3 * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, 2000,
4 * 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009 by Larry Wall
7 * You may distribute under the terms of either the GNU General Public
8 * License or the Artistic License, as specified in the README file.
13 * 'I wonder what the Entish is for "yes" and "no",' he thought.
16 * [p.480 of _The Lord of the Rings_, III/iv: "Treebeard"]
22 * This file contains the code that creates, manipulates and destroys
23 * scalar values (SVs). The other types (AV, HV, GV, etc.) reuse the
24 * structure of an SV, so their creation and destruction is handled
25 * here; higher-level functions are in av.c, hv.c, and so on. Opcode
26 * level functions (eg. substr, split, join) for each of the types are
36 # if __STDC_VERSION__ >= 199901L && !defined(VMS)
47 /* Missing proto on LynxOS */
48 char *gconvert(double, int, int, char *);
51 #ifdef PERL_UTF8_CACHE_ASSERT
52 /* if adding more checks watch out for the following tests:
53 * t/op/index.t t/op/length.t t/op/pat.t t/op/substr.t
54 * lib/utf8.t lib/Unicode/Collate/t/index.t
57 # define ASSERT_UTF8_CACHE(cache) \
58 STMT_START { if (cache) { assert((cache)[0] <= (cache)[1]); \
59 assert((cache)[2] <= (cache)[3]); \
60 assert((cache)[3] <= (cache)[1]);} \
63 # define ASSERT_UTF8_CACHE(cache) NOOP
66 #ifdef PERL_OLD_COPY_ON_WRITE
67 #define SV_COW_NEXT_SV(sv) INT2PTR(SV *,SvUVX(sv))
68 #define SV_COW_NEXT_SV_SET(current,next) SvUV_set(current, PTR2UV(next))
69 /* This is a pessimistic view. Scalar must be purely a read-write PV to copy-
73 /* ============================================================================
75 =head1 Allocation and deallocation of SVs.
77 An SV (or AV, HV, etc.) is allocated in two parts: the head (struct
78 sv, av, hv...) contains type and reference count information, and for
79 many types, a pointer to the body (struct xrv, xpv, xpviv...), which
80 contains fields specific to each type. Some types store all they need
81 in the head, so don't have a body.
83 In all but the most memory-paranoid configurations (ex: PURIFY), heads
84 and bodies are allocated out of arenas, which by default are
85 approximately 4K chunks of memory parcelled up into N heads or bodies.
86 Sv-bodies are allocated by their sv-type, guaranteeing size
87 consistency needed to allocate safely from arrays.
89 For SV-heads, the first slot in each arena is reserved, and holds a
90 link to the next arena, some flags, and a note of the number of slots.
91 Snaked through each arena chain is a linked list of free items; when
92 this becomes empty, an extra arena is allocated and divided up into N
93 items which are threaded into the free list.
95 SV-bodies are similar, but they use arena-sets by default, which
96 separate the link and info from the arena itself, and reclaim the 1st
97 slot in the arena. SV-bodies are further described later.
99 The following global variables are associated with arenas:
101 PL_sv_arenaroot pointer to list of SV arenas
102 PL_sv_root pointer to list of free SV structures
104 PL_body_arenas head of linked-list of body arenas
105 PL_body_roots[] array of pointers to list of free bodies of svtype
106 arrays are indexed by the svtype needed
108 A few special SV heads are not allocated from an arena, but are
109 instead directly created in the interpreter structure, eg PL_sv_undef.
110 The size of arenas can be changed from the default by setting
111 PERL_ARENA_SIZE appropriately at compile time.
113 The SV arena serves the secondary purpose of allowing still-live SVs
114 to be located and destroyed during final cleanup.
116 At the lowest level, the macros new_SV() and del_SV() grab and free
117 an SV head. (If debugging with -DD, del_SV() calls the function S_del_sv()
118 to return the SV to the free list with error checking.) new_SV() calls
119 more_sv() / sv_add_arena() to add an extra arena if the free list is empty.
120 SVs in the free list have their SvTYPE field set to all ones.
122 At the time of very final cleanup, sv_free_arenas() is called from
123 perl_destruct() to physically free all the arenas allocated since the
124 start of the interpreter.
126 The function visit() scans the SV arenas list, and calls a specified
127 function for each SV it finds which is still live - ie which has an SvTYPE
128 other than all 1's, and a non-zero SvREFCNT. visit() is used by the
129 following functions (specified as [function that calls visit()] / [function
130 called by visit() for each SV]):
132 sv_report_used() / do_report_used()
133 dump all remaining SVs (debugging aid)
135 sv_clean_objs() / do_clean_objs(),do_clean_named_objs(),
136 do_clean_named_io_objs()
137 Attempt to free all objects pointed to by RVs,
138 and try to do the same for all objects indirectly
139 referenced by typeglobs too. Called once from
140 perl_destruct(), prior to calling sv_clean_all()
143 sv_clean_all() / do_clean_all()
144 SvREFCNT_dec(sv) each remaining SV, possibly
145 triggering an sv_free(). It also sets the
146 SVf_BREAK flag on the SV to indicate that the
147 refcnt has been artificially lowered, and thus
148 stopping sv_free() from giving spurious warnings
149 about SVs which unexpectedly have a refcnt
150 of zero. called repeatedly from perl_destruct()
151 until there are no SVs left.
153 =head2 Arena allocator API Summary
155 Private API to rest of sv.c
159 new_XPVNV(), del_XPVGV(),
164 sv_report_used(), sv_clean_objs(), sv_clean_all(), sv_free_arenas()
168 * ========================================================================= */
171 * "A time to plant, and a time to uproot what was planted..."
175 # define MEM_LOG_NEW_SV(sv, file, line, func) \
176 Perl_mem_log_new_sv(sv, file, line, func)
177 # define MEM_LOG_DEL_SV(sv, file, line, func) \
178 Perl_mem_log_del_sv(sv, file, line, func)
180 # define MEM_LOG_NEW_SV(sv, file, line, func) NOOP
181 # define MEM_LOG_DEL_SV(sv, file, line, func) NOOP
184 #ifdef DEBUG_LEAKING_SCALARS
185 # define FREE_SV_DEBUG_FILE(sv) Safefree((sv)->sv_debug_file)
186 # define DEBUG_SV_SERIAL(sv) \
187 DEBUG_m(PerlIO_printf(Perl_debug_log, "0x%"UVxf": (%05ld) del_SV\n", \
188 PTR2UV(sv), (long)(sv)->sv_debug_serial))
190 # define FREE_SV_DEBUG_FILE(sv)
191 # define DEBUG_SV_SERIAL(sv) NOOP
195 # define SvARENA_CHAIN(sv) ((sv)->sv_u.svu_rv)
196 # define SvARENA_CHAIN_SET(sv,val) (sv)->sv_u.svu_rv = MUTABLE_SV((val))
197 /* Whilst I'd love to do this, it seems that things like to check on
199 # define POSION_SV_HEAD(sv) PoisonNew(sv, 1, struct STRUCT_SV)
201 # define POSION_SV_HEAD(sv) PoisonNew(&SvANY(sv), 1, void *), \
202 PoisonNew(&SvREFCNT(sv), 1, U32)
204 # define SvARENA_CHAIN(sv) SvANY(sv)
205 # define SvARENA_CHAIN_SET(sv,val) SvANY(sv) = (void *)(val)
206 # define POSION_SV_HEAD(sv)
209 /* Mark an SV head as unused, and add to free list.
211 * If SVf_BREAK is set, skip adding it to the free list, as this SV had
212 * its refcount artificially decremented during global destruction, so
213 * there may be dangling pointers to it. The last thing we want in that
214 * case is for it to be reused. */
216 #define plant_SV(p) \
218 const U32 old_flags = SvFLAGS(p); \
219 MEM_LOG_DEL_SV(p, __FILE__, __LINE__, FUNCTION__); \
220 DEBUG_SV_SERIAL(p); \
221 FREE_SV_DEBUG_FILE(p); \
223 SvFLAGS(p) = SVTYPEMASK; \
224 if (!(old_flags & SVf_BREAK)) { \
225 SvARENA_CHAIN_SET(p, PL_sv_root); \
231 #define uproot_SV(p) \
234 PL_sv_root = MUTABLE_SV(SvARENA_CHAIN(p)); \
239 /* make some more SVs by adding another arena */
246 char *chunk; /* must use New here to match call to */
247 Newx(chunk,PERL_ARENA_SIZE,char); /* Safefree() in sv_free_arenas() */
248 sv_add_arena(chunk, PERL_ARENA_SIZE, 0);
253 /* new_SV(): return a new, empty SV head */
255 #ifdef DEBUG_LEAKING_SCALARS
256 /* provide a real function for a debugger to play with */
258 S_new_SV(pTHX_ const char *file, int line, const char *func)
265 sv = S_more_sv(aTHX);
269 sv->sv_debug_optype = PL_op ? PL_op->op_type : 0;
270 sv->sv_debug_line = (U16) (PL_parser && PL_parser->copline != NOLINE
276 sv->sv_debug_inpad = 0;
277 sv->sv_debug_parent = NULL;
278 sv->sv_debug_file = PL_curcop ? savepv(CopFILE(PL_curcop)): NULL;
280 sv->sv_debug_serial = PL_sv_serial++;
282 MEM_LOG_NEW_SV(sv, file, line, func);
283 DEBUG_m(PerlIO_printf(Perl_debug_log, "0x%"UVxf": (%05ld) new_SV (from %s:%d [%s])\n",
284 PTR2UV(sv), (long)sv->sv_debug_serial, file, line, func));
288 # define new_SV(p) (p)=S_new_SV(aTHX_ __FILE__, __LINE__, FUNCTION__)
296 (p) = S_more_sv(aTHX); \
300 MEM_LOG_NEW_SV(p, __FILE__, __LINE__, FUNCTION__); \
305 /* del_SV(): return an empty SV head to the free list */
318 S_del_sv(pTHX_ SV *p)
322 PERL_ARGS_ASSERT_DEL_SV;
327 for (sva = PL_sv_arenaroot; sva; sva = MUTABLE_SV(SvANY(sva))) {
328 const SV * const sv = sva + 1;
329 const SV * const svend = &sva[SvREFCNT(sva)];
330 if (p >= sv && p < svend) {
336 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL),
337 "Attempt to free non-arena SV: 0x%"UVxf
338 pTHX__FORMAT, PTR2UV(p) pTHX__VALUE);
345 #else /* ! DEBUGGING */
347 #define del_SV(p) plant_SV(p)
349 #endif /* DEBUGGING */
353 =head1 SV Manipulation Functions
355 =for apidoc sv_add_arena
357 Given a chunk of memory, link it to the head of the list of arenas,
358 and split it into a list of free SVs.
364 S_sv_add_arena(pTHX_ char *const ptr, const U32 size, const U32 flags)
367 SV *const sva = MUTABLE_SV(ptr);
371 PERL_ARGS_ASSERT_SV_ADD_ARENA;
373 /* The first SV in an arena isn't an SV. */
374 SvANY(sva) = (void *) PL_sv_arenaroot; /* ptr to next arena */
375 SvREFCNT(sva) = size / sizeof(SV); /* number of SV slots */
376 SvFLAGS(sva) = flags; /* FAKE if not to be freed */
378 PL_sv_arenaroot = sva;
379 PL_sv_root = sva + 1;
381 svend = &sva[SvREFCNT(sva) - 1];
384 SvARENA_CHAIN_SET(sv, (sv + 1));
388 /* Must always set typemask because it's always checked in on cleanup
389 when the arenas are walked looking for objects. */
390 SvFLAGS(sv) = SVTYPEMASK;
393 SvARENA_CHAIN_SET(sv, 0);
397 SvFLAGS(sv) = SVTYPEMASK;
400 /* visit(): call the named function for each non-free SV in the arenas
401 * whose flags field matches the flags/mask args. */
404 S_visit(pTHX_ SVFUNC_t f, const U32 flags, const U32 mask)
410 PERL_ARGS_ASSERT_VISIT;
412 for (sva = PL_sv_arenaroot; sva; sva = MUTABLE_SV(SvANY(sva))) {
413 register const SV * const svend = &sva[SvREFCNT(sva)];
415 for (sv = sva + 1; sv < svend; ++sv) {
416 if (SvTYPE(sv) != (svtype)SVTYPEMASK
417 && (sv->sv_flags & mask) == flags
430 /* called by sv_report_used() for each live SV */
433 do_report_used(pTHX_ SV *const sv)
435 if (SvTYPE(sv) != (svtype)SVTYPEMASK) {
436 PerlIO_printf(Perl_debug_log, "****\n");
443 =for apidoc sv_report_used
445 Dump the contents of all SVs not yet freed (debugging aid).
451 Perl_sv_report_used(pTHX)
454 visit(do_report_used, 0, 0);
460 /* called by sv_clean_objs() for each live SV */
463 do_clean_objs(pTHX_ SV *const ref)
468 SV * const target = SvRV(ref);
469 if (SvOBJECT(target)) {
470 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning object ref:\n "), sv_dump(ref)));
471 if (SvWEAKREF(ref)) {
472 sv_del_backref(target, ref);
478 SvREFCNT_dec(target);
483 /* XXX Might want to check arrays, etc. */
487 /* clear any slots in a GV which hold objects - except IO;
488 * called by sv_clean_objs() for each live GV */
491 do_clean_named_objs(pTHX_ SV *const sv)
495 assert(SvTYPE(sv) == SVt_PVGV);
496 assert(isGV_with_GP(sv));
500 /* freeing GP entries may indirectly free the current GV;
501 * hold onto it while we mess with the GP slots */
504 if ( ((obj = GvSV(sv) )) && SvOBJECT(obj)) {
505 DEBUG_D((PerlIO_printf(Perl_debug_log,
506 "Cleaning named glob SV object:\n "), sv_dump(obj)));
510 if ( ((obj = MUTABLE_SV(GvAV(sv)) )) && SvOBJECT(obj)) {
511 DEBUG_D((PerlIO_printf(Perl_debug_log,
512 "Cleaning named glob AV object:\n "), sv_dump(obj)));
516 if ( ((obj = MUTABLE_SV(GvHV(sv)) )) && SvOBJECT(obj)) {
517 DEBUG_D((PerlIO_printf(Perl_debug_log,
518 "Cleaning named glob HV object:\n "), sv_dump(obj)));
522 if ( ((obj = MUTABLE_SV(GvCV(sv)) )) && SvOBJECT(obj)) {
523 DEBUG_D((PerlIO_printf(Perl_debug_log,
524 "Cleaning named glob CV object:\n "), sv_dump(obj)));
528 SvREFCNT_dec(sv); /* undo the inc above */
531 /* clear any IO slots in a GV which hold objects (except stderr, defout);
532 * called by sv_clean_objs() for each live GV */
535 do_clean_named_io_objs(pTHX_ SV *const sv)
539 assert(SvTYPE(sv) == SVt_PVGV);
540 assert(isGV_with_GP(sv));
541 if (!GvGP(sv) || sv == (SV*)PL_stderrgv || sv == (SV*)PL_defoutgv)
545 if ( ((obj = MUTABLE_SV(GvIO(sv)) )) && SvOBJECT(obj)) {
546 DEBUG_D((PerlIO_printf(Perl_debug_log,
547 "Cleaning named glob IO object:\n "), sv_dump(obj)));
551 SvREFCNT_dec(sv); /* undo the inc above */
554 /* Void wrapper to pass to visit() */
556 do_curse(pTHX_ SV * const sv) {
557 if ((PL_stderrgv && GvGP(PL_stderrgv) && (SV*)GvIO(PL_stderrgv) == sv)
558 || (PL_defoutgv && GvGP(PL_defoutgv) && (SV*)GvIO(PL_defoutgv) == sv))
564 =for apidoc sv_clean_objs
566 Attempt to destroy all objects not yet freed.
572 Perl_sv_clean_objs(pTHX)
576 PL_in_clean_objs = TRUE;
577 visit(do_clean_objs, SVf_ROK, SVf_ROK);
578 /* Some barnacles may yet remain, clinging to typeglobs.
579 * Run the non-IO destructors first: they may want to output
580 * error messages, close files etc */
581 visit(do_clean_named_objs, SVt_PVGV|SVpgv_GP, SVTYPEMASK|SVp_POK|SVpgv_GP);
582 visit(do_clean_named_io_objs, SVt_PVGV|SVpgv_GP, SVTYPEMASK|SVp_POK|SVpgv_GP);
583 /* And if there are some very tenacious barnacles clinging to arrays,
584 closures, or what have you.... */
585 visit(do_curse, SVs_OBJECT, SVs_OBJECT);
586 olddef = PL_defoutgv;
587 PL_defoutgv = NULL; /* disable skip of PL_defoutgv */
588 if (olddef && isGV_with_GP(olddef))
589 do_clean_named_io_objs(aTHX_ MUTABLE_SV(olddef));
590 olderr = PL_stderrgv;
591 PL_stderrgv = NULL; /* disable skip of PL_stderrgv */
592 if (olderr && isGV_with_GP(olderr))
593 do_clean_named_io_objs(aTHX_ MUTABLE_SV(olderr));
594 SvREFCNT_dec(olddef);
595 PL_in_clean_objs = FALSE;
598 /* called by sv_clean_all() for each live SV */
601 do_clean_all(pTHX_ SV *const sv)
604 if (sv == (const SV *) PL_fdpid || sv == (const SV *)PL_strtab) {
605 /* don't clean pid table and strtab */
608 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning loops: SV at 0x%"UVxf"\n", PTR2UV(sv)) ));
609 SvFLAGS(sv) |= SVf_BREAK;
614 =for apidoc sv_clean_all
616 Decrement the refcnt of each remaining SV, possibly triggering a
617 cleanup. This function may have to be called multiple times to free
618 SVs which are in complex self-referential hierarchies.
624 Perl_sv_clean_all(pTHX)
628 PL_in_clean_all = TRUE;
629 cleaned = visit(do_clean_all, 0,0);
634 ARENASETS: a meta-arena implementation which separates arena-info
635 into struct arena_set, which contains an array of struct
636 arena_descs, each holding info for a single arena. By separating
637 the meta-info from the arena, we recover the 1st slot, formerly
638 borrowed for list management. The arena_set is about the size of an
639 arena, avoiding the needless malloc overhead of a naive linked-list.
641 The cost is 1 arena-set malloc per ~320 arena-mallocs, + the unused
642 memory in the last arena-set (1/2 on average). In trade, we get
643 back the 1st slot in each arena (ie 1.7% of a CV-arena, less for
644 smaller types). The recovery of the wasted space allows use of
645 small arenas for large, rare body types, by changing array* fields
646 in body_details_by_type[] below.
649 char *arena; /* the raw storage, allocated aligned */
650 size_t size; /* its size ~4k typ */
651 svtype utype; /* bodytype stored in arena */
656 /* Get the maximum number of elements in set[] such that struct arena_set
657 will fit within PERL_ARENA_SIZE, which is probably just under 4K, and
658 therefore likely to be 1 aligned memory page. */
660 #define ARENAS_PER_SET ((PERL_ARENA_SIZE - sizeof(struct arena_set*) \
661 - 2 * sizeof(int)) / sizeof (struct arena_desc))
664 struct arena_set* next;
665 unsigned int set_size; /* ie ARENAS_PER_SET */
666 unsigned int curr; /* index of next available arena-desc */
667 struct arena_desc set[ARENAS_PER_SET];
671 =for apidoc sv_free_arenas
673 Deallocate the memory used by all arenas. Note that all the individual SV
674 heads and bodies within the arenas must already have been freed.
679 Perl_sv_free_arenas(pTHX)
686 /* Free arenas here, but be careful about fake ones. (We assume
687 contiguity of the fake ones with the corresponding real ones.) */
689 for (sva = PL_sv_arenaroot; sva; sva = svanext) {
690 svanext = MUTABLE_SV(SvANY(sva));
691 while (svanext && SvFAKE(svanext))
692 svanext = MUTABLE_SV(SvANY(svanext));
699 struct arena_set *aroot = (struct arena_set*) PL_body_arenas;
702 struct arena_set *current = aroot;
705 assert(aroot->set[i].arena);
706 Safefree(aroot->set[i].arena);
714 i = PERL_ARENA_ROOTS_SIZE;
716 PL_body_roots[i] = 0;
723 Here are mid-level routines that manage the allocation of bodies out
724 of the various arenas. There are 5 kinds of arenas:
726 1. SV-head arenas, which are discussed and handled above
727 2. regular body arenas
728 3. arenas for reduced-size bodies
731 Arena types 2 & 3 are chained by body-type off an array of
732 arena-root pointers, which is indexed by svtype. Some of the
733 larger/less used body types are malloced singly, since a large
734 unused block of them is wasteful. Also, several svtypes dont have
735 bodies; the data fits into the sv-head itself. The arena-root
736 pointer thus has a few unused root-pointers (which may be hijacked
737 later for arena types 4,5)
739 3 differs from 2 as an optimization; some body types have several
740 unused fields in the front of the structure (which are kept in-place
741 for consistency). These bodies can be allocated in smaller chunks,
742 because the leading fields arent accessed. Pointers to such bodies
743 are decremented to point at the unused 'ghost' memory, knowing that
744 the pointers are used with offsets to the real memory.
747 =head1 SV-Body Allocation
749 Allocation of SV-bodies is similar to SV-heads, differing as follows;
750 the allocation mechanism is used for many body types, so is somewhat
751 more complicated, it uses arena-sets, and has no need for still-live
754 At the outermost level, (new|del)_X*V macros return bodies of the
755 appropriate type. These macros call either (new|del)_body_type or
756 (new|del)_body_allocated macro pairs, depending on specifics of the
757 type. Most body types use the former pair, the latter pair is used to
758 allocate body types with "ghost fields".
760 "ghost fields" are fields that are unused in certain types, and
761 consequently don't need to actually exist. They are declared because
762 they're part of a "base type", which allows use of functions as
763 methods. The simplest examples are AVs and HVs, 2 aggregate types
764 which don't use the fields which support SCALAR semantics.
766 For these types, the arenas are carved up into appropriately sized
767 chunks, we thus avoid wasted memory for those unaccessed members.
768 When bodies are allocated, we adjust the pointer back in memory by the
769 size of the part not allocated, so it's as if we allocated the full
770 structure. (But things will all go boom if you write to the part that
771 is "not there", because you'll be overwriting the last members of the
772 preceding structure in memory.)
774 We calculate the correction using the STRUCT_OFFSET macro on the first
775 member present. If the allocated structure is smaller (no initial NV
776 actually allocated) then the net effect is to subtract the size of the NV
777 from the pointer, to return a new pointer as if an initial NV were actually
778 allocated. (We were using structures named *_allocated for this, but
779 this turned out to be a subtle bug, because a structure without an NV
780 could have a lower alignment constraint, but the compiler is allowed to
781 optimised accesses based on the alignment constraint of the actual pointer
782 to the full structure, for example, using a single 64 bit load instruction
783 because it "knows" that two adjacent 32 bit members will be 8-byte aligned.)
785 This is the same trick as was used for NV and IV bodies. Ironically it
786 doesn't need to be used for NV bodies any more, because NV is now at
787 the start of the structure. IV bodies don't need it either, because
788 they are no longer allocated.
790 In turn, the new_body_* allocators call S_new_body(), which invokes
791 new_body_inline macro, which takes a lock, and takes a body off the
792 linked list at PL_body_roots[sv_type], calling Perl_more_bodies() if
793 necessary to refresh an empty list. Then the lock is released, and
794 the body is returned.
796 Perl_more_bodies allocates a new arena, and carves it up into an array of N
797 bodies, which it strings into a linked list. It looks up arena-size
798 and body-size from the body_details table described below, thus
799 supporting the multiple body-types.
801 If PURIFY is defined, or PERL_ARENA_SIZE=0, arenas are not used, and
802 the (new|del)_X*V macros are mapped directly to malloc/free.
804 For each sv-type, struct body_details bodies_by_type[] carries
805 parameters which control these aspects of SV handling:
807 Arena_size determines whether arenas are used for this body type, and if
808 so, how big they are. PURIFY or PERL_ARENA_SIZE=0 set this field to
809 zero, forcing individual mallocs and frees.
811 Body_size determines how big a body is, and therefore how many fit into
812 each arena. Offset carries the body-pointer adjustment needed for
813 "ghost fields", and is used in *_allocated macros.
815 But its main purpose is to parameterize info needed in
816 Perl_sv_upgrade(). The info here dramatically simplifies the function
817 vs the implementation in 5.8.8, making it table-driven. All fields
818 are used for this, except for arena_size.
820 For the sv-types that have no bodies, arenas are not used, so those
821 PL_body_roots[sv_type] are unused, and can be overloaded. In
822 something of a special case, SVt_NULL is borrowed for HE arenas;
823 PL_body_roots[HE_SVSLOT=SVt_NULL] is filled by S_more_he, but the
824 bodies_by_type[SVt_NULL] slot is not used, as the table is not
829 struct body_details {
830 U8 body_size; /* Size to allocate */
831 U8 copy; /* Size of structure to copy (may be shorter) */
833 unsigned int type : 4; /* We have space for a sanity check. */
834 unsigned int cant_upgrade : 1; /* Cannot upgrade this type */
835 unsigned int zero_nv : 1; /* zero the NV when upgrading from this */
836 unsigned int arena : 1; /* Allocated from an arena */
837 size_t arena_size; /* Size of arena to allocate */
845 /* With -DPURFIY we allocate everything directly, and don't use arenas.
846 This seems a rather elegant way to simplify some of the code below. */
847 #define HASARENA FALSE
849 #define HASARENA TRUE
851 #define NOARENA FALSE
853 /* Size the arenas to exactly fit a given number of bodies. A count
854 of 0 fits the max number bodies into a PERL_ARENA_SIZE.block,
855 simplifying the default. If count > 0, the arena is sized to fit
856 only that many bodies, allowing arenas to be used for large, rare
857 bodies (XPVFM, XPVIO) without undue waste. The arena size is
858 limited by PERL_ARENA_SIZE, so we can safely oversize the
861 #define FIT_ARENA0(body_size) \
862 ((size_t)(PERL_ARENA_SIZE / body_size) * body_size)
863 #define FIT_ARENAn(count,body_size) \
864 ( count * body_size <= PERL_ARENA_SIZE) \
865 ? count * body_size \
866 : FIT_ARENA0 (body_size)
867 #define FIT_ARENA(count,body_size) \
869 ? FIT_ARENAn (count, body_size) \
870 : FIT_ARENA0 (body_size)
872 /* Calculate the length to copy. Specifically work out the length less any
873 final padding the compiler needed to add. See the comment in sv_upgrade
874 for why copying the padding proved to be a bug. */
876 #define copy_length(type, last_member) \
877 STRUCT_OFFSET(type, last_member) \
878 + sizeof (((type*)SvANY((const SV *)0))->last_member)
880 static const struct body_details bodies_by_type[] = {
881 /* HEs use this offset for their arena. */
882 { 0, 0, 0, SVt_NULL, FALSE, NONV, NOARENA, 0 },
884 /* The bind placeholder pretends to be an RV for now.
885 Also it's marked as "can't upgrade" to stop anyone using it before it's
887 { 0, 0, 0, SVt_BIND, TRUE, NONV, NOARENA, 0 },
889 /* IVs are in the head, so the allocation size is 0. */
891 sizeof(IV), /* This is used to copy out the IV body. */
892 STRUCT_OFFSET(XPVIV, xiv_iv), SVt_IV, FALSE, NONV,
893 NOARENA /* IVS don't need an arena */, 0
896 { sizeof(NV), sizeof(NV),
897 STRUCT_OFFSET(XPVNV, xnv_u),
898 SVt_NV, FALSE, HADNV, HASARENA, FIT_ARENA(0, sizeof(NV)) },
900 { sizeof(XPV) - STRUCT_OFFSET(XPV, xpv_cur),
901 copy_length(XPV, xpv_len) - STRUCT_OFFSET(XPV, xpv_cur),
902 + STRUCT_OFFSET(XPV, xpv_cur),
903 SVt_PV, FALSE, NONV, HASARENA,
904 FIT_ARENA(0, sizeof(XPV) - STRUCT_OFFSET(XPV, xpv_cur)) },
906 { sizeof(XPVIV) - STRUCT_OFFSET(XPV, xpv_cur),
907 copy_length(XPVIV, xiv_u) - STRUCT_OFFSET(XPV, xpv_cur),
908 + STRUCT_OFFSET(XPV, xpv_cur),
909 SVt_PVIV, FALSE, NONV, HASARENA,
910 FIT_ARENA(0, sizeof(XPVIV) - STRUCT_OFFSET(XPV, xpv_cur)) },
912 { sizeof(XPVNV) - STRUCT_OFFSET(XPV, xpv_cur),
913 copy_length(XPVNV, xnv_u) - STRUCT_OFFSET(XPV, xpv_cur),
914 + STRUCT_OFFSET(XPV, xpv_cur),
915 SVt_PVNV, FALSE, HADNV, HASARENA,
916 FIT_ARENA(0, sizeof(XPVNV) - STRUCT_OFFSET(XPV, xpv_cur)) },
918 { sizeof(XPVMG), copy_length(XPVMG, xnv_u), 0, SVt_PVMG, FALSE, HADNV,
919 HASARENA, FIT_ARENA(0, sizeof(XPVMG)) },
924 SVt_REGEXP, FALSE, NONV, HASARENA,
925 FIT_ARENA(0, sizeof(regexp))
928 { sizeof(XPVGV), sizeof(XPVGV), 0, SVt_PVGV, TRUE, HADNV,
929 HASARENA, FIT_ARENA(0, sizeof(XPVGV)) },
931 { sizeof(XPVLV), sizeof(XPVLV), 0, SVt_PVLV, TRUE, HADNV,
932 HASARENA, FIT_ARENA(0, sizeof(XPVLV)) },
935 copy_length(XPVAV, xav_alloc),
937 SVt_PVAV, TRUE, NONV, HASARENA,
938 FIT_ARENA(0, sizeof(XPVAV)) },
941 copy_length(XPVHV, xhv_max),
943 SVt_PVHV, TRUE, NONV, HASARENA,
944 FIT_ARENA(0, sizeof(XPVHV)) },
949 SVt_PVCV, TRUE, NONV, HASARENA,
950 FIT_ARENA(0, sizeof(XPVCV)) },
955 SVt_PVFM, TRUE, NONV, NOARENA,
956 FIT_ARENA(20, sizeof(XPVFM)) },
961 SVt_PVIO, TRUE, NONV, HASARENA,
962 FIT_ARENA(24, sizeof(XPVIO)) },
965 #define new_body_allocated(sv_type) \
966 (void *)((char *)S_new_body(aTHX_ sv_type) \
967 - bodies_by_type[sv_type].offset)
969 /* return a thing to the free list */
971 #define del_body(thing, root) \
973 void ** const thing_copy = (void **)thing; \
974 *thing_copy = *root; \
975 *root = (void*)thing_copy; \
980 #define new_XNV() safemalloc(sizeof(XPVNV))
981 #define new_XPVNV() safemalloc(sizeof(XPVNV))
982 #define new_XPVMG() safemalloc(sizeof(XPVMG))
984 #define del_XPVGV(p) safefree(p)
988 #define new_XNV() new_body_allocated(SVt_NV)
989 #define new_XPVNV() new_body_allocated(SVt_PVNV)
990 #define new_XPVMG() new_body_allocated(SVt_PVMG)
992 #define del_XPVGV(p) del_body(p + bodies_by_type[SVt_PVGV].offset, \
993 &PL_body_roots[SVt_PVGV])
997 /* no arena for you! */
999 #define new_NOARENA(details) \
1000 safemalloc((details)->body_size + (details)->offset)
1001 #define new_NOARENAZ(details) \
1002 safecalloc((details)->body_size + (details)->offset, 1)
1005 Perl_more_bodies (pTHX_ const svtype sv_type, const size_t body_size,
1006 const size_t arena_size)
1009 void ** const root = &PL_body_roots[sv_type];
1010 struct arena_desc *adesc;
1011 struct arena_set *aroot = (struct arena_set *) PL_body_arenas;
1015 const size_t good_arena_size = Perl_malloc_good_size(arena_size);
1016 #if defined(DEBUGGING) && !defined(PERL_GLOBAL_STRUCT_PRIVATE)
1017 static bool done_sanity_check;
1019 /* PERL_GLOBAL_STRUCT_PRIVATE cannot coexist with global
1020 * variables like done_sanity_check. */
1021 if (!done_sanity_check) {
1022 unsigned int i = SVt_LAST;
1024 done_sanity_check = TRUE;
1027 assert (bodies_by_type[i].type == i);
1033 /* may need new arena-set to hold new arena */
1034 if (!aroot || aroot->curr >= aroot->set_size) {
1035 struct arena_set *newroot;
1036 Newxz(newroot, 1, struct arena_set);
1037 newroot->set_size = ARENAS_PER_SET;
1038 newroot->next = aroot;
1040 PL_body_arenas = (void *) newroot;
1041 DEBUG_m(PerlIO_printf(Perl_debug_log, "new arenaset %p\n", (void*)aroot));
1044 /* ok, now have arena-set with at least 1 empty/available arena-desc */
1045 curr = aroot->curr++;
1046 adesc = &(aroot->set[curr]);
1047 assert(!adesc->arena);
1049 Newx(adesc->arena, good_arena_size, char);
1050 adesc->size = good_arena_size;
1051 adesc->utype = sv_type;
1052 DEBUG_m(PerlIO_printf(Perl_debug_log, "arena %d added: %p size %"UVuf"\n",
1053 curr, (void*)adesc->arena, (UV)good_arena_size));
1055 start = (char *) adesc->arena;
1057 /* Get the address of the byte after the end of the last body we can fit.
1058 Remember, this is integer division: */
1059 end = start + good_arena_size / body_size * body_size;
1061 /* computed count doesn't reflect the 1st slot reservation */
1062 #if defined(MYMALLOC) || defined(HAS_MALLOC_GOOD_SIZE)
1063 DEBUG_m(PerlIO_printf(Perl_debug_log,
1064 "arena %p end %p arena-size %d (from %d) type %d "
1066 (void*)start, (void*)end, (int)good_arena_size,
1067 (int)arena_size, sv_type, (int)body_size,
1068 (int)good_arena_size / (int)body_size));
1070 DEBUG_m(PerlIO_printf(Perl_debug_log,
1071 "arena %p end %p arena-size %d type %d size %d ct %d\n",
1072 (void*)start, (void*)end,
1073 (int)arena_size, sv_type, (int)body_size,
1074 (int)good_arena_size / (int)body_size));
1076 *root = (void *)start;
1079 /* Where the next body would start: */
1080 char * const next = start + body_size;
1083 /* This is the last body: */
1084 assert(next == end);
1086 *(void **)start = 0;
1090 *(void**) start = (void *)next;
1095 /* grab a new thing from the free list, allocating more if necessary.
1096 The inline version is used for speed in hot routines, and the
1097 function using it serves the rest (unless PURIFY).
1099 #define new_body_inline(xpv, sv_type) \
1101 void ** const r3wt = &PL_body_roots[sv_type]; \
1102 xpv = (PTR_TBL_ENT_t*) (*((void **)(r3wt)) \
1103 ? *((void **)(r3wt)) : Perl_more_bodies(aTHX_ sv_type, \
1104 bodies_by_type[sv_type].body_size,\
1105 bodies_by_type[sv_type].arena_size)); \
1106 *(r3wt) = *(void**)(xpv); \
1112 S_new_body(pTHX_ const svtype sv_type)
1116 new_body_inline(xpv, sv_type);
1122 static const struct body_details fake_rv =
1123 { 0, 0, 0, SVt_IV, FALSE, NONV, NOARENA, 0 };
1126 =for apidoc sv_upgrade
1128 Upgrade an SV to a more complex form. Generally adds a new body type to the
1129 SV, then copies across as much information as possible from the old body.
1130 It croaks if the SV is already in a more complex form than requested. You
1131 generally want to use the C<SvUPGRADE> macro wrapper, which checks the type
1132 before calling C<sv_upgrade>, and hence does not croak. See also
1139 Perl_sv_upgrade(pTHX_ register SV *const sv, svtype new_type)
1144 const svtype old_type = SvTYPE(sv);
1145 const struct body_details *new_type_details;
1146 const struct body_details *old_type_details
1147 = bodies_by_type + old_type;
1148 SV *referant = NULL;
1150 PERL_ARGS_ASSERT_SV_UPGRADE;
1152 if (old_type == new_type)
1155 /* This clause was purposefully added ahead of the early return above to
1156 the shared string hackery for (sort {$a <=> $b} keys %hash), with the
1157 inference by Nick I-S that it would fix other troublesome cases. See
1158 changes 7162, 7163 (f130fd4589cf5fbb24149cd4db4137c8326f49c1 and parent)
1160 Given that shared hash key scalars are no longer PVIV, but PV, there is
1161 no longer need to unshare so as to free up the IVX slot for its proper
1162 purpose. So it's safe to move the early return earlier. */
1164 if (new_type != SVt_PV && SvIsCOW(sv)) {
1165 sv_force_normal_flags(sv, 0);
1168 old_body = SvANY(sv);
1170 /* Copying structures onto other structures that have been neatly zeroed
1171 has a subtle gotcha. Consider XPVMG
1173 +------+------+------+------+------+-------+-------+
1174 | NV | CUR | LEN | IV | MAGIC | STASH |
1175 +------+------+------+------+------+-------+-------+
1176 0 4 8 12 16 20 24 28
1178 where NVs are aligned to 8 bytes, so that sizeof that structure is
1179 actually 32 bytes long, with 4 bytes of padding at the end:
1181 +------+------+------+------+------+-------+-------+------+
1182 | NV | CUR | LEN | IV | MAGIC | STASH | ??? |
1183 +------+------+------+------+------+-------+-------+------+
1184 0 4 8 12 16 20 24 28 32
1186 so what happens if you allocate memory for this structure:
1188 +------+------+------+------+------+-------+-------+------+------+...
1189 | NV | CUR | LEN | IV | MAGIC | STASH | GP | NAME |
1190 +------+------+------+------+------+-------+-------+------+------+...
1191 0 4 8 12 16 20 24 28 32 36
1193 zero it, then copy sizeof(XPVMG) bytes on top of it? Not quite what you
1194 expect, because you copy the area marked ??? onto GP. Now, ??? may have
1195 started out as zero once, but it's quite possible that it isn't. So now,
1196 rather than a nicely zeroed GP, you have it pointing somewhere random.
1199 (In fact, GP ends up pointing at a previous GP structure, because the
1200 principle cause of the padding in XPVMG getting garbage is a copy of
1201 sizeof(XPVMG) bytes from a XPVGV structure in sv_unglob. Right now
1202 this happens to be moot because XPVGV has been re-ordered, with GP
1203 no longer after STASH)
1205 So we are careful and work out the size of used parts of all the
1213 referant = SvRV(sv);
1214 old_type_details = &fake_rv;
1215 if (new_type == SVt_NV)
1216 new_type = SVt_PVNV;
1218 if (new_type < SVt_PVIV) {
1219 new_type = (new_type == SVt_NV)
1220 ? SVt_PVNV : SVt_PVIV;
1225 if (new_type < SVt_PVNV) {
1226 new_type = SVt_PVNV;
1230 assert(new_type > SVt_PV);
1231 assert(SVt_IV < SVt_PV);
1232 assert(SVt_NV < SVt_PV);
1239 /* Because the XPVMG of PL_mess_sv isn't allocated from the arena,
1240 there's no way that it can be safely upgraded, because perl.c
1241 expects to Safefree(SvANY(PL_mess_sv)) */
1242 assert(sv != PL_mess_sv);
1243 /* This flag bit is used to mean other things in other scalar types.
1244 Given that it only has meaning inside the pad, it shouldn't be set
1245 on anything that can get upgraded. */
1246 assert(!SvPAD_TYPED(sv));
1249 if (old_type_details->cant_upgrade)
1250 Perl_croak(aTHX_ "Can't upgrade %s (%" UVuf ") to %" UVuf,
1251 sv_reftype(sv, 0), (UV) old_type, (UV) new_type);
1254 if (old_type > new_type)
1255 Perl_croak(aTHX_ "sv_upgrade from type %d down to type %d",
1256 (int)old_type, (int)new_type);
1258 new_type_details = bodies_by_type + new_type;
1260 SvFLAGS(sv) &= ~SVTYPEMASK;
1261 SvFLAGS(sv) |= new_type;
1263 /* This can't happen, as SVt_NULL is <= all values of new_type, so one of
1264 the return statements above will have triggered. */
1265 assert (new_type != SVt_NULL);
1268 assert(old_type == SVt_NULL);
1269 SvANY(sv) = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
1273 assert(old_type == SVt_NULL);
1274 SvANY(sv) = new_XNV();
1279 assert(new_type_details->body_size);
1282 assert(new_type_details->arena);
1283 assert(new_type_details->arena_size);
1284 /* This points to the start of the allocated area. */
1285 new_body_inline(new_body, new_type);
1286 Zero(new_body, new_type_details->body_size, char);
1287 new_body = ((char *)new_body) - new_type_details->offset;
1289 /* We always allocated the full length item with PURIFY. To do this
1290 we fake things so that arena is false for all 16 types.. */
1291 new_body = new_NOARENAZ(new_type_details);
1293 SvANY(sv) = new_body;
1294 if (new_type == SVt_PVAV) {
1298 if (old_type_details->body_size) {
1301 /* It will have been zeroed when the new body was allocated.
1302 Lets not write to it, in case it confuses a write-back
1308 #ifndef NODEFAULT_SHAREKEYS
1309 HvSHAREKEYS_on(sv); /* key-sharing on by default */
1311 HvMAX(sv) = 7; /* (start with 8 buckets) */
1314 /* SVt_NULL isn't the only thing upgraded to AV or HV.
1315 The target created by newSVrv also is, and it can have magic.
1316 However, it never has SvPVX set.
1318 if (old_type == SVt_IV) {
1320 } else if (old_type >= SVt_PV) {
1321 assert(SvPVX_const(sv) == 0);
1324 if (old_type >= SVt_PVMG) {
1325 SvMAGIC_set(sv, ((XPVMG*)old_body)->xmg_u.xmg_magic);
1326 SvSTASH_set(sv, ((XPVMG*)old_body)->xmg_stash);
1328 sv->sv_u.svu_array = NULL; /* or svu_hash */
1334 /* This ensures that SvTHINKFIRST(sv) is true, and hence that
1335 sv_force_normal_flags(sv) is called. */
1338 /* XXX Is this still needed? Was it ever needed? Surely as there is
1339 no route from NV to PVIV, NOK can never be true */
1340 assert(!SvNOKp(sv));
1351 assert(new_type_details->body_size);
1352 /* We always allocated the full length item with PURIFY. To do this
1353 we fake things so that arena is false for all 16 types.. */
1354 if(new_type_details->arena) {
1355 /* This points to the start of the allocated area. */
1356 new_body_inline(new_body, new_type);
1357 Zero(new_body, new_type_details->body_size, char);
1358 new_body = ((char *)new_body) - new_type_details->offset;
1360 new_body = new_NOARENAZ(new_type_details);
1362 SvANY(sv) = new_body;
1364 if (old_type_details->copy) {
1365 /* There is now the potential for an upgrade from something without
1366 an offset (PVNV or PVMG) to something with one (PVCV, PVFM) */
1367 int offset = old_type_details->offset;
1368 int length = old_type_details->copy;
1370 if (new_type_details->offset > old_type_details->offset) {
1371 const int difference
1372 = new_type_details->offset - old_type_details->offset;
1373 offset += difference;
1374 length -= difference;
1376 assert (length >= 0);
1378 Copy((char *)old_body + offset, (char *)new_body + offset, length,
1382 #ifndef NV_ZERO_IS_ALLBITS_ZERO
1383 /* If NV 0.0 is stores as all bits 0 then Zero() already creates a
1384 * correct 0.0 for us. Otherwise, if the old body didn't have an
1385 * NV slot, but the new one does, then we need to initialise the
1386 * freshly created NV slot with whatever the correct bit pattern is
1388 if (old_type_details->zero_nv && !new_type_details->zero_nv
1389 && !isGV_with_GP(sv))
1393 if (new_type == SVt_PVIO) {
1394 IO * const io = MUTABLE_IO(sv);
1395 GV *iogv = gv_fetchpvs("IO::File::", GV_ADD, SVt_PVHV);
1398 /* Clear the stashcache because a new IO could overrule a package
1400 hv_clear(PL_stashcache);
1402 SvSTASH_set(io, MUTABLE_HV(SvREFCNT_inc(GvHV(iogv))));
1403 IoPAGE_LEN(sv) = 60;
1405 if (old_type < SVt_PV) {
1406 /* referant will be NULL unless the old type was SVt_IV emulating
1408 sv->sv_u.svu_rv = referant;
1412 Perl_croak(aTHX_ "panic: sv_upgrade to unknown type %lu",
1413 (unsigned long)new_type);
1416 if (old_type > SVt_IV) {
1420 /* Note that there is an assumption that all bodies of types that
1421 can be upgraded came from arenas. Only the more complex non-
1422 upgradable types are allowed to be directly malloc()ed. */
1423 assert(old_type_details->arena);
1424 del_body((void*)((char*)old_body + old_type_details->offset),
1425 &PL_body_roots[old_type]);
1431 =for apidoc sv_backoff
1433 Remove any string offset. You should normally use the C<SvOOK_off> macro
1440 Perl_sv_backoff(pTHX_ register SV *const sv)
1443 const char * const s = SvPVX_const(sv);
1445 PERL_ARGS_ASSERT_SV_BACKOFF;
1446 PERL_UNUSED_CONTEXT;
1449 assert(SvTYPE(sv) != SVt_PVHV);
1450 assert(SvTYPE(sv) != SVt_PVAV);
1452 SvOOK_offset(sv, delta);
1454 SvLEN_set(sv, SvLEN(sv) + delta);
1455 SvPV_set(sv, SvPVX(sv) - delta);
1456 Move(s, SvPVX(sv), SvCUR(sv)+1, char);
1457 SvFLAGS(sv) &= ~SVf_OOK;
1464 Expands the character buffer in the SV. If necessary, uses C<sv_unref> and
1465 upgrades the SV to C<SVt_PV>. Returns a pointer to the character buffer.
1466 Use the C<SvGROW> wrapper instead.
1472 Perl_sv_grow(pTHX_ register SV *const sv, register STRLEN newlen)
1476 PERL_ARGS_ASSERT_SV_GROW;
1478 if (PL_madskills && newlen >= 0x100000) {
1479 PerlIO_printf(Perl_debug_log,
1480 "Allocation too large: %"UVxf"\n", (UV)newlen);
1482 #ifdef HAS_64K_LIMIT
1483 if (newlen >= 0x10000) {
1484 PerlIO_printf(Perl_debug_log,
1485 "Allocation too large: %"UVxf"\n", (UV)newlen);
1488 #endif /* HAS_64K_LIMIT */
1491 if (SvTYPE(sv) < SVt_PV) {
1492 sv_upgrade(sv, SVt_PV);
1493 s = SvPVX_mutable(sv);
1495 else if (SvOOK(sv)) { /* pv is offset? */
1497 s = SvPVX_mutable(sv);
1498 if (newlen > SvLEN(sv))
1499 newlen += 10 * (newlen - SvCUR(sv)); /* avoid copy each time */
1500 #ifdef HAS_64K_LIMIT
1501 if (newlen >= 0x10000)
1506 s = SvPVX_mutable(sv);
1508 if (newlen > SvLEN(sv)) { /* need more room? */
1509 STRLEN minlen = SvCUR(sv);
1510 minlen += (minlen >> PERL_STRLEN_EXPAND_SHIFT) + 10;
1511 if (newlen < minlen)
1513 #ifndef Perl_safesysmalloc_size
1514 newlen = PERL_STRLEN_ROUNDUP(newlen);
1516 if (SvLEN(sv) && s) {
1517 s = (char*)saferealloc(s, newlen);
1520 s = (char*)safemalloc(newlen);
1521 if (SvPVX_const(sv) && SvCUR(sv)) {
1522 Move(SvPVX_const(sv), s, (newlen < SvCUR(sv)) ? newlen : SvCUR(sv), char);
1526 #ifdef Perl_safesysmalloc_size
1527 /* Do this here, do it once, do it right, and then we will never get
1528 called back into sv_grow() unless there really is some growing
1530 SvLEN_set(sv, Perl_safesysmalloc_size(s));
1532 SvLEN_set(sv, newlen);
1539 =for apidoc sv_setiv
1541 Copies an integer into the given SV, upgrading first if necessary.
1542 Does not handle 'set' magic. See also C<sv_setiv_mg>.
1548 Perl_sv_setiv(pTHX_ register SV *const sv, const IV i)
1552 PERL_ARGS_ASSERT_SV_SETIV;
1554 SV_CHECK_THINKFIRST_COW_DROP(sv);
1555 switch (SvTYPE(sv)) {
1558 sv_upgrade(sv, SVt_IV);
1561 sv_upgrade(sv, SVt_PVIV);
1565 if (!isGV_with_GP(sv))
1572 /* diag_listed_as: Can't coerce %s to %s in %s */
1573 Perl_croak(aTHX_ "Can't coerce %s to integer in %s", sv_reftype(sv,0),
1577 (void)SvIOK_only(sv); /* validate number */
1583 =for apidoc sv_setiv_mg
1585 Like C<sv_setiv>, but also handles 'set' magic.
1591 Perl_sv_setiv_mg(pTHX_ register SV *const sv, const IV i)
1593 PERL_ARGS_ASSERT_SV_SETIV_MG;
1600 =for apidoc sv_setuv
1602 Copies an unsigned integer into the given SV, upgrading first if necessary.
1603 Does not handle 'set' magic. See also C<sv_setuv_mg>.
1609 Perl_sv_setuv(pTHX_ register SV *const sv, const UV u)
1611 PERL_ARGS_ASSERT_SV_SETUV;
1613 /* With these two if statements:
1614 u=1.49 s=0.52 cu=72.49 cs=10.64 scripts=270 tests=20865
1617 u=1.35 s=0.47 cu=73.45 cs=11.43 scripts=270 tests=20865
1619 If you wish to remove them, please benchmark to see what the effect is
1621 if (u <= (UV)IV_MAX) {
1622 sv_setiv(sv, (IV)u);
1631 =for apidoc sv_setuv_mg
1633 Like C<sv_setuv>, but also handles 'set' magic.
1639 Perl_sv_setuv_mg(pTHX_ register SV *const sv, const UV u)
1641 PERL_ARGS_ASSERT_SV_SETUV_MG;
1648 =for apidoc sv_setnv
1650 Copies a double into the given SV, upgrading first if necessary.
1651 Does not handle 'set' magic. See also C<sv_setnv_mg>.
1657 Perl_sv_setnv(pTHX_ register SV *const sv, const NV num)
1661 PERL_ARGS_ASSERT_SV_SETNV;
1663 SV_CHECK_THINKFIRST_COW_DROP(sv);
1664 switch (SvTYPE(sv)) {
1667 sv_upgrade(sv, SVt_NV);
1671 sv_upgrade(sv, SVt_PVNV);
1675 if (!isGV_with_GP(sv))
1682 /* diag_listed_as: Can't coerce %s to %s in %s */
1683 Perl_croak(aTHX_ "Can't coerce %s to number in %s", sv_reftype(sv,0),
1688 (void)SvNOK_only(sv); /* validate number */
1693 =for apidoc sv_setnv_mg
1695 Like C<sv_setnv>, but also handles 'set' magic.
1701 Perl_sv_setnv_mg(pTHX_ register SV *const sv, const NV num)
1703 PERL_ARGS_ASSERT_SV_SETNV_MG;
1709 /* Print an "isn't numeric" warning, using a cleaned-up,
1710 * printable version of the offending string
1714 S_not_a_number(pTHX_ SV *const sv)
1721 PERL_ARGS_ASSERT_NOT_A_NUMBER;
1724 dsv = newSVpvs_flags("", SVs_TEMP);
1725 pv = sv_uni_display(dsv, sv, 10, UNI_DISPLAY_ISPRINT);
1728 const char * const limit = tmpbuf + sizeof(tmpbuf) - 8;
1729 /* each *s can expand to 4 chars + "...\0",
1730 i.e. need room for 8 chars */
1732 const char *s = SvPVX_const(sv);
1733 const char * const end = s + SvCUR(sv);
1734 for ( ; s < end && d < limit; s++ ) {
1736 if (ch & 128 && !isPRINT_LC(ch)) {
1745 else if (ch == '\r') {
1749 else if (ch == '\f') {
1753 else if (ch == '\\') {
1757 else if (ch == '\0') {
1761 else if (isPRINT_LC(ch))
1778 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
1779 /* diag_listed_as: Argument "%s" isn't numeric%s */
1780 "Argument \"%s\" isn't numeric in %s", pv,
1783 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
1784 /* diag_listed_as: Argument "%s" isn't numeric%s */
1785 "Argument \"%s\" isn't numeric", pv);
1789 =for apidoc looks_like_number
1791 Test if the content of an SV looks like a number (or is a number).
1792 C<Inf> and C<Infinity> are treated as numbers (so will not issue a
1793 non-numeric warning), even if your atof() doesn't grok them. Get-magic is
1800 Perl_looks_like_number(pTHX_ SV *const sv)
1802 register const char *sbegin;
1805 PERL_ARGS_ASSERT_LOOKS_LIKE_NUMBER;
1807 if (SvPOK(sv) || SvPOKp(sv)) {
1808 sbegin = SvPV_nomg_const(sv, len);
1811 return SvFLAGS(sv) & (SVf_NOK|SVp_NOK|SVf_IOK|SVp_IOK);
1812 return grok_number(sbegin, len, NULL);
1816 S_glob_2number(pTHX_ GV * const gv)
1818 SV *const buffer = sv_newmortal();
1820 PERL_ARGS_ASSERT_GLOB_2NUMBER;
1822 gv_efullname3(buffer, gv, "*");
1824 /* We know that all GVs stringify to something that is not-a-number,
1825 so no need to test that. */
1826 if (ckWARN(WARN_NUMERIC))
1827 not_a_number(buffer);
1828 /* We just want something true to return, so that S_sv_2iuv_common
1829 can tail call us and return true. */
1833 /* Actually, ISO C leaves conversion of UV to IV undefined, but
1834 until proven guilty, assume that things are not that bad... */
1839 As 64 bit platforms often have an NV that doesn't preserve all bits of
1840 an IV (an assumption perl has been based on to date) it becomes necessary
1841 to remove the assumption that the NV always carries enough precision to
1842 recreate the IV whenever needed, and that the NV is the canonical form.
1843 Instead, IV/UV and NV need to be given equal rights. So as to not lose
1844 precision as a side effect of conversion (which would lead to insanity
1845 and the dragon(s) in t/op/numconvert.t getting very angry) the intent is
1846 1) to distinguish between IV/UV/NV slots that have cached a valid
1847 conversion where precision was lost and IV/UV/NV slots that have a
1848 valid conversion which has lost no precision
1849 2) to ensure that if a numeric conversion to one form is requested that
1850 would lose precision, the precise conversion (or differently
1851 imprecise conversion) is also performed and cached, to prevent
1852 requests for different numeric formats on the same SV causing
1853 lossy conversion chains. (lossless conversion chains are perfectly
1858 SvIOKp is true if the IV slot contains a valid value
1859 SvIOK is true only if the IV value is accurate (UV if SvIOK_UV true)
1860 SvNOKp is true if the NV slot contains a valid value
1861 SvNOK is true only if the NV value is accurate
1864 while converting from PV to NV, check to see if converting that NV to an
1865 IV(or UV) would lose accuracy over a direct conversion from PV to
1866 IV(or UV). If it would, cache both conversions, return NV, but mark
1867 SV as IOK NOKp (ie not NOK).
1869 While converting from PV to IV, check to see if converting that IV to an
1870 NV would lose accuracy over a direct conversion from PV to NV. If it
1871 would, cache both conversions, flag similarly.
1873 Before, the SV value "3.2" could become NV=3.2 IV=3 NOK, IOK quite
1874 correctly because if IV & NV were set NV *always* overruled.
1875 Now, "3.2" will become NV=3.2 IV=3 NOK, IOKp, because the flag's meaning
1876 changes - now IV and NV together means that the two are interchangeable:
1877 SvIVX == (IV) SvNVX && SvNVX == (NV) SvIVX;
1879 The benefit of this is that operations such as pp_add know that if
1880 SvIOK is true for both left and right operands, then integer addition
1881 can be used instead of floating point (for cases where the result won't
1882 overflow). Before, floating point was always used, which could lead to
1883 loss of precision compared with integer addition.
1885 * making IV and NV equal status should make maths accurate on 64 bit
1887 * may speed up maths somewhat if pp_add and friends start to use
1888 integers when possible instead of fp. (Hopefully the overhead in
1889 looking for SvIOK and checking for overflow will not outweigh the
1890 fp to integer speedup)
1891 * will slow down integer operations (callers of SvIV) on "inaccurate"
1892 values, as the change from SvIOK to SvIOKp will cause a call into
1893 sv_2iv each time rather than a macro access direct to the IV slot
1894 * should speed up number->string conversion on integers as IV is
1895 favoured when IV and NV are equally accurate
1897 ####################################################################
1898 You had better be using SvIOK_notUV if you want an IV for arithmetic:
1899 SvIOK is true if (IV or UV), so you might be getting (IV)SvUV.
1900 On the other hand, SvUOK is true iff UV.
1901 ####################################################################
1903 Your mileage will vary depending your CPU's relative fp to integer
1907 #ifndef NV_PRESERVES_UV
1908 # define IS_NUMBER_UNDERFLOW_IV 1
1909 # define IS_NUMBER_UNDERFLOW_UV 2
1910 # define IS_NUMBER_IV_AND_UV 2
1911 # define IS_NUMBER_OVERFLOW_IV 4
1912 # define IS_NUMBER_OVERFLOW_UV 5
1914 /* sv_2iuv_non_preserve(): private routine for use by sv_2iv() and sv_2uv() */
1916 /* For sv_2nv these three cases are "SvNOK and don't bother casting" */
1918 S_sv_2iuv_non_preserve(pTHX_ register SV *const sv
1926 PERL_ARGS_ASSERT_SV_2IUV_NON_PRESERVE;
1928 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));
1929 if (SvNVX(sv) < (NV)IV_MIN) {
1930 (void)SvIOKp_on(sv);
1932 SvIV_set(sv, IV_MIN);
1933 return IS_NUMBER_UNDERFLOW_IV;
1935 if (SvNVX(sv) > (NV)UV_MAX) {
1936 (void)SvIOKp_on(sv);
1939 SvUV_set(sv, UV_MAX);
1940 return IS_NUMBER_OVERFLOW_UV;
1942 (void)SvIOKp_on(sv);
1944 /* Can't use strtol etc to convert this string. (See truth table in
1946 if (SvNVX(sv) <= (UV)IV_MAX) {
1947 SvIV_set(sv, I_V(SvNVX(sv)));
1948 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
1949 SvIOK_on(sv); /* Integer is precise. NOK, IOK */
1951 /* Integer is imprecise. NOK, IOKp */
1953 return SvNVX(sv) < 0 ? IS_NUMBER_UNDERFLOW_UV : IS_NUMBER_IV_AND_UV;
1956 SvUV_set(sv, U_V(SvNVX(sv)));
1957 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
1958 if (SvUVX(sv) == UV_MAX) {
1959 /* As we know that NVs don't preserve UVs, UV_MAX cannot
1960 possibly be preserved by NV. Hence, it must be overflow.
1962 return IS_NUMBER_OVERFLOW_UV;
1964 SvIOK_on(sv); /* Integer is precise. NOK, UOK */
1966 /* Integer is imprecise. NOK, IOKp */
1968 return IS_NUMBER_OVERFLOW_IV;
1970 #endif /* !NV_PRESERVES_UV*/
1973 S_sv_2iuv_common(pTHX_ SV *const sv)
1977 PERL_ARGS_ASSERT_SV_2IUV_COMMON;
1980 /* erm. not sure. *should* never get NOKp (without NOK) from sv_2nv
1981 * without also getting a cached IV/UV from it at the same time
1982 * (ie PV->NV conversion should detect loss of accuracy and cache
1983 * IV or UV at same time to avoid this. */
1984 /* IV-over-UV optimisation - choose to cache IV if possible */
1986 if (SvTYPE(sv) == SVt_NV)
1987 sv_upgrade(sv, SVt_PVNV);
1989 (void)SvIOKp_on(sv); /* Must do this first, to clear any SvOOK */
1990 /* < not <= as for NV doesn't preserve UV, ((NV)IV_MAX+1) will almost
1991 certainly cast into the IV range at IV_MAX, whereas the correct
1992 answer is the UV IV_MAX +1. Hence < ensures that dodgy boundary
1994 #if defined(NAN_COMPARE_BROKEN) && defined(Perl_isnan)
1995 if (Perl_isnan(SvNVX(sv))) {
2001 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2002 SvIV_set(sv, I_V(SvNVX(sv)));
2003 if (SvNVX(sv) == (NV) SvIVX(sv)
2004 #ifndef NV_PRESERVES_UV
2005 && (((UV)1 << NV_PRESERVES_UV_BITS) >
2006 (UV)(SvIVX(sv) > 0 ? SvIVX(sv) : -SvIVX(sv)))
2007 /* Don't flag it as "accurately an integer" if the number
2008 came from a (by definition imprecise) NV operation, and
2009 we're outside the range of NV integer precision */
2013 SvIOK_on(sv); /* Can this go wrong with rounding? NWC */
2015 /* scalar has trailing garbage, eg "42a" */
2017 DEBUG_c(PerlIO_printf(Perl_debug_log,
2018 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (precise)\n",
2024 /* IV not precise. No need to convert from PV, as NV
2025 conversion would already have cached IV if it detected
2026 that PV->IV would be better than PV->NV->IV
2027 flags already correct - don't set public IOK. */
2028 DEBUG_c(PerlIO_printf(Perl_debug_log,
2029 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (imprecise)\n",
2034 /* Can the above go wrong if SvIVX == IV_MIN and SvNVX < IV_MIN,
2035 but the cast (NV)IV_MIN rounds to a the value less (more
2036 negative) than IV_MIN which happens to be equal to SvNVX ??
2037 Analogous to 0xFFFFFFFFFFFFFFFF rounding up to NV (2**64) and
2038 NV rounding back to 0xFFFFFFFFFFFFFFFF, so UVX == UV(NVX) and
2039 (NV)UVX == NVX are both true, but the values differ. :-(
2040 Hopefully for 2s complement IV_MIN is something like
2041 0x8000000000000000 which will be exact. NWC */
2044 SvUV_set(sv, U_V(SvNVX(sv)));
2046 (SvNVX(sv) == (NV) SvUVX(sv))
2047 #ifndef NV_PRESERVES_UV
2048 /* Make sure it's not 0xFFFFFFFFFFFFFFFF */
2049 /*&& (SvUVX(sv) != UV_MAX) irrelevant with code below */
2050 && (((UV)1 << NV_PRESERVES_UV_BITS) > SvUVX(sv))
2051 /* Don't flag it as "accurately an integer" if the number
2052 came from a (by definition imprecise) NV operation, and
2053 we're outside the range of NV integer precision */
2059 DEBUG_c(PerlIO_printf(Perl_debug_log,
2060 "0x%"UVxf" 2iv(%"UVuf" => %"IVdf") (as unsigned)\n",
2066 else if (SvPOKp(sv) && SvLEN(sv)) {
2068 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2069 /* We want to avoid a possible problem when we cache an IV/ a UV which
2070 may be later translated to an NV, and the resulting NV is not
2071 the same as the direct translation of the initial string
2072 (eg 123.456 can shortcut to the IV 123 with atol(), but we must
2073 be careful to ensure that the value with the .456 is around if the
2074 NV value is requested in the future).
2076 This means that if we cache such an IV/a UV, we need to cache the
2077 NV as well. Moreover, we trade speed for space, and do not
2078 cache the NV if we are sure it's not needed.
2081 /* SVt_PVNV is one higher than SVt_PVIV, hence this order */
2082 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2083 == IS_NUMBER_IN_UV) {
2084 /* It's definitely an integer, only upgrade to PVIV */
2085 if (SvTYPE(sv) < SVt_PVIV)
2086 sv_upgrade(sv, SVt_PVIV);
2088 } else if (SvTYPE(sv) < SVt_PVNV)
2089 sv_upgrade(sv, SVt_PVNV);
2091 /* If NVs preserve UVs then we only use the UV value if we know that
2092 we aren't going to call atof() below. If NVs don't preserve UVs
2093 then the value returned may have more precision than atof() will
2094 return, even though value isn't perfectly accurate. */
2095 if ((numtype & (IS_NUMBER_IN_UV
2096 #ifdef NV_PRESERVES_UV
2099 )) == IS_NUMBER_IN_UV) {
2100 /* This won't turn off the public IOK flag if it was set above */
2101 (void)SvIOKp_on(sv);
2103 if (!(numtype & IS_NUMBER_NEG)) {
2105 if (value <= (UV)IV_MAX) {
2106 SvIV_set(sv, (IV)value);
2108 /* it didn't overflow, and it was positive. */
2109 SvUV_set(sv, value);
2113 /* 2s complement assumption */
2114 if (value <= (UV)IV_MIN) {
2115 SvIV_set(sv, -(IV)value);
2117 /* Too negative for an IV. This is a double upgrade, but
2118 I'm assuming it will be rare. */
2119 if (SvTYPE(sv) < SVt_PVNV)
2120 sv_upgrade(sv, SVt_PVNV);
2124 SvNV_set(sv, -(NV)value);
2125 SvIV_set(sv, IV_MIN);
2129 /* For !NV_PRESERVES_UV and IS_NUMBER_IN_UV and IS_NUMBER_NOT_INT we
2130 will be in the previous block to set the IV slot, and the next
2131 block to set the NV slot. So no else here. */
2133 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2134 != IS_NUMBER_IN_UV) {
2135 /* It wasn't an (integer that doesn't overflow the UV). */
2136 SvNV_set(sv, Atof(SvPVX_const(sv)));
2138 if (! numtype && ckWARN(WARN_NUMERIC))
2141 #if defined(USE_LONG_DOUBLE)
2142 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%" PERL_PRIgldbl ")\n",
2143 PTR2UV(sv), SvNVX(sv)));
2145 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"NVgf")\n",
2146 PTR2UV(sv), SvNVX(sv)));
2149 #ifdef NV_PRESERVES_UV
2150 (void)SvIOKp_on(sv);
2152 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2153 SvIV_set(sv, I_V(SvNVX(sv)));
2154 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
2157 NOOP; /* Integer is imprecise. NOK, IOKp */
2159 /* UV will not work better than IV */
2161 if (SvNVX(sv) > (NV)UV_MAX) {
2163 /* Integer is inaccurate. NOK, IOKp, is UV */
2164 SvUV_set(sv, UV_MAX);
2166 SvUV_set(sv, U_V(SvNVX(sv)));
2167 /* 0xFFFFFFFFFFFFFFFF not an issue in here, NVs
2168 NV preservse UV so can do correct comparison. */
2169 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
2172 NOOP; /* Integer is imprecise. NOK, IOKp, is UV */
2177 #else /* NV_PRESERVES_UV */
2178 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2179 == (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT)) {
2180 /* The IV/UV slot will have been set from value returned by
2181 grok_number above. The NV slot has just been set using
2184 assert (SvIOKp(sv));
2186 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2187 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2188 /* Small enough to preserve all bits. */
2189 (void)SvIOKp_on(sv);
2191 SvIV_set(sv, I_V(SvNVX(sv)));
2192 if ((NV)(SvIVX(sv)) == SvNVX(sv))
2194 /* Assumption: first non-preserved integer is < IV_MAX,
2195 this NV is in the preserved range, therefore: */
2196 if (!(U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))
2198 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);
2202 0 0 already failed to read UV.
2203 0 1 already failed to read UV.
2204 1 0 you won't get here in this case. IV/UV
2205 slot set, public IOK, Atof() unneeded.
2206 1 1 already read UV.
2207 so there's no point in sv_2iuv_non_preserve() attempting
2208 to use atol, strtol, strtoul etc. */
2210 sv_2iuv_non_preserve (sv, numtype);
2212 sv_2iuv_non_preserve (sv);
2216 #endif /* NV_PRESERVES_UV */
2217 /* It might be more code efficient to go through the entire logic above
2218 and conditionally set with SvIOKp_on() rather than SvIOK(), but it
2219 gets complex and potentially buggy, so more programmer efficient
2220 to do it this way, by turning off the public flags: */
2222 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK);
2226 if (isGV_with_GP(sv))
2227 return glob_2number(MUTABLE_GV(sv));
2229 if (!SvPADTMP(sv)) {
2230 if (!PL_localizing && ckWARN(WARN_UNINITIALIZED))
2233 if (SvTYPE(sv) < SVt_IV)
2234 /* Typically the caller expects that sv_any is not NULL now. */
2235 sv_upgrade(sv, SVt_IV);
2236 /* Return 0 from the caller. */
2243 =for apidoc sv_2iv_flags
2245 Return the integer value of an SV, doing any necessary string
2246 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2247 Normally used via the C<SvIV(sv)> and C<SvIVx(sv)> macros.
2253 Perl_sv_2iv_flags(pTHX_ register SV *const sv, const I32 flags)
2258 if (SvGMAGICAL(sv) || SvVALID(sv)) {
2259 /* FBMs use the space for SvIVX and SvNVX for other purposes, and use
2260 the same flag bit as SVf_IVisUV, so must not let them cache IVs.
2261 In practice they are extremely unlikely to actually get anywhere
2262 accessible by user Perl code - the only way that I'm aware of is when
2263 a constant subroutine which is used as the second argument to index.
2265 if (flags & SV_GMAGIC)
2270 return I_V(SvNVX(sv));
2272 if (SvPOKp(sv) && SvLEN(sv)) {
2275 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2277 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2278 == IS_NUMBER_IN_UV) {
2279 /* It's definitely an integer */
2280 if (numtype & IS_NUMBER_NEG) {
2281 if (value < (UV)IV_MIN)
2284 if (value < (UV)IV_MAX)
2289 if (ckWARN(WARN_NUMERIC))
2292 return I_V(Atof(SvPVX_const(sv)));
2297 assert(SvTYPE(sv) >= SVt_PVMG);
2298 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
2299 } else if (SvTHINKFIRST(sv)) {
2304 if (flags & SV_SKIP_OVERLOAD)
2306 tmpstr = AMG_CALLunary(sv, numer_amg);
2307 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2308 return SvIV(tmpstr);
2311 return PTR2IV(SvRV(sv));
2314 sv_force_normal_flags(sv, 0);
2316 if (SvREADONLY(sv) && !SvOK(sv)) {
2317 if (ckWARN(WARN_UNINITIALIZED))
2323 if (S_sv_2iuv_common(aTHX_ sv))
2326 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"IVdf")\n",
2327 PTR2UV(sv),SvIVX(sv)));
2328 return SvIsUV(sv) ? (IV)SvUVX(sv) : SvIVX(sv);
2332 =for apidoc sv_2uv_flags
2334 Return the unsigned integer value of an SV, doing any necessary string
2335 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2336 Normally used via the C<SvUV(sv)> and C<SvUVx(sv)> macros.
2342 Perl_sv_2uv_flags(pTHX_ register SV *const sv, const I32 flags)
2347 if (SvGMAGICAL(sv) || SvVALID(sv)) {
2348 /* FBMs use the space for SvIVX and SvNVX for other purposes, and use
2349 the same flag bit as SVf_IVisUV, so must not let them cache IVs. */
2350 if (flags & SV_GMAGIC)
2355 return U_V(SvNVX(sv));
2356 if (SvPOKp(sv) && SvLEN(sv)) {
2359 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2361 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2362 == IS_NUMBER_IN_UV) {
2363 /* It's definitely an integer */
2364 if (!(numtype & IS_NUMBER_NEG))
2368 if (ckWARN(WARN_NUMERIC))
2371 return U_V(Atof(SvPVX_const(sv)));
2376 assert(SvTYPE(sv) >= SVt_PVMG);
2377 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
2378 } else if (SvTHINKFIRST(sv)) {
2383 if (flags & SV_SKIP_OVERLOAD)
2385 tmpstr = AMG_CALLunary(sv, numer_amg);
2386 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2387 return SvUV(tmpstr);
2390 return PTR2UV(SvRV(sv));
2393 sv_force_normal_flags(sv, 0);
2395 if (SvREADONLY(sv) && !SvOK(sv)) {
2396 if (ckWARN(WARN_UNINITIALIZED))
2402 if (S_sv_2iuv_common(aTHX_ sv))
2406 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2uv(%"UVuf")\n",
2407 PTR2UV(sv),SvUVX(sv)));
2408 return SvIsUV(sv) ? SvUVX(sv) : (UV)SvIVX(sv);
2412 =for apidoc sv_2nv_flags
2414 Return the num value of an SV, doing any necessary string or integer
2415 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2416 Normally used via the C<SvNV(sv)> and C<SvNVx(sv)> macros.
2422 Perl_sv_2nv_flags(pTHX_ register SV *const sv, const I32 flags)
2427 if (SvGMAGICAL(sv) || SvVALID(sv)) {
2428 /* FBMs use the space for SvIVX and SvNVX for other purposes, and use
2429 the same flag bit as SVf_IVisUV, so must not let them cache NVs. */
2430 if (flags & SV_GMAGIC)
2434 if ((SvPOKp(sv) && SvLEN(sv)) && !SvIOKp(sv)) {
2435 if (!SvIOKp(sv) && ckWARN(WARN_NUMERIC) &&
2436 !grok_number(SvPVX_const(sv), SvCUR(sv), NULL))
2438 return Atof(SvPVX_const(sv));
2442 return (NV)SvUVX(sv);
2444 return (NV)SvIVX(sv);
2449 assert(SvTYPE(sv) >= SVt_PVMG);
2450 /* This falls through to the report_uninit near the end of the
2452 } else if (SvTHINKFIRST(sv)) {
2457 if (flags & SV_SKIP_OVERLOAD)
2459 tmpstr = AMG_CALLunary(sv, numer_amg);
2460 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2461 return SvNV(tmpstr);
2464 return PTR2NV(SvRV(sv));
2467 sv_force_normal_flags(sv, 0);
2469 if (SvREADONLY(sv) && !SvOK(sv)) {
2470 if (ckWARN(WARN_UNINITIALIZED))
2475 if (SvTYPE(sv) < SVt_NV) {
2476 /* The logic to use SVt_PVNV if necessary is in sv_upgrade. */
2477 sv_upgrade(sv, SVt_NV);
2478 #ifdef USE_LONG_DOUBLE
2480 STORE_NUMERIC_LOCAL_SET_STANDARD();
2481 PerlIO_printf(Perl_debug_log,
2482 "0x%"UVxf" num(%" PERL_PRIgldbl ")\n",
2483 PTR2UV(sv), SvNVX(sv));
2484 RESTORE_NUMERIC_LOCAL();
2488 STORE_NUMERIC_LOCAL_SET_STANDARD();
2489 PerlIO_printf(Perl_debug_log, "0x%"UVxf" num(%"NVgf")\n",
2490 PTR2UV(sv), SvNVX(sv));
2491 RESTORE_NUMERIC_LOCAL();
2495 else if (SvTYPE(sv) < SVt_PVNV)
2496 sv_upgrade(sv, SVt_PVNV);
2501 SvNV_set(sv, SvIsUV(sv) ? (NV)SvUVX(sv) : (NV)SvIVX(sv));
2502 #ifdef NV_PRESERVES_UV
2508 /* Only set the public NV OK flag if this NV preserves the IV */
2509 /* Check it's not 0xFFFFFFFFFFFFFFFF */
2511 SvIsUV(sv) ? ((SvUVX(sv) != UV_MAX)&&(SvUVX(sv) == U_V(SvNVX(sv))))
2512 : (SvIVX(sv) == I_V(SvNVX(sv))))
2518 else if (SvPOKp(sv) && SvLEN(sv)) {
2520 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2521 if (!SvIOKp(sv) && !numtype && ckWARN(WARN_NUMERIC))
2523 #ifdef NV_PRESERVES_UV
2524 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2525 == IS_NUMBER_IN_UV) {
2526 /* It's definitely an integer */
2527 SvNV_set(sv, (numtype & IS_NUMBER_NEG) ? -(NV)value : (NV)value);
2529 SvNV_set(sv, Atof(SvPVX_const(sv)));
2535 SvNV_set(sv, Atof(SvPVX_const(sv)));
2536 /* Only set the public NV OK flag if this NV preserves the value in
2537 the PV at least as well as an IV/UV would.
2538 Not sure how to do this 100% reliably. */
2539 /* if that shift count is out of range then Configure's test is
2540 wonky. We shouldn't be in here with NV_PRESERVES_UV_BITS ==
2542 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2543 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2544 SvNOK_on(sv); /* Definitely small enough to preserve all bits */
2545 } else if (!(numtype & IS_NUMBER_IN_UV)) {
2546 /* Can't use strtol etc to convert this string, so don't try.
2547 sv_2iv and sv_2uv will use the NV to convert, not the PV. */
2550 /* value has been set. It may not be precise. */
2551 if ((numtype & IS_NUMBER_NEG) && (value > (UV)IV_MIN)) {
2552 /* 2s complement assumption for (UV)IV_MIN */
2553 SvNOK_on(sv); /* Integer is too negative. */
2558 if (numtype & IS_NUMBER_NEG) {
2559 SvIV_set(sv, -(IV)value);
2560 } else if (value <= (UV)IV_MAX) {
2561 SvIV_set(sv, (IV)value);
2563 SvUV_set(sv, value);
2567 if (numtype & IS_NUMBER_NOT_INT) {
2568 /* I believe that even if the original PV had decimals,
2569 they are lost beyond the limit of the FP precision.
2570 However, neither is canonical, so both only get p
2571 flags. NWC, 2000/11/25 */
2572 /* Both already have p flags, so do nothing */
2574 const NV nv = SvNVX(sv);
2575 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2576 if (SvIVX(sv) == I_V(nv)) {
2579 /* It had no "." so it must be integer. */
2583 /* between IV_MAX and NV(UV_MAX).
2584 Could be slightly > UV_MAX */
2586 if (numtype & IS_NUMBER_NOT_INT) {
2587 /* UV and NV both imprecise. */
2589 const UV nv_as_uv = U_V(nv);
2591 if (value == nv_as_uv && SvUVX(sv) != UV_MAX) {
2600 /* It might be more code efficient to go through the entire logic above
2601 and conditionally set with SvNOKp_on() rather than SvNOK(), but it
2602 gets complex and potentially buggy, so more programmer efficient
2603 to do it this way, by turning off the public flags: */
2605 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK);
2606 #endif /* NV_PRESERVES_UV */
2609 if (isGV_with_GP(sv)) {
2610 glob_2number(MUTABLE_GV(sv));
2614 if (!PL_localizing && !SvPADTMP(sv) && ckWARN(WARN_UNINITIALIZED))
2616 assert (SvTYPE(sv) >= SVt_NV);
2617 /* Typically the caller expects that sv_any is not NULL now. */
2618 /* XXX Ilya implies that this is a bug in callers that assume this
2619 and ideally should be fixed. */
2622 #if defined(USE_LONG_DOUBLE)
2624 STORE_NUMERIC_LOCAL_SET_STANDARD();
2625 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2nv(%" PERL_PRIgldbl ")\n",
2626 PTR2UV(sv), SvNVX(sv));
2627 RESTORE_NUMERIC_LOCAL();
2631 STORE_NUMERIC_LOCAL_SET_STANDARD();
2632 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 1nv(%"NVgf")\n",
2633 PTR2UV(sv), SvNVX(sv));
2634 RESTORE_NUMERIC_LOCAL();
2643 Return an SV with the numeric value of the source SV, doing any necessary
2644 reference or overload conversion. You must use the C<SvNUM(sv)> macro to
2645 access this function.
2651 Perl_sv_2num(pTHX_ register SV *const sv)
2653 PERL_ARGS_ASSERT_SV_2NUM;
2658 SV * const tmpsv = AMG_CALLunary(sv, numer_amg);
2659 TAINT_IF(tmpsv && SvTAINTED(tmpsv));
2660 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
2661 return sv_2num(tmpsv);
2663 return sv_2mortal(newSVuv(PTR2UV(SvRV(sv))));
2666 /* uiv_2buf(): private routine for use by sv_2pv_flags(): print an IV or
2667 * UV as a string towards the end of buf, and return pointers to start and
2670 * We assume that buf is at least TYPE_CHARS(UV) long.
2674 S_uiv_2buf(char *const buf, const IV iv, UV uv, const int is_uv, char **const peob)
2676 char *ptr = buf + TYPE_CHARS(UV);
2677 char * const ebuf = ptr;
2680 PERL_ARGS_ASSERT_UIV_2BUF;
2692 *--ptr = '0' + (char)(uv % 10);
2701 =for apidoc sv_2pv_flags
2703 Returns a pointer to the string value of an SV, and sets *lp to its length.
2704 If flags includes SV_GMAGIC, does an mg_get() first. Coerces sv to a
2705 string if necessary. Normally invoked via the C<SvPV_flags> macro.
2706 C<sv_2pv()> and C<sv_2pv_nomg> usually end up here too.
2712 Perl_sv_2pv_flags(pTHX_ register SV *const sv, STRLEN *const lp, const I32 flags)
2722 if (SvGMAGICAL(sv)) {
2723 if (flags & SV_GMAGIC)
2728 if (flags & SV_MUTABLE_RETURN)
2729 return SvPVX_mutable(sv);
2730 if (flags & SV_CONST_RETURN)
2731 return (char *)SvPVX_const(sv);
2734 if (SvIOKp(sv) || SvNOKp(sv)) {
2735 char tbuf[64]; /* Must fit sprintf/Gconvert of longest IV/NV */
2740 ? my_snprintf(tbuf, sizeof(tbuf), "%"UVuf, (UV)SvUVX(sv))
2741 : my_snprintf(tbuf, sizeof(tbuf), "%"IVdf, (IV)SvIVX(sv));
2742 } else if(SvNVX(sv) == 0.0) {
2747 Gconvert(SvNVX(sv), NV_DIG, 0, tbuf);
2754 SvUPGRADE(sv, SVt_PV);
2757 s = SvGROW_mutable(sv, len + 1);
2760 return (char*)memcpy(s, tbuf, len + 1);
2766 assert(SvTYPE(sv) >= SVt_PVMG);
2767 /* This falls through to the report_uninit near the end of the
2769 } else if (SvTHINKFIRST(sv)) {
2774 if (flags & SV_SKIP_OVERLOAD)
2776 tmpstr = AMG_CALLunary(sv, string_amg);
2777 TAINT_IF(tmpstr && SvTAINTED(tmpstr));
2778 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2780 /* char *pv = lp ? SvPV(tmpstr, *lp) : SvPV_nolen(tmpstr);
2784 if ((SvFLAGS(tmpstr) & (SVf_POK)) == SVf_POK) {
2785 if (flags & SV_CONST_RETURN) {
2786 pv = (char *) SvPVX_const(tmpstr);
2788 pv = (flags & SV_MUTABLE_RETURN)
2789 ? SvPVX_mutable(tmpstr) : SvPVX(tmpstr);
2792 *lp = SvCUR(tmpstr);
2794 pv = sv_2pv_flags(tmpstr, lp, flags);
2807 SV *const referent = SvRV(sv);
2811 retval = buffer = savepvn("NULLREF", len);
2812 } else if (SvTYPE(referent) == SVt_REGEXP && (
2813 !(PL_curcop->cop_hints & HINT_NO_AMAGIC)
2814 || amagic_is_enabled(string_amg)
2816 REGEXP * const re = (REGEXP *)MUTABLE_PTR(referent);
2821 /* If the regex is UTF-8 we want the containing scalar to
2822 have an UTF-8 flag too */
2828 if ((seen_evals = RX_SEEN_EVALS(re)))
2829 PL_reginterp_cnt += seen_evals;
2832 *lp = RX_WRAPLEN(re);
2834 return RX_WRAPPED(re);
2836 const char *const typestr = sv_reftype(referent, 0);
2837 const STRLEN typelen = strlen(typestr);
2838 UV addr = PTR2UV(referent);
2839 const char *stashname = NULL;
2840 STRLEN stashnamelen = 0; /* hush, gcc */
2841 const char *buffer_end;
2843 if (SvOBJECT(referent)) {
2844 const HEK *const name = HvNAME_HEK(SvSTASH(referent));
2847 stashname = HEK_KEY(name);
2848 stashnamelen = HEK_LEN(name);
2850 if (HEK_UTF8(name)) {
2856 stashname = "__ANON__";
2859 len = stashnamelen + 1 /* = */ + typelen + 3 /* (0x */
2860 + 2 * sizeof(UV) + 2 /* )\0 */;
2862 len = typelen + 3 /* (0x */
2863 + 2 * sizeof(UV) + 2 /* )\0 */;
2866 Newx(buffer, len, char);
2867 buffer_end = retval = buffer + len;
2869 /* Working backwards */
2873 *--retval = PL_hexdigit[addr & 15];
2874 } while (addr >>= 4);
2880 memcpy(retval, typestr, typelen);
2884 retval -= stashnamelen;
2885 memcpy(retval, stashname, stashnamelen);
2887 /* retval may not necessarily have reached the start of the
2889 assert (retval >= buffer);
2891 len = buffer_end - retval - 1; /* -1 for that \0 */
2899 if (SvREADONLY(sv) && !SvOK(sv)) {
2902 if (flags & SV_UNDEF_RETURNS_NULL)
2904 if (ckWARN(WARN_UNINITIALIZED))
2909 if (SvIOK(sv) || ((SvIOKp(sv) && !SvNOKp(sv)))) {
2910 /* I'm assuming that if both IV and NV are equally valid then
2911 converting the IV is going to be more efficient */
2912 const U32 isUIOK = SvIsUV(sv);
2913 char buf[TYPE_CHARS(UV)];
2917 if (SvTYPE(sv) < SVt_PVIV)
2918 sv_upgrade(sv, SVt_PVIV);
2919 ptr = uiv_2buf(buf, SvIVX(sv), SvUVX(sv), isUIOK, &ebuf);
2921 /* inlined from sv_setpvn */
2922 s = SvGROW_mutable(sv, len + 1);
2923 Move(ptr, s, len, char);
2927 else if (SvNOKp(sv)) {
2928 if (SvTYPE(sv) < SVt_PVNV)
2929 sv_upgrade(sv, SVt_PVNV);
2930 if (SvNVX(sv) == 0.0) {
2931 s = SvGROW_mutable(sv, 2);
2936 /* The +20 is pure guesswork. Configure test needed. --jhi */
2937 s = SvGROW_mutable(sv, NV_DIG + 20);
2938 /* some Xenix systems wipe out errno here */
2939 Gconvert(SvNVX(sv), NV_DIG, 0, s);
2949 if (isGV_with_GP(sv)) {
2950 GV *const gv = MUTABLE_GV(sv);
2951 SV *const buffer = sv_newmortal();
2953 gv_efullname3(buffer, gv, "*");
2955 assert(SvPOK(buffer));
2957 *lp = SvCUR(buffer);
2959 if ( SvUTF8(buffer) ) SvUTF8_on(sv);
2960 return SvPVX(buffer);
2965 if (flags & SV_UNDEF_RETURNS_NULL)
2967 if (!PL_localizing && !SvPADTMP(sv) && ckWARN(WARN_UNINITIALIZED))
2969 if (SvTYPE(sv) < SVt_PV)
2970 /* Typically the caller expects that sv_any is not NULL now. */
2971 sv_upgrade(sv, SVt_PV);
2975 const STRLEN len = s - SvPVX_const(sv);
2981 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
2982 PTR2UV(sv),SvPVX_const(sv)));
2983 if (flags & SV_CONST_RETURN)
2984 return (char *)SvPVX_const(sv);
2985 if (flags & SV_MUTABLE_RETURN)
2986 return SvPVX_mutable(sv);
2991 =for apidoc sv_copypv
2993 Copies a stringified representation of the source SV into the
2994 destination SV. Automatically performs any necessary mg_get and
2995 coercion of numeric values into strings. Guaranteed to preserve
2996 UTF8 flag even from overloaded objects. Similar in nature to
2997 sv_2pv[_flags] but operates directly on an SV instead of just the
2998 string. Mostly uses sv_2pv_flags to do its work, except when that
2999 would lose the UTF-8'ness of the PV.
3005 Perl_sv_copypv(pTHX_ SV *const dsv, register SV *const ssv)
3008 const char * const s = SvPV_const(ssv,len);
3010 PERL_ARGS_ASSERT_SV_COPYPV;
3012 sv_setpvn(dsv,s,len);
3020 =for apidoc sv_2pvbyte
3022 Return a pointer to the byte-encoded representation of the SV, and set *lp
3023 to its length. May cause the SV to be downgraded from UTF-8 as a
3026 Usually accessed via the C<SvPVbyte> macro.
3032 Perl_sv_2pvbyte(pTHX_ register SV *sv, STRLEN *const lp)
3034 PERL_ARGS_ASSERT_SV_2PVBYTE;
3036 if ((SvTHINKFIRST(sv) && !SvIsCOW(sv)) || isGV_with_GP(sv)) {
3037 SV *sv2 = sv_newmortal();
3041 else SvGETMAGIC(sv);
3042 sv_utf8_downgrade(sv,0);
3043 return lp ? SvPV_nomg(sv,*lp) : SvPV_nomg_nolen(sv);
3047 =for apidoc sv_2pvutf8
3049 Return a pointer to the UTF-8-encoded representation of the SV, and set *lp
3050 to its length. May cause the SV to be upgraded to UTF-8 as a side-effect.
3052 Usually accessed via the C<SvPVutf8> macro.
3058 Perl_sv_2pvutf8(pTHX_ register SV *sv, STRLEN *const lp)
3060 PERL_ARGS_ASSERT_SV_2PVUTF8;
3062 if ((SvTHINKFIRST(sv) && !SvIsCOW(sv)) || isGV_with_GP(sv))
3063 sv = sv_mortalcopy(sv);
3064 sv_utf8_upgrade(sv);
3065 return lp ? SvPV(sv,*lp) : SvPV_nolen(sv);
3070 =for apidoc sv_2bool
3072 This macro is only used by sv_true() or its macro equivalent, and only if
3073 the latter's argument is neither SvPOK, SvIOK nor SvNOK.
3074 It calls sv_2bool_flags with the SV_GMAGIC flag.
3076 =for apidoc sv_2bool_flags
3078 This function is only used by sv_true() and friends, and only if
3079 the latter's argument is neither SvPOK, SvIOK nor SvNOK. If the flags
3080 contain SV_GMAGIC, then it does an mg_get() first.
3087 Perl_sv_2bool_flags(pTHX_ register SV *const sv, const I32 flags)
3091 PERL_ARGS_ASSERT_SV_2BOOL_FLAGS;
3093 if(flags & SV_GMAGIC) SvGETMAGIC(sv);
3099 SV * const tmpsv = AMG_CALLunary(sv, bool__amg);
3100 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
3101 return cBOOL(SvTRUE(tmpsv));
3103 return SvRV(sv) != 0;
3106 register XPV* const Xpvtmp = (XPV*)SvANY(sv);
3108 (*sv->sv_u.svu_pv > '0' ||
3109 Xpvtmp->xpv_cur > 1 ||
3110 (Xpvtmp->xpv_cur && *sv->sv_u.svu_pv != '0')))
3117 return SvIVX(sv) != 0;
3120 return SvNVX(sv) != 0.0;
3122 if (isGV_with_GP(sv))
3132 =for apidoc sv_utf8_upgrade
3134 Converts the PV of an SV to its UTF-8-encoded form.
3135 Forces the SV to string form if it is not already.
3136 Will C<mg_get> on C<sv> if appropriate.
3137 Always sets the SvUTF8 flag to avoid future validity checks even
3138 if the whole string is the same in UTF-8 as not.
3139 Returns the number of bytes in the converted string
3141 This is not as a general purpose byte encoding to Unicode interface:
3142 use the Encode extension for that.
3144 =for apidoc sv_utf8_upgrade_nomg
3146 Like sv_utf8_upgrade, but doesn't do magic on C<sv>.
3148 =for apidoc sv_utf8_upgrade_flags
3150 Converts the PV of an SV to its UTF-8-encoded form.
3151 Forces the SV to string form if it is not already.
3152 Always sets the SvUTF8 flag to avoid future validity checks even
3153 if all the bytes are invariant in UTF-8.
3154 If C<flags> has C<SV_GMAGIC> bit set,
3155 will C<mg_get> on C<sv> if appropriate, else not.
3156 Returns the number of bytes in the converted string
3157 C<sv_utf8_upgrade> and
3158 C<sv_utf8_upgrade_nomg> are implemented in terms of this function.
3160 This is not as a general purpose byte encoding to Unicode interface:
3161 use the Encode extension for that.
3165 The grow version is currently not externally documented. It adds a parameter,
3166 extra, which is the number of unused bytes the string of 'sv' is guaranteed to
3167 have free after it upon return. This allows the caller to reserve extra space
3168 that it intends to fill, to avoid extra grows.
3170 Also externally undocumented for the moment is the flag SV_FORCE_UTF8_UPGRADE,
3171 which can be used to tell this function to not first check to see if there are
3172 any characters that are different in UTF-8 (variant characters) which would
3173 force it to allocate a new string to sv, but to assume there are. Typically
3174 this flag is used by a routine that has already parsed the string to find that
3175 there are such characters, and passes this information on so that the work
3176 doesn't have to be repeated.
3178 (One might think that the calling routine could pass in the position of the
3179 first such variant, so it wouldn't have to be found again. But that is not the
3180 case, because typically when the caller is likely to use this flag, it won't be
3181 calling this routine unless it finds something that won't fit into a byte.
3182 Otherwise it tries to not upgrade and just use bytes. But some things that
3183 do fit into a byte are variants in utf8, and the caller may not have been
3184 keeping track of these.)
3186 If the routine itself changes the string, it adds a trailing NUL. Such a NUL
3187 isn't guaranteed due to having other routines do the work in some input cases,
3188 or if the input is already flagged as being in utf8.
3190 The speed of this could perhaps be improved for many cases if someone wanted to
3191 write a fast function that counts the number of variant characters in a string,
3192 especially if it could return the position of the first one.
3197 Perl_sv_utf8_upgrade_flags_grow(pTHX_ register SV *const sv, const I32 flags, STRLEN extra)
3201 PERL_ARGS_ASSERT_SV_UTF8_UPGRADE_FLAGS_GROW;
3203 if (sv == &PL_sv_undef)
3207 if (SvREADONLY(sv) && (SvPOKp(sv) || SvIOKp(sv) || SvNOKp(sv))) {
3208 (void) sv_2pv_flags(sv,&len, flags);
3210 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3214 (void) SvPV_force_flags(sv,len,flags & SV_GMAGIC);
3219 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3224 sv_force_normal_flags(sv, 0);
3227 if (PL_encoding && !(flags & SV_UTF8_NO_ENCODING)) {
3228 sv_recode_to_utf8(sv, PL_encoding);
3229 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3233 if (SvCUR(sv) == 0) {
3234 if (extra) SvGROW(sv, extra);
3235 } else { /* Assume Latin-1/EBCDIC */
3236 /* This function could be much more efficient if we
3237 * had a FLAG in SVs to signal if there are any variant
3238 * chars in the PV. Given that there isn't such a flag
3239 * make the loop as fast as possible (although there are certainly ways
3240 * to speed this up, eg. through vectorization) */
3241 U8 * s = (U8 *) SvPVX_const(sv);
3242 U8 * e = (U8 *) SvEND(sv);
3244 STRLEN two_byte_count = 0;
3246 if (flags & SV_FORCE_UTF8_UPGRADE) goto must_be_utf8;
3248 /* See if really will need to convert to utf8. We mustn't rely on our
3249 * incoming SV being well formed and having a trailing '\0', as certain
3250 * code in pp_formline can send us partially built SVs. */
3254 if (NATIVE_IS_INVARIANT(ch)) continue;
3256 t--; /* t already incremented; re-point to first variant */
3261 /* utf8 conversion not needed because all are invariants. Mark as
3262 * UTF-8 even if no variant - saves scanning loop */
3264 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3269 /* Here, the string should be converted to utf8, either because of an
3270 * input flag (two_byte_count = 0), or because a character that
3271 * requires 2 bytes was found (two_byte_count = 1). t points either to
3272 * the beginning of the string (if we didn't examine anything), or to
3273 * the first variant. In either case, everything from s to t - 1 will
3274 * occupy only 1 byte each on output.
3276 * There are two main ways to convert. One is to create a new string
3277 * and go through the input starting from the beginning, appending each
3278 * converted value onto the new string as we go along. It's probably
3279 * best to allocate enough space in the string for the worst possible
3280 * case rather than possibly running out of space and having to
3281 * reallocate and then copy what we've done so far. Since everything
3282 * from s to t - 1 is invariant, the destination can be initialized
3283 * with these using a fast memory copy
3285 * The other way is to figure out exactly how big the string should be
3286 * by parsing the entire input. Then you don't have to make it big
3287 * enough to handle the worst possible case, and more importantly, if
3288 * the string you already have is large enough, you don't have to
3289 * allocate a new string, you can copy the last character in the input
3290 * string to the final position(s) that will be occupied by the
3291 * converted string and go backwards, stopping at t, since everything
3292 * before that is invariant.
3294 * There are advantages and disadvantages to each method.
3296 * In the first method, we can allocate a new string, do the memory
3297 * copy from the s to t - 1, and then proceed through the rest of the
3298 * string byte-by-byte.
3300 * In the second method, we proceed through the rest of the input
3301 * string just calculating how big the converted string will be. Then
3302 * there are two cases:
3303 * 1) if the string has enough extra space to handle the converted
3304 * value. We go backwards through the string, converting until we
3305 * get to the position we are at now, and then stop. If this
3306 * position is far enough along in the string, this method is
3307 * faster than the other method. If the memory copy were the same
3308 * speed as the byte-by-byte loop, that position would be about
3309 * half-way, as at the half-way mark, parsing to the end and back
3310 * is one complete string's parse, the same amount as starting
3311 * over and going all the way through. Actually, it would be
3312 * somewhat less than half-way, as it's faster to just count bytes
3313 * than to also copy, and we don't have the overhead of allocating
3314 * a new string, changing the scalar to use it, and freeing the
3315 * existing one. But if the memory copy is fast, the break-even
3316 * point is somewhere after half way. The counting loop could be
3317 * sped up by vectorization, etc, to move the break-even point
3318 * further towards the beginning.
3319 * 2) if the string doesn't have enough space to handle the converted
3320 * value. A new string will have to be allocated, and one might
3321 * as well, given that, start from the beginning doing the first
3322 * method. We've spent extra time parsing the string and in
3323 * exchange all we've gotten is that we know precisely how big to
3324 * make the new one. Perl is more optimized for time than space,
3325 * so this case is a loser.
3326 * So what I've decided to do is not use the 2nd method unless it is
3327 * guaranteed that a new string won't have to be allocated, assuming
3328 * the worst case. I also decided not to put any more conditions on it
3329 * than this, for now. It seems likely that, since the worst case is
3330 * twice as big as the unknown portion of the string (plus 1), we won't
3331 * be guaranteed enough space, causing us to go to the first method,
3332 * unless the string is short, or the first variant character is near
3333 * the end of it. In either of these cases, it seems best to use the
3334 * 2nd method. The only circumstance I can think of where this would
3335 * be really slower is if the string had once had much more data in it
3336 * than it does now, but there is still a substantial amount in it */
3339 STRLEN invariant_head = t - s;
3340 STRLEN size = invariant_head + (e - t) * 2 + 1 + extra;
3341 if (SvLEN(sv) < size) {
3343 /* Here, have decided to allocate a new string */
3348 Newx(dst, size, U8);
3350 /* If no known invariants at the beginning of the input string,
3351 * set so starts from there. Otherwise, can use memory copy to
3352 * get up to where we are now, and then start from here */
3354 if (invariant_head <= 0) {
3357 Copy(s, dst, invariant_head, char);
3358 d = dst + invariant_head;
3362 const UV uv = NATIVE8_TO_UNI(*t++);
3363 if (UNI_IS_INVARIANT(uv))
3364 *d++ = (U8)UNI_TO_NATIVE(uv);
3366 *d++ = (U8)UTF8_EIGHT_BIT_HI(uv);
3367 *d++ = (U8)UTF8_EIGHT_BIT_LO(uv);
3371 SvPV_free(sv); /* No longer using pre-existing string */
3372 SvPV_set(sv, (char*)dst);
3373 SvCUR_set(sv, d - dst);
3374 SvLEN_set(sv, size);
3377 /* Here, have decided to get the exact size of the string.
3378 * Currently this happens only when we know that there is
3379 * guaranteed enough space to fit the converted string, so
3380 * don't have to worry about growing. If two_byte_count is 0,
3381 * then t points to the first byte of the string which hasn't
3382 * been examined yet. Otherwise two_byte_count is 1, and t
3383 * points to the first byte in the string that will expand to
3384 * two. Depending on this, start examining at t or 1 after t.
3387 U8 *d = t + two_byte_count;
3390 /* Count up the remaining bytes that expand to two */
3393 const U8 chr = *d++;
3394 if (! NATIVE_IS_INVARIANT(chr)) two_byte_count++;
3397 /* The string will expand by just the number of bytes that
3398 * occupy two positions. But we are one afterwards because of
3399 * the increment just above. This is the place to put the
3400 * trailing NUL, and to set the length before we decrement */
3402 d += two_byte_count;
3403 SvCUR_set(sv, d - s);
3407 /* Having decremented d, it points to the position to put the
3408 * very last byte of the expanded string. Go backwards through
3409 * the string, copying and expanding as we go, stopping when we
3410 * get to the part that is invariant the rest of the way down */
3414 const U8 ch = NATIVE8_TO_UNI(*e--);
3415 if (UNI_IS_INVARIANT(ch)) {
3416 *d-- = UNI_TO_NATIVE(ch);
3418 *d-- = (U8)UTF8_EIGHT_BIT_LO(ch);
3419 *d-- = (U8)UTF8_EIGHT_BIT_HI(ch);
3424 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
3425 /* Update pos. We do it at the end rather than during
3426 * the upgrade, to avoid slowing down the common case
3427 * (upgrade without pos) */
3428 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
3430 I32 pos = mg->mg_len;
3431 if (pos > 0 && (U32)pos > invariant_head) {
3432 U8 *d = (U8*) SvPVX(sv) + invariant_head;
3433 STRLEN n = (U32)pos - invariant_head;
3435 if (UTF8_IS_START(*d))
3440 mg->mg_len = d - (U8*)SvPVX(sv);
3443 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
3444 magic_setutf8(sv,mg); /* clear UTF8 cache */
3449 /* Mark as UTF-8 even if no variant - saves scanning loop */
3455 =for apidoc sv_utf8_downgrade
3457 Attempts to convert the PV of an SV from characters to bytes.
3458 If the PV contains a character that cannot fit
3459 in a byte, this conversion will fail;
3460 in this case, either returns false or, if C<fail_ok> is not
3463 This is not as a general purpose Unicode to byte encoding interface:
3464 use the Encode extension for that.
3470 Perl_sv_utf8_downgrade(pTHX_ register SV *const sv, const bool fail_ok)
3474 PERL_ARGS_ASSERT_SV_UTF8_DOWNGRADE;
3476 if (SvPOKp(sv) && SvUTF8(sv)) {
3480 int mg_flags = SV_GMAGIC;
3483 sv_force_normal_flags(sv, 0);
3485 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
3487 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
3489 I32 pos = mg->mg_len;
3491 sv_pos_b2u(sv, &pos);
3492 mg_flags = 0; /* sv_pos_b2u does get magic */
3496 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
3497 magic_setutf8(sv,mg); /* clear UTF8 cache */
3500 s = (U8 *) SvPV_flags(sv, len, mg_flags);
3502 if (!utf8_to_bytes(s, &len)) {
3507 Perl_croak(aTHX_ "Wide character in %s",
3510 Perl_croak(aTHX_ "Wide character");
3521 =for apidoc sv_utf8_encode
3523 Converts the PV of an SV to UTF-8, but then turns the C<SvUTF8>
3524 flag off so that it looks like octets again.
3530 Perl_sv_utf8_encode(pTHX_ register SV *const sv)
3532 PERL_ARGS_ASSERT_SV_UTF8_ENCODE;
3534 if (SvREADONLY(sv)) {
3535 sv_force_normal_flags(sv, 0);
3537 (void) sv_utf8_upgrade(sv);
3542 =for apidoc sv_utf8_decode
3544 If the PV of the SV is an octet sequence in UTF-8
3545 and contains a multiple-byte character, the C<SvUTF8> flag is turned on
3546 so that it looks like a character. If the PV contains only single-byte
3547 characters, the C<SvUTF8> flag stays off.
3548 Scans PV for validity and returns false if the PV is invalid UTF-8.
3554 Perl_sv_utf8_decode(pTHX_ register SV *const sv)
3556 PERL_ARGS_ASSERT_SV_UTF8_DECODE;
3559 const U8 *start, *c;
3562 /* The octets may have got themselves encoded - get them back as
3565 if (!sv_utf8_downgrade(sv, TRUE))
3568 /* it is actually just a matter of turning the utf8 flag on, but
3569 * we want to make sure everything inside is valid utf8 first.
3571 c = start = (const U8 *) SvPVX_const(sv);
3572 if (!is_utf8_string(c, SvCUR(sv)))
3574 e = (const U8 *) SvEND(sv);
3577 if (!UTF8_IS_INVARIANT(ch)) {
3582 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
3583 /* adjust pos to the start of a UTF8 char sequence */
3584 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
3586 I32 pos = mg->mg_len;
3588 for (c = start + pos; c > start; c--) {
3589 if (UTF8_IS_START(*c))
3592 mg->mg_len = c - start;
3595 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
3596 magic_setutf8(sv,mg); /* clear UTF8 cache */
3603 =for apidoc sv_setsv
3605 Copies the contents of the source SV C<ssv> into the destination SV
3606 C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3607 function if the source SV needs to be reused. Does not handle 'set' magic.
3608 Loosely speaking, it performs a copy-by-value, obliterating any previous
3609 content of the destination.
3611 You probably want to use one of the assortment of wrappers, such as
3612 C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3613 C<SvSetMagicSV_nosteal>.
3615 =for apidoc sv_setsv_flags
3617 Copies the contents of the source SV C<ssv> into the destination SV
3618 C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3619 function if the source SV needs to be reused. Does not handle 'set' magic.
3620 Loosely speaking, it performs a copy-by-value, obliterating any previous
3621 content of the destination.
3622 If the C<flags> parameter has the C<SV_GMAGIC> bit set, will C<mg_get> on
3623 C<ssv> if appropriate, else not. If the C<flags>
3624 parameter has the C<NOSTEAL> bit set then the
3625 buffers of temps will not be stolen. <sv_setsv>
3626 and C<sv_setsv_nomg> are implemented in terms of this function.
3628 You probably want to use one of the assortment of wrappers, such as
3629 C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3630 C<SvSetMagicSV_nosteal>.
3632 This is the primary function for copying scalars, and most other
3633 copy-ish functions and macros use this underneath.
3639 S_glob_assign_glob(pTHX_ SV *const dstr, SV *const sstr, const int dtype)
3641 I32 mro_changes = 0; /* 1 = method, 2 = isa, 3 = recursive isa */
3642 HV *old_stash = NULL;
3644 PERL_ARGS_ASSERT_GLOB_ASSIGN_GLOB;
3646 if (dtype != SVt_PVGV && !isGV_with_GP(dstr)) {
3647 const char * const name = GvNAME(sstr);
3648 const STRLEN len = GvNAMELEN(sstr);
3650 if (dtype >= SVt_PV) {
3656 SvUPGRADE(dstr, SVt_PVGV);
3657 (void)SvOK_off(dstr);
3658 /* We have to turn this on here, even though we turn it off
3659 below, as GvSTASH will fail an assertion otherwise. */
3660 isGV_with_GP_on(dstr);
3662 GvSTASH(dstr) = GvSTASH(sstr);
3664 Perl_sv_add_backref(aTHX_ MUTABLE_SV(GvSTASH(dstr)), dstr);
3665 gv_name_set(MUTABLE_GV(dstr), name, len,
3666 GV_ADD | (GvNAMEUTF8(sstr) ? SVf_UTF8 : 0 ));
3667 SvFAKE_on(dstr); /* can coerce to non-glob */
3670 if(GvGP(MUTABLE_GV(sstr))) {
3671 /* If source has method cache entry, clear it */
3673 SvREFCNT_dec(GvCV(sstr));
3674 GvCV_set(sstr, NULL);
3677 /* If source has a real method, then a method is
3680 GvCV((const GV *)sstr) && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3686 /* If dest already had a real method, that's a change as well */
3688 !mro_changes && GvGP(MUTABLE_GV(dstr)) && GvCVu((const GV *)dstr)
3689 && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3694 /* We don't need to check the name of the destination if it was not a
3695 glob to begin with. */
3696 if(dtype == SVt_PVGV) {
3697 const char * const name = GvNAME((const GV *)dstr);
3700 /* The stash may have been detached from the symbol table, so
3702 && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3703 && GvAV((const GV *)sstr)
3707 const STRLEN len = GvNAMELEN(dstr);
3708 if ((len > 1 && name[len-2] == ':' && name[len-1] == ':')
3709 || (len == 1 && name[0] == ':')) {
3712 /* Set aside the old stash, so we can reset isa caches on
3714 if((old_stash = GvHV(dstr)))
3715 /* Make sure we do not lose it early. */
3716 SvREFCNT_inc_simple_void_NN(
3717 sv_2mortal((SV *)old_stash)
3723 gp_free(MUTABLE_GV(dstr));
3724 isGV_with_GP_off(dstr); /* SvOK_off does not like globs. */
3725 (void)SvOK_off(dstr);
3726 isGV_with_GP_on(dstr);
3727 GvINTRO_off(dstr); /* one-shot flag */
3728 GvGP_set(dstr, gp_ref(GvGP(sstr)));
3729 if (SvTAINTED(sstr))
3731 if (GvIMPORTED(dstr) != GVf_IMPORTED
3732 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3734 GvIMPORTED_on(dstr);
3737 if(mro_changes == 2) {
3739 SV * const sref = (SV *)GvAV((const GV *)dstr);
3740 if (SvSMAGICAL(sref) && (mg = mg_find(sref, PERL_MAGIC_isa))) {
3741 if (SvTYPE(mg->mg_obj) != SVt_PVAV) {
3742 AV * const ary = newAV();
3743 av_push(ary, mg->mg_obj); /* takes the refcount */
3744 mg->mg_obj = (SV *)ary;
3746 av_push((AV *)mg->mg_obj, SvREFCNT_inc_simple_NN(dstr));
3748 else sv_magic(sref, dstr, PERL_MAGIC_isa, NULL, 0);
3749 mro_isa_changed_in(GvSTASH(dstr));
3751 else if(mro_changes == 3) {
3752 HV * const stash = GvHV(dstr);
3753 if(old_stash ? (HV *)HvENAME_get(old_stash) : stash)
3759 else if(mro_changes) mro_method_changed_in(GvSTASH(dstr));
3764 S_glob_assign_ref(pTHX_ SV *const dstr, SV *const sstr)
3766 SV * const sref = SvREFCNT_inc(SvRV(sstr));
3768 const int intro = GvINTRO(dstr);
3771 const U32 stype = SvTYPE(sref);
3773 PERL_ARGS_ASSERT_GLOB_ASSIGN_REF;
3776 GvINTRO_off(dstr); /* one-shot flag */
3777 GvLINE(dstr) = CopLINE(PL_curcop);
3778 GvEGV(dstr) = MUTABLE_GV(dstr);
3783 location = (SV **) &(GvGP(dstr)->gp_cv); /* XXX bypassing GvCV_set */
3784 import_flag = GVf_IMPORTED_CV;
3787 location = (SV **) &GvHV(dstr);
3788 import_flag = GVf_IMPORTED_HV;
3791 location = (SV **) &GvAV(dstr);
3792 import_flag = GVf_IMPORTED_AV;
3795 location = (SV **) &GvIOp(dstr);
3798 location = (SV **) &GvFORM(dstr);
3801 location = &GvSV(dstr);
3802 import_flag = GVf_IMPORTED_SV;
3805 if (stype == SVt_PVCV) {
3806 /*if (GvCVGEN(dstr) && (GvCV(dstr) != (const CV *)sref || GvCVGEN(dstr))) {*/
3807 if (GvCVGEN(dstr)) {
3808 SvREFCNT_dec(GvCV(dstr));
3809 GvCV_set(dstr, NULL);
3810 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3813 SAVEGENERICSV(*location);
3817 if (stype == SVt_PVCV && (*location != sref || GvCVGEN(dstr))) {
3818 CV* const cv = MUTABLE_CV(*location);
3820 if (!GvCVGEN((const GV *)dstr) &&
3821 (CvROOT(cv) || CvXSUB(cv)) &&
3822 /* redundant check that avoids creating the extra SV
3823 most of the time: */
3824 (CvCONST(cv) || ckWARN(WARN_REDEFINE)))
3826 SV * const new_const_sv =
3827 CvCONST((const CV *)sref)
3828 ? cv_const_sv((const CV *)sref)
3830 report_redefined_cv(
3831 sv_2mortal(Perl_newSVpvf(aTHX_
3834 HvNAME_HEK(GvSTASH((const GV *)dstr))
3836 HEKfARG(GvENAME_HEK(MUTABLE_GV(dstr)))
3839 CvCONST((const CV *)sref) ? &new_const_sv : NULL
3843 cv_ckproto_len_flags(cv, (const GV *)dstr,
3844 SvPOK(sref) ? CvPROTO(sref) : NULL,
3845 SvPOK(sref) ? CvPROTOLEN(sref) : 0,
3846 SvPOK(sref) ? SvUTF8(sref) : 0);
3848 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3849 GvASSUMECV_on(dstr);
3850 if(GvSTASH(dstr)) mro_method_changed_in(GvSTASH(dstr)); /* sub foo { 1 } sub bar { 2 } *bar = \&foo */
3853 if (import_flag && !(GvFLAGS(dstr) & import_flag)
3854 && CopSTASH_ne(PL_curcop, GvSTASH(dstr))) {
3855 GvFLAGS(dstr) |= import_flag;
3857 if (stype == SVt_PVHV) {
3858 const char * const name = GvNAME((GV*)dstr);
3859 const STRLEN len = GvNAMELEN(dstr);
3862 (len > 1 && name[len-2] == ':' && name[len-1] == ':')
3863 || (len == 1 && name[0] == ':')
3865 && (!dref || HvENAME_get(dref))
3868 (HV *)sref, (HV *)dref,
3874 stype == SVt_PVAV && sref != dref
3875 && strEQ(GvNAME((GV*)dstr), "ISA")
3876 /* The stash may have been detached from the symbol table, so
3877 check its name before doing anything. */
3878 && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3881 MAGIC * const omg = dref && SvSMAGICAL(dref)
3882 ? mg_find(dref, PERL_MAGIC_isa)
3884 if (SvSMAGICAL(sref) && (mg = mg_find(sref, PERL_MAGIC_isa))) {
3885 if (SvTYPE(mg->mg_obj) != SVt_PVAV) {
3886 AV * const ary = newAV();
3887 av_push(ary, mg->mg_obj); /* takes the refcount */
3888 mg->mg_obj = (SV *)ary;
3891 if (SvTYPE(omg->mg_obj) == SVt_PVAV) {
3892 SV **svp = AvARRAY((AV *)omg->mg_obj);
3893 I32 items = AvFILLp((AV *)omg->mg_obj) + 1;
3897 SvREFCNT_inc_simple_NN(*svp++)
3903 SvREFCNT_inc_simple_NN(omg->mg_obj)
3907 av_push((AV *)mg->mg_obj,SvREFCNT_inc_simple_NN(dstr));
3912 sref, omg ? omg->mg_obj : dstr, PERL_MAGIC_isa, NULL, 0
3914 mg = mg_find(sref, PERL_MAGIC_isa);
3916 /* Since the *ISA assignment could have affected more than
3917 one stash, don't call mro_isa_changed_in directly, but let
3918 magic_clearisa do it for us, as it already has the logic for
3919 dealing with globs vs arrays of globs. */
3921 Perl_magic_clearisa(aTHX_ NULL, mg);
3926 if (SvTAINTED(sstr))
3932 Perl_sv_setsv_flags(pTHX_ SV *dstr, register SV* sstr, const I32 flags)
3935 register U32 sflags;
3937 register svtype stype;
3939 PERL_ARGS_ASSERT_SV_SETSV_FLAGS;
3944 if (SvIS_FREED(dstr)) {
3945 Perl_croak(aTHX_ "panic: attempt to copy value %" SVf
3946 " to a freed scalar %p", SVfARG(sstr), (void *)dstr);
3948 SV_CHECK_THINKFIRST_COW_DROP(dstr);
3950 sstr = &PL_sv_undef;
3951 if (SvIS_FREED(sstr)) {
3952 Perl_croak(aTHX_ "panic: attempt to copy freed scalar %p to %p",
3953 (void*)sstr, (void*)dstr);
3955 stype = SvTYPE(sstr);
3956 dtype = SvTYPE(dstr);
3958 (void)SvAMAGIC_off(dstr);
3961 /* need to nuke the magic */
3962 sv_unmagic(dstr, PERL_MAGIC_vstring);
3965 /* There's a lot of redundancy below but we're going for speed here */
3970 if (dtype != SVt_PVGV && dtype != SVt_PVLV) {
3971 (void)SvOK_off(dstr);
3979 sv_upgrade(dstr, SVt_IV);
3983 sv_upgrade(dstr, SVt_PVIV);
3987 goto end_of_first_switch;
3989 (void)SvIOK_only(dstr);
3990 SvIV_set(dstr, SvIVX(sstr));
3993 /* SvTAINTED can only be true if the SV has taint magic, which in
3994 turn means that the SV type is PVMG (or greater). This is the
3995 case statement for SVt_IV, so this cannot be true (whatever gcov
3997 assert(!SvTAINTED(sstr));
4002 if (dtype < SVt_PV && dtype != SVt_IV)
4003 sv_upgrade(dstr, SVt_IV);
4011 sv_upgrade(dstr, SVt_NV);
4015 sv_upgrade(dstr, SVt_PVNV);
4019 goto end_of_first_switch;
4021 SvNV_set(dstr, SvNVX(sstr));
4022 (void)SvNOK_only(dstr);
4023 /* SvTAINTED can only be true if the SV has taint magic, which in
4024 turn means that the SV type is PVMG (or greater). This is the
4025 case statement for SVt_NV, so this cannot be true (whatever gcov
4027 assert(!SvTAINTED(sstr));
4033 #ifdef PERL_OLD_COPY_ON_WRITE
4034 if ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS) {
4035 if (dtype < SVt_PVIV)
4036 sv_upgrade(dstr, SVt_PVIV);
4043 sv_upgrade(dstr, SVt_PV);
4046 if (dtype < SVt_PVIV)
4047 sv_upgrade(dstr, SVt_PVIV);
4050 if (dtype < SVt_PVNV)
4051 sv_upgrade(dstr, SVt_PVNV);
4055 const char * const type = sv_reftype(sstr,0);
4057 /* diag_listed_as: Bizarre copy of %s */
4058 Perl_croak(aTHX_ "Bizarre copy of %s in %s", type, OP_DESC(PL_op));
4060 Perl_croak(aTHX_ "Bizarre copy of %s", type);
4065 if (dtype < SVt_REGEXP)
4066 sv_upgrade(dstr, SVt_REGEXP);
4069 /* case SVt_BIND: */
4073 if (SvGMAGICAL(sstr) && (flags & SV_GMAGIC)) {
4075 if (SvTYPE(sstr) != stype)
4076 stype = SvTYPE(sstr);
4078 if (isGV_with_GP(sstr) && dtype <= SVt_PVLV) {
4079 glob_assign_glob(dstr, sstr, dtype);
4082 if (stype == SVt_PVLV)
4083 SvUPGRADE(dstr, SVt_PVNV);
4085 SvUPGRADE(dstr, (svtype)stype);
4087 end_of_first_switch:
4089 /* dstr may have been upgraded. */
4090 dtype = SvTYPE(dstr);
4091 sflags = SvFLAGS(sstr);
4093 if (dtype == SVt_PVCV || dtype == SVt_PVFM) {
4094 /* Assigning to a subroutine sets the prototype. */
4097 const char *const ptr = SvPV_const(sstr, len);
4099 SvGROW(dstr, len + 1);
4100 Copy(ptr, SvPVX(dstr), len + 1, char);
4101 SvCUR_set(dstr, len);
4103 SvFLAGS(dstr) |= sflags & SVf_UTF8;
4104 CvAUTOLOAD_off(dstr);
4108 } else if (dtype == SVt_PVAV || dtype == SVt_PVHV) {
4109 const char * const type = sv_reftype(dstr,0);
4111 /* diag_listed_as: Cannot copy to %s */
4112 Perl_croak(aTHX_ "Cannot copy to %s in %s", type, OP_DESC(PL_op));
4114 Perl_croak(aTHX_ "Cannot copy to %s", type);
4115 } else if (sflags & SVf_ROK) {
4116 if (isGV_with_GP(dstr)
4117 && SvTYPE(SvRV(sstr)) == SVt_PVGV && isGV_with_GP(SvRV(sstr))) {
4120 if (GvIMPORTED(dstr) != GVf_IMPORTED
4121 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
4123 GvIMPORTED_on(dstr);
4128 glob_assign_glob(dstr, sstr, dtype);
4132 if (dtype >= SVt_PV) {
4133 if (isGV_with_GP(dstr)) {
4134 glob_assign_ref(dstr, sstr);
4137 if (SvPVX_const(dstr)) {
4143 (void)SvOK_off(dstr);
4144 SvRV_set(dstr, SvREFCNT_inc(SvRV(sstr)));
4145 SvFLAGS(dstr) |= sflags & SVf_ROK;
4146 assert(!(sflags & SVp_NOK));
4147 assert(!(sflags & SVp_IOK));
4148 assert(!(sflags & SVf_NOK));
4149 assert(!(sflags & SVf_IOK));
4151 else if (isGV_with_GP(dstr)) {
4152 if (!(sflags & SVf_OK)) {
4153 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
4154 "Undefined value assigned to typeglob");
4157 GV *gv = gv_fetchsv_nomg(sstr, GV_ADD, SVt_PVGV);
4158 if (dstr != (const SV *)gv) {
4159 const char * const name = GvNAME((const GV *)dstr);
4160 const STRLEN len = GvNAMELEN(dstr);
4161 HV *old_stash = NULL;
4162 bool reset_isa = FALSE;
4163 if ((len > 1 && name[len-2] == ':' && name[len-1] == ':')
4164 || (len == 1 && name[0] == ':')) {
4165 /* Set aside the old stash, so we can reset isa caches
4166 on its subclasses. */
4167 if((old_stash = GvHV(dstr))) {
4168 /* Make sure we do not lose it early. */
4169 SvREFCNT_inc_simple_void_NN(
4170 sv_2mortal((SV *)old_stash)
4177 gp_free(MUTABLE_GV(dstr));
4178 GvGP_set(dstr, gp_ref(GvGP(gv)));
4181 HV * const stash = GvHV(dstr);
4183 old_stash ? (HV *)HvENAME_get(old_stash) : stash
4193 else if (dtype == SVt_REGEXP && stype == SVt_REGEXP) {
4194 reg_temp_copy((REGEXP*)dstr, (REGEXP*)sstr);
4196 else if (sflags & SVp_POK) {
4200 * Check to see if we can just swipe the string. If so, it's a
4201 * possible small lose on short strings, but a big win on long ones.
4202 * It might even be a win on short strings if SvPVX_const(dstr)
4203 * has to be allocated and SvPVX_const(sstr) has to be freed.
4204 * Likewise if we can set up COW rather than doing an actual copy, we
4205 * drop to the else clause, as the swipe code and the COW setup code
4206 * have much in common.
4209 /* Whichever path we take through the next code, we want this true,
4210 and doing it now facilitates the COW check. */
4211 (void)SvPOK_only(dstr);
4214 /* If we're already COW then this clause is not true, and if COW
4215 is allowed then we drop down to the else and make dest COW
4216 with us. If caller hasn't said that we're allowed to COW
4217 shared hash keys then we don't do the COW setup, even if the
4218 source scalar is a shared hash key scalar. */
4219 (((flags & SV_COW_SHARED_HASH_KEYS)
4220 ? (sflags & (SVf_FAKE|SVf_READONLY)) != (SVf_FAKE|SVf_READONLY)
4221 : 1 /* If making a COW copy is forbidden then the behaviour we
4222 desire is as if the source SV isn't actually already
4223 COW, even if it is. So we act as if the source flags
4224 are not COW, rather than actually testing them. */
4226 #ifndef PERL_OLD_COPY_ON_WRITE
4227 /* The change that added SV_COW_SHARED_HASH_KEYS makes the logic
4228 when PERL_OLD_COPY_ON_WRITE is defined a little wrong.
4229 Conceptually PERL_OLD_COPY_ON_WRITE being defined should
4230 override SV_COW_SHARED_HASH_KEYS, because it means "always COW"
4231 but in turn, it's somewhat dead code, never expected to go
4232 live, but more kept as a placeholder on how to do it better
4233 in a newer implementation. */
4234 /* If we are COW and dstr is a suitable target then we drop down
4235 into the else and make dest a COW of us. */
4236 || (SvFLAGS(dstr) & CAN_COW_MASK) != CAN_COW_FLAGS
4241 (sflags & SVs_TEMP) && /* slated for free anyway? */
4242 !(sflags & SVf_OOK) && /* and not involved in OOK hack? */
4243 (!(flags & SV_NOSTEAL)) &&
4244 /* and we're allowed to steal temps */
4245 SvREFCNT(sstr) == 1 && /* and no other references to it? */
4246 SvLEN(sstr)) /* and really is a string */
4247 #ifdef PERL_OLD_COPY_ON_WRITE
4248 && ((flags & SV_COW_SHARED_HASH_KEYS)
4249 ? (!((sflags & CAN_COW_MASK) == CAN_COW_FLAGS
4250 && (SvFLAGS(dstr) & CAN_COW_MASK) == CAN_COW_FLAGS
4251 && SvTYPE(sstr) >= SVt_PVIV && SvTYPE(sstr) != SVt_PVFM))
4255 /* Failed the swipe test, and it's not a shared hash key either.
4256 Have to copy the string. */
4257 STRLEN len = SvCUR(sstr);
4258 SvGROW(dstr, len + 1); /* inlined from sv_setpvn */
4259 Move(SvPVX_const(sstr),SvPVX(dstr),len,char);
4260 SvCUR_set(dstr, len);
4261 *SvEND(dstr) = '\0';
4263 /* If PERL_OLD_COPY_ON_WRITE is not defined, then isSwipe will always
4265 /* Either it's a shared hash key, or it's suitable for
4266 copy-on-write or we can swipe the string. */
4268 PerlIO_printf(Perl_debug_log, "Copy on write: sstr --> dstr\n");
4272 #ifdef PERL_OLD_COPY_ON_WRITE
4274 if ((sflags & (SVf_FAKE | SVf_READONLY))
4275 != (SVf_FAKE | SVf_READONLY)) {
4276 SvREADONLY_on(sstr);
4278 /* Make the source SV into a loop of 1.
4279 (about to become 2) */
4280 SV_COW_NEXT_SV_SET(sstr, sstr);
4284 /* Initial code is common. */
4285 if (SvPVX_const(dstr)) { /* we know that dtype >= SVt_PV */
4290 /* making another shared SV. */
4291 STRLEN cur = SvCUR(sstr);
4292 STRLEN len = SvLEN(sstr);
4293 #ifdef PERL_OLD_COPY_ON_WRITE
4295 assert (SvTYPE(dstr) >= SVt_PVIV);
4296 /* SvIsCOW_normal */
4297 /* splice us in between source and next-after-source. */
4298 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4299 SV_COW_NEXT_SV_SET(sstr, dstr);
4300 SvPV_set(dstr, SvPVX_mutable(sstr));
4304 /* SvIsCOW_shared_hash */
4305 DEBUG_C(PerlIO_printf(Perl_debug_log,
4306 "Copy on write: Sharing hash\n"));
4308 assert (SvTYPE(dstr) >= SVt_PV);
4310 HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)))));
4312 SvLEN_set(dstr, len);
4313 SvCUR_set(dstr, cur);
4314 SvREADONLY_on(dstr);
4318 { /* Passes the swipe test. */
4319 SvPV_set(dstr, SvPVX_mutable(sstr));
4320 SvLEN_set(dstr, SvLEN(sstr));
4321 SvCUR_set(dstr, SvCUR(sstr));
4324 (void)SvOK_off(sstr); /* NOTE: nukes most SvFLAGS on sstr */
4325 SvPV_set(sstr, NULL);
4331 if (sflags & SVp_NOK) {
4332 SvNV_set(dstr, SvNVX(sstr));
4334 if (sflags & SVp_IOK) {
4335 SvIV_set(dstr, SvIVX(sstr));
4336 /* Must do this otherwise some other overloaded use of 0x80000000
4337 gets confused. I guess SVpbm_VALID */
4338 if (sflags & SVf_IVisUV)
4341 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_NOK|SVp_NOK|SVf_UTF8);
4343 const MAGIC * const smg = SvVSTRING_mg(sstr);
4345 sv_magic(dstr, NULL, PERL_MAGIC_vstring,
4346 smg->mg_ptr, smg->mg_len);
4347 SvRMAGICAL_on(dstr);
4351 else if (sflags & (SVp_IOK|SVp_NOK)) {
4352 (void)SvOK_off(dstr);
4353 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_IVisUV|SVf_NOK|SVp_NOK);
4354 if (sflags & SVp_IOK) {
4355 /* XXXX Do we want to set IsUV for IV(ROK)? Be extra safe... */
4356 SvIV_set(dstr, SvIVX(sstr));
4358 if (sflags & SVp_NOK) {
4359 SvNV_set(dstr, SvNVX(sstr));
4363 if (isGV_with_GP(sstr)) {
4364 gv_efullname3(dstr, MUTABLE_GV(sstr), "*");
4367 (void)SvOK_off(dstr);
4369 if (SvTAINTED(sstr))
4374 =for apidoc sv_setsv_mg
4376 Like C<sv_setsv>, but also handles 'set' magic.
4382 Perl_sv_setsv_mg(pTHX_ SV *const dstr, register SV *const sstr)
4384 PERL_ARGS_ASSERT_SV_SETSV_MG;
4386 sv_setsv(dstr,sstr);
4390 #ifdef PERL_OLD_COPY_ON_WRITE
4392 Perl_sv_setsv_cow(pTHX_ SV *dstr, SV *sstr)
4394 STRLEN cur = SvCUR(sstr);
4395 STRLEN len = SvLEN(sstr);
4396 register char *new_pv;
4398 PERL_ARGS_ASSERT_SV_SETSV_COW;
4401 PerlIO_printf(Perl_debug_log, "Fast copy on write: %p -> %p\n",
4402 (void*)sstr, (void*)dstr);
4409 if (SvTHINKFIRST(dstr))
4410 sv_force_normal_flags(dstr, SV_COW_DROP_PV);
4411 else if (SvPVX_const(dstr))
4412 Safefree(SvPVX_const(dstr));
4416 SvUPGRADE(dstr, SVt_PVIV);
4418 assert (SvPOK(sstr));
4419 assert (SvPOKp(sstr));
4420 assert (!SvIOK(sstr));
4421 assert (!SvIOKp(sstr));
4422 assert (!SvNOK(sstr));
4423 assert (!SvNOKp(sstr));
4425 if (SvIsCOW(sstr)) {
4427 if (SvLEN(sstr) == 0) {
4428 /* source is a COW shared hash key. */
4429 DEBUG_C(PerlIO_printf(Perl_debug_log,
4430 "Fast copy on write: Sharing hash\n"));
4431 new_pv = HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr))));
4434 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4436 assert ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS);
4437 SvUPGRADE(sstr, SVt_PVIV);
4438 SvREADONLY_on(sstr);
4440 DEBUG_C(PerlIO_printf(Perl_debug_log,
4441 "Fast copy on write: Converting sstr to COW\n"));
4442 SV_COW_NEXT_SV_SET(dstr, sstr);
4444 SV_COW_NEXT_SV_SET(sstr, dstr);
4445 new_pv = SvPVX_mutable(sstr);
4448 SvPV_set(dstr, new_pv);
4449 SvFLAGS(dstr) = (SVt_PVIV|SVf_POK|SVp_POK|SVf_FAKE|SVf_READONLY);
4452 SvLEN_set(dstr, len);
4453 SvCUR_set(dstr, cur);
4462 =for apidoc sv_setpvn
4464 Copies a string into an SV. The C<len> parameter indicates the number of
4465 bytes to be copied. If the C<ptr> argument is NULL the SV will become
4466 undefined. Does not handle 'set' magic. See C<sv_setpvn_mg>.
4472 Perl_sv_setpvn(pTHX_ register SV *const sv, register const char *const ptr, register const STRLEN len)
4475 register char *dptr;
4477 PERL_ARGS_ASSERT_SV_SETPVN;
4479 SV_CHECK_THINKFIRST_COW_DROP(sv);
4485 /* len is STRLEN which is unsigned, need to copy to signed */
4488 Perl_croak(aTHX_ "panic: sv_setpvn called with negative strlen %"
4491 SvUPGRADE(sv, SVt_PV);
4493 dptr = SvGROW(sv, len + 1);
4494 Move(ptr,dptr,len,char);
4497 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4499 if (SvTYPE(sv) == SVt_PVCV) CvAUTOLOAD_off(sv);
4503 =for apidoc sv_setpvn_mg
4505 Like C<sv_setpvn>, but also handles 'set' magic.
4511 Perl_sv_setpvn_mg(pTHX_ register SV *const sv, register const char *const ptr, register const STRLEN len)
4513 PERL_ARGS_ASSERT_SV_SETPVN_MG;
4515 sv_setpvn(sv,ptr,len);
4520 =for apidoc sv_setpv
4522 Copies a string into an SV. The string must be null-terminated. Does not
4523 handle 'set' magic. See C<sv_setpv_mg>.
4529 Perl_sv_setpv(pTHX_ register SV *const sv, register const char *const ptr)
4532 register STRLEN len;
4534 PERL_ARGS_ASSERT_SV_SETPV;
4536 SV_CHECK_THINKFIRST_COW_DROP(sv);
4542 SvUPGRADE(sv, SVt_PV);
4544 SvGROW(sv, len + 1);
4545 Move(ptr,SvPVX(sv),len+1,char);
4547 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4549 if (SvTYPE(sv) == SVt_PVCV) CvAUTOLOAD_off(sv);
4553 =for apidoc sv_setpv_mg
4555 Like C<sv_setpv>, but also handles 'set' magic.
4561 Perl_sv_setpv_mg(pTHX_ register SV *const sv, register const char *const ptr)
4563 PERL_ARGS_ASSERT_SV_SETPV_MG;
4570 Perl_sv_sethek(pTHX_ register SV *const sv, const HEK *const hek)
4574 PERL_ARGS_ASSERT_SV_SETHEK;
4580 if (HEK_LEN(hek) == HEf_SVKEY) {
4581 sv_setsv(sv, *(SV**)HEK_KEY(hek));
4584 const int flags = HEK_FLAGS(hek);
4585 if (flags & HVhek_WASUTF8) {
4586 STRLEN utf8_len = HEK_LEN(hek);
4587 char *as_utf8 = (char *)bytes_to_utf8((U8*)HEK_KEY(hek), &utf8_len);
4588 sv_usepvn_flags(sv, as_utf8, utf8_len, SV_HAS_TRAILING_NUL);
4591 } else if (flags & (HVhek_REHASH|HVhek_UNSHARED)) {
4592 sv_setpvn(sv, HEK_KEY(hek), HEK_LEN(hek));
4595 else SvUTF8_off(sv);
4599 SV_CHECK_THINKFIRST_COW_DROP(sv);
4600 SvUPGRADE(sv, SVt_PV);
4601 SvPV_set(sv,(char *)HEK_KEY(share_hek_hek(hek)));
4602 SvCUR_set(sv, HEK_LEN(hek));
4609 else SvUTF8_off(sv);
4617 =for apidoc sv_usepvn_flags
4619 Tells an SV to use C<ptr> to find its string value. Normally the
4620 string is stored inside the SV but sv_usepvn allows the SV to use an
4621 outside string. The C<ptr> should point to memory that was allocated
4622 by C<malloc>. It must be the start of a mallocked block
4623 of memory, and not a pointer to the middle of it. The
4624 string length, C<len>, must be supplied. By default
4625 this function will realloc (i.e. move) the memory pointed to by C<ptr>,
4626 so that pointer should not be freed or used by the programmer after
4627 giving it to sv_usepvn, and neither should any pointers from "behind"
4628 that pointer (e.g. ptr + 1) be used.
4630 If C<flags> & SV_SMAGIC is true, will call SvSETMAGIC. If C<flags> &
4631 SV_HAS_TRAILING_NUL is true, then C<ptr[len]> must be NUL, and the realloc
4632 will be skipped (i.e. the buffer is actually at least 1 byte longer than
4633 C<len>, and already meets the requirements for storing in C<SvPVX>).
4639 Perl_sv_usepvn_flags(pTHX_ SV *const sv, char *ptr, const STRLEN len, const U32 flags)
4644 PERL_ARGS_ASSERT_SV_USEPVN_FLAGS;
4646 SV_CHECK_THINKFIRST_COW_DROP(sv);
4647 SvUPGRADE(sv, SVt_PV);
4650 if (flags & SV_SMAGIC)
4654 if (SvPVX_const(sv))
4658 if (flags & SV_HAS_TRAILING_NUL)
4659 assert(ptr[len] == '\0');
4662 allocate = (flags & SV_HAS_TRAILING_NUL)
4664 #ifdef Perl_safesysmalloc_size
4667 PERL_STRLEN_ROUNDUP(len + 1);
4669 if (flags & SV_HAS_TRAILING_NUL) {
4670 /* It's long enough - do nothing.
4671 Specifically Perl_newCONSTSUB is relying on this. */
4674 /* Force a move to shake out bugs in callers. */
4675 char *new_ptr = (char*)safemalloc(allocate);
4676 Copy(ptr, new_ptr, len, char);
4677 PoisonFree(ptr,len,char);
4681 ptr = (char*) saferealloc (ptr, allocate);
4684 #ifdef Perl_safesysmalloc_size
4685 SvLEN_set(sv, Perl_safesysmalloc_size(ptr));
4687 SvLEN_set(sv, allocate);
4691 if (!(flags & SV_HAS_TRAILING_NUL)) {
4694 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4696 if (flags & SV_SMAGIC)
4700 #ifdef PERL_OLD_COPY_ON_WRITE
4701 /* Need to do this *after* making the SV normal, as we need the buffer
4702 pointer to remain valid until after we've copied it. If we let go too early,
4703 another thread could invalidate it by unsharing last of the same hash key
4704 (which it can do by means other than releasing copy-on-write Svs)
4705 or by changing the other copy-on-write SVs in the loop. */
4707 S_sv_release_COW(pTHX_ register SV *sv, const char *pvx, SV *after)
4709 PERL_ARGS_ASSERT_SV_RELEASE_COW;
4711 { /* this SV was SvIsCOW_normal(sv) */
4712 /* we need to find the SV pointing to us. */
4713 SV *current = SV_COW_NEXT_SV(after);
4715 if (current == sv) {
4716 /* The SV we point to points back to us (there were only two of us
4718 Hence other SV is no longer copy on write either. */
4720 SvREADONLY_off(after);
4722 /* We need to follow the pointers around the loop. */
4724 while ((next = SV_COW_NEXT_SV(current)) != sv) {
4727 /* don't loop forever if the structure is bust, and we have
4728 a pointer into a closed loop. */
4729 assert (current != after);
4730 assert (SvPVX_const(current) == pvx);
4732 /* Make the SV before us point to the SV after us. */
4733 SV_COW_NEXT_SV_SET(current, after);
4739 =for apidoc sv_force_normal_flags
4741 Undo various types of fakery on an SV: if the PV is a shared string, make
4742 a private copy; if we're a ref, stop refing; if we're a glob, downgrade to
4743 an xpvmg; if we're a copy-on-write scalar, this is the on-write time when
4744 we do the copy, and is also used locally. If C<SV_COW_DROP_PV> is set
4745 then a copy-on-write scalar drops its PV buffer (if any) and becomes
4746 SvPOK_off rather than making a copy. (Used where this
4747 scalar is about to be set to some other value.) In addition,
4748 the C<flags> parameter gets passed to C<sv_unref_flags()>
4749 when unreffing. C<sv_force_normal> calls this function
4750 with flags set to 0.
4756 Perl_sv_force_normal_flags(pTHX_ register SV *const sv, const U32 flags)
4760 PERL_ARGS_ASSERT_SV_FORCE_NORMAL_FLAGS;
4762 #ifdef PERL_OLD_COPY_ON_WRITE
4763 if (SvREADONLY(sv)) {
4765 const char * const pvx = SvPVX_const(sv);
4766 const STRLEN len = SvLEN(sv);
4767 const STRLEN cur = SvCUR(sv);
4768 /* next COW sv in the loop. If len is 0 then this is a shared-hash
4769 key scalar, so we mustn't attempt to call SV_COW_NEXT_SV(), as
4770 we'll fail an assertion. */
4771 SV * const next = len ? SV_COW_NEXT_SV(sv) : 0;
4774 PerlIO_printf(Perl_debug_log,
4775 "Copy on write: Force normal %ld\n",
4781 /* This SV doesn't own the buffer, so need to Newx() a new one: */
4784 if (flags & SV_COW_DROP_PV) {
4785 /* OK, so we don't need to copy our buffer. */
4788 SvGROW(sv, cur + 1);
4789 Move(pvx,SvPVX(sv),cur,char);
4794 sv_release_COW(sv, pvx, next);
4796 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4802 else if (IN_PERL_RUNTIME)
4803 Perl_croak_no_modify(aTHX);
4806 if (SvREADONLY(sv)) {
4808 const char * const pvx = SvPVX_const(sv);
4809 const STRLEN len = SvCUR(sv);
4814 if (flags & SV_COW_DROP_PV) {
4815 /* OK, so we don't need to copy our buffer. */
4818 SvGROW(sv, len + 1);
4819 Move(pvx,SvPVX(sv),len,char);
4822 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4824 else if (IN_PERL_RUNTIME)
4825 Perl_croak_no_modify(aTHX);
4829 sv_unref_flags(sv, flags);
4830 else if (SvFAKE(sv) && isGV_with_GP(sv))
4831 sv_unglob(sv, flags);
4832 else if (SvFAKE(sv) && SvTYPE(sv) == SVt_REGEXP) {
4833 /* Need to downgrade the REGEXP to a simple(r) scalar. This is analogous
4834 to sv_unglob. We only need it here, so inline it. */
4835 const svtype new_type = SvMAGIC(sv) || SvSTASH(sv) ? SVt_PVMG : SVt_PV;
4836 SV *const temp = newSV_type(new_type);
4837 void *const temp_p = SvANY(sv);
4839 if (new_type == SVt_PVMG) {
4840 SvMAGIC_set(temp, SvMAGIC(sv));
4841 SvMAGIC_set(sv, NULL);
4842 SvSTASH_set(temp, SvSTASH(sv));
4843 SvSTASH_set(sv, NULL);
4845 SvCUR_set(temp, SvCUR(sv));
4846 /* Remember that SvPVX is in the head, not the body. */
4848 SvLEN_set(temp, SvLEN(sv));
4849 /* This signals "buffer is owned by someone else" in sv_clear,
4850 which is the least effort way to stop it freeing the buffer.
4852 SvLEN_set(sv, SvLEN(sv)+1);
4854 /* Their buffer is already owned by someone else. */
4855 SvPVX(sv) = savepvn(SvPVX(sv), SvCUR(sv));
4856 SvLEN_set(temp, SvCUR(sv)+1);
4859 /* Now swap the rest of the bodies. */
4861 SvFLAGS(sv) &= ~(SVf_FAKE|SVTYPEMASK);
4862 SvFLAGS(sv) |= new_type;
4863 SvANY(sv) = SvANY(temp);
4865 SvFLAGS(temp) &= ~(SVTYPEMASK);
4866 SvFLAGS(temp) |= SVt_REGEXP|SVf_FAKE;
4867 SvANY(temp) = temp_p;
4876 Efficient removal of characters from the beginning of the string buffer.
4877 SvPOK(sv) must be true and the C<ptr> must be a pointer to somewhere inside
4878 the string buffer. The C<ptr> becomes the first character of the adjusted
4879 string. Uses the "OOK hack".
4881 Beware: after this function returns, C<ptr> and SvPVX_const(sv) may no longer
4882 refer to the same chunk of data.
4884 The unfortunate similarity of this function's name to that of Perl's C<chop>
4885 operator is strictly coincidental. This function works from the left;
4886 C<chop> works from the right.
4892 Perl_sv_chop(pTHX_ register SV *const sv, register const char *const ptr)
4903 PERL_ARGS_ASSERT_SV_CHOP;
4905 if (!ptr || !SvPOKp(sv))
4907 delta = ptr - SvPVX_const(sv);
4909 /* Nothing to do. */
4912 max_delta = SvLEN(sv) ? SvLEN(sv) : SvCUR(sv);
4913 if (delta > max_delta)
4914 Perl_croak(aTHX_ "panic: sv_chop ptr=%p, start=%p, end=%p",
4915 ptr, SvPVX_const(sv), SvPVX_const(sv) + max_delta);
4916 /* SvPVX(sv) may move in SV_CHECK_THINKFIRST(sv), so don't use ptr any more */
4917 SV_CHECK_THINKFIRST(sv);
4920 if (!SvLEN(sv)) { /* make copy of shared string */
4921 const char *pvx = SvPVX_const(sv);
4922 const STRLEN len = SvCUR(sv);
4923 SvGROW(sv, len + 1);
4924 Move(pvx,SvPVX(sv),len,char);
4930 SvOOK_offset(sv, old_delta);
4932 SvLEN_set(sv, SvLEN(sv) - delta);
4933 SvCUR_set(sv, SvCUR(sv) - delta);
4934 SvPV_set(sv, SvPVX(sv) + delta);
4936 p = (U8 *)SvPVX_const(sv);
4939 /* how many bytes were evacuated? we will fill them with sentinel
4940 bytes, except for the part holding the new offset of course. */
4943 evacn += (old_delta < 0x100 ? 1 : 1 + sizeof(STRLEN));
4945 assert(evacn <= delta + old_delta);
4951 if (delta < 0x100) {
4955 p -= sizeof(STRLEN);
4956 Copy((U8*)&delta, p, sizeof(STRLEN), U8);
4960 /* Fill the preceding buffer with sentinals to verify that no-one is
4970 =for apidoc sv_catpvn
4972 Concatenates the string onto the end of the string which is in the SV. The
4973 C<len> indicates number of bytes to copy. If the SV has the UTF-8
4974 status set, then the bytes appended should be valid UTF-8.
4975 Handles 'get' magic, but not 'set' magic. See C<sv_catpvn_mg>.
4977 =for apidoc sv_catpvn_flags
4979 Concatenates the string onto the end of the string which is in the SV. The
4980 C<len> indicates number of bytes to copy. If the SV has the UTF-8
4981 status set, then the bytes appended should be valid UTF-8.
4982 If C<flags> has the C<SV_SMAGIC> bit set, will
4983 C<mg_set> on C<dsv> afterwards if appropriate.
4984 C<sv_catpvn> and C<sv_catpvn_nomg> are implemented
4985 in terms of this function.
4991 Perl_sv_catpvn_flags(pTHX_ register SV *const dsv, register const char *sstr, register const STRLEN slen, const I32 flags)
4995 const char * const dstr = SvPV_force_flags(dsv, dlen, flags);
4997 PERL_ARGS_ASSERT_SV_CATPVN_FLAGS;
4998 assert((flags & (SV_CATBYTES|SV_CATUTF8)) != (SV_CATBYTES|SV_CATUTF8));
5000 if (!(flags & SV_CATBYTES) || !SvUTF8(dsv)) {
5001 if (flags & SV_CATUTF8 && !SvUTF8(dsv)) {
5002 sv_utf8_upgrade_flags_grow(dsv, 0, slen + 1);
5005 else SvGROW(dsv, dlen + slen + 1);
5007 sstr = SvPVX_const(dsv);
5008 Move(sstr, SvPVX(dsv) + dlen, slen, char);
5009 SvCUR_set(dsv, SvCUR(dsv) + slen);
5012 /* We inline bytes_to_utf8, to avoid an extra malloc. */
5013 const char * const send = sstr + slen;
5016 /* Something this code does not account for, which I think is
5017 impossible; it would require the same pv to be treated as
5018 bytes *and* utf8, which would indicate a bug elsewhere. */
5019 assert(sstr != dstr);
5021 SvGROW(dsv, dlen + slen * 2 + 1);
5022 d = (U8 *)SvPVX(dsv) + dlen;
5024 while (sstr < send) {
5025 const UV uv = NATIVE_TO_ASCII((U8)*sstr++);
5026 if (UNI_IS_INVARIANT(uv))
5027 *d++ = (U8)UTF_TO_NATIVE(uv);
5029 *d++ = (U8)UTF8_EIGHT_BIT_HI(uv);
5030 *d++ = (U8)UTF8_EIGHT_BIT_LO(uv);
5033 SvCUR_set(dsv, d-(const U8 *)SvPVX(dsv));
5036 (void)SvPOK_only_UTF8(dsv); /* validate pointer */
5038 if (flags & SV_SMAGIC)
5043 =for apidoc sv_catsv
5045 Concatenates the string from SV C<ssv> onto the end of the string in
5046 SV C<dsv>. Modifies C<dsv> but not C<ssv>. Handles 'get' magic, but
5047 not 'set' magic. See C<sv_catsv_mg>.
5049 =for apidoc sv_catsv_flags
5051 Concatenates the string from SV C<ssv> onto the end of the string in
5052 SV C<dsv>. Modifies C<dsv> but not C<ssv>. If C<flags> has C<SV_GMAGIC>
5053 bit set, will C<mg_get> on the C<ssv>, if appropriate, before
5054 reading it. If the C<flags> contain C<SV_SMAGIC>, C<mg_set> will be
5055 called on the modified SV afterward, if appropriate. C<sv_catsv>
5056 and C<sv_catsv_nomg> are implemented in terms of this function.
5061 Perl_sv_catsv_flags(pTHX_ SV *const dsv, register SV *const ssv, const I32 flags)
5065 PERL_ARGS_ASSERT_SV_CATSV_FLAGS;
5069 const char *spv = SvPV_flags_const(ssv, slen, flags);
5071 if (SvGMAGICAL(dsv) && (flags & SV_GMAGIC))
5073 sv_catpvn_flags(dsv, spv, slen,
5074 DO_UTF8(ssv) ? SV_CATUTF8 : SV_CATBYTES);
5077 if (flags & SV_SMAGIC)
5082 =for apidoc sv_catpv
5084 Concatenates the string onto the end of the string which is in the SV.
5085 If the SV has the UTF-8 status set, then the bytes appended should be
5086 valid UTF-8. Handles 'get' magic, but not 'set' magic. See C<sv_catpv_mg>.
5091 Perl_sv_catpv(pTHX_ register SV *const sv, register const char *ptr)
5094 register STRLEN len;
5098 PERL_ARGS_ASSERT_SV_CATPV;
5102 junk = SvPV_force(sv, tlen);
5104 SvGROW(sv, tlen + len + 1);
5106 ptr = SvPVX_const(sv);
5107 Move(ptr,SvPVX(sv)+tlen,len+1,char);
5108 SvCUR_set(sv, SvCUR(sv) + len);
5109 (void)SvPOK_only_UTF8(sv); /* validate pointer */
5114 =for apidoc sv_catpv_flags
5116 Concatenates the string onto the end of the string which is in the SV.
5117 If the SV has the UTF-8 status set, then the bytes appended should
5118 be valid UTF-8. If C<flags> has the C<SV_SMAGIC> bit set, will C<mg_set>
5119 on the modified SV if appropriate.
5125 Perl_sv_catpv_flags(pTHX_ SV *dstr, const char *sstr, const I32 flags)
5127 PERL_ARGS_ASSERT_SV_CATPV_FLAGS;
5128 sv_catpvn_flags(dstr, sstr, strlen(sstr), flags);
5132 =for apidoc sv_catpv_mg
5134 Like C<sv_catpv>, but also handles 'set' magic.
5140 Perl_sv_catpv_mg(pTHX_ register SV *const sv, register const char *const ptr)
5142 PERL_ARGS_ASSERT_SV_CATPV_MG;
5151 Creates a new SV. A non-zero C<len> parameter indicates the number of
5152 bytes of preallocated string space the SV should have. An extra byte for a
5153 trailing NUL is also reserved. (SvPOK is not set for the SV even if string
5154 space is allocated.) The reference count for the new SV is set to 1.
5156 In 5.9.3, newSV() replaces the older NEWSV() API, and drops the first
5157 parameter, I<x>, a debug aid which allowed callers to identify themselves.
5158 This aid has been superseded by a new build option, PERL_MEM_LOG (see
5159 L<perlhacktips/PERL_MEM_LOG>). The older API is still there for use in XS
5160 modules supporting older perls.
5166 Perl_newSV(pTHX_ const STRLEN len)
5173 sv_upgrade(sv, SVt_PV);
5174 SvGROW(sv, len + 1);
5179 =for apidoc sv_magicext
5181 Adds magic to an SV, upgrading it if necessary. Applies the
5182 supplied vtable and returns a pointer to the magic added.
5184 Note that C<sv_magicext> will allow things that C<sv_magic> will not.
5185 In particular, you can add magic to SvREADONLY SVs, and add more than
5186 one instance of the same 'how'.
5188 If C<namlen> is greater than zero then a C<savepvn> I<copy> of C<name> is
5189 stored, if C<namlen> is zero then C<name> is stored as-is and - as another
5190 special case - if C<(name && namlen == HEf_SVKEY)> then C<name> is assumed
5191 to contain an C<SV*> and is stored as-is with its REFCNT incremented.
5193 (This is now used as a subroutine by C<sv_magic>.)
5198 Perl_sv_magicext(pTHX_ SV *const sv, SV *const obj, const int how,
5199 const MGVTBL *const vtable, const char *const name, const I32 namlen)
5204 PERL_ARGS_ASSERT_SV_MAGICEXT;
5206 SvUPGRADE(sv, SVt_PVMG);
5207 Newxz(mg, 1, MAGIC);
5208 mg->mg_moremagic = SvMAGIC(sv);
5209 SvMAGIC_set(sv, mg);
5211 /* Sometimes a magic contains a reference loop, where the sv and
5212 object refer to each other. To prevent a reference loop that
5213 would prevent such objects being freed, we look for such loops
5214 and if we find one we avoid incrementing the object refcount.
5216 Note we cannot do this to avoid self-tie loops as intervening RV must
5217 have its REFCNT incremented to keep it in existence.
5220 if (!obj || obj == sv ||
5221 how == PERL_MAGIC_arylen ||
5222 how == PERL_MAGIC_symtab ||
5223 (SvTYPE(obj) == SVt_PVGV &&
5224 (GvSV(obj) == sv || GvHV(obj) == (const HV *)sv
5225 || GvAV(obj) == (const AV *)sv || GvCV(obj) == (const CV *)sv
5226 || GvIOp(obj) == (const IO *)sv || GvFORM(obj) == (const CV *)sv)))
5231 mg->mg_obj = SvREFCNT_inc_simple(obj);
5232 mg->mg_flags |= MGf_REFCOUNTED;
5235 /* Normal self-ties simply pass a null object, and instead of
5236 using mg_obj directly, use the SvTIED_obj macro to produce a
5237 new RV as needed. For glob "self-ties", we are tieing the PVIO
5238 with an RV obj pointing to the glob containing the PVIO. In
5239 this case, to avoid a reference loop, we need to weaken the
5243 if (how == PERL_MAGIC_tiedscalar && SvTYPE(sv) == SVt_PVIO &&
5244 obj && SvROK(obj) && GvIO(SvRV(obj)) == (const IO *)sv)
5250 mg->mg_len = namlen;
5253 mg->mg_ptr = savepvn(name, namlen);
5254 else if (namlen == HEf_SVKEY) {
5255 /* Yes, this is casting away const. This is only for the case of
5256 HEf_SVKEY. I think we need to document this aberation of the
5257 constness of the API, rather than making name non-const, as
5258 that change propagating outwards a long way. */
5259 mg->mg_ptr = (char*)SvREFCNT_inc_simple_NN((SV *)name);
5261 mg->mg_ptr = (char *) name;
5263 mg->mg_virtual = (MGVTBL *) vtable;
5267 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
5272 =for apidoc sv_magic
5274 Adds magic to an SV. First upgrades C<sv> to type C<SVt_PVMG> if
5275 necessary, then adds a new magic item of type C<how> to the head of the
5278 See C<sv_magicext> (which C<sv_magic> now calls) for a description of the
5279 handling of the C<name> and C<namlen> arguments.
5281 You need to use C<sv_magicext> to add magic to SvREADONLY SVs and also
5282 to add more than one instance of the same 'how'.
5288 Perl_sv_magic(pTHX_ register SV *const sv, SV *const obj, const int how,
5289 const char *const name, const I32 namlen)
5292 const MGVTBL *vtable;
5295 unsigned int vtable_index;
5297 PERL_ARGS_ASSERT_SV_MAGIC;
5299 if (how < 0 || (unsigned)how > C_ARRAY_LENGTH(PL_magic_data)
5300 || ((flags = PL_magic_data[how]),
5301 (vtable_index = flags & PERL_MAGIC_VTABLE_MASK)
5302 > magic_vtable_max))
5303 Perl_croak(aTHX_ "Don't know how to handle magic of type \\%o", how);
5305 /* PERL_MAGIC_ext is reserved for use by extensions not perl internals.
5306 Useful for attaching extension internal data to perl vars.
5307 Note that multiple extensions may clash if magical scalars
5308 etc holding private data from one are passed to another. */
5310 vtable = (vtable_index == magic_vtable_max)
5311 ? NULL : PL_magic_vtables + vtable_index;
5313 #ifdef PERL_OLD_COPY_ON_WRITE
5315 sv_force_normal_flags(sv, 0);
5317 if (SvREADONLY(sv)) {
5319 /* its okay to attach magic to shared strings */
5323 && !PERL_MAGIC_TYPE_READONLY_ACCEPTABLE(how)
5326 Perl_croak_no_modify(aTHX);
5329 if (SvMAGICAL(sv) || (how == PERL_MAGIC_taint && SvTYPE(sv) >= SVt_PVMG)) {
5330 if (SvMAGIC(sv) && (mg = mg_find(sv, how))) {
5331 /* sv_magic() refuses to add a magic of the same 'how' as an
5334 if (how == PERL_MAGIC_taint) {
5336 /* Any scalar which already had taint magic on which someone
5337 (erroneously?) did SvIOK_on() or similar will now be
5338 incorrectly sporting public "OK" flags. */
5339 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
5345 /* Rest of work is done else where */
5346 mg = sv_magicext(sv,obj,how,vtable,name,namlen);
5349 case PERL_MAGIC_taint:
5352 case PERL_MAGIC_ext:
5353 case PERL_MAGIC_dbfile:
5360 S_sv_unmagicext_flags(pTHX_ SV *const sv, const int type, MGVTBL *vtbl, const U32 flags)
5367 if (SvTYPE(sv) < SVt_PVMG || !SvMAGIC(sv))
5369 mgp = &(((XPVMG*) SvANY(sv))->xmg_u.xmg_magic);
5370 for (mg = *mgp; mg; mg = *mgp) {
5371 const MGVTBL* const virt = mg->mg_virtual;
5372 if (mg->mg_type == type && (!flags || virt == vtbl)) {
5373 *mgp = mg->mg_moremagic;
5374 if (virt && virt->svt_free)
5375 virt->svt_free(aTHX_ sv, mg);
5376 if (mg->mg_ptr && mg->mg_type != PERL_MAGIC_regex_global) {
5378 Safefree(mg->mg_ptr);
5379 else if (mg->mg_len == HEf_SVKEY)
5380 SvREFCNT_dec(MUTABLE_SV(mg->mg_ptr));
5381 else if (mg->mg_type == PERL_MAGIC_utf8)
5382 Safefree(mg->mg_ptr);
5384 if (mg->mg_flags & MGf_REFCOUNTED)
5385 SvREFCNT_dec(mg->mg_obj);
5389 mgp = &mg->mg_moremagic;
5392 if (SvMAGICAL(sv)) /* if we're under save_magic, wait for restore_magic; */
5393 mg_magical(sv); /* else fix the flags now */
5397 SvFLAGS(sv) |= (SvFLAGS(sv) & (SVp_IOK|SVp_NOK|SVp_POK)) >> PRIVSHIFT;
5403 =for apidoc sv_unmagic
5405 Removes all magic of type C<type> from an SV.
5411 Perl_sv_unmagic(pTHX_ SV *const sv, const int type)
5413 PERL_ARGS_ASSERT_SV_UNMAGIC;
5414 return S_sv_unmagicext_flags(aTHX_ sv, type, NULL, 0);
5418 =for apidoc sv_unmagicext
5420 Removes all magic of type C<type> with the specified C<vtbl> from an SV.
5426 Perl_sv_unmagicext(pTHX_ SV *const sv, const int type, MGVTBL *vtbl)
5428 PERL_ARGS_ASSERT_SV_UNMAGICEXT;
5429 return S_sv_unmagicext_flags(aTHX_ sv, type, vtbl, 1);
5433 =for apidoc sv_rvweaken
5435 Weaken a reference: set the C<SvWEAKREF> flag on this RV; give the
5436 referred-to SV C<PERL_MAGIC_backref> magic if it hasn't already; and
5437 push a back-reference to this RV onto the array of backreferences
5438 associated with that magic. If the RV is magical, set magic will be
5439 called after the RV is cleared.
5445 Perl_sv_rvweaken(pTHX_ SV *const sv)
5449 PERL_ARGS_ASSERT_SV_RVWEAKEN;
5451 if (!SvOK(sv)) /* let undefs pass */
5454 Perl_croak(aTHX_ "Can't weaken a nonreference");
5455 else if (SvWEAKREF(sv)) {
5456 Perl_ck_warner(aTHX_ packWARN(WARN_MISC), "Reference is already weak");
5459 else if (SvREADONLY(sv)) croak_no_modify();
5461 Perl_sv_add_backref(aTHX_ tsv, sv);
5467 /* Give tsv backref magic if it hasn't already got it, then push a
5468 * back-reference to sv onto the array associated with the backref magic.
5470 * As an optimisation, if there's only one backref and it's not an AV,
5471 * store it directly in the HvAUX or mg_obj slot, avoiding the need to
5472 * allocate an AV. (Whether the slot holds an AV tells us whether this is
5476 /* A discussion about the backreferences array and its refcount:
5478 * The AV holding the backreferences is pointed to either as the mg_obj of
5479 * PERL_MAGIC_backref, or in the specific case of a HV, from the
5480 * xhv_backreferences field. The array is created with a refcount
5481 * of 2. This means that if during global destruction the array gets
5482 * picked on before its parent to have its refcount decremented by the
5483 * random zapper, it won't actually be freed, meaning it's still there for
5484 * when its parent gets freed.
5486 * When the parent SV is freed, the extra ref is killed by
5487 * Perl_sv_kill_backrefs. The other ref is killed, in the case of magic,
5488 * by mg_free() / MGf_REFCOUNTED, or for a hash, by Perl_hv_kill_backrefs.
5490 * When a single backref SV is stored directly, it is not reference
5495 Perl_sv_add_backref(pTHX_ SV *const tsv, SV *const sv)
5502 PERL_ARGS_ASSERT_SV_ADD_BACKREF;
5504 /* find slot to store array or singleton backref */
5506 if (SvTYPE(tsv) == SVt_PVHV) {
5507 svp = (SV**)Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5510 (SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL))))
5512 sv_magic(tsv, NULL, PERL_MAGIC_backref, NULL, 0);
5513 mg = mg_find(tsv, PERL_MAGIC_backref);
5515 svp = &(mg->mg_obj);
5518 /* create or retrieve the array */
5520 if ( (!*svp && SvTYPE(sv) == SVt_PVAV)
5521 || (*svp && SvTYPE(*svp) != SVt_PVAV)
5526 SvREFCNT_inc_simple_void(av);
5527 /* av now has a refcnt of 2; see discussion above */
5529 /* move single existing backref to the array */
5531 AvARRAY(av)[++AvFILLp(av)] = *svp; /* av_push() */
5535 mg->mg_flags |= MGf_REFCOUNTED;
5538 av = MUTABLE_AV(*svp);
5541 /* optimisation: store single backref directly in HvAUX or mg_obj */
5545 /* push new backref */
5546 assert(SvTYPE(av) == SVt_PVAV);
5547 if (AvFILLp(av) >= AvMAX(av)) {
5548 av_extend(av, AvFILLp(av)+1);
5550 AvARRAY(av)[++AvFILLp(av)] = sv; /* av_push() */
5553 /* delete a back-reference to ourselves from the backref magic associated
5554 * with the SV we point to.
5558 Perl_sv_del_backref(pTHX_ SV *const tsv, SV *const sv)
5563 PERL_ARGS_ASSERT_SV_DEL_BACKREF;
5565 if (SvTYPE(tsv) == SVt_PVHV) {
5567 svp = (SV**)Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5571 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
5572 svp = mg ? &(mg->mg_obj) : NULL;
5576 Perl_croak(aTHX_ "panic: del_backref, svp=0");
5578 /* It's possible that sv is being freed recursively part way through the
5579 freeing of tsv. If this happens, the backreferences array of tsv has
5580 already been freed, and so svp will be NULL. If this is the case,
5581 we should not panic. Instead, nothing needs doing, so return. */
5582 if (PL_phase == PERL_PHASE_DESTRUCT && SvREFCNT(tsv) == 0)
5584 Perl_croak(aTHX_ "panic: del_backref, *svp=%p phase=%s refcnt=%" UVuf,
5585 *svp, PL_phase_names[PL_phase], SvREFCNT(tsv));
5588 if (SvTYPE(*svp) == SVt_PVAV) {
5592 AV * const av = (AV*)*svp;
5594 assert(!SvIS_FREED(av));
5598 /* for an SV with N weak references to it, if all those
5599 * weak refs are deleted, then sv_del_backref will be called
5600 * N times and O(N^2) compares will be done within the backref
5601 * array. To ameliorate this potential slowness, we:
5602 * 1) make sure this code is as tight as possible;
5603 * 2) when looking for SV, look for it at both the head and tail of the
5604 * array first before searching the rest, since some create/destroy
5605 * patterns will cause the backrefs to be freed in order.
5612 SV **p = &svp[fill];
5613 SV *const topsv = *p;
5620 /* We weren't the last entry.
5621 An unordered list has this property that you
5622 can take the last element off the end to fill
5623 the hole, and it's still an unordered list :-)
5629 break; /* should only be one */
5636 AvFILLp(av) = fill-1;
5638 else if (SvIS_FREED(*svp) && PL_phase == PERL_PHASE_DESTRUCT) {
5639 /* freed AV; skip */
5642 /* optimisation: only a single backref, stored directly */
5644 Perl_croak(aTHX_ "panic: del_backref, *svp=%p, sv=%p", *svp, sv);
5651 Perl_sv_kill_backrefs(pTHX_ SV *const sv, AV *const av)
5657 PERL_ARGS_ASSERT_SV_KILL_BACKREFS;
5662 /* after multiple passes through Perl_sv_clean_all() for a thinngy
5663 * that has badly leaked, the backref array may have gotten freed,
5664 * since we only protect it against 1 round of cleanup */
5665 if (SvIS_FREED(av)) {
5666 if (PL_in_clean_all) /* All is fair */
5669 "panic: magic_killbackrefs (freed backref AV/SV)");
5673 is_array = (SvTYPE(av) == SVt_PVAV);
5675 assert(!SvIS_FREED(av));
5678 last = svp + AvFILLp(av);
5681 /* optimisation: only a single backref, stored directly */
5687 while (svp <= last) {
5689 SV *const referrer = *svp;
5690 if (SvWEAKREF(referrer)) {
5691 /* XXX Should we check that it hasn't changed? */
5692 assert(SvROK(referrer));
5693 SvRV_set(referrer, 0);
5695 SvWEAKREF_off(referrer);
5696 SvSETMAGIC(referrer);
5697 } else if (SvTYPE(referrer) == SVt_PVGV ||
5698 SvTYPE(referrer) == SVt_PVLV) {
5699 assert(SvTYPE(sv) == SVt_PVHV); /* stash backref */
5700 /* You lookin' at me? */
5701 assert(GvSTASH(referrer));
5702 assert(GvSTASH(referrer) == (const HV *)sv);
5703 GvSTASH(referrer) = 0;
5704 } else if (SvTYPE(referrer) == SVt_PVCV ||
5705 SvTYPE(referrer) == SVt_PVFM) {
5706 if (SvTYPE(sv) == SVt_PVHV) { /* stash backref */
5707 /* You lookin' at me? */
5708 assert(CvSTASH(referrer));
5709 assert(CvSTASH(referrer) == (const HV *)sv);
5710 SvANY(MUTABLE_CV(referrer))->xcv_stash = 0;
5713 assert(SvTYPE(sv) == SVt_PVGV);
5714 /* You lookin' at me? */
5715 assert(CvGV(referrer));
5716 assert(CvGV(referrer) == (const GV *)sv);
5717 anonymise_cv_maybe(MUTABLE_GV(sv),
5718 MUTABLE_CV(referrer));
5723 "panic: magic_killbackrefs (flags=%"UVxf")",
5724 (UV)SvFLAGS(referrer));
5735 SvREFCNT_dec(av); /* remove extra count added by sv_add_backref() */
5741 =for apidoc sv_insert
5743 Inserts a string at the specified offset/length within the SV. Similar to
5744 the Perl substr() function. Handles get magic.
5746 =for apidoc sv_insert_flags
5748 Same as C<sv_insert>, but the extra C<flags> are passed to the
5749 C<SvPV_force_flags> that applies to C<bigstr>.
5755 Perl_sv_insert_flags(pTHX_ SV *const bigstr, const STRLEN offset, const STRLEN len, const char *const little, const STRLEN littlelen, const U32 flags)
5760 register char *midend;
5761 register char *bigend;
5762 register SSize_t i; /* better be sizeof(STRLEN) or bad things happen */
5765 PERL_ARGS_ASSERT_SV_INSERT_FLAGS;
5768 Perl_croak(aTHX_ "Can't modify nonexistent substring");
5769 SvPV_force_flags(bigstr, curlen, flags);
5770 (void)SvPOK_only_UTF8(bigstr);
5771 if (offset + len > curlen) {
5772 SvGROW(bigstr, offset+len+1);
5773 Zero(SvPVX(bigstr)+curlen, offset+len-curlen, char);
5774 SvCUR_set(bigstr, offset+len);
5778 i = littlelen - len;
5779 if (i > 0) { /* string might grow */
5780 big = SvGROW(bigstr, SvCUR(bigstr) + i + 1);
5781 mid = big + offset + len;
5782 midend = bigend = big + SvCUR(bigstr);
5785 while (midend > mid) /* shove everything down */
5786 *--bigend = *--midend;
5787 Move(little,big+offset,littlelen,char);
5788 SvCUR_set(bigstr, SvCUR(bigstr) + i);
5793 Move(little,SvPVX(bigstr)+offset,len,char);
5798 big = SvPVX(bigstr);
5801 bigend = big + SvCUR(bigstr);
5803 if (midend > bigend)
5804 Perl_croak(aTHX_ "panic: sv_insert, midend=%p, bigend=%p",
5807 if (mid - big > bigend - midend) { /* faster to shorten from end */
5809 Move(little, mid, littlelen,char);
5812 i = bigend - midend;
5814 Move(midend, mid, i,char);
5818 SvCUR_set(bigstr, mid - big);
5820 else if ((i = mid - big)) { /* faster from front */
5821 midend -= littlelen;
5823 Move(big, midend - i, i, char);
5824 sv_chop(bigstr,midend-i);
5826 Move(little, mid, littlelen,char);
5828 else if (littlelen) {
5829 midend -= littlelen;
5830 sv_chop(bigstr,midend);
5831 Move(little,midend,littlelen,char);
5834 sv_chop(bigstr,midend);
5840 =for apidoc sv_replace
5842 Make the first argument a copy of the second, then delete the original.
5843 The target SV physically takes over ownership of the body of the source SV
5844 and inherits its flags; however, the target keeps any magic it owns,
5845 and any magic in the source is discarded.
5846 Note that this is a rather specialist SV copying operation; most of the
5847 time you'll want to use C<sv_setsv> or one of its many macro front-ends.
5853 Perl_sv_replace(pTHX_ register SV *const sv, register SV *const nsv)
5856 const U32 refcnt = SvREFCNT(sv);
5858 PERL_ARGS_ASSERT_SV_REPLACE;
5860 SV_CHECK_THINKFIRST_COW_DROP(sv);
5861 if (SvREFCNT(nsv) != 1) {
5862 Perl_croak(aTHX_ "panic: reference miscount on nsv in sv_replace()"
5863 " (%" UVuf " != 1)", (UV) SvREFCNT(nsv));
5865 if (SvMAGICAL(sv)) {
5869 sv_upgrade(nsv, SVt_PVMG);
5870 SvMAGIC_set(nsv, SvMAGIC(sv));
5871 SvFLAGS(nsv) |= SvMAGICAL(sv);
5873 SvMAGIC_set(sv, NULL);
5877 assert(!SvREFCNT(sv));
5878 #ifdef DEBUG_LEAKING_SCALARS
5879 sv->sv_flags = nsv->sv_flags;
5880 sv->sv_any = nsv->sv_any;
5881 sv->sv_refcnt = nsv->sv_refcnt;
5882 sv->sv_u = nsv->sv_u;
5884 StructCopy(nsv,sv,SV);
5886 if(SvTYPE(sv) == SVt_IV) {
5888 = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
5892 #ifdef PERL_OLD_COPY_ON_WRITE
5893 if (SvIsCOW_normal(nsv)) {
5894 /* We need to follow the pointers around the loop to make the
5895 previous SV point to sv, rather than nsv. */
5898 while ((next = SV_COW_NEXT_SV(current)) != nsv) {
5901 assert(SvPVX_const(current) == SvPVX_const(nsv));
5903 /* Make the SV before us point to the SV after us. */
5905 PerlIO_printf(Perl_debug_log, "previous is\n");
5907 PerlIO_printf(Perl_debug_log,
5908 "move it from 0x%"UVxf" to 0x%"UVxf"\n",
5909 (UV) SV_COW_NEXT_SV(current), (UV) sv);
5911 SV_COW_NEXT_SV_SET(current, sv);
5914 SvREFCNT(sv) = refcnt;
5915 SvFLAGS(nsv) |= SVTYPEMASK; /* Mark as freed */
5920 /* We're about to free a GV which has a CV that refers back to us.
5921 * If that CV will outlive us, make it anonymous (i.e. fix up its CvGV
5925 S_anonymise_cv_maybe(pTHX_ GV *gv, CV* cv)
5930 PERL_ARGS_ASSERT_ANONYMISE_CV_MAYBE;
5933 assert(SvREFCNT(gv) == 0);
5934 assert(isGV(gv) && isGV_with_GP(gv));
5936 assert(!CvANON(cv));
5937 assert(CvGV(cv) == gv);
5939 /* will the CV shortly be freed by gp_free() ? */
5940 if (GvCV(gv) == cv && GvGP(gv)->gp_refcnt < 2 && SvREFCNT(cv) < 2) {
5941 SvANY(cv)->xcv_gv = NULL;
5945 /* if not, anonymise: */
5946 gvname = (GvSTASH(gv) && HvNAME(GvSTASH(gv)) && HvENAME(GvSTASH(gv)))
5947 ? newSVhek(HvENAME_HEK(GvSTASH(gv)))
5948 : newSVpvn_flags( "__ANON__", 8, 0 );
5949 sv_catpvs(gvname, "::__ANON__");
5950 anongv = gv_fetchsv(gvname, GV_ADDMULTI, SVt_PVCV);
5951 SvREFCNT_dec(gvname);
5955 SvANY(cv)->xcv_gv = MUTABLE_GV(SvREFCNT_inc(anongv));
5960 =for apidoc sv_clear
5962 Clear an SV: call any destructors, free up any memory used by the body,
5963 and free the body itself. The SV's head is I<not> freed, although
5964 its type is set to all 1's so that it won't inadvertently be assumed
5965 to be live during global destruction etc.
5966 This function should only be called when REFCNT is zero. Most of the time
5967 you'll want to call C<sv_free()> (or its macro wrapper C<SvREFCNT_dec>)
5974 Perl_sv_clear(pTHX_ SV *const orig_sv)
5979 const struct body_details *sv_type_details;
5982 register SV *sv = orig_sv;
5985 PERL_ARGS_ASSERT_SV_CLEAR;
5987 /* within this loop, sv is the SV currently being freed, and
5988 * iter_sv is the most recent AV or whatever that's being iterated
5989 * over to provide more SVs */
5995 assert(SvREFCNT(sv) == 0);
5996 assert(SvTYPE(sv) != (svtype)SVTYPEMASK);
5998 if (type <= SVt_IV) {
5999 /* See the comment in sv.h about the collusion between this
6000 * early return and the overloading of the NULL slots in the
6004 SvFLAGS(sv) &= SVf_BREAK;
6005 SvFLAGS(sv) |= SVTYPEMASK;
6009 assert(!SvOBJECT(sv) || type >= SVt_PVMG); /* objs are always >= MG */
6011 if (type >= SVt_PVMG) {
6013 if (!curse(sv, 1)) goto get_next_sv;
6014 type = SvTYPE(sv); /* destructor may have changed it */
6016 /* Free back-references before magic, in case the magic calls
6017 * Perl code that has weak references to sv. */
6018 if (type == SVt_PVHV) {
6019 Perl_hv_kill_backrefs(aTHX_ MUTABLE_HV(sv));
6023 else if (type == SVt_PVMG && SvPAD_OUR(sv)) {
6024 SvREFCNT_dec(SvOURSTASH(sv));
6025 } else if (SvMAGIC(sv)) {
6026 /* Free back-references before other types of magic. */
6027 sv_unmagic(sv, PERL_MAGIC_backref);
6030 if (type == SVt_PVMG && SvPAD_TYPED(sv))
6031 SvREFCNT_dec(SvSTASH(sv));
6034 /* case SVt_BIND: */
6037 IoIFP(sv) != PerlIO_stdin() &&
6038 IoIFP(sv) != PerlIO_stdout() &&
6039 IoIFP(sv) != PerlIO_stderr() &&
6040 !(IoFLAGS(sv) & IOf_FAKE_DIRP))
6042 io_close(MUTABLE_IO(sv), FALSE);
6044 if (IoDIRP(sv) && !(IoFLAGS(sv) & IOf_FAKE_DIRP))
6045 PerlDir_close(IoDIRP(sv));
6046 IoDIRP(sv) = (DIR*)NULL;
6047 Safefree(IoTOP_NAME(sv));
6048 Safefree(IoFMT_NAME(sv));
6049 Safefree(IoBOTTOM_NAME(sv));
6050 if ((const GV *)sv == PL_statgv)
6054 /* FIXME for plugins */
6055 pregfree2((REGEXP*) sv);
6059 cv_undef(MUTABLE_CV(sv));
6060 /* If we're in a stash, we don't own a reference to it.
6061 * However it does have a back reference to us, which needs to
6063 if ((stash = CvSTASH(sv)))
6064 sv_del_backref(MUTABLE_SV(stash), sv);
6067 if (PL_last_swash_hv == (const HV *)sv) {
6068 PL_last_swash_hv = NULL;
6070 if (HvTOTALKEYS((HV*)sv) > 0) {
6072 /* this statement should match the one at the beginning of
6073 * hv_undef_flags() */
6074 if ( PL_phase != PERL_PHASE_DESTRUCT
6075 && (name = HvNAME((HV*)sv)))
6078 (void)hv_delete(PL_stashcache, name,
6079 HvNAMEUTF8((HV*)sv) ? -HvNAMELEN_get((HV*)sv) : HvNAMELEN_get((HV*)sv), G_DISCARD);
6080 hv_name_set((HV*)sv, NULL, 0, 0);
6083 /* save old iter_sv in unused SvSTASH field */
6084 assert(!SvOBJECT(sv));
6085 SvSTASH(sv) = (HV*)iter_sv;
6088 /* XXX ideally we should save the old value of hash_index
6089 * too, but I can't think of any place to hide it. The
6090 * effect of not saving it is that for freeing hashes of
6091 * hashes, we become quadratic in scanning the HvARRAY of
6092 * the top hash looking for new entries to free; but
6093 * hopefully this will be dwarfed by the freeing of all
6094 * the nested hashes. */
6096 next_sv = Perl_hfree_next_entry(aTHX_ (HV*)sv, &hash_index);
6097 goto get_next_sv; /* process this new sv */
6099 /* free empty hash */
6100 Perl_hv_undef_flags(aTHX_ MUTABLE_HV(sv), HV_NAME_SETALL);
6101 assert(!HvARRAY((HV*)sv));
6105 AV* av = MUTABLE_AV(sv);
6106 if (PL_comppad == av) {
6110 if (AvREAL(av) && AvFILLp(av) > -1) {
6111 next_sv = AvARRAY(av)[AvFILLp(av)--];
6112 /* save old iter_sv in top-most slot of AV,
6113 * and pray that it doesn't get wiped in the meantime */
6114 AvARRAY(av)[AvMAX(av)] = iter_sv;
6116 goto get_next_sv; /* process this new sv */
6118 Safefree(AvALLOC(av));
6123 if (LvTYPE(sv) == 'T') { /* for tie: return HE to pool */
6124 SvREFCNT_dec(HeKEY_sv((HE*)LvTARG(sv)));
6125 HeNEXT((HE*)LvTARG(sv)) = PL_hv_fetch_ent_mh;
6126 PL_hv_fetch_ent_mh = (HE*)LvTARG(sv);
6128 else if (LvTYPE(sv) != 't') /* unless tie: unrefcnted fake SV** */
6129 SvREFCNT_dec(LvTARG(sv));
6131 if (isGV_with_GP(sv)) {
6132 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
6133 && HvENAME_get(stash))
6134 mro_method_changed_in(stash);
6135 gp_free(MUTABLE_GV(sv));
6137 unshare_hek(GvNAME_HEK(sv));
6138 /* If we're in a stash, we don't own a reference to it.
6139 * However it does have a back reference to us, which
6140 * needs to be cleared. */
6141 if (!SvVALID(sv) && (stash = GvSTASH(sv)))
6142 sv_del_backref(MUTABLE_SV(stash), sv);
6144 /* FIXME. There are probably more unreferenced pointers to SVs
6145 * in the interpreter struct that we should check and tidy in
6146 * a similar fashion to this: */
6147 /* See also S_sv_unglob, which does the same thing. */
6148 if ((const GV *)sv == PL_last_in_gv)
6149 PL_last_in_gv = NULL;
6150 else if ((const GV *)sv == PL_statgv)
6157 /* Don't bother with SvOOK_off(sv); as we're only going to
6161 SvOOK_offset(sv, offset);
6162 SvPV_set(sv, SvPVX_mutable(sv) - offset);
6163 /* Don't even bother with turning off the OOK flag. */
6168 SV * const target = SvRV(sv);
6170 sv_del_backref(target, sv);
6175 #ifdef PERL_OLD_COPY_ON_WRITE
6176 else if (SvPVX_const(sv)
6177 && !(SvTYPE(sv) == SVt_PVIO
6178 && !(IoFLAGS(sv) & IOf_FAKE_DIRP)))
6182 PerlIO_printf(Perl_debug_log, "Copy on write: clear\n");
6186 sv_release_COW(sv, SvPVX_const(sv), SV_COW_NEXT_SV(sv));
6188 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
6192 } else if (SvLEN(sv)) {
6193 Safefree(SvPVX_const(sv));
6197 else if (SvPVX_const(sv) && SvLEN(sv)
6198 && !(SvTYPE(sv) == SVt_PVIO
6199 && !(IoFLAGS(sv) & IOf_FAKE_DIRP)))
6200 Safefree(SvPVX_mutable(sv));
6201 else if (SvPVX_const(sv) && SvIsCOW(sv)) {
6202 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
6213 SvFLAGS(sv) &= SVf_BREAK;
6214 SvFLAGS(sv) |= SVTYPEMASK;
6216 sv_type_details = bodies_by_type + type;
6217 if (sv_type_details->arena) {
6218 del_body(((char *)SvANY(sv) + sv_type_details->offset),
6219 &PL_body_roots[type]);
6221 else if (sv_type_details->body_size) {
6222 safefree(SvANY(sv));
6226 /* caller is responsible for freeing the head of the original sv */
6227 if (sv != orig_sv && !SvREFCNT(sv))
6230 /* grab and free next sv, if any */
6238 else if (!iter_sv) {
6240 } else if (SvTYPE(iter_sv) == SVt_PVAV) {
6241 AV *const av = (AV*)iter_sv;
6242 if (AvFILLp(av) > -1) {
6243 sv = AvARRAY(av)[AvFILLp(av)--];
6245 else { /* no more elements of current AV to free */
6248 /* restore previous value, squirrelled away */
6249 iter_sv = AvARRAY(av)[AvMAX(av)];
6250 Safefree(AvALLOC(av));
6253 } else if (SvTYPE(iter_sv) == SVt_PVHV) {
6254 sv = Perl_hfree_next_entry(aTHX_ (HV*)iter_sv, &hash_index);
6255 if (!sv && !HvTOTALKEYS((HV *)iter_sv)) {
6256 /* no more elements of current HV to free */
6259 /* Restore previous value of iter_sv, squirrelled away */
6260 assert(!SvOBJECT(sv));
6261 iter_sv = (SV*)SvSTASH(sv);
6263 /* ideally we should restore the old hash_index here,
6264 * but we don't currently save the old value */
6267 /* free any remaining detritus from the hash struct */
6268 Perl_hv_undef_flags(aTHX_ MUTABLE_HV(sv), HV_NAME_SETALL);
6269 assert(!HvARRAY((HV*)sv));
6274 /* unrolled SvREFCNT_dec and sv_free2 follows: */
6278 if (!SvREFCNT(sv)) {
6282 if (--(SvREFCNT(sv)))
6286 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEBUGGING),
6287 "Attempt to free temp prematurely: SV 0x%"UVxf
6288 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
6292 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
6293 /* make sure SvREFCNT(sv)==0 happens very seldom */
6294 SvREFCNT(sv) = (~(U32)0)/2;
6303 /* This routine curses the sv itself, not the object referenced by sv. So
6304 sv does not have to be ROK. */
6307 S_curse(pTHX_ SV * const sv, const bool check_refcnt) {
6310 PERL_ARGS_ASSERT_CURSE;
6311 assert(SvOBJECT(sv));
6313 if (PL_defstash && /* Still have a symbol table? */
6320 stash = SvSTASH(sv);
6321 destructor = StashHANDLER(stash,DESTROY);
6323 /* A constant subroutine can have no side effects, so
6324 don't bother calling it. */
6325 && !CvCONST(destructor)
6326 /* Don't bother calling an empty destructor or one that
6327 returns immediately. */
6328 && (CvISXSUB(destructor)
6329 || (CvSTART(destructor)
6330 && (CvSTART(destructor)->op_next->op_type
6332 && (CvSTART(destructor)->op_next->op_type
6334 || CvSTART(destructor)->op_next->op_next->op_type
6340 SV* const tmpref = newRV(sv);
6341 SvREADONLY_on(tmpref); /* DESTROY() could be naughty */
6343 PUSHSTACKi(PERLSI_DESTROY);
6348 call_sv(MUTABLE_SV(destructor),
6349 G_DISCARD|G_EVAL|G_KEEPERR|G_VOID);
6353 if(SvREFCNT(tmpref) < 2) {
6354 /* tmpref is not kept alive! */
6356 SvRV_set(tmpref, NULL);
6359 SvREFCNT_dec(tmpref);
6361 } while (SvOBJECT(sv) && SvSTASH(sv) != stash);
6364 if (check_refcnt && SvREFCNT(sv)) {
6365 if (PL_in_clean_objs)
6367 "DESTROY created new reference to dead object '%"HEKf"'",
6368 HEKfARG(HvNAME_HEK(stash)));
6369 /* DESTROY gave object new lease on life */
6375 SvREFCNT_dec(SvSTASH(sv)); /* possibly of changed persuasion */
6376 SvOBJECT_off(sv); /* Curse the object. */
6377 if (SvTYPE(sv) != SVt_PVIO)
6378 --PL_sv_objcount;/* XXX Might want something more general */
6384 =for apidoc sv_newref
6386 Increment an SV's reference count. Use the C<SvREFCNT_inc()> wrapper
6393 Perl_sv_newref(pTHX_ SV *const sv)
6395 PERL_UNUSED_CONTEXT;
6404 Decrement an SV's reference count, and if it drops to zero, call
6405 C<sv_clear> to invoke destructors and free up any memory used by
6406 the body; finally, deallocate the SV's head itself.
6407 Normally called via a wrapper macro C<SvREFCNT_dec>.
6413 Perl_sv_free(pTHX_ SV *const sv)
6418 if (SvREFCNT(sv) == 0) {
6419 if (SvFLAGS(sv) & SVf_BREAK)
6420 /* this SV's refcnt has been artificially decremented to
6421 * trigger cleanup */
6423 if (PL_in_clean_all) /* All is fair */
6425 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
6426 /* make sure SvREFCNT(sv)==0 happens very seldom */
6427 SvREFCNT(sv) = (~(U32)0)/2;
6430 if (ckWARN_d(WARN_INTERNAL)) {
6431 #ifdef DEBUG_LEAKING_SCALARS_FORK_DUMP
6432 Perl_dump_sv_child(aTHX_ sv);
6434 #ifdef DEBUG_LEAKING_SCALARS
6437 #ifdef DEBUG_LEAKING_SCALARS_ABORT
6438 if (PL_warnhook == PERL_WARNHOOK_FATAL
6439 || ckDEAD(packWARN(WARN_INTERNAL))) {
6440 /* Don't let Perl_warner cause us to escape our fate: */
6444 /* This may not return: */
6445 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
6446 "Attempt to free unreferenced scalar: SV 0x%"UVxf
6447 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
6450 #ifdef DEBUG_LEAKING_SCALARS_ABORT
6455 if (--(SvREFCNT(sv)) > 0)
6457 Perl_sv_free2(aTHX_ sv);
6461 Perl_sv_free2(pTHX_ SV *const sv)
6465 PERL_ARGS_ASSERT_SV_FREE2;
6469 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEBUGGING),
6470 "Attempt to free temp prematurely: SV 0x%"UVxf
6471 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
6475 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
6476 /* make sure SvREFCNT(sv)==0 happens very seldom */
6477 SvREFCNT(sv) = (~(U32)0)/2;
6488 Returns the length of the string in the SV. Handles magic and type
6489 coercion. See also C<SvCUR>, which gives raw access to the xpv_cur slot.
6495 Perl_sv_len(pTHX_ register SV *const sv)
6503 len = mg_length(sv);
6505 (void)SvPV_const(sv, len);
6510 =for apidoc sv_len_utf8
6512 Returns the number of characters in the string in an SV, counting wide
6513 UTF-8 bytes as a single character. Handles magic and type coercion.
6519 * The length is cached in PERL_MAGIC_utf8, in the mg_len field. Also the
6520 * mg_ptr is used, by sv_pos_u2b() and sv_pos_b2u() - see the comments below.
6521 * (Note that the mg_len is not the length of the mg_ptr field.
6522 * This allows the cache to store the character length of the string without
6523 * needing to malloc() extra storage to attach to the mg_ptr.)
6528 Perl_sv_len_utf8(pTHX_ register SV *const sv)
6534 return mg_length(sv);
6538 const U8 *s = (U8*)SvPV_const(sv, len);
6542 MAGIC *mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_utf8) : NULL;
6544 if (mg && (mg->mg_len != -1 || mg->mg_ptr)) {
6545 if (mg->mg_len != -1)
6548 /* We can use the offset cache for a headstart.
6549 The longer value is stored in the first pair. */
6550 STRLEN *cache = (STRLEN *) mg->mg_ptr;
6552 ulen = cache[0] + Perl_utf8_length(aTHX_ s + cache[1],
6556 if (PL_utf8cache < 0) {
6557 const STRLEN real = Perl_utf8_length(aTHX_ s, s + len);
6558 assert_uft8_cache_coherent("sv_len_utf8", ulen, real, sv);
6562 ulen = Perl_utf8_length(aTHX_ s, s + len);
6563 utf8_mg_len_cache_update(sv, &mg, ulen);
6567 return Perl_utf8_length(aTHX_ s, s + len);
6571 /* Walk forwards to find the byte corresponding to the passed in UTF-8
6574 S_sv_pos_u2b_forwards(const U8 *const start, const U8 *const send,
6575 STRLEN *const uoffset_p, bool *const at_end)
6577 const U8 *s = start;
6578 STRLEN uoffset = *uoffset_p;
6580 PERL_ARGS_ASSERT_SV_POS_U2B_FORWARDS;
6582 while (s < send && uoffset) {
6589 else if (s > send) {
6591 /* This is the existing behaviour. Possibly it should be a croak, as
6592 it's actually a bounds error */
6595 *uoffset_p -= uoffset;
6599 /* Given the length of the string in both bytes and UTF-8 characters, decide
6600 whether to walk forwards or backwards to find the byte corresponding to
6601 the passed in UTF-8 offset. */
6603 S_sv_pos_u2b_midway(const U8 *const start, const U8 *send,
6604 STRLEN uoffset, const STRLEN uend)
6606 STRLEN backw = uend - uoffset;
6608 PERL_ARGS_ASSERT_SV_POS_U2B_MIDWAY;
6610 if (uoffset < 2 * backw) {
6611 /* The assumption is that going forwards is twice the speed of going
6612 forward (that's where the 2 * backw comes from).
6613 (The real figure of course depends on the UTF-8 data.) */
6614 const U8 *s = start;
6616 while (s < send && uoffset--)
6626 while (UTF8_IS_CONTINUATION(*send))
6629 return send - start;
6632 /* For the string representation of the given scalar, find the byte
6633 corresponding to the passed in UTF-8 offset. uoffset0 and boffset0
6634 give another position in the string, *before* the sought offset, which
6635 (which is always true, as 0, 0 is a valid pair of positions), which should
6636 help reduce the amount of linear searching.
6637 If *mgp is non-NULL, it should point to the UTF-8 cache magic, which
6638 will be used to reduce the amount of linear searching. The cache will be
6639 created if necessary, and the found value offered to it for update. */
6641 S_sv_pos_u2b_cached(pTHX_ SV *const sv, MAGIC **const mgp, const U8 *const start,
6642 const U8 *const send, STRLEN uoffset,
6643 STRLEN uoffset0, STRLEN boffset0)
6645 STRLEN boffset = 0; /* Actually always set, but let's keep gcc happy. */
6647 bool at_end = FALSE;
6649 PERL_ARGS_ASSERT_SV_POS_U2B_CACHED;
6651 assert (uoffset >= uoffset0);
6658 && (*mgp || (SvTYPE(sv) >= SVt_PVMG &&
6659 (*mgp = mg_find(sv, PERL_MAGIC_utf8))))) {
6660 if ((*mgp)->mg_ptr) {
6661 STRLEN *cache = (STRLEN *) (*mgp)->mg_ptr;
6662 if (cache[0] == uoffset) {
6663 /* An exact match. */
6666 if (cache[2] == uoffset) {
6667 /* An exact match. */
6671 if (cache[0] < uoffset) {
6672 /* The cache already knows part of the way. */
6673 if (cache[0] > uoffset0) {
6674 /* The cache knows more than the passed in pair */
6675 uoffset0 = cache[0];
6676 boffset0 = cache[1];
6678 if ((*mgp)->mg_len != -1) {
6679 /* And we know the end too. */
6681 + sv_pos_u2b_midway(start + boffset0, send,
6683 (*mgp)->mg_len - uoffset0);
6685 uoffset -= uoffset0;
6687 + sv_pos_u2b_forwards(start + boffset0,
6688 send, &uoffset, &at_end);
6689 uoffset += uoffset0;
6692 else if (cache[2] < uoffset) {
6693 /* We're between the two cache entries. */
6694 if (cache[2] > uoffset0) {
6695 /* and the cache knows more than the passed in pair */
6696 uoffset0 = cache[2];
6697 boffset0 = cache[3];
6701 + sv_pos_u2b_midway(start + boffset0,
6704 cache[0] - uoffset0);
6707 + sv_pos_u2b_midway(start + boffset0,
6710 cache[2] - uoffset0);
6714 else if ((*mgp)->mg_len != -1) {
6715 /* If we can take advantage of a passed in offset, do so. */
6716 /* In fact, offset0 is either 0, or less than offset, so don't
6717 need to worry about the other possibility. */
6719 + sv_pos_u2b_midway(start + boffset0, send,
6721 (*mgp)->mg_len - uoffset0);
6726 if (!found || PL_utf8cache < 0) {
6727 STRLEN real_boffset;
6728 uoffset -= uoffset0;
6729 real_boffset = boffset0 + sv_pos_u2b_forwards(start + boffset0,
6730 send, &uoffset, &at_end);
6731 uoffset += uoffset0;
6733 if (found && PL_utf8cache < 0)
6734 assert_uft8_cache_coherent("sv_pos_u2b_cache", boffset,
6736 boffset = real_boffset;
6741 utf8_mg_len_cache_update(sv, mgp, uoffset);
6743 utf8_mg_pos_cache_update(sv, mgp, boffset, uoffset, send - start);
6750 =for apidoc sv_pos_u2b_flags
6752 Converts the value pointed to by offsetp from a count of UTF-8 chars from
6753 the start of the string, to a count of the equivalent number of bytes; if
6754 lenp is non-zero, it does the same to lenp, but this time starting from
6755 the offset, rather than from the start
6756 of the string. Handles type coercion.
6757 I<flags> is passed to C<SvPV_flags>, and usually should be
6758 C<SV_GMAGIC|SV_CONST_RETURN> to handle magic.
6764 * sv_pos_u2b_flags() uses, like sv_pos_b2u(), the mg_ptr of the potential
6765 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6766 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
6771 Perl_sv_pos_u2b_flags(pTHX_ SV *const sv, STRLEN uoffset, STRLEN *const lenp,
6778 PERL_ARGS_ASSERT_SV_POS_U2B_FLAGS;
6780 start = (U8*)SvPV_flags(sv, len, flags);
6782 const U8 * const send = start + len;
6784 boffset = sv_pos_u2b_cached(sv, &mg, start, send, uoffset, 0, 0);
6787 && *lenp /* don't bother doing work for 0, as its bytes equivalent
6788 is 0, and *lenp is already set to that. */) {
6789 /* Convert the relative offset to absolute. */
6790 const STRLEN uoffset2 = uoffset + *lenp;
6791 const STRLEN boffset2
6792 = sv_pos_u2b_cached(sv, &mg, start, send, uoffset2,
6793 uoffset, boffset) - boffset;
6807 =for apidoc sv_pos_u2b
6809 Converts the value pointed to by offsetp from a count of UTF-8 chars from
6810 the start of the string, to a count of the equivalent number of bytes; if
6811 lenp is non-zero, it does the same to lenp, but this time starting from
6812 the offset, rather than from the start of the string. Handles magic and
6815 Use C<sv_pos_u2b_flags> in preference, which correctly handles strings longer
6822 * sv_pos_u2b() uses, like sv_pos_b2u(), the mg_ptr of the potential
6823 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6824 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
6828 /* This function is subject to size and sign problems */
6831 Perl_sv_pos_u2b(pTHX_ register SV *const sv, I32 *const offsetp, I32 *const lenp)
6833 PERL_ARGS_ASSERT_SV_POS_U2B;
6836 STRLEN ulen = (STRLEN)*lenp;
6837 *offsetp = (I32)sv_pos_u2b_flags(sv, (STRLEN)*offsetp, &ulen,
6838 SV_GMAGIC|SV_CONST_RETURN);
6841 *offsetp = (I32)sv_pos_u2b_flags(sv, (STRLEN)*offsetp, NULL,
6842 SV_GMAGIC|SV_CONST_RETURN);
6847 S_utf8_mg_len_cache_update(pTHX_ SV *const sv, MAGIC **const mgp,
6850 PERL_ARGS_ASSERT_UTF8_MG_LEN_CACHE_UPDATE;
6854 if (!*mgp && (SvTYPE(sv) < SVt_PVMG ||
6855 !(*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
6856 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, &PL_vtbl_utf8, 0, 0);
6860 (*mgp)->mg_len = ulen;
6861 /* For now, treat "overflowed" as "still unknown". See RT #72924. */
6862 if (ulen != (STRLEN) (*mgp)->mg_len)
6863 (*mgp)->mg_len = -1;
6866 /* Create and update the UTF8 magic offset cache, with the proffered utf8/
6867 byte length pairing. The (byte) length of the total SV is passed in too,
6868 as blen, because for some (more esoteric) SVs, the call to SvPV_const()
6869 may not have updated SvCUR, so we can't rely on reading it directly.
6871 The proffered utf8/byte length pairing isn't used if the cache already has
6872 two pairs, and swapping either for the proffered pair would increase the
6873 RMS of the intervals between known byte offsets.
6875 The cache itself consists of 4 STRLEN values
6876 0: larger UTF-8 offset
6877 1: corresponding byte offset
6878 2: smaller UTF-8 offset
6879 3: corresponding byte offset
6881 Unused cache pairs have the value 0, 0.
6882 Keeping the cache "backwards" means that the invariant of
6883 cache[0] >= cache[2] is maintained even with empty slots, which means that
6884 the code that uses it doesn't need to worry if only 1 entry has actually
6885 been set to non-zero. It also makes the "position beyond the end of the
6886 cache" logic much simpler, as the first slot is always the one to start
6890 S_utf8_mg_pos_cache_update(pTHX_ SV *const sv, MAGIC **const mgp, const STRLEN byte,
6891 const STRLEN utf8, const STRLEN blen)
6895 PERL_ARGS_ASSERT_UTF8_MG_POS_CACHE_UPDATE;
6900 if (!*mgp && (SvTYPE(sv) < SVt_PVMG ||
6901 !(*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
6902 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, (MGVTBL*)&PL_vtbl_utf8, 0,
6904 (*mgp)->mg_len = -1;
6908 if (!(cache = (STRLEN *)(*mgp)->mg_ptr)) {
6909 Newxz(cache, PERL_MAGIC_UTF8_CACHESIZE * 2, STRLEN);
6910 (*mgp)->mg_ptr = (char *) cache;
6914 if (PL_utf8cache < 0 && SvPOKp(sv)) {
6915 /* SvPOKp() because it's possible that sv has string overloading, and
6916 therefore is a reference, hence SvPVX() is actually a pointer.
6917 This cures the (very real) symptoms of RT 69422, but I'm not actually
6918 sure whether we should even be caching the results of UTF-8
6919 operations on overloading, given that nothing stops overloading
6920 returning a different value every time it's called. */
6921 const U8 *start = (const U8 *) SvPVX_const(sv);
6922 const STRLEN realutf8 = utf8_length(start, start + byte);
6924 assert_uft8_cache_coherent("utf8_mg_pos_cache_update", utf8, realutf8,
6928 /* Cache is held with the later position first, to simplify the code
6929 that deals with unbounded ends. */
6931 ASSERT_UTF8_CACHE(cache);
6932 if (cache[1] == 0) {
6933 /* Cache is totally empty */
6936 } else if (cache[3] == 0) {
6937 if (byte > cache[1]) {
6938 /* New one is larger, so goes first. */
6939 cache[2] = cache[0];
6940 cache[3] = cache[1];
6948 #define THREEWAY_SQUARE(a,b,c,d) \
6949 ((float)((d) - (c))) * ((float)((d) - (c))) \
6950 + ((float)((c) - (b))) * ((float)((c) - (b))) \
6951 + ((float)((b) - (a))) * ((float)((b) - (a)))
6953 /* Cache has 2 slots in use, and we know three potential pairs.
6954 Keep the two that give the lowest RMS distance. Do the
6955 calculation in bytes simply because we always know the byte
6956 length. squareroot has the same ordering as the positive value,
6957 so don't bother with the actual square root. */
6958 const float existing = THREEWAY_SQUARE(0, cache[3], cache[1], blen);
6959 if (byte > cache[1]) {
6960 /* New position is after the existing pair of pairs. */
6961 const float keep_earlier
6962 = THREEWAY_SQUARE(0, cache[3], byte, blen);
6963 const float keep_later
6964 = THREEWAY_SQUARE(0, cache[1], byte, blen);
6966 if (keep_later < keep_earlier) {
6967 if (keep_later < existing) {
6968 cache[2] = cache[0];
6969 cache[3] = cache[1];
6975 if (keep_earlier < existing) {
6981 else if (byte > cache[3]) {
6982 /* New position is between the existing pair of pairs. */
6983 const float keep_earlier
6984 = THREEWAY_SQUARE(0, cache[3], byte, blen);
6985 const float keep_later
6986 = THREEWAY_SQUARE(0, byte, cache[1], blen);
6988 if (keep_later < keep_earlier) {
6989 if (keep_later < existing) {
6995 if (keep_earlier < existing) {
7002 /* New position is before the existing pair of pairs. */
7003 const float keep_earlier
7004 = THREEWAY_SQUARE(0, byte, cache[3], blen);
7005 const float keep_later
7006 = THREEWAY_SQUARE(0, byte, cache[1], blen);
7008 if (keep_later < keep_earlier) {
7009 if (keep_later < existing) {
7015 if (keep_earlier < existing) {
7016 cache[0] = cache[2];
7017 cache[1] = cache[3];
7024 ASSERT_UTF8_CACHE(cache);
7027 /* We already know all of the way, now we may be able to walk back. The same
7028 assumption is made as in S_sv_pos_u2b_midway(), namely that walking
7029 backward is half the speed of walking forward. */
7031 S_sv_pos_b2u_midway(pTHX_ const U8 *const s, const U8 *const target,
7032 const U8 *end, STRLEN endu)
7034 const STRLEN forw = target - s;
7035 STRLEN backw = end - target;
7037 PERL_ARGS_ASSERT_SV_POS_B2U_MIDWAY;
7039 if (forw < 2 * backw) {
7040 return utf8_length(s, target);
7043 while (end > target) {
7045 while (UTF8_IS_CONTINUATION(*end)) {
7054 =for apidoc sv_pos_b2u
7056 Converts the value pointed to by offsetp from a count of bytes from the
7057 start of the string, to a count of the equivalent number of UTF-8 chars.
7058 Handles magic and type coercion.
7064 * sv_pos_b2u() uses, like sv_pos_u2b(), the mg_ptr of the potential
7065 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
7070 Perl_sv_pos_b2u(pTHX_ register SV *const sv, I32 *const offsetp)
7073 const STRLEN byte = *offsetp;
7074 STRLEN len = 0; /* Actually always set, but let's keep gcc happy. */
7080 PERL_ARGS_ASSERT_SV_POS_B2U;
7085 s = (const U8*)SvPV_const(sv, blen);
7088 Perl_croak(aTHX_ "panic: sv_pos_b2u: bad byte offset, blen=%"UVuf
7089 ", byte=%"UVuf, (UV)blen, (UV)byte);
7095 && SvTYPE(sv) >= SVt_PVMG
7096 && (mg = mg_find(sv, PERL_MAGIC_utf8)))
7099 STRLEN * const cache = (STRLEN *) mg->mg_ptr;
7100 if (cache[1] == byte) {
7101 /* An exact match. */
7102 *offsetp = cache[0];
7105 if (cache[3] == byte) {
7106 /* An exact match. */
7107 *offsetp = cache[2];
7111 if (cache[1] < byte) {
7112 /* We already know part of the way. */
7113 if (mg->mg_len != -1) {
7114 /* Actually, we know the end too. */
7116 + S_sv_pos_b2u_midway(aTHX_ s + cache[1], send,
7117 s + blen, mg->mg_len - cache[0]);
7119 len = cache[0] + utf8_length(s + cache[1], send);
7122 else if (cache[3] < byte) {
7123 /* We're between the two cached pairs, so we do the calculation
7124 offset by the byte/utf-8 positions for the earlier pair,
7125 then add the utf-8 characters from the string start to
7127 len = S_sv_pos_b2u_midway(aTHX_ s + cache[3], send,
7128 s + cache[1], cache[0] - cache[2])
7132 else { /* cache[3] > byte */
7133 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + cache[3],
7137 ASSERT_UTF8_CACHE(cache);
7139 } else if (mg->mg_len != -1) {
7140 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + blen, mg->mg_len);
7144 if (!found || PL_utf8cache < 0) {
7145 const STRLEN real_len = utf8_length(s, send);
7147 if (found && PL_utf8cache < 0)
7148 assert_uft8_cache_coherent("sv_pos_b2u", len, real_len, sv);
7155 utf8_mg_len_cache_update(sv, &mg, len);
7157 utf8_mg_pos_cache_update(sv, &mg, byte, len, blen);
7162 S_assert_uft8_cache_coherent(pTHX_ const char *const func, STRLEN from_cache,
7163 STRLEN real, SV *const sv)
7165 PERL_ARGS_ASSERT_ASSERT_UFT8_CACHE_COHERENT;
7167 /* As this is debugging only code, save space by keeping this test here,
7168 rather than inlining it in all the callers. */
7169 if (from_cache == real)
7172 /* Need to turn the assertions off otherwise we may recurse infinitely
7173 while printing error messages. */
7174 SAVEI8(PL_utf8cache);
7176 Perl_croak(aTHX_ "panic: %s cache %"UVuf" real %"UVuf" for %"SVf,
7177 func, (UV) from_cache, (UV) real, SVfARG(sv));
7183 Returns a boolean indicating whether the strings in the two SVs are
7184 identical. Is UTF-8 and 'use bytes' aware, handles get magic, and will
7185 coerce its args to strings if necessary.
7187 =for apidoc sv_eq_flags
7189 Returns a boolean indicating whether the strings in the two SVs are
7190 identical. Is UTF-8 and 'use bytes' aware and coerces its args to strings
7191 if necessary. If the flags include SV_GMAGIC, it handles get-magic, too.
7197 Perl_sv_eq_flags(pTHX_ register SV *sv1, register SV *sv2, const U32 flags)
7205 SV* svrecode = NULL;
7212 /* if pv1 and pv2 are the same, second SvPV_const call may
7213 * invalidate pv1 (if we are handling magic), so we may need to
7215 if (sv1 == sv2 && flags & SV_GMAGIC
7216 && (SvTHINKFIRST(sv1) || SvGMAGICAL(sv1))) {
7217 pv1 = SvPV_const(sv1, cur1);
7218 sv1 = newSVpvn_flags(pv1, cur1, SVs_TEMP | SvUTF8(sv2));
7220 pv1 = SvPV_flags_const(sv1, cur1, flags);
7228 pv2 = SvPV_flags_const(sv2, cur2, flags);
7230 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
7231 /* Differing utf8ness.
7232 * Do not UTF8size the comparands as a side-effect. */
7235 svrecode = newSVpvn(pv2, cur2);
7236 sv_recode_to_utf8(svrecode, PL_encoding);
7237 pv2 = SvPV_const(svrecode, cur2);
7240 svrecode = newSVpvn(pv1, cur1);
7241 sv_recode_to_utf8(svrecode, PL_encoding);
7242 pv1 = SvPV_const(svrecode, cur1);
7244 /* Now both are in UTF-8. */
7246 SvREFCNT_dec(svrecode);
7252 /* sv1 is the UTF-8 one */
7253 return bytes_cmp_utf8((const U8*)pv2, cur2,
7254 (const U8*)pv1, cur1) == 0;
7257 /* sv2 is the UTF-8 one */
7258 return bytes_cmp_utf8((const U8*)pv1, cur1,
7259 (const U8*)pv2, cur2) == 0;
7265 eq = (pv1 == pv2) || memEQ(pv1, pv2, cur1);
7267 SvREFCNT_dec(svrecode);
7275 Compares the strings in two SVs. Returns -1, 0, or 1 indicating whether the
7276 string in C<sv1> is less than, equal to, or greater than the string in
7277 C<sv2>. Is UTF-8 and 'use bytes' aware, handles get magic, and will
7278 coerce its args to strings if necessary. See also C<sv_cmp_locale>.
7280 =for apidoc sv_cmp_flags
7282 Compares the strings in two SVs. Returns -1, 0, or 1 indicating whether the
7283 string in C<sv1> is less than, equal to, or greater than the string in
7284 C<sv2>. Is UTF-8 and 'use bytes' aware and will coerce its args to strings
7285 if necessary. If the flags include SV_GMAGIC, it handles get magic. See
7286 also C<sv_cmp_locale_flags>.
7292 Perl_sv_cmp(pTHX_ register SV *const sv1, register SV *const sv2)
7294 return sv_cmp_flags(sv1, sv2, SV_GMAGIC);
7298 Perl_sv_cmp_flags(pTHX_ register SV *const sv1, register SV *const sv2,
7303 const char *pv1, *pv2;
7306 SV *svrecode = NULL;
7313 pv1 = SvPV_flags_const(sv1, cur1, flags);
7320 pv2 = SvPV_flags_const(sv2, cur2, flags);
7322 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
7323 /* Differing utf8ness.
7324 * Do not UTF8size the comparands as a side-effect. */
7327 svrecode = newSVpvn(pv2, cur2);
7328 sv_recode_to_utf8(svrecode, PL_encoding);
7329 pv2 = SvPV_const(svrecode, cur2);
7332 const int retval = -bytes_cmp_utf8((const U8*)pv2, cur2,
7333 (const U8*)pv1, cur1);
7334 return retval ? retval < 0 ? -1 : +1 : 0;
7339 svrecode = newSVpvn(pv1, cur1);
7340 sv_recode_to_utf8(svrecode, PL_encoding);
7341 pv1 = SvPV_const(svrecode, cur1);
7344 const int retval = bytes_cmp_utf8((const U8*)pv1, cur1,
7345 (const U8*)pv2, cur2);
7346 return retval ? retval < 0 ? -1 : +1 : 0;
7352 cmp = cur2 ? -1 : 0;
7356 const I32 retval = memcmp((const void*)pv1, (const void*)pv2, cur1 < cur2 ? cur1 : cur2);
7359 cmp = retval < 0 ? -1 : 1;
7360 } else if (cur1 == cur2) {
7363 cmp = cur1 < cur2 ? -1 : 1;
7367 SvREFCNT_dec(svrecode);
7375 =for apidoc sv_cmp_locale
7377 Compares the strings in two SVs in a locale-aware manner. Is UTF-8 and
7378 'use bytes' aware, handles get magic, and will coerce its args to strings
7379 if necessary. See also C<sv_cmp>.
7381 =for apidoc sv_cmp_locale_flags
7383 Compares the strings in two SVs in a locale-aware manner. Is UTF-8 and
7384 'use bytes' aware and will coerce its args to strings if necessary. If the
7385 flags contain SV_GMAGIC, it handles get magic. See also C<sv_cmp_flags>.
7391 Perl_sv_cmp_locale(pTHX_ register SV *const sv1, register SV *const sv2)
7393 return sv_cmp_locale_flags(sv1, sv2, SV_GMAGIC);
7397 Perl_sv_cmp_locale_flags(pTHX_ register SV *const sv1, register SV *const sv2,
7401 #ifdef USE_LOCALE_COLLATE
7407 if (PL_collation_standard)
7411 pv1 = sv1 ? sv_collxfrm_flags(sv1, &len1, flags) : (char *) NULL;
7413 pv2 = sv2 ? sv_collxfrm_flags(sv2, &len2, flags) : (char *) NULL;
7415 if (!pv1 || !len1) {
7426 retval = memcmp((void*)pv1, (void*)pv2, len1 < len2 ? len1 : len2);
7429 return retval < 0 ? -1 : 1;
7432 * When the result of collation is equality, that doesn't mean
7433 * that there are no differences -- some locales exclude some
7434 * characters from consideration. So to avoid false equalities,
7435 * we use the raw string as a tiebreaker.
7441 #endif /* USE_LOCALE_COLLATE */
7443 return sv_cmp(sv1, sv2);
7447 #ifdef USE_LOCALE_COLLATE
7450 =for apidoc sv_collxfrm
7452 This calls C<sv_collxfrm_flags> with the SV_GMAGIC flag. See
7453 C<sv_collxfrm_flags>.
7455 =for apidoc sv_collxfrm_flags
7457 Add Collate Transform magic to an SV if it doesn't already have it. If the
7458 flags contain SV_GMAGIC, it handles get-magic.
7460 Any scalar variable may carry PERL_MAGIC_collxfrm magic that contains the
7461 scalar data of the variable, but transformed to such a format that a normal
7462 memory comparison can be used to compare the data according to the locale
7469 Perl_sv_collxfrm_flags(pTHX_ SV *const sv, STRLEN *const nxp, const I32 flags)
7474 PERL_ARGS_ASSERT_SV_COLLXFRM_FLAGS;
7476 mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_collxfrm) : (MAGIC *) NULL;
7477 if (!mg || !mg->mg_ptr || *(U32*)mg->mg_ptr != PL_collation_ix) {
7483 Safefree(mg->mg_ptr);
7484 s = SvPV_flags_const(sv, len, flags);
7485 if ((xf = mem_collxfrm(s, len, &xlen))) {
7487 #ifdef PERL_OLD_COPY_ON_WRITE
7489 sv_force_normal_flags(sv, 0);
7491 mg = sv_magicext(sv, 0, PERL_MAGIC_collxfrm, &PL_vtbl_collxfrm,
7505 if (mg && mg->mg_ptr) {
7507 return mg->mg_ptr + sizeof(PL_collation_ix);
7515 #endif /* USE_LOCALE_COLLATE */
7518 S_sv_gets_append_to_utf8(pTHX_ SV *const sv, PerlIO *const fp, I32 append)
7520 SV * const tsv = newSV(0);
7523 sv_gets(tsv, fp, 0);
7524 sv_utf8_upgrade_nomg(tsv);
7525 SvCUR_set(sv,append);
7528 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7532 S_sv_gets_read_record(pTHX_ SV *const sv, PerlIO *const fp, I32 append)
7535 const U32 recsize = SvUV(SvRV(PL_rs)); /* RsRECORD() guarantees > 0. */
7536 /* Grab the size of the record we're getting */
7537 char *const buffer = SvGROW(sv, (STRLEN)(recsize + append + 1)) + append;
7544 /* VMS wants read instead of fread, because fread doesn't respect */
7545 /* RMS record boundaries. This is not necessarily a good thing to be */
7546 /* doing, but we've got no other real choice - except avoid stdio
7547 as implementation - perhaps write a :vms layer ?
7549 fd = PerlIO_fileno(fp);
7551 bytesread = PerlLIO_read(fd, buffer, recsize);
7553 else /* in-memory file from PerlIO::Scalar */
7556 bytesread = PerlIO_read(fp, buffer, recsize);
7561 SvCUR_set(sv, bytesread + append);
7562 buffer[bytesread] = '\0';
7563 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7569 Get a line from the filehandle and store it into the SV, optionally
7570 appending to the currently-stored string.
7576 Perl_sv_gets(pTHX_ register SV *const sv, register PerlIO *const fp, I32 append)
7581 register STDCHAR rslast;
7582 register STDCHAR *bp;
7587 PERL_ARGS_ASSERT_SV_GETS;
7589 if (SvTHINKFIRST(sv))
7590 sv_force_normal_flags(sv, append ? 0 : SV_COW_DROP_PV);
7591 /* XXX. If you make this PVIV, then copy on write can copy scalars read
7593 However, perlbench says it's slower, because the existing swipe code
7594 is faster than copy on write.
7595 Swings and roundabouts. */
7596 SvUPGRADE(sv, SVt_PV);
7601 if (PerlIO_isutf8(fp)) {
7603 sv_utf8_upgrade_nomg(sv);
7604 sv_pos_u2b(sv,&append,0);
7606 } else if (SvUTF8(sv)) {
7607 return S_sv_gets_append_to_utf8(aTHX_ sv, fp, append);
7615 if (PerlIO_isutf8(fp))
7618 if (IN_PERL_COMPILETIME) {
7619 /* we always read code in line mode */
7623 else if (RsSNARF(PL_rs)) {
7624 /* If it is a regular disk file use size from stat() as estimate
7625 of amount we are going to read -- may result in mallocing
7626 more memory than we really need if the layers below reduce
7627 the size we read (e.g. CRLF or a gzip layer).
7630 if (!PerlLIO_fstat(PerlIO_fileno(fp), &st) && S_ISREG(st.st_mode)) {
7631 const Off_t offset = PerlIO_tell(fp);
7632 if (offset != (Off_t) -1 && st.st_size + append > offset) {
7633 (void) SvGROW(sv, (STRLEN)((st.st_size - offset) + append + 1));
7639 else if (RsRECORD(PL_rs)) {
7640 return S_sv_gets_read_record(aTHX_ sv, fp, append);
7642 else if (RsPARA(PL_rs)) {
7648 /* Get $/ i.e. PL_rs into same encoding as stream wants */
7649 if (PerlIO_isutf8(fp)) {
7650 rsptr = SvPVutf8(PL_rs, rslen);
7653 if (SvUTF8(PL_rs)) {
7654 if (!sv_utf8_downgrade(PL_rs, TRUE)) {
7655 Perl_croak(aTHX_ "Wide character in $/");
7658 rsptr = SvPV_const(PL_rs, rslen);
7662 rslast = rslen ? rsptr[rslen - 1] : '\0';
7664 if (rspara) { /* have to do this both before and after */
7665 do { /* to make sure file boundaries work right */
7668 i = PerlIO_getc(fp);
7672 PerlIO_ungetc(fp,i);
7678 /* See if we know enough about I/O mechanism to cheat it ! */
7680 /* This used to be #ifdef test - it is made run-time test for ease
7681 of abstracting out stdio interface. One call should be cheap
7682 enough here - and may even be a macro allowing compile
7686 if (PerlIO_fast_gets(fp)) {
7689 * We're going to steal some values from the stdio struct
7690 * and put EVERYTHING in the innermost loop into registers.
7692 register STDCHAR *ptr;
7696 #if defined(VMS) && defined(PERLIO_IS_STDIO)
7697 /* An ungetc()d char is handled separately from the regular
7698 * buffer, so we getc() it back out and stuff it in the buffer.
7700 i = PerlIO_getc(fp);
7701 if (i == EOF) return 0;
7702 *(--((*fp)->_ptr)) = (unsigned char) i;
7706 /* Here is some breathtakingly efficient cheating */
7708 cnt = PerlIO_get_cnt(fp); /* get count into register */
7709 /* make sure we have the room */
7710 if ((I32)(SvLEN(sv) - append) <= cnt + 1) {
7711 /* Not room for all of it
7712 if we are looking for a separator and room for some
7714 if (rslen && cnt > 80 && (I32)SvLEN(sv) > append) {
7715 /* just process what we have room for */
7716 shortbuffered = cnt - SvLEN(sv) + append + 1;
7717 cnt -= shortbuffered;
7721 /* remember that cnt can be negative */
7722 SvGROW(sv, (STRLEN)(append + (cnt <= 0 ? 2 : (cnt + 1))));
7727 bp = (STDCHAR*)SvPVX_const(sv) + append; /* move these two too to registers */
7728 ptr = (STDCHAR*)PerlIO_get_ptr(fp);
7729 DEBUG_P(PerlIO_printf(Perl_debug_log,
7730 "Screamer: entering, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7731 DEBUG_P(PerlIO_printf(Perl_debug_log,
7732 "Screamer: entering: PerlIO * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7733 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7734 PTR2UV(PerlIO_has_base(fp) ? PerlIO_get_base(fp) : 0)));
7739 while (cnt > 0) { /* this | eat */
7741 if ((*bp++ = *ptr++) == rslast) /* really | dust */
7742 goto thats_all_folks; /* screams | sed :-) */
7746 Copy(ptr, bp, cnt, char); /* this | eat */
7747 bp += cnt; /* screams | dust */
7748 ptr += cnt; /* louder | sed :-) */
7750 assert (!shortbuffered);
7751 goto cannot_be_shortbuffered;
7755 if (shortbuffered) { /* oh well, must extend */
7756 cnt = shortbuffered;
7758 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
7760 SvGROW(sv, SvLEN(sv) + append + cnt + 2);
7761 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
7765 cannot_be_shortbuffered:
7766 DEBUG_P(PerlIO_printf(Perl_debug_log,
7767 "Screamer: going to getc, ptr=%"UVuf", cnt=%ld\n",
7768 PTR2UV(ptr),(long)cnt));
7769 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* deregisterize cnt and ptr */
7771 DEBUG_Pv(PerlIO_printf(Perl_debug_log,
7772 "Screamer: pre: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7773 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7774 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7776 /* This used to call 'filbuf' in stdio form, but as that behaves like
7777 getc when cnt <= 0 we use PerlIO_getc here to avoid introducing
7778 another abstraction. */
7779 i = PerlIO_getc(fp); /* get more characters */
7781 DEBUG_Pv(PerlIO_printf(Perl_debug_log,
7782 "Screamer: post: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7783 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7784 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7786 cnt = PerlIO_get_cnt(fp);
7787 ptr = (STDCHAR*)PerlIO_get_ptr(fp); /* reregisterize cnt and ptr */
7788 DEBUG_P(PerlIO_printf(Perl_debug_log,
7789 "Screamer: after getc, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7791 if (i == EOF) /* all done for ever? */
7792 goto thats_really_all_folks;
7794 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
7796 SvGROW(sv, bpx + cnt + 2);
7797 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
7799 *bp++ = (STDCHAR)i; /* store character from PerlIO_getc */
7801 if (rslen && (STDCHAR)i == rslast) /* all done for now? */
7802 goto thats_all_folks;
7806 if ((rslen > 1 && (STRLEN)(bp - (STDCHAR*)SvPVX_const(sv)) < rslen) ||
7807 memNE((char*)bp - rslen, rsptr, rslen))
7808 goto screamer; /* go back to the fray */
7809 thats_really_all_folks:
7811 cnt += shortbuffered;
7812 DEBUG_P(PerlIO_printf(Perl_debug_log,
7813 "Screamer: quitting, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7814 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* put these back or we're in trouble */
7815 DEBUG_P(PerlIO_printf(Perl_debug_log,
7816 "Screamer: end: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7817 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7818 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7820 SvCUR_set(sv, bp - (STDCHAR*)SvPVX_const(sv)); /* set length */
7821 DEBUG_P(PerlIO_printf(Perl_debug_log,
7822 "Screamer: done, len=%ld, string=|%.*s|\n",
7823 (long)SvCUR(sv),(int)SvCUR(sv),SvPVX_const(sv)));
7827 /*The big, slow, and stupid way. */
7828 #ifdef USE_HEAP_INSTEAD_OF_STACK /* Even slower way. */
7829 STDCHAR *buf = NULL;
7830 Newx(buf, 8192, STDCHAR);
7838 register const STDCHAR * const bpe = buf + sizeof(buf);
7840 while ((i = PerlIO_getc(fp)) != EOF && (*bp++ = (STDCHAR)i) != rslast && bp < bpe)
7841 ; /* keep reading */
7845 cnt = PerlIO_read(fp,(char*)buf, sizeof(buf));
7846 /* Accommodate broken VAXC compiler, which applies U8 cast to
7847 * both args of ?: operator, causing EOF to change into 255
7850 i = (U8)buf[cnt - 1];
7856 cnt = 0; /* we do need to re-set the sv even when cnt <= 0 */
7858 sv_catpvn(sv, (char *) buf, cnt);
7860 sv_setpvn(sv, (char *) buf, cnt);
7862 if (i != EOF && /* joy */
7864 SvCUR(sv) < rslen ||
7865 memNE(SvPVX_const(sv) + SvCUR(sv) - rslen, rsptr, rslen)))
7869 * If we're reading from a TTY and we get a short read,
7870 * indicating that the user hit his EOF character, we need
7871 * to notice it now, because if we try to read from the TTY
7872 * again, the EOF condition will disappear.
7874 * The comparison of cnt to sizeof(buf) is an optimization
7875 * that prevents unnecessary calls to feof().
7879 if (!(cnt < (I32)sizeof(buf) && PerlIO_eof(fp)))
7883 #ifdef USE_HEAP_INSTEAD_OF_STACK
7888 if (rspara) { /* have to do this both before and after */
7889 while (i != EOF) { /* to make sure file boundaries work right */
7890 i = PerlIO_getc(fp);
7892 PerlIO_ungetc(fp,i);
7898 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7904 Auto-increment of the value in the SV, doing string to numeric conversion
7905 if necessary. Handles 'get' magic and operator overloading.
7911 Perl_sv_inc(pTHX_ register SV *const sv)
7920 =for apidoc sv_inc_nomg
7922 Auto-increment of the value in the SV, doing string to numeric conversion
7923 if necessary. Handles operator overloading. Skips handling 'get' magic.
7929 Perl_sv_inc_nomg(pTHX_ register SV *const sv)
7937 if (SvTHINKFIRST(sv)) {
7938 if (SvIsCOW(sv) || isGV_with_GP(sv))
7939 sv_force_normal_flags(sv, 0);
7940 if (SvREADONLY(sv)) {
7941 if (IN_PERL_RUNTIME)
7942 Perl_croak_no_modify(aTHX);
7946 if (SvAMAGIC(sv) && AMG_CALLunary(sv, inc_amg))
7948 i = PTR2IV(SvRV(sv));
7953 flags = SvFLAGS(sv);
7954 if ((flags & (SVp_NOK|SVp_IOK)) == SVp_NOK) {
7955 /* It's (privately or publicly) a float, but not tested as an
7956 integer, so test it to see. */
7958 flags = SvFLAGS(sv);
7960 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
7961 /* It's publicly an integer, or privately an integer-not-float */
7962 #ifdef PERL_PRESERVE_IVUV
7966 if (SvUVX(sv) == UV_MAX)
7967 sv_setnv(sv, UV_MAX_P1);
7969 (void)SvIOK_only_UV(sv);
7970 SvUV_set(sv, SvUVX(sv) + 1);
7972 if (SvIVX(sv) == IV_MAX)
7973 sv_setuv(sv, (UV)IV_MAX + 1);
7975 (void)SvIOK_only(sv);
7976 SvIV_set(sv, SvIVX(sv) + 1);
7981 if (flags & SVp_NOK) {
7982 const NV was = SvNVX(sv);
7983 if (NV_OVERFLOWS_INTEGERS_AT &&
7984 was >= NV_OVERFLOWS_INTEGERS_AT) {
7985 /* diag_listed_as: Lost precision when %s %f by 1 */
7986 Perl_ck_warner(aTHX_ packWARN(WARN_IMPRECISION),
7987 "Lost precision when incrementing %" NVff " by 1",
7990 (void)SvNOK_only(sv);
7991 SvNV_set(sv, was + 1.0);
7995 if (!(flags & SVp_POK) || !*SvPVX_const(sv)) {
7996 if ((flags & SVTYPEMASK) < SVt_PVIV)
7997 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV ? SVt_PVIV : SVt_IV));
7998 (void)SvIOK_only(sv);
8003 while (isALPHA(*d)) d++;
8004 while (isDIGIT(*d)) d++;
8005 if (d < SvEND(sv)) {
8006 #ifdef PERL_PRESERVE_IVUV
8007 /* Got to punt this as an integer if needs be, but we don't issue
8008 warnings. Probably ought to make the sv_iv_please() that does
8009 the conversion if possible, and silently. */
8010 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
8011 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
8012 /* Need to try really hard to see if it's an integer.
8013 9.22337203685478e+18 is an integer.
8014 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
8015 so $a="9.22337203685478e+18"; $a+0; $a++
8016 needs to be the same as $a="9.22337203685478e+18"; $a++
8023 /* sv_2iv *should* have made this an NV */
8024 if (flags & SVp_NOK) {
8025 (void)SvNOK_only(sv);
8026 SvNV_set(sv, SvNVX(sv) + 1.0);
8029 /* I don't think we can get here. Maybe I should assert this
8030 And if we do get here I suspect that sv_setnv will croak. NWC
8032 #if defined(USE_LONG_DOUBLE)
8033 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",
8034 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
8036 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_inc punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
8037 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
8040 #endif /* PERL_PRESERVE_IVUV */
8041 sv_setnv(sv,Atof(SvPVX_const(sv)) + 1.0);
8045 while (d >= SvPVX_const(sv)) {
8053 /* MKS: The original code here died if letters weren't consecutive.
8054 * at least it didn't have to worry about non-C locales. The
8055 * new code assumes that ('z'-'a')==('Z'-'A'), letters are
8056 * arranged in order (although not consecutively) and that only
8057 * [A-Za-z] are accepted by isALPHA in the C locale.
8059 if (*d != 'z' && *d != 'Z') {
8060 do { ++*d; } while (!isALPHA(*d));
8063 *(d--) -= 'z' - 'a';
8068 *(d--) -= 'z' - 'a' + 1;
8072 /* oh,oh, the number grew */
8073 SvGROW(sv, SvCUR(sv) + 2);
8074 SvCUR_set(sv, SvCUR(sv) + 1);
8075 for (d = SvPVX(sv) + SvCUR(sv); d > SvPVX_const(sv); d--)
8086 Auto-decrement of the value in the SV, doing string to numeric conversion
8087 if necessary. Handles 'get' magic and operator overloading.
8093 Perl_sv_dec(pTHX_ register SV *const sv)
8103 =for apidoc sv_dec_nomg
8105 Auto-decrement of the value in the SV, doing string to numeric conversion
8106 if necessary. Handles operator overloading. Skips handling 'get' magic.
8112 Perl_sv_dec_nomg(pTHX_ register SV *const sv)
8119 if (SvTHINKFIRST(sv)) {
8120 if (SvIsCOW(sv) || isGV_with_GP(sv))
8121 sv_force_normal_flags(sv, 0);
8122 if (SvREADONLY(sv)) {
8123 if (IN_PERL_RUNTIME)
8124 Perl_croak_no_modify(aTHX);
8128 if (SvAMAGIC(sv) && AMG_CALLunary(sv, dec_amg))
8130 i = PTR2IV(SvRV(sv));
8135 /* Unlike sv_inc we don't have to worry about string-never-numbers
8136 and keeping them magic. But we mustn't warn on punting */
8137 flags = SvFLAGS(sv);
8138 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
8139 /* It's publicly an integer, or privately an integer-not-float */
8140 #ifdef PERL_PRESERVE_IVUV
8144 if (SvUVX(sv) == 0) {
8145 (void)SvIOK_only(sv);
8149 (void)SvIOK_only_UV(sv);
8150 SvUV_set(sv, SvUVX(sv) - 1);
8153 if (SvIVX(sv) == IV_MIN) {
8154 sv_setnv(sv, (NV)IV_MIN);
8158 (void)SvIOK_only(sv);
8159 SvIV_set(sv, SvIVX(sv) - 1);
8164 if (flags & SVp_NOK) {
8167 const NV was = SvNVX(sv);
8168 if (NV_OVERFLOWS_INTEGERS_AT &&
8169 was <= -NV_OVERFLOWS_INTEGERS_AT) {
8170 /* diag_listed_as: Lost precision when %s %f by 1 */
8171 Perl_ck_warner(aTHX_ packWARN(WARN_IMPRECISION),
8172 "Lost precision when decrementing %" NVff " by 1",
8175 (void)SvNOK_only(sv);
8176 SvNV_set(sv, was - 1.0);
8180 if (!(flags & SVp_POK)) {
8181 if ((flags & SVTYPEMASK) < SVt_PVIV)
8182 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV) ? SVt_PVIV : SVt_IV);
8184 (void)SvIOK_only(sv);
8187 #ifdef PERL_PRESERVE_IVUV
8189 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
8190 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
8191 /* Need to try really hard to see if it's an integer.
8192 9.22337203685478e+18 is an integer.
8193 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
8194 so $a="9.22337203685478e+18"; $a+0; $a--
8195 needs to be the same as $a="9.22337203685478e+18"; $a--
8202 /* sv_2iv *should* have made this an NV */
8203 if (flags & SVp_NOK) {
8204 (void)SvNOK_only(sv);
8205 SvNV_set(sv, SvNVX(sv) - 1.0);
8208 /* I don't think we can get here. Maybe I should assert this
8209 And if we do get here I suspect that sv_setnv will croak. NWC
8211 #if defined(USE_LONG_DOUBLE)
8212 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",
8213 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
8215 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_dec punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
8216 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
8220 #endif /* PERL_PRESERVE_IVUV */
8221 sv_setnv(sv,Atof(SvPVX_const(sv)) - 1.0); /* punt */
8224 /* this define is used to eliminate a chunk of duplicated but shared logic
8225 * it has the suffix __SV_C to signal that it isnt API, and isnt meant to be
8226 * used anywhere but here - yves
8228 #define PUSH_EXTEND_MORTAL__SV_C(AnSv) \
8231 PL_tmps_stack[++PL_tmps_ix] = (AnSv); \
8235 =for apidoc sv_mortalcopy
8237 Creates a new SV which is a copy of the original SV (using C<sv_setsv>).
8238 The new SV is marked as mortal. It will be destroyed "soon", either by an
8239 explicit call to FREETMPS, or by an implicit call at places such as
8240 statement boundaries. See also C<sv_newmortal> and C<sv_2mortal>.
8245 /* Make a string that will exist for the duration of the expression
8246 * evaluation. Actually, it may have to last longer than that, but
8247 * hopefully we won't free it until it has been assigned to a
8248 * permanent location. */
8251 Perl_sv_mortalcopy(pTHX_ SV *const oldstr)
8257 sv_setsv(sv,oldstr);
8258 PUSH_EXTEND_MORTAL__SV_C(sv);
8264 =for apidoc sv_newmortal
8266 Creates a new null SV which is mortal. The reference count of the SV is
8267 set to 1. It will be destroyed "soon", either by an explicit call to
8268 FREETMPS, or by an implicit call at places such as statement boundaries.
8269 See also C<sv_mortalcopy> and C<sv_2mortal>.
8275 Perl_sv_newmortal(pTHX)
8281 SvFLAGS(sv) = SVs_TEMP;
8282 PUSH_EXTEND_MORTAL__SV_C(sv);
8288 =for apidoc newSVpvn_flags
8290 Creates a new SV and copies a string into it. The reference count for the
8291 SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
8292 string. You are responsible for ensuring that the source string is at least
8293 C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
8294 Currently the only flag bits accepted are C<SVf_UTF8> and C<SVs_TEMP>.
8295 If C<SVs_TEMP> is set, then C<sv_2mortal()> is called on the result before
8296 returning. If C<SVf_UTF8> is set, C<s>
8297 is considered to be in UTF-8 and the
8298 C<SVf_UTF8> flag will be set on the new SV.
8299 C<newSVpvn_utf8()> is a convenience wrapper for this function, defined as
8301 #define newSVpvn_utf8(s, len, u) \
8302 newSVpvn_flags((s), (len), (u) ? SVf_UTF8 : 0)
8308 Perl_newSVpvn_flags(pTHX_ const char *const s, const STRLEN len, const U32 flags)
8313 /* All the flags we don't support must be zero.
8314 And we're new code so I'm going to assert this from the start. */
8315 assert(!(flags & ~(SVf_UTF8|SVs_TEMP)));
8317 sv_setpvn(sv,s,len);
8319 /* This code used to a sv_2mortal(), however we now unroll the call to sv_2mortal()
8320 * and do what it does ourselves here.
8321 * Since we have asserted that flags can only have the SVf_UTF8 and/or SVs_TEMP flags
8322 * set above we can use it to enable the sv flags directly (bypassing SvTEMP_on), which
8323 * in turn means we dont need to mask out the SVf_UTF8 flag below, which means that we
8324 * eliminate quite a few steps than it looks - Yves (explaining patch by gfx)
8327 SvFLAGS(sv) |= flags;
8329 if(flags & SVs_TEMP){
8330 PUSH_EXTEND_MORTAL__SV_C(sv);
8337 =for apidoc sv_2mortal
8339 Marks an existing SV as mortal. The SV will be destroyed "soon", either
8340 by an explicit call to FREETMPS, or by an implicit call at places such as
8341 statement boundaries. SvTEMP() is turned on which means that the SV's
8342 string buffer can be "stolen" if this SV is copied. See also C<sv_newmortal>
8343 and C<sv_mortalcopy>.
8349 Perl_sv_2mortal(pTHX_ register SV *const sv)
8354 if (SvREADONLY(sv) && SvIMMORTAL(sv))
8356 PUSH_EXTEND_MORTAL__SV_C(sv);
8364 Creates a new SV and copies a string into it. The reference count for the
8365 SV is set to 1. If C<len> is zero, Perl will compute the length using
8366 strlen(). For efficiency, consider using C<newSVpvn> instead.
8372 Perl_newSVpv(pTHX_ const char *const s, const STRLEN len)
8378 sv_setpvn(sv, s, len || s == NULL ? len : strlen(s));
8383 =for apidoc newSVpvn
8385 Creates a new SV and copies a string into it. The reference count for the
8386 SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
8387 string. You are responsible for ensuring that the source string is at least
8388 C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
8394 Perl_newSVpvn(pTHX_ const char *const s, const STRLEN len)
8400 sv_setpvn(sv,s,len);
8405 =for apidoc newSVhek
8407 Creates a new SV from the hash key structure. It will generate scalars that
8408 point to the shared string table where possible. Returns a new (undefined)
8409 SV if the hek is NULL.
8415 Perl_newSVhek(pTHX_ const HEK *const hek)
8425 if (HEK_LEN(hek) == HEf_SVKEY) {
8426 return newSVsv(*(SV**)HEK_KEY(hek));
8428 const int flags = HEK_FLAGS(hek);
8429 if (flags & HVhek_WASUTF8) {
8431 Andreas would like keys he put in as utf8 to come back as utf8
8433 STRLEN utf8_len = HEK_LEN(hek);
8434 SV * const sv = newSV_type(SVt_PV);
8435 char *as_utf8 = (char *)bytes_to_utf8 ((U8*)HEK_KEY(hek), &utf8_len);
8436 /* bytes_to_utf8() allocates a new string, which we can repurpose: */
8437 sv_usepvn_flags(sv, as_utf8, utf8_len, SV_HAS_TRAILING_NUL);
8440 } else if (flags & (HVhek_REHASH|HVhek_UNSHARED)) {
8441 /* We don't have a pointer to the hv, so we have to replicate the
8442 flag into every HEK. This hv is using custom a hasing
8443 algorithm. Hence we can't return a shared string scalar, as
8444 that would contain the (wrong) hash value, and might get passed
8445 into an hv routine with a regular hash.
8446 Similarly, a hash that isn't using shared hash keys has to have
8447 the flag in every key so that we know not to try to call
8448 share_hek_hek on it. */
8450 SV * const sv = newSVpvn (HEK_KEY(hek), HEK_LEN(hek));
8455 /* This will be overwhelminly the most common case. */
8457 /* Inline most of newSVpvn_share(), because share_hek_hek() is far
8458 more efficient than sharepvn(). */
8462 sv_upgrade(sv, SVt_PV);
8463 SvPV_set(sv, (char *)HEK_KEY(share_hek_hek(hek)));
8464 SvCUR_set(sv, HEK_LEN(hek));
8477 =for apidoc newSVpvn_share
8479 Creates a new SV with its SvPVX_const pointing to a shared string in the string
8480 table. If the string does not already exist in the table, it is
8481 created first. Turns on READONLY and FAKE. If the C<hash> parameter
8482 is non-zero, that value is used; otherwise the hash is computed.
8483 The string's hash can later be retrieved from the SV
8484 with the C<SvSHARED_HASH()> macro. The idea here is
8485 that as the string table is used for shared hash keys these strings will have
8486 SvPVX_const == HeKEY and hash lookup will avoid string compare.
8492 Perl_newSVpvn_share(pTHX_ const char *src, I32 len, U32 hash)
8496 bool is_utf8 = FALSE;
8497 const char *const orig_src = src;
8500 STRLEN tmplen = -len;
8502 /* See the note in hv.c:hv_fetch() --jhi */
8503 src = (char*)bytes_from_utf8((const U8*)src, &tmplen, &is_utf8);
8507 PERL_HASH(hash, src, len);
8509 /* The logic for this is inlined in S_mro_get_linear_isa_dfs(), so if it
8510 changes here, update it there too. */
8511 sv_upgrade(sv, SVt_PV);
8512 SvPV_set(sv, sharepvn(src, is_utf8?-len:len, hash));
8520 if (src != orig_src)
8526 =for apidoc newSVpv_share
8528 Like C<newSVpvn_share>, but takes a nul-terminated string instead of a
8535 Perl_newSVpv_share(pTHX_ const char *src, U32 hash)
8537 return newSVpvn_share(src, strlen(src), hash);
8540 #if defined(PERL_IMPLICIT_CONTEXT)
8542 /* pTHX_ magic can't cope with varargs, so this is a no-context
8543 * version of the main function, (which may itself be aliased to us).
8544 * Don't access this version directly.
8548 Perl_newSVpvf_nocontext(const char *const pat, ...)
8554 PERL_ARGS_ASSERT_NEWSVPVF_NOCONTEXT;
8556 va_start(args, pat);
8557 sv = vnewSVpvf(pat, &args);
8564 =for apidoc newSVpvf
8566 Creates a new SV and initializes it with the string formatted like
8573 Perl_newSVpvf(pTHX_ const char *const pat, ...)
8578 PERL_ARGS_ASSERT_NEWSVPVF;
8580 va_start(args, pat);
8581 sv = vnewSVpvf(pat, &args);
8586 /* backend for newSVpvf() and newSVpvf_nocontext() */
8589 Perl_vnewSVpvf(pTHX_ const char *const pat, va_list *const args)
8594 PERL_ARGS_ASSERT_VNEWSVPVF;
8597 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
8604 Creates a new SV and copies a floating point value into it.
8605 The reference count for the SV is set to 1.
8611 Perl_newSVnv(pTHX_ const NV n)
8624 Creates a new SV and copies an integer into it. The reference count for the
8631 Perl_newSViv(pTHX_ const IV i)
8644 Creates a new SV and copies an unsigned integer into it.
8645 The reference count for the SV is set to 1.
8651 Perl_newSVuv(pTHX_ const UV u)
8662 =for apidoc newSV_type
8664 Creates a new SV, of the type specified. The reference count for the new SV
8671 Perl_newSV_type(pTHX_ const svtype type)
8676 sv_upgrade(sv, type);
8681 =for apidoc newRV_noinc
8683 Creates an RV wrapper for an SV. The reference count for the original
8684 SV is B<not> incremented.
8690 Perl_newRV_noinc(pTHX_ SV *const tmpRef)
8693 register SV *sv = newSV_type(SVt_IV);
8695 PERL_ARGS_ASSERT_NEWRV_NOINC;
8698 SvRV_set(sv, tmpRef);
8703 /* newRV_inc is the official function name to use now.
8704 * newRV_inc is in fact #defined to newRV in sv.h
8708 Perl_newRV(pTHX_ SV *const sv)
8712 PERL_ARGS_ASSERT_NEWRV;
8714 return newRV_noinc(SvREFCNT_inc_simple_NN(sv));
8720 Creates a new SV which is an exact duplicate of the original SV.
8727 Perl_newSVsv(pTHX_ register SV *const old)
8734 if (SvTYPE(old) == (svtype)SVTYPEMASK) {
8735 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL), "semi-panic: attempt to dup freed string");
8739 /* SV_GMAGIC is the default for sv_setv()
8740 SV_NOSTEAL prevents TEMP buffers being, well, stolen, and saves games
8741 with SvTEMP_off and SvTEMP_on round a call to sv_setsv. */
8742 sv_setsv_flags(sv, old, SV_GMAGIC | SV_NOSTEAL);
8747 =for apidoc sv_reset
8749 Underlying implementation for the C<reset> Perl function.
8750 Note that the perl-level function is vaguely deprecated.
8756 Perl_sv_reset(pTHX_ register const char *s, HV *const stash)
8759 char todo[PERL_UCHAR_MAX+1];
8761 PERL_ARGS_ASSERT_SV_RESET;
8766 if (!*s) { /* reset ?? searches */
8767 MAGIC * const mg = mg_find((const SV *)stash, PERL_MAGIC_symtab);
8769 const U32 count = mg->mg_len / sizeof(PMOP**);
8770 PMOP **pmp = (PMOP**) mg->mg_ptr;
8771 PMOP *const *const end = pmp + count;
8775 SvREADONLY_off(PL_regex_pad[(*pmp)->op_pmoffset]);
8777 (*pmp)->op_pmflags &= ~PMf_USED;
8785 /* reset variables */
8787 if (!HvARRAY(stash))
8790 Zero(todo, 256, char);
8793 I32 i = (unsigned char)*s;
8797 max = (unsigned char)*s++;
8798 for ( ; i <= max; i++) {
8801 for (i = 0; i <= (I32) HvMAX(stash); i++) {
8803 for (entry = HvARRAY(stash)[i];
8805 entry = HeNEXT(entry))
8810 if (!todo[(U8)*HeKEY(entry)])
8812 gv = MUTABLE_GV(HeVAL(entry));
8815 if (SvTHINKFIRST(sv)) {
8816 if (!SvREADONLY(sv) && SvROK(sv))
8818 /* XXX Is this continue a bug? Why should THINKFIRST
8819 exempt us from resetting arrays and hashes? */
8823 if (SvTYPE(sv) >= SVt_PV) {
8825 if (SvPVX_const(sv) != NULL)
8833 if (GvHV(gv) && !HvNAME_get(GvHV(gv))) {
8835 Perl_die(aTHX_ "Can't reset %%ENV on this system");
8838 # if defined(USE_ENVIRON_ARRAY)
8841 # endif /* USE_ENVIRON_ARRAY */
8852 Using various gambits, try to get an IO from an SV: the IO slot if its a
8853 GV; or the recursive result if we're an RV; or the IO slot of the symbol
8854 named after the PV if we're a string.
8856 'Get' magic is ignored on the sv passed in, but will be called on
8857 C<SvRV(sv)> if sv is an RV.
8863 Perl_sv_2io(pTHX_ SV *const sv)
8868 PERL_ARGS_ASSERT_SV_2IO;
8870 switch (SvTYPE(sv)) {
8872 io = MUTABLE_IO(sv);
8876 if (isGV_with_GP(sv)) {
8877 gv = MUTABLE_GV(sv);
8880 Perl_croak(aTHX_ "Bad filehandle: %"HEKf,
8881 HEKfARG(GvNAME_HEK(gv)));
8887 Perl_croak(aTHX_ PL_no_usym, "filehandle");
8889 SvGETMAGIC(SvRV(sv));
8890 return sv_2io(SvRV(sv));
8892 gv = gv_fetchsv_nomg(sv, 0, SVt_PVIO);
8899 if (SvGMAGICAL(sv)) {
8900 newsv = sv_newmortal();
8901 sv_setsv_nomg(newsv, sv);
8903 Perl_croak(aTHX_ "Bad filehandle: %"SVf, SVfARG(newsv));
8913 Using various gambits, try to get a CV from an SV; in addition, try if
8914 possible to set C<*st> and C<*gvp> to the stash and GV associated with it.
8915 The flags in C<lref> are passed to gv_fetchsv.
8921 Perl_sv_2cv(pTHX_ SV *sv, HV **const st, GV **const gvp, const I32 lref)
8927 PERL_ARGS_ASSERT_SV_2CV;
8934 switch (SvTYPE(sv)) {
8938 return MUTABLE_CV(sv);
8948 sv = amagic_deref_call(sv, to_cv_amg);
8951 if (SvTYPE(sv) == SVt_PVCV) {
8952 cv = MUTABLE_CV(sv);
8957 else if(SvGETMAGIC(sv), isGV_with_GP(sv))
8958 gv = MUTABLE_GV(sv);
8960 Perl_croak(aTHX_ "Not a subroutine reference");
8962 else if (isGV_with_GP(sv)) {
8963 gv = MUTABLE_GV(sv);
8966 gv = gv_fetchsv_nomg(sv, lref, SVt_PVCV);
8973 /* Some flags to gv_fetchsv mean don't really create the GV */
8974 if (!isGV_with_GP(gv)) {
8979 if (lref & ~GV_ADDMG && !GvCVu(gv)) {
8983 gv_efullname3(tmpsv, gv, NULL);
8984 /* XXX this is probably not what they think they're getting.
8985 * It has the same effect as "sub name;", i.e. just a forward
8987 newSUB(start_subparse(FALSE, 0),
8988 newSVOP(OP_CONST, 0, tmpsv),
8992 Perl_croak(aTHX_ "Unable to create sub named \"%"SVf"\"",
8993 SVfARG(SvOK(sv) ? sv : &PL_sv_no));
9002 Returns true if the SV has a true value by Perl's rules.
9003 Use the C<SvTRUE> macro instead, which may call C<sv_true()> or may
9004 instead use an in-line version.
9010 Perl_sv_true(pTHX_ register SV *const sv)
9015 register const XPV* const tXpv = (XPV*)SvANY(sv);
9017 (tXpv->xpv_cur > 1 ||
9018 (tXpv->xpv_cur && *sv->sv_u.svu_pv != '0')))
9025 return SvIVX(sv) != 0;
9028 return SvNVX(sv) != 0.0;
9030 return sv_2bool(sv);
9036 =for apidoc sv_pvn_force
9038 Get a sensible string out of the SV somehow.
9039 A private implementation of the C<SvPV_force> macro for compilers which
9040 can't cope with complex macro expressions. Always use the macro instead.
9042 =for apidoc sv_pvn_force_flags
9044 Get a sensible string out of the SV somehow.
9045 If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<sv> if
9046 appropriate, else not. C<sv_pvn_force> and C<sv_pvn_force_nomg> are
9047 implemented in terms of this function.
9048 You normally want to use the various wrapper macros instead: see
9049 C<SvPV_force> and C<SvPV_force_nomg>
9055 Perl_sv_pvn_force_flags(pTHX_ SV *const sv, STRLEN *const lp, const I32 flags)
9059 PERL_ARGS_ASSERT_SV_PVN_FORCE_FLAGS;
9061 if (flags & SV_GMAGIC) SvGETMAGIC(sv);
9062 if (SvTHINKFIRST(sv) && !SvROK(sv))
9063 sv_force_normal_flags(sv, 0);
9073 if (SvREADONLY(sv) && !(flags & SV_MUTABLE_RETURN)) {
9074 const char * const ref = sv_reftype(sv,0);
9076 Perl_croak(aTHX_ "Can't coerce readonly %s to string in %s",
9077 ref, OP_DESC(PL_op));
9079 Perl_croak(aTHX_ "Can't coerce readonly %s to string", ref);
9081 if ((SvTYPE(sv) > SVt_PVLV && SvTYPE(sv) != SVt_PVFM)
9082 || isGV_with_GP(sv))
9083 /* diag_listed_as: Can't coerce %s to %s in %s */
9084 Perl_croak(aTHX_ "Can't coerce %s to string in %s", sv_reftype(sv,0),
9086 s = sv_2pv_flags(sv, &len, flags &~ SV_GMAGIC);
9090 if (s != SvPVX_const(sv)) { /* Almost, but not quite, sv_setpvn() */
9093 SvUPGRADE(sv, SVt_PV); /* Never FALSE */
9094 SvGROW(sv, len + 1);
9095 Move(s,SvPVX(sv),len,char);
9097 SvPVX(sv)[len] = '\0';
9100 SvPOK_on(sv); /* validate pointer */
9102 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
9103 PTR2UV(sv),SvPVX_const(sv)));
9106 return SvPVX_mutable(sv);
9110 =for apidoc sv_pvbyten_force
9112 The backend for the C<SvPVbytex_force> macro. Always use the macro
9119 Perl_sv_pvbyten_force(pTHX_ SV *const sv, STRLEN *const lp)
9121 PERL_ARGS_ASSERT_SV_PVBYTEN_FORCE;
9123 sv_pvn_force(sv,lp);
9124 sv_utf8_downgrade(sv,0);
9130 =for apidoc sv_pvutf8n_force
9132 The backend for the C<SvPVutf8x_force> macro. Always use the macro
9139 Perl_sv_pvutf8n_force(pTHX_ SV *const sv, STRLEN *const lp)
9141 PERL_ARGS_ASSERT_SV_PVUTF8N_FORCE;
9143 sv_pvn_force(sv,lp);
9144 sv_utf8_upgrade(sv);
9150 =for apidoc sv_reftype
9152 Returns a string describing what the SV is a reference to.
9158 Perl_sv_reftype(pTHX_ const SV *const sv, const int ob)
9160 PERL_ARGS_ASSERT_SV_REFTYPE;
9161 if (ob && SvOBJECT(sv)) {
9162 return SvPV_nolen_const(sv_ref(NULL, sv, ob));
9165 switch (SvTYPE(sv)) {
9180 case SVt_PVLV: return (char *) (SvROK(sv) ? "REF"
9181 /* tied lvalues should appear to be
9182 * scalars for backwards compatibility */
9183 : (LvTYPE(sv) == 't' || LvTYPE(sv) == 'T')
9184 ? "SCALAR" : "LVALUE");
9185 case SVt_PVAV: return "ARRAY";
9186 case SVt_PVHV: return "HASH";
9187 case SVt_PVCV: return "CODE";
9188 case SVt_PVGV: return (char *) (isGV_with_GP(sv)
9189 ? "GLOB" : "SCALAR");
9190 case SVt_PVFM: return "FORMAT";
9191 case SVt_PVIO: return "IO";
9192 case SVt_BIND: return "BIND";
9193 case SVt_REGEXP: return "REGEXP";
9194 default: return "UNKNOWN";
9202 Returns a SV describing what the SV passed in is a reference to.
9208 Perl_sv_ref(pTHX_ register SV *dst, const SV *const sv, const int ob)
9210 PERL_ARGS_ASSERT_SV_REF;
9213 dst = sv_newmortal();
9215 if (ob && SvOBJECT(sv)) {
9216 HvNAME_get(SvSTASH(sv))
9217 ? sv_sethek(dst, HvNAME_HEK(SvSTASH(sv)))
9218 : sv_setpvn(dst, "__ANON__", 8);
9221 const char * reftype = sv_reftype(sv, 0);
9222 sv_setpv(dst, reftype);
9228 =for apidoc sv_isobject
9230 Returns a boolean indicating whether the SV is an RV pointing to a blessed
9231 object. If the SV is not an RV, or if the object is not blessed, then this
9238 Perl_sv_isobject(pTHX_ SV *sv)
9254 Returns a boolean indicating whether the SV is blessed into the specified
9255 class. This does not check for subtypes; use C<sv_derived_from> to verify
9256 an inheritance relationship.
9262 Perl_sv_isa(pTHX_ SV *sv, const char *const name)
9266 PERL_ARGS_ASSERT_SV_ISA;
9276 hvname = HvNAME_get(SvSTASH(sv));
9280 return strEQ(hvname, name);
9286 Creates a new SV for the RV, C<rv>, to point to. If C<rv> is not an RV then
9287 it will be upgraded to one. If C<classname> is non-null then the new SV will
9288 be blessed in the specified package. The new SV is returned and its
9289 reference count is 1.
9295 Perl_newSVrv(pTHX_ SV *const rv, const char *const classname)
9300 PERL_ARGS_ASSERT_NEWSVRV;
9304 SV_CHECK_THINKFIRST_COW_DROP(rv);
9305 (void)SvAMAGIC_off(rv);
9307 if (SvTYPE(rv) >= SVt_PVMG) {
9308 const U32 refcnt = SvREFCNT(rv);
9312 SvREFCNT(rv) = refcnt;
9314 sv_upgrade(rv, SVt_IV);
9315 } else if (SvROK(rv)) {
9316 SvREFCNT_dec(SvRV(rv));
9318 prepare_SV_for_RV(rv);
9326 HV* const stash = gv_stashpv(classname, GV_ADD);
9327 (void)sv_bless(rv, stash);
9333 =for apidoc sv_setref_pv
9335 Copies a pointer into a new SV, optionally blessing the SV. The C<rv>
9336 argument will be upgraded to an RV. That RV will be modified to point to
9337 the new SV. If the C<pv> argument is NULL then C<PL_sv_undef> will be placed
9338 into the SV. The C<classname> argument indicates the package for the
9339 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9340 will have a reference count of 1, and the RV will be returned.
9342 Do not use with other Perl types such as HV, AV, SV, CV, because those
9343 objects will become corrupted by the pointer copy process.
9345 Note that C<sv_setref_pvn> copies the string while this copies the pointer.
9351 Perl_sv_setref_pv(pTHX_ SV *const rv, const char *const classname, void *const pv)
9355 PERL_ARGS_ASSERT_SV_SETREF_PV;
9358 sv_setsv(rv, &PL_sv_undef);
9362 sv_setiv(newSVrv(rv,classname), PTR2IV(pv));
9367 =for apidoc sv_setref_iv
9369 Copies an integer into a new SV, optionally blessing the SV. The C<rv>
9370 argument will be upgraded to an RV. That RV will be modified to point to
9371 the new SV. The C<classname> argument indicates the package for the
9372 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9373 will have a reference count of 1, and the RV will be returned.
9379 Perl_sv_setref_iv(pTHX_ SV *const rv, const char *const classname, const IV iv)
9381 PERL_ARGS_ASSERT_SV_SETREF_IV;
9383 sv_setiv(newSVrv(rv,classname), iv);
9388 =for apidoc sv_setref_uv
9390 Copies an unsigned integer into a new SV, optionally blessing the SV. The C<rv>
9391 argument will be upgraded to an RV. That RV will be modified to point to
9392 the new SV. The C<classname> argument indicates the package for the
9393 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9394 will have a reference count of 1, and the RV will be returned.
9400 Perl_sv_setref_uv(pTHX_ SV *const rv, const char *const classname, const UV uv)
9402 PERL_ARGS_ASSERT_SV_SETREF_UV;
9404 sv_setuv(newSVrv(rv,classname), uv);
9409 =for apidoc sv_setref_nv
9411 Copies a double into a new SV, optionally blessing the SV. The C<rv>
9412 argument will be upgraded to an RV. That RV will be modified to point to
9413 the new SV. The C<classname> argument indicates the package for the
9414 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9415 will have a reference count of 1, and the RV will be returned.
9421 Perl_sv_setref_nv(pTHX_ SV *const rv, const char *const classname, const NV nv)
9423 PERL_ARGS_ASSERT_SV_SETREF_NV;
9425 sv_setnv(newSVrv(rv,classname), nv);
9430 =for apidoc sv_setref_pvn
9432 Copies a string into a new SV, optionally blessing the SV. The length of the
9433 string must be specified with C<n>. The C<rv> argument will be upgraded to
9434 an RV. That RV will be modified to point to the new SV. The C<classname>
9435 argument indicates the package for the blessing. Set C<classname> to
9436 C<NULL> to avoid the blessing. The new SV will have a reference count
9437 of 1, and the RV will be returned.
9439 Note that C<sv_setref_pv> copies the pointer while this copies the string.
9445 Perl_sv_setref_pvn(pTHX_ SV *const rv, const char *const classname,
9446 const char *const pv, const STRLEN n)
9448 PERL_ARGS_ASSERT_SV_SETREF_PVN;
9450 sv_setpvn(newSVrv(rv,classname), pv, n);
9455 =for apidoc sv_bless
9457 Blesses an SV into a specified package. The SV must be an RV. The package
9458 must be designated by its stash (see C<gv_stashpv()>). The reference count
9459 of the SV is unaffected.
9465 Perl_sv_bless(pTHX_ SV *const sv, HV *const stash)
9470 PERL_ARGS_ASSERT_SV_BLESS;
9473 Perl_croak(aTHX_ "Can't bless non-reference value");
9475 if (SvFLAGS(tmpRef) & (SVs_OBJECT|SVf_READONLY)) {
9476 if (SvIsCOW(tmpRef))
9477 sv_force_normal_flags(tmpRef, 0);
9478 if (SvREADONLY(tmpRef))
9479 Perl_croak_no_modify(aTHX);
9480 if (SvOBJECT(tmpRef)) {
9481 if (SvTYPE(tmpRef) != SVt_PVIO)
9483 SvREFCNT_dec(SvSTASH(tmpRef));
9486 SvOBJECT_on(tmpRef);
9487 if (SvTYPE(tmpRef) != SVt_PVIO)
9489 SvUPGRADE(tmpRef, SVt_PVMG);
9490 SvSTASH_set(tmpRef, MUTABLE_HV(SvREFCNT_inc_simple(stash)));
9495 (void)SvAMAGIC_off(sv);
9497 if(SvSMAGICAL(tmpRef))
9498 if(mg_find(tmpRef, PERL_MAGIC_ext) || mg_find(tmpRef, PERL_MAGIC_uvar))
9506 /* Downgrades a PVGV to a PVMG. If it's actually a PVLV, we leave the type
9507 * as it is after unglobbing it.
9510 PERL_STATIC_INLINE void
9511 S_sv_unglob(pTHX_ SV *const sv, U32 flags)
9516 SV * const temp = flags & SV_COW_DROP_PV ? NULL : sv_newmortal();
9518 PERL_ARGS_ASSERT_SV_UNGLOB;
9520 assert(SvTYPE(sv) == SVt_PVGV || SvTYPE(sv) == SVt_PVLV);
9522 if (!(flags & SV_COW_DROP_PV))
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 if (flags & SV_COW_DROP_PV) SvOK_off(sv);
9555 else sv_setsv_flags(sv, temp, 0);
9557 if ((const GV *)sv == PL_last_in_gv)
9558 PL_last_in_gv = NULL;
9559 else if ((const GV *)sv == PL_statgv)
9564 =for apidoc sv_unref_flags
9566 Unsets the RV status of the SV, and decrements the reference count of
9567 whatever was being referenced by the RV. This can almost be thought of
9568 as a reversal of C<newSVrv>. The C<cflags> argument can contain
9569 C<SV_IMMEDIATE_UNREF> to force the reference count to be decremented
9570 (otherwise the decrementing is conditional on the reference count being
9571 different from one or the reference being a readonly SV).
9578 Perl_sv_unref_flags(pTHX_ SV *const ref, const U32 flags)
9580 SV* const target = SvRV(ref);
9582 PERL_ARGS_ASSERT_SV_UNREF_FLAGS;
9584 if (SvWEAKREF(ref)) {
9585 sv_del_backref(target, ref);
9587 SvRV_set(ref, NULL);
9590 SvRV_set(ref, NULL);
9592 /* You can't have a || SvREADONLY(target) here, as $a = $$a, where $a was
9593 assigned to as BEGIN {$a = \"Foo"} will fail. */
9594 if (SvREFCNT(target) != 1 || (flags & SV_IMMEDIATE_UNREF))
9595 SvREFCNT_dec(target);
9596 else /* XXX Hack, but hard to make $a=$a->[1] work otherwise */
9597 sv_2mortal(target); /* Schedule for freeing later */
9601 =for apidoc sv_untaint
9603 Untaint an SV. Use C<SvTAINTED_off> instead.
9609 Perl_sv_untaint(pTHX_ SV *const sv)
9611 PERL_ARGS_ASSERT_SV_UNTAINT;
9613 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
9614 MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
9621 =for apidoc sv_tainted
9623 Test an SV for taintedness. Use C<SvTAINTED> instead.
9629 Perl_sv_tainted(pTHX_ SV *const sv)
9631 PERL_ARGS_ASSERT_SV_TAINTED;
9633 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
9634 const MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
9635 if (mg && (mg->mg_len & 1) )
9642 =for apidoc sv_setpviv
9644 Copies an integer into the given SV, also updating its string value.
9645 Does not handle 'set' magic. See C<sv_setpviv_mg>.
9651 Perl_sv_setpviv(pTHX_ SV *const sv, const IV iv)
9653 char buf[TYPE_CHARS(UV)];
9655 char * const ptr = uiv_2buf(buf, iv, 0, 0, &ebuf);
9657 PERL_ARGS_ASSERT_SV_SETPVIV;
9659 sv_setpvn(sv, ptr, ebuf - ptr);
9663 =for apidoc sv_setpviv_mg
9665 Like C<sv_setpviv>, but also handles 'set' magic.
9671 Perl_sv_setpviv_mg(pTHX_ SV *const sv, const IV iv)
9673 PERL_ARGS_ASSERT_SV_SETPVIV_MG;
9679 #if defined(PERL_IMPLICIT_CONTEXT)
9681 /* pTHX_ magic can't cope with varargs, so this is a no-context
9682 * version of the main function, (which may itself be aliased to us).
9683 * Don't access this version directly.
9687 Perl_sv_setpvf_nocontext(SV *const sv, const char *const pat, ...)
9692 PERL_ARGS_ASSERT_SV_SETPVF_NOCONTEXT;
9694 va_start(args, pat);
9695 sv_vsetpvf(sv, pat, &args);
9699 /* pTHX_ magic can't cope with varargs, so this is a no-context
9700 * version of the main function, (which may itself be aliased to us).
9701 * Don't access this version directly.
9705 Perl_sv_setpvf_mg_nocontext(SV *const sv, const char *const pat, ...)
9710 PERL_ARGS_ASSERT_SV_SETPVF_MG_NOCONTEXT;
9712 va_start(args, pat);
9713 sv_vsetpvf_mg(sv, pat, &args);
9719 =for apidoc sv_setpvf
9721 Works like C<sv_catpvf> but copies the text into the SV instead of
9722 appending it. Does not handle 'set' magic. See C<sv_setpvf_mg>.
9728 Perl_sv_setpvf(pTHX_ SV *const sv, const char *const pat, ...)
9732 PERL_ARGS_ASSERT_SV_SETPVF;
9734 va_start(args, pat);
9735 sv_vsetpvf(sv, pat, &args);
9740 =for apidoc sv_vsetpvf
9742 Works like C<sv_vcatpvf> but copies the text into the SV instead of
9743 appending it. Does not handle 'set' magic. See C<sv_vsetpvf_mg>.
9745 Usually used via its frontend C<sv_setpvf>.
9751 Perl_sv_vsetpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9753 PERL_ARGS_ASSERT_SV_VSETPVF;
9755 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9759 =for apidoc sv_setpvf_mg
9761 Like C<sv_setpvf>, but also handles 'set' magic.
9767 Perl_sv_setpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
9771 PERL_ARGS_ASSERT_SV_SETPVF_MG;
9773 va_start(args, pat);
9774 sv_vsetpvf_mg(sv, pat, &args);
9779 =for apidoc sv_vsetpvf_mg
9781 Like C<sv_vsetpvf>, but also handles 'set' magic.
9783 Usually used via its frontend C<sv_setpvf_mg>.
9789 Perl_sv_vsetpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9791 PERL_ARGS_ASSERT_SV_VSETPVF_MG;
9793 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9797 #if defined(PERL_IMPLICIT_CONTEXT)
9799 /* pTHX_ magic can't cope with varargs, so this is a no-context
9800 * version of the main function, (which may itself be aliased to us).
9801 * Don't access this version directly.
9805 Perl_sv_catpvf_nocontext(SV *const sv, const char *const pat, ...)
9810 PERL_ARGS_ASSERT_SV_CATPVF_NOCONTEXT;
9812 va_start(args, pat);
9813 sv_vcatpvf(sv, pat, &args);
9817 /* pTHX_ magic can't cope with varargs, so this is a no-context
9818 * version of the main function, (which may itself be aliased to us).
9819 * Don't access this version directly.
9823 Perl_sv_catpvf_mg_nocontext(SV *const sv, const char *const pat, ...)
9828 PERL_ARGS_ASSERT_SV_CATPVF_MG_NOCONTEXT;
9830 va_start(args, pat);
9831 sv_vcatpvf_mg(sv, pat, &args);
9837 =for apidoc sv_catpvf
9839 Processes its arguments like C<sprintf> and appends the formatted
9840 output to an SV. If the appended data contains "wide" characters
9841 (including, but not limited to, SVs with a UTF-8 PV formatted with %s,
9842 and characters >255 formatted with %c), the original SV might get
9843 upgraded to UTF-8. Handles 'get' magic, but not 'set' magic. See
9844 C<sv_catpvf_mg>. If the original SV was UTF-8, the pattern should be
9845 valid UTF-8; if the original SV was bytes, the pattern should be too.
9850 Perl_sv_catpvf(pTHX_ SV *const sv, const char *const pat, ...)
9854 PERL_ARGS_ASSERT_SV_CATPVF;
9856 va_start(args, pat);
9857 sv_vcatpvf(sv, pat, &args);
9862 =for apidoc sv_vcatpvf
9864 Processes its arguments like C<vsprintf> and appends the formatted output
9865 to an SV. Does not handle 'set' magic. See C<sv_vcatpvf_mg>.
9867 Usually used via its frontend C<sv_catpvf>.
9873 Perl_sv_vcatpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9875 PERL_ARGS_ASSERT_SV_VCATPVF;
9877 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9881 =for apidoc sv_catpvf_mg
9883 Like C<sv_catpvf>, but also handles 'set' magic.
9889 Perl_sv_catpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
9893 PERL_ARGS_ASSERT_SV_CATPVF_MG;
9895 va_start(args, pat);
9896 sv_vcatpvf_mg(sv, pat, &args);
9901 =for apidoc sv_vcatpvf_mg
9903 Like C<sv_vcatpvf>, but also handles 'set' magic.
9905 Usually used via its frontend C<sv_catpvf_mg>.
9911 Perl_sv_vcatpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9913 PERL_ARGS_ASSERT_SV_VCATPVF_MG;
9915 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9920 =for apidoc sv_vsetpvfn
9922 Works like C<sv_vcatpvfn> but copies the text into the SV instead of
9925 Usually used via one of its frontends C<sv_vsetpvf> and C<sv_vsetpvf_mg>.
9931 Perl_sv_vsetpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
9932 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
9934 PERL_ARGS_ASSERT_SV_VSETPVFN;
9937 sv_vcatpvfn(sv, pat, patlen, args, svargs, svmax, maybe_tainted);
9942 * Warn of missing argument to sprintf, and then return a defined value
9943 * to avoid inappropriate "use of uninit" warnings [perl #71000].
9945 #define WARN_MISSING WARN_UNINITIALIZED /* Not sure we want a new category */
9947 S_vcatpvfn_missing_argument(pTHX) {
9948 if (ckWARN(WARN_MISSING)) {
9949 Perl_warner(aTHX_ packWARN(WARN_MISSING), "Missing argument in %s",
9950 PL_op ? OP_DESC(PL_op) : "sv_vcatpvfn()");
9957 S_expect_number(pTHX_ char **const pattern)
9962 PERL_ARGS_ASSERT_EXPECT_NUMBER;
9964 switch (**pattern) {
9965 case '1': case '2': case '3':
9966 case '4': case '5': case '6':
9967 case '7': case '8': case '9':
9968 var = *(*pattern)++ - '0';
9969 while (isDIGIT(**pattern)) {
9970 const I32 tmp = var * 10 + (*(*pattern)++ - '0');
9972 Perl_croak(aTHX_ "Integer overflow in format string for %s", (PL_op ? OP_DESC(PL_op) : "sv_vcatpvfn"));
9980 S_F0convert(NV nv, char *const endbuf, STRLEN *const len)
9982 const int neg = nv < 0;
9985 PERL_ARGS_ASSERT_F0CONVERT;
9993 if (uv & 1 && uv == nv)
9994 uv--; /* Round to even */
9996 const unsigned dig = uv % 10;
10009 =for apidoc sv_vcatpvfn
10011 Processes its arguments like C<vsprintf> and appends the formatted output
10012 to an SV. Uses an array of SVs if the C style variable argument list is
10013 missing (NULL). When running with taint checks enabled, indicates via
10014 C<maybe_tainted> if results are untrustworthy (often due to the use of
10017 Usually used via one of its frontends C<sv_vcatpvf> and C<sv_vcatpvf_mg>.
10023 #define VECTORIZE_ARGS vecsv = va_arg(*args, SV*);\
10024 vecstr = (U8*)SvPV_const(vecsv,veclen);\
10025 vec_utf8 = DO_UTF8(vecsv);
10027 /* XXX maybe_tainted is never assigned to, so the doc above is lying. */
10030 Perl_sv_vcatpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
10031 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
10036 const char *patend;
10039 static const char nullstr[] = "(null)";
10041 bool has_utf8 = DO_UTF8(sv); /* has the result utf8? */
10042 const bool pat_utf8 = has_utf8; /* the pattern is in utf8? */
10044 /* Times 4: a decimal digit takes more than 3 binary digits.
10045 * NV_DIG: mantissa takes than many decimal digits.
10046 * Plus 32: Playing safe. */
10047 char ebuf[IV_DIG * 4 + NV_DIG + 32];
10048 /* large enough for "%#.#f" --chip */
10049 /* what about long double NVs? --jhi */
10051 PERL_ARGS_ASSERT_SV_VCATPVFN;
10052 PERL_UNUSED_ARG(maybe_tainted);
10054 /* no matter what, this is a string now */
10055 (void)SvPV_force(sv, origlen);
10057 /* special-case "", "%s", and "%-p" (SVf - see below) */
10060 if (patlen == 2 && pat[0] == '%' && pat[1] == 's') {
10062 const char * const s = va_arg(*args, char*);
10063 sv_catpv(sv, s ? s : nullstr);
10065 else if (svix < svmax) {
10066 sv_catsv(sv, *svargs);
10069 S_vcatpvfn_missing_argument(aTHX);
10072 if (args && patlen == 3 && pat[0] == '%' &&
10073 pat[1] == '-' && pat[2] == 'p') {
10074 argsv = MUTABLE_SV(va_arg(*args, void*));
10075 sv_catsv(sv, argsv);
10079 #ifndef USE_LONG_DOUBLE
10080 /* special-case "%.<number>[gf]" */
10081 if ( !args && patlen <= 5 && pat[0] == '%' && pat[1] == '.'
10082 && (pat[patlen-1] == 'g' || pat[patlen-1] == 'f') ) {
10083 unsigned digits = 0;
10087 while (*pp >= '0' && *pp <= '9')
10088 digits = 10 * digits + (*pp++ - '0');
10089 if (pp - pat == (int)patlen - 1 && svix < svmax) {
10090 const NV nv = SvNV(*svargs);
10092 /* Add check for digits != 0 because it seems that some
10093 gconverts are buggy in this case, and we don't yet have
10094 a Configure test for this. */
10095 if (digits && digits < sizeof(ebuf) - NV_DIG - 10) {
10096 /* 0, point, slack */
10097 Gconvert(nv, (int)digits, 0, ebuf);
10098 sv_catpv(sv, ebuf);
10099 if (*ebuf) /* May return an empty string for digits==0 */
10102 } else if (!digits) {
10105 if ((p = F0convert(nv, ebuf + sizeof ebuf, &l))) {
10106 sv_catpvn(sv, p, l);
10112 #endif /* !USE_LONG_DOUBLE */
10114 if (!args && svix < svmax && DO_UTF8(*svargs))
10117 patend = (char*)pat + patlen;
10118 for (p = (char*)pat; p < patend; p = q) {
10121 bool vectorize = FALSE;
10122 bool vectorarg = FALSE;
10123 bool vec_utf8 = FALSE;
10129 bool has_precis = FALSE;
10131 const I32 osvix = svix;
10132 bool is_utf8 = FALSE; /* is this item utf8? */
10133 #ifdef HAS_LDBL_SPRINTF_BUG
10134 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
10135 with sfio - Allen <allens@cpan.org> */
10136 bool fix_ldbl_sprintf_bug = FALSE;
10140 U8 utf8buf[UTF8_MAXBYTES+1];
10141 STRLEN esignlen = 0;
10143 const char *eptr = NULL;
10144 const char *fmtstart;
10147 const U8 *vecstr = NULL;
10154 /* we need a long double target in case HAS_LONG_DOUBLE but
10155 not USE_LONG_DOUBLE
10157 #if defined(HAS_LONG_DOUBLE) && LONG_DOUBLESIZE > DOUBLESIZE
10165 const char *dotstr = ".";
10166 STRLEN dotstrlen = 1;
10167 I32 efix = 0; /* explicit format parameter index */
10168 I32 ewix = 0; /* explicit width index */
10169 I32 epix = 0; /* explicit precision index */
10170 I32 evix = 0; /* explicit vector index */
10171 bool asterisk = FALSE;
10173 /* echo everything up to the next format specification */
10174 for (q = p; q < patend && *q != '%'; ++q) ;
10176 if (has_utf8 && !pat_utf8)
10177 sv_catpvn_utf8_upgrade(sv, p, q - p, nsv);
10179 sv_catpvn(sv, p, q - p);
10188 We allow format specification elements in this order:
10189 \d+\$ explicit format parameter index
10191 v|\*(\d+\$)?v vector with optional (optionally specified) arg
10192 0 flag (as above): repeated to allow "v02"
10193 \d+|\*(\d+\$)? width using optional (optionally specified) arg
10194 \.(\d*|\*(\d+\$)?) precision using optional (optionally specified) arg
10196 [%bcdefginopsuxDFOUX] format (mandatory)
10201 As of perl5.9.3, printf format checking is on by default.
10202 Internally, perl uses %p formats to provide an escape to
10203 some extended formatting. This block deals with those
10204 extensions: if it does not match, (char*)q is reset and
10205 the normal format processing code is used.
10207 Currently defined extensions are:
10208 %p include pointer address (standard)
10209 %-p (SVf) include an SV (previously %_)
10210 %-<num>p include an SV with precision <num>
10212 %3p include a HEK with precision of 256
10213 %<num>p (where num != 2 or 3) reserved for future
10216 Robin Barker 2005-07-14 (but modified since)
10218 %1p (VDf) removed. RMB 2007-10-19
10225 n = expect_number(&q);
10227 if (sv) { /* SVf */
10232 argsv = MUTABLE_SV(va_arg(*args, void*));
10233 eptr = SvPV_const(argsv, elen);
10234 if (DO_UTF8(argsv))
10238 else if (n==2 || n==3) { /* HEKf */
10239 HEK * const hek = va_arg(*args, HEK *);
10240 eptr = HEK_KEY(hek);
10241 elen = HEK_LEN(hek);
10242 if (HEK_UTF8(hek)) is_utf8 = TRUE;
10243 if (n==3) precis = 256, has_precis = TRUE;
10247 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL),
10248 "internal %%<num>p might conflict with future printf extensions");
10254 if ( (width = expect_number(&q)) ) {
10269 if (plus == '+' && *q == ' ') /* '+' over ' ' */
10298 if ( (ewix = expect_number(&q)) )
10307 if ((vectorarg = asterisk)) {
10320 width = expect_number(&q);
10323 if (vectorize && vectorarg) {
10324 /* vectorizing, but not with the default "." */
10326 vecsv = va_arg(*args, SV*);
10328 vecsv = (evix > 0 && evix <= svmax)
10329 ? svargs[evix-1] : S_vcatpvfn_missing_argument(aTHX);
10331 vecsv = svix < svmax
10332 ? svargs[svix++] : S_vcatpvfn_missing_argument(aTHX);
10334 dotstr = SvPV_const(vecsv, dotstrlen);
10335 /* Keep the DO_UTF8 test *after* the SvPV call, else things go
10336 bad with tied or overloaded values that return UTF8. */
10337 if (DO_UTF8(vecsv))
10339 else if (has_utf8) {
10340 vecsv = sv_mortalcopy(vecsv);
10341 sv_utf8_upgrade(vecsv);
10342 dotstr = SvPV_const(vecsv, dotstrlen);
10349 i = va_arg(*args, int);
10351 i = (ewix ? ewix <= svmax : svix < svmax) ?
10352 SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
10354 width = (i < 0) ? -i : i;
10364 if ( ((epix = expect_number(&q))) && (*q++ != '$') )
10366 /* XXX: todo, support specified precision parameter */
10370 i = va_arg(*args, int);
10372 i = (ewix ? ewix <= svmax : svix < svmax)
10373 ? SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
10375 has_precis = !(i < 0);
10379 while (isDIGIT(*q))
10380 precis = precis * 10 + (*q++ - '0');
10389 else if (efix ? (efix > 0 && efix <= svmax) : svix < svmax) {
10390 vecsv = svargs[efix ? efix-1 : svix++];
10391 vecstr = (U8*)SvPV_const(vecsv,veclen);
10392 vec_utf8 = DO_UTF8(vecsv);
10394 /* if this is a version object, we need to convert
10395 * back into v-string notation and then let the
10396 * vectorize happen normally
10398 if (sv_isobject(vecsv) && sv_derived_from(vecsv, "version")) {
10399 char *version = savesvpv(vecsv);
10400 if ( hv_exists(MUTABLE_HV(SvRV(vecsv)), "alpha", 5 ) ) {
10401 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
10402 "vector argument not supported with alpha versions");
10405 vecsv = sv_newmortal();
10406 scan_vstring(version, version + veclen, vecsv);
10407 vecstr = (U8*)SvPV_const(vecsv, veclen);
10408 vec_utf8 = DO_UTF8(vecsv);
10422 case 'I': /* Ix, I32x, and I64x */
10424 if (q[1] == '6' && q[2] == '4') {
10430 if (q[1] == '3' && q[2] == '2') {
10440 #if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
10452 #if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
10453 if (*q == 'l') { /* lld, llf */
10462 if (*++q == 'h') { /* hhd, hhu */
10491 if (!vectorize && !args) {
10493 const I32 i = efix-1;
10494 argsv = (i >= 0 && i < svmax)
10495 ? svargs[i] : S_vcatpvfn_missing_argument(aTHX);
10497 argsv = (svix >= 0 && svix < svmax)
10498 ? svargs[svix++] : S_vcatpvfn_missing_argument(aTHX);
10502 switch (c = *q++) {
10509 uv = (args) ? va_arg(*args, int) : SvIV(argsv);
10511 (!UNI_IS_INVARIANT(uv) && SvUTF8(sv)))
10513 eptr = (char*)utf8buf;
10514 elen = uvchr_to_utf8((U8*)eptr, uv) - utf8buf;
10528 eptr = va_arg(*args, char*);
10530 elen = strlen(eptr);
10532 eptr = (char *)nullstr;
10533 elen = sizeof nullstr - 1;
10537 eptr = SvPV_const(argsv, elen);
10538 if (DO_UTF8(argsv)) {
10539 STRLEN old_precis = precis;
10540 if (has_precis && precis < elen) {
10541 STRLEN ulen = sv_len_utf8(argsv);
10542 I32 p = precis > ulen ? ulen : precis;
10543 sv_pos_u2b(argsv, &p, 0); /* sticks at end */
10546 if (width) { /* fudge width (can't fudge elen) */
10547 if (has_precis && precis < elen)
10548 width += precis - old_precis;
10550 width += elen - sv_len_utf8(argsv);
10557 if (has_precis && precis < elen)
10564 if (alt || vectorize)
10566 uv = PTR2UV(args ? va_arg(*args, void*) : argsv);
10587 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
10596 esignbuf[esignlen++] = plus;
10600 case 'c': iv = (char)va_arg(*args, int); break;
10601 case 'h': iv = (short)va_arg(*args, int); break;
10602 case 'l': iv = va_arg(*args, long); break;
10603 case 'V': iv = va_arg(*args, IV); break;
10604 case 'z': iv = va_arg(*args, SSize_t); break;
10605 case 't': iv = va_arg(*args, ptrdiff_t); break;
10606 default: iv = va_arg(*args, int); break;
10608 case 'j': iv = va_arg(*args, intmax_t); break;
10612 iv = va_arg(*args, Quad_t); break;
10619 IV tiv = SvIV(argsv); /* work around GCC bug #13488 */
10621 case 'c': iv = (char)tiv; break;
10622 case 'h': iv = (short)tiv; break;
10623 case 'l': iv = (long)tiv; break;
10625 default: iv = tiv; break;
10628 iv = (Quad_t)tiv; break;
10634 if ( !vectorize ) /* we already set uv above */
10639 esignbuf[esignlen++] = plus;
10643 esignbuf[esignlen++] = '-';
10687 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
10698 case 'c': uv = (unsigned char)va_arg(*args, unsigned); break;
10699 case 'h': uv = (unsigned short)va_arg(*args, unsigned); break;
10700 case 'l': uv = va_arg(*args, unsigned long); break;
10701 case 'V': uv = va_arg(*args, UV); break;
10702 case 'z': uv = va_arg(*args, Size_t); break;
10703 case 't': uv = va_arg(*args, ptrdiff_t); break; /* will sign extend, but there is no uptrdiff_t, so oh well */
10705 case 'j': uv = va_arg(*args, uintmax_t); break;
10707 default: uv = va_arg(*args, unsigned); break;
10710 uv = va_arg(*args, Uquad_t); break;
10717 UV tuv = SvUV(argsv); /* work around GCC bug #13488 */
10719 case 'c': uv = (unsigned char)tuv; break;
10720 case 'h': uv = (unsigned short)tuv; break;
10721 case 'l': uv = (unsigned long)tuv; break;
10723 default: uv = tuv; break;
10726 uv = (Uquad_t)tuv; break;
10735 char *ptr = ebuf + sizeof ebuf;
10736 bool tempalt = uv ? alt : FALSE; /* Vectors can't change alt */
10742 p = (char *)((c == 'X') ? PL_hexdigit + 16 : PL_hexdigit);
10746 } while (uv >>= 4);
10748 esignbuf[esignlen++] = '0';
10749 esignbuf[esignlen++] = c; /* 'x' or 'X' */
10755 *--ptr = '0' + dig;
10756 } while (uv >>= 3);
10757 if (alt && *ptr != '0')
10763 *--ptr = '0' + dig;
10764 } while (uv >>= 1);
10766 esignbuf[esignlen++] = '0';
10767 esignbuf[esignlen++] = c;
10770 default: /* it had better be ten or less */
10773 *--ptr = '0' + dig;
10774 } while (uv /= base);
10777 elen = (ebuf + sizeof ebuf) - ptr;
10781 zeros = precis - elen;
10782 else if (precis == 0 && elen == 1 && *eptr == '0'
10783 && !(base == 8 && alt)) /* "%#.0o" prints "0" */
10786 /* a precision nullifies the 0 flag. */
10793 /* FLOATING POINT */
10796 c = 'f'; /* maybe %F isn't supported here */
10798 case 'e': case 'E':
10800 case 'g': case 'G':
10804 /* This is evil, but floating point is even more evil */
10806 /* for SV-style calling, we can only get NV
10807 for C-style calling, we assume %f is double;
10808 for simplicity we allow any of %Lf, %llf, %qf for long double
10812 #if defined(USE_LONG_DOUBLE)
10816 /* [perl #20339] - we should accept and ignore %lf rather than die */
10820 #if defined(USE_LONG_DOUBLE)
10821 intsize = args ? 0 : 'q';
10825 #if defined(HAS_LONG_DOUBLE)
10838 /* now we need (long double) if intsize == 'q', else (double) */
10840 #if LONG_DOUBLESIZE > DOUBLESIZE
10842 va_arg(*args, long double) :
10843 va_arg(*args, double)
10845 va_arg(*args, double)
10850 /* nv * 0 will be NaN for NaN, +Inf and -Inf, and 0 for anything
10851 else. frexp() has some unspecified behaviour for those three */
10852 if (c != 'e' && c != 'E' && (nv * 0) == 0) {
10854 /* FIXME: if HAS_LONG_DOUBLE but not USE_LONG_DOUBLE this
10855 will cast our (long double) to (double) */
10856 (void)Perl_frexp(nv, &i);
10857 if (i == PERL_INT_MIN)
10858 Perl_die(aTHX_ "panic: frexp");
10860 need = BIT_DIGITS(i);
10862 need += has_precis ? precis : 6; /* known default */
10867 #ifdef HAS_LDBL_SPRINTF_BUG
10868 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
10869 with sfio - Allen <allens@cpan.org> */
10872 # define MY_DBL_MAX DBL_MAX
10873 # else /* XXX guessing! HUGE_VAL may be defined as infinity, so not using */
10874 # if DOUBLESIZE >= 8
10875 # define MY_DBL_MAX 1.7976931348623157E+308L
10877 # define MY_DBL_MAX 3.40282347E+38L
10881 # ifdef HAS_LDBL_SPRINTF_BUG_LESS1 /* only between -1L & 1L - Allen */
10882 # define MY_DBL_MAX_BUG 1L
10884 # define MY_DBL_MAX_BUG MY_DBL_MAX
10888 # define MY_DBL_MIN DBL_MIN
10889 # else /* XXX guessing! -Allen */
10890 # if DOUBLESIZE >= 8
10891 # define MY_DBL_MIN 2.2250738585072014E-308L
10893 # define MY_DBL_MIN 1.17549435E-38L
10897 if ((intsize == 'q') && (c == 'f') &&
10898 ((nv < MY_DBL_MAX_BUG) && (nv > -MY_DBL_MAX_BUG)) &&
10899 (need < DBL_DIG)) {
10900 /* it's going to be short enough that
10901 * long double precision is not needed */
10903 if ((nv <= 0L) && (nv >= -0L))
10904 fix_ldbl_sprintf_bug = TRUE; /* 0 is 0 - easiest */
10906 /* would use Perl_fp_class as a double-check but not
10907 * functional on IRIX - see perl.h comments */
10909 if ((nv >= MY_DBL_MIN) || (nv <= -MY_DBL_MIN)) {
10910 /* It's within the range that a double can represent */
10911 #if defined(DBL_MAX) && !defined(DBL_MIN)
10912 if ((nv >= ((long double)1/DBL_MAX)) ||
10913 (nv <= (-(long double)1/DBL_MAX)))
10915 fix_ldbl_sprintf_bug = TRUE;
10918 if (fix_ldbl_sprintf_bug == TRUE) {
10928 # undef MY_DBL_MAX_BUG
10931 #endif /* HAS_LDBL_SPRINTF_BUG */
10933 need += 20; /* fudge factor */
10934 if (PL_efloatsize < need) {
10935 Safefree(PL_efloatbuf);
10936 PL_efloatsize = need + 20; /* more fudge */
10937 Newx(PL_efloatbuf, PL_efloatsize, char);
10938 PL_efloatbuf[0] = '\0';
10941 if ( !(width || left || plus || alt) && fill != '0'
10942 && has_precis && intsize != 'q' ) { /* Shortcuts */
10943 /* See earlier comment about buggy Gconvert when digits,
10945 if ( c == 'g' && precis) {
10946 Gconvert((NV)nv, (int)precis, 0, PL_efloatbuf);
10947 /* May return an empty string for digits==0 */
10948 if (*PL_efloatbuf) {
10949 elen = strlen(PL_efloatbuf);
10950 goto float_converted;
10952 } else if ( c == 'f' && !precis) {
10953 if ((eptr = F0convert(nv, ebuf + sizeof ebuf, &elen)))
10958 char *ptr = ebuf + sizeof ebuf;
10961 /* FIXME: what to do if HAS_LONG_DOUBLE but not PERL_PRIfldbl? */
10962 #if defined(HAS_LONG_DOUBLE) && defined(PERL_PRIfldbl)
10963 if (intsize == 'q') {
10964 /* Copy the one or more characters in a long double
10965 * format before the 'base' ([efgEFG]) character to
10966 * the format string. */
10967 static char const prifldbl[] = PERL_PRIfldbl;
10968 char const *p = prifldbl + sizeof(prifldbl) - 3;
10969 while (p >= prifldbl) { *--ptr = *p--; }
10974 do { *--ptr = '0' + (base % 10); } while (base /= 10);
10979 do { *--ptr = '0' + (base % 10); } while (base /= 10);
10991 /* No taint. Otherwise we are in the strange situation
10992 * where printf() taints but print($float) doesn't.
10994 #if defined(HAS_LONG_DOUBLE)
10995 elen = ((intsize == 'q')
10996 ? my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, nv)
10997 : my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, (double)nv));
10999 elen = my_sprintf(PL_efloatbuf, ptr, nv);
11003 eptr = PL_efloatbuf;
11011 i = SvCUR(sv) - origlen;
11014 case 'c': *(va_arg(*args, char*)) = i; break;
11015 case 'h': *(va_arg(*args, short*)) = i; break;
11016 default: *(va_arg(*args, int*)) = i; break;
11017 case 'l': *(va_arg(*args, long*)) = i; break;
11018 case 'V': *(va_arg(*args, IV*)) = i; break;
11019 case 'z': *(va_arg(*args, SSize_t*)) = i; break;
11020 case 't': *(va_arg(*args, ptrdiff_t*)) = i; break;
11022 case 'j': *(va_arg(*args, intmax_t*)) = i; break;
11026 *(va_arg(*args, Quad_t*)) = i; break;
11033 sv_setuv_mg(argsv, has_utf8 ? (UV)sv_len_utf8(sv) : (UV)i);
11034 continue; /* not "break" */
11041 && (PL_op->op_type == OP_PRTF || PL_op->op_type == OP_SPRINTF)
11042 && ckWARN(WARN_PRINTF))
11044 SV * const msg = sv_newmortal();
11045 Perl_sv_setpvf(aTHX_ msg, "Invalid conversion in %sprintf: ",
11046 (PL_op->op_type == OP_PRTF) ? "" : "s");
11047 if (fmtstart < patend) {
11048 const char * const fmtend = q < patend ? q : patend;
11050 sv_catpvs(msg, "\"%");
11051 for (f = fmtstart; f < fmtend; f++) {
11053 sv_catpvn(msg, f, 1);
11055 Perl_sv_catpvf(aTHX_ msg,
11056 "\\%03"UVof, (UV)*f & 0xFF);
11059 sv_catpvs(msg, "\"");
11061 sv_catpvs(msg, "end of string");
11063 Perl_warner(aTHX_ packWARN(WARN_PRINTF), "%"SVf, SVfARG(msg)); /* yes, this is reentrant */
11066 /* output mangled stuff ... */
11072 /* ... right here, because formatting flags should not apply */
11073 SvGROW(sv, SvCUR(sv) + elen + 1);
11075 Copy(eptr, p, elen, char);
11078 SvCUR_set(sv, p - SvPVX_const(sv));
11080 continue; /* not "break" */
11083 if (is_utf8 != has_utf8) {
11086 sv_utf8_upgrade(sv);
11089 const STRLEN old_elen = elen;
11090 SV * const nsv = newSVpvn_flags(eptr, elen, SVs_TEMP);
11091 sv_utf8_upgrade(nsv);
11092 eptr = SvPVX_const(nsv);
11095 if (width) { /* fudge width (can't fudge elen) */
11096 width += elen - old_elen;
11102 have = esignlen + zeros + elen;
11104 Perl_croak_nocontext("%s", PL_memory_wrap);
11106 need = (have > width ? have : width);
11109 if (need >= (((STRLEN)~0) - SvCUR(sv) - dotstrlen - 1))
11110 Perl_croak_nocontext("%s", PL_memory_wrap);
11111 SvGROW(sv, SvCUR(sv) + need + dotstrlen + 1);
11113 if (esignlen && fill == '0') {
11115 for (i = 0; i < (int)esignlen; i++)
11116 *p++ = esignbuf[i];
11118 if (gap && !left) {
11119 memset(p, fill, gap);
11122 if (esignlen && fill != '0') {
11124 for (i = 0; i < (int)esignlen; i++)
11125 *p++ = esignbuf[i];
11129 for (i = zeros; i; i--)
11133 Copy(eptr, p, elen, char);
11137 memset(p, ' ', gap);
11142 Copy(dotstr, p, dotstrlen, char);
11146 vectorize = FALSE; /* done iterating over vecstr */
11153 SvCUR_set(sv, p - SvPVX_const(sv));
11162 /* =========================================================================
11164 =head1 Cloning an interpreter
11166 All the macros and functions in this section are for the private use of
11167 the main function, perl_clone().
11169 The foo_dup() functions make an exact copy of an existing foo thingy.
11170 During the course of a cloning, a hash table is used to map old addresses
11171 to new addresses. The table is created and manipulated with the
11172 ptr_table_* functions.
11176 * =========================================================================*/
11179 #if defined(USE_ITHREADS)
11181 /* XXX Remove this so it doesn't have to go thru the macro and return for nothing */
11182 #ifndef GpREFCNT_inc
11183 # define GpREFCNT_inc(gp) ((gp) ? (++(gp)->gp_refcnt, (gp)) : (GP*)NULL)
11187 /* Certain cases in Perl_ss_dup have been merged, by relying on the fact
11188 that currently av_dup, gv_dup and hv_dup are the same as sv_dup.
11189 If this changes, please unmerge ss_dup.
11190 Likewise, sv_dup_inc_multiple() relies on this fact. */
11191 #define sv_dup_inc_NN(s,t) SvREFCNT_inc_NN(sv_dup_inc(s,t))
11192 #define av_dup(s,t) MUTABLE_AV(sv_dup((const SV *)s,t))
11193 #define av_dup_inc(s,t) MUTABLE_AV(sv_dup_inc((const SV *)s,t))
11194 #define hv_dup(s,t) MUTABLE_HV(sv_dup((const SV *)s,t))
11195 #define hv_dup_inc(s,t) MUTABLE_HV(sv_dup_inc((const SV *)s,t))
11196 #define cv_dup(s,t) MUTABLE_CV(sv_dup((const SV *)s,t))
11197 #define cv_dup_inc(s,t) MUTABLE_CV(sv_dup_inc((const SV *)s,t))
11198 #define io_dup(s,t) MUTABLE_IO(sv_dup((const SV *)s,t))
11199 #define io_dup_inc(s,t) MUTABLE_IO(sv_dup_inc((const SV *)s,t))
11200 #define gv_dup(s,t) MUTABLE_GV(sv_dup((const SV *)s,t))
11201 #define gv_dup_inc(s,t) MUTABLE_GV(sv_dup_inc((const SV *)s,t))
11202 #define SAVEPV(p) ((p) ? savepv(p) : NULL)
11203 #define SAVEPVN(p,n) ((p) ? savepvn(p,n) : NULL)
11205 /* clone a parser */
11208 Perl_parser_dup(pTHX_ const yy_parser *const proto, CLONE_PARAMS *const param)
11212 PERL_ARGS_ASSERT_PARSER_DUP;
11217 /* look for it in the table first */
11218 parser = (yy_parser *)ptr_table_fetch(PL_ptr_table, proto);
11222 /* create anew and remember what it is */
11223 Newxz(parser, 1, yy_parser);
11224 ptr_table_store(PL_ptr_table, proto, parser);
11226 /* XXX these not yet duped */
11227 parser->old_parser = NULL;
11228 parser->stack = NULL;
11230 parser->stack_size = 0;
11231 /* XXX parser->stack->state = 0; */
11233 /* XXX eventually, just Copy() most of the parser struct ? */
11235 parser->lex_brackets = proto->lex_brackets;
11236 parser->lex_casemods = proto->lex_casemods;
11237 parser->lex_brackstack = savepvn(proto->lex_brackstack,
11238 (proto->lex_brackets < 120 ? 120 : proto->lex_brackets));
11239 parser->lex_casestack = savepvn(proto->lex_casestack,
11240 (proto->lex_casemods < 12 ? 12 : proto->lex_casemods));
11241 parser->lex_defer = proto->lex_defer;
11242 parser->lex_dojoin = proto->lex_dojoin;
11243 parser->lex_expect = proto->lex_expect;
11244 parser->lex_formbrack = proto->lex_formbrack;
11245 parser->lex_inpat = proto->lex_inpat;
11246 parser->lex_inwhat = proto->lex_inwhat;
11247 parser->lex_op = proto->lex_op;
11248 parser->lex_repl = sv_dup_inc(proto->lex_repl, param);
11249 parser->lex_starts = proto->lex_starts;
11250 parser->lex_stuff = sv_dup_inc(proto->lex_stuff, param);
11251 parser->multi_close = proto->multi_close;
11252 parser->multi_open = proto->multi_open;
11253 parser->multi_start = proto->multi_start;
11254 parser->multi_end = proto->multi_end;
11255 parser->pending_ident = proto->pending_ident;
11256 parser->preambled = proto->preambled;
11257 parser->sublex_info = proto->sublex_info; /* XXX not quite right */
11258 parser->linestr = sv_dup_inc(proto->linestr, param);
11259 parser->expect = proto->expect;
11260 parser->copline = proto->copline;
11261 parser->last_lop_op = proto->last_lop_op;
11262 parser->lex_state = proto->lex_state;
11263 parser->rsfp = fp_dup(proto->rsfp, '<', param);
11264 /* rsfp_filters entries have fake IoDIRP() */
11265 parser->rsfp_filters= av_dup_inc(proto->rsfp_filters, param);
11266 parser->in_my = proto->in_my;
11267 parser->in_my_stash = hv_dup(proto->in_my_stash, param);
11268 parser->error_count = proto->error_count;
11271 parser->linestr = sv_dup_inc(proto->linestr, param);
11274 char * const ols = SvPVX(proto->linestr);
11275 char * const ls = SvPVX(parser->linestr);
11277 parser->bufptr = ls + (proto->bufptr >= ols ?
11278 proto->bufptr - ols : 0);
11279 parser->oldbufptr = ls + (proto->oldbufptr >= ols ?
11280 proto->oldbufptr - ols : 0);
11281 parser->oldoldbufptr= ls + (proto->oldoldbufptr >= ols ?
11282 proto->oldoldbufptr - ols : 0);
11283 parser->linestart = ls + (proto->linestart >= ols ?
11284 proto->linestart - ols : 0);
11285 parser->last_uni = ls + (proto->last_uni >= ols ?
11286 proto->last_uni - ols : 0);
11287 parser->last_lop = ls + (proto->last_lop >= ols ?
11288 proto->last_lop - ols : 0);
11290 parser->bufend = ls + SvCUR(parser->linestr);
11293 Copy(proto->tokenbuf, parser->tokenbuf, 256, char);
11297 parser->endwhite = proto->endwhite;
11298 parser->faketokens = proto->faketokens;
11299 parser->lasttoke = proto->lasttoke;
11300 parser->nextwhite = proto->nextwhite;
11301 parser->realtokenstart = proto->realtokenstart;
11302 parser->skipwhite = proto->skipwhite;
11303 parser->thisclose = proto->thisclose;
11304 parser->thismad = proto->thismad;
11305 parser->thisopen = proto->thisopen;
11306 parser->thisstuff = proto->thisstuff;
11307 parser->thistoken = proto->thistoken;
11308 parser->thiswhite = proto->thiswhite;
11310 Copy(proto->nexttoke, parser->nexttoke, 5, NEXTTOKE);
11311 parser->curforce = proto->curforce;
11313 Copy(proto->nextval, parser->nextval, 5, YYSTYPE);
11314 Copy(proto->nexttype, parser->nexttype, 5, I32);
11315 parser->nexttoke = proto->nexttoke;
11318 /* XXX should clone saved_curcop here, but we aren't passed
11319 * proto_perl; so do it in perl_clone_using instead */
11325 /* duplicate a file handle */
11328 Perl_fp_dup(pTHX_ PerlIO *const fp, const char type, CLONE_PARAMS *const param)
11332 PERL_ARGS_ASSERT_FP_DUP;
11333 PERL_UNUSED_ARG(type);
11336 return (PerlIO*)NULL;
11338 /* look for it in the table first */
11339 ret = (PerlIO*)ptr_table_fetch(PL_ptr_table, fp);
11343 /* create anew and remember what it is */
11344 ret = PerlIO_fdupopen(aTHX_ fp, param, PERLIO_DUP_CLONE);
11345 ptr_table_store(PL_ptr_table, fp, ret);
11349 /* duplicate a directory handle */
11352 Perl_dirp_dup(pTHX_ DIR *const dp, CLONE_PARAMS *const param)
11358 register const Direntry_t *dirent;
11359 char smallbuf[256];
11365 PERL_UNUSED_CONTEXT;
11366 PERL_ARGS_ASSERT_DIRP_DUP;
11371 /* look for it in the table first */
11372 ret = (DIR*)ptr_table_fetch(PL_ptr_table, dp);
11378 PERL_UNUSED_ARG(param);
11382 /* open the current directory (so we can switch back) */
11383 if (!(pwd = PerlDir_open("."))) return (DIR *)NULL;
11385 /* chdir to our dir handle and open the present working directory */
11386 if (fchdir(my_dirfd(dp)) < 0 || !(ret = PerlDir_open("."))) {
11387 PerlDir_close(pwd);
11388 return (DIR *)NULL;
11390 /* Now we should have two dir handles pointing to the same dir. */
11392 /* Be nice to the calling code and chdir back to where we were. */
11393 fchdir(my_dirfd(pwd)); /* If this fails, then what? */
11395 /* We have no need of the pwd handle any more. */
11396 PerlDir_close(pwd);
11399 # define d_namlen(d) (d)->d_namlen
11401 # define d_namlen(d) strlen((d)->d_name)
11403 /* Iterate once through dp, to get the file name at the current posi-
11404 tion. Then step back. */
11405 pos = PerlDir_tell(dp);
11406 if ((dirent = PerlDir_read(dp))) {
11407 len = d_namlen(dirent);
11408 if (len <= sizeof smallbuf) name = smallbuf;
11409 else Newx(name, len, char);
11410 Move(dirent->d_name, name, len, char);
11412 PerlDir_seek(dp, pos);
11414 /* Iterate through the new dir handle, till we find a file with the
11416 if (!dirent) /* just before the end */
11418 pos = PerlDir_tell(ret);
11419 if (PerlDir_read(ret)) continue; /* not there yet */
11420 PerlDir_seek(ret, pos); /* step back */
11424 const long pos0 = PerlDir_tell(ret);
11426 pos = PerlDir_tell(ret);
11427 if ((dirent = PerlDir_read(ret))) {
11428 if (len == d_namlen(dirent)
11429 && memEQ(name, dirent->d_name, len)) {
11431 PerlDir_seek(ret, pos); /* step back */
11434 /* else we are not there yet; keep iterating */
11436 else { /* This is not meant to happen. The best we can do is
11437 reset the iterator to the beginning. */
11438 PerlDir_seek(ret, pos0);
11445 if (name && name != smallbuf)
11450 ret = win32_dirp_dup(dp, param);
11453 /* pop it in the pointer table */
11455 ptr_table_store(PL_ptr_table, dp, ret);
11460 /* duplicate a typeglob */
11463 Perl_gp_dup(pTHX_ GP *const gp, CLONE_PARAMS *const param)
11467 PERL_ARGS_ASSERT_GP_DUP;
11471 /* look for it in the table first */
11472 ret = (GP*)ptr_table_fetch(PL_ptr_table, gp);
11476 /* create anew and remember what it is */
11478 ptr_table_store(PL_ptr_table, gp, ret);
11481 /* ret->gp_refcnt must be 0 before any other dups are called. We're relying
11482 on Newxz() to do this for us. */
11483 ret->gp_sv = sv_dup_inc(gp->gp_sv, param);
11484 ret->gp_io = io_dup_inc(gp->gp_io, param);
11485 ret->gp_form = cv_dup_inc(gp->gp_form, param);
11486 ret->gp_av = av_dup_inc(gp->gp_av, param);
11487 ret->gp_hv = hv_dup_inc(gp->gp_hv, param);
11488 ret->gp_egv = gv_dup(gp->gp_egv, param);/* GvEGV is not refcounted */
11489 ret->gp_cv = cv_dup_inc(gp->gp_cv, param);
11490 ret->gp_cvgen = gp->gp_cvgen;
11491 ret->gp_line = gp->gp_line;
11492 ret->gp_file_hek = hek_dup(gp->gp_file_hek, param);
11496 /* duplicate a chain of magic */
11499 Perl_mg_dup(pTHX_ MAGIC *mg, CLONE_PARAMS *const param)
11501 MAGIC *mgret = NULL;
11502 MAGIC **mgprev_p = &mgret;
11504 PERL_ARGS_ASSERT_MG_DUP;
11506 for (; mg; mg = mg->mg_moremagic) {
11509 if ((param->flags & CLONEf_JOIN_IN)
11510 && mg->mg_type == PERL_MAGIC_backref)
11511 /* when joining, we let the individual SVs add themselves to
11512 * backref as needed. */
11515 Newx(nmg, 1, MAGIC);
11517 mgprev_p = &(nmg->mg_moremagic);
11519 /* There was a comment "XXX copy dynamic vtable?" but as we don't have
11520 dynamic vtables, I'm not sure why Sarathy wrote it. The comment dates
11521 from the original commit adding Perl_mg_dup() - revision 4538.
11522 Similarly there is the annotation "XXX random ptr?" next to the
11523 assignment to nmg->mg_ptr. */
11526 /* FIXME for plugins
11527 if (nmg->mg_type == PERL_MAGIC_qr) {
11528 nmg->mg_obj = MUTABLE_SV(CALLREGDUPE((REGEXP*)nmg->mg_obj, param));
11532 nmg->mg_obj = (nmg->mg_flags & MGf_REFCOUNTED)
11533 ? nmg->mg_type == PERL_MAGIC_backref
11534 /* The backref AV has its reference
11535 * count deliberately bumped by 1 */
11536 ? SvREFCNT_inc(av_dup_inc((const AV *)
11537 nmg->mg_obj, param))
11538 : sv_dup_inc(nmg->mg_obj, param)
11539 : sv_dup(nmg->mg_obj, param);
11541 if (nmg->mg_ptr && nmg->mg_type != PERL_MAGIC_regex_global) {
11542 if (nmg->mg_len > 0) {
11543 nmg->mg_ptr = SAVEPVN(nmg->mg_ptr, nmg->mg_len);
11544 if (nmg->mg_type == PERL_MAGIC_overload_table &&
11545 AMT_AMAGIC((AMT*)nmg->mg_ptr))
11547 AMT * const namtp = (AMT*)nmg->mg_ptr;
11548 sv_dup_inc_multiple((SV**)(namtp->table),
11549 (SV**)(namtp->table), NofAMmeth, param);
11552 else if (nmg->mg_len == HEf_SVKEY)
11553 nmg->mg_ptr = (char*)sv_dup_inc((const SV *)nmg->mg_ptr, param);
11555 if ((nmg->mg_flags & MGf_DUP) && nmg->mg_virtual && nmg->mg_virtual->svt_dup) {
11556 nmg->mg_virtual->svt_dup(aTHX_ nmg, param);
11562 #endif /* USE_ITHREADS */
11564 struct ptr_tbl_arena {
11565 struct ptr_tbl_arena *next;
11566 struct ptr_tbl_ent array[1023/3]; /* as ptr_tbl_ent has 3 pointers. */
11569 /* create a new pointer-mapping table */
11572 Perl_ptr_table_new(pTHX)
11575 PERL_UNUSED_CONTEXT;
11577 Newx(tbl, 1, PTR_TBL_t);
11578 tbl->tbl_max = 511;
11579 tbl->tbl_items = 0;
11580 tbl->tbl_arena = NULL;
11581 tbl->tbl_arena_next = NULL;
11582 tbl->tbl_arena_end = NULL;
11583 Newxz(tbl->tbl_ary, tbl->tbl_max + 1, PTR_TBL_ENT_t*);
11587 #define PTR_TABLE_HASH(ptr) \
11588 ((PTR2UV(ptr) >> 3) ^ (PTR2UV(ptr) >> (3 + 7)) ^ (PTR2UV(ptr) >> (3 + 17)))
11590 /* map an existing pointer using a table */
11592 STATIC PTR_TBL_ENT_t *
11593 S_ptr_table_find(PTR_TBL_t *const tbl, const void *const sv)
11595 PTR_TBL_ENT_t *tblent;
11596 const UV hash = PTR_TABLE_HASH(sv);
11598 PERL_ARGS_ASSERT_PTR_TABLE_FIND;
11600 tblent = tbl->tbl_ary[hash & tbl->tbl_max];
11601 for (; tblent; tblent = tblent->next) {
11602 if (tblent->oldval == sv)
11609 Perl_ptr_table_fetch(pTHX_ PTR_TBL_t *const tbl, const void *const sv)
11611 PTR_TBL_ENT_t const *const tblent = ptr_table_find(tbl, sv);
11613 PERL_ARGS_ASSERT_PTR_TABLE_FETCH;
11614 PERL_UNUSED_CONTEXT;
11616 return tblent ? tblent->newval : NULL;
11619 /* add a new entry to a pointer-mapping table */
11622 Perl_ptr_table_store(pTHX_ PTR_TBL_t *const tbl, const void *const oldsv, void *const newsv)
11624 PTR_TBL_ENT_t *tblent = ptr_table_find(tbl, oldsv);
11626 PERL_ARGS_ASSERT_PTR_TABLE_STORE;
11627 PERL_UNUSED_CONTEXT;
11630 tblent->newval = newsv;
11632 const UV entry = PTR_TABLE_HASH(oldsv) & tbl->tbl_max;
11634 if (tbl->tbl_arena_next == tbl->tbl_arena_end) {
11635 struct ptr_tbl_arena *new_arena;
11637 Newx(new_arena, 1, struct ptr_tbl_arena);
11638 new_arena->next = tbl->tbl_arena;
11639 tbl->tbl_arena = new_arena;
11640 tbl->tbl_arena_next = new_arena->array;
11641 tbl->tbl_arena_end = new_arena->array
11642 + sizeof(new_arena->array) / sizeof(new_arena->array[0]);
11645 tblent = tbl->tbl_arena_next++;
11647 tblent->oldval = oldsv;
11648 tblent->newval = newsv;
11649 tblent->next = tbl->tbl_ary[entry];
11650 tbl->tbl_ary[entry] = tblent;
11652 if (tblent->next && tbl->tbl_items > tbl->tbl_max)
11653 ptr_table_split(tbl);
11657 /* double the hash bucket size of an existing ptr table */
11660 Perl_ptr_table_split(pTHX_ PTR_TBL_t *const tbl)
11662 PTR_TBL_ENT_t **ary = tbl->tbl_ary;
11663 const UV oldsize = tbl->tbl_max + 1;
11664 UV newsize = oldsize * 2;
11667 PERL_ARGS_ASSERT_PTR_TABLE_SPLIT;
11668 PERL_UNUSED_CONTEXT;
11670 Renew(ary, newsize, PTR_TBL_ENT_t*);
11671 Zero(&ary[oldsize], newsize-oldsize, PTR_TBL_ENT_t*);
11672 tbl->tbl_max = --newsize;
11673 tbl->tbl_ary = ary;
11674 for (i=0; i < oldsize; i++, ary++) {
11675 PTR_TBL_ENT_t **entp = ary;
11676 PTR_TBL_ENT_t *ent = *ary;
11677 PTR_TBL_ENT_t **curentp;
11680 curentp = ary + oldsize;
11682 if ((newsize & PTR_TABLE_HASH(ent->oldval)) != i) {
11684 ent->next = *curentp;
11694 /* remove all the entries from a ptr table */
11695 /* Deprecated - will be removed post 5.14 */
11698 Perl_ptr_table_clear(pTHX_ PTR_TBL_t *const tbl)
11700 if (tbl && tbl->tbl_items) {
11701 struct ptr_tbl_arena *arena = tbl->tbl_arena;
11703 Zero(tbl->tbl_ary, tbl->tbl_max + 1, struct ptr_tbl_ent **);
11706 struct ptr_tbl_arena *next = arena->next;
11712 tbl->tbl_items = 0;
11713 tbl->tbl_arena = NULL;
11714 tbl->tbl_arena_next = NULL;
11715 tbl->tbl_arena_end = NULL;
11719 /* clear and free a ptr table */
11722 Perl_ptr_table_free(pTHX_ PTR_TBL_t *const tbl)
11724 struct ptr_tbl_arena *arena;
11730 arena = tbl->tbl_arena;
11733 struct ptr_tbl_arena *next = arena->next;
11739 Safefree(tbl->tbl_ary);
11743 #if defined(USE_ITHREADS)
11746 Perl_rvpv_dup(pTHX_ SV *const dstr, const SV *const sstr, CLONE_PARAMS *const param)
11748 PERL_ARGS_ASSERT_RVPV_DUP;
11751 if (SvWEAKREF(sstr)) {
11752 SvRV_set(dstr, sv_dup(SvRV_const(sstr), param));
11753 if (param->flags & CLONEf_JOIN_IN) {
11754 /* if joining, we add any back references individually rather
11755 * than copying the whole backref array */
11756 Perl_sv_add_backref(aTHX_ SvRV(dstr), dstr);
11760 SvRV_set(dstr, sv_dup_inc(SvRV_const(sstr), param));
11762 else if (SvPVX_const(sstr)) {
11763 /* Has something there */
11765 /* Normal PV - clone whole allocated space */
11766 SvPV_set(dstr, SAVEPVN(SvPVX_const(sstr), SvLEN(sstr)-1));
11767 if (SvREADONLY(sstr) && SvFAKE(sstr)) {
11768 /* Not that normal - actually sstr is copy on write.
11769 But we are a true, independent SV, so: */
11770 SvREADONLY_off(dstr);
11775 /* Special case - not normally malloced for some reason */
11776 if (isGV_with_GP(sstr)) {
11777 /* Don't need to do anything here. */
11779 else if ((SvREADONLY(sstr) && SvFAKE(sstr))) {
11780 /* A "shared" PV - clone it as "shared" PV */
11782 HEK_KEY(hek_dup(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)),
11786 /* Some other special case - random pointer */
11787 SvPV_set(dstr, (char *) SvPVX_const(sstr));
11792 /* Copy the NULL */
11793 SvPV_set(dstr, NULL);
11797 /* duplicate a list of SVs. source and dest may point to the same memory. */
11799 S_sv_dup_inc_multiple(pTHX_ SV *const *source, SV **dest,
11800 SSize_t items, CLONE_PARAMS *const param)
11802 PERL_ARGS_ASSERT_SV_DUP_INC_MULTIPLE;
11804 while (items-- > 0) {
11805 *dest++ = sv_dup_inc(*source++, param);
11811 /* duplicate an SV of any type (including AV, HV etc) */
11814 S_sv_dup_common(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
11819 PERL_ARGS_ASSERT_SV_DUP_COMMON;
11821 if (SvTYPE(sstr) == (svtype)SVTYPEMASK) {
11822 #ifdef DEBUG_LEAKING_SCALARS_ABORT
11827 /* look for it in the table first */
11828 dstr = MUTABLE_SV(ptr_table_fetch(PL_ptr_table, sstr));
11832 if(param->flags & CLONEf_JOIN_IN) {
11833 /** We are joining here so we don't want do clone
11834 something that is bad **/
11835 if (SvTYPE(sstr) == SVt_PVHV) {
11836 const HEK * const hvname = HvNAME_HEK(sstr);
11838 /** don't clone stashes if they already exist **/
11839 dstr = MUTABLE_SV(gv_stashpvn(HEK_KEY(hvname), HEK_LEN(hvname),
11840 HEK_UTF8(hvname) ? SVf_UTF8 : 0));
11841 ptr_table_store(PL_ptr_table, sstr, dstr);
11845 else if (SvTYPE(sstr) == SVt_PVGV && !SvFAKE(sstr)) {
11846 HV *stash = GvSTASH(sstr);
11847 const HEK * hvname;
11848 if (stash && (hvname = HvNAME_HEK(stash))) {
11849 /** don't clone GVs if they already exist **/
11851 stash = gv_stashpvn(HEK_KEY(hvname), HEK_LEN(hvname),
11852 HEK_UTF8(hvname) ? SVf_UTF8 : 0);
11854 stash, GvNAME(sstr),
11860 if (svp && *svp && SvTYPE(*svp) == SVt_PVGV) {
11861 ptr_table_store(PL_ptr_table, sstr, *svp);
11868 /* create anew and remember what it is */
11871 #ifdef DEBUG_LEAKING_SCALARS
11872 dstr->sv_debug_optype = sstr->sv_debug_optype;
11873 dstr->sv_debug_line = sstr->sv_debug_line;
11874 dstr->sv_debug_inpad = sstr->sv_debug_inpad;
11875 dstr->sv_debug_parent = (SV*)sstr;
11876 FREE_SV_DEBUG_FILE(dstr);
11877 dstr->sv_debug_file = savepv(sstr->sv_debug_file);
11880 ptr_table_store(PL_ptr_table, sstr, dstr);
11883 SvFLAGS(dstr) = SvFLAGS(sstr);
11884 SvFLAGS(dstr) &= ~SVf_OOK; /* don't propagate OOK hack */
11885 SvREFCNT(dstr) = 0; /* must be before any other dups! */
11888 if (SvANY(sstr) && PL_watch_pvx && SvPVX_const(sstr) == PL_watch_pvx)
11889 PerlIO_printf(Perl_debug_log, "watch at %p hit, found string \"%s\"\n",
11890 (void*)PL_watch_pvx, SvPVX_const(sstr));
11893 /* don't clone objects whose class has asked us not to */
11894 if (SvOBJECT(sstr) && ! (SvFLAGS(SvSTASH(sstr)) & SVphv_CLONEABLE)) {
11899 switch (SvTYPE(sstr)) {
11901 SvANY(dstr) = NULL;
11904 SvANY(dstr) = (XPVIV*)((char*)&(dstr->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
11906 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
11908 SvIV_set(dstr, SvIVX(sstr));
11912 SvANY(dstr) = new_XNV();
11913 SvNV_set(dstr, SvNVX(sstr));
11915 /* case SVt_BIND: */
11918 /* These are all the types that need complex bodies allocating. */
11920 const svtype sv_type = SvTYPE(sstr);
11921 const struct body_details *const sv_type_details
11922 = bodies_by_type + sv_type;
11926 Perl_croak(aTHX_ "Bizarre SvTYPE [%" IVdf "]", (IV)SvTYPE(sstr));
11941 assert(sv_type_details->body_size);
11942 if (sv_type_details->arena) {
11943 new_body_inline(new_body, sv_type);
11945 = (void*)((char*)new_body - sv_type_details->offset);
11947 new_body = new_NOARENA(sv_type_details);
11951 SvANY(dstr) = new_body;
11954 Copy(((char*)SvANY(sstr)) + sv_type_details->offset,
11955 ((char*)SvANY(dstr)) + sv_type_details->offset,
11956 sv_type_details->copy, char);
11958 Copy(((char*)SvANY(sstr)),
11959 ((char*)SvANY(dstr)),
11960 sv_type_details->body_size + sv_type_details->offset, char);
11963 if (sv_type != SVt_PVAV && sv_type != SVt_PVHV
11964 && !isGV_with_GP(dstr)
11965 && !(sv_type == SVt_PVIO && !(IoFLAGS(dstr) & IOf_FAKE_DIRP)))
11966 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
11968 /* The Copy above means that all the source (unduplicated) pointers
11969 are now in the destination. We can check the flags and the
11970 pointers in either, but it's possible that there's less cache
11971 missing by always going for the destination.
11972 FIXME - instrument and check that assumption */
11973 if (sv_type >= SVt_PVMG) {
11974 if ((sv_type == SVt_PVMG) && SvPAD_OUR(dstr)) {
11975 SvOURSTASH_set(dstr, hv_dup_inc(SvOURSTASH(dstr), param));
11976 } else if (SvMAGIC(dstr))
11977 SvMAGIC_set(dstr, mg_dup(SvMAGIC(dstr), param));
11979 SvSTASH_set(dstr, hv_dup_inc(SvSTASH(dstr), param));
11982 /* The cast silences a GCC warning about unhandled types. */
11983 switch ((int)sv_type) {
11993 /* FIXME for plugins */
11994 re_dup_guts((REGEXP*) sstr, (REGEXP*) dstr, param);
11997 /* XXX LvTARGOFF sometimes holds PMOP* when DEBUGGING */
11998 if (LvTYPE(dstr) == 't') /* for tie: unrefcnted fake (SV**) */
11999 LvTARG(dstr) = dstr;
12000 else if (LvTYPE(dstr) == 'T') /* for tie: fake HE */
12001 LvTARG(dstr) = MUTABLE_SV(he_dup((HE*)LvTARG(dstr), 0, param));
12003 LvTARG(dstr) = sv_dup_inc(LvTARG(dstr), param);
12005 /* non-GP case already handled above */
12006 if(isGV_with_GP(sstr)) {
12007 GvNAME_HEK(dstr) = hek_dup(GvNAME_HEK(dstr), param);
12008 /* Don't call sv_add_backref here as it's going to be
12009 created as part of the magic cloning of the symbol
12010 table--unless this is during a join and the stash
12011 is not actually being cloned. */
12012 /* Danger Will Robinson - GvGP(dstr) isn't initialised
12013 at the point of this comment. */
12014 GvSTASH(dstr) = hv_dup(GvSTASH(dstr), param);
12015 if (param->flags & CLONEf_JOIN_IN)
12016 Perl_sv_add_backref(aTHX_ MUTABLE_SV(GvSTASH(dstr)), dstr);
12017 GvGP_set(dstr, gp_dup(GvGP(sstr), param));
12018 (void)GpREFCNT_inc(GvGP(dstr));
12022 /* PL_parser->rsfp_filters entries have fake IoDIRP() */
12023 if(IoFLAGS(dstr) & IOf_FAKE_DIRP) {
12024 /* I have no idea why fake dirp (rsfps)
12025 should be treated differently but otherwise
12026 we end up with leaks -- sky*/
12027 IoTOP_GV(dstr) = gv_dup_inc(IoTOP_GV(dstr), param);
12028 IoFMT_GV(dstr) = gv_dup_inc(IoFMT_GV(dstr), param);
12029 IoBOTTOM_GV(dstr) = gv_dup_inc(IoBOTTOM_GV(dstr), param);
12031 IoTOP_GV(dstr) = gv_dup(IoTOP_GV(dstr), param);
12032 IoFMT_GV(dstr) = gv_dup(IoFMT_GV(dstr), param);
12033 IoBOTTOM_GV(dstr) = gv_dup(IoBOTTOM_GV(dstr), param);
12034 if (IoDIRP(dstr)) {
12035 IoDIRP(dstr) = dirp_dup(IoDIRP(dstr), param);
12038 /* IoDIRP(dstr) is already a copy of IoDIRP(sstr) */
12040 IoIFP(dstr) = fp_dup(IoIFP(sstr), IoTYPE(dstr), param);
12042 if (IoOFP(dstr) == IoIFP(sstr))
12043 IoOFP(dstr) = IoIFP(dstr);
12045 IoOFP(dstr) = fp_dup(IoOFP(dstr), IoTYPE(dstr), param);
12046 IoTOP_NAME(dstr) = SAVEPV(IoTOP_NAME(dstr));
12047 IoFMT_NAME(dstr) = SAVEPV(IoFMT_NAME(dstr));
12048 IoBOTTOM_NAME(dstr) = SAVEPV(IoBOTTOM_NAME(dstr));
12051 /* avoid cloning an empty array */
12052 if (AvARRAY((const AV *)sstr) && AvFILLp((const AV *)sstr) >= 0) {
12053 SV **dst_ary, **src_ary;
12054 SSize_t items = AvFILLp((const AV *)sstr) + 1;
12056 src_ary = AvARRAY((const AV *)sstr);
12057 Newxz(dst_ary, AvMAX((const AV *)sstr)+1, SV*);
12058 ptr_table_store(PL_ptr_table, src_ary, dst_ary);
12059 AvARRAY(MUTABLE_AV(dstr)) = dst_ary;
12060 AvALLOC((const AV *)dstr) = dst_ary;
12061 if (AvREAL((const AV *)sstr)) {
12062 dst_ary = sv_dup_inc_multiple(src_ary, dst_ary, items,
12066 while (items-- > 0)
12067 *dst_ary++ = sv_dup(*src_ary++, param);
12069 items = AvMAX((const AV *)sstr) - AvFILLp((const AV *)sstr);
12070 while (items-- > 0) {
12071 *dst_ary++ = &PL_sv_undef;
12075 AvARRAY(MUTABLE_AV(dstr)) = NULL;
12076 AvALLOC((const AV *)dstr) = (SV**)NULL;
12077 AvMAX( (const AV *)dstr) = -1;
12078 AvFILLp((const AV *)dstr) = -1;
12082 if (HvARRAY((const HV *)sstr)) {
12084 const bool sharekeys = !!HvSHAREKEYS(sstr);
12085 XPVHV * const dxhv = (XPVHV*)SvANY(dstr);
12086 XPVHV * const sxhv = (XPVHV*)SvANY(sstr);
12088 Newx(darray, PERL_HV_ARRAY_ALLOC_BYTES(dxhv->xhv_max+1)
12089 + (SvOOK(sstr) ? sizeof(struct xpvhv_aux) : 0),
12091 HvARRAY(dstr) = (HE**)darray;
12092 while (i <= sxhv->xhv_max) {
12093 const HE * const source = HvARRAY(sstr)[i];
12094 HvARRAY(dstr)[i] = source
12095 ? he_dup(source, sharekeys, param) : 0;
12099 const struct xpvhv_aux * const saux = HvAUX(sstr);
12100 struct xpvhv_aux * const daux = HvAUX(dstr);
12101 /* This flag isn't copied. */
12104 if (saux->xhv_name_count) {
12105 HEK ** const sname = saux->xhv_name_u.xhvnameu_names;
12107 = saux->xhv_name_count < 0
12108 ? -saux->xhv_name_count
12109 : saux->xhv_name_count;
12110 HEK **shekp = sname + count;
12112 Newx(daux->xhv_name_u.xhvnameu_names, count, HEK *);
12113 dhekp = daux->xhv_name_u.xhvnameu_names + count;
12114 while (shekp-- > sname) {
12116 *dhekp = hek_dup(*shekp, param);
12120 daux->xhv_name_u.xhvnameu_name
12121 = hek_dup(saux->xhv_name_u.xhvnameu_name,
12124 daux->xhv_name_count = saux->xhv_name_count;
12126 daux->xhv_riter = saux->xhv_riter;
12127 daux->xhv_eiter = saux->xhv_eiter
12128 ? he_dup(saux->xhv_eiter,
12129 cBOOL(HvSHAREKEYS(sstr)), param) : 0;
12130 /* backref array needs refcnt=2; see sv_add_backref */
12131 daux->xhv_backreferences =
12132 (param->flags & CLONEf_JOIN_IN)
12133 /* when joining, we let the individual GVs and
12134 * CVs add themselves to backref as
12135 * needed. This avoids pulling in stuff
12136 * that isn't required, and simplifies the
12137 * case where stashes aren't cloned back
12138 * if they already exist in the parent
12141 : saux->xhv_backreferences
12142 ? (SvTYPE(saux->xhv_backreferences) == SVt_PVAV)
12143 ? MUTABLE_AV(SvREFCNT_inc(
12144 sv_dup_inc((const SV *)
12145 saux->xhv_backreferences, param)))
12146 : MUTABLE_AV(sv_dup((const SV *)
12147 saux->xhv_backreferences, param))
12150 daux->xhv_mro_meta = saux->xhv_mro_meta
12151 ? mro_meta_dup(saux->xhv_mro_meta, param)
12154 /* Record stashes for possible cloning in Perl_clone(). */
12156 av_push(param->stashes, dstr);
12160 HvARRAY(MUTABLE_HV(dstr)) = NULL;
12163 if (!(param->flags & CLONEf_COPY_STACKS)) {
12168 /* NOTE: not refcounted */
12169 SvANY(MUTABLE_CV(dstr))->xcv_stash =
12170 hv_dup(CvSTASH(dstr), param);
12171 if ((param->flags & CLONEf_JOIN_IN) && CvSTASH(dstr))
12172 Perl_sv_add_backref(aTHX_ MUTABLE_SV(CvSTASH(dstr)), dstr);
12173 if (!CvISXSUB(dstr)) {
12175 CvROOT(dstr) = OpREFCNT_inc(CvROOT(dstr));
12177 } else if (CvCONST(dstr)) {
12178 CvXSUBANY(dstr).any_ptr =
12179 sv_dup_inc((const SV *)CvXSUBANY(dstr).any_ptr, param);
12181 if (CvDYNFILE(dstr)) CvFILE(dstr) = SAVEPV(CvFILE(dstr));
12182 /* don't dup if copying back - CvGV isn't refcounted, so the
12183 * duped GV may never be freed. A bit of a hack! DAPM */
12184 SvANY(MUTABLE_CV(dstr))->xcv_gv =
12186 ? gv_dup_inc(CvGV(sstr), param)
12187 : (param->flags & CLONEf_JOIN_IN)
12189 : gv_dup(CvGV(sstr), param);
12191 CvPADLIST(dstr) = padlist_dup(CvPADLIST(sstr), param);
12193 CvWEAKOUTSIDE(sstr)
12194 ? cv_dup( CvOUTSIDE(dstr), param)
12195 : cv_dup_inc(CvOUTSIDE(dstr), param);
12201 if (SvOBJECT(dstr) && SvTYPE(dstr) != SVt_PVIO)
12208 Perl_sv_dup_inc(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
12210 PERL_ARGS_ASSERT_SV_DUP_INC;
12211 return sstr ? SvREFCNT_inc(sv_dup_common(sstr, param)) : NULL;
12215 Perl_sv_dup(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
12217 SV *dstr = sstr ? sv_dup_common(sstr, param) : NULL;
12218 PERL_ARGS_ASSERT_SV_DUP;
12220 /* Track every SV that (at least initially) had a reference count of 0.
12221 We need to do this by holding an actual reference to it in this array.
12222 If we attempt to cheat, turn AvREAL_off(), and store only pointers
12223 (akin to the stashes hash, and the perl stack), we come unstuck if
12224 a weak reference (or other SV legitimately SvREFCNT() == 0 for this
12225 thread) is manipulated in a CLONE method, because CLONE runs before the
12226 unreferenced array is walked to find SVs still with SvREFCNT() == 0
12227 (and fix things up by giving each a reference via the temps stack).
12228 Instead, during CLONE, if the 0-referenced SV has SvREFCNT_inc() and
12229 then SvREFCNT_dec(), it will be cleaned up (and added to the free list)
12230 before the walk of unreferenced happens and a reference to that is SV
12231 added to the temps stack. At which point we have the same SV considered
12232 to be in use, and free to be re-used. Not good.
12234 if (dstr && !(param->flags & CLONEf_COPY_STACKS) && !SvREFCNT(dstr)) {
12235 assert(param->unreferenced);
12236 av_push(param->unreferenced, SvREFCNT_inc(dstr));
12242 /* duplicate a context */
12245 Perl_cx_dup(pTHX_ PERL_CONTEXT *cxs, I32 ix, I32 max, CLONE_PARAMS* param)
12247 PERL_CONTEXT *ncxs;
12249 PERL_ARGS_ASSERT_CX_DUP;
12252 return (PERL_CONTEXT*)NULL;
12254 /* look for it in the table first */
12255 ncxs = (PERL_CONTEXT*)ptr_table_fetch(PL_ptr_table, cxs);
12259 /* create anew and remember what it is */
12260 Newx(ncxs, max + 1, PERL_CONTEXT);
12261 ptr_table_store(PL_ptr_table, cxs, ncxs);
12262 Copy(cxs, ncxs, max + 1, PERL_CONTEXT);
12265 PERL_CONTEXT * const ncx = &ncxs[ix];
12266 if (CxTYPE(ncx) == CXt_SUBST) {
12267 Perl_croak(aTHX_ "Cloning substitution context is unimplemented");
12270 switch (CxTYPE(ncx)) {
12272 ncx->blk_sub.cv = (ncx->blk_sub.olddepth == 0
12273 ? cv_dup_inc(ncx->blk_sub.cv, param)
12274 : cv_dup(ncx->blk_sub.cv,param));
12275 ncx->blk_sub.argarray = (CxHASARGS(ncx)
12276 ? av_dup_inc(ncx->blk_sub.argarray,
12279 ncx->blk_sub.savearray = av_dup_inc(ncx->blk_sub.savearray,
12281 ncx->blk_sub.oldcomppad = (PAD*)ptr_table_fetch(PL_ptr_table,
12282 ncx->blk_sub.oldcomppad);
12285 ncx->blk_eval.old_namesv = sv_dup_inc(ncx->blk_eval.old_namesv,
12287 ncx->blk_eval.cur_text = sv_dup(ncx->blk_eval.cur_text, param);
12289 case CXt_LOOP_LAZYSV:
12290 ncx->blk_loop.state_u.lazysv.end
12291 = sv_dup_inc(ncx->blk_loop.state_u.lazysv.end, param);
12292 /* We are taking advantage of av_dup_inc and sv_dup_inc
12293 actually being the same function, and order equivalence of
12295 We can assert the later [but only at run time :-(] */
12296 assert ((void *) &ncx->blk_loop.state_u.ary.ary ==
12297 (void *) &ncx->blk_loop.state_u.lazysv.cur);
12299 ncx->blk_loop.state_u.ary.ary
12300 = av_dup_inc(ncx->blk_loop.state_u.ary.ary, param);
12301 case CXt_LOOP_LAZYIV:
12302 case CXt_LOOP_PLAIN:
12303 if (CxPADLOOP(ncx)) {
12304 ncx->blk_loop.itervar_u.oldcomppad
12305 = (PAD*)ptr_table_fetch(PL_ptr_table,
12306 ncx->blk_loop.itervar_u.oldcomppad);
12308 ncx->blk_loop.itervar_u.gv
12309 = gv_dup((const GV *)ncx->blk_loop.itervar_u.gv,
12314 ncx->blk_format.cv = cv_dup(ncx->blk_format.cv, param);
12315 ncx->blk_format.gv = gv_dup(ncx->blk_format.gv, param);
12316 ncx->blk_format.dfoutgv = gv_dup_inc(ncx->blk_format.dfoutgv,
12329 /* duplicate a stack info structure */
12332 Perl_si_dup(pTHX_ PERL_SI *si, CLONE_PARAMS* param)
12336 PERL_ARGS_ASSERT_SI_DUP;
12339 return (PERL_SI*)NULL;
12341 /* look for it in the table first */
12342 nsi = (PERL_SI*)ptr_table_fetch(PL_ptr_table, si);
12346 /* create anew and remember what it is */
12347 Newxz(nsi, 1, PERL_SI);
12348 ptr_table_store(PL_ptr_table, si, nsi);
12350 nsi->si_stack = av_dup_inc(si->si_stack, param);
12351 nsi->si_cxix = si->si_cxix;
12352 nsi->si_cxmax = si->si_cxmax;
12353 nsi->si_cxstack = cx_dup(si->si_cxstack, si->si_cxix, si->si_cxmax, param);
12354 nsi->si_type = si->si_type;
12355 nsi->si_prev = si_dup(si->si_prev, param);
12356 nsi->si_next = si_dup(si->si_next, param);
12357 nsi->si_markoff = si->si_markoff;
12362 #define POPINT(ss,ix) ((ss)[--(ix)].any_i32)
12363 #define TOPINT(ss,ix) ((ss)[ix].any_i32)
12364 #define POPLONG(ss,ix) ((ss)[--(ix)].any_long)
12365 #define TOPLONG(ss,ix) ((ss)[ix].any_long)
12366 #define POPIV(ss,ix) ((ss)[--(ix)].any_iv)
12367 #define TOPIV(ss,ix) ((ss)[ix].any_iv)
12368 #define POPUV(ss,ix) ((ss)[--(ix)].any_uv)
12369 #define TOPUV(ss,ix) ((ss)[ix].any_uv)
12370 #define POPBOOL(ss,ix) ((ss)[--(ix)].any_bool)
12371 #define TOPBOOL(ss,ix) ((ss)[ix].any_bool)
12372 #define POPPTR(ss,ix) ((ss)[--(ix)].any_ptr)
12373 #define TOPPTR(ss,ix) ((ss)[ix].any_ptr)
12374 #define POPDPTR(ss,ix) ((ss)[--(ix)].any_dptr)
12375 #define TOPDPTR(ss,ix) ((ss)[ix].any_dptr)
12376 #define POPDXPTR(ss,ix) ((ss)[--(ix)].any_dxptr)
12377 #define TOPDXPTR(ss,ix) ((ss)[ix].any_dxptr)
12380 #define pv_dup_inc(p) SAVEPV(p)
12381 #define pv_dup(p) SAVEPV(p)
12382 #define svp_dup_inc(p,pp) any_dup(p,pp)
12384 /* map any object to the new equivent - either something in the
12385 * ptr table, or something in the interpreter structure
12389 Perl_any_dup(pTHX_ void *v, const PerlInterpreter *proto_perl)
12393 PERL_ARGS_ASSERT_ANY_DUP;
12396 return (void*)NULL;
12398 /* look for it in the table first */
12399 ret = ptr_table_fetch(PL_ptr_table, v);
12403 /* see if it is part of the interpreter structure */
12404 if (v >= (void*)proto_perl && v < (void*)(proto_perl+1))
12405 ret = (void*)(((char*)aTHX) + (((char*)v) - (char*)proto_perl));
12413 /* duplicate the save stack */
12416 Perl_ss_dup(pTHX_ PerlInterpreter *proto_perl, CLONE_PARAMS* param)
12419 ANY * const ss = proto_perl->Isavestack;
12420 const I32 max = proto_perl->Isavestack_max;
12421 I32 ix = proto_perl->Isavestack_ix;
12434 void (*dptr) (void*);
12435 void (*dxptr) (pTHX_ void*);
12437 PERL_ARGS_ASSERT_SS_DUP;
12439 Newxz(nss, max, ANY);
12442 const UV uv = POPUV(ss,ix);
12443 const U8 type = (U8)uv & SAVE_MASK;
12445 TOPUV(nss,ix) = uv;
12447 case SAVEt_CLEARSV:
12449 case SAVEt_HELEM: /* hash element */
12450 sv = (const SV *)POPPTR(ss,ix);
12451 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12453 case SAVEt_ITEM: /* normal string */
12454 case SAVEt_GVSV: /* scalar slot in GV */
12455 case SAVEt_SV: /* scalar reference */
12456 sv = (const SV *)POPPTR(ss,ix);
12457 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12460 case SAVEt_MORTALIZESV:
12461 sv = (const SV *)POPPTR(ss,ix);
12462 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12464 case SAVEt_SHARED_PVREF: /* char* in shared space */
12465 c = (char*)POPPTR(ss,ix);
12466 TOPPTR(nss,ix) = savesharedpv(c);
12467 ptr = POPPTR(ss,ix);
12468 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12470 case SAVEt_GENERIC_SVREF: /* generic sv */
12471 case SAVEt_SVREF: /* scalar reference */
12472 sv = (const SV *)POPPTR(ss,ix);
12473 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12474 ptr = POPPTR(ss,ix);
12475 TOPPTR(nss,ix) = svp_dup_inc((SV**)ptr, proto_perl);/* XXXXX */
12477 case SAVEt_HV: /* hash reference */
12478 case SAVEt_AV: /* array reference */
12479 sv = (const SV *) POPPTR(ss,ix);
12480 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12482 case SAVEt_COMPPAD:
12484 sv = (const SV *) POPPTR(ss,ix);
12485 TOPPTR(nss,ix) = sv_dup(sv, param);
12487 case SAVEt_INT: /* int reference */
12488 ptr = POPPTR(ss,ix);
12489 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12490 intval = (int)POPINT(ss,ix);
12491 TOPINT(nss,ix) = intval;
12493 case SAVEt_LONG: /* long reference */
12494 ptr = POPPTR(ss,ix);
12495 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12496 longval = (long)POPLONG(ss,ix);
12497 TOPLONG(nss,ix) = longval;
12499 case SAVEt_I32: /* I32 reference */
12500 ptr = POPPTR(ss,ix);
12501 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12503 TOPINT(nss,ix) = i;
12505 case SAVEt_IV: /* IV reference */
12506 ptr = POPPTR(ss,ix);
12507 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12509 TOPIV(nss,ix) = iv;
12511 case SAVEt_HPTR: /* HV* reference */
12512 case SAVEt_APTR: /* AV* reference */
12513 case SAVEt_SPTR: /* SV* reference */
12514 ptr = POPPTR(ss,ix);
12515 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12516 sv = (const SV *)POPPTR(ss,ix);
12517 TOPPTR(nss,ix) = sv_dup(sv, param);
12519 case SAVEt_VPTR: /* random* reference */
12520 ptr = POPPTR(ss,ix);
12521 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12523 case SAVEt_INT_SMALL:
12524 case SAVEt_I32_SMALL:
12525 case SAVEt_I16: /* I16 reference */
12526 case SAVEt_I8: /* I8 reference */
12528 ptr = POPPTR(ss,ix);
12529 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12531 case SAVEt_GENERIC_PVREF: /* generic char* */
12532 case SAVEt_PPTR: /* char* reference */
12533 ptr = POPPTR(ss,ix);
12534 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12535 c = (char*)POPPTR(ss,ix);
12536 TOPPTR(nss,ix) = pv_dup(c);
12538 case SAVEt_GP: /* scalar reference */
12539 gp = (GP*)POPPTR(ss,ix);
12540 TOPPTR(nss,ix) = gp = gp_dup(gp, param);
12541 (void)GpREFCNT_inc(gp);
12542 gv = (const GV *)POPPTR(ss,ix);
12543 TOPPTR(nss,ix) = gv_dup_inc(gv, param);
12546 ptr = POPPTR(ss,ix);
12547 if (ptr && (((OP*)ptr)->op_private & OPpREFCOUNTED)) {
12548 /* these are assumed to be refcounted properly */
12550 switch (((OP*)ptr)->op_type) {
12552 case OP_LEAVESUBLV:
12556 case OP_LEAVEWRITE:
12557 TOPPTR(nss,ix) = ptr;
12560 (void) OpREFCNT_inc(o);
12564 TOPPTR(nss,ix) = NULL;
12569 TOPPTR(nss,ix) = NULL;
12571 case SAVEt_FREECOPHH:
12572 ptr = POPPTR(ss,ix);
12573 TOPPTR(nss,ix) = cophh_copy((COPHH *)ptr);
12576 hv = (const HV *)POPPTR(ss,ix);
12577 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
12579 TOPINT(nss,ix) = i;
12582 c = (char*)POPPTR(ss,ix);
12583 TOPPTR(nss,ix) = pv_dup_inc(c);
12585 case SAVEt_STACK_POS: /* Position on Perl stack */
12587 TOPINT(nss,ix) = i;
12589 case SAVEt_DESTRUCTOR:
12590 ptr = POPPTR(ss,ix);
12591 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
12592 dptr = POPDPTR(ss,ix);
12593 TOPDPTR(nss,ix) = DPTR2FPTR(void (*)(void*),
12594 any_dup(FPTR2DPTR(void *, dptr),
12597 case SAVEt_DESTRUCTOR_X:
12598 ptr = POPPTR(ss,ix);
12599 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
12600 dxptr = POPDXPTR(ss,ix);
12601 TOPDXPTR(nss,ix) = DPTR2FPTR(void (*)(pTHX_ void*),
12602 any_dup(FPTR2DPTR(void *, dxptr),
12605 case SAVEt_REGCONTEXT:
12607 ix -= uv >> SAVE_TIGHT_SHIFT;
12609 case SAVEt_AELEM: /* array element */
12610 sv = (const SV *)POPPTR(ss,ix);
12611 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12613 TOPINT(nss,ix) = i;
12614 av = (const AV *)POPPTR(ss,ix);
12615 TOPPTR(nss,ix) = av_dup_inc(av, param);
12618 ptr = POPPTR(ss,ix);
12619 TOPPTR(nss,ix) = ptr;
12622 ptr = POPPTR(ss,ix);
12623 ptr = cophh_copy((COPHH*)ptr);
12624 TOPPTR(nss,ix) = ptr;
12626 TOPINT(nss,ix) = i;
12627 if (i & HINT_LOCALIZE_HH) {
12628 hv = (const HV *)POPPTR(ss,ix);
12629 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
12632 case SAVEt_PADSV_AND_MORTALIZE:
12633 longval = (long)POPLONG(ss,ix);
12634 TOPLONG(nss,ix) = longval;
12635 ptr = POPPTR(ss,ix);
12636 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12637 sv = (const SV *)POPPTR(ss,ix);
12638 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12640 case SAVEt_SET_SVFLAGS:
12642 TOPINT(nss,ix) = i;
12644 TOPINT(nss,ix) = i;
12645 sv = (const SV *)POPPTR(ss,ix);
12646 TOPPTR(nss,ix) = sv_dup(sv, param);
12648 case SAVEt_RE_STATE:
12650 const struct re_save_state *const old_state
12651 = (struct re_save_state *)
12652 (ss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
12653 struct re_save_state *const new_state
12654 = (struct re_save_state *)
12655 (nss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
12657 Copy(old_state, new_state, 1, struct re_save_state);
12658 ix -= SAVESTACK_ALLOC_FOR_RE_SAVE_STATE;
12660 new_state->re_state_bostr
12661 = pv_dup(old_state->re_state_bostr);
12662 new_state->re_state_reginput
12663 = pv_dup(old_state->re_state_reginput);
12664 new_state->re_state_regeol
12665 = pv_dup(old_state->re_state_regeol);
12666 new_state->re_state_regoffs
12667 = (regexp_paren_pair*)
12668 any_dup(old_state->re_state_regoffs, proto_perl);
12669 new_state->re_state_reglastparen
12670 = (U32*) any_dup(old_state->re_state_reglastparen,
12672 new_state->re_state_reglastcloseparen
12673 = (U32*)any_dup(old_state->re_state_reglastcloseparen,
12675 /* XXX This just has to be broken. The old save_re_context
12676 code did SAVEGENERICPV(PL_reg_start_tmp);
12677 PL_reg_start_tmp is char **.
12678 Look above to what the dup code does for
12679 SAVEt_GENERIC_PVREF
12680 It can never have worked.
12681 So this is merely a faithful copy of the exiting bug: */
12682 new_state->re_state_reg_start_tmp
12683 = (char **) pv_dup((char *)
12684 old_state->re_state_reg_start_tmp);
12685 /* I assume that it only ever "worked" because no-one called
12686 (pseudo)fork while the regexp engine had re-entered itself.
12688 #ifdef PERL_OLD_COPY_ON_WRITE
12689 new_state->re_state_nrs
12690 = sv_dup(old_state->re_state_nrs, param);
12692 new_state->re_state_reg_magic
12693 = (MAGIC*) any_dup(old_state->re_state_reg_magic,
12695 new_state->re_state_reg_oldcurpm
12696 = (PMOP*) any_dup(old_state->re_state_reg_oldcurpm,
12698 new_state->re_state_reg_curpm
12699 = (PMOP*) any_dup(old_state->re_state_reg_curpm,
12701 new_state->re_state_reg_oldsaved
12702 = pv_dup(old_state->re_state_reg_oldsaved);
12703 new_state->re_state_reg_poscache
12704 = pv_dup(old_state->re_state_reg_poscache);
12705 new_state->re_state_reg_starttry
12706 = pv_dup(old_state->re_state_reg_starttry);
12709 case SAVEt_COMPILE_WARNINGS:
12710 ptr = POPPTR(ss,ix);
12711 TOPPTR(nss,ix) = DUP_WARNINGS((STRLEN*)ptr);
12714 ptr = POPPTR(ss,ix);
12715 TOPPTR(nss,ix) = parser_dup((const yy_parser*)ptr, param);
12719 "panic: ss_dup inconsistency (%"IVdf")", (IV) type);
12727 /* if sv is a stash, call $class->CLONE_SKIP(), and set the SVphv_CLONEABLE
12728 * flag to the result. This is done for each stash before cloning starts,
12729 * so we know which stashes want their objects cloned */
12732 do_mark_cloneable_stash(pTHX_ SV *const sv)
12734 const HEK * const hvname = HvNAME_HEK((const HV *)sv);
12736 GV* const cloner = gv_fetchmethod_autoload(MUTABLE_HV(sv), "CLONE_SKIP", 0);
12737 SvFLAGS(sv) |= SVphv_CLONEABLE; /* clone objects by default */
12738 if (cloner && GvCV(cloner)) {
12745 mXPUSHs(newSVhek(hvname));
12747 call_sv(MUTABLE_SV(GvCV(cloner)), G_SCALAR);
12754 SvFLAGS(sv) &= ~SVphv_CLONEABLE;
12762 =for apidoc perl_clone
12764 Create and return a new interpreter by cloning the current one.
12766 perl_clone takes these flags as parameters:
12768 CLONEf_COPY_STACKS - is used to, well, copy the stacks also,
12769 without it we only clone the data and zero the stacks,
12770 with it we copy the stacks and the new perl interpreter is
12771 ready to run at the exact same point as the previous one.
12772 The pseudo-fork code uses COPY_STACKS while the
12773 threads->create doesn't.
12775 CLONEf_KEEP_PTR_TABLE -
12776 perl_clone keeps a ptr_table with the pointer of the old
12777 variable as a key and the new variable as a value,
12778 this allows it to check if something has been cloned and not
12779 clone it again but rather just use the value and increase the
12780 refcount. If KEEP_PTR_TABLE is not set then perl_clone will kill
12781 the ptr_table using the function
12782 C<ptr_table_free(PL_ptr_table); PL_ptr_table = NULL;>,
12783 reason to keep it around is if you want to dup some of your own
12784 variable who are outside the graph perl scans, example of this
12785 code is in threads.xs create.
12787 CLONEf_CLONE_HOST -
12788 This is a win32 thing, it is ignored on unix, it tells perls
12789 win32host code (which is c++) to clone itself, this is needed on
12790 win32 if you want to run two threads at the same time,
12791 if you just want to do some stuff in a separate perl interpreter
12792 and then throw it away and return to the original one,
12793 you don't need to do anything.
12798 /* XXX the above needs expanding by someone who actually understands it ! */
12799 EXTERN_C PerlInterpreter *
12800 perl_clone_host(PerlInterpreter* proto_perl, UV flags);
12803 perl_clone(PerlInterpreter *proto_perl, UV flags)
12806 #ifdef PERL_IMPLICIT_SYS
12808 PERL_ARGS_ASSERT_PERL_CLONE;
12810 /* perlhost.h so we need to call into it
12811 to clone the host, CPerlHost should have a c interface, sky */
12813 if (flags & CLONEf_CLONE_HOST) {
12814 return perl_clone_host(proto_perl,flags);
12816 return perl_clone_using(proto_perl, flags,
12818 proto_perl->IMemShared,
12819 proto_perl->IMemParse,
12821 proto_perl->IStdIO,
12825 proto_perl->IProc);
12829 perl_clone_using(PerlInterpreter *proto_perl, UV flags,
12830 struct IPerlMem* ipM, struct IPerlMem* ipMS,
12831 struct IPerlMem* ipMP, struct IPerlEnv* ipE,
12832 struct IPerlStdIO* ipStd, struct IPerlLIO* ipLIO,
12833 struct IPerlDir* ipD, struct IPerlSock* ipS,
12834 struct IPerlProc* ipP)
12836 /* XXX many of the string copies here can be optimized if they're
12837 * constants; they need to be allocated as common memory and just
12838 * their pointers copied. */
12841 CLONE_PARAMS clone_params;
12842 CLONE_PARAMS* const param = &clone_params;
12844 PerlInterpreter * const my_perl = (PerlInterpreter*)(*ipM->pMalloc)(ipM, sizeof(PerlInterpreter));
12846 PERL_ARGS_ASSERT_PERL_CLONE_USING;
12847 #else /* !PERL_IMPLICIT_SYS */
12849 CLONE_PARAMS clone_params;
12850 CLONE_PARAMS* param = &clone_params;
12851 PerlInterpreter * const my_perl = (PerlInterpreter*)PerlMem_malloc(sizeof(PerlInterpreter));
12853 PERL_ARGS_ASSERT_PERL_CLONE;
12854 #endif /* PERL_IMPLICIT_SYS */
12856 /* for each stash, determine whether its objects should be cloned */
12857 S_visit(proto_perl, do_mark_cloneable_stash, SVt_PVHV, SVTYPEMASK);
12858 PERL_SET_THX(my_perl);
12861 PoisonNew(my_perl, 1, PerlInterpreter);
12864 PL_defstash = NULL; /* may be used by perl malloc() */
12867 PL_scopestack_name = 0;
12869 PL_savestack_ix = 0;
12870 PL_savestack_max = -1;
12871 PL_sig_pending = 0;
12873 Zero(&PL_debug_pad, 1, struct perl_debug_pad);
12874 # ifdef DEBUG_LEAKING_SCALARS
12875 PL_sv_serial = (((UV)my_perl >> 2) & 0xfff) * 1000000;
12877 #else /* !DEBUGGING */
12878 Zero(my_perl, 1, PerlInterpreter);
12879 #endif /* DEBUGGING */
12881 #ifdef PERL_IMPLICIT_SYS
12882 /* host pointers */
12884 PL_MemShared = ipMS;
12885 PL_MemParse = ipMP;
12892 #endif /* PERL_IMPLICIT_SYS */
12894 param->flags = flags;
12895 /* Nothing in the core code uses this, but we make it available to
12896 extensions (using mg_dup). */
12897 param->proto_perl = proto_perl;
12898 /* Likely nothing will use this, but it is initialised to be consistent
12899 with Perl_clone_params_new(). */
12900 param->new_perl = my_perl;
12901 param->unreferenced = NULL;
12903 INIT_TRACK_MEMPOOL(my_perl->Imemory_debug_header, my_perl);
12905 PL_body_arenas = NULL;
12906 Zero(&PL_body_roots, 1, PL_body_roots);
12909 PL_sv_objcount = 0;
12911 PL_sv_arenaroot = NULL;
12913 PL_debug = proto_perl->Idebug;
12915 PL_hash_seed = proto_perl->Ihash_seed;
12916 PL_rehash_seed = proto_perl->Irehash_seed;
12918 SvANY(&PL_sv_undef) = NULL;
12919 SvREFCNT(&PL_sv_undef) = (~(U32)0)/2;
12920 SvFLAGS(&PL_sv_undef) = SVf_READONLY|SVt_NULL;
12921 SvREFCNT(&PL_sv_no) = (~(U32)0)/2;
12922 SvFLAGS(&PL_sv_no) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
12923 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
12925 SvANY(&PL_sv_yes) = new_XPVNV();
12926 SvREFCNT(&PL_sv_yes) = (~(U32)0)/2;
12927 SvFLAGS(&PL_sv_yes) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
12928 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
12930 /* dbargs array probably holds garbage */
12933 PL_compiling = proto_perl->Icompiling;
12935 #ifdef PERL_DEBUG_READONLY_OPS
12940 /* pseudo environmental stuff */
12941 PL_origargc = proto_perl->Iorigargc;
12942 PL_origargv = proto_perl->Iorigargv;
12944 /* Set tainting stuff before PerlIO_debug can possibly get called */
12945 PL_tainting = proto_perl->Itainting;
12946 PL_taint_warn = proto_perl->Itaint_warn;
12948 PL_minus_c = proto_perl->Iminus_c;
12950 PL_localpatches = proto_perl->Ilocalpatches;
12951 PL_splitstr = proto_perl->Isplitstr;
12952 PL_minus_n = proto_perl->Iminus_n;
12953 PL_minus_p = proto_perl->Iminus_p;
12954 PL_minus_l = proto_perl->Iminus_l;
12955 PL_minus_a = proto_perl->Iminus_a;
12956 PL_minus_E = proto_perl->Iminus_E;
12957 PL_minus_F = proto_perl->Iminus_F;
12958 PL_doswitches = proto_perl->Idoswitches;
12959 PL_dowarn = proto_perl->Idowarn;
12960 PL_sawampersand = proto_perl->Isawampersand;
12961 PL_unsafe = proto_perl->Iunsafe;
12962 PL_perldb = proto_perl->Iperldb;
12963 PL_perl_destruct_level = proto_perl->Iperl_destruct_level;
12964 PL_exit_flags = proto_perl->Iexit_flags;
12966 /* XXX time(&PL_basetime) when asked for? */
12967 PL_basetime = proto_perl->Ibasetime;
12969 PL_maxsysfd = proto_perl->Imaxsysfd;
12970 PL_statusvalue = proto_perl->Istatusvalue;
12972 PL_statusvalue_vms = proto_perl->Istatusvalue_vms;
12974 PL_statusvalue_posix = proto_perl->Istatusvalue_posix;
12977 /* RE engine related */
12978 Zero(&PL_reg_state, 1, struct re_save_state);
12979 PL_reginterp_cnt = 0;
12980 PL_regmatch_slab = NULL;
12982 PL_sub_generation = proto_perl->Isub_generation;
12984 /* funky return mechanisms */
12985 PL_forkprocess = proto_perl->Iforkprocess;
12987 /* internal state */
12988 PL_maxo = proto_perl->Imaxo;
12990 PL_main_start = proto_perl->Imain_start;
12991 PL_eval_root = proto_perl->Ieval_root;
12992 PL_eval_start = proto_perl->Ieval_start;
12994 PL_filemode = proto_perl->Ifilemode;
12995 PL_lastfd = proto_perl->Ilastfd;
12996 PL_oldname = proto_perl->Ioldname; /* XXX not quite right */
12999 PL_gensym = proto_perl->Igensym;
13001 PL_laststatval = proto_perl->Ilaststatval;
13002 PL_laststype = proto_perl->Ilaststype;
13005 PL_profiledata = NULL;
13007 PL_generation = proto_perl->Igeneration;
13009 PL_in_clean_objs = proto_perl->Iin_clean_objs;
13010 PL_in_clean_all = proto_perl->Iin_clean_all;
13012 PL_uid = proto_perl->Iuid;
13013 PL_euid = proto_perl->Ieuid;
13014 PL_gid = proto_perl->Igid;
13015 PL_egid = proto_perl->Iegid;
13016 PL_nomemok = proto_perl->Inomemok;
13017 PL_an = proto_perl->Ian;
13018 PL_evalseq = proto_perl->Ievalseq;
13019 PL_origenviron = proto_perl->Iorigenviron; /* XXX not quite right */
13020 PL_origalen = proto_perl->Iorigalen;
13022 PL_sighandlerp = proto_perl->Isighandlerp;
13024 PL_runops = proto_perl->Irunops;
13026 PL_subline = proto_perl->Isubline;
13029 PL_cryptseen = proto_perl->Icryptseen;
13032 PL_hints = proto_perl->Ihints;
13034 PL_amagic_generation = proto_perl->Iamagic_generation;
13036 #ifdef USE_LOCALE_COLLATE
13037 PL_collation_ix = proto_perl->Icollation_ix;
13038 PL_collation_standard = proto_perl->Icollation_standard;
13039 PL_collxfrm_base = proto_perl->Icollxfrm_base;
13040 PL_collxfrm_mult = proto_perl->Icollxfrm_mult;
13041 #endif /* USE_LOCALE_COLLATE */
13043 #ifdef USE_LOCALE_NUMERIC
13044 PL_numeric_standard = proto_perl->Inumeric_standard;
13045 PL_numeric_local = proto_perl->Inumeric_local;
13046 #endif /* !USE_LOCALE_NUMERIC */
13048 /* Did the locale setup indicate UTF-8? */
13049 PL_utf8locale = proto_perl->Iutf8locale;
13050 /* Unicode features (see perlrun/-C) */
13051 PL_unicode = proto_perl->Iunicode;
13053 /* Pre-5.8 signals control */
13054 PL_signals = proto_perl->Isignals;
13056 /* times() ticks per second */
13057 PL_clocktick = proto_perl->Iclocktick;
13059 /* Recursion stopper for PerlIO_find_layer */
13060 PL_in_load_module = proto_perl->Iin_load_module;
13062 /* sort() routine */
13063 PL_sort_RealCmp = proto_perl->Isort_RealCmp;
13065 /* Not really needed/useful since the reenrant_retint is "volatile",
13066 * but do it for consistency's sake. */
13067 PL_reentrant_retint = proto_perl->Ireentrant_retint;
13069 /* Hooks to shared SVs and locks. */
13070 PL_sharehook = proto_perl->Isharehook;
13071 PL_lockhook = proto_perl->Ilockhook;
13072 PL_unlockhook = proto_perl->Iunlockhook;
13073 PL_threadhook = proto_perl->Ithreadhook;
13074 PL_destroyhook = proto_perl->Idestroyhook;
13075 PL_signalhook = proto_perl->Isignalhook;
13077 PL_globhook = proto_perl->Iglobhook;
13079 #ifdef THREADS_HAVE_PIDS
13080 PL_ppid = proto_perl->Ippid;
13084 PL_last_swash_hv = NULL; /* reinits on demand */
13085 PL_last_swash_klen = 0;
13086 PL_last_swash_key[0]= '\0';
13087 PL_last_swash_tmps = (U8*)NULL;
13088 PL_last_swash_slen = 0;
13090 PL_glob_index = proto_perl->Iglob_index;
13091 PL_srand_called = proto_perl->Isrand_called;
13093 if (flags & CLONEf_COPY_STACKS) {
13094 /* next allocation will be PL_tmps_stack[PL_tmps_ix+1] */
13095 PL_tmps_ix = proto_perl->Itmps_ix;
13096 PL_tmps_max = proto_perl->Itmps_max;
13097 PL_tmps_floor = proto_perl->Itmps_floor;
13099 /* next push_scope()/ENTER sets PL_scopestack[PL_scopestack_ix]
13100 * NOTE: unlike the others! */
13101 PL_scopestack_ix = proto_perl->Iscopestack_ix;
13102 PL_scopestack_max = proto_perl->Iscopestack_max;
13104 /* next SSPUSHFOO() sets PL_savestack[PL_savestack_ix]
13105 * NOTE: unlike the others! */
13106 PL_savestack_ix = proto_perl->Isavestack_ix;
13107 PL_savestack_max = proto_perl->Isavestack_max;
13110 PL_start_env = proto_perl->Istart_env; /* XXXXXX */
13111 PL_top_env = &PL_start_env;
13113 PL_op = proto_perl->Iop;
13116 PL_Xpv = (XPV*)NULL;
13117 my_perl->Ina = proto_perl->Ina;
13119 PL_statbuf = proto_perl->Istatbuf;
13120 PL_statcache = proto_perl->Istatcache;
13123 PL_timesbuf = proto_perl->Itimesbuf;
13126 PL_tainted = proto_perl->Itainted;
13127 PL_curpm = proto_perl->Icurpm; /* XXX No PMOP ref count */
13129 PL_chopset = proto_perl->Ichopset; /* XXX never deallocated */
13131 PL_restartjmpenv = proto_perl->Irestartjmpenv;
13132 PL_restartop = proto_perl->Irestartop;
13133 PL_in_eval = proto_perl->Iin_eval;
13134 PL_delaymagic = proto_perl->Idelaymagic;
13135 PL_phase = proto_perl->Iphase;
13136 PL_localizing = proto_perl->Ilocalizing;
13138 PL_hv_fetch_ent_mh = NULL;
13139 PL_modcount = proto_perl->Imodcount;
13140 PL_lastgotoprobe = NULL;
13141 PL_dumpindent = proto_perl->Idumpindent;
13143 PL_efloatbuf = NULL; /* reinits on demand */
13144 PL_efloatsize = 0; /* reinits on demand */
13148 PL_regdummy = proto_perl->Iregdummy;
13149 PL_colorset = 0; /* reinits PL_colors[] */
13150 /*PL_colors[6] = {0,0,0,0,0,0};*/
13152 /* Pluggable optimizer */
13153 PL_peepp = proto_perl->Ipeepp;
13154 PL_rpeepp = proto_perl->Irpeepp;
13155 /* op_free() hook */
13156 PL_opfreehook = proto_perl->Iopfreehook;
13158 #ifdef USE_REENTRANT_API
13159 /* XXX: things like -Dm will segfault here in perlio, but doing
13160 * PERL_SET_CONTEXT(proto_perl);
13161 * breaks too many other things
13163 Perl_reentrant_init(aTHX);
13166 /* create SV map for pointer relocation */
13167 PL_ptr_table = ptr_table_new();
13169 /* initialize these special pointers as early as possible */
13170 ptr_table_store(PL_ptr_table, &proto_perl->Isv_undef, &PL_sv_undef);
13172 SvANY(&PL_sv_no) = new_XPVNV();
13173 SvPV_set(&PL_sv_no, savepvn(PL_No, 0));
13174 SvCUR_set(&PL_sv_no, 0);
13175 SvLEN_set(&PL_sv_no, 1);
13176 SvIV_set(&PL_sv_no, 0);
13177 SvNV_set(&PL_sv_no, 0);
13178 ptr_table_store(PL_ptr_table, &proto_perl->Isv_no, &PL_sv_no);
13180 SvPV_set(&PL_sv_yes, savepvn(PL_Yes, 1));
13181 SvCUR_set(&PL_sv_yes, 1);
13182 SvLEN_set(&PL_sv_yes, 2);
13183 SvIV_set(&PL_sv_yes, 1);
13184 SvNV_set(&PL_sv_yes, 1);
13185 ptr_table_store(PL_ptr_table, &proto_perl->Isv_yes, &PL_sv_yes);
13187 /* create (a non-shared!) shared string table */
13188 PL_strtab = newHV();
13189 HvSHAREKEYS_off(PL_strtab);
13190 hv_ksplit(PL_strtab, HvTOTALKEYS(proto_perl->Istrtab));
13191 ptr_table_store(PL_ptr_table, proto_perl->Istrtab, PL_strtab);
13193 /* These two PVs will be free'd special way so must set them same way op.c does */
13194 PL_compiling.cop_stashpv = savesharedpv(PL_compiling.cop_stashpv);
13195 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_stashpv, PL_compiling.cop_stashpv);
13197 PL_compiling.cop_file = savesharedpv(PL_compiling.cop_file);
13198 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_file, PL_compiling.cop_file);
13200 ptr_table_store(PL_ptr_table, &proto_perl->Icompiling, &PL_compiling);
13201 PL_compiling.cop_warnings = DUP_WARNINGS(PL_compiling.cop_warnings);
13202 CopHINTHASH_set(&PL_compiling, cophh_copy(CopHINTHASH_get(&PL_compiling)));
13203 PL_curcop = (COP*)any_dup(proto_perl->Icurcop, proto_perl);
13205 param->stashes = newAV(); /* Setup array of objects to call clone on */
13206 /* This makes no difference to the implementation, as it always pushes
13207 and shifts pointers to other SVs without changing their reference
13208 count, with the array becoming empty before it is freed. However, it
13209 makes it conceptually clear what is going on, and will avoid some
13210 work inside av.c, filling slots between AvFILL() and AvMAX() with
13211 &PL_sv_undef, and SvREFCNT_dec()ing those. */
13212 AvREAL_off(param->stashes);
13214 if (!(flags & CLONEf_COPY_STACKS)) {
13215 param->unreferenced = newAV();
13218 #ifdef PERLIO_LAYERS
13219 /* Clone PerlIO tables as soon as we can handle general xx_dup() */
13220 PerlIO_clone(aTHX_ proto_perl, param);
13223 PL_envgv = gv_dup(proto_perl->Ienvgv, param);
13224 PL_incgv = gv_dup(proto_perl->Iincgv, param);
13225 PL_hintgv = gv_dup(proto_perl->Ihintgv, param);
13226 PL_origfilename = SAVEPV(proto_perl->Iorigfilename);
13227 PL_diehook = sv_dup_inc(proto_perl->Idiehook, param);
13228 PL_warnhook = sv_dup_inc(proto_perl->Iwarnhook, param);
13231 PL_patchlevel = sv_dup_inc(proto_perl->Ipatchlevel, param);
13232 PL_apiversion = sv_dup_inc(proto_perl->Iapiversion, param);
13233 PL_inplace = SAVEPV(proto_perl->Iinplace);
13234 PL_e_script = sv_dup_inc(proto_perl->Ie_script, param);
13236 /* magical thingies */
13237 PL_formfeed = sv_dup(proto_perl->Iformfeed, param);
13239 PL_encoding = sv_dup(proto_perl->Iencoding, param);
13241 sv_setpvs(PERL_DEBUG_PAD(0), ""); /* For regex debugging. */
13242 sv_setpvs(PERL_DEBUG_PAD(1), ""); /* ext/re needs these */
13243 sv_setpvs(PERL_DEBUG_PAD(2), ""); /* even without DEBUGGING. */
13246 /* Clone the regex array */
13247 /* ORANGE FIXME for plugins, probably in the SV dup code.
13248 newSViv(PTR2IV(CALLREGDUPE(
13249 INT2PTR(REGEXP *, SvIVX(regex)), param))))
13251 PL_regex_padav = av_dup_inc(proto_perl->Iregex_padav, param);
13252 PL_regex_pad = AvARRAY(PL_regex_padav);
13254 /* shortcuts to various I/O objects */
13255 PL_ofsgv = gv_dup_inc(proto_perl->Iofsgv, param);
13256 PL_stdingv = gv_dup(proto_perl->Istdingv, param);
13257 PL_stderrgv = gv_dup(proto_perl->Istderrgv, param);
13258 PL_defgv = gv_dup(proto_perl->Idefgv, param);
13259 PL_argvgv = gv_dup(proto_perl->Iargvgv, param);
13260 PL_argvoutgv = gv_dup(proto_perl->Iargvoutgv, param);
13261 PL_argvout_stack = av_dup_inc(proto_perl->Iargvout_stack, param);
13263 /* shortcuts to regexp stuff */
13264 PL_replgv = gv_dup(proto_perl->Ireplgv, param);
13266 /* shortcuts to misc objects */
13267 PL_errgv = gv_dup(proto_perl->Ierrgv, param);
13269 /* shortcuts to debugging objects */
13270 PL_DBgv = gv_dup(proto_perl->IDBgv, param);
13271 PL_DBline = gv_dup(proto_perl->IDBline, param);
13272 PL_DBsub = gv_dup(proto_perl->IDBsub, param);
13273 PL_DBsingle = sv_dup(proto_perl->IDBsingle, param);
13274 PL_DBtrace = sv_dup(proto_perl->IDBtrace, param);
13275 PL_DBsignal = sv_dup(proto_perl->IDBsignal, param);
13277 /* symbol tables */
13278 PL_defstash = hv_dup_inc(proto_perl->Idefstash, param);
13279 PL_curstash = hv_dup_inc(proto_perl->Icurstash, param);
13280 PL_debstash = hv_dup(proto_perl->Idebstash, param);
13281 PL_globalstash = hv_dup(proto_perl->Iglobalstash, param);
13282 PL_curstname = sv_dup_inc(proto_perl->Icurstname, param);
13284 PL_beginav = av_dup_inc(proto_perl->Ibeginav, param);
13285 PL_beginav_save = av_dup_inc(proto_perl->Ibeginav_save, param);
13286 PL_checkav_save = av_dup_inc(proto_perl->Icheckav_save, param);
13287 PL_unitcheckav = av_dup_inc(proto_perl->Iunitcheckav, param);
13288 PL_unitcheckav_save = av_dup_inc(proto_perl->Iunitcheckav_save, param);
13289 PL_endav = av_dup_inc(proto_perl->Iendav, param);
13290 PL_checkav = av_dup_inc(proto_perl->Icheckav, param);
13291 PL_initav = av_dup_inc(proto_perl->Iinitav, param);
13293 PL_isarev = hv_dup_inc(proto_perl->Iisarev, param);
13295 /* subprocess state */
13296 PL_fdpid = av_dup_inc(proto_perl->Ifdpid, param);
13298 if (proto_perl->Iop_mask)
13299 PL_op_mask = SAVEPVN(proto_perl->Iop_mask, PL_maxo);
13302 /* PL_asserting = proto_perl->Iasserting; */
13304 /* current interpreter roots */
13305 PL_main_cv = cv_dup_inc(proto_perl->Imain_cv, param);
13307 PL_main_root = OpREFCNT_inc(proto_perl->Imain_root);
13310 /* runtime control stuff */
13311 PL_curcopdb = (COP*)any_dup(proto_perl->Icurcopdb, proto_perl);
13313 PL_preambleav = av_dup_inc(proto_perl->Ipreambleav, param);
13315 PL_ors_sv = sv_dup_inc(proto_perl->Iors_sv, param);
13317 /* interpreter atexit processing */
13318 PL_exitlistlen = proto_perl->Iexitlistlen;
13319 if (PL_exitlistlen) {
13320 Newx(PL_exitlist, PL_exitlistlen, PerlExitListEntry);
13321 Copy(proto_perl->Iexitlist, PL_exitlist, PL_exitlistlen, PerlExitListEntry);
13324 PL_exitlist = (PerlExitListEntry*)NULL;
13326 PL_my_cxt_size = proto_perl->Imy_cxt_size;
13327 if (PL_my_cxt_size) {
13328 Newx(PL_my_cxt_list, PL_my_cxt_size, void *);
13329 Copy(proto_perl->Imy_cxt_list, PL_my_cxt_list, PL_my_cxt_size, void *);
13330 #ifdef PERL_GLOBAL_STRUCT_PRIVATE
13331 Newx(PL_my_cxt_keys, PL_my_cxt_size, const char *);
13332 Copy(proto_perl->Imy_cxt_keys, PL_my_cxt_keys, PL_my_cxt_size, char *);
13336 PL_my_cxt_list = (void**)NULL;
13337 #ifdef PERL_GLOBAL_STRUCT_PRIVATE
13338 PL_my_cxt_keys = (const char**)NULL;
13341 PL_modglobal = hv_dup_inc(proto_perl->Imodglobal, param);
13342 PL_custom_op_names = hv_dup_inc(proto_perl->Icustom_op_names,param);
13343 PL_custom_op_descs = hv_dup_inc(proto_perl->Icustom_op_descs,param);
13344 PL_custom_ops = hv_dup_inc(proto_perl->Icustom_ops, param);
13346 PL_compcv = cv_dup(proto_perl->Icompcv, param);
13348 PAD_CLONE_VARS(proto_perl, param);
13350 #ifdef HAVE_INTERP_INTERN
13351 sys_intern_dup(&proto_perl->Isys_intern, &PL_sys_intern);
13354 PL_DBcv = cv_dup(proto_perl->IDBcv, param);
13356 #ifdef PERL_USES_PL_PIDSTATUS
13357 PL_pidstatus = newHV(); /* XXX flag for cloning? */
13359 PL_osname = SAVEPV(proto_perl->Iosname);
13360 PL_parser = parser_dup(proto_perl->Iparser, param);
13362 /* XXX this only works if the saved cop has already been cloned */
13363 if (proto_perl->Iparser) {
13364 PL_parser->saved_curcop = (COP*)any_dup(
13365 proto_perl->Iparser->saved_curcop,
13369 PL_subname = sv_dup_inc(proto_perl->Isubname, param);
13371 #ifdef USE_LOCALE_COLLATE
13372 PL_collation_name = SAVEPV(proto_perl->Icollation_name);
13373 #endif /* USE_LOCALE_COLLATE */
13375 #ifdef USE_LOCALE_NUMERIC
13376 PL_numeric_name = SAVEPV(proto_perl->Inumeric_name);
13377 PL_numeric_radix_sv = sv_dup_inc(proto_perl->Inumeric_radix_sv, param);
13378 #endif /* !USE_LOCALE_NUMERIC */
13380 /* utf8 character classes */
13381 PL_utf8_alnum = sv_dup_inc(proto_perl->Iutf8_alnum, param);
13382 PL_utf8_alpha = sv_dup_inc(proto_perl->Iutf8_alpha, param);
13383 PL_utf8_space = sv_dup_inc(proto_perl->Iutf8_space, param);
13384 PL_utf8_graph = sv_dup_inc(proto_perl->Iutf8_graph, param);
13385 PL_utf8_digit = sv_dup_inc(proto_perl->Iutf8_digit, param);
13386 PL_utf8_upper = sv_dup_inc(proto_perl->Iutf8_upper, param);
13387 PL_utf8_lower = sv_dup_inc(proto_perl->Iutf8_lower, param);
13388 PL_utf8_print = sv_dup_inc(proto_perl->Iutf8_print, param);
13389 PL_utf8_punct = sv_dup_inc(proto_perl->Iutf8_punct, param);
13390 PL_utf8_xdigit = sv_dup_inc(proto_perl->Iutf8_xdigit, param);
13391 PL_utf8_mark = sv_dup_inc(proto_perl->Iutf8_mark, param);
13392 PL_utf8_X_begin = sv_dup_inc(proto_perl->Iutf8_X_begin, param);
13393 PL_utf8_X_extend = sv_dup_inc(proto_perl->Iutf8_X_extend, param);
13394 PL_utf8_X_prepend = sv_dup_inc(proto_perl->Iutf8_X_prepend, param);
13395 PL_utf8_X_non_hangul = sv_dup_inc(proto_perl->Iutf8_X_non_hangul, param);
13396 PL_utf8_X_L = sv_dup_inc(proto_perl->Iutf8_X_L, param);
13397 PL_utf8_X_LV = sv_dup_inc(proto_perl->Iutf8_X_LV, param);
13398 PL_utf8_X_LVT = sv_dup_inc(proto_perl->Iutf8_X_LVT, param);
13399 PL_utf8_X_T = sv_dup_inc(proto_perl->Iutf8_X_T, param);
13400 PL_utf8_X_V = sv_dup_inc(proto_perl->Iutf8_X_V, param);
13401 PL_utf8_X_LV_LVT_V = sv_dup_inc(proto_perl->Iutf8_X_LV_LVT_V, param);
13402 PL_utf8_toupper = sv_dup_inc(proto_perl->Iutf8_toupper, param);
13403 PL_utf8_totitle = sv_dup_inc(proto_perl->Iutf8_totitle, param);
13404 PL_utf8_tolower = sv_dup_inc(proto_perl->Iutf8_tolower, param);
13405 PL_utf8_tofold = sv_dup_inc(proto_perl->Iutf8_tofold, param);
13406 PL_utf8_idstart = sv_dup_inc(proto_perl->Iutf8_idstart, param);
13407 PL_utf8_xidstart = sv_dup_inc(proto_perl->Iutf8_xidstart, param);
13408 PL_utf8_perl_idstart = sv_dup_inc(proto_perl->Iutf8_perl_idstart, param);
13409 PL_utf8_idcont = sv_dup_inc(proto_perl->Iutf8_idcont, param);
13410 PL_utf8_xidcont = sv_dup_inc(proto_perl->Iutf8_xidcont, param);
13411 PL_utf8_foldable = sv_dup_inc(proto_perl->Iutf8_foldable, param);
13414 if (proto_perl->Ipsig_pend) {
13415 Newxz(PL_psig_pend, SIG_SIZE, int);
13418 PL_psig_pend = (int*)NULL;
13421 if (proto_perl->Ipsig_name) {
13422 Newx(PL_psig_name, 2 * SIG_SIZE, SV*);
13423 sv_dup_inc_multiple(proto_perl->Ipsig_name, PL_psig_name, 2 * SIG_SIZE,
13425 PL_psig_ptr = PL_psig_name + SIG_SIZE;
13428 PL_psig_ptr = (SV**)NULL;
13429 PL_psig_name = (SV**)NULL;
13432 if (flags & CLONEf_COPY_STACKS) {
13433 Newx(PL_tmps_stack, PL_tmps_max, SV*);
13434 sv_dup_inc_multiple(proto_perl->Itmps_stack, PL_tmps_stack,
13435 PL_tmps_ix+1, param);
13437 /* next PUSHMARK() sets *(PL_markstack_ptr+1) */
13438 i = proto_perl->Imarkstack_max - proto_perl->Imarkstack;
13439 Newxz(PL_markstack, i, I32);
13440 PL_markstack_max = PL_markstack + (proto_perl->Imarkstack_max
13441 - proto_perl->Imarkstack);
13442 PL_markstack_ptr = PL_markstack + (proto_perl->Imarkstack_ptr
13443 - proto_perl->Imarkstack);
13444 Copy(proto_perl->Imarkstack, PL_markstack,
13445 PL_markstack_ptr - PL_markstack + 1, I32);
13447 /* next push_scope()/ENTER sets PL_scopestack[PL_scopestack_ix]
13448 * NOTE: unlike the others! */
13449 Newxz(PL_scopestack, PL_scopestack_max, I32);
13450 Copy(proto_perl->Iscopestack, PL_scopestack, PL_scopestack_ix, I32);
13453 Newxz(PL_scopestack_name, PL_scopestack_max, const char *);
13454 Copy(proto_perl->Iscopestack_name, PL_scopestack_name, PL_scopestack_ix, const char *);
13456 /* NOTE: si_dup() looks at PL_markstack */
13457 PL_curstackinfo = si_dup(proto_perl->Icurstackinfo, param);
13459 /* PL_curstack = PL_curstackinfo->si_stack; */
13460 PL_curstack = av_dup(proto_perl->Icurstack, param);
13461 PL_mainstack = av_dup(proto_perl->Imainstack, param);
13463 /* next PUSHs() etc. set *(PL_stack_sp+1) */
13464 PL_stack_base = AvARRAY(PL_curstack);
13465 PL_stack_sp = PL_stack_base + (proto_perl->Istack_sp
13466 - proto_perl->Istack_base);
13467 PL_stack_max = PL_stack_base + AvMAX(PL_curstack);
13469 /*Newxz(PL_savestack, PL_savestack_max, ANY);*/
13470 PL_savestack = ss_dup(proto_perl, param);
13474 ENTER; /* perl_destruct() wants to LEAVE; */
13477 PL_statgv = gv_dup(proto_perl->Istatgv, param);
13478 PL_statname = sv_dup_inc(proto_perl->Istatname, param);
13480 PL_rs = sv_dup_inc(proto_perl->Irs, param);
13481 PL_last_in_gv = gv_dup(proto_perl->Ilast_in_gv, param);
13482 PL_defoutgv = gv_dup_inc(proto_perl->Idefoutgv, param);
13483 PL_toptarget = sv_dup_inc(proto_perl->Itoptarget, param);
13484 PL_bodytarget = sv_dup_inc(proto_perl->Ibodytarget, param);
13485 PL_formtarget = sv_dup(proto_perl->Iformtarget, param);
13487 PL_errors = sv_dup_inc(proto_perl->Ierrors, param);
13489 PL_sortcop = (OP*)any_dup(proto_perl->Isortcop, proto_perl);
13490 PL_sortstash = hv_dup(proto_perl->Isortstash, param);
13491 PL_firstgv = gv_dup(proto_perl->Ifirstgv, param);
13492 PL_secondgv = gv_dup(proto_perl->Isecondgv, param);
13494 PL_stashcache = newHV();
13496 PL_watchaddr = (char **) ptr_table_fetch(PL_ptr_table,
13497 proto_perl->Iwatchaddr);
13498 PL_watchok = PL_watchaddr ? * PL_watchaddr : NULL;
13499 if (PL_debug && PL_watchaddr) {
13500 PerlIO_printf(Perl_debug_log,
13501 "WATCHING: %"UVxf" cloned as %"UVxf" with value %"UVxf"\n",
13502 PTR2UV(proto_perl->Iwatchaddr), PTR2UV(PL_watchaddr),
13503 PTR2UV(PL_watchok));
13506 PL_registered_mros = hv_dup_inc(proto_perl->Iregistered_mros, param);
13507 PL_blockhooks = av_dup_inc(proto_perl->Iblockhooks, param);
13508 PL_utf8_foldclosures = hv_dup_inc(proto_perl->Iutf8_foldclosures, param);
13510 /* Call the ->CLONE method, if it exists, for each of the stashes
13511 identified by sv_dup() above.
13513 while(av_len(param->stashes) != -1) {
13514 HV* const stash = MUTABLE_HV(av_shift(param->stashes));
13515 GV* const cloner = gv_fetchmethod_autoload(stash, "CLONE", 0);
13516 if (cloner && GvCV(cloner)) {
13521 mXPUSHs(newSVhek(HvNAME_HEK(stash)));
13523 call_sv(MUTABLE_SV(GvCV(cloner)), G_DISCARD);
13529 if (!(flags & CLONEf_KEEP_PTR_TABLE)) {
13530 ptr_table_free(PL_ptr_table);
13531 PL_ptr_table = NULL;
13534 if (!(flags & CLONEf_COPY_STACKS)) {
13535 unreferenced_to_tmp_stack(param->unreferenced);
13538 SvREFCNT_dec(param->stashes);
13540 /* orphaned? eg threads->new inside BEGIN or use */
13541 if (PL_compcv && ! SvREFCNT(PL_compcv)) {
13542 SvREFCNT_inc_simple_void(PL_compcv);
13543 SAVEFREESV(PL_compcv);
13550 S_unreferenced_to_tmp_stack(pTHX_ AV *const unreferenced)
13552 PERL_ARGS_ASSERT_UNREFERENCED_TO_TMP_STACK;
13554 if (AvFILLp(unreferenced) > -1) {
13555 SV **svp = AvARRAY(unreferenced);
13556 SV **const last = svp + AvFILLp(unreferenced);
13560 if (SvREFCNT(*svp) == 1)
13562 } while (++svp <= last);
13564 EXTEND_MORTAL(count);
13565 svp = AvARRAY(unreferenced);
13568 if (SvREFCNT(*svp) == 1) {
13569 /* Our reference is the only one to this SV. This means that
13570 in this thread, the scalar effectively has a 0 reference.
13571 That doesn't work (cleanup never happens), so donate our
13572 reference to it onto the save stack. */
13573 PL_tmps_stack[++PL_tmps_ix] = *svp;
13575 /* As an optimisation, because we are already walking the
13576 entire array, instead of above doing either
13577 SvREFCNT_inc(*svp) or *svp = &PL_sv_undef, we can instead
13578 release our reference to the scalar, so that at the end of
13579 the array owns zero references to the scalars it happens to
13580 point to. We are effectively converting the array from
13581 AvREAL() on to AvREAL() off. This saves the av_clear()
13582 (triggered by the SvREFCNT_dec(unreferenced) below) from
13583 walking the array a second time. */
13584 SvREFCNT_dec(*svp);
13587 } while (++svp <= last);
13588 AvREAL_off(unreferenced);
13590 SvREFCNT_dec(unreferenced);
13594 Perl_clone_params_del(CLONE_PARAMS *param)
13596 /* This seemingly funky ordering keeps the build with PERL_GLOBAL_STRUCT
13598 PerlInterpreter *const to = param->new_perl;
13600 PerlInterpreter *const was = PERL_GET_THX;
13602 PERL_ARGS_ASSERT_CLONE_PARAMS_DEL;
13608 SvREFCNT_dec(param->stashes);
13609 if (param->unreferenced)
13610 unreferenced_to_tmp_stack(param->unreferenced);
13620 Perl_clone_params_new(PerlInterpreter *const from, PerlInterpreter *const to)
13623 /* Need to play this game, as newAV() can call safesysmalloc(), and that
13624 does a dTHX; to get the context from thread local storage.
13625 FIXME - under PERL_CORE Newx(), Safefree() and friends should expand to
13626 a version that passes in my_perl. */
13627 PerlInterpreter *const was = PERL_GET_THX;
13628 CLONE_PARAMS *param;
13630 PERL_ARGS_ASSERT_CLONE_PARAMS_NEW;
13636 /* Given that we've set the context, we can do this unshared. */
13637 Newx(param, 1, CLONE_PARAMS);
13640 param->proto_perl = from;
13641 param->new_perl = to;
13642 param->stashes = (AV *)Perl_newSV_type(to, SVt_PVAV);
13643 AvREAL_off(param->stashes);
13644 param->unreferenced = (AV *)Perl_newSV_type(to, SVt_PVAV);
13652 #endif /* USE_ITHREADS */
13655 =head1 Unicode Support
13657 =for apidoc sv_recode_to_utf8
13659 The encoding is assumed to be an Encode object, on entry the PV
13660 of the sv is assumed to be octets in that encoding, and the sv
13661 will be converted into Unicode (and UTF-8).
13663 If the sv already is UTF-8 (or if it is not POK), or if the encoding
13664 is not a reference, nothing is done to the sv. If the encoding is not
13665 an C<Encode::XS> Encoding object, bad things will happen.
13666 (See F<lib/encoding.pm> and L<Encode>.)
13668 The PV of the sv is returned.
13673 Perl_sv_recode_to_utf8(pTHX_ SV *sv, SV *encoding)
13677 PERL_ARGS_ASSERT_SV_RECODE_TO_UTF8;
13679 if (SvPOK(sv) && !SvUTF8(sv) && !IN_BYTES && SvROK(encoding)) {
13693 Passing sv_yes is wrong - it needs to be or'ed set of constants
13694 for Encode::XS, while UTf-8 decode (currently) assumes a true value means
13695 remove converted chars from source.
13697 Both will default the value - let them.
13699 XPUSHs(&PL_sv_yes);
13702 call_method("decode", G_SCALAR);
13706 s = SvPV_const(uni, len);
13707 if (s != SvPVX_const(sv)) {
13708 SvGROW(sv, len + 1);
13709 Move(s, SvPVX(sv), len + 1, char);
13710 SvCUR_set(sv, len);
13714 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
13715 /* clear pos and any utf8 cache */
13716 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
13719 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
13720 magic_setutf8(sv,mg); /* clear UTF8 cache */
13725 return SvPOKp(sv) ? SvPVX(sv) : NULL;
13729 =for apidoc sv_cat_decode
13731 The encoding is assumed to be an Encode object, the PV of the ssv is
13732 assumed to be octets in that encoding and decoding the input starts
13733 from the position which (PV + *offset) pointed to. The dsv will be
13734 concatenated the decoded UTF-8 string from ssv. Decoding will terminate
13735 when the string tstr appears in decoding output or the input ends on
13736 the PV of the ssv. The value which the offset points will be modified
13737 to the last input position on the ssv.
13739 Returns TRUE if the terminator was found, else returns FALSE.
13744 Perl_sv_cat_decode(pTHX_ SV *dsv, SV *encoding,
13745 SV *ssv, int *offset, char *tstr, int tlen)
13750 PERL_ARGS_ASSERT_SV_CAT_DECODE;
13752 if (SvPOK(ssv) && SvPOK(dsv) && SvROK(encoding) && offset) {
13763 offsv = newSViv(*offset);
13765 mXPUSHp(tstr, tlen);
13767 call_method("cat_decode", G_SCALAR);
13769 ret = SvTRUE(TOPs);
13770 *offset = SvIV(offsv);
13776 Perl_croak(aTHX_ "Invalid argument to sv_cat_decode");
13781 /* ---------------------------------------------------------------------
13783 * support functions for report_uninit()
13786 /* the maxiumum size of array or hash where we will scan looking
13787 * for the undefined element that triggered the warning */
13789 #define FUV_MAX_SEARCH_SIZE 1000
13791 /* Look for an entry in the hash whose value has the same SV as val;
13792 * If so, return a mortal copy of the key. */
13795 S_find_hash_subscript(pTHX_ const HV *const hv, const SV *const val)
13798 register HE **array;
13801 PERL_ARGS_ASSERT_FIND_HASH_SUBSCRIPT;
13803 if (!hv || SvMAGICAL(hv) || !HvARRAY(hv) ||
13804 (HvTOTALKEYS(hv) > FUV_MAX_SEARCH_SIZE))
13807 array = HvARRAY(hv);
13809 for (i=HvMAX(hv); i>0; i--) {
13810 register HE *entry;
13811 for (entry = array[i]; entry; entry = HeNEXT(entry)) {
13812 if (HeVAL(entry) != val)
13814 if ( HeVAL(entry) == &PL_sv_undef ||
13815 HeVAL(entry) == &PL_sv_placeholder)
13819 if (HeKLEN(entry) == HEf_SVKEY)
13820 return sv_mortalcopy(HeKEY_sv(entry));
13821 return sv_2mortal(newSVhek(HeKEY_hek(entry)));
13827 /* Look for an entry in the array whose value has the same SV as val;
13828 * If so, return the index, otherwise return -1. */
13831 S_find_array_subscript(pTHX_ const AV *const av, const SV *const val)
13835 PERL_ARGS_ASSERT_FIND_ARRAY_SUBSCRIPT;
13837 if (!av || SvMAGICAL(av) || !AvARRAY(av) ||
13838 (AvFILLp(av) > FUV_MAX_SEARCH_SIZE))
13841 if (val != &PL_sv_undef) {
13842 SV ** const svp = AvARRAY(av);
13845 for (i=AvFILLp(av); i>=0; i--)
13852 /* S_varname(): return the name of a variable, optionally with a subscript.
13853 * If gv is non-zero, use the name of that global, along with gvtype (one
13854 * of "$", "@", "%"); otherwise use the name of the lexical at pad offset
13855 * targ. Depending on the value of the subscript_type flag, return:
13858 #define FUV_SUBSCRIPT_NONE 1 /* "@foo" */
13859 #define FUV_SUBSCRIPT_ARRAY 2 /* "$foo[aindex]" */
13860 #define FUV_SUBSCRIPT_HASH 3 /* "$foo{keyname}" */
13861 #define FUV_SUBSCRIPT_WITHIN 4 /* "within @foo" */
13864 Perl_varname(pTHX_ const GV *const gv, const char gvtype, PADOFFSET targ,
13865 const SV *const keyname, I32 aindex, int subscript_type)
13868 SV * const name = sv_newmortal();
13869 if (gv && isGV(gv)) {
13871 buffer[0] = gvtype;
13874 /* as gv_fullname4(), but add literal '^' for $^FOO names */
13876 gv_fullname4(name, gv, buffer, 0);
13878 if ((unsigned int)SvPVX(name)[1] <= 26) {
13880 buffer[1] = SvPVX(name)[1] + 'A' - 1;
13882 /* Swap the 1 unprintable control character for the 2 byte pretty
13883 version - ie substr($name, 1, 1) = $buffer; */
13884 sv_insert(name, 1, 1, buffer, 2);
13888 CV * const cv = gv ? (CV *)gv : find_runcv(NULL);
13892 assert(!cv || SvTYPE(cv) == SVt_PVCV);
13894 if (!cv || !CvPADLIST(cv))
13896 av = MUTABLE_AV((*av_fetch(CvPADLIST(cv), 0, FALSE)));
13897 sv = *av_fetch(av, targ, FALSE);
13898 sv_setsv(name, sv);
13901 if (subscript_type == FUV_SUBSCRIPT_HASH) {
13902 SV * const sv = newSV(0);
13903 *SvPVX(name) = '$';
13904 Perl_sv_catpvf(aTHX_ name, "{%s}",
13905 pv_pretty(sv, SvPVX_const(keyname), SvCUR(keyname), 32, NULL, NULL,
13906 PERL_PV_PRETTY_DUMP | PERL_PV_ESCAPE_UNI_DETECT ));
13909 else if (subscript_type == FUV_SUBSCRIPT_ARRAY) {
13910 *SvPVX(name) = '$';
13911 Perl_sv_catpvf(aTHX_ name, "[%"IVdf"]", (IV)aindex);
13913 else if (subscript_type == FUV_SUBSCRIPT_WITHIN) {
13914 /* We know that name has no magic, so can use 0 instead of SV_GMAGIC */
13915 Perl_sv_insert_flags(aTHX_ name, 0, 0, STR_WITH_LEN("within "), 0);
13923 =for apidoc find_uninit_var
13925 Find the name of the undefined variable (if any) that caused the operator
13926 to issue a "Use of uninitialized value" warning.
13927 If match is true, only return a name if its value matches uninit_sv.
13928 So roughly speaking, if a unary operator (such as OP_COS) generates a
13929 warning, then following the direct child of the op may yield an
13930 OP_PADSV or OP_GV that gives the name of the undefined variable. On the
13931 other hand, with OP_ADD there are two branches to follow, so we only print
13932 the variable name if we get an exact match.
13934 The name is returned as a mortal SV.
13936 Assumes that PL_op is the op that originally triggered the error, and that
13937 PL_comppad/PL_curpad points to the currently executing pad.
13943 S_find_uninit_var(pTHX_ const OP *const obase, const SV *const uninit_sv,
13949 const OP *o, *o2, *kid;
13951 if (!obase || (match && (!uninit_sv || uninit_sv == &PL_sv_undef ||
13952 uninit_sv == &PL_sv_placeholder)))
13955 switch (obase->op_type) {
13962 const bool pad = (obase->op_type == OP_PADAV || obase->op_type == OP_PADHV);
13963 const bool hash = (obase->op_type == OP_PADHV || obase->op_type == OP_RV2HV);
13966 int subscript_type = FUV_SUBSCRIPT_WITHIN;
13968 if (pad) { /* @lex, %lex */
13969 sv = PAD_SVl(obase->op_targ);
13973 if (cUNOPx(obase)->op_first->op_type == OP_GV) {
13974 /* @global, %global */
13975 gv = cGVOPx_gv(cUNOPx(obase)->op_first);
13978 sv = hash ? MUTABLE_SV(GvHV(gv)): MUTABLE_SV(GvAV(gv));
13980 else if (obase == PL_op) /* @{expr}, %{expr} */
13981 return find_uninit_var(cUNOPx(obase)->op_first,
13983 else /* @{expr}, %{expr} as a sub-expression */
13987 /* attempt to find a match within the aggregate */
13989 keysv = find_hash_subscript((const HV*)sv, uninit_sv);
13991 subscript_type = FUV_SUBSCRIPT_HASH;
13994 index = find_array_subscript((const AV *)sv, uninit_sv);
13996 subscript_type = FUV_SUBSCRIPT_ARRAY;
13999 if (match && subscript_type == FUV_SUBSCRIPT_WITHIN)
14002 return varname(gv, hash ? '%' : '@', obase->op_targ,
14003 keysv, index, subscript_type);
14007 if (cUNOPx(obase)->op_first->op_type == OP_GV) {
14009 gv = cGVOPx_gv(cUNOPx(obase)->op_first);
14010 if (!gv || !GvSTASH(gv))
14012 if (match && (GvSV(gv) != uninit_sv))
14014 return varname(gv, '$', 0, NULL, 0, FUV_SUBSCRIPT_NONE);
14017 return find_uninit_var(cUNOPx(obase)->op_first, uninit_sv, 1);
14020 if (match && PAD_SVl(obase->op_targ) != uninit_sv)
14022 return varname(NULL, '$', obase->op_targ,
14023 NULL, 0, FUV_SUBSCRIPT_NONE);
14026 gv = cGVOPx_gv(obase);
14027 if (!gv || (match && GvSV(gv) != uninit_sv) || !GvSTASH(gv))
14029 return varname(gv, '$', 0, NULL, 0, FUV_SUBSCRIPT_NONE);
14031 case OP_AELEMFAST_LEX:
14034 AV *av = MUTABLE_AV(PAD_SV(obase->op_targ));
14035 if (!av || SvRMAGICAL(av))
14037 svp = av_fetch(av, (I32)obase->op_private, FALSE);
14038 if (!svp || *svp != uninit_sv)
14041 return varname(NULL, '$', obase->op_targ,
14042 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
14045 gv = cGVOPx_gv(obase);
14050 AV *const av = GvAV(gv);
14051 if (!av || SvRMAGICAL(av))
14053 svp = av_fetch(av, (I32)obase->op_private, FALSE);
14054 if (!svp || *svp != uninit_sv)
14057 return varname(gv, '$', 0,
14058 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
14063 o = cUNOPx(obase)->op_first;
14064 if (!o || o->op_type != OP_NULL ||
14065 ! (o->op_targ == OP_AELEM || o->op_targ == OP_HELEM))
14067 return find_uninit_var(cBINOPo->op_last, uninit_sv, match);
14072 bool negate = FALSE;
14074 if (PL_op == obase)
14075 /* $a[uninit_expr] or $h{uninit_expr} */
14076 return find_uninit_var(cBINOPx(obase)->op_last, uninit_sv, match);
14079 o = cBINOPx(obase)->op_first;
14080 kid = cBINOPx(obase)->op_last;
14082 /* get the av or hv, and optionally the gv */
14084 if (o->op_type == OP_PADAV || o->op_type == OP_PADHV) {
14085 sv = PAD_SV(o->op_targ);
14087 else if ((o->op_type == OP_RV2AV || o->op_type == OP_RV2HV)
14088 && cUNOPo->op_first->op_type == OP_GV)
14090 gv = cGVOPx_gv(cUNOPo->op_first);
14094 == OP_RV2HV ? MUTABLE_SV(GvHV(gv)) : MUTABLE_SV(GvAV(gv));
14099 if (kid && kid->op_type == OP_NEGATE) {
14101 kid = cUNOPx(kid)->op_first;
14104 if (kid && kid->op_type == OP_CONST && SvOK(cSVOPx_sv(kid))) {
14105 /* index is constant */
14108 kidsv = sv_2mortal(newSVpvs("-"));
14109 sv_catsv(kidsv, cSVOPx_sv(kid));
14112 kidsv = cSVOPx_sv(kid);
14116 if (obase->op_type == OP_HELEM) {
14117 HE* he = hv_fetch_ent(MUTABLE_HV(sv), kidsv, 0, 0);
14118 if (!he || HeVAL(he) != uninit_sv)
14122 SV * const * const svp = av_fetch(MUTABLE_AV(sv),
14123 negate ? - SvIV(cSVOPx_sv(kid)) : SvIV(cSVOPx_sv(kid)),
14125 if (!svp || *svp != uninit_sv)
14129 if (obase->op_type == OP_HELEM)
14130 return varname(gv, '%', o->op_targ,
14131 kidsv, 0, FUV_SUBSCRIPT_HASH);
14133 return varname(gv, '@', o->op_targ, NULL,
14134 negate ? - SvIV(cSVOPx_sv(kid)) : SvIV(cSVOPx_sv(kid)),
14135 FUV_SUBSCRIPT_ARRAY);
14138 /* index is an expression;
14139 * attempt to find a match within the aggregate */
14140 if (obase->op_type == OP_HELEM) {
14141 SV * const keysv = find_hash_subscript((const HV*)sv, uninit_sv);
14143 return varname(gv, '%', o->op_targ,
14144 keysv, 0, FUV_SUBSCRIPT_HASH);
14148 = find_array_subscript((const AV *)sv, uninit_sv);
14150 return varname(gv, '@', o->op_targ,
14151 NULL, index, FUV_SUBSCRIPT_ARRAY);
14156 (o->op_type == OP_PADAV || o->op_type == OP_RV2AV)
14158 o->op_targ, NULL, 0, FUV_SUBSCRIPT_WITHIN);
14164 /* only examine RHS */
14165 return find_uninit_var(cBINOPx(obase)->op_first, uninit_sv, match);
14168 o = cUNOPx(obase)->op_first;
14169 if (o->op_type == OP_PUSHMARK)
14172 if (!o->op_sibling) {
14173 /* one-arg version of open is highly magical */
14175 if (o->op_type == OP_GV) { /* open FOO; */
14177 if (match && GvSV(gv) != uninit_sv)
14179 return varname(gv, '$', 0,
14180 NULL, 0, FUV_SUBSCRIPT_NONE);
14182 /* other possibilities not handled are:
14183 * open $x; or open my $x; should return '${*$x}'
14184 * open expr; should return '$'.expr ideally
14190 /* ops where $_ may be an implicit arg */
14195 if ( !(obase->op_flags & OPf_STACKED)) {
14196 if (uninit_sv == ((obase->op_private & OPpTARGET_MY)
14197 ? PAD_SVl(obase->op_targ)
14200 sv = sv_newmortal();
14201 sv_setpvs(sv, "$_");
14210 match = 1; /* print etc can return undef on defined args */
14211 /* skip filehandle as it can't produce 'undef' warning */
14212 o = cUNOPx(obase)->op_first;
14213 if ((obase->op_flags & OPf_STACKED) && o->op_type == OP_PUSHMARK)
14214 o = o->op_sibling->op_sibling;
14218 case OP_ENTEREVAL: /* could be eval $undef or $x='$undef'; eval $x */
14219 case OP_CUSTOM: /* XS or custom code could trigger random warnings */
14221 /* the following ops are capable of returning PL_sv_undef even for
14222 * defined arg(s) */
14241 case OP_GETPEERNAME:
14289 case OP_SMARTMATCH:
14298 /* XXX tmp hack: these two may call an XS sub, and currently
14299 XS subs don't have a SUB entry on the context stack, so CV and
14300 pad determination goes wrong, and BAD things happen. So, just
14301 don't try to determine the value under those circumstances.
14302 Need a better fix at dome point. DAPM 11/2007 */
14308 GV * const gv = gv_fetchpvs(".", GV_NOTQUAL, SVt_PV);
14309 if (gv && GvSV(gv) == uninit_sv)
14310 return newSVpvs_flags("$.", SVs_TEMP);
14315 /* def-ness of rval pos() is independent of the def-ness of its arg */
14316 if ( !(obase->op_flags & OPf_MOD))
14321 if (SvROK(PL_rs) && uninit_sv == SvRV(PL_rs))
14322 return newSVpvs_flags("${$/}", SVs_TEMP);
14327 if (!(obase->op_flags & OPf_KIDS))
14329 o = cUNOPx(obase)->op_first;
14335 /* This loop checks all the kid ops, skipping any that cannot pos-
14336 * sibly be responsible for the uninitialized value; i.e., defined
14337 * constants and ops that return nothing. If there is only one op
14338 * left that is not skipped, then we *know* it is responsible for
14339 * the uninitialized value. If there is more than one op left, we
14340 * have to look for an exact match in the while() loop below.
14343 for (kid=o; kid; kid = kid->op_sibling) {
14345 const OPCODE type = kid->op_type;
14346 if ( (type == OP_CONST && SvOK(cSVOPx_sv(kid)))
14347 || (type == OP_NULL && ! (kid->op_flags & OPf_KIDS))
14348 || (type == OP_PUSHMARK)
14352 if (o2) { /* more than one found */
14359 return find_uninit_var(o2, uninit_sv, match);
14361 /* scan all args */
14363 sv = find_uninit_var(o, uninit_sv, 1);
14375 =for apidoc report_uninit
14377 Print appropriate "Use of uninitialized variable" warning.
14383 Perl_report_uninit(pTHX_ const SV *uninit_sv)
14387 SV* varname = NULL;
14388 if (uninit_sv && PL_curpad) {
14389 varname = find_uninit_var(PL_op, uninit_sv,0);
14391 sv_insert(varname, 0, 0, " ", 1);
14393 /* diag_listed_as: Use of uninitialized value%s */
14394 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit_sv,
14395 SVfARG(varname ? varname : &PL_sv_no),
14396 " in ", OP_DESC(PL_op));
14399 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
14405 * c-indentation-style: bsd
14406 * c-basic-offset: 4
14407 * indent-tabs-mode: t
14410 * ex: set ts=8 sts=4 sw=4 noet: