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 You generally want to use the C<SvUPGRADE> macro wrapper. See also C<svtype>.
1136 Perl_sv_upgrade(pTHX_ register SV *const sv, svtype new_type)
1141 const svtype old_type = SvTYPE(sv);
1142 const struct body_details *new_type_details;
1143 const struct body_details *old_type_details
1144 = bodies_by_type + old_type;
1145 SV *referant = NULL;
1147 PERL_ARGS_ASSERT_SV_UPGRADE;
1149 if (old_type == new_type)
1152 /* This clause was purposefully added ahead of the early return above to
1153 the shared string hackery for (sort {$a <=> $b} keys %hash), with the
1154 inference by Nick I-S that it would fix other troublesome cases. See
1155 changes 7162, 7163 (f130fd4589cf5fbb24149cd4db4137c8326f49c1 and parent)
1157 Given that shared hash key scalars are no longer PVIV, but PV, there is
1158 no longer need to unshare so as to free up the IVX slot for its proper
1159 purpose. So it's safe to move the early return earlier. */
1161 if (new_type != SVt_PV && SvIsCOW(sv)) {
1162 sv_force_normal_flags(sv, 0);
1165 old_body = SvANY(sv);
1167 /* Copying structures onto other structures that have been neatly zeroed
1168 has a subtle gotcha. Consider XPVMG
1170 +------+------+------+------+------+-------+-------+
1171 | NV | CUR | LEN | IV | MAGIC | STASH |
1172 +------+------+------+------+------+-------+-------+
1173 0 4 8 12 16 20 24 28
1175 where NVs are aligned to 8 bytes, so that sizeof that structure is
1176 actually 32 bytes long, with 4 bytes of padding at the end:
1178 +------+------+------+------+------+-------+-------+------+
1179 | NV | CUR | LEN | IV | MAGIC | STASH | ??? |
1180 +------+------+------+------+------+-------+-------+------+
1181 0 4 8 12 16 20 24 28 32
1183 so what happens if you allocate memory for this structure:
1185 +------+------+------+------+------+-------+-------+------+------+...
1186 | NV | CUR | LEN | IV | MAGIC | STASH | GP | NAME |
1187 +------+------+------+------+------+-------+-------+------+------+...
1188 0 4 8 12 16 20 24 28 32 36
1190 zero it, then copy sizeof(XPVMG) bytes on top of it? Not quite what you
1191 expect, because you copy the area marked ??? onto GP. Now, ??? may have
1192 started out as zero once, but it's quite possible that it isn't. So now,
1193 rather than a nicely zeroed GP, you have it pointing somewhere random.
1196 (In fact, GP ends up pointing at a previous GP structure, because the
1197 principle cause of the padding in XPVMG getting garbage is a copy of
1198 sizeof(XPVMG) bytes from a XPVGV structure in sv_unglob. Right now
1199 this happens to be moot because XPVGV has been re-ordered, with GP
1200 no longer after STASH)
1202 So we are careful and work out the size of used parts of all the
1210 referant = SvRV(sv);
1211 old_type_details = &fake_rv;
1212 if (new_type == SVt_NV)
1213 new_type = SVt_PVNV;
1215 if (new_type < SVt_PVIV) {
1216 new_type = (new_type == SVt_NV)
1217 ? SVt_PVNV : SVt_PVIV;
1222 if (new_type < SVt_PVNV) {
1223 new_type = SVt_PVNV;
1227 assert(new_type > SVt_PV);
1228 assert(SVt_IV < SVt_PV);
1229 assert(SVt_NV < SVt_PV);
1236 /* Because the XPVMG of PL_mess_sv isn't allocated from the arena,
1237 there's no way that it can be safely upgraded, because perl.c
1238 expects to Safefree(SvANY(PL_mess_sv)) */
1239 assert(sv != PL_mess_sv);
1240 /* This flag bit is used to mean other things in other scalar types.
1241 Given that it only has meaning inside the pad, it shouldn't be set
1242 on anything that can get upgraded. */
1243 assert(!SvPAD_TYPED(sv));
1246 if (old_type_details->cant_upgrade)
1247 Perl_croak(aTHX_ "Can't upgrade %s (%" UVuf ") to %" UVuf,
1248 sv_reftype(sv, 0), (UV) old_type, (UV) new_type);
1251 if (old_type > new_type)
1252 Perl_croak(aTHX_ "sv_upgrade from type %d down to type %d",
1253 (int)old_type, (int)new_type);
1255 new_type_details = bodies_by_type + new_type;
1257 SvFLAGS(sv) &= ~SVTYPEMASK;
1258 SvFLAGS(sv) |= new_type;
1260 /* This can't happen, as SVt_NULL is <= all values of new_type, so one of
1261 the return statements above will have triggered. */
1262 assert (new_type != SVt_NULL);
1265 assert(old_type == SVt_NULL);
1266 SvANY(sv) = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
1270 assert(old_type == SVt_NULL);
1271 SvANY(sv) = new_XNV();
1276 assert(new_type_details->body_size);
1279 assert(new_type_details->arena);
1280 assert(new_type_details->arena_size);
1281 /* This points to the start of the allocated area. */
1282 new_body_inline(new_body, new_type);
1283 Zero(new_body, new_type_details->body_size, char);
1284 new_body = ((char *)new_body) - new_type_details->offset;
1286 /* We always allocated the full length item with PURIFY. To do this
1287 we fake things so that arena is false for all 16 types.. */
1288 new_body = new_NOARENAZ(new_type_details);
1290 SvANY(sv) = new_body;
1291 if (new_type == SVt_PVAV) {
1295 if (old_type_details->body_size) {
1298 /* It will have been zeroed when the new body was allocated.
1299 Lets not write to it, in case it confuses a write-back
1305 #ifndef NODEFAULT_SHAREKEYS
1306 HvSHAREKEYS_on(sv); /* key-sharing on by default */
1308 HvMAX(sv) = 7; /* (start with 8 buckets) */
1311 /* SVt_NULL isn't the only thing upgraded to AV or HV.
1312 The target created by newSVrv also is, and it can have magic.
1313 However, it never has SvPVX set.
1315 if (old_type == SVt_IV) {
1317 } else if (old_type >= SVt_PV) {
1318 assert(SvPVX_const(sv) == 0);
1321 if (old_type >= SVt_PVMG) {
1322 SvMAGIC_set(sv, ((XPVMG*)old_body)->xmg_u.xmg_magic);
1323 SvSTASH_set(sv, ((XPVMG*)old_body)->xmg_stash);
1325 sv->sv_u.svu_array = NULL; /* or svu_hash */
1331 /* This ensures that SvTHINKFIRST(sv) is true, and hence that
1332 sv_force_normal_flags(sv) is called. */
1335 /* XXX Is this still needed? Was it ever needed? Surely as there is
1336 no route from NV to PVIV, NOK can never be true */
1337 assert(!SvNOKp(sv));
1348 assert(new_type_details->body_size);
1349 /* We always allocated the full length item with PURIFY. To do this
1350 we fake things so that arena is false for all 16 types.. */
1351 if(new_type_details->arena) {
1352 /* This points to the start of the allocated area. */
1353 new_body_inline(new_body, new_type);
1354 Zero(new_body, new_type_details->body_size, char);
1355 new_body = ((char *)new_body) - new_type_details->offset;
1357 new_body = new_NOARENAZ(new_type_details);
1359 SvANY(sv) = new_body;
1361 if (old_type_details->copy) {
1362 /* There is now the potential for an upgrade from something without
1363 an offset (PVNV or PVMG) to something with one (PVCV, PVFM) */
1364 int offset = old_type_details->offset;
1365 int length = old_type_details->copy;
1367 if (new_type_details->offset > old_type_details->offset) {
1368 const int difference
1369 = new_type_details->offset - old_type_details->offset;
1370 offset += difference;
1371 length -= difference;
1373 assert (length >= 0);
1375 Copy((char *)old_body + offset, (char *)new_body + offset, length,
1379 #ifndef NV_ZERO_IS_ALLBITS_ZERO
1380 /* If NV 0.0 is stores as all bits 0 then Zero() already creates a
1381 * correct 0.0 for us. Otherwise, if the old body didn't have an
1382 * NV slot, but the new one does, then we need to initialise the
1383 * freshly created NV slot with whatever the correct bit pattern is
1385 if (old_type_details->zero_nv && !new_type_details->zero_nv
1386 && !isGV_with_GP(sv))
1390 if (new_type == SVt_PVIO) {
1391 IO * const io = MUTABLE_IO(sv);
1392 GV *iogv = gv_fetchpvs("IO::File::", GV_ADD, SVt_PVHV);
1395 /* Clear the stashcache because a new IO could overrule a package
1397 hv_clear(PL_stashcache);
1399 SvSTASH_set(io, MUTABLE_HV(SvREFCNT_inc(GvHV(iogv))));
1400 IoPAGE_LEN(sv) = 60;
1402 if (old_type < SVt_PV) {
1403 /* referant will be NULL unless the old type was SVt_IV emulating
1405 sv->sv_u.svu_rv = referant;
1409 Perl_croak(aTHX_ "panic: sv_upgrade to unknown type %lu",
1410 (unsigned long)new_type);
1413 if (old_type > SVt_IV) {
1417 /* Note that there is an assumption that all bodies of types that
1418 can be upgraded came from arenas. Only the more complex non-
1419 upgradable types are allowed to be directly malloc()ed. */
1420 assert(old_type_details->arena);
1421 del_body((void*)((char*)old_body + old_type_details->offset),
1422 &PL_body_roots[old_type]);
1428 =for apidoc sv_backoff
1430 Remove any string offset. You should normally use the C<SvOOK_off> macro
1437 Perl_sv_backoff(pTHX_ register SV *const sv)
1440 const char * const s = SvPVX_const(sv);
1442 PERL_ARGS_ASSERT_SV_BACKOFF;
1443 PERL_UNUSED_CONTEXT;
1446 assert(SvTYPE(sv) != SVt_PVHV);
1447 assert(SvTYPE(sv) != SVt_PVAV);
1449 SvOOK_offset(sv, delta);
1451 SvLEN_set(sv, SvLEN(sv) + delta);
1452 SvPV_set(sv, SvPVX(sv) - delta);
1453 Move(s, SvPVX(sv), SvCUR(sv)+1, char);
1454 SvFLAGS(sv) &= ~SVf_OOK;
1461 Expands the character buffer in the SV. If necessary, uses C<sv_unref> and
1462 upgrades the SV to C<SVt_PV>. Returns a pointer to the character buffer.
1463 Use the C<SvGROW> wrapper instead.
1469 Perl_sv_grow(pTHX_ register SV *const sv, register STRLEN newlen)
1473 PERL_ARGS_ASSERT_SV_GROW;
1475 if (PL_madskills && newlen >= 0x100000) {
1476 PerlIO_printf(Perl_debug_log,
1477 "Allocation too large: %"UVxf"\n", (UV)newlen);
1479 #ifdef HAS_64K_LIMIT
1480 if (newlen >= 0x10000) {
1481 PerlIO_printf(Perl_debug_log,
1482 "Allocation too large: %"UVxf"\n", (UV)newlen);
1485 #endif /* HAS_64K_LIMIT */
1488 if (SvTYPE(sv) < SVt_PV) {
1489 sv_upgrade(sv, SVt_PV);
1490 s = SvPVX_mutable(sv);
1492 else if (SvOOK(sv)) { /* pv is offset? */
1494 s = SvPVX_mutable(sv);
1495 if (newlen > SvLEN(sv))
1496 newlen += 10 * (newlen - SvCUR(sv)); /* avoid copy each time */
1497 #ifdef HAS_64K_LIMIT
1498 if (newlen >= 0x10000)
1503 s = SvPVX_mutable(sv);
1505 if (newlen > SvLEN(sv)) { /* need more room? */
1506 STRLEN minlen = SvCUR(sv);
1507 minlen += (minlen >> PERL_STRLEN_EXPAND_SHIFT) + 10;
1508 if (newlen < minlen)
1510 #ifndef Perl_safesysmalloc_size
1511 newlen = PERL_STRLEN_ROUNDUP(newlen);
1513 if (SvLEN(sv) && s) {
1514 s = (char*)saferealloc(s, newlen);
1517 s = (char*)safemalloc(newlen);
1518 if (SvPVX_const(sv) && SvCUR(sv)) {
1519 Move(SvPVX_const(sv), s, (newlen < SvCUR(sv)) ? newlen : SvCUR(sv), char);
1523 #ifdef Perl_safesysmalloc_size
1524 /* Do this here, do it once, do it right, and then we will never get
1525 called back into sv_grow() unless there really is some growing
1527 SvLEN_set(sv, Perl_safesysmalloc_size(s));
1529 SvLEN_set(sv, newlen);
1536 =for apidoc sv_setiv
1538 Copies an integer into the given SV, upgrading first if necessary.
1539 Does not handle 'set' magic. See also C<sv_setiv_mg>.
1545 Perl_sv_setiv(pTHX_ register SV *const sv, const IV i)
1549 PERL_ARGS_ASSERT_SV_SETIV;
1551 SV_CHECK_THINKFIRST_COW_DROP(sv);
1552 switch (SvTYPE(sv)) {
1555 sv_upgrade(sv, SVt_IV);
1558 sv_upgrade(sv, SVt_PVIV);
1562 if (!isGV_with_GP(sv))
1569 /* diag_listed_as: Can't coerce %s to %s in %s */
1570 Perl_croak(aTHX_ "Can't coerce %s to integer in %s", sv_reftype(sv,0),
1574 (void)SvIOK_only(sv); /* validate number */
1580 =for apidoc sv_setiv_mg
1582 Like C<sv_setiv>, but also handles 'set' magic.
1588 Perl_sv_setiv_mg(pTHX_ register SV *const sv, const IV i)
1590 PERL_ARGS_ASSERT_SV_SETIV_MG;
1597 =for apidoc sv_setuv
1599 Copies an unsigned integer into the given SV, upgrading first if necessary.
1600 Does not handle 'set' magic. See also C<sv_setuv_mg>.
1606 Perl_sv_setuv(pTHX_ register SV *const sv, const UV u)
1608 PERL_ARGS_ASSERT_SV_SETUV;
1610 /* With these two if statements:
1611 u=1.49 s=0.52 cu=72.49 cs=10.64 scripts=270 tests=20865
1614 u=1.35 s=0.47 cu=73.45 cs=11.43 scripts=270 tests=20865
1616 If you wish to remove them, please benchmark to see what the effect is
1618 if (u <= (UV)IV_MAX) {
1619 sv_setiv(sv, (IV)u);
1628 =for apidoc sv_setuv_mg
1630 Like C<sv_setuv>, but also handles 'set' magic.
1636 Perl_sv_setuv_mg(pTHX_ register SV *const sv, const UV u)
1638 PERL_ARGS_ASSERT_SV_SETUV_MG;
1645 =for apidoc sv_setnv
1647 Copies a double into the given SV, upgrading first if necessary.
1648 Does not handle 'set' magic. See also C<sv_setnv_mg>.
1654 Perl_sv_setnv(pTHX_ register SV *const sv, const NV num)
1658 PERL_ARGS_ASSERT_SV_SETNV;
1660 SV_CHECK_THINKFIRST_COW_DROP(sv);
1661 switch (SvTYPE(sv)) {
1664 sv_upgrade(sv, SVt_NV);
1668 sv_upgrade(sv, SVt_PVNV);
1672 if (!isGV_with_GP(sv))
1679 /* diag_listed_as: Can't coerce %s to %s in %s */
1680 Perl_croak(aTHX_ "Can't coerce %s to number in %s", sv_reftype(sv,0),
1685 (void)SvNOK_only(sv); /* validate number */
1690 =for apidoc sv_setnv_mg
1692 Like C<sv_setnv>, but also handles 'set' magic.
1698 Perl_sv_setnv_mg(pTHX_ register SV *const sv, const NV num)
1700 PERL_ARGS_ASSERT_SV_SETNV_MG;
1706 /* Print an "isn't numeric" warning, using a cleaned-up,
1707 * printable version of the offending string
1711 S_not_a_number(pTHX_ SV *const sv)
1718 PERL_ARGS_ASSERT_NOT_A_NUMBER;
1721 dsv = newSVpvs_flags("", SVs_TEMP);
1722 pv = sv_uni_display(dsv, sv, 10, UNI_DISPLAY_ISPRINT);
1725 const char * const limit = tmpbuf + sizeof(tmpbuf) - 8;
1726 /* each *s can expand to 4 chars + "...\0",
1727 i.e. need room for 8 chars */
1729 const char *s = SvPVX_const(sv);
1730 const char * const end = s + SvCUR(sv);
1731 for ( ; s < end && d < limit; s++ ) {
1733 if (ch & 128 && !isPRINT_LC(ch)) {
1742 else if (ch == '\r') {
1746 else if (ch == '\f') {
1750 else if (ch == '\\') {
1754 else if (ch == '\0') {
1758 else if (isPRINT_LC(ch))
1775 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
1776 "Argument \"%s\" isn't numeric in %s", pv,
1779 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
1780 "Argument \"%s\" isn't numeric", pv);
1784 =for apidoc looks_like_number
1786 Test if the content of an SV looks like a number (or is a number).
1787 C<Inf> and C<Infinity> are treated as numbers (so will not issue a
1788 non-numeric warning), even if your atof() doesn't grok them. Get-magic is
1795 Perl_looks_like_number(pTHX_ SV *const sv)
1797 register const char *sbegin;
1800 PERL_ARGS_ASSERT_LOOKS_LIKE_NUMBER;
1802 if (SvPOK(sv) || SvPOKp(sv)) {
1803 sbegin = SvPV_nomg_const(sv, len);
1806 return SvFLAGS(sv) & (SVf_NOK|SVp_NOK|SVf_IOK|SVp_IOK);
1807 return grok_number(sbegin, len, NULL);
1811 S_glob_2number(pTHX_ GV * const gv)
1813 const U32 wasfake = SvFLAGS(gv) & SVf_FAKE;
1814 SV *const buffer = sv_newmortal();
1816 PERL_ARGS_ASSERT_GLOB_2NUMBER;
1818 /* FAKE globs can get coerced, so need to turn this off temporarily if it
1821 gv_efullname3(buffer, gv, "*");
1822 SvFLAGS(gv) |= wasfake;
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 string
2706 Normally invoked via the C<SvPV_flags> macro. C<sv_2pv()> and C<sv_2pv_nomg>
2707 usually end up here too.
2713 Perl_sv_2pv_flags(pTHX_ register SV *const sv, STRLEN *const lp, const I32 flags)
2723 if (SvGMAGICAL(sv)) {
2724 if (flags & SV_GMAGIC)
2729 if (flags & SV_MUTABLE_RETURN)
2730 return SvPVX_mutable(sv);
2731 if (flags & SV_CONST_RETURN)
2732 return (char *)SvPVX_const(sv);
2735 if (SvIOKp(sv) || SvNOKp(sv)) {
2736 char tbuf[64]; /* Must fit sprintf/Gconvert of longest IV/NV */
2741 ? my_snprintf(tbuf, sizeof(tbuf), "%"UVuf, (UV)SvUVX(sv))
2742 : my_snprintf(tbuf, sizeof(tbuf), "%"IVdf, (IV)SvIVX(sv));
2743 } else if(SvNVX(sv) == 0.0) {
2748 Gconvert(SvNVX(sv), NV_DIG, 0, tbuf);
2755 SvUPGRADE(sv, SVt_PV);
2758 s = SvGROW_mutable(sv, len + 1);
2761 return (char*)memcpy(s, tbuf, len + 1);
2767 assert(SvTYPE(sv) >= SVt_PVMG);
2768 /* This falls through to the report_uninit near the end of the
2770 } else if (SvTHINKFIRST(sv)) {
2775 if (flags & SV_SKIP_OVERLOAD)
2777 tmpstr = AMG_CALLunary(sv, string_amg);
2778 TAINT_IF(tmpstr && SvTAINTED(tmpstr));
2779 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2781 /* char *pv = lp ? SvPV(tmpstr, *lp) : SvPV_nolen(tmpstr);
2785 if ((SvFLAGS(tmpstr) & (SVf_POK)) == SVf_POK) {
2786 if (flags & SV_CONST_RETURN) {
2787 pv = (char *) SvPVX_const(tmpstr);
2789 pv = (flags & SV_MUTABLE_RETURN)
2790 ? SvPVX_mutable(tmpstr) : SvPVX(tmpstr);
2793 *lp = SvCUR(tmpstr);
2795 pv = sv_2pv_flags(tmpstr, lp, flags);
2808 SV *const referent = SvRV(sv);
2812 retval = buffer = savepvn("NULLREF", len);
2813 } else if (SvTYPE(referent) == SVt_REGEXP) {
2814 REGEXP * const re = (REGEXP *)MUTABLE_PTR(referent);
2819 /* If the regex is UTF-8 we want the containing scalar to
2820 have an UTF-8 flag too */
2826 if ((seen_evals = RX_SEEN_EVALS(re)))
2827 PL_reginterp_cnt += seen_evals;
2830 *lp = RX_WRAPLEN(re);
2832 return RX_WRAPPED(re);
2834 const char *const typestr = sv_reftype(referent, 0);
2835 const STRLEN typelen = strlen(typestr);
2836 UV addr = PTR2UV(referent);
2837 const char *stashname = NULL;
2838 STRLEN stashnamelen = 0; /* hush, gcc */
2839 const char *buffer_end;
2841 if (SvOBJECT(referent)) {
2842 const HEK *const name = HvNAME_HEK(SvSTASH(referent));
2845 stashname = HEK_KEY(name);
2846 stashnamelen = HEK_LEN(name);
2848 if (HEK_UTF8(name)) {
2854 stashname = "__ANON__";
2857 len = stashnamelen + 1 /* = */ + typelen + 3 /* (0x */
2858 + 2 * sizeof(UV) + 2 /* )\0 */;
2860 len = typelen + 3 /* (0x */
2861 + 2 * sizeof(UV) + 2 /* )\0 */;
2864 Newx(buffer, len, char);
2865 buffer_end = retval = buffer + len;
2867 /* Working backwards */
2871 *--retval = PL_hexdigit[addr & 15];
2872 } while (addr >>= 4);
2878 memcpy(retval, typestr, typelen);
2882 retval -= stashnamelen;
2883 memcpy(retval, stashname, stashnamelen);
2885 /* retval may not necessarily have reached the start of the
2887 assert (retval >= buffer);
2889 len = buffer_end - retval - 1; /* -1 for that \0 */
2897 if (SvREADONLY(sv) && !SvOK(sv)) {
2900 if (flags & SV_UNDEF_RETURNS_NULL)
2902 if (ckWARN(WARN_UNINITIALIZED))
2907 if (SvIOK(sv) || ((SvIOKp(sv) && !SvNOKp(sv)))) {
2908 /* I'm assuming that if both IV and NV are equally valid then
2909 converting the IV is going to be more efficient */
2910 const U32 isUIOK = SvIsUV(sv);
2911 char buf[TYPE_CHARS(UV)];
2915 if (SvTYPE(sv) < SVt_PVIV)
2916 sv_upgrade(sv, SVt_PVIV);
2917 ptr = uiv_2buf(buf, SvIVX(sv), SvUVX(sv), isUIOK, &ebuf);
2919 /* inlined from sv_setpvn */
2920 s = SvGROW_mutable(sv, len + 1);
2921 Move(ptr, s, len, char);
2925 else if (SvNOKp(sv)) {
2926 if (SvTYPE(sv) < SVt_PVNV)
2927 sv_upgrade(sv, SVt_PVNV);
2928 if (SvNVX(sv) == 0.0) {
2929 s = SvGROW_mutable(sv, 2);
2934 /* The +20 is pure guesswork. Configure test needed. --jhi */
2935 s = SvGROW_mutable(sv, NV_DIG + 20);
2936 /* some Xenix systems wipe out errno here */
2937 Gconvert(SvNVX(sv), NV_DIG, 0, s);
2947 if (isGV_with_GP(sv)) {
2948 GV *const gv = MUTABLE_GV(sv);
2949 const U32 wasfake = SvFLAGS(gv) & SVf_FAKE;
2950 SV *const buffer = sv_newmortal();
2952 /* FAKE globs can get coerced, so need to turn this off temporarily
2955 gv_efullname3(buffer, gv, "*");
2956 SvFLAGS(gv) |= wasfake;
2958 if (SvPOK(buffer)) {
2960 *lp = SvCUR(buffer);
2962 if ( SvUTF8(buffer) ) SvUTF8_on(sv);
2963 return SvPVX(buffer);
2974 if (flags & SV_UNDEF_RETURNS_NULL)
2976 if (!PL_localizing && !SvPADTMP(sv) && ckWARN(WARN_UNINITIALIZED))
2978 if (SvTYPE(sv) < SVt_PV)
2979 /* Typically the caller expects that sv_any is not NULL now. */
2980 sv_upgrade(sv, SVt_PV);
2984 const STRLEN len = s - SvPVX_const(sv);
2990 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
2991 PTR2UV(sv),SvPVX_const(sv)));
2992 if (flags & SV_CONST_RETURN)
2993 return (char *)SvPVX_const(sv);
2994 if (flags & SV_MUTABLE_RETURN)
2995 return SvPVX_mutable(sv);
3000 =for apidoc sv_copypv
3002 Copies a stringified representation of the source SV into the
3003 destination SV. Automatically performs any necessary mg_get and
3004 coercion of numeric values into strings. Guaranteed to preserve
3005 UTF8 flag even from overloaded objects. Similar in nature to
3006 sv_2pv[_flags] but operates directly on an SV instead of just the
3007 string. Mostly uses sv_2pv_flags to do its work, except when that
3008 would lose the UTF-8'ness of the PV.
3014 Perl_sv_copypv(pTHX_ SV *const dsv, register SV *const ssv)
3017 const char * const s = SvPV_const(ssv,len);
3019 PERL_ARGS_ASSERT_SV_COPYPV;
3021 sv_setpvn(dsv,s,len);
3029 =for apidoc sv_2pvbyte
3031 Return a pointer to the byte-encoded representation of the SV, and set *lp
3032 to its length. May cause the SV to be downgraded from UTF-8 as a
3035 Usually accessed via the C<SvPVbyte> macro.
3041 Perl_sv_2pvbyte(pTHX_ register SV *const sv, STRLEN *const lp)
3043 PERL_ARGS_ASSERT_SV_2PVBYTE;
3046 sv_utf8_downgrade(sv,0);
3047 return lp ? SvPV_nomg(sv,*lp) : SvPV_nomg_nolen(sv);
3051 =for apidoc sv_2pvutf8
3053 Return a pointer to the UTF-8-encoded representation of the SV, and set *lp
3054 to its length. May cause the SV to be upgraded to UTF-8 as a side-effect.
3056 Usually accessed via the C<SvPVutf8> macro.
3062 Perl_sv_2pvutf8(pTHX_ register SV *const sv, STRLEN *const lp)
3064 PERL_ARGS_ASSERT_SV_2PVUTF8;
3066 sv_utf8_upgrade(sv);
3067 return lp ? SvPV(sv,*lp) : SvPV_nolen(sv);
3072 =for apidoc sv_2bool
3074 This macro is only used by sv_true() or its macro equivalent, and only if
3075 the latter's argument is neither SvPOK, SvIOK nor SvNOK.
3076 It calls sv_2bool_flags with the SV_GMAGIC flag.
3078 =for apidoc sv_2bool_flags
3080 This function is only used by sv_true() and friends, and only if
3081 the latter's argument is neither SvPOK, SvIOK nor SvNOK. If the flags
3082 contain SV_GMAGIC, then it does an mg_get() first.
3089 Perl_sv_2bool_flags(pTHX_ register SV *const sv, const I32 flags)
3093 PERL_ARGS_ASSERT_SV_2BOOL_FLAGS;
3095 if(flags & SV_GMAGIC) SvGETMAGIC(sv);
3101 SV * const tmpsv = AMG_CALLunary(sv, bool__amg);
3102 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
3103 return cBOOL(SvTRUE(tmpsv));
3105 return SvRV(sv) != 0;
3108 register XPV* const Xpvtmp = (XPV*)SvANY(sv);
3110 (*sv->sv_u.svu_pv > '0' ||
3111 Xpvtmp->xpv_cur > 1 ||
3112 (Xpvtmp->xpv_cur && *sv->sv_u.svu_pv != '0')))
3119 return SvIVX(sv) != 0;
3122 return SvNVX(sv) != 0.0;
3124 if (isGV_with_GP(sv))
3134 =for apidoc sv_utf8_upgrade
3136 Converts the PV of an SV to its UTF-8-encoded form.
3137 Forces the SV to string form if it is not already.
3138 Will C<mg_get> on C<sv> if appropriate.
3139 Always sets the SvUTF8 flag to avoid future validity checks even
3140 if the whole string is the same in UTF-8 as not.
3141 Returns the number of bytes in the converted string
3143 This is not as a general purpose byte encoding to Unicode interface:
3144 use the Encode extension for that.
3146 =for apidoc sv_utf8_upgrade_nomg
3148 Like sv_utf8_upgrade, but doesn't do magic on C<sv>
3150 =for apidoc sv_utf8_upgrade_flags
3152 Converts the PV of an SV to its UTF-8-encoded form.
3153 Forces the SV to string form if it is not already.
3154 Always sets the SvUTF8 flag to avoid future validity checks even
3155 if all the bytes are invariant in UTF-8. If C<flags> has C<SV_GMAGIC> bit set,
3156 will C<mg_get> on C<sv> if appropriate, else not.
3157 Returns the number of bytes in the converted string
3158 C<sv_utf8_upgrade> and
3159 C<sv_utf8_upgrade_nomg> are implemented in terms of this function.
3161 This is not as a general purpose byte encoding to Unicode interface:
3162 use the Encode extension for that.
3166 The grow version is currently not externally documented. It adds a parameter,
3167 extra, which is the number of unused bytes the string of 'sv' is guaranteed to
3168 have free after it upon return. This allows the caller to reserve extra space
3169 that it intends to fill, to avoid extra grows.
3171 Also externally undocumented for the moment is the flag SV_FORCE_UTF8_UPGRADE,
3172 which can be used to tell this function to not first check to see if there are
3173 any characters that are different in UTF-8 (variant characters) which would
3174 force it to allocate a new string to sv, but to assume there are. Typically
3175 this flag is used by a routine that has already parsed the string to find that
3176 there are such characters, and passes this information on so that the work
3177 doesn't have to be repeated.
3179 (One might think that the calling routine could pass in the position of the
3180 first such variant, so it wouldn't have to be found again. But that is not the
3181 case, because typically when the caller is likely to use this flag, it won't be
3182 calling this routine unless it finds something that won't fit into a byte.
3183 Otherwise it tries to not upgrade and just use bytes. But some things that
3184 do fit into a byte are variants in utf8, and the caller may not have been
3185 keeping track of these.)
3187 If the routine itself changes the string, it adds a trailing NUL. Such a NUL
3188 isn't guaranteed due to having other routines do the work in some input cases,
3189 or if the input is already flagged as being in utf8.
3191 The speed of this could perhaps be improved for many cases if someone wanted to
3192 write a fast function that counts the number of variant characters in a string,
3193 especially if it could return the position of the first one.
3198 Perl_sv_utf8_upgrade_flags_grow(pTHX_ register SV *const sv, const I32 flags, STRLEN extra)
3202 PERL_ARGS_ASSERT_SV_UTF8_UPGRADE_FLAGS_GROW;
3204 if (sv == &PL_sv_undef)
3208 if (SvREADONLY(sv) && (SvPOKp(sv) || SvIOKp(sv) || SvNOKp(sv))) {
3209 (void) sv_2pv_flags(sv,&len, flags);
3211 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3215 (void) SvPV_force_flags(sv,len,flags & SV_GMAGIC);
3220 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3225 sv_force_normal_flags(sv, 0);
3228 if (PL_encoding && !(flags & SV_UTF8_NO_ENCODING)) {
3229 sv_recode_to_utf8(sv, PL_encoding);
3230 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3234 if (SvCUR(sv) == 0) {
3235 if (extra) SvGROW(sv, extra);
3236 } else { /* Assume Latin-1/EBCDIC */
3237 /* This function could be much more efficient if we
3238 * had a FLAG in SVs to signal if there are any variant
3239 * chars in the PV. Given that there isn't such a flag
3240 * make the loop as fast as possible (although there are certainly ways
3241 * to speed this up, eg. through vectorization) */
3242 U8 * s = (U8 *) SvPVX_const(sv);
3243 U8 * e = (U8 *) SvEND(sv);
3245 STRLEN two_byte_count = 0;
3247 if (flags & SV_FORCE_UTF8_UPGRADE) goto must_be_utf8;
3249 /* See if really will need to convert to utf8. We mustn't rely on our
3250 * incoming SV being well formed and having a trailing '\0', as certain
3251 * code in pp_formline can send us partially built SVs. */
3255 if (NATIVE_IS_INVARIANT(ch)) continue;
3257 t--; /* t already incremented; re-point to first variant */
3262 /* utf8 conversion not needed because all are invariants. Mark as
3263 * UTF-8 even if no variant - saves scanning loop */
3265 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3270 /* Here, the string should be converted to utf8, either because of an
3271 * input flag (two_byte_count = 0), or because a character that
3272 * requires 2 bytes was found (two_byte_count = 1). t points either to
3273 * the beginning of the string (if we didn't examine anything), or to
3274 * the first variant. In either case, everything from s to t - 1 will
3275 * occupy only 1 byte each on output.
3277 * There are two main ways to convert. One is to create a new string
3278 * and go through the input starting from the beginning, appending each
3279 * converted value onto the new string as we go along. It's probably
3280 * best to allocate enough space in the string for the worst possible
3281 * case rather than possibly running out of space and having to
3282 * reallocate and then copy what we've done so far. Since everything
3283 * from s to t - 1 is invariant, the destination can be initialized
3284 * with these using a fast memory copy
3286 * The other way is to figure out exactly how big the string should be
3287 * by parsing the entire input. Then you don't have to make it big
3288 * enough to handle the worst possible case, and more importantly, if
3289 * the string you already have is large enough, you don't have to
3290 * allocate a new string, you can copy the last character in the input
3291 * string to the final position(s) that will be occupied by the
3292 * converted string and go backwards, stopping at t, since everything
3293 * before that is invariant.
3295 * There are advantages and disadvantages to each method.
3297 * In the first method, we can allocate a new string, do the memory
3298 * copy from the s to t - 1, and then proceed through the rest of the
3299 * string byte-by-byte.
3301 * In the second method, we proceed through the rest of the input
3302 * string just calculating how big the converted string will be. Then
3303 * there are two cases:
3304 * 1) if the string has enough extra space to handle the converted
3305 * value. We go backwards through the string, converting until we
3306 * get to the position we are at now, and then stop. If this
3307 * position is far enough along in the string, this method is
3308 * faster than the other method. If the memory copy were the same
3309 * speed as the byte-by-byte loop, that position would be about
3310 * half-way, as at the half-way mark, parsing to the end and back
3311 * is one complete string's parse, the same amount as starting
3312 * over and going all the way through. Actually, it would be
3313 * somewhat less than half-way, as it's faster to just count bytes
3314 * than to also copy, and we don't have the overhead of allocating
3315 * a new string, changing the scalar to use it, and freeing the
3316 * existing one. But if the memory copy is fast, the break-even
3317 * point is somewhere after half way. The counting loop could be
3318 * sped up by vectorization, etc, to move the break-even point
3319 * further towards the beginning.
3320 * 2) if the string doesn't have enough space to handle the converted
3321 * value. A new string will have to be allocated, and one might
3322 * as well, given that, start from the beginning doing the first
3323 * method. We've spent extra time parsing the string and in
3324 * exchange all we've gotten is that we know precisely how big to
3325 * make the new one. Perl is more optimized for time than space,
3326 * so this case is a loser.
3327 * So what I've decided to do is not use the 2nd method unless it is
3328 * guaranteed that a new string won't have to be allocated, assuming
3329 * the worst case. I also decided not to put any more conditions on it
3330 * than this, for now. It seems likely that, since the worst case is
3331 * twice as big as the unknown portion of the string (plus 1), we won't
3332 * be guaranteed enough space, causing us to go to the first method,
3333 * unless the string is short, or the first variant character is near
3334 * the end of it. In either of these cases, it seems best to use the
3335 * 2nd method. The only circumstance I can think of where this would
3336 * be really slower is if the string had once had much more data in it
3337 * than it does now, but there is still a substantial amount in it */
3340 STRLEN invariant_head = t - s;
3341 STRLEN size = invariant_head + (e - t) * 2 + 1 + extra;
3342 if (SvLEN(sv) < size) {
3344 /* Here, have decided to allocate a new string */
3349 Newx(dst, size, U8);
3351 /* If no known invariants at the beginning of the input string,
3352 * set so starts from there. Otherwise, can use memory copy to
3353 * get up to where we are now, and then start from here */
3355 if (invariant_head <= 0) {
3358 Copy(s, dst, invariant_head, char);
3359 d = dst + invariant_head;
3363 const UV uv = NATIVE8_TO_UNI(*t++);
3364 if (UNI_IS_INVARIANT(uv))
3365 *d++ = (U8)UNI_TO_NATIVE(uv);
3367 *d++ = (U8)UTF8_EIGHT_BIT_HI(uv);
3368 *d++ = (U8)UTF8_EIGHT_BIT_LO(uv);
3372 SvPV_free(sv); /* No longer using pre-existing string */
3373 SvPV_set(sv, (char*)dst);
3374 SvCUR_set(sv, d - dst);
3375 SvLEN_set(sv, size);
3378 /* Here, have decided to get the exact size of the string.
3379 * Currently this happens only when we know that there is
3380 * guaranteed enough space to fit the converted string, so
3381 * don't have to worry about growing. If two_byte_count is 0,
3382 * then t points to the first byte of the string which hasn't
3383 * been examined yet. Otherwise two_byte_count is 1, and t
3384 * points to the first byte in the string that will expand to
3385 * two. Depending on this, start examining at t or 1 after t.
3388 U8 *d = t + two_byte_count;
3391 /* Count up the remaining bytes that expand to two */
3394 const U8 chr = *d++;
3395 if (! NATIVE_IS_INVARIANT(chr)) two_byte_count++;
3398 /* The string will expand by just the number of bytes that
3399 * occupy two positions. But we are one afterwards because of
3400 * the increment just above. This is the place to put the
3401 * trailing NUL, and to set the length before we decrement */
3403 d += two_byte_count;
3404 SvCUR_set(sv, d - s);
3408 /* Having decremented d, it points to the position to put the
3409 * very last byte of the expanded string. Go backwards through
3410 * the string, copying and expanding as we go, stopping when we
3411 * get to the part that is invariant the rest of the way down */
3415 const U8 ch = NATIVE8_TO_UNI(*e--);
3416 if (UNI_IS_INVARIANT(ch)) {
3417 *d-- = UNI_TO_NATIVE(ch);
3419 *d-- = (U8)UTF8_EIGHT_BIT_LO(ch);
3420 *d-- = (U8)UTF8_EIGHT_BIT_HI(ch);
3425 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
3426 /* Update pos. We do it at the end rather than during
3427 * the upgrade, to avoid slowing down the common case
3428 * (upgrade without pos) */
3429 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
3431 I32 pos = mg->mg_len;
3432 if (pos > 0 && (U32)pos > invariant_head) {
3433 U8 *d = (U8*) SvPVX(sv) + invariant_head;
3434 STRLEN n = (U32)pos - invariant_head;
3436 if (UTF8_IS_START(*d))
3441 mg->mg_len = d - (U8*)SvPVX(sv);
3444 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
3445 magic_setutf8(sv,mg); /* clear UTF8 cache */
3450 /* Mark as UTF-8 even if no variant - saves scanning loop */
3456 =for apidoc sv_utf8_downgrade
3458 Attempts to convert the PV of an SV from characters to bytes.
3459 If the PV contains a character that cannot fit
3460 in a byte, this conversion will fail;
3461 in this case, either returns false or, if C<fail_ok> is not
3464 This is not as a general purpose Unicode to byte encoding interface:
3465 use the Encode extension for that.
3471 Perl_sv_utf8_downgrade(pTHX_ register SV *const sv, const bool fail_ok)
3475 PERL_ARGS_ASSERT_SV_UTF8_DOWNGRADE;
3477 if (SvPOKp(sv) && SvUTF8(sv)) {
3481 int mg_flags = SV_GMAGIC;
3484 sv_force_normal_flags(sv, 0);
3486 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
3488 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
3490 I32 pos = mg->mg_len;
3492 sv_pos_b2u(sv, &pos);
3493 mg_flags = 0; /* sv_pos_b2u does get magic */
3497 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
3498 magic_setutf8(sv,mg); /* clear UTF8 cache */
3501 s = (U8 *) SvPV_flags(sv, len, mg_flags);
3503 if (!utf8_to_bytes(s, &len)) {
3508 Perl_croak(aTHX_ "Wide character in %s",
3511 Perl_croak(aTHX_ "Wide character");
3522 =for apidoc sv_utf8_encode
3524 Converts the PV of an SV to UTF-8, but then turns the C<SvUTF8>
3525 flag off so that it looks like octets again.
3531 Perl_sv_utf8_encode(pTHX_ register SV *const sv)
3533 PERL_ARGS_ASSERT_SV_UTF8_ENCODE;
3536 sv_force_normal_flags(sv, 0);
3538 if (SvREADONLY(sv)) {
3539 Perl_croak_no_modify(aTHX);
3541 (void) sv_utf8_upgrade(sv);
3546 =for apidoc sv_utf8_decode
3548 If the PV of the SV is an octet sequence in UTF-8
3549 and contains a multiple-byte character, the C<SvUTF8> flag is turned on
3550 so that it looks like a character. If the PV contains only single-byte
3551 characters, the C<SvUTF8> flag stays off.
3552 Scans PV for validity and returns false if the PV is invalid UTF-8.
3558 Perl_sv_utf8_decode(pTHX_ register SV *const sv)
3560 PERL_ARGS_ASSERT_SV_UTF8_DECODE;
3563 const U8 *start, *c;
3566 /* The octets may have got themselves encoded - get them back as
3569 if (!sv_utf8_downgrade(sv, TRUE))
3572 /* it is actually just a matter of turning the utf8 flag on, but
3573 * we want to make sure everything inside is valid utf8 first.
3575 c = start = (const U8 *) SvPVX_const(sv);
3576 if (!is_utf8_string(c, SvCUR(sv)+1))
3578 e = (const U8 *) SvEND(sv);
3581 if (!UTF8_IS_INVARIANT(ch)) {
3586 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
3587 /* adjust pos to the start of a UTF8 char sequence */
3588 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
3590 I32 pos = mg->mg_len;
3592 for (c = start + pos; c > start; c--) {
3593 if (UTF8_IS_START(*c))
3596 mg->mg_len = c - start;
3599 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
3600 magic_setutf8(sv,mg); /* clear UTF8 cache */
3607 =for apidoc sv_setsv
3609 Copies the contents of the source SV C<ssv> into the destination SV
3610 C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3611 function if the source SV needs to be reused. Does not handle 'set' magic.
3612 Loosely speaking, it performs a copy-by-value, obliterating any previous
3613 content of the destination.
3615 You probably want to use one of the assortment of wrappers, such as
3616 C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3617 C<SvSetMagicSV_nosteal>.
3619 =for apidoc sv_setsv_flags
3621 Copies the contents of the source SV C<ssv> into the destination SV
3622 C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3623 function if the source SV needs to be reused. Does not handle 'set' magic.
3624 Loosely speaking, it performs a copy-by-value, obliterating any previous
3625 content of the destination.
3626 If the C<flags> parameter has the C<SV_GMAGIC> bit set, will C<mg_get> on
3627 C<ssv> if appropriate, else not. If the C<flags> parameter has the
3628 C<NOSTEAL> bit set then the buffers of temps will not be stolen. <sv_setsv>
3629 and C<sv_setsv_nomg> are implemented in terms of this function.
3631 You probably want to use one of the assortment of wrappers, such as
3632 C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3633 C<SvSetMagicSV_nosteal>.
3635 This is the primary function for copying scalars, and most other
3636 copy-ish functions and macros use this underneath.
3642 S_glob_assign_glob(pTHX_ SV *const dstr, SV *const sstr, const int dtype)
3644 I32 mro_changes = 0; /* 1 = method, 2 = isa, 3 = recursive isa */
3645 HV *old_stash = NULL;
3647 PERL_ARGS_ASSERT_GLOB_ASSIGN_GLOB;
3649 if (dtype != SVt_PVGV && !isGV_with_GP(dstr)) {
3650 const char * const name = GvNAME(sstr);
3651 const STRLEN len = GvNAMELEN(sstr);
3653 if (dtype >= SVt_PV) {
3659 SvUPGRADE(dstr, SVt_PVGV);
3660 (void)SvOK_off(dstr);
3661 /* FIXME - why are we doing this, then turning it off and on again
3663 isGV_with_GP_on(dstr);
3665 GvSTASH(dstr) = GvSTASH(sstr);
3667 Perl_sv_add_backref(aTHX_ MUTABLE_SV(GvSTASH(dstr)), dstr);
3668 gv_name_set(MUTABLE_GV(dstr), name, len,
3669 GV_ADD | (GvNAMEUTF8(sstr) ? SVf_UTF8 : 0 ));
3670 SvFAKE_on(dstr); /* can coerce to non-glob */
3673 if(GvGP(MUTABLE_GV(sstr))) {
3674 /* If source has method cache entry, clear it */
3676 SvREFCNT_dec(GvCV(sstr));
3677 GvCV_set(sstr, NULL);
3680 /* If source has a real method, then a method is
3683 GvCV((const GV *)sstr) && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3689 /* If dest already had a real method, that's a change as well */
3691 !mro_changes && GvGP(MUTABLE_GV(dstr)) && GvCVu((const GV *)dstr)
3692 && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3697 /* We don’t need to check the name of the destination if it was not a
3698 glob to begin with. */
3699 if(dtype == SVt_PVGV) {
3700 const char * const name = GvNAME((const GV *)dstr);
3703 /* The stash may have been detached from the symbol table, so
3705 && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3706 && GvAV((const GV *)sstr)
3710 const STRLEN len = GvNAMELEN(dstr);
3711 if ((len > 1 && name[len-2] == ':' && name[len-1] == ':')
3712 || (len == 1 && name[0] == ':')) {
3715 /* Set aside the old stash, so we can reset isa caches on
3717 if((old_stash = GvHV(dstr)))
3718 /* Make sure we do not lose it early. */
3719 SvREFCNT_inc_simple_void_NN(
3720 sv_2mortal((SV *)old_stash)
3726 gp_free(MUTABLE_GV(dstr));
3727 isGV_with_GP_off(dstr);
3728 (void)SvOK_off(dstr);
3729 isGV_with_GP_on(dstr);
3730 GvINTRO_off(dstr); /* one-shot flag */
3731 GvGP_set(dstr, gp_ref(GvGP(sstr)));
3732 if (SvTAINTED(sstr))
3734 if (GvIMPORTED(dstr) != GVf_IMPORTED
3735 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3737 GvIMPORTED_on(dstr);
3740 if(mro_changes == 2) {
3742 SV * const sref = (SV *)GvAV((const GV *)dstr);
3743 if (SvSMAGICAL(sref) && (mg = mg_find(sref, PERL_MAGIC_isa))) {
3744 if (SvTYPE(mg->mg_obj) != SVt_PVAV) {
3745 AV * const ary = newAV();
3746 av_push(ary, mg->mg_obj); /* takes the refcount */
3747 mg->mg_obj = (SV *)ary;
3749 av_push((AV *)mg->mg_obj, SvREFCNT_inc_simple_NN(dstr));
3751 else sv_magic(sref, dstr, PERL_MAGIC_isa, NULL, 0);
3752 mro_isa_changed_in(GvSTASH(dstr));
3754 else if(mro_changes == 3) {
3755 HV * const stash = GvHV(dstr);
3756 if(old_stash ? (HV *)HvENAME_get(old_stash) : stash)
3762 else if(mro_changes) mro_method_changed_in(GvSTASH(dstr));
3767 S_glob_assign_ref(pTHX_ SV *const dstr, SV *const sstr)
3769 SV * const sref = SvREFCNT_inc(SvRV(sstr));
3771 const int intro = GvINTRO(dstr);
3774 const U32 stype = SvTYPE(sref);
3776 PERL_ARGS_ASSERT_GLOB_ASSIGN_REF;
3779 GvINTRO_off(dstr); /* one-shot flag */
3780 GvLINE(dstr) = CopLINE(PL_curcop);
3781 GvEGV(dstr) = MUTABLE_GV(dstr);
3786 location = (SV **) &(GvGP(dstr)->gp_cv); /* XXX bypassing GvCV_set */
3787 import_flag = GVf_IMPORTED_CV;
3790 location = (SV **) &GvHV(dstr);
3791 import_flag = GVf_IMPORTED_HV;
3794 location = (SV **) &GvAV(dstr);
3795 import_flag = GVf_IMPORTED_AV;
3798 location = (SV **) &GvIOp(dstr);
3801 location = (SV **) &GvFORM(dstr);
3804 location = &GvSV(dstr);
3805 import_flag = GVf_IMPORTED_SV;
3808 if (stype == SVt_PVCV) {
3809 /*if (GvCVGEN(dstr) && (GvCV(dstr) != (const CV *)sref || GvCVGEN(dstr))) {*/
3810 if (GvCVGEN(dstr)) {
3811 SvREFCNT_dec(GvCV(dstr));
3812 GvCV_set(dstr, NULL);
3813 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3816 SAVEGENERICSV(*location);
3820 if (stype == SVt_PVCV && (*location != sref || GvCVGEN(dstr))) {
3821 CV* const cv = MUTABLE_CV(*location);
3823 if (!GvCVGEN((const GV *)dstr) &&
3824 (CvROOT(cv) || CvXSUB(cv)))
3826 /* Redefining a sub - warning is mandatory if
3827 it was a const and its value changed. */
3828 if (CvCONST(cv) && CvCONST((const CV *)sref)
3830 == cv_const_sv((const CV *)sref)) {
3832 /* They are 2 constant subroutines generated from
3833 the same constant. This probably means that
3834 they are really the "same" proxy subroutine
3835 instantiated in 2 places. Most likely this is
3836 when a constant is exported twice. Don't warn.
3839 else if (ckWARN(WARN_REDEFINE)
3841 && (!CvCONST((const CV *)sref)
3842 || sv_cmp(cv_const_sv(cv),
3843 cv_const_sv((const CV *)
3845 Perl_warner(aTHX_ packWARN(WARN_REDEFINE),
3848 ? "Constant subroutine %"HEKf
3849 "::%"HEKf" redefined"
3850 : "Subroutine %"HEKf"::%"HEKf
3853 HvNAME_HEK(GvSTASH((const GV *)dstr))
3855 HEKfARG(GvENAME_HEK(MUTABLE_GV(dstr))));
3859 cv_ckproto_len_flags(cv, (const GV *)dstr,
3860 SvPOK(sref) ? CvPROTO(sref) : NULL,
3861 SvPOK(sref) ? CvPROTOLEN(sref) : 0,
3862 SvPOK(sref) ? SvUTF8(sref) : 0);
3864 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3865 GvASSUMECV_on(dstr);
3866 if(GvSTASH(dstr)) mro_method_changed_in(GvSTASH(dstr)); /* sub foo { 1 } sub bar { 2 } *bar = \&foo */
3869 if (import_flag && !(GvFLAGS(dstr) & import_flag)
3870 && CopSTASH_ne(PL_curcop, GvSTASH(dstr))) {
3871 GvFLAGS(dstr) |= import_flag;
3873 if (stype == SVt_PVHV) {
3874 const char * const name = GvNAME((GV*)dstr);
3875 const STRLEN len = GvNAMELEN(dstr);
3878 (len > 1 && name[len-2] == ':' && name[len-1] == ':')
3879 || (len == 1 && name[0] == ':')
3881 && (!dref || HvENAME_get(dref))
3884 (HV *)sref, (HV *)dref,
3890 stype == SVt_PVAV && sref != dref
3891 && strEQ(GvNAME((GV*)dstr), "ISA")
3892 /* The stash may have been detached from the symbol table, so
3893 check its name before doing anything. */
3894 && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3897 MAGIC * const omg = dref && SvSMAGICAL(dref)
3898 ? mg_find(dref, PERL_MAGIC_isa)
3900 if (SvSMAGICAL(sref) && (mg = mg_find(sref, PERL_MAGIC_isa))) {
3901 if (SvTYPE(mg->mg_obj) != SVt_PVAV) {
3902 AV * const ary = newAV();
3903 av_push(ary, mg->mg_obj); /* takes the refcount */
3904 mg->mg_obj = (SV *)ary;
3907 if (SvTYPE(omg->mg_obj) == SVt_PVAV) {
3908 SV **svp = AvARRAY((AV *)omg->mg_obj);
3909 I32 items = AvFILLp((AV *)omg->mg_obj) + 1;
3913 SvREFCNT_inc_simple_NN(*svp++)
3919 SvREFCNT_inc_simple_NN(omg->mg_obj)
3923 av_push((AV *)mg->mg_obj,SvREFCNT_inc_simple_NN(dstr));
3928 sref, omg ? omg->mg_obj : dstr, PERL_MAGIC_isa, NULL, 0
3930 mg = mg_find(sref, PERL_MAGIC_isa);
3932 /* Since the *ISA assignment could have affected more than
3933 one stash, don’t call mro_isa_changed_in directly, but let
3934 magic_clearisa do it for us, as it already has the logic for
3935 dealing with globs vs arrays of globs. */
3937 Perl_magic_clearisa(aTHX_ NULL, mg);
3942 if (SvTAINTED(sstr))
3948 Perl_sv_setsv_flags(pTHX_ SV *dstr, register SV* sstr, const I32 flags)
3951 register U32 sflags;
3953 register svtype stype;
3955 PERL_ARGS_ASSERT_SV_SETSV_FLAGS;
3960 if (SvIS_FREED(dstr)) {
3961 Perl_croak(aTHX_ "panic: attempt to copy value %" SVf
3962 " to a freed scalar %p", SVfARG(sstr), (void *)dstr);
3964 SV_CHECK_THINKFIRST_COW_DROP(dstr);
3966 sstr = &PL_sv_undef;
3967 if (SvIS_FREED(sstr)) {
3968 Perl_croak(aTHX_ "panic: attempt to copy freed scalar %p to %p",
3969 (void*)sstr, (void*)dstr);
3971 stype = SvTYPE(sstr);
3972 dtype = SvTYPE(dstr);
3974 (void)SvAMAGIC_off(dstr);
3977 /* need to nuke the magic */
3981 /* There's a lot of redundancy below but we're going for speed here */
3986 if (dtype != SVt_PVGV && dtype != SVt_PVLV) {
3987 (void)SvOK_off(dstr);
3995 sv_upgrade(dstr, SVt_IV);
3999 sv_upgrade(dstr, SVt_PVIV);
4003 goto end_of_first_switch;
4005 (void)SvIOK_only(dstr);
4006 SvIV_set(dstr, SvIVX(sstr));
4009 /* SvTAINTED can only be true if the SV has taint magic, which in
4010 turn means that the SV type is PVMG (or greater). This is the
4011 case statement for SVt_IV, so this cannot be true (whatever gcov
4013 assert(!SvTAINTED(sstr));
4018 if (dtype < SVt_PV && dtype != SVt_IV)
4019 sv_upgrade(dstr, SVt_IV);
4027 sv_upgrade(dstr, SVt_NV);
4031 sv_upgrade(dstr, SVt_PVNV);
4035 goto end_of_first_switch;
4037 SvNV_set(dstr, SvNVX(sstr));
4038 (void)SvNOK_only(dstr);
4039 /* SvTAINTED can only be true if the SV has taint magic, which in
4040 turn means that the SV type is PVMG (or greater). This is the
4041 case statement for SVt_NV, so this cannot be true (whatever gcov
4043 assert(!SvTAINTED(sstr));
4049 #ifdef PERL_OLD_COPY_ON_WRITE
4050 if ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS) {
4051 if (dtype < SVt_PVIV)
4052 sv_upgrade(dstr, SVt_PVIV);
4059 sv_upgrade(dstr, SVt_PV);
4062 if (dtype < SVt_PVIV)
4063 sv_upgrade(dstr, SVt_PVIV);
4066 if (dtype < SVt_PVNV)
4067 sv_upgrade(dstr, SVt_PVNV);
4071 const char * const type = sv_reftype(sstr,0);
4073 Perl_croak(aTHX_ "Bizarre copy of %s in %s", type, OP_DESC(PL_op));
4075 Perl_croak(aTHX_ "Bizarre copy of %s", type);
4080 if (dtype < SVt_REGEXP)
4081 sv_upgrade(dstr, SVt_REGEXP);
4084 /* case SVt_BIND: */
4088 if (SvGMAGICAL(sstr) && (flags & SV_GMAGIC)) {
4090 if (SvTYPE(sstr) != stype)
4091 stype = SvTYPE(sstr);
4093 if (isGV_with_GP(sstr) && dtype <= SVt_PVLV) {
4094 glob_assign_glob(dstr, sstr, dtype);
4097 if (stype == SVt_PVLV)
4098 SvUPGRADE(dstr, SVt_PVNV);
4100 SvUPGRADE(dstr, (svtype)stype);
4102 end_of_first_switch:
4104 /* dstr may have been upgraded. */
4105 dtype = SvTYPE(dstr);
4106 sflags = SvFLAGS(sstr);
4108 if (dtype == SVt_PVCV || dtype == SVt_PVFM) {
4109 /* Assigning to a subroutine sets the prototype. */
4112 const char *const ptr = SvPV_const(sstr, len);
4114 SvGROW(dstr, len + 1);
4115 Copy(ptr, SvPVX(dstr), len + 1, char);
4116 SvCUR_set(dstr, len);
4118 SvFLAGS(dstr) |= sflags & SVf_UTF8;
4119 CvAUTOLOAD_off(dstr);
4123 } else if (dtype == SVt_PVAV || dtype == SVt_PVHV) {
4124 const char * const type = sv_reftype(dstr,0);
4126 Perl_croak(aTHX_ "Cannot copy to %s in %s", type, OP_DESC(PL_op));
4128 Perl_croak(aTHX_ "Cannot copy to %s", type);
4129 } else if (sflags & SVf_ROK) {
4130 if (isGV_with_GP(dstr)
4131 && SvTYPE(SvRV(sstr)) == SVt_PVGV && isGV_with_GP(SvRV(sstr))) {
4134 if (GvIMPORTED(dstr) != GVf_IMPORTED
4135 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
4137 GvIMPORTED_on(dstr);
4142 glob_assign_glob(dstr, sstr, dtype);
4146 if (dtype >= SVt_PV) {
4147 if (isGV_with_GP(dstr)) {
4148 glob_assign_ref(dstr, sstr);
4151 if (SvPVX_const(dstr)) {
4157 (void)SvOK_off(dstr);
4158 SvRV_set(dstr, SvREFCNT_inc(SvRV(sstr)));
4159 SvFLAGS(dstr) |= sflags & SVf_ROK;
4160 assert(!(sflags & SVp_NOK));
4161 assert(!(sflags & SVp_IOK));
4162 assert(!(sflags & SVf_NOK));
4163 assert(!(sflags & SVf_IOK));
4165 else if (isGV_with_GP(dstr)) {
4166 if (!(sflags & SVf_OK)) {
4167 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
4168 "Undefined value assigned to typeglob");
4171 GV *gv = gv_fetchsv_nomg(sstr, GV_ADD, SVt_PVGV);
4172 if (dstr != (const SV *)gv) {
4173 const char * const name = GvNAME((const GV *)dstr);
4174 const STRLEN len = GvNAMELEN(dstr);
4175 HV *old_stash = NULL;
4176 bool reset_isa = FALSE;
4177 if ((len > 1 && name[len-2] == ':' && name[len-1] == ':')
4178 || (len == 1 && name[0] == ':')) {
4179 /* Set aside the old stash, so we can reset isa caches
4180 on its subclasses. */
4181 if((old_stash = GvHV(dstr))) {
4182 /* Make sure we do not lose it early. */
4183 SvREFCNT_inc_simple_void_NN(
4184 sv_2mortal((SV *)old_stash)
4191 gp_free(MUTABLE_GV(dstr));
4192 GvGP_set(dstr, gp_ref(GvGP(gv)));
4195 HV * const stash = GvHV(dstr);
4197 old_stash ? (HV *)HvENAME_get(old_stash) : stash
4207 else if (dtype == SVt_REGEXP && stype == SVt_REGEXP) {
4208 reg_temp_copy((REGEXP*)dstr, (REGEXP*)sstr);
4210 else if (sflags & SVp_POK) {
4214 * Check to see if we can just swipe the string. If so, it's a
4215 * possible small lose on short strings, but a big win on long ones.
4216 * It might even be a win on short strings if SvPVX_const(dstr)
4217 * has to be allocated and SvPVX_const(sstr) has to be freed.
4218 * Likewise if we can set up COW rather than doing an actual copy, we
4219 * drop to the else clause, as the swipe code and the COW setup code
4220 * have much in common.
4223 /* Whichever path we take through the next code, we want this true,
4224 and doing it now facilitates the COW check. */
4225 (void)SvPOK_only(dstr);
4228 /* If we're already COW then this clause is not true, and if COW
4229 is allowed then we drop down to the else and make dest COW
4230 with us. If caller hasn't said that we're allowed to COW
4231 shared hash keys then we don't do the COW setup, even if the
4232 source scalar is a shared hash key scalar. */
4233 (((flags & SV_COW_SHARED_HASH_KEYS)
4234 ? (sflags & (SVf_FAKE|SVf_READONLY)) != (SVf_FAKE|SVf_READONLY)
4235 : 1 /* If making a COW copy is forbidden then the behaviour we
4236 desire is as if the source SV isn't actually already
4237 COW, even if it is. So we act as if the source flags
4238 are not COW, rather than actually testing them. */
4240 #ifndef PERL_OLD_COPY_ON_WRITE
4241 /* The change that added SV_COW_SHARED_HASH_KEYS makes the logic
4242 when PERL_OLD_COPY_ON_WRITE is defined a little wrong.
4243 Conceptually PERL_OLD_COPY_ON_WRITE being defined should
4244 override SV_COW_SHARED_HASH_KEYS, because it means "always COW"
4245 but in turn, it's somewhat dead code, never expected to go
4246 live, but more kept as a placeholder on how to do it better
4247 in a newer implementation. */
4248 /* If we are COW and dstr is a suitable target then we drop down
4249 into the else and make dest a COW of us. */
4250 || (SvFLAGS(dstr) & CAN_COW_MASK) != CAN_COW_FLAGS
4255 (sflags & SVs_TEMP) && /* slated for free anyway? */
4256 !(sflags & SVf_OOK) && /* and not involved in OOK hack? */
4257 (!(flags & SV_NOSTEAL)) &&
4258 /* and we're allowed to steal temps */
4259 SvREFCNT(sstr) == 1 && /* and no other references to it? */
4260 SvLEN(sstr)) /* and really is a string */
4261 #ifdef PERL_OLD_COPY_ON_WRITE
4262 && ((flags & SV_COW_SHARED_HASH_KEYS)
4263 ? (!((sflags & CAN_COW_MASK) == CAN_COW_FLAGS
4264 && (SvFLAGS(dstr) & CAN_COW_MASK) == CAN_COW_FLAGS
4265 && SvTYPE(sstr) >= SVt_PVIV && SvTYPE(sstr) != SVt_PVFM))
4269 /* Failed the swipe test, and it's not a shared hash key either.
4270 Have to copy the string. */
4271 STRLEN len = SvCUR(sstr);
4272 SvGROW(dstr, len + 1); /* inlined from sv_setpvn */
4273 Move(SvPVX_const(sstr),SvPVX(dstr),len,char);
4274 SvCUR_set(dstr, len);
4275 *SvEND(dstr) = '\0';
4277 /* If PERL_OLD_COPY_ON_WRITE is not defined, then isSwipe will always
4279 /* Either it's a shared hash key, or it's suitable for
4280 copy-on-write or we can swipe the string. */
4282 PerlIO_printf(Perl_debug_log, "Copy on write: sstr --> dstr\n");
4286 #ifdef PERL_OLD_COPY_ON_WRITE
4288 if ((sflags & (SVf_FAKE | SVf_READONLY))
4289 != (SVf_FAKE | SVf_READONLY)) {
4290 SvREADONLY_on(sstr);
4292 /* Make the source SV into a loop of 1.
4293 (about to become 2) */
4294 SV_COW_NEXT_SV_SET(sstr, sstr);
4298 /* Initial code is common. */
4299 if (SvPVX_const(dstr)) { /* we know that dtype >= SVt_PV */
4304 /* making another shared SV. */
4305 STRLEN cur = SvCUR(sstr);
4306 STRLEN len = SvLEN(sstr);
4307 #ifdef PERL_OLD_COPY_ON_WRITE
4309 assert (SvTYPE(dstr) >= SVt_PVIV);
4310 /* SvIsCOW_normal */
4311 /* splice us in between source and next-after-source. */
4312 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4313 SV_COW_NEXT_SV_SET(sstr, dstr);
4314 SvPV_set(dstr, SvPVX_mutable(sstr));
4318 /* SvIsCOW_shared_hash */
4319 DEBUG_C(PerlIO_printf(Perl_debug_log,
4320 "Copy on write: Sharing hash\n"));
4322 assert (SvTYPE(dstr) >= SVt_PV);
4324 HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)))));
4326 SvLEN_set(dstr, len);
4327 SvCUR_set(dstr, cur);
4328 SvREADONLY_on(dstr);
4332 { /* Passes the swipe test. */
4333 SvPV_set(dstr, SvPVX_mutable(sstr));
4334 SvLEN_set(dstr, SvLEN(sstr));
4335 SvCUR_set(dstr, SvCUR(sstr));
4338 (void)SvOK_off(sstr); /* NOTE: nukes most SvFLAGS on sstr */
4339 SvPV_set(sstr, NULL);
4345 if (sflags & SVp_NOK) {
4346 SvNV_set(dstr, SvNVX(sstr));
4348 if (sflags & SVp_IOK) {
4349 SvIV_set(dstr, SvIVX(sstr));
4350 /* Must do this otherwise some other overloaded use of 0x80000000
4351 gets confused. I guess SVpbm_VALID */
4352 if (sflags & SVf_IVisUV)
4355 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_NOK|SVp_NOK|SVf_UTF8);
4357 const MAGIC * const smg = SvVSTRING_mg(sstr);
4359 sv_magic(dstr, NULL, PERL_MAGIC_vstring,
4360 smg->mg_ptr, smg->mg_len);
4361 SvRMAGICAL_on(dstr);
4365 else if (sflags & (SVp_IOK|SVp_NOK)) {
4366 (void)SvOK_off(dstr);
4367 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_IVisUV|SVf_NOK|SVp_NOK);
4368 if (sflags & SVp_IOK) {
4369 /* XXXX Do we want to set IsUV for IV(ROK)? Be extra safe... */
4370 SvIV_set(dstr, SvIVX(sstr));
4372 if (sflags & SVp_NOK) {
4373 SvNV_set(dstr, SvNVX(sstr));
4377 if (isGV_with_GP(sstr)) {
4378 /* This stringification rule for globs is spread in 3 places.
4379 This feels bad. FIXME. */
4380 const U32 wasfake = sflags & SVf_FAKE;
4382 /* FAKE globs can get coerced, so need to turn this off
4383 temporarily if it is on. */
4385 gv_efullname3(dstr, MUTABLE_GV(sstr), "*");
4386 SvFLAGS(sstr) |= wasfake;
4389 (void)SvOK_off(dstr);
4391 if (SvTAINTED(sstr))
4396 =for apidoc sv_setsv_mg
4398 Like C<sv_setsv>, but also handles 'set' magic.
4404 Perl_sv_setsv_mg(pTHX_ SV *const dstr, register SV *const sstr)
4406 PERL_ARGS_ASSERT_SV_SETSV_MG;
4408 sv_setsv(dstr,sstr);
4412 #ifdef PERL_OLD_COPY_ON_WRITE
4414 Perl_sv_setsv_cow(pTHX_ SV *dstr, SV *sstr)
4416 STRLEN cur = SvCUR(sstr);
4417 STRLEN len = SvLEN(sstr);
4418 register char *new_pv;
4420 PERL_ARGS_ASSERT_SV_SETSV_COW;
4423 PerlIO_printf(Perl_debug_log, "Fast copy on write: %p -> %p\n",
4424 (void*)sstr, (void*)dstr);
4431 if (SvTHINKFIRST(dstr))
4432 sv_force_normal_flags(dstr, SV_COW_DROP_PV);
4433 else if (SvPVX_const(dstr))
4434 Safefree(SvPVX_const(dstr));
4438 SvUPGRADE(dstr, SVt_PVIV);
4440 assert (SvPOK(sstr));
4441 assert (SvPOKp(sstr));
4442 assert (!SvIOK(sstr));
4443 assert (!SvIOKp(sstr));
4444 assert (!SvNOK(sstr));
4445 assert (!SvNOKp(sstr));
4447 if (SvIsCOW(sstr)) {
4449 if (SvLEN(sstr) == 0) {
4450 /* source is a COW shared hash key. */
4451 DEBUG_C(PerlIO_printf(Perl_debug_log,
4452 "Fast copy on write: Sharing hash\n"));
4453 new_pv = HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr))));
4456 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4458 assert ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS);
4459 SvUPGRADE(sstr, SVt_PVIV);
4460 SvREADONLY_on(sstr);
4462 DEBUG_C(PerlIO_printf(Perl_debug_log,
4463 "Fast copy on write: Converting sstr to COW\n"));
4464 SV_COW_NEXT_SV_SET(dstr, sstr);
4466 SV_COW_NEXT_SV_SET(sstr, dstr);
4467 new_pv = SvPVX_mutable(sstr);
4470 SvPV_set(dstr, new_pv);
4471 SvFLAGS(dstr) = (SVt_PVIV|SVf_POK|SVp_POK|SVf_FAKE|SVf_READONLY);
4474 SvLEN_set(dstr, len);
4475 SvCUR_set(dstr, cur);
4484 =for apidoc sv_setpvn
4486 Copies a string into an SV. The C<len> parameter indicates the number of
4487 bytes to be copied. If the C<ptr> argument is NULL the SV will become
4488 undefined. Does not handle 'set' magic. See C<sv_setpvn_mg>.
4494 Perl_sv_setpvn(pTHX_ register SV *const sv, register const char *const ptr, register const STRLEN len)
4497 register char *dptr;
4499 PERL_ARGS_ASSERT_SV_SETPVN;
4501 SV_CHECK_THINKFIRST_COW_DROP(sv);
4507 /* len is STRLEN which is unsigned, need to copy to signed */
4510 Perl_croak(aTHX_ "panic: sv_setpvn called with negative strlen");
4512 SvUPGRADE(sv, SVt_PV);
4514 dptr = SvGROW(sv, len + 1);
4515 Move(ptr,dptr,len,char);
4518 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4520 if (SvTYPE(sv) == SVt_PVCV) CvAUTOLOAD_off(sv);
4524 =for apidoc sv_setpvn_mg
4526 Like C<sv_setpvn>, but also handles 'set' magic.
4532 Perl_sv_setpvn_mg(pTHX_ register SV *const sv, register const char *const ptr, register const STRLEN len)
4534 PERL_ARGS_ASSERT_SV_SETPVN_MG;
4536 sv_setpvn(sv,ptr,len);
4541 =for apidoc sv_setpv
4543 Copies a string into an SV. The string must be null-terminated. Does not
4544 handle 'set' magic. See C<sv_setpv_mg>.
4550 Perl_sv_setpv(pTHX_ register SV *const sv, register const char *const ptr)
4553 register STRLEN len;
4555 PERL_ARGS_ASSERT_SV_SETPV;
4557 SV_CHECK_THINKFIRST_COW_DROP(sv);
4563 SvUPGRADE(sv, SVt_PV);
4565 SvGROW(sv, len + 1);
4566 Move(ptr,SvPVX(sv),len+1,char);
4568 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4570 if (SvTYPE(sv) == SVt_PVCV) CvAUTOLOAD_off(sv);
4574 =for apidoc sv_setpv_mg
4576 Like C<sv_setpv>, but also handles 'set' magic.
4582 Perl_sv_setpv_mg(pTHX_ register SV *const sv, register const char *const ptr)
4584 PERL_ARGS_ASSERT_SV_SETPV_MG;
4591 Perl_sv_sethek(pTHX_ register SV *const sv, const HEK *const hek)
4595 PERL_ARGS_ASSERT_SV_SETHEK;
4601 if (HEK_LEN(hek) == HEf_SVKEY) {
4602 sv_setsv(sv, *(SV**)HEK_KEY(hek));
4605 const int flags = HEK_FLAGS(hek);
4606 if (flags & HVhek_WASUTF8) {
4607 STRLEN utf8_len = HEK_LEN(hek);
4608 char *as_utf8 = (char *)bytes_to_utf8((U8*)HEK_KEY(hek), &utf8_len);
4609 sv_usepvn_flags(sv, as_utf8, utf8_len, SV_HAS_TRAILING_NUL);
4612 } else if (flags & (HVhek_REHASH|HVhek_UNSHARED)) {
4613 sv_setpvn(sv, HEK_KEY(hek), HEK_LEN(hek));
4619 sv_upgrade(sv, SVt_PV);
4620 sv_usepvn_flags(sv, (char *)HEK_KEY(share_hek_hek(hek)), HEK_LEN(hek), SV_HAS_TRAILING_NUL);
4634 =for apidoc sv_usepvn_flags
4636 Tells an SV to use C<ptr> to find its string value. Normally the
4637 string is stored inside the SV but sv_usepvn allows the SV to use an
4638 outside string. The C<ptr> should point to memory that was allocated
4639 by C<malloc>. The string length, C<len>, must be supplied. By default
4640 this function will realloc (i.e. move) the memory pointed to by C<ptr>,
4641 so that pointer should not be freed or used by the programmer after
4642 giving it to sv_usepvn, and neither should any pointers from "behind"
4643 that pointer (e.g. ptr + 1) be used.
4645 If C<flags> & SV_SMAGIC is true, will call SvSETMAGIC. If C<flags> &
4646 SV_HAS_TRAILING_NUL is true, then C<ptr[len]> must be NUL, and the realloc
4647 will be skipped. (i.e. the buffer is actually at least 1 byte longer than
4648 C<len>, and already meets the requirements for storing in C<SvPVX>)
4654 Perl_sv_usepvn_flags(pTHX_ SV *const sv, char *ptr, const STRLEN len, const U32 flags)
4659 PERL_ARGS_ASSERT_SV_USEPVN_FLAGS;
4661 SV_CHECK_THINKFIRST_COW_DROP(sv);
4662 SvUPGRADE(sv, SVt_PV);
4665 if (flags & SV_SMAGIC)
4669 if (SvPVX_const(sv))
4673 if (flags & SV_HAS_TRAILING_NUL)
4674 assert(ptr[len] == '\0');
4677 allocate = (flags & SV_HAS_TRAILING_NUL)
4679 #ifdef Perl_safesysmalloc_size
4682 PERL_STRLEN_ROUNDUP(len + 1);
4684 if (flags & SV_HAS_TRAILING_NUL) {
4685 /* It's long enough - do nothing.
4686 Specifically Perl_newCONSTSUB is relying on this. */
4689 /* Force a move to shake out bugs in callers. */
4690 char *new_ptr = (char*)safemalloc(allocate);
4691 Copy(ptr, new_ptr, len, char);
4692 PoisonFree(ptr,len,char);
4696 ptr = (char*) saferealloc (ptr, allocate);
4699 #ifdef Perl_safesysmalloc_size
4700 SvLEN_set(sv, Perl_safesysmalloc_size(ptr));
4702 SvLEN_set(sv, allocate);
4706 if (!(flags & SV_HAS_TRAILING_NUL)) {
4709 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4711 if (flags & SV_SMAGIC)
4715 #ifdef PERL_OLD_COPY_ON_WRITE
4716 /* Need to do this *after* making the SV normal, as we need the buffer
4717 pointer to remain valid until after we've copied it. If we let go too early,
4718 another thread could invalidate it by unsharing last of the same hash key
4719 (which it can do by means other than releasing copy-on-write Svs)
4720 or by changing the other copy-on-write SVs in the loop. */
4722 S_sv_release_COW(pTHX_ register SV *sv, const char *pvx, SV *after)
4724 PERL_ARGS_ASSERT_SV_RELEASE_COW;
4726 { /* this SV was SvIsCOW_normal(sv) */
4727 /* we need to find the SV pointing to us. */
4728 SV *current = SV_COW_NEXT_SV(after);
4730 if (current == sv) {
4731 /* The SV we point to points back to us (there were only two of us
4733 Hence other SV is no longer copy on write either. */
4735 SvREADONLY_off(after);
4737 /* We need to follow the pointers around the loop. */
4739 while ((next = SV_COW_NEXT_SV(current)) != sv) {
4742 /* don't loop forever if the structure is bust, and we have
4743 a pointer into a closed loop. */
4744 assert (current != after);
4745 assert (SvPVX_const(current) == pvx);
4747 /* Make the SV before us point to the SV after us. */
4748 SV_COW_NEXT_SV_SET(current, after);
4754 =for apidoc sv_force_normal_flags
4756 Undo various types of fakery on an SV: if the PV is a shared string, make
4757 a private copy; if we're a ref, stop refing; if we're a glob, downgrade to
4758 an xpvmg; if we're a copy-on-write scalar, this is the on-write time when
4759 we do the copy, and is also used locally. If C<SV_COW_DROP_PV> is set
4760 then a copy-on-write scalar drops its PV buffer (if any) and becomes
4761 SvPOK_off rather than making a copy. (Used where this scalar is about to be
4762 set to some other value.) In addition, the C<flags> parameter gets passed to
4763 C<sv_unref_flags()> when unreffing. C<sv_force_normal> calls this function
4764 with flags set to 0.
4770 Perl_sv_force_normal_flags(pTHX_ register SV *const sv, const U32 flags)
4774 PERL_ARGS_ASSERT_SV_FORCE_NORMAL_FLAGS;
4776 #ifdef PERL_OLD_COPY_ON_WRITE
4777 if (SvREADONLY(sv)) {
4779 const char * const pvx = SvPVX_const(sv);
4780 const STRLEN len = SvLEN(sv);
4781 const STRLEN cur = SvCUR(sv);
4782 /* next COW sv in the loop. If len is 0 then this is a shared-hash
4783 key scalar, so we mustn't attempt to call SV_COW_NEXT_SV(), as
4784 we'll fail an assertion. */
4785 SV * const next = len ? SV_COW_NEXT_SV(sv) : 0;
4788 PerlIO_printf(Perl_debug_log,
4789 "Copy on write: Force normal %ld\n",
4795 /* This SV doesn't own the buffer, so need to Newx() a new one: */
4798 if (flags & SV_COW_DROP_PV) {
4799 /* OK, so we don't need to copy our buffer. */
4802 SvGROW(sv, cur + 1);
4803 Move(pvx,SvPVX(sv),cur,char);
4808 sv_release_COW(sv, pvx, next);
4810 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4816 else if (IN_PERL_RUNTIME)
4817 Perl_croak_no_modify(aTHX);
4820 if (SvREADONLY(sv)) {
4821 if (SvFAKE(sv) && !isGV_with_GP(sv)) {
4822 const char * const pvx = SvPVX_const(sv);
4823 const STRLEN len = SvCUR(sv);
4828 SvGROW(sv, len + 1);
4829 Move(pvx,SvPVX(sv),len,char);
4831 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4833 else if (IN_PERL_RUNTIME)
4834 Perl_croak_no_modify(aTHX);
4838 sv_unref_flags(sv, flags);
4839 else if (SvFAKE(sv) && isGV_with_GP(sv))
4841 else if (SvFAKE(sv) && SvTYPE(sv) == SVt_REGEXP) {
4842 /* Need to downgrade the REGEXP to a simple(r) scalar. This is analogous
4843 to sv_unglob. We only need it here, so inline it. */
4844 const svtype new_type = SvMAGIC(sv) || SvSTASH(sv) ? SVt_PVMG : SVt_PV;
4845 SV *const temp = newSV_type(new_type);
4846 void *const temp_p = SvANY(sv);
4848 if (new_type == SVt_PVMG) {
4849 SvMAGIC_set(temp, SvMAGIC(sv));
4850 SvMAGIC_set(sv, NULL);
4851 SvSTASH_set(temp, SvSTASH(sv));
4852 SvSTASH_set(sv, NULL);
4854 SvCUR_set(temp, SvCUR(sv));
4855 /* Remember that SvPVX is in the head, not the body. */
4857 SvLEN_set(temp, SvLEN(sv));
4858 /* This signals "buffer is owned by someone else" in sv_clear,
4859 which is the least effort way to stop it freeing the buffer.
4861 SvLEN_set(sv, SvLEN(sv)+1);
4863 /* Their buffer is already owned by someone else. */
4864 SvPVX(sv) = savepvn(SvPVX(sv), SvCUR(sv));
4865 SvLEN_set(temp, SvCUR(sv)+1);
4868 /* Now swap the rest of the bodies. */
4870 SvFLAGS(sv) &= ~(SVf_FAKE|SVTYPEMASK);
4871 SvFLAGS(sv) |= new_type;
4872 SvANY(sv) = SvANY(temp);
4874 SvFLAGS(temp) &= ~(SVTYPEMASK);
4875 SvFLAGS(temp) |= SVt_REGEXP|SVf_FAKE;
4876 SvANY(temp) = temp_p;
4885 Efficient removal of characters from the beginning of the string buffer.
4886 SvPOK(sv) must be true and the C<ptr> must be a pointer to somewhere inside
4887 the string buffer. The C<ptr> becomes the first character of the adjusted
4888 string. Uses the "OOK hack".
4890 Beware: after this function returns, C<ptr> and SvPVX_const(sv) may no longer
4891 refer to the same chunk of data.
4893 The unfortunate similarity of this function's name to that of Perl's C<chop>
4894 operator is strictly coincidental. This function works from the left;
4895 C<chop> works from the right.
4901 Perl_sv_chop(pTHX_ register SV *const sv, register const char *const ptr)
4912 PERL_ARGS_ASSERT_SV_CHOP;
4914 if (!ptr || !SvPOKp(sv))
4916 delta = ptr - SvPVX_const(sv);
4918 /* Nothing to do. */
4921 max_delta = SvLEN(sv) ? SvLEN(sv) : SvCUR(sv);
4922 if (delta > max_delta)
4923 Perl_croak(aTHX_ "panic: sv_chop ptr=%p, start=%p, end=%p",
4924 ptr, SvPVX_const(sv), SvPVX_const(sv) + max_delta);
4925 /* SvPVX(sv) may move in SV_CHECK_THINKFIRST(sv), so don't use ptr any more */
4926 SV_CHECK_THINKFIRST(sv);
4929 if (!SvLEN(sv)) { /* make copy of shared string */
4930 const char *pvx = SvPVX_const(sv);
4931 const STRLEN len = SvCUR(sv);
4932 SvGROW(sv, len + 1);
4933 Move(pvx,SvPVX(sv),len,char);
4936 SvFLAGS(sv) |= SVf_OOK;
4939 SvOOK_offset(sv, old_delta);
4941 SvLEN_set(sv, SvLEN(sv) - delta);
4942 SvCUR_set(sv, SvCUR(sv) - delta);
4943 SvPV_set(sv, SvPVX(sv) + delta);
4945 p = (U8 *)SvPVX_const(sv);
4948 /* how many bytes were evacuated? we will fill them with sentinel
4949 bytes, except for the part holding the new offset of course. */
4952 evacn += (old_delta < 0x100 ? 1 : 1 + sizeof(STRLEN));
4954 assert(evacn <= delta + old_delta);
4960 if (delta < 0x100) {
4964 p -= sizeof(STRLEN);
4965 Copy((U8*)&delta, p, sizeof(STRLEN), U8);
4969 /* Fill the preceding buffer with sentinals to verify that no-one is
4979 =for apidoc sv_catpvn
4981 Concatenates the string onto the end of the string which is in the SV. The
4982 C<len> indicates number of bytes to copy. If the SV has the UTF-8
4983 status set, then the bytes appended should be valid UTF-8.
4984 Handles 'get' magic, but not 'set' magic. See C<sv_catpvn_mg>.
4986 =for apidoc sv_catpvn_flags
4988 Concatenates the string onto the end of the string which is in the SV. The
4989 C<len> indicates number of bytes to copy. If the SV has the UTF-8
4990 status set, then the bytes appended should be valid UTF-8.
4991 If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<dsv> if
4992 appropriate, else not. C<sv_catpvn> and C<sv_catpvn_nomg> are implemented
4993 in terms of this function.
4999 Perl_sv_catpvn_flags(pTHX_ register SV *const dsv, register const char *sstr, register const STRLEN slen, const I32 flags)
5003 const char * const dstr = SvPV_force_flags(dsv, dlen, flags);
5005 PERL_ARGS_ASSERT_SV_CATPVN_FLAGS;
5006 assert((flags & (SV_CATBYTES|SV_CATUTF8)) != (SV_CATBYTES|SV_CATUTF8));
5008 if (!(flags & SV_CATBYTES) || !SvUTF8(dsv)) {
5009 if (flags & SV_CATUTF8 && !SvUTF8(dsv)) {
5010 sv_utf8_upgrade_flags_grow(dsv, 0, slen + 1);
5013 else SvGROW(dsv, dlen + slen + 1);
5015 sstr = SvPVX_const(dsv);
5016 Move(sstr, SvPVX(dsv) + dlen, slen, char);
5017 SvCUR_set(dsv, SvCUR(dsv) + slen);
5020 /* We inline bytes_to_utf8, to avoid an extra malloc. */
5021 const char * const send = sstr + slen;
5024 /* Something this code does not account for, which I think is
5025 impossible; it would require the same pv to be treated as
5026 bytes *and* utf8, which would indicate a bug elsewhere. */
5027 assert(sstr != dstr);
5029 SvGROW(dsv, dlen + slen * 2 + 1);
5030 d = (U8 *)SvPVX(dsv) + dlen;
5032 while (sstr < send) {
5033 const UV uv = NATIVE_TO_ASCII((U8)*sstr++);
5034 if (UNI_IS_INVARIANT(uv))
5035 *d++ = (U8)UTF_TO_NATIVE(uv);
5037 *d++ = (U8)UTF8_EIGHT_BIT_HI(uv);
5038 *d++ = (U8)UTF8_EIGHT_BIT_LO(uv);
5041 SvCUR_set(dsv, d-(const U8 *)SvPVX(dsv));
5044 (void)SvPOK_only_UTF8(dsv); /* validate pointer */
5046 if (flags & SV_SMAGIC)
5051 =for apidoc sv_catsv
5053 Concatenates the string from SV C<ssv> onto the end of the string in
5054 SV C<dsv>. Modifies C<dsv> but not C<ssv>. Handles 'get' magic, but
5055 not 'set' magic. See C<sv_catsv_mg>.
5057 =for apidoc sv_catsv_flags
5059 Concatenates the string from SV C<ssv> onto the end of the string in
5060 SV C<dsv>. Modifies C<dsv> but not C<ssv>. If C<flags> has C<SV_GMAGIC>
5061 bit set, will C<mg_get> on the SVs if appropriate, else not. C<sv_catsv>
5062 and C<sv_catsv_nomg> are implemented in terms of this function.
5067 Perl_sv_catsv_flags(pTHX_ SV *const dsv, register SV *const ssv, const I32 flags)
5071 PERL_ARGS_ASSERT_SV_CATSV_FLAGS;
5075 const char *spv = SvPV_flags_const(ssv, slen, flags);
5077 if (SvGMAGICAL(dsv) && (flags & SV_GMAGIC))
5079 sv_catpvn_flags(dsv, spv, slen,
5080 DO_UTF8(ssv) ? SV_CATUTF8 : SV_CATBYTES);
5083 if (flags & SV_SMAGIC)
5088 =for apidoc sv_catpv
5090 Concatenates the string onto the end of the string which is in the SV.
5091 If the SV has the UTF-8 status set, then the bytes appended should be
5092 valid UTF-8. Handles 'get' magic, but not 'set' magic. See C<sv_catpv_mg>.
5097 Perl_sv_catpv(pTHX_ register SV *const sv, register const char *ptr)
5100 register STRLEN len;
5104 PERL_ARGS_ASSERT_SV_CATPV;
5108 junk = SvPV_force(sv, tlen);
5110 SvGROW(sv, tlen + len + 1);
5112 ptr = SvPVX_const(sv);
5113 Move(ptr,SvPVX(sv)+tlen,len+1,char);
5114 SvCUR_set(sv, SvCUR(sv) + len);
5115 (void)SvPOK_only_UTF8(sv); /* validate pointer */
5120 =for apidoc sv_catpv_flags
5122 Concatenates the string onto the end of the string which is in the SV.
5123 If the SV has the UTF-8 status set, then the bytes appended should
5124 be valid UTF-8. If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get>
5125 on the SVs if appropriate, else not.
5131 Perl_sv_catpv_flags(pTHX_ SV *dstr, const char *sstr, const I32 flags)
5133 PERL_ARGS_ASSERT_SV_CATPV_FLAGS;
5134 sv_catpvn_flags(dstr, sstr, strlen(sstr), flags);
5138 =for apidoc sv_catpv_mg
5140 Like C<sv_catpv>, but also handles 'set' magic.
5146 Perl_sv_catpv_mg(pTHX_ register SV *const sv, register const char *const ptr)
5148 PERL_ARGS_ASSERT_SV_CATPV_MG;
5157 Creates a new SV. A non-zero C<len> parameter indicates the number of
5158 bytes of preallocated string space the SV should have. An extra byte for a
5159 trailing NUL is also reserved. (SvPOK is not set for the SV even if string
5160 space is allocated.) The reference count for the new SV is set to 1.
5162 In 5.9.3, newSV() replaces the older NEWSV() API, and drops the first
5163 parameter, I<x>, a debug aid which allowed callers to identify themselves.
5164 This aid has been superseded by a new build option, PERL_MEM_LOG (see
5165 L<perlhacktips/PERL_MEM_LOG>). The older API is still there for use in XS
5166 modules supporting older perls.
5172 Perl_newSV(pTHX_ const STRLEN len)
5179 sv_upgrade(sv, SVt_PV);
5180 SvGROW(sv, len + 1);
5185 =for apidoc sv_magicext
5187 Adds magic to an SV, upgrading it if necessary. Applies the
5188 supplied vtable and returns a pointer to the magic added.
5190 Note that C<sv_magicext> will allow things that C<sv_magic> will not.
5191 In particular, you can add magic to SvREADONLY SVs, and add more than
5192 one instance of the same 'how'.
5194 If C<namlen> is greater than zero then a C<savepvn> I<copy> of C<name> is
5195 stored, if C<namlen> is zero then C<name> is stored as-is and - as another
5196 special case - if C<(name && namlen == HEf_SVKEY)> then C<name> is assumed
5197 to contain an C<SV*> and is stored as-is with its REFCNT incremented.
5199 (This is now used as a subroutine by C<sv_magic>.)
5204 Perl_sv_magicext(pTHX_ SV *const sv, SV *const obj, const int how,
5205 const MGVTBL *const vtable, const char *const name, const I32 namlen)
5210 PERL_ARGS_ASSERT_SV_MAGICEXT;
5212 SvUPGRADE(sv, SVt_PVMG);
5213 Newxz(mg, 1, MAGIC);
5214 mg->mg_moremagic = SvMAGIC(sv);
5215 SvMAGIC_set(sv, mg);
5217 /* Sometimes a magic contains a reference loop, where the sv and
5218 object refer to each other. To prevent a reference loop that
5219 would prevent such objects being freed, we look for such loops
5220 and if we find one we avoid incrementing the object refcount.
5222 Note we cannot do this to avoid self-tie loops as intervening RV must
5223 have its REFCNT incremented to keep it in existence.
5226 if (!obj || obj == sv ||
5227 how == PERL_MAGIC_arylen ||
5228 how == PERL_MAGIC_symtab ||
5229 (SvTYPE(obj) == SVt_PVGV &&
5230 (GvSV(obj) == sv || GvHV(obj) == (const HV *)sv
5231 || GvAV(obj) == (const AV *)sv || GvCV(obj) == (const CV *)sv
5232 || GvIOp(obj) == (const IO *)sv || GvFORM(obj) == (const CV *)sv)))
5237 mg->mg_obj = SvREFCNT_inc_simple(obj);
5238 mg->mg_flags |= MGf_REFCOUNTED;
5241 /* Normal self-ties simply pass a null object, and instead of
5242 using mg_obj directly, use the SvTIED_obj macro to produce a
5243 new RV as needed. For glob "self-ties", we are tieing the PVIO
5244 with an RV obj pointing to the glob containing the PVIO. In
5245 this case, to avoid a reference loop, we need to weaken the
5249 if (how == PERL_MAGIC_tiedscalar && SvTYPE(sv) == SVt_PVIO &&
5250 obj && SvROK(obj) && GvIO(SvRV(obj)) == (const IO *)sv)
5256 mg->mg_len = namlen;
5259 mg->mg_ptr = savepvn(name, namlen);
5260 else if (namlen == HEf_SVKEY) {
5261 /* Yes, this is casting away const. This is only for the case of
5262 HEf_SVKEY. I think we need to document this aberation of the
5263 constness of the API, rather than making name non-const, as
5264 that change propagating outwards a long way. */
5265 mg->mg_ptr = (char*)SvREFCNT_inc_simple_NN((SV *)name);
5267 mg->mg_ptr = (char *) name;
5269 mg->mg_virtual = (MGVTBL *) vtable;
5273 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
5278 =for apidoc sv_magic
5280 Adds magic to an SV. First upgrades C<sv> to type C<SVt_PVMG> if necessary,
5281 then adds a new magic item of type C<how> to the head of the magic list.
5283 See C<sv_magicext> (which C<sv_magic> now calls) for a description of the
5284 handling of the C<name> and C<namlen> arguments.
5286 You need to use C<sv_magicext> to add magic to SvREADONLY SVs and also
5287 to add more than one instance of the same 'how'.
5293 Perl_sv_magic(pTHX_ register SV *const sv, SV *const obj, const int how,
5294 const char *const name, const I32 namlen)
5297 const MGVTBL *vtable;
5300 unsigned int vtable_index;
5302 PERL_ARGS_ASSERT_SV_MAGIC;
5304 if (how < 0 || (unsigned)how > C_ARRAY_LENGTH(PL_magic_data)
5305 || ((flags = PL_magic_data[how]),
5306 (vtable_index = flags & PERL_MAGIC_VTABLE_MASK)
5307 > magic_vtable_max))
5308 Perl_croak(aTHX_ "Don't know how to handle magic of type \\%o", how);
5310 /* PERL_MAGIC_ext is reserved for use by extensions not perl internals.
5311 Useful for attaching extension internal data to perl vars.
5312 Note that multiple extensions may clash if magical scalars
5313 etc holding private data from one are passed to another. */
5315 vtable = (vtable_index == magic_vtable_max)
5316 ? NULL : PL_magic_vtables + vtable_index;
5318 #ifdef PERL_OLD_COPY_ON_WRITE
5320 sv_force_normal_flags(sv, 0);
5322 if (SvREADONLY(sv)) {
5324 /* its okay to attach magic to shared strings; the subsequent
5325 * upgrade to PVMG will unshare the string */
5326 !(SvFAKE(sv) && SvTYPE(sv) < SVt_PVMG)
5329 && !PERL_MAGIC_TYPE_READONLY_ACCEPTABLE(how)
5332 Perl_croak_no_modify(aTHX);
5335 if (SvMAGICAL(sv) || (how == PERL_MAGIC_taint && SvTYPE(sv) >= SVt_PVMG)) {
5336 if (SvMAGIC(sv) && (mg = mg_find(sv, how))) {
5337 /* sv_magic() refuses to add a magic of the same 'how' as an
5340 if (how == PERL_MAGIC_taint) {
5342 /* Any scalar which already had taint magic on which someone
5343 (erroneously?) did SvIOK_on() or similar will now be
5344 incorrectly sporting public "OK" flags. */
5345 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
5351 /* Rest of work is done else where */
5352 mg = sv_magicext(sv,obj,how,vtable,name,namlen);
5355 case PERL_MAGIC_taint:
5358 case PERL_MAGIC_ext:
5359 case PERL_MAGIC_dbfile:
5366 S_sv_unmagicext_flags(pTHX_ SV *const sv, const int type, MGVTBL *vtbl, const U32 flags)
5373 if (SvTYPE(sv) < SVt_PVMG || !SvMAGIC(sv))
5375 mgp = &(((XPVMG*) SvANY(sv))->xmg_u.xmg_magic);
5376 for (mg = *mgp; mg; mg = *mgp) {
5377 const MGVTBL* const virt = mg->mg_virtual;
5378 if (mg->mg_type == type && (!flags || virt == vtbl)) {
5379 *mgp = mg->mg_moremagic;
5380 if (virt && virt->svt_free)
5381 virt->svt_free(aTHX_ sv, mg);
5382 if (mg->mg_ptr && mg->mg_type != PERL_MAGIC_regex_global) {
5384 Safefree(mg->mg_ptr);
5385 else if (mg->mg_len == HEf_SVKEY)
5386 SvREFCNT_dec(MUTABLE_SV(mg->mg_ptr));
5387 else if (mg->mg_type == PERL_MAGIC_utf8)
5388 Safefree(mg->mg_ptr);
5390 if (mg->mg_flags & MGf_REFCOUNTED)
5391 SvREFCNT_dec(mg->mg_obj);
5395 mgp = &mg->mg_moremagic;
5398 if (SvMAGICAL(sv)) /* if we're under save_magic, wait for restore_magic; */
5399 mg_magical(sv); /* else fix the flags now */
5403 SvFLAGS(sv) |= (SvFLAGS(sv) & (SVp_IOK|SVp_NOK|SVp_POK)) >> PRIVSHIFT;
5409 =for apidoc sv_unmagic
5411 Removes all magic of type C<type> from an SV.
5417 Perl_sv_unmagic(pTHX_ SV *const sv, const int type)
5419 PERL_ARGS_ASSERT_SV_UNMAGIC;
5420 return S_sv_unmagicext_flags(aTHX_ sv, type, NULL, 0);
5424 =for apidoc sv_unmagicext
5426 Removes all magic of type C<type> with the specified C<vtbl> from an SV.
5432 Perl_sv_unmagicext(pTHX_ SV *const sv, const int type, MGVTBL *vtbl)
5434 PERL_ARGS_ASSERT_SV_UNMAGICEXT;
5435 return S_sv_unmagicext_flags(aTHX_ sv, type, vtbl, 1);
5439 =for apidoc sv_rvweaken
5441 Weaken a reference: set the C<SvWEAKREF> flag on this RV; give the
5442 referred-to SV C<PERL_MAGIC_backref> magic if it hasn't already; and
5443 push a back-reference to this RV onto the array of backreferences
5444 associated with that magic. If the RV is magical, set magic will be
5445 called after the RV is cleared.
5451 Perl_sv_rvweaken(pTHX_ SV *const sv)
5455 PERL_ARGS_ASSERT_SV_RVWEAKEN;
5457 if (!SvOK(sv)) /* let undefs pass */
5460 Perl_croak(aTHX_ "Can't weaken a nonreference");
5461 else if (SvWEAKREF(sv)) {
5462 Perl_ck_warner(aTHX_ packWARN(WARN_MISC), "Reference is already weak");
5465 else if (SvREADONLY(sv)) croak_no_modify();
5467 Perl_sv_add_backref(aTHX_ tsv, sv);
5473 /* Give tsv backref magic if it hasn't already got it, then push a
5474 * back-reference to sv onto the array associated with the backref magic.
5476 * As an optimisation, if there's only one backref and it's not an AV,
5477 * store it directly in the HvAUX or mg_obj slot, avoiding the need to
5478 * allocate an AV. (Whether the slot holds an AV tells us whether this is
5482 /* A discussion about the backreferences array and its refcount:
5484 * The AV holding the backreferences is pointed to either as the mg_obj of
5485 * PERL_MAGIC_backref, or in the specific case of a HV, from the
5486 * xhv_backreferences field. The array is created with a refcount
5487 * of 2. This means that if during global destruction the array gets
5488 * picked on before its parent to have its refcount decremented by the
5489 * random zapper, it won't actually be freed, meaning it's still there for
5490 * when its parent gets freed.
5492 * When the parent SV is freed, the extra ref is killed by
5493 * Perl_sv_kill_backrefs. The other ref is killed, in the case of magic,
5494 * by mg_free() / MGf_REFCOUNTED, or for a hash, by Perl_hv_kill_backrefs.
5496 * When a single backref SV is stored directly, it is not reference
5501 Perl_sv_add_backref(pTHX_ SV *const tsv, SV *const sv)
5508 PERL_ARGS_ASSERT_SV_ADD_BACKREF;
5510 /* find slot to store array or singleton backref */
5512 if (SvTYPE(tsv) == SVt_PVHV) {
5513 svp = (SV**)Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5516 (SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL))))
5518 sv_magic(tsv, NULL, PERL_MAGIC_backref, NULL, 0);
5519 mg = mg_find(tsv, PERL_MAGIC_backref);
5521 svp = &(mg->mg_obj);
5524 /* create or retrieve the array */
5526 if ( (!*svp && SvTYPE(sv) == SVt_PVAV)
5527 || (*svp && SvTYPE(*svp) != SVt_PVAV)
5532 SvREFCNT_inc_simple_void(av);
5533 /* av now has a refcnt of 2; see discussion above */
5535 /* move single existing backref to the array */
5537 AvARRAY(av)[++AvFILLp(av)] = *svp; /* av_push() */
5541 mg->mg_flags |= MGf_REFCOUNTED;
5544 av = MUTABLE_AV(*svp);
5547 /* optimisation: store single backref directly in HvAUX or mg_obj */
5551 /* push new backref */
5552 assert(SvTYPE(av) == SVt_PVAV);
5553 if (AvFILLp(av) >= AvMAX(av)) {
5554 av_extend(av, AvFILLp(av)+1);
5556 AvARRAY(av)[++AvFILLp(av)] = sv; /* av_push() */
5559 /* delete a back-reference to ourselves from the backref magic associated
5560 * with the SV we point to.
5564 Perl_sv_del_backref(pTHX_ SV *const tsv, SV *const sv)
5569 PERL_ARGS_ASSERT_SV_DEL_BACKREF;
5571 if (SvTYPE(tsv) == SVt_PVHV) {
5573 svp = (SV**)Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5577 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
5578 svp = mg ? &(mg->mg_obj) : NULL;
5582 Perl_croak(aTHX_ "panic: del_backref");
5584 if (SvTYPE(*svp) == SVt_PVAV) {
5588 AV * const av = (AV*)*svp;
5590 assert(!SvIS_FREED(av));
5594 /* for an SV with N weak references to it, if all those
5595 * weak refs are deleted, then sv_del_backref will be called
5596 * N times and O(N^2) compares will be done within the backref
5597 * array. To ameliorate this potential slowness, we:
5598 * 1) make sure this code is as tight as possible;
5599 * 2) when looking for SV, look for it at both the head and tail of the
5600 * array first before searching the rest, since some create/destroy
5601 * patterns will cause the backrefs to be freed in order.
5608 SV **p = &svp[fill];
5609 SV *const topsv = *p;
5616 /* We weren't the last entry.
5617 An unordered list has this property that you
5618 can take the last element off the end to fill
5619 the hole, and it's still an unordered list :-)
5625 break; /* should only be one */
5632 AvFILLp(av) = fill-1;
5635 /* optimisation: only a single backref, stored directly */
5637 Perl_croak(aTHX_ "panic: del_backref");
5644 Perl_sv_kill_backrefs(pTHX_ SV *const sv, AV *const av)
5650 PERL_ARGS_ASSERT_SV_KILL_BACKREFS;
5655 /* after multiple passes through Perl_sv_clean_all() for a thinngy
5656 * that has badly leaked, the backref array may have gotten freed,
5657 * since we only protect it against 1 round of cleanup */
5658 if (SvIS_FREED(av)) {
5659 if (PL_in_clean_all) /* All is fair */
5662 "panic: magic_killbackrefs (freed backref AV/SV)");
5666 is_array = (SvTYPE(av) == SVt_PVAV);
5668 assert(!SvIS_FREED(av));
5671 last = svp + AvFILLp(av);
5674 /* optimisation: only a single backref, stored directly */
5680 while (svp <= last) {
5682 SV *const referrer = *svp;
5683 if (SvWEAKREF(referrer)) {
5684 /* XXX Should we check that it hasn't changed? */
5685 assert(SvROK(referrer));
5686 SvRV_set(referrer, 0);
5688 SvWEAKREF_off(referrer);
5689 SvSETMAGIC(referrer);
5690 } else if (SvTYPE(referrer) == SVt_PVGV ||
5691 SvTYPE(referrer) == SVt_PVLV) {
5692 assert(SvTYPE(sv) == SVt_PVHV); /* stash backref */
5693 /* You lookin' at me? */
5694 assert(GvSTASH(referrer));
5695 assert(GvSTASH(referrer) == (const HV *)sv);
5696 GvSTASH(referrer) = 0;
5697 } else if (SvTYPE(referrer) == SVt_PVCV ||
5698 SvTYPE(referrer) == SVt_PVFM) {
5699 if (SvTYPE(sv) == SVt_PVHV) { /* stash backref */
5700 /* You lookin' at me? */
5701 assert(CvSTASH(referrer));
5702 assert(CvSTASH(referrer) == (const HV *)sv);
5703 SvANY(MUTABLE_CV(referrer))->xcv_stash = 0;
5706 assert(SvTYPE(sv) == SVt_PVGV);
5707 /* You lookin' at me? */
5708 assert(CvGV(referrer));
5709 assert(CvGV(referrer) == (const GV *)sv);
5710 anonymise_cv_maybe(MUTABLE_GV(sv),
5711 MUTABLE_CV(referrer));
5716 "panic: magic_killbackrefs (flags=%"UVxf")",
5717 (UV)SvFLAGS(referrer));
5728 SvREFCNT_dec(av); /* remove extra count added by sv_add_backref() */
5734 =for apidoc sv_insert
5736 Inserts a string at the specified offset/length within the SV. Similar to
5737 the Perl substr() function. Handles get magic.
5739 =for apidoc sv_insert_flags
5741 Same as C<sv_insert>, but the extra C<flags> are passed the C<SvPV_force_flags> that applies to C<bigstr>.
5747 Perl_sv_insert_flags(pTHX_ SV *const bigstr, const STRLEN offset, const STRLEN len, const char *const little, const STRLEN littlelen, const U32 flags)
5752 register char *midend;
5753 register char *bigend;
5754 register SSize_t i; /* better be sizeof(STRLEN) or bad things happen */
5757 PERL_ARGS_ASSERT_SV_INSERT_FLAGS;
5760 Perl_croak(aTHX_ "Can't modify non-existent substring");
5761 SvPV_force_flags(bigstr, curlen, flags);
5762 (void)SvPOK_only_UTF8(bigstr);
5763 if (offset + len > curlen) {
5764 SvGROW(bigstr, offset+len+1);
5765 Zero(SvPVX(bigstr)+curlen, offset+len-curlen, char);
5766 SvCUR_set(bigstr, offset+len);
5770 i = littlelen - len;
5771 if (i > 0) { /* string might grow */
5772 big = SvGROW(bigstr, SvCUR(bigstr) + i + 1);
5773 mid = big + offset + len;
5774 midend = bigend = big + SvCUR(bigstr);
5777 while (midend > mid) /* shove everything down */
5778 *--bigend = *--midend;
5779 Move(little,big+offset,littlelen,char);
5780 SvCUR_set(bigstr, SvCUR(bigstr) + i);
5785 Move(little,SvPVX(bigstr)+offset,len,char);
5790 big = SvPVX(bigstr);
5793 bigend = big + SvCUR(bigstr);
5795 if (midend > bigend)
5796 Perl_croak(aTHX_ "panic: sv_insert");
5798 if (mid - big > bigend - midend) { /* faster to shorten from end */
5800 Move(little, mid, littlelen,char);
5803 i = bigend - midend;
5805 Move(midend, mid, i,char);
5809 SvCUR_set(bigstr, mid - big);
5811 else if ((i = mid - big)) { /* faster from front */
5812 midend -= littlelen;
5814 Move(big, midend - i, i, char);
5815 sv_chop(bigstr,midend-i);
5817 Move(little, mid, littlelen,char);
5819 else if (littlelen) {
5820 midend -= littlelen;
5821 sv_chop(bigstr,midend);
5822 Move(little,midend,littlelen,char);
5825 sv_chop(bigstr,midend);
5831 =for apidoc sv_replace
5833 Make the first argument a copy of the second, then delete the original.
5834 The target SV physically takes over ownership of the body of the source SV
5835 and inherits its flags; however, the target keeps any magic it owns,
5836 and any magic in the source is discarded.
5837 Note that this is a rather specialist SV copying operation; most of the
5838 time you'll want to use C<sv_setsv> or one of its many macro front-ends.
5844 Perl_sv_replace(pTHX_ register SV *const sv, register SV *const nsv)
5847 const U32 refcnt = SvREFCNT(sv);
5849 PERL_ARGS_ASSERT_SV_REPLACE;
5851 SV_CHECK_THINKFIRST_COW_DROP(sv);
5852 if (SvREFCNT(nsv) != 1) {
5853 Perl_croak(aTHX_ "panic: reference miscount on nsv in sv_replace()"
5854 " (%" UVuf " != 1)", (UV) SvREFCNT(nsv));
5856 if (SvMAGICAL(sv)) {
5860 sv_upgrade(nsv, SVt_PVMG);
5861 SvMAGIC_set(nsv, SvMAGIC(sv));
5862 SvFLAGS(nsv) |= SvMAGICAL(sv);
5864 SvMAGIC_set(sv, NULL);
5868 assert(!SvREFCNT(sv));
5869 #ifdef DEBUG_LEAKING_SCALARS
5870 sv->sv_flags = nsv->sv_flags;
5871 sv->sv_any = nsv->sv_any;
5872 sv->sv_refcnt = nsv->sv_refcnt;
5873 sv->sv_u = nsv->sv_u;
5875 StructCopy(nsv,sv,SV);
5877 if(SvTYPE(sv) == SVt_IV) {
5879 = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
5883 #ifdef PERL_OLD_COPY_ON_WRITE
5884 if (SvIsCOW_normal(nsv)) {
5885 /* We need to follow the pointers around the loop to make the
5886 previous SV point to sv, rather than nsv. */
5889 while ((next = SV_COW_NEXT_SV(current)) != nsv) {
5892 assert(SvPVX_const(current) == SvPVX_const(nsv));
5894 /* Make the SV before us point to the SV after us. */
5896 PerlIO_printf(Perl_debug_log, "previous is\n");
5898 PerlIO_printf(Perl_debug_log,
5899 "move it from 0x%"UVxf" to 0x%"UVxf"\n",
5900 (UV) SV_COW_NEXT_SV(current), (UV) sv);
5902 SV_COW_NEXT_SV_SET(current, sv);
5905 SvREFCNT(sv) = refcnt;
5906 SvFLAGS(nsv) |= SVTYPEMASK; /* Mark as freed */
5911 /* We're about to free a GV which has a CV that refers back to us.
5912 * If that CV will outlive us, make it anonymous (i.e. fix up its CvGV
5916 S_anonymise_cv_maybe(pTHX_ GV *gv, CV* cv)
5921 PERL_ARGS_ASSERT_ANONYMISE_CV_MAYBE;
5924 assert(SvREFCNT(gv) == 0);
5925 assert(isGV(gv) && isGV_with_GP(gv));
5927 assert(!CvANON(cv));
5928 assert(CvGV(cv) == gv);
5930 /* will the CV shortly be freed by gp_free() ? */
5931 if (GvCV(gv) == cv && GvGP(gv)->gp_refcnt < 2 && SvREFCNT(cv) < 2) {
5932 SvANY(cv)->xcv_gv = NULL;
5936 /* if not, anonymise: */
5937 gvname = (GvSTASH(gv) && HvNAME(GvSTASH(gv)) && HvENAME(GvSTASH(gv)))
5938 ? newSVhek(HvENAME_HEK(GvSTASH(gv)))
5939 : newSVpvn_flags( "__ANON__", 8, 0 );
5940 sv_catpvs(gvname, "::__ANON__");
5941 anongv = gv_fetchsv(gvname, GV_ADDMULTI, SVt_PVCV);
5942 SvREFCNT_dec(gvname);
5946 SvANY(cv)->xcv_gv = MUTABLE_GV(SvREFCNT_inc(anongv));
5951 =for apidoc sv_clear
5953 Clear an SV: call any destructors, free up any memory used by the body,
5954 and free the body itself. The SV's head is I<not> freed, although
5955 its type is set to all 1's so that it won't inadvertently be assumed
5956 to be live during global destruction etc.
5957 This function should only be called when REFCNT is zero. Most of the time
5958 you'll want to call C<sv_free()> (or its macro wrapper C<SvREFCNT_dec>)
5965 Perl_sv_clear(pTHX_ SV *const orig_sv)
5970 const struct body_details *sv_type_details;
5973 register SV *sv = orig_sv;
5976 PERL_ARGS_ASSERT_SV_CLEAR;
5978 /* within this loop, sv is the SV currently being freed, and
5979 * iter_sv is the most recent AV or whatever that's being iterated
5980 * over to provide more SVs */
5986 assert(SvREFCNT(sv) == 0);
5987 assert(SvTYPE(sv) != (svtype)SVTYPEMASK);
5989 if (type <= SVt_IV) {
5990 /* See the comment in sv.h about the collusion between this
5991 * early return and the overloading of the NULL slots in the
5995 SvFLAGS(sv) &= SVf_BREAK;
5996 SvFLAGS(sv) |= SVTYPEMASK;
6000 assert(!SvOBJECT(sv) || type >= SVt_PVMG); /* objs are always >= MG */
6002 if (type >= SVt_PVMG) {
6004 if (!curse(sv, 1)) goto get_next_sv;
6005 type = SvTYPE(sv); /* destructor may have changed it */
6007 /* Free back-references before magic, in case the magic calls
6008 * Perl code that has weak references to sv. */
6009 if (type == SVt_PVHV) {
6010 Perl_hv_kill_backrefs(aTHX_ MUTABLE_HV(sv));
6014 else if (type == SVt_PVMG && SvPAD_OUR(sv)) {
6015 SvREFCNT_dec(SvOURSTASH(sv));
6016 } else if (SvMAGIC(sv)) {
6017 /* Free back-references before other types of magic. */
6018 sv_unmagic(sv, PERL_MAGIC_backref);
6021 if (type == SVt_PVMG && SvPAD_TYPED(sv))
6022 SvREFCNT_dec(SvSTASH(sv));
6025 /* case SVt_BIND: */
6028 IoIFP(sv) != PerlIO_stdin() &&
6029 IoIFP(sv) != PerlIO_stdout() &&
6030 IoIFP(sv) != PerlIO_stderr() &&
6031 !(IoFLAGS(sv) & IOf_FAKE_DIRP))
6033 io_close(MUTABLE_IO(sv), FALSE);
6035 if (IoDIRP(sv) && !(IoFLAGS(sv) & IOf_FAKE_DIRP))
6036 PerlDir_close(IoDIRP(sv));
6037 IoDIRP(sv) = (DIR*)NULL;
6038 Safefree(IoTOP_NAME(sv));
6039 Safefree(IoFMT_NAME(sv));
6040 Safefree(IoBOTTOM_NAME(sv));
6043 /* FIXME for plugins */
6044 pregfree2((REGEXP*) sv);
6048 cv_undef(MUTABLE_CV(sv));
6049 /* If we're in a stash, we don't own a reference to it.
6050 * However it does have a back reference to us, which needs to
6052 if ((stash = CvSTASH(sv)))
6053 sv_del_backref(MUTABLE_SV(stash), sv);
6056 if (PL_last_swash_hv == (const HV *)sv) {
6057 PL_last_swash_hv = NULL;
6059 if (HvTOTALKEYS((HV*)sv) > 0) {
6061 /* this statement should match the one at the beginning of
6062 * hv_undef_flags() */
6063 if ( PL_phase != PERL_PHASE_DESTRUCT
6064 && (name = HvNAME((HV*)sv)))
6067 (void)hv_delete(PL_stashcache, name,
6068 HvNAMEUTF8((HV*)sv) ? -HvNAMELEN_get((HV*)sv) : HvNAMELEN_get((HV*)sv), G_DISCARD);
6069 hv_name_set((HV*)sv, NULL, 0, 0);
6072 /* save old iter_sv in unused SvSTASH field */
6073 assert(!SvOBJECT(sv));
6074 SvSTASH(sv) = (HV*)iter_sv;
6077 /* XXX ideally we should save the old value of hash_index
6078 * too, but I can't think of any place to hide it. The
6079 * effect of not saving it is that for freeing hashes of
6080 * hashes, we become quadratic in scanning the HvARRAY of
6081 * the top hash looking for new entries to free; but
6082 * hopefully this will be dwarfed by the freeing of all
6083 * the nested hashes. */
6085 next_sv = Perl_hfree_next_entry(aTHX_ (HV*)sv, &hash_index);
6086 goto get_next_sv; /* process this new sv */
6088 /* free empty hash */
6089 Perl_hv_undef_flags(aTHX_ MUTABLE_HV(sv), HV_NAME_SETALL);
6090 assert(!HvARRAY((HV*)sv));
6094 AV* av = MUTABLE_AV(sv);
6095 if (PL_comppad == av) {
6099 if (AvREAL(av) && AvFILLp(av) > -1) {
6100 next_sv = AvARRAY(av)[AvFILLp(av)--];
6101 /* save old iter_sv in top-most slot of AV,
6102 * and pray that it doesn't get wiped in the meantime */
6103 AvARRAY(av)[AvMAX(av)] = iter_sv;
6105 goto get_next_sv; /* process this new sv */
6107 Safefree(AvALLOC(av));
6112 if (LvTYPE(sv) == 'T') { /* for tie: return HE to pool */
6113 SvREFCNT_dec(HeKEY_sv((HE*)LvTARG(sv)));
6114 HeNEXT((HE*)LvTARG(sv)) = PL_hv_fetch_ent_mh;
6115 PL_hv_fetch_ent_mh = (HE*)LvTARG(sv);
6117 else if (LvTYPE(sv) != 't') /* unless tie: unrefcnted fake SV** */
6118 SvREFCNT_dec(LvTARG(sv));
6120 if (isGV_with_GP(sv)) {
6121 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
6122 && HvENAME_get(stash))
6123 mro_method_changed_in(stash);
6124 gp_free(MUTABLE_GV(sv));
6126 unshare_hek(GvNAME_HEK(sv));
6127 /* If we're in a stash, we don't own a reference to it.
6128 * However it does have a back reference to us, which
6129 * needs to be cleared. */
6130 if (!SvVALID(sv) && (stash = GvSTASH(sv)))
6131 sv_del_backref(MUTABLE_SV(stash), sv);
6133 /* FIXME. There are probably more unreferenced pointers to SVs
6134 * in the interpreter struct that we should check and tidy in
6135 * a similar fashion to this: */
6136 if ((const GV *)sv == PL_last_in_gv)
6137 PL_last_in_gv = NULL;
6143 /* Don't bother with SvOOK_off(sv); as we're only going to
6147 SvOOK_offset(sv, offset);
6148 SvPV_set(sv, SvPVX_mutable(sv) - offset);
6149 /* Don't even bother with turning off the OOK flag. */
6154 SV * const target = SvRV(sv);
6156 sv_del_backref(target, sv);
6161 #ifdef PERL_OLD_COPY_ON_WRITE
6162 else if (SvPVX_const(sv)
6163 && !(SvTYPE(sv) == SVt_PVIO
6164 && !(IoFLAGS(sv) & IOf_FAKE_DIRP)))
6168 PerlIO_printf(Perl_debug_log, "Copy on write: clear\n");
6172 sv_release_COW(sv, SvPVX_const(sv), SV_COW_NEXT_SV(sv));
6174 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
6178 } else if (SvLEN(sv)) {
6179 Safefree(SvPVX_const(sv));
6183 else if (SvPVX_const(sv) && SvLEN(sv)
6184 && !(SvTYPE(sv) == SVt_PVIO
6185 && !(IoFLAGS(sv) & IOf_FAKE_DIRP)))
6186 Safefree(SvPVX_mutable(sv));
6187 else if (SvPVX_const(sv) && SvREADONLY(sv) && SvFAKE(sv)) {
6188 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
6199 SvFLAGS(sv) &= SVf_BREAK;
6200 SvFLAGS(sv) |= SVTYPEMASK;
6202 sv_type_details = bodies_by_type + type;
6203 if (sv_type_details->arena) {
6204 del_body(((char *)SvANY(sv) + sv_type_details->offset),
6205 &PL_body_roots[type]);
6207 else if (sv_type_details->body_size) {
6208 safefree(SvANY(sv));
6212 /* caller is responsible for freeing the head of the original sv */
6213 if (sv != orig_sv && !SvREFCNT(sv))
6216 /* grab and free next sv, if any */
6224 else if (!iter_sv) {
6226 } else if (SvTYPE(iter_sv) == SVt_PVAV) {
6227 AV *const av = (AV*)iter_sv;
6228 if (AvFILLp(av) > -1) {
6229 sv = AvARRAY(av)[AvFILLp(av)--];
6231 else { /* no more elements of current AV to free */
6234 /* restore previous value, squirrelled away */
6235 iter_sv = AvARRAY(av)[AvMAX(av)];
6236 Safefree(AvALLOC(av));
6239 } else if (SvTYPE(iter_sv) == SVt_PVHV) {
6240 sv = Perl_hfree_next_entry(aTHX_ (HV*)iter_sv, &hash_index);
6241 if (!sv && !HvTOTALKEYS((HV *)iter_sv)) {
6242 /* no more elements of current HV to free */
6245 /* Restore previous value of iter_sv, squirrelled away */
6246 assert(!SvOBJECT(sv));
6247 iter_sv = (SV*)SvSTASH(sv);
6249 /* ideally we should restore the old hash_index here,
6250 * but we don't currently save the old value */
6253 /* free any remaining detritus from the hash struct */
6254 Perl_hv_undef_flags(aTHX_ MUTABLE_HV(sv), HV_NAME_SETALL);
6255 assert(!HvARRAY((HV*)sv));
6260 /* unrolled SvREFCNT_dec and sv_free2 follows: */
6264 if (!SvREFCNT(sv)) {
6268 if (--(SvREFCNT(sv)))
6272 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEBUGGING),
6273 "Attempt to free temp prematurely: SV 0x%"UVxf
6274 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
6278 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
6279 /* make sure SvREFCNT(sv)==0 happens very seldom */
6280 SvREFCNT(sv) = (~(U32)0)/2;
6289 /* This routine curses the sv itself, not the object referenced by sv. So
6290 sv does not have to be ROK. */
6293 S_curse(pTHX_ SV * const sv, const bool check_refcnt) {
6296 PERL_ARGS_ASSERT_CURSE;
6297 assert(SvOBJECT(sv));
6299 if (PL_defstash && /* Still have a symbol table? */
6306 stash = SvSTASH(sv);
6307 destructor = StashHANDLER(stash,DESTROY);
6309 /* A constant subroutine can have no side effects, so
6310 don't bother calling it. */
6311 && !CvCONST(destructor)
6312 /* Don't bother calling an empty destructor */
6313 && (CvISXSUB(destructor)
6314 || (CvSTART(destructor)
6315 && (CvSTART(destructor)->op_next->op_type
6318 SV* const tmpref = newRV(sv);
6319 SvREADONLY_on(tmpref); /* DESTROY() could be naughty */
6321 PUSHSTACKi(PERLSI_DESTROY);
6326 call_sv(MUTABLE_SV(destructor),
6327 G_DISCARD|G_EVAL|G_KEEPERR|G_VOID);
6331 if(SvREFCNT(tmpref) < 2) {
6332 /* tmpref is not kept alive! */
6334 SvRV_set(tmpref, NULL);
6337 SvREFCNT_dec(tmpref);
6339 } while (SvOBJECT(sv) && SvSTASH(sv) != stash);
6342 if (check_refcnt && SvREFCNT(sv)) {
6343 if (PL_in_clean_objs)
6345 "DESTROY created new reference to dead object '%"HEKf"'",
6346 HEKfARG(HvNAME_HEK(stash)));
6347 /* DESTROY gave object new lease on life */
6353 SvREFCNT_dec(SvSTASH(sv)); /* possibly of changed persuasion */
6354 SvOBJECT_off(sv); /* Curse the object. */
6355 if (SvTYPE(sv) != SVt_PVIO)
6356 --PL_sv_objcount;/* XXX Might want something more general */
6362 =for apidoc sv_newref
6364 Increment an SV's reference count. Use the C<SvREFCNT_inc()> wrapper
6371 Perl_sv_newref(pTHX_ SV *const sv)
6373 PERL_UNUSED_CONTEXT;
6382 Decrement an SV's reference count, and if it drops to zero, call
6383 C<sv_clear> to invoke destructors and free up any memory used by
6384 the body; finally, deallocate the SV's head itself.
6385 Normally called via a wrapper macro C<SvREFCNT_dec>.
6391 Perl_sv_free(pTHX_ SV *const sv)
6396 if (SvREFCNT(sv) == 0) {
6397 if (SvFLAGS(sv) & SVf_BREAK)
6398 /* this SV's refcnt has been artificially decremented to
6399 * trigger cleanup */
6401 if (PL_in_clean_all) /* All is fair */
6403 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
6404 /* make sure SvREFCNT(sv)==0 happens very seldom */
6405 SvREFCNT(sv) = (~(U32)0)/2;
6408 if (ckWARN_d(WARN_INTERNAL)) {
6409 #ifdef DEBUG_LEAKING_SCALARS_FORK_DUMP
6410 Perl_dump_sv_child(aTHX_ sv);
6412 #ifdef DEBUG_LEAKING_SCALARS
6415 #ifdef DEBUG_LEAKING_SCALARS_ABORT
6416 if (PL_warnhook == PERL_WARNHOOK_FATAL
6417 || ckDEAD(packWARN(WARN_INTERNAL))) {
6418 /* Don't let Perl_warner cause us to escape our fate: */
6422 /* This may not return: */
6423 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
6424 "Attempt to free unreferenced scalar: SV 0x%"UVxf
6425 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
6428 #ifdef DEBUG_LEAKING_SCALARS_ABORT
6433 if (--(SvREFCNT(sv)) > 0)
6435 Perl_sv_free2(aTHX_ sv);
6439 Perl_sv_free2(pTHX_ SV *const sv)
6443 PERL_ARGS_ASSERT_SV_FREE2;
6447 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEBUGGING),
6448 "Attempt to free temp prematurely: SV 0x%"UVxf
6449 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
6453 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
6454 /* make sure SvREFCNT(sv)==0 happens very seldom */
6455 SvREFCNT(sv) = (~(U32)0)/2;
6466 Returns the length of the string in the SV. Handles magic and type
6467 coercion. See also C<SvCUR>, which gives raw access to the xpv_cur slot.
6473 Perl_sv_len(pTHX_ register SV *const sv)
6481 len = mg_length(sv);
6483 (void)SvPV_const(sv, len);
6488 =for apidoc sv_len_utf8
6490 Returns the number of characters in the string in an SV, counting wide
6491 UTF-8 bytes as a single character. Handles magic and type coercion.
6497 * The length is cached in PERL_MAGIC_utf8, in the mg_len field. Also the
6498 * mg_ptr is used, by sv_pos_u2b() and sv_pos_b2u() - see the comments below.
6499 * (Note that the mg_len is not the length of the mg_ptr field.
6500 * This allows the cache to store the character length of the string without
6501 * needing to malloc() extra storage to attach to the mg_ptr.)
6506 Perl_sv_len_utf8(pTHX_ register SV *const sv)
6512 return mg_length(sv);
6516 const U8 *s = (U8*)SvPV_const(sv, len);
6520 MAGIC *mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_utf8) : NULL;
6522 if (mg && (mg->mg_len != -1 || mg->mg_ptr)) {
6523 if (mg->mg_len != -1)
6526 /* We can use the offset cache for a headstart.
6527 The longer value is stored in the first pair. */
6528 STRLEN *cache = (STRLEN *) mg->mg_ptr;
6530 ulen = cache[0] + Perl_utf8_length(aTHX_ s + cache[1],
6534 if (PL_utf8cache < 0) {
6535 const STRLEN real = Perl_utf8_length(aTHX_ s, s + len);
6536 assert_uft8_cache_coherent("sv_len_utf8", ulen, real, sv);
6540 ulen = Perl_utf8_length(aTHX_ s, s + len);
6541 utf8_mg_len_cache_update(sv, &mg, ulen);
6545 return Perl_utf8_length(aTHX_ s, s + len);
6549 /* Walk forwards to find the byte corresponding to the passed in UTF-8
6552 S_sv_pos_u2b_forwards(const U8 *const start, const U8 *const send,
6553 STRLEN *const uoffset_p, bool *const at_end)
6555 const U8 *s = start;
6556 STRLEN uoffset = *uoffset_p;
6558 PERL_ARGS_ASSERT_SV_POS_U2B_FORWARDS;
6560 while (s < send && uoffset) {
6567 else if (s > send) {
6569 /* This is the existing behaviour. Possibly it should be a croak, as
6570 it's actually a bounds error */
6573 *uoffset_p -= uoffset;
6577 /* Given the length of the string in both bytes and UTF-8 characters, decide
6578 whether to walk forwards or backwards to find the byte corresponding to
6579 the passed in UTF-8 offset. */
6581 S_sv_pos_u2b_midway(const U8 *const start, const U8 *send,
6582 STRLEN uoffset, const STRLEN uend)
6584 STRLEN backw = uend - uoffset;
6586 PERL_ARGS_ASSERT_SV_POS_U2B_MIDWAY;
6588 if (uoffset < 2 * backw) {
6589 /* The assumption is that going forwards is twice the speed of going
6590 forward (that's where the 2 * backw comes from).
6591 (The real figure of course depends on the UTF-8 data.) */
6592 const U8 *s = start;
6594 while (s < send && uoffset--)
6604 while (UTF8_IS_CONTINUATION(*send))
6607 return send - start;
6610 /* For the string representation of the given scalar, find the byte
6611 corresponding to the passed in UTF-8 offset. uoffset0 and boffset0
6612 give another position in the string, *before* the sought offset, which
6613 (which is always true, as 0, 0 is a valid pair of positions), which should
6614 help reduce the amount of linear searching.
6615 If *mgp is non-NULL, it should point to the UTF-8 cache magic, which
6616 will be used to reduce the amount of linear searching. The cache will be
6617 created if necessary, and the found value offered to it for update. */
6619 S_sv_pos_u2b_cached(pTHX_ SV *const sv, MAGIC **const mgp, const U8 *const start,
6620 const U8 *const send, STRLEN uoffset,
6621 STRLEN uoffset0, STRLEN boffset0)
6623 STRLEN boffset = 0; /* Actually always set, but let's keep gcc happy. */
6625 bool at_end = FALSE;
6627 PERL_ARGS_ASSERT_SV_POS_U2B_CACHED;
6629 assert (uoffset >= uoffset0);
6636 && (*mgp || (SvTYPE(sv) >= SVt_PVMG &&
6637 (*mgp = mg_find(sv, PERL_MAGIC_utf8))))) {
6638 if ((*mgp)->mg_ptr) {
6639 STRLEN *cache = (STRLEN *) (*mgp)->mg_ptr;
6640 if (cache[0] == uoffset) {
6641 /* An exact match. */
6644 if (cache[2] == uoffset) {
6645 /* An exact match. */
6649 if (cache[0] < uoffset) {
6650 /* The cache already knows part of the way. */
6651 if (cache[0] > uoffset0) {
6652 /* The cache knows more than the passed in pair */
6653 uoffset0 = cache[0];
6654 boffset0 = cache[1];
6656 if ((*mgp)->mg_len != -1) {
6657 /* And we know the end too. */
6659 + sv_pos_u2b_midway(start + boffset0, send,
6661 (*mgp)->mg_len - uoffset0);
6663 uoffset -= uoffset0;
6665 + sv_pos_u2b_forwards(start + boffset0,
6666 send, &uoffset, &at_end);
6667 uoffset += uoffset0;
6670 else if (cache[2] < uoffset) {
6671 /* We're between the two cache entries. */
6672 if (cache[2] > uoffset0) {
6673 /* and the cache knows more than the passed in pair */
6674 uoffset0 = cache[2];
6675 boffset0 = cache[3];
6679 + sv_pos_u2b_midway(start + boffset0,
6682 cache[0] - uoffset0);
6685 + sv_pos_u2b_midway(start + boffset0,
6688 cache[2] - uoffset0);
6692 else if ((*mgp)->mg_len != -1) {
6693 /* If we can take advantage of a passed in offset, do so. */
6694 /* In fact, offset0 is either 0, or less than offset, so don't
6695 need to worry about the other possibility. */
6697 + sv_pos_u2b_midway(start + boffset0, send,
6699 (*mgp)->mg_len - uoffset0);
6704 if (!found || PL_utf8cache < 0) {
6705 STRLEN real_boffset;
6706 uoffset -= uoffset0;
6707 real_boffset = boffset0 + sv_pos_u2b_forwards(start + boffset0,
6708 send, &uoffset, &at_end);
6709 uoffset += uoffset0;
6711 if (found && PL_utf8cache < 0)
6712 assert_uft8_cache_coherent("sv_pos_u2b_cache", boffset,
6714 boffset = real_boffset;
6719 utf8_mg_len_cache_update(sv, mgp, uoffset);
6721 utf8_mg_pos_cache_update(sv, mgp, boffset, uoffset, send - start);
6728 =for apidoc sv_pos_u2b_flags
6730 Converts the value pointed to by offsetp from a count of UTF-8 chars from
6731 the start of the string, to a count of the equivalent number of bytes; if
6732 lenp is non-zero, it does the same to lenp, but this time starting from
6733 the offset, rather than from the start of the string. Handles type coercion.
6734 I<flags> is passed to C<SvPV_flags>, and usually should be
6735 C<SV_GMAGIC|SV_CONST_RETURN> to handle magic.
6741 * sv_pos_u2b_flags() uses, like sv_pos_b2u(), the mg_ptr of the potential
6742 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6743 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
6748 Perl_sv_pos_u2b_flags(pTHX_ SV *const sv, STRLEN uoffset, STRLEN *const lenp,
6755 PERL_ARGS_ASSERT_SV_POS_U2B_FLAGS;
6757 start = (U8*)SvPV_flags(sv, len, flags);
6759 const U8 * const send = start + len;
6761 boffset = sv_pos_u2b_cached(sv, &mg, start, send, uoffset, 0, 0);
6764 && *lenp /* don't bother doing work for 0, as its bytes equivalent
6765 is 0, and *lenp is already set to that. */) {
6766 /* Convert the relative offset to absolute. */
6767 const STRLEN uoffset2 = uoffset + *lenp;
6768 const STRLEN boffset2
6769 = sv_pos_u2b_cached(sv, &mg, start, send, uoffset2,
6770 uoffset, boffset) - boffset;
6784 =for apidoc sv_pos_u2b
6786 Converts the value pointed to by offsetp from a count of UTF-8 chars from
6787 the start of the string, to a count of the equivalent number of bytes; if
6788 lenp is non-zero, it does the same to lenp, but this time starting from
6789 the offset, rather than from the start of the string. Handles magic and
6792 Use C<sv_pos_u2b_flags> in preference, which correctly handles strings longer
6799 * sv_pos_u2b() uses, like sv_pos_b2u(), the mg_ptr of the potential
6800 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6801 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
6805 /* This function is subject to size and sign problems */
6808 Perl_sv_pos_u2b(pTHX_ register SV *const sv, I32 *const offsetp, I32 *const lenp)
6810 PERL_ARGS_ASSERT_SV_POS_U2B;
6813 STRLEN ulen = (STRLEN)*lenp;
6814 *offsetp = (I32)sv_pos_u2b_flags(sv, (STRLEN)*offsetp, &ulen,
6815 SV_GMAGIC|SV_CONST_RETURN);
6818 *offsetp = (I32)sv_pos_u2b_flags(sv, (STRLEN)*offsetp, NULL,
6819 SV_GMAGIC|SV_CONST_RETURN);
6824 S_utf8_mg_len_cache_update(pTHX_ SV *const sv, MAGIC **const mgp,
6827 PERL_ARGS_ASSERT_UTF8_MG_LEN_CACHE_UPDATE;
6831 if (!*mgp && (SvTYPE(sv) < SVt_PVMG ||
6832 !(*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
6833 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, &PL_vtbl_utf8, 0, 0);
6837 (*mgp)->mg_len = ulen;
6838 /* For now, treat "overflowed" as "still unknown". See RT #72924. */
6839 if (ulen != (STRLEN) (*mgp)->mg_len)
6840 (*mgp)->mg_len = -1;
6843 /* Create and update the UTF8 magic offset cache, with the proffered utf8/
6844 byte length pairing. The (byte) length of the total SV is passed in too,
6845 as blen, because for some (more esoteric) SVs, the call to SvPV_const()
6846 may not have updated SvCUR, so we can't rely on reading it directly.
6848 The proffered utf8/byte length pairing isn't used if the cache already has
6849 two pairs, and swapping either for the proffered pair would increase the
6850 RMS of the intervals between known byte offsets.
6852 The cache itself consists of 4 STRLEN values
6853 0: larger UTF-8 offset
6854 1: corresponding byte offset
6855 2: smaller UTF-8 offset
6856 3: corresponding byte offset
6858 Unused cache pairs have the value 0, 0.
6859 Keeping the cache "backwards" means that the invariant of
6860 cache[0] >= cache[2] is maintained even with empty slots, which means that
6861 the code that uses it doesn't need to worry if only 1 entry has actually
6862 been set to non-zero. It also makes the "position beyond the end of the
6863 cache" logic much simpler, as the first slot is always the one to start
6867 S_utf8_mg_pos_cache_update(pTHX_ SV *const sv, MAGIC **const mgp, const STRLEN byte,
6868 const STRLEN utf8, const STRLEN blen)
6872 PERL_ARGS_ASSERT_UTF8_MG_POS_CACHE_UPDATE;
6877 if (!*mgp && (SvTYPE(sv) < SVt_PVMG ||
6878 !(*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
6879 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, (MGVTBL*)&PL_vtbl_utf8, 0,
6881 (*mgp)->mg_len = -1;
6885 if (!(cache = (STRLEN *)(*mgp)->mg_ptr)) {
6886 Newxz(cache, PERL_MAGIC_UTF8_CACHESIZE * 2, STRLEN);
6887 (*mgp)->mg_ptr = (char *) cache;
6891 if (PL_utf8cache < 0 && SvPOKp(sv)) {
6892 /* SvPOKp() because it's possible that sv has string overloading, and
6893 therefore is a reference, hence SvPVX() is actually a pointer.
6894 This cures the (very real) symptoms of RT 69422, but I'm not actually
6895 sure whether we should even be caching the results of UTF-8
6896 operations on overloading, given that nothing stops overloading
6897 returning a different value every time it's called. */
6898 const U8 *start = (const U8 *) SvPVX_const(sv);
6899 const STRLEN realutf8 = utf8_length(start, start + byte);
6901 assert_uft8_cache_coherent("utf8_mg_pos_cache_update", utf8, realutf8,
6905 /* Cache is held with the later position first, to simplify the code
6906 that deals with unbounded ends. */
6908 ASSERT_UTF8_CACHE(cache);
6909 if (cache[1] == 0) {
6910 /* Cache is totally empty */
6913 } else if (cache[3] == 0) {
6914 if (byte > cache[1]) {
6915 /* New one is larger, so goes first. */
6916 cache[2] = cache[0];
6917 cache[3] = cache[1];
6925 #define THREEWAY_SQUARE(a,b,c,d) \
6926 ((float)((d) - (c))) * ((float)((d) - (c))) \
6927 + ((float)((c) - (b))) * ((float)((c) - (b))) \
6928 + ((float)((b) - (a))) * ((float)((b) - (a)))
6930 /* Cache has 2 slots in use, and we know three potential pairs.
6931 Keep the two that give the lowest RMS distance. Do the
6932 calculation in bytes simply because we always know the byte
6933 length. squareroot has the same ordering as the positive value,
6934 so don't bother with the actual square root. */
6935 const float existing = THREEWAY_SQUARE(0, cache[3], cache[1], blen);
6936 if (byte > cache[1]) {
6937 /* New position is after the existing pair of pairs. */
6938 const float keep_earlier
6939 = THREEWAY_SQUARE(0, cache[3], byte, blen);
6940 const float keep_later
6941 = THREEWAY_SQUARE(0, cache[1], byte, blen);
6943 if (keep_later < keep_earlier) {
6944 if (keep_later < existing) {
6945 cache[2] = cache[0];
6946 cache[3] = cache[1];
6952 if (keep_earlier < existing) {
6958 else if (byte > cache[3]) {
6959 /* New position is between the existing pair of pairs. */
6960 const float keep_earlier
6961 = THREEWAY_SQUARE(0, cache[3], byte, blen);
6962 const float keep_later
6963 = THREEWAY_SQUARE(0, byte, cache[1], blen);
6965 if (keep_later < keep_earlier) {
6966 if (keep_later < existing) {
6972 if (keep_earlier < existing) {
6979 /* New position is before the existing pair of pairs. */
6980 const float keep_earlier
6981 = THREEWAY_SQUARE(0, byte, cache[3], blen);
6982 const float keep_later
6983 = THREEWAY_SQUARE(0, byte, cache[1], blen);
6985 if (keep_later < keep_earlier) {
6986 if (keep_later < existing) {
6992 if (keep_earlier < existing) {
6993 cache[0] = cache[2];
6994 cache[1] = cache[3];
7001 ASSERT_UTF8_CACHE(cache);
7004 /* We already know all of the way, now we may be able to walk back. The same
7005 assumption is made as in S_sv_pos_u2b_midway(), namely that walking
7006 backward is half the speed of walking forward. */
7008 S_sv_pos_b2u_midway(pTHX_ const U8 *const s, const U8 *const target,
7009 const U8 *end, STRLEN endu)
7011 const STRLEN forw = target - s;
7012 STRLEN backw = end - target;
7014 PERL_ARGS_ASSERT_SV_POS_B2U_MIDWAY;
7016 if (forw < 2 * backw) {
7017 return utf8_length(s, target);
7020 while (end > target) {
7022 while (UTF8_IS_CONTINUATION(*end)) {
7031 =for apidoc sv_pos_b2u
7033 Converts the value pointed to by offsetp from a count of bytes from the
7034 start of the string, to a count of the equivalent number of UTF-8 chars.
7035 Handles magic and type coercion.
7041 * sv_pos_b2u() uses, like sv_pos_u2b(), the mg_ptr of the potential
7042 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
7047 Perl_sv_pos_b2u(pTHX_ register SV *const sv, I32 *const offsetp)
7050 const STRLEN byte = *offsetp;
7051 STRLEN len = 0; /* Actually always set, but let's keep gcc happy. */
7057 PERL_ARGS_ASSERT_SV_POS_B2U;
7062 s = (const U8*)SvPV_const(sv, blen);
7065 Perl_croak(aTHX_ "panic: sv_pos_b2u: bad byte offset");
7071 && SvTYPE(sv) >= SVt_PVMG
7072 && (mg = mg_find(sv, PERL_MAGIC_utf8)))
7075 STRLEN * const cache = (STRLEN *) mg->mg_ptr;
7076 if (cache[1] == byte) {
7077 /* An exact match. */
7078 *offsetp = cache[0];
7081 if (cache[3] == byte) {
7082 /* An exact match. */
7083 *offsetp = cache[2];
7087 if (cache[1] < byte) {
7088 /* We already know part of the way. */
7089 if (mg->mg_len != -1) {
7090 /* Actually, we know the end too. */
7092 + S_sv_pos_b2u_midway(aTHX_ s + cache[1], send,
7093 s + blen, mg->mg_len - cache[0]);
7095 len = cache[0] + utf8_length(s + cache[1], send);
7098 else if (cache[3] < byte) {
7099 /* We're between the two cached pairs, so we do the calculation
7100 offset by the byte/utf-8 positions for the earlier pair,
7101 then add the utf-8 characters from the string start to
7103 len = S_sv_pos_b2u_midway(aTHX_ s + cache[3], send,
7104 s + cache[1], cache[0] - cache[2])
7108 else { /* cache[3] > byte */
7109 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + cache[3],
7113 ASSERT_UTF8_CACHE(cache);
7115 } else if (mg->mg_len != -1) {
7116 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + blen, mg->mg_len);
7120 if (!found || PL_utf8cache < 0) {
7121 const STRLEN real_len = utf8_length(s, send);
7123 if (found && PL_utf8cache < 0)
7124 assert_uft8_cache_coherent("sv_pos_b2u", len, real_len, sv);
7131 utf8_mg_len_cache_update(sv, &mg, len);
7133 utf8_mg_pos_cache_update(sv, &mg, byte, len, blen);
7138 S_assert_uft8_cache_coherent(pTHX_ const char *const func, STRLEN from_cache,
7139 STRLEN real, SV *const sv)
7141 PERL_ARGS_ASSERT_ASSERT_UFT8_CACHE_COHERENT;
7143 /* As this is debugging only code, save space by keeping this test here,
7144 rather than inlining it in all the callers. */
7145 if (from_cache == real)
7148 /* Need to turn the assertions off otherwise we may recurse infinitely
7149 while printing error messages. */
7150 SAVEI8(PL_utf8cache);
7152 Perl_croak(aTHX_ "panic: %s cache %"UVuf" real %"UVuf" for %"SVf,
7153 func, (UV) from_cache, (UV) real, SVfARG(sv));
7159 Returns a boolean indicating whether the strings in the two SVs are
7160 identical. Is UTF-8 and 'use bytes' aware, handles get magic, and will
7161 coerce its args to strings if necessary.
7163 =for apidoc sv_eq_flags
7165 Returns a boolean indicating whether the strings in the two SVs are
7166 identical. Is UTF-8 and 'use bytes' aware and coerces its args to strings
7167 if necessary. If the flags include SV_GMAGIC, it handles get-magic, too.
7173 Perl_sv_eq_flags(pTHX_ register SV *sv1, register SV *sv2, const U32 flags)
7182 SV* svrecode = NULL;
7189 /* if pv1 and pv2 are the same, second SvPV_const call may
7190 * invalidate pv1 (if we are handling magic), so we may need to
7192 if (sv1 == sv2 && flags & SV_GMAGIC
7193 && (SvTHINKFIRST(sv1) || SvGMAGICAL(sv1))) {
7194 pv1 = SvPV_const(sv1, cur1);
7195 sv1 = newSVpvn_flags(pv1, cur1, SVs_TEMP | SvUTF8(sv2));
7197 pv1 = SvPV_flags_const(sv1, cur1, flags);
7205 pv2 = SvPV_flags_const(sv2, cur2, flags);
7207 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
7208 /* Differing utf8ness.
7209 * Do not UTF8size the comparands as a side-effect. */
7212 svrecode = newSVpvn(pv2, cur2);
7213 sv_recode_to_utf8(svrecode, PL_encoding);
7214 pv2 = SvPV_const(svrecode, cur2);
7217 svrecode = newSVpvn(pv1, cur1);
7218 sv_recode_to_utf8(svrecode, PL_encoding);
7219 pv1 = SvPV_const(svrecode, cur1);
7221 /* Now both are in UTF-8. */
7223 SvREFCNT_dec(svrecode);
7229 /* sv1 is the UTF-8 one */
7230 return bytes_cmp_utf8((const U8*)pv2, cur2,
7231 (const U8*)pv1, cur1) == 0;
7234 /* sv2 is the UTF-8 one */
7235 return bytes_cmp_utf8((const U8*)pv1, cur1,
7236 (const U8*)pv2, cur2) == 0;
7242 eq = (pv1 == pv2) || memEQ(pv1, pv2, cur1);
7244 SvREFCNT_dec(svrecode);
7254 Compares the strings in two SVs. Returns -1, 0, or 1 indicating whether the
7255 string in C<sv1> is less than, equal to, or greater than the string in
7256 C<sv2>. Is UTF-8 and 'use bytes' aware, handles get magic, and will
7257 coerce its args to strings if necessary. See also C<sv_cmp_locale>.
7259 =for apidoc sv_cmp_flags
7261 Compares the strings in two SVs. Returns -1, 0, or 1 indicating whether the
7262 string in C<sv1> is less than, equal to, or greater than the string in
7263 C<sv2>. Is UTF-8 and 'use bytes' aware and will coerce its args to strings
7264 if necessary. If the flags include SV_GMAGIC, it handles get magic. See
7265 also C<sv_cmp_locale_flags>.
7271 Perl_sv_cmp(pTHX_ register SV *const sv1, register SV *const sv2)
7273 return sv_cmp_flags(sv1, sv2, SV_GMAGIC);
7277 Perl_sv_cmp_flags(pTHX_ register SV *const sv1, register SV *const sv2,
7282 const char *pv1, *pv2;
7285 SV *svrecode = NULL;
7292 pv1 = SvPV_flags_const(sv1, cur1, flags);
7299 pv2 = SvPV_flags_const(sv2, cur2, flags);
7301 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
7302 /* Differing utf8ness.
7303 * Do not UTF8size the comparands as a side-effect. */
7306 svrecode = newSVpvn(pv2, cur2);
7307 sv_recode_to_utf8(svrecode, PL_encoding);
7308 pv2 = SvPV_const(svrecode, cur2);
7311 const int retval = -bytes_cmp_utf8((const U8*)pv2, cur2,
7312 (const U8*)pv1, cur1);
7313 return retval ? retval < 0 ? -1 : +1 : 0;
7318 svrecode = newSVpvn(pv1, cur1);
7319 sv_recode_to_utf8(svrecode, PL_encoding);
7320 pv1 = SvPV_const(svrecode, cur1);
7323 const int retval = bytes_cmp_utf8((const U8*)pv1, cur1,
7324 (const U8*)pv2, cur2);
7325 return retval ? retval < 0 ? -1 : +1 : 0;
7331 cmp = cur2 ? -1 : 0;
7335 const I32 retval = memcmp((const void*)pv1, (const void*)pv2, cur1 < cur2 ? cur1 : cur2);
7338 cmp = retval < 0 ? -1 : 1;
7339 } else if (cur1 == cur2) {
7342 cmp = cur1 < cur2 ? -1 : 1;
7346 SvREFCNT_dec(svrecode);
7354 =for apidoc sv_cmp_locale
7356 Compares the strings in two SVs in a locale-aware manner. Is UTF-8 and
7357 'use bytes' aware, handles get magic, and will coerce its args to strings
7358 if necessary. See also C<sv_cmp>.
7360 =for apidoc sv_cmp_locale_flags
7362 Compares the strings in two SVs in a locale-aware manner. Is UTF-8 and
7363 'use bytes' aware and will coerce its args to strings if necessary. If the
7364 flags contain SV_GMAGIC, it handles get magic. See also C<sv_cmp_flags>.
7370 Perl_sv_cmp_locale(pTHX_ register SV *const sv1, register SV *const sv2)
7372 return sv_cmp_locale_flags(sv1, sv2, SV_GMAGIC);
7376 Perl_sv_cmp_locale_flags(pTHX_ register SV *const sv1, register SV *const sv2,
7380 #ifdef USE_LOCALE_COLLATE
7386 if (PL_collation_standard)
7390 pv1 = sv1 ? sv_collxfrm_flags(sv1, &len1, flags) : (char *) NULL;
7392 pv2 = sv2 ? sv_collxfrm_flags(sv2, &len2, flags) : (char *) NULL;
7394 if (!pv1 || !len1) {
7405 retval = memcmp((void*)pv1, (void*)pv2, len1 < len2 ? len1 : len2);
7408 return retval < 0 ? -1 : 1;
7411 * When the result of collation is equality, that doesn't mean
7412 * that there are no differences -- some locales exclude some
7413 * characters from consideration. So to avoid false equalities,
7414 * we use the raw string as a tiebreaker.
7420 #endif /* USE_LOCALE_COLLATE */
7422 return sv_cmp(sv1, sv2);
7426 #ifdef USE_LOCALE_COLLATE
7429 =for apidoc sv_collxfrm
7431 This calls C<sv_collxfrm_flags> with the SV_GMAGIC flag. See
7432 C<sv_collxfrm_flags>.
7434 =for apidoc sv_collxfrm_flags
7436 Add Collate Transform magic to an SV if it doesn't already have it. If the
7437 flags contain SV_GMAGIC, it handles get-magic.
7439 Any scalar variable may carry PERL_MAGIC_collxfrm magic that contains the
7440 scalar data of the variable, but transformed to such a format that a normal
7441 memory comparison can be used to compare the data according to the locale
7448 Perl_sv_collxfrm_flags(pTHX_ SV *const sv, STRLEN *const nxp, const I32 flags)
7453 PERL_ARGS_ASSERT_SV_COLLXFRM_FLAGS;
7455 mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_collxfrm) : (MAGIC *) NULL;
7456 if (!mg || !mg->mg_ptr || *(U32*)mg->mg_ptr != PL_collation_ix) {
7462 Safefree(mg->mg_ptr);
7463 s = SvPV_flags_const(sv, len, flags);
7464 if ((xf = mem_collxfrm(s, len, &xlen))) {
7466 #ifdef PERL_OLD_COPY_ON_WRITE
7468 sv_force_normal_flags(sv, 0);
7470 mg = sv_magicext(sv, 0, PERL_MAGIC_collxfrm, &PL_vtbl_collxfrm,
7484 if (mg && mg->mg_ptr) {
7486 return mg->mg_ptr + sizeof(PL_collation_ix);
7494 #endif /* USE_LOCALE_COLLATE */
7497 S_sv_gets_append_to_utf8(pTHX_ SV *const sv, PerlIO *const fp, I32 append)
7499 SV * const tsv = newSV(0);
7502 sv_gets(tsv, fp, 0);
7503 sv_utf8_upgrade_nomg(tsv);
7504 SvCUR_set(sv,append);
7507 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7511 S_sv_gets_read_record(pTHX_ SV *const sv, PerlIO *const fp, I32 append)
7514 const U32 recsize = SvUV(SvRV(PL_rs)); /* RsRECORD() guarantees > 0. */
7515 /* Grab the size of the record we're getting */
7516 char *const buffer = SvGROW(sv, (STRLEN)(recsize + append + 1)) + append;
7523 /* VMS wants read instead of fread, because fread doesn't respect */
7524 /* RMS record boundaries. This is not necessarily a good thing to be */
7525 /* doing, but we've got no other real choice - except avoid stdio
7526 as implementation - perhaps write a :vms layer ?
7528 fd = PerlIO_fileno(fp);
7530 bytesread = PerlLIO_read(fd, buffer, recsize);
7532 else /* in-memory file from PerlIO::Scalar */
7535 bytesread = PerlIO_read(fp, buffer, recsize);
7540 SvCUR_set(sv, bytesread + append);
7541 buffer[bytesread] = '\0';
7542 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7548 Get a line from the filehandle and store it into the SV, optionally
7549 appending to the currently-stored string.
7555 Perl_sv_gets(pTHX_ register SV *const sv, register PerlIO *const fp, I32 append)
7560 register STDCHAR rslast;
7561 register STDCHAR *bp;
7566 PERL_ARGS_ASSERT_SV_GETS;
7568 if (SvTHINKFIRST(sv))
7569 sv_force_normal_flags(sv, append ? 0 : SV_COW_DROP_PV);
7570 /* XXX. If you make this PVIV, then copy on write can copy scalars read
7572 However, perlbench says it's slower, because the existing swipe code
7573 is faster than copy on write.
7574 Swings and roundabouts. */
7575 SvUPGRADE(sv, SVt_PV);
7580 if (PerlIO_isutf8(fp)) {
7582 sv_utf8_upgrade_nomg(sv);
7583 sv_pos_u2b(sv,&append,0);
7585 } else if (SvUTF8(sv)) {
7586 return S_sv_gets_append_to_utf8(aTHX_ sv, fp, append);
7594 if (PerlIO_isutf8(fp))
7597 if (IN_PERL_COMPILETIME) {
7598 /* we always read code in line mode */
7602 else if (RsSNARF(PL_rs)) {
7603 /* If it is a regular disk file use size from stat() as estimate
7604 of amount we are going to read -- may result in mallocing
7605 more memory than we really need if the layers below reduce
7606 the size we read (e.g. CRLF or a gzip layer).
7609 if (!PerlLIO_fstat(PerlIO_fileno(fp), &st) && S_ISREG(st.st_mode)) {
7610 const Off_t offset = PerlIO_tell(fp);
7611 if (offset != (Off_t) -1 && st.st_size + append > offset) {
7612 (void) SvGROW(sv, (STRLEN)((st.st_size - offset) + append + 1));
7618 else if (RsRECORD(PL_rs)) {
7619 return S_sv_gets_read_record(aTHX_ sv, fp, append);
7621 else if (RsPARA(PL_rs)) {
7627 /* Get $/ i.e. PL_rs into same encoding as stream wants */
7628 if (PerlIO_isutf8(fp)) {
7629 rsptr = SvPVutf8(PL_rs, rslen);
7632 if (SvUTF8(PL_rs)) {
7633 if (!sv_utf8_downgrade(PL_rs, TRUE)) {
7634 Perl_croak(aTHX_ "Wide character in $/");
7637 rsptr = SvPV_const(PL_rs, rslen);
7641 rslast = rslen ? rsptr[rslen - 1] : '\0';
7643 if (rspara) { /* have to do this both before and after */
7644 do { /* to make sure file boundaries work right */
7647 i = PerlIO_getc(fp);
7651 PerlIO_ungetc(fp,i);
7657 /* See if we know enough about I/O mechanism to cheat it ! */
7659 /* This used to be #ifdef test - it is made run-time test for ease
7660 of abstracting out stdio interface. One call should be cheap
7661 enough here - and may even be a macro allowing compile
7665 if (PerlIO_fast_gets(fp)) {
7668 * We're going to steal some values from the stdio struct
7669 * and put EVERYTHING in the innermost loop into registers.
7671 register STDCHAR *ptr;
7675 #if defined(VMS) && defined(PERLIO_IS_STDIO)
7676 /* An ungetc()d char is handled separately from the regular
7677 * buffer, so we getc() it back out and stuff it in the buffer.
7679 i = PerlIO_getc(fp);
7680 if (i == EOF) return 0;
7681 *(--((*fp)->_ptr)) = (unsigned char) i;
7685 /* Here is some breathtakingly efficient cheating */
7687 cnt = PerlIO_get_cnt(fp); /* get count into register */
7688 /* make sure we have the room */
7689 if ((I32)(SvLEN(sv) - append) <= cnt + 1) {
7690 /* Not room for all of it
7691 if we are looking for a separator and room for some
7693 if (rslen && cnt > 80 && (I32)SvLEN(sv) > append) {
7694 /* just process what we have room for */
7695 shortbuffered = cnt - SvLEN(sv) + append + 1;
7696 cnt -= shortbuffered;
7700 /* remember that cnt can be negative */
7701 SvGROW(sv, (STRLEN)(append + (cnt <= 0 ? 2 : (cnt + 1))));
7706 bp = (STDCHAR*)SvPVX_const(sv) + append; /* move these two too to registers */
7707 ptr = (STDCHAR*)PerlIO_get_ptr(fp);
7708 DEBUG_P(PerlIO_printf(Perl_debug_log,
7709 "Screamer: entering, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7710 DEBUG_P(PerlIO_printf(Perl_debug_log,
7711 "Screamer: entering: PerlIO * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7712 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7713 PTR2UV(PerlIO_has_base(fp) ? PerlIO_get_base(fp) : 0)));
7718 while (cnt > 0) { /* this | eat */
7720 if ((*bp++ = *ptr++) == rslast) /* really | dust */
7721 goto thats_all_folks; /* screams | sed :-) */
7725 Copy(ptr, bp, cnt, char); /* this | eat */
7726 bp += cnt; /* screams | dust */
7727 ptr += cnt; /* louder | sed :-) */
7729 assert (!shortbuffered);
7730 goto cannot_be_shortbuffered;
7734 if (shortbuffered) { /* oh well, must extend */
7735 cnt = shortbuffered;
7737 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
7739 SvGROW(sv, SvLEN(sv) + append + cnt + 2);
7740 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
7744 cannot_be_shortbuffered:
7745 DEBUG_P(PerlIO_printf(Perl_debug_log,
7746 "Screamer: going to getc, ptr=%"UVuf", cnt=%ld\n",
7747 PTR2UV(ptr),(long)cnt));
7748 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* deregisterize cnt and ptr */
7750 DEBUG_Pv(PerlIO_printf(Perl_debug_log,
7751 "Screamer: pre: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7752 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7753 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7755 /* This used to call 'filbuf' in stdio form, but as that behaves like
7756 getc when cnt <= 0 we use PerlIO_getc here to avoid introducing
7757 another abstraction. */
7758 i = PerlIO_getc(fp); /* get more characters */
7760 DEBUG_Pv(PerlIO_printf(Perl_debug_log,
7761 "Screamer: post: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7762 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7763 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7765 cnt = PerlIO_get_cnt(fp);
7766 ptr = (STDCHAR*)PerlIO_get_ptr(fp); /* reregisterize cnt and ptr */
7767 DEBUG_P(PerlIO_printf(Perl_debug_log,
7768 "Screamer: after getc, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7770 if (i == EOF) /* all done for ever? */
7771 goto thats_really_all_folks;
7773 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
7775 SvGROW(sv, bpx + cnt + 2);
7776 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
7778 *bp++ = (STDCHAR)i; /* store character from PerlIO_getc */
7780 if (rslen && (STDCHAR)i == rslast) /* all done for now? */
7781 goto thats_all_folks;
7785 if ((rslen > 1 && (STRLEN)(bp - (STDCHAR*)SvPVX_const(sv)) < rslen) ||
7786 memNE((char*)bp - rslen, rsptr, rslen))
7787 goto screamer; /* go back to the fray */
7788 thats_really_all_folks:
7790 cnt += shortbuffered;
7791 DEBUG_P(PerlIO_printf(Perl_debug_log,
7792 "Screamer: quitting, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7793 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* put these back or we're in trouble */
7794 DEBUG_P(PerlIO_printf(Perl_debug_log,
7795 "Screamer: end: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7796 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7797 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7799 SvCUR_set(sv, bp - (STDCHAR*)SvPVX_const(sv)); /* set length */
7800 DEBUG_P(PerlIO_printf(Perl_debug_log,
7801 "Screamer: done, len=%ld, string=|%.*s|\n",
7802 (long)SvCUR(sv),(int)SvCUR(sv),SvPVX_const(sv)));
7806 /*The big, slow, and stupid way. */
7807 #ifdef USE_HEAP_INSTEAD_OF_STACK /* Even slower way. */
7808 STDCHAR *buf = NULL;
7809 Newx(buf, 8192, STDCHAR);
7817 register const STDCHAR * const bpe = buf + sizeof(buf);
7819 while ((i = PerlIO_getc(fp)) != EOF && (*bp++ = (STDCHAR)i) != rslast && bp < bpe)
7820 ; /* keep reading */
7824 cnt = PerlIO_read(fp,(char*)buf, sizeof(buf));
7825 /* Accommodate broken VAXC compiler, which applies U8 cast to
7826 * both args of ?: operator, causing EOF to change into 255
7829 i = (U8)buf[cnt - 1];
7835 cnt = 0; /* we do need to re-set the sv even when cnt <= 0 */
7837 sv_catpvn(sv, (char *) buf, cnt);
7839 sv_setpvn(sv, (char *) buf, cnt);
7841 if (i != EOF && /* joy */
7843 SvCUR(sv) < rslen ||
7844 memNE(SvPVX_const(sv) + SvCUR(sv) - rslen, rsptr, rslen)))
7848 * If we're reading from a TTY and we get a short read,
7849 * indicating that the user hit his EOF character, we need
7850 * to notice it now, because if we try to read from the TTY
7851 * again, the EOF condition will disappear.
7853 * The comparison of cnt to sizeof(buf) is an optimization
7854 * that prevents unnecessary calls to feof().
7858 if (!(cnt < (I32)sizeof(buf) && PerlIO_eof(fp)))
7862 #ifdef USE_HEAP_INSTEAD_OF_STACK
7867 if (rspara) { /* have to do this both before and after */
7868 while (i != EOF) { /* to make sure file boundaries work right */
7869 i = PerlIO_getc(fp);
7871 PerlIO_ungetc(fp,i);
7877 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7883 Auto-increment of the value in the SV, doing string to numeric conversion
7884 if necessary. Handles 'get' magic and operator overloading.
7890 Perl_sv_inc(pTHX_ register SV *const sv)
7899 =for apidoc sv_inc_nomg
7901 Auto-increment of the value in the SV, doing string to numeric conversion
7902 if necessary. Handles operator overloading. Skips handling 'get' magic.
7908 Perl_sv_inc_nomg(pTHX_ register SV *const sv)
7916 if (SvTHINKFIRST(sv)) {
7917 if (SvIsCOW(sv) || isGV_with_GP(sv))
7918 sv_force_normal_flags(sv, 0);
7919 if (SvREADONLY(sv)) {
7920 if (IN_PERL_RUNTIME)
7921 Perl_croak_no_modify(aTHX);
7925 if (SvAMAGIC(sv) && AMG_CALLunary(sv, inc_amg))
7927 i = PTR2IV(SvRV(sv));
7932 flags = SvFLAGS(sv);
7933 if ((flags & (SVp_NOK|SVp_IOK)) == SVp_NOK) {
7934 /* It's (privately or publicly) a float, but not tested as an
7935 integer, so test it to see. */
7937 flags = SvFLAGS(sv);
7939 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
7940 /* It's publicly an integer, or privately an integer-not-float */
7941 #ifdef PERL_PRESERVE_IVUV
7945 if (SvUVX(sv) == UV_MAX)
7946 sv_setnv(sv, UV_MAX_P1);
7948 (void)SvIOK_only_UV(sv);
7949 SvUV_set(sv, SvUVX(sv) + 1);
7951 if (SvIVX(sv) == IV_MAX)
7952 sv_setuv(sv, (UV)IV_MAX + 1);
7954 (void)SvIOK_only(sv);
7955 SvIV_set(sv, SvIVX(sv) + 1);
7960 if (flags & SVp_NOK) {
7961 const NV was = SvNVX(sv);
7962 if (NV_OVERFLOWS_INTEGERS_AT &&
7963 was >= NV_OVERFLOWS_INTEGERS_AT) {
7964 Perl_ck_warner(aTHX_ packWARN(WARN_IMPRECISION),
7965 "Lost precision when incrementing %" NVff " by 1",
7968 (void)SvNOK_only(sv);
7969 SvNV_set(sv, was + 1.0);
7973 if (!(flags & SVp_POK) || !*SvPVX_const(sv)) {
7974 if ((flags & SVTYPEMASK) < SVt_PVIV)
7975 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV ? SVt_PVIV : SVt_IV));
7976 (void)SvIOK_only(sv);
7981 while (isALPHA(*d)) d++;
7982 while (isDIGIT(*d)) d++;
7983 if (d < SvEND(sv)) {
7984 #ifdef PERL_PRESERVE_IVUV
7985 /* Got to punt this as an integer if needs be, but we don't issue
7986 warnings. Probably ought to make the sv_iv_please() that does
7987 the conversion if possible, and silently. */
7988 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
7989 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
7990 /* Need to try really hard to see if it's an integer.
7991 9.22337203685478e+18 is an integer.
7992 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
7993 so $a="9.22337203685478e+18"; $a+0; $a++
7994 needs to be the same as $a="9.22337203685478e+18"; $a++
8001 /* sv_2iv *should* have made this an NV */
8002 if (flags & SVp_NOK) {
8003 (void)SvNOK_only(sv);
8004 SvNV_set(sv, SvNVX(sv) + 1.0);
8007 /* I don't think we can get here. Maybe I should assert this
8008 And if we do get here I suspect that sv_setnv will croak. NWC
8010 #if defined(USE_LONG_DOUBLE)
8011 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",
8012 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
8014 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_inc punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
8015 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
8018 #endif /* PERL_PRESERVE_IVUV */
8019 sv_setnv(sv,Atof(SvPVX_const(sv)) + 1.0);
8023 while (d >= SvPVX_const(sv)) {
8031 /* MKS: The original code here died if letters weren't consecutive.
8032 * at least it didn't have to worry about non-C locales. The
8033 * new code assumes that ('z'-'a')==('Z'-'A'), letters are
8034 * arranged in order (although not consecutively) and that only
8035 * [A-Za-z] are accepted by isALPHA in the C locale.
8037 if (*d != 'z' && *d != 'Z') {
8038 do { ++*d; } while (!isALPHA(*d));
8041 *(d--) -= 'z' - 'a';
8046 *(d--) -= 'z' - 'a' + 1;
8050 /* oh,oh, the number grew */
8051 SvGROW(sv, SvCUR(sv) + 2);
8052 SvCUR_set(sv, SvCUR(sv) + 1);
8053 for (d = SvPVX(sv) + SvCUR(sv); d > SvPVX_const(sv); d--)
8064 Auto-decrement of the value in the SV, doing string to numeric conversion
8065 if necessary. Handles 'get' magic and operator overloading.
8071 Perl_sv_dec(pTHX_ register SV *const sv)
8081 =for apidoc sv_dec_nomg
8083 Auto-decrement of the value in the SV, doing string to numeric conversion
8084 if necessary. Handles operator overloading. Skips handling 'get' magic.
8090 Perl_sv_dec_nomg(pTHX_ register SV *const sv)
8097 if (SvTHINKFIRST(sv)) {
8098 if (SvIsCOW(sv) || isGV_with_GP(sv))
8099 sv_force_normal_flags(sv, 0);
8100 if (SvREADONLY(sv)) {
8101 if (IN_PERL_RUNTIME)
8102 Perl_croak_no_modify(aTHX);
8106 if (SvAMAGIC(sv) && AMG_CALLunary(sv, dec_amg))
8108 i = PTR2IV(SvRV(sv));
8113 /* Unlike sv_inc we don't have to worry about string-never-numbers
8114 and keeping them magic. But we mustn't warn on punting */
8115 flags = SvFLAGS(sv);
8116 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
8117 /* It's publicly an integer, or privately an integer-not-float */
8118 #ifdef PERL_PRESERVE_IVUV
8122 if (SvUVX(sv) == 0) {
8123 (void)SvIOK_only(sv);
8127 (void)SvIOK_only_UV(sv);
8128 SvUV_set(sv, SvUVX(sv) - 1);
8131 if (SvIVX(sv) == IV_MIN) {
8132 sv_setnv(sv, (NV)IV_MIN);
8136 (void)SvIOK_only(sv);
8137 SvIV_set(sv, SvIVX(sv) - 1);
8142 if (flags & SVp_NOK) {
8145 const NV was = SvNVX(sv);
8146 if (NV_OVERFLOWS_INTEGERS_AT &&
8147 was <= -NV_OVERFLOWS_INTEGERS_AT) {
8148 Perl_ck_warner(aTHX_ packWARN(WARN_IMPRECISION),
8149 "Lost precision when decrementing %" NVff " by 1",
8152 (void)SvNOK_only(sv);
8153 SvNV_set(sv, was - 1.0);
8157 if (!(flags & SVp_POK)) {
8158 if ((flags & SVTYPEMASK) < SVt_PVIV)
8159 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV) ? SVt_PVIV : SVt_IV);
8161 (void)SvIOK_only(sv);
8164 #ifdef PERL_PRESERVE_IVUV
8166 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
8167 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
8168 /* Need to try really hard to see if it's an integer.
8169 9.22337203685478e+18 is an integer.
8170 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
8171 so $a="9.22337203685478e+18"; $a+0; $a--
8172 needs to be the same as $a="9.22337203685478e+18"; $a--
8179 /* sv_2iv *should* have made this an NV */
8180 if (flags & SVp_NOK) {
8181 (void)SvNOK_only(sv);
8182 SvNV_set(sv, SvNVX(sv) - 1.0);
8185 /* I don't think we can get here. Maybe I should assert this
8186 And if we do get here I suspect that sv_setnv will croak. NWC
8188 #if defined(USE_LONG_DOUBLE)
8189 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",
8190 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
8192 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_dec punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
8193 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
8197 #endif /* PERL_PRESERVE_IVUV */
8198 sv_setnv(sv,Atof(SvPVX_const(sv)) - 1.0); /* punt */
8201 /* this define is used to eliminate a chunk of duplicated but shared logic
8202 * it has the suffix __SV_C to signal that it isnt API, and isnt meant to be
8203 * used anywhere but here - yves
8205 #define PUSH_EXTEND_MORTAL__SV_C(AnSv) \
8208 PL_tmps_stack[++PL_tmps_ix] = (AnSv); \
8212 =for apidoc sv_mortalcopy
8214 Creates a new SV which is a copy of the original SV (using C<sv_setsv>).
8215 The new SV is marked as mortal. It will be destroyed "soon", either by an
8216 explicit call to FREETMPS, or by an implicit call at places such as
8217 statement boundaries. See also C<sv_newmortal> and C<sv_2mortal>.
8222 /* Make a string that will exist for the duration of the expression
8223 * evaluation. Actually, it may have to last longer than that, but
8224 * hopefully we won't free it until it has been assigned to a
8225 * permanent location. */
8228 Perl_sv_mortalcopy(pTHX_ SV *const oldstr)
8234 sv_setsv(sv,oldstr);
8235 PUSH_EXTEND_MORTAL__SV_C(sv);
8241 =for apidoc sv_newmortal
8243 Creates a new null SV which is mortal. The reference count of the SV is
8244 set to 1. It will be destroyed "soon", either by an explicit call to
8245 FREETMPS, or by an implicit call at places such as statement boundaries.
8246 See also C<sv_mortalcopy> and C<sv_2mortal>.
8252 Perl_sv_newmortal(pTHX)
8258 SvFLAGS(sv) = SVs_TEMP;
8259 PUSH_EXTEND_MORTAL__SV_C(sv);
8265 =for apidoc newSVpvn_flags
8267 Creates a new SV and copies a string into it. The reference count for the
8268 SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
8269 string. You are responsible for ensuring that the source string is at least
8270 C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
8271 Currently the only flag bits accepted are C<SVf_UTF8> and C<SVs_TEMP>.
8272 If C<SVs_TEMP> is set, then C<sv_2mortal()> is called on the result before
8273 returning. If C<SVf_UTF8> is set, C<s> is considered to be in UTF-8 and the
8274 C<SVf_UTF8> flag will be set on the new SV.
8275 C<newSVpvn_utf8()> is a convenience wrapper for this function, defined as
8277 #define newSVpvn_utf8(s, len, u) \
8278 newSVpvn_flags((s), (len), (u) ? SVf_UTF8 : 0)
8284 Perl_newSVpvn_flags(pTHX_ const char *const s, const STRLEN len, const U32 flags)
8289 /* All the flags we don't support must be zero.
8290 And we're new code so I'm going to assert this from the start. */
8291 assert(!(flags & ~(SVf_UTF8|SVs_TEMP)));
8293 sv_setpvn(sv,s,len);
8295 /* This code used to a sv_2mortal(), however we now unroll the call to sv_2mortal()
8296 * and do what it does ourselves here.
8297 * Since we have asserted that flags can only have the SVf_UTF8 and/or SVs_TEMP flags
8298 * set above we can use it to enable the sv flags directly (bypassing SvTEMP_on), which
8299 * in turn means we dont need to mask out the SVf_UTF8 flag below, which means that we
8300 * eliminate quite a few steps than it looks - Yves (explaining patch by gfx)
8303 SvFLAGS(sv) |= flags;
8305 if(flags & SVs_TEMP){
8306 PUSH_EXTEND_MORTAL__SV_C(sv);
8313 =for apidoc sv_2mortal
8315 Marks an existing SV as mortal. The SV will be destroyed "soon", either
8316 by an explicit call to FREETMPS, or by an implicit call at places such as
8317 statement boundaries. SvTEMP() is turned on which means that the SV's
8318 string buffer can be "stolen" if this SV is copied. See also C<sv_newmortal>
8319 and C<sv_mortalcopy>.
8325 Perl_sv_2mortal(pTHX_ register SV *const sv)
8330 if (SvREADONLY(sv) && SvIMMORTAL(sv))
8332 PUSH_EXTEND_MORTAL__SV_C(sv);
8340 Creates a new SV and copies a string into it. The reference count for the
8341 SV is set to 1. If C<len> is zero, Perl will compute the length using
8342 strlen(). For efficiency, consider using C<newSVpvn> instead.
8348 Perl_newSVpv(pTHX_ const char *const s, const STRLEN len)
8354 sv_setpvn(sv, s, len || s == NULL ? len : strlen(s));
8359 =for apidoc newSVpvn
8361 Creates a new SV and copies a string into it. The reference count for the
8362 SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
8363 string. You are responsible for ensuring that the source string is at least
8364 C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
8370 Perl_newSVpvn(pTHX_ const char *const s, const STRLEN len)
8376 sv_setpvn(sv,s,len);
8381 =for apidoc newSVhek
8383 Creates a new SV from the hash key structure. It will generate scalars that
8384 point to the shared string table where possible. Returns a new (undefined)
8385 SV if the hek is NULL.
8391 Perl_newSVhek(pTHX_ const HEK *const hek)
8401 if (HEK_LEN(hek) == HEf_SVKEY) {
8402 return newSVsv(*(SV**)HEK_KEY(hek));
8404 const int flags = HEK_FLAGS(hek);
8405 if (flags & HVhek_WASUTF8) {
8407 Andreas would like keys he put in as utf8 to come back as utf8
8409 STRLEN utf8_len = HEK_LEN(hek);
8410 SV * const sv = newSV_type(SVt_PV);
8411 char *as_utf8 = (char *)bytes_to_utf8 ((U8*)HEK_KEY(hek), &utf8_len);
8412 /* bytes_to_utf8() allocates a new string, which we can repurpose: */
8413 sv_usepvn_flags(sv, as_utf8, utf8_len, SV_HAS_TRAILING_NUL);
8416 } else if (flags & (HVhek_REHASH|HVhek_UNSHARED)) {
8417 /* We don't have a pointer to the hv, so we have to replicate the
8418 flag into every HEK. This hv is using custom a hasing
8419 algorithm. Hence we can't return a shared string scalar, as
8420 that would contain the (wrong) hash value, and might get passed
8421 into an hv routine with a regular hash.
8422 Similarly, a hash that isn't using shared hash keys has to have
8423 the flag in every key so that we know not to try to call
8424 share_hek_hek on it. */
8426 SV * const sv = newSVpvn (HEK_KEY(hek), HEK_LEN(hek));
8431 /* This will be overwhelminly the most common case. */
8433 /* Inline most of newSVpvn_share(), because share_hek_hek() is far
8434 more efficient than sharepvn(). */
8438 sv_upgrade(sv, SVt_PV);
8439 SvPV_set(sv, (char *)HEK_KEY(share_hek_hek(hek)));
8440 SvCUR_set(sv, HEK_LEN(hek));
8453 =for apidoc newSVpvn_share
8455 Creates a new SV with its SvPVX_const pointing to a shared string in the string
8456 table. If the string does not already exist in the table, it is created
8457 first. Turns on READONLY and FAKE. If the C<hash> parameter is non-zero, that
8458 value is used; otherwise the hash is computed. The string's hash can be later
8459 be retrieved from the SV with the C<SvSHARED_HASH()> macro. The idea here is
8460 that as the string table is used for shared hash keys these strings will have
8461 SvPVX_const == HeKEY and hash lookup will avoid string compare.
8467 Perl_newSVpvn_share(pTHX_ const char *src, I32 len, U32 hash)
8471 bool is_utf8 = FALSE;
8472 const char *const orig_src = src;
8475 STRLEN tmplen = -len;
8477 /* See the note in hv.c:hv_fetch() --jhi */
8478 src = (char*)bytes_from_utf8((const U8*)src, &tmplen, &is_utf8);
8482 PERL_HASH(hash, src, len);
8484 /* The logic for this is inlined in S_mro_get_linear_isa_dfs(), so if it
8485 changes here, update it there too. */
8486 sv_upgrade(sv, SVt_PV);
8487 SvPV_set(sv, sharepvn(src, is_utf8?-len:len, hash));
8495 if (src != orig_src)
8501 =for apidoc newSVpv_share
8503 Like C<newSVpvn_share>, but takes a nul-terminated string instead of a
8510 Perl_newSVpv_share(pTHX_ const char *src, U32 hash)
8512 return newSVpvn_share(src, strlen(src), hash);
8515 #if defined(PERL_IMPLICIT_CONTEXT)
8517 /* pTHX_ magic can't cope with varargs, so this is a no-context
8518 * version of the main function, (which may itself be aliased to us).
8519 * Don't access this version directly.
8523 Perl_newSVpvf_nocontext(const char *const pat, ...)
8529 PERL_ARGS_ASSERT_NEWSVPVF_NOCONTEXT;
8531 va_start(args, pat);
8532 sv = vnewSVpvf(pat, &args);
8539 =for apidoc newSVpvf
8541 Creates a new SV and initializes it with the string formatted like
8548 Perl_newSVpvf(pTHX_ const char *const pat, ...)
8553 PERL_ARGS_ASSERT_NEWSVPVF;
8555 va_start(args, pat);
8556 sv = vnewSVpvf(pat, &args);
8561 /* backend for newSVpvf() and newSVpvf_nocontext() */
8564 Perl_vnewSVpvf(pTHX_ const char *const pat, va_list *const args)
8569 PERL_ARGS_ASSERT_VNEWSVPVF;
8572 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
8579 Creates a new SV and copies a floating point value into it.
8580 The reference count for the SV is set to 1.
8586 Perl_newSVnv(pTHX_ const NV n)
8599 Creates a new SV and copies an integer into it. The reference count for the
8606 Perl_newSViv(pTHX_ const IV i)
8619 Creates a new SV and copies an unsigned integer into it.
8620 The reference count for the SV is set to 1.
8626 Perl_newSVuv(pTHX_ const UV u)
8637 =for apidoc newSV_type
8639 Creates a new SV, of the type specified. The reference count for the new SV
8646 Perl_newSV_type(pTHX_ const svtype type)
8651 sv_upgrade(sv, type);
8656 =for apidoc newRV_noinc
8658 Creates an RV wrapper for an SV. The reference count for the original
8659 SV is B<not> incremented.
8665 Perl_newRV_noinc(pTHX_ SV *const tmpRef)
8668 register SV *sv = newSV_type(SVt_IV);
8670 PERL_ARGS_ASSERT_NEWRV_NOINC;
8673 SvRV_set(sv, tmpRef);
8678 /* newRV_inc is the official function name to use now.
8679 * newRV_inc is in fact #defined to newRV in sv.h
8683 Perl_newRV(pTHX_ SV *const sv)
8687 PERL_ARGS_ASSERT_NEWRV;
8689 return newRV_noinc(SvREFCNT_inc_simple_NN(sv));
8695 Creates a new SV which is an exact duplicate of the original SV.
8702 Perl_newSVsv(pTHX_ register SV *const old)
8709 if (SvTYPE(old) == (svtype)SVTYPEMASK) {
8710 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL), "semi-panic: attempt to dup freed string");
8714 /* SV_GMAGIC is the default for sv_setv()
8715 SV_NOSTEAL prevents TEMP buffers being, well, stolen, and saves games
8716 with SvTEMP_off and SvTEMP_on round a call to sv_setsv. */
8717 sv_setsv_flags(sv, old, SV_GMAGIC | SV_NOSTEAL);
8722 =for apidoc sv_reset
8724 Underlying implementation for the C<reset> Perl function.
8725 Note that the perl-level function is vaguely deprecated.
8731 Perl_sv_reset(pTHX_ register const char *s, HV *const stash)
8734 char todo[PERL_UCHAR_MAX+1];
8736 PERL_ARGS_ASSERT_SV_RESET;
8741 if (!*s) { /* reset ?? searches */
8742 MAGIC * const mg = mg_find((const SV *)stash, PERL_MAGIC_symtab);
8744 const U32 count = mg->mg_len / sizeof(PMOP**);
8745 PMOP **pmp = (PMOP**) mg->mg_ptr;
8746 PMOP *const *const end = pmp + count;
8750 SvREADONLY_off(PL_regex_pad[(*pmp)->op_pmoffset]);
8752 (*pmp)->op_pmflags &= ~PMf_USED;
8760 /* reset variables */
8762 if (!HvARRAY(stash))
8765 Zero(todo, 256, char);
8768 I32 i = (unsigned char)*s;
8772 max = (unsigned char)*s++;
8773 for ( ; i <= max; i++) {
8776 for (i = 0; i <= (I32) HvMAX(stash); i++) {
8778 for (entry = HvARRAY(stash)[i];
8780 entry = HeNEXT(entry))
8785 if (!todo[(U8)*HeKEY(entry)])
8787 gv = MUTABLE_GV(HeVAL(entry));
8790 if (SvTHINKFIRST(sv)) {
8791 if (!SvREADONLY(sv) && SvROK(sv))
8793 /* XXX Is this continue a bug? Why should THINKFIRST
8794 exempt us from resetting arrays and hashes? */
8798 if (SvTYPE(sv) >= SVt_PV) {
8800 if (SvPVX_const(sv) != NULL)
8808 if (GvHV(gv) && !HvNAME_get(GvHV(gv))) {
8810 Perl_die(aTHX_ "Can't reset %%ENV on this system");
8813 # if defined(USE_ENVIRON_ARRAY)
8816 # endif /* USE_ENVIRON_ARRAY */
8827 Using various gambits, try to get an IO from an SV: the IO slot if its a
8828 GV; or the recursive result if we're an RV; or the IO slot of the symbol
8829 named after the PV if we're a string.
8835 Perl_sv_2io(pTHX_ SV *const sv)
8840 PERL_ARGS_ASSERT_SV_2IO;
8842 switch (SvTYPE(sv)) {
8844 io = MUTABLE_IO(sv);
8848 if (isGV_with_GP(sv)) {
8849 gv = MUTABLE_GV(sv);
8852 Perl_croak(aTHX_ "Bad filehandle: %"HEKf,
8853 HEKfARG(GvNAME_HEK(gv)));
8859 Perl_croak(aTHX_ PL_no_usym, "filehandle");
8861 return sv_2io(SvRV(sv));
8862 gv = gv_fetchsv(sv, 0, SVt_PVIO);
8868 Perl_croak(aTHX_ "Bad filehandle: %"SVf, SVfARG(sv));
8877 Using various gambits, try to get a CV from an SV; in addition, try if
8878 possible to set C<*st> and C<*gvp> to the stash and GV associated with it.
8879 The flags in C<lref> are passed to gv_fetchsv.
8885 Perl_sv_2cv(pTHX_ SV *sv, HV **const st, GV **const gvp, const I32 lref)
8891 PERL_ARGS_ASSERT_SV_2CV;
8898 switch (SvTYPE(sv)) {
8902 return MUTABLE_CV(sv);
8912 sv = amagic_deref_call(sv, to_cv_amg);
8913 /* At this point I'd like to do SPAGAIN, but really I need to
8914 force it upon my callers. Hmmm. This is a mess... */
8917 if (SvTYPE(sv) == SVt_PVCV) {
8918 cv = MUTABLE_CV(sv);
8923 else if(isGV_with_GP(sv))
8924 gv = MUTABLE_GV(sv);
8926 Perl_croak(aTHX_ "Not a subroutine reference");
8928 else if (isGV_with_GP(sv)) {
8929 gv = MUTABLE_GV(sv);
8932 gv = gv_fetchsv_nomg(sv, lref, SVt_PVCV);
8939 /* Some flags to gv_fetchsv mean don't really create the GV */
8940 if (!isGV_with_GP(gv)) {
8945 if (lref & ~GV_ADDMG && !GvCVu(gv)) {
8949 gv_efullname3(tmpsv, gv, NULL);
8950 /* XXX this is probably not what they think they're getting.
8951 * It has the same effect as "sub name;", i.e. just a forward
8953 newSUB(start_subparse(FALSE, 0),
8954 newSVOP(OP_CONST, 0, tmpsv),
8958 Perl_croak(aTHX_ "Unable to create sub named \"%"SVf"\"",
8959 SVfARG(SvOK(sv) ? sv : &PL_sv_no));
8968 Returns true if the SV has a true value by Perl's rules.
8969 Use the C<SvTRUE> macro instead, which may call C<sv_true()> or may
8970 instead use an in-line version.
8976 Perl_sv_true(pTHX_ register SV *const sv)
8981 register const XPV* const tXpv = (XPV*)SvANY(sv);
8983 (tXpv->xpv_cur > 1 ||
8984 (tXpv->xpv_cur && *sv->sv_u.svu_pv != '0')))
8991 return SvIVX(sv) != 0;
8994 return SvNVX(sv) != 0.0;
8996 return sv_2bool(sv);
9002 =for apidoc sv_pvn_force
9004 Get a sensible string out of the SV somehow.
9005 A private implementation of the C<SvPV_force> macro for compilers which
9006 can't cope with complex macro expressions. Always use the macro instead.
9008 =for apidoc sv_pvn_force_flags
9010 Get a sensible string out of the SV somehow.
9011 If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<sv> if
9012 appropriate, else not. C<sv_pvn_force> and C<sv_pvn_force_nomg> are
9013 implemented in terms of this function.
9014 You normally want to use the various wrapper macros instead: see
9015 C<SvPV_force> and C<SvPV_force_nomg>
9021 Perl_sv_pvn_force_flags(pTHX_ SV *const sv, STRLEN *const lp, const I32 flags)
9025 PERL_ARGS_ASSERT_SV_PVN_FORCE_FLAGS;
9027 if (SvTHINKFIRST(sv) && !SvROK(sv))
9028 sv_force_normal_flags(sv, 0);
9038 if (SvREADONLY(sv) && !(flags & SV_MUTABLE_RETURN)) {
9039 const char * const ref = sv_reftype(sv,0);
9041 Perl_croak(aTHX_ "Can't coerce readonly %s to string in %s",
9042 ref, OP_DESC(PL_op));
9044 Perl_croak(aTHX_ "Can't coerce readonly %s to string", ref);
9046 if ((SvTYPE(sv) > SVt_PVLV && SvTYPE(sv) != SVt_PVFM)
9047 || isGV_with_GP(sv))
9048 /* diag_listed_as: Can't coerce %s to %s in %s */
9049 Perl_croak(aTHX_ "Can't coerce %s to string in %s", sv_reftype(sv,0),
9051 s = sv_2pv_flags(sv, &len, flags);
9055 if (s != SvPVX_const(sv)) { /* Almost, but not quite, sv_setpvn() */
9058 SvUPGRADE(sv, SVt_PV); /* Never FALSE */
9059 SvGROW(sv, len + 1);
9060 Move(s,SvPVX(sv),len,char);
9062 SvPVX(sv)[len] = '\0';
9065 SvPOK_on(sv); /* validate pointer */
9067 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
9068 PTR2UV(sv),SvPVX_const(sv)));
9071 return SvPVX_mutable(sv);
9075 =for apidoc sv_pvbyten_force
9077 The backend for the C<SvPVbytex_force> macro. Always use the macro instead.
9083 Perl_sv_pvbyten_force(pTHX_ SV *const sv, STRLEN *const lp)
9085 PERL_ARGS_ASSERT_SV_PVBYTEN_FORCE;
9087 sv_pvn_force(sv,lp);
9088 sv_utf8_downgrade(sv,0);
9094 =for apidoc sv_pvutf8n_force
9096 The backend for the C<SvPVutf8x_force> macro. Always use the macro instead.
9102 Perl_sv_pvutf8n_force(pTHX_ SV *const sv, STRLEN *const lp)
9104 PERL_ARGS_ASSERT_SV_PVUTF8N_FORCE;
9106 sv_pvn_force(sv,lp);
9107 sv_utf8_upgrade(sv);
9113 =for apidoc sv_reftype
9115 Returns a string describing what the SV is a reference to.
9121 Perl_sv_reftype(pTHX_ const SV *const sv, const int ob)
9123 PERL_ARGS_ASSERT_SV_REFTYPE;
9124 if (ob && SvOBJECT(sv)) {
9125 return SvPV_nolen_const(sv_ref(NULL, sv, ob));
9128 switch (SvTYPE(sv)) {
9143 case SVt_PVLV: return (char *) (SvROK(sv) ? "REF"
9144 /* tied lvalues should appear to be
9145 * scalars for backwards compatibility */
9146 : (LvTYPE(sv) == 't' || LvTYPE(sv) == 'T')
9147 ? "SCALAR" : "LVALUE");
9148 case SVt_PVAV: return "ARRAY";
9149 case SVt_PVHV: return "HASH";
9150 case SVt_PVCV: return "CODE";
9151 case SVt_PVGV: return (char *) (isGV_with_GP(sv)
9152 ? "GLOB" : "SCALAR");
9153 case SVt_PVFM: return "FORMAT";
9154 case SVt_PVIO: return "IO";
9155 case SVt_BIND: return "BIND";
9156 case SVt_REGEXP: return "REGEXP";
9157 default: return "UNKNOWN";
9165 Returns a SV describing what the SV passed in is a reference to.
9171 Perl_sv_ref(pTHX_ register SV *dst, const SV *const sv, const int ob)
9173 PERL_ARGS_ASSERT_SV_REF;
9176 dst = sv_newmortal();
9178 if (ob && SvOBJECT(sv)) {
9179 HvNAME_get(SvSTASH(sv))
9180 ? sv_sethek(dst, HvNAME_HEK(SvSTASH(sv)))
9181 : sv_setpvn(dst, "__ANON__", 8);
9184 const char * reftype = sv_reftype(sv, 0);
9185 sv_setpv(dst, reftype);
9191 =for apidoc sv_isobject
9193 Returns a boolean indicating whether the SV is an RV pointing to a blessed
9194 object. If the SV is not an RV, or if the object is not blessed, then this
9201 Perl_sv_isobject(pTHX_ SV *sv)
9217 Returns a boolean indicating whether the SV is blessed into the specified
9218 class. This does not check for subtypes; use C<sv_derived_from> to verify
9219 an inheritance relationship.
9225 Perl_sv_isa(pTHX_ SV *sv, const char *const name)
9229 PERL_ARGS_ASSERT_SV_ISA;
9239 hvname = HvNAME_get(SvSTASH(sv));
9243 return strEQ(hvname, name);
9249 Creates a new SV for the RV, C<rv>, to point to. If C<rv> is not an RV then
9250 it will be upgraded to one. If C<classname> is non-null then the new SV will
9251 be blessed in the specified package. The new SV is returned and its
9252 reference count is 1.
9258 Perl_newSVrv(pTHX_ SV *const rv, const char *const classname)
9263 PERL_ARGS_ASSERT_NEWSVRV;
9267 SV_CHECK_THINKFIRST_COW_DROP(rv);
9268 (void)SvAMAGIC_off(rv);
9270 if (SvTYPE(rv) >= SVt_PVMG) {
9271 const U32 refcnt = SvREFCNT(rv);
9275 SvREFCNT(rv) = refcnt;
9277 sv_upgrade(rv, SVt_IV);
9278 } else if (SvROK(rv)) {
9279 SvREFCNT_dec(SvRV(rv));
9281 prepare_SV_for_RV(rv);
9289 HV* const stash = gv_stashpv(classname, GV_ADD);
9290 (void)sv_bless(rv, stash);
9296 =for apidoc sv_setref_pv
9298 Copies a pointer into a new SV, optionally blessing the SV. The C<rv>
9299 argument will be upgraded to an RV. That RV will be modified to point to
9300 the new SV. If the C<pv> argument is NULL then C<PL_sv_undef> will be placed
9301 into the SV. The C<classname> argument indicates the package for the
9302 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9303 will have a reference count of 1, and the RV will be returned.
9305 Do not use with other Perl types such as HV, AV, SV, CV, because those
9306 objects will become corrupted by the pointer copy process.
9308 Note that C<sv_setref_pvn> copies the string while this copies the pointer.
9314 Perl_sv_setref_pv(pTHX_ SV *const rv, const char *const classname, void *const pv)
9318 PERL_ARGS_ASSERT_SV_SETREF_PV;
9321 sv_setsv(rv, &PL_sv_undef);
9325 sv_setiv(newSVrv(rv,classname), PTR2IV(pv));
9330 =for apidoc sv_setref_iv
9332 Copies an integer into a new SV, optionally blessing the SV. The C<rv>
9333 argument will be upgraded to an RV. That RV will be modified to point to
9334 the new SV. The C<classname> argument indicates the package for the
9335 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9336 will have a reference count of 1, and the RV will be returned.
9342 Perl_sv_setref_iv(pTHX_ SV *const rv, const char *const classname, const IV iv)
9344 PERL_ARGS_ASSERT_SV_SETREF_IV;
9346 sv_setiv(newSVrv(rv,classname), iv);
9351 =for apidoc sv_setref_uv
9353 Copies an unsigned integer into a new SV, optionally blessing the SV. The C<rv>
9354 argument will be upgraded to an RV. That RV will be modified to point to
9355 the new SV. The C<classname> argument indicates the package for the
9356 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9357 will have a reference count of 1, and the RV will be returned.
9363 Perl_sv_setref_uv(pTHX_ SV *const rv, const char *const classname, const UV uv)
9365 PERL_ARGS_ASSERT_SV_SETREF_UV;
9367 sv_setuv(newSVrv(rv,classname), uv);
9372 =for apidoc sv_setref_nv
9374 Copies a double into a new SV, optionally blessing the SV. The C<rv>
9375 argument will be upgraded to an RV. That RV will be modified to point to
9376 the new SV. The C<classname> argument indicates the package for the
9377 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9378 will have a reference count of 1, and the RV will be returned.
9384 Perl_sv_setref_nv(pTHX_ SV *const rv, const char *const classname, const NV nv)
9386 PERL_ARGS_ASSERT_SV_SETREF_NV;
9388 sv_setnv(newSVrv(rv,classname), nv);
9393 =for apidoc sv_setref_pvn
9395 Copies a string into a new SV, optionally blessing the SV. The length of the
9396 string must be specified with C<n>. The C<rv> argument will be upgraded to
9397 an RV. That RV will be modified to point to the new SV. The C<classname>
9398 argument indicates the package for the blessing. Set C<classname> to
9399 C<NULL> to avoid the blessing. The new SV will have a reference count
9400 of 1, and the RV will be returned.
9402 Note that C<sv_setref_pv> copies the pointer while this copies the string.
9408 Perl_sv_setref_pvn(pTHX_ SV *const rv, const char *const classname,
9409 const char *const pv, const STRLEN n)
9411 PERL_ARGS_ASSERT_SV_SETREF_PVN;
9413 sv_setpvn(newSVrv(rv,classname), pv, n);
9418 =for apidoc sv_bless
9420 Blesses an SV into a specified package. The SV must be an RV. The package
9421 must be designated by its stash (see C<gv_stashpv()>). The reference count
9422 of the SV is unaffected.
9428 Perl_sv_bless(pTHX_ SV *const sv, HV *const stash)
9433 PERL_ARGS_ASSERT_SV_BLESS;
9436 Perl_croak(aTHX_ "Can't bless non-reference value");
9438 if (SvFLAGS(tmpRef) & (SVs_OBJECT|SVf_READONLY)) {
9439 if (SvIsCOW(tmpRef))
9440 sv_force_normal_flags(tmpRef, 0);
9441 if (SvREADONLY(tmpRef))
9442 Perl_croak_no_modify(aTHX);
9443 if (SvOBJECT(tmpRef)) {
9444 if (SvTYPE(tmpRef) != SVt_PVIO)
9446 SvREFCNT_dec(SvSTASH(tmpRef));
9449 SvOBJECT_on(tmpRef);
9450 if (SvTYPE(tmpRef) != SVt_PVIO)
9452 SvUPGRADE(tmpRef, SVt_PVMG);
9453 SvSTASH_set(tmpRef, MUTABLE_HV(SvREFCNT_inc_simple(stash)));
9458 (void)SvAMAGIC_off(sv);
9460 if(SvSMAGICAL(tmpRef))
9461 if(mg_find(tmpRef, PERL_MAGIC_ext) || mg_find(tmpRef, PERL_MAGIC_uvar))
9469 /* Downgrades a PVGV to a PVMG. If it’s actually a PVLV, we leave the type
9470 * as it is after unglobbing it.
9474 S_sv_unglob(pTHX_ SV *const sv)
9479 SV * const temp = sv_newmortal();
9481 PERL_ARGS_ASSERT_SV_UNGLOB;
9483 assert(SvTYPE(sv) == SVt_PVGV || SvTYPE(sv) == SVt_PVLV);
9485 gv_efullname3(temp, MUTABLE_GV(sv), "*");
9488 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
9489 && HvNAME_get(stash))
9490 mro_method_changed_in(stash);
9491 gp_free(MUTABLE_GV(sv));
9494 sv_del_backref(MUTABLE_SV(GvSTASH(sv)), sv);
9498 if (GvNAME_HEK(sv)) {
9499 unshare_hek(GvNAME_HEK(sv));
9501 isGV_with_GP_off(sv);
9503 if(SvTYPE(sv) == SVt_PVGV) {
9504 /* need to keep SvANY(sv) in the right arena */
9505 xpvmg = new_XPVMG();
9506 StructCopy(SvANY(sv), xpvmg, XPVMG);
9507 del_XPVGV(SvANY(sv));
9510 SvFLAGS(sv) &= ~SVTYPEMASK;
9511 SvFLAGS(sv) |= SVt_PVMG;
9514 /* Intentionally not calling any local SET magic, as this isn't so much a
9515 set operation as merely an internal storage change. */
9516 sv_setsv_flags(sv, temp, 0);
9520 =for apidoc sv_unref_flags
9522 Unsets the RV status of the SV, and decrements the reference count of
9523 whatever was being referenced by the RV. This can almost be thought of
9524 as a reversal of C<newSVrv>. The C<cflags> argument can contain
9525 C<SV_IMMEDIATE_UNREF> to force the reference count to be decremented
9526 (otherwise the decrementing is conditional on the reference count being
9527 different from one or the reference being a readonly SV).
9534 Perl_sv_unref_flags(pTHX_ SV *const ref, const U32 flags)
9536 SV* const target = SvRV(ref);
9538 PERL_ARGS_ASSERT_SV_UNREF_FLAGS;
9540 if (SvWEAKREF(ref)) {
9541 sv_del_backref(target, ref);
9543 SvRV_set(ref, NULL);
9546 SvRV_set(ref, NULL);
9548 /* You can't have a || SvREADONLY(target) here, as $a = $$a, where $a was
9549 assigned to as BEGIN {$a = \"Foo"} will fail. */
9550 if (SvREFCNT(target) != 1 || (flags & SV_IMMEDIATE_UNREF))
9551 SvREFCNT_dec(target);
9552 else /* XXX Hack, but hard to make $a=$a->[1] work otherwise */
9553 sv_2mortal(target); /* Schedule for freeing later */
9557 =for apidoc sv_untaint
9559 Untaint an SV. Use C<SvTAINTED_off> instead.
9565 Perl_sv_untaint(pTHX_ SV *const sv)
9567 PERL_ARGS_ASSERT_SV_UNTAINT;
9569 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
9570 MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
9577 =for apidoc sv_tainted
9579 Test an SV for taintedness. Use C<SvTAINTED> instead.
9585 Perl_sv_tainted(pTHX_ SV *const sv)
9587 PERL_ARGS_ASSERT_SV_TAINTED;
9589 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
9590 const MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
9591 if (mg && (mg->mg_len & 1) )
9598 =for apidoc sv_setpviv
9600 Copies an integer into the given SV, also updating its string value.
9601 Does not handle 'set' magic. See C<sv_setpviv_mg>.
9607 Perl_sv_setpviv(pTHX_ SV *const sv, const IV iv)
9609 char buf[TYPE_CHARS(UV)];
9611 char * const ptr = uiv_2buf(buf, iv, 0, 0, &ebuf);
9613 PERL_ARGS_ASSERT_SV_SETPVIV;
9615 sv_setpvn(sv, ptr, ebuf - ptr);
9619 =for apidoc sv_setpviv_mg
9621 Like C<sv_setpviv>, but also handles 'set' magic.
9627 Perl_sv_setpviv_mg(pTHX_ SV *const sv, const IV iv)
9629 PERL_ARGS_ASSERT_SV_SETPVIV_MG;
9635 #if defined(PERL_IMPLICIT_CONTEXT)
9637 /* pTHX_ magic can't cope with varargs, so this is a no-context
9638 * version of the main function, (which may itself be aliased to us).
9639 * Don't access this version directly.
9643 Perl_sv_setpvf_nocontext(SV *const sv, const char *const pat, ...)
9648 PERL_ARGS_ASSERT_SV_SETPVF_NOCONTEXT;
9650 va_start(args, pat);
9651 sv_vsetpvf(sv, pat, &args);
9655 /* pTHX_ magic can't cope with varargs, so this is a no-context
9656 * version of the main function, (which may itself be aliased to us).
9657 * Don't access this version directly.
9661 Perl_sv_setpvf_mg_nocontext(SV *const sv, const char *const pat, ...)
9666 PERL_ARGS_ASSERT_SV_SETPVF_MG_NOCONTEXT;
9668 va_start(args, pat);
9669 sv_vsetpvf_mg(sv, pat, &args);
9675 =for apidoc sv_setpvf
9677 Works like C<sv_catpvf> but copies the text into the SV instead of
9678 appending it. Does not handle 'set' magic. See C<sv_setpvf_mg>.
9684 Perl_sv_setpvf(pTHX_ SV *const sv, const char *const pat, ...)
9688 PERL_ARGS_ASSERT_SV_SETPVF;
9690 va_start(args, pat);
9691 sv_vsetpvf(sv, pat, &args);
9696 =for apidoc sv_vsetpvf
9698 Works like C<sv_vcatpvf> but copies the text into the SV instead of
9699 appending it. Does not handle 'set' magic. See C<sv_vsetpvf_mg>.
9701 Usually used via its frontend C<sv_setpvf>.
9707 Perl_sv_vsetpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9709 PERL_ARGS_ASSERT_SV_VSETPVF;
9711 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9715 =for apidoc sv_setpvf_mg
9717 Like C<sv_setpvf>, but also handles 'set' magic.
9723 Perl_sv_setpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
9727 PERL_ARGS_ASSERT_SV_SETPVF_MG;
9729 va_start(args, pat);
9730 sv_vsetpvf_mg(sv, pat, &args);
9735 =for apidoc sv_vsetpvf_mg
9737 Like C<sv_vsetpvf>, but also handles 'set' magic.
9739 Usually used via its frontend C<sv_setpvf_mg>.
9745 Perl_sv_vsetpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9747 PERL_ARGS_ASSERT_SV_VSETPVF_MG;
9749 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9753 #if defined(PERL_IMPLICIT_CONTEXT)
9755 /* pTHX_ magic can't cope with varargs, so this is a no-context
9756 * version of the main function, (which may itself be aliased to us).
9757 * Don't access this version directly.
9761 Perl_sv_catpvf_nocontext(SV *const sv, const char *const pat, ...)
9766 PERL_ARGS_ASSERT_SV_CATPVF_NOCONTEXT;
9768 va_start(args, pat);
9769 sv_vcatpvf(sv, pat, &args);
9773 /* pTHX_ magic can't cope with varargs, so this is a no-context
9774 * version of the main function, (which may itself be aliased to us).
9775 * Don't access this version directly.
9779 Perl_sv_catpvf_mg_nocontext(SV *const sv, const char *const pat, ...)
9784 PERL_ARGS_ASSERT_SV_CATPVF_MG_NOCONTEXT;
9786 va_start(args, pat);
9787 sv_vcatpvf_mg(sv, pat, &args);
9793 =for apidoc sv_catpvf
9795 Processes its arguments like C<sprintf> and appends the formatted
9796 output to an SV. If the appended data contains "wide" characters
9797 (including, but not limited to, SVs with a UTF-8 PV formatted with %s,
9798 and characters >255 formatted with %c), the original SV might get
9799 upgraded to UTF-8. Handles 'get' magic, but not 'set' magic. See
9800 C<sv_catpvf_mg>. If the original SV was UTF-8, the pattern should be
9801 valid UTF-8; if the original SV was bytes, the pattern should be too.
9806 Perl_sv_catpvf(pTHX_ SV *const sv, const char *const pat, ...)
9810 PERL_ARGS_ASSERT_SV_CATPVF;
9812 va_start(args, pat);
9813 sv_vcatpvf(sv, pat, &args);
9818 =for apidoc sv_vcatpvf
9820 Processes its arguments like C<vsprintf> and appends the formatted output
9821 to an SV. Does not handle 'set' magic. See C<sv_vcatpvf_mg>.
9823 Usually used via its frontend C<sv_catpvf>.
9829 Perl_sv_vcatpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9831 PERL_ARGS_ASSERT_SV_VCATPVF;
9833 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9837 =for apidoc sv_catpvf_mg
9839 Like C<sv_catpvf>, but also handles 'set' magic.
9845 Perl_sv_catpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
9849 PERL_ARGS_ASSERT_SV_CATPVF_MG;
9851 va_start(args, pat);
9852 sv_vcatpvf_mg(sv, pat, &args);
9857 =for apidoc sv_vcatpvf_mg
9859 Like C<sv_vcatpvf>, but also handles 'set' magic.
9861 Usually used via its frontend C<sv_catpvf_mg>.
9867 Perl_sv_vcatpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9869 PERL_ARGS_ASSERT_SV_VCATPVF_MG;
9871 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9876 =for apidoc sv_vsetpvfn
9878 Works like C<sv_vcatpvfn> but copies the text into the SV instead of
9881 Usually used via one of its frontends C<sv_vsetpvf> and C<sv_vsetpvf_mg>.
9887 Perl_sv_vsetpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
9888 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
9890 PERL_ARGS_ASSERT_SV_VSETPVFN;
9893 sv_vcatpvfn(sv, pat, patlen, args, svargs, svmax, maybe_tainted);
9898 * Warn of missing argument to sprintf, and then return a defined value
9899 * to avoid inappropriate "use of uninit" warnings [perl #71000].
9901 #define WARN_MISSING WARN_UNINITIALIZED /* Not sure we want a new category */
9903 S_vcatpvfn_missing_argument(pTHX) {
9904 if (ckWARN(WARN_MISSING)) {
9905 Perl_warner(aTHX_ packWARN(WARN_MISSING), "Missing argument in %s",
9906 PL_op ? OP_DESC(PL_op) : "sv_vcatpvfn()");
9913 S_expect_number(pTHX_ char **const pattern)
9918 PERL_ARGS_ASSERT_EXPECT_NUMBER;
9920 switch (**pattern) {
9921 case '1': case '2': case '3':
9922 case '4': case '5': case '6':
9923 case '7': case '8': case '9':
9924 var = *(*pattern)++ - '0';
9925 while (isDIGIT(**pattern)) {
9926 const I32 tmp = var * 10 + (*(*pattern)++ - '0');
9928 Perl_croak(aTHX_ "Integer overflow in format string for %s", (PL_op ? OP_DESC(PL_op) : "sv_vcatpvfn"));
9936 S_F0convert(NV nv, char *const endbuf, STRLEN *const len)
9938 const int neg = nv < 0;
9941 PERL_ARGS_ASSERT_F0CONVERT;
9949 if (uv & 1 && uv == nv)
9950 uv--; /* Round to even */
9952 const unsigned dig = uv % 10;
9965 =for apidoc sv_vcatpvfn
9967 Processes its arguments like C<vsprintf> and appends the formatted output
9968 to an SV. Uses an array of SVs if the C style variable argument list is
9969 missing (NULL). When running with taint checks enabled, indicates via
9970 C<maybe_tainted> if results are untrustworthy (often due to the use of
9973 Usually used via one of its frontends C<sv_vcatpvf> and C<sv_vcatpvf_mg>.
9979 #define VECTORIZE_ARGS vecsv = va_arg(*args, SV*);\
9980 vecstr = (U8*)SvPV_const(vecsv,veclen);\
9981 vec_utf8 = DO_UTF8(vecsv);
9983 /* XXX maybe_tainted is never assigned to, so the doc above is lying. */
9986 Perl_sv_vcatpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
9987 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
9995 static const char nullstr[] = "(null)";
9997 bool has_utf8 = DO_UTF8(sv); /* has the result utf8? */
9998 const bool pat_utf8 = has_utf8; /* the pattern is in utf8? */
10000 /* Times 4: a decimal digit takes more than 3 binary digits.
10001 * NV_DIG: mantissa takes than many decimal digits.
10002 * Plus 32: Playing safe. */
10003 char ebuf[IV_DIG * 4 + NV_DIG + 32];
10004 /* large enough for "%#.#f" --chip */
10005 /* what about long double NVs? --jhi */
10007 PERL_ARGS_ASSERT_SV_VCATPVFN;
10008 PERL_UNUSED_ARG(maybe_tainted);
10010 /* no matter what, this is a string now */
10011 (void)SvPV_force(sv, origlen);
10013 /* special-case "", "%s", and "%-p" (SVf - see below) */
10016 if (patlen == 2 && pat[0] == '%' && pat[1] == 's') {
10018 const char * const s = va_arg(*args, char*);
10019 sv_catpv(sv, s ? s : nullstr);
10021 else if (svix < svmax) {
10022 sv_catsv(sv, *svargs);
10025 S_vcatpvfn_missing_argument(aTHX);
10028 if (args && patlen == 3 && pat[0] == '%' &&
10029 pat[1] == '-' && pat[2] == 'p') {
10030 argsv = MUTABLE_SV(va_arg(*args, void*));
10031 sv_catsv(sv, argsv);
10035 #ifndef USE_LONG_DOUBLE
10036 /* special-case "%.<number>[gf]" */
10037 if ( !args && patlen <= 5 && pat[0] == '%' && pat[1] == '.'
10038 && (pat[patlen-1] == 'g' || pat[patlen-1] == 'f') ) {
10039 unsigned digits = 0;
10043 while (*pp >= '0' && *pp <= '9')
10044 digits = 10 * digits + (*pp++ - '0');
10045 if (pp - pat == (int)patlen - 1 && svix < svmax) {
10046 const NV nv = SvNV(*svargs);
10048 /* Add check for digits != 0 because it seems that some
10049 gconverts are buggy in this case, and we don't yet have
10050 a Configure test for this. */
10051 if (digits && digits < sizeof(ebuf) - NV_DIG - 10) {
10052 /* 0, point, slack */
10053 Gconvert(nv, (int)digits, 0, ebuf);
10054 sv_catpv(sv, ebuf);
10055 if (*ebuf) /* May return an empty string for digits==0 */
10058 } else if (!digits) {
10061 if ((p = F0convert(nv, ebuf + sizeof ebuf, &l))) {
10062 sv_catpvn(sv, p, l);
10068 #endif /* !USE_LONG_DOUBLE */
10070 if (!args && svix < svmax && DO_UTF8(*svargs))
10073 patend = (char*)pat + patlen;
10074 for (p = (char*)pat; p < patend; p = q) {
10077 bool vectorize = FALSE;
10078 bool vectorarg = FALSE;
10079 bool vec_utf8 = FALSE;
10085 bool has_precis = FALSE;
10087 const I32 osvix = svix;
10088 bool is_utf8 = FALSE; /* is this item utf8? */
10089 #ifdef HAS_LDBL_SPRINTF_BUG
10090 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
10091 with sfio - Allen <allens@cpan.org> */
10092 bool fix_ldbl_sprintf_bug = FALSE;
10096 U8 utf8buf[UTF8_MAXBYTES+1];
10097 STRLEN esignlen = 0;
10099 const char *eptr = NULL;
10100 const char *fmtstart;
10103 const U8 *vecstr = NULL;
10110 /* we need a long double target in case HAS_LONG_DOUBLE but
10111 not USE_LONG_DOUBLE
10113 #if defined(HAS_LONG_DOUBLE) && LONG_DOUBLESIZE > DOUBLESIZE
10121 const char *dotstr = ".";
10122 STRLEN dotstrlen = 1;
10123 I32 efix = 0; /* explicit format parameter index */
10124 I32 ewix = 0; /* explicit width index */
10125 I32 epix = 0; /* explicit precision index */
10126 I32 evix = 0; /* explicit vector index */
10127 bool asterisk = FALSE;
10129 /* echo everything up to the next format specification */
10130 for (q = p; q < patend && *q != '%'; ++q) ;
10132 if (has_utf8 && !pat_utf8)
10133 sv_catpvn_utf8_upgrade(sv, p, q - p, nsv);
10135 sv_catpvn(sv, p, q - p);
10144 We allow format specification elements in this order:
10145 \d+\$ explicit format parameter index
10147 v|\*(\d+\$)?v vector with optional (optionally specified) arg
10148 0 flag (as above): repeated to allow "v02"
10149 \d+|\*(\d+\$)? width using optional (optionally specified) arg
10150 \.(\d*|\*(\d+\$)?) precision using optional (optionally specified) arg
10152 [%bcdefginopsuxDFOUX] format (mandatory)
10157 As of perl5.9.3, printf format checking is on by default.
10158 Internally, perl uses %p formats to provide an escape to
10159 some extended formatting. This block deals with those
10160 extensions: if it does not match, (char*)q is reset and
10161 the normal format processing code is used.
10163 Currently defined extensions are:
10164 %p include pointer address (standard)
10165 %-p (SVf) include an SV (previously %_)
10166 %-<num>p include an SV with precision <num>
10168 %3p include a HEK with precision of 256
10169 %<num>p (where num != 2 or 3) reserved for future
10172 Robin Barker 2005-07-14 (but modified since)
10174 %1p (VDf) removed. RMB 2007-10-19
10181 n = expect_number(&q);
10183 if (sv) { /* SVf */
10188 argsv = MUTABLE_SV(va_arg(*args, void*));
10189 eptr = SvPV_const(argsv, elen);
10190 if (DO_UTF8(argsv))
10194 else if (n==2 || n==3) { /* HEKf */
10195 HEK * const hek = va_arg(*args, HEK *);
10196 eptr = HEK_KEY(hek);
10197 elen = HEK_LEN(hek);
10198 if (HEK_UTF8(hek)) is_utf8 = TRUE;
10199 if (n==3) precis = 256, has_precis = TRUE;
10203 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL),
10204 "internal %%<num>p might conflict with future printf extensions");
10210 if ( (width = expect_number(&q)) ) {
10225 if (plus == '+' && *q == ' ') /* '+' over ' ' */
10254 if ( (ewix = expect_number(&q)) )
10263 if ((vectorarg = asterisk)) {
10276 width = expect_number(&q);
10279 if (vectorize && vectorarg) {
10280 /* vectorizing, but not with the default "." */
10282 vecsv = va_arg(*args, SV*);
10284 vecsv = (evix > 0 && evix <= svmax)
10285 ? svargs[evix-1] : S_vcatpvfn_missing_argument(aTHX);
10287 vecsv = svix < svmax
10288 ? svargs[svix++] : S_vcatpvfn_missing_argument(aTHX);
10290 dotstr = SvPV_const(vecsv, dotstrlen);
10291 /* Keep the DO_UTF8 test *after* the SvPV call, else things go
10292 bad with tied or overloaded values that return UTF8. */
10293 if (DO_UTF8(vecsv))
10295 else if (has_utf8) {
10296 vecsv = sv_mortalcopy(vecsv);
10297 sv_utf8_upgrade(vecsv);
10298 dotstr = SvPV_const(vecsv, dotstrlen);
10305 i = va_arg(*args, int);
10307 i = (ewix ? ewix <= svmax : svix < svmax) ?
10308 SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
10310 width = (i < 0) ? -i : i;
10320 if ( ((epix = expect_number(&q))) && (*q++ != '$') )
10322 /* XXX: todo, support specified precision parameter */
10326 i = va_arg(*args, int);
10328 i = (ewix ? ewix <= svmax : svix < svmax)
10329 ? SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
10331 has_precis = !(i < 0);
10335 while (isDIGIT(*q))
10336 precis = precis * 10 + (*q++ - '0');
10345 else if (efix ? (efix > 0 && efix <= svmax) : svix < svmax) {
10346 vecsv = svargs[efix ? efix-1 : svix++];
10347 vecstr = (U8*)SvPV_const(vecsv,veclen);
10348 vec_utf8 = DO_UTF8(vecsv);
10350 /* if this is a version object, we need to convert
10351 * back into v-string notation and then let the
10352 * vectorize happen normally
10354 if (sv_derived_from(vecsv, "version")) {
10355 char *version = savesvpv(vecsv);
10356 if ( hv_exists(MUTABLE_HV(SvRV(vecsv)), "alpha", 5 ) ) {
10357 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
10358 "vector argument not supported with alpha versions");
10361 vecsv = sv_newmortal();
10362 scan_vstring(version, version + veclen, vecsv);
10363 vecstr = (U8*)SvPV_const(vecsv, veclen);
10364 vec_utf8 = DO_UTF8(vecsv);
10378 case 'I': /* Ix, I32x, and I64x */
10380 if (q[1] == '6' && q[2] == '4') {
10386 if (q[1] == '3' && q[2] == '2') {
10396 #if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
10408 #if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
10409 if (*q == 'l') { /* lld, llf */
10418 if (*++q == 'h') { /* hhd, hhu */
10447 if (!vectorize && !args) {
10449 const I32 i = efix-1;
10450 argsv = (i >= 0 && i < svmax)
10451 ? svargs[i] : S_vcatpvfn_missing_argument(aTHX);
10453 argsv = (svix >= 0 && svix < svmax)
10454 ? svargs[svix++] : S_vcatpvfn_missing_argument(aTHX);
10458 switch (c = *q++) {
10465 uv = (args) ? va_arg(*args, int) : SvIV(argsv);
10467 (!UNI_IS_INVARIANT(uv) && SvUTF8(sv)))
10469 eptr = (char*)utf8buf;
10470 elen = uvchr_to_utf8((U8*)eptr, uv) - utf8buf;
10484 eptr = va_arg(*args, char*);
10486 elen = strlen(eptr);
10488 eptr = (char *)nullstr;
10489 elen = sizeof nullstr - 1;
10493 eptr = SvPV_const(argsv, elen);
10494 if (DO_UTF8(argsv)) {
10495 STRLEN old_precis = precis;
10496 if (has_precis && precis < elen) {
10497 STRLEN ulen = sv_len_utf8(argsv);
10498 I32 p = precis > ulen ? ulen : precis;
10499 sv_pos_u2b(argsv, &p, 0); /* sticks at end */
10502 if (width) { /* fudge width (can't fudge elen) */
10503 if (has_precis && precis < elen)
10504 width += precis - old_precis;
10506 width += elen - sv_len_utf8(argsv);
10513 if (has_precis && precis < elen)
10520 if (alt || vectorize)
10522 uv = PTR2UV(args ? va_arg(*args, void*) : argsv);
10543 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
10552 esignbuf[esignlen++] = plus;
10556 case 'c': iv = (char)va_arg(*args, int); break;
10557 case 'h': iv = (short)va_arg(*args, int); break;
10558 case 'l': iv = va_arg(*args, long); break;
10559 case 'V': iv = va_arg(*args, IV); break;
10560 case 'z': iv = va_arg(*args, SSize_t); break;
10561 case 't': iv = va_arg(*args, ptrdiff_t); break;
10562 default: iv = va_arg(*args, int); break;
10564 case 'j': iv = va_arg(*args, intmax_t); break;
10568 iv = va_arg(*args, Quad_t); break;
10575 IV tiv = SvIV(argsv); /* work around GCC bug #13488 */
10577 case 'c': iv = (char)tiv; break;
10578 case 'h': iv = (short)tiv; break;
10579 case 'l': iv = (long)tiv; break;
10581 default: iv = tiv; break;
10584 iv = (Quad_t)tiv; break;
10590 if ( !vectorize ) /* we already set uv above */
10595 esignbuf[esignlen++] = plus;
10599 esignbuf[esignlen++] = '-';
10643 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
10654 case 'c': uv = (unsigned char)va_arg(*args, unsigned); break;
10655 case 'h': uv = (unsigned short)va_arg(*args, unsigned); break;
10656 case 'l': uv = va_arg(*args, unsigned long); break;
10657 case 'V': uv = va_arg(*args, UV); break;
10658 case 'z': uv = va_arg(*args, Size_t); break;
10659 case 't': uv = va_arg(*args, ptrdiff_t); break; /* will sign extend, but there is no uptrdiff_t, so oh well */
10661 case 'j': uv = va_arg(*args, uintmax_t); break;
10663 default: uv = va_arg(*args, unsigned); break;
10666 uv = va_arg(*args, Uquad_t); break;
10673 UV tuv = SvUV(argsv); /* work around GCC bug #13488 */
10675 case 'c': uv = (unsigned char)tuv; break;
10676 case 'h': uv = (unsigned short)tuv; break;
10677 case 'l': uv = (unsigned long)tuv; break;
10679 default: uv = tuv; break;
10682 uv = (Uquad_t)tuv; break;
10691 char *ptr = ebuf + sizeof ebuf;
10692 bool tempalt = uv ? alt : FALSE; /* Vectors can't change alt */
10698 p = (char *)((c == 'X') ? PL_hexdigit + 16 : PL_hexdigit);
10702 } while (uv >>= 4);
10704 esignbuf[esignlen++] = '0';
10705 esignbuf[esignlen++] = c; /* 'x' or 'X' */
10711 *--ptr = '0' + dig;
10712 } while (uv >>= 3);
10713 if (alt && *ptr != '0')
10719 *--ptr = '0' + dig;
10720 } while (uv >>= 1);
10722 esignbuf[esignlen++] = '0';
10723 esignbuf[esignlen++] = c;
10726 default: /* it had better be ten or less */
10729 *--ptr = '0' + dig;
10730 } while (uv /= base);
10733 elen = (ebuf + sizeof ebuf) - ptr;
10737 zeros = precis - elen;
10738 else if (precis == 0 && elen == 1 && *eptr == '0'
10739 && !(base == 8 && alt)) /* "%#.0o" prints "0" */
10742 /* a precision nullifies the 0 flag. */
10749 /* FLOATING POINT */
10752 c = 'f'; /* maybe %F isn't supported here */
10754 case 'e': case 'E':
10756 case 'g': case 'G':
10760 /* This is evil, but floating point is even more evil */
10762 /* for SV-style calling, we can only get NV
10763 for C-style calling, we assume %f is double;
10764 for simplicity we allow any of %Lf, %llf, %qf for long double
10768 #if defined(USE_LONG_DOUBLE)
10772 /* [perl #20339] - we should accept and ignore %lf rather than die */
10776 #if defined(USE_LONG_DOUBLE)
10777 intsize = args ? 0 : 'q';
10781 #if defined(HAS_LONG_DOUBLE)
10794 /* now we need (long double) if intsize == 'q', else (double) */
10796 #if LONG_DOUBLESIZE > DOUBLESIZE
10798 va_arg(*args, long double) :
10799 va_arg(*args, double)
10801 va_arg(*args, double)
10806 /* nv * 0 will be NaN for NaN, +Inf and -Inf, and 0 for anything
10807 else. frexp() has some unspecified behaviour for those three */
10808 if (c != 'e' && c != 'E' && (nv * 0) == 0) {
10810 /* FIXME: if HAS_LONG_DOUBLE but not USE_LONG_DOUBLE this
10811 will cast our (long double) to (double) */
10812 (void)Perl_frexp(nv, &i);
10813 if (i == PERL_INT_MIN)
10814 Perl_die(aTHX_ "panic: frexp");
10816 need = BIT_DIGITS(i);
10818 need += has_precis ? precis : 6; /* known default */
10823 #ifdef HAS_LDBL_SPRINTF_BUG
10824 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
10825 with sfio - Allen <allens@cpan.org> */
10828 # define MY_DBL_MAX DBL_MAX
10829 # else /* XXX guessing! HUGE_VAL may be defined as infinity, so not using */
10830 # if DOUBLESIZE >= 8
10831 # define MY_DBL_MAX 1.7976931348623157E+308L
10833 # define MY_DBL_MAX 3.40282347E+38L
10837 # ifdef HAS_LDBL_SPRINTF_BUG_LESS1 /* only between -1L & 1L - Allen */
10838 # define MY_DBL_MAX_BUG 1L
10840 # define MY_DBL_MAX_BUG MY_DBL_MAX
10844 # define MY_DBL_MIN DBL_MIN
10845 # else /* XXX guessing! -Allen */
10846 # if DOUBLESIZE >= 8
10847 # define MY_DBL_MIN 2.2250738585072014E-308L
10849 # define MY_DBL_MIN 1.17549435E-38L
10853 if ((intsize == 'q') && (c == 'f') &&
10854 ((nv < MY_DBL_MAX_BUG) && (nv > -MY_DBL_MAX_BUG)) &&
10855 (need < DBL_DIG)) {
10856 /* it's going to be short enough that
10857 * long double precision is not needed */
10859 if ((nv <= 0L) && (nv >= -0L))
10860 fix_ldbl_sprintf_bug = TRUE; /* 0 is 0 - easiest */
10862 /* would use Perl_fp_class as a double-check but not
10863 * functional on IRIX - see perl.h comments */
10865 if ((nv >= MY_DBL_MIN) || (nv <= -MY_DBL_MIN)) {
10866 /* It's within the range that a double can represent */
10867 #if defined(DBL_MAX) && !defined(DBL_MIN)
10868 if ((nv >= ((long double)1/DBL_MAX)) ||
10869 (nv <= (-(long double)1/DBL_MAX)))
10871 fix_ldbl_sprintf_bug = TRUE;
10874 if (fix_ldbl_sprintf_bug == TRUE) {
10884 # undef MY_DBL_MAX_BUG
10887 #endif /* HAS_LDBL_SPRINTF_BUG */
10889 need += 20; /* fudge factor */
10890 if (PL_efloatsize < need) {
10891 Safefree(PL_efloatbuf);
10892 PL_efloatsize = need + 20; /* more fudge */
10893 Newx(PL_efloatbuf, PL_efloatsize, char);
10894 PL_efloatbuf[0] = '\0';
10897 if ( !(width || left || plus || alt) && fill != '0'
10898 && has_precis && intsize != 'q' ) { /* Shortcuts */
10899 /* See earlier comment about buggy Gconvert when digits,
10901 if ( c == 'g' && precis) {
10902 Gconvert((NV)nv, (int)precis, 0, PL_efloatbuf);
10903 /* May return an empty string for digits==0 */
10904 if (*PL_efloatbuf) {
10905 elen = strlen(PL_efloatbuf);
10906 goto float_converted;
10908 } else if ( c == 'f' && !precis) {
10909 if ((eptr = F0convert(nv, ebuf + sizeof ebuf, &elen)))
10914 char *ptr = ebuf + sizeof ebuf;
10917 /* FIXME: what to do if HAS_LONG_DOUBLE but not PERL_PRIfldbl? */
10918 #if defined(HAS_LONG_DOUBLE) && defined(PERL_PRIfldbl)
10919 if (intsize == 'q') {
10920 /* Copy the one or more characters in a long double
10921 * format before the 'base' ([efgEFG]) character to
10922 * the format string. */
10923 static char const prifldbl[] = PERL_PRIfldbl;
10924 char const *p = prifldbl + sizeof(prifldbl) - 3;
10925 while (p >= prifldbl) { *--ptr = *p--; }
10930 do { *--ptr = '0' + (base % 10); } while (base /= 10);
10935 do { *--ptr = '0' + (base % 10); } while (base /= 10);
10947 /* No taint. Otherwise we are in the strange situation
10948 * where printf() taints but print($float) doesn't.
10950 #if defined(HAS_LONG_DOUBLE)
10951 elen = ((intsize == 'q')
10952 ? my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, nv)
10953 : my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, (double)nv));
10955 elen = my_sprintf(PL_efloatbuf, ptr, nv);
10959 eptr = PL_efloatbuf;
10967 i = SvCUR(sv) - origlen;
10970 case 'c': *(va_arg(*args, char*)) = i; break;
10971 case 'h': *(va_arg(*args, short*)) = i; break;
10972 default: *(va_arg(*args, int*)) = i; break;
10973 case 'l': *(va_arg(*args, long*)) = i; break;
10974 case 'V': *(va_arg(*args, IV*)) = i; break;
10975 case 'z': *(va_arg(*args, SSize_t*)) = i; break;
10976 case 't': *(va_arg(*args, ptrdiff_t*)) = i; break;
10978 case 'j': *(va_arg(*args, intmax_t*)) = i; break;
10982 *(va_arg(*args, Quad_t*)) = i; break;
10989 sv_setuv_mg(argsv, (UV)i);
10990 continue; /* not "break" */
10997 && (PL_op->op_type == OP_PRTF || PL_op->op_type == OP_SPRINTF)
10998 && ckWARN(WARN_PRINTF))
11000 SV * const msg = sv_newmortal();
11001 Perl_sv_setpvf(aTHX_ msg, "Invalid conversion in %sprintf: ",
11002 (PL_op->op_type == OP_PRTF) ? "" : "s");
11003 if (fmtstart < patend) {
11004 const char * const fmtend = q < patend ? q : patend;
11006 sv_catpvs(msg, "\"%");
11007 for (f = fmtstart; f < fmtend; f++) {
11009 sv_catpvn(msg, f, 1);
11011 Perl_sv_catpvf(aTHX_ msg,
11012 "\\%03"UVof, (UV)*f & 0xFF);
11015 sv_catpvs(msg, "\"");
11017 sv_catpvs(msg, "end of string");
11019 Perl_warner(aTHX_ packWARN(WARN_PRINTF), "%"SVf, SVfARG(msg)); /* yes, this is reentrant */
11022 /* output mangled stuff ... */
11028 /* ... right here, because formatting flags should not apply */
11029 SvGROW(sv, SvCUR(sv) + elen + 1);
11031 Copy(eptr, p, elen, char);
11034 SvCUR_set(sv, p - SvPVX_const(sv));
11036 continue; /* not "break" */
11039 if (is_utf8 != has_utf8) {
11042 sv_utf8_upgrade(sv);
11045 const STRLEN old_elen = elen;
11046 SV * const nsv = newSVpvn_flags(eptr, elen, SVs_TEMP);
11047 sv_utf8_upgrade(nsv);
11048 eptr = SvPVX_const(nsv);
11051 if (width) { /* fudge width (can't fudge elen) */
11052 width += elen - old_elen;
11058 have = esignlen + zeros + elen;
11060 Perl_croak_nocontext("%s", PL_memory_wrap);
11062 need = (have > width ? have : width);
11065 if (need >= (((STRLEN)~0) - SvCUR(sv) - dotstrlen - 1))
11066 Perl_croak_nocontext("%s", PL_memory_wrap);
11067 SvGROW(sv, SvCUR(sv) + need + dotstrlen + 1);
11069 if (esignlen && fill == '0') {
11071 for (i = 0; i < (int)esignlen; i++)
11072 *p++ = esignbuf[i];
11074 if (gap && !left) {
11075 memset(p, fill, gap);
11078 if (esignlen && fill != '0') {
11080 for (i = 0; i < (int)esignlen; i++)
11081 *p++ = esignbuf[i];
11085 for (i = zeros; i; i--)
11089 Copy(eptr, p, elen, char);
11093 memset(p, ' ', gap);
11098 Copy(dotstr, p, dotstrlen, char);
11102 vectorize = FALSE; /* done iterating over vecstr */
11109 SvCUR_set(sv, p - SvPVX_const(sv));
11118 /* =========================================================================
11120 =head1 Cloning an interpreter
11122 All the macros and functions in this section are for the private use of
11123 the main function, perl_clone().
11125 The foo_dup() functions make an exact copy of an existing foo thingy.
11126 During the course of a cloning, a hash table is used to map old addresses
11127 to new addresses. The table is created and manipulated with the
11128 ptr_table_* functions.
11132 * =========================================================================*/
11135 #if defined(USE_ITHREADS)
11137 /* XXX Remove this so it doesn't have to go thru the macro and return for nothing */
11138 #ifndef GpREFCNT_inc
11139 # define GpREFCNT_inc(gp) ((gp) ? (++(gp)->gp_refcnt, (gp)) : (GP*)NULL)
11143 /* Certain cases in Perl_ss_dup have been merged, by relying on the fact
11144 that currently av_dup, gv_dup and hv_dup are the same as sv_dup.
11145 If this changes, please unmerge ss_dup.
11146 Likewise, sv_dup_inc_multiple() relies on this fact. */
11147 #define sv_dup_inc_NN(s,t) SvREFCNT_inc_NN(sv_dup_inc(s,t))
11148 #define av_dup(s,t) MUTABLE_AV(sv_dup((const SV *)s,t))
11149 #define av_dup_inc(s,t) MUTABLE_AV(sv_dup_inc((const SV *)s,t))
11150 #define hv_dup(s,t) MUTABLE_HV(sv_dup((const SV *)s,t))
11151 #define hv_dup_inc(s,t) MUTABLE_HV(sv_dup_inc((const SV *)s,t))
11152 #define cv_dup(s,t) MUTABLE_CV(sv_dup((const SV *)s,t))
11153 #define cv_dup_inc(s,t) MUTABLE_CV(sv_dup_inc((const SV *)s,t))
11154 #define io_dup(s,t) MUTABLE_IO(sv_dup((const SV *)s,t))
11155 #define io_dup_inc(s,t) MUTABLE_IO(sv_dup_inc((const SV *)s,t))
11156 #define gv_dup(s,t) MUTABLE_GV(sv_dup((const SV *)s,t))
11157 #define gv_dup_inc(s,t) MUTABLE_GV(sv_dup_inc((const SV *)s,t))
11158 #define SAVEPV(p) ((p) ? savepv(p) : NULL)
11159 #define SAVEPVN(p,n) ((p) ? savepvn(p,n) : NULL)
11161 /* clone a parser */
11164 Perl_parser_dup(pTHX_ const yy_parser *const proto, CLONE_PARAMS *const param)
11168 PERL_ARGS_ASSERT_PARSER_DUP;
11173 /* look for it in the table first */
11174 parser = (yy_parser *)ptr_table_fetch(PL_ptr_table, proto);
11178 /* create anew and remember what it is */
11179 Newxz(parser, 1, yy_parser);
11180 ptr_table_store(PL_ptr_table, proto, parser);
11182 /* XXX these not yet duped */
11183 parser->old_parser = NULL;
11184 parser->stack = NULL;
11186 parser->stack_size = 0;
11187 /* XXX parser->stack->state = 0; */
11189 /* XXX eventually, just Copy() most of the parser struct ? */
11191 parser->lex_brackets = proto->lex_brackets;
11192 parser->lex_casemods = proto->lex_casemods;
11193 parser->lex_brackstack = savepvn(proto->lex_brackstack,
11194 (proto->lex_brackets < 120 ? 120 : proto->lex_brackets));
11195 parser->lex_casestack = savepvn(proto->lex_casestack,
11196 (proto->lex_casemods < 12 ? 12 : proto->lex_casemods));
11197 parser->lex_defer = proto->lex_defer;
11198 parser->lex_dojoin = proto->lex_dojoin;
11199 parser->lex_expect = proto->lex_expect;
11200 parser->lex_formbrack = proto->lex_formbrack;
11201 parser->lex_inpat = proto->lex_inpat;
11202 parser->lex_inwhat = proto->lex_inwhat;
11203 parser->lex_op = proto->lex_op;
11204 parser->lex_repl = sv_dup_inc(proto->lex_repl, param);
11205 parser->lex_starts = proto->lex_starts;
11206 parser->lex_stuff = sv_dup_inc(proto->lex_stuff, param);
11207 parser->multi_close = proto->multi_close;
11208 parser->multi_open = proto->multi_open;
11209 parser->multi_start = proto->multi_start;
11210 parser->multi_end = proto->multi_end;
11211 parser->pending_ident = proto->pending_ident;
11212 parser->preambled = proto->preambled;
11213 parser->sublex_info = proto->sublex_info; /* XXX not quite right */
11214 parser->linestr = sv_dup_inc(proto->linestr, param);
11215 parser->expect = proto->expect;
11216 parser->copline = proto->copline;
11217 parser->last_lop_op = proto->last_lop_op;
11218 parser->lex_state = proto->lex_state;
11219 parser->rsfp = fp_dup(proto->rsfp, '<', param);
11220 /* rsfp_filters entries have fake IoDIRP() */
11221 parser->rsfp_filters= av_dup_inc(proto->rsfp_filters, param);
11222 parser->in_my = proto->in_my;
11223 parser->in_my_stash = hv_dup(proto->in_my_stash, param);
11224 parser->error_count = proto->error_count;
11227 parser->linestr = sv_dup_inc(proto->linestr, param);
11230 char * const ols = SvPVX(proto->linestr);
11231 char * const ls = SvPVX(parser->linestr);
11233 parser->bufptr = ls + (proto->bufptr >= ols ?
11234 proto->bufptr - ols : 0);
11235 parser->oldbufptr = ls + (proto->oldbufptr >= ols ?
11236 proto->oldbufptr - ols : 0);
11237 parser->oldoldbufptr= ls + (proto->oldoldbufptr >= ols ?
11238 proto->oldoldbufptr - ols : 0);
11239 parser->linestart = ls + (proto->linestart >= ols ?
11240 proto->linestart - ols : 0);
11241 parser->last_uni = ls + (proto->last_uni >= ols ?
11242 proto->last_uni - ols : 0);
11243 parser->last_lop = ls + (proto->last_lop >= ols ?
11244 proto->last_lop - ols : 0);
11246 parser->bufend = ls + SvCUR(parser->linestr);
11249 Copy(proto->tokenbuf, parser->tokenbuf, 256, char);
11253 parser->endwhite = proto->endwhite;
11254 parser->faketokens = proto->faketokens;
11255 parser->lasttoke = proto->lasttoke;
11256 parser->nextwhite = proto->nextwhite;
11257 parser->realtokenstart = proto->realtokenstart;
11258 parser->skipwhite = proto->skipwhite;
11259 parser->thisclose = proto->thisclose;
11260 parser->thismad = proto->thismad;
11261 parser->thisopen = proto->thisopen;
11262 parser->thisstuff = proto->thisstuff;
11263 parser->thistoken = proto->thistoken;
11264 parser->thiswhite = proto->thiswhite;
11266 Copy(proto->nexttoke, parser->nexttoke, 5, NEXTTOKE);
11267 parser->curforce = proto->curforce;
11269 Copy(proto->nextval, parser->nextval, 5, YYSTYPE);
11270 Copy(proto->nexttype, parser->nexttype, 5, I32);
11271 parser->nexttoke = proto->nexttoke;
11274 /* XXX should clone saved_curcop here, but we aren't passed
11275 * proto_perl; so do it in perl_clone_using instead */
11281 /* duplicate a file handle */
11284 Perl_fp_dup(pTHX_ PerlIO *const fp, const char type, CLONE_PARAMS *const param)
11288 PERL_ARGS_ASSERT_FP_DUP;
11289 PERL_UNUSED_ARG(type);
11292 return (PerlIO*)NULL;
11294 /* look for it in the table first */
11295 ret = (PerlIO*)ptr_table_fetch(PL_ptr_table, fp);
11299 /* create anew and remember what it is */
11300 ret = PerlIO_fdupopen(aTHX_ fp, param, PERLIO_DUP_CLONE);
11301 ptr_table_store(PL_ptr_table, fp, ret);
11305 /* duplicate a directory handle */
11308 Perl_dirp_dup(pTHX_ DIR *const dp, CLONE_PARAMS *const param)
11314 register const Direntry_t *dirent;
11315 char smallbuf[256];
11321 PERL_UNUSED_CONTEXT;
11322 PERL_ARGS_ASSERT_DIRP_DUP;
11327 /* look for it in the table first */
11328 ret = (DIR*)ptr_table_fetch(PL_ptr_table, dp);
11334 PERL_UNUSED_ARG(param);
11338 /* open the current directory (so we can switch back) */
11339 if (!(pwd = PerlDir_open("."))) return (DIR *)NULL;
11341 /* chdir to our dir handle and open the present working directory */
11342 if (fchdir(my_dirfd(dp)) < 0 || !(ret = PerlDir_open("."))) {
11343 PerlDir_close(pwd);
11344 return (DIR *)NULL;
11346 /* Now we should have two dir handles pointing to the same dir. */
11348 /* Be nice to the calling code and chdir back to where we were. */
11349 fchdir(my_dirfd(pwd)); /* If this fails, then what? */
11351 /* We have no need of the pwd handle any more. */
11352 PerlDir_close(pwd);
11355 # define d_namlen(d) (d)->d_namlen
11357 # define d_namlen(d) strlen((d)->d_name)
11359 /* Iterate once through dp, to get the file name at the current posi-
11360 tion. Then step back. */
11361 pos = PerlDir_tell(dp);
11362 if ((dirent = PerlDir_read(dp))) {
11363 len = d_namlen(dirent);
11364 if (len <= sizeof smallbuf) name = smallbuf;
11365 else Newx(name, len, char);
11366 Move(dirent->d_name, name, len, char);
11368 PerlDir_seek(dp, pos);
11370 /* Iterate through the new dir handle, till we find a file with the
11372 if (!dirent) /* just before the end */
11374 pos = PerlDir_tell(ret);
11375 if (PerlDir_read(ret)) continue; /* not there yet */
11376 PerlDir_seek(ret, pos); /* step back */
11380 const long pos0 = PerlDir_tell(ret);
11382 pos = PerlDir_tell(ret);
11383 if ((dirent = PerlDir_read(ret))) {
11384 if (len == d_namlen(dirent)
11385 && memEQ(name, dirent->d_name, len)) {
11387 PerlDir_seek(ret, pos); /* step back */
11390 /* else we are not there yet; keep iterating */
11392 else { /* This is not meant to happen. The best we can do is
11393 reset the iterator to the beginning. */
11394 PerlDir_seek(ret, pos0);
11401 if (name && name != smallbuf)
11406 ret = win32_dirp_dup(dp, param);
11409 /* pop it in the pointer table */
11411 ptr_table_store(PL_ptr_table, dp, ret);
11416 /* duplicate a typeglob */
11419 Perl_gp_dup(pTHX_ GP *const gp, CLONE_PARAMS *const param)
11423 PERL_ARGS_ASSERT_GP_DUP;
11427 /* look for it in the table first */
11428 ret = (GP*)ptr_table_fetch(PL_ptr_table, gp);
11432 /* create anew and remember what it is */
11434 ptr_table_store(PL_ptr_table, gp, ret);
11437 /* ret->gp_refcnt must be 0 before any other dups are called. We're relying
11438 on Newxz() to do this for us. */
11439 ret->gp_sv = sv_dup_inc(gp->gp_sv, param);
11440 ret->gp_io = io_dup_inc(gp->gp_io, param);
11441 ret->gp_form = cv_dup_inc(gp->gp_form, param);
11442 ret->gp_av = av_dup_inc(gp->gp_av, param);
11443 ret->gp_hv = hv_dup_inc(gp->gp_hv, param);
11444 ret->gp_egv = gv_dup(gp->gp_egv, param);/* GvEGV is not refcounted */
11445 ret->gp_cv = cv_dup_inc(gp->gp_cv, param);
11446 ret->gp_cvgen = gp->gp_cvgen;
11447 ret->gp_line = gp->gp_line;
11448 ret->gp_file_hek = hek_dup(gp->gp_file_hek, param);
11452 /* duplicate a chain of magic */
11455 Perl_mg_dup(pTHX_ MAGIC *mg, CLONE_PARAMS *const param)
11457 MAGIC *mgret = NULL;
11458 MAGIC **mgprev_p = &mgret;
11460 PERL_ARGS_ASSERT_MG_DUP;
11462 for (; mg; mg = mg->mg_moremagic) {
11465 if ((param->flags & CLONEf_JOIN_IN)
11466 && mg->mg_type == PERL_MAGIC_backref)
11467 /* when joining, we let the individual SVs add themselves to
11468 * backref as needed. */
11471 Newx(nmg, 1, MAGIC);
11473 mgprev_p = &(nmg->mg_moremagic);
11475 /* There was a comment "XXX copy dynamic vtable?" but as we don't have
11476 dynamic vtables, I'm not sure why Sarathy wrote it. The comment dates
11477 from the original commit adding Perl_mg_dup() - revision 4538.
11478 Similarly there is the annotation "XXX random ptr?" next to the
11479 assignment to nmg->mg_ptr. */
11482 /* FIXME for plugins
11483 if (nmg->mg_type == PERL_MAGIC_qr) {
11484 nmg->mg_obj = MUTABLE_SV(CALLREGDUPE((REGEXP*)nmg->mg_obj, param));
11488 nmg->mg_obj = (nmg->mg_flags & MGf_REFCOUNTED)
11489 ? nmg->mg_type == PERL_MAGIC_backref
11490 /* The backref AV has its reference
11491 * count deliberately bumped by 1 */
11492 ? SvREFCNT_inc(av_dup_inc((const AV *)
11493 nmg->mg_obj, param))
11494 : sv_dup_inc(nmg->mg_obj, param)
11495 : sv_dup(nmg->mg_obj, param);
11497 if (nmg->mg_ptr && nmg->mg_type != PERL_MAGIC_regex_global) {
11498 if (nmg->mg_len > 0) {
11499 nmg->mg_ptr = SAVEPVN(nmg->mg_ptr, nmg->mg_len);
11500 if (nmg->mg_type == PERL_MAGIC_overload_table &&
11501 AMT_AMAGIC((AMT*)nmg->mg_ptr))
11503 AMT * const namtp = (AMT*)nmg->mg_ptr;
11504 sv_dup_inc_multiple((SV**)(namtp->table),
11505 (SV**)(namtp->table), NofAMmeth, param);
11508 else if (nmg->mg_len == HEf_SVKEY)
11509 nmg->mg_ptr = (char*)sv_dup_inc((const SV *)nmg->mg_ptr, param);
11511 if ((nmg->mg_flags & MGf_DUP) && nmg->mg_virtual && nmg->mg_virtual->svt_dup) {
11512 nmg->mg_virtual->svt_dup(aTHX_ nmg, param);
11518 #endif /* USE_ITHREADS */
11520 struct ptr_tbl_arena {
11521 struct ptr_tbl_arena *next;
11522 struct ptr_tbl_ent array[1023/3]; /* as ptr_tbl_ent has 3 pointers. */
11525 /* create a new pointer-mapping table */
11528 Perl_ptr_table_new(pTHX)
11531 PERL_UNUSED_CONTEXT;
11533 Newx(tbl, 1, PTR_TBL_t);
11534 tbl->tbl_max = 511;
11535 tbl->tbl_items = 0;
11536 tbl->tbl_arena = NULL;
11537 tbl->tbl_arena_next = NULL;
11538 tbl->tbl_arena_end = NULL;
11539 Newxz(tbl->tbl_ary, tbl->tbl_max + 1, PTR_TBL_ENT_t*);
11543 #define PTR_TABLE_HASH(ptr) \
11544 ((PTR2UV(ptr) >> 3) ^ (PTR2UV(ptr) >> (3 + 7)) ^ (PTR2UV(ptr) >> (3 + 17)))
11546 /* map an existing pointer using a table */
11548 STATIC PTR_TBL_ENT_t *
11549 S_ptr_table_find(PTR_TBL_t *const tbl, const void *const sv)
11551 PTR_TBL_ENT_t *tblent;
11552 const UV hash = PTR_TABLE_HASH(sv);
11554 PERL_ARGS_ASSERT_PTR_TABLE_FIND;
11556 tblent = tbl->tbl_ary[hash & tbl->tbl_max];
11557 for (; tblent; tblent = tblent->next) {
11558 if (tblent->oldval == sv)
11565 Perl_ptr_table_fetch(pTHX_ PTR_TBL_t *const tbl, const void *const sv)
11567 PTR_TBL_ENT_t const *const tblent = ptr_table_find(tbl, sv);
11569 PERL_ARGS_ASSERT_PTR_TABLE_FETCH;
11570 PERL_UNUSED_CONTEXT;
11572 return tblent ? tblent->newval : NULL;
11575 /* add a new entry to a pointer-mapping table */
11578 Perl_ptr_table_store(pTHX_ PTR_TBL_t *const tbl, const void *const oldsv, void *const newsv)
11580 PTR_TBL_ENT_t *tblent = ptr_table_find(tbl, oldsv);
11582 PERL_ARGS_ASSERT_PTR_TABLE_STORE;
11583 PERL_UNUSED_CONTEXT;
11586 tblent->newval = newsv;
11588 const UV entry = PTR_TABLE_HASH(oldsv) & tbl->tbl_max;
11590 if (tbl->tbl_arena_next == tbl->tbl_arena_end) {
11591 struct ptr_tbl_arena *new_arena;
11593 Newx(new_arena, 1, struct ptr_tbl_arena);
11594 new_arena->next = tbl->tbl_arena;
11595 tbl->tbl_arena = new_arena;
11596 tbl->tbl_arena_next = new_arena->array;
11597 tbl->tbl_arena_end = new_arena->array
11598 + sizeof(new_arena->array) / sizeof(new_arena->array[0]);
11601 tblent = tbl->tbl_arena_next++;
11603 tblent->oldval = oldsv;
11604 tblent->newval = newsv;
11605 tblent->next = tbl->tbl_ary[entry];
11606 tbl->tbl_ary[entry] = tblent;
11608 if (tblent->next && tbl->tbl_items > tbl->tbl_max)
11609 ptr_table_split(tbl);
11613 /* double the hash bucket size of an existing ptr table */
11616 Perl_ptr_table_split(pTHX_ PTR_TBL_t *const tbl)
11618 PTR_TBL_ENT_t **ary = tbl->tbl_ary;
11619 const UV oldsize = tbl->tbl_max + 1;
11620 UV newsize = oldsize * 2;
11623 PERL_ARGS_ASSERT_PTR_TABLE_SPLIT;
11624 PERL_UNUSED_CONTEXT;
11626 Renew(ary, newsize, PTR_TBL_ENT_t*);
11627 Zero(&ary[oldsize], newsize-oldsize, PTR_TBL_ENT_t*);
11628 tbl->tbl_max = --newsize;
11629 tbl->tbl_ary = ary;
11630 for (i=0; i < oldsize; i++, ary++) {
11631 PTR_TBL_ENT_t **entp = ary;
11632 PTR_TBL_ENT_t *ent = *ary;
11633 PTR_TBL_ENT_t **curentp;
11636 curentp = ary + oldsize;
11638 if ((newsize & PTR_TABLE_HASH(ent->oldval)) != i) {
11640 ent->next = *curentp;
11650 /* remove all the entries from a ptr table */
11651 /* Deprecated - will be removed post 5.14 */
11654 Perl_ptr_table_clear(pTHX_ PTR_TBL_t *const tbl)
11656 if (tbl && tbl->tbl_items) {
11657 struct ptr_tbl_arena *arena = tbl->tbl_arena;
11659 Zero(tbl->tbl_ary, tbl->tbl_max + 1, struct ptr_tbl_ent **);
11662 struct ptr_tbl_arena *next = arena->next;
11668 tbl->tbl_items = 0;
11669 tbl->tbl_arena = NULL;
11670 tbl->tbl_arena_next = NULL;
11671 tbl->tbl_arena_end = NULL;
11675 /* clear and free a ptr table */
11678 Perl_ptr_table_free(pTHX_ PTR_TBL_t *const tbl)
11680 struct ptr_tbl_arena *arena;
11686 arena = tbl->tbl_arena;
11689 struct ptr_tbl_arena *next = arena->next;
11695 Safefree(tbl->tbl_ary);
11699 #if defined(USE_ITHREADS)
11702 Perl_rvpv_dup(pTHX_ SV *const dstr, const SV *const sstr, CLONE_PARAMS *const param)
11704 PERL_ARGS_ASSERT_RVPV_DUP;
11707 if (SvWEAKREF(sstr)) {
11708 SvRV_set(dstr, sv_dup(SvRV_const(sstr), param));
11709 if (param->flags & CLONEf_JOIN_IN) {
11710 /* if joining, we add any back references individually rather
11711 * than copying the whole backref array */
11712 Perl_sv_add_backref(aTHX_ SvRV(dstr), dstr);
11716 SvRV_set(dstr, sv_dup_inc(SvRV_const(sstr), param));
11718 else if (SvPVX_const(sstr)) {
11719 /* Has something there */
11721 /* Normal PV - clone whole allocated space */
11722 SvPV_set(dstr, SAVEPVN(SvPVX_const(sstr), SvLEN(sstr)-1));
11723 if (SvREADONLY(sstr) && SvFAKE(sstr)) {
11724 /* Not that normal - actually sstr is copy on write.
11725 But we are a true, independent SV, so: */
11726 SvREADONLY_off(dstr);
11731 /* Special case - not normally malloced for some reason */
11732 if (isGV_with_GP(sstr)) {
11733 /* Don't need to do anything here. */
11735 else if ((SvREADONLY(sstr) && SvFAKE(sstr))) {
11736 /* A "shared" PV - clone it as "shared" PV */
11738 HEK_KEY(hek_dup(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)),
11742 /* Some other special case - random pointer */
11743 SvPV_set(dstr, (char *) SvPVX_const(sstr));
11748 /* Copy the NULL */
11749 SvPV_set(dstr, NULL);
11753 /* duplicate a list of SVs. source and dest may point to the same memory. */
11755 S_sv_dup_inc_multiple(pTHX_ SV *const *source, SV **dest,
11756 SSize_t items, CLONE_PARAMS *const param)
11758 PERL_ARGS_ASSERT_SV_DUP_INC_MULTIPLE;
11760 while (items-- > 0) {
11761 *dest++ = sv_dup_inc(*source++, param);
11767 /* duplicate an SV of any type (including AV, HV etc) */
11770 S_sv_dup_common(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
11775 PERL_ARGS_ASSERT_SV_DUP_COMMON;
11777 if (SvTYPE(sstr) == (svtype)SVTYPEMASK) {
11778 #ifdef DEBUG_LEAKING_SCALARS_ABORT
11783 /* look for it in the table first */
11784 dstr = MUTABLE_SV(ptr_table_fetch(PL_ptr_table, sstr));
11788 if(param->flags & CLONEf_JOIN_IN) {
11789 /** We are joining here so we don't want do clone
11790 something that is bad **/
11791 if (SvTYPE(sstr) == SVt_PVHV) {
11792 const HEK * const hvname = HvNAME_HEK(sstr);
11794 /** don't clone stashes if they already exist **/
11795 dstr = MUTABLE_SV(gv_stashpvn(HEK_KEY(hvname), HEK_LEN(hvname),
11796 HEK_UTF8(hvname) ? SVf_UTF8 : 0));
11797 ptr_table_store(PL_ptr_table, sstr, dstr);
11803 /* create anew and remember what it is */
11806 #ifdef DEBUG_LEAKING_SCALARS
11807 dstr->sv_debug_optype = sstr->sv_debug_optype;
11808 dstr->sv_debug_line = sstr->sv_debug_line;
11809 dstr->sv_debug_inpad = sstr->sv_debug_inpad;
11810 dstr->sv_debug_parent = (SV*)sstr;
11811 FREE_SV_DEBUG_FILE(dstr);
11812 dstr->sv_debug_file = savepv(sstr->sv_debug_file);
11815 ptr_table_store(PL_ptr_table, sstr, dstr);
11818 SvFLAGS(dstr) = SvFLAGS(sstr);
11819 SvFLAGS(dstr) &= ~SVf_OOK; /* don't propagate OOK hack */
11820 SvREFCNT(dstr) = 0; /* must be before any other dups! */
11823 if (SvANY(sstr) && PL_watch_pvx && SvPVX_const(sstr) == PL_watch_pvx)
11824 PerlIO_printf(Perl_debug_log, "watch at %p hit, found string \"%s\"\n",
11825 (void*)PL_watch_pvx, SvPVX_const(sstr));
11828 /* don't clone objects whose class has asked us not to */
11829 if (SvOBJECT(sstr) && ! (SvFLAGS(SvSTASH(sstr)) & SVphv_CLONEABLE)) {
11834 switch (SvTYPE(sstr)) {
11836 SvANY(dstr) = NULL;
11839 SvANY(dstr) = (XPVIV*)((char*)&(dstr->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
11841 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
11843 SvIV_set(dstr, SvIVX(sstr));
11847 SvANY(dstr) = new_XNV();
11848 SvNV_set(dstr, SvNVX(sstr));
11850 /* case SVt_BIND: */
11853 /* These are all the types that need complex bodies allocating. */
11855 const svtype sv_type = SvTYPE(sstr);
11856 const struct body_details *const sv_type_details
11857 = bodies_by_type + sv_type;
11861 Perl_croak(aTHX_ "Bizarre SvTYPE [%" IVdf "]", (IV)SvTYPE(sstr));
11876 assert(sv_type_details->body_size);
11877 if (sv_type_details->arena) {
11878 new_body_inline(new_body, sv_type);
11880 = (void*)((char*)new_body - sv_type_details->offset);
11882 new_body = new_NOARENA(sv_type_details);
11886 SvANY(dstr) = new_body;
11889 Copy(((char*)SvANY(sstr)) + sv_type_details->offset,
11890 ((char*)SvANY(dstr)) + sv_type_details->offset,
11891 sv_type_details->copy, char);
11893 Copy(((char*)SvANY(sstr)),
11894 ((char*)SvANY(dstr)),
11895 sv_type_details->body_size + sv_type_details->offset, char);
11898 if (sv_type != SVt_PVAV && sv_type != SVt_PVHV
11899 && !isGV_with_GP(dstr)
11900 && !(sv_type == SVt_PVIO && !(IoFLAGS(dstr) & IOf_FAKE_DIRP)))
11901 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
11903 /* The Copy above means that all the source (unduplicated) pointers
11904 are now in the destination. We can check the flags and the
11905 pointers in either, but it's possible that there's less cache
11906 missing by always going for the destination.
11907 FIXME - instrument and check that assumption */
11908 if (sv_type >= SVt_PVMG) {
11909 if ((sv_type == SVt_PVMG) && SvPAD_OUR(dstr)) {
11910 SvOURSTASH_set(dstr, hv_dup_inc(SvOURSTASH(dstr), param));
11911 } else if (SvMAGIC(dstr))
11912 SvMAGIC_set(dstr, mg_dup(SvMAGIC(dstr), param));
11914 SvSTASH_set(dstr, hv_dup_inc(SvSTASH(dstr), param));
11917 /* The cast silences a GCC warning about unhandled types. */
11918 switch ((int)sv_type) {
11928 /* FIXME for plugins */
11929 re_dup_guts((REGEXP*) sstr, (REGEXP*) dstr, param);
11932 /* XXX LvTARGOFF sometimes holds PMOP* when DEBUGGING */
11933 if (LvTYPE(dstr) == 't') /* for tie: unrefcnted fake (SV**) */
11934 LvTARG(dstr) = dstr;
11935 else if (LvTYPE(dstr) == 'T') /* for tie: fake HE */
11936 LvTARG(dstr) = MUTABLE_SV(he_dup((HE*)LvTARG(dstr), 0, param));
11938 LvTARG(dstr) = sv_dup_inc(LvTARG(dstr), param);
11940 /* non-GP case already handled above */
11941 if(isGV_with_GP(sstr)) {
11942 GvNAME_HEK(dstr) = hek_dup(GvNAME_HEK(dstr), param);
11943 /* Don't call sv_add_backref here as it's going to be
11944 created as part of the magic cloning of the symbol
11945 table--unless this is during a join and the stash
11946 is not actually being cloned. */
11947 /* Danger Will Robinson - GvGP(dstr) isn't initialised
11948 at the point of this comment. */
11949 GvSTASH(dstr) = hv_dup(GvSTASH(dstr), param);
11950 if (param->flags & CLONEf_JOIN_IN)
11951 Perl_sv_add_backref(aTHX_ MUTABLE_SV(GvSTASH(dstr)), dstr);
11952 GvGP_set(dstr, gp_dup(GvGP(sstr), param));
11953 (void)GpREFCNT_inc(GvGP(dstr));
11957 /* PL_parser->rsfp_filters entries have fake IoDIRP() */
11958 if(IoFLAGS(dstr) & IOf_FAKE_DIRP) {
11959 /* I have no idea why fake dirp (rsfps)
11960 should be treated differently but otherwise
11961 we end up with leaks -- sky*/
11962 IoTOP_GV(dstr) = gv_dup_inc(IoTOP_GV(dstr), param);
11963 IoFMT_GV(dstr) = gv_dup_inc(IoFMT_GV(dstr), param);
11964 IoBOTTOM_GV(dstr) = gv_dup_inc(IoBOTTOM_GV(dstr), param);
11966 IoTOP_GV(dstr) = gv_dup(IoTOP_GV(dstr), param);
11967 IoFMT_GV(dstr) = gv_dup(IoFMT_GV(dstr), param);
11968 IoBOTTOM_GV(dstr) = gv_dup(IoBOTTOM_GV(dstr), param);
11969 if (IoDIRP(dstr)) {
11970 IoDIRP(dstr) = dirp_dup(IoDIRP(dstr), param);
11973 /* IoDIRP(dstr) is already a copy of IoDIRP(sstr) */
11975 IoIFP(dstr) = fp_dup(IoIFP(sstr), IoTYPE(dstr), param);
11977 if (IoOFP(dstr) == IoIFP(sstr))
11978 IoOFP(dstr) = IoIFP(dstr);
11980 IoOFP(dstr) = fp_dup(IoOFP(dstr), IoTYPE(dstr), param);
11981 IoTOP_NAME(dstr) = SAVEPV(IoTOP_NAME(dstr));
11982 IoFMT_NAME(dstr) = SAVEPV(IoFMT_NAME(dstr));
11983 IoBOTTOM_NAME(dstr) = SAVEPV(IoBOTTOM_NAME(dstr));
11986 /* avoid cloning an empty array */
11987 if (AvARRAY((const AV *)sstr) && AvFILLp((const AV *)sstr) >= 0) {
11988 SV **dst_ary, **src_ary;
11989 SSize_t items = AvFILLp((const AV *)sstr) + 1;
11991 src_ary = AvARRAY((const AV *)sstr);
11992 Newxz(dst_ary, AvMAX((const AV *)sstr)+1, SV*);
11993 ptr_table_store(PL_ptr_table, src_ary, dst_ary);
11994 AvARRAY(MUTABLE_AV(dstr)) = dst_ary;
11995 AvALLOC((const AV *)dstr) = dst_ary;
11996 if (AvREAL((const AV *)sstr)) {
11997 dst_ary = sv_dup_inc_multiple(src_ary, dst_ary, items,
12001 while (items-- > 0)
12002 *dst_ary++ = sv_dup(*src_ary++, param);
12004 items = AvMAX((const AV *)sstr) - AvFILLp((const AV *)sstr);
12005 while (items-- > 0) {
12006 *dst_ary++ = &PL_sv_undef;
12010 AvARRAY(MUTABLE_AV(dstr)) = NULL;
12011 AvALLOC((const AV *)dstr) = (SV**)NULL;
12012 AvMAX( (const AV *)dstr) = -1;
12013 AvFILLp((const AV *)dstr) = -1;
12017 if (HvARRAY((const HV *)sstr)) {
12019 const bool sharekeys = !!HvSHAREKEYS(sstr);
12020 XPVHV * const dxhv = (XPVHV*)SvANY(dstr);
12021 XPVHV * const sxhv = (XPVHV*)SvANY(sstr);
12023 Newx(darray, PERL_HV_ARRAY_ALLOC_BYTES(dxhv->xhv_max+1)
12024 + (SvOOK(sstr) ? sizeof(struct xpvhv_aux) : 0),
12026 HvARRAY(dstr) = (HE**)darray;
12027 while (i <= sxhv->xhv_max) {
12028 const HE * const source = HvARRAY(sstr)[i];
12029 HvARRAY(dstr)[i] = source
12030 ? he_dup(source, sharekeys, param) : 0;
12034 const struct xpvhv_aux * const saux = HvAUX(sstr);
12035 struct xpvhv_aux * const daux = HvAUX(dstr);
12036 /* This flag isn't copied. */
12037 /* SvOOK_on(hv) attacks the IV flags. */
12038 SvFLAGS(dstr) |= SVf_OOK;
12040 if (saux->xhv_name_count) {
12041 HEK ** const sname = saux->xhv_name_u.xhvnameu_names;
12043 = saux->xhv_name_count < 0
12044 ? -saux->xhv_name_count
12045 : saux->xhv_name_count;
12046 HEK **shekp = sname + count;
12048 Newx(daux->xhv_name_u.xhvnameu_names, count, HEK *);
12049 dhekp = daux->xhv_name_u.xhvnameu_names + count;
12050 while (shekp-- > sname) {
12052 *dhekp = hek_dup(*shekp, param);
12056 daux->xhv_name_u.xhvnameu_name
12057 = hek_dup(saux->xhv_name_u.xhvnameu_name,
12060 daux->xhv_name_count = saux->xhv_name_count;
12062 daux->xhv_riter = saux->xhv_riter;
12063 daux->xhv_eiter = saux->xhv_eiter
12064 ? he_dup(saux->xhv_eiter,
12065 cBOOL(HvSHAREKEYS(sstr)), param) : 0;
12066 /* backref array needs refcnt=2; see sv_add_backref */
12067 daux->xhv_backreferences =
12068 (param->flags & CLONEf_JOIN_IN)
12069 /* when joining, we let the individual GVs and
12070 * CVs add themselves to backref as
12071 * needed. This avoids pulling in stuff
12072 * that isn't required, and simplifies the
12073 * case where stashes aren't cloned back
12074 * if they already exist in the parent
12077 : saux->xhv_backreferences
12078 ? (SvTYPE(saux->xhv_backreferences) == SVt_PVAV)
12079 ? MUTABLE_AV(SvREFCNT_inc(
12080 sv_dup_inc((const SV *)
12081 saux->xhv_backreferences, param)))
12082 : MUTABLE_AV(sv_dup((const SV *)
12083 saux->xhv_backreferences, param))
12086 daux->xhv_mro_meta = saux->xhv_mro_meta
12087 ? mro_meta_dup(saux->xhv_mro_meta, param)
12090 /* Record stashes for possible cloning in Perl_clone(). */
12092 av_push(param->stashes, dstr);
12096 HvARRAY(MUTABLE_HV(dstr)) = NULL;
12099 if (!(param->flags & CLONEf_COPY_STACKS)) {
12104 /* NOTE: not refcounted */
12105 SvANY(MUTABLE_CV(dstr))->xcv_stash =
12106 hv_dup(CvSTASH(dstr), param);
12107 if ((param->flags & CLONEf_JOIN_IN) && CvSTASH(dstr))
12108 Perl_sv_add_backref(aTHX_ MUTABLE_SV(CvSTASH(dstr)), dstr);
12109 if (!CvISXSUB(dstr)) {
12111 CvROOT(dstr) = OpREFCNT_inc(CvROOT(dstr));
12113 } else if (CvCONST(dstr)) {
12114 CvXSUBANY(dstr).any_ptr =
12115 sv_dup_inc((const SV *)CvXSUBANY(dstr).any_ptr, param);
12117 if (CvDYNFILE(dstr)) CvFILE(dstr) = SAVEPV(CvFILE(dstr));
12118 /* don't dup if copying back - CvGV isn't refcounted, so the
12119 * duped GV may never be freed. A bit of a hack! DAPM */
12120 SvANY(MUTABLE_CV(dstr))->xcv_gv =
12122 ? gv_dup_inc(CvGV(sstr), param)
12123 : (param->flags & CLONEf_JOIN_IN)
12125 : gv_dup(CvGV(sstr), param);
12127 CvPADLIST(dstr) = padlist_dup(CvPADLIST(sstr), param);
12129 CvWEAKOUTSIDE(sstr)
12130 ? cv_dup( CvOUTSIDE(dstr), param)
12131 : cv_dup_inc(CvOUTSIDE(dstr), param);
12137 if (SvOBJECT(dstr) && SvTYPE(dstr) != SVt_PVIO)
12144 Perl_sv_dup_inc(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
12146 PERL_ARGS_ASSERT_SV_DUP_INC;
12147 return sstr ? SvREFCNT_inc(sv_dup_common(sstr, param)) : NULL;
12151 Perl_sv_dup(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
12153 SV *dstr = sstr ? sv_dup_common(sstr, param) : NULL;
12154 PERL_ARGS_ASSERT_SV_DUP;
12156 /* Track every SV that (at least initially) had a reference count of 0.
12157 We need to do this by holding an actual reference to it in this array.
12158 If we attempt to cheat, turn AvREAL_off(), and store only pointers
12159 (akin to the stashes hash, and the perl stack), we come unstuck if
12160 a weak reference (or other SV legitimately SvREFCNT() == 0 for this
12161 thread) is manipulated in a CLONE method, because CLONE runs before the
12162 unreferenced array is walked to find SVs still with SvREFCNT() == 0
12163 (and fix things up by giving each a reference via the temps stack).
12164 Instead, during CLONE, if the 0-referenced SV has SvREFCNT_inc() and
12165 then SvREFCNT_dec(), it will be cleaned up (and added to the free list)
12166 before the walk of unreferenced happens and a reference to that is SV
12167 added to the temps stack. At which point we have the same SV considered
12168 to be in use, and free to be re-used. Not good.
12170 if (dstr && !(param->flags & CLONEf_COPY_STACKS) && !SvREFCNT(dstr)) {
12171 assert(param->unreferenced);
12172 av_push(param->unreferenced, SvREFCNT_inc(dstr));
12178 /* duplicate a context */
12181 Perl_cx_dup(pTHX_ PERL_CONTEXT *cxs, I32 ix, I32 max, CLONE_PARAMS* param)
12183 PERL_CONTEXT *ncxs;
12185 PERL_ARGS_ASSERT_CX_DUP;
12188 return (PERL_CONTEXT*)NULL;
12190 /* look for it in the table first */
12191 ncxs = (PERL_CONTEXT*)ptr_table_fetch(PL_ptr_table, cxs);
12195 /* create anew and remember what it is */
12196 Newx(ncxs, max + 1, PERL_CONTEXT);
12197 ptr_table_store(PL_ptr_table, cxs, ncxs);
12198 Copy(cxs, ncxs, max + 1, PERL_CONTEXT);
12201 PERL_CONTEXT * const ncx = &ncxs[ix];
12202 if (CxTYPE(ncx) == CXt_SUBST) {
12203 Perl_croak(aTHX_ "Cloning substitution context is unimplemented");
12206 switch (CxTYPE(ncx)) {
12208 ncx->blk_sub.cv = (ncx->blk_sub.olddepth == 0
12209 ? cv_dup_inc(ncx->blk_sub.cv, param)
12210 : cv_dup(ncx->blk_sub.cv,param));
12211 ncx->blk_sub.argarray = (CxHASARGS(ncx)
12212 ? av_dup_inc(ncx->blk_sub.argarray,
12215 ncx->blk_sub.savearray = av_dup_inc(ncx->blk_sub.savearray,
12217 ncx->blk_sub.oldcomppad = (PAD*)ptr_table_fetch(PL_ptr_table,
12218 ncx->blk_sub.oldcomppad);
12221 ncx->blk_eval.old_namesv = sv_dup_inc(ncx->blk_eval.old_namesv,
12223 ncx->blk_eval.cur_text = sv_dup(ncx->blk_eval.cur_text, param);
12225 case CXt_LOOP_LAZYSV:
12226 ncx->blk_loop.state_u.lazysv.end
12227 = sv_dup_inc(ncx->blk_loop.state_u.lazysv.end, param);
12228 /* We are taking advantage of av_dup_inc and sv_dup_inc
12229 actually being the same function, and order equivalence of
12231 We can assert the later [but only at run time :-(] */
12232 assert ((void *) &ncx->blk_loop.state_u.ary.ary ==
12233 (void *) &ncx->blk_loop.state_u.lazysv.cur);
12235 ncx->blk_loop.state_u.ary.ary
12236 = av_dup_inc(ncx->blk_loop.state_u.ary.ary, param);
12237 case CXt_LOOP_LAZYIV:
12238 case CXt_LOOP_PLAIN:
12239 if (CxPADLOOP(ncx)) {
12240 ncx->blk_loop.itervar_u.oldcomppad
12241 = (PAD*)ptr_table_fetch(PL_ptr_table,
12242 ncx->blk_loop.itervar_u.oldcomppad);
12244 ncx->blk_loop.itervar_u.gv
12245 = gv_dup((const GV *)ncx->blk_loop.itervar_u.gv,
12250 ncx->blk_format.cv = cv_dup(ncx->blk_format.cv, param);
12251 ncx->blk_format.gv = gv_dup(ncx->blk_format.gv, param);
12252 ncx->blk_format.dfoutgv = gv_dup_inc(ncx->blk_format.dfoutgv,
12265 /* duplicate a stack info structure */
12268 Perl_si_dup(pTHX_ PERL_SI *si, CLONE_PARAMS* param)
12272 PERL_ARGS_ASSERT_SI_DUP;
12275 return (PERL_SI*)NULL;
12277 /* look for it in the table first */
12278 nsi = (PERL_SI*)ptr_table_fetch(PL_ptr_table, si);
12282 /* create anew and remember what it is */
12283 Newxz(nsi, 1, PERL_SI);
12284 ptr_table_store(PL_ptr_table, si, nsi);
12286 nsi->si_stack = av_dup_inc(si->si_stack, param);
12287 nsi->si_cxix = si->si_cxix;
12288 nsi->si_cxmax = si->si_cxmax;
12289 nsi->si_cxstack = cx_dup(si->si_cxstack, si->si_cxix, si->si_cxmax, param);
12290 nsi->si_type = si->si_type;
12291 nsi->si_prev = si_dup(si->si_prev, param);
12292 nsi->si_next = si_dup(si->si_next, param);
12293 nsi->si_markoff = si->si_markoff;
12298 #define POPINT(ss,ix) ((ss)[--(ix)].any_i32)
12299 #define TOPINT(ss,ix) ((ss)[ix].any_i32)
12300 #define POPLONG(ss,ix) ((ss)[--(ix)].any_long)
12301 #define TOPLONG(ss,ix) ((ss)[ix].any_long)
12302 #define POPIV(ss,ix) ((ss)[--(ix)].any_iv)
12303 #define TOPIV(ss,ix) ((ss)[ix].any_iv)
12304 #define POPUV(ss,ix) ((ss)[--(ix)].any_uv)
12305 #define TOPUV(ss,ix) ((ss)[ix].any_uv)
12306 #define POPBOOL(ss,ix) ((ss)[--(ix)].any_bool)
12307 #define TOPBOOL(ss,ix) ((ss)[ix].any_bool)
12308 #define POPPTR(ss,ix) ((ss)[--(ix)].any_ptr)
12309 #define TOPPTR(ss,ix) ((ss)[ix].any_ptr)
12310 #define POPDPTR(ss,ix) ((ss)[--(ix)].any_dptr)
12311 #define TOPDPTR(ss,ix) ((ss)[ix].any_dptr)
12312 #define POPDXPTR(ss,ix) ((ss)[--(ix)].any_dxptr)
12313 #define TOPDXPTR(ss,ix) ((ss)[ix].any_dxptr)
12316 #define pv_dup_inc(p) SAVEPV(p)
12317 #define pv_dup(p) SAVEPV(p)
12318 #define svp_dup_inc(p,pp) any_dup(p,pp)
12320 /* map any object to the new equivent - either something in the
12321 * ptr table, or something in the interpreter structure
12325 Perl_any_dup(pTHX_ void *v, const PerlInterpreter *proto_perl)
12329 PERL_ARGS_ASSERT_ANY_DUP;
12332 return (void*)NULL;
12334 /* look for it in the table first */
12335 ret = ptr_table_fetch(PL_ptr_table, v);
12339 /* see if it is part of the interpreter structure */
12340 if (v >= (void*)proto_perl && v < (void*)(proto_perl+1))
12341 ret = (void*)(((char*)aTHX) + (((char*)v) - (char*)proto_perl));
12349 /* duplicate the save stack */
12352 Perl_ss_dup(pTHX_ PerlInterpreter *proto_perl, CLONE_PARAMS* param)
12355 ANY * const ss = proto_perl->Isavestack;
12356 const I32 max = proto_perl->Isavestack_max;
12357 I32 ix = proto_perl->Isavestack_ix;
12370 void (*dptr) (void*);
12371 void (*dxptr) (pTHX_ void*);
12373 PERL_ARGS_ASSERT_SS_DUP;
12375 Newxz(nss, max, ANY);
12378 const UV uv = POPUV(ss,ix);
12379 const U8 type = (U8)uv & SAVE_MASK;
12381 TOPUV(nss,ix) = uv;
12383 case SAVEt_CLEARSV:
12385 case SAVEt_HELEM: /* hash element */
12386 sv = (const SV *)POPPTR(ss,ix);
12387 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12389 case SAVEt_ITEM: /* normal string */
12390 case SAVEt_GVSV: /* scalar slot in GV */
12391 case SAVEt_SV: /* scalar reference */
12392 sv = (const SV *)POPPTR(ss,ix);
12393 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12396 case SAVEt_MORTALIZESV:
12397 sv = (const SV *)POPPTR(ss,ix);
12398 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12400 case SAVEt_SHARED_PVREF: /* char* in shared space */
12401 c = (char*)POPPTR(ss,ix);
12402 TOPPTR(nss,ix) = savesharedpv(c);
12403 ptr = POPPTR(ss,ix);
12404 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12406 case SAVEt_GENERIC_SVREF: /* generic sv */
12407 case SAVEt_SVREF: /* scalar reference */
12408 sv = (const SV *)POPPTR(ss,ix);
12409 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12410 ptr = POPPTR(ss,ix);
12411 TOPPTR(nss,ix) = svp_dup_inc((SV**)ptr, proto_perl);/* XXXXX */
12413 case SAVEt_HV: /* hash reference */
12414 case SAVEt_AV: /* array reference */
12415 sv = (const SV *) POPPTR(ss,ix);
12416 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12418 case SAVEt_COMPPAD:
12420 sv = (const SV *) POPPTR(ss,ix);
12421 TOPPTR(nss,ix) = sv_dup(sv, param);
12423 case SAVEt_INT: /* int reference */
12424 ptr = POPPTR(ss,ix);
12425 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12426 intval = (int)POPINT(ss,ix);
12427 TOPINT(nss,ix) = intval;
12429 case SAVEt_LONG: /* long reference */
12430 ptr = POPPTR(ss,ix);
12431 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12432 longval = (long)POPLONG(ss,ix);
12433 TOPLONG(nss,ix) = longval;
12435 case SAVEt_I32: /* I32 reference */
12436 ptr = POPPTR(ss,ix);
12437 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12439 TOPINT(nss,ix) = i;
12441 case SAVEt_IV: /* IV reference */
12442 ptr = POPPTR(ss,ix);
12443 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12445 TOPIV(nss,ix) = iv;
12447 case SAVEt_HPTR: /* HV* reference */
12448 case SAVEt_APTR: /* AV* reference */
12449 case SAVEt_SPTR: /* SV* reference */
12450 ptr = POPPTR(ss,ix);
12451 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12452 sv = (const SV *)POPPTR(ss,ix);
12453 TOPPTR(nss,ix) = sv_dup(sv, param);
12455 case SAVEt_VPTR: /* random* reference */
12456 ptr = POPPTR(ss,ix);
12457 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12459 case SAVEt_INT_SMALL:
12460 case SAVEt_I32_SMALL:
12461 case SAVEt_I16: /* I16 reference */
12462 case SAVEt_I8: /* I8 reference */
12464 ptr = POPPTR(ss,ix);
12465 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12467 case SAVEt_GENERIC_PVREF: /* generic char* */
12468 case SAVEt_PPTR: /* char* reference */
12469 ptr = POPPTR(ss,ix);
12470 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12471 c = (char*)POPPTR(ss,ix);
12472 TOPPTR(nss,ix) = pv_dup(c);
12474 case SAVEt_GP: /* scalar reference */
12475 gp = (GP*)POPPTR(ss,ix);
12476 TOPPTR(nss,ix) = gp = gp_dup(gp, param);
12477 (void)GpREFCNT_inc(gp);
12478 gv = (const GV *)POPPTR(ss,ix);
12479 TOPPTR(nss,ix) = gv_dup_inc(gv, param);
12482 ptr = POPPTR(ss,ix);
12483 if (ptr && (((OP*)ptr)->op_private & OPpREFCOUNTED)) {
12484 /* these are assumed to be refcounted properly */
12486 switch (((OP*)ptr)->op_type) {
12488 case OP_LEAVESUBLV:
12492 case OP_LEAVEWRITE:
12493 TOPPTR(nss,ix) = ptr;
12496 (void) OpREFCNT_inc(o);
12500 TOPPTR(nss,ix) = NULL;
12505 TOPPTR(nss,ix) = NULL;
12507 case SAVEt_FREECOPHH:
12508 ptr = POPPTR(ss,ix);
12509 TOPPTR(nss,ix) = cophh_copy((COPHH *)ptr);
12512 hv = (const HV *)POPPTR(ss,ix);
12513 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
12515 TOPINT(nss,ix) = i;
12518 c = (char*)POPPTR(ss,ix);
12519 TOPPTR(nss,ix) = pv_dup_inc(c);
12521 case SAVEt_STACK_POS: /* Position on Perl stack */
12523 TOPINT(nss,ix) = i;
12525 case SAVEt_DESTRUCTOR:
12526 ptr = POPPTR(ss,ix);
12527 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
12528 dptr = POPDPTR(ss,ix);
12529 TOPDPTR(nss,ix) = DPTR2FPTR(void (*)(void*),
12530 any_dup(FPTR2DPTR(void *, dptr),
12533 case SAVEt_DESTRUCTOR_X:
12534 ptr = POPPTR(ss,ix);
12535 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
12536 dxptr = POPDXPTR(ss,ix);
12537 TOPDXPTR(nss,ix) = DPTR2FPTR(void (*)(pTHX_ void*),
12538 any_dup(FPTR2DPTR(void *, dxptr),
12541 case SAVEt_REGCONTEXT:
12543 ix -= uv >> SAVE_TIGHT_SHIFT;
12545 case SAVEt_AELEM: /* array element */
12546 sv = (const SV *)POPPTR(ss,ix);
12547 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12549 TOPINT(nss,ix) = i;
12550 av = (const AV *)POPPTR(ss,ix);
12551 TOPPTR(nss,ix) = av_dup_inc(av, param);
12554 ptr = POPPTR(ss,ix);
12555 TOPPTR(nss,ix) = ptr;
12558 ptr = POPPTR(ss,ix);
12559 ptr = cophh_copy((COPHH*)ptr);
12560 TOPPTR(nss,ix) = ptr;
12562 TOPINT(nss,ix) = i;
12563 if (i & HINT_LOCALIZE_HH) {
12564 hv = (const HV *)POPPTR(ss,ix);
12565 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
12568 case SAVEt_PADSV_AND_MORTALIZE:
12569 longval = (long)POPLONG(ss,ix);
12570 TOPLONG(nss,ix) = longval;
12571 ptr = POPPTR(ss,ix);
12572 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12573 sv = (const SV *)POPPTR(ss,ix);
12574 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12576 case SAVEt_SET_SVFLAGS:
12578 TOPINT(nss,ix) = i;
12580 TOPINT(nss,ix) = i;
12581 sv = (const SV *)POPPTR(ss,ix);
12582 TOPPTR(nss,ix) = sv_dup(sv, param);
12584 case SAVEt_RE_STATE:
12586 const struct re_save_state *const old_state
12587 = (struct re_save_state *)
12588 (ss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
12589 struct re_save_state *const new_state
12590 = (struct re_save_state *)
12591 (nss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
12593 Copy(old_state, new_state, 1, struct re_save_state);
12594 ix -= SAVESTACK_ALLOC_FOR_RE_SAVE_STATE;
12596 new_state->re_state_bostr
12597 = pv_dup(old_state->re_state_bostr);
12598 new_state->re_state_reginput
12599 = pv_dup(old_state->re_state_reginput);
12600 new_state->re_state_regeol
12601 = pv_dup(old_state->re_state_regeol);
12602 new_state->re_state_regoffs
12603 = (regexp_paren_pair*)
12604 any_dup(old_state->re_state_regoffs, proto_perl);
12605 new_state->re_state_reglastparen
12606 = (U32*) any_dup(old_state->re_state_reglastparen,
12608 new_state->re_state_reglastcloseparen
12609 = (U32*)any_dup(old_state->re_state_reglastcloseparen,
12611 /* XXX This just has to be broken. The old save_re_context
12612 code did SAVEGENERICPV(PL_reg_start_tmp);
12613 PL_reg_start_tmp is char **.
12614 Look above to what the dup code does for
12615 SAVEt_GENERIC_PVREF
12616 It can never have worked.
12617 So this is merely a faithful copy of the exiting bug: */
12618 new_state->re_state_reg_start_tmp
12619 = (char **) pv_dup((char *)
12620 old_state->re_state_reg_start_tmp);
12621 /* I assume that it only ever "worked" because no-one called
12622 (pseudo)fork while the regexp engine had re-entered itself.
12624 #ifdef PERL_OLD_COPY_ON_WRITE
12625 new_state->re_state_nrs
12626 = sv_dup(old_state->re_state_nrs, param);
12628 new_state->re_state_reg_magic
12629 = (MAGIC*) any_dup(old_state->re_state_reg_magic,
12631 new_state->re_state_reg_oldcurpm
12632 = (PMOP*) any_dup(old_state->re_state_reg_oldcurpm,
12634 new_state->re_state_reg_curpm
12635 = (PMOP*) any_dup(old_state->re_state_reg_curpm,
12637 new_state->re_state_reg_oldsaved
12638 = pv_dup(old_state->re_state_reg_oldsaved);
12639 new_state->re_state_reg_poscache
12640 = pv_dup(old_state->re_state_reg_poscache);
12641 new_state->re_state_reg_starttry
12642 = pv_dup(old_state->re_state_reg_starttry);
12645 case SAVEt_COMPILE_WARNINGS:
12646 ptr = POPPTR(ss,ix);
12647 TOPPTR(nss,ix) = DUP_WARNINGS((STRLEN*)ptr);
12650 ptr = POPPTR(ss,ix);
12651 TOPPTR(nss,ix) = parser_dup((const yy_parser*)ptr, param);
12655 "panic: ss_dup inconsistency (%"IVdf")", (IV) type);
12663 /* if sv is a stash, call $class->CLONE_SKIP(), and set the SVphv_CLONEABLE
12664 * flag to the result. This is done for each stash before cloning starts,
12665 * so we know which stashes want their objects cloned */
12668 do_mark_cloneable_stash(pTHX_ SV *const sv)
12670 const HEK * const hvname = HvNAME_HEK((const HV *)sv);
12672 GV* const cloner = gv_fetchmethod_autoload(MUTABLE_HV(sv), "CLONE_SKIP", 0);
12673 SvFLAGS(sv) |= SVphv_CLONEABLE; /* clone objects by default */
12674 if (cloner && GvCV(cloner)) {
12681 mXPUSHs(newSVhek(hvname));
12683 call_sv(MUTABLE_SV(GvCV(cloner)), G_SCALAR);
12690 SvFLAGS(sv) &= ~SVphv_CLONEABLE;
12698 =for apidoc perl_clone
12700 Create and return a new interpreter by cloning the current one.
12702 perl_clone takes these flags as parameters:
12704 CLONEf_COPY_STACKS - is used to, well, copy the stacks also,
12705 without it we only clone the data and zero the stacks,
12706 with it we copy the stacks and the new perl interpreter is
12707 ready to run at the exact same point as the previous one.
12708 The pseudo-fork code uses COPY_STACKS while the
12709 threads->create doesn't.
12711 CLONEf_KEEP_PTR_TABLE
12712 perl_clone keeps a ptr_table with the pointer of the old
12713 variable as a key and the new variable as a value,
12714 this allows it to check if something has been cloned and not
12715 clone it again but rather just use the value and increase the
12716 refcount. If KEEP_PTR_TABLE is not set then perl_clone will kill
12717 the ptr_table using the function
12718 C<ptr_table_free(PL_ptr_table); PL_ptr_table = NULL;>,
12719 reason to keep it around is if you want to dup some of your own
12720 variable who are outside the graph perl scans, example of this
12721 code is in threads.xs create
12724 This is a win32 thing, it is ignored on unix, it tells perls
12725 win32host code (which is c++) to clone itself, this is needed on
12726 win32 if you want to run two threads at the same time,
12727 if you just want to do some stuff in a separate perl interpreter
12728 and then throw it away and return to the original one,
12729 you don't need to do anything.
12734 /* XXX the above needs expanding by someone who actually understands it ! */
12735 EXTERN_C PerlInterpreter *
12736 perl_clone_host(PerlInterpreter* proto_perl, UV flags);
12739 perl_clone(PerlInterpreter *proto_perl, UV flags)
12742 #ifdef PERL_IMPLICIT_SYS
12744 PERL_ARGS_ASSERT_PERL_CLONE;
12746 /* perlhost.h so we need to call into it
12747 to clone the host, CPerlHost should have a c interface, sky */
12749 if (flags & CLONEf_CLONE_HOST) {
12750 return perl_clone_host(proto_perl,flags);
12752 return perl_clone_using(proto_perl, flags,
12754 proto_perl->IMemShared,
12755 proto_perl->IMemParse,
12757 proto_perl->IStdIO,
12761 proto_perl->IProc);
12765 perl_clone_using(PerlInterpreter *proto_perl, UV flags,
12766 struct IPerlMem* ipM, struct IPerlMem* ipMS,
12767 struct IPerlMem* ipMP, struct IPerlEnv* ipE,
12768 struct IPerlStdIO* ipStd, struct IPerlLIO* ipLIO,
12769 struct IPerlDir* ipD, struct IPerlSock* ipS,
12770 struct IPerlProc* ipP)
12772 /* XXX many of the string copies here can be optimized if they're
12773 * constants; they need to be allocated as common memory and just
12774 * their pointers copied. */
12777 CLONE_PARAMS clone_params;
12778 CLONE_PARAMS* const param = &clone_params;
12780 PerlInterpreter * const my_perl = (PerlInterpreter*)(*ipM->pMalloc)(ipM, sizeof(PerlInterpreter));
12782 PERL_ARGS_ASSERT_PERL_CLONE_USING;
12783 #else /* !PERL_IMPLICIT_SYS */
12785 CLONE_PARAMS clone_params;
12786 CLONE_PARAMS* param = &clone_params;
12787 PerlInterpreter * const my_perl = (PerlInterpreter*)PerlMem_malloc(sizeof(PerlInterpreter));
12789 PERL_ARGS_ASSERT_PERL_CLONE;
12790 #endif /* PERL_IMPLICIT_SYS */
12792 /* for each stash, determine whether its objects should be cloned */
12793 S_visit(proto_perl, do_mark_cloneable_stash, SVt_PVHV, SVTYPEMASK);
12794 PERL_SET_THX(my_perl);
12797 PoisonNew(my_perl, 1, PerlInterpreter);
12800 PL_defstash = NULL; /* may be used by perl malloc() */
12803 PL_scopestack_name = 0;
12805 PL_savestack_ix = 0;
12806 PL_savestack_max = -1;
12807 PL_sig_pending = 0;
12809 Zero(&PL_debug_pad, 1, struct perl_debug_pad);
12810 # ifdef DEBUG_LEAKING_SCALARS
12811 PL_sv_serial = (((UV)my_perl >> 2) & 0xfff) * 1000000;
12813 #else /* !DEBUGGING */
12814 Zero(my_perl, 1, PerlInterpreter);
12815 #endif /* DEBUGGING */
12817 #ifdef PERL_IMPLICIT_SYS
12818 /* host pointers */
12820 PL_MemShared = ipMS;
12821 PL_MemParse = ipMP;
12828 #endif /* PERL_IMPLICIT_SYS */
12830 param->flags = flags;
12831 /* Nothing in the core code uses this, but we make it available to
12832 extensions (using mg_dup). */
12833 param->proto_perl = proto_perl;
12834 /* Likely nothing will use this, but it is initialised to be consistent
12835 with Perl_clone_params_new(). */
12836 param->new_perl = my_perl;
12837 param->unreferenced = NULL;
12839 INIT_TRACK_MEMPOOL(my_perl->Imemory_debug_header, my_perl);
12841 PL_body_arenas = NULL;
12842 Zero(&PL_body_roots, 1, PL_body_roots);
12845 PL_sv_objcount = 0;
12847 PL_sv_arenaroot = NULL;
12849 PL_debug = proto_perl->Idebug;
12851 PL_hash_seed = proto_perl->Ihash_seed;
12852 PL_rehash_seed = proto_perl->Irehash_seed;
12854 SvANY(&PL_sv_undef) = NULL;
12855 SvREFCNT(&PL_sv_undef) = (~(U32)0)/2;
12856 SvFLAGS(&PL_sv_undef) = SVf_READONLY|SVt_NULL;
12857 SvREFCNT(&PL_sv_no) = (~(U32)0)/2;
12858 SvFLAGS(&PL_sv_no) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
12859 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
12861 SvANY(&PL_sv_yes) = new_XPVNV();
12862 SvREFCNT(&PL_sv_yes) = (~(U32)0)/2;
12863 SvFLAGS(&PL_sv_yes) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
12864 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
12866 /* dbargs array probably holds garbage */
12869 PL_compiling = proto_perl->Icompiling;
12871 #ifdef PERL_DEBUG_READONLY_OPS
12876 /* pseudo environmental stuff */
12877 PL_origargc = proto_perl->Iorigargc;
12878 PL_origargv = proto_perl->Iorigargv;
12880 /* Set tainting stuff before PerlIO_debug can possibly get called */
12881 PL_tainting = proto_perl->Itainting;
12882 PL_taint_warn = proto_perl->Itaint_warn;
12884 PL_minus_c = proto_perl->Iminus_c;
12886 PL_localpatches = proto_perl->Ilocalpatches;
12887 PL_splitstr = proto_perl->Isplitstr;
12888 PL_minus_n = proto_perl->Iminus_n;
12889 PL_minus_p = proto_perl->Iminus_p;
12890 PL_minus_l = proto_perl->Iminus_l;
12891 PL_minus_a = proto_perl->Iminus_a;
12892 PL_minus_E = proto_perl->Iminus_E;
12893 PL_minus_F = proto_perl->Iminus_F;
12894 PL_doswitches = proto_perl->Idoswitches;
12895 PL_dowarn = proto_perl->Idowarn;
12896 PL_sawampersand = proto_perl->Isawampersand;
12897 PL_unsafe = proto_perl->Iunsafe;
12898 PL_perldb = proto_perl->Iperldb;
12899 PL_perl_destruct_level = proto_perl->Iperl_destruct_level;
12900 PL_exit_flags = proto_perl->Iexit_flags;
12902 /* XXX time(&PL_basetime) when asked for? */
12903 PL_basetime = proto_perl->Ibasetime;
12905 PL_maxsysfd = proto_perl->Imaxsysfd;
12906 PL_statusvalue = proto_perl->Istatusvalue;
12908 PL_statusvalue_vms = proto_perl->Istatusvalue_vms;
12910 PL_statusvalue_posix = proto_perl->Istatusvalue_posix;
12913 /* RE engine related */
12914 Zero(&PL_reg_state, 1, struct re_save_state);
12915 PL_reginterp_cnt = 0;
12916 PL_regmatch_slab = NULL;
12918 PL_sub_generation = proto_perl->Isub_generation;
12920 /* funky return mechanisms */
12921 PL_forkprocess = proto_perl->Iforkprocess;
12923 /* internal state */
12924 PL_maxo = proto_perl->Imaxo;
12926 PL_main_start = proto_perl->Imain_start;
12927 PL_eval_root = proto_perl->Ieval_root;
12928 PL_eval_start = proto_perl->Ieval_start;
12930 PL_filemode = proto_perl->Ifilemode;
12931 PL_lastfd = proto_perl->Ilastfd;
12932 PL_oldname = proto_perl->Ioldname; /* XXX not quite right */
12935 PL_gensym = proto_perl->Igensym;
12937 PL_laststatval = proto_perl->Ilaststatval;
12938 PL_laststype = proto_perl->Ilaststype;
12941 PL_profiledata = NULL;
12943 PL_generation = proto_perl->Igeneration;
12945 PL_in_clean_objs = proto_perl->Iin_clean_objs;
12946 PL_in_clean_all = proto_perl->Iin_clean_all;
12948 PL_uid = proto_perl->Iuid;
12949 PL_euid = proto_perl->Ieuid;
12950 PL_gid = proto_perl->Igid;
12951 PL_egid = proto_perl->Iegid;
12952 PL_nomemok = proto_perl->Inomemok;
12953 PL_an = proto_perl->Ian;
12954 PL_evalseq = proto_perl->Ievalseq;
12955 PL_origenviron = proto_perl->Iorigenviron; /* XXX not quite right */
12956 PL_origalen = proto_perl->Iorigalen;
12958 PL_sighandlerp = proto_perl->Isighandlerp;
12960 PL_runops = proto_perl->Irunops;
12962 PL_subline = proto_perl->Isubline;
12965 PL_cryptseen = proto_perl->Icryptseen;
12968 PL_hints = proto_perl->Ihints;
12970 PL_amagic_generation = proto_perl->Iamagic_generation;
12972 #ifdef USE_LOCALE_COLLATE
12973 PL_collation_ix = proto_perl->Icollation_ix;
12974 PL_collation_standard = proto_perl->Icollation_standard;
12975 PL_collxfrm_base = proto_perl->Icollxfrm_base;
12976 PL_collxfrm_mult = proto_perl->Icollxfrm_mult;
12977 #endif /* USE_LOCALE_COLLATE */
12979 #ifdef USE_LOCALE_NUMERIC
12980 PL_numeric_standard = proto_perl->Inumeric_standard;
12981 PL_numeric_local = proto_perl->Inumeric_local;
12982 #endif /* !USE_LOCALE_NUMERIC */
12984 /* Did the locale setup indicate UTF-8? */
12985 PL_utf8locale = proto_perl->Iutf8locale;
12986 /* Unicode features (see perlrun/-C) */
12987 PL_unicode = proto_perl->Iunicode;
12989 /* Pre-5.8 signals control */
12990 PL_signals = proto_perl->Isignals;
12992 /* times() ticks per second */
12993 PL_clocktick = proto_perl->Iclocktick;
12995 /* Recursion stopper for PerlIO_find_layer */
12996 PL_in_load_module = proto_perl->Iin_load_module;
12998 /* sort() routine */
12999 PL_sort_RealCmp = proto_perl->Isort_RealCmp;
13001 /* Not really needed/useful since the reenrant_retint is "volatile",
13002 * but do it for consistency's sake. */
13003 PL_reentrant_retint = proto_perl->Ireentrant_retint;
13005 /* Hooks to shared SVs and locks. */
13006 PL_sharehook = proto_perl->Isharehook;
13007 PL_lockhook = proto_perl->Ilockhook;
13008 PL_unlockhook = proto_perl->Iunlockhook;
13009 PL_threadhook = proto_perl->Ithreadhook;
13010 PL_destroyhook = proto_perl->Idestroyhook;
13011 PL_signalhook = proto_perl->Isignalhook;
13013 #ifdef THREADS_HAVE_PIDS
13014 PL_ppid = proto_perl->Ippid;
13018 PL_last_swash_hv = NULL; /* reinits on demand */
13019 PL_last_swash_klen = 0;
13020 PL_last_swash_key[0]= '\0';
13021 PL_last_swash_tmps = (U8*)NULL;
13022 PL_last_swash_slen = 0;
13024 PL_glob_index = proto_perl->Iglob_index;
13025 PL_srand_called = proto_perl->Isrand_called;
13027 if (flags & CLONEf_COPY_STACKS) {
13028 /* next allocation will be PL_tmps_stack[PL_tmps_ix+1] */
13029 PL_tmps_ix = proto_perl->Itmps_ix;
13030 PL_tmps_max = proto_perl->Itmps_max;
13031 PL_tmps_floor = proto_perl->Itmps_floor;
13033 /* next push_scope()/ENTER sets PL_scopestack[PL_scopestack_ix]
13034 * NOTE: unlike the others! */
13035 PL_scopestack_ix = proto_perl->Iscopestack_ix;
13036 PL_scopestack_max = proto_perl->Iscopestack_max;
13038 /* next SSPUSHFOO() sets PL_savestack[PL_savestack_ix]
13039 * NOTE: unlike the others! */
13040 PL_savestack_ix = proto_perl->Isavestack_ix;
13041 PL_savestack_max = proto_perl->Isavestack_max;
13044 PL_start_env = proto_perl->Istart_env; /* XXXXXX */
13045 PL_top_env = &PL_start_env;
13047 PL_op = proto_perl->Iop;
13050 PL_Xpv = (XPV*)NULL;
13051 my_perl->Ina = proto_perl->Ina;
13053 PL_statbuf = proto_perl->Istatbuf;
13054 PL_statcache = proto_perl->Istatcache;
13057 PL_timesbuf = proto_perl->Itimesbuf;
13060 PL_tainted = proto_perl->Itainted;
13061 PL_curpm = proto_perl->Icurpm; /* XXX No PMOP ref count */
13063 PL_chopset = proto_perl->Ichopset; /* XXX never deallocated */
13065 PL_restartjmpenv = proto_perl->Irestartjmpenv;
13066 PL_restartop = proto_perl->Irestartop;
13067 PL_in_eval = proto_perl->Iin_eval;
13068 PL_delaymagic = proto_perl->Idelaymagic;
13069 PL_phase = proto_perl->Iphase;
13070 PL_localizing = proto_perl->Ilocalizing;
13072 PL_hv_fetch_ent_mh = NULL;
13073 PL_modcount = proto_perl->Imodcount;
13074 PL_lastgotoprobe = NULL;
13075 PL_dumpindent = proto_perl->Idumpindent;
13077 PL_efloatbuf = NULL; /* reinits on demand */
13078 PL_efloatsize = 0; /* reinits on demand */
13082 PL_regdummy = proto_perl->Iregdummy;
13083 PL_colorset = 0; /* reinits PL_colors[] */
13084 /*PL_colors[6] = {0,0,0,0,0,0};*/
13086 /* Pluggable optimizer */
13087 PL_peepp = proto_perl->Ipeepp;
13088 PL_rpeepp = proto_perl->Irpeepp;
13089 /* op_free() hook */
13090 PL_opfreehook = proto_perl->Iopfreehook;
13092 #ifdef USE_REENTRANT_API
13093 /* XXX: things like -Dm will segfault here in perlio, but doing
13094 * PERL_SET_CONTEXT(proto_perl);
13095 * breaks too many other things
13097 Perl_reentrant_init(aTHX);
13100 /* create SV map for pointer relocation */
13101 PL_ptr_table = ptr_table_new();
13103 /* initialize these special pointers as early as possible */
13104 ptr_table_store(PL_ptr_table, &proto_perl->Isv_undef, &PL_sv_undef);
13106 SvANY(&PL_sv_no) = new_XPVNV();
13107 SvPV_set(&PL_sv_no, savepvn(PL_No, 0));
13108 SvCUR_set(&PL_sv_no, 0);
13109 SvLEN_set(&PL_sv_no, 1);
13110 SvIV_set(&PL_sv_no, 0);
13111 SvNV_set(&PL_sv_no, 0);
13112 ptr_table_store(PL_ptr_table, &proto_perl->Isv_no, &PL_sv_no);
13114 SvPV_set(&PL_sv_yes, savepvn(PL_Yes, 1));
13115 SvCUR_set(&PL_sv_yes, 1);
13116 SvLEN_set(&PL_sv_yes, 2);
13117 SvIV_set(&PL_sv_yes, 1);
13118 SvNV_set(&PL_sv_yes, 1);
13119 ptr_table_store(PL_ptr_table, &proto_perl->Isv_yes, &PL_sv_yes);
13121 /* create (a non-shared!) shared string table */
13122 PL_strtab = newHV();
13123 HvSHAREKEYS_off(PL_strtab);
13124 hv_ksplit(PL_strtab, HvTOTALKEYS(proto_perl->Istrtab));
13125 ptr_table_store(PL_ptr_table, proto_perl->Istrtab, PL_strtab);
13127 /* These two PVs will be free'd special way so must set them same way op.c does */
13128 PL_compiling.cop_stashpv = savesharedpv(PL_compiling.cop_stashpv);
13129 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_stashpv, PL_compiling.cop_stashpv);
13131 PL_compiling.cop_file = savesharedpv(PL_compiling.cop_file);
13132 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_file, PL_compiling.cop_file);
13134 ptr_table_store(PL_ptr_table, &proto_perl->Icompiling, &PL_compiling);
13135 PL_compiling.cop_warnings = DUP_WARNINGS(PL_compiling.cop_warnings);
13136 CopHINTHASH_set(&PL_compiling, cophh_copy(CopHINTHASH_get(&PL_compiling)));
13137 PL_curcop = (COP*)any_dup(proto_perl->Icurcop, proto_perl);
13139 param->stashes = newAV(); /* Setup array of objects to call clone on */
13140 /* This makes no difference to the implementation, as it always pushes
13141 and shifts pointers to other SVs without changing their reference
13142 count, with the array becoming empty before it is freed. However, it
13143 makes it conceptually clear what is going on, and will avoid some
13144 work inside av.c, filling slots between AvFILL() and AvMAX() with
13145 &PL_sv_undef, and SvREFCNT_dec()ing those. */
13146 AvREAL_off(param->stashes);
13148 if (!(flags & CLONEf_COPY_STACKS)) {
13149 param->unreferenced = newAV();
13152 #ifdef PERLIO_LAYERS
13153 /* Clone PerlIO tables as soon as we can handle general xx_dup() */
13154 PerlIO_clone(aTHX_ proto_perl, param);
13157 PL_envgv = gv_dup(proto_perl->Ienvgv, param);
13158 PL_incgv = gv_dup(proto_perl->Iincgv, param);
13159 PL_hintgv = gv_dup(proto_perl->Ihintgv, param);
13160 PL_origfilename = SAVEPV(proto_perl->Iorigfilename);
13161 PL_diehook = sv_dup_inc(proto_perl->Idiehook, param);
13162 PL_warnhook = sv_dup_inc(proto_perl->Iwarnhook, param);
13165 PL_patchlevel = sv_dup_inc(proto_perl->Ipatchlevel, param);
13166 PL_apiversion = sv_dup_inc(proto_perl->Iapiversion, param);
13167 PL_inplace = SAVEPV(proto_perl->Iinplace);
13168 PL_e_script = sv_dup_inc(proto_perl->Ie_script, param);
13170 /* magical thingies */
13171 PL_formfeed = sv_dup(proto_perl->Iformfeed, param);
13173 PL_encoding = sv_dup(proto_perl->Iencoding, param);
13175 sv_setpvs(PERL_DEBUG_PAD(0), ""); /* For regex debugging. */
13176 sv_setpvs(PERL_DEBUG_PAD(1), ""); /* ext/re needs these */
13177 sv_setpvs(PERL_DEBUG_PAD(2), ""); /* even without DEBUGGING. */
13180 /* Clone the regex array */
13181 /* ORANGE FIXME for plugins, probably in the SV dup code.
13182 newSViv(PTR2IV(CALLREGDUPE(
13183 INT2PTR(REGEXP *, SvIVX(regex)), param))))
13185 PL_regex_padav = av_dup_inc(proto_perl->Iregex_padav, param);
13186 PL_regex_pad = AvARRAY(PL_regex_padav);
13188 /* shortcuts to various I/O objects */
13189 PL_ofsgv = gv_dup_inc(proto_perl->Iofsgv, param);
13190 PL_stdingv = gv_dup(proto_perl->Istdingv, param);
13191 PL_stderrgv = gv_dup(proto_perl->Istderrgv, param);
13192 PL_defgv = gv_dup(proto_perl->Idefgv, param);
13193 PL_argvgv = gv_dup(proto_perl->Iargvgv, param);
13194 PL_argvoutgv = gv_dup(proto_perl->Iargvoutgv, param);
13195 PL_argvout_stack = av_dup_inc(proto_perl->Iargvout_stack, param);
13197 /* shortcuts to regexp stuff */
13198 PL_replgv = gv_dup(proto_perl->Ireplgv, param);
13200 /* shortcuts to misc objects */
13201 PL_errgv = gv_dup(proto_perl->Ierrgv, param);
13203 /* shortcuts to debugging objects */
13204 PL_DBgv = gv_dup(proto_perl->IDBgv, param);
13205 PL_DBline = gv_dup(proto_perl->IDBline, param);
13206 PL_DBsub = gv_dup(proto_perl->IDBsub, param);
13207 PL_DBsingle = sv_dup(proto_perl->IDBsingle, param);
13208 PL_DBtrace = sv_dup(proto_perl->IDBtrace, param);
13209 PL_DBsignal = sv_dup(proto_perl->IDBsignal, param);
13211 /* symbol tables */
13212 PL_defstash = hv_dup_inc(proto_perl->Idefstash, param);
13213 PL_curstash = hv_dup(proto_perl->Icurstash, param);
13214 PL_debstash = hv_dup(proto_perl->Idebstash, param);
13215 PL_globalstash = hv_dup(proto_perl->Iglobalstash, param);
13216 PL_curstname = sv_dup_inc(proto_perl->Icurstname, param);
13218 PL_beginav = av_dup_inc(proto_perl->Ibeginav, param);
13219 PL_beginav_save = av_dup_inc(proto_perl->Ibeginav_save, param);
13220 PL_checkav_save = av_dup_inc(proto_perl->Icheckav_save, param);
13221 PL_unitcheckav = av_dup_inc(proto_perl->Iunitcheckav, param);
13222 PL_unitcheckav_save = av_dup_inc(proto_perl->Iunitcheckav_save, param);
13223 PL_endav = av_dup_inc(proto_perl->Iendav, param);
13224 PL_checkav = av_dup_inc(proto_perl->Icheckav, param);
13225 PL_initav = av_dup_inc(proto_perl->Iinitav, param);
13227 PL_isarev = hv_dup_inc(proto_perl->Iisarev, param);
13229 /* subprocess state */
13230 PL_fdpid = av_dup_inc(proto_perl->Ifdpid, param);
13232 if (proto_perl->Iop_mask)
13233 PL_op_mask = SAVEPVN(proto_perl->Iop_mask, PL_maxo);
13236 /* PL_asserting = proto_perl->Iasserting; */
13238 /* current interpreter roots */
13239 PL_main_cv = cv_dup_inc(proto_perl->Imain_cv, param);
13241 PL_main_root = OpREFCNT_inc(proto_perl->Imain_root);
13244 /* runtime control stuff */
13245 PL_curcopdb = (COP*)any_dup(proto_perl->Icurcopdb, proto_perl);
13247 PL_preambleav = av_dup_inc(proto_perl->Ipreambleav, param);
13249 PL_ors_sv = sv_dup_inc(proto_perl->Iors_sv, param);
13251 /* interpreter atexit processing */
13252 PL_exitlistlen = proto_perl->Iexitlistlen;
13253 if (PL_exitlistlen) {
13254 Newx(PL_exitlist, PL_exitlistlen, PerlExitListEntry);
13255 Copy(proto_perl->Iexitlist, PL_exitlist, PL_exitlistlen, PerlExitListEntry);
13258 PL_exitlist = (PerlExitListEntry*)NULL;
13260 PL_my_cxt_size = proto_perl->Imy_cxt_size;
13261 if (PL_my_cxt_size) {
13262 Newx(PL_my_cxt_list, PL_my_cxt_size, void *);
13263 Copy(proto_perl->Imy_cxt_list, PL_my_cxt_list, PL_my_cxt_size, void *);
13264 #ifdef PERL_GLOBAL_STRUCT_PRIVATE
13265 Newx(PL_my_cxt_keys, PL_my_cxt_size, const char *);
13266 Copy(proto_perl->Imy_cxt_keys, PL_my_cxt_keys, PL_my_cxt_size, char *);
13270 PL_my_cxt_list = (void**)NULL;
13271 #ifdef PERL_GLOBAL_STRUCT_PRIVATE
13272 PL_my_cxt_keys = (const char**)NULL;
13275 PL_modglobal = hv_dup_inc(proto_perl->Imodglobal, param);
13276 PL_custom_op_names = hv_dup_inc(proto_perl->Icustom_op_names,param);
13277 PL_custom_op_descs = hv_dup_inc(proto_perl->Icustom_op_descs,param);
13278 PL_custom_ops = hv_dup_inc(proto_perl->Icustom_ops, param);
13280 PL_compcv = cv_dup(proto_perl->Icompcv, param);
13282 PAD_CLONE_VARS(proto_perl, param);
13284 #ifdef HAVE_INTERP_INTERN
13285 sys_intern_dup(&proto_perl->Isys_intern, &PL_sys_intern);
13288 PL_DBcv = cv_dup(proto_perl->IDBcv, param);
13290 #ifdef PERL_USES_PL_PIDSTATUS
13291 PL_pidstatus = newHV(); /* XXX flag for cloning? */
13293 PL_osname = SAVEPV(proto_perl->Iosname);
13294 PL_parser = parser_dup(proto_perl->Iparser, param);
13296 /* XXX this only works if the saved cop has already been cloned */
13297 if (proto_perl->Iparser) {
13298 PL_parser->saved_curcop = (COP*)any_dup(
13299 proto_perl->Iparser->saved_curcop,
13303 PL_subname = sv_dup_inc(proto_perl->Isubname, param);
13305 #ifdef USE_LOCALE_COLLATE
13306 PL_collation_name = SAVEPV(proto_perl->Icollation_name);
13307 #endif /* USE_LOCALE_COLLATE */
13309 #ifdef USE_LOCALE_NUMERIC
13310 PL_numeric_name = SAVEPV(proto_perl->Inumeric_name);
13311 PL_numeric_radix_sv = sv_dup_inc(proto_perl->Inumeric_radix_sv, param);
13312 #endif /* !USE_LOCALE_NUMERIC */
13314 /* utf8 character classes */
13315 PL_utf8_alnum = sv_dup_inc(proto_perl->Iutf8_alnum, param);
13316 PL_utf8_alpha = sv_dup_inc(proto_perl->Iutf8_alpha, param);
13317 PL_utf8_space = sv_dup_inc(proto_perl->Iutf8_space, param);
13318 PL_utf8_graph = sv_dup_inc(proto_perl->Iutf8_graph, param);
13319 PL_utf8_digit = sv_dup_inc(proto_perl->Iutf8_digit, param);
13320 PL_utf8_upper = sv_dup_inc(proto_perl->Iutf8_upper, param);
13321 PL_utf8_lower = sv_dup_inc(proto_perl->Iutf8_lower, param);
13322 PL_utf8_print = sv_dup_inc(proto_perl->Iutf8_print, param);
13323 PL_utf8_punct = sv_dup_inc(proto_perl->Iutf8_punct, param);
13324 PL_utf8_xdigit = sv_dup_inc(proto_perl->Iutf8_xdigit, param);
13325 PL_utf8_mark = sv_dup_inc(proto_perl->Iutf8_mark, param);
13326 PL_utf8_X_begin = sv_dup_inc(proto_perl->Iutf8_X_begin, param);
13327 PL_utf8_X_extend = sv_dup_inc(proto_perl->Iutf8_X_extend, param);
13328 PL_utf8_X_prepend = sv_dup_inc(proto_perl->Iutf8_X_prepend, param);
13329 PL_utf8_X_non_hangul = sv_dup_inc(proto_perl->Iutf8_X_non_hangul, param);
13330 PL_utf8_X_L = sv_dup_inc(proto_perl->Iutf8_X_L, param);
13331 PL_utf8_X_LV = sv_dup_inc(proto_perl->Iutf8_X_LV, param);
13332 PL_utf8_X_LVT = sv_dup_inc(proto_perl->Iutf8_X_LVT, param);
13333 PL_utf8_X_T = sv_dup_inc(proto_perl->Iutf8_X_T, param);
13334 PL_utf8_X_V = sv_dup_inc(proto_perl->Iutf8_X_V, param);
13335 PL_utf8_X_LV_LVT_V = sv_dup_inc(proto_perl->Iutf8_X_LV_LVT_V, param);
13336 PL_utf8_toupper = sv_dup_inc(proto_perl->Iutf8_toupper, param);
13337 PL_utf8_totitle = sv_dup_inc(proto_perl->Iutf8_totitle, param);
13338 PL_utf8_tolower = sv_dup_inc(proto_perl->Iutf8_tolower, param);
13339 PL_utf8_tofold = sv_dup_inc(proto_perl->Iutf8_tofold, param);
13340 PL_utf8_idstart = sv_dup_inc(proto_perl->Iutf8_idstart, param);
13341 PL_utf8_xidstart = sv_dup_inc(proto_perl->Iutf8_xidstart, param);
13342 PL_utf8_perl_idstart = sv_dup_inc(proto_perl->Iutf8_perl_idstart, param);
13343 PL_utf8_idcont = sv_dup_inc(proto_perl->Iutf8_idcont, param);
13344 PL_utf8_xidcont = sv_dup_inc(proto_perl->Iutf8_xidcont, param);
13345 PL_utf8_foldable = sv_dup_inc(proto_perl->Iutf8_foldable, param);
13348 if (proto_perl->Ipsig_pend) {
13349 Newxz(PL_psig_pend, SIG_SIZE, int);
13352 PL_psig_pend = (int*)NULL;
13355 if (proto_perl->Ipsig_name) {
13356 Newx(PL_psig_name, 2 * SIG_SIZE, SV*);
13357 sv_dup_inc_multiple(proto_perl->Ipsig_name, PL_psig_name, 2 * SIG_SIZE,
13359 PL_psig_ptr = PL_psig_name + SIG_SIZE;
13362 PL_psig_ptr = (SV**)NULL;
13363 PL_psig_name = (SV**)NULL;
13366 if (flags & CLONEf_COPY_STACKS) {
13367 Newx(PL_tmps_stack, PL_tmps_max, SV*);
13368 sv_dup_inc_multiple(proto_perl->Itmps_stack, PL_tmps_stack,
13369 PL_tmps_ix+1, param);
13371 /* next PUSHMARK() sets *(PL_markstack_ptr+1) */
13372 i = proto_perl->Imarkstack_max - proto_perl->Imarkstack;
13373 Newxz(PL_markstack, i, I32);
13374 PL_markstack_max = PL_markstack + (proto_perl->Imarkstack_max
13375 - proto_perl->Imarkstack);
13376 PL_markstack_ptr = PL_markstack + (proto_perl->Imarkstack_ptr
13377 - proto_perl->Imarkstack);
13378 Copy(proto_perl->Imarkstack, PL_markstack,
13379 PL_markstack_ptr - PL_markstack + 1, I32);
13381 /* next push_scope()/ENTER sets PL_scopestack[PL_scopestack_ix]
13382 * NOTE: unlike the others! */
13383 Newxz(PL_scopestack, PL_scopestack_max, I32);
13384 Copy(proto_perl->Iscopestack, PL_scopestack, PL_scopestack_ix, I32);
13387 Newxz(PL_scopestack_name, PL_scopestack_max, const char *);
13388 Copy(proto_perl->Iscopestack_name, PL_scopestack_name, PL_scopestack_ix, const char *);
13390 /* NOTE: si_dup() looks at PL_markstack */
13391 PL_curstackinfo = si_dup(proto_perl->Icurstackinfo, param);
13393 /* PL_curstack = PL_curstackinfo->si_stack; */
13394 PL_curstack = av_dup(proto_perl->Icurstack, param);
13395 PL_mainstack = av_dup(proto_perl->Imainstack, param);
13397 /* next PUSHs() etc. set *(PL_stack_sp+1) */
13398 PL_stack_base = AvARRAY(PL_curstack);
13399 PL_stack_sp = PL_stack_base + (proto_perl->Istack_sp
13400 - proto_perl->Istack_base);
13401 PL_stack_max = PL_stack_base + AvMAX(PL_curstack);
13403 /*Newxz(PL_savestack, PL_savestack_max, ANY);*/
13404 PL_savestack = ss_dup(proto_perl, param);
13408 ENTER; /* perl_destruct() wants to LEAVE; */
13411 PL_statgv = gv_dup(proto_perl->Istatgv, param);
13412 PL_statname = sv_dup_inc(proto_perl->Istatname, param);
13414 PL_rs = sv_dup_inc(proto_perl->Irs, param);
13415 PL_last_in_gv = gv_dup(proto_perl->Ilast_in_gv, param);
13416 PL_defoutgv = gv_dup_inc(proto_perl->Idefoutgv, param);
13417 PL_toptarget = sv_dup_inc(proto_perl->Itoptarget, param);
13418 PL_bodytarget = sv_dup_inc(proto_perl->Ibodytarget, param);
13419 PL_formtarget = sv_dup(proto_perl->Iformtarget, param);
13421 PL_errors = sv_dup_inc(proto_perl->Ierrors, param);
13423 PL_sortcop = (OP*)any_dup(proto_perl->Isortcop, proto_perl);
13424 PL_sortstash = hv_dup(proto_perl->Isortstash, param);
13425 PL_firstgv = gv_dup(proto_perl->Ifirstgv, param);
13426 PL_secondgv = gv_dup(proto_perl->Isecondgv, param);
13428 PL_stashcache = newHV();
13430 PL_watchaddr = (char **) ptr_table_fetch(PL_ptr_table,
13431 proto_perl->Iwatchaddr);
13432 PL_watchok = PL_watchaddr ? * PL_watchaddr : NULL;
13433 if (PL_debug && PL_watchaddr) {
13434 PerlIO_printf(Perl_debug_log,
13435 "WATCHING: %"UVxf" cloned as %"UVxf" with value %"UVxf"\n",
13436 PTR2UV(proto_perl->Iwatchaddr), PTR2UV(PL_watchaddr),
13437 PTR2UV(PL_watchok));
13440 PL_registered_mros = hv_dup_inc(proto_perl->Iregistered_mros, param);
13441 PL_blockhooks = av_dup_inc(proto_perl->Iblockhooks, param);
13442 PL_utf8_foldclosures = hv_dup_inc(proto_perl->Iutf8_foldclosures, param);
13444 /* Call the ->CLONE method, if it exists, for each of the stashes
13445 identified by sv_dup() above.
13447 while(av_len(param->stashes) != -1) {
13448 HV* const stash = MUTABLE_HV(av_shift(param->stashes));
13449 GV* const cloner = gv_fetchmethod_autoload(stash, "CLONE", 0);
13450 if (cloner && GvCV(cloner)) {
13455 mXPUSHs(newSVhek(HvNAME_HEK(stash)));
13457 call_sv(MUTABLE_SV(GvCV(cloner)), G_DISCARD);
13463 if (!(flags & CLONEf_KEEP_PTR_TABLE)) {
13464 ptr_table_free(PL_ptr_table);
13465 PL_ptr_table = NULL;
13468 if (!(flags & CLONEf_COPY_STACKS)) {
13469 unreferenced_to_tmp_stack(param->unreferenced);
13472 SvREFCNT_dec(param->stashes);
13474 /* orphaned? eg threads->new inside BEGIN or use */
13475 if (PL_compcv && ! SvREFCNT(PL_compcv)) {
13476 SvREFCNT_inc_simple_void(PL_compcv);
13477 SAVEFREESV(PL_compcv);
13484 S_unreferenced_to_tmp_stack(pTHX_ AV *const unreferenced)
13486 PERL_ARGS_ASSERT_UNREFERENCED_TO_TMP_STACK;
13488 if (AvFILLp(unreferenced) > -1) {
13489 SV **svp = AvARRAY(unreferenced);
13490 SV **const last = svp + AvFILLp(unreferenced);
13494 if (SvREFCNT(*svp) == 1)
13496 } while (++svp <= last);
13498 EXTEND_MORTAL(count);
13499 svp = AvARRAY(unreferenced);
13502 if (SvREFCNT(*svp) == 1) {
13503 /* Our reference is the only one to this SV. This means that
13504 in this thread, the scalar effectively has a 0 reference.
13505 That doesn't work (cleanup never happens), so donate our
13506 reference to it onto the save stack. */
13507 PL_tmps_stack[++PL_tmps_ix] = *svp;
13509 /* As an optimisation, because we are already walking the
13510 entire array, instead of above doing either
13511 SvREFCNT_inc(*svp) or *svp = &PL_sv_undef, we can instead
13512 release our reference to the scalar, so that at the end of
13513 the array owns zero references to the scalars it happens to
13514 point to. We are effectively converting the array from
13515 AvREAL() on to AvREAL() off. This saves the av_clear()
13516 (triggered by the SvREFCNT_dec(unreferenced) below) from
13517 walking the array a second time. */
13518 SvREFCNT_dec(*svp);
13521 } while (++svp <= last);
13522 AvREAL_off(unreferenced);
13524 SvREFCNT_dec(unreferenced);
13528 Perl_clone_params_del(CLONE_PARAMS *param)
13530 /* This seemingly funky ordering keeps the build with PERL_GLOBAL_STRUCT
13532 PerlInterpreter *const to = param->new_perl;
13534 PerlInterpreter *const was = PERL_GET_THX;
13536 PERL_ARGS_ASSERT_CLONE_PARAMS_DEL;
13542 SvREFCNT_dec(param->stashes);
13543 if (param->unreferenced)
13544 unreferenced_to_tmp_stack(param->unreferenced);
13554 Perl_clone_params_new(PerlInterpreter *const from, PerlInterpreter *const to)
13557 /* Need to play this game, as newAV() can call safesysmalloc(), and that
13558 does a dTHX; to get the context from thread local storage.
13559 FIXME - under PERL_CORE Newx(), Safefree() and friends should expand to
13560 a version that passes in my_perl. */
13561 PerlInterpreter *const was = PERL_GET_THX;
13562 CLONE_PARAMS *param;
13564 PERL_ARGS_ASSERT_CLONE_PARAMS_NEW;
13570 /* Given that we've set the context, we can do this unshared. */
13571 Newx(param, 1, CLONE_PARAMS);
13574 param->proto_perl = from;
13575 param->new_perl = to;
13576 param->stashes = (AV *)Perl_newSV_type(to, SVt_PVAV);
13577 AvREAL_off(param->stashes);
13578 param->unreferenced = (AV *)Perl_newSV_type(to, SVt_PVAV);
13586 #endif /* USE_ITHREADS */
13589 =head1 Unicode Support
13591 =for apidoc sv_recode_to_utf8
13593 The encoding is assumed to be an Encode object, on entry the PV
13594 of the sv is assumed to be octets in that encoding, and the sv
13595 will be converted into Unicode (and UTF-8).
13597 If the sv already is UTF-8 (or if it is not POK), or if the encoding
13598 is not a reference, nothing is done to the sv. If the encoding is not
13599 an C<Encode::XS> Encoding object, bad things will happen.
13600 (See F<lib/encoding.pm> and L<Encode>).
13602 The PV of the sv is returned.
13607 Perl_sv_recode_to_utf8(pTHX_ SV *sv, SV *encoding)
13611 PERL_ARGS_ASSERT_SV_RECODE_TO_UTF8;
13613 if (SvPOK(sv) && !SvUTF8(sv) && !IN_BYTES && SvROK(encoding)) {
13627 Passing sv_yes is wrong - it needs to be or'ed set of constants
13628 for Encode::XS, while UTf-8 decode (currently) assumes a true value means
13629 remove converted chars from source.
13631 Both will default the value - let them.
13633 XPUSHs(&PL_sv_yes);
13636 call_method("decode", G_SCALAR);
13640 s = SvPV_const(uni, len);
13641 if (s != SvPVX_const(sv)) {
13642 SvGROW(sv, len + 1);
13643 Move(s, SvPVX(sv), len + 1, char);
13644 SvCUR_set(sv, len);
13648 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
13649 /* clear pos and any utf8 cache */
13650 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
13653 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
13654 magic_setutf8(sv,mg); /* clear UTF8 cache */
13659 return SvPOKp(sv) ? SvPVX(sv) : NULL;
13663 =for apidoc sv_cat_decode
13665 The encoding is assumed to be an Encode object, the PV of the ssv is
13666 assumed to be octets in that encoding and decoding the input starts
13667 from the position which (PV + *offset) pointed to. The dsv will be
13668 concatenated the decoded UTF-8 string from ssv. Decoding will terminate
13669 when the string tstr appears in decoding output or the input ends on
13670 the PV of the ssv. The value which the offset points will be modified
13671 to the last input position on the ssv.
13673 Returns TRUE if the terminator was found, else returns FALSE.
13678 Perl_sv_cat_decode(pTHX_ SV *dsv, SV *encoding,
13679 SV *ssv, int *offset, char *tstr, int tlen)
13684 PERL_ARGS_ASSERT_SV_CAT_DECODE;
13686 if (SvPOK(ssv) && SvPOK(dsv) && SvROK(encoding) && offset) {
13697 offsv = newSViv(*offset);
13699 mXPUSHp(tstr, tlen);
13701 call_method("cat_decode", G_SCALAR);
13703 ret = SvTRUE(TOPs);
13704 *offset = SvIV(offsv);
13710 Perl_croak(aTHX_ "Invalid argument to sv_cat_decode");
13715 /* ---------------------------------------------------------------------
13717 * support functions for report_uninit()
13720 /* the maxiumum size of array or hash where we will scan looking
13721 * for the undefined element that triggered the warning */
13723 #define FUV_MAX_SEARCH_SIZE 1000
13725 /* Look for an entry in the hash whose value has the same SV as val;
13726 * If so, return a mortal copy of the key. */
13729 S_find_hash_subscript(pTHX_ const HV *const hv, const SV *const val)
13732 register HE **array;
13735 PERL_ARGS_ASSERT_FIND_HASH_SUBSCRIPT;
13737 if (!hv || SvMAGICAL(hv) || !HvARRAY(hv) ||
13738 (HvTOTALKEYS(hv) > FUV_MAX_SEARCH_SIZE))
13741 array = HvARRAY(hv);
13743 for (i=HvMAX(hv); i>0; i--) {
13744 register HE *entry;
13745 for (entry = array[i]; entry; entry = HeNEXT(entry)) {
13746 if (HeVAL(entry) != val)
13748 if ( HeVAL(entry) == &PL_sv_undef ||
13749 HeVAL(entry) == &PL_sv_placeholder)
13753 if (HeKLEN(entry) == HEf_SVKEY)
13754 return sv_mortalcopy(HeKEY_sv(entry));
13755 return sv_2mortal(newSVhek(HeKEY_hek(entry)));
13761 /* Look for an entry in the array whose value has the same SV as val;
13762 * If so, return the index, otherwise return -1. */
13765 S_find_array_subscript(pTHX_ const AV *const av, const SV *const val)
13769 PERL_ARGS_ASSERT_FIND_ARRAY_SUBSCRIPT;
13771 if (!av || SvMAGICAL(av) || !AvARRAY(av) ||
13772 (AvFILLp(av) > FUV_MAX_SEARCH_SIZE))
13775 if (val != &PL_sv_undef) {
13776 SV ** const svp = AvARRAY(av);
13779 for (i=AvFILLp(av); i>=0; i--)
13786 /* S_varname(): return the name of a variable, optionally with a subscript.
13787 * If gv is non-zero, use the name of that global, along with gvtype (one
13788 * of "$", "@", "%"); otherwise use the name of the lexical at pad offset
13789 * targ. Depending on the value of the subscript_type flag, return:
13792 #define FUV_SUBSCRIPT_NONE 1 /* "@foo" */
13793 #define FUV_SUBSCRIPT_ARRAY 2 /* "$foo[aindex]" */
13794 #define FUV_SUBSCRIPT_HASH 3 /* "$foo{keyname}" */
13795 #define FUV_SUBSCRIPT_WITHIN 4 /* "within @foo" */
13798 S_varname(pTHX_ const GV *const gv, const char gvtype, PADOFFSET targ,
13799 const SV *const keyname, I32 aindex, int subscript_type)
13802 SV * const name = sv_newmortal();
13805 buffer[0] = gvtype;
13808 /* as gv_fullname4(), but add literal '^' for $^FOO names */
13810 gv_fullname4(name, gv, buffer, 0);
13812 if ((unsigned int)SvPVX(name)[1] <= 26) {
13814 buffer[1] = SvPVX(name)[1] + 'A' - 1;
13816 /* Swap the 1 unprintable control character for the 2 byte pretty
13817 version - ie substr($name, 1, 1) = $buffer; */
13818 sv_insert(name, 1, 1, buffer, 2);
13822 CV * const cv = find_runcv(NULL);
13826 if (!cv || !CvPADLIST(cv))
13828 av = MUTABLE_AV((*av_fetch(CvPADLIST(cv), 0, FALSE)));
13829 sv = *av_fetch(av, targ, FALSE);
13830 sv_setsv(name, sv);
13833 if (subscript_type == FUV_SUBSCRIPT_HASH) {
13834 SV * const sv = newSV(0);
13835 *SvPVX(name) = '$';
13836 Perl_sv_catpvf(aTHX_ name, "{%s}",
13837 pv_display(sv,SvPVX_const(keyname), SvCUR(keyname), 0, 32));
13840 else if (subscript_type == FUV_SUBSCRIPT_ARRAY) {
13841 *SvPVX(name) = '$';
13842 Perl_sv_catpvf(aTHX_ name, "[%"IVdf"]", (IV)aindex);
13844 else if (subscript_type == FUV_SUBSCRIPT_WITHIN) {
13845 /* We know that name has no magic, so can use 0 instead of SV_GMAGIC */
13846 Perl_sv_insert_flags(aTHX_ name, 0, 0, STR_WITH_LEN("within "), 0);
13854 =for apidoc find_uninit_var
13856 Find the name of the undefined variable (if any) that caused the operator o
13857 to issue a "Use of uninitialized value" warning.
13858 If match is true, only return a name if it's value matches uninit_sv.
13859 So roughly speaking, if a unary operator (such as OP_COS) generates a
13860 warning, then following the direct child of the op may yield an
13861 OP_PADSV or OP_GV that gives the name of the undefined variable. On the
13862 other hand, with OP_ADD there are two branches to follow, so we only print
13863 the variable name if we get an exact match.
13865 The name is returned as a mortal SV.
13867 Assumes that PL_op is the op that originally triggered the error, and that
13868 PL_comppad/PL_curpad points to the currently executing pad.
13874 S_find_uninit_var(pTHX_ const OP *const obase, const SV *const uninit_sv,
13880 const OP *o, *o2, *kid;
13882 if (!obase || (match && (!uninit_sv || uninit_sv == &PL_sv_undef ||
13883 uninit_sv == &PL_sv_placeholder)))
13886 switch (obase->op_type) {
13893 const bool pad = (obase->op_type == OP_PADAV || obase->op_type == OP_PADHV);
13894 const bool hash = (obase->op_type == OP_PADHV || obase->op_type == OP_RV2HV);
13897 int subscript_type = FUV_SUBSCRIPT_WITHIN;
13899 if (pad) { /* @lex, %lex */
13900 sv = PAD_SVl(obase->op_targ);
13904 if (cUNOPx(obase)->op_first->op_type == OP_GV) {
13905 /* @global, %global */
13906 gv = cGVOPx_gv(cUNOPx(obase)->op_first);
13909 sv = hash ? MUTABLE_SV(GvHV(gv)): MUTABLE_SV(GvAV(gv));
13911 else /* @{expr}, %{expr} */
13912 return find_uninit_var(cUNOPx(obase)->op_first,
13916 /* attempt to find a match within the aggregate */
13918 keysv = find_hash_subscript((const HV*)sv, uninit_sv);
13920 subscript_type = FUV_SUBSCRIPT_HASH;
13923 index = find_array_subscript((const AV *)sv, uninit_sv);
13925 subscript_type = FUV_SUBSCRIPT_ARRAY;
13928 if (match && subscript_type == FUV_SUBSCRIPT_WITHIN)
13931 return varname(gv, hash ? '%' : '@', obase->op_targ,
13932 keysv, index, subscript_type);
13936 if (cUNOPx(obase)->op_first->op_type == OP_GV) {
13938 gv = cGVOPx_gv(cUNOPx(obase)->op_first);
13939 if (!gv || !GvSTASH(gv))
13941 if (match && (GvSV(gv) != uninit_sv))
13943 return varname(gv, '$', 0, NULL, 0, FUV_SUBSCRIPT_NONE);
13946 return find_uninit_var(cUNOPx(obase)->op_first, uninit_sv, 1);
13949 if (match && PAD_SVl(obase->op_targ) != uninit_sv)
13951 return varname(NULL, '$', obase->op_targ,
13952 NULL, 0, FUV_SUBSCRIPT_NONE);
13955 gv = cGVOPx_gv(obase);
13956 if (!gv || (match && GvSV(gv) != uninit_sv) || !GvSTASH(gv))
13958 return varname(gv, '$', 0, NULL, 0, FUV_SUBSCRIPT_NONE);
13960 case OP_AELEMFAST_LEX:
13963 AV *av = MUTABLE_AV(PAD_SV(obase->op_targ));
13964 if (!av || SvRMAGICAL(av))
13966 svp = av_fetch(av, (I32)obase->op_private, FALSE);
13967 if (!svp || *svp != uninit_sv)
13970 return varname(NULL, '$', obase->op_targ,
13971 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
13974 gv = cGVOPx_gv(obase);
13979 AV *const av = GvAV(gv);
13980 if (!av || SvRMAGICAL(av))
13982 svp = av_fetch(av, (I32)obase->op_private, FALSE);
13983 if (!svp || *svp != uninit_sv)
13986 return varname(gv, '$', 0,
13987 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
13992 o = cUNOPx(obase)->op_first;
13993 if (!o || o->op_type != OP_NULL ||
13994 ! (o->op_targ == OP_AELEM || o->op_targ == OP_HELEM))
13996 return find_uninit_var(cBINOPo->op_last, uninit_sv, match);
14001 bool negate = FALSE;
14003 if (PL_op == obase)
14004 /* $a[uninit_expr] or $h{uninit_expr} */
14005 return find_uninit_var(cBINOPx(obase)->op_last, uninit_sv, match);
14008 o = cBINOPx(obase)->op_first;
14009 kid = cBINOPx(obase)->op_last;
14011 /* get the av or hv, and optionally the gv */
14013 if (o->op_type == OP_PADAV || o->op_type == OP_PADHV) {
14014 sv = PAD_SV(o->op_targ);
14016 else if ((o->op_type == OP_RV2AV || o->op_type == OP_RV2HV)
14017 && cUNOPo->op_first->op_type == OP_GV)
14019 gv = cGVOPx_gv(cUNOPo->op_first);
14023 == OP_RV2HV ? MUTABLE_SV(GvHV(gv)) : MUTABLE_SV(GvAV(gv));
14028 if (kid && kid->op_type == OP_NEGATE) {
14030 kid = cUNOPx(kid)->op_first;
14033 if (kid && kid->op_type == OP_CONST && SvOK(cSVOPx_sv(kid))) {
14034 /* index is constant */
14037 kidsv = sv_2mortal(newSVpvs("-"));
14038 sv_catsv(kidsv, cSVOPx_sv(kid));
14041 kidsv = cSVOPx_sv(kid);
14045 if (obase->op_type == OP_HELEM) {
14046 HE* he = hv_fetch_ent(MUTABLE_HV(sv), kidsv, 0, 0);
14047 if (!he || HeVAL(he) != uninit_sv)
14051 SV * const * const svp = av_fetch(MUTABLE_AV(sv),
14052 negate ? - SvIV(cSVOPx_sv(kid)) : SvIV(cSVOPx_sv(kid)),
14054 if (!svp || *svp != uninit_sv)
14058 if (obase->op_type == OP_HELEM)
14059 return varname(gv, '%', o->op_targ,
14060 kidsv, 0, FUV_SUBSCRIPT_HASH);
14062 return varname(gv, '@', o->op_targ, NULL,
14063 negate ? - SvIV(cSVOPx_sv(kid)) : SvIV(cSVOPx_sv(kid)),
14064 FUV_SUBSCRIPT_ARRAY);
14067 /* index is an expression;
14068 * attempt to find a match within the aggregate */
14069 if (obase->op_type == OP_HELEM) {
14070 SV * const keysv = find_hash_subscript((const HV*)sv, uninit_sv);
14072 return varname(gv, '%', o->op_targ,
14073 keysv, 0, FUV_SUBSCRIPT_HASH);
14077 = find_array_subscript((const AV *)sv, uninit_sv);
14079 return varname(gv, '@', o->op_targ,
14080 NULL, index, FUV_SUBSCRIPT_ARRAY);
14085 (o->op_type == OP_PADAV || o->op_type == OP_RV2AV)
14087 o->op_targ, NULL, 0, FUV_SUBSCRIPT_WITHIN);
14093 /* only examine RHS */
14094 return find_uninit_var(cBINOPx(obase)->op_first, uninit_sv, match);
14097 o = cUNOPx(obase)->op_first;
14098 if (o->op_type == OP_PUSHMARK)
14101 if (!o->op_sibling) {
14102 /* one-arg version of open is highly magical */
14104 if (o->op_type == OP_GV) { /* open FOO; */
14106 if (match && GvSV(gv) != uninit_sv)
14108 return varname(gv, '$', 0,
14109 NULL, 0, FUV_SUBSCRIPT_NONE);
14111 /* other possibilities not handled are:
14112 * open $x; or open my $x; should return '${*$x}'
14113 * open expr; should return '$'.expr ideally
14119 /* ops where $_ may be an implicit arg */
14123 if ( !(obase->op_flags & OPf_STACKED)) {
14124 if (uninit_sv == ((obase->op_private & OPpTARGET_MY)
14125 ? PAD_SVl(obase->op_targ)
14128 sv = sv_newmortal();
14129 sv_setpvs(sv, "$_");
14138 match = 1; /* print etc can return undef on defined args */
14139 /* skip filehandle as it can't produce 'undef' warning */
14140 o = cUNOPx(obase)->op_first;
14141 if ((obase->op_flags & OPf_STACKED) && o->op_type == OP_PUSHMARK)
14142 o = o->op_sibling->op_sibling;
14146 case OP_ENTEREVAL: /* could be eval $undef or $x='$undef'; eval $x */
14147 case OP_CUSTOM: /* XS or custom code could trigger random warnings */
14149 /* the following ops are capable of returning PL_sv_undef even for
14150 * defined arg(s) */
14169 case OP_GETPEERNAME:
14217 case OP_SMARTMATCH:
14226 /* XXX tmp hack: these two may call an XS sub, and currently
14227 XS subs don't have a SUB entry on the context stack, so CV and
14228 pad determination goes wrong, and BAD things happen. So, just
14229 don't try to determine the value under those circumstances.
14230 Need a better fix at dome point. DAPM 11/2007 */
14236 GV * const gv = gv_fetchpvs(".", GV_NOTQUAL, SVt_PV);
14237 if (gv && GvSV(gv) == uninit_sv)
14238 return newSVpvs_flags("$.", SVs_TEMP);
14243 /* def-ness of rval pos() is independent of the def-ness of its arg */
14244 if ( !(obase->op_flags & OPf_MOD))
14249 if (SvROK(PL_rs) && uninit_sv == SvRV(PL_rs))
14250 return newSVpvs_flags("${$/}", SVs_TEMP);
14255 if (!(obase->op_flags & OPf_KIDS))
14257 o = cUNOPx(obase)->op_first;
14263 /* if all except one arg are constant, or have no side-effects,
14264 * or are optimized away, then it's unambiguous */
14266 for (kid=o; kid; kid = kid->op_sibling) {
14268 const OPCODE type = kid->op_type;
14269 if ( (type == OP_CONST && SvOK(cSVOPx_sv(kid)))
14270 || (type == OP_NULL && ! (kid->op_flags & OPf_KIDS))
14271 || (type == OP_PUSHMARK)
14273 /* @$a and %$a, but not @a or %a */
14274 (type == OP_RV2AV || type == OP_RV2HV)
14275 && cUNOPx(kid)->op_first
14276 && cUNOPx(kid)->op_first->op_type != OP_GV
14281 if (o2) { /* more than one found */
14288 return find_uninit_var(o2, uninit_sv, match);
14290 /* scan all args */
14292 sv = find_uninit_var(o, uninit_sv, 1);
14304 =for apidoc report_uninit
14306 Print appropriate "Use of uninitialized variable" warning
14312 Perl_report_uninit(pTHX_ const SV *uninit_sv)
14316 SV* varname = NULL;
14317 if (uninit_sv && PL_curpad) {
14318 varname = find_uninit_var(PL_op, uninit_sv,0);
14320 sv_insert(varname, 0, 0, " ", 1);
14322 /* diag_listed_as: Use of uninitialized value%s */
14323 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit_sv,
14324 SVfARG(varname ? varname : &PL_sv_no),
14325 " in ", OP_DESC(PL_op));
14328 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
14334 * c-indentation-style: bsd
14335 * c-basic-offset: 4
14336 * indent-tabs-mode: t
14339 * ex: set ts=8 sts=4 sw=4 noet: