_setpvf_nocontext(SV *const sv, const char *const pat, ...)
9684 PERL_ARGS_ASSERT_SV_SETPVF_NOCONTEXT;
9686 va_start(args, pat);
9687 sv_vsetpvf(sv, pat, &args);
9691 /* pTHX_ magic can't cope with varargs, so this is a no-context
9692 * version of the main function, (which may itself be aliased to us).
9693 * Don't access this version directly.
9697 Perl_sv_setpvf_mg_nocontext(SV *const sv, const char *const pat, ...)
9702 PERL_ARGS_ASSERT_SV_SETPVF_MG_NOCONTEXT;
9704 va_start(args, pat);
9705 sv_vsetpvf_mg(sv, pat, &args);
9711 =for apidoc sv_setpvf
9713 Works like C<sv_catpvf> but copies the text into the SV instead of
9714 appending it. Does not handle 'set' magic. See C<sv_setpvf_mg>.
9720 Perl_sv_setpvf(pTHX_ SV *const sv, const char *const pat, ...)
9724 PERL_ARGS_ASSERT_SV_SETPVF;
9726 va_start(args, pat);
9727 sv_vsetpvf(sv, pat, &args);
9732 =for apidoc sv_vsetpvf
9734 Works like C<sv_vcatpvf> but copies the text into the SV instead of
9735 appending it. Does not handle 'set' magic. See C<sv_vsetpvf_mg>.
9737 Usually used via its frontend C<sv_setpvf>.
9743 Perl_sv_vsetpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9745 PERL_ARGS_ASSERT_SV_VSETPVF;
9747 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9751 =for apidoc sv_setpvf_mg
9753 Like C<sv_setpvf>, but also handles 'set' magic.
9759 Perl_sv_setpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
9763 PERL_ARGS_ASSERT_SV_SETPVF_MG;
9765 va_start(args, pat);
9766 sv_vsetpvf_mg(sv, pat, &args);
9771 =for apidoc sv_vsetpvf_mg
9773 Like C<sv_vsetpvf>, but also handles 'set' magic.
9775 Usually used via its frontend C<sv_setpvf_mg>.
9781 Perl_sv_vsetpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9783 PERL_ARGS_ASSERT_SV_VSETPVF_MG;
9785 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9789 #if defined(PERL_IMPLICIT_CONTEXT)
9791 /* pTHX_ magic can't cope with varargs, so this is a no-context
9792 * version of the main function, (which may itself be aliased to us).
9793 * Don't access this version directly.
9797 Perl_sv_catpvf_nocontext(SV *const sv, const char *const pat, ...)
9802 PERL_ARGS_ASSERT_SV_CATPVF_NOCONTEXT;
9804 va_start(args, pat);
9805 sv_vcatpvf(sv, pat, &args);
9809 /* pTHX_ magic can't cope with varargs, so this is a no-context
9810 * version of the main function, (which may itself be aliased to us).
9811 * Don't access this version directly.
9815 Perl_sv_catpvf_mg_nocontext(SV *const sv, const char *const pat, ...)
9820 PERL_ARGS_ASSERT_SV_CATPVF_MG_NOCONTEXT;
9822 va_start(args, pat);
9823 sv_vcatpvf_mg(sv, pat, &args);
9829 =for apidoc sv_catpvf
9831 Processes its arguments like C<sprintf> and appends the formatted
9832 output to an SV. If the appended data contains "wide" characters
9833 (including, but not limited to, SVs with a UTF-8 PV formatted with %s,
9834 and characters >255 formatted with %c), the original SV might get
9835 upgraded to UTF-8. Handles 'get' magic, but not 'set' magic. See
9836 C<sv_catpvf_mg>. If the original SV was UTF-8, the pattern should be
9837 valid UTF-8; if the original SV was bytes, the pattern should be too.
9842 Perl_sv_catpvf(pTHX_ SV *const sv, const char *const pat, ...)
9846 PERL_ARGS_ASSERT_SV_CATPVF;
9848 va_start(args, pat);
9849 sv_vcatpvf(sv, pat, &args);
9854 =for apidoc sv_vcatpvf
9856 Processes its arguments like C<vsprintf> and appends the formatted output
9857 to an SV. Does not handle 'set' magic. See C<sv_vcatpvf_mg>.
9859 Usually used via its frontend C<sv_catpvf>.
9865 Perl_sv_vcatpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9867 PERL_ARGS_ASSERT_SV_VCATPVF;
9869 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9873 =for apidoc sv_catpvf_mg
9875 Like C<sv_catpvf>, but also handles 'set' magic.
9881 Perl_sv_catpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
9885 PERL_ARGS_ASSERT_SV_CATPVF_MG;
9887 va_start(args, pat);
9888 sv_vcatpvf_mg(sv, pat, &args);
9893 =for apidoc sv_vcatpvf_mg
9895 Like C<sv_vcatpvf>, but also handles 'set' magic.
9897 Usually used via its frontend C<sv_catpvf_mg>.
9903 Perl_sv_vcatpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9905 PERL_ARGS_ASSERT_SV_VCATPVF_MG;
9907 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9912 =for apidoc sv_vsetpvfn
9914 Works like C<sv_vcatpvfn> but copies the text into the SV instead of
9917 Usually used via one of its frontends C<sv_vsetpvf> and C<sv_vsetpvf_mg>.
9923 Perl_sv_vsetpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
9924 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
9926 PERL_ARGS_ASSERT_SV_VSETPVFN;
9929 sv_vcatpvfn(sv, pat, patlen, args, svargs, svmax, maybe_tainted);
9934 * Warn of missing argument to sprintf, and then return a defined value
9935 * to avoid inappropriate "use of uninit" warnings [perl #71000].
9937 #define WARN_MISSING WARN_UNINITIALIZED /* Not sure we want a new category */
9939 S_vcatpvfn_missing_argument(pTHX) {
9940 if (ckWARN(WARN_MISSING)) {
9941 Perl_warner(aTHX_ packWARN(WARN_MISSING), "Missing argument in %s",
9942 PL_op ? OP_DESC(PL_op) : "sv_vcatpvfn()");
9949 S_expect_number(pTHX_ char **const pattern)
9954 PERL_ARGS_ASSERT_EXPECT_NUMBER;
9956 switch (**pattern) {
9957 case '1': case '2': case '3':
9958 case '4': case '5': case '6':
9959 case '7': case '8': case '9':
9960 var = *(*pattern)++ - '0';
9961 while (isDIGIT(**pattern)) {
9962 const I32 tmp = var * 10 + (*(*pattern)++ - '0');
9964 Perl_croak(aTHX_ "Integer overflow in format string for %s", (PL_op ? OP_DESC(PL_op) : "sv_vcatpvfn"));
9972 S_F0convert(NV nv, char *const endbuf, STRLEN *const len)
9974 const int neg = nv < 0;
9977 PERL_ARGS_ASSERT_F0CONVERT;
9985 if (uv & 1 && uv == nv)
9986 uv--; /* Round to even */
9988 const unsigned dig = uv % 10;
10001 =for apidoc sv_vcatpvfn
10003 Processes its arguments like C<vsprintf> and appends the formatted output
10004 to an SV. Uses an array of SVs if the C style variable argument list is
10005 missing (NULL). When running with taint checks enabled, indicates via
10006 C<maybe_tainted> if results are untrustworthy (often due to the use of
10009 Usually used via one of its frontends C<sv_vcatpvf> and C<sv_vcatpvf_mg>.
10015 #define VECTORIZE_ARGS vecsv = va_arg(*args, SV*);\
10016 vecstr = (U8*)SvPV_const(vecsv,veclen);\
10017 vec_utf8 = DO_UTF8(vecsv);
10019 /* XXX maybe_tainted is never assigned to, so the doc above is lying. */
10022 Perl_sv_vcatpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
10023 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
10028 const char *patend;
10031 static const char nullstr[] = "(null)";
10033 bool has_utf8 = DO_UTF8(sv); /* has the result utf8? */
10034 const bool pat_utf8 = has_utf8; /* the pattern is in utf8? */
10036 /* Times 4: a decimal digit takes more than 3 binary digits.
10037 * NV_DIG: mantissa takes than many decimal digits.
10038 * Plus 32: Playing safe. */
10039 char ebuf[IV_DIG * 4 + NV_DIG + 32];
10040 /* large enough for "%#.#f" --chip */
10041 /* what about long double NVs? --jhi */
10043 PERL_ARGS_ASSERT_SV_VCATPVFN;
10044 PERL_UNUSED_ARG(maybe_tainted);
10046 /* no matter what, this is a string now */
10047 (void)SvPV_force(sv, origlen);
10049 /* special-case "", "%s", and "%-p" (SVf - see below) */
10052 if (patlen == 2 && pat[0] == '%' && pat[1] == 's') {
10054 const char * const s = va_arg(*args, char*);
10055 sv_catpv(sv, s ? s : nullstr);
10057 else if (svix < svmax) {
10058 sv_catsv(sv, *svargs);
10061 S_vcatpvfn_missing_argument(aTHX);
10064 if (args && patlen == 3 && pat[0] == '%' &&
10065 pat[1] == '-' && pat[2] == 'p') {
10066 argsv = MUTABLE_SV(va_arg(*args, void*));
10067 sv_catsv(sv, argsv);
10071 #ifndef USE_LONG_DOUBLE
10072 /* special-case "%.<number>[gf]" */
10073 if ( !args && patlen <= 5 && pat[0] == '%' && pat[1] == '.'
10074 && (pat[patlen-1] == 'g' || pat[patlen-1] == 'f') ) {
10075 unsigned digits = 0;
10079 while (*pp >= '0' && *pp <= '9')
10080 digits = 10 * digits + (*pp++ - '0');
10081 if (pp - pat == (int)patlen - 1 && svix < svmax) {
10082 const NV nv = SvNV(*svargs);
10084 /* Add check for digits != 0 because it seems that some
10085 gconverts are buggy in this case, and we don't yet have
10086 a Configure test for this. */
10087 if (digits && digits < sizeof(ebuf) - NV_DIG - 10) {
10088 /* 0, point, slack */
10089 Gconvert(nv, (int)digits, 0, ebuf);
10090 sv_catpv(sv, ebuf);
10091 if (*ebuf) /* May return an empty string for digits==0 */
10094 } else if (!digits) {
10097 if ((p = F0convert(nv, ebuf + sizeof ebuf, &l))) {
10098 sv_catpvn(sv, p, l);
10104 #endif /* !USE_LONG_DOUBLE */
10106 if (!args && svix < svmax && DO_UTF8(*svargs))
10109 patend = (char*)pat + patlen;
10110 for (p = (char*)pat; p < patend; p = q) {
10113 bool vectorize = FALSE;
10114 bool vectorarg = FALSE;
10115 bool vec_utf8 = FALSE;
10121 bool has_precis = FALSE;
10123 const I32 osvix = svix;
10124 bool is_utf8 = FALSE; /* is this item utf8? */
10125 #ifdef HAS_LDBL_SPRINTF_BUG
10126 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
10127 with sfio - Allen <allens@cpan.org> */
10128 bool fix_ldbl_sprintf_bug = FALSE;
10132 U8 utf8buf[UTF8_MAXBYTES+1];
10133 STRLEN esignlen = 0;
10135 const char *eptr = NULL;
10136 const char *fmtstart;
10139 const U8 *vecstr = NULL;
10146 /* we need a long double target in case HAS_LONG_DOUBLE but
10147 not USE_LONG_DOUBLE
10149 #if defined(HAS_LONG_DOUBLE) && LONG_DOUBLESIZE > DOUBLESIZE
10157 const char *dotstr = ".";
10158 STRLEN dotstrlen = 1;
10159 I32 efix = 0; /* explicit format parameter index */
10160 I32 ewix = 0; /* explicit width index */
10161 I32 epix = 0; /* explicit precision index */
10162 I32 evix = 0; /* explicit vector index */
10163 bool asterisk = FALSE;
10165 /* echo everything up to the next format specification */
10166 for (q = p; q < patend && *q != '%'; ++q) ;
10168 if (has_utf8 && !pat_utf8)
10169 sv_catpvn_utf8_upgrade(sv, p, q - p, nsv);
10171 sv_catpvn(sv, p, q - p);
10180 We allow format specification elements in this order:
10181 \d+\$ explicit format parameter index
10183 v|\*(\d+\$)?v vector with optional (optionally specified) arg
10184 0 flag (as above): repeated to allow "v02"
10185 \d+|\*(\d+\$)? width using optional (optionally specified) arg
10186 \.(\d*|\*(\d+\$)?) precision using optional (optionally specified) arg
10188 [%bcdefginopsuxDFOUX] format (mandatory)
10193 As of perl5.9.3, printf format checking is on by default.
10194 Internally, perl uses %p formats to provide an escape to
10195 some extended formatting. This block deals with those
10196 extensions: if it does not match, (char*)q is reset and
10197 the normal format processing code is used.
10199 Currently defined extensions are:
10200 %p include pointer address (standard)
10201 %-p (SVf) include an SV (previously %_)
10202 %-<num>p include an SV with precision <num>
10203 %<num>p reserved for future extensions
10205 Robin Barker 2005-07-14
10207 %1p (VDf) removed. RMB 2007-10-19
10214 n = expect_number(&q);
10216 if (sv) { /* SVf */
10221 argsv = MUTABLE_SV(va_arg(*args, void*));
10222 eptr = SvPV_const(argsv, elen);
10223 if (DO_UTF8(argsv))
10228 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL),
10229 "internal %%<num>p might conflict with future printf extensions");
10235 if ( (width = expect_number(&q)) ) {
10250 if (plus == '+' && *q == ' ') /* '+' over ' ' */
10279 if ( (ewix = expect_number(&q)) )
10288 if ((vectorarg = asterisk)) {
10301 width = expect_number(&q);
10304 if (vectorize && vectorarg) {
10305 /* vectorizing, but not with the default "." */
10307 vecsv = va_arg(*args, SV*);
10309 vecsv = (evix > 0 && evix <= svmax)
10310 ? svargs[evix-1] : S_vcatpvfn_missing_argument(aTHX);
10312 vecsv = svix < svmax
10313 ? svargs[svix++] : S_vcatpvfn_missing_argument(aTHX);
10315 dotstr = SvPV_const(vecsv, dotstrlen);
10316 /* Keep the DO_UTF8 test *after* the SvPV call, else things go
10317 bad with tied or overloaded values that return UTF8. */
10318 if (DO_UTF8(vecsv))
10320 else if (has_utf8) {
10321 vecsv = sv_mortalcopy(vecsv);
10322 sv_utf8_upgrade(vecsv);
10323 dotstr = SvPV_const(vecsv, dotstrlen);
10330 i = va_arg(*args, int);
10332 i = (ewix ? ewix <= svmax : svix < svmax) ?
10333 SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
10335 width = (i < 0) ? -i : i;
10345 if ( ((epix = expect_number(&q))) && (*q++ != '$') )
10347 /* XXX: todo, support specified precision parameter */
10351 i = va_arg(*args, int);
10353 i = (ewix ? ewix <= svmax : svix < svmax)
10354 ? SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
10356 has_precis = !(i < 0);
10360 while (isDIGIT(*q))
10361 precis = precis * 10 + (*q++ - '0');
10370 else if (efix ? (efix > 0 && efix <= svmax) : svix < svmax) {
10371 vecsv = svargs[efix ? efix-1 : svix++];
10372 vecstr = (U8*)SvPV_const(vecsv,veclen);
10373 vec_utf8 = DO_UTF8(vecsv);
10375 /* if this is a version object, we need to convert
10376 * back into v-string notation and then let the
10377 * vectorize happen normally
10379 if (sv_derived_from(vecsv, "version")) {
10380 char *version = savesvpv(vecsv);
10381 if ( hv_exists(MUTABLE_HV(SvRV(vecsv)), "alpha", 5 ) ) {
10382 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
10383 "vector argument not supported with alpha versions");
10386 vecsv = sv_newmortal();
10387 scan_vstring(version, version + veclen, vecsv);
10388 vecstr = (U8*)SvPV_const(vecsv, veclen);
10389 vec_utf8 = DO_UTF8(vecsv);
10403 case 'I': /* Ix, I32x, and I64x */
10405 if (q[1] == '6' && q[2] == '4') {
10411 if (q[1] == '3' && q[2] == '2') {
10421 #if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
10433 #if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
10434 if (*q == 'l') { /* lld, llf */
10443 if (*++q == 'h') { /* hhd, hhu */
10472 if (!vectorize && !args) {
10474 const I32 i = efix-1;
10475 argsv = (i >= 0 && i < svmax)
10476 ? svargs[i] : S_vcatpvfn_missing_argument(aTHX);
10478 argsv = (svix >= 0 && svix < svmax)
10479 ? svargs[svix++] : S_vcatpvfn_missing_argument(aTHX);
10483 switch (c = *q++) {
10490 uv = (args) ? va_arg(*args, int) : SvIV(argsv);
10492 (!UNI_IS_INVARIANT(uv) && SvUTF8(sv)))
10494 eptr = (char*)utf8buf;
10495 elen = uvchr_to_utf8((U8*)eptr, uv) - utf8buf;
10509 eptr = va_arg(*args, char*);
10511 elen = strlen(eptr);
10513 eptr = (char *)nullstr;
10514 elen = sizeof nullstr - 1;
10518 eptr = SvPV_const(argsv, elen);
10519 if (DO_UTF8(argsv)) {
10520 STRLEN old_precis = precis;
10521 if (has_precis && precis < elen) {
10522 STRLEN ulen = sv_len_utf8(argsv);
10523 I32 p = precis > ulen ? ulen : precis;
10524 sv_pos_u2b(argsv, &p, 0); /* sticks at end */
10527 if (width) { /* fudge width (can't fudge elen) */
10528 if (has_precis && precis < elen)
10529 width += precis - old_precis;
10531 width += elen - sv_len_utf8(argsv);
10538 if (has_precis && precis < elen)
10545 if (alt || vectorize)
10547 uv = PTR2UV(args ? va_arg(*args, void*) : argsv);
10568 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
10577 esignbuf[esignlen++] = plus;
10581 case 'c': iv = (char)va_arg(*args, int); break;
10582 case 'h': iv = (short)va_arg(*args, int); break;
10583 case 'l': iv = va_arg(*args, long); break;
10584 case 'V': iv = va_arg(*args, IV); break;
10585 case 'z': iv = va_arg(*args, SSize_t); break;
10586 case 't': iv = va_arg(*args, ptrdiff_t); break;
10587 default: iv = va_arg(*args, int); break;
10589 case 'j': iv = va_arg(*args, intmax_t); break;
10593 iv = va_arg(*args, Quad_t); break;
10600 IV tiv = SvIV(argsv); /* work around GCC bug #13488 */
10602 case 'c': iv = (char)tiv; break;
10603 case 'h': iv = (short)tiv; break;
10604 case 'l': iv = (long)tiv; break;
10606 default: iv = tiv; break;
10609 iv = (Quad_t)tiv; break;
10615 if ( !vectorize ) /* we already set uv above */
10620 esignbuf[esignlen++] = plus;
10624 esignbuf[esignlen++] = '-';
10668 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
10679 case 'c': uv = (unsigned char)va_arg(*args, unsigned); break;
10680 case 'h': uv = (unsigned short)va_arg(*args, unsigned); break;
10681 case 'l': uv = va_arg(*args, unsigned long); break;
10682 case 'V': uv = va_arg(*args, UV); break;
10683 case 'z': uv = va_arg(*args, Size_t); break;
10684 case 't': uv = va_arg(*args, ptrdiff_t); break; /* will sign extend, but there is no uptrdiff_t, so oh well */
10686 case 'j': uv = va_arg(*args, uintmax_t); break;
10688 default: uv = va_arg(*args, unsigned); break;
10691 uv = va_arg(*args, Uquad_t); break;
10698 UV tuv = SvUV(argsv); /* work around GCC bug #13488 */
10700 case 'c': uv = (unsigned char)tuv; break;
10701 case 'h': uv = (unsigned short)tuv; break;
10702 case 'l': uv = (unsigned long)tuv; break;
10704 default: uv = tuv; break;
10707 uv = (Uquad_t)tuv; break;
10716 char *ptr = ebuf + sizeof ebuf;
10717 bool tempalt = uv ? alt : FALSE; /* Vectors can't change alt */
10723 p = (char *)((c == 'X') ? PL_hexdigit + 16 : PL_hexdigit);
10727 } while (uv >>= 4);
10729 esignbuf[esignlen++] = '0';
10730 esignbuf[esignlen++] = c; /* 'x' or 'X' */
10736 *--ptr = '0' + dig;
10737 } while (uv >>= 3);
10738 if (alt && *ptr != '0')
10744 *--ptr = '0' + dig;
10745 } while (uv >>= 1);
10747 esignbuf[esignlen++] = '0';
10748 esignbuf[esignlen++] = c;
10751 default: /* it had better be ten or less */
10754 *--ptr = '0' + dig;
10755 } while (uv /= base);
10758 elen = (ebuf + sizeof ebuf) - ptr;
10762 zeros = precis - elen;
10763 else if (precis == 0 && elen == 1 && *eptr == '0'
10764 && !(base == 8 && alt)) /* "%#.0o" prints "0" */
10767 /* a precision nullifies the 0 flag. */
10774 /* FLOATING POINT */
10777 c = 'f'; /* maybe %F isn't supported here */
10779 case 'e': case 'E':
10781 case 'g': case 'G':
10785 /* This is evil, but floating point is even more evil */
10787 /* for SV-style calling, we can only get NV
10788 for C-style calling, we assume %f is double;
10789 for simplicity we allow any of %Lf, %llf, %qf for long double
10793 #if defined(USE_LONG_DOUBLE)
10797 /* [perl #20339] - we should accept and ignore %lf rather than die */
10801 #if defined(USE_LONG_DOUBLE)
10802 intsize = args ? 0 : 'q';
10806 #if defined(HAS_LONG_DOUBLE)
10819 /* now we need (long double) if intsize == 'q', else (double) */
10821 #if LONG_DOUBLESIZE > DOUBLESIZE
10823 va_arg(*args, long double) :
10824 va_arg(*args, double)
10826 va_arg(*args, double)
10831 /* nv * 0 will be NaN for NaN, +Inf and -Inf, and 0 for anything
10832 else. frexp() has some unspecified behaviour for those three */
10833 if (c != 'e' && c != 'E' && (nv * 0) == 0) {
10835 /* FIXME: if HAS_LONG_DOUBLE but not USE_LONG_DOUBLE this
10836 will cast our (long double) to (double) */
10837 (void)Perl_frexp(nv, &i);
10838 if (i == PERL_INT_MIN)
10839 Perl_die(aTHX_ "panic: frexp");
10841 need = BIT_DIGITS(i);
10843 need += has_precis ? precis : 6; /* known default */
10848 #ifdef HAS_LDBL_SPRINTF_BUG
10849 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
10850 with sfio - Allen <allens@cpan.org> */
10853 # define MY_DBL_MAX DBL_MAX
10854 # else /* XXX guessing! HUGE_VAL may be defined as infinity, so not using */
10855 # if DOUBLESIZE >= 8
10856 # define MY_DBL_MAX 1.7976931348623157E+308L
10858 # define MY_DBL_MAX 3.40282347E+38L
10862 # ifdef HAS_LDBL_SPRINTF_BUG_LESS1 /* only between -1L & 1L - Allen */
10863 # define MY_DBL_MAX_BUG 1L
10865 # define MY_DBL_MAX_BUG MY_DBL_MAX
10869 # define MY_DBL_MIN DBL_MIN
10870 # else /* XXX guessing! -Allen */
10871 # if DOUBLESIZE >= 8
10872 # define MY_DBL_MIN 2.2250738585072014E-308L
10874 # define MY_DBL_MIN 1.17549435E-38L
10878 if ((intsize == 'q') && (c == 'f') &&
10879 ((nv < MY_DBL_MAX_BUG) && (nv > -MY_DBL_MAX_BUG)) &&
10880 (need < DBL_DIG)) {
10881 /* it's going to be short enough that
10882 * long double precision is not needed */
10884 if ((nv <= 0L) && (nv >= -0L))
10885 fix_ldbl_sprintf_bug = TRUE; /* 0 is 0 - easiest */
10887 /* would use Perl_fp_class as a double-check but not
10888 * functional on IRIX - see perl.h comments */
10890 if ((nv >= MY_DBL_MIN) || (nv <= -MY_DBL_MIN)) {
10891 /* It's within the range that a double can represent */
10892 #if defined(DBL_MAX) && !defined(DBL_MIN)
10893 if ((nv >= ((long double)1/DBL_MAX)) ||
10894 (nv <= (-(long double)1/DBL_MAX)))
10896 fix_ldbl_sprintf_bug = TRUE;
10899 if (fix_ldbl_sprintf_bug == TRUE) {
10909 # undef MY_DBL_MAX_BUG
10912 #endif /* HAS_LDBL_SPRINTF_BUG */
10914 need += 20; /* fudge factor */
10915 if (PL_efloatsize < need) {
10916 Safefree(PL_efloatbuf);
10917 PL_efloatsize = need + 20; /* more fudge */
10918 Newx(PL_efloatbuf, PL_efloatsize, char);
10919 PL_efloatbuf[0] = '\0';
10922 if ( !(width || left || plus || alt) && fill != '0'
10923 && has_precis && intsize != 'q' ) { /* Shortcuts */
10924 /* See earlier comment about buggy Gconvert when digits,
10926 if ( c == 'g' && precis) {
10927 Gconvert((NV)nv, (int)precis, 0, PL_efloatbuf);
10928 /* May return an empty string for digits==0 */
10929 if (*PL_efloatbuf) {
10930 elen = strlen(PL_efloatbuf);
10931 goto float_converted;
10933 } else if ( c == 'f' && !precis) {
10934 if ((eptr = F0convert(nv, ebuf + sizeof ebuf, &elen)))
10939 char *ptr = ebuf + sizeof ebuf;
10942 /* FIXME: what to do if HAS_LONG_DOUBLE but not PERL_PRIfldbl? */
10943 #if defined(HAS_LONG_DOUBLE) && defined(PERL_PRIfldbl)
10944 if (intsize == 'q') {
10945 /* Copy the one or more characters in a long double
10946 * format before the 'base' ([efgEFG]) character to
10947 * the format string. */
10948 static char const prifldbl[] = PERL_PRIfldbl;
10949 char const *p = prifldbl + sizeof(prifldbl) - 3;
10950 while (p >= prifldbl) { *--ptr = *p--; }
10955 do { *--ptr = '0' + (base % 10); } while (base /= 10);
10960 do { *--ptr = '0' + (base % 10); } while (base /= 10);
10972 /* No taint. Otherwise we are in the strange situation
10973 * where printf() taints but print($float) doesn't.
10975 #if defined(HAS_LONG_DOUBLE)
10976 elen = ((intsize == 'q')
10977 ? my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, nv)
10978 : my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, (double)nv));
10980 elen = my_sprintf(PL_efloatbuf, ptr, nv);
10984 eptr = PL_efloatbuf;
10992 i = SvCUR(sv) - origlen;
10995 case 'c': *(va_arg(*args, char*)) = i; break;
10996 case 'h': *(va_arg(*args, short*)) = i; break;
10997 default: *(va_arg(*args, int*)) = i; break;
10998 case 'l': *(va_arg(*args, long*)) = i; break;
10999 case 'V': *(va_arg(*args, IV*)) = i; break;
11000 case 'z': *(va_arg(*args, SSize_t*)) = i; break;
11001 case 't': *(va_arg(*args, ptrdiff_t*)) = i; break;
11003 case 'j': *(va_arg(*args, intmax_t*)) = i; break;
11007 *(va_arg(*args, Quad_t*)) = i; break;
11014 sv_setuv_mg(argsv, (UV)i);
11015 continue; /* not "break" */
11022 && (PL_op->op_type == OP_PRTF || PL_op->op_type == OP_SPRINTF)
11023 && ckWARN(WARN_PRINTF))
11025 SV * const msg = sv_newmortal();
11026 Perl_sv_setpvf(aTHX_ msg, "Invalid conversion in %sprintf: ",
11027 (PL_op->op_type == OP_PRTF) ? "" : "s");
11028 if (fmtstart < patend) {
11029 const char * const fmtend = q < patend ? q : patend;
11031 sv_catpvs(msg, "\"%");
11032 for (f = fmtstart; f < fmtend; f++) {
11034 sv_catpvn(msg, f, 1);
11036 Perl_sv_catpvf(aTHX_ msg,
11037 "\\%03"UVof, (UV)*f & 0xFF);
11040 sv_catpvs(msg, "\"");
11042 sv_catpvs(msg, "end of string");
11044 Perl_warner(aTHX_ packWARN(WARN_PRINTF), "%"SVf, SVfARG(msg)); /* yes, this is reentrant */
11047 /* output mangled stuff ... */
11053 /* ... right here, because formatting flags should not apply */
11054 SvGROW(sv, SvCUR(sv) + elen + 1);
11056 Copy(eptr, p, elen, char);
11059 SvCUR_set(sv, p - SvPVX_const(sv));
11061 continue; /* not "break" */
11064 if (is_utf8 != has_utf8) {
11067 sv_utf8_upgrade(sv);
11070 const STRLEN old_elen = elen;
11071 SV * const nsv = newSVpvn_flags(eptr, elen, SVs_TEMP);
11072 sv_utf8_upgrade(nsv);
11073 eptr = SvPVX_const(nsv);
11076 if (width) { /* fudge width (can't fudge elen) */
11077 width += elen - old_elen;
11083 have = esignlen + zeros + elen;
11085 Perl_croak_nocontext("%s", PL_memory_wrap);
11087 need = (have > width ? have : width);
11090 if (need >= (((STRLEN)~0) - SvCUR(sv) - dotstrlen - 1))
11091 Perl_croak_nocontext("%s", PL_memory_wrap);
11092 SvGROW(sv, SvCUR(sv) + need + dotstrlen + 1);
11094 if (esignlen && fill == '0') {
11096 for (i = 0; i < (int)esignlen; i++)
11097 *p++ = esignbuf[i];
11099 if (gap && !left) {
11100 memset(p, fill, gap);
11103 if (esignlen && fill != '0') {
11105 for (i = 0; i < (int)esignlen; i++)
11106 *p++ = esignbuf[i];
11110 for (i = zeros; i; i--)
11114 Copy(eptr, p, elen, char);
11118 memset(p, ' ', gap);
11123 Copy(dotstr, p, dotstrlen, char);
11127 vectorize = FALSE; /* done iterating over vecstr */
11134 SvCUR_set(sv, p - SvPVX_const(sv));
11143 /* =========================================================================
11145 =head1 Cloning an interpreter
11147 All the macros and functions in this section are for the private use of
11148 the main function, perl_clone().
11150 The foo_dup() functions make an exact copy of an existing foo thingy.
11151 During the course of a cloning, a hash table is used to map old addresses
11152 to new addresses. The table is created and manipulated with the
11153 ptr_table_* functions.
11157 * =========================================================================*/
11160 #if defined(USE_ITHREADS)
11162 /* XXX Remove this so it doesn't have to go thru the macro and return for nothing */
11163 #ifndef GpREFCNT_inc
11164 # define GpREFCNT_inc(gp) ((gp) ? (++(gp)->gp_refcnt, (gp)) : (GP*)NULL)
11168 /* Certain cases in Perl_ss_dup have been merged, by relying on the fact
11169 that currently av_dup, gv_dup and hv_dup are the same as sv_dup.
11170 If this changes, please unmerge ss_dup.
11171 Likewise, sv_dup_inc_multiple() relies on this fact. */
11172 #define sv_dup_inc_NN(s,t) SvREFCNT_inc_NN(sv_dup_inc(s,t))
11173 #define av_dup(s,t) MUTABLE_AV(sv_dup((const SV *)s,t))
11174 #define av_dup_inc(s,t) MUTABLE_AV(sv_dup_inc((const SV *)s,t))
11175 #define hv_dup(s,t) MUTABLE_HV(sv_dup((const SV *)s,t))
11176 #define hv_dup_inc(s,t) MUTABLE_HV(sv_dup_inc((const SV *)s,t))
11177 #define cv_dup(s,t) MUTABLE_CV(sv_dup((const SV *)s,t))
11178 #define cv_dup_inc(s,t) MUTABLE_CV(sv_dup_inc((const SV *)s,t))
11179 #define io_dup(s,t) MUTABLE_IO(sv_dup((const SV *)s,t))
11180 #define io_dup_inc(s,t) MUTABLE_IO(sv_dup_inc((const SV *)s,t))
11181 #define gv_dup(s,t) MUTABLE_GV(sv_dup((const SV *)s,t))
11182 #define gv_dup_inc(s,t) MUTABLE_GV(sv_dup_inc((const SV *)s,t))
11183 #define SAVEPV(p) ((p) ? savepv(p) : NULL)
11184 #define SAVEPVN(p,n) ((p) ? savepvn(p,n) : NULL)
11186 /* clone a parser */
11189 Perl_parser_dup(pTHX_ const yy_parser *const proto, CLONE_PARAMS *const param)
11193 PERL_ARGS_ASSERT_PARSER_DUP;
11198 /* look for it in the table first */
11199 parser = (yy_parser *)ptr_table_fetch(PL_ptr_table, proto);
11203 /* create anew and remember what it is */
11204 Newxz(parser, 1, yy_parser);
11205 ptr_table_store(PL_ptr_table, proto, parser);
11207 /* XXX these not yet duped */
11208 parser->old_parser = NULL;
11209 parser->stack = NULL;
11211 parser->stack_size = 0;
11212 /* XXX parser->stack->state = 0; */
11214 /* XXX eventually, just Copy() most of the parser struct ? */
11216 parser->lex_brackets = proto->lex_brackets;
11217 parser->lex_casemods = proto->lex_casemods;
11218 parser->lex_brackstack = savepvn(proto->lex_brackstack,
11219 (proto->lex_brackets < 120 ? 120 : proto->lex_brackets));
11220 parser->lex_casestack = savepvn(proto->lex_casestack,
11221 (proto->lex_casemods < 12 ? 12 : proto->lex_casemods));
11222 parser->lex_defer = proto->lex_defer;
11223 parser->lex_dojoin = proto->lex_dojoin;
11224 parser->lex_expect = proto->lex_expect;
11225 parser->lex_formbrack = proto->lex_formbrack;
11226 parser->lex_inpat = proto->lex_inpat;
11227 parser->lex_inwhat = proto->lex_inwhat;
11228 parser->lex_op = proto->lex_op;
11229 parser->lex_repl = sv_dup_inc(proto->lex_repl, param);
11230 parser->lex_starts = proto->lex_starts;
11231 parser->lex_stuff = sv_dup_inc(proto->lex_stuff, param);
11232 parser->multi_close = proto->multi_close;
11233 parser->multi_open = proto->multi_open;
11234 parser->multi_start = proto->multi_start;
11235 parser->multi_end = proto->multi_end;
11236 parser->pending_ident = proto->pending_ident;
11237 parser->preambled = proto->preambled;
11238 parser->sublex_info = proto->sublex_info; /* XXX not quite right */
11239 parser->linestr = sv_dup_inc(proto->linestr, param);
11240 parser->expect = proto->expect;
11241 parser->copline = proto->copline;
11242 parser->last_lop_op = proto->last_lop_op;
11243 parser->lex_state = proto->lex_state;
11244 parser->rsfp = fp_dup(proto->rsfp, '<', param);
11245 /* rsfp_filters entries have fake IoDIRP() */
11246 parser->rsfp_filters= av_dup_inc(proto->rsfp_filters, param);
11247 parser->in_my = proto->in_my;
11248 parser->in_my_stash = hv_dup(proto->in_my_stash, param);
11249 parser->error_count = proto->error_count;
11252 parser->linestr = sv_dup_inc(proto->linestr, param);
11255 char * const ols = SvPVX(proto->linestr);
11256 char * const ls = SvPVX(parser->linestr);
11258 parser->bufptr = ls + (proto->bufptr >= ols ?
11259 proto->bufptr - ols : 0);
11260 parser->oldbufptr = ls + (proto->oldbufptr >= ols ?
11261 proto->oldbufptr - ols : 0);
11262 parser->oldoldbufptr= ls + (proto->oldoldbufptr >= ols ?
11263 proto->oldoldbufptr - ols : 0);
11264 parser->linestart = ls + (proto->linestart >= ols ?
11265 proto->linestart - ols : 0);
11266 parser->last_uni = ls + (proto->last_uni >= ols ?
11267 proto->last_uni - ols : 0);
11268 parser->last_lop = ls + (proto->last_lop >= ols ?
11269 proto->last_lop - ols : 0);
11271 parser->bufend = ls + SvCUR(parser->linestr);
11274 Copy(proto->tokenbuf, parser->tokenbuf, 256, char);
11278 parser->endwhite = proto->endwhite;
11279 parser->faketokens = proto->faketokens;
11280 parser->lasttoke = proto->lasttoke;
11281 parser->nextwhite = proto->nextwhite;
11282 parser->realtokenstart = proto->realtokenstart;
11283 parser->skipwhite = proto->skipwhite;
11284 parser->thisclose = proto->thisclose;
11285 parser->thismad = proto->thismad;
11286 parser->thisopen = proto->thisopen;
11287 parser->thisstuff = proto->thisstuff;
11288 parser->thistoken = proto->thistoken;
11289 parser->thiswhite = proto->thiswhite;
11291 Copy(proto->nexttoke, parser->nexttoke, 5, NEXTTOKE);
11292 parser->curforce = proto->curforce;
11294 Copy(proto->nextval, parser->nextval, 5, YYSTYPE);
11295 Copy(proto->nexttype, parser->nexttype, 5, I32);
11296 parser->nexttoke = proto->nexttoke;
11299 /* XXX should clone saved_curcop here, but we aren't passed
11300 * proto_perl; so do it in perl_clone_using instead */
11306 /* duplicate a file handle */
11309 Perl_fp_dup(pTHX_ PerlIO *const fp, const char type, CLONE_PARAMS *const param)
11313 PERL_ARGS_ASSERT_FP_DUP;
11314 PERL_UNUSED_ARG(type);
11317 return (PerlIO*)NULL;
11319 /* look for it in the table first */
11320 ret = (PerlIO*)ptr_table_fetch(PL_ptr_table, fp);
11324 /* create anew and remember what it is */
11325 ret = PerlIO_fdupopen(aTHX_ fp, param, PERLIO_DUP_CLONE);
11326 ptr_table_store(PL_ptr_table, fp, ret);
11330 /* duplicate a directory handle */
11333 Perl_dirp_dup(pTHX_ DIR *const dp, CLONE_PARAMS *const param)
11339 register const Direntry_t *dirent;
11340 char smallbuf[256];
11346 PERL_UNUSED_CONTEXT;
11347 PERL_ARGS_ASSERT_DIRP_DUP;
11352 /* look for it in the table first */
11353 ret = (DIR*)ptr_table_fetch(PL_ptr_table, dp);
11359 PERL_UNUSED_ARG(param);
11363 /* open the current directory (so we can switch back) */
11364 if (!(pwd = PerlDir_open("."))) return (DIR *)NULL;
11366 /* chdir to our dir handle and open the present working directory */
11367 if (fchdir(my_dirfd(dp)) < 0 || !(ret = PerlDir_open("."))) {
11368 PerlDir_close(pwd);
11369 return (DIR *)NULL;
11371 /* Now we should have two dir handles pointing to the same dir. */
11373 /* Be nice to the calling code and chdir back to where we were. */
11374 fchdir(my_dirfd(pwd)); /* If this fails, then what? */
11376 /* We have no need of the pwd handle any more. */
11377 PerlDir_close(pwd);
11380 # define d_namlen(d) (d)->d_namlen
11382 # define d_namlen(d) strlen((d)->d_name)
11384 /* Iterate once through dp, to get the file name at the current posi-
11385 tion. Then step back. */
11386 pos = PerlDir_tell(dp);
11387 if ((dirent = PerlDir_read(dp))) {
11388 len = d_namlen(dirent);
11389 if (len <= sizeof smallbuf) name = smallbuf;
11390 else Newx(name, len, char);
11391 Move(dirent->d_name, name, len, char);
11393 PerlDir_seek(dp, pos);
11395 /* Iterate through the new dir handle, till we find a file with the
11397 if (!dirent) /* just before the end */
11399 pos = PerlDir_tell(ret);
11400 if (PerlDir_read(ret)) continue; /* not there yet */
11401 PerlDir_seek(ret, pos); /* step back */
11405 const long pos0 = PerlDir_tell(ret);
11407 pos = PerlDir_tell(ret);
11408 if ((dirent = PerlDir_read(ret))) {
11409 if (len == d_namlen(dirent)
11410 && memEQ(name, dirent->d_name, len)) {
11412 PerlDir_seek(ret, pos); /* step back */
11415 /* else we are not there yet; keep iterating */
11417 else { /* This is not meant to happen. The best we can do is
11418 reset the iterator to the beginning. */
11419 PerlDir_seek(ret, pos0);
11426 if (name && name != smallbuf)
11431 ret = win32_dirp_dup(dp, param);
11434 /* pop it in the pointer table */
11436 ptr_table_store(PL_ptr_table, dp, ret);
11441 /* duplicate a typeglob */
11444 Perl_gp_dup(pTHX_ GP *const gp, CLONE_PARAMS *const param)
11448 PERL_ARGS_ASSERT_GP_DUP;
11452 /* look for it in the table first */
11453 ret = (GP*)ptr_table_fetch(PL_ptr_table, gp);
11457 /* create anew and remember what it is */
11459 ptr_table_store(PL_ptr_table, gp, ret);
11462 /* ret->gp_refcnt must be 0 before any other dups are called. We're relying
11463 on Newxz() to do this for us. */
11464 ret->gp_sv = sv_dup_inc(gp->gp_sv, param);
11465 ret->gp_io = io_dup_inc(gp->gp_io, param);
11466 ret->gp_form = cv_dup_inc(gp->gp_form, param);
11467 ret->gp_av = av_dup_inc(gp->gp_av, param);
11468 ret->gp_hv = hv_dup_inc(gp->gp_hv, param);
11469 ret->gp_egv = gv_dup(gp->gp_egv, param);/* GvEGV is not refcounted */
11470 ret->gp_cv = cv_dup_inc(gp->gp_cv, param);
11471 ret->gp_cvgen = gp->gp_cvgen;
11472 ret->gp_line = gp->gp_line;
11473 ret->gp_file_hek = hek_dup(gp->gp_file_hek, param);
11477 /* duplicate a chain of magic */
11480 Perl_mg_dup(pTHX_ MAGIC *mg, CLONE_PARAMS *const param)
11482 MAGIC *mgret = NULL;
11483 MAGIC **mgprev_p = &mgret;
11485 PERL_ARGS_ASSERT_MG_DUP;
11487 for (; mg; mg = mg->mg_moremagic) {
11490 if ((param->flags & CLONEf_JOIN_IN)
11491 && mg->mg_type == PERL_MAGIC_backref)
11492 /* when joining, we let the individual SVs add themselves to
11493 * backref as needed. */
11496 Newx(nmg, 1, MAGIC);
11498 mgprev_p = &(nmg->mg_moremagic);
11500 /* There was a comment "XXX copy dynamic vtable?" but as we don't have
11501 dynamic vtables, I'm not sure why Sarathy wrote it. The comment dates
11502 from the original commit adding Perl_mg_dup() - revision 4538.
11503 Similarly there is the annotation "XXX random ptr?" next to the
11504 assignment to nmg->mg_ptr. */
11507 /* FIXME for plugins
11508 if (nmg->mg_type == PERL_MAGIC_qr) {
11509 nmg->mg_obj = MUTABLE_SV(CALLREGDUPE((REGEXP*)nmg->mg_obj, param));
11513 nmg->mg_obj = (nmg->mg_flags & MGf_REFCOUNTED)
11514 ? nmg->mg_type == PERL_MAGIC_backref
11515 /* The backref AV has its reference
11516 * count deliberately bumped by 1 */
11517 ? SvREFCNT_inc(av_dup_inc((const AV *)
11518 nmg->mg_obj, param))
11519 : sv_dup_inc(nmg->mg_obj, param)
11520 : sv_dup(nmg->mg_obj, param);
11522 if (nmg->mg_ptr && nmg->mg_type != PERL_MAGIC_regex_global) {
11523 if (nmg->mg_len > 0) {
11524 nmg->mg_ptr = SAVEPVN(nmg->mg_ptr, nmg->mg_len);
11525 if (nmg->mg_type == PERL_MAGIC_overload_table &&
11526 AMT_AMAGIC((AMT*)nmg->mg_ptr))
11528 AMT * const namtp = (AMT*)nmg->mg_ptr;
11529 sv_dup_inc_multiple((SV**)(namtp->table),
11530 (SV**)(namtp->table), NofAMmeth, param);
11533 else if (nmg->mg_len == HEf_SVKEY)
11534 nmg->mg_ptr = (char*)sv_dup_inc((const SV *)nmg->mg_ptr, param);
11536 if ((nmg->mg_flags & MGf_DUP) && nmg->mg_virtual && nmg->mg_virtual->svt_dup) {
11537 nmg->mg_virtual->svt_dup(aTHX_ nmg, param);
11543 #endif /* USE_ITHREADS */
11545 struct ptr_tbl_arena {
11546 struct ptr_tbl_arena *next;
11547 struct ptr_tbl_ent array[1023/3]; /* as ptr_tbl_ent has 3 pointers. */
11550 /* create a new pointer-mapping table */
11553 Perl_ptr_table_new(pTHX)
11556 PERL_UNUSED_CONTEXT;
11558 Newx(tbl, 1, PTR_TBL_t);
11559 tbl->tbl_max = 511;
11560 tbl->tbl_items = 0;
11561 tbl->tbl_arena = NULL;
11562 tbl->tbl_arena_next = NULL;
11563 tbl->tbl_arena_end = NULL;
11564 Newxz(tbl->tbl_ary, tbl->tbl_max + 1, PTR_TBL_ENT_t*);
11568 #define PTR_TABLE_HASH(ptr) \
11569 ((PTR2UV(ptr) >> 3) ^ (PTR2UV(ptr) >> (3 + 7)) ^ (PTR2UV(ptr) >> (3 + 17)))
11571 /* map an existing pointer using a table */
11573 STATIC PTR_TBL_ENT_t *
11574 S_ptr_table_find(PTR_TBL_t *const tbl, const void *const sv)
11576 PTR_TBL_ENT_t *tblent;
11577 const UV hash = PTR_TABLE_HASH(sv);
11579 PERL_ARGS_ASSERT_PTR_TABLE_FIND;
11581 tblent = tbl->tbl_ary[hash & tbl->tbl_max];
11582 for (; tblent; tblent = tblent->next) {
11583 if (tblent->oldval == sv)
11590 Perl_ptr_table_fetch(pTHX_ PTR_TBL_t *const tbl, const void *const sv)
11592 PTR_TBL_ENT_t const *const tblent = ptr_table_find(tbl, sv);
11594 PERL_ARGS_ASSERT_PTR_TABLE_FETCH;
11595 PERL_UNUSED_CONTEXT;
11597 return tblent ? tblent->newval : NULL;
11600 /* add a new entry to a pointer-mapping table */
11603 Perl_ptr_table_store(pTHX_ PTR_TBL_t *const tbl, const void *const oldsv, void *const newsv)
11605 PTR_TBL_ENT_t *tblent = ptr_table_find(tbl, oldsv);
11607 PERL_ARGS_ASSERT_PTR_TABLE_STORE;
11608 PERL_UNUSED_CONTEXT;
11611 tblent->newval = newsv;
11613 const UV entry = PTR_TABLE_HASH(oldsv) & tbl->tbl_max;
11615 if (tbl->tbl_arena_next == tbl->tbl_arena_end) {
11616 struct ptr_tbl_arena *new_arena;
11618 Newx(new_arena, 1, struct ptr_tbl_arena);
11619 new_arena->next = tbl->tbl_arena;
11620 tbl->tbl_arena = new_arena;
11621 tbl->tbl_arena_next = new_arena->array;
11622 tbl->tbl_arena_end = new_arena->array
11623 + sizeof(new_arena->array) / sizeof(new_arena->array[0]);
11626 tblent = tbl->tbl_arena_next++;
11628 tblent->oldval = oldsv;
11629 tblent->newval = newsv;
11630 tblent->next = tbl->tbl_ary[entry];
11631 tbl->tbl_ary[entry] = tblent;
11633 if (tblent->next && tbl->tbl_items > tbl->tbl_max)
11634 ptr_table_split(tbl);
11638 /* double the hash bucket size of an existing ptr table */
11641 Perl_ptr_table_split(pTHX_ PTR_TBL_t *const tbl)
11643 PTR_TBL_ENT_t **ary = tbl->tbl_ary;
11644 const UV oldsize = tbl->tbl_max + 1;
11645 UV newsize = oldsize * 2;
11648 PERL_ARGS_ASSERT_PTR_TABLE_SPLIT;
11649 PERL_UNUSED_CONTEXT;
11651 Renew(ary, newsize, PTR_TBL_ENT_t*);
11652 Zero(&ary[oldsize], newsize-oldsize, PTR_TBL_ENT_t*);
11653 tbl->tbl_max = --newsize;
11654 tbl->tbl_ary = ary;
11655 for (i=0; i < oldsize; i++, ary++) {
11656 PTR_TBL_ENT_t **entp = ary;
11657 PTR_TBL_ENT_t *ent = *ary;
11658 PTR_TBL_ENT_t **curentp;
11661 curentp = ary + oldsize;
11663 if ((newsize & PTR_TABLE_HASH(ent->oldval)) != i) {
11665 ent->next = *curentp;
11675 /* remove all the entries from a ptr table */
11676 /* Deprecated - will be removed post 5.14 */
11679 Perl_ptr_table_clear(pTHX_ PTR_TBL_t *const tbl)
11681 if (tbl && tbl->tbl_items) {
11682 struct ptr_tbl_arena *arena = tbl->tbl_arena;
11684 Zero(tbl->tbl_ary, tbl->tbl_max + 1, struct ptr_tbl_ent **);
11687 struct ptr_tbl_arena *next = arena->next;
11693 tbl->tbl_items = 0;
11694 tbl->tbl_arena = NULL;
11695 tbl->tbl_arena_next = NULL;
11696 tbl->tbl_arena_end = NULL;
11700 /* clear and free a ptr table */
11703 Perl_ptr_table_free(pTHX_ PTR_TBL_t *const tbl)
11705 struct ptr_tbl_arena *arena;
11711 arena = tbl->tbl_arena;
11714 struct ptr_tbl_arena *next = arena->next;
11720 Safefree(tbl->tbl_ary);
11724 #if defined(USE_ITHREADS)
11727 Perl_rvpv_dup(pTHX_ SV *const dstr, const SV *const sstr, CLONE_PARAMS *const param)
11729 PERL_ARGS_ASSERT_RVPV_DUP;
11732 if (SvWEAKREF(sstr)) {
11733 SvRV_set(dstr, sv_dup(SvRV_const(sstr), param));
11734 if (param->flags & CLONEf_JOIN_IN) {
11735 /* if joining, we add any back references individually rather
11736 * than copying the whole backref array */
11737 Perl_sv_add_backref(aTHX_ SvRV(dstr), dstr);
11741 SvRV_set(dstr, sv_dup_inc(SvRV_const(sstr), param));
11743 else if (SvPVX_const(sstr)) {
11744 /* Has something there */
11746 /* Normal PV - clone whole allocated space */
11747 SvPV_set(dstr, SAVEPVN(SvPVX_const(sstr), SvLEN(sstr)-1));
11748 if (SvREADONLY(sstr) && SvFAKE(sstr)) {
11749 /* Not that normal - actually sstr is copy on write.
11750 But we are a true, independent SV, so: */
11751 SvREADONLY_off(dstr);
11756 /* Special case - not normally malloced for some reason */
11757 if (isGV_with_GP(sstr)) {
11758 /* Don't need to do anything here. */
11760 else if ((SvREADONLY(sstr) && SvFAKE(sstr))) {
11761 /* A "shared" PV - clone it as "shared" PV */
11763 HEK_KEY(hek_dup(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)),
11767 /* Some other special case - random pointer */
11768 SvPV_set(dstr, (char *) SvPVX_const(sstr));
11773 /* Copy the NULL */
11774 SvPV_set(dstr, NULL);
11778 /* duplicate a list of SVs. source and dest may point to the same memory. */
11780 S_sv_dup_inc_multiple(pTHX_ SV *const *source, SV **dest,
11781 SSize_t items, CLONE_PARAMS *const param)
11783 PERL_ARGS_ASSERT_SV_DUP_INC_MULTIPLE;
11785 while (items-- > 0) {
11786 *dest++ = sv_dup_inc(*source++, param);
11792 /* duplicate an SV of any type (including AV, HV etc) */
11795 S_sv_dup_common(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
11800 PERL_ARGS_ASSERT_SV_DUP_COMMON;
11802 if (SvTYPE(sstr) == SVTYPEMASK) {
11803 #ifdef DEBUG_LEAKING_SCALARS_ABORT
11808 /* look for it in the table first */
11809 dstr = MUTABLE_SV(ptr_table_fetch(PL_ptr_table, sstr));
11813 if(param->flags & CLONEf_JOIN_IN) {
11814 /** We are joining here so we don't want do clone
11815 something that is bad **/
11816 if (SvTYPE(sstr) == SVt_PVHV) {
11817 const HEK * const hvname = HvNAME_HEK(sstr);
11819 /** don't clone stashes if they already exist **/
11820 dstr = MUTABLE_SV(gv_stashpvn(HEK_KEY(hvname), HEK_LEN(hvname), 0));
11821 ptr_table_store(PL_ptr_table, sstr, dstr);
11827 /* create anew and remember what it is */
11830 #ifdef DEBUG_LEAKING_SCALARS
11831 dstr->sv_debug_optype = sstr->sv_debug_optype;
11832 dstr->sv_debug_line = sstr->sv_debug_line;
11833 dstr->sv_debug_inpad = sstr->sv_debug_inpad;
11834 dstr->sv_debug_parent = (SV*)sstr;
11835 FREE_SV_DEBUG_FILE(dstr);
11836 dstr->sv_debug_file = savepv(sstr->sv_debug_file);
11839 ptr_table_store(PL_ptr_table, sstr, dstr);
11842 SvFLAGS(dstr) = SvFLAGS(sstr);
11843 SvFLAGS(dstr) &= ~SVf_OOK; /* don't propagate OOK hack */
11844 SvREFCNT(dstr) = 0; /* must be before any other dups! */
11847 if (SvANY(sstr) && PL_watch_pvx && SvPVX_const(sstr) == PL_watch_pvx)
11848 PerlIO_printf(Perl_debug_log, "watch at %p hit, found string \"%s\"\n",
11849 (void*)PL_watch_pvx, SvPVX_const(sstr));
11852 /* don't clone objects whose class has asked us not to */
11853 if (SvOBJECT(sstr) && ! (SvFLAGS(SvSTASH(sstr)) & SVphv_CLONEABLE)) {
11858 switch (SvTYPE(sstr)) {
11860 SvANY(dstr) = NULL;
11863 SvANY(dstr) = (XPVIV*)((char*)&(dstr->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
11865 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
11867 SvIV_set(dstr, SvIVX(sstr));
11871 SvANY(dstr) = new_XNV();
11872 SvNV_set(dstr, SvNVX(sstr));
11874 /* case SVt_BIND: */
11877 /* These are all the types that need complex bodies allocating. */
11879 const svtype sv_type = SvTYPE(sstr);
11880 const struct body_details *const sv_type_details
11881 = bodies_by_type + sv_type;
11885 Perl_croak(aTHX_ "Bizarre SvTYPE [%" IVdf "]", (IV)SvTYPE(sstr));
11900 assert(sv_type_details->body_size);
11901 if (sv_type_details->arena) {
11902 new_body_inline(new_body, sv_type);
11904 = (void*)((char*)new_body - sv_type_details->offset);
11906 new_body = new_NOARENA(sv_type_details);
11910 SvANY(dstr) = new_body;
11913 Copy(((char*)SvANY(sstr)) + sv_type_details->offset,
11914 ((char*)SvANY(dstr)) + sv_type_details->offset,
11915 sv_type_details->copy, char);
11917 Copy(((char*)SvANY(sstr)),
11918 ((char*)SvANY(dstr)),
11919 sv_type_details->body_size + sv_type_details->offset, char);
11922 if (sv_type != SVt_PVAV && sv_type != SVt_PVHV
11923 && !isGV_with_GP(dstr)
11924 && !(sv_type == SVt_PVIO && !(IoFLAGS(dstr) & IOf_FAKE_DIRP)))
11925 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
11927 /* The Copy above means that all the source (unduplicated) pointers
11928 are now in the destination. We can check the flags and the
11929 pointers in either, but it's possible that there's less cache
11930 missing by always going for the destination.
11931 FIXME - instrument and check that assumption */
11932 if (sv_type >= SVt_PVMG) {
11933 if ((sv_type == SVt_PVMG) && SvPAD_OUR(dstr)) {
11934 SvOURSTASH_set(dstr, hv_dup_inc(SvOURSTASH(dstr), param));
11935 } else if (SvMAGIC(dstr))
11936 SvMAGIC_set(dstr, mg_dup(SvMAGIC(dstr), param));
11938 SvSTASH_set(dstr, hv_dup_inc(SvSTASH(dstr), param));
11941 /* The cast silences a GCC warning about unhandled types. */
11942 switch ((int)sv_type) {
11952 /* FIXME for plugins */
11953 re_dup_guts((REGEXP*) sstr, (REGEXP*) dstr, param);
11956 /* XXX LvTARGOFF sometimes holds PMOP* when DEBUGGING */
11957 if (LvTYPE(dstr) == 't') /* for tie: unrefcnted fake (SV**) */
11958 LvTARG(dstr) = dstr;
11959 else if (LvTYPE(dstr) == 'T') /* for tie: fake HE */
11960 LvTARG(dstr) = MUTABLE_SV(he_dup((HE*)LvTARG(dstr), 0, param));
11962 LvTARG(dstr) = sv_dup_inc(LvTARG(dstr), param);
11964 /* non-GP case already handled above */
11965 if(isGV_with_GP(sstr)) {
11966 GvNAME_HEK(dstr) = hek_dup(GvNAME_HEK(dstr), param);
11967 /* Don't call sv_add_backref here as it's going to be
11968 created as part of the magic cloning of the symbol
11969 table--unless this is during a join and the stash
11970 is not actually being cloned. */
11971 /* Danger Will Robinson - GvGP(dstr) isn't initialised
11972 at the point of this comment. */
11973 GvSTASH(dstr) = hv_dup(GvSTASH(dstr), param);
11974 if (param->flags & CLONEf_JOIN_IN)
11975 Perl_sv_add_backref(aTHX_ MUTABLE_SV(GvSTASH(dstr)), dstr);
11976 GvGP_set(dstr, gp_dup(GvGP(sstr), param));
11977 (void)GpREFCNT_inc(GvGP(dstr));
11981 /* PL_parser->rsfp_filters entries have fake IoDIRP() */
11982 if(IoFLAGS(dstr) & IOf_FAKE_DIRP) {
11983 /* I have no idea why fake dirp (rsfps)
11984 should be treated differently but otherwise
11985 we end up with leaks -- sky*/
11986 IoTOP_GV(dstr) = gv_dup_inc(IoTOP_GV(dstr), param);
11987 IoFMT_GV(dstr) = gv_dup_inc(IoFMT_GV(dstr), param);
11988 IoBOTTOM_GV(dstr) = gv_dup_inc(IoBOTTOM_GV(dstr), param);
11990 IoTOP_GV(dstr) = gv_dup(IoTOP_GV(dstr), param);
11991 IoFMT_GV(dstr) = gv_dup(IoFMT_GV(dstr), param);
11992 IoBOTTOM_GV(dstr) = gv_dup(IoBOTTOM_GV(dstr), param);
11993 if (IoDIRP(dstr)) {
11994 IoDIRP(dstr) = dirp_dup(IoDIRP(dstr), param);
11997 /* IoDIRP(dstr) is already a copy of IoDIRP(sstr) */
11999 IoIFP(dstr) = fp_dup(IoIFP(sstr), IoTYPE(dstr), param);
12001 if (IoOFP(dstr) == IoIFP(sstr))
12002 IoOFP(dstr) = IoIFP(dstr);
12004 IoOFP(dstr) = fp_dup(IoOFP(dstr), IoTYPE(dstr), param);
12005 IoTOP_NAME(dstr) = SAVEPV(IoTOP_NAME(dstr));
12006 IoFMT_NAME(dstr) = SAVEPV(IoFMT_NAME(dstr));
12007 IoBOTTOM_NAME(dstr) = SAVEPV(IoBOTTOM_NAME(dstr));
12010 /* avoid cloning an empty array */
12011 if (AvARRAY((const AV *)sstr) && AvFILLp((const AV *)sstr) >= 0) {
12012 SV **dst_ary, **src_ary;
12013 SSize_t items = AvFILLp((const AV *)sstr) + 1;
12015 src_ary = AvARRAY((const AV *)sstr);
12016 Newxz(dst_ary, AvMAX((const AV *)sstr)+1, SV*);
12017 ptr_table_store(PL_ptr_table, src_ary, dst_ary);
12018 AvARRAY(MUTABLE_AV(dstr)) = dst_ary;
12019 AvALLOC((const AV *)dstr) = dst_ary;
12020 if (AvREAL((const AV *)sstr)) {
12021 dst_ary = sv_dup_inc_multiple(src_ary, dst_ary, items,
12025 while (items-- > 0)
12026 *dst_ary++ = sv_dup(*src_ary++, param);
12028 items = AvMAX((const AV *)sstr) - AvFILLp((const AV *)sstr);
12029 while (items-- > 0) {
12030 *dst_ary++ = &PL_sv_undef;
12034 AvARRAY(MUTABLE_AV(dstr)) = NULL;
12035 AvALLOC((const AV *)dstr) = (SV**)NULL;
12036 AvMAX( (const AV *)dstr) = -1;
12037 AvFILLp((const AV *)dstr) = -1;
12041 if (HvARRAY((const HV *)sstr)) {
12043 const bool sharekeys = !!HvSHAREKEYS(sstr);
12044 XPVHV * const dxhv = (XPVHV*)SvANY(dstr);
12045 XPVHV * const sxhv = (XPVHV*)SvANY(sstr);
12047 Newx(darray, PERL_HV_ARRAY_ALLOC_BYTES(dxhv->xhv_max+1)
12048 + (SvOOK(sstr) ? sizeof(struct xpvhv_aux) : 0),
12050 HvARRAY(dstr) = (HE**)darray;
12051 while (i <= sxhv->xhv_max) {
12052 const HE * const source = HvARRAY(sstr)[i];
12053 HvARRAY(dstr)[i] = source
12054 ? he_dup(source, sharekeys, param) : 0;
12058 const struct xpvhv_aux * const saux = HvAUX(sstr);
12059 struct xpvhv_aux * const daux = HvAUX(dstr);
12060 /* This flag isn't copied. */
12061 /* SvOOK_on(hv) attacks the IV flags. */
12062 SvFLAGS(dstr) |= SVf_OOK;
12064 if (saux->xhv_name_count) {
12065 HEK ** const sname = saux->xhv_name_u.xhvnameu_names;
12067 = saux->xhv_name_count < 0
12068 ? -saux->xhv_name_count
12069 : saux->xhv_name_count;
12070 HEK **shekp = sname + count;
12072 Newx(daux->xhv_name_u.xhvnameu_names, count, HEK *);
12073 dhekp = daux->xhv_name_u.xhvnameu_names + count;
12074 while (shekp-- > sname) {
12076 *dhekp = hek_dup(*shekp, param);
12080 daux->xhv_name_u.xhvnameu_name
12081 = hek_dup(saux->xhv_name_u.xhvnameu_name,
12084 daux->xhv_name_count = saux->xhv_name_count;
12086 daux->xhv_riter = saux->xhv_riter;
12087 daux->xhv_eiter = saux->xhv_eiter
12088 ? he_dup(saux->xhv_eiter,
12089 cBOOL(HvSHAREKEYS(sstr)), param) : 0;
12090 /* backref array needs refcnt=2; see sv_add_backref */
12091 daux->xhv_backreferences =
12092 (param->flags & CLONEf_JOIN_IN)
12093 /* when joining, we let the individual GVs and
12094 * CVs add themselves to backref as
12095 * needed. This avoids pulling in stuff
12096 * that isn't required, and simplifies the
12097 * case where stashes aren't cloned back
12098 * if they already exist in the parent
12101 : saux->xhv_backreferences
12102 ? (SvTYPE(saux->xhv_backreferences) == SVt_PVAV)
12103 ? MUTABLE_AV(SvREFCNT_inc(
12104 sv_dup_inc((const SV *)
12105 saux->xhv_backreferences, param)))
12106 : MUTABLE_AV(sv_dup((const SV *)
12107 saux->xhv_backreferences, param))
12110 daux->xhv_mro_meta = saux->xhv_mro_meta
12111 ? mro_meta_dup(saux->xhv_mro_meta, param)
12114 /* Record stashes for possible cloning in Perl_clone(). */
12116 av_push(param->stashes, dstr);
12120 HvARRAY(MUTABLE_HV(dstr)) = NULL;
12123 if (!(param->flags & CLONEf_COPY_STACKS)) {
12128 /* NOTE: not refcounted */
12129 SvANY(MUTABLE_CV(dstr))->xcv_stash =
12130 hv_dup(CvSTASH(dstr), param);
12131 if ((param->flags & CLONEf_JOIN_IN) && CvSTASH(dstr))
12132 Perl_sv_add_backref(aTHX_ MUTABLE_SV(CvSTASH(dstr)), dstr);
12133 if (!CvISXSUB(dstr)) {
12135 CvROOT(dstr) = OpREFCNT_inc(CvROOT(dstr));
12137 CvFILE(dstr) = SAVEPV(CvFILE(dstr));
12138 } else if (CvCONST(dstr)) {
12139 CvXSUBANY(dstr).any_ptr =
12140 sv_dup_inc((const SV *)CvXSUBANY(dstr).any_ptr, param);
12142 /* don't dup if copying back - CvGV isn't refcounted, so the
12143 * duped GV may never be freed. A bit of a hack! DAPM */
12144 SvANY(MUTABLE_CV(dstr))->xcv_gv =
12146 ? gv_dup_inc(CvGV(sstr), param)
12147 : (param->flags & CLONEf_JOIN_IN)
12149 : gv_dup(CvGV(sstr), param);
12151 CvPADLIST(dstr) = padlist_dup(CvPADLIST(sstr), param);
12153 CvWEAKOUTSIDE(sstr)
12154 ? cv_dup( CvOUTSIDE(dstr), param)
12155 : cv_dup_inc(CvOUTSIDE(dstr), param);
12161 if (SvOBJECT(dstr) && SvTYPE(dstr) != SVt_PVIO)
12168 Perl_sv_dup_inc(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
12170 PERL_ARGS_ASSERT_SV_DUP_INC;
12171 return sstr ? SvREFCNT_inc(sv_dup_common(sstr, param)) : NULL;
12175 Perl_sv_dup(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
12177 SV *dstr = sstr ? sv_dup_common(sstr, param) : NULL;
12178 PERL_ARGS_ASSERT_SV_DUP;
12180 /* Track every SV that (at least initially) had a reference count of 0.
12181 We need to do this by holding an actual reference to it in this array.
12182 If we attempt to cheat, turn AvREAL_off(), and store only pointers
12183 (akin to the stashes hash, and the perl stack), we come unstuck if
12184 a weak reference (or other SV legitimately SvREFCNT() == 0 for this
12185 thread) is manipulated in a CLONE method, because CLONE runs before the
12186 unreferenced array is walked to find SVs still with SvREFCNT() == 0
12187 (and fix things up by giving each a reference via the temps stack).
12188 Instead, during CLONE, if the 0-referenced SV has SvREFCNT_inc() and
12189 then SvREFCNT_dec(), it will be cleaned up (and added to the free list)
12190 before the walk of unreferenced happens and a reference to that is SV
12191 added to the temps stack. At which point we have the same SV considered
12192 to be in use, and free to be re-used. Not good.
12194 if (dstr && !(param->flags & CLONEf_COPY_STACKS) && !SvREFCNT(dstr)) {
12195 assert(param->unreferenced);
12196 av_push(param->unreferenced, SvREFCNT_inc(dstr));
12202 /* duplicate a context */
12205 Perl_cx_dup(pTHX_ PERL_CONTEXT *cxs, I32 ix, I32 max, CLONE_PARAMS* param)
12207 PERL_CONTEXT *ncxs;
12209 PERL_ARGS_ASSERT_CX_DUP;
12212 return (PERL_CONTEXT*)NULL;
12214 /* look for it in the table first */
12215 ncxs = (PERL_CONTEXT*)ptr_table_fetch(PL_ptr_table, cxs);
12219 /* create anew and remember what it is */
12220 Newx(ncxs, max + 1, PERL_CONTEXT);
12221 ptr_table_store(PL_ptr_table, cxs, ncxs);
12222 Copy(cxs, ncxs, max + 1, PERL_CONTEXT);
12225 PERL_CONTEXT * const ncx = &ncxs[ix];
12226 if (CxTYPE(ncx) == CXt_SUBST) {
12227 Perl_croak(aTHX_ "Cloning substitution context is unimplemented");
12230 switch (CxTYPE(ncx)) {
12232 ncx->blk_sub.cv = (ncx->blk_sub.olddepth == 0
12233 ? cv_dup_inc(ncx->blk_sub.cv, param)
12234 : cv_dup(ncx->blk_sub.cv,param));
12235 ncx->blk_sub.argarray = (CxHASARGS(ncx)
12236 ? av_dup_inc(ncx->blk_sub.argarray,
12239 ncx->blk_sub.savearray = av_dup_inc(ncx->blk_sub.savearray,
12241 ncx->blk_sub.oldcomppad = (PAD*)ptr_table_fetch(PL_ptr_table,
12242 ncx->blk_sub.oldcomppad);
12245 ncx->blk_eval.old_namesv = sv_dup_inc(ncx->blk_eval.old_namesv,
12247 ncx->blk_eval.cur_text = sv_dup(ncx->blk_eval.cur_text, param);
12249 case CXt_LOOP_LAZYSV:
12250 ncx->blk_loop.state_u.lazysv.end
12251 = sv_dup_inc(ncx->blk_loop.state_u.lazysv.end, param);
12252 /* We are taking advantage of av_dup_inc and sv_dup_inc
12253 actually being the same function, and order equivalence of
12255 We can assert the later [but only at run time :-(] */
12256 assert ((void *) &ncx->blk_loop.state_u.ary.ary ==
12257 (void *) &ncx->blk_loop.state_u.lazysv.cur);
12259 ncx->blk_loop.state_u.ary.ary
12260 = av_dup_inc(ncx->blk_loop.state_u.ary.ary, param);
12261 case CXt_LOOP_LAZYIV:
12262 case CXt_LOOP_PLAIN:
12263 if (CxPADLOOP(ncx)) {
12264 ncx->blk_loop.itervar_u.oldcomppad
12265 = (PAD*)ptr_table_fetch(PL_ptr_table,
12266 ncx->blk_loop.itervar_u.oldcomppad);
12268 ncx->blk_loop.itervar_u.gv
12269 = gv_dup((const GV *)ncx->blk_loop.itervar_u.gv,
12274 ncx->blk_format.cv = cv_dup(ncx->blk_format.cv, param);
12275 ncx->blk_format.gv = gv_dup(ncx->blk_format.gv, param);
12276 ncx->blk_format.dfoutgv = gv_dup_inc(ncx->blk_format.dfoutgv,
12289 /* duplicate a stack info structure */
12292 Perl_si_dup(pTHX_ PERL_SI *si, CLONE_PARAMS* param)
12296 PERL_ARGS_ASSERT_SI_DUP;
12299 return (PERL_SI*)NULL;
12301 /* look for it in the table first */
12302 nsi = (PERL_SI*)ptr_table_fetch(PL_ptr_table, si);
12306 /* create anew and remember what it is */
12307 Newxz(nsi, 1, PERL_SI);
12308 ptr_table_store(PL_ptr_table, si, nsi);
12310 nsi->si_stack = av_dup_inc(si->si_stack, param);
12311 nsi->si_cxix = si->si_cxix;
12312 nsi->si_cxmax = si->si_cxmax;
12313 nsi->si_cxstack = cx_dup(si->si_cxstack, si->si_cxix, si->si_cxmax, param);
12314 nsi->si_type = si->si_type;
12315 nsi->si_prev = si_dup(si->si_prev, param);
12316 nsi->si_next = si_dup(si->si_next, param);
12317 nsi->si_markoff = si->si_markoff;
12322 #define POPINT(ss,ix) ((ss)[--(ix)].any_i32)
12323 #define TOPINT(ss,ix) ((ss)[ix].any_i32)
12324 #define POPLONG(ss,ix) ((ss)[--(ix)].any_long)
12325 #define TOPLONG(ss,ix) ((ss)[ix].any_long)
12326 #define POPIV(ss,ix) ((ss)[--(ix)].any_iv)
12327 #define TOPIV(ss,ix) ((ss)[ix].any_iv)
12328 #define POPUV(ss,ix) ((ss)[--(ix)].any_uv)
12329 #define TOPUV(ss,ix) ((ss)[ix].any_uv)
12330 #define POPBOOL(ss,ix) ((ss)[--(ix)].any_bool)
12331 #define TOPBOOL(ss,ix) ((ss)[ix].any_bool)
12332 #define POPPTR(ss,ix) ((ss)[--(ix)].any_ptr)
12333 #define TOPPTR(ss,ix) ((ss)[ix].any_ptr)
12334 #define POPDPTR(ss,ix) ((ss)[--(ix)].any_dptr)
12335 #define TOPDPTR(ss,ix) ((ss)[ix].any_dptr)
12336 #define POPDXPTR(ss,ix) ((ss)[--(ix)].any_dxptr)
12337 #define TOPDXPTR(ss,ix) ((ss)[ix].any_dxptr)
12340 #define pv_dup_inc(p) SAVEPV(p)
12341 #define pv_dup(p) SAVEPV(p)
12342 #define svp_dup_inc(p,pp) any_dup(p,pp)
12344 /* map any object to the new equivent - either something in the
12345 * ptr table, or something in the interpreter structure
12349 Perl_any_dup(pTHX_ void *v, const PerlInterpreter *proto_perl)
12353 PERL_ARGS_ASSERT_ANY_DUP;
12356 return (void*)NULL;
12358 /* look for it in the table first */
12359 ret = ptr_table_fetch(PL_ptr_table, v);
12363 /* see if it is part of the interpreter structure */
12364 if (v >= (void*)proto_perl && v < (void*)(proto_perl+1))
12365 ret = (void*)(((char*)aTHX) + (((char*)v) - (char*)proto_perl));
12373 /* duplicate the save stack */
12376 Perl_ss_dup(pTHX_ PerlInterpreter *proto_perl, CLONE_PARAMS* param)
12379 ANY * const ss = proto_perl->Isavestack;
12380 const I32 max = proto_perl->Isavestack_max;
12381 I32 ix = proto_perl->Isavestack_ix;
12394 void (*dptr) (void*);
12395 void (*dxptr) (pTHX_ void*);
12397 PERL_ARGS_ASSERT_SS_DUP;
12399 Newxz(nss, max, ANY);
12402 const UV uv = POPUV(ss,ix);
12403 const U8 type = (U8)uv & SAVE_MASK;
12405 TOPUV(nss,ix) = uv;
12407 case SAVEt_CLEARSV:
12409 case SAVEt_HELEM: /* hash element */
12410 sv = (const SV *)POPPTR(ss,ix);
12411 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12413 case SAVEt_ITEM: /* normal string */
12414 case SAVEt_GVSV: /* scalar slot in GV */
12415 case SAVEt_SV: /* scalar reference */
12416 sv = (const SV *)POPPTR(ss,ix);
12417 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12420 case SAVEt_MORTALIZESV:
12421 sv = (const SV *)POPPTR(ss,ix);
12422 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12424 case SAVEt_SHARED_PVREF: /* char* in shared space */
12425 c = (char*)POPPTR(ss,ix);
12426 TOPPTR(nss,ix) = savesharedpv(c);
12427 ptr = POPPTR(ss,ix);
12428 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12430 case SAVEt_GENERIC_SVREF: /* generic sv */
12431 case SAVEt_SVREF: /* scalar reference */
12432 sv = (const SV *)POPPTR(ss,ix);
12433 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12434 ptr = POPPTR(ss,ix);
12435 TOPPTR(nss,ix) = svp_dup_inc((SV**)ptr, proto_perl);/* XXXXX */
12437 case SAVEt_HV: /* hash reference */
12438 case SAVEt_AV: /* array reference */
12439 sv = (const SV *) POPPTR(ss,ix);
12440 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12442 case SAVEt_COMPPAD:
12444 sv = (const SV *) POPPTR(ss,ix);
12445 TOPPTR(nss,ix) = sv_dup(sv, param);
12447 case SAVEt_INT: /* int reference */
12448 ptr = POPPTR(ss,ix);
12449 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12450 intval = (int)POPINT(ss,ix);
12451 TOPINT(nss,ix) = intval;
12453 case SAVEt_LONG: /* long reference */
12454 ptr = POPPTR(ss,ix);
12455 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12456 longval = (long)POPLONG(ss,ix);
12457 TOPLONG(nss,ix) = longval;
12459 case SAVEt_I32: /* I32 reference */
12460 case SAVEt_COP_ARYBASE: /* call CopARYBASE_set */
12461 ptr = POPPTR(ss,ix);
12462 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12464 TOPINT(nss,ix) = i;
12466 case SAVEt_IV: /* IV reference */
12467 ptr = POPPTR(ss,ix);
12468 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12470 TOPIV(nss,ix) = iv;
12472 case SAVEt_HPTR: /* HV* reference */
12473 case SAVEt_APTR: /* AV* reference */
12474 case SAVEt_SPTR: /* SV* reference */
12475 ptr = POPPTR(ss,ix);
12476 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12477 sv = (const SV *)POPPTR(ss,ix);
12478 TOPPTR(nss,ix) = sv_dup(sv, param);
12480 case SAVEt_VPTR: /* random* reference */
12481 ptr = POPPTR(ss,ix);
12482 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12484 case SAVEt_INT_SMALL:
12485 case SAVEt_I32_SMALL:
12486 case SAVEt_I16: /* I16 reference */
12487 case SAVEt_I8: /* I8 reference */
12489 ptr = POPPTR(ss,ix);
12490 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12492 case SAVEt_GENERIC_PVREF: /* generic char* */
12493 case SAVEt_PPTR: /* char* reference */
12494 ptr = POPPTR(ss,ix);
12495 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12496 c = (char*)POPPTR(ss,ix);
12497 TOPPTR(nss,ix) = pv_dup(c);
12499 case SAVEt_GP: /* scalar reference */
12500 gp = (GP*)POPPTR(ss,ix);
12501 TOPPTR(nss,ix) = gp = gp_dup(gp, param);
12502 (void)GpREFCNT_inc(gp);
12503 gv = (const GV *)POPPTR(ss,ix);
12504 TOPPTR(nss,ix) = gv_dup_inc(gv, param);
12507 ptr = POPPTR(ss,ix);
12508 if (ptr && (((OP*)ptr)->op_private & OPpREFCOUNTED)) {
12509 /* these are assumed to be refcounted properly */
12511 switch (((OP*)ptr)->op_type) {
12513 case OP_LEAVESUBLV:
12517 case OP_LEAVEWRITE:
12518 TOPPTR(nss,ix) = ptr;
12521 (void) OpREFCNT_inc(o);
12525 TOPPTR(nss,ix) = NULL;
12530 TOPPTR(nss,ix) = NULL;
12532 case SAVEt_FREECOPHH:
12533 ptr = POPPTR(ss,ix);
12534 TOPPTR(nss,ix) = cophh_copy((COPHH *)ptr);
12537 hv = (const HV *)POPPTR(ss,ix);
12538 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
12540 TOPINT(nss,ix) = i;
12543 c = (char*)POPPTR(ss,ix);
12544 TOPPTR(nss,ix) = pv_dup_inc(c);
12546 case SAVEt_STACK_POS: /* Position on Perl stack */
12548 TOPINT(nss,ix) = i;
12550 case SAVEt_DESTRUCTOR:
12551 ptr = POPPTR(ss,ix);
12552 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
12553 dptr = POPDPTR(ss,ix);
12554 TOPDPTR(nss,ix) = DPTR2FPTR(void (*)(void*),
12555 any_dup(FPTR2DPTR(void *, dptr),
12558 case SAVEt_DESTRUCTOR_X:
12559 ptr = POPPTR(ss,ix);
12560 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
12561 dxptr = POPDXPTR(ss,ix);
12562 TOPDXPTR(nss,ix) = DPTR2FPTR(void (*)(pTHX_ void*),
12563 any_dup(FPTR2DPTR(void *, dxptr),
12566 case SAVEt_REGCONTEXT:
12568 ix -= uv >> SAVE_TIGHT_SHIFT;
12570 case SAVEt_AELEM: /* array element */
12571 sv = (const SV *)POPPTR(ss,ix);
12572 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12574 TOPINT(nss,ix) = i;
12575 av = (const AV *)POPPTR(ss,ix);
12576 TOPPTR(nss,ix) = av_dup_inc(av, param);
12579 ptr = POPPTR(ss,ix);
12580 TOPPTR(nss,ix) = ptr;
12583 ptr = POPPTR(ss,ix);
12584 ptr = cophh_copy((COPHH*)ptr);
12585 TOPPTR(nss,ix) = ptr;
12587 TOPINT(nss,ix) = i;
12588 if (i & HINT_LOCALIZE_HH) {
12589 hv = (const HV *)POPPTR(ss,ix);
12590 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
12593 case SAVEt_PADSV_AND_MORTALIZE:
12594 longval = (long)POPLONG(ss,ix);
12595 TOPLONG(nss,ix) = longval;
12596 ptr = POPPTR(ss,ix);
12597 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12598 sv = (const SV *)POPPTR(ss,ix);
12599 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12601 case SAVEt_SET_SVFLAGS:
12603 TOPINT(nss,ix) = i;
12605 TOPINT(nss,ix) = i;
12606 sv = (const SV *)POPPTR(ss,ix);
12607 TOPPTR(nss,ix) = sv_dup(sv, param);
12609 case SAVEt_RE_STATE:
12611 const struct re_save_state *const old_state
12612 = (struct re_save_state *)
12613 (ss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
12614 struct re_save_state *const new_state
12615 = (struct re_save_state *)
12616 (nss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
12618 Copy(old_state, new_state, 1, struct re_save_state);
12619 ix -= SAVESTACK_ALLOC_FOR_RE_SAVE_STATE;
12621 new_state->re_state_bostr
12622 = pv_dup(old_state->re_state_bostr);
12623 new_state->re_state_reginput
12624 = pv_dup(old_state->re_state_reginput);
12625 new_state->re_state_regeol
12626 = pv_dup(old_state->re_state_regeol);
12627 new_state->re_state_regoffs
12628 = (regexp_paren_pair*)
12629 any_dup(old_state->re_state_regoffs, proto_perl);
12630 new_state->re_state_reglastparen
12631 = (U32*) any_dup(old_state->re_state_reglastparen,
12633 new_state->re_state_reglastcloseparen
12634 = (U32*)any_dup(old_state->re_state_reglastcloseparen,
12636 /* XXX This just has to be broken. The old save_re_context
12637 code did SAVEGENERICPV(PL_reg_start_tmp);
12638 PL_reg_start_tmp is char **.
12639 Look above to what the dup code does for
12640 SAVEt_GENERIC_PVREF
12641 It can never have worked.
12642 So this is merely a faithful copy of the exiting bug: */
12643 new_state->re_state_reg_start_tmp
12644 = (char **) pv_dup((char *)
12645 old_state->re_state_reg_start_tmp);
12646 /* I assume that it only ever "worked" because no-one called
12647 (pseudo)fork while the regexp engine had re-entered itself.
12649 #ifdef PERL_OLD_COPY_ON_WRITE
12650 new_state->re_state_nrs
12651 = sv_dup(old_state->re_state_nrs, param);
12653 new_state->re_state_reg_magic
12654 = (MAGIC*) any_dup(old_state->re_state_reg_magic,
12656 new_state->re_state_reg_oldcurpm
12657 = (PMOP*) any_dup(old_state->re_state_reg_oldcurpm,
12659 new_state->re_state_reg_curpm
12660 = (PMOP*) any_dup(old_state->re_state_reg_curpm,
12662 new_state->re_state_reg_oldsaved
12663 = pv_dup(old_state->re_state_reg_oldsaved);
12664 new_state->re_state_reg_poscache
12665 = pv_dup(old_state->re_state_reg_poscache);
12666 new_state->re_state_reg_starttry
12667 = pv_dup(old_state->re_state_reg_starttry);
12670 case SAVEt_COMPILE_WARNINGS:
12671 ptr = POPPTR(ss,ix);
12672 TOPPTR(nss,ix) = DUP_WARNINGS((STRLEN*)ptr);
12675 ptr = POPPTR(ss,ix);
12676 TOPPTR(nss,ix) = parser_dup((const yy_parser*)ptr, param);
12680 "panic: ss_dup inconsistency (%"IVdf")", (IV) type);
12688 /* if sv is a stash, call $class->CLONE_SKIP(), and set the SVphv_CLONEABLE
12689 * flag to the result. This is done for each stash before cloning starts,
12690 * so we know which stashes want their objects cloned */
12693 do_mark_cloneable_stash(pTHX_ SV *const sv)
12695 const HEK * const hvname = HvNAME_HEK((const HV *)sv);
12697 GV* const cloner = gv_fetchmethod_autoload(MUTABLE_HV(sv), "CLONE_SKIP", 0);
12698 SvFLAGS(sv) |= SVphv_CLONEABLE; /* clone objects by default */
12699 if (cloner && GvCV(cloner)) {
12706 mXPUSHs(newSVhek(hvname));
12708 call_sv(MUTABLE_SV(GvCV(cloner)), G_SCALAR);
12715 SvFLAGS(sv) &= ~SVphv_CLONEABLE;
12723 =for apidoc perl_clone
12725 Create and return a new interpreter by cloning the current one.
12727 perl_clone takes these flags as parameters:
12729 CLONEf_COPY_STACKS - is used to, well, copy the stacks also,
12730 without it we only clone the data and zero the stacks,
12731 with it we copy the stacks and the new perl interpreter is
12732 ready to run at the exact same point as the previous one.
12733 The pseudo-fork code uses COPY_STACKS while the
12734 threads->create doesn't.
12736 CLONEf_KEEP_PTR_TABLE
12737 perl_clone keeps a ptr_table with the pointer of the old
12738 variable as a key and the new variable as a value,
12739 this allows it to check if something has been cloned and not
12740 clone it again but rather just use the value and increase the
12741 refcount. If KEEP_PTR_TABLE is not set then perl_clone will kill
12742 the ptr_table using the function
12743 C<ptr_table_free(PL_ptr_table); PL_ptr_table = NULL;>,
12744 reason to keep it around is if you want to dup some of your own
12745 variable who are outside the graph perl scans, example of this
12746 code is in threads.xs create
12749 This is a win32 thing, it is ignored on unix, it tells perls
12750 win32host code (which is c++) to clone itself, this is needed on
12751 win32 if you want to run two threads at the same time,
12752 if you just want to do some stuff in a separate perl interpreter
12753 and then throw it away and return to the original one,
12754 you don't need to do anything.
12759 /* XXX the above needs expanding by someone who actually understands it ! */
12760 EXTERN_C PerlInterpreter *
12761 perl_clone_host(PerlInterpreter* proto_perl, UV flags);
12764 perl_clone(PerlInterpreter *proto_perl, UV flags)
12767 #ifdef PERL_IMPLICIT_SYS
12769 PERL_ARGS_ASSERT_PERL_CLONE;
12771 /* perlhost.h so we need to call into it
12772 to clone the host, CPerlHost should have a c interface, sky */
12774 if (flags & CLONEf_CLONE_HOST) {
12775 return perl_clone_host(proto_perl,flags);
12777 return perl_clone_using(proto_perl, flags,
12779 proto_perl->IMemShared,
12780 proto_perl->IMemParse,
12782 proto_perl->IStdIO,
12786 proto_perl->IProc);
12790 perl_clone_using(PerlInterpreter *proto_perl, UV flags,
12791 struct IPerlMem* ipM, struct IPerlMem* ipMS,
12792 struct IPerlMem* ipMP, struct IPerlEnv* ipE,
12793 struct IPerlStdIO* ipStd, struct IPerlLIO* ipLIO,
12794 struct IPerlDir* ipD, struct IPerlSock* ipS,
12795 struct IPerlProc* ipP)
12797 /* XXX many of the string copies here can be optimized if they're
12798 * constants; they need to be allocated as common memory and just
12799 * their pointers copied. */
12802 CLONE_PARAMS clone_params;
12803 CLONE_PARAMS* const param = &clone_params;
12805 PerlInterpreter * const my_perl = (PerlInterpreter*)(*ipM->pMalloc)(ipM, sizeof(PerlInterpreter));
12807 PERL_ARGS_ASSERT_PERL_CLONE_USING;
12808 #else /* !PERL_IMPLICIT_SYS */
12810 CLONE_PARAMS clone_params;
12811 CLONE_PARAMS* param = &clone_params;
12812 PerlInterpreter * const my_perl = (PerlInterpreter*)PerlMem_malloc(sizeof(PerlInterpreter));
12814 PERL_ARGS_ASSERT_PERL_CLONE;
12815 #endif /* PERL_IMPLICIT_SYS */
12817 /* for each stash, determine whether its objects should be cloned */
12818 S_visit(proto_perl, do_mark_cloneable_stash, SVt_PVHV, SVTYPEMASK);
12819 PERL_SET_THX(my_perl);
12822 PoisonNew(my_perl, 1, PerlInterpreter);
12825 PL_defstash = NULL; /* may be used by perl malloc() */
12828 PL_scopestack_name = 0;
12830 PL_savestack_ix = 0;
12831 PL_savestack_max = -1;
12832 PL_sig_pending = 0;
12834 Zero(&PL_debug_pad, 1, struct perl_debug_pad);
12835 # ifdef DEBUG_LEAKING_SCALARS
12836 PL_sv_serial = (((UV)my_perl >> 2) & 0xfff) * 1000000;
12838 #else /* !DEBUGGING */
12839 Zero(my_perl, 1, PerlInterpreter);
12840 #endif /* DEBUGGING */
12842 #ifdef PERL_IMPLICIT_SYS
12843 /* host pointers */
12845 PL_MemShared = ipMS;
12846 PL_MemParse = ipMP;
12853 #endif /* PERL_IMPLICIT_SYS */
12855 param->flags = flags;
12856 /* Nothing in the core code uses this, but we make it available to
12857 extensions (using mg_dup). */
12858 param->proto_perl = proto_perl;
12859 /* Likely nothing will use this, but it is initialised to be consistent
12860 with Perl_clone_params_new(). */
12861 param->new_perl = my_perl;
12862 param->unreferenced = NULL;
12864 INIT_TRACK_MEMPOOL(my_perl->Imemory_debug_header, my_perl);
12866 PL_body_arenas = NULL;
12867 Zero(&PL_body_roots, 1, PL_body_roots);
12870 PL_sv_objcount = 0;
12872 PL_sv_arenaroot = NULL;
12874 PL_debug = proto_perl->Idebug;
12876 PL_hash_seed = proto_perl->Ihash_seed;
12877 PL_rehash_seed = proto_perl->Irehash_seed;
12879 SvANY(&PL_sv_undef) = NULL;
12880 SvREFCNT(&PL_sv_undef) = (~(U32)0)/2;
12881 SvFLAGS(&PL_sv_undef) = SVf_READONLY|SVt_NULL;
12882 SvREFCNT(&PL_sv_no) = (~(U32)0)/2;
12883 SvFLAGS(&PL_sv_no) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
12884 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
12886 SvANY(&PL_sv_yes) = new_XPVNV();
12887 SvREFCNT(&PL_sv_yes) = (~(U32)0)/2;
12888 SvFLAGS(&PL_sv_yes) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
12889 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
12891 /* dbargs array probably holds garbage */
12894 PL_compiling = proto_perl->Icompiling;
12896 #ifdef PERL_DEBUG_READONLY_OPS
12901 /* pseudo environmental stuff */
12902 PL_origargc = proto_perl->Iorigargc;
12903 PL_origargv = proto_perl->Iorigargv;
12905 /* Set tainting stuff before PerlIO_debug can possibly get called */
12906 PL_tainting = proto_perl->Itainting;
12907 PL_taint_warn = proto_perl->Itaint_warn;
12909 PL_minus_c = proto_perl->Iminus_c;
12911 PL_localpatches = proto_perl->Ilocalpatches;
12912 PL_splitstr = proto_perl->Isplitstr;
12913 PL_minus_n = proto_perl->Iminus_n;
12914 PL_minus_p = proto_perl->Iminus_p;
12915 PL_minus_l = proto_perl->Iminus_l;
12916 PL_minus_a = proto_perl->Iminus_a;
12917 PL_minus_E = proto_perl->Iminus_E;
12918 PL_minus_F = proto_perl->Iminus_F;
12919 PL_doswitches = proto_perl->Idoswitches;
12920 PL_dowarn = proto_perl->Idowarn;
12921 PL_sawampersand = proto_perl->Isawampersand;
12922 PL_unsafe = proto_perl->Iunsafe;
12923 PL_perldb = proto_perl->Iperldb;
12924 PL_perl_destruct_level = proto_perl->Iperl_destruct_level;
12925 PL_exit_flags = proto_perl->Iexit_flags;
12927 /* XXX time(&PL_basetime) when asked for? */
12928 PL_basetime = proto_perl->Ibasetime;
12930 PL_maxsysfd = proto_perl->Imaxsysfd;
12931 PL_statusvalue = proto_perl->Istatusvalue;
12933 PL_statusvalue_vms = proto_perl->Istatusvalue_vms;
12935 PL_statusvalue_posix = proto_perl->Istatusvalue_posix;
12938 /* RE engine related */
12939 Zero(&PL_reg_state, 1, struct re_save_state);
12940 PL_reginterp_cnt = 0;
12941 PL_regmatch_slab = NULL;
12943 PL_sub_generation = proto_perl->Isub_generation;
12945 /* funky return mechanisms */
12946 PL_forkprocess = proto_perl->Iforkprocess;
12948 /* internal state */
12949 PL_maxo = proto_perl->Imaxo;
12951 PL_main_start = proto_perl->Imain_start;
12952 PL_eval_root = proto_perl->Ieval_root;
12953 PL_eval_start = proto_perl->Ieval_start;
12955 PL_filemode = proto_perl->Ifilemode;
12956 PL_lastfd = proto_perl->Ilastfd;
12957 PL_oldname = proto_perl->Ioldname; /* XXX not quite right */
12960 PL_gensym = proto_perl->Igensym;
12962 PL_laststatval = proto_perl->Ilaststatval;
12963 PL_laststype = proto_perl->Ilaststype;
12966 PL_profiledata = NULL;
12968 PL_generation = proto_perl->Igeneration;
12970 PL_in_clean_objs = proto_perl->Iin_clean_objs;
12971 PL_in_clean_all = proto_perl->Iin_clean_all;
12973 PL_uid = proto_perl->Iuid;
12974 PL_euid = proto_perl->Ieuid;
12975 PL_gid = proto_perl->Igid;
12976 PL_egid = proto_perl->Iegid;
12977 PL_nomemok = proto_perl->Inomemok;
12978 PL_an = proto_perl->Ian;
12979 PL_evalseq = proto_perl->Ievalseq;
12980 PL_origenviron = proto_perl->Iorigenviron; /* XXX not quite right */
12981 PL_origalen = proto_perl->Iorigalen;
12983 PL_sighandlerp = proto_perl->Isighandlerp;
12985 PL_runops = proto_perl->Irunops;
12987 PL_subline = proto_perl->Isubline;
12990 PL_cryptseen = proto_perl->Icryptseen;
12993 PL_hints = proto_perl->Ihints;
12995 PL_amagic_generation = proto_perl->Iamagic_generation;
12997 #ifdef USE_LOCALE_COLLATE
12998 PL_collation_ix = proto_perl->Icollation_ix;
12999 PL_collation_standard = proto_perl->Icollation_standard;
13000 PL_collxfrm_base = proto_perl->Icollxfrm_base;
13001 PL_collxfrm_mult = proto_perl->Icollxfrm_mult;
13002 #endif /* USE_LOCALE_COLLATE */
13004 #ifdef USE_LOCALE_NUMERIC
13005 PL_numeric_standard = proto_perl->Inumeric_standard;
13006 PL_numeric_local = proto_perl->Inumeric_local;
13007 #endif /* !USE_LOCALE_NUMERIC */
13009 /* Did the locale setup indicate UTF-8? */
13010 PL_utf8locale = proto_perl->Iutf8locale;
13011 /* Unicode features (see perlrun/-C) */
13012 PL_unicode = proto_perl->Iunicode;
13014 /* Pre-5.8 signals control */
13015 PL_signals = proto_perl->Isignals;
13017 /* times() ticks per second */
13018 PL_clocktick = proto_perl->Iclocktick;
13020 /* Recursion stopper for PerlIO_find_layer */
13021 PL_in_load_module = proto_perl->Iin_load_module;
13023 /* sort() routine */
13024 PL_sort_RealCmp = proto_perl->Isort_RealCmp;
13026 /* Not really needed/useful since the reenrant_retint is "volatile",
13027 * but do it for consistency's sake. */
13028 PL_reentrant_retint = proto_perl->Ireentrant_retint;
13030 /* Hooks to shared SVs and locks. */
13031 PL_sharehook = proto_perl->Isharehook;
13032 PL_lockhook = proto_perl->Ilockhook;
13033 PL_unlockhook = proto_perl->Iunlockhook;
13034 PL_threadhook = proto_perl->Ithreadhook;
13035 PL_destroyhook = proto_perl->Idestroyhook;
13036 PL_signalhook = proto_perl->Isignalhook;
13038 #ifdef THREADS_HAVE_PIDS
13039 PL_ppid = proto_perl->Ippid;
13043 PL_last_swash_hv = NULL; /* reinits on demand */
13044 PL_last_swash_klen = 0;
13045 PL_last_swash_key[0]= '\0';
13046 PL_last_swash_tmps = (U8*)NULL;
13047 PL_last_swash_slen = 0;
13049 PL_glob_index = proto_perl->Iglob_index;
13050 PL_srand_called = proto_perl->Isrand_called;
13052 if (flags & CLONEf_COPY_STACKS) {
13053 /* next allocation will be PL_tmps_stack[PL_tmps_ix+1] */
13054 PL_tmps_ix = proto_perl->Itmps_ix;
13055 PL_tmps_max = proto_perl->Itmps_max;
13056 PL_tmps_floor = proto_perl->Itmps_floor;
13058 /* next push_scope()/ENTER sets PL_scopestack[PL_scopestack_ix]
13059 * NOTE: unlike the others! */
13060 PL_scopestack_ix = proto_perl->Iscopestack_ix;
13061 PL_scopestack_max = proto_perl->Iscopestack_max;
13063 /* next SSPUSHFOO() sets PL_savestack[PL_savestack_ix]
13064 * NOTE: unlike the others! */
13065 PL_savestack_ix = proto_perl->Isavestack_ix;
13066 PL_savestack_max = proto_perl->Isavestack_max;
13069 PL_start_env = proto_perl->Istart_env; /* XXXXXX */
13070 PL_top_env = &PL_start_env;
13072 PL_op = proto_perl->Iop;
13075 PL_Xpv = (XPV*)NULL;
13076 my_perl->Ina = proto_perl->Ina;
13078 PL_statbuf = proto_perl->Istatbuf;
13079 PL_statcache = proto_perl->Istatcache;
13082 PL_timesbuf = proto_perl->Itimesbuf;
13085 PL_tainted = proto_perl->Itainted;
13086 PL_curpm = proto_perl->Icurpm; /* XXX No PMOP ref count */
13088 PL_chopset = proto_perl->Ichopset; /* XXX never deallocated */
13090 PL_restartjmpenv = proto_perl->Irestartjmpenv;
13091 PL_restartop = proto_perl->Irestartop;
13092 PL_in_eval = proto_perl->Iin_eval;
13093 PL_delaymagic = proto_perl->Idelaymagic;
13094 PL_phase = proto_perl->Iphase;
13095 PL_localizing = proto_perl->Ilocalizing;
13097 PL_hv_fetch_ent_mh = NULL;
13098 PL_modcount = proto_perl->Imodcount;
13099 PL_lastgotoprobe = NULL;
13100 PL_dumpindent = proto_perl->Idumpindent;
13102 PL_efloatbuf = NULL; /* reinits on demand */
13103 PL_efloatsize = 0; /* reinits on demand */
13107 PL_screamfirst = NULL;
13108 PL_screamnext = NULL;
13109 PL_maxscream = -1; /* reinits on demand */
13110 PL_lastscream = NULL;
13113 PL_regdummy = proto_perl->Iregdummy;
13114 PL_colorset = 0; /* reinits PL_colors[] */
13115 /*PL_colors[6] = {0,0,0,0,0,0};*/
13117 /* Pluggable optimizer */
13118 PL_peepp = proto_perl->Ipeepp;
13119 PL_rpeepp = proto_perl->Irpeepp;
13120 /* op_free() hook */
13121 PL_opfreehook = proto_perl->Iopfreehook;
13123 #ifdef USE_REENTRANT_API
13124 /* XXX: things like -Dm will segfault here in perlio, but doing
13125 * PERL_SET_CONTEXT(proto_perl);
13126 * breaks too many other things
13128 Perl_reentrant_init(aTHX);
13131 /* create SV map for pointer relocation */
13132 PL_ptr_table = ptr_table_new();
13134 /* initialize these special pointers as early as possible */
13135 ptr_table_store(PL_ptr_table, &proto_perl->Isv_undef, &PL_sv_undef);
13137 SvANY(&PL_sv_no) = new_XPVNV();
13138 SvPV_set(&PL_sv_no, savepvn(PL_No, 0));
13139 SvCUR_set(&PL_sv_no, 0);
13140 SvLEN_set(&PL_sv_no, 1);
13141 SvIV_set(&PL_sv_no, 0);
13142 SvNV_set(&PL_sv_no, 0);
13143 ptr_table_store(PL_ptr_table, &proto_perl->Isv_no, &PL_sv_no);
13145 SvPV_set(&PL_sv_yes, savepvn(PL_Yes, 1));
13146 SvCUR_set(&PL_sv_yes, 1);
13147 SvLEN_set(&PL_sv_yes, 2);
13148 SvIV_set(&PL_sv_yes, 1);
13149 SvNV_set(&PL_sv_yes, 1);
13150 ptr_table_store(PL_ptr_table, &proto_perl->Isv_yes, &PL_sv_yes);
13152 /* create (a non-shared!) shared string table */
13153 PL_strtab = newHV();
13154 HvSHAREKEYS_off(PL_strtab);
13155 hv_ksplit(PL_strtab, HvTOTALKEYS(proto_perl->Istrtab));
13156 ptr_table_store(PL_ptr_table, proto_perl->Istrtab, PL_strtab);
13158 /* These two PVs will be free'd special way so must set them same way op.c does */
13159 PL_compiling.cop_stashpv = savesharedpv(PL_compiling.cop_stashpv);
13160 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_stashpv, PL_compiling.cop_stashpv);
13162 PL_compiling.cop_file = savesharedpv(PL_compiling.cop_file);
13163 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_file, PL_compiling.cop_file);
13165 ptr_table_store(PL_ptr_table, &proto_perl->Icompiling, &PL_compiling);
13166 PL_compiling.cop_warnings = DUP_WARNINGS(PL_compiling.cop_warnings);
13167 CopHINTHASH_set(&PL_compiling, cophh_copy(CopHINTHASH_get(&PL_compiling)));
13168 PL_curcop = (COP*)any_dup(proto_perl->Icurcop, proto_perl);
13170 param->stashes = newAV(); /* Setup array of objects to call clone on */
13171 /* This makes no difference to the implementation, as it always pushes
13172 and shifts pointers to other SVs without changing their reference
13173 count, with the array becoming empty before it is freed. However, it
13174 makes it conceptually clear what is going on, and will avoid some
13175 work inside av.c, filling slots between AvFILL() and AvMAX() with
13176 &PL_sv_undef, and SvREFCNT_dec()ing those. */
13177 AvREAL_off(param->stashes);
13179 if (!(flags & CLONEf_COPY_STACKS)) {
13180 param->unreferenced = newAV();
13183 #ifdef PERLIO_LAYERS
13184 /* Clone PerlIO tables as soon as we can handle general xx_dup() */
13185 PerlIO_clone(aTHX_ proto_perl, param);
13188 PL_envgv = gv_dup(proto_perl->Ienvgv, param);
13189 PL_incgv = gv_dup(proto_perl->Iincgv, param);
13190 PL_hintgv = gv_dup(proto_perl->Ihintgv, param);
13191 PL_origfilename = SAVEPV(proto_perl->Iorigfilename);
13192 PL_diehook = sv_dup_inc(proto_perl->Idiehook, param);
13193 PL_warnhook = sv_dup_inc(proto_perl->Iwarnhook, param);
13196 PL_patchlevel = sv_dup_inc(proto_perl->Ipatchlevel, param);
13197 PL_apiversion = sv_dup_inc(proto_perl->Iapiversion, param);
13198 PL_inplace = SAVEPV(proto_perl->Iinplace);
13199 PL_e_script = sv_dup_inc(proto_perl->Ie_script, param);
13201 /* magical thingies */
13202 PL_formfeed = sv_dup(proto_perl->Iformfeed, param);
13204 PL_encoding = sv_dup(proto_perl->Iencoding, param);
13206 sv_setpvs(PERL_DEBUG_PAD(0), ""); /* For regex debugging. */
13207 sv_setpvs(PERL_DEBUG_PAD(1), ""); /* ext/re needs these */
13208 sv_setpvs(PERL_DEBUG_PAD(2), ""); /* even without DEBUGGING. */
13211 /* Clone the regex array */
13212 /* ORANGE FIXME for plugins, probably in the SV dup code.
13213 newSViv(PTR2IV(CALLREGDUPE(
13214 INT2PTR(REGEXP *, SvIVX(regex)), param))))
13216 PL_regex_padav = av_dup_inc(proto_perl->Iregex_padav, param);
13217 PL_regex_pad = AvARRAY(PL_regex_padav);
13219 /* shortcuts to various I/O objects */
13220 PL_ofsgv = gv_dup_inc(proto_perl->Iofsgv, param);
13221 PL_stdingv = gv_dup(proto_perl->Istdingv, param);
13222 PL_stderrgv = gv_dup(proto_perl->Istderrgv, param);
13223 PL_defgv = gv_dup(proto_perl->Idefgv, param);
13224 PL_argvgv = gv_dup(proto_perl->Iargvgv, param);
13225 PL_argvoutgv = gv_dup(proto_perl->Iargvoutgv, param);
13226 PL_argvout_stack = av_dup_inc(proto_perl->Iargvout_stack, param);
13228 /* shortcuts to regexp stuff */
13229 PL_replgv = gv_dup(proto_perl->Ireplgv, param);
13231 /* shortcuts to misc objects */
13232 PL_errgv = gv_dup(proto_perl->Ierrgv, param);
13234 /* shortcuts to debugging objects */
13235 PL_DBgv = gv_dup(proto_perl->IDBgv, param);
13236 PL_DBline = gv_dup(proto_perl->IDBline, param);
13237 PL_DBsub = gv_dup(proto_perl->IDBsub, param);
13238 PL_DBsingle = sv_dup(proto_perl->IDBsingle, param);
13239 PL_DBtrace = sv_dup(proto_perl->IDBtrace, param);
13240 PL_DBsignal = sv_dup(proto_perl->IDBsignal, param);
13242 /* symbol tables */
13243 PL_defstash = hv_dup_inc(proto_perl->Idefstash, param);
13244 PL_curstash = hv_dup(proto_perl->Icurstash, param);
13245 PL_debstash = hv_dup(proto_perl->Idebstash, param);
13246 PL_globalstash = hv_dup(proto_perl->Iglobalstash, param);
13247 PL_curstname = sv_dup_inc(proto_perl->Icurstname, param);
13249 PL_beginav = av_dup_inc(proto_perl->Ibeginav, param);
13250 PL_beginav_save = av_dup_inc(proto_perl->Ibeginav_save, param);
13251 PL_checkav_save = av_dup_inc(proto_perl->Icheckav_save, param);
13252 PL_unitcheckav = av_dup_inc(proto_perl->Iunitcheckav, param);
13253 PL_unitcheckav_save = av_dup_inc(proto_perl->Iunitcheckav_save, param);
13254 PL_endav = av_dup_inc(proto_perl->Iendav, param);
13255 PL_checkav = av_dup_inc(proto_perl->Icheckav, param);
13256 PL_initav = av_dup_inc(proto_perl->Iinitav, param);
13258 PL_isarev = hv_dup_inc(proto_perl->Iisarev, param);
13260 /* subprocess state */
13261 PL_fdpid = av_dup_inc(proto_perl->Ifdpid, param);
13263 if (proto_perl->Iop_mask)
13264 PL_op_mask = SAVEPVN(proto_perl->Iop_mask, PL_maxo);
13267 /* PL_asserting = proto_perl->Iasserting; */
13269 /* current interpreter roots */
13270 PL_main_cv = cv_dup_inc(proto_perl->Imain_cv, param);
13272 PL_main_root = OpREFCNT_inc(proto_perl->Imain_root);
13275 /* runtime control stuff */
13276 PL_curcopdb = (COP*)any_dup(proto_perl->Icurcopdb, proto_perl);
13278 PL_preambleav = av_dup_inc(proto_perl->Ipreambleav, param);
13280 PL_ors_sv = sv_dup_inc(proto_perl->Iors_sv, param);
13282 /* interpreter atexit processing */
13283 PL_exitlistlen = proto_perl->Iexitlistlen;
13284 if (PL_exitlistlen) {
13285 Newx(PL_exitlist, PL_exitlistlen, PerlExitListEntry);
13286 Copy(proto_perl->Iexitlist, PL_exitlist, PL_exitlistlen, PerlExitListEntry);
13289 PL_exitlist = (PerlExitListEntry*)NULL;
13291 PL_my_cxt_size = proto_perl->Imy_cxt_size;
13292 if (PL_my_cxt_size) {
13293 Newx(PL_my_cxt_list, PL_my_cxt_size, void *);
13294 Copy(proto_perl->Imy_cxt_list, PL_my_cxt_list, PL_my_cxt_size, void *);
13295 #ifdef PERL_GLOBAL_STRUCT_PRIVATE
13296 Newx(PL_my_cxt_keys, PL_my_cxt_size, const char *);
13297 Copy(proto_perl->Imy_cxt_keys, PL_my_cxt_keys, PL_my_cxt_size, char *);
13301 PL_my_cxt_list = (void**)NULL;
13302 #ifdef PERL_GLOBAL_STRUCT_PRIVATE
13303 PL_my_cxt_keys = (const char**)NULL;
13306 PL_modglobal = hv_dup_inc(proto_perl->Imodglobal, param);
13307 PL_custom_op_names = hv_dup_inc(proto_perl->Icustom_op_names,param);
13308 PL_custom_op_descs = hv_dup_inc(proto_perl->Icustom_op_descs,param);
13309 PL_custom_ops = hv_dup_inc(proto_perl->Icustom_ops, param);
13311 PL_compcv = cv_dup(proto_perl->Icompcv, param);
13313 PAD_CLONE_VARS(proto_perl, param);
13315 #ifdef HAVE_INTERP_INTERN
13316 sys_intern_dup(&proto_perl->Isys_intern, &PL_sys_intern);
13319 PL_DBcv = cv_dup(proto_perl->IDBcv, param);
13321 #ifdef PERL_USES_PL_PIDSTATUS
13322 PL_pidstatus = newHV(); /* XXX flag for cloning? */
13324 PL_osname = SAVEPV(proto_perl->Iosname);
13325 PL_parser = parser_dup(proto_perl->Iparser, param);
13327 /* XXX this only works if the saved cop has already been cloned */
13328 if (proto_perl->Iparser) {
13329 PL_parser->saved_curcop = (COP*)any_dup(
13330 proto_perl->Iparser->saved_curcop,
13334 PL_subname = sv_dup_inc(proto_perl->Isubname, param);
13336 #ifdef USE_LOCALE_COLLATE
13337 PL_collation_name = SAVEPV(proto_perl->Icollation_name);
13338 #endif /* USE_LOCALE_COLLATE */
13340 #ifdef USE_LOCALE_NUMERIC
13341 PL_numeric_name = SAVEPV(proto_perl->Inumeric_name);
13342 PL_numeric_radix_sv = sv_dup_inc(proto_perl->Inumeric_radix_sv, param);
13343 #endif /* !USE_LOCALE_NUMERIC */
13345 /* utf8 character classes */
13346 PL_utf8_alnum = sv_dup_inc(proto_perl->Iutf8_alnum, param);
13347 PL_utf8_ascii = sv_dup_inc(proto_perl->Iutf8_ascii, param);
13348 PL_utf8_alpha = sv_dup_inc(proto_perl->Iutf8_alpha, param);
13349 PL_utf8_space = sv_dup_inc(proto_perl->Iutf8_space, param);
13350 PL_utf8_cntrl = sv_dup_inc(proto_perl->Iutf8_cntrl, param);
13351 PL_utf8_graph = sv_dup_inc(proto_perl->Iutf8_graph, param);
13352 PL_utf8_digit = sv_dup_inc(proto_perl->Iutf8_digit, param);
13353 PL_utf8_upper = sv_dup_inc(proto_perl->Iutf8_upper, param);
13354 PL_utf8_lower = sv_dup_inc(proto_perl->Iutf8_lower, param);
13355 PL_utf8_print = sv_dup_inc(proto_perl->Iutf8_print, param);
13356 PL_utf8_punct = sv_dup_inc(proto_perl->Iutf8_punct, param);
13357 PL_utf8_xdigit = sv_dup_inc(proto_perl->Iutf8_xdigit, param);
13358 PL_utf8_mark = sv_dup_inc(proto_perl->Iutf8_mark, param);
13359 PL_utf8_X_begin = sv_dup_inc(proto_perl->Iutf8_X_begin, param);
13360 PL_utf8_X_extend = sv_dup_inc(proto_perl->Iutf8_X_extend, param);
13361 PL_utf8_X_prepend = sv_dup_inc(proto_perl->Iutf8_X_prepend, param);
13362 PL_utf8_X_non_hangul = sv_dup_inc(proto_perl->Iutf8_X_non_hangul, param);
13363 PL_utf8_X_L = sv_dup_inc(proto_perl->Iutf8_X_L, param);
13364 PL_utf8_X_LV = sv_dup_inc(proto_perl->Iutf8_X_LV, param);
13365 PL_utf8_X_LVT = sv_dup_inc(proto_perl->Iutf8_X_LVT, param);
13366 PL_utf8_X_T = sv_dup_inc(proto_perl->Iutf8_X_T, param);
13367 PL_utf8_X_V = sv_dup_inc(proto_perl->Iutf8_X_V, param);
13368 PL_utf8_X_LV_LVT_V = sv_dup_inc(proto_perl->Iutf8_X_LV_LVT_V, param);
13369 PL_utf8_toupper = sv_dup_inc(proto_perl->Iutf8_toupper, param);
13370 PL_utf8_totitle = sv_dup_inc(proto_perl->Iutf8_totitle, param);
13371 PL_utf8_tolower = sv_dup_inc(proto_perl->Iutf8_tolower, param);
13372 PL_utf8_tofold = sv_dup_inc(proto_perl->Iutf8_tofold, param);
13373 PL_utf8_idstart = sv_dup_inc(proto_perl->Iutf8_idstart, param);
13374 PL_utf8_xidstart = sv_dup_inc(proto_perl->Iutf8_xidstart, param);
13375 PL_utf8_idcont = sv_dup_inc(proto_perl->Iutf8_idcont, param);
13376 PL_utf8_xidcont = sv_dup_inc(proto_perl->Iutf8_xidcont, param);
13377 PL_utf8_foldable = hv_dup_inc(proto_perl->Iutf8_foldable, param);
13380 if (proto_perl->Ipsig_pend) {
13381 Newxz(PL_psig_pend, SIG_SIZE, int);
13384 PL_psig_pend = (int*)NULL;
13387 if (proto_perl->Ipsig_name) {
13388 Newx(PL_psig_name, 2 * SIG_SIZE, SV*);
13389 sv_dup_inc_multiple(proto_perl->Ipsig_name, PL_psig_name, 2 * SIG_SIZE,
13391 PL_psig_ptr = PL_psig_name + SIG_SIZE;
13394 PL_psig_ptr = (SV**)NULL;
13395 PL_psig_name = (SV**)NULL;
13398 if (flags & CLONEf_COPY_STACKS) {
13399 Newx(PL_tmps_stack, PL_tmps_max, SV*);
13400 sv_dup_inc_multiple(proto_perl->Itmps_stack, PL_tmps_stack,
13401 PL_tmps_ix+1, param);
13403 /* next PUSHMARK() sets *(PL_markstack_ptr+1) */
13404 i = proto_perl->Imarkstack_max - proto_perl->Imarkstack;
13405 Newxz(PL_markstack, i, I32);
13406 PL_markstack_max = PL_markstack + (proto_perl->Imarkstack_max
13407 - proto_perl->Imarkstack);
13408 PL_markstack_ptr = PL_markstack + (proto_perl->Imarkstack_ptr
13409 - proto_perl->Imarkstack);
13410 Copy(proto_perl->Imarkstack, PL_markstack,
13411 PL_markstack_ptr - PL_markstack + 1, I32);
13413 /* next push_scope()/ENTER sets PL_scopestack[PL_scopestack_ix]
13414 * NOTE: unlike the others! */
13415 Newxz(PL_scopestack, PL_scopestack_max, I32);
13416 Copy(proto_perl->Iscopestack, PL_scopestack, PL_scopestack_ix, I32);
13419 Newxz(PL_scopestack_name, PL_scopestack_max, const char *);
13420 Copy(proto_perl->Iscopestack_name, PL_scopestack_name, PL_scopestack_ix, const char *);
13422 /* NOTE: si_dup() looks at PL_markstack */
13423 PL_curstackinfo = si_dup(proto_perl->Icurstackinfo, param);
13425 /* PL_curstack = PL_curstackinfo->si_stack; */
13426 PL_curstack = av_dup(proto_perl->Icurstack, param);
13427 PL_mainstack = av_dup(proto_perl->Imainstack, param);
13429 /* next PUSHs() etc. set *(PL_stack_sp+1) */
13430 PL_stack_base = AvARRAY(PL_curstack);
13431 PL_stack_sp = PL_stack_base + (proto_perl->Istack_sp
13432 - proto_perl->Istack_base);
13433 PL_stack_max = PL_stack_base + AvMAX(PL_curstack);
13435 /*Newxz(PL_savestack, PL_savestack_max, ANY);*/
13436 PL_savestack = ss_dup(proto_perl, param);
13440 ENTER; /* perl_destruct() wants to LEAVE; */
13443 PL_statgv = gv_dup(proto_perl->Istatgv, param);
13444 PL_statname = sv_dup_inc(proto_perl->Istatname, param);
13446 PL_rs = sv_dup_inc(proto_perl->Irs, param);
13447 PL_last_in_gv = gv_dup(proto_perl->Ilast_in_gv, param);
13448 PL_defoutgv = gv_dup_inc(proto_perl->Idefoutgv, param);
13449 PL_toptarget = sv_dup_inc(proto_perl->Itoptarget, param);
13450 PL_bodytarget = sv_dup_inc(proto_perl->Ibodytarget, param);
13451 PL_formtarget = sv_dup(proto_perl->Iformtarget, param);
13453 PL_errors = sv_dup_inc(proto_perl->Ierrors, param);
13455 PL_sortcop = (OP*)any_dup(proto_perl->Isortcop, proto_perl);
13456 PL_sortstash = hv_dup(proto_perl->Isortstash, param);
13457 PL_firstgv = gv_dup(proto_perl->Ifirstgv, param);
13458 PL_secondgv = gv_dup(proto_perl->Isecondgv, param);
13460 PL_stashcache = newHV();
13462 PL_watchaddr = (char **) ptr_table_fetch(PL_ptr_table,
13463 proto_perl->Iwatchaddr);
13464 PL_watchok = PL_watchaddr ? * PL_watchaddr : NULL;
13465 if (PL_debug && PL_watchaddr) {
13466 PerlIO_printf(Perl_debug_log,
13467 "WATCHING: %"UVxf" cloned as %"UVxf" with value %"UVxf"\n",
13468 PTR2UV(proto_perl->Iwatchaddr), PTR2UV(PL_watchaddr),
13469 PTR2UV(PL_watchok));
13472 PL_registered_mros = hv_dup_inc(proto_perl->Iregistered_mros, param);
13473 PL_blockhooks = av_dup_inc(proto_perl->Iblockhooks, param);
13474 PL_utf8_foldclosures = hv_dup_inc(proto_perl->Iutf8_foldclosures, param);
13476 /* Call the ->CLONE method, if it exists, for each of the stashes
13477 identified by sv_dup() above.
13479 while(av_len(param->stashes) != -1) {
13480 HV* const stash = MUTABLE_HV(av_shift(param->stashes));
13481 GV* const cloner = gv_fetchmethod_autoload(stash, "CLONE", 0);
13482 if (cloner && GvCV(cloner)) {
13487 mXPUSHs(newSVhek(HvNAME_HEK(stash)));
13489 call_sv(MUTABLE_SV(GvCV(cloner)), G_DISCARD);
13495 if (!(flags & CLONEf_KEEP_PTR_TABLE)) {
13496 ptr_table_free(PL_ptr_table);
13497 PL_ptr_table = NULL;
13500 if (!(flags & CLONEf_COPY_STACKS)) {
13501 unreferenced_to_tmp_stack(param->unreferenced);
13504 SvREFCNT_dec(param->stashes);
13506 /* orphaned? eg threads->new inside BEGIN or use */
13507 if (PL_compcv && ! SvREFCNT(PL_compcv)) {
13508 SvREFCNT_inc_simple_void(PL_compcv);
13509 SAVEFREESV(PL_compcv);
13516 S_unreferenced_to_tmp_stack(pTHX_ AV *const unreferenced)
13518 PERL_ARGS_ASSERT_UNREFERENCED_TO_TMP_STACK;
13520 if (AvFILLp(unreferenced) > -1) {
13521 SV **svp = AvARRAY(unreferenced);
13522 SV **const last = svp + AvFILLp(unreferenced);
13526 if (SvREFCNT(*svp) == 1)
13528 } while (++svp <= last);
13530 EXTEND_MORTAL(count);
13531 svp = AvARRAY(unreferenced);
13534 if (SvREFCNT(*svp) == 1) {
13535 /* Our reference is the only one to this SV. This means that
13536 in this thread, the scalar effectively has a 0 reference.
13537 That doesn't work (cleanup never happens), so donate our
13538 reference to it onto the save stack. */
13539 PL_tmps_stack[++PL_tmps_ix] = *svp;
13541 /* As an optimisation, because we are already walking the
13542 entire array, instead of above doing either
13543 SvREFCNT_inc(*svp) or *svp = &PL_sv_undef, we can instead
13544 release our reference to the scalar, so that at the end of
13545 the array owns zero references to the scalars it happens to
13546 point to. We are effectively converting the array from
13547 AvREAL() on to AvREAL() off. This saves the av_clear()
13548 (triggered by the SvREFCNT_dec(unreferenced) below) from
13549 walking the array a second time. */
13550 SvREFCNT_dec(*svp);
13553 } while (++svp <= last);
13554 AvREAL_off(unreferenced);
13556 SvREFCNT_dec(unreferenced);
13560 Perl_clone_params_del(CLONE_PARAMS *param)
13562 /* This seemingly funky ordering keeps the build with PERL_GLOBAL_STRUCT
13564 PerlInterpreter *const to = param->new_perl;
13566 PerlInterpreter *const was = PERL_GET_THX;
13568 PERL_ARGS_ASSERT_CLONE_PARAMS_DEL;
13574 SvREFCNT_dec(param->stashes);
13575 if (param->unreferenced)
13576 unreferenced_to_tmp_stack(param->unreferenced);
13586 Perl_clone_params_new(PerlInterpreter *const from, PerlInterpreter *const to)
13589 /* Need to play this game, as newAV() can call safesysmalloc(), and that
13590 does a dTHX; to get the context from thread local storage.
13591 FIXME - under PERL_CORE Newx(), Safefree() and friends should expand to
13592 a version that passes in my_perl. */
13593 PerlInterpreter *const was = PERL_GET_THX;
13594 CLONE_PARAMS *param;
13596 PERL_ARGS_ASSERT_CLONE_PARAMS_NEW;
13602 /* Given that we've set the context, we can do this unshared. */
13603 Newx(param, 1, CLONE_PARAMS);
13606 param->proto_perl = from;
13607 param->new_perl = to;
13608 param->stashes = (AV *)Perl_newSV_type(to, SVt_PVAV);
13609 AvREAL_off(param->stashes);
13610 param->unreferenced = (AV *)Perl_newSV_type(to, SVt_PVAV);
13618 #endif /* USE_ITHREADS */
13621 =head1 Unicode Support
13623 =for apidoc sv_recode_to_utf8
13625 The encoding is assumed to be an Encode object, on entry the PV
13626 of the sv is assumed to be octets in that encoding, and the sv
13627 will be converted into Unicode (and UTF-8).
13629 If the sv already is UTF-8 (or if it is not POK), or if the encoding
13630 is not a reference, nothing is done to the sv. If the encoding is not
13631 an C<Encode::XS> Encoding object, bad things will happen.
13632 (See F<lib/encoding.pm> and L<Encode>).
13634 The PV of the sv is returned.
13639 Perl_sv_recode_to_utf8(pTHX_ SV *sv, SV *encoding)
13643 PERL_ARGS_ASSERT_SV_RECODE_TO_UTF8;
13645 if (SvPOK(sv) && !SvUTF8(sv) && !IN_BYTES && SvROK(encoding)) {
13659 Passing sv_yes is wrong - it needs to be or'ed set of constants
13660 for Encode::XS, while UTf-8 decode (currently) assumes a true value means
13661 remove converted chars from source.
13663 Both will default the value - let them.
13665 XPUSHs(&PL_sv_yes);
13668 call_method("decode", G_SCALAR);
13672 s = SvPV_const(uni, len);
13673 if (s != SvPVX_const(sv)) {
13674 SvGROW(sv, len + 1);
13675 Move(s, SvPVX(sv), len + 1, char);
13676 SvCUR_set(sv, len);
13680 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
13681 /* clear pos and any utf8 cache */
13682 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
13685 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
13686 magic_setutf8(sv,mg); /* clear UTF8 cache */
13691 return SvPOKp(sv) ? SvPVX(sv) : NULL;
13695 =for apidoc sv_cat_decode
13697 The encoding is assumed to be an Encode object, the PV of the ssv is
13698 assumed to be octets in that encoding and decoding the input starts
13699 from the position which (PV + *offset) pointed to. The dsv will be
13700 concatenated the decoded UTF-8 string from ssv. Decoding will terminate
13701 when the string tstr appears in decoding output or the input ends on
13702 the PV of the ssv. The value which the offset points will be modified
13703 to the last input position on the ssv.
13705 Returns TRUE if the terminator was found, else returns FALSE.
13710 Perl_sv_cat_decode(pTHX_ SV *dsv, SV *encoding,
13711 SV *ssv, int *offset, char *tstr, int tlen)
13716 PERL_ARGS_ASSERT_SV_CAT_DECODE;
13718 if (SvPOK(ssv) && SvPOK(dsv) && SvROK(encoding) && offset) {
13729 offsv = newSViv(*offset);
13731 mXPUSHp(tstr, tlen);
13733 call_method("cat_decode", G_SCALAR);
13735 ret = SvTRUE(TOPs);
13736 *offset = SvIV(offsv);
13742 Perl_croak(aTHX_ "Invalid argument to sv_cat_decode");
13747 /* ---------------------------------------------------------------------
13749 * support functions for report_uninit()
13752 /* the maxiumum size of array or hash where we will scan looking
13753 * for the undefined element that triggered the warning */
13755 #define FUV_MAX_SEARCH_SIZE 1000
13757 /* Look for an entry in the hash whose value has the same SV as val;
13758 * If so, return a mortal copy of the key. */
13761 S_find_hash_subscript(pTHX_ const HV *const hv, const SV *const val)
13764 register HE **array;
13767 PERL_ARGS_ASSERT_FIND_HASH_SUBSCRIPT;
13769 if (!hv || SvMAGICAL(hv) || !HvARRAY(hv) ||
13770 (HvTOTALKEYS(hv) > FUV_MAX_SEARCH_SIZE))
13773 array = HvARRAY(hv);
13775 for (i=HvMAX(hv); i>0; i--) {
13776 register HE *entry;
13777 for (entry = array[i]; entry; entry = HeNEXT(entry)) {
13778 if (HeVAL(entry) != val)
13780 if ( HeVAL(entry) == &PL_sv_undef ||
13781 HeVAL(entry) == &PL_sv_placeholder)
13785 if (HeKLEN(entry) == HEf_SVKEY)
13786 return sv_mortalcopy(HeKEY_sv(entry));
13787 return sv_2mortal(newSVhek(HeKEY_hek(entry)));
13793 /* Look for an entry in the array whose value has the same SV as val;
13794 * If so, return the index, otherwise return -1. */
13797 S_find_array_subscript(pTHX_ const AV *const av, const SV *const val)
13801 PERL_ARGS_ASSERT_FIND_ARRAY_SUBSCRIPT;
13803 if (!av || SvMAGICAL(av) || !AvARRAY(av) ||
13804 (AvFILLp(av) > FUV_MAX_SEARCH_SIZE))
13807 if (val != &PL_sv_undef) {
13808 SV ** const svp = AvARRAY(av);
13811 for (i=AvFILLp(av); i>=0; i--)
13818 /* S_varname(): return the name of a variable, optionally with a subscript.
13819 * If gv is non-zero, use the name of that global, along with gvtype (one
13820 * of "$", "@", "%"); otherwise use the name of the lexical at pad offset
13821 * targ. Depending on the value of the subscript_type flag, return:
13824 #define FUV_SUBSCRIPT_NONE 1 /* "@foo" */
13825 #define FUV_SUBSCRIPT_ARRAY 2 /* "$foo[aindex]" */
13826 #define FUV_SUBSCRIPT_HASH 3 /* "$foo{keyname}" */
13827 #define FUV_SUBSCRIPT_WITHIN 4 /* "within @foo" */
13830 S_varname(pTHX_ const GV *const gv, const char gvtype, PADOFFSET targ,
13831 const SV *const keyname, I32 aindex, int subscript_type)
13834 SV * const name = sv_newmortal();
13837 buffer[0] = gvtype;
13840 /* as gv_fullname4(), but add literal '^' for $^FOO names */
13842 gv_fullname4(name, gv, buffer, 0);
13844 if ((unsigned int)SvPVX(name)[1] <= 26) {
13846 buffer[1] = SvPVX(name)[1] + 'A' - 1;
13848 /* Swap the 1 unprintable control character for the 2 byte pretty
13849 version - ie substr($name, 1, 1) = $buffer; */
13850 sv_insert(name, 1, 1, buffer, 2);
13854 CV * const cv = find_runcv(NULL);
13858 if (!cv || !CvPADLIST(cv))
13860 av = MUTABLE_AV((*av_fetch(CvPADLIST(cv), 0, FALSE)));
13861 sv = *av_fetch(av, targ, FALSE);
13862 sv_setpvn(name, SvPV_nolen_const(sv), SvCUR(sv));
13865 if (subscript_type == FUV_SUBSCRIPT_HASH) {
13866 SV * const sv = newSV(0);
13867 *SvPVX(name) = '$';
13868 Perl_sv_catpvf(aTHX_ name, "{%s}",
13869 pv_display(sv,SvPVX_const(keyname), SvCUR(keyname), 0, 32));
13872 else if (subscript_type == FUV_SUBSCRIPT_ARRAY) {
13873 *SvPVX(name) = '$';
13874 Perl_sv_catpvf(aTHX_ name, "[%"IVdf"]", (IV)aindex);
13876 else if (subscript_type == FUV_SUBSCRIPT_WITHIN) {
13877 /* We know that name has no magic, so can use 0 instead of SV_GMAGIC */
13878 Perl_sv_insert_flags(aTHX_ name, 0, 0, STR_WITH_LEN("within "), 0);
13886 =for apidoc find_uninit_var
13888 Find the name of the undefined variable (if any) that caused the operator o
13889 to issue a "Use of uninitialized value" warning.
13890 If match is true, only return a name if it's value matches uninit_sv.
13891 So roughly speaking, if a unary operator (such as OP_COS) generates a
13892 warning, then following the direct child of the op may yield an
13893 OP_PADSV or OP_GV that gives the name of the undefined variable. On the
13894 other hand, with OP_ADD there are two branches to follow, so we only print
13895 the variable name if we get an exact match.
13897 The name is returned as a mortal SV.
13899 Assumes that PL_op is the op that originally triggered the error, and that
13900 PL_comppad/PL_curpad points to the currently executing pad.
13906 S_find_uninit_var(pTHX_ const OP *const obase, const SV *const uninit_sv,
13912 const OP *o, *o2, *kid;
13914 if (!obase || (match && (!uninit_sv || uninit_sv == &PL_sv_undef ||
13915 uninit_sv == &PL_sv_placeholder)))
13918 switch (obase->op_type) {
13925 const bool pad = (obase->op_type == OP_PADAV || obase->op_type == OP_PADHV);
13926 const bool hash = (obase->op_type == OP_PADHV || obase->op_type == OP_RV2HV);
13929 int subscript_type = FUV_SUBSCRIPT_WITHIN;
13931 if (pad) { /* @lex, %lex */
13932 sv = PAD_SVl(obase->op_targ);
13936 if (cUNOPx(obase)->op_first->op_type == OP_GV) {
13937 /* @global, %global */
13938 gv = cGVOPx_gv(cUNOPx(obase)->op_first);
13941 sv = hash ? MUTABLE_SV(GvHV(gv)): MUTABLE_SV(GvAV(gv));
13943 else /* @{expr}, %{expr} */
13944 return find_uninit_var(cUNOPx(obase)->op_first,
13948 /* attempt to find a match within the aggregate */
13950 keysv = find_hash_subscript((const HV*)sv, uninit_sv);
13952 subscript_type = FUV_SUBSCRIPT_HASH;
13955 index = find_array_subscript((const AV *)sv, uninit_sv);
13957 subscript_type = FUV_SUBSCRIPT_ARRAY;
13960 if (match && subscript_type == FUV_SUBSCRIPT_WITHIN)
13963 return varname(gv, hash ? '%' : '@', obase->op_targ,
13964 keysv, index, subscript_type);
13968 if (match && PAD_SVl(obase->op_targ) != uninit_sv)
13970 return varname(NULL, '$', obase->op_targ,
13971 NULL, 0, FUV_SUBSCRIPT_NONE);
13974 gv = cGVOPx_gv(obase);
13975 if (!gv || (match && GvSV(gv) != uninit_sv) || !GvSTASH(gv))
13977 return varname(gv, '$', 0, NULL, 0, FUV_SUBSCRIPT_NONE);
13980 if (obase->op_flags & OPf_SPECIAL) { /* lexical array */
13983 AV *av = MUTABLE_AV(PAD_SV(obase->op_targ));
13984 if (!av || SvRMAGICAL(av))
13986 svp = av_fetch(av, (I32)obase->op_private, FALSE);
13987 if (!svp || *svp != uninit_sv)
13990 return varname(NULL, '$', obase->op_targ,
13991 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
13994 gv = cGVOPx_gv(obase);
13999 AV *const av = GvAV(gv);
14000 if (!av || SvRMAGICAL(av))
14002 svp = av_fetch(av, (I32)obase->op_private, FALSE);
14003 if (!svp || *svp != uninit_sv)
14006 return varname(gv, '$', 0,
14007 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
14012 o = cUNOPx(obase)->op_first;
14013 if (!o || o->op_type != OP_NULL ||
14014 ! (o->op_targ == OP_AELEM || o->op_targ == OP_HELEM))
14016 return find_uninit_var(cBINOPo->op_last, uninit_sv, match);
14020 if (PL_op == obase)
14021 /* $a[uninit_expr] or $h{uninit_expr} */
14022 return find_uninit_var(cBINOPx(obase)->op_last, uninit_sv, match);
14025 o = cBINOPx(obase)->op_first;
14026 kid = cBINOPx(obase)->op_last;
14028 /* get the av or hv, and optionally the gv */
14030 if (o->op_type == OP_PADAV || o->op_type == OP_PADHV) {
14031 sv = PAD_SV(o->op_targ);
14033 else if ((o->op_type == OP_RV2AV || o->op_type == OP_RV2HV)
14034 && cUNOPo->op_first->op_type == OP_GV)
14036 gv = cGVOPx_gv(cUNOPo->op_first);
14040 == OP_RV2HV ? MUTABLE_SV(GvHV(gv)) : MUTABLE_SV(GvAV(gv));
14045 if (kid && kid->op_type == OP_CONST && SvOK(cSVOPx_sv(kid))) {
14046 /* index is constant */
14050 if (obase->op_type == OP_HELEM) {
14051 HE* he = hv_fetch_ent(MUTABLE_HV(sv), cSVOPx_sv(kid), 0, 0);
14052 if (!he || HeVAL(he) != uninit_sv)
14056 SV * const * const svp = av_fetch(MUTABLE_AV(sv), SvIV(cSVOPx_sv(kid)), FALSE);
14057 if (!svp || *svp != uninit_sv)
14061 if (obase->op_type == OP_HELEM)
14062 return varname(gv, '%', o->op_targ,
14063 cSVOPx_sv(kid), 0, FUV_SUBSCRIPT_HASH);
14065 return varname(gv, '@', o->op_targ, NULL,
14066 SvIV(cSVOPx_sv(kid)), FUV_SUBSCRIPT_ARRAY);
14069 /* index is an expression;
14070 * attempt to find a match within the aggregate */
14071 if (obase->op_type == OP_HELEM) {
14072 SV * const keysv = find_hash_subscript((const HV*)sv, uninit_sv);
14074 return varname(gv, '%', o->op_targ,
14075 keysv, 0, FUV_SUBSCRIPT_HASH);
14079 = find_array_subscript((const AV *)sv, uninit_sv);
14081 return varname(gv, '@', o->op_targ,
14082 NULL, index, FUV_SUBSCRIPT_ARRAY);
14087 (o->op_type == OP_PADAV || o->op_type == OP_RV2AV)
14089 o->op_targ, NULL, 0, FUV_SUBSCRIPT_WITHIN);
14094 /* only examine RHS */
14095 return find_uninit_var(cBINOPx(obase)->op_first, uninit_sv, match);
14098 o = cUNOPx(obase)->op_first;
14099 if (o->op_type == OP_PUSHMARK)
14102 if (!o->op_sibling) {
14103 /* one-arg version of open is highly magical */
14105 if (o->op_type == OP_GV) { /* open FOO; */
14107 if (match && GvSV(gv) != uninit_sv)
14109 return varname(gv, '$', 0,
14110 NULL, 0, FUV_SUBSCRIPT_NONE);
14112 /* other possibilities not handled are:
14113 * open $x; or open my $x; should return '${*$x}'
14114 * open expr; should return '$'.expr ideally
14120 /* ops where $_ may be an implicit arg */
14124 if ( !(obase->op_flags & OPf_STACKED)) {
14125 if (uninit_sv == ((obase->op_private & OPpTARGET_MY)
14126 ? PAD_SVl(obase->op_targ)
14129 sv = sv_newmortal();
14130 sv_setpvs(sv, "$_");
14139 match = 1; /* print etc can return undef on defined args */
14140 /* skip filehandle as it can't produce 'undef' warning */
14141 o = cUNOPx(obase)->op_first;
14142 if ((obase->op_flags & OPf_STACKED) && o->op_type == OP_PUSHMARK)
14143 o = o->op_sibling->op_sibling;
14147 case OP_ENTEREVAL: /* could be eval $undef or $x='$undef'; eval $x */
14149 case OP_CUSTOM: /* XS or custom code could trigger random warnings */
14151 /* the following ops are capable of returning PL_sv_undef even for
14152 * defined arg(s) */
14171 case OP_GETPEERNAME:
14219 case OP_SMARTMATCH:
14228 /* XXX tmp hack: these two may call an XS sub, and currently
14229 XS subs don't have a SUB entry on the context stack, so CV and
14230 pad determination goes wrong, and BAD things happen. So, just
14231 don't try to determine the value under those circumstances.
14232 Need a better fix at dome point. DAPM 11/2007 */
14238 GV * const gv = gv_fetchpvs(".", GV_NOTQUAL, SVt_PV);
14239 if (gv && GvSV(gv) == uninit_sv)
14240 return newSVpvs_flags("$.", SVs_TEMP);
14245 /* def-ness of rval pos() is independent of the def-ness of its arg */
14246 if ( !(obase->op_flags & OPf_MOD))
14251 if (SvROK(PL_rs) && uninit_sv == SvRV(PL_rs))
14252 return newSVpvs_flags("${$/}", SVs_TEMP);
14257 if (!(obase->op_flags & OPf_KIDS))
14259 o = cUNOPx(obase)->op_first;
14265 /* if all except one arg are constant, or have no side-effects,
14266 * or are optimized away, then it's unambiguous */
14268 for (kid=o; kid; kid = kid->op_sibling) {
14270 const OPCODE type = kid->op_type;
14271 if ( (type == OP_CONST && SvOK(cSVOPx_sv(kid)))
14272 || (type == OP_NULL && ! (kid->op_flags & OPf_KIDS))
14273 || (type == OP_PUSHMARK)
14275 /* @$a and %$a, but not @a or %a */
14276 (type == OP_RV2AV || type == OP_RV2HV)
14277 && cUNOPx(kid)->op_first
14278 && cUNOPx(kid)->op_first->op_type != OP_GV
14283 if (o2) { /* more than one found */
14290 return find_uninit_var(o2, uninit_sv, match);
14292 /* scan all args */
14294 sv = find_uninit_var(o, uninit_sv, 1);
14306 =for apidoc report_uninit
14308 Print appropriate "Use of uninitialized variable" warning
14314 Perl_report_uninit(pTHX_ const SV *uninit_sv)
14318 SV* varname = NULL;
14320 varname = find_uninit_var(PL_op, uninit_sv,0);
14322 sv_insert(varname, 0, 0, " ", 1);
14324 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
14325 varname ? SvPV_nolen_const(varname) : "",
14326 " in ", OP_DESC(PL_op));
14329 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
14335 * c-indentation-style: bsd
14336 * c-basic-offset: 4
14337 * indent-tabs-mode: t
14340 * ex: set ts=8 sts=4 sw=4 noet: