3 * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, 2000,
4 * 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009 by Larry Wall
7 * You may distribute under the terms of either the GNU General Public
8 * License or the Artistic License, as specified in the README file.
13 * 'I wonder what the Entish is for "yes" and "no",' he thought.
16 * [p.480 of _The Lord of the Rings_, III/iv: "Treebeard"]
22 * This file contains the code that creates, manipulates and destroys
23 * scalar values (SVs). The other types (AV, HV, GV, etc.) reuse the
24 * structure of an SV, so their creation and destruction is handled
25 * here; higher-level functions are in av.c, hv.c, and so on. Opcode
26 * level functions (eg. substr, split, join) for each of the types are
36 # if __STDC_VERSION__ >= 199901L && !defined(VMS)
47 /* Missing proto on LynxOS */
48 char *gconvert(double, int, int, char *);
51 #ifdef PERL_UTF8_CACHE_ASSERT
52 /* if adding more checks watch out for the following tests:
53 * t/op/index.t t/op/length.t t/op/pat.t t/op/substr.t
54 * lib/utf8.t lib/Unicode/Collate/t/index.t
57 # define ASSERT_UTF8_CACHE(cache) \
58 STMT_START { if (cache) { assert((cache)[0] <= (cache)[1]); \
59 assert((cache)[2] <= (cache)[3]); \
60 assert((cache)[3] <= (cache)[1]);} \
63 # define ASSERT_UTF8_CACHE(cache) NOOP
66 #ifdef PERL_OLD_COPY_ON_WRITE
67 #define SV_COW_NEXT_SV(sv) INT2PTR(SV *,SvUVX(sv))
68 #define SV_COW_NEXT_SV_SET(current,next) SvUV_set(current, PTR2UV(next))
69 /* This is a pessimistic view. Scalar must be purely a read-write PV to copy-
73 /* ============================================================================
75 =head1 Allocation and deallocation of SVs.
77 An SV (or AV, HV, etc.) is allocated in two parts: the head (struct
78 sv, av, hv...) contains type and reference count information, and for
79 many types, a pointer to the body (struct xrv, xpv, xpviv...), which
80 contains fields specific to each type. Some types store all they need
81 in the head, so don't have a body.
83 In all but the most memory-paranoid configurations (ex: PURIFY), heads
84 and bodies are allocated out of arenas, which by default are
85 approximately 4K chunks of memory parcelled up into N heads or bodies.
86 Sv-bodies are allocated by their sv-type, guaranteeing size
87 consistency needed to allocate safely from arrays.
89 For SV-heads, the first slot in each arena is reserved, and holds a
90 link to the next arena, some flags, and a note of the number of slots.
91 Snaked through each arena chain is a linked list of free items; when
92 this becomes empty, an extra arena is allocated and divided up into N
93 items which are threaded into the free list.
95 SV-bodies are similar, but they use arena-sets by default, which
96 separate the link and info from the arena itself, and reclaim the 1st
97 slot in the arena. SV-bodies are further described later.
99 The following global variables are associated with arenas:
101 PL_sv_arenaroot pointer to list of SV arenas
102 PL_sv_root pointer to list of free SV structures
104 PL_body_arenas head of linked-list of body arenas
105 PL_body_roots[] array of pointers to list of free bodies of svtype
106 arrays are indexed by the svtype needed
108 A few special SV heads are not allocated from an arena, but are
109 instead directly created in the interpreter structure, eg PL_sv_undef.
110 The size of arenas can be changed from the default by setting
111 PERL_ARENA_SIZE appropriately at compile time.
113 The SV arena serves the secondary purpose of allowing still-live SVs
114 to be located and destroyed during final cleanup.
116 At the lowest level, the macros new_SV() and del_SV() grab and free
117 an SV head. (If debugging with -DD, del_SV() calls the function S_del_sv()
118 to return the SV to the free list with error checking.) new_SV() calls
119 more_sv() / sv_add_arena() to add an extra arena if the free list is empty.
120 SVs in the free list have their SvTYPE field set to all ones.
122 At the time of very final cleanup, sv_free_arenas() is called from
123 perl_destruct() to physically free all the arenas allocated since the
124 start of the interpreter.
126 The function visit() scans the SV arenas list, and calls a specified
127 function for each SV it finds which is still live - ie which has an SvTYPE
128 other than all 1's, and a non-zero SvREFCNT. visit() is used by the
129 following functions (specified as [function that calls visit()] / [function
130 called by visit() for each SV]):
132 sv_report_used() / do_report_used()
133 dump all remaining SVs (debugging aid)
135 sv_clean_objs() / do_clean_objs(),do_clean_named_objs(),
136 do_clean_named_io_objs()
137 Attempt to free all objects pointed to by RVs,
138 and try to do the same for all objects indirectly
139 referenced by typeglobs too. Called once from
140 perl_destruct(), prior to calling sv_clean_all()
143 sv_clean_all() / do_clean_all()
144 SvREFCNT_dec(sv) each remaining SV, possibly
145 triggering an sv_free(). It also sets the
146 SVf_BREAK flag on the SV to indicate that the
147 refcnt has been artificially lowered, and thus
148 stopping sv_free() from giving spurious warnings
149 about SVs which unexpectedly have a refcnt
150 of zero. called repeatedly from perl_destruct()
151 until there are no SVs left.
153 =head2 Arena allocator API Summary
155 Private API to rest of sv.c
159 new_XPVNV(), del_XPVGV(),
164 sv_report_used(), sv_clean_objs(), sv_clean_all(), sv_free_arenas()
168 * ========================================================================= */
171 * "A time to plant, and a time to uproot what was planted..."
175 # define MEM_LOG_NEW_SV(sv, file, line, func) \
176 Perl_mem_log_new_sv(sv, file, line, func)
177 # define MEM_LOG_DEL_SV(sv, file, line, func) \
178 Perl_mem_log_del_sv(sv, file, line, func)
180 # define MEM_LOG_NEW_SV(sv, file, line, func) NOOP
181 # define MEM_LOG_DEL_SV(sv, file, line, func) NOOP
184 #ifdef DEBUG_LEAKING_SCALARS
185 # define FREE_SV_DEBUG_FILE(sv) Safefree((sv)->sv_debug_file)
186 # define DEBUG_SV_SERIAL(sv) \
187 DEBUG_m(PerlIO_printf(Perl_debug_log, "0x%"UVxf": (%05ld) del_SV\n", \
188 PTR2UV(sv), (long)(sv)->sv_debug_serial))
190 # define FREE_SV_DEBUG_FILE(sv)
191 # define DEBUG_SV_SERIAL(sv) NOOP
195 # define SvARENA_CHAIN(sv) ((sv)->sv_u.svu_rv)
196 # define SvARENA_CHAIN_SET(sv,val) (sv)->sv_u.svu_rv = MUTABLE_SV((val))
197 /* Whilst I'd love to do this, it seems that things like to check on
199 # define POSION_SV_HEAD(sv) PoisonNew(sv, 1, struct STRUCT_SV)
201 # define POSION_SV_HEAD(sv) PoisonNew(&SvANY(sv), 1, void *), \
202 PoisonNew(&SvREFCNT(sv), 1, U32)
204 # define SvARENA_CHAIN(sv) SvANY(sv)
205 # define SvARENA_CHAIN_SET(sv,val) SvANY(sv) = (void *)(val)
206 # define POSION_SV_HEAD(sv)
209 /* Mark an SV head as unused, and add to free list.
211 * If SVf_BREAK is set, skip adding it to the free list, as this SV had
212 * its refcount artificially decremented during global destruction, so
213 * there may be dangling pointers to it. The last thing we want in that
214 * case is for it to be reused. */
216 #define plant_SV(p) \
218 const U32 old_flags = SvFLAGS(p); \
219 MEM_LOG_DEL_SV(p, __FILE__, __LINE__, FUNCTION__); \
220 DEBUG_SV_SERIAL(p); \
221 FREE_SV_DEBUG_FILE(p); \
223 SvFLAGS(p) = SVTYPEMASK; \
224 if (!(old_flags & SVf_BREAK)) { \
225 SvARENA_CHAIN_SET(p, PL_sv_root); \
231 #define uproot_SV(p) \
234 PL_sv_root = MUTABLE_SV(SvARENA_CHAIN(p)); \
239 /* make some more SVs by adding another arena */
246 char *chunk; /* must use New here to match call to */
247 Newx(chunk,PERL_ARENA_SIZE,char); /* Safefree() in sv_free_arenas() */
248 sv_add_arena(chunk, PERL_ARENA_SIZE, 0);
253 /* new_SV(): return a new, empty SV head */
255 #ifdef DEBUG_LEAKING_SCALARS
256 /* provide a real function for a debugger to play with */
258 S_new_SV(pTHX_ const char *file, int line, const char *func)
265 sv = S_more_sv(aTHX);
269 sv->sv_debug_optype = PL_op ? PL_op->op_type : 0;
270 sv->sv_debug_line = (U16) (PL_parser && PL_parser->copline != NOLINE
276 sv->sv_debug_inpad = 0;
277 sv->sv_debug_parent = NULL;
278 sv->sv_debug_file = PL_curcop ? savepv(CopFILE(PL_curcop)): NULL;
280 sv->sv_debug_serial = PL_sv_serial++;
282 MEM_LOG_NEW_SV(sv, file, line, func);
283 DEBUG_m(PerlIO_printf(Perl_debug_log, "0x%"UVxf": (%05ld) new_SV (from %s:%d [%s])\n",
284 PTR2UV(sv), (long)sv->sv_debug_serial, file, line, func));
288 # define new_SV(p) (p)=S_new_SV(aTHX_ __FILE__, __LINE__, FUNCTION__)
296 (p) = S_more_sv(aTHX); \
300 MEM_LOG_NEW_SV(p, __FILE__, __LINE__, FUNCTION__); \
305 /* del_SV(): return an empty SV head to the free list */
318 S_del_sv(pTHX_ SV *p)
322 PERL_ARGS_ASSERT_DEL_SV;
327 for (sva = PL_sv_arenaroot; sva; sva = MUTABLE_SV(SvANY(sva))) {
328 const SV * const sv = sva + 1;
329 const SV * const svend = &sva[SvREFCNT(sva)];
330 if (p >= sv && p < svend) {
336 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL),
337 "Attempt to free non-arena SV: 0x%"UVxf
338 pTHX__FORMAT, PTR2UV(p) pTHX__VALUE);
345 #else /* ! DEBUGGING */
347 #define del_SV(p) plant_SV(p)
349 #endif /* DEBUGGING */
353 =head1 SV Manipulation Functions
355 =for apidoc sv_add_arena
357 Given a chunk of memory, link it to the head of the list of arenas,
358 and split it into a list of free SVs.
364 S_sv_add_arena(pTHX_ char *const ptr, const U32 size, const U32 flags)
367 SV *const sva = MUTABLE_SV(ptr);
371 PERL_ARGS_ASSERT_SV_ADD_ARENA;
373 /* The first SV in an arena isn't an SV. */
374 SvANY(sva) = (void *) PL_sv_arenaroot; /* ptr to next arena */
375 SvREFCNT(sva) = size / sizeof(SV); /* number of SV slots */
376 SvFLAGS(sva) = flags; /* FAKE if not to be freed */
378 PL_sv_arenaroot = sva;
379 PL_sv_root = sva + 1;
381 svend = &sva[SvREFCNT(sva) - 1];
384 SvARENA_CHAIN_SET(sv, (sv + 1));
388 /* Must always set typemask because it's always checked in on cleanup
389 when the arenas are walked looking for objects. */
390 SvFLAGS(sv) = SVTYPEMASK;
393 SvARENA_CHAIN_SET(sv, 0);
397 SvFLAGS(sv) = SVTYPEMASK;
400 /* visit(): call the named function for each non-free SV in the arenas
401 * whose flags field matches the flags/mask args. */
404 S_visit(pTHX_ SVFUNC_t f, const U32 flags, const U32 mask)
410 PERL_ARGS_ASSERT_VISIT;
412 for (sva = PL_sv_arenaroot; sva; sva = MUTABLE_SV(SvANY(sva))) {
413 register const SV * const svend = &sva[SvREFCNT(sva)];
415 for (sv = sva + 1; sv < svend; ++sv) {
416 if (SvTYPE(sv) != (svtype)SVTYPEMASK
417 && (sv->sv_flags & mask) == flags
430 /* called by sv_report_used() for each live SV */
433 do_report_used(pTHX_ SV *const sv)
435 if (SvTYPE(sv) != (svtype)SVTYPEMASK) {
436 PerlIO_printf(Perl_debug_log, "****\n");
443 =for apidoc sv_report_used
445 Dump the contents of all SVs not yet freed. (Debugging aid).
451 Perl_sv_report_used(pTHX)
454 visit(do_report_used, 0, 0);
460 /* called by sv_clean_objs() for each live SV */
463 do_clean_objs(pTHX_ SV *const ref)
468 SV * const target = SvRV(ref);
469 if (SvOBJECT(target)) {
470 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning object ref:\n "), sv_dump(ref)));
471 if (SvWEAKREF(ref)) {
472 sv_del_backref(target, ref);
478 SvREFCNT_dec(target);
483 /* XXX Might want to check arrays, etc. */
487 /* clear any slots in a GV which hold objects - except IO;
488 * called by sv_clean_objs() for each live GV */
491 do_clean_named_objs(pTHX_ SV *const sv)
495 assert(SvTYPE(sv) == SVt_PVGV);
496 assert(isGV_with_GP(sv));
500 /* freeing GP entries may indirectly free the current GV;
501 * hold onto it while we mess with the GP slots */
504 if ( ((obj = GvSV(sv) )) && SvOBJECT(obj)) {
505 DEBUG_D((PerlIO_printf(Perl_debug_log,
506 "Cleaning named glob SV object:\n "), sv_dump(obj)));
510 if ( ((obj = MUTABLE_SV(GvAV(sv)) )) && SvOBJECT(obj)) {
511 DEBUG_D((PerlIO_printf(Perl_debug_log,
512 "Cleaning named glob AV object:\n "), sv_dump(obj)));
516 if ( ((obj = MUTABLE_SV(GvHV(sv)) )) && SvOBJECT(obj)) {
517 DEBUG_D((PerlIO_printf(Perl_debug_log,
518 "Cleaning named glob HV object:\n "), sv_dump(obj)));
522 if ( ((obj = MUTABLE_SV(GvCV(sv)) )) && SvOBJECT(obj)) {
523 DEBUG_D((PerlIO_printf(Perl_debug_log,
524 "Cleaning named glob CV object:\n "), sv_dump(obj)));
528 SvREFCNT_dec(sv); /* undo the inc above */
531 /* clear any IO slots in a GV which hold objects (except stderr, defout);
532 * called by sv_clean_objs() for each live GV */
535 do_clean_named_io_objs(pTHX_ SV *const sv)
539 assert(SvTYPE(sv) == SVt_PVGV);
540 assert(isGV_with_GP(sv));
541 if (!GvGP(sv) || sv == (SV*)PL_stderrgv || sv == (SV*)PL_defoutgv)
545 if ( ((obj = MUTABLE_SV(GvIO(sv)) )) && SvOBJECT(obj)) {
546 DEBUG_D((PerlIO_printf(Perl_debug_log,
547 "Cleaning named glob IO object:\n "), sv_dump(obj)));
551 SvREFCNT_dec(sv); /* undo the inc above */
554 /* Void wrapper to pass to visit() */
556 do_curse(pTHX_ SV * const sv) {
557 if ((PL_stderrgv && GvGP(PL_stderrgv) && (SV*)GvIO(PL_stderrgv) == sv)
558 || (PL_defoutgv && GvGP(PL_defoutgv) && (SV*)GvIO(PL_defoutgv) == sv))
564 =for apidoc sv_clean_objs
566 Attempt to destroy all objects not yet freed
572 Perl_sv_clean_objs(pTHX)
576 PL_in_clean_objs = TRUE;
577 visit(do_clean_objs, SVf_ROK, SVf_ROK);
578 /* Some barnacles may yet remain, clinging to typeglobs.
579 * Run the non-IO destructors first: they may want to output
580 * error messages, close files etc */
581 visit(do_clean_named_objs, SVt_PVGV|SVpgv_GP, SVTYPEMASK|SVp_POK|SVpgv_GP);
582 visit(do_clean_named_io_objs, SVt_PVGV|SVpgv_GP, SVTYPEMASK|SVp_POK|SVpgv_GP);
583 /* And if there are some very tenacious barnacles clinging to arrays,
584 closures, or what have you.... */
585 visit(do_curse, SVs_OBJECT, SVs_OBJECT);
586 olddef = PL_defoutgv;
587 PL_defoutgv = NULL; /* disable skip of PL_defoutgv */
588 if (olddef && isGV_with_GP(olddef))
589 do_clean_named_io_objs(aTHX_ MUTABLE_SV(olddef));
590 olderr = PL_stderrgv;
591 PL_stderrgv = NULL; /* disable skip of PL_stderrgv */
592 if (olderr && isGV_with_GP(olderr))
593 do_clean_named_io_objs(aTHX_ MUTABLE_SV(olderr));
594 SvREFCNT_dec(olddef);
595 PL_in_clean_objs = FALSE;
598 /* called by sv_clean_all() for each live SV */
601 do_clean_all(pTHX_ SV *const sv)
604 if (sv == (const SV *) PL_fdpid || sv == (const SV *)PL_strtab) {
605 /* don't clean pid table and strtab */
608 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning loops: SV at 0x%"UVxf"\n", PTR2UV(sv)) ));
609 SvFLAGS(sv) |= SVf_BREAK;
614 =for apidoc sv_clean_all
616 Decrement the refcnt of each remaining SV, possibly triggering a
617 cleanup. This function may have to be called multiple times to free
618 SVs which are in complex self-referential hierarchies.
624 Perl_sv_clean_all(pTHX)
628 PL_in_clean_all = TRUE;
629 cleaned = visit(do_clean_all, 0,0);
634 ARENASETS: a meta-arena implementation which separates arena-info
635 into struct arena_set, which contains an array of struct
636 arena_descs, each holding info for a single arena. By separating
637 the meta-info from the arena, we recover the 1st slot, formerly
638 borrowed for list management. The arena_set is about the size of an
639 arena, avoiding the needless malloc overhead of a naive linked-list.
641 The cost is 1 arena-set malloc per ~320 arena-mallocs, + the unused
642 memory in the last arena-set (1/2 on average). In trade, we get
643 back the 1st slot in each arena (ie 1.7% of a CV-arena, less for
644 smaller types). The recovery of the wasted space allows use of
645 small arenas for large, rare body types, by changing array* fields
646 in body_details_by_type[] below.
649 char *arena; /* the raw storage, allocated aligned */
650 size_t size; /* its size ~4k typ */
651 svtype utype; /* bodytype stored in arena */
656 /* Get the maximum number of elements in set[] such that struct arena_set
657 will fit within PERL_ARENA_SIZE, which is probably just under 4K, and
658 therefore likely to be 1 aligned memory page. */
660 #define ARENAS_PER_SET ((PERL_ARENA_SIZE - sizeof(struct arena_set*) \
661 - 2 * sizeof(int)) / sizeof (struct arena_desc))
664 struct arena_set* next;
665 unsigned int set_size; /* ie ARENAS_PER_SET */
666 unsigned int curr; /* index of next available arena-desc */
667 struct arena_desc set[ARENAS_PER_SET];
671 =for apidoc sv_free_arenas
673 Deallocate the memory used by all arenas. Note that all the individual SV
674 heads and bodies within the arenas must already have been freed.
679 Perl_sv_free_arenas(pTHX)
686 /* Free arenas here, but be careful about fake ones. (We assume
687 contiguity of the fake ones with the corresponding real ones.) */
689 for (sva = PL_sv_arenaroot; sva; sva = svanext) {
690 svanext = MUTABLE_SV(SvANY(sva));
691 while (svanext && SvFAKE(svanext))
692 svanext = MUTABLE_SV(SvANY(svanext));
699 struct arena_set *aroot = (struct arena_set*) PL_body_arenas;
702 struct arena_set *current = aroot;
705 assert(aroot->set[i].arena);
706 Safefree(aroot->set[i].arena);
714 i = PERL_ARENA_ROOTS_SIZE;
716 PL_body_roots[i] = 0;
723 Here are mid-level routines that manage the allocation of bodies out
724 of the various arenas. There are 5 kinds of arenas:
726 1. SV-head arenas, which are discussed and handled above
727 2. regular body arenas
728 3. arenas for reduced-size bodies
731 Arena types 2 & 3 are chained by body-type off an array of
732 arena-root pointers, which is indexed by svtype. Some of the
733 larger/less used body types are malloced singly, since a large
734 unused block of them is wasteful. Also, several svtypes dont have
735 bodies; the data fits into the sv-head itself. The arena-root
736 pointer thus has a few unused root-pointers (which may be hijacked
737 later for arena types 4,5)
739 3 differs from 2 as an optimization; some body types have several
740 unused fields in the front of the structure (which are kept in-place
741 for consistency). These bodies can be allocated in smaller chunks,
742 because the leading fields arent accessed. Pointers to such bodies
743 are decremented to point at the unused 'ghost' memory, knowing that
744 the pointers are used with offsets to the real memory.
747 =head1 SV-Body Allocation
749 Allocation of SV-bodies is similar to SV-heads, differing as follows;
750 the allocation mechanism is used for many body types, so is somewhat
751 more complicated, it uses arena-sets, and has no need for still-live
754 At the outermost level, (new|del)_X*V macros return bodies of the
755 appropriate type. These macros call either (new|del)_body_type or
756 (new|del)_body_allocated macro pairs, depending on specifics of the
757 type. Most body types use the former pair, the latter pair is used to
758 allocate body types with "ghost fields".
760 "ghost fields" are fields that are unused in certain types, and
761 consequently don't need to actually exist. They are declared because
762 they're part of a "base type", which allows use of functions as
763 methods. The simplest examples are AVs and HVs, 2 aggregate types
764 which don't use the fields which support SCALAR semantics.
766 For these types, the arenas are carved up into appropriately sized
767 chunks, we thus avoid wasted memory for those unaccessed members.
768 When bodies are allocated, we adjust the pointer back in memory by the
769 size of the part not allocated, so it's as if we allocated the full
770 structure. (But things will all go boom if you write to the part that
771 is "not there", because you'll be overwriting the last members of the
772 preceding structure in memory.)
774 We calculate the correction using the STRUCT_OFFSET macro on the first
775 member present. If the allocated structure is smaller (no initial NV
776 actually allocated) then the net effect is to subtract the size of the NV
777 from the pointer, to return a new pointer as if an initial NV were actually
778 allocated. (We were using structures named *_allocated for this, but
779 this turned out to be a subtle bug, because a structure without an NV
780 could have a lower alignment constraint, but the compiler is allowed to
781 optimised accesses based on the alignment constraint of the actual pointer
782 to the full structure, for example, using a single 64 bit load instruction
783 because it "knows" that two adjacent 32 bit members will be 8-byte aligned.)
785 This is the same trick as was used for NV and IV bodies. Ironically it
786 doesn't need to be used for NV bodies any more, because NV is now at
787 the start of the structure. IV bodies don't need it either, because
788 they are no longer allocated.
790 In turn, the new_body_* allocators call S_new_body(), which invokes
791 new_body_inline macro, which takes a lock, and takes a body off the
792 linked list at PL_body_roots[sv_type], calling Perl_more_bodies() if
793 necessary to refresh an empty list. Then the lock is released, and
794 the body is returned.
796 Perl_more_bodies allocates a new arena, and carves it up into an array of N
797 bodies, which it strings into a linked list. It looks up arena-size
798 and body-size from the body_details table described below, thus
799 supporting the multiple body-types.
801 If PURIFY is defined, or PERL_ARENA_SIZE=0, arenas are not used, and
802 the (new|del)_X*V macros are mapped directly to malloc/free.
804 For each sv-type, struct body_details bodies_by_type[] carries
805 parameters which control these aspects of SV handling:
807 Arena_size determines whether arenas are used for this body type, and if
808 so, how big they are. PURIFY or PERL_ARENA_SIZE=0 set this field to
809 zero, forcing individual mallocs and frees.
811 Body_size determines how big a body is, and therefore how many fit into
812 each arena. Offset carries the body-pointer adjustment needed for
813 "ghost fields", and is used in *_allocated macros.
815 But its main purpose is to parameterize info needed in
816 Perl_sv_upgrade(). The info here dramatically simplifies the function
817 vs the implementation in 5.8.8, making it table-driven. All fields
818 are used for this, except for arena_size.
820 For the sv-types that have no bodies, arenas are not used, so those
821 PL_body_roots[sv_type] are unused, and can be overloaded. In
822 something of a special case, SVt_NULL is borrowed for HE arenas;
823 PL_body_roots[HE_SVSLOT=SVt_NULL] is filled by S_more_he, but the
824 bodies_by_type[SVt_NULL] slot is not used, as the table is not
829 struct body_details {
830 U8 body_size; /* Size to allocate */
831 U8 copy; /* Size of structure to copy (may be shorter) */
833 unsigned int type : 4; /* We have space for a sanity check. */
834 unsigned int cant_upgrade : 1; /* Cannot upgrade this type */
835 unsigned int zero_nv : 1; /* zero the NV when upgrading from this */
836 unsigned int arena : 1; /* Allocated from an arena */
837 size_t arena_size; /* Size of arena to allocate */
845 /* With -DPURFIY we allocate everything directly, and don't use arenas.
846 This seems a rather elegant way to simplify some of the code below. */
847 #define HASARENA FALSE
849 #define HASARENA TRUE
851 #define NOARENA FALSE
853 /* Size the arenas to exactly fit a given number of bodies. A count
854 of 0 fits the max number bodies into a PERL_ARENA_SIZE.block,
855 simplifying the default. If count > 0, the arena is sized to fit
856 only that many bodies, allowing arenas to be used for large, rare
857 bodies (XPVFM, XPVIO) without undue waste. The arena size is
858 limited by PERL_ARENA_SIZE, so we can safely oversize the
861 #define FIT_ARENA0(body_size) \
862 ((size_t)(PERL_ARENA_SIZE / body_size) * body_size)
863 #define FIT_ARENAn(count,body_size) \
864 ( count * body_size <= PERL_ARENA_SIZE) \
865 ? count * body_size \
866 : FIT_ARENA0 (body_size)
867 #define FIT_ARENA(count,body_size) \
869 ? FIT_ARENAn (count, body_size) \
870 : FIT_ARENA0 (body_size)
872 /* Calculate the length to copy. Specifically work out the length less any
873 final padding the compiler needed to add. See the comment in sv_upgrade
874 for why copying the padding proved to be a bug. */
876 #define copy_length(type, last_member) \
877 STRUCT_OFFSET(type, last_member) \
878 + sizeof (((type*)SvANY((const SV *)0))->last_member)
880 static const struct body_details bodies_by_type[] = {
881 /* HEs use this offset for their arena. */
882 { 0, 0, 0, SVt_NULL, FALSE, NONV, NOARENA, 0 },
884 /* The bind placeholder pretends to be an RV for now.
885 Also it's marked as "can't upgrade" to stop anyone using it before it's
887 { 0, 0, 0, SVt_BIND, TRUE, NONV, NOARENA, 0 },
889 /* IVs are in the head, so the allocation size is 0. */
891 sizeof(IV), /* This is used to copy out the IV body. */
892 STRUCT_OFFSET(XPVIV, xiv_iv), SVt_IV, FALSE, NONV,
893 NOARENA /* IVS don't need an arena */, 0
896 { sizeof(NV), sizeof(NV),
897 STRUCT_OFFSET(XPVNV, xnv_u),
898 SVt_NV, FALSE, HADNV, HASARENA, FIT_ARENA(0, sizeof(NV)) },
900 { sizeof(XPV) - STRUCT_OFFSET(XPV, xpv_cur),
901 copy_length(XPV, xpv_len) - STRUCT_OFFSET(XPV, xpv_cur),
902 + STRUCT_OFFSET(XPV, xpv_cur),
903 SVt_PV, FALSE, NONV, HASARENA,
904 FIT_ARENA(0, sizeof(XPV) - STRUCT_OFFSET(XPV, xpv_cur)) },
906 { sizeof(XPVIV) - STRUCT_OFFSET(XPV, xpv_cur),
907 copy_length(XPVIV, xiv_u) - STRUCT_OFFSET(XPV, xpv_cur),
908 + STRUCT_OFFSET(XPV, xpv_cur),
909 SVt_PVIV, FALSE, NONV, HASARENA,
910 FIT_ARENA(0, sizeof(XPVIV) - STRUCT_OFFSET(XPV, xpv_cur)) },
912 { sizeof(XPVNV) - STRUCT_OFFSET(XPV, xpv_cur),
913 copy_length(XPVNV, xnv_u) - STRUCT_OFFSET(XPV, xpv_cur),
914 + STRUCT_OFFSET(XPV, xpv_cur),
915 SVt_PVNV, FALSE, HADNV, HASARENA,
916 FIT_ARENA(0, sizeof(XPVNV) - STRUCT_OFFSET(XPV, xpv_cur)) },
918 { sizeof(XPVMG), copy_length(XPVMG, xnv_u), 0, SVt_PVMG, FALSE, HADNV,
919 HASARENA, FIT_ARENA(0, sizeof(XPVMG)) },
924 SVt_REGEXP, FALSE, NONV, HASARENA,
925 FIT_ARENA(0, sizeof(regexp))
928 { sizeof(XPVGV), sizeof(XPVGV), 0, SVt_PVGV, TRUE, HADNV,
929 HASARENA, FIT_ARENA(0, sizeof(XPVGV)) },
931 { sizeof(XPVLV), sizeof(XPVLV), 0, SVt_PVLV, TRUE, HADNV,
932 HASARENA, FIT_ARENA(0, sizeof(XPVLV)) },
935 copy_length(XPVAV, xav_alloc),
937 SVt_PVAV, TRUE, NONV, HASARENA,
938 FIT_ARENA(0, sizeof(XPVAV)) },
941 copy_length(XPVHV, xhv_max),
943 SVt_PVHV, TRUE, NONV, HASARENA,
944 FIT_ARENA(0, sizeof(XPVHV)) },
949 SVt_PVCV, TRUE, NONV, HASARENA,
950 FIT_ARENA(0, sizeof(XPVCV)) },
955 SVt_PVFM, TRUE, NONV, NOARENA,
956 FIT_ARENA(20, sizeof(XPVFM)) },
961 SVt_PVIO, TRUE, NONV, HASARENA,
962 FIT_ARENA(24, sizeof(XPVIO)) },
965 #define new_body_allocated(sv_type) \
966 (void *)((char *)S_new_body(aTHX_ sv_type) \
967 - bodies_by_type[sv_type].offset)
969 /* return a thing to the free list */
971 #define del_body(thing, root) \
973 void ** const thing_copy = (void **)thing; \
974 *thing_copy = *root; \
975 *root = (void*)thing_copy; \
980 #define new_XNV() safemalloc(sizeof(XPVNV))
981 #define new_XPVNV() safemalloc(sizeof(XPVNV))
982 #define new_XPVMG() safemalloc(sizeof(XPVMG))
984 #define del_XPVGV(p) safefree(p)
988 #define new_XNV() new_body_allocated(SVt_NV)
989 #define new_XPVNV() new_body_allocated(SVt_PVNV)
990 #define new_XPVMG() new_body_allocated(SVt_PVMG)
992 #define del_XPVGV(p) del_body(p + bodies_by_type[SVt_PVGV].offset, \
993 &PL_body_roots[SVt_PVGV])
997 /* no arena for you! */
999 #define new_NOARENA(details) \
1000 safemalloc((details)->body_size + (details)->offset)
1001 #define new_NOARENAZ(details) \
1002 safecalloc((details)->body_size + (details)->offset, 1)
1005 Perl_more_bodies (pTHX_ const svtype sv_type, const size_t body_size,
1006 const size_t arena_size)
1009 void ** const root = &PL_body_roots[sv_type];
1010 struct arena_desc *adesc;
1011 struct arena_set *aroot = (struct arena_set *) PL_body_arenas;
1015 const size_t good_arena_size = Perl_malloc_good_size(arena_size);
1016 #if defined(DEBUGGING) && !defined(PERL_GLOBAL_STRUCT_PRIVATE)
1017 static bool done_sanity_check;
1019 /* PERL_GLOBAL_STRUCT_PRIVATE cannot coexist with global
1020 * variables like done_sanity_check. */
1021 if (!done_sanity_check) {
1022 unsigned int i = SVt_LAST;
1024 done_sanity_check = TRUE;
1027 assert (bodies_by_type[i].type == i);
1033 /* may need new arena-set to hold new arena */
1034 if (!aroot || aroot->curr >= aroot->set_size) {
1035 struct arena_set *newroot;
1036 Newxz(newroot, 1, struct arena_set);
1037 newroot->set_size = ARENAS_PER_SET;
1038 newroot->next = aroot;
1040 PL_body_arenas = (void *) newroot;
1041 DEBUG_m(PerlIO_printf(Perl_debug_log, "new arenaset %p\n", (void*)aroot));
1044 /* ok, now have arena-set with at least 1 empty/available arena-desc */
1045 curr = aroot->curr++;
1046 adesc = &(aroot->set[curr]);
1047 assert(!adesc->arena);
1049 Newx(adesc->arena, good_arena_size, char);
1050 adesc->size = good_arena_size;
1051 adesc->utype = sv_type;
1052 DEBUG_m(PerlIO_printf(Perl_debug_log, "arena %d added: %p size %"UVuf"\n",
1053 curr, (void*)adesc->arena, (UV)good_arena_size));
1055 start = (char *) adesc->arena;
1057 /* Get the address of the byte after the end of the last body we can fit.
1058 Remember, this is integer division: */
1059 end = start + good_arena_size / body_size * body_size;
1061 /* computed count doesn't reflect the 1st slot reservation */
1062 #if defined(MYMALLOC) || defined(HAS_MALLOC_GOOD_SIZE)
1063 DEBUG_m(PerlIO_printf(Perl_debug_log,
1064 "arena %p end %p arena-size %d (from %d) type %d "
1066 (void*)start, (void*)end, (int)good_arena_size,
1067 (int)arena_size, sv_type, (int)body_size,
1068 (int)good_arena_size / (int)body_size));
1070 DEBUG_m(PerlIO_printf(Perl_debug_log,
1071 "arena %p end %p arena-size %d type %d size %d ct %d\n",
1072 (void*)start, (void*)end,
1073 (int)arena_size, sv_type, (int)body_size,
1074 (int)good_arena_size / (int)body_size));
1076 *root = (void *)start;
1079 /* Where the next body would start: */
1080 char * const next = start + body_size;
1083 /* This is the last body: */
1084 assert(next == end);
1086 *(void **)start = 0;
1090 *(void**) start = (void *)next;
1095 /* grab a new thing from the free list, allocating more if necessary.
1096 The inline version is used for speed in hot routines, and the
1097 function using it serves the rest (unless PURIFY).
1099 #define new_body_inline(xpv, sv_type) \
1101 void ** const r3wt = &PL_body_roots[sv_type]; \
1102 xpv = (PTR_TBL_ENT_t*) (*((void **)(r3wt)) \
1103 ? *((void **)(r3wt)) : Perl_more_bodies(aTHX_ sv_type, \
1104 bodies_by_type[sv_type].body_size,\
1105 bodies_by_type[sv_type].arena_size)); \
1106 *(r3wt) = *(void**)(xpv); \
1112 S_new_body(pTHX_ const svtype sv_type)
1116 new_body_inline(xpv, sv_type);
1122 static const struct body_details fake_rv =
1123 { 0, 0, 0, SVt_IV, FALSE, NONV, NOARENA, 0 };
1126 =for apidoc sv_upgrade
1128 Upgrade an SV to a more complex form. Generally adds a new body type to the
1129 SV, then copies across as much information as possible from the old body.
1130 It croaks if the SV is already in a more complex form than requested. You
1131 generally want to use the C<SvUPGRADE> macro wrapper, which checks the type
1132 before calling C<sv_upgrade>, and hence does not croak. See also
1139 Perl_sv_upgrade(pTHX_ register SV *const sv, svtype new_type)
1144 const svtype old_type = SvTYPE(sv);
1145 const struct body_details *new_type_details;
1146 const struct body_details *old_type_details
1147 = bodies_by_type + old_type;
1148 SV *referant = NULL;
1150 PERL_ARGS_ASSERT_SV_UPGRADE;
1152 if (old_type == new_type)
1155 /* This clause was purposefully added ahead of the early return above to
1156 the shared string hackery for (sort {$a <=> $b} keys %hash), with the
1157 inference by Nick I-S that it would fix other troublesome cases. See
1158 changes 7162, 7163 (f130fd4589cf5fbb24149cd4db4137c8326f49c1 and parent)
1160 Given that shared hash key scalars are no longer PVIV, but PV, there is
1161 no longer need to unshare so as to free up the IVX slot for its proper
1162 purpose. So it's safe to move the early return earlier. */
1164 if (new_type != SVt_PV && SvIsCOW(sv)) {
1165 sv_force_normal_flags(sv, 0);
1168 old_body = SvANY(sv);
1170 /* Copying structures onto other structures that have been neatly zeroed
1171 has a subtle gotcha. Consider XPVMG
1173 +------+------+------+------+------+-------+-------+
1174 | NV | CUR | LEN | IV | MAGIC | STASH |
1175 +------+------+------+------+------+-------+-------+
1176 0 4 8 12 16 20 24 28
1178 where NVs are aligned to 8 bytes, so that sizeof that structure is
1179 actually 32 bytes long, with 4 bytes of padding at the end:
1181 +------+------+------+------+------+-------+-------+------+
1182 | NV | CUR | LEN | IV | MAGIC | STASH | ??? |
1183 +------+------+------+------+------+-------+-------+------+
1184 0 4 8 12 16 20 24 28 32
1186 so what happens if you allocate memory for this structure:
1188 +------+------+------+------+------+-------+-------+------+------+...
1189 | NV | CUR | LEN | IV | MAGIC | STASH | GP | NAME |
1190 +------+------+------+------+------+-------+-------+------+------+...
1191 0 4 8 12 16 20 24 28 32 36
1193 zero it, then copy sizeof(XPVMG) bytes on top of it? Not quite what you
1194 expect, because you copy the area marked ??? onto GP. Now, ??? may have
1195 started out as zero once, but it's quite possible that it isn't. So now,
1196 rather than a nicely zeroed GP, you have it pointing somewhere random.
1199 (In fact, GP ends up pointing at a previous GP structure, because the
1200 principle cause of the padding in XPVMG getting garbage is a copy of
1201 sizeof(XPVMG) bytes from a XPVGV structure in sv_unglob. Right now
1202 this happens to be moot because XPVGV has been re-ordered, with GP
1203 no longer after STASH)
1205 So we are careful and work out the size of used parts of all the
1213 referant = SvRV(sv);
1214 old_type_details = &fake_rv;
1215 if (new_type == SVt_NV)
1216 new_type = SVt_PVNV;
1218 if (new_type < SVt_PVIV) {
1219 new_type = (new_type == SVt_NV)
1220 ? SVt_PVNV : SVt_PVIV;
1225 if (new_type < SVt_PVNV) {
1226 new_type = SVt_PVNV;
1230 assert(new_type > SVt_PV);
1231 assert(SVt_IV < SVt_PV);
1232 assert(SVt_NV < SVt_PV);
1239 /* Because the XPVMG of PL_mess_sv isn't allocated from the arena,
1240 there's no way that it can be safely upgraded, because perl.c
1241 expects to Safefree(SvANY(PL_mess_sv)) */
1242 assert(sv != PL_mess_sv);
1243 /* This flag bit is used to mean other things in other scalar types.
1244 Given that it only has meaning inside the pad, it shouldn't be set
1245 on anything that can get upgraded. */
1246 assert(!SvPAD_TYPED(sv));
1249 if (old_type_details->cant_upgrade)
1250 Perl_croak(aTHX_ "Can't upgrade %s (%" UVuf ") to %" UVuf,
1251 sv_reftype(sv, 0), (UV) old_type, (UV) new_type);
1254 if (old_type > new_type)
1255 Perl_croak(aTHX_ "sv_upgrade from type %d down to type %d",
1256 (int)old_type, (int)new_type);
1258 new_type_details = bodies_by_type + new_type;
1260 SvFLAGS(sv) &= ~SVTYPEMASK;
1261 SvFLAGS(sv) |= new_type;
1263 /* This can't happen, as SVt_NULL is <= all values of new_type, so one of
1264 the return statements above will have triggered. */
1265 assert (new_type != SVt_NULL);
1268 assert(old_type == SVt_NULL);
1269 SvANY(sv) = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
1273 assert(old_type == SVt_NULL);
1274 SvANY(sv) = new_XNV();
1279 assert(new_type_details->body_size);
1282 assert(new_type_details->arena);
1283 assert(new_type_details->arena_size);
1284 /* This points to the start of the allocated area. */
1285 new_body_inline(new_body, new_type);
1286 Zero(new_body, new_type_details->body_size, char);
1287 new_body = ((char *)new_body) - new_type_details->offset;
1289 /* We always allocated the full length item with PURIFY. To do this
1290 we fake things so that arena is false for all 16 types.. */
1291 new_body = new_NOARENAZ(new_type_details);
1293 SvANY(sv) = new_body;
1294 if (new_type == SVt_PVAV) {
1298 if (old_type_details->body_size) {
1301 /* It will have been zeroed when the new body was allocated.
1302 Lets not write to it, in case it confuses a write-back
1308 #ifndef NODEFAULT_SHAREKEYS
1309 HvSHAREKEYS_on(sv); /* key-sharing on by default */
1311 HvMAX(sv) = 7; /* (start with 8 buckets) */
1314 /* SVt_NULL isn't the only thing upgraded to AV or HV.
1315 The target created by newSVrv also is, and it can have magic.
1316 However, it never has SvPVX set.
1318 if (old_type == SVt_IV) {
1320 } else if (old_type >= SVt_PV) {
1321 assert(SvPVX_const(sv) == 0);
1324 if (old_type >= SVt_PVMG) {
1325 SvMAGIC_set(sv, ((XPVMG*)old_body)->xmg_u.xmg_magic);
1326 SvSTASH_set(sv, ((XPVMG*)old_body)->xmg_stash);
1328 sv->sv_u.svu_array = NULL; /* or svu_hash */
1334 /* This ensures that SvTHINKFIRST(sv) is true, and hence that
1335 sv_force_normal_flags(sv) is called. */
1338 /* XXX Is this still needed? Was it ever needed? Surely as there is
1339 no route from NV to PVIV, NOK can never be true */
1340 assert(!SvNOKp(sv));
1351 assert(new_type_details->body_size);
1352 /* We always allocated the full length item with PURIFY. To do this
1353 we fake things so that arena is false for all 16 types.. */
1354 if(new_type_details->arena) {
1355 /* This points to the start of the allocated area. */
1356 new_body_inline(new_body, new_type);
1357 Zero(new_body, new_type_details->body_size, char);
1358 new_body = ((char *)new_body) - new_type_details->offset;
1360 new_body = new_NOARENAZ(new_type_details);
1362 SvANY(sv) = new_body;
1364 if (old_type_details->copy) {
1365 /* There is now the potential for an upgrade from something without
1366 an offset (PVNV or PVMG) to something with one (PVCV, PVFM) */
1367 int offset = old_type_details->offset;
1368 int length = old_type_details->copy;
1370 if (new_type_details->offset > old_type_details->offset) {
1371 const int difference
1372 = new_type_details->offset - old_type_details->offset;
1373 offset += difference;
1374 length -= difference;
1376 assert (length >= 0);
1378 Copy((char *)old_body + offset, (char *)new_body + offset, length,
1382 #ifndef NV_ZERO_IS_ALLBITS_ZERO
1383 /* If NV 0.0 is stores as all bits 0 then Zero() already creates a
1384 * correct 0.0 for us. Otherwise, if the old body didn't have an
1385 * NV slot, but the new one does, then we need to initialise the
1386 * freshly created NV slot with whatever the correct bit pattern is
1388 if (old_type_details->zero_nv && !new_type_details->zero_nv
1389 && !isGV_with_GP(sv))
1393 if (new_type == SVt_PVIO) {
1394 IO * const io = MUTABLE_IO(sv);
1395 GV *iogv = gv_fetchpvs("IO::File::", GV_ADD, SVt_PVHV);
1398 /* Clear the stashcache because a new IO could overrule a package
1400 hv_clear(PL_stashcache);
1402 SvSTASH_set(io, MUTABLE_HV(SvREFCNT_inc(GvHV(iogv))));
1403 IoPAGE_LEN(sv) = 60;
1405 if (old_type < SVt_PV) {
1406 /* referant will be NULL unless the old type was SVt_IV emulating
1408 sv->sv_u.svu_rv = referant;
1412 Perl_croak(aTHX_ "panic: sv_upgrade to unknown type %lu",
1413 (unsigned long)new_type);
1416 if (old_type > SVt_IV) {
1420 /* Note that there is an assumption that all bodies of types that
1421 can be upgraded came from arenas. Only the more complex non-
1422 upgradable types are allowed to be directly malloc()ed. */
1423 assert(old_type_details->arena);
1424 del_body((void*)((char*)old_body + old_type_details->offset),
1425 &PL_body_roots[old_type]);
1431 =for apidoc sv_backoff
1433 Remove any string offset. You should normally use the C<SvOOK_off> macro
1440 Perl_sv_backoff(pTHX_ register SV *const sv)
1443 const char * const s = SvPVX_const(sv);
1445 PERL_ARGS_ASSERT_SV_BACKOFF;
1446 PERL_UNUSED_CONTEXT;
1449 assert(SvTYPE(sv) != SVt_PVHV);
1450 assert(SvTYPE(sv) != SVt_PVAV);
1452 SvOOK_offset(sv, delta);
1454 SvLEN_set(sv, SvLEN(sv) + delta);
1455 SvPV_set(sv, SvPVX(sv) - delta);
1456 Move(s, SvPVX(sv), SvCUR(sv)+1, char);
1457 SvFLAGS(sv) &= ~SVf_OOK;
1464 Expands the character buffer in the SV. If necessary, uses C<sv_unref> and
1465 upgrades the SV to C<SVt_PV>. Returns a pointer to the character buffer.
1466 Use the C<SvGROW> wrapper instead.
1472 Perl_sv_grow(pTHX_ register SV *const sv, register STRLEN newlen)
1476 PERL_ARGS_ASSERT_SV_GROW;
1478 if (PL_madskills && newlen >= 0x100000) {
1479 PerlIO_printf(Perl_debug_log,
1480 "Allocation too large: %"UVxf"\n", (UV)newlen);
1482 #ifdef HAS_64K_LIMIT
1483 if (newlen >= 0x10000) {
1484 PerlIO_printf(Perl_debug_log,
1485 "Allocation too large: %"UVxf"\n", (UV)newlen);
1488 #endif /* HAS_64K_LIMIT */
1491 if (SvTYPE(sv) < SVt_PV) {
1492 sv_upgrade(sv, SVt_PV);
1493 s = SvPVX_mutable(sv);
1495 else if (SvOOK(sv)) { /* pv is offset? */
1497 s = SvPVX_mutable(sv);
1498 if (newlen > SvLEN(sv))
1499 newlen += 10 * (newlen - SvCUR(sv)); /* avoid copy each time */
1500 #ifdef HAS_64K_LIMIT
1501 if (newlen >= 0x10000)
1506 s = SvPVX_mutable(sv);
1508 if (newlen > SvLEN(sv)) { /* need more room? */
1509 STRLEN minlen = SvCUR(sv);
1510 minlen += (minlen >> PERL_STRLEN_EXPAND_SHIFT) + 10;
1511 if (newlen < minlen)
1513 #ifndef Perl_safesysmalloc_size
1514 newlen = PERL_STRLEN_ROUNDUP(newlen);
1516 if (SvLEN(sv) && s) {
1517 s = (char*)saferealloc(s, newlen);
1520 s = (char*)safemalloc(newlen);
1521 if (SvPVX_const(sv) && SvCUR(sv)) {
1522 Move(SvPVX_const(sv), s, (newlen < SvCUR(sv)) ? newlen : SvCUR(sv), char);
1526 #ifdef Perl_safesysmalloc_size
1527 /* Do this here, do it once, do it right, and then we will never get
1528 called back into sv_grow() unless there really is some growing
1530 SvLEN_set(sv, Perl_safesysmalloc_size(s));
1532 SvLEN_set(sv, newlen);
1539 =for apidoc sv_setiv
1541 Copies an integer into the given SV, upgrading first if necessary.
1542 Does not handle 'set' magic. See also C<sv_setiv_mg>.
1548 Perl_sv_setiv(pTHX_ register SV *const sv, const IV i)
1552 PERL_ARGS_ASSERT_SV_SETIV;
1554 SV_CHECK_THINKFIRST_COW_DROP(sv);
1555 switch (SvTYPE(sv)) {
1558 sv_upgrade(sv, SVt_IV);
1561 sv_upgrade(sv, SVt_PVIV);
1565 if (!isGV_with_GP(sv))
1572 /* diag_listed_as: Can't coerce %s to %s in %s */
1573 Perl_croak(aTHX_ "Can't coerce %s to integer in %s", sv_reftype(sv,0),
1577 (void)SvIOK_only(sv); /* validate number */
1583 =for apidoc sv_setiv_mg
1585 Like C<sv_setiv>, but also handles 'set' magic.
1591 Perl_sv_setiv_mg(pTHX_ register SV *const sv, const IV i)
1593 PERL_ARGS_ASSERT_SV_SETIV_MG;
1600 =for apidoc sv_setuv
1602 Copies an unsigned integer into the given SV, upgrading first if necessary.
1603 Does not handle 'set' magic. See also C<sv_setuv_mg>.
1609 Perl_sv_setuv(pTHX_ register SV *const sv, const UV u)
1611 PERL_ARGS_ASSERT_SV_SETUV;
1613 /* With these two if statements:
1614 u=1.49 s=0.52 cu=72.49 cs=10.64 scripts=270 tests=20865
1617 u=1.35 s=0.47 cu=73.45 cs=11.43 scripts=270 tests=20865
1619 If you wish to remove them, please benchmark to see what the effect is
1621 if (u <= (UV)IV_MAX) {
1622 sv_setiv(sv, (IV)u);
1631 =for apidoc sv_setuv_mg
1633 Like C<sv_setuv>, but also handles 'set' magic.
1639 Perl_sv_setuv_mg(pTHX_ register SV *const sv, const UV u)
1641 PERL_ARGS_ASSERT_SV_SETUV_MG;
1648 =for apidoc sv_setnv
1650 Copies a double into the given SV, upgrading first if necessary.
1651 Does not handle 'set' magic. See also C<sv_setnv_mg>.
1657 Perl_sv_setnv(pTHX_ register SV *const sv, const NV num)
1661 PERL_ARGS_ASSERT_SV_SETNV;
1663 SV_CHECK_THINKFIRST_COW_DROP(sv);
1664 switch (SvTYPE(sv)) {
1667 sv_upgrade(sv, SVt_NV);
1671 sv_upgrade(sv, SVt_PVNV);
1675 if (!isGV_with_GP(sv))
1682 /* diag_listed_as: Can't coerce %s to %s in %s */
1683 Perl_croak(aTHX_ "Can't coerce %s to number in %s", sv_reftype(sv,0),
1688 (void)SvNOK_only(sv); /* validate number */
1693 =for apidoc sv_setnv_mg
1695 Like C<sv_setnv>, but also handles 'set' magic.
1701 Perl_sv_setnv_mg(pTHX_ register SV *const sv, const NV num)
1703 PERL_ARGS_ASSERT_SV_SETNV_MG;
1709 /* Print an "isn't numeric" warning, using a cleaned-up,
1710 * printable version of the offending string
1714 S_not_a_number(pTHX_ SV *const sv)
1721 PERL_ARGS_ASSERT_NOT_A_NUMBER;
1724 dsv = newSVpvs_flags("", SVs_TEMP);
1725 pv = sv_uni_display(dsv, sv, 10, UNI_DISPLAY_ISPRINT);
1728 const char * const limit = tmpbuf + sizeof(tmpbuf) - 8;
1729 /* each *s can expand to 4 chars + "...\0",
1730 i.e. need room for 8 chars */
1732 const char *s = SvPVX_const(sv);
1733 const char * const end = s + SvCUR(sv);
1734 for ( ; s < end && d < limit; s++ ) {
1736 if (ch & 128 && !isPRINT_LC(ch)) {
1745 else if (ch == '\r') {
1749 else if (ch == '\f') {
1753 else if (ch == '\\') {
1757 else if (ch == '\0') {
1761 else if (isPRINT_LC(ch))
1778 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
1779 "Argument \"%s\" isn't numeric in %s", pv,
1782 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
1783 "Argument \"%s\" isn't numeric", pv);
1787 =for apidoc looks_like_number
1789 Test if the content of an SV looks like a number (or is a number).
1790 C<Inf> and C<Infinity> are treated as numbers (so will not issue a
1791 non-numeric warning), even if your atof() doesn't grok them. Get-magic is
1798 Perl_looks_like_number(pTHX_ SV *const sv)
1800 register const char *sbegin;
1803 PERL_ARGS_ASSERT_LOOKS_LIKE_NUMBER;
1805 if (SvPOK(sv) || SvPOKp(sv)) {
1806 sbegin = SvPV_nomg_const(sv, len);
1809 return SvFLAGS(sv) & (SVf_NOK|SVp_NOK|SVf_IOK|SVp_IOK);
1810 return grok_number(sbegin, len, NULL);
1814 S_glob_2number(pTHX_ GV * const gv)
1816 SV *const buffer = sv_newmortal();
1818 PERL_ARGS_ASSERT_GLOB_2NUMBER;
1820 gv_efullname3(buffer, gv, "*");
1822 /* We know that all GVs stringify to something that is not-a-number,
1823 so no need to test that. */
1824 if (ckWARN(WARN_NUMERIC))
1825 not_a_number(buffer);
1826 /* We just want something true to return, so that S_sv_2iuv_common
1827 can tail call us and return true. */
1831 /* Actually, ISO C leaves conversion of UV to IV undefined, but
1832 until proven guilty, assume that things are not that bad... */
1837 As 64 bit platforms often have an NV that doesn't preserve all bits of
1838 an IV (an assumption perl has been based on to date) it becomes necessary
1839 to remove the assumption that the NV always carries enough precision to
1840 recreate the IV whenever needed, and that the NV is the canonical form.
1841 Instead, IV/UV and NV need to be given equal rights. So as to not lose
1842 precision as a side effect of conversion (which would lead to insanity
1843 and the dragon(s) in t/op/numconvert.t getting very angry) the intent is
1844 1) to distinguish between IV/UV/NV slots that have cached a valid
1845 conversion where precision was lost and IV/UV/NV slots that have a
1846 valid conversion which has lost no precision
1847 2) to ensure that if a numeric conversion to one form is requested that
1848 would lose precision, the precise conversion (or differently
1849 imprecise conversion) is also performed and cached, to prevent
1850 requests for different numeric formats on the same SV causing
1851 lossy conversion chains. (lossless conversion chains are perfectly
1856 SvIOKp is true if the IV slot contains a valid value
1857 SvIOK is true only if the IV value is accurate (UV if SvIOK_UV true)
1858 SvNOKp is true if the NV slot contains a valid value
1859 SvNOK is true only if the NV value is accurate
1862 while converting from PV to NV, check to see if converting that NV to an
1863 IV(or UV) would lose accuracy over a direct conversion from PV to
1864 IV(or UV). If it would, cache both conversions, return NV, but mark
1865 SV as IOK NOKp (ie not NOK).
1867 While converting from PV to IV, check to see if converting that IV to an
1868 NV would lose accuracy over a direct conversion from PV to NV. If it
1869 would, cache both conversions, flag similarly.
1871 Before, the SV value "3.2" could become NV=3.2 IV=3 NOK, IOK quite
1872 correctly because if IV & NV were set NV *always* overruled.
1873 Now, "3.2" will become NV=3.2 IV=3 NOK, IOKp, because the flag's meaning
1874 changes - now IV and NV together means that the two are interchangeable:
1875 SvIVX == (IV) SvNVX && SvNVX == (NV) SvIVX;
1877 The benefit of this is that operations such as pp_add know that if
1878 SvIOK is true for both left and right operands, then integer addition
1879 can be used instead of floating point (for cases where the result won't
1880 overflow). Before, floating point was always used, which could lead to
1881 loss of precision compared with integer addition.
1883 * making IV and NV equal status should make maths accurate on 64 bit
1885 * may speed up maths somewhat if pp_add and friends start to use
1886 integers when possible instead of fp. (Hopefully the overhead in
1887 looking for SvIOK and checking for overflow will not outweigh the
1888 fp to integer speedup)
1889 * will slow down integer operations (callers of SvIV) on "inaccurate"
1890 values, as the change from SvIOK to SvIOKp will cause a call into
1891 sv_2iv each time rather than a macro access direct to the IV slot
1892 * should speed up number->string conversion on integers as IV is
1893 favoured when IV and NV are equally accurate
1895 ####################################################################
1896 You had better be using SvIOK_notUV if you want an IV for arithmetic:
1897 SvIOK is true if (IV or UV), so you might be getting (IV)SvUV.
1898 On the other hand, SvUOK is true iff UV.
1899 ####################################################################
1901 Your mileage will vary depending your CPU's relative fp to integer
1905 #ifndef NV_PRESERVES_UV
1906 # define IS_NUMBER_UNDERFLOW_IV 1
1907 # define IS_NUMBER_UNDERFLOW_UV 2
1908 # define IS_NUMBER_IV_AND_UV 2
1909 # define IS_NUMBER_OVERFLOW_IV 4
1910 # define IS_NUMBER_OVERFLOW_UV 5
1912 /* sv_2iuv_non_preserve(): private routine for use by sv_2iv() and sv_2uv() */
1914 /* For sv_2nv these three cases are "SvNOK and don't bother casting" */
1916 S_sv_2iuv_non_preserve(pTHX_ register SV *const sv
1924 PERL_ARGS_ASSERT_SV_2IUV_NON_PRESERVE;
1926 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));
1927 if (SvNVX(sv) < (NV)IV_MIN) {
1928 (void)SvIOKp_on(sv);
1930 SvIV_set(sv, IV_MIN);
1931 return IS_NUMBER_UNDERFLOW_IV;
1933 if (SvNVX(sv) > (NV)UV_MAX) {
1934 (void)SvIOKp_on(sv);
1937 SvUV_set(sv, UV_MAX);
1938 return IS_NUMBER_OVERFLOW_UV;
1940 (void)SvIOKp_on(sv);
1942 /* Can't use strtol etc to convert this string. (See truth table in
1944 if (SvNVX(sv) <= (UV)IV_MAX) {
1945 SvIV_set(sv, I_V(SvNVX(sv)));
1946 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
1947 SvIOK_on(sv); /* Integer is precise. NOK, IOK */
1949 /* Integer is imprecise. NOK, IOKp */
1951 return SvNVX(sv) < 0 ? IS_NUMBER_UNDERFLOW_UV : IS_NUMBER_IV_AND_UV;
1954 SvUV_set(sv, U_V(SvNVX(sv)));
1955 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
1956 if (SvUVX(sv) == UV_MAX) {
1957 /* As we know that NVs don't preserve UVs, UV_MAX cannot
1958 possibly be preserved by NV. Hence, it must be overflow.
1960 return IS_NUMBER_OVERFLOW_UV;
1962 SvIOK_on(sv); /* Integer is precise. NOK, UOK */
1964 /* Integer is imprecise. NOK, IOKp */
1966 return IS_NUMBER_OVERFLOW_IV;
1968 #endif /* !NV_PRESERVES_UV*/
1971 S_sv_2iuv_common(pTHX_ SV *const sv)
1975 PERL_ARGS_ASSERT_SV_2IUV_COMMON;
1978 /* erm. not sure. *should* never get NOKp (without NOK) from sv_2nv
1979 * without also getting a cached IV/UV from it at the same time
1980 * (ie PV->NV conversion should detect loss of accuracy and cache
1981 * IV or UV at same time to avoid this. */
1982 /* IV-over-UV optimisation - choose to cache IV if possible */
1984 if (SvTYPE(sv) == SVt_NV)
1985 sv_upgrade(sv, SVt_PVNV);
1987 (void)SvIOKp_on(sv); /* Must do this first, to clear any SvOOK */
1988 /* < not <= as for NV doesn't preserve UV, ((NV)IV_MAX+1) will almost
1989 certainly cast into the IV range at IV_MAX, whereas the correct
1990 answer is the UV IV_MAX +1. Hence < ensures that dodgy boundary
1992 #if defined(NAN_COMPARE_BROKEN) && defined(Perl_isnan)
1993 if (Perl_isnan(SvNVX(sv))) {
1999 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2000 SvIV_set(sv, I_V(SvNVX(sv)));
2001 if (SvNVX(sv) == (NV) SvIVX(sv)
2002 #ifndef NV_PRESERVES_UV
2003 && (((UV)1 << NV_PRESERVES_UV_BITS) >
2004 (UV)(SvIVX(sv) > 0 ? SvIVX(sv) : -SvIVX(sv)))
2005 /* Don't flag it as "accurately an integer" if the number
2006 came from a (by definition imprecise) NV operation, and
2007 we're outside the range of NV integer precision */
2011 SvIOK_on(sv); /* Can this go wrong with rounding? NWC */
2013 /* scalar has trailing garbage, eg "42a" */
2015 DEBUG_c(PerlIO_printf(Perl_debug_log,
2016 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (precise)\n",
2022 /* IV not precise. No need to convert from PV, as NV
2023 conversion would already have cached IV if it detected
2024 that PV->IV would be better than PV->NV->IV
2025 flags already correct - don't set public IOK. */
2026 DEBUG_c(PerlIO_printf(Perl_debug_log,
2027 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (imprecise)\n",
2032 /* Can the above go wrong if SvIVX == IV_MIN and SvNVX < IV_MIN,
2033 but the cast (NV)IV_MIN rounds to a the value less (more
2034 negative) than IV_MIN which happens to be equal to SvNVX ??
2035 Analogous to 0xFFFFFFFFFFFFFFFF rounding up to NV (2**64) and
2036 NV rounding back to 0xFFFFFFFFFFFFFFFF, so UVX == UV(NVX) and
2037 (NV)UVX == NVX are both true, but the values differ. :-(
2038 Hopefully for 2s complement IV_MIN is something like
2039 0x8000000000000000 which will be exact. NWC */
2042 SvUV_set(sv, U_V(SvNVX(sv)));
2044 (SvNVX(sv) == (NV) SvUVX(sv))
2045 #ifndef NV_PRESERVES_UV
2046 /* Make sure it's not 0xFFFFFFFFFFFFFFFF */
2047 /*&& (SvUVX(sv) != UV_MAX) irrelevant with code below */
2048 && (((UV)1 << NV_PRESERVES_UV_BITS) > SvUVX(sv))
2049 /* Don't flag it as "accurately an integer" if the number
2050 came from a (by definition imprecise) NV operation, and
2051 we're outside the range of NV integer precision */
2057 DEBUG_c(PerlIO_printf(Perl_debug_log,
2058 "0x%"UVxf" 2iv(%"UVuf" => %"IVdf") (as unsigned)\n",
2064 else if (SvPOKp(sv) && SvLEN(sv)) {
2066 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2067 /* We want to avoid a possible problem when we cache an IV/ a UV which
2068 may be later translated to an NV, and the resulting NV is not
2069 the same as the direct translation of the initial string
2070 (eg 123.456 can shortcut to the IV 123 with atol(), but we must
2071 be careful to ensure that the value with the .456 is around if the
2072 NV value is requested in the future).
2074 This means that if we cache such an IV/a UV, we need to cache the
2075 NV as well. Moreover, we trade speed for space, and do not
2076 cache the NV if we are sure it's not needed.
2079 /* SVt_PVNV is one higher than SVt_PVIV, hence this order */
2080 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2081 == IS_NUMBER_IN_UV) {
2082 /* It's definitely an integer, only upgrade to PVIV */
2083 if (SvTYPE(sv) < SVt_PVIV)
2084 sv_upgrade(sv, SVt_PVIV);
2086 } else if (SvTYPE(sv) < SVt_PVNV)
2087 sv_upgrade(sv, SVt_PVNV);
2089 /* If NVs preserve UVs then we only use the UV value if we know that
2090 we aren't going to call atof() below. If NVs don't preserve UVs
2091 then the value returned may have more precision than atof() will
2092 return, even though value isn't perfectly accurate. */
2093 if ((numtype & (IS_NUMBER_IN_UV
2094 #ifdef NV_PRESERVES_UV
2097 )) == IS_NUMBER_IN_UV) {
2098 /* This won't turn off the public IOK flag if it was set above */
2099 (void)SvIOKp_on(sv);
2101 if (!(numtype & IS_NUMBER_NEG)) {
2103 if (value <= (UV)IV_MAX) {
2104 SvIV_set(sv, (IV)value);
2106 /* it didn't overflow, and it was positive. */
2107 SvUV_set(sv, value);
2111 /* 2s complement assumption */
2112 if (value <= (UV)IV_MIN) {
2113 SvIV_set(sv, -(IV)value);
2115 /* Too negative for an IV. This is a double upgrade, but
2116 I'm assuming it will be rare. */
2117 if (SvTYPE(sv) < SVt_PVNV)
2118 sv_upgrade(sv, SVt_PVNV);
2122 SvNV_set(sv, -(NV)value);
2123 SvIV_set(sv, IV_MIN);
2127 /* For !NV_PRESERVES_UV and IS_NUMBER_IN_UV and IS_NUMBER_NOT_INT we
2128 will be in the previous block to set the IV slot, and the next
2129 block to set the NV slot. So no else here. */
2131 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2132 != IS_NUMBER_IN_UV) {
2133 /* It wasn't an (integer that doesn't overflow the UV). */
2134 SvNV_set(sv, Atof(SvPVX_const(sv)));
2136 if (! numtype && ckWARN(WARN_NUMERIC))
2139 #if defined(USE_LONG_DOUBLE)
2140 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%" PERL_PRIgldbl ")\n",
2141 PTR2UV(sv), SvNVX(sv)));
2143 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"NVgf")\n",
2144 PTR2UV(sv), SvNVX(sv)));
2147 #ifdef NV_PRESERVES_UV
2148 (void)SvIOKp_on(sv);
2150 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2151 SvIV_set(sv, I_V(SvNVX(sv)));
2152 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
2155 NOOP; /* Integer is imprecise. NOK, IOKp */
2157 /* UV will not work better than IV */
2159 if (SvNVX(sv) > (NV)UV_MAX) {
2161 /* Integer is inaccurate. NOK, IOKp, is UV */
2162 SvUV_set(sv, UV_MAX);
2164 SvUV_set(sv, U_V(SvNVX(sv)));
2165 /* 0xFFFFFFFFFFFFFFFF not an issue in here, NVs
2166 NV preservse UV so can do correct comparison. */
2167 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
2170 NOOP; /* Integer is imprecise. NOK, IOKp, is UV */
2175 #else /* NV_PRESERVES_UV */
2176 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2177 == (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT)) {
2178 /* The IV/UV slot will have been set from value returned by
2179 grok_number above. The NV slot has just been set using
2182 assert (SvIOKp(sv));
2184 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2185 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2186 /* Small enough to preserve all bits. */
2187 (void)SvIOKp_on(sv);
2189 SvIV_set(sv, I_V(SvNVX(sv)));
2190 if ((NV)(SvIVX(sv)) == SvNVX(sv))
2192 /* Assumption: first non-preserved integer is < IV_MAX,
2193 this NV is in the preserved range, therefore: */
2194 if (!(U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))
2196 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);
2200 0 0 already failed to read UV.
2201 0 1 already failed to read UV.
2202 1 0 you won't get here in this case. IV/UV
2203 slot set, public IOK, Atof() unneeded.
2204 1 1 already read UV.
2205 so there's no point in sv_2iuv_non_preserve() attempting
2206 to use atol, strtol, strtoul etc. */
2208 sv_2iuv_non_preserve (sv, numtype);
2210 sv_2iuv_non_preserve (sv);
2214 #endif /* NV_PRESERVES_UV */
2215 /* It might be more code efficient to go through the entire logic above
2216 and conditionally set with SvIOKp_on() rather than SvIOK(), but it
2217 gets complex and potentially buggy, so more programmer efficient
2218 to do it this way, by turning off the public flags: */
2220 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK);
2224 if (isGV_with_GP(sv))
2225 return glob_2number(MUTABLE_GV(sv));
2227 if (!SvPADTMP(sv)) {
2228 if (!PL_localizing && ckWARN(WARN_UNINITIALIZED))
2231 if (SvTYPE(sv) < SVt_IV)
2232 /* Typically the caller expects that sv_any is not NULL now. */
2233 sv_upgrade(sv, SVt_IV);
2234 /* Return 0 from the caller. */
2241 =for apidoc sv_2iv_flags
2243 Return the integer value of an SV, doing any necessary string
2244 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2245 Normally used via the C<SvIV(sv)> and C<SvIVx(sv)> macros.
2251 Perl_sv_2iv_flags(pTHX_ register SV *const sv, const I32 flags)
2256 if (SvGMAGICAL(sv) || SvVALID(sv)) {
2257 /* FBMs use the space for SvIVX and SvNVX for other purposes, and use
2258 the same flag bit as SVf_IVisUV, so must not let them cache IVs.
2259 In practice they are extremely unlikely to actually get anywhere
2260 accessible by user Perl code - the only way that I'm aware of is when
2261 a constant subroutine which is used as the second argument to index.
2263 if (flags & SV_GMAGIC)
2268 return I_V(SvNVX(sv));
2270 if (SvPOKp(sv) && SvLEN(sv)) {
2273 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2275 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2276 == IS_NUMBER_IN_UV) {
2277 /* It's definitely an integer */
2278 if (numtype & IS_NUMBER_NEG) {
2279 if (value < (UV)IV_MIN)
2282 if (value < (UV)IV_MAX)
2287 if (ckWARN(WARN_NUMERIC))
2290 return I_V(Atof(SvPVX_const(sv)));
2295 assert(SvTYPE(sv) >= SVt_PVMG);
2296 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
2297 } else if (SvTHINKFIRST(sv)) {
2302 if (flags & SV_SKIP_OVERLOAD)
2304 tmpstr = AMG_CALLunary(sv, numer_amg);
2305 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2306 return SvIV(tmpstr);
2309 return PTR2IV(SvRV(sv));
2312 sv_force_normal_flags(sv, 0);
2314 if (SvREADONLY(sv) && !SvOK(sv)) {
2315 if (ckWARN(WARN_UNINITIALIZED))
2321 if (S_sv_2iuv_common(aTHX_ sv))
2324 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"IVdf")\n",
2325 PTR2UV(sv),SvIVX(sv)));
2326 return SvIsUV(sv) ? (IV)SvUVX(sv) : SvIVX(sv);
2330 =for apidoc sv_2uv_flags
2332 Return the unsigned integer value of an SV, doing any necessary string
2333 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2334 Normally used via the C<SvUV(sv)> and C<SvUVx(sv)> macros.
2340 Perl_sv_2uv_flags(pTHX_ register SV *const sv, const I32 flags)
2345 if (SvGMAGICAL(sv) || SvVALID(sv)) {
2346 /* FBMs use the space for SvIVX and SvNVX for other purposes, and use
2347 the same flag bit as SVf_IVisUV, so must not let them cache IVs. */
2348 if (flags & SV_GMAGIC)
2353 return U_V(SvNVX(sv));
2354 if (SvPOKp(sv) && SvLEN(sv)) {
2357 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2359 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2360 == IS_NUMBER_IN_UV) {
2361 /* It's definitely an integer */
2362 if (!(numtype & IS_NUMBER_NEG))
2366 if (ckWARN(WARN_NUMERIC))
2369 return U_V(Atof(SvPVX_const(sv)));
2374 assert(SvTYPE(sv) >= SVt_PVMG);
2375 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
2376 } else if (SvTHINKFIRST(sv)) {
2381 if (flags & SV_SKIP_OVERLOAD)
2383 tmpstr = AMG_CALLunary(sv, numer_amg);
2384 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2385 return SvUV(tmpstr);
2388 return PTR2UV(SvRV(sv));
2391 sv_force_normal_flags(sv, 0);
2393 if (SvREADONLY(sv) && !SvOK(sv)) {
2394 if (ckWARN(WARN_UNINITIALIZED))
2400 if (S_sv_2iuv_common(aTHX_ sv))
2404 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2uv(%"UVuf")\n",
2405 PTR2UV(sv),SvUVX(sv)));
2406 return SvIsUV(sv) ? SvUVX(sv) : (UV)SvIVX(sv);
2410 =for apidoc sv_2nv_flags
2412 Return the num value of an SV, doing any necessary string or integer
2413 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2414 Normally used via the C<SvNV(sv)> and C<SvNVx(sv)> macros.
2420 Perl_sv_2nv_flags(pTHX_ register SV *const sv, const I32 flags)
2425 if (SvGMAGICAL(sv) || SvVALID(sv)) {
2426 /* FBMs use the space for SvIVX and SvNVX for other purposes, and use
2427 the same flag bit as SVf_IVisUV, so must not let them cache NVs. */
2428 if (flags & SV_GMAGIC)
2432 if ((SvPOKp(sv) && SvLEN(sv)) && !SvIOKp(sv)) {
2433 if (!SvIOKp(sv) && ckWARN(WARN_NUMERIC) &&
2434 !grok_number(SvPVX_const(sv), SvCUR(sv), NULL))
2436 return Atof(SvPVX_const(sv));
2440 return (NV)SvUVX(sv);
2442 return (NV)SvIVX(sv);
2447 assert(SvTYPE(sv) >= SVt_PVMG);
2448 /* This falls through to the report_uninit near the end of the
2450 } else if (SvTHINKFIRST(sv)) {
2455 if (flags & SV_SKIP_OVERLOAD)
2457 tmpstr = AMG_CALLunary(sv, numer_amg);
2458 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2459 return SvNV(tmpstr);
2462 return PTR2NV(SvRV(sv));
2465 sv_force_normal_flags(sv, 0);
2467 if (SvREADONLY(sv) && !SvOK(sv)) {
2468 if (ckWARN(WARN_UNINITIALIZED))
2473 if (SvTYPE(sv) < SVt_NV) {
2474 /* The logic to use SVt_PVNV if necessary is in sv_upgrade. */
2475 sv_upgrade(sv, SVt_NV);
2476 #ifdef USE_LONG_DOUBLE
2478 STORE_NUMERIC_LOCAL_SET_STANDARD();
2479 PerlIO_printf(Perl_debug_log,
2480 "0x%"UVxf" num(%" PERL_PRIgldbl ")\n",
2481 PTR2UV(sv), SvNVX(sv));
2482 RESTORE_NUMERIC_LOCAL();
2486 STORE_NUMERIC_LOCAL_SET_STANDARD();
2487 PerlIO_printf(Perl_debug_log, "0x%"UVxf" num(%"NVgf")\n",
2488 PTR2UV(sv), SvNVX(sv));
2489 RESTORE_NUMERIC_LOCAL();
2493 else if (SvTYPE(sv) < SVt_PVNV)
2494 sv_upgrade(sv, SVt_PVNV);
2499 SvNV_set(sv, SvIsUV(sv) ? (NV)SvUVX(sv) : (NV)SvIVX(sv));
2500 #ifdef NV_PRESERVES_UV
2506 /* Only set the public NV OK flag if this NV preserves the IV */
2507 /* Check it's not 0xFFFFFFFFFFFFFFFF */
2509 SvIsUV(sv) ? ((SvUVX(sv) != UV_MAX)&&(SvUVX(sv) == U_V(SvNVX(sv))))
2510 : (SvIVX(sv) == I_V(SvNVX(sv))))
2516 else if (SvPOKp(sv) && SvLEN(sv)) {
2518 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2519 if (!SvIOKp(sv) && !numtype && ckWARN(WARN_NUMERIC))
2521 #ifdef NV_PRESERVES_UV
2522 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2523 == IS_NUMBER_IN_UV) {
2524 /* It's definitely an integer */
2525 SvNV_set(sv, (numtype & IS_NUMBER_NEG) ? -(NV)value : (NV)value);
2527 SvNV_set(sv, Atof(SvPVX_const(sv)));
2533 SvNV_set(sv, Atof(SvPVX_const(sv)));
2534 /* Only set the public NV OK flag if this NV preserves the value in
2535 the PV at least as well as an IV/UV would.
2536 Not sure how to do this 100% reliably. */
2537 /* if that shift count is out of range then Configure's test is
2538 wonky. We shouldn't be in here with NV_PRESERVES_UV_BITS ==
2540 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2541 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2542 SvNOK_on(sv); /* Definitely small enough to preserve all bits */
2543 } else if (!(numtype & IS_NUMBER_IN_UV)) {
2544 /* Can't use strtol etc to convert this string, so don't try.
2545 sv_2iv and sv_2uv will use the NV to convert, not the PV. */
2548 /* value has been set. It may not be precise. */
2549 if ((numtype & IS_NUMBER_NEG) && (value > (UV)IV_MIN)) {
2550 /* 2s complement assumption for (UV)IV_MIN */
2551 SvNOK_on(sv); /* Integer is too negative. */
2556 if (numtype & IS_NUMBER_NEG) {
2557 SvIV_set(sv, -(IV)value);
2558 } else if (value <= (UV)IV_MAX) {
2559 SvIV_set(sv, (IV)value);
2561 SvUV_set(sv, value);
2565 if (numtype & IS_NUMBER_NOT_INT) {
2566 /* I believe that even if the original PV had decimals,
2567 they are lost beyond the limit of the FP precision.
2568 However, neither is canonical, so both only get p
2569 flags. NWC, 2000/11/25 */
2570 /* Both already have p flags, so do nothing */
2572 const NV nv = SvNVX(sv);
2573 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2574 if (SvIVX(sv) == I_V(nv)) {
2577 /* It had no "." so it must be integer. */
2581 /* between IV_MAX and NV(UV_MAX).
2582 Could be slightly > UV_MAX */
2584 if (numtype & IS_NUMBER_NOT_INT) {
2585 /* UV and NV both imprecise. */
2587 const UV nv_as_uv = U_V(nv);
2589 if (value == nv_as_uv && SvUVX(sv) != UV_MAX) {
2598 /* It might be more code efficient to go through the entire logic above
2599 and conditionally set with SvNOKp_on() rather than SvNOK(), but it
2600 gets complex and potentially buggy, so more programmer efficient
2601 to do it this way, by turning off the public flags: */
2603 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK);
2604 #endif /* NV_PRESERVES_UV */
2607 if (isGV_with_GP(sv)) {
2608 glob_2number(MUTABLE_GV(sv));
2612 if (!PL_localizing && !SvPADTMP(sv) && ckWARN(WARN_UNINITIALIZED))
2614 assert (SvTYPE(sv) >= SVt_NV);
2615 /* Typically the caller expects that sv_any is not NULL now. */
2616 /* XXX Ilya implies that this is a bug in callers that assume this
2617 and ideally should be fixed. */
2620 #if defined(USE_LONG_DOUBLE)
2622 STORE_NUMERIC_LOCAL_SET_STANDARD();
2623 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2nv(%" PERL_PRIgldbl ")\n",
2624 PTR2UV(sv), SvNVX(sv));
2625 RESTORE_NUMERIC_LOCAL();
2629 STORE_NUMERIC_LOCAL_SET_STANDARD();
2630 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 1nv(%"NVgf")\n",
2631 PTR2UV(sv), SvNVX(sv));
2632 RESTORE_NUMERIC_LOCAL();
2641 Return an SV with the numeric value of the source SV, doing any necessary
2642 reference or overload conversion. You must use the C<SvNUM(sv)> macro to
2643 access this function.
2649 Perl_sv_2num(pTHX_ register SV *const sv)
2651 PERL_ARGS_ASSERT_SV_2NUM;
2656 SV * const tmpsv = AMG_CALLunary(sv, numer_amg);
2657 TAINT_IF(tmpsv && SvTAINTED(tmpsv));
2658 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
2659 return sv_2num(tmpsv);
2661 return sv_2mortal(newSVuv(PTR2UV(SvRV(sv))));
2664 /* uiv_2buf(): private routine for use by sv_2pv_flags(): print an IV or
2665 * UV as a string towards the end of buf, and return pointers to start and
2668 * We assume that buf is at least TYPE_CHARS(UV) long.
2672 S_uiv_2buf(char *const buf, const IV iv, UV uv, const int is_uv, char **const peob)
2674 char *ptr = buf + TYPE_CHARS(UV);
2675 char * const ebuf = ptr;
2678 PERL_ARGS_ASSERT_UIV_2BUF;
2690 *--ptr = '0' + (char)(uv % 10);
2699 =for apidoc sv_2pv_flags
2701 Returns a pointer to the string value of an SV, and sets *lp to its length.
2702 If flags includes SV_GMAGIC, does an mg_get() first. Coerces sv to a string
2704 Normally invoked via the C<SvPV_flags> macro. C<sv_2pv()> and C<sv_2pv_nomg>
2705 usually end up here too.
2711 Perl_sv_2pv_flags(pTHX_ register SV *const sv, STRLEN *const lp, const I32 flags)
2721 if (SvGMAGICAL(sv)) {
2722 if (flags & SV_GMAGIC)
2727 if (flags & SV_MUTABLE_RETURN)
2728 return SvPVX_mutable(sv);
2729 if (flags & SV_CONST_RETURN)
2730 return (char *)SvPVX_const(sv);
2733 if (SvIOKp(sv) || SvNOKp(sv)) {
2734 char tbuf[64]; /* Must fit sprintf/Gconvert of longest IV/NV */
2739 ? my_snprintf(tbuf, sizeof(tbuf), "%"UVuf, (UV)SvUVX(sv))
2740 : my_snprintf(tbuf, sizeof(tbuf), "%"IVdf, (IV)SvIVX(sv));
2741 } else if(SvNVX(sv) == 0.0) {
2746 Gconvert(SvNVX(sv), NV_DIG, 0, tbuf);
2753 SvUPGRADE(sv, SVt_PV);
2756 s = SvGROW_mutable(sv, len + 1);
2759 return (char*)memcpy(s, tbuf, len + 1);
2765 assert(SvTYPE(sv) >= SVt_PVMG);
2766 /* This falls through to the report_uninit near the end of the
2768 } else if (SvTHINKFIRST(sv)) {
2773 if (flags & SV_SKIP_OVERLOAD)
2775 tmpstr = AMG_CALLunary(sv, string_amg);
2776 TAINT_IF(tmpstr && SvTAINTED(tmpstr));
2777 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2779 /* char *pv = lp ? SvPV(tmpstr, *lp) : SvPV_nolen(tmpstr);
2783 if ((SvFLAGS(tmpstr) & (SVf_POK)) == SVf_POK) {
2784 if (flags & SV_CONST_RETURN) {
2785 pv = (char *) SvPVX_const(tmpstr);
2787 pv = (flags & SV_MUTABLE_RETURN)
2788 ? SvPVX_mutable(tmpstr) : SvPVX(tmpstr);
2791 *lp = SvCUR(tmpstr);
2793 pv = sv_2pv_flags(tmpstr, lp, flags);
2806 SV *const referent = SvRV(sv);
2810 retval = buffer = savepvn("NULLREF", len);
2811 } else if (SvTYPE(referent) == SVt_REGEXP) {
2812 REGEXP * const re = (REGEXP *)MUTABLE_PTR(referent);
2817 /* If the regex is UTF-8 we want the containing scalar to
2818 have an UTF-8 flag too */
2824 if ((seen_evals = RX_SEEN_EVALS(re)))
2825 PL_reginterp_cnt += seen_evals;
2828 *lp = RX_WRAPLEN(re);
2830 return RX_WRAPPED(re);
2832 const char *const typestr = sv_reftype(referent, 0);
2833 const STRLEN typelen = strlen(typestr);
2834 UV addr = PTR2UV(referent);
2835 const char *stashname = NULL;
2836 STRLEN stashnamelen = 0; /* hush, gcc */
2837 const char *buffer_end;
2839 if (SvOBJECT(referent)) {
2840 const HEK *const name = HvNAME_HEK(SvSTASH(referent));
2843 stashname = HEK_KEY(name);
2844 stashnamelen = HEK_LEN(name);
2846 if (HEK_UTF8(name)) {
2852 stashname = "__ANON__";
2855 len = stashnamelen + 1 /* = */ + typelen + 3 /* (0x */
2856 + 2 * sizeof(UV) + 2 /* )\0 */;
2858 len = typelen + 3 /* (0x */
2859 + 2 * sizeof(UV) + 2 /* )\0 */;
2862 Newx(buffer, len, char);
2863 buffer_end = retval = buffer + len;
2865 /* Working backwards */
2869 *--retval = PL_hexdigit[addr & 15];
2870 } while (addr >>= 4);
2876 memcpy(retval, typestr, typelen);
2880 retval -= stashnamelen;
2881 memcpy(retval, stashname, stashnamelen);
2883 /* retval may not necessarily have reached the start of the
2885 assert (retval >= buffer);
2887 len = buffer_end - retval - 1; /* -1 for that \0 */
2895 if (SvREADONLY(sv) && !SvOK(sv)) {
2898 if (flags & SV_UNDEF_RETURNS_NULL)
2900 if (ckWARN(WARN_UNINITIALIZED))
2905 if (SvIOK(sv) || ((SvIOKp(sv) && !SvNOKp(sv)))) {
2906 /* I'm assuming that if both IV and NV are equally valid then
2907 converting the IV is going to be more efficient */
2908 const U32 isUIOK = SvIsUV(sv);
2909 char buf[TYPE_CHARS(UV)];
2913 if (SvTYPE(sv) < SVt_PVIV)
2914 sv_upgrade(sv, SVt_PVIV);
2915 ptr = uiv_2buf(buf, SvIVX(sv), SvUVX(sv), isUIOK, &ebuf);
2917 /* inlined from sv_setpvn */
2918 s = SvGROW_mutable(sv, len + 1);
2919 Move(ptr, s, len, char);
2923 else if (SvNOKp(sv)) {
2924 if (SvTYPE(sv) < SVt_PVNV)
2925 sv_upgrade(sv, SVt_PVNV);
2926 if (SvNVX(sv) == 0.0) {
2927 s = SvGROW_mutable(sv, 2);
2932 /* The +20 is pure guesswork. Configure test needed. --jhi */
2933 s = SvGROW_mutable(sv, NV_DIG + 20);
2934 /* some Xenix systems wipe out errno here */
2935 Gconvert(SvNVX(sv), NV_DIG, 0, s);
2945 if (isGV_with_GP(sv)) {
2946 GV *const gv = MUTABLE_GV(sv);
2947 SV *const buffer = sv_newmortal();
2949 gv_efullname3(buffer, gv, "*");
2951 assert(SvPOK(buffer));
2953 *lp = SvCUR(buffer);
2955 if ( SvUTF8(buffer) ) SvUTF8_on(sv);
2956 return SvPVX(buffer);
2961 if (flags & SV_UNDEF_RETURNS_NULL)
2963 if (!PL_localizing && !SvPADTMP(sv) && ckWARN(WARN_UNINITIALIZED))
2965 if (SvTYPE(sv) < SVt_PV)
2966 /* Typically the caller expects that sv_any is not NULL now. */
2967 sv_upgrade(sv, SVt_PV);
2971 const STRLEN len = s - SvPVX_const(sv);
2977 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
2978 PTR2UV(sv),SvPVX_const(sv)));
2979 if (flags & SV_CONST_RETURN)
2980 return (char *)SvPVX_const(sv);
2981 if (flags & SV_MUTABLE_RETURN)
2982 return SvPVX_mutable(sv);
2987 =for apidoc sv_copypv
2989 Copies a stringified representation of the source SV into the
2990 destination SV. Automatically performs any necessary mg_get and
2991 coercion of numeric values into strings. Guaranteed to preserve
2992 UTF8 flag even from overloaded objects. Similar in nature to
2993 sv_2pv[_flags] but operates directly on an SV instead of just the
2994 string. Mostly uses sv_2pv_flags to do its work, except when that
2995 would lose the UTF-8'ness of the PV.
3001 Perl_sv_copypv(pTHX_ SV *const dsv, register SV *const ssv)
3004 const char * const s = SvPV_const(ssv,len);
3006 PERL_ARGS_ASSERT_SV_COPYPV;
3008 sv_setpvn(dsv,s,len);
3016 =for apidoc sv_2pvbyte
3018 Return a pointer to the byte-encoded representation of the SV, and set *lp
3019 to its length. May cause the SV to be downgraded from UTF-8 as a
3022 Usually accessed via the C<SvPVbyte> macro.
3028 Perl_sv_2pvbyte(pTHX_ register SV *const sv, STRLEN *const lp)
3030 PERL_ARGS_ASSERT_SV_2PVBYTE;
3033 sv_utf8_downgrade(sv,0);
3034 return lp ? SvPV_nomg(sv,*lp) : SvPV_nomg_nolen(sv);
3038 =for apidoc sv_2pvutf8
3040 Return a pointer to the UTF-8-encoded representation of the SV, and set *lp
3041 to its length. May cause the SV to be upgraded to UTF-8 as a side-effect.
3043 Usually accessed via the C<SvPVutf8> macro.
3049 Perl_sv_2pvutf8(pTHX_ register SV *const sv, STRLEN *const lp)
3051 PERL_ARGS_ASSERT_SV_2PVUTF8;
3053 sv_utf8_upgrade(sv);
3054 return lp ? SvPV(sv,*lp) : SvPV_nolen(sv);
3059 =for apidoc sv_2bool
3061 This macro is only used by sv_true() or its macro equivalent, and only if
3062 the latter's argument is neither SvPOK, SvIOK nor SvNOK.
3063 It calls sv_2bool_flags with the SV_GMAGIC flag.
3065 =for apidoc sv_2bool_flags
3067 This function is only used by sv_true() and friends, and only if
3068 the latter's argument is neither SvPOK, SvIOK nor SvNOK. If the flags
3069 contain SV_GMAGIC, then it does an mg_get() first.
3076 Perl_sv_2bool_flags(pTHX_ register SV *const sv, const I32 flags)
3080 PERL_ARGS_ASSERT_SV_2BOOL_FLAGS;
3082 if(flags & SV_GMAGIC) SvGETMAGIC(sv);
3088 SV * const tmpsv = AMG_CALLunary(sv, bool__amg);
3089 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
3090 return cBOOL(SvTRUE(tmpsv));
3092 return SvRV(sv) != 0;
3095 register XPV* const Xpvtmp = (XPV*)SvANY(sv);
3097 (*sv->sv_u.svu_pv > '0' ||
3098 Xpvtmp->xpv_cur > 1 ||
3099 (Xpvtmp->xpv_cur && *sv->sv_u.svu_pv != '0')))
3106 return SvIVX(sv) != 0;
3109 return SvNVX(sv) != 0.0;
3111 if (isGV_with_GP(sv))
3121 =for apidoc sv_utf8_upgrade
3123 Converts the PV of an SV to its UTF-8-encoded form.
3124 Forces the SV to string form if it is not already.
3125 Will C<mg_get> on C<sv> if appropriate.
3126 Always sets the SvUTF8 flag to avoid future validity checks even
3127 if the whole string is the same in UTF-8 as not.
3128 Returns the number of bytes in the converted string
3130 This is not as a general purpose byte encoding to Unicode interface:
3131 use the Encode extension for that.
3133 =for apidoc sv_utf8_upgrade_nomg
3135 Like sv_utf8_upgrade, but doesn't do magic on C<sv>
3137 =for apidoc sv_utf8_upgrade_flags
3139 Converts the PV of an SV to its UTF-8-encoded form.
3140 Forces the SV to string form if it is not already.
3141 Always sets the SvUTF8 flag to avoid future validity checks even
3142 if all the bytes are invariant in UTF-8. If C<flags> has C<SV_GMAGIC> bit set,
3143 will C<mg_get> on C<sv> if appropriate, else not.
3144 Returns the number of bytes in the converted string
3145 C<sv_utf8_upgrade> and
3146 C<sv_utf8_upgrade_nomg> are implemented in terms of this function.
3148 This is not as a general purpose byte encoding to Unicode interface:
3149 use the Encode extension for that.
3153 The grow version is currently not externally documented. It adds a parameter,
3154 extra, which is the number of unused bytes the string of 'sv' is guaranteed to
3155 have free after it upon return. This allows the caller to reserve extra space
3156 that it intends to fill, to avoid extra grows.
3158 Also externally undocumented for the moment is the flag SV_FORCE_UTF8_UPGRADE,
3159 which can be used to tell this function to not first check to see if there are
3160 any characters that are different in UTF-8 (variant characters) which would
3161 force it to allocate a new string to sv, but to assume there are. Typically
3162 this flag is used by a routine that has already parsed the string to find that
3163 there are such characters, and passes this information on so that the work
3164 doesn't have to be repeated.
3166 (One might think that the calling routine could pass in the position of the
3167 first such variant, so it wouldn't have to be found again. But that is not the
3168 case, because typically when the caller is likely to use this flag, it won't be
3169 calling this routine unless it finds something that won't fit into a byte.
3170 Otherwise it tries to not upgrade and just use bytes. But some things that
3171 do fit into a byte are variants in utf8, and the caller may not have been
3172 keeping track of these.)
3174 If the routine itself changes the string, it adds a trailing NUL. Such a NUL
3175 isn't guaranteed due to having other routines do the work in some input cases,
3176 or if the input is already flagged as being in utf8.
3178 The speed of this could perhaps be improved for many cases if someone wanted to
3179 write a fast function that counts the number of variant characters in a string,
3180 especially if it could return the position of the first one.
3185 Perl_sv_utf8_upgrade_flags_grow(pTHX_ register SV *const sv, const I32 flags, STRLEN extra)
3189 PERL_ARGS_ASSERT_SV_UTF8_UPGRADE_FLAGS_GROW;
3191 if (sv == &PL_sv_undef)
3195 if (SvREADONLY(sv) && (SvPOKp(sv) || SvIOKp(sv) || SvNOKp(sv))) {
3196 (void) sv_2pv_flags(sv,&len, flags);
3198 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3202 (void) SvPV_force_flags(sv,len,flags & SV_GMAGIC);
3207 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3212 sv_force_normal_flags(sv, 0);
3215 if (PL_encoding && !(flags & SV_UTF8_NO_ENCODING)) {
3216 sv_recode_to_utf8(sv, PL_encoding);
3217 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3221 if (SvCUR(sv) == 0) {
3222 if (extra) SvGROW(sv, extra);
3223 } else { /* Assume Latin-1/EBCDIC */
3224 /* This function could be much more efficient if we
3225 * had a FLAG in SVs to signal if there are any variant
3226 * chars in the PV. Given that there isn't such a flag
3227 * make the loop as fast as possible (although there are certainly ways
3228 * to speed this up, eg. through vectorization) */
3229 U8 * s = (U8 *) SvPVX_const(sv);
3230 U8 * e = (U8 *) SvEND(sv);
3232 STRLEN two_byte_count = 0;
3234 if (flags & SV_FORCE_UTF8_UPGRADE) goto must_be_utf8;
3236 /* See if really will need to convert to utf8. We mustn't rely on our
3237 * incoming SV being well formed and having a trailing '\0', as certain
3238 * code in pp_formline can send us partially built SVs. */
3242 if (NATIVE_IS_INVARIANT(ch)) continue;
3244 t--; /* t already incremented; re-point to first variant */
3249 /* utf8 conversion not needed because all are invariants. Mark as
3250 * UTF-8 even if no variant - saves scanning loop */
3252 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3257 /* Here, the string should be converted to utf8, either because of an
3258 * input flag (two_byte_count = 0), or because a character that
3259 * requires 2 bytes was found (two_byte_count = 1). t points either to
3260 * the beginning of the string (if we didn't examine anything), or to
3261 * the first variant. In either case, everything from s to t - 1 will
3262 * occupy only 1 byte each on output.
3264 * There are two main ways to convert. One is to create a new string
3265 * and go through the input starting from the beginning, appending each
3266 * converted value onto the new string as we go along. It's probably
3267 * best to allocate enough space in the string for the worst possible
3268 * case rather than possibly running out of space and having to
3269 * reallocate and then copy what we've done so far. Since everything
3270 * from s to t - 1 is invariant, the destination can be initialized
3271 * with these using a fast memory copy
3273 * The other way is to figure out exactly how big the string should be
3274 * by parsing the entire input. Then you don't have to make it big
3275 * enough to handle the worst possible case, and more importantly, if
3276 * the string you already have is large enough, you don't have to
3277 * allocate a new string, you can copy the last character in the input
3278 * string to the final position(s) that will be occupied by the
3279 * converted string and go backwards, stopping at t, since everything
3280 * before that is invariant.
3282 * There are advantages and disadvantages to each method.
3284 * In the first method, we can allocate a new string, do the memory
3285 * copy from the s to t - 1, and then proceed through the rest of the
3286 * string byte-by-byte.
3288 * In the second method, we proceed through the rest of the input
3289 * string just calculating how big the converted string will be. Then
3290 * there are two cases:
3291 * 1) if the string has enough extra space to handle the converted
3292 * value. We go backwards through the string, converting until we
3293 * get to the position we are at now, and then stop. If this
3294 * position is far enough along in the string, this method is
3295 * faster than the other method. If the memory copy were the same
3296 * speed as the byte-by-byte loop, that position would be about
3297 * half-way, as at the half-way mark, parsing to the end and back
3298 * is one complete string's parse, the same amount as starting
3299 * over and going all the way through. Actually, it would be
3300 * somewhat less than half-way, as it's faster to just count bytes
3301 * than to also copy, and we don't have the overhead of allocating
3302 * a new string, changing the scalar to use it, and freeing the
3303 * existing one. But if the memory copy is fast, the break-even
3304 * point is somewhere after half way. The counting loop could be
3305 * sped up by vectorization, etc, to move the break-even point
3306 * further towards the beginning.
3307 * 2) if the string doesn't have enough space to handle the converted
3308 * value. A new string will have to be allocated, and one might
3309 * as well, given that, start from the beginning doing the first
3310 * method. We've spent extra time parsing the string and in
3311 * exchange all we've gotten is that we know precisely how big to
3312 * make the new one. Perl is more optimized for time than space,
3313 * so this case is a loser.
3314 * So what I've decided to do is not use the 2nd method unless it is
3315 * guaranteed that a new string won't have to be allocated, assuming
3316 * the worst case. I also decided not to put any more conditions on it
3317 * than this, for now. It seems likely that, since the worst case is
3318 * twice as big as the unknown portion of the string (plus 1), we won't
3319 * be guaranteed enough space, causing us to go to the first method,
3320 * unless the string is short, or the first variant character is near
3321 * the end of it. In either of these cases, it seems best to use the
3322 * 2nd method. The only circumstance I can think of where this would
3323 * be really slower is if the string had once had much more data in it
3324 * than it does now, but there is still a substantial amount in it */
3327 STRLEN invariant_head = t - s;
3328 STRLEN size = invariant_head + (e - t) * 2 + 1 + extra;
3329 if (SvLEN(sv) < size) {
3331 /* Here, have decided to allocate a new string */
3336 Newx(dst, size, U8);
3338 /* If no known invariants at the beginning of the input string,
3339 * set so starts from there. Otherwise, can use memory copy to
3340 * get up to where we are now, and then start from here */
3342 if (invariant_head <= 0) {
3345 Copy(s, dst, invariant_head, char);
3346 d = dst + invariant_head;
3350 const UV uv = NATIVE8_TO_UNI(*t++);
3351 if (UNI_IS_INVARIANT(uv))
3352 *d++ = (U8)UNI_TO_NATIVE(uv);
3354 *d++ = (U8)UTF8_EIGHT_BIT_HI(uv);
3355 *d++ = (U8)UTF8_EIGHT_BIT_LO(uv);
3359 SvPV_free(sv); /* No longer using pre-existing string */
3360 SvPV_set(sv, (char*)dst);
3361 SvCUR_set(sv, d - dst);
3362 SvLEN_set(sv, size);
3365 /* Here, have decided to get the exact size of the string.
3366 * Currently this happens only when we know that there is
3367 * guaranteed enough space to fit the converted string, so
3368 * don't have to worry about growing. If two_byte_count is 0,
3369 * then t points to the first byte of the string which hasn't
3370 * been examined yet. Otherwise two_byte_count is 1, and t
3371 * points to the first byte in the string that will expand to
3372 * two. Depending on this, start examining at t or 1 after t.
3375 U8 *d = t + two_byte_count;
3378 /* Count up the remaining bytes that expand to two */
3381 const U8 chr = *d++;
3382 if (! NATIVE_IS_INVARIANT(chr)) two_byte_count++;
3385 /* The string will expand by just the number of bytes that
3386 * occupy two positions. But we are one afterwards because of
3387 * the increment just above. This is the place to put the
3388 * trailing NUL, and to set the length before we decrement */
3390 d += two_byte_count;
3391 SvCUR_set(sv, d - s);
3395 /* Having decremented d, it points to the position to put the
3396 * very last byte of the expanded string. Go backwards through
3397 * the string, copying and expanding as we go, stopping when we
3398 * get to the part that is invariant the rest of the way down */
3402 const U8 ch = NATIVE8_TO_UNI(*e--);
3403 if (UNI_IS_INVARIANT(ch)) {
3404 *d-- = UNI_TO_NATIVE(ch);
3406 *d-- = (U8)UTF8_EIGHT_BIT_LO(ch);
3407 *d-- = (U8)UTF8_EIGHT_BIT_HI(ch);
3412 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
3413 /* Update pos. We do it at the end rather than during
3414 * the upgrade, to avoid slowing down the common case
3415 * (upgrade without pos) */
3416 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
3418 I32 pos = mg->mg_len;
3419 if (pos > 0 && (U32)pos > invariant_head) {
3420 U8 *d = (U8*) SvPVX(sv) + invariant_head;
3421 STRLEN n = (U32)pos - invariant_head;
3423 if (UTF8_IS_START(*d))
3428 mg->mg_len = d - (U8*)SvPVX(sv);
3431 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
3432 magic_setutf8(sv,mg); /* clear UTF8 cache */
3437 /* Mark as UTF-8 even if no variant - saves scanning loop */
3443 =for apidoc sv_utf8_downgrade
3445 Attempts to convert the PV of an SV from characters to bytes.
3446 If the PV contains a character that cannot fit
3447 in a byte, this conversion will fail;
3448 in this case, either returns false or, if C<fail_ok> is not
3451 This is not as a general purpose Unicode to byte encoding interface:
3452 use the Encode extension for that.
3458 Perl_sv_utf8_downgrade(pTHX_ register SV *const sv, const bool fail_ok)
3462 PERL_ARGS_ASSERT_SV_UTF8_DOWNGRADE;
3464 if (SvPOKp(sv) && SvUTF8(sv)) {
3468 int mg_flags = SV_GMAGIC;
3471 sv_force_normal_flags(sv, 0);
3473 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
3475 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
3477 I32 pos = mg->mg_len;
3479 sv_pos_b2u(sv, &pos);
3480 mg_flags = 0; /* sv_pos_b2u does get magic */
3484 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
3485 magic_setutf8(sv,mg); /* clear UTF8 cache */
3488 s = (U8 *) SvPV_flags(sv, len, mg_flags);
3490 if (!utf8_to_bytes(s, &len)) {
3495 Perl_croak(aTHX_ "Wide character in %s",
3498 Perl_croak(aTHX_ "Wide character");
3509 =for apidoc sv_utf8_encode
3511 Converts the PV of an SV to UTF-8, but then turns the C<SvUTF8>
3512 flag off so that it looks like octets again.
3518 Perl_sv_utf8_encode(pTHX_ register SV *const sv)
3520 PERL_ARGS_ASSERT_SV_UTF8_ENCODE;
3523 sv_force_normal_flags(sv, 0);
3525 if (SvREADONLY(sv)) {
3526 Perl_croak_no_modify(aTHX);
3528 (void) sv_utf8_upgrade(sv);
3533 =for apidoc sv_utf8_decode
3535 If the PV of the SV is an octet sequence in UTF-8
3536 and contains a multiple-byte character, the C<SvUTF8> flag is turned on
3537 so that it looks like a character. If the PV contains only single-byte
3538 characters, the C<SvUTF8> flag stays off.
3539 Scans PV for validity and returns false if the PV is invalid UTF-8.
3545 Perl_sv_utf8_decode(pTHX_ register SV *const sv)
3547 PERL_ARGS_ASSERT_SV_UTF8_DECODE;
3550 const U8 *start, *c;
3553 /* The octets may have got themselves encoded - get them back as
3556 if (!sv_utf8_downgrade(sv, TRUE))
3559 /* it is actually just a matter of turning the utf8 flag on, but
3560 * we want to make sure everything inside is valid utf8 first.
3562 c = start = (const U8 *) SvPVX_const(sv);
3563 if (!is_utf8_string(c, SvCUR(sv)+1))
3565 e = (const U8 *) SvEND(sv);
3568 if (!UTF8_IS_INVARIANT(ch)) {
3573 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
3574 /* adjust pos to the start of a UTF8 char sequence */
3575 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
3577 I32 pos = mg->mg_len;
3579 for (c = start + pos; c > start; c--) {
3580 if (UTF8_IS_START(*c))
3583 mg->mg_len = c - start;
3586 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
3587 magic_setutf8(sv,mg); /* clear UTF8 cache */
3594 =for apidoc sv_setsv
3596 Copies the contents of the source SV C<ssv> into the destination SV
3597 C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3598 function if the source SV needs to be reused. Does not handle 'set' magic.
3599 Loosely speaking, it performs a copy-by-value, obliterating any previous
3600 content of the destination.
3602 You probably want to use one of the assortment of wrappers, such as
3603 C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3604 C<SvSetMagicSV_nosteal>.
3606 =for apidoc sv_setsv_flags
3608 Copies the contents of the source SV C<ssv> into the destination SV
3609 C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3610 function if the source SV needs to be reused. Does not handle 'set' magic.
3611 Loosely speaking, it performs a copy-by-value, obliterating any previous
3612 content of the destination.
3613 If the C<flags> parameter has the C<SV_GMAGIC> bit set, will C<mg_get> on
3614 C<ssv> if appropriate, else not. If the C<flags> parameter has the
3615 C<NOSTEAL> bit set then the buffers of temps will not be stolen. <sv_setsv>
3616 and C<sv_setsv_nomg> are implemented in terms of this function.
3618 You probably want to use one of the assortment of wrappers, such as
3619 C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3620 C<SvSetMagicSV_nosteal>.
3622 This is the primary function for copying scalars, and most other
3623 copy-ish functions and macros use this underneath.
3629 S_glob_assign_glob(pTHX_ SV *const dstr, SV *const sstr, const int dtype)
3631 I32 mro_changes = 0; /* 1 = method, 2 = isa, 3 = recursive isa */
3632 HV *old_stash = NULL;
3634 PERL_ARGS_ASSERT_GLOB_ASSIGN_GLOB;
3636 if (dtype != SVt_PVGV && !isGV_with_GP(dstr)) {
3637 const char * const name = GvNAME(sstr);
3638 const STRLEN len = GvNAMELEN(sstr);
3640 if (dtype >= SVt_PV) {
3646 SvUPGRADE(dstr, SVt_PVGV);
3647 (void)SvOK_off(dstr);
3648 /* FIXME - why are we doing this, then turning it off and on again
3650 isGV_with_GP_on(dstr);
3652 GvSTASH(dstr) = GvSTASH(sstr);
3654 Perl_sv_add_backref(aTHX_ MUTABLE_SV(GvSTASH(dstr)), dstr);
3655 gv_name_set(MUTABLE_GV(dstr), name, len,
3656 GV_ADD | (GvNAMEUTF8(sstr) ? SVf_UTF8 : 0 ));
3657 SvFAKE_on(dstr); /* can coerce to non-glob */
3660 if(GvGP(MUTABLE_GV(sstr))) {
3661 /* If source has method cache entry, clear it */
3663 SvREFCNT_dec(GvCV(sstr));
3664 GvCV_set(sstr, NULL);
3667 /* If source has a real method, then a method is
3670 GvCV((const GV *)sstr) && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3676 /* If dest already had a real method, that's a change as well */
3678 !mro_changes && GvGP(MUTABLE_GV(dstr)) && GvCVu((const GV *)dstr)
3679 && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3684 /* We don’t need to check the name of the destination if it was not a
3685 glob to begin with. */
3686 if(dtype == SVt_PVGV) {
3687 const char * const name = GvNAME((const GV *)dstr);
3690 /* The stash may have been detached from the symbol table, so
3692 && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3693 && GvAV((const GV *)sstr)
3697 const STRLEN len = GvNAMELEN(dstr);
3698 if ((len > 1 && name[len-2] == ':' && name[len-1] == ':')
3699 || (len == 1 && name[0] == ':')) {
3702 /* Set aside the old stash, so we can reset isa caches on
3704 if((old_stash = GvHV(dstr)))
3705 /* Make sure we do not lose it early. */
3706 SvREFCNT_inc_simple_void_NN(
3707 sv_2mortal((SV *)old_stash)
3713 gp_free(MUTABLE_GV(dstr));
3714 isGV_with_GP_off(dstr);
3715 (void)SvOK_off(dstr);
3716 isGV_with_GP_on(dstr);
3717 GvINTRO_off(dstr); /* one-shot flag */
3718 GvGP_set(dstr, gp_ref(GvGP(sstr)));
3719 if (SvTAINTED(sstr))
3721 if (GvIMPORTED(dstr) != GVf_IMPORTED
3722 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3724 GvIMPORTED_on(dstr);
3727 if(mro_changes == 2) {
3729 SV * const sref = (SV *)GvAV((const GV *)dstr);
3730 if (SvSMAGICAL(sref) && (mg = mg_find(sref, PERL_MAGIC_isa))) {
3731 if (SvTYPE(mg->mg_obj) != SVt_PVAV) {
3732 AV * const ary = newAV();
3733 av_push(ary, mg->mg_obj); /* takes the refcount */
3734 mg->mg_obj = (SV *)ary;
3736 av_push((AV *)mg->mg_obj, SvREFCNT_inc_simple_NN(dstr));
3738 else sv_magic(sref, dstr, PERL_MAGIC_isa, NULL, 0);
3739 mro_isa_changed_in(GvSTASH(dstr));
3741 else if(mro_changes == 3) {
3742 HV * const stash = GvHV(dstr);
3743 if(old_stash ? (HV *)HvENAME_get(old_stash) : stash)
3749 else if(mro_changes) mro_method_changed_in(GvSTASH(dstr));
3754 S_glob_assign_ref(pTHX_ SV *const dstr, SV *const sstr)
3756 SV * const sref = SvREFCNT_inc(SvRV(sstr));
3758 const int intro = GvINTRO(dstr);
3761 const U32 stype = SvTYPE(sref);
3763 PERL_ARGS_ASSERT_GLOB_ASSIGN_REF;
3766 GvINTRO_off(dstr); /* one-shot flag */
3767 GvLINE(dstr) = CopLINE(PL_curcop);
3768 GvEGV(dstr) = MUTABLE_GV(dstr);
3773 location = (SV **) &(GvGP(dstr)->gp_cv); /* XXX bypassing GvCV_set */
3774 import_flag = GVf_IMPORTED_CV;
3777 location = (SV **) &GvHV(dstr);
3778 import_flag = GVf_IMPORTED_HV;
3781 location = (SV **) &GvAV(dstr);
3782 import_flag = GVf_IMPORTED_AV;
3785 location = (SV **) &GvIOp(dstr);
3788 location = (SV **) &GvFORM(dstr);
3791 location = &GvSV(dstr);
3792 import_flag = GVf_IMPORTED_SV;
3795 if (stype == SVt_PVCV) {
3796 /*if (GvCVGEN(dstr) && (GvCV(dstr) != (const CV *)sref || GvCVGEN(dstr))) {*/
3797 if (GvCVGEN(dstr)) {
3798 SvREFCNT_dec(GvCV(dstr));
3799 GvCV_set(dstr, NULL);
3800 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3803 SAVEGENERICSV(*location);
3807 if (stype == SVt_PVCV && (*location != sref || GvCVGEN(dstr))) {
3808 CV* const cv = MUTABLE_CV(*location);
3810 if (!GvCVGEN((const GV *)dstr) &&
3811 (CvROOT(cv) || CvXSUB(cv)))
3813 /* Redefining a sub - warning is mandatory if
3814 it was a const and its value changed. */
3815 if (CvCONST(cv) && CvCONST((const CV *)sref)
3817 == cv_const_sv((const CV *)sref)) {
3819 /* They are 2 constant subroutines generated from
3820 the same constant. This probably means that
3821 they are really the "same" proxy subroutine
3822 instantiated in 2 places. Most likely this is
3823 when a constant is exported twice. Don't warn.
3826 else if (ckWARN(WARN_REDEFINE)
3828 && (!CvCONST((const CV *)sref)
3829 || sv_cmp(cv_const_sv(cv),
3830 cv_const_sv((const CV *)
3832 Perl_warner(aTHX_ packWARN(WARN_REDEFINE),
3835 ? "Constant subroutine %"HEKf
3836 "::%"HEKf" redefined"
3837 : "Subroutine %"HEKf"::%"HEKf
3840 HvNAME_HEK(GvSTASH((const GV *)dstr))
3842 HEKfARG(GvENAME_HEK(MUTABLE_GV(dstr))));
3846 cv_ckproto_len_flags(cv, (const GV *)dstr,
3847 SvPOK(sref) ? CvPROTO(sref) : NULL,
3848 SvPOK(sref) ? CvPROTOLEN(sref) : 0,
3849 SvPOK(sref) ? SvUTF8(sref) : 0);
3851 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3852 GvASSUMECV_on(dstr);
3853 if(GvSTASH(dstr)) mro_method_changed_in(GvSTASH(dstr)); /* sub foo { 1 } sub bar { 2 } *bar = \&foo */
3856 if (import_flag && !(GvFLAGS(dstr) & import_flag)
3857 && CopSTASH_ne(PL_curcop, GvSTASH(dstr))) {
3858 GvFLAGS(dstr) |= import_flag;
3860 if (stype == SVt_PVHV) {
3861 const char * const name = GvNAME((GV*)dstr);
3862 const STRLEN len = GvNAMELEN(dstr);
3865 (len > 1 && name[len-2] == ':' && name[len-1] == ':')
3866 || (len == 1 && name[0] == ':')
3868 && (!dref || HvENAME_get(dref))
3871 (HV *)sref, (HV *)dref,
3877 stype == SVt_PVAV && sref != dref
3878 && strEQ(GvNAME((GV*)dstr), "ISA")
3879 /* The stash may have been detached from the symbol table, so
3880 check its name before doing anything. */
3881 && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3884 MAGIC * const omg = dref && SvSMAGICAL(dref)
3885 ? mg_find(dref, PERL_MAGIC_isa)
3887 if (SvSMAGICAL(sref) && (mg = mg_find(sref, PERL_MAGIC_isa))) {
3888 if (SvTYPE(mg->mg_obj) != SVt_PVAV) {
3889 AV * const ary = newAV();
3890 av_push(ary, mg->mg_obj); /* takes the refcount */
3891 mg->mg_obj = (SV *)ary;
3894 if (SvTYPE(omg->mg_obj) == SVt_PVAV) {
3895 SV **svp = AvARRAY((AV *)omg->mg_obj);
3896 I32 items = AvFILLp((AV *)omg->mg_obj) + 1;
3900 SvREFCNT_inc_simple_NN(*svp++)
3906 SvREFCNT_inc_simple_NN(omg->mg_obj)
3910 av_push((AV *)mg->mg_obj,SvREFCNT_inc_simple_NN(dstr));
3915 sref, omg ? omg->mg_obj : dstr, PERL_MAGIC_isa, NULL, 0
3917 mg = mg_find(sref, PERL_MAGIC_isa);
3919 /* Since the *ISA assignment could have affected more than
3920 one stash, don’t call mro_isa_changed_in directly, but let
3921 magic_clearisa do it for us, as it already has the logic for
3922 dealing with globs vs arrays of globs. */
3924 Perl_magic_clearisa(aTHX_ NULL, mg);
3929 if (SvTAINTED(sstr))
3935 Perl_sv_setsv_flags(pTHX_ SV *dstr, register SV* sstr, const I32 flags)
3938 register U32 sflags;
3940 register svtype stype;
3942 PERL_ARGS_ASSERT_SV_SETSV_FLAGS;
3947 if (SvIS_FREED(dstr)) {
3948 Perl_croak(aTHX_ "panic: attempt to copy value %" SVf
3949 " to a freed scalar %p", SVfARG(sstr), (void *)dstr);
3951 SV_CHECK_THINKFIRST_COW_DROP(dstr);
3953 sstr = &PL_sv_undef;
3954 if (SvIS_FREED(sstr)) {
3955 Perl_croak(aTHX_ "panic: attempt to copy freed scalar %p to %p",
3956 (void*)sstr, (void*)dstr);
3958 stype = SvTYPE(sstr);
3959 dtype = SvTYPE(dstr);
3961 (void)SvAMAGIC_off(dstr);
3964 /* need to nuke the magic */
3968 /* There's a lot of redundancy below but we're going for speed here */
3973 if (dtype != SVt_PVGV && dtype != SVt_PVLV) {
3974 (void)SvOK_off(dstr);
3982 sv_upgrade(dstr, SVt_IV);
3986 sv_upgrade(dstr, SVt_PVIV);
3990 goto end_of_first_switch;
3992 (void)SvIOK_only(dstr);
3993 SvIV_set(dstr, SvIVX(sstr));
3996 /* SvTAINTED can only be true if the SV has taint magic, which in
3997 turn means that the SV type is PVMG (or greater). This is the
3998 case statement for SVt_IV, so this cannot be true (whatever gcov
4000 assert(!SvTAINTED(sstr));
4005 if (dtype < SVt_PV && dtype != SVt_IV)
4006 sv_upgrade(dstr, SVt_IV);
4014 sv_upgrade(dstr, SVt_NV);
4018 sv_upgrade(dstr, SVt_PVNV);
4022 goto end_of_first_switch;
4024 SvNV_set(dstr, SvNVX(sstr));
4025 (void)SvNOK_only(dstr);
4026 /* SvTAINTED can only be true if the SV has taint magic, which in
4027 turn means that the SV type is PVMG (or greater). This is the
4028 case statement for SVt_NV, so this cannot be true (whatever gcov
4030 assert(!SvTAINTED(sstr));
4036 #ifdef PERL_OLD_COPY_ON_WRITE
4037 if ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS) {
4038 if (dtype < SVt_PVIV)
4039 sv_upgrade(dstr, SVt_PVIV);
4046 sv_upgrade(dstr, SVt_PV);
4049 if (dtype < SVt_PVIV)
4050 sv_upgrade(dstr, SVt_PVIV);
4053 if (dtype < SVt_PVNV)
4054 sv_upgrade(dstr, SVt_PVNV);
4058 const char * const type = sv_reftype(sstr,0);
4060 Perl_croak(aTHX_ "Bizarre copy of %s in %s", type, OP_DESC(PL_op));
4062 Perl_croak(aTHX_ "Bizarre copy of %s", type);
4067 if (dtype < SVt_REGEXP)
4068 sv_upgrade(dstr, SVt_REGEXP);
4071 /* case SVt_BIND: */
4075 if (SvGMAGICAL(sstr) && (flags & SV_GMAGIC)) {
4077 if (SvTYPE(sstr) != stype)
4078 stype = SvTYPE(sstr);
4080 if (isGV_with_GP(sstr) && dtype <= SVt_PVLV) {
4081 glob_assign_glob(dstr, sstr, dtype);
4084 if (stype == SVt_PVLV)
4085 SvUPGRADE(dstr, SVt_PVNV);
4087 SvUPGRADE(dstr, (svtype)stype);
4089 end_of_first_switch:
4091 /* dstr may have been upgraded. */
4092 dtype = SvTYPE(dstr);
4093 sflags = SvFLAGS(sstr);
4095 if (dtype == SVt_PVCV || dtype == SVt_PVFM) {
4096 /* Assigning to a subroutine sets the prototype. */
4099 const char *const ptr = SvPV_const(sstr, len);
4101 SvGROW(dstr, len + 1);
4102 Copy(ptr, SvPVX(dstr), len + 1, char);
4103 SvCUR_set(dstr, len);
4105 SvFLAGS(dstr) |= sflags & SVf_UTF8;
4106 CvAUTOLOAD_off(dstr);
4110 } else if (dtype == SVt_PVAV || dtype == SVt_PVHV) {
4111 const char * const type = sv_reftype(dstr,0);
4113 Perl_croak(aTHX_ "Cannot copy to %s in %s", type, OP_DESC(PL_op));
4115 Perl_croak(aTHX_ "Cannot copy to %s", type);
4116 } else if (sflags & SVf_ROK) {
4117 if (isGV_with_GP(dstr)
4118 && SvTYPE(SvRV(sstr)) == SVt_PVGV && isGV_with_GP(SvRV(sstr))) {
4121 if (GvIMPORTED(dstr) != GVf_IMPORTED
4122 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
4124 GvIMPORTED_on(dstr);
4129 glob_assign_glob(dstr, sstr, dtype);
4133 if (dtype >= SVt_PV) {
4134 if (isGV_with_GP(dstr)) {
4135 glob_assign_ref(dstr, sstr);
4138 if (SvPVX_const(dstr)) {
4144 (void)SvOK_off(dstr);
4145 SvRV_set(dstr, SvREFCNT_inc(SvRV(sstr)));
4146 SvFLAGS(dstr) |= sflags & SVf_ROK;
4147 assert(!(sflags & SVp_NOK));
4148 assert(!(sflags & SVp_IOK));
4149 assert(!(sflags & SVf_NOK));
4150 assert(!(sflags & SVf_IOK));
4152 else if (isGV_with_GP(dstr)) {
4153 if (!(sflags & SVf_OK)) {
4154 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
4155 "Undefined value assigned to typeglob");
4158 GV *gv = gv_fetchsv_nomg(sstr, GV_ADD, SVt_PVGV);
4159 if (dstr != (const SV *)gv) {
4160 const char * const name = GvNAME((const GV *)dstr);
4161 const STRLEN len = GvNAMELEN(dstr);
4162 HV *old_stash = NULL;
4163 bool reset_isa = FALSE;
4164 if ((len > 1 && name[len-2] == ':' && name[len-1] == ':')
4165 || (len == 1 && name[0] == ':')) {
4166 /* Set aside the old stash, so we can reset isa caches
4167 on its subclasses. */
4168 if((old_stash = GvHV(dstr))) {
4169 /* Make sure we do not lose it early. */
4170 SvREFCNT_inc_simple_void_NN(
4171 sv_2mortal((SV *)old_stash)
4178 gp_free(MUTABLE_GV(dstr));
4179 GvGP_set(dstr, gp_ref(GvGP(gv)));
4182 HV * const stash = GvHV(dstr);
4184 old_stash ? (HV *)HvENAME_get(old_stash) : stash
4194 else if (dtype == SVt_REGEXP && stype == SVt_REGEXP) {
4195 reg_temp_copy((REGEXP*)dstr, (REGEXP*)sstr);
4197 else if (sflags & SVp_POK) {
4201 * Check to see if we can just swipe the string. If so, it's a
4202 * possible small lose on short strings, but a big win on long ones.
4203 * It might even be a win on short strings if SvPVX_const(dstr)
4204 * has to be allocated and SvPVX_const(sstr) has to be freed.
4205 * Likewise if we can set up COW rather than doing an actual copy, we
4206 * drop to the else clause, as the swipe code and the COW setup code
4207 * have much in common.
4210 /* Whichever path we take through the next code, we want this true,
4211 and doing it now facilitates the COW check. */
4212 (void)SvPOK_only(dstr);
4215 /* If we're already COW then this clause is not true, and if COW
4216 is allowed then we drop down to the else and make dest COW
4217 with us. If caller hasn't said that we're allowed to COW
4218 shared hash keys then we don't do the COW setup, even if the
4219 source scalar is a shared hash key scalar. */
4220 (((flags & SV_COW_SHARED_HASH_KEYS)
4221 ? (sflags & (SVf_FAKE|SVf_READONLY)) != (SVf_FAKE|SVf_READONLY)
4222 : 1 /* If making a COW copy is forbidden then the behaviour we
4223 desire is as if the source SV isn't actually already
4224 COW, even if it is. So we act as if the source flags
4225 are not COW, rather than actually testing them. */
4227 #ifndef PERL_OLD_COPY_ON_WRITE
4228 /* The change that added SV_COW_SHARED_HASH_KEYS makes the logic
4229 when PERL_OLD_COPY_ON_WRITE is defined a little wrong.
4230 Conceptually PERL_OLD_COPY_ON_WRITE being defined should
4231 override SV_COW_SHARED_HASH_KEYS, because it means "always COW"
4232 but in turn, it's somewhat dead code, never expected to go
4233 live, but more kept as a placeholder on how to do it better
4234 in a newer implementation. */
4235 /* If we are COW and dstr is a suitable target then we drop down
4236 into the else and make dest a COW of us. */
4237 || (SvFLAGS(dstr) & CAN_COW_MASK) != CAN_COW_FLAGS
4242 (sflags & SVs_TEMP) && /* slated for free anyway? */
4243 !(sflags & SVf_OOK) && /* and not involved in OOK hack? */
4244 (!(flags & SV_NOSTEAL)) &&
4245 /* and we're allowed to steal temps */
4246 SvREFCNT(sstr) == 1 && /* and no other references to it? */
4247 SvLEN(sstr)) /* and really is a string */
4248 #ifdef PERL_OLD_COPY_ON_WRITE
4249 && ((flags & SV_COW_SHARED_HASH_KEYS)
4250 ? (!((sflags & CAN_COW_MASK) == CAN_COW_FLAGS
4251 && (SvFLAGS(dstr) & CAN_COW_MASK) == CAN_COW_FLAGS
4252 && SvTYPE(sstr) >= SVt_PVIV && SvTYPE(sstr) != SVt_PVFM))
4256 /* Failed the swipe test, and it's not a shared hash key either.
4257 Have to copy the string. */
4258 STRLEN len = SvCUR(sstr);
4259 SvGROW(dstr, len + 1); /* inlined from sv_setpvn */
4260 Move(SvPVX_const(sstr),SvPVX(dstr),len,char);
4261 SvCUR_set(dstr, len);
4262 *SvEND(dstr) = '\0';
4264 /* If PERL_OLD_COPY_ON_WRITE is not defined, then isSwipe will always
4266 /* Either it's a shared hash key, or it's suitable for
4267 copy-on-write or we can swipe the string. */
4269 PerlIO_printf(Perl_debug_log, "Copy on write: sstr --> dstr\n");
4273 #ifdef PERL_OLD_COPY_ON_WRITE
4275 if ((sflags & (SVf_FAKE | SVf_READONLY))
4276 != (SVf_FAKE | SVf_READONLY)) {
4277 SvREADONLY_on(sstr);
4279 /* Make the source SV into a loop of 1.
4280 (about to become 2) */
4281 SV_COW_NEXT_SV_SET(sstr, sstr);
4285 /* Initial code is common. */
4286 if (SvPVX_const(dstr)) { /* we know that dtype >= SVt_PV */
4291 /* making another shared SV. */
4292 STRLEN cur = SvCUR(sstr);
4293 STRLEN len = SvLEN(sstr);
4294 #ifdef PERL_OLD_COPY_ON_WRITE
4296 assert (SvTYPE(dstr) >= SVt_PVIV);
4297 /* SvIsCOW_normal */
4298 /* splice us in between source and next-after-source. */
4299 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4300 SV_COW_NEXT_SV_SET(sstr, dstr);
4301 SvPV_set(dstr, SvPVX_mutable(sstr));
4305 /* SvIsCOW_shared_hash */
4306 DEBUG_C(PerlIO_printf(Perl_debug_log,
4307 "Copy on write: Sharing hash\n"));
4309 assert (SvTYPE(dstr) >= SVt_PV);
4311 HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)))));
4313 SvLEN_set(dstr, len);
4314 SvCUR_set(dstr, cur);
4315 SvREADONLY_on(dstr);
4319 { /* Passes the swipe test. */
4320 SvPV_set(dstr, SvPVX_mutable(sstr));
4321 SvLEN_set(dstr, SvLEN(sstr));
4322 SvCUR_set(dstr, SvCUR(sstr));
4325 (void)SvOK_off(sstr); /* NOTE: nukes most SvFLAGS on sstr */
4326 SvPV_set(sstr, NULL);
4332 if (sflags & SVp_NOK) {
4333 SvNV_set(dstr, SvNVX(sstr));
4335 if (sflags & SVp_IOK) {
4336 SvIV_set(dstr, SvIVX(sstr));
4337 /* Must do this otherwise some other overloaded use of 0x80000000
4338 gets confused. I guess SVpbm_VALID */
4339 if (sflags & SVf_IVisUV)
4342 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_NOK|SVp_NOK|SVf_UTF8);
4344 const MAGIC * const smg = SvVSTRING_mg(sstr);
4346 sv_magic(dstr, NULL, PERL_MAGIC_vstring,
4347 smg->mg_ptr, smg->mg_len);
4348 SvRMAGICAL_on(dstr);
4352 else if (sflags & (SVp_IOK|SVp_NOK)) {
4353 (void)SvOK_off(dstr);
4354 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_IVisUV|SVf_NOK|SVp_NOK);
4355 if (sflags & SVp_IOK) {
4356 /* XXXX Do we want to set IsUV for IV(ROK)? Be extra safe... */
4357 SvIV_set(dstr, SvIVX(sstr));
4359 if (sflags & SVp_NOK) {
4360 SvNV_set(dstr, SvNVX(sstr));
4364 if (isGV_with_GP(sstr)) {
4365 gv_efullname3(dstr, MUTABLE_GV(sstr), "*");
4368 (void)SvOK_off(dstr);
4370 if (SvTAINTED(sstr))
4375 =for apidoc sv_setsv_mg
4377 Like C<sv_setsv>, but also handles 'set' magic.
4383 Perl_sv_setsv_mg(pTHX_ SV *const dstr, register SV *const sstr)
4385 PERL_ARGS_ASSERT_SV_SETSV_MG;
4387 sv_setsv(dstr,sstr);
4391 #ifdef PERL_OLD_COPY_ON_WRITE
4393 Perl_sv_setsv_cow(pTHX_ SV *dstr, SV *sstr)
4395 STRLEN cur = SvCUR(sstr);
4396 STRLEN len = SvLEN(sstr);
4397 register char *new_pv;
4399 PERL_ARGS_ASSERT_SV_SETSV_COW;
4402 PerlIO_printf(Perl_debug_log, "Fast copy on write: %p -> %p\n",
4403 (void*)sstr, (void*)dstr);
4410 if (SvTHINKFIRST(dstr))
4411 sv_force_normal_flags(dstr, SV_COW_DROP_PV);
4412 else if (SvPVX_const(dstr))
4413 Safefree(SvPVX_const(dstr));
4417 SvUPGRADE(dstr, SVt_PVIV);
4419 assert (SvPOK(sstr));
4420 assert (SvPOKp(sstr));
4421 assert (!SvIOK(sstr));
4422 assert (!SvIOKp(sstr));
4423 assert (!SvNOK(sstr));
4424 assert (!SvNOKp(sstr));
4426 if (SvIsCOW(sstr)) {
4428 if (SvLEN(sstr) == 0) {
4429 /* source is a COW shared hash key. */
4430 DEBUG_C(PerlIO_printf(Perl_debug_log,
4431 "Fast copy on write: Sharing hash\n"));
4432 new_pv = HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr))));
4435 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4437 assert ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS);
4438 SvUPGRADE(sstr, SVt_PVIV);
4439 SvREADONLY_on(sstr);
4441 DEBUG_C(PerlIO_printf(Perl_debug_log,
4442 "Fast copy on write: Converting sstr to COW\n"));
4443 SV_COW_NEXT_SV_SET(dstr, sstr);
4445 SV_COW_NEXT_SV_SET(sstr, dstr);
4446 new_pv = SvPVX_mutable(sstr);
4449 SvPV_set(dstr, new_pv);
4450 SvFLAGS(dstr) = (SVt_PVIV|SVf_POK|SVp_POK|SVf_FAKE|SVf_READONLY);
4453 SvLEN_set(dstr, len);
4454 SvCUR_set(dstr, cur);
4463 =for apidoc sv_setpvn
4465 Copies a string into an SV. The C<len> parameter indicates the number of
4466 bytes to be copied. If the C<ptr> argument is NULL the SV will become
4467 undefined. Does not handle 'set' magic. See C<sv_setpvn_mg>.
4473 Perl_sv_setpvn(pTHX_ register SV *const sv, register const char *const ptr, register const STRLEN len)
4476 register char *dptr;
4478 PERL_ARGS_ASSERT_SV_SETPVN;
4480 SV_CHECK_THINKFIRST_COW_DROP(sv);
4486 /* len is STRLEN which is unsigned, need to copy to signed */
4489 Perl_croak(aTHX_ "panic: sv_setpvn called with negative strlen");
4491 SvUPGRADE(sv, SVt_PV);
4493 dptr = SvGROW(sv, len + 1);
4494 Move(ptr,dptr,len,char);
4497 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4499 if (SvTYPE(sv) == SVt_PVCV) CvAUTOLOAD_off(sv);
4503 =for apidoc sv_setpvn_mg
4505 Like C<sv_setpvn>, but also handles 'set' magic.
4511 Perl_sv_setpvn_mg(pTHX_ register SV *const sv, register const char *const ptr, register const STRLEN len)
4513 PERL_ARGS_ASSERT_SV_SETPVN_MG;
4515 sv_setpvn(sv,ptr,len);
4520 =for apidoc sv_setpv
4522 Copies a string into an SV. The string must be null-terminated. Does not
4523 handle 'set' magic. See C<sv_setpv_mg>.
4529 Perl_sv_setpv(pTHX_ register SV *const sv, register const char *const ptr)
4532 register STRLEN len;
4534 PERL_ARGS_ASSERT_SV_SETPV;
4536 SV_CHECK_THINKFIRST_COW_DROP(sv);
4542 SvUPGRADE(sv, SVt_PV);
4544 SvGROW(sv, len + 1);
4545 Move(ptr,SvPVX(sv),len+1,char);
4547 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4549 if (SvTYPE(sv) == SVt_PVCV) CvAUTOLOAD_off(sv);
4553 =for apidoc sv_setpv_mg
4555 Like C<sv_setpv>, but also handles 'set' magic.
4561 Perl_sv_setpv_mg(pTHX_ register SV *const sv, register const char *const ptr)
4563 PERL_ARGS_ASSERT_SV_SETPV_MG;
4570 Perl_sv_sethek(pTHX_ register SV *const sv, const HEK *const hek)
4574 PERL_ARGS_ASSERT_SV_SETHEK;
4580 if (HEK_LEN(hek) == HEf_SVKEY) {
4581 sv_setsv(sv, *(SV**)HEK_KEY(hek));
4584 const int flags = HEK_FLAGS(hek);
4585 if (flags & HVhek_WASUTF8) {
4586 STRLEN utf8_len = HEK_LEN(hek);
4587 char *as_utf8 = (char *)bytes_to_utf8((U8*)HEK_KEY(hek), &utf8_len);
4588 sv_usepvn_flags(sv, as_utf8, utf8_len, SV_HAS_TRAILING_NUL);
4591 } else if (flags & (HVhek_REHASH|HVhek_UNSHARED)) {
4592 sv_setpvn(sv, HEK_KEY(hek), HEK_LEN(hek));
4595 else SvUTF8_off(sv);
4599 SvUPGRADE(sv, SVt_PV);
4600 sv_usepvn_flags(sv, (char *)HEK_KEY(share_hek_hek(hek)), HEK_LEN(hek), SV_HAS_TRAILING_NUL);
4607 else SvUTF8_off(sv);
4615 =for apidoc sv_usepvn_flags
4617 Tells an SV to use C<ptr> to find its string value. Normally the
4618 string is stored inside the SV but sv_usepvn allows the SV to use an
4619 outside string. The C<ptr> should point to memory that was allocated
4620 by C<malloc>. The string length, C<len>, must be supplied. By default
4621 this function will realloc (i.e. move) the memory pointed to by C<ptr>,
4622 so that pointer should not be freed or used by the programmer after
4623 giving it to sv_usepvn, and neither should any pointers from "behind"
4624 that pointer (e.g. ptr + 1) be used.
4626 If C<flags> & SV_SMAGIC is true, will call SvSETMAGIC. If C<flags> &
4627 SV_HAS_TRAILING_NUL is true, then C<ptr[len]> must be NUL, and the realloc
4628 will be skipped. (i.e. the buffer is actually at least 1 byte longer than
4629 C<len>, and already meets the requirements for storing in C<SvPVX>)
4635 Perl_sv_usepvn_flags(pTHX_ SV *const sv, char *ptr, const STRLEN len, const U32 flags)
4640 PERL_ARGS_ASSERT_SV_USEPVN_FLAGS;
4642 SV_CHECK_THINKFIRST_COW_DROP(sv);
4643 SvUPGRADE(sv, SVt_PV);
4646 if (flags & SV_SMAGIC)
4650 if (SvPVX_const(sv))
4654 if (flags & SV_HAS_TRAILING_NUL)
4655 assert(ptr[len] == '\0');
4658 allocate = (flags & SV_HAS_TRAILING_NUL)
4660 #ifdef Perl_safesysmalloc_size
4663 PERL_STRLEN_ROUNDUP(len + 1);
4665 if (flags & SV_HAS_TRAILING_NUL) {
4666 /* It's long enough - do nothing.
4667 Specifically Perl_newCONSTSUB is relying on this. */
4670 /* Force a move to shake out bugs in callers. */
4671 char *new_ptr = (char*)safemalloc(allocate);
4672 Copy(ptr, new_ptr, len, char);
4673 PoisonFree(ptr,len,char);
4677 ptr = (char*) saferealloc (ptr, allocate);
4680 #ifdef Perl_safesysmalloc_size
4681 SvLEN_set(sv, Perl_safesysmalloc_size(ptr));
4683 SvLEN_set(sv, allocate);
4687 if (!(flags & SV_HAS_TRAILING_NUL)) {
4690 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4692 if (flags & SV_SMAGIC)
4696 #ifdef PERL_OLD_COPY_ON_WRITE
4697 /* Need to do this *after* making the SV normal, as we need the buffer
4698 pointer to remain valid until after we've copied it. If we let go too early,
4699 another thread could invalidate it by unsharing last of the same hash key
4700 (which it can do by means other than releasing copy-on-write Svs)
4701 or by changing the other copy-on-write SVs in the loop. */
4703 S_sv_release_COW(pTHX_ register SV *sv, const char *pvx, SV *after)
4705 PERL_ARGS_ASSERT_SV_RELEASE_COW;
4707 { /* this SV was SvIsCOW_normal(sv) */
4708 /* we need to find the SV pointing to us. */
4709 SV *current = SV_COW_NEXT_SV(after);
4711 if (current == sv) {
4712 /* The SV we point to points back to us (there were only two of us
4714 Hence other SV is no longer copy on write either. */
4716 SvREADONLY_off(after);
4718 /* We need to follow the pointers around the loop. */
4720 while ((next = SV_COW_NEXT_SV(current)) != sv) {
4723 /* don't loop forever if the structure is bust, and we have
4724 a pointer into a closed loop. */
4725 assert (current != after);
4726 assert (SvPVX_const(current) == pvx);
4728 /* Make the SV before us point to the SV after us. */
4729 SV_COW_NEXT_SV_SET(current, after);
4735 =for apidoc sv_force_normal_flags
4737 Undo various types of fakery on an SV: if the PV is a shared string, make
4738 a private copy; if we're a ref, stop refing; if we're a glob, downgrade to
4739 an xpvmg; if we're a copy-on-write scalar, this is the on-write time when
4740 we do the copy, and is also used locally. If C<SV_COW_DROP_PV> is set
4741 then a copy-on-write scalar drops its PV buffer (if any) and becomes
4742 SvPOK_off rather than making a copy. (Used where this scalar is about to be
4743 set to some other value.) In addition, the C<flags> parameter gets passed to
4744 C<sv_unref_flags()> when unreffing. C<sv_force_normal> calls this function
4745 with flags set to 0.
4751 Perl_sv_force_normal_flags(pTHX_ register SV *const sv, const U32 flags)
4755 PERL_ARGS_ASSERT_SV_FORCE_NORMAL_FLAGS;
4757 #ifdef PERL_OLD_COPY_ON_WRITE
4758 if (SvREADONLY(sv)) {
4760 const char * const pvx = SvPVX_const(sv);
4761 const STRLEN len = SvLEN(sv);
4762 const STRLEN cur = SvCUR(sv);
4763 /* next COW sv in the loop. If len is 0 then this is a shared-hash
4764 key scalar, so we mustn't attempt to call SV_COW_NEXT_SV(), as
4765 we'll fail an assertion. */
4766 SV * const next = len ? SV_COW_NEXT_SV(sv) : 0;
4769 PerlIO_printf(Perl_debug_log,
4770 "Copy on write: Force normal %ld\n",
4776 /* This SV doesn't own the buffer, so need to Newx() a new one: */
4779 if (flags & SV_COW_DROP_PV) {
4780 /* OK, so we don't need to copy our buffer. */
4783 SvGROW(sv, cur + 1);
4784 Move(pvx,SvPVX(sv),cur,char);
4789 sv_release_COW(sv, pvx, next);
4791 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4797 else if (IN_PERL_RUNTIME)
4798 Perl_croak_no_modify(aTHX);
4801 if (SvREADONLY(sv)) {
4802 if (SvFAKE(sv) && !isGV_with_GP(sv)) {
4803 const char * const pvx = SvPVX_const(sv);
4804 const STRLEN len = SvCUR(sv);
4809 SvGROW(sv, len + 1);
4810 Move(pvx,SvPVX(sv),len,char);
4812 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4814 else if (IN_PERL_RUNTIME)
4815 Perl_croak_no_modify(aTHX);
4819 sv_unref_flags(sv, flags);
4820 else if (SvFAKE(sv) && isGV_with_GP(sv))
4822 else if (SvFAKE(sv) && SvTYPE(sv) == SVt_REGEXP) {
4823 /* Need to downgrade the REGEXP to a simple(r) scalar. This is analogous
4824 to sv_unglob. We only need it here, so inline it. */
4825 const svtype new_type = SvMAGIC(sv) || SvSTASH(sv) ? SVt_PVMG : SVt_PV;
4826 SV *const temp = newSV_type(new_type);
4827 void *const temp_p = SvANY(sv);
4829 if (new_type == SVt_PVMG) {
4830 SvMAGIC_set(temp, SvMAGIC(sv));
4831 SvMAGIC_set(sv, NULL);
4832 SvSTASH_set(temp, SvSTASH(sv));
4833 SvSTASH_set(sv, NULL);
4835 SvCUR_set(temp, SvCUR(sv));
4836 /* Remember that SvPVX is in the head, not the body. */
4838 SvLEN_set(temp, SvLEN(sv));
4839 /* This signals "buffer is owned by someone else" in sv_clear,
4840 which is the least effort way to stop it freeing the buffer.
4842 SvLEN_set(sv, SvLEN(sv)+1);
4844 /* Their buffer is already owned by someone else. */
4845 SvPVX(sv) = savepvn(SvPVX(sv), SvCUR(sv));
4846 SvLEN_set(temp, SvCUR(sv)+1);
4849 /* Now swap the rest of the bodies. */
4851 SvFLAGS(sv) &= ~(SVf_FAKE|SVTYPEMASK);
4852 SvFLAGS(sv) |= new_type;
4853 SvANY(sv) = SvANY(temp);
4855 SvFLAGS(temp) &= ~(SVTYPEMASK);
4856 SvFLAGS(temp) |= SVt_REGEXP|SVf_FAKE;
4857 SvANY(temp) = temp_p;
4866 Efficient removal of characters from the beginning of the string buffer.
4867 SvPOK(sv) must be true and the C<ptr> must be a pointer to somewhere inside
4868 the string buffer. The C<ptr> becomes the first character of the adjusted
4869 string. Uses the "OOK hack".
4871 Beware: after this function returns, C<ptr> and SvPVX_const(sv) may no longer
4872 refer to the same chunk of data.
4874 The unfortunate similarity of this function's name to that of Perl's C<chop>
4875 operator is strictly coincidental. This function works from the left;
4876 C<chop> works from the right.
4882 Perl_sv_chop(pTHX_ register SV *const sv, register const char *const ptr)
4893 PERL_ARGS_ASSERT_SV_CHOP;
4895 if (!ptr || !SvPOKp(sv))
4897 delta = ptr - SvPVX_const(sv);
4899 /* Nothing to do. */
4902 max_delta = SvLEN(sv) ? SvLEN(sv) : SvCUR(sv);
4903 if (delta > max_delta)
4904 Perl_croak(aTHX_ "panic: sv_chop ptr=%p, start=%p, end=%p",
4905 ptr, SvPVX_const(sv), SvPVX_const(sv) + max_delta);
4906 /* SvPVX(sv) may move in SV_CHECK_THINKFIRST(sv), so don't use ptr any more */
4907 SV_CHECK_THINKFIRST(sv);
4910 if (!SvLEN(sv)) { /* make copy of shared string */
4911 const char *pvx = SvPVX_const(sv);
4912 const STRLEN len = SvCUR(sv);
4913 SvGROW(sv, len + 1);
4914 Move(pvx,SvPVX(sv),len,char);
4917 SvFLAGS(sv) |= SVf_OOK;
4920 SvOOK_offset(sv, old_delta);
4922 SvLEN_set(sv, SvLEN(sv) - delta);
4923 SvCUR_set(sv, SvCUR(sv) - delta);
4924 SvPV_set(sv, SvPVX(sv) + delta);
4926 p = (U8 *)SvPVX_const(sv);
4929 /* how many bytes were evacuated? we will fill them with sentinel
4930 bytes, except for the part holding the new offset of course. */
4933 evacn += (old_delta < 0x100 ? 1 : 1 + sizeof(STRLEN));
4935 assert(evacn <= delta + old_delta);
4941 if (delta < 0x100) {
4945 p -= sizeof(STRLEN);
4946 Copy((U8*)&delta, p, sizeof(STRLEN), U8);
4950 /* Fill the preceding buffer with sentinals to verify that no-one is
4960 =for apidoc sv_catpvn
4962 Concatenates the string onto the end of the string which is in the SV. The
4963 C<len> indicates number of bytes to copy. If the SV has the UTF-8
4964 status set, then the bytes appended should be valid UTF-8.
4965 Handles 'get' magic, but not 'set' magic. See C<sv_catpvn_mg>.
4967 =for apidoc sv_catpvn_flags
4969 Concatenates the string onto the end of the string which is in the SV. The
4970 C<len> indicates number of bytes to copy. If the SV has the UTF-8
4971 status set, then the bytes appended should be valid UTF-8.
4972 If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<dsv> if
4973 appropriate, else not. C<sv_catpvn> and C<sv_catpvn_nomg> are implemented
4974 in terms of this function.
4980 Perl_sv_catpvn_flags(pTHX_ register SV *const dsv, register const char *sstr, register const STRLEN slen, const I32 flags)
4984 const char * const dstr = SvPV_force_flags(dsv, dlen, flags);
4986 PERL_ARGS_ASSERT_SV_CATPVN_FLAGS;
4987 assert((flags & (SV_CATBYTES|SV_CATUTF8)) != (SV_CATBYTES|SV_CATUTF8));
4989 if (!(flags & SV_CATBYTES) || !SvUTF8(dsv)) {
4990 if (flags & SV_CATUTF8 && !SvUTF8(dsv)) {
4991 sv_utf8_upgrade_flags_grow(dsv, 0, slen + 1);
4994 else SvGROW(dsv, dlen + slen + 1);
4996 sstr = SvPVX_const(dsv);
4997 Move(sstr, SvPVX(dsv) + dlen, slen, char);
4998 SvCUR_set(dsv, SvCUR(dsv) + slen);
5001 /* We inline bytes_to_utf8, to avoid an extra malloc. */
5002 const char * const send = sstr + slen;
5005 /* Something this code does not account for, which I think is
5006 impossible; it would require the same pv to be treated as
5007 bytes *and* utf8, which would indicate a bug elsewhere. */
5008 assert(sstr != dstr);
5010 SvGROW(dsv, dlen + slen * 2 + 1);
5011 d = (U8 *)SvPVX(dsv) + dlen;
5013 while (sstr < send) {
5014 const UV uv = NATIVE_TO_ASCII((U8)*sstr++);
5015 if (UNI_IS_INVARIANT(uv))
5016 *d++ = (U8)UTF_TO_NATIVE(uv);
5018 *d++ = (U8)UTF8_EIGHT_BIT_HI(uv);
5019 *d++ = (U8)UTF8_EIGHT_BIT_LO(uv);
5022 SvCUR_set(dsv, d-(const U8 *)SvPVX(dsv));
5025 (void)SvPOK_only_UTF8(dsv); /* validate pointer */
5027 if (flags & SV_SMAGIC)
5032 =for apidoc sv_catsv
5034 Concatenates the string from SV C<ssv> onto the end of the string in
5035 SV C<dsv>. Modifies C<dsv> but not C<ssv>. Handles 'get' magic, but
5036 not 'set' magic. See C<sv_catsv_mg>.
5038 =for apidoc sv_catsv_flags
5040 Concatenates the string from SV C<ssv> onto the end of the string in
5041 SV C<dsv>. Modifies C<dsv> but not C<ssv>. If C<flags> has C<SV_GMAGIC>
5042 bit set, will C<mg_get> on the SVs if appropriate, else not. C<sv_catsv>
5043 and C<sv_catsv_nomg> are implemented in terms of this function.
5048 Perl_sv_catsv_flags(pTHX_ SV *const dsv, register SV *const ssv, const I32 flags)
5052 PERL_ARGS_ASSERT_SV_CATSV_FLAGS;
5056 const char *spv = SvPV_flags_const(ssv, slen, flags);
5058 if (SvGMAGICAL(dsv) && (flags & SV_GMAGIC))
5060 sv_catpvn_flags(dsv, spv, slen,
5061 DO_UTF8(ssv) ? SV_CATUTF8 : SV_CATBYTES);
5064 if (flags & SV_SMAGIC)
5069 =for apidoc sv_catpv
5071 Concatenates the string onto the end of the string which is in the SV.
5072 If the SV has the UTF-8 status set, then the bytes appended should be
5073 valid UTF-8. Handles 'get' magic, but not 'set' magic. See C<sv_catpv_mg>.
5078 Perl_sv_catpv(pTHX_ register SV *const sv, register const char *ptr)
5081 register STRLEN len;
5085 PERL_ARGS_ASSERT_SV_CATPV;
5089 junk = SvPV_force(sv, tlen);
5091 SvGROW(sv, tlen + len + 1);
5093 ptr = SvPVX_const(sv);
5094 Move(ptr,SvPVX(sv)+tlen,len+1,char);
5095 SvCUR_set(sv, SvCUR(sv) + len);
5096 (void)SvPOK_only_UTF8(sv); /* validate pointer */
5101 =for apidoc sv_catpv_flags
5103 Concatenates the string onto the end of the string which is in the SV.
5104 If the SV has the UTF-8 status set, then the bytes appended should
5105 be valid UTF-8. If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get>
5106 on the SVs if appropriate, else not.
5112 Perl_sv_catpv_flags(pTHX_ SV *dstr, const char *sstr, const I32 flags)
5114 PERL_ARGS_ASSERT_SV_CATPV_FLAGS;
5115 sv_catpvn_flags(dstr, sstr, strlen(sstr), flags);
5119 =for apidoc sv_catpv_mg
5121 Like C<sv_catpv>, but also handles 'set' magic.
5127 Perl_sv_catpv_mg(pTHX_ register SV *const sv, register const char *const ptr)
5129 PERL_ARGS_ASSERT_SV_CATPV_MG;
5138 Creates a new SV. A non-zero C<len> parameter indicates the number of
5139 bytes of preallocated string space the SV should have. An extra byte for a
5140 trailing NUL is also reserved. (SvPOK is not set for the SV even if string
5141 space is allocated.) The reference count for the new SV is set to 1.
5143 In 5.9.3, newSV() replaces the older NEWSV() API, and drops the first
5144 parameter, I<x>, a debug aid which allowed callers to identify themselves.
5145 This aid has been superseded by a new build option, PERL_MEM_LOG (see
5146 L<perlhacktips/PERL_MEM_LOG>). The older API is still there for use in XS
5147 modules supporting older perls.
5153 Perl_newSV(pTHX_ const STRLEN len)
5160 sv_upgrade(sv, SVt_PV);
5161 SvGROW(sv, len + 1);
5166 =for apidoc sv_magicext
5168 Adds magic to an SV, upgrading it if necessary. Applies the
5169 supplied vtable and returns a pointer to the magic added.
5171 Note that C<sv_magicext> will allow things that C<sv_magic> will not.
5172 In particular, you can add magic to SvREADONLY SVs, and add more than
5173 one instance of the same 'how'.
5175 If C<namlen> is greater than zero then a C<savepvn> I<copy> of C<name> is
5176 stored, if C<namlen> is zero then C<name> is stored as-is and - as another
5177 special case - if C<(name && namlen == HEf_SVKEY)> then C<name> is assumed
5178 to contain an C<SV*> and is stored as-is with its REFCNT incremented.
5180 (This is now used as a subroutine by C<sv_magic>.)
5185 Perl_sv_magicext(pTHX_ SV *const sv, SV *const obj, const int how,
5186 const MGVTBL *const vtable, const char *const name, const I32 namlen)
5191 PERL_ARGS_ASSERT_SV_MAGICEXT;
5193 SvUPGRADE(sv, SVt_PVMG);
5194 Newxz(mg, 1, MAGIC);
5195 mg->mg_moremagic = SvMAGIC(sv);
5196 SvMAGIC_set(sv, mg);
5198 /* Sometimes a magic contains a reference loop, where the sv and
5199 object refer to each other. To prevent a reference loop that
5200 would prevent such objects being freed, we look for such loops
5201 and if we find one we avoid incrementing the object refcount.
5203 Note we cannot do this to avoid self-tie loops as intervening RV must
5204 have its REFCNT incremented to keep it in existence.
5207 if (!obj || obj == sv ||
5208 how == PERL_MAGIC_arylen ||
5209 how == PERL_MAGIC_symtab ||
5210 (SvTYPE(obj) == SVt_PVGV &&
5211 (GvSV(obj) == sv || GvHV(obj) == (const HV *)sv
5212 || GvAV(obj) == (const AV *)sv || GvCV(obj) == (const CV *)sv
5213 || GvIOp(obj) == (const IO *)sv || GvFORM(obj) == (const CV *)sv)))
5218 mg->mg_obj = SvREFCNT_inc_simple(obj);
5219 mg->mg_flags |= MGf_REFCOUNTED;
5222 /* Normal self-ties simply pass a null object, and instead of
5223 using mg_obj directly, use the SvTIED_obj macro to produce a
5224 new RV as needed. For glob "self-ties", we are tieing the PVIO
5225 with an RV obj pointing to the glob containing the PVIO. In
5226 this case, to avoid a reference loop, we need to weaken the
5230 if (how == PERL_MAGIC_tiedscalar && SvTYPE(sv) == SVt_PVIO &&
5231 obj && SvROK(obj) && GvIO(SvRV(obj)) == (const IO *)sv)
5237 mg->mg_len = namlen;
5240 mg->mg_ptr = savepvn(name, namlen);
5241 else if (namlen == HEf_SVKEY) {
5242 /* Yes, this is casting away const. This is only for the case of
5243 HEf_SVKEY. I think we need to document this aberation of the
5244 constness of the API, rather than making name non-const, as
5245 that change propagating outwards a long way. */
5246 mg->mg_ptr = (char*)SvREFCNT_inc_simple_NN((SV *)name);
5248 mg->mg_ptr = (char *) name;
5250 mg->mg_virtual = (MGVTBL *) vtable;
5254 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
5259 =for apidoc sv_magic
5261 Adds magic to an SV. First upgrades C<sv> to type C<SVt_PVMG> if necessary,
5262 then adds a new magic item of type C<how> to the head of the magic list.
5264 See C<sv_magicext> (which C<sv_magic> now calls) for a description of the
5265 handling of the C<name> and C<namlen> arguments.
5267 You need to use C<sv_magicext> to add magic to SvREADONLY SVs and also
5268 to add more than one instance of the same 'how'.
5274 Perl_sv_magic(pTHX_ register SV *const sv, SV *const obj, const int how,
5275 const char *const name, const I32 namlen)
5278 const MGVTBL *vtable;
5281 unsigned int vtable_index;
5283 PERL_ARGS_ASSERT_SV_MAGIC;
5285 if (how < 0 || (unsigned)how > C_ARRAY_LENGTH(PL_magic_data)
5286 || ((flags = PL_magic_data[how]),
5287 (vtable_index = flags & PERL_MAGIC_VTABLE_MASK)
5288 > magic_vtable_max))
5289 Perl_croak(aTHX_ "Don't know how to handle magic of type \\%o", how);
5291 /* PERL_MAGIC_ext is reserved for use by extensions not perl internals.
5292 Useful for attaching extension internal data to perl vars.
5293 Note that multiple extensions may clash if magical scalars
5294 etc holding private data from one are passed to another. */
5296 vtable = (vtable_index == magic_vtable_max)
5297 ? NULL : PL_magic_vtables + vtable_index;
5299 #ifdef PERL_OLD_COPY_ON_WRITE
5301 sv_force_normal_flags(sv, 0);
5303 if (SvREADONLY(sv)) {
5305 /* its okay to attach magic to shared strings; the subsequent
5306 * upgrade to PVMG will unshare the string */
5307 !(SvFAKE(sv) && SvTYPE(sv) < SVt_PVMG)
5310 && !PERL_MAGIC_TYPE_READONLY_ACCEPTABLE(how)
5313 Perl_croak_no_modify(aTHX);
5316 if (SvMAGICAL(sv) || (how == PERL_MAGIC_taint && SvTYPE(sv) >= SVt_PVMG)) {
5317 if (SvMAGIC(sv) && (mg = mg_find(sv, how))) {
5318 /* sv_magic() refuses to add a magic of the same 'how' as an
5321 if (how == PERL_MAGIC_taint) {
5323 /* Any scalar which already had taint magic on which someone
5324 (erroneously?) did SvIOK_on() or similar will now be
5325 incorrectly sporting public "OK" flags. */
5326 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
5332 /* Rest of work is done else where */
5333 mg = sv_magicext(sv,obj,how,vtable,name,namlen);
5336 case PERL_MAGIC_taint:
5339 case PERL_MAGIC_ext:
5340 case PERL_MAGIC_dbfile:
5347 S_sv_unmagicext_flags(pTHX_ SV *const sv, const int type, MGVTBL *vtbl, const U32 flags)
5354 if (SvTYPE(sv) < SVt_PVMG || !SvMAGIC(sv))
5356 mgp = &(((XPVMG*) SvANY(sv))->xmg_u.xmg_magic);
5357 for (mg = *mgp; mg; mg = *mgp) {
5358 const MGVTBL* const virt = mg->mg_virtual;
5359 if (mg->mg_type == type && (!flags || virt == vtbl)) {
5360 *mgp = mg->mg_moremagic;
5361 if (virt && virt->svt_free)
5362 virt->svt_free(aTHX_ sv, mg);
5363 if (mg->mg_ptr && mg->mg_type != PERL_MAGIC_regex_global) {
5365 Safefree(mg->mg_ptr);
5366 else if (mg->mg_len == HEf_SVKEY)
5367 SvREFCNT_dec(MUTABLE_SV(mg->mg_ptr));
5368 else if (mg->mg_type == PERL_MAGIC_utf8)
5369 Safefree(mg->mg_ptr);
5371 if (mg->mg_flags & MGf_REFCOUNTED)
5372 SvREFCNT_dec(mg->mg_obj);
5376 mgp = &mg->mg_moremagic;
5379 if (SvMAGICAL(sv)) /* if we're under save_magic, wait for restore_magic; */
5380 mg_magical(sv); /* else fix the flags now */
5384 SvFLAGS(sv) |= (SvFLAGS(sv) & (SVp_IOK|SVp_NOK|SVp_POK)) >> PRIVSHIFT;
5390 =for apidoc sv_unmagic
5392 Removes all magic of type C<type> from an SV.
5398 Perl_sv_unmagic(pTHX_ SV *const sv, const int type)
5400 PERL_ARGS_ASSERT_SV_UNMAGIC;
5401 return S_sv_unmagicext_flags(aTHX_ sv, type, NULL, 0);
5405 =for apidoc sv_unmagicext
5407 Removes all magic of type C<type> with the specified C<vtbl> from an SV.
5413 Perl_sv_unmagicext(pTHX_ SV *const sv, const int type, MGVTBL *vtbl)
5415 PERL_ARGS_ASSERT_SV_UNMAGICEXT;
5416 return S_sv_unmagicext_flags(aTHX_ sv, type, vtbl, 1);
5420 =for apidoc sv_rvweaken
5422 Weaken a reference: set the C<SvWEAKREF> flag on this RV; give the
5423 referred-to SV C<PERL_MAGIC_backref> magic if it hasn't already; and
5424 push a back-reference to this RV onto the array of backreferences
5425 associated with that magic. If the RV is magical, set magic will be
5426 called after the RV is cleared.
5432 Perl_sv_rvweaken(pTHX_ SV *const sv)
5436 PERL_ARGS_ASSERT_SV_RVWEAKEN;
5438 if (!SvOK(sv)) /* let undefs pass */
5441 Perl_croak(aTHX_ "Can't weaken a nonreference");
5442 else if (SvWEAKREF(sv)) {
5443 Perl_ck_warner(aTHX_ packWARN(WARN_MISC), "Reference is already weak");
5446 else if (SvREADONLY(sv)) croak_no_modify();
5448 Perl_sv_add_backref(aTHX_ tsv, sv);
5454 /* Give tsv backref magic if it hasn't already got it, then push a
5455 * back-reference to sv onto the array associated with the backref magic.
5457 * As an optimisation, if there's only one backref and it's not an AV,
5458 * store it directly in the HvAUX or mg_obj slot, avoiding the need to
5459 * allocate an AV. (Whether the slot holds an AV tells us whether this is
5463 /* A discussion about the backreferences array and its refcount:
5465 * The AV holding the backreferences is pointed to either as the mg_obj of
5466 * PERL_MAGIC_backref, or in the specific case of a HV, from the
5467 * xhv_backreferences field. The array is created with a refcount
5468 * of 2. This means that if during global destruction the array gets
5469 * picked on before its parent to have its refcount decremented by the
5470 * random zapper, it won't actually be freed, meaning it's still there for
5471 * when its parent gets freed.
5473 * When the parent SV is freed, the extra ref is killed by
5474 * Perl_sv_kill_backrefs. The other ref is killed, in the case of magic,
5475 * by mg_free() / MGf_REFCOUNTED, or for a hash, by Perl_hv_kill_backrefs.
5477 * When a single backref SV is stored directly, it is not reference
5482 Perl_sv_add_backref(pTHX_ SV *const tsv, SV *const sv)
5489 PERL_ARGS_ASSERT_SV_ADD_BACKREF;
5491 /* find slot to store array or singleton backref */
5493 if (SvTYPE(tsv) == SVt_PVHV) {
5494 svp = (SV**)Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5497 (SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL))))
5499 sv_magic(tsv, NULL, PERL_MAGIC_backref, NULL, 0);
5500 mg = mg_find(tsv, PERL_MAGIC_backref);
5502 svp = &(mg->mg_obj);
5505 /* create or retrieve the array */
5507 if ( (!*svp && SvTYPE(sv) == SVt_PVAV)
5508 || (*svp && SvTYPE(*svp) != SVt_PVAV)
5513 SvREFCNT_inc_simple_void(av);
5514 /* av now has a refcnt of 2; see discussion above */
5516 /* move single existing backref to the array */
5518 AvARRAY(av)[++AvFILLp(av)] = *svp; /* av_push() */
5522 mg->mg_flags |= MGf_REFCOUNTED;
5525 av = MUTABLE_AV(*svp);
5528 /* optimisation: store single backref directly in HvAUX or mg_obj */
5532 /* push new backref */
5533 assert(SvTYPE(av) == SVt_PVAV);
5534 if (AvFILLp(av) >= AvMAX(av)) {
5535 av_extend(av, AvFILLp(av)+1);
5537 AvARRAY(av)[++AvFILLp(av)] = sv; /* av_push() */
5540 /* delete a back-reference to ourselves from the backref magic associated
5541 * with the SV we point to.
5545 Perl_sv_del_backref(pTHX_ SV *const tsv, SV *const sv)
5550 PERL_ARGS_ASSERT_SV_DEL_BACKREF;
5552 if (SvTYPE(tsv) == SVt_PVHV) {
5554 svp = (SV**)Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5558 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
5559 svp = mg ? &(mg->mg_obj) : NULL;
5563 Perl_croak(aTHX_ "panic: del_backref");
5565 if (SvTYPE(*svp) == SVt_PVAV) {
5569 AV * const av = (AV*)*svp;
5571 assert(!SvIS_FREED(av));
5575 /* for an SV with N weak references to it, if all those
5576 * weak refs are deleted, then sv_del_backref will be called
5577 * N times and O(N^2) compares will be done within the backref
5578 * array. To ameliorate this potential slowness, we:
5579 * 1) make sure this code is as tight as possible;
5580 * 2) when looking for SV, look for it at both the head and tail of the
5581 * array first before searching the rest, since some create/destroy
5582 * patterns will cause the backrefs to be freed in order.
5589 SV **p = &svp[fill];
5590 SV *const topsv = *p;
5597 /* We weren't the last entry.
5598 An unordered list has this property that you
5599 can take the last element off the end to fill
5600 the hole, and it's still an unordered list :-)
5606 break; /* should only be one */
5613 AvFILLp(av) = fill-1;
5616 /* optimisation: only a single backref, stored directly */
5618 Perl_croak(aTHX_ "panic: del_backref");
5625 Perl_sv_kill_backrefs(pTHX_ SV *const sv, AV *const av)
5631 PERL_ARGS_ASSERT_SV_KILL_BACKREFS;
5636 /* after multiple passes through Perl_sv_clean_all() for a thinngy
5637 * that has badly leaked, the backref array may have gotten freed,
5638 * since we only protect it against 1 round of cleanup */
5639 if (SvIS_FREED(av)) {
5640 if (PL_in_clean_all) /* All is fair */
5643 "panic: magic_killbackrefs (freed backref AV/SV)");
5647 is_array = (SvTYPE(av) == SVt_PVAV);
5649 assert(!SvIS_FREED(av));
5652 last = svp + AvFILLp(av);
5655 /* optimisation: only a single backref, stored directly */
5661 while (svp <= last) {
5663 SV *const referrer = *svp;
5664 if (SvWEAKREF(referrer)) {
5665 /* XXX Should we check that it hasn't changed? */
5666 assert(SvROK(referrer));
5667 SvRV_set(referrer, 0);
5669 SvWEAKREF_off(referrer);
5670 SvSETMAGIC(referrer);
5671 } else if (SvTYPE(referrer) == SVt_PVGV ||
5672 SvTYPE(referrer) == SVt_PVLV) {
5673 assert(SvTYPE(sv) == SVt_PVHV); /* stash backref */
5674 /* You lookin' at me? */
5675 assert(GvSTASH(referrer));
5676 assert(GvSTASH(referrer) == (const HV *)sv);
5677 GvSTASH(referrer) = 0;
5678 } else if (SvTYPE(referrer) == SVt_PVCV ||
5679 SvTYPE(referrer) == SVt_PVFM) {
5680 if (SvTYPE(sv) == SVt_PVHV) { /* stash backref */
5681 /* You lookin' at me? */
5682 assert(CvSTASH(referrer));
5683 assert(CvSTASH(referrer) == (const HV *)sv);
5684 SvANY(MUTABLE_CV(referrer))->xcv_stash = 0;
5687 assert(SvTYPE(sv) == SVt_PVGV);
5688 /* You lookin' at me? */
5689 assert(CvGV(referrer));
5690 assert(CvGV(referrer) == (const GV *)sv);
5691 anonymise_cv_maybe(MUTABLE_GV(sv),
5692 MUTABLE_CV(referrer));
5697 "panic: magic_killbackrefs (flags=%"UVxf")",
5698 (UV)SvFLAGS(referrer));
5709 SvREFCNT_dec(av); /* remove extra count added by sv_add_backref() */
5715 =for apidoc sv_insert
5717 Inserts a string at the specified offset/length within the SV. Similar to
5718 the Perl substr() function. Handles get magic.
5720 =for apidoc sv_insert_flags
5722 Same as C<sv_insert>, but the extra C<flags> are passed the C<SvPV_force_flags> that applies to C<bigstr>.
5728 Perl_sv_insert_flags(pTHX_ SV *const bigstr, const STRLEN offset, const STRLEN len, const char *const little, const STRLEN littlelen, const U32 flags)
5733 register char *midend;
5734 register char *bigend;
5735 register SSize_t i; /* better be sizeof(STRLEN) or bad things happen */
5738 PERL_ARGS_ASSERT_SV_INSERT_FLAGS;
5741 Perl_croak(aTHX_ "Can't modify non-existent substring");
5742 SvPV_force_flags(bigstr, curlen, flags);
5743 (void)SvPOK_only_UTF8(bigstr);
5744 if (offset + len > curlen) {
5745 SvGROW(bigstr, offset+len+1);
5746 Zero(SvPVX(bigstr)+curlen, offset+len-curlen, char);
5747 SvCUR_set(bigstr, offset+len);
5751 i = littlelen - len;
5752 if (i > 0) { /* string might grow */
5753 big = SvGROW(bigstr, SvCUR(bigstr) + i + 1);
5754 mid = big + offset + len;
5755 midend = bigend = big + SvCUR(bigstr);
5758 while (midend > mid) /* shove everything down */
5759 *--bigend = *--midend;
5760 Move(little,big+offset,littlelen,char);
5761 SvCUR_set(bigstr, SvCUR(bigstr) + i);
5766 Move(little,SvPVX(bigstr)+offset,len,char);
5771 big = SvPVX(bigstr);
5774 bigend = big + SvCUR(bigstr);
5776 if (midend > bigend)
5777 Perl_croak(aTHX_ "panic: sv_insert");
5779 if (mid - big > bigend - midend) { /* faster to shorten from end */
5781 Move(little, mid, littlelen,char);
5784 i = bigend - midend;
5786 Move(midend, mid, i,char);
5790 SvCUR_set(bigstr, mid - big);
5792 else if ((i = mid - big)) { /* faster from front */
5793 midend -= littlelen;
5795 Move(big, midend - i, i, char);
5796 sv_chop(bigstr,midend-i);
5798 Move(little, mid, littlelen,char);
5800 else if (littlelen) {
5801 midend -= littlelen;
5802 sv_chop(bigstr,midend);
5803 Move(little,midend,littlelen,char);
5806 sv_chop(bigstr,midend);
5812 =for apidoc sv_replace
5814 Make the first argument a copy of the second, then delete the original.
5815 The target SV physically takes over ownership of the body of the source SV
5816 and inherits its flags; however, the target keeps any magic it owns,
5817 and any magic in the source is discarded.
5818 Note that this is a rather specialist SV copying operation; most of the
5819 time you'll want to use C<sv_setsv> or one of its many macro front-ends.
5825 Perl_sv_replace(pTHX_ register SV *const sv, register SV *const nsv)
5828 const U32 refcnt = SvREFCNT(sv);
5830 PERL_ARGS_ASSERT_SV_REPLACE;
5832 SV_CHECK_THINKFIRST_COW_DROP(sv);
5833 if (SvREFCNT(nsv) != 1) {
5834 Perl_croak(aTHX_ "panic: reference miscount on nsv in sv_replace()"
5835 " (%" UVuf " != 1)", (UV) SvREFCNT(nsv));
5837 if (SvMAGICAL(sv)) {
5841 sv_upgrade(nsv, SVt_PVMG);
5842 SvMAGIC_set(nsv, SvMAGIC(sv));
5843 SvFLAGS(nsv) |= SvMAGICAL(sv);
5845 SvMAGIC_set(sv, NULL);
5849 assert(!SvREFCNT(sv));
5850 #ifdef DEBUG_LEAKING_SCALARS
5851 sv->sv_flags = nsv->sv_flags;
5852 sv->sv_any = nsv->sv_any;
5853 sv->sv_refcnt = nsv->sv_refcnt;
5854 sv->sv_u = nsv->sv_u;
5856 StructCopy(nsv,sv,SV);
5858 if(SvTYPE(sv) == SVt_IV) {
5860 = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
5864 #ifdef PERL_OLD_COPY_ON_WRITE
5865 if (SvIsCOW_normal(nsv)) {
5866 /* We need to follow the pointers around the loop to make the
5867 previous SV point to sv, rather than nsv. */
5870 while ((next = SV_COW_NEXT_SV(current)) != nsv) {
5873 assert(SvPVX_const(current) == SvPVX_const(nsv));
5875 /* Make the SV before us point to the SV after us. */
5877 PerlIO_printf(Perl_debug_log, "previous is\n");
5879 PerlIO_printf(Perl_debug_log,
5880 "move it from 0x%"UVxf" to 0x%"UVxf"\n",
5881 (UV) SV_COW_NEXT_SV(current), (UV) sv);
5883 SV_COW_NEXT_SV_SET(current, sv);
5886 SvREFCNT(sv) = refcnt;
5887 SvFLAGS(nsv) |= SVTYPEMASK; /* Mark as freed */
5892 /* We're about to free a GV which has a CV that refers back to us.
5893 * If that CV will outlive us, make it anonymous (i.e. fix up its CvGV
5897 S_anonymise_cv_maybe(pTHX_ GV *gv, CV* cv)
5902 PERL_ARGS_ASSERT_ANONYMISE_CV_MAYBE;
5905 assert(SvREFCNT(gv) == 0);
5906 assert(isGV(gv) && isGV_with_GP(gv));
5908 assert(!CvANON(cv));
5909 assert(CvGV(cv) == gv);
5911 /* will the CV shortly be freed by gp_free() ? */
5912 if (GvCV(gv) == cv && GvGP(gv)->gp_refcnt < 2 && SvREFCNT(cv) < 2) {
5913 SvANY(cv)->xcv_gv = NULL;
5917 /* if not, anonymise: */
5918 gvname = (GvSTASH(gv) && HvNAME(GvSTASH(gv)) && HvENAME(GvSTASH(gv)))
5919 ? newSVhek(HvENAME_HEK(GvSTASH(gv)))
5920 : newSVpvn_flags( "__ANON__", 8, 0 );
5921 sv_catpvs(gvname, "::__ANON__");
5922 anongv = gv_fetchsv(gvname, GV_ADDMULTI, SVt_PVCV);
5923 SvREFCNT_dec(gvname);
5927 SvANY(cv)->xcv_gv = MUTABLE_GV(SvREFCNT_inc(anongv));
5932 =for apidoc sv_clear
5934 Clear an SV: call any destructors, free up any memory used by the body,
5935 and free the body itself. The SV's head is I<not> freed, although
5936 its type is set to all 1's so that it won't inadvertently be assumed
5937 to be live during global destruction etc.
5938 This function should only be called when REFCNT is zero. Most of the time
5939 you'll want to call C<sv_free()> (or its macro wrapper C<SvREFCNT_dec>)
5946 Perl_sv_clear(pTHX_ SV *const orig_sv)
5951 const struct body_details *sv_type_details;
5954 register SV *sv = orig_sv;
5957 PERL_ARGS_ASSERT_SV_CLEAR;
5959 /* within this loop, sv is the SV currently being freed, and
5960 * iter_sv is the most recent AV or whatever that's being iterated
5961 * over to provide more SVs */
5967 assert(SvREFCNT(sv) == 0);
5968 assert(SvTYPE(sv) != (svtype)SVTYPEMASK);
5970 if (type <= SVt_IV) {
5971 /* See the comment in sv.h about the collusion between this
5972 * early return and the overloading of the NULL slots in the
5976 SvFLAGS(sv) &= SVf_BREAK;
5977 SvFLAGS(sv) |= SVTYPEMASK;
5981 assert(!SvOBJECT(sv) || type >= SVt_PVMG); /* objs are always >= MG */
5983 if (type >= SVt_PVMG) {
5985 if (!curse(sv, 1)) goto get_next_sv;
5986 type = SvTYPE(sv); /* destructor may have changed it */
5988 /* Free back-references before magic, in case the magic calls
5989 * Perl code that has weak references to sv. */
5990 if (type == SVt_PVHV) {
5991 Perl_hv_kill_backrefs(aTHX_ MUTABLE_HV(sv));
5995 else if (type == SVt_PVMG && SvPAD_OUR(sv)) {
5996 SvREFCNT_dec(SvOURSTASH(sv));
5997 } else if (SvMAGIC(sv)) {
5998 /* Free back-references before other types of magic. */
5999 sv_unmagic(sv, PERL_MAGIC_backref);
6002 if (type == SVt_PVMG && SvPAD_TYPED(sv))
6003 SvREFCNT_dec(SvSTASH(sv));
6006 /* case SVt_BIND: */
6009 IoIFP(sv) != PerlIO_stdin() &&
6010 IoIFP(sv) != PerlIO_stdout() &&
6011 IoIFP(sv) != PerlIO_stderr() &&
6012 !(IoFLAGS(sv) & IOf_FAKE_DIRP))
6014 io_close(MUTABLE_IO(sv), FALSE);
6016 if (IoDIRP(sv) && !(IoFLAGS(sv) & IOf_FAKE_DIRP))
6017 PerlDir_close(IoDIRP(sv));
6018 IoDIRP(sv) = (DIR*)NULL;
6019 Safefree(IoTOP_NAME(sv));
6020 Safefree(IoFMT_NAME(sv));
6021 Safefree(IoBOTTOM_NAME(sv));
6024 /* FIXME for plugins */
6025 pregfree2((REGEXP*) sv);
6029 cv_undef(MUTABLE_CV(sv));
6030 /* If we're in a stash, we don't own a reference to it.
6031 * However it does have a back reference to us, which needs to
6033 if ((stash = CvSTASH(sv)))
6034 sv_del_backref(MUTABLE_SV(stash), sv);
6037 if (PL_last_swash_hv == (const HV *)sv) {
6038 PL_last_swash_hv = NULL;
6040 if (HvTOTALKEYS((HV*)sv) > 0) {
6042 /* this statement should match the one at the beginning of
6043 * hv_undef_flags() */
6044 if ( PL_phase != PERL_PHASE_DESTRUCT
6045 && (name = HvNAME((HV*)sv)))
6048 (void)hv_delete(PL_stashcache, name,
6049 HvNAMEUTF8((HV*)sv) ? -HvNAMELEN_get((HV*)sv) : HvNAMELEN_get((HV*)sv), G_DISCARD);
6050 hv_name_set((HV*)sv, NULL, 0, 0);
6053 /* save old iter_sv in unused SvSTASH field */
6054 assert(!SvOBJECT(sv));
6055 SvSTASH(sv) = (HV*)iter_sv;
6058 /* XXX ideally we should save the old value of hash_index
6059 * too, but I can't think of any place to hide it. The
6060 * effect of not saving it is that for freeing hashes of
6061 * hashes, we become quadratic in scanning the HvARRAY of
6062 * the top hash looking for new entries to free; but
6063 * hopefully this will be dwarfed by the freeing of all
6064 * the nested hashes. */
6066 next_sv = Perl_hfree_next_entry(aTHX_ (HV*)sv, &hash_index);
6067 goto get_next_sv; /* process this new sv */
6069 /* free empty hash */
6070 Perl_hv_undef_flags(aTHX_ MUTABLE_HV(sv), HV_NAME_SETALL);
6071 assert(!HvARRAY((HV*)sv));
6075 AV* av = MUTABLE_AV(sv);
6076 if (PL_comppad == av) {
6080 if (AvREAL(av) && AvFILLp(av) > -1) {
6081 next_sv = AvARRAY(av)[AvFILLp(av)--];
6082 /* save old iter_sv in top-most slot of AV,
6083 * and pray that it doesn't get wiped in the meantime */
6084 AvARRAY(av)[AvMAX(av)] = iter_sv;
6086 goto get_next_sv; /* process this new sv */
6088 Safefree(AvALLOC(av));
6093 if (LvTYPE(sv) == 'T') { /* for tie: return HE to pool */
6094 SvREFCNT_dec(HeKEY_sv((HE*)LvTARG(sv)));
6095 HeNEXT((HE*)LvTARG(sv)) = PL_hv_fetch_ent_mh;
6096 PL_hv_fetch_ent_mh = (HE*)LvTARG(sv);
6098 else if (LvTYPE(sv) != 't') /* unless tie: unrefcnted fake SV** */
6099 SvREFCNT_dec(LvTARG(sv));
6101 if (isGV_with_GP(sv)) {
6102 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
6103 && HvENAME_get(stash))
6104 mro_method_changed_in(stash);
6105 gp_free(MUTABLE_GV(sv));
6107 unshare_hek(GvNAME_HEK(sv));
6108 /* If we're in a stash, we don't own a reference to it.
6109 * However it does have a back reference to us, which
6110 * needs to be cleared. */
6111 if (!SvVALID(sv) && (stash = GvSTASH(sv)))
6112 sv_del_backref(MUTABLE_SV(stash), sv);
6114 /* FIXME. There are probably more unreferenced pointers to SVs
6115 * in the interpreter struct that we should check and tidy in
6116 * a similar fashion to this: */
6117 if ((const GV *)sv == PL_last_in_gv)
6118 PL_last_in_gv = NULL;
6124 /* Don't bother with SvOOK_off(sv); as we're only going to
6128 SvOOK_offset(sv, offset);
6129 SvPV_set(sv, SvPVX_mutable(sv) - offset);
6130 /* Don't even bother with turning off the OOK flag. */
6135 SV * const target = SvRV(sv);
6137 sv_del_backref(target, sv);
6142 #ifdef PERL_OLD_COPY_ON_WRITE
6143 else if (SvPVX_const(sv)
6144 && !(SvTYPE(sv) == SVt_PVIO
6145 && !(IoFLAGS(sv) & IOf_FAKE_DIRP)))
6149 PerlIO_printf(Perl_debug_log, "Copy on write: clear\n");
6153 sv_release_COW(sv, SvPVX_const(sv), SV_COW_NEXT_SV(sv));
6155 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
6159 } else if (SvLEN(sv)) {
6160 Safefree(SvPVX_const(sv));
6164 else if (SvPVX_const(sv) && SvLEN(sv)
6165 && !(SvTYPE(sv) == SVt_PVIO
6166 && !(IoFLAGS(sv) & IOf_FAKE_DIRP)))
6167 Safefree(SvPVX_mutable(sv));
6168 else if (SvPVX_const(sv) && SvREADONLY(sv) && SvFAKE(sv)) {
6169 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
6180 SvFLAGS(sv) &= SVf_BREAK;
6181 SvFLAGS(sv) |= SVTYPEMASK;
6183 sv_type_details = bodies_by_type + type;
6184 if (sv_type_details->arena) {
6185 del_body(((char *)SvANY(sv) + sv_type_details->offset),
6186 &PL_body_roots[type]);
6188 else if (sv_type_details->body_size) {
6189 safefree(SvANY(sv));
6193 /* caller is responsible for freeing the head of the original sv */
6194 if (sv != orig_sv && !SvREFCNT(sv))
6197 /* grab and free next sv, if any */
6205 else if (!iter_sv) {
6207 } else if (SvTYPE(iter_sv) == SVt_PVAV) {
6208 AV *const av = (AV*)iter_sv;
6209 if (AvFILLp(av) > -1) {
6210 sv = AvARRAY(av)[AvFILLp(av)--];
6212 else { /* no more elements of current AV to free */
6215 /* restore previous value, squirrelled away */
6216 iter_sv = AvARRAY(av)[AvMAX(av)];
6217 Safefree(AvALLOC(av));
6220 } else if (SvTYPE(iter_sv) == SVt_PVHV) {
6221 sv = Perl_hfree_next_entry(aTHX_ (HV*)iter_sv, &hash_index);
6222 if (!sv && !HvTOTALKEYS((HV *)iter_sv)) {
6223 /* no more elements of current HV to free */
6226 /* Restore previous value of iter_sv, squirrelled away */
6227 assert(!SvOBJECT(sv));
6228 iter_sv = (SV*)SvSTASH(sv);
6230 /* ideally we should restore the old hash_index here,
6231 * but we don't currently save the old value */
6234 /* free any remaining detritus from the hash struct */
6235 Perl_hv_undef_flags(aTHX_ MUTABLE_HV(sv), HV_NAME_SETALL);
6236 assert(!HvARRAY((HV*)sv));
6241 /* unrolled SvREFCNT_dec and sv_free2 follows: */
6245 if (!SvREFCNT(sv)) {
6249 if (--(SvREFCNT(sv)))
6253 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEBUGGING),
6254 "Attempt to free temp prematurely: SV 0x%"UVxf
6255 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
6259 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
6260 /* make sure SvREFCNT(sv)==0 happens very seldom */
6261 SvREFCNT(sv) = (~(U32)0)/2;
6270 /* This routine curses the sv itself, not the object referenced by sv. So
6271 sv does not have to be ROK. */
6274 S_curse(pTHX_ SV * const sv, const bool check_refcnt) {
6277 PERL_ARGS_ASSERT_CURSE;
6278 assert(SvOBJECT(sv));
6280 if (PL_defstash && /* Still have a symbol table? */
6287 stash = SvSTASH(sv);
6288 destructor = StashHANDLER(stash,DESTROY);
6290 /* A constant subroutine can have no side effects, so
6291 don't bother calling it. */
6292 && !CvCONST(destructor)
6293 /* Don't bother calling an empty destructor */
6294 && (CvISXSUB(destructor)
6295 || (CvSTART(destructor)
6296 && (CvSTART(destructor)->op_next->op_type
6299 SV* const tmpref = newRV(sv);
6300 SvREADONLY_on(tmpref); /* DESTROY() could be naughty */
6302 PUSHSTACKi(PERLSI_DESTROY);
6307 call_sv(MUTABLE_SV(destructor),
6308 G_DISCARD|G_EVAL|G_KEEPERR|G_VOID);
6312 if(SvREFCNT(tmpref) < 2) {
6313 /* tmpref is not kept alive! */
6315 SvRV_set(tmpref, NULL);
6318 SvREFCNT_dec(tmpref);
6320 } while (SvOBJECT(sv) && SvSTASH(sv) != stash);
6323 if (check_refcnt && SvREFCNT(sv)) {
6324 if (PL_in_clean_objs)
6326 "DESTROY created new reference to dead object '%"HEKf"'",
6327 HEKfARG(HvNAME_HEK(stash)));
6328 /* DESTROY gave object new lease on life */
6334 SvREFCNT_dec(SvSTASH(sv)); /* possibly of changed persuasion */
6335 SvOBJECT_off(sv); /* Curse the object. */
6336 if (SvTYPE(sv) != SVt_PVIO)
6337 --PL_sv_objcount;/* XXX Might want something more general */
6343 =for apidoc sv_newref
6345 Increment an SV's reference count. Use the C<SvREFCNT_inc()> wrapper
6352 Perl_sv_newref(pTHX_ SV *const sv)
6354 PERL_UNUSED_CONTEXT;
6363 Decrement an SV's reference count, and if it drops to zero, call
6364 C<sv_clear> to invoke destructors and free up any memory used by
6365 the body; finally, deallocate the SV's head itself.
6366 Normally called via a wrapper macro C<SvREFCNT_dec>.
6372 Perl_sv_free(pTHX_ SV *const sv)
6377 if (SvREFCNT(sv) == 0) {
6378 if (SvFLAGS(sv) & SVf_BREAK)
6379 /* this SV's refcnt has been artificially decremented to
6380 * trigger cleanup */
6382 if (PL_in_clean_all) /* All is fair */
6384 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
6385 /* make sure SvREFCNT(sv)==0 happens very seldom */
6386 SvREFCNT(sv) = (~(U32)0)/2;
6389 if (ckWARN_d(WARN_INTERNAL)) {
6390 #ifdef DEBUG_LEAKING_SCALARS_FORK_DUMP
6391 Perl_dump_sv_child(aTHX_ sv);
6393 #ifdef DEBUG_LEAKING_SCALARS
6396 #ifdef DEBUG_LEAKING_SCALARS_ABORT
6397 if (PL_warnhook == PERL_WARNHOOK_FATAL
6398 || ckDEAD(packWARN(WARN_INTERNAL))) {
6399 /* Don't let Perl_warner cause us to escape our fate: */
6403 /* This may not return: */
6404 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
6405 "Attempt to free unreferenced scalar: SV 0x%"UVxf
6406 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
6409 #ifdef DEBUG_LEAKING_SCALARS_ABORT
6414 if (--(SvREFCNT(sv)) > 0)
6416 Perl_sv_free2(aTHX_ sv);
6420 Perl_sv_free2(pTHX_ SV *const sv)
6424 PERL_ARGS_ASSERT_SV_FREE2;
6428 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEBUGGING),
6429 "Attempt to free temp prematurely: SV 0x%"UVxf
6430 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
6434 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
6435 /* make sure SvREFCNT(sv)==0 happens very seldom */
6436 SvREFCNT(sv) = (~(U32)0)/2;
6447 Returns the length of the string in the SV. Handles magic and type
6448 coercion. See also C<SvCUR>, which gives raw access to the xpv_cur slot.
6454 Perl_sv_len(pTHX_ register SV *const sv)
6462 len = mg_length(sv);
6464 (void)SvPV_const(sv, len);
6469 =for apidoc sv_len_utf8
6471 Returns the number of characters in the string in an SV, counting wide
6472 UTF-8 bytes as a single character. Handles magic and type coercion.
6478 * The length is cached in PERL_MAGIC_utf8, in the mg_len field. Also the
6479 * mg_ptr is used, by sv_pos_u2b() and sv_pos_b2u() - see the comments below.
6480 * (Note that the mg_len is not the length of the mg_ptr field.
6481 * This allows the cache to store the character length of the string without
6482 * needing to malloc() extra storage to attach to the mg_ptr.)
6487 Perl_sv_len_utf8(pTHX_ register SV *const sv)
6493 return mg_length(sv);
6497 const U8 *s = (U8*)SvPV_const(sv, len);
6501 MAGIC *mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_utf8) : NULL;
6503 if (mg && (mg->mg_len != -1 || mg->mg_ptr)) {
6504 if (mg->mg_len != -1)
6507 /* We can use the offset cache for a headstart.
6508 The longer value is stored in the first pair. */
6509 STRLEN *cache = (STRLEN *) mg->mg_ptr;
6511 ulen = cache[0] + Perl_utf8_length(aTHX_ s + cache[1],
6515 if (PL_utf8cache < 0) {
6516 const STRLEN real = Perl_utf8_length(aTHX_ s, s + len);
6517 assert_uft8_cache_coherent("sv_len_utf8", ulen, real, sv);
6521 ulen = Perl_utf8_length(aTHX_ s, s + len);
6522 utf8_mg_len_cache_update(sv, &mg, ulen);
6526 return Perl_utf8_length(aTHX_ s, s + len);
6530 /* Walk forwards to find the byte corresponding to the passed in UTF-8
6533 S_sv_pos_u2b_forwards(const U8 *const start, const U8 *const send,
6534 STRLEN *const uoffset_p, bool *const at_end)
6536 const U8 *s = start;
6537 STRLEN uoffset = *uoffset_p;
6539 PERL_ARGS_ASSERT_SV_POS_U2B_FORWARDS;
6541 while (s < send && uoffset) {
6548 else if (s > send) {
6550 /* This is the existing behaviour. Possibly it should be a croak, as
6551 it's actually a bounds error */
6554 *uoffset_p -= uoffset;
6558 /* Given the length of the string in both bytes and UTF-8 characters, decide
6559 whether to walk forwards or backwards to find the byte corresponding to
6560 the passed in UTF-8 offset. */
6562 S_sv_pos_u2b_midway(const U8 *const start, const U8 *send,
6563 STRLEN uoffset, const STRLEN uend)
6565 STRLEN backw = uend - uoffset;
6567 PERL_ARGS_ASSERT_SV_POS_U2B_MIDWAY;
6569 if (uoffset < 2 * backw) {
6570 /* The assumption is that going forwards is twice the speed of going
6571 forward (that's where the 2 * backw comes from).
6572 (The real figure of course depends on the UTF-8 data.) */
6573 const U8 *s = start;
6575 while (s < send && uoffset--)
6585 while (UTF8_IS_CONTINUATION(*send))
6588 return send - start;
6591 /* For the string representation of the given scalar, find the byte
6592 corresponding to the passed in UTF-8 offset. uoffset0 and boffset0
6593 give another position in the string, *before* the sought offset, which
6594 (which is always true, as 0, 0 is a valid pair of positions), which should
6595 help reduce the amount of linear searching.
6596 If *mgp is non-NULL, it should point to the UTF-8 cache magic, which
6597 will be used to reduce the amount of linear searching. The cache will be
6598 created if necessary, and the found value offered to it for update. */
6600 S_sv_pos_u2b_cached(pTHX_ SV *const sv, MAGIC **const mgp, const U8 *const start,
6601 const U8 *const send, STRLEN uoffset,
6602 STRLEN uoffset0, STRLEN boffset0)
6604 STRLEN boffset = 0; /* Actually always set, but let's keep gcc happy. */
6606 bool at_end = FALSE;
6608 PERL_ARGS_ASSERT_SV_POS_U2B_CACHED;
6610 assert (uoffset >= uoffset0);
6617 && (*mgp || (SvTYPE(sv) >= SVt_PVMG &&
6618 (*mgp = mg_find(sv, PERL_MAGIC_utf8))))) {
6619 if ((*mgp)->mg_ptr) {
6620 STRLEN *cache = (STRLEN *) (*mgp)->mg_ptr;
6621 if (cache[0] == uoffset) {
6622 /* An exact match. */
6625 if (cache[2] == uoffset) {
6626 /* An exact match. */
6630 if (cache[0] < uoffset) {
6631 /* The cache already knows part of the way. */
6632 if (cache[0] > uoffset0) {
6633 /* The cache knows more than the passed in pair */
6634 uoffset0 = cache[0];
6635 boffset0 = cache[1];
6637 if ((*mgp)->mg_len != -1) {
6638 /* And we know the end too. */
6640 + sv_pos_u2b_midway(start + boffset0, send,
6642 (*mgp)->mg_len - uoffset0);
6644 uoffset -= uoffset0;
6646 + sv_pos_u2b_forwards(start + boffset0,
6647 send, &uoffset, &at_end);
6648 uoffset += uoffset0;
6651 else if (cache[2] < uoffset) {
6652 /* We're between the two cache entries. */
6653 if (cache[2] > uoffset0) {
6654 /* and the cache knows more than the passed in pair */
6655 uoffset0 = cache[2];
6656 boffset0 = cache[3];
6660 + sv_pos_u2b_midway(start + boffset0,
6663 cache[0] - uoffset0);
6666 + sv_pos_u2b_midway(start + boffset0,
6669 cache[2] - uoffset0);
6673 else if ((*mgp)->mg_len != -1) {
6674 /* If we can take advantage of a passed in offset, do so. */
6675 /* In fact, offset0 is either 0, or less than offset, so don't
6676 need to worry about the other possibility. */
6678 + sv_pos_u2b_midway(start + boffset0, send,
6680 (*mgp)->mg_len - uoffset0);
6685 if (!found || PL_utf8cache < 0) {
6686 STRLEN real_boffset;
6687 uoffset -= uoffset0;
6688 real_boffset = boffset0 + sv_pos_u2b_forwards(start + boffset0,
6689 send, &uoffset, &at_end);
6690 uoffset += uoffset0;
6692 if (found && PL_utf8cache < 0)
6693 assert_uft8_cache_coherent("sv_pos_u2b_cache", boffset,
6695 boffset = real_boffset;
6700 utf8_mg_len_cache_update(sv, mgp, uoffset);
6702 utf8_mg_pos_cache_update(sv, mgp, boffset, uoffset, send - start);
6709 =for apidoc sv_pos_u2b_flags
6711 Converts the value pointed to by offsetp from a count of UTF-8 chars from
6712 the start of the string, to a count of the equivalent number of bytes; if
6713 lenp is non-zero, it does the same to lenp, but this time starting from
6714 the offset, rather than from the start of the string. Handles type coercion.
6715 I<flags> is passed to C<SvPV_flags>, and usually should be
6716 C<SV_GMAGIC|SV_CONST_RETURN> to handle magic.
6722 * sv_pos_u2b_flags() uses, like sv_pos_b2u(), the mg_ptr of the potential
6723 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6724 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
6729 Perl_sv_pos_u2b_flags(pTHX_ SV *const sv, STRLEN uoffset, STRLEN *const lenp,
6736 PERL_ARGS_ASSERT_SV_POS_U2B_FLAGS;
6738 start = (U8*)SvPV_flags(sv, len, flags);
6740 const U8 * const send = start + len;
6742 boffset = sv_pos_u2b_cached(sv, &mg, start, send, uoffset, 0, 0);
6745 && *lenp /* don't bother doing work for 0, as its bytes equivalent
6746 is 0, and *lenp is already set to that. */) {
6747 /* Convert the relative offset to absolute. */
6748 const STRLEN uoffset2 = uoffset + *lenp;
6749 const STRLEN boffset2
6750 = sv_pos_u2b_cached(sv, &mg, start, send, uoffset2,
6751 uoffset, boffset) - boffset;
6765 =for apidoc sv_pos_u2b
6767 Converts the value pointed to by offsetp from a count of UTF-8 chars from
6768 the start of the string, to a count of the equivalent number of bytes; if
6769 lenp is non-zero, it does the same to lenp, but this time starting from
6770 the offset, rather than from the start of the string. Handles magic and
6773 Use C<sv_pos_u2b_flags> in preference, which correctly handles strings longer
6780 * sv_pos_u2b() uses, like sv_pos_b2u(), the mg_ptr of the potential
6781 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6782 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
6786 /* This function is subject to size and sign problems */
6789 Perl_sv_pos_u2b(pTHX_ register SV *const sv, I32 *const offsetp, I32 *const lenp)
6791 PERL_ARGS_ASSERT_SV_POS_U2B;
6794 STRLEN ulen = (STRLEN)*lenp;
6795 *offsetp = (I32)sv_pos_u2b_flags(sv, (STRLEN)*offsetp, &ulen,
6796 SV_GMAGIC|SV_CONST_RETURN);
6799 *offsetp = (I32)sv_pos_u2b_flags(sv, (STRLEN)*offsetp, NULL,
6800 SV_GMAGIC|SV_CONST_RETURN);
6805 S_utf8_mg_len_cache_update(pTHX_ SV *const sv, MAGIC **const mgp,
6808 PERL_ARGS_ASSERT_UTF8_MG_LEN_CACHE_UPDATE;
6812 if (!*mgp && (SvTYPE(sv) < SVt_PVMG ||
6813 !(*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
6814 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, &PL_vtbl_utf8, 0, 0);
6818 (*mgp)->mg_len = ulen;
6819 /* For now, treat "overflowed" as "still unknown". See RT #72924. */
6820 if (ulen != (STRLEN) (*mgp)->mg_len)
6821 (*mgp)->mg_len = -1;
6824 /* Create and update the UTF8 magic offset cache, with the proffered utf8/
6825 byte length pairing. The (byte) length of the total SV is passed in too,
6826 as blen, because for some (more esoteric) SVs, the call to SvPV_const()
6827 may not have updated SvCUR, so we can't rely on reading it directly.
6829 The proffered utf8/byte length pairing isn't used if the cache already has
6830 two pairs, and swapping either for the proffered pair would increase the
6831 RMS of the intervals between known byte offsets.
6833 The cache itself consists of 4 STRLEN values
6834 0: larger UTF-8 offset
6835 1: corresponding byte offset
6836 2: smaller UTF-8 offset
6837 3: corresponding byte offset
6839 Unused cache pairs have the value 0, 0.
6840 Keeping the cache "backwards" means that the invariant of
6841 cache[0] >= cache[2] is maintained even with empty slots, which means that
6842 the code that uses it doesn't need to worry if only 1 entry has actually
6843 been set to non-zero. It also makes the "position beyond the end of the
6844 cache" logic much simpler, as the first slot is always the one to start
6848 S_utf8_mg_pos_cache_update(pTHX_ SV *const sv, MAGIC **const mgp, const STRLEN byte,
6849 const STRLEN utf8, const STRLEN blen)
6853 PERL_ARGS_ASSERT_UTF8_MG_POS_CACHE_UPDATE;
6858 if (!*mgp && (SvTYPE(sv) < SVt_PVMG ||
6859 !(*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
6860 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, (MGVTBL*)&PL_vtbl_utf8, 0,
6862 (*mgp)->mg_len = -1;
6866 if (!(cache = (STRLEN *)(*mgp)->mg_ptr)) {
6867 Newxz(cache, PERL_MAGIC_UTF8_CACHESIZE * 2, STRLEN);
6868 (*mgp)->mg_ptr = (char *) cache;
6872 if (PL_utf8cache < 0 && SvPOKp(sv)) {
6873 /* SvPOKp() because it's possible that sv has string overloading, and
6874 therefore is a reference, hence SvPVX() is actually a pointer.
6875 This cures the (very real) symptoms of RT 69422, but I'm not actually
6876 sure whether we should even be caching the results of UTF-8
6877 operations on overloading, given that nothing stops overloading
6878 returning a different value every time it's called. */
6879 const U8 *start = (const U8 *) SvPVX_const(sv);
6880 const STRLEN realutf8 = utf8_length(start, start + byte);
6882 assert_uft8_cache_coherent("utf8_mg_pos_cache_update", utf8, realutf8,
6886 /* Cache is held with the later position first, to simplify the code
6887 that deals with unbounded ends. */
6889 ASSERT_UTF8_CACHE(cache);
6890 if (cache[1] == 0) {
6891 /* Cache is totally empty */
6894 } else if (cache[3] == 0) {
6895 if (byte > cache[1]) {
6896 /* New one is larger, so goes first. */
6897 cache[2] = cache[0];
6898 cache[3] = cache[1];
6906 #define THREEWAY_SQUARE(a,b,c,d) \
6907 ((float)((d) - (c))) * ((float)((d) - (c))) \
6908 + ((float)((c) - (b))) * ((float)((c) - (b))) \
6909 + ((float)((b) - (a))) * ((float)((b) - (a)))
6911 /* Cache has 2 slots in use, and we know three potential pairs.
6912 Keep the two that give the lowest RMS distance. Do the
6913 calculation in bytes simply because we always know the byte
6914 length. squareroot has the same ordering as the positive value,
6915 so don't bother with the actual square root. */
6916 const float existing = THREEWAY_SQUARE(0, cache[3], cache[1], blen);
6917 if (byte > cache[1]) {
6918 /* New position is after the existing pair of pairs. */
6919 const float keep_earlier
6920 = THREEWAY_SQUARE(0, cache[3], byte, blen);
6921 const float keep_later
6922 = THREEWAY_SQUARE(0, cache[1], byte, blen);
6924 if (keep_later < keep_earlier) {
6925 if (keep_later < existing) {
6926 cache[2] = cache[0];
6927 cache[3] = cache[1];
6933 if (keep_earlier < existing) {
6939 else if (byte > cache[3]) {
6940 /* New position is between the existing pair of pairs. */
6941 const float keep_earlier
6942 = THREEWAY_SQUARE(0, cache[3], byte, blen);
6943 const float keep_later
6944 = THREEWAY_SQUARE(0, byte, cache[1], blen);
6946 if (keep_later < keep_earlier) {
6947 if (keep_later < existing) {
6953 if (keep_earlier < existing) {
6960 /* New position is before the existing pair of pairs. */
6961 const float keep_earlier
6962 = THREEWAY_SQUARE(0, byte, cache[3], blen);
6963 const float keep_later
6964 = THREEWAY_SQUARE(0, byte, cache[1], blen);
6966 if (keep_later < keep_earlier) {
6967 if (keep_later < existing) {
6973 if (keep_earlier < existing) {
6974 cache[0] = cache[2];
6975 cache[1] = cache[3];
6982 ASSERT_UTF8_CACHE(cache);
6985 /* We already know all of the way, now we may be able to walk back. The same
6986 assumption is made as in S_sv_pos_u2b_midway(), namely that walking
6987 backward is half the speed of walking forward. */
6989 S_sv_pos_b2u_midway(pTHX_ const U8 *const s, const U8 *const target,
6990 const U8 *end, STRLEN endu)
6992 const STRLEN forw = target - s;
6993 STRLEN backw = end - target;
6995 PERL_ARGS_ASSERT_SV_POS_B2U_MIDWAY;
6997 if (forw < 2 * backw) {
6998 return utf8_length(s, target);
7001 while (end > target) {
7003 while (UTF8_IS_CONTINUATION(*end)) {
7012 =for apidoc sv_pos_b2u
7014 Converts the value pointed to by offsetp from a count of bytes from the
7015 start of the string, to a count of the equivalent number of UTF-8 chars.
7016 Handles magic and type coercion.
7022 * sv_pos_b2u() uses, like sv_pos_u2b(), the mg_ptr of the potential
7023 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
7028 Perl_sv_pos_b2u(pTHX_ register SV *const sv, I32 *const offsetp)
7031 const STRLEN byte = *offsetp;
7032 STRLEN len = 0; /* Actually always set, but let's keep gcc happy. */
7038 PERL_ARGS_ASSERT_SV_POS_B2U;
7043 s = (const U8*)SvPV_const(sv, blen);
7046 Perl_croak(aTHX_ "panic: sv_pos_b2u: bad byte offset");
7052 && SvTYPE(sv) >= SVt_PVMG
7053 && (mg = mg_find(sv, PERL_MAGIC_utf8)))
7056 STRLEN * const cache = (STRLEN *) mg->mg_ptr;
7057 if (cache[1] == byte) {
7058 /* An exact match. */
7059 *offsetp = cache[0];
7062 if (cache[3] == byte) {
7063 /* An exact match. */
7064 *offsetp = cache[2];
7068 if (cache[1] < byte) {
7069 /* We already know part of the way. */
7070 if (mg->mg_len != -1) {
7071 /* Actually, we know the end too. */
7073 + S_sv_pos_b2u_midway(aTHX_ s + cache[1], send,
7074 s + blen, mg->mg_len - cache[0]);
7076 len = cache[0] + utf8_length(s + cache[1], send);
7079 else if (cache[3] < byte) {
7080 /* We're between the two cached pairs, so we do the calculation
7081 offset by the byte/utf-8 positions for the earlier pair,
7082 then add the utf-8 characters from the string start to
7084 len = S_sv_pos_b2u_midway(aTHX_ s + cache[3], send,
7085 s + cache[1], cache[0] - cache[2])
7089 else { /* cache[3] > byte */
7090 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + cache[3],
7094 ASSERT_UTF8_CACHE(cache);
7096 } else if (mg->mg_len != -1) {
7097 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + blen, mg->mg_len);
7101 if (!found || PL_utf8cache < 0) {
7102 const STRLEN real_len = utf8_length(s, send);
7104 if (found && PL_utf8cache < 0)
7105 assert_uft8_cache_coherent("sv_pos_b2u", len, real_len, sv);
7112 utf8_mg_len_cache_update(sv, &mg, len);
7114 utf8_mg_pos_cache_update(sv, &mg, byte, len, blen);
7119 S_assert_uft8_cache_coherent(pTHX_ const char *const func, STRLEN from_cache,
7120 STRLEN real, SV *const sv)
7122 PERL_ARGS_ASSERT_ASSERT_UFT8_CACHE_COHERENT;
7124 /* As this is debugging only code, save space by keeping this test here,
7125 rather than inlining it in all the callers. */
7126 if (from_cache == real)
7129 /* Need to turn the assertions off otherwise we may recurse infinitely
7130 while printing error messages. */
7131 SAVEI8(PL_utf8cache);
7133 Perl_croak(aTHX_ "panic: %s cache %"UVuf" real %"UVuf" for %"SVf,
7134 func, (UV) from_cache, (UV) real, SVfARG(sv));
7140 Returns a boolean indicating whether the strings in the two SVs are
7141 identical. Is UTF-8 and 'use bytes' aware, handles get magic, and will
7142 coerce its args to strings if necessary.
7144 =for apidoc sv_eq_flags
7146 Returns a boolean indicating whether the strings in the two SVs are
7147 identical. Is UTF-8 and 'use bytes' aware and coerces its args to strings
7148 if necessary. If the flags include SV_GMAGIC, it handles get-magic, too.
7154 Perl_sv_eq_flags(pTHX_ register SV *sv1, register SV *sv2, const U32 flags)
7163 SV* svrecode = NULL;
7170 /* if pv1 and pv2 are the same, second SvPV_const call may
7171 * invalidate pv1 (if we are handling magic), so we may need to
7173 if (sv1 == sv2 && flags & SV_GMAGIC
7174 && (SvTHINKFIRST(sv1) || SvGMAGICAL(sv1))) {
7175 pv1 = SvPV_const(sv1, cur1);
7176 sv1 = newSVpvn_flags(pv1, cur1, SVs_TEMP | SvUTF8(sv2));
7178 pv1 = SvPV_flags_const(sv1, cur1, flags);
7186 pv2 = SvPV_flags_const(sv2, cur2, flags);
7188 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
7189 /* Differing utf8ness.
7190 * Do not UTF8size the comparands as a side-effect. */
7193 svrecode = newSVpvn(pv2, cur2);
7194 sv_recode_to_utf8(svrecode, PL_encoding);
7195 pv2 = SvPV_const(svrecode, cur2);
7198 svrecode = newSVpvn(pv1, cur1);
7199 sv_recode_to_utf8(svrecode, PL_encoding);
7200 pv1 = SvPV_const(svrecode, cur1);
7202 /* Now both are in UTF-8. */
7204 SvREFCNT_dec(svrecode);
7210 /* sv1 is the UTF-8 one */
7211 return bytes_cmp_utf8((const U8*)pv2, cur2,
7212 (const U8*)pv1, cur1) == 0;
7215 /* sv2 is the UTF-8 one */
7216 return bytes_cmp_utf8((const U8*)pv1, cur1,
7217 (const U8*)pv2, cur2) == 0;
7223 eq = (pv1 == pv2) || memEQ(pv1, pv2, cur1);
7225 SvREFCNT_dec(svrecode);
7235 Compares the strings in two SVs. Returns -1, 0, or 1 indicating whether the
7236 string in C<sv1> is less than, equal to, or greater than the string in
7237 C<sv2>. Is UTF-8 and 'use bytes' aware, handles get magic, and will
7238 coerce its args to strings if necessary. See also C<sv_cmp_locale>.
7240 =for apidoc sv_cmp_flags
7242 Compares the strings in two SVs. Returns -1, 0, or 1 indicating whether the
7243 string in C<sv1> is less than, equal to, or greater than the string in
7244 C<sv2>. Is UTF-8 and 'use bytes' aware and will coerce its args to strings
7245 if necessary. If the flags include SV_GMAGIC, it handles get magic. See
7246 also C<sv_cmp_locale_flags>.
7252 Perl_sv_cmp(pTHX_ register SV *const sv1, register SV *const sv2)
7254 return sv_cmp_flags(sv1, sv2, SV_GMAGIC);
7258 Perl_sv_cmp_flags(pTHX_ register SV *const sv1, register SV *const sv2,
7263 const char *pv1, *pv2;
7266 SV *svrecode = NULL;
7273 pv1 = SvPV_flags_const(sv1, cur1, flags);
7280 pv2 = SvPV_flags_const(sv2, cur2, flags);
7282 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
7283 /* Differing utf8ness.
7284 * Do not UTF8size the comparands as a side-effect. */
7287 svrecode = newSVpvn(pv2, cur2);
7288 sv_recode_to_utf8(svrecode, PL_encoding);
7289 pv2 = SvPV_const(svrecode, cur2);
7292 const int retval = -bytes_cmp_utf8((const U8*)pv2, cur2,
7293 (const U8*)pv1, cur1);
7294 return retval ? retval < 0 ? -1 : +1 : 0;
7299 svrecode = newSVpvn(pv1, cur1);
7300 sv_recode_to_utf8(svrecode, PL_encoding);
7301 pv1 = SvPV_const(svrecode, cur1);
7304 const int retval = bytes_cmp_utf8((const U8*)pv1, cur1,
7305 (const U8*)pv2, cur2);
7306 return retval ? retval < 0 ? -1 : +1 : 0;
7312 cmp = cur2 ? -1 : 0;
7316 const I32 retval = memcmp((const void*)pv1, (const void*)pv2, cur1 < cur2 ? cur1 : cur2);
7319 cmp = retval < 0 ? -1 : 1;
7320 } else if (cur1 == cur2) {
7323 cmp = cur1 < cur2 ? -1 : 1;
7327 SvREFCNT_dec(svrecode);
7335 =for apidoc sv_cmp_locale
7337 Compares the strings in two SVs in a locale-aware manner. Is UTF-8 and
7338 'use bytes' aware, handles get magic, and will coerce its args to strings
7339 if necessary. See also C<sv_cmp>.
7341 =for apidoc sv_cmp_locale_flags
7343 Compares the strings in two SVs in a locale-aware manner. Is UTF-8 and
7344 'use bytes' aware and will coerce its args to strings if necessary. If the
7345 flags contain SV_GMAGIC, it handles get magic. See also C<sv_cmp_flags>.
7351 Perl_sv_cmp_locale(pTHX_ register SV *const sv1, register SV *const sv2)
7353 return sv_cmp_locale_flags(sv1, sv2, SV_GMAGIC);
7357 Perl_sv_cmp_locale_flags(pTHX_ register SV *const sv1, register SV *const sv2,
7361 #ifdef USE_LOCALE_COLLATE
7367 if (PL_collation_standard)
7371 pv1 = sv1 ? sv_collxfrm_flags(sv1, &len1, flags) : (char *) NULL;
7373 pv2 = sv2 ? sv_collxfrm_flags(sv2, &len2, flags) : (char *) NULL;
7375 if (!pv1 || !len1) {
7386 retval = memcmp((void*)pv1, (void*)pv2, len1 < len2 ? len1 : len2);
7389 return retval < 0 ? -1 : 1;
7392 * When the result of collation is equality, that doesn't mean
7393 * that there are no differences -- some locales exclude some
7394 * characters from consideration. So to avoid false equalities,
7395 * we use the raw string as a tiebreaker.
7401 #endif /* USE_LOCALE_COLLATE */
7403 return sv_cmp(sv1, sv2);
7407 #ifdef USE_LOCALE_COLLATE
7410 =for apidoc sv_collxfrm
7412 This calls C<sv_collxfrm_flags> with the SV_GMAGIC flag. See
7413 C<sv_collxfrm_flags>.
7415 =for apidoc sv_collxfrm_flags
7417 Add Collate Transform magic to an SV if it doesn't already have it. If the
7418 flags contain SV_GMAGIC, it handles get-magic.
7420 Any scalar variable may carry PERL_MAGIC_collxfrm magic that contains the
7421 scalar data of the variable, but transformed to such a format that a normal
7422 memory comparison can be used to compare the data according to the locale
7429 Perl_sv_collxfrm_flags(pTHX_ SV *const sv, STRLEN *const nxp, const I32 flags)
7434 PERL_ARGS_ASSERT_SV_COLLXFRM_FLAGS;
7436 mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_collxfrm) : (MAGIC *) NULL;
7437 if (!mg || !mg->mg_ptr || *(U32*)mg->mg_ptr != PL_collation_ix) {
7443 Safefree(mg->mg_ptr);
7444 s = SvPV_flags_const(sv, len, flags);
7445 if ((xf = mem_collxfrm(s, len, &xlen))) {
7447 #ifdef PERL_OLD_COPY_ON_WRITE
7449 sv_force_normal_flags(sv, 0);
7451 mg = sv_magicext(sv, 0, PERL_MAGIC_collxfrm, &PL_vtbl_collxfrm,
7465 if (mg && mg->mg_ptr) {
7467 return mg->mg_ptr + sizeof(PL_collation_ix);
7475 #endif /* USE_LOCALE_COLLATE */
7478 S_sv_gets_append_to_utf8(pTHX_ SV *const sv, PerlIO *const fp, I32 append)
7480 SV * const tsv = newSV(0);
7483 sv_gets(tsv, fp, 0);
7484 sv_utf8_upgrade_nomg(tsv);
7485 SvCUR_set(sv,append);
7488 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7492 S_sv_gets_read_record(pTHX_ SV *const sv, PerlIO *const fp, I32 append)
7495 const U32 recsize = SvUV(SvRV(PL_rs)); /* RsRECORD() guarantees > 0. */
7496 /* Grab the size of the record we're getting */
7497 char *const buffer = SvGROW(sv, (STRLEN)(recsize + append + 1)) + append;
7504 /* VMS wants read instead of fread, because fread doesn't respect */
7505 /* RMS record boundaries. This is not necessarily a good thing to be */
7506 /* doing, but we've got no other real choice - except avoid stdio
7507 as implementation - perhaps write a :vms layer ?
7509 fd = PerlIO_fileno(fp);
7511 bytesread = PerlLIO_read(fd, buffer, recsize);
7513 else /* in-memory file from PerlIO::Scalar */
7516 bytesread = PerlIO_read(fp, buffer, recsize);
7521 SvCUR_set(sv, bytesread + append);
7522 buffer[bytesread] = '\0';
7523 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7529 Get a line from the filehandle and store it into the SV, optionally
7530 appending to the currently-stored string.
7536 Perl_sv_gets(pTHX_ register SV *const sv, register PerlIO *const fp, I32 append)
7541 register STDCHAR rslast;
7542 register STDCHAR *bp;
7547 PERL_ARGS_ASSERT_SV_GETS;
7549 if (SvTHINKFIRST(sv))
7550 sv_force_normal_flags(sv, append ? 0 : SV_COW_DROP_PV);
7551 /* XXX. If you make this PVIV, then copy on write can copy scalars read
7553 However, perlbench says it's slower, because the existing swipe code
7554 is faster than copy on write.
7555 Swings and roundabouts. */
7556 SvUPGRADE(sv, SVt_PV);
7561 if (PerlIO_isutf8(fp)) {
7563 sv_utf8_upgrade_nomg(sv);
7564 sv_pos_u2b(sv,&append,0);
7566 } else if (SvUTF8(sv)) {
7567 return S_sv_gets_append_to_utf8(aTHX_ sv, fp, append);
7575 if (PerlIO_isutf8(fp))
7578 if (IN_PERL_COMPILETIME) {
7579 /* we always read code in line mode */
7583 else if (RsSNARF(PL_rs)) {
7584 /* If it is a regular disk file use size from stat() as estimate
7585 of amount we are going to read -- may result in mallocing
7586 more memory than we really need if the layers below reduce
7587 the size we read (e.g. CRLF or a gzip layer).
7590 if (!PerlLIO_fstat(PerlIO_fileno(fp), &st) && S_ISREG(st.st_mode)) {
7591 const Off_t offset = PerlIO_tell(fp);
7592 if (offset != (Off_t) -1 && st.st_size + append > offset) {
7593 (void) SvGROW(sv, (STRLEN)((st.st_size - offset) + append + 1));
7599 else if (RsRECORD(PL_rs)) {
7600 return S_sv_gets_read_record(aTHX_ sv, fp, append);
7602 else if (RsPARA(PL_rs)) {
7608 /* Get $/ i.e. PL_rs into same encoding as stream wants */
7609 if (PerlIO_isutf8(fp)) {
7610 rsptr = SvPVutf8(PL_rs, rslen);
7613 if (SvUTF8(PL_rs)) {
7614 if (!sv_utf8_downgrade(PL_rs, TRUE)) {
7615 Perl_croak(aTHX_ "Wide character in $/");
7618 rsptr = SvPV_const(PL_rs, rslen);
7622 rslast = rslen ? rsptr[rslen - 1] : '\0';
7624 if (rspara) { /* have to do this both before and after */
7625 do { /* to make sure file boundaries work right */
7628 i = PerlIO_getc(fp);
7632 PerlIO_ungetc(fp,i);
7638 /* See if we know enough about I/O mechanism to cheat it ! */
7640 /* This used to be #ifdef test - it is made run-time test for ease
7641 of abstracting out stdio interface. One call should be cheap
7642 enough here - and may even be a macro allowing compile
7646 if (PerlIO_fast_gets(fp)) {
7649 * We're going to steal some values from the stdio struct
7650 * and put EVERYTHING in the innermost loop into registers.
7652 register STDCHAR *ptr;
7656 #if defined(VMS) && defined(PERLIO_IS_STDIO)
7657 /* An ungetc()d char is handled separately from the regular
7658 * buffer, so we getc() it back out and stuff it in the buffer.
7660 i = PerlIO_getc(fp);
7661 if (i == EOF) return 0;
7662 *(--((*fp)->_ptr)) = (unsigned char) i;
7666 /* Here is some breathtakingly efficient cheating */
7668 cnt = PerlIO_get_cnt(fp); /* get count into register */
7669 /* make sure we have the room */
7670 if ((I32)(SvLEN(sv) - append) <= cnt + 1) {
7671 /* Not room for all of it
7672 if we are looking for a separator and room for some
7674 if (rslen && cnt > 80 && (I32)SvLEN(sv) > append) {
7675 /* just process what we have room for */
7676 shortbuffered = cnt - SvLEN(sv) + append + 1;
7677 cnt -= shortbuffered;
7681 /* remember that cnt can be negative */
7682 SvGROW(sv, (STRLEN)(append + (cnt <= 0 ? 2 : (cnt + 1))));
7687 bp = (STDCHAR*)SvPVX_const(sv) + append; /* move these two too to registers */
7688 ptr = (STDCHAR*)PerlIO_get_ptr(fp);
7689 DEBUG_P(PerlIO_printf(Perl_debug_log,
7690 "Screamer: entering, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7691 DEBUG_P(PerlIO_printf(Perl_debug_log,
7692 "Screamer: entering: PerlIO * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7693 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7694 PTR2UV(PerlIO_has_base(fp) ? PerlIO_get_base(fp) : 0)));
7699 while (cnt > 0) { /* this | eat */
7701 if ((*bp++ = *ptr++) == rslast) /* really | dust */
7702 goto thats_all_folks; /* screams | sed :-) */
7706 Copy(ptr, bp, cnt, char); /* this | eat */
7707 bp += cnt; /* screams | dust */
7708 ptr += cnt; /* louder | sed :-) */
7710 assert (!shortbuffered);
7711 goto cannot_be_shortbuffered;
7715 if (shortbuffered) { /* oh well, must extend */
7716 cnt = shortbuffered;
7718 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
7720 SvGROW(sv, SvLEN(sv) + append + cnt + 2);
7721 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
7725 cannot_be_shortbuffered:
7726 DEBUG_P(PerlIO_printf(Perl_debug_log,
7727 "Screamer: going to getc, ptr=%"UVuf", cnt=%ld\n",
7728 PTR2UV(ptr),(long)cnt));
7729 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* deregisterize cnt and ptr */
7731 DEBUG_Pv(PerlIO_printf(Perl_debug_log,
7732 "Screamer: pre: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7733 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7734 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7736 /* This used to call 'filbuf' in stdio form, but as that behaves like
7737 getc when cnt <= 0 we use PerlIO_getc here to avoid introducing
7738 another abstraction. */
7739 i = PerlIO_getc(fp); /* get more characters */
7741 DEBUG_Pv(PerlIO_printf(Perl_debug_log,
7742 "Screamer: post: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7743 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7744 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7746 cnt = PerlIO_get_cnt(fp);
7747 ptr = (STDCHAR*)PerlIO_get_ptr(fp); /* reregisterize cnt and ptr */
7748 DEBUG_P(PerlIO_printf(Perl_debug_log,
7749 "Screamer: after getc, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7751 if (i == EOF) /* all done for ever? */
7752 goto thats_really_all_folks;
7754 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
7756 SvGROW(sv, bpx + cnt + 2);
7757 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
7759 *bp++ = (STDCHAR)i; /* store character from PerlIO_getc */
7761 if (rslen && (STDCHAR)i == rslast) /* all done for now? */
7762 goto thats_all_folks;
7766 if ((rslen > 1 && (STRLEN)(bp - (STDCHAR*)SvPVX_const(sv)) < rslen) ||
7767 memNE((char*)bp - rslen, rsptr, rslen))
7768 goto screamer; /* go back to the fray */
7769 thats_really_all_folks:
7771 cnt += shortbuffered;
7772 DEBUG_P(PerlIO_printf(Perl_debug_log,
7773 "Screamer: quitting, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7774 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* put these back or we're in trouble */
7775 DEBUG_P(PerlIO_printf(Perl_debug_log,
7776 "Screamer: end: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7777 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7778 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7780 SvCUR_set(sv, bp - (STDCHAR*)SvPVX_const(sv)); /* set length */
7781 DEBUG_P(PerlIO_printf(Perl_debug_log,
7782 "Screamer: done, len=%ld, string=|%.*s|\n",
7783 (long)SvCUR(sv),(int)SvCUR(sv),SvPVX_const(sv)));
7787 /*The big, slow, and stupid way. */
7788 #ifdef USE_HEAP_INSTEAD_OF_STACK /* Even slower way. */
7789 STDCHAR *buf = NULL;
7790 Newx(buf, 8192, STDCHAR);
7798 register const STDCHAR * const bpe = buf + sizeof(buf);
7800 while ((i = PerlIO_getc(fp)) != EOF && (*bp++ = (STDCHAR)i) != rslast && bp < bpe)
7801 ; /* keep reading */
7805 cnt = PerlIO_read(fp,(char*)buf, sizeof(buf));
7806 /* Accommodate broken VAXC compiler, which applies U8 cast to
7807 * both args of ?: operator, causing EOF to change into 255
7810 i = (U8)buf[cnt - 1];
7816 cnt = 0; /* we do need to re-set the sv even when cnt <= 0 */
7818 sv_catpvn(sv, (char *) buf, cnt);
7820 sv_setpvn(sv, (char *) buf, cnt);
7822 if (i != EOF && /* joy */
7824 SvCUR(sv) < rslen ||
7825 memNE(SvPVX_const(sv) + SvCUR(sv) - rslen, rsptr, rslen)))
7829 * If we're reading from a TTY and we get a short read,
7830 * indicating that the user hit his EOF character, we need
7831 * to notice it now, because if we try to read from the TTY
7832 * again, the EOF condition will disappear.
7834 * The comparison of cnt to sizeof(buf) is an optimization
7835 * that prevents unnecessary calls to feof().
7839 if (!(cnt < (I32)sizeof(buf) && PerlIO_eof(fp)))
7843 #ifdef USE_HEAP_INSTEAD_OF_STACK
7848 if (rspara) { /* have to do this both before and after */
7849 while (i != EOF) { /* to make sure file boundaries work right */
7850 i = PerlIO_getc(fp);
7852 PerlIO_ungetc(fp,i);
7858 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7864 Auto-increment of the value in the SV, doing string to numeric conversion
7865 if necessary. Handles 'get' magic and operator overloading.
7871 Perl_sv_inc(pTHX_ register SV *const sv)
7880 =for apidoc sv_inc_nomg
7882 Auto-increment of the value in the SV, doing string to numeric conversion
7883 if necessary. Handles operator overloading. Skips handling 'get' magic.
7889 Perl_sv_inc_nomg(pTHX_ register SV *const sv)
7897 if (SvTHINKFIRST(sv)) {
7898 if (SvIsCOW(sv) || isGV_with_GP(sv))
7899 sv_force_normal_flags(sv, 0);
7900 if (SvREADONLY(sv)) {
7901 if (IN_PERL_RUNTIME)
7902 Perl_croak_no_modify(aTHX);
7906 if (SvAMAGIC(sv) && AMG_CALLunary(sv, inc_amg))
7908 i = PTR2IV(SvRV(sv));
7913 flags = SvFLAGS(sv);
7914 if ((flags & (SVp_NOK|SVp_IOK)) == SVp_NOK) {
7915 /* It's (privately or publicly) a float, but not tested as an
7916 integer, so test it to see. */
7918 flags = SvFLAGS(sv);
7920 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
7921 /* It's publicly an integer, or privately an integer-not-float */
7922 #ifdef PERL_PRESERVE_IVUV
7926 if (SvUVX(sv) == UV_MAX)
7927 sv_setnv(sv, UV_MAX_P1);
7929 (void)SvIOK_only_UV(sv);
7930 SvUV_set(sv, SvUVX(sv) + 1);
7932 if (SvIVX(sv) == IV_MAX)
7933 sv_setuv(sv, (UV)IV_MAX + 1);
7935 (void)SvIOK_only(sv);
7936 SvIV_set(sv, SvIVX(sv) + 1);
7941 if (flags & SVp_NOK) {
7942 const NV was = SvNVX(sv);
7943 if (NV_OVERFLOWS_INTEGERS_AT &&
7944 was >= NV_OVERFLOWS_INTEGERS_AT) {
7945 Perl_ck_warner(aTHX_ packWARN(WARN_IMPRECISION),
7946 "Lost precision when incrementing %" NVff " by 1",
7949 (void)SvNOK_only(sv);
7950 SvNV_set(sv, was + 1.0);
7954 if (!(flags & SVp_POK) || !*SvPVX_const(sv)) {
7955 if ((flags & SVTYPEMASK) < SVt_PVIV)
7956 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV ? SVt_PVIV : SVt_IV));
7957 (void)SvIOK_only(sv);
7962 while (isALPHA(*d)) d++;
7963 while (isDIGIT(*d)) d++;
7964 if (d < SvEND(sv)) {
7965 #ifdef PERL_PRESERVE_IVUV
7966 /* Got to punt this as an integer if needs be, but we don't issue
7967 warnings. Probably ought to make the sv_iv_please() that does
7968 the conversion if possible, and silently. */
7969 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
7970 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
7971 /* Need to try really hard to see if it's an integer.
7972 9.22337203685478e+18 is an integer.
7973 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
7974 so $a="9.22337203685478e+18"; $a+0; $a++
7975 needs to be the same as $a="9.22337203685478e+18"; $a++
7982 /* sv_2iv *should* have made this an NV */
7983 if (flags & SVp_NOK) {
7984 (void)SvNOK_only(sv);
7985 SvNV_set(sv, SvNVX(sv) + 1.0);
7988 /* I don't think we can get here. Maybe I should assert this
7989 And if we do get here I suspect that sv_setnv will croak. NWC
7991 #if defined(USE_LONG_DOUBLE)
7992 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",
7993 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7995 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_inc punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
7996 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7999 #endif /* PERL_PRESERVE_IVUV */
8000 sv_setnv(sv,Atof(SvPVX_const(sv)) + 1.0);
8004 while (d >= SvPVX_const(sv)) {
8012 /* MKS: The original code here died if letters weren't consecutive.
8013 * at least it didn't have to worry about non-C locales. The
8014 * new code assumes that ('z'-'a')==('Z'-'A'), letters are
8015 * arranged in order (although not consecutively) and that only
8016 * [A-Za-z] are accepted by isALPHA in the C locale.
8018 if (*d != 'z' && *d != 'Z') {
8019 do { ++*d; } while (!isALPHA(*d));
8022 *(d--) -= 'z' - 'a';
8027 *(d--) -= 'z' - 'a' + 1;
8031 /* oh,oh, the number grew */
8032 SvGROW(sv, SvCUR(sv) + 2);
8033 SvCUR_set(sv, SvCUR(sv) + 1);
8034 for (d = SvPVX(sv) + SvCUR(sv); d > SvPVX_const(sv); d--)
8045 Auto-decrement of the value in the SV, doing string to numeric conversion
8046 if necessary. Handles 'get' magic and operator overloading.
8052 Perl_sv_dec(pTHX_ register SV *const sv)
8062 =for apidoc sv_dec_nomg
8064 Auto-decrement of the value in the SV, doing string to numeric conversion
8065 if necessary. Handles operator overloading. Skips handling 'get' magic.
8071 Perl_sv_dec_nomg(pTHX_ register SV *const sv)
8078 if (SvTHINKFIRST(sv)) {
8079 if (SvIsCOW(sv) || isGV_with_GP(sv))
8080 sv_force_normal_flags(sv, 0);
8081 if (SvREADONLY(sv)) {
8082 if (IN_PERL_RUNTIME)
8083 Perl_croak_no_modify(aTHX);
8087 if (SvAMAGIC(sv) && AMG_CALLunary(sv, dec_amg))
8089 i = PTR2IV(SvRV(sv));
8094 /* Unlike sv_inc we don't have to worry about string-never-numbers
8095 and keeping them magic. But we mustn't warn on punting */
8096 flags = SvFLAGS(sv);
8097 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
8098 /* It's publicly an integer, or privately an integer-not-float */
8099 #ifdef PERL_PRESERVE_IVUV
8103 if (SvUVX(sv) == 0) {
8104 (void)SvIOK_only(sv);
8108 (void)SvIOK_only_UV(sv);
8109 SvUV_set(sv, SvUVX(sv) - 1);
8112 if (SvIVX(sv) == IV_MIN) {
8113 sv_setnv(sv, (NV)IV_MIN);
8117 (void)SvIOK_only(sv);
8118 SvIV_set(sv, SvIVX(sv) - 1);
8123 if (flags & SVp_NOK) {
8126 const NV was = SvNVX(sv);
8127 if (NV_OVERFLOWS_INTEGERS_AT &&
8128 was <= -NV_OVERFLOWS_INTEGERS_AT) {
8129 Perl_ck_warner(aTHX_ packWARN(WARN_IMPRECISION),
8130 "Lost precision when decrementing %" NVff " by 1",
8133 (void)SvNOK_only(sv);
8134 SvNV_set(sv, was - 1.0);
8138 if (!(flags & SVp_POK)) {
8139 if ((flags & SVTYPEMASK) < SVt_PVIV)
8140 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV) ? SVt_PVIV : SVt_IV);
8142 (void)SvIOK_only(sv);
8145 #ifdef PERL_PRESERVE_IVUV
8147 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
8148 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
8149 /* Need to try really hard to see if it's an integer.
8150 9.22337203685478e+18 is an integer.
8151 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
8152 so $a="9.22337203685478e+18"; $a+0; $a--
8153 needs to be the same as $a="9.22337203685478e+18"; $a--
8160 /* sv_2iv *should* have made this an NV */
8161 if (flags & SVp_NOK) {
8162 (void)SvNOK_only(sv);
8163 SvNV_set(sv, SvNVX(sv) - 1.0);
8166 /* I don't think we can get here. Maybe I should assert this
8167 And if we do get here I suspect that sv_setnv will croak. NWC
8169 #if defined(USE_LONG_DOUBLE)
8170 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",
8171 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
8173 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_dec punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
8174 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
8178 #endif /* PERL_PRESERVE_IVUV */
8179 sv_setnv(sv,Atof(SvPVX_const(sv)) - 1.0); /* punt */
8182 /* this define is used to eliminate a chunk of duplicated but shared logic
8183 * it has the suffix __SV_C to signal that it isnt API, and isnt meant to be
8184 * used anywhere but here - yves
8186 #define PUSH_EXTEND_MORTAL__SV_C(AnSv) \
8189 PL_tmps_stack[++PL_tmps_ix] = (AnSv); \
8193 =for apidoc sv_mortalcopy
8195 Creates a new SV which is a copy of the original SV (using C<sv_setsv>).
8196 The new SV is marked as mortal. It will be destroyed "soon", either by an
8197 explicit call to FREETMPS, or by an implicit call at places such as
8198 statement boundaries. See also C<sv_newmortal> and C<sv_2mortal>.
8203 /* Make a string that will exist for the duration of the expression
8204 * evaluation. Actually, it may have to last longer than that, but
8205 * hopefully we won't free it until it has been assigned to a
8206 * permanent location. */
8209 Perl_sv_mortalcopy(pTHX_ SV *const oldstr)
8215 sv_setsv(sv,oldstr);
8216 PUSH_EXTEND_MORTAL__SV_C(sv);
8222 =for apidoc sv_newmortal
8224 Creates a new null SV which is mortal. The reference count of the SV is
8225 set to 1. It will be destroyed "soon", either by an explicit call to
8226 FREETMPS, or by an implicit call at places such as statement boundaries.
8227 See also C<sv_mortalcopy> and C<sv_2mortal>.
8233 Perl_sv_newmortal(pTHX)
8239 SvFLAGS(sv) = SVs_TEMP;
8240 PUSH_EXTEND_MORTAL__SV_C(sv);
8246 =for apidoc newSVpvn_flags
8248 Creates a new SV and copies a string into it. The reference count for the
8249 SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
8250 string. You are responsible for ensuring that the source string is at least
8251 C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
8252 Currently the only flag bits accepted are C<SVf_UTF8> and C<SVs_TEMP>.
8253 If C<SVs_TEMP> is set, then C<sv_2mortal()> is called on the result before
8254 returning. If C<SVf_UTF8> is set, C<s> is considered to be in UTF-8 and the
8255 C<SVf_UTF8> flag will be set on the new SV.
8256 C<newSVpvn_utf8()> is a convenience wrapper for this function, defined as
8258 #define newSVpvn_utf8(s, len, u) \
8259 newSVpvn_flags((s), (len), (u) ? SVf_UTF8 : 0)
8265 Perl_newSVpvn_flags(pTHX_ const char *const s, const STRLEN len, const U32 flags)
8270 /* All the flags we don't support must be zero.
8271 And we're new code so I'm going to assert this from the start. */
8272 assert(!(flags & ~(SVf_UTF8|SVs_TEMP)));
8274 sv_setpvn(sv,s,len);
8276 /* This code used to a sv_2mortal(), however we now unroll the call to sv_2mortal()
8277 * and do what it does ourselves here.
8278 * Since we have asserted that flags can only have the SVf_UTF8 and/or SVs_TEMP flags
8279 * set above we can use it to enable the sv flags directly (bypassing SvTEMP_on), which
8280 * in turn means we dont need to mask out the SVf_UTF8 flag below, which means that we
8281 * eliminate quite a few steps than it looks - Yves (explaining patch by gfx)
8284 SvFLAGS(sv) |= flags;
8286 if(flags & SVs_TEMP){
8287 PUSH_EXTEND_MORTAL__SV_C(sv);
8294 =for apidoc sv_2mortal
8296 Marks an existing SV as mortal. The SV will be destroyed "soon", either
8297 by an explicit call to FREETMPS, or by an implicit call at places such as
8298 statement boundaries. SvTEMP() is turned on which means that the SV's
8299 string buffer can be "stolen" if this SV is copied. See also C<sv_newmortal>
8300 and C<sv_mortalcopy>.
8306 Perl_sv_2mortal(pTHX_ register SV *const sv)
8311 if (SvREADONLY(sv) && SvIMMORTAL(sv))
8313 PUSH_EXTEND_MORTAL__SV_C(sv);
8321 Creates a new SV and copies a string into it. The reference count for the
8322 SV is set to 1. If C<len> is zero, Perl will compute the length using
8323 strlen(). For efficiency, consider using C<newSVpvn> instead.
8329 Perl_newSVpv(pTHX_ const char *const s, const STRLEN len)
8335 sv_setpvn(sv, s, len || s == NULL ? len : strlen(s));
8340 =for apidoc newSVpvn
8342 Creates a new SV and copies a string into it. The reference count for the
8343 SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
8344 string. You are responsible for ensuring that the source string is at least
8345 C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
8351 Perl_newSVpvn(pTHX_ const char *const s, const STRLEN len)
8357 sv_setpvn(sv,s,len);
8362 =for apidoc newSVhek
8364 Creates a new SV from the hash key structure. It will generate scalars that
8365 point to the shared string table where possible. Returns a new (undefined)
8366 SV if the hek is NULL.
8372 Perl_newSVhek(pTHX_ const HEK *const hek)
8382 if (HEK_LEN(hek) == HEf_SVKEY) {
8383 return newSVsv(*(SV**)HEK_KEY(hek));
8385 const int flags = HEK_FLAGS(hek);
8386 if (flags & HVhek_WASUTF8) {
8388 Andreas would like keys he put in as utf8 to come back as utf8
8390 STRLEN utf8_len = HEK_LEN(hek);
8391 SV * const sv = newSV_type(SVt_PV);
8392 char *as_utf8 = (char *)bytes_to_utf8 ((U8*)HEK_KEY(hek), &utf8_len);
8393 /* bytes_to_utf8() allocates a new string, which we can repurpose: */
8394 sv_usepvn_flags(sv, as_utf8, utf8_len, SV_HAS_TRAILING_NUL);
8397 } else if (flags & (HVhek_REHASH|HVhek_UNSHARED)) {
8398 /* We don't have a pointer to the hv, so we have to replicate the
8399 flag into every HEK. This hv is using custom a hasing
8400 algorithm. Hence we can't return a shared string scalar, as
8401 that would contain the (wrong) hash value, and might get passed
8402 into an hv routine with a regular hash.
8403 Similarly, a hash that isn't using shared hash keys has to have
8404 the flag in every key so that we know not to try to call
8405 share_hek_hek on it. */
8407 SV * const sv = newSVpvn (HEK_KEY(hek), HEK_LEN(hek));
8412 /* This will be overwhelminly the most common case. */
8414 /* Inline most of newSVpvn_share(), because share_hek_hek() is far
8415 more efficient than sharepvn(). */
8419 sv_upgrade(sv, SVt_PV);
8420 SvPV_set(sv, (char *)HEK_KEY(share_hek_hek(hek)));
8421 SvCUR_set(sv, HEK_LEN(hek));
8434 =for apidoc newSVpvn_share
8436 Creates a new SV with its SvPVX_const pointing to a shared string in the string
8437 table. If the string does not already exist in the table, it is created
8438 first. Turns on READONLY and FAKE. If the C<hash> parameter is non-zero, that
8439 value is used; otherwise the hash is computed. The string's hash can be later
8440 be retrieved from the SV with the C<SvSHARED_HASH()> macro. The idea here is
8441 that as the string table is used for shared hash keys these strings will have
8442 SvPVX_const == HeKEY and hash lookup will avoid string compare.
8448 Perl_newSVpvn_share(pTHX_ const char *src, I32 len, U32 hash)
8452 bool is_utf8 = FALSE;
8453 const char *const orig_src = src;
8456 STRLEN tmplen = -len;
8458 /* See the note in hv.c:hv_fetch() --jhi */
8459 src = (char*)bytes_from_utf8((const U8*)src, &tmplen, &is_utf8);
8463 PERL_HASH(hash, src, len);
8465 /* The logic for this is inlined in S_mro_get_linear_isa_dfs(), so if it
8466 changes here, update it there too. */
8467 sv_upgrade(sv, SVt_PV);
8468 SvPV_set(sv, sharepvn(src, is_utf8?-len:len, hash));
8476 if (src != orig_src)
8482 =for apidoc newSVpv_share
8484 Like C<newSVpvn_share>, but takes a nul-terminated string instead of a
8491 Perl_newSVpv_share(pTHX_ const char *src, U32 hash)
8493 return newSVpvn_share(src, strlen(src), hash);
8496 #if defined(PERL_IMPLICIT_CONTEXT)
8498 /* pTHX_ magic can't cope with varargs, so this is a no-context
8499 * version of the main function, (which may itself be aliased to us).
8500 * Don't access this version directly.
8504 Perl_newSVpvf_nocontext(const char *const pat, ...)
8510 PERL_ARGS_ASSERT_NEWSVPVF_NOCONTEXT;
8512 va_start(args, pat);
8513 sv = vnewSVpvf(pat, &args);
8520 =for apidoc newSVpvf
8522 Creates a new SV and initializes it with the string formatted like
8529 Perl_newSVpvf(pTHX_ const char *const pat, ...)
8534 PERL_ARGS_ASSERT_NEWSVPVF;
8536 va_start(args, pat);
8537 sv = vnewSVpvf(pat, &args);
8542 /* backend for newSVpvf() and newSVpvf_nocontext() */
8545 Perl_vnewSVpvf(pTHX_ const char *const pat, va_list *const args)
8550 PERL_ARGS_ASSERT_VNEWSVPVF;
8553 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
8560 Creates a new SV and copies a floating point value into it.
8561 The reference count for the SV is set to 1.
8567 Perl_newSVnv(pTHX_ const NV n)
8580 Creates a new SV and copies an integer into it. The reference count for the
8587 Perl_newSViv(pTHX_ const IV i)
8600 Creates a new SV and copies an unsigned integer into it.
8601 The reference count for the SV is set to 1.
8607 Perl_newSVuv(pTHX_ const UV u)
8618 =for apidoc newSV_type
8620 Creates a new SV, of the type specified. The reference count for the new SV
8627 Perl_newSV_type(pTHX_ const svtype type)
8632 sv_upgrade(sv, type);
8637 =for apidoc newRV_noinc
8639 Creates an RV wrapper for an SV. The reference count for the original
8640 SV is B<not> incremented.
8646 Perl_newRV_noinc(pTHX_ SV *const tmpRef)
8649 register SV *sv = newSV_type(SVt_IV);
8651 PERL_ARGS_ASSERT_NEWRV_NOINC;
8654 SvRV_set(sv, tmpRef);
8659 /* newRV_inc is the official function name to use now.
8660 * newRV_inc is in fact #defined to newRV in sv.h
8664 Perl_newRV(pTHX_ SV *const sv)
8668 PERL_ARGS_ASSERT_NEWRV;
8670 return newRV_noinc(SvREFCNT_inc_simple_NN(sv));
8676 Creates a new SV which is an exact duplicate of the original SV.
8683 Perl_newSVsv(pTHX_ register SV *const old)
8690 if (SvTYPE(old) == (svtype)SVTYPEMASK) {
8691 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL), "semi-panic: attempt to dup freed string");
8695 /* SV_GMAGIC is the default for sv_setv()
8696 SV_NOSTEAL prevents TEMP buffers being, well, stolen, and saves games
8697 with SvTEMP_off and SvTEMP_on round a call to sv_setsv. */
8698 sv_setsv_flags(sv, old, SV_GMAGIC | SV_NOSTEAL);
8703 =for apidoc sv_reset
8705 Underlying implementation for the C<reset> Perl function.
8706 Note that the perl-level function is vaguely deprecated.
8712 Perl_sv_reset(pTHX_ register const char *s, HV *const stash)
8715 char todo[PERL_UCHAR_MAX+1];
8717 PERL_ARGS_ASSERT_SV_RESET;
8722 if (!*s) { /* reset ?? searches */
8723 MAGIC * const mg = mg_find((const SV *)stash, PERL_MAGIC_symtab);
8725 const U32 count = mg->mg_len / sizeof(PMOP**);
8726 PMOP **pmp = (PMOP**) mg->mg_ptr;
8727 PMOP *const *const end = pmp + count;
8731 SvREADONLY_off(PL_regex_pad[(*pmp)->op_pmoffset]);
8733 (*pmp)->op_pmflags &= ~PMf_USED;
8741 /* reset variables */
8743 if (!HvARRAY(stash))
8746 Zero(todo, 256, char);
8749 I32 i = (unsigned char)*s;
8753 max = (unsigned char)*s++;
8754 for ( ; i <= max; i++) {
8757 for (i = 0; i <= (I32) HvMAX(stash); i++) {
8759 for (entry = HvARRAY(stash)[i];
8761 entry = HeNEXT(entry))
8766 if (!todo[(U8)*HeKEY(entry)])
8768 gv = MUTABLE_GV(HeVAL(entry));
8771 if (SvTHINKFIRST(sv)) {
8772 if (!SvREADONLY(sv) && SvROK(sv))
8774 /* XXX Is this continue a bug? Why should THINKFIRST
8775 exempt us from resetting arrays and hashes? */
8779 if (SvTYPE(sv) >= SVt_PV) {
8781 if (SvPVX_const(sv) != NULL)
8789 if (GvHV(gv) && !HvNAME_get(GvHV(gv))) {
8791 Perl_die(aTHX_ "Can't reset %%ENV on this system");
8794 # if defined(USE_ENVIRON_ARRAY)
8797 # endif /* USE_ENVIRON_ARRAY */
8808 Using various gambits, try to get an IO from an SV: the IO slot if its a
8809 GV; or the recursive result if we're an RV; or the IO slot of the symbol
8810 named after the PV if we're a string.
8816 Perl_sv_2io(pTHX_ SV *const sv)
8821 PERL_ARGS_ASSERT_SV_2IO;
8823 switch (SvTYPE(sv)) {
8825 io = MUTABLE_IO(sv);
8829 if (isGV_with_GP(sv)) {
8830 gv = MUTABLE_GV(sv);
8833 Perl_croak(aTHX_ "Bad filehandle: %"HEKf,
8834 HEKfARG(GvNAME_HEK(gv)));
8840 Perl_croak(aTHX_ PL_no_usym, "filehandle");
8842 return sv_2io(SvRV(sv));
8843 gv = gv_fetchsv(sv, 0, SVt_PVIO);
8849 Perl_croak(aTHX_ "Bad filehandle: %"SVf, SVfARG(sv));
8858 Using various gambits, try to get a CV from an SV; in addition, try if
8859 possible to set C<*st> and C<*gvp> to the stash and GV associated with it.
8860 The flags in C<lref> are passed to gv_fetchsv.
8866 Perl_sv_2cv(pTHX_ SV *sv, HV **const st, GV **const gvp, const I32 lref)
8872 PERL_ARGS_ASSERT_SV_2CV;
8879 switch (SvTYPE(sv)) {
8883 return MUTABLE_CV(sv);
8893 sv = amagic_deref_call(sv, to_cv_amg);
8894 /* At this point I'd like to do SPAGAIN, but really I need to
8895 force it upon my callers. Hmmm. This is a mess... */
8898 if (SvTYPE(sv) == SVt_PVCV) {
8899 cv = MUTABLE_CV(sv);
8904 else if(isGV_with_GP(sv))
8905 gv = MUTABLE_GV(sv);
8907 Perl_croak(aTHX_ "Not a subroutine reference");
8909 else if (isGV_with_GP(sv)) {
8910 gv = MUTABLE_GV(sv);
8913 gv = gv_fetchsv_nomg(sv, lref, SVt_PVCV);
8920 /* Some flags to gv_fetchsv mean don't really create the GV */
8921 if (!isGV_with_GP(gv)) {
8926 if (lref & ~GV_ADDMG && !GvCVu(gv)) {
8930 gv_efullname3(tmpsv, gv, NULL);
8931 /* XXX this is probably not what they think they're getting.
8932 * It has the same effect as "sub name;", i.e. just a forward
8934 newSUB(start_subparse(FALSE, 0),
8935 newSVOP(OP_CONST, 0, tmpsv),
8939 Perl_croak(aTHX_ "Unable to create sub named \"%"SVf"\"",
8940 SVfARG(SvOK(sv) ? sv : &PL_sv_no));
8949 Returns true if the SV has a true value by Perl's rules.
8950 Use the C<SvTRUE> macro instead, which may call C<sv_true()> or may
8951 instead use an in-line version.
8957 Perl_sv_true(pTHX_ register SV *const sv)
8962 register const XPV* const tXpv = (XPV*)SvANY(sv);
8964 (tXpv->xpv_cur > 1 ||
8965 (tXpv->xpv_cur && *sv->sv_u.svu_pv != '0')))
8972 return SvIVX(sv) != 0;
8975 return SvNVX(sv) != 0.0;
8977 return sv_2bool(sv);
8983 =for apidoc sv_pvn_force
8985 Get a sensible string out of the SV somehow.
8986 A private implementation of the C<SvPV_force> macro for compilers which
8987 can't cope with complex macro expressions. Always use the macro instead.
8989 =for apidoc sv_pvn_force_flags
8991 Get a sensible string out of the SV somehow.
8992 If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<sv> if
8993 appropriate, else not. C<sv_pvn_force> and C<sv_pvn_force_nomg> are
8994 implemented in terms of this function.
8995 You normally want to use the various wrapper macros instead: see
8996 C<SvPV_force> and C<SvPV_force_nomg>
9002 Perl_sv_pvn_force_flags(pTHX_ SV *const sv, STRLEN *const lp, const I32 flags)
9006 PERL_ARGS_ASSERT_SV_PVN_FORCE_FLAGS;
9008 if (SvTHINKFIRST(sv) && !SvROK(sv))
9009 sv_force_normal_flags(sv, 0);
9019 if (SvREADONLY(sv) && !(flags & SV_MUTABLE_RETURN)) {
9020 const char * const ref = sv_reftype(sv,0);
9022 Perl_croak(aTHX_ "Can't coerce readonly %s to string in %s",
9023 ref, OP_DESC(PL_op));
9025 Perl_croak(aTHX_ "Can't coerce readonly %s to string", ref);
9027 if ((SvTYPE(sv) > SVt_PVLV && SvTYPE(sv) != SVt_PVFM)
9028 || isGV_with_GP(sv))
9029 /* diag_listed_as: Can't coerce %s to %s in %s */
9030 Perl_croak(aTHX_ "Can't coerce %s to string in %s", sv_reftype(sv,0),
9032 s = sv_2pv_flags(sv, &len, flags);
9036 if (s != SvPVX_const(sv)) { /* Almost, but not quite, sv_setpvn() */
9039 SvUPGRADE(sv, SVt_PV); /* Never FALSE */
9040 SvGROW(sv, len + 1);
9041 Move(s,SvPVX(sv),len,char);
9043 SvPVX(sv)[len] = '\0';
9046 SvPOK_on(sv); /* validate pointer */
9048 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
9049 PTR2UV(sv),SvPVX_const(sv)));
9052 return SvPVX_mutable(sv);
9056 =for apidoc sv_pvbyten_force
9058 The backend for the C<SvPVbytex_force> macro. Always use the macro instead.
9064 Perl_sv_pvbyten_force(pTHX_ SV *const sv, STRLEN *const lp)
9066 PERL_ARGS_ASSERT_SV_PVBYTEN_FORCE;
9068 sv_pvn_force(sv,lp);
9069 sv_utf8_downgrade(sv,0);
9075 =for apidoc sv_pvutf8n_force
9077 The backend for the C<SvPVutf8x_force> macro. Always use the macro instead.
9083 Perl_sv_pvutf8n_force(pTHX_ SV *const sv, STRLEN *const lp)
9085 PERL_ARGS_ASSERT_SV_PVUTF8N_FORCE;
9087 sv_pvn_force(sv,lp);
9088 sv_utf8_upgrade(sv);
9094 =for apidoc sv_reftype
9096 Returns a string describing what the SV is a reference to.
9102 Perl_sv_reftype(pTHX_ const SV *const sv, const int ob)
9104 PERL_ARGS_ASSERT_SV_REFTYPE;
9105 if (ob && SvOBJECT(sv)) {
9106 return SvPV_nolen_const(sv_ref(NULL, sv, ob));
9109 switch (SvTYPE(sv)) {
9124 case SVt_PVLV: return (char *) (SvROK(sv) ? "REF"
9125 /* tied lvalues should appear to be
9126 * scalars for backwards compatibility */
9127 : (LvTYPE(sv) == 't' || LvTYPE(sv) == 'T')
9128 ? "SCALAR" : "LVALUE");
9129 case SVt_PVAV: return "ARRAY";
9130 case SVt_PVHV: return "HASH";
9131 case SVt_PVCV: return "CODE";
9132 case SVt_PVGV: return (char *) (isGV_with_GP(sv)
9133 ? "GLOB" : "SCALAR");
9134 case SVt_PVFM: return "FORMAT";
9135 case SVt_PVIO: return "IO";
9136 case SVt_BIND: return "BIND";
9137 case SVt_REGEXP: return "REGEXP";
9138 default: return "UNKNOWN";
9146 Returns a SV describing what the SV passed in is a reference to.
9152 Perl_sv_ref(pTHX_ register SV *dst, const SV *const sv, const int ob)
9154 PERL_ARGS_ASSERT_SV_REF;
9157 dst = sv_newmortal();
9159 if (ob && SvOBJECT(sv)) {
9160 HvNAME_get(SvSTASH(sv))
9161 ? sv_sethek(dst, HvNAME_HEK(SvSTASH(sv)))
9162 : sv_setpvn(dst, "__ANON__", 8);
9165 const char * reftype = sv_reftype(sv, 0);
9166 sv_setpv(dst, reftype);
9172 =for apidoc sv_isobject
9174 Returns a boolean indicating whether the SV is an RV pointing to a blessed
9175 object. If the SV is not an RV, or if the object is not blessed, then this
9182 Perl_sv_isobject(pTHX_ SV *sv)
9198 Returns a boolean indicating whether the SV is blessed into the specified
9199 class. This does not check for subtypes; use C<sv_derived_from> to verify
9200 an inheritance relationship.
9206 Perl_sv_isa(pTHX_ SV *sv, const char *const name)
9210 PERL_ARGS_ASSERT_SV_ISA;
9220 hvname = HvNAME_get(SvSTASH(sv));
9224 return strEQ(hvname, name);
9230 Creates a new SV for the RV, C<rv>, to point to. If C<rv> is not an RV then
9231 it will be upgraded to one. If C<classname> is non-null then the new SV will
9232 be blessed in the specified package. The new SV is returned and its
9233 reference count is 1.
9239 Perl_newSVrv(pTHX_ SV *const rv, const char *const classname)
9244 PERL_ARGS_ASSERT_NEWSVRV;
9248 SV_CHECK_THINKFIRST_COW_DROP(rv);
9249 (void)SvAMAGIC_off(rv);
9251 if (SvTYPE(rv) >= SVt_PVMG) {
9252 const U32 refcnt = SvREFCNT(rv);
9256 SvREFCNT(rv) = refcnt;
9258 sv_upgrade(rv, SVt_IV);
9259 } else if (SvROK(rv)) {
9260 SvREFCNT_dec(SvRV(rv));
9262 prepare_SV_for_RV(rv);
9270 HV* const stash = gv_stashpv(classname, GV_ADD);
9271 (void)sv_bless(rv, stash);
9277 =for apidoc sv_setref_pv
9279 Copies a pointer into a new SV, optionally blessing the SV. The C<rv>
9280 argument will be upgraded to an RV. That RV will be modified to point to
9281 the new SV. If the C<pv> argument is NULL then C<PL_sv_undef> will be placed
9282 into the SV. The C<classname> argument indicates the package for the
9283 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9284 will have a reference count of 1, and the RV will be returned.
9286 Do not use with other Perl types such as HV, AV, SV, CV, because those
9287 objects will become corrupted by the pointer copy process.
9289 Note that C<sv_setref_pvn> copies the string while this copies the pointer.
9295 Perl_sv_setref_pv(pTHX_ SV *const rv, const char *const classname, void *const pv)
9299 PERL_ARGS_ASSERT_SV_SETREF_PV;
9302 sv_setsv(rv, &PL_sv_undef);
9306 sv_setiv(newSVrv(rv,classname), PTR2IV(pv));
9311 =for apidoc sv_setref_iv
9313 Copies an integer into a new SV, optionally blessing the SV. The C<rv>
9314 argument will be upgraded to an RV. That RV will be modified to point to
9315 the new SV. The C<classname> argument indicates the package for the
9316 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9317 will have a reference count of 1, and the RV will be returned.
9323 Perl_sv_setref_iv(pTHX_ SV *const rv, const char *const classname, const IV iv)
9325 PERL_ARGS_ASSERT_SV_SETREF_IV;
9327 sv_setiv(newSVrv(rv,classname), iv);
9332 =for apidoc sv_setref_uv
9334 Copies an unsigned integer into a new SV, optionally blessing the SV. The C<rv>
9335 argument will be upgraded to an RV. That RV will be modified to point to
9336 the new SV. The C<classname> argument indicates the package for the
9337 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9338 will have a reference count of 1, and the RV will be returned.
9344 Perl_sv_setref_uv(pTHX_ SV *const rv, const char *const classname, const UV uv)
9346 PERL_ARGS_ASSERT_SV_SETREF_UV;
9348 sv_setuv(newSVrv(rv,classname), uv);
9353 =for apidoc sv_setref_nv
9355 Copies a double into a new SV, optionally blessing the SV. The C<rv>
9356 argument will be upgraded to an RV. That RV will be modified to point to
9357 the new SV. The C<classname> argument indicates the package for the
9358 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9359 will have a reference count of 1, and the RV will be returned.
9365 Perl_sv_setref_nv(pTHX_ SV *const rv, const char *const classname, const NV nv)
9367 PERL_ARGS_ASSERT_SV_SETREF_NV;
9369 sv_setnv(newSVrv(rv,classname), nv);
9374 =for apidoc sv_setref_pvn
9376 Copies a string into a new SV, optionally blessing the SV. The length of the
9377 string must be specified with C<n>. The C<rv> argument will be upgraded to
9378 an RV. That RV will be modified to point to the new SV. The C<classname>
9379 argument indicates the package for the blessing. Set C<classname> to
9380 C<NULL> to avoid the blessing. The new SV will have a reference count
9381 of 1, and the RV will be returned.
9383 Note that C<sv_setref_pv> copies the pointer while this copies the string.
9389 Perl_sv_setref_pvn(pTHX_ SV *const rv, const char *const classname,
9390 const char *const pv, const STRLEN n)
9392 PERL_ARGS_ASSERT_SV_SETREF_PVN;
9394 sv_setpvn(newSVrv(rv,classname), pv, n);
9399 =for apidoc sv_bless
9401 Blesses an SV into a specified package. The SV must be an RV. The package
9402 must be designated by its stash (see C<gv_stashpv()>). The reference count
9403 of the SV is unaffected.
9409 Perl_sv_bless(pTHX_ SV *const sv, HV *const stash)
9414 PERL_ARGS_ASSERT_SV_BLESS;
9417 Perl_croak(aTHX_ "Can't bless non-reference value");
9419 if (SvFLAGS(tmpRef) & (SVs_OBJECT|SVf_READONLY)) {
9420 if (SvIsCOW(tmpRef))
9421 sv_force_normal_flags(tmpRef, 0);
9422 if (SvREADONLY(tmpRef))
9423 Perl_croak_no_modify(aTHX);
9424 if (SvOBJECT(tmpRef)) {
9425 if (SvTYPE(tmpRef) != SVt_PVIO)
9427 SvREFCNT_dec(SvSTASH(tmpRef));
9430 SvOBJECT_on(tmpRef);
9431 if (SvTYPE(tmpRef) != SVt_PVIO)
9433 SvUPGRADE(tmpRef, SVt_PVMG);
9434 SvSTASH_set(tmpRef, MUTABLE_HV(SvREFCNT_inc_simple(stash)));
9439 (void)SvAMAGIC_off(sv);
9441 if(SvSMAGICAL(tmpRef))
9442 if(mg_find(tmpRef, PERL_MAGIC_ext) || mg_find(tmpRef, PERL_MAGIC_uvar))
9450 /* Downgrades a PVGV to a PVMG. If it’s actually a PVLV, we leave the type
9451 * as it is after unglobbing it.
9455 S_sv_unglob(pTHX_ SV *const sv)
9460 SV * const temp = sv_newmortal();
9462 PERL_ARGS_ASSERT_SV_UNGLOB;
9464 assert(SvTYPE(sv) == SVt_PVGV || SvTYPE(sv) == SVt_PVLV);
9466 gv_efullname3(temp, MUTABLE_GV(sv), "*");
9469 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
9470 && HvNAME_get(stash))
9471 mro_method_changed_in(stash);
9472 gp_free(MUTABLE_GV(sv));
9475 sv_del_backref(MUTABLE_SV(GvSTASH(sv)), sv);
9479 if (GvNAME_HEK(sv)) {
9480 unshare_hek(GvNAME_HEK(sv));
9482 isGV_with_GP_off(sv);
9484 if(SvTYPE(sv) == SVt_PVGV) {
9485 /* need to keep SvANY(sv) in the right arena */
9486 xpvmg = new_XPVMG();
9487 StructCopy(SvANY(sv), xpvmg, XPVMG);
9488 del_XPVGV(SvANY(sv));
9491 SvFLAGS(sv) &= ~SVTYPEMASK;
9492 SvFLAGS(sv) |= SVt_PVMG;
9495 /* Intentionally not calling any local SET magic, as this isn't so much a
9496 set operation as merely an internal storage change. */
9497 sv_setsv_flags(sv, temp, 0);
9501 =for apidoc sv_unref_flags
9503 Unsets the RV status of the SV, and decrements the reference count of
9504 whatever was being referenced by the RV. This can almost be thought of
9505 as a reversal of C<newSVrv>. The C<cflags> argument can contain
9506 C<SV_IMMEDIATE_UNREF> to force the reference count to be decremented
9507 (otherwise the decrementing is conditional on the reference count being
9508 different from one or the reference being a readonly SV).
9515 Perl_sv_unref_flags(pTHX_ SV *const ref, const U32 flags)
9517 SV* const target = SvRV(ref);
9519 PERL_ARGS_ASSERT_SV_UNREF_FLAGS;
9521 if (SvWEAKREF(ref)) {
9522 sv_del_backref(target, ref);
9524 SvRV_set(ref, NULL);
9527 SvRV_set(ref, NULL);
9529 /* You can't have a || SvREADONLY(target) here, as $a = $$a, where $a was
9530 assigned to as BEGIN {$a = \"Foo"} will fail. */
9531 if (SvREFCNT(target) != 1 || (flags & SV_IMMEDIATE_UNREF))
9532 SvREFCNT_dec(target);
9533 else /* XXX Hack, but hard to make $a=$a->[1] work otherwise */
9534 sv_2mortal(target); /* Schedule for freeing later */
9538 =for apidoc sv_untaint
9540 Untaint an SV. Use C<SvTAINTED_off> instead.
9546 Perl_sv_untaint(pTHX_ SV *const sv)
9548 PERL_ARGS_ASSERT_SV_UNTAINT;
9550 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
9551 MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
9558 =for apidoc sv_tainted
9560 Test an SV for taintedness. Use C<SvTAINTED> instead.
9566 Perl_sv_tainted(pTHX_ SV *const sv)
9568 PERL_ARGS_ASSERT_SV_TAINTED;
9570 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
9571 const MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
9572 if (mg && (mg->mg_len & 1) )
9579 =for apidoc sv_setpviv
9581 Copies an integer into the given SV, also updating its string value.
9582 Does not handle 'set' magic. See C<sv_setpviv_mg>.
9588 Perl_sv_setpviv(pTHX_ SV *const sv, const IV iv)
9590 char buf[TYPE_CHARS(UV)];
9592 char * const ptr = uiv_2buf(buf, iv, 0, 0, &ebuf);
9594 PERL_ARGS_ASSERT_SV_SETPVIV;
9596 sv_setpvn(sv, ptr, ebuf - ptr);
9600 =for apidoc sv_setpviv_mg
9602 Like C<sv_setpviv>, but also handles 'set' magic.
9608 Perl_sv_setpviv_mg(pTHX_ SV *const sv, const IV iv)
9610 PERL_ARGS_ASSERT_SV_SETPVIV_MG;
9616 #if defined(PERL_IMPLICIT_CONTEXT)
9618 /* pTHX_ magic can't cope with varargs, so this is a no-context
9619 * version of the main function, (which may itself be aliased to us).
9620 * Don't access this version directly.
9624 Perl_sv_setpvf_nocontext(SV *const sv, const char *const pat, ...)
9629 PERL_ARGS_ASSERT_SV_SETPVF_NOCONTEXT;
9631 va_start(args, pat);
9632 sv_vsetpvf(sv, pat, &args);
9636 /* pTHX_ magic can't cope with varargs, so this is a no-context
9637 * version of the main function, (which may itself be aliased to us).
9638 * Don't access this version directly.
9642 Perl_sv_setpvf_mg_nocontext(SV *const sv, const char *const pat, ...)
9647 PERL_ARGS_ASSERT_SV_SETPVF_MG_NOCONTEXT;
9649 va_start(args, pat);
9650 sv_vsetpvf_mg(sv, pat, &args);
9656 =for apidoc sv_setpvf
9658 Works like C<sv_catpvf> but copies the text into the SV instead of
9659 appending it. Does not handle 'set' magic. See C<sv_setpvf_mg>.
9665 Perl_sv_setpvf(pTHX_ SV *const sv, const char *const pat, ...)
9669 PERL_ARGS_ASSERT_SV_SETPVF;
9671 va_start(args, pat);
9672 sv_vsetpvf(sv, pat, &args);
9677 =for apidoc sv_vsetpvf
9679 Works like C<sv_vcatpvf> but copies the text into the SV instead of
9680 appending it. Does not handle 'set' magic. See C<sv_vsetpvf_mg>.
9682 Usually used via its frontend C<sv_setpvf>.
9688 Perl_sv_vsetpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9690 PERL_ARGS_ASSERT_SV_VSETPVF;
9692 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9696 =for apidoc sv_setpvf_mg
9698 Like C<sv_setpvf>, but also handles 'set' magic.
9704 Perl_sv_setpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
9708 PERL_ARGS_ASSERT_SV_SETPVF_MG;
9710 va_start(args, pat);
9711 sv_vsetpvf_mg(sv, pat, &args);
9716 =for apidoc sv_vsetpvf_mg
9718 Like C<sv_vsetpvf>, but also handles 'set' magic.
9720 Usually used via its frontend C<sv_setpvf_mg>.
9726 Perl_sv_vsetpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9728 PERL_ARGS_ASSERT_SV_VSETPVF_MG;
9730 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9734 #if defined(PERL_IMPLICIT_CONTEXT)
9736 /* pTHX_ magic can't cope with varargs, so this is a no-context
9737 * version of the main function, (which may itself be aliased to us).
9738 * Don't access this version directly.
9742 Perl_sv_catpvf_nocontext(SV *const sv, const char *const pat, ...)
9747 PERL_ARGS_ASSERT_SV_CATPVF_NOCONTEXT;
9749 va_start(args, pat);
9750 sv_vcatpvf(sv, pat, &args);
9754 /* pTHX_ magic can't cope with varargs, so this is a no-context
9755 * version of the main function, (which may itself be aliased to us).
9756 * Don't access this version directly.
9760 Perl_sv_catpvf_mg_nocontext(SV *const sv, const char *const pat, ...)
9765 PERL_ARGS_ASSERT_SV_CATPVF_MG_NOCONTEXT;
9767 va_start(args, pat);
9768 sv_vcatpvf_mg(sv, pat, &args);
9774 =for apidoc sv_catpvf
9776 Processes its arguments like C<sprintf> and appends the formatted
9777 output to an SV. If the appended data contains "wide" characters
9778 (including, but not limited to, SVs with a UTF-8 PV formatted with %s,
9779 and characters >255 formatted with %c), the original SV might get
9780 upgraded to UTF-8. Handles 'get' magic, but not 'set' magic. See
9781 C<sv_catpvf_mg>. If the original SV was UTF-8, the pattern should be
9782 valid UTF-8; if the original SV was bytes, the pattern should be too.
9787 Perl_sv_catpvf(pTHX_ SV *const sv, const char *const pat, ...)
9791 PERL_ARGS_ASSERT_SV_CATPVF;
9793 va_start(args, pat);
9794 sv_vcatpvf(sv, pat, &args);
9799 =for apidoc sv_vcatpvf
9801 Processes its arguments like C<vsprintf> and appends the formatted output
9802 to an SV. Does not handle 'set' magic. See C<sv_vcatpvf_mg>.
9804 Usually used via its frontend C<sv_catpvf>.
9810 Perl_sv_vcatpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9812 PERL_ARGS_ASSERT_SV_VCATPVF;
9814 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9818 =for apidoc sv_catpvf_mg
9820 Like C<sv_catpvf>, but also handles 'set' magic.
9826 Perl_sv_catpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
9830 PERL_ARGS_ASSERT_SV_CATPVF_MG;
9832 va_start(args, pat);
9833 sv_vcatpvf_mg(sv, pat, &args);
9838 =for apidoc sv_vcatpvf_mg
9840 Like C<sv_vcatpvf>, but also handles 'set' magic.
9842 Usually used via its frontend C<sv_catpvf_mg>.
9848 Perl_sv_vcatpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9850 PERL_ARGS_ASSERT_SV_VCATPVF_MG;
9852 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9857 =for apidoc sv_vsetpvfn
9859 Works like C<sv_vcatpvfn> but copies the text into the SV instead of
9862 Usually used via one of its frontends C<sv_vsetpvf> and C<sv_vsetpvf_mg>.
9868 Perl_sv_vsetpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
9869 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
9871 PERL_ARGS_ASSERT_SV_VSETPVFN;
9874 sv_vcatpvfn(sv, pat, patlen, args, svargs, svmax, maybe_tainted);
9879 * Warn of missing argument to sprintf, and then return a defined value
9880 * to avoid inappropriate "use of uninit" warnings [perl #71000].
9882 #define WARN_MISSING WARN_UNINITIALIZED /* Not sure we want a new category */
9884 S_vcatpvfn_missing_argument(pTHX) {
9885 if (ckWARN(WARN_MISSING)) {
9886 Perl_warner(aTHX_ packWARN(WARN_MISSING), "Missing argument in %s",
9887 PL_op ? OP_DESC(PL_op) : "sv_vcatpvfn()");
9894 S_expect_number(pTHX_ char **const pattern)
9899 PERL_ARGS_ASSERT_EXPECT_NUMBER;
9901 switch (**pattern) {
9902 case '1': case '2': case '3':
9903 case '4': case '5': case '6':
9904 case '7': case '8': case '9':
9905 var = *(*pattern)++ - '0';
9906 while (isDIGIT(**pattern)) {
9907 const I32 tmp = var * 10 + (*(*pattern)++ - '0');
9909 Perl_croak(aTHX_ "Integer overflow in format string for %s", (PL_op ? OP_DESC(PL_op) : "sv_vcatpvfn"));
9917 S_F0convert(NV nv, char *const endbuf, STRLEN *const len)
9919 const int neg = nv < 0;
9922 PERL_ARGS_ASSERT_F0CONVERT;
9930 if (uv & 1 && uv == nv)
9931 uv--; /* Round to even */
9933 const unsigned dig = uv % 10;
9946 =for apidoc sv_vcatpvfn
9948 Processes its arguments like C<vsprintf> and appends the formatted output
9949 to an SV. Uses an array of SVs if the C style variable argument list is
9950 missing (NULL). When running with taint checks enabled, indicates via
9951 C<maybe_tainted> if results are untrustworthy (often due to the use of
9954 Usually used via one of its frontends C<sv_vcatpvf> and C<sv_vcatpvf_mg>.
9960 #define VECTORIZE_ARGS vecsv = va_arg(*args, SV*);\
9961 vecstr = (U8*)SvPV_const(vecsv,veclen);\
9962 vec_utf8 = DO_UTF8(vecsv);
9964 /* XXX maybe_tainted is never assigned to, so the doc above is lying. */
9967 Perl_sv_vcatpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
9968 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
9976 static const char nullstr[] = "(null)";
9978 bool has_utf8 = DO_UTF8(sv); /* has the result utf8? */
9979 const bool pat_utf8 = has_utf8; /* the pattern is in utf8? */
9981 /* Times 4: a decimal digit takes more than 3 binary digits.
9982 * NV_DIG: mantissa takes than many decimal digits.
9983 * Plus 32: Playing safe. */
9984 char ebuf[IV_DIG * 4 + NV_DIG + 32];
9985 /* large enough for "%#.#f" --chip */
9986 /* what about long double NVs? --jhi */
9988 PERL_ARGS_ASSERT_SV_VCATPVFN;
9989 PERL_UNUSED_ARG(maybe_tainted);
9991 /* no matter what, this is a string now */
9992 (void)SvPV_force(sv, origlen);
9994 /* special-case "", "%s", and "%-p" (SVf - see below) */
9997 if (patlen == 2 && pat[0] == '%' && pat[1] == 's') {
9999 const char * const s = va_arg(*args, char*);
10000 sv_catpv(sv, s ? s : nullstr);
10002 else if (svix < svmax) {
10003 sv_catsv(sv, *svargs);
10006 S_vcatpvfn_missing_argument(aTHX);
10009 if (args && patlen == 3 && pat[0] == '%' &&
10010 pat[1] == '-' && pat[2] == 'p') {
10011 argsv = MUTABLE_SV(va_arg(*args, void*));
10012 sv_catsv(sv, argsv);
10016 #ifndef USE_LONG_DOUBLE
10017 /* special-case "%.<number>[gf]" */
10018 if ( !args && patlen <= 5 && pat[0] == '%' && pat[1] == '.'
10019 && (pat[patlen-1] == 'g' || pat[patlen-1] == 'f') ) {
10020 unsigned digits = 0;
10024 while (*pp >= '0' && *pp <= '9')
10025 digits = 10 * digits + (*pp++ - '0');
10026 if (pp - pat == (int)patlen - 1 && svix < svmax) {
10027 const NV nv = SvNV(*svargs);
10029 /* Add check for digits != 0 because it seems that some
10030 gconverts are buggy in this case, and we don't yet have
10031 a Configure test for this. */
10032 if (digits && digits < sizeof(ebuf) - NV_DIG - 10) {
10033 /* 0, point, slack */
10034 Gconvert(nv, (int)digits, 0, ebuf);
10035 sv_catpv(sv, ebuf);
10036 if (*ebuf) /* May return an empty string for digits==0 */
10039 } else if (!digits) {
10042 if ((p = F0convert(nv, ebuf + sizeof ebuf, &l))) {
10043 sv_catpvn(sv, p, l);
10049 #endif /* !USE_LONG_DOUBLE */
10051 if (!args && svix < svmax && DO_UTF8(*svargs))
10054 patend = (char*)pat + patlen;
10055 for (p = (char*)pat; p < patend; p = q) {
10058 bool vectorize = FALSE;
10059 bool vectorarg = FALSE;
10060 bool vec_utf8 = FALSE;
10066 bool has_precis = FALSE;
10068 const I32 osvix = svix;
10069 bool is_utf8 = FALSE; /* is this item utf8? */
10070 #ifdef HAS_LDBL_SPRINTF_BUG
10071 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
10072 with sfio - Allen <allens@cpan.org> */
10073 bool fix_ldbl_sprintf_bug = FALSE;
10077 U8 utf8buf[UTF8_MAXBYTES+1];
10078 STRLEN esignlen = 0;
10080 const char *eptr = NULL;
10081 const char *fmtstart;
10084 const U8 *vecstr = NULL;
10091 /* we need a long double target in case HAS_LONG_DOUBLE but
10092 not USE_LONG_DOUBLE
10094 #if defined(HAS_LONG_DOUBLE) && LONG_DOUBLESIZE > DOUBLESIZE
10102 const char *dotstr = ".";
10103 STRLEN dotstrlen = 1;
10104 I32 efix = 0; /* explicit format parameter index */
10105 I32 ewix = 0; /* explicit width index */
10106 I32 epix = 0; /* explicit precision index */
10107 I32 evix = 0; /* explicit vector index */
10108 bool asterisk = FALSE;
10110 /* echo everything up to the next format specification */
10111 for (q = p; q < patend && *q != '%'; ++q) ;
10113 if (has_utf8 && !pat_utf8)
10114 sv_catpvn_utf8_upgrade(sv, p, q - p, nsv);
10116 sv_catpvn(sv, p, q - p);
10125 We allow format specification elements in this order:
10126 \d+\$ explicit format parameter index
10128 v|\*(\d+\$)?v vector with optional (optionally specified) arg
10129 0 flag (as above): repeated to allow "v02"
10130 \d+|\*(\d+\$)? width using optional (optionally specified) arg
10131 \.(\d*|\*(\d+\$)?) precision using optional (optionally specified) arg
10133 [%bcdefginopsuxDFOUX] format (mandatory)
10138 As of perl5.9.3, printf format checking is on by default.
10139 Internally, perl uses %p formats to provide an escape to
10140 some extended formatting. This block deals with those
10141 extensions: if it does not match, (char*)q is reset and
10142 the normal format processing code is used.
10144 Currently defined extensions are:
10145 %p include pointer address (standard)
10146 %-p (SVf) include an SV (previously %_)
10147 %-<num>p include an SV with precision <num>
10149 %3p include a HEK with precision of 256
10150 %<num>p (where num != 2 or 3) reserved for future
10153 Robin Barker 2005-07-14 (but modified since)
10155 %1p (VDf) removed. RMB 2007-10-19
10162 n = expect_number(&q);
10164 if (sv) { /* SVf */
10169 argsv = MUTABLE_SV(va_arg(*args, void*));
10170 eptr = SvPV_const(argsv, elen);
10171 if (DO_UTF8(argsv))
10175 else if (n==2 || n==3) { /* HEKf */
10176 HEK * const hek = va_arg(*args, HEK *);
10177 eptr = HEK_KEY(hek);
10178 elen = HEK_LEN(hek);
10179 if (HEK_UTF8(hek)) is_utf8 = TRUE;
10180 if (n==3) precis = 256, has_precis = TRUE;
10184 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL),
10185 "internal %%<num>p might conflict with future printf extensions");
10191 if ( (width = expect_number(&q)) ) {
10206 if (plus == '+' && *q == ' ') /* '+' over ' ' */
10235 if ( (ewix = expect_number(&q)) )
10244 if ((vectorarg = asterisk)) {
10257 width = expect_number(&q);
10260 if (vectorize && vectorarg) {
10261 /* vectorizing, but not with the default "." */
10263 vecsv = va_arg(*args, SV*);
10265 vecsv = (evix > 0 && evix <= svmax)
10266 ? svargs[evix-1] : S_vcatpvfn_missing_argument(aTHX);
10268 vecsv = svix < svmax
10269 ? svargs[svix++] : S_vcatpvfn_missing_argument(aTHX);
10271 dotstr = SvPV_const(vecsv, dotstrlen);
10272 /* Keep the DO_UTF8 test *after* the SvPV call, else things go
10273 bad with tied or overloaded values that return UTF8. */
10274 if (DO_UTF8(vecsv))
10276 else if (has_utf8) {
10277 vecsv = sv_mortalcopy(vecsv);
10278 sv_utf8_upgrade(vecsv);
10279 dotstr = SvPV_const(vecsv, dotstrlen);
10286 i = va_arg(*args, int);
10288 i = (ewix ? ewix <= svmax : svix < svmax) ?
10289 SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
10291 width = (i < 0) ? -i : i;
10301 if ( ((epix = expect_number(&q))) && (*q++ != '$') )
10303 /* XXX: todo, support specified precision parameter */
10307 i = va_arg(*args, int);
10309 i = (ewix ? ewix <= svmax : svix < svmax)
10310 ? SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
10312 has_precis = !(i < 0);
10316 while (isDIGIT(*q))
10317 precis = precis * 10 + (*q++ - '0');
10326 else if (efix ? (efix > 0 && efix <= svmax) : svix < svmax) {
10327 vecsv = svargs[efix ? efix-1 : svix++];
10328 vecstr = (U8*)SvPV_const(vecsv,veclen);
10329 vec_utf8 = DO_UTF8(vecsv);
10331 /* if this is a version object, we need to convert
10332 * back into v-string notation and then let the
10333 * vectorize happen normally
10335 if (sv_derived_from(vecsv, "version")) {
10336 char *version = savesvpv(vecsv);
10337 if ( hv_exists(MUTABLE_HV(SvRV(vecsv)), "alpha", 5 ) ) {
10338 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
10339 "vector argument not supported with alpha versions");
10342 vecsv = sv_newmortal();
10343 scan_vstring(version, version + veclen, vecsv);
10344 vecstr = (U8*)SvPV_const(vecsv, veclen);
10345 vec_utf8 = DO_UTF8(vecsv);
10359 case 'I': /* Ix, I32x, and I64x */
10361 if (q[1] == '6' && q[2] == '4') {
10367 if (q[1] == '3' && q[2] == '2') {
10377 #if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
10389 #if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
10390 if (*q == 'l') { /* lld, llf */
10399 if (*++q == 'h') { /* hhd, hhu */
10428 if (!vectorize && !args) {
10430 const I32 i = efix-1;
10431 argsv = (i >= 0 && i < svmax)
10432 ? svargs[i] : S_vcatpvfn_missing_argument(aTHX);
10434 argsv = (svix >= 0 && svix < svmax)
10435 ? svargs[svix++] : S_vcatpvfn_missing_argument(aTHX);
10439 switch (c = *q++) {
10446 uv = (args) ? va_arg(*args, int) : SvIV(argsv);
10448 (!UNI_IS_INVARIANT(uv) && SvUTF8(sv)))
10450 eptr = (char*)utf8buf;
10451 elen = uvchr_to_utf8((U8*)eptr, uv) - utf8buf;
10465 eptr = va_arg(*args, char*);
10467 elen = strlen(eptr);
10469 eptr = (char *)nullstr;
10470 elen = sizeof nullstr - 1;
10474 eptr = SvPV_const(argsv, elen);
10475 if (DO_UTF8(argsv)) {
10476 STRLEN old_precis = precis;
10477 if (has_precis && precis < elen) {
10478 STRLEN ulen = sv_len_utf8(argsv);
10479 I32 p = precis > ulen ? ulen : precis;
10480 sv_pos_u2b(argsv, &p, 0); /* sticks at end */
10483 if (width) { /* fudge width (can't fudge elen) */
10484 if (has_precis && precis < elen)
10485 width += precis - old_precis;
10487 width += elen - sv_len_utf8(argsv);
10494 if (has_precis && precis < elen)
10501 if (alt || vectorize)
10503 uv = PTR2UV(args ? va_arg(*args, void*) : argsv);
10524 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
10533 esignbuf[esignlen++] = plus;
10537 case 'c': iv = (char)va_arg(*args, int); break;
10538 case 'h': iv = (short)va_arg(*args, int); break;
10539 case 'l': iv = va_arg(*args, long); break;
10540 case 'V': iv = va_arg(*args, IV); break;
10541 case 'z': iv = va_arg(*args, SSize_t); break;
10542 case 't': iv = va_arg(*args, ptrdiff_t); break;
10543 default: iv = va_arg(*args, int); break;
10545 case 'j': iv = va_arg(*args, intmax_t); break;
10549 iv = va_arg(*args, Quad_t); break;
10556 IV tiv = SvIV(argsv); /* work around GCC bug #13488 */
10558 case 'c': iv = (char)tiv; break;
10559 case 'h': iv = (short)tiv; break;
10560 case 'l': iv = (long)tiv; break;
10562 default: iv = tiv; break;
10565 iv = (Quad_t)tiv; break;
10571 if ( !vectorize ) /* we already set uv above */
10576 esignbuf[esignlen++] = plus;
10580 esignbuf[esignlen++] = '-';
10624 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
10635 case 'c': uv = (unsigned char)va_arg(*args, unsigned); break;
10636 case 'h': uv = (unsigned short)va_arg(*args, unsigned); break;
10637 case 'l': uv = va_arg(*args, unsigned long); break;
10638 case 'V': uv = va_arg(*args, UV); break;
10639 case 'z': uv = va_arg(*args, Size_t); break;
10640 case 't': uv = va_arg(*args, ptrdiff_t); break; /* will sign extend, but there is no uptrdiff_t, so oh well */
10642 case 'j': uv = va_arg(*args, uintmax_t); break;
10644 default: uv = va_arg(*args, unsigned); break;
10647 uv = va_arg(*args, Uquad_t); break;
10654 UV tuv = SvUV(argsv); /* work around GCC bug #13488 */
10656 case 'c': uv = (unsigned char)tuv; break;
10657 case 'h': uv = (unsigned short)tuv; break;
10658 case 'l': uv = (unsigned long)tuv; break;
10660 default: uv = tuv; break;
10663 uv = (Uquad_t)tuv; break;
10672 char *ptr = ebuf + sizeof ebuf;
10673 bool tempalt = uv ? alt : FALSE; /* Vectors can't change alt */
10679 p = (char *)((c == 'X') ? PL_hexdigit + 16 : PL_hexdigit);
10683 } while (uv >>= 4);
10685 esignbuf[esignlen++] = '0';
10686 esignbuf[esignlen++] = c; /* 'x' or 'X' */
10692 *--ptr = '0' + dig;
10693 } while (uv >>= 3);
10694 if (alt && *ptr != '0')
10700 *--ptr = '0' + dig;
10701 } while (uv >>= 1);
10703 esignbuf[esignlen++] = '0';
10704 esignbuf[esignlen++] = c;
10707 default: /* it had better be ten or less */
10710 *--ptr = '0' + dig;
10711 } while (uv /= base);
10714 elen = (ebuf + sizeof ebuf) - ptr;
10718 zeros = precis - elen;
10719 else if (precis == 0 && elen == 1 && *eptr == '0'
10720 && !(base == 8 && alt)) /* "%#.0o" prints "0" */
10723 /* a precision nullifies the 0 flag. */
10730 /* FLOATING POINT */
10733 c = 'f'; /* maybe %F isn't supported here */
10735 case 'e': case 'E':
10737 case 'g': case 'G':
10741 /* This is evil, but floating point is even more evil */
10743 /* for SV-style calling, we can only get NV
10744 for C-style calling, we assume %f is double;
10745 for simplicity we allow any of %Lf, %llf, %qf for long double
10749 #if defined(USE_LONG_DOUBLE)
10753 /* [perl #20339] - we should accept and ignore %lf rather than die */
10757 #if defined(USE_LONG_DOUBLE)
10758 intsize = args ? 0 : 'q';
10762 #if defined(HAS_LONG_DOUBLE)
10775 /* now we need (long double) if intsize == 'q', else (double) */
10777 #if LONG_DOUBLESIZE > DOUBLESIZE
10779 va_arg(*args, long double) :
10780 va_arg(*args, double)
10782 va_arg(*args, double)
10787 /* nv * 0 will be NaN for NaN, +Inf and -Inf, and 0 for anything
10788 else. frexp() has some unspecified behaviour for those three */
10789 if (c != 'e' && c != 'E' && (nv * 0) == 0) {
10791 /* FIXME: if HAS_LONG_DOUBLE but not USE_LONG_DOUBLE this
10792 will cast our (long double) to (double) */
10793 (void)Perl_frexp(nv, &i);
10794 if (i == PERL_INT_MIN)
10795 Perl_die(aTHX_ "panic: frexp");
10797 need = BIT_DIGITS(i);
10799 need += has_precis ? precis : 6; /* known default */
10804 #ifdef HAS_LDBL_SPRINTF_BUG
10805 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
10806 with sfio - Allen <allens@cpan.org> */
10809 # define MY_DBL_MAX DBL_MAX
10810 # else /* XXX guessing! HUGE_VAL may be defined as infinity, so not using */
10811 # if DOUBLESIZE >= 8
10812 # define MY_DBL_MAX 1.7976931348623157E+308L
10814 # define MY_DBL_MAX 3.40282347E+38L
10818 # ifdef HAS_LDBL_SPRINTF_BUG_LESS1 /* only between -1L & 1L - Allen */
10819 # define MY_DBL_MAX_BUG 1L
10821 # define MY_DBL_MAX_BUG MY_DBL_MAX
10825 # define MY_DBL_MIN DBL_MIN
10826 # else /* XXX guessing! -Allen */
10827 # if DOUBLESIZE >= 8
10828 # define MY_DBL_MIN 2.2250738585072014E-308L
10830 # define MY_DBL_MIN 1.17549435E-38L
10834 if ((intsize == 'q') && (c == 'f') &&
10835 ((nv < MY_DBL_MAX_BUG) && (nv > -MY_DBL_MAX_BUG)) &&
10836 (need < DBL_DIG)) {
10837 /* it's going to be short enough that
10838 * long double precision is not needed */
10840 if ((nv <= 0L) && (nv >= -0L))
10841 fix_ldbl_sprintf_bug = TRUE; /* 0 is 0 - easiest */
10843 /* would use Perl_fp_class as a double-check but not
10844 * functional on IRIX - see perl.h comments */
10846 if ((nv >= MY_DBL_MIN) || (nv <= -MY_DBL_MIN)) {
10847 /* It's within the range that a double can represent */
10848 #if defined(DBL_MAX) && !defined(DBL_MIN)
10849 if ((nv >= ((long double)1/DBL_MAX)) ||
10850 (nv <= (-(long double)1/DBL_MAX)))
10852 fix_ldbl_sprintf_bug = TRUE;
10855 if (fix_ldbl_sprintf_bug == TRUE) {
10865 # undef MY_DBL_MAX_BUG
10868 #endif /* HAS_LDBL_SPRINTF_BUG */
10870 need += 20; /* fudge factor */
10871 if (PL_efloatsize < need) {
10872 Safefree(PL_efloatbuf);
10873 PL_efloatsize = need + 20; /* more fudge */
10874 Newx(PL_efloatbuf, PL_efloatsize, char);
10875 PL_efloatbuf[0] = '\0';
10878 if ( !(width || left || plus || alt) && fill != '0'
10879 && has_precis && intsize != 'q' ) { /* Shortcuts */
10880 /* See earlier comment about buggy Gconvert when digits,
10882 if ( c == 'g' && precis) {
10883 Gconvert((NV)nv, (int)precis, 0, PL_efloatbuf);
10884 /* May return an empty string for digits==0 */
10885 if (*PL_efloatbuf) {
10886 elen = strlen(PL_efloatbuf);
10887 goto float_converted;
10889 } else if ( c == 'f' && !precis) {
10890 if ((eptr = F0convert(nv, ebuf + sizeof ebuf, &elen)))
10895 char *ptr = ebuf + sizeof ebuf;
10898 /* FIXME: what to do if HAS_LONG_DOUBLE but not PERL_PRIfldbl? */
10899 #if defined(HAS_LONG_DOUBLE) && defined(PERL_PRIfldbl)
10900 if (intsize == 'q') {
10901 /* Copy the one or more characters in a long double
10902 * format before the 'base' ([efgEFG]) character to
10903 * the format string. */
10904 static char const prifldbl[] = PERL_PRIfldbl;
10905 char const *p = prifldbl + sizeof(prifldbl) - 3;
10906 while (p >= prifldbl) { *--ptr = *p--; }
10911 do { *--ptr = '0' + (base % 10); } while (base /= 10);
10916 do { *--ptr = '0' + (base % 10); } while (base /= 10);
10928 /* No taint. Otherwise we are in the strange situation
10929 * where printf() taints but print($float) doesn't.
10931 #if defined(HAS_LONG_DOUBLE)
10932 elen = ((intsize == 'q')
10933 ? my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, nv)
10934 : my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, (double)nv));
10936 elen = my_sprintf(PL_efloatbuf, ptr, nv);
10940 eptr = PL_efloatbuf;
10948 i = SvCUR(sv) - origlen;
10951 case 'c': *(va_arg(*args, char*)) = i; break;
10952 case 'h': *(va_arg(*args, short*)) = i; break;
10953 default: *(va_arg(*args, int*)) = i; break;
10954 case 'l': *(va_arg(*args, long*)) = i; break;
10955 case 'V': *(va_arg(*args, IV*)) = i; break;
10956 case 'z': *(va_arg(*args, SSize_t*)) = i; break;
10957 case 't': *(va_arg(*args, ptrdiff_t*)) = i; break;
10959 case 'j': *(va_arg(*args, intmax_t*)) = i; break;
10963 *(va_arg(*args, Quad_t*)) = i; break;
10970 sv_setuv_mg(argsv, (UV)i);
10971 continue; /* not "break" */
10978 && (PL_op->op_type == OP_PRTF || PL_op->op_type == OP_SPRINTF)
10979 && ckWARN(WARN_PRINTF))
10981 SV * const msg = sv_newmortal();
10982 Perl_sv_setpvf(aTHX_ msg, "Invalid conversion in %sprintf: ",
10983 (PL_op->op_type == OP_PRTF) ? "" : "s");
10984 if (fmtstart < patend) {
10985 const char * const fmtend = q < patend ? q : patend;
10987 sv_catpvs(msg, "\"%");
10988 for (f = fmtstart; f < fmtend; f++) {
10990 sv_catpvn(msg, f, 1);
10992 Perl_sv_catpvf(aTHX_ msg,
10993 "\\%03"UVof, (UV)*f & 0xFF);
10996 sv_catpvs(msg, "\"");
10998 sv_catpvs(msg, "end of string");
11000 Perl_warner(aTHX_ packWARN(WARN_PRINTF), "%"SVf, SVfARG(msg)); /* yes, this is reentrant */
11003 /* output mangled stuff ... */
11009 /* ... right here, because formatting flags should not apply */
11010 SvGROW(sv, SvCUR(sv) + elen + 1);
11012 Copy(eptr, p, elen, char);
11015 SvCUR_set(sv, p - SvPVX_const(sv));
11017 continue; /* not "break" */
11020 if (is_utf8 != has_utf8) {
11023 sv_utf8_upgrade(sv);
11026 const STRLEN old_elen = elen;
11027 SV * const nsv = newSVpvn_flags(eptr, elen, SVs_TEMP);
11028 sv_utf8_upgrade(nsv);
11029 eptr = SvPVX_const(nsv);
11032 if (width) { /* fudge width (can't fudge elen) */
11033 width += elen - old_elen;
11039 have = esignlen + zeros + elen;
11041 Perl_croak_nocontext("%s", PL_memory_wrap);
11043 need = (have > width ? have : width);
11046 if (need >= (((STRLEN)~0) - SvCUR(sv) - dotstrlen - 1))
11047 Perl_croak_nocontext("%s", PL_memory_wrap);
11048 SvGROW(sv, SvCUR(sv) + need + dotstrlen + 1);
11050 if (esignlen && fill == '0') {
11052 for (i = 0; i < (int)esignlen; i++)
11053 *p++ = esignbuf[i];
11055 if (gap && !left) {
11056 memset(p, fill, gap);
11059 if (esignlen && fill != '0') {
11061 for (i = 0; i < (int)esignlen; i++)
11062 *p++ = esignbuf[i];
11066 for (i = zeros; i; i--)
11070 Copy(eptr, p, elen, char);
11074 memset(p, ' ', gap);
11079 Copy(dotstr, p, dotstrlen, char);
11083 vectorize = FALSE; /* done iterating over vecstr */
11090 SvCUR_set(sv, p - SvPVX_const(sv));
11099 /* =========================================================================
11101 =head1 Cloning an interpreter
11103 All the macros and functions in this section are for the private use of
11104 the main function, perl_clone().
11106 The foo_dup() functions make an exact copy of an existing foo thingy.
11107 During the course of a cloning, a hash table is used to map old addresses
11108 to new addresses. The table is created and manipulated with the
11109 ptr_table_* functions.
11113 * =========================================================================*/
11116 #if defined(USE_ITHREADS)
11118 /* XXX Remove this so it doesn't have to go thru the macro and return for nothing */
11119 #ifndef GpREFCNT_inc
11120 # define GpREFCNT_inc(gp) ((gp) ? (++(gp)->gp_refcnt, (gp)) : (GP*)NULL)
11124 /* Certain cases in Perl_ss_dup have been merged, by relying on the fact
11125 that currently av_dup, gv_dup and hv_dup are the same as sv_dup.
11126 If this changes, please unmerge ss_dup.
11127 Likewise, sv_dup_inc_multiple() relies on this fact. */
11128 #define sv_dup_inc_NN(s,t) SvREFCNT_inc_NN(sv_dup_inc(s,t))
11129 #define av_dup(s,t) MUTABLE_AV(sv_dup((const SV *)s,t))
11130 #define av_dup_inc(s,t) MUTABLE_AV(sv_dup_inc((const SV *)s,t))
11131 #define hv_dup(s,t) MUTABLE_HV(sv_dup((const SV *)s,t))
11132 #define hv_dup_inc(s,t) MUTABLE_HV(sv_dup_inc((const SV *)s,t))
11133 #define cv_dup(s,t) MUTABLE_CV(sv_dup((const SV *)s,t))
11134 #define cv_dup_inc(s,t) MUTABLE_CV(sv_dup_inc((const SV *)s,t))
11135 #define io_dup(s,t) MUTABLE_IO(sv_dup((const SV *)s,t))
11136 #define io_dup_inc(s,t) MUTABLE_IO(sv_dup_inc((const SV *)s,t))
11137 #define gv_dup(s,t) MUTABLE_GV(sv_dup((const SV *)s,t))
11138 #define gv_dup_inc(s,t) MUTABLE_GV(sv_dup_inc((const SV *)s,t))
11139 #define SAVEPV(p) ((p) ? savepv(p) : NULL)
11140 #define SAVEPVN(p,n) ((p) ? savepvn(p,n) : NULL)
11142 /* clone a parser */
11145 Perl_parser_dup(pTHX_ const yy_parser *const proto, CLONE_PARAMS *const param)
11149 PERL_ARGS_ASSERT_PARSER_DUP;
11154 /* look for it in the table first */
11155 parser = (yy_parser *)ptr_table_fetch(PL_ptr_table, proto);
11159 /* create anew and remember what it is */
11160 Newxz(parser, 1, yy_parser);
11161 ptr_table_store(PL_ptr_table, proto, parser);
11163 /* XXX these not yet duped */
11164 parser->old_parser = NULL;
11165 parser->stack = NULL;
11167 parser->stack_size = 0;
11168 /* XXX parser->stack->state = 0; */
11170 /* XXX eventually, just Copy() most of the parser struct ? */
11172 parser->lex_brackets = proto->lex_brackets;
11173 parser->lex_casemods = proto->lex_casemods;
11174 parser->lex_brackstack = savepvn(proto->lex_brackstack,
11175 (proto->lex_brackets < 120 ? 120 : proto->lex_brackets));
11176 parser->lex_casestack = savepvn(proto->lex_casestack,
11177 (proto->lex_casemods < 12 ? 12 : proto->lex_casemods));
11178 parser->lex_defer = proto->lex_defer;
11179 parser->lex_dojoin = proto->lex_dojoin;
11180 parser->lex_expect = proto->lex_expect;
11181 parser->lex_formbrack = proto->lex_formbrack;
11182 parser->lex_inpat = proto->lex_inpat;
11183 parser->lex_inwhat = proto->lex_inwhat;
11184 parser->lex_op = proto->lex_op;
11185 parser->lex_repl = sv_dup_inc(proto->lex_repl, param);
11186 parser->lex_starts = proto->lex_starts;
11187 parser->lex_stuff = sv_dup_inc(proto->lex_stuff, param);
11188 parser->multi_close = proto->multi_close;
11189 parser->multi_open = proto->multi_open;
11190 parser->multi_start = proto->multi_start;
11191 parser->multi_end = proto->multi_end;
11192 parser->pending_ident = proto->pending_ident;
11193 parser->preambled = proto->preambled;
11194 parser->sublex_info = proto->sublex_info; /* XXX not quite right */
11195 parser->linestr = sv_dup_inc(proto->linestr, param);
11196 parser->expect = proto->expect;
11197 parser->copline = proto->copline;
11198 parser->last_lop_op = proto->last_lop_op;
11199 parser->lex_state = proto->lex_state;
11200 parser->rsfp = fp_dup(proto->rsfp, '<', param);
11201 /* rsfp_filters entries have fake IoDIRP() */
11202 parser->rsfp_filters= av_dup_inc(proto->rsfp_filters, param);
11203 parser->in_my = proto->in_my;
11204 parser->in_my_stash = hv_dup(proto->in_my_stash, param);
11205 parser->error_count = proto->error_count;
11208 parser->linestr = sv_dup_inc(proto->linestr, param);
11211 char * const ols = SvPVX(proto->linestr);
11212 char * const ls = SvPVX(parser->linestr);
11214 parser->bufptr = ls + (proto->bufptr >= ols ?
11215 proto->bufptr - ols : 0);
11216 parser->oldbufptr = ls + (proto->oldbufptr >= ols ?
11217 proto->oldbufptr - ols : 0);
11218 parser->oldoldbufptr= ls + (proto->oldoldbufptr >= ols ?
11219 proto->oldoldbufptr - ols : 0);
11220 parser->linestart = ls + (proto->linestart >= ols ?
11221 proto->linestart - ols : 0);
11222 parser->last_uni = ls + (proto->last_uni >= ols ?
11223 proto->last_uni - ols : 0);
11224 parser->last_lop = ls + (proto->last_lop >= ols ?
11225 proto->last_lop - ols : 0);
11227 parser->bufend = ls + SvCUR(parser->linestr);
11230 Copy(proto->tokenbuf, parser->tokenbuf, 256, char);
11234 parser->endwhite = proto->endwhite;
11235 parser->faketokens = proto->faketokens;
11236 parser->lasttoke = proto->lasttoke;
11237 parser->nextwhite = proto->nextwhite;
11238 parser->realtokenstart = proto->realtokenstart;
11239 parser->skipwhite = proto->skipwhite;
11240 parser->thisclose = proto->thisclose;
11241 parser->thismad = proto->thismad;
11242 parser->thisopen = proto->thisopen;
11243 parser->thisstuff = proto->thisstuff;
11244 parser->thistoken = proto->thistoken;
11245 parser->thiswhite = proto->thiswhite;
11247 Copy(proto->nexttoke, parser->nexttoke, 5, NEXTTOKE);
11248 parser->curforce = proto->curforce;
11250 Copy(proto->nextval, parser->nextval, 5, YYSTYPE);
11251 Copy(proto->nexttype, parser->nexttype, 5, I32);
11252 parser->nexttoke = proto->nexttoke;
11255 /* XXX should clone saved_curcop here, but we aren't passed
11256 * proto_perl; so do it in perl_clone_using instead */
11262 /* duplicate a file handle */
11265 Perl_fp_dup(pTHX_ PerlIO *const fp, const char type, CLONE_PARAMS *const param)
11269 PERL_ARGS_ASSERT_FP_DUP;
11270 PERL_UNUSED_ARG(type);
11273 return (PerlIO*)NULL;
11275 /* look for it in the table first */
11276 ret = (PerlIO*)ptr_table_fetch(PL_ptr_table, fp);
11280 /* create anew and remember what it is */
11281 ret = PerlIO_fdupopen(aTHX_ fp, param, PERLIO_DUP_CLONE);
11282 ptr_table_store(PL_ptr_table, fp, ret);
11286 /* duplicate a directory handle */
11289 Perl_dirp_dup(pTHX_ DIR *const dp, CLONE_PARAMS *const param)
11295 register const Direntry_t *dirent;
11296 char smallbuf[256];
11302 PERL_UNUSED_CONTEXT;
11303 PERL_ARGS_ASSERT_DIRP_DUP;
11308 /* look for it in the table first */
11309 ret = (DIR*)ptr_table_fetch(PL_ptr_table, dp);
11315 PERL_UNUSED_ARG(param);
11319 /* open the current directory (so we can switch back) */
11320 if (!(pwd = PerlDir_open("."))) return (DIR *)NULL;
11322 /* chdir to our dir handle and open the present working directory */
11323 if (fchdir(my_dirfd(dp)) < 0 || !(ret = PerlDir_open("."))) {
11324 PerlDir_close(pwd);
11325 return (DIR *)NULL;
11327 /* Now we should have two dir handles pointing to the same dir. */
11329 /* Be nice to the calling code and chdir back to where we were. */
11330 fchdir(my_dirfd(pwd)); /* If this fails, then what? */
11332 /* We have no need of the pwd handle any more. */
11333 PerlDir_close(pwd);
11336 # define d_namlen(d) (d)->d_namlen
11338 # define d_namlen(d) strlen((d)->d_name)
11340 /* Iterate once through dp, to get the file name at the current posi-
11341 tion. Then step back. */
11342 pos = PerlDir_tell(dp);
11343 if ((dirent = PerlDir_read(dp))) {
11344 len = d_namlen(dirent);
11345 if (len <= sizeof smallbuf) name = smallbuf;
11346 else Newx(name, len, char);
11347 Move(dirent->d_name, name, len, char);
11349 PerlDir_seek(dp, pos);
11351 /* Iterate through the new dir handle, till we find a file with the
11353 if (!dirent) /* just before the end */
11355 pos = PerlDir_tell(ret);
11356 if (PerlDir_read(ret)) continue; /* not there yet */
11357 PerlDir_seek(ret, pos); /* step back */
11361 const long pos0 = PerlDir_tell(ret);
11363 pos = PerlDir_tell(ret);
11364 if ((dirent = PerlDir_read(ret))) {
11365 if (len == d_namlen(dirent)
11366 && memEQ(name, dirent->d_name, len)) {
11368 PerlDir_seek(ret, pos); /* step back */
11371 /* else we are not there yet; keep iterating */
11373 else { /* This is not meant to happen. The best we can do is
11374 reset the iterator to the beginning. */
11375 PerlDir_seek(ret, pos0);
11382 if (name && name != smallbuf)
11387 ret = win32_dirp_dup(dp, param);
11390 /* pop it in the pointer table */
11392 ptr_table_store(PL_ptr_table, dp, ret);
11397 /* duplicate a typeglob */
11400 Perl_gp_dup(pTHX_ GP *const gp, CLONE_PARAMS *const param)
11404 PERL_ARGS_ASSERT_GP_DUP;
11408 /* look for it in the table first */
11409 ret = (GP*)ptr_table_fetch(PL_ptr_table, gp);
11413 /* create anew and remember what it is */
11415 ptr_table_store(PL_ptr_table, gp, ret);
11418 /* ret->gp_refcnt must be 0 before any other dups are called. We're relying
11419 on Newxz() to do this for us. */
11420 ret->gp_sv = sv_dup_inc(gp->gp_sv, param);
11421 ret->gp_io = io_dup_inc(gp->gp_io, param);
11422 ret->gp_form = cv_dup_inc(gp->gp_form, param);
11423 ret->gp_av = av_dup_inc(gp->gp_av, param);
11424 ret->gp_hv = hv_dup_inc(gp->gp_hv, param);
11425 ret->gp_egv = gv_dup(gp->gp_egv, param);/* GvEGV is not refcounted */
11426 ret->gp_cv = cv_dup_inc(gp->gp_cv, param);
11427 ret->gp_cvgen = gp->gp_cvgen;
11428 ret->gp_line = gp->gp_line;
11429 ret->gp_file_hek = hek_dup(gp->gp_file_hek, param);
11433 /* duplicate a chain of magic */
11436 Perl_mg_dup(pTHX_ MAGIC *mg, CLONE_PARAMS *const param)
11438 MAGIC *mgret = NULL;
11439 MAGIC **mgprev_p = &mgret;
11441 PERL_ARGS_ASSERT_MG_DUP;
11443 for (; mg; mg = mg->mg_moremagic) {
11446 if ((param->flags & CLONEf_JOIN_IN)
11447 && mg->mg_type == PERL_MAGIC_backref)
11448 /* when joining, we let the individual SVs add themselves to
11449 * backref as needed. */
11452 Newx(nmg, 1, MAGIC);
11454 mgprev_p = &(nmg->mg_moremagic);
11456 /* There was a comment "XXX copy dynamic vtable?" but as we don't have
11457 dynamic vtables, I'm not sure why Sarathy wrote it. The comment dates
11458 from the original commit adding Perl_mg_dup() - revision 4538.
11459 Similarly there is the annotation "XXX random ptr?" next to the
11460 assignment to nmg->mg_ptr. */
11463 /* FIXME for plugins
11464 if (nmg->mg_type == PERL_MAGIC_qr) {
11465 nmg->mg_obj = MUTABLE_SV(CALLREGDUPE((REGEXP*)nmg->mg_obj, param));
11469 nmg->mg_obj = (nmg->mg_flags & MGf_REFCOUNTED)
11470 ? nmg->mg_type == PERL_MAGIC_backref
11471 /* The backref AV has its reference
11472 * count deliberately bumped by 1 */
11473 ? SvREFCNT_inc(av_dup_inc((const AV *)
11474 nmg->mg_obj, param))
11475 : sv_dup_inc(nmg->mg_obj, param)
11476 : sv_dup(nmg->mg_obj, param);
11478 if (nmg->mg_ptr && nmg->mg_type != PERL_MAGIC_regex_global) {
11479 if (nmg->mg_len > 0) {
11480 nmg->mg_ptr = SAVEPVN(nmg->mg_ptr, nmg->mg_len);
11481 if (nmg->mg_type == PERL_MAGIC_overload_table &&
11482 AMT_AMAGIC((AMT*)nmg->mg_ptr))
11484 AMT * const namtp = (AMT*)nmg->mg_ptr;
11485 sv_dup_inc_multiple((SV**)(namtp->table),
11486 (SV**)(namtp->table), NofAMmeth, param);
11489 else if (nmg->mg_len == HEf_SVKEY)
11490 nmg->mg_ptr = (char*)sv_dup_inc((const SV *)nmg->mg_ptr, param);
11492 if ((nmg->mg_flags & MGf_DUP) && nmg->mg_virtual && nmg->mg_virtual->svt_dup) {
11493 nmg->mg_virtual->svt_dup(aTHX_ nmg, param);
11499 #endif /* USE_ITHREADS */
11501 struct ptr_tbl_arena {
11502 struct ptr_tbl_arena *next;
11503 struct ptr_tbl_ent array[1023/3]; /* as ptr_tbl_ent has 3 pointers. */
11506 /* create a new pointer-mapping table */
11509 Perl_ptr_table_new(pTHX)
11512 PERL_UNUSED_CONTEXT;
11514 Newx(tbl, 1, PTR_TBL_t);
11515 tbl->tbl_max = 511;
11516 tbl->tbl_items = 0;
11517 tbl->tbl_arena = NULL;
11518 tbl->tbl_arena_next = NULL;
11519 tbl->tbl_arena_end = NULL;
11520 Newxz(tbl->tbl_ary, tbl->tbl_max + 1, PTR_TBL_ENT_t*);
11524 #define PTR_TABLE_HASH(ptr) \
11525 ((PTR2UV(ptr) >> 3) ^ (PTR2UV(ptr) >> (3 + 7)) ^ (PTR2UV(ptr) >> (3 + 17)))
11527 /* map an existing pointer using a table */
11529 STATIC PTR_TBL_ENT_t *
11530 S_ptr_table_find(PTR_TBL_t *const tbl, const void *const sv)
11532 PTR_TBL_ENT_t *tblent;
11533 const UV hash = PTR_TABLE_HASH(sv);
11535 PERL_ARGS_ASSERT_PTR_TABLE_FIND;
11537 tblent = tbl->tbl_ary[hash & tbl->tbl_max];
11538 for (; tblent; tblent = tblent->next) {
11539 if (tblent->oldval == sv)
11546 Perl_ptr_table_fetch(pTHX_ PTR_TBL_t *const tbl, const void *const sv)
11548 PTR_TBL_ENT_t const *const tblent = ptr_table_find(tbl, sv);
11550 PERL_ARGS_ASSERT_PTR_TABLE_FETCH;
11551 PERL_UNUSED_CONTEXT;
11553 return tblent ? tblent->newval : NULL;
11556 /* add a new entry to a pointer-mapping table */
11559 Perl_ptr_table_store(pTHX_ PTR_TBL_t *const tbl, const void *const oldsv, void *const newsv)
11561 PTR_TBL_ENT_t *tblent = ptr_table_find(tbl, oldsv);
11563 PERL_ARGS_ASSERT_PTR_TABLE_STORE;
11564 PERL_UNUSED_CONTEXT;
11567 tblent->newval = newsv;
11569 const UV entry = PTR_TABLE_HASH(oldsv) & tbl->tbl_max;
11571 if (tbl->tbl_arena_next == tbl->tbl_arena_end) {
11572 struct ptr_tbl_arena *new_arena;
11574 Newx(new_arena, 1, struct ptr_tbl_arena);
11575 new_arena->next = tbl->tbl_arena;
11576 tbl->tbl_arena = new_arena;
11577 tbl->tbl_arena_next = new_arena->array;
11578 tbl->tbl_arena_end = new_arena->array
11579 + sizeof(new_arena->array) / sizeof(new_arena->array[0]);
11582 tblent = tbl->tbl_arena_next++;
11584 tblent->oldval = oldsv;
11585 tblent->newval = newsv;
11586 tblent->next = tbl->tbl_ary[entry];
11587 tbl->tbl_ary[entry] = tblent;
11589 if (tblent->next && tbl->tbl_items > tbl->tbl_max)
11590 ptr_table_split(tbl);
11594 /* double the hash bucket size of an existing ptr table */
11597 Perl_ptr_table_split(pTHX_ PTR_TBL_t *const tbl)
11599 PTR_TBL_ENT_t **ary = tbl->tbl_ary;
11600 const UV oldsize = tbl->tbl_max + 1;
11601 UV newsize = oldsize * 2;
11604 PERL_ARGS_ASSERT_PTR_TABLE_SPLIT;
11605 PERL_UNUSED_CONTEXT;
11607 Renew(ary, newsize, PTR_TBL_ENT_t*);
11608 Zero(&ary[oldsize], newsize-oldsize, PTR_TBL_ENT_t*);
11609 tbl->tbl_max = --newsize;
11610 tbl->tbl_ary = ary;
11611 for (i=0; i < oldsize; i++, ary++) {
11612 PTR_TBL_ENT_t **entp = ary;
11613 PTR_TBL_ENT_t *ent = *ary;
11614 PTR_TBL_ENT_t **curentp;
11617 curentp = ary + oldsize;
11619 if ((newsize & PTR_TABLE_HASH(ent->oldval)) != i) {
11621 ent->next = *curentp;
11631 /* remove all the entries from a ptr table */
11632 /* Deprecated - will be removed post 5.14 */
11635 Perl_ptr_table_clear(pTHX_ PTR_TBL_t *const tbl)
11637 if (tbl && tbl->tbl_items) {
11638 struct ptr_tbl_arena *arena = tbl->tbl_arena;
11640 Zero(tbl->tbl_ary, tbl->tbl_max + 1, struct ptr_tbl_ent **);
11643 struct ptr_tbl_arena *next = arena->next;
11649 tbl->tbl_items = 0;
11650 tbl->tbl_arena = NULL;
11651 tbl->tbl_arena_next = NULL;
11652 tbl->tbl_arena_end = NULL;
11656 /* clear and free a ptr table */
11659 Perl_ptr_table_free(pTHX_ PTR_TBL_t *const tbl)
11661 struct ptr_tbl_arena *arena;
11667 arena = tbl->tbl_arena;
11670 struct ptr_tbl_arena *next = arena->next;
11676 Safefree(tbl->tbl_ary);
11680 #if defined(USE_ITHREADS)
11683 Perl_rvpv_dup(pTHX_ SV *const dstr, const SV *const sstr, CLONE_PARAMS *const param)
11685 PERL_ARGS_ASSERT_RVPV_DUP;
11688 if (SvWEAKREF(sstr)) {
11689 SvRV_set(dstr, sv_dup(SvRV_const(sstr), param));
11690 if (param->flags & CLONEf_JOIN_IN) {
11691 /* if joining, we add any back references individually rather
11692 * than copying the whole backref array */
11693 Perl_sv_add_backref(aTHX_ SvRV(dstr), dstr);
11697 SvRV_set(dstr, sv_dup_inc(SvRV_const(sstr), param));
11699 else if (SvPVX_const(sstr)) {
11700 /* Has something there */
11702 /* Normal PV - clone whole allocated space */
11703 SvPV_set(dstr, SAVEPVN(SvPVX_const(sstr), SvLEN(sstr)-1));
11704 if (SvREADONLY(sstr) && SvFAKE(sstr)) {
11705 /* Not that normal - actually sstr is copy on write.
11706 But we are a true, independent SV, so: */
11707 SvREADONLY_off(dstr);
11712 /* Special case - not normally malloced for some reason */
11713 if (isGV_with_GP(sstr)) {
11714 /* Don't need to do anything here. */
11716 else if ((SvREADONLY(sstr) && SvFAKE(sstr))) {
11717 /* A "shared" PV - clone it as "shared" PV */
11719 HEK_KEY(hek_dup(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)),
11723 /* Some other special case - random pointer */
11724 SvPV_set(dstr, (char *) SvPVX_const(sstr));
11729 /* Copy the NULL */
11730 SvPV_set(dstr, NULL);
11734 /* duplicate a list of SVs. source and dest may point to the same memory. */
11736 S_sv_dup_inc_multiple(pTHX_ SV *const *source, SV **dest,
11737 SSize_t items, CLONE_PARAMS *const param)
11739 PERL_ARGS_ASSERT_SV_DUP_INC_MULTIPLE;
11741 while (items-- > 0) {
11742 *dest++ = sv_dup_inc(*source++, param);
11748 /* duplicate an SV of any type (including AV, HV etc) */
11751 S_sv_dup_common(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
11756 PERL_ARGS_ASSERT_SV_DUP_COMMON;
11758 if (SvTYPE(sstr) == (svtype)SVTYPEMASK) {
11759 #ifdef DEBUG_LEAKING_SCALARS_ABORT
11764 /* look for it in the table first */
11765 dstr = MUTABLE_SV(ptr_table_fetch(PL_ptr_table, sstr));
11769 if(param->flags & CLONEf_JOIN_IN) {
11770 /** We are joining here so we don't want do clone
11771 something that is bad **/
11772 if (SvTYPE(sstr) == SVt_PVHV) {
11773 const HEK * const hvname = HvNAME_HEK(sstr);
11775 /** don't clone stashes if they already exist **/
11776 dstr = MUTABLE_SV(gv_stashpvn(HEK_KEY(hvname), HEK_LEN(hvname),
11777 HEK_UTF8(hvname) ? SVf_UTF8 : 0));
11778 ptr_table_store(PL_ptr_table, sstr, dstr);
11784 /* create anew and remember what it is */
11787 #ifdef DEBUG_LEAKING_SCALARS
11788 dstr->sv_debug_optype = sstr->sv_debug_optype;
11789 dstr->sv_debug_line = sstr->sv_debug_line;
11790 dstr->sv_debug_inpad = sstr->sv_debug_inpad;
11791 dstr->sv_debug_parent = (SV*)sstr;
11792 FREE_SV_DEBUG_FILE(dstr);
11793 dstr->sv_debug_file = savepv(sstr->sv_debug_file);
11796 ptr_table_store(PL_ptr_table, sstr, dstr);
11799 SvFLAGS(dstr) = SvFLAGS(sstr);
11800 SvFLAGS(dstr) &= ~SVf_OOK; /* don't propagate OOK hack */
11801 SvREFCNT(dstr) = 0; /* must be before any other dups! */
11804 if (SvANY(sstr) && PL_watch_pvx && SvPVX_const(sstr) == PL_watch_pvx)
11805 PerlIO_printf(Perl_debug_log, "watch at %p hit, found string \"%s\"\n",
11806 (void*)PL_watch_pvx, SvPVX_const(sstr));
11809 /* don't clone objects whose class has asked us not to */
11810 if (SvOBJECT(sstr) && ! (SvFLAGS(SvSTASH(sstr)) & SVphv_CLONEABLE)) {
11815 switch (SvTYPE(sstr)) {
11817 SvANY(dstr) = NULL;
11820 SvANY(dstr) = (XPVIV*)((char*)&(dstr->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
11822 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
11824 SvIV_set(dstr, SvIVX(sstr));
11828 SvANY(dstr) = new_XNV();
11829 SvNV_set(dstr, SvNVX(sstr));
11831 /* case SVt_BIND: */
11834 /* These are all the types that need complex bodies allocating. */
11836 const svtype sv_type = SvTYPE(sstr);
11837 const struct body_details *const sv_type_details
11838 = bodies_by_type + sv_type;
11842 Perl_croak(aTHX_ "Bizarre SvTYPE [%" IVdf "]", (IV)SvTYPE(sstr));
11857 assert(sv_type_details->body_size);
11858 if (sv_type_details->arena) {
11859 new_body_inline(new_body, sv_type);
11861 = (void*)((char*)new_body - sv_type_details->offset);
11863 new_body = new_NOARENA(sv_type_details);
11867 SvANY(dstr) = new_body;
11870 Copy(((char*)SvANY(sstr)) + sv_type_details->offset,
11871 ((char*)SvANY(dstr)) + sv_type_details->offset,
11872 sv_type_details->copy, char);
11874 Copy(((char*)SvANY(sstr)),
11875 ((char*)SvANY(dstr)),
11876 sv_type_details->body_size + sv_type_details->offset, char);
11879 if (sv_type != SVt_PVAV && sv_type != SVt_PVHV
11880 && !isGV_with_GP(dstr)
11881 && !(sv_type == SVt_PVIO && !(IoFLAGS(dstr) & IOf_FAKE_DIRP)))
11882 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
11884 /* The Copy above means that all the source (unduplicated) pointers
11885 are now in the destination. We can check the flags and the
11886 pointers in either, but it's possible that there's less cache
11887 missing by always going for the destination.
11888 FIXME - instrument and check that assumption */
11889 if (sv_type >= SVt_PVMG) {
11890 if ((sv_type == SVt_PVMG) && SvPAD_OUR(dstr)) {
11891 SvOURSTASH_set(dstr, hv_dup_inc(SvOURSTASH(dstr), param));
11892 } else if (SvMAGIC(dstr))
11893 SvMAGIC_set(dstr, mg_dup(SvMAGIC(dstr), param));
11895 SvSTASH_set(dstr, hv_dup_inc(SvSTASH(dstr), param));
11898 /* The cast silences a GCC warning about unhandled types. */
11899 switch ((int)sv_type) {
11909 /* FIXME for plugins */
11910 re_dup_guts((REGEXP*) sstr, (REGEXP*) dstr, param);
11913 /* XXX LvTARGOFF sometimes holds PMOP* when DEBUGGING */
11914 if (LvTYPE(dstr) == 't') /* for tie: unrefcnted fake (SV**) */
11915 LvTARG(dstr) = dstr;
11916 else if (LvTYPE(dstr) == 'T') /* for tie: fake HE */
11917 LvTARG(dstr) = MUTABLE_SV(he_dup((HE*)LvTARG(dstr), 0, param));
11919 LvTARG(dstr) = sv_dup_inc(LvTARG(dstr), param);
11921 /* non-GP case already handled above */
11922 if(isGV_with_GP(sstr)) {
11923 GvNAME_HEK(dstr) = hek_dup(GvNAME_HEK(dstr), param);
11924 /* Don't call sv_add_backref here as it's going to be
11925 created as part of the magic cloning of the symbol
11926 table--unless this is during a join and the stash
11927 is not actually being cloned. */
11928 /* Danger Will Robinson - GvGP(dstr) isn't initialised
11929 at the point of this comment. */
11930 GvSTASH(dstr) = hv_dup(GvSTASH(dstr), param);
11931 if (param->flags & CLONEf_JOIN_IN)
11932 Perl_sv_add_backref(aTHX_ MUTABLE_SV(GvSTASH(dstr)), dstr);
11933 GvGP_set(dstr, gp_dup(GvGP(sstr), param));
11934 (void)GpREFCNT_inc(GvGP(dstr));
11938 /* PL_parser->rsfp_filters entries have fake IoDIRP() */
11939 if(IoFLAGS(dstr) & IOf_FAKE_DIRP) {
11940 /* I have no idea why fake dirp (rsfps)
11941 should be treated differently but otherwise
11942 we end up with leaks -- sky*/
11943 IoTOP_GV(dstr) = gv_dup_inc(IoTOP_GV(dstr), param);
11944 IoFMT_GV(dstr) = gv_dup_inc(IoFMT_GV(dstr), param);
11945 IoBOTTOM_GV(dstr) = gv_dup_inc(IoBOTTOM_GV(dstr), param);
11947 IoTOP_GV(dstr) = gv_dup(IoTOP_GV(dstr), param);
11948 IoFMT_GV(dstr) = gv_dup(IoFMT_GV(dstr), param);
11949 IoBOTTOM_GV(dstr) = gv_dup(IoBOTTOM_GV(dstr), param);
11950 if (IoDIRP(dstr)) {
11951 IoDIRP(dstr) = dirp_dup(IoDIRP(dstr), param);
11954 /* IoDIRP(dstr) is already a copy of IoDIRP(sstr) */
11956 IoIFP(dstr) = fp_dup(IoIFP(sstr), IoTYPE(dstr), param);
11958 if (IoOFP(dstr) == IoIFP(sstr))
11959 IoOFP(dstr) = IoIFP(dstr);
11961 IoOFP(dstr) = fp_dup(IoOFP(dstr), IoTYPE(dstr), param);
11962 IoTOP_NAME(dstr) = SAVEPV(IoTOP_NAME(dstr));
11963 IoFMT_NAME(dstr) = SAVEPV(IoFMT_NAME(dstr));
11964 IoBOTTOM_NAME(dstr) = SAVEPV(IoBOTTOM_NAME(dstr));
11967 /* avoid cloning an empty array */
11968 if (AvARRAY((const AV *)sstr) && AvFILLp((const AV *)sstr) >= 0) {
11969 SV **dst_ary, **src_ary;
11970 SSize_t items = AvFILLp((const AV *)sstr) + 1;
11972 src_ary = AvARRAY((const AV *)sstr);
11973 Newxz(dst_ary, AvMAX((const AV *)sstr)+1, SV*);
11974 ptr_table_store(PL_ptr_table, src_ary, dst_ary);
11975 AvARRAY(MUTABLE_AV(dstr)) = dst_ary;
11976 AvALLOC((const AV *)dstr) = dst_ary;
11977 if (AvREAL((const AV *)sstr)) {
11978 dst_ary = sv_dup_inc_multiple(src_ary, dst_ary, items,
11982 while (items-- > 0)
11983 *dst_ary++ = sv_dup(*src_ary++, param);
11985 items = AvMAX((const AV *)sstr) - AvFILLp((const AV *)sstr);
11986 while (items-- > 0) {
11987 *dst_ary++ = &PL_sv_undef;
11991 AvARRAY(MUTABLE_AV(dstr)) = NULL;
11992 AvALLOC((const AV *)dstr) = (SV**)NULL;
11993 AvMAX( (const AV *)dstr) = -1;
11994 AvFILLp((const AV *)dstr) = -1;
11998 if (HvARRAY((const HV *)sstr)) {
12000 const bool sharekeys = !!HvSHAREKEYS(sstr);
12001 XPVHV * const dxhv = (XPVHV*)SvANY(dstr);
12002 XPVHV * const sxhv = (XPVHV*)SvANY(sstr);
12004 Newx(darray, PERL_HV_ARRAY_ALLOC_BYTES(dxhv->xhv_max+1)
12005 + (SvOOK(sstr) ? sizeof(struct xpvhv_aux) : 0),
12007 HvARRAY(dstr) = (HE**)darray;
12008 while (i <= sxhv->xhv_max) {
12009 const HE * const source = HvARRAY(sstr)[i];
12010 HvARRAY(dstr)[i] = source
12011 ? he_dup(source, sharekeys, param) : 0;
12015 const struct xpvhv_aux * const saux = HvAUX(sstr);
12016 struct xpvhv_aux * const daux = HvAUX(dstr);
12017 /* This flag isn't copied. */
12018 /* SvOOK_on(hv) attacks the IV flags. */
12019 SvFLAGS(dstr) |= SVf_OOK;
12021 if (saux->xhv_name_count) {
12022 HEK ** const sname = saux->xhv_name_u.xhvnameu_names;
12024 = saux->xhv_name_count < 0
12025 ? -saux->xhv_name_count
12026 : saux->xhv_name_count;
12027 HEK **shekp = sname + count;
12029 Newx(daux->xhv_name_u.xhvnameu_names, count, HEK *);
12030 dhekp = daux->xhv_name_u.xhvnameu_names + count;
12031 while (shekp-- > sname) {
12033 *dhekp = hek_dup(*shekp, param);
12037 daux->xhv_name_u.xhvnameu_name
12038 = hek_dup(saux->xhv_name_u.xhvnameu_name,
12041 daux->xhv_name_count = saux->xhv_name_count;
12043 daux->xhv_riter = saux->xhv_riter;
12044 daux->xhv_eiter = saux->xhv_eiter
12045 ? he_dup(saux->xhv_eiter,
12046 cBOOL(HvSHAREKEYS(sstr)), param) : 0;
12047 /* backref array needs refcnt=2; see sv_add_backref */
12048 daux->xhv_backreferences =
12049 (param->flags & CLONEf_JOIN_IN)
12050 /* when joining, we let the individual GVs and
12051 * CVs add themselves to backref as
12052 * needed. This avoids pulling in stuff
12053 * that isn't required, and simplifies the
12054 * case where stashes aren't cloned back
12055 * if they already exist in the parent
12058 : saux->xhv_backreferences
12059 ? (SvTYPE(saux->xhv_backreferences) == SVt_PVAV)
12060 ? MUTABLE_AV(SvREFCNT_inc(
12061 sv_dup_inc((const SV *)
12062 saux->xhv_backreferences, param)))
12063 : MUTABLE_AV(sv_dup((const SV *)
12064 saux->xhv_backreferences, param))
12067 daux->xhv_mro_meta = saux->xhv_mro_meta
12068 ? mro_meta_dup(saux->xhv_mro_meta, param)
12071 /* Record stashes for possible cloning in Perl_clone(). */
12073 av_push(param->stashes, dstr);
12077 HvARRAY(MUTABLE_HV(dstr)) = NULL;
12080 if (!(param->flags & CLONEf_COPY_STACKS)) {
12085 /* NOTE: not refcounted */
12086 SvANY(MUTABLE_CV(dstr))->xcv_stash =
12087 hv_dup(CvSTASH(dstr), param);
12088 if ((param->flags & CLONEf_JOIN_IN) && CvSTASH(dstr))
12089 Perl_sv_add_backref(aTHX_ MUTABLE_SV(CvSTASH(dstr)), dstr);
12090 if (!CvISXSUB(dstr)) {
12092 CvROOT(dstr) = OpREFCNT_inc(CvROOT(dstr));
12094 } else if (CvCONST(dstr)) {
12095 CvXSUBANY(dstr).any_ptr =
12096 sv_dup_inc((const SV *)CvXSUBANY(dstr).any_ptr, param);
12098 if (CvDYNFILE(dstr)) CvFILE(dstr) = SAVEPV(CvFILE(dstr));
12099 /* don't dup if copying back - CvGV isn't refcounted, so the
12100 * duped GV may never be freed. A bit of a hack! DAPM */
12101 SvANY(MUTABLE_CV(dstr))->xcv_gv =
12103 ? gv_dup_inc(CvGV(sstr), param)
12104 : (param->flags & CLONEf_JOIN_IN)
12106 : gv_dup(CvGV(sstr), param);
12108 CvPADLIST(dstr) = padlist_dup(CvPADLIST(sstr), param);
12110 CvWEAKOUTSIDE(sstr)
12111 ? cv_dup( CvOUTSIDE(dstr), param)
12112 : cv_dup_inc(CvOUTSIDE(dstr), param);
12118 if (SvOBJECT(dstr) && SvTYPE(dstr) != SVt_PVIO)
12125 Perl_sv_dup_inc(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
12127 PERL_ARGS_ASSERT_SV_DUP_INC;
12128 return sstr ? SvREFCNT_inc(sv_dup_common(sstr, param)) : NULL;
12132 Perl_sv_dup(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
12134 SV *dstr = sstr ? sv_dup_common(sstr, param) : NULL;
12135 PERL_ARGS_ASSERT_SV_DUP;
12137 /* Track every SV that (at least initially) had a reference count of 0.
12138 We need to do this by holding an actual reference to it in this array.
12139 If we attempt to cheat, turn AvREAL_off(), and store only pointers
12140 (akin to the stashes hash, and the perl stack), we come unstuck if
12141 a weak reference (or other SV legitimately SvREFCNT() == 0 for this
12142 thread) is manipulated in a CLONE method, because CLONE runs before the
12143 unreferenced array is walked to find SVs still with SvREFCNT() == 0
12144 (and fix things up by giving each a reference via the temps stack).
12145 Instead, during CLONE, if the 0-referenced SV has SvREFCNT_inc() and
12146 then SvREFCNT_dec(), it will be cleaned up (and added to the free list)
12147 before the walk of unreferenced happens and a reference to that is SV
12148 added to the temps stack. At which point we have the same SV considered
12149 to be in use, and free to be re-used. Not good.
12151 if (dstr && !(param->flags & CLONEf_COPY_STACKS) && !SvREFCNT(dstr)) {
12152 assert(param->unreferenced);
12153 av_push(param->unreferenced, SvREFCNT_inc(dstr));
12159 /* duplicate a context */
12162 Perl_cx_dup(pTHX_ PERL_CONTEXT *cxs, I32 ix, I32 max, CLONE_PARAMS* param)
12164 PERL_CONTEXT *ncxs;
12166 PERL_ARGS_ASSERT_CX_DUP;
12169 return (PERL_CONTEXT*)NULL;
12171 /* look for it in the table first */
12172 ncxs = (PERL_CONTEXT*)ptr_table_fetch(PL_ptr_table, cxs);
12176 /* create anew and remember what it is */
12177 Newx(ncxs, max + 1, PERL_CONTEXT);
12178 ptr_table_store(PL_ptr_table, cxs, ncxs);
12179 Copy(cxs, ncxs, max + 1, PERL_CONTEXT);
12182 PERL_CONTEXT * const ncx = &ncxs[ix];
12183 if (CxTYPE(ncx) == CXt_SUBST) {
12184 Perl_croak(aTHX_ "Cloning substitution context is unimplemented");
12187 switch (CxTYPE(ncx)) {
12189 ncx->blk_sub.cv = (ncx->blk_sub.olddepth == 0
12190 ? cv_dup_inc(ncx->blk_sub.cv, param)
12191 : cv_dup(ncx->blk_sub.cv,param));
12192 ncx->blk_sub.argarray = (CxHASARGS(ncx)
12193 ? av_dup_inc(ncx->blk_sub.argarray,
12196 ncx->blk_sub.savearray = av_dup_inc(ncx->blk_sub.savearray,
12198 ncx->blk_sub.oldcomppad = (PAD*)ptr_table_fetch(PL_ptr_table,
12199 ncx->blk_sub.oldcomppad);
12202 ncx->blk_eval.old_namesv = sv_dup_inc(ncx->blk_eval.old_namesv,
12204 ncx->blk_eval.cur_text = sv_dup(ncx->blk_eval.cur_text, param);
12206 case CXt_LOOP_LAZYSV:
12207 ncx->blk_loop.state_u.lazysv.end
12208 = sv_dup_inc(ncx->blk_loop.state_u.lazysv.end, param);
12209 /* We are taking advantage of av_dup_inc and sv_dup_inc
12210 actually being the same function, and order equivalence of
12212 We can assert the later [but only at run time :-(] */
12213 assert ((void *) &ncx->blk_loop.state_u.ary.ary ==
12214 (void *) &ncx->blk_loop.state_u.lazysv.cur);
12216 ncx->blk_loop.state_u.ary.ary
12217 = av_dup_inc(ncx->blk_loop.state_u.ary.ary, param);
12218 case CXt_LOOP_LAZYIV:
12219 case CXt_LOOP_PLAIN:
12220 if (CxPADLOOP(ncx)) {
12221 ncx->blk_loop.itervar_u.oldcomppad
12222 = (PAD*)ptr_table_fetch(PL_ptr_table,
12223 ncx->blk_loop.itervar_u.oldcomppad);
12225 ncx->blk_loop.itervar_u.gv
12226 = gv_dup((const GV *)ncx->blk_loop.itervar_u.gv,
12231 ncx->blk_format.cv = cv_dup(ncx->blk_format.cv, param);
12232 ncx->blk_format.gv = gv_dup(ncx->blk_format.gv, param);
12233 ncx->blk_format.dfoutgv = gv_dup_inc(ncx->blk_format.dfoutgv,
12246 /* duplicate a stack info structure */
12249 Perl_si_dup(pTHX_ PERL_SI *si, CLONE_PARAMS* param)
12253 PERL_ARGS_ASSERT_SI_DUP;
12256 return (PERL_SI*)NULL;
12258 /* look for it in the table first */
12259 nsi = (PERL_SI*)ptr_table_fetch(PL_ptr_table, si);
12263 /* create anew and remember what it is */
12264 Newxz(nsi, 1, PERL_SI);
12265 ptr_table_store(PL_ptr_table, si, nsi);
12267 nsi->si_stack = av_dup_inc(si->si_stack, param);
12268 nsi->si_cxix = si->si_cxix;
12269 nsi->si_cxmax = si->si_cxmax;
12270 nsi->si_cxstack = cx_dup(si->si_cxstack, si->si_cxix, si->si_cxmax, param);
12271 nsi->si_type = si->si_type;
12272 nsi->si_prev = si_dup(si->si_prev, param);
12273 nsi->si_next = si_dup(si->si_next, param);
12274 nsi->si_markoff = si->si_markoff;
12279 #define POPINT(ss,ix) ((ss)[--(ix)].any_i32)
12280 #define TOPINT(ss,ix) ((ss)[ix].any_i32)
12281 #define POPLONG(ss,ix) ((ss)[--(ix)].any_long)
12282 #define TOPLONG(ss,ix) ((ss)[ix].any_long)
12283 #define POPIV(ss,ix) ((ss)[--(ix)].any_iv)
12284 #define TOPIV(ss,ix) ((ss)[ix].any_iv)
12285 #define POPUV(ss,ix) ((ss)[--(ix)].any_uv)
12286 #define TOPUV(ss,ix) ((ss)[ix].any_uv)
12287 #define POPBOOL(ss,ix) ((ss)[--(ix)].any_bool)
12288 #define TOPBOOL(ss,ix) ((ss)[ix].any_bool)
12289 #define POPPTR(ss,ix) ((ss)[--(ix)].any_ptr)
12290 #define TOPPTR(ss,ix) ((ss)[ix].any_ptr)
12291 #define POPDPTR(ss,ix) ((ss)[--(ix)].any_dptr)
12292 #define TOPDPTR(ss,ix) ((ss)[ix].any_dptr)
12293 #define POPDXPTR(ss,ix) ((ss)[--(ix)].any_dxptr)
12294 #define TOPDXPTR(ss,ix) ((ss)[ix].any_dxptr)
12297 #define pv_dup_inc(p) SAVEPV(p)
12298 #define pv_dup(p) SAVEPV(p)
12299 #define svp_dup_inc(p,pp) any_dup(p,pp)
12301 /* map any object to the new equivent - either something in the
12302 * ptr table, or something in the interpreter structure
12306 Perl_any_dup(pTHX_ void *v, const PerlInterpreter *proto_perl)
12310 PERL_ARGS_ASSERT_ANY_DUP;
12313 return (void*)NULL;
12315 /* look for it in the table first */
12316 ret = ptr_table_fetch(PL_ptr_table, v);
12320 /* see if it is part of the interpreter structure */
12321 if (v >= (void*)proto_perl && v < (void*)(proto_perl+1))
12322 ret = (void*)(((char*)aTHX) + (((char*)v) - (char*)proto_perl));
12330 /* duplicate the save stack */
12333 Perl_ss_dup(pTHX_ PerlInterpreter *proto_perl, CLONE_PARAMS* param)
12336 ANY * const ss = proto_perl->Isavestack;
12337 const I32 max = proto_perl->Isavestack_max;
12338 I32 ix = proto_perl->Isavestack_ix;
12351 void (*dptr) (void*);
12352 void (*dxptr) (pTHX_ void*);
12354 PERL_ARGS_ASSERT_SS_DUP;
12356 Newxz(nss, max, ANY);
12359 const UV uv = POPUV(ss,ix);
12360 const U8 type = (U8)uv & SAVE_MASK;
12362 TOPUV(nss,ix) = uv;
12364 case SAVEt_CLEARSV:
12366 case SAVEt_HELEM: /* hash element */
12367 sv = (const SV *)POPPTR(ss,ix);
12368 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12370 case SAVEt_ITEM: /* normal string */
12371 case SAVEt_GVSV: /* scalar slot in GV */
12372 case SAVEt_SV: /* scalar reference */
12373 sv = (const SV *)POPPTR(ss,ix);
12374 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12377 case SAVEt_MORTALIZESV:
12378 sv = (const SV *)POPPTR(ss,ix);
12379 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12381 case SAVEt_SHARED_PVREF: /* char* in shared space */
12382 c = (char*)POPPTR(ss,ix);
12383 TOPPTR(nss,ix) = savesharedpv(c);
12384 ptr = POPPTR(ss,ix);
12385 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12387 case SAVEt_GENERIC_SVREF: /* generic sv */
12388 case SAVEt_SVREF: /* scalar reference */
12389 sv = (const SV *)POPPTR(ss,ix);
12390 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12391 ptr = POPPTR(ss,ix);
12392 TOPPTR(nss,ix) = svp_dup_inc((SV**)ptr, proto_perl);/* XXXXX */
12394 case SAVEt_HV: /* hash reference */
12395 case SAVEt_AV: /* array reference */
12396 sv = (const SV *) POPPTR(ss,ix);
12397 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12399 case SAVEt_COMPPAD:
12401 sv = (const SV *) POPPTR(ss,ix);
12402 TOPPTR(nss,ix) = sv_dup(sv, param);
12404 case SAVEt_INT: /* int reference */
12405 ptr = POPPTR(ss,ix);
12406 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12407 intval = (int)POPINT(ss,ix);
12408 TOPINT(nss,ix) = intval;
12410 case SAVEt_LONG: /* long reference */
12411 ptr = POPPTR(ss,ix);
12412 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12413 longval = (long)POPLONG(ss,ix);
12414 TOPLONG(nss,ix) = longval;
12416 case SAVEt_I32: /* I32 reference */
12417 ptr = POPPTR(ss,ix);
12418 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12420 TOPINT(nss,ix) = i;
12422 case SAVEt_IV: /* IV reference */
12423 ptr = POPPTR(ss,ix);
12424 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12426 TOPIV(nss,ix) = iv;
12428 case SAVEt_HPTR: /* HV* reference */
12429 case SAVEt_APTR: /* AV* reference */
12430 case SAVEt_SPTR: /* SV* reference */
12431 ptr = POPPTR(ss,ix);
12432 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12433 sv = (const SV *)POPPTR(ss,ix);
12434 TOPPTR(nss,ix) = sv_dup(sv, param);
12436 case SAVEt_VPTR: /* random* reference */
12437 ptr = POPPTR(ss,ix);
12438 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12440 case SAVEt_INT_SMALL:
12441 case SAVEt_I32_SMALL:
12442 case SAVEt_I16: /* I16 reference */
12443 case SAVEt_I8: /* I8 reference */
12445 ptr = POPPTR(ss,ix);
12446 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12448 case SAVEt_GENERIC_PVREF: /* generic char* */
12449 case SAVEt_PPTR: /* char* reference */
12450 ptr = POPPTR(ss,ix);
12451 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12452 c = (char*)POPPTR(ss,ix);
12453 TOPPTR(nss,ix) = pv_dup(c);
12455 case SAVEt_GP: /* scalar reference */
12456 gp = (GP*)POPPTR(ss,ix);
12457 TOPPTR(nss,ix) = gp = gp_dup(gp, param);
12458 (void)GpREFCNT_inc(gp);
12459 gv = (const GV *)POPPTR(ss,ix);
12460 TOPPTR(nss,ix) = gv_dup_inc(gv, param);
12463 ptr = POPPTR(ss,ix);
12464 if (ptr && (((OP*)ptr)->op_private & OPpREFCOUNTED)) {
12465 /* these are assumed to be refcounted properly */
12467 switch (((OP*)ptr)->op_type) {
12469 case OP_LEAVESUBLV:
12473 case OP_LEAVEWRITE:
12474 TOPPTR(nss,ix) = ptr;
12477 (void) OpREFCNT_inc(o);
12481 TOPPTR(nss,ix) = NULL;
12486 TOPPTR(nss,ix) = NULL;
12488 case SAVEt_FREECOPHH:
12489 ptr = POPPTR(ss,ix);
12490 TOPPTR(nss,ix) = cophh_copy((COPHH *)ptr);
12493 hv = (const HV *)POPPTR(ss,ix);
12494 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
12496 TOPINT(nss,ix) = i;
12499 c = (char*)POPPTR(ss,ix);
12500 TOPPTR(nss,ix) = pv_dup_inc(c);
12502 case SAVEt_STACK_POS: /* Position on Perl stack */
12504 TOPINT(nss,ix) = i;
12506 case SAVEt_DESTRUCTOR:
12507 ptr = POPPTR(ss,ix);
12508 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
12509 dptr = POPDPTR(ss,ix);
12510 TOPDPTR(nss,ix) = DPTR2FPTR(void (*)(void*),
12511 any_dup(FPTR2DPTR(void *, dptr),
12514 case SAVEt_DESTRUCTOR_X:
12515 ptr = POPPTR(ss,ix);
12516 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
12517 dxptr = POPDXPTR(ss,ix);
12518 TOPDXPTR(nss,ix) = DPTR2FPTR(void (*)(pTHX_ void*),
12519 any_dup(FPTR2DPTR(void *, dxptr),
12522 case SAVEt_REGCONTEXT:
12524 ix -= uv >> SAVE_TIGHT_SHIFT;
12526 case SAVEt_AELEM: /* array element */
12527 sv = (const SV *)POPPTR(ss,ix);
12528 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12530 TOPINT(nss,ix) = i;
12531 av = (const AV *)POPPTR(ss,ix);
12532 TOPPTR(nss,ix) = av_dup_inc(av, param);
12535 ptr = POPPTR(ss,ix);
12536 TOPPTR(nss,ix) = ptr;
12539 ptr = POPPTR(ss,ix);
12540 ptr = cophh_copy((COPHH*)ptr);
12541 TOPPTR(nss,ix) = ptr;
12543 TOPINT(nss,ix) = i;
12544 if (i & HINT_LOCALIZE_HH) {
12545 hv = (const HV *)POPPTR(ss,ix);
12546 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
12549 case SAVEt_PADSV_AND_MORTALIZE:
12550 longval = (long)POPLONG(ss,ix);
12551 TOPLONG(nss,ix) = longval;
12552 ptr = POPPTR(ss,ix);
12553 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12554 sv = (const SV *)POPPTR(ss,ix);
12555 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12557 case SAVEt_SET_SVFLAGS:
12559 TOPINT(nss,ix) = i;
12561 TOPINT(nss,ix) = i;
12562 sv = (const SV *)POPPTR(ss,ix);
12563 TOPPTR(nss,ix) = sv_dup(sv, param);
12565 case SAVEt_RE_STATE:
12567 const struct re_save_state *const old_state
12568 = (struct re_save_state *)
12569 (ss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
12570 struct re_save_state *const new_state
12571 = (struct re_save_state *)
12572 (nss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
12574 Copy(old_state, new_state, 1, struct re_save_state);
12575 ix -= SAVESTACK_ALLOC_FOR_RE_SAVE_STATE;
12577 new_state->re_state_bostr
12578 = pv_dup(old_state->re_state_bostr);
12579 new_state->re_state_reginput
12580 = pv_dup(old_state->re_state_reginput);
12581 new_state->re_state_regeol
12582 = pv_dup(old_state->re_state_regeol);
12583 new_state->re_state_regoffs
12584 = (regexp_paren_pair*)
12585 any_dup(old_state->re_state_regoffs, proto_perl);
12586 new_state->re_state_reglastparen
12587 = (U32*) any_dup(old_state->re_state_reglastparen,
12589 new_state->re_state_reglastcloseparen
12590 = (U32*)any_dup(old_state->re_state_reglastcloseparen,
12592 /* XXX This just has to be broken. The old save_re_context
12593 code did SAVEGENERICPV(PL_reg_start_tmp);
12594 PL_reg_start_tmp is char **.
12595 Look above to what the dup code does for
12596 SAVEt_GENERIC_PVREF
12597 It can never have worked.
12598 So this is merely a faithful copy of the exiting bug: */
12599 new_state->re_state_reg_start_tmp
12600 = (char **) pv_dup((char *)
12601 old_state->re_state_reg_start_tmp);
12602 /* I assume that it only ever "worked" because no-one called
12603 (pseudo)fork while the regexp engine had re-entered itself.
12605 #ifdef PERL_OLD_COPY_ON_WRITE
12606 new_state->re_state_nrs
12607 = sv_dup(old_state->re_state_nrs, param);
12609 new_state->re_state_reg_magic
12610 = (MAGIC*) any_dup(old_state->re_state_reg_magic,
12612 new_state->re_state_reg_oldcurpm
12613 = (PMOP*) any_dup(old_state->re_state_reg_oldcurpm,
12615 new_state->re_state_reg_curpm
12616 = (PMOP*) any_dup(old_state->re_state_reg_curpm,
12618 new_state->re_state_reg_oldsaved
12619 = pv_dup(old_state->re_state_reg_oldsaved);
12620 new_state->re_state_reg_poscache
12621 = pv_dup(old_state->re_state_reg_poscache);
12622 new_state->re_state_reg_starttry
12623 = pv_dup(old_state->re_state_reg_starttry);
12626 case SAVEt_COMPILE_WARNINGS:
12627 ptr = POPPTR(ss,ix);
12628 TOPPTR(nss,ix) = DUP_WARNINGS((STRLEN*)ptr);
12631 ptr = POPPTR(ss,ix);
12632 TOPPTR(nss,ix) = parser_dup((const yy_parser*)ptr, param);
12636 "panic: ss_dup inconsistency (%"IVdf")", (IV) type);
12644 /* if sv is a stash, call $class->CLONE_SKIP(), and set the SVphv_CLONEABLE
12645 * flag to the result. This is done for each stash before cloning starts,
12646 * so we know which stashes want their objects cloned */
12649 do_mark_cloneable_stash(pTHX_ SV *const sv)
12651 const HEK * const hvname = HvNAME_HEK((const HV *)sv);
12653 GV* const cloner = gv_fetchmethod_autoload(MUTABLE_HV(sv), "CLONE_SKIP", 0);
12654 SvFLAGS(sv) |= SVphv_CLONEABLE; /* clone objects by default */
12655 if (cloner && GvCV(cloner)) {
12662 mXPUSHs(newSVhek(hvname));
12664 call_sv(MUTABLE_SV(GvCV(cloner)), G_SCALAR);
12671 SvFLAGS(sv) &= ~SVphv_CLONEABLE;
12679 =for apidoc perl_clone
12681 Create and return a new interpreter by cloning the current one.
12683 perl_clone takes these flags as parameters:
12685 CLONEf_COPY_STACKS - is used to, well, copy the stacks also,
12686 without it we only clone the data and zero the stacks,
12687 with it we copy the stacks and the new perl interpreter is
12688 ready to run at the exact same point as the previous one.
12689 The pseudo-fork code uses COPY_STACKS while the
12690 threads->create doesn't.
12692 CLONEf_KEEP_PTR_TABLE
12693 perl_clone keeps a ptr_table with the pointer of the old
12694 variable as a key and the new variable as a value,
12695 this allows it to check if something has been cloned and not
12696 clone it again but rather just use the value and increase the
12697 refcount. If KEEP_PTR_TABLE is not set then perl_clone will kill
12698 the ptr_table using the function
12699 C<ptr_table_free(PL_ptr_table); PL_ptr_table = NULL;>,
12700 reason to keep it around is if you want to dup some of your own
12701 variable who are outside the graph perl scans, example of this
12702 code is in threads.xs create
12705 This is a win32 thing, it is ignored on unix, it tells perls
12706 win32host code (which is c++) to clone itself, this is needed on
12707 win32 if you want to run two threads at the same time,
12708 if you just want to do some stuff in a separate perl interpreter
12709 and then throw it away and return to the original one,
12710 you don't need to do anything.
12715 /* XXX the above needs expanding by someone who actually understands it ! */
12716 EXTERN_C PerlInterpreter *
12717 perl_clone_host(PerlInterpreter* proto_perl, UV flags);
12720 perl_clone(PerlInterpreter *proto_perl, UV flags)
12723 #ifdef PERL_IMPLICIT_SYS
12725 PERL_ARGS_ASSERT_PERL_CLONE;
12727 /* perlhost.h so we need to call into it
12728 to clone the host, CPerlHost should have a c interface, sky */
12730 if (flags & CLONEf_CLONE_HOST) {
12731 return perl_clone_host(proto_perl,flags);
12733 return perl_clone_using(proto_perl, flags,
12735 proto_perl->IMemShared,
12736 proto_perl->IMemParse,
12738 proto_perl->IStdIO,
12742 proto_perl->IProc);
12746 perl_clone_using(PerlInterpreter *proto_perl, UV flags,
12747 struct IPerlMem* ipM, struct IPerlMem* ipMS,
12748 struct IPerlMem* ipMP, struct IPerlEnv* ipE,
12749 struct IPerlStdIO* ipStd, struct IPerlLIO* ipLIO,
12750 struct IPerlDir* ipD, struct IPerlSock* ipS,
12751 struct IPerlProc* ipP)
12753 /* XXX many of the string copies here can be optimized if they're
12754 * constants; they need to be allocated as common memory and just
12755 * their pointers copied. */
12758 CLONE_PARAMS clone_params;
12759 CLONE_PARAMS* const param = &clone_params;
12761 PerlInterpreter * const my_perl = (PerlInterpreter*)(*ipM->pMalloc)(ipM, sizeof(PerlInterpreter));
12763 PERL_ARGS_ASSERT_PERL_CLONE_USING;
12764 #else /* !PERL_IMPLICIT_SYS */
12766 CLONE_PARAMS clone_params;
12767 CLONE_PARAMS* param = &clone_params;
12768 PerlInterpreter * const my_perl = (PerlInterpreter*)PerlMem_malloc(sizeof(PerlInterpreter));
12770 PERL_ARGS_ASSERT_PERL_CLONE;
12771 #endif /* PERL_IMPLICIT_SYS */
12773 /* for each stash, determine whether its objects should be cloned */
12774 S_visit(proto_perl, do_mark_cloneable_stash, SVt_PVHV, SVTYPEMASK);
12775 PERL_SET_THX(my_perl);
12778 PoisonNew(my_perl, 1, PerlInterpreter);
12781 PL_defstash = NULL; /* may be used by perl malloc() */
12784 PL_scopestack_name = 0;
12786 PL_savestack_ix = 0;
12787 PL_savestack_max = -1;
12788 PL_sig_pending = 0;
12790 Zero(&PL_debug_pad, 1, struct perl_debug_pad);
12791 # ifdef DEBUG_LEAKING_SCALARS
12792 PL_sv_serial = (((UV)my_perl >> 2) & 0xfff) * 1000000;
12794 #else /* !DEBUGGING */
12795 Zero(my_perl, 1, PerlInterpreter);
12796 #endif /* DEBUGGING */
12798 #ifdef PERL_IMPLICIT_SYS
12799 /* host pointers */
12801 PL_MemShared = ipMS;
12802 PL_MemParse = ipMP;
12809 #endif /* PERL_IMPLICIT_SYS */
12811 param->flags = flags;
12812 /* Nothing in the core code uses this, but we make it available to
12813 extensions (using mg_dup). */
12814 param->proto_perl = proto_perl;
12815 /* Likely nothing will use this, but it is initialised to be consistent
12816 with Perl_clone_params_new(). */
12817 param->new_perl = my_perl;
12818 param->unreferenced = NULL;
12820 INIT_TRACK_MEMPOOL(my_perl->Imemory_debug_header, my_perl);
12822 PL_body_arenas = NULL;
12823 Zero(&PL_body_roots, 1, PL_body_roots);
12826 PL_sv_objcount = 0;
12828 PL_sv_arenaroot = NULL;
12830 PL_debug = proto_perl->Idebug;
12832 PL_hash_seed = proto_perl->Ihash_seed;
12833 PL_rehash_seed = proto_perl->Irehash_seed;
12835 SvANY(&PL_sv_undef) = NULL;
12836 SvREFCNT(&PL_sv_undef) = (~(U32)0)/2;
12837 SvFLAGS(&PL_sv_undef) = SVf_READONLY|SVt_NULL;
12838 SvREFCNT(&PL_sv_no) = (~(U32)0)/2;
12839 SvFLAGS(&PL_sv_no) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
12840 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
12842 SvANY(&PL_sv_yes) = new_XPVNV();
12843 SvREFCNT(&PL_sv_yes) = (~(U32)0)/2;
12844 SvFLAGS(&PL_sv_yes) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
12845 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
12847 /* dbargs array probably holds garbage */
12850 PL_compiling = proto_perl->Icompiling;
12852 #ifdef PERL_DEBUG_READONLY_OPS
12857 /* pseudo environmental stuff */
12858 PL_origargc = proto_perl->Iorigargc;
12859 PL_origargv = proto_perl->Iorigargv;
12861 /* Set tainting stuff before PerlIO_debug can possibly get called */
12862 PL_tainting = proto_perl->Itainting;
12863 PL_taint_warn = proto_perl->Itaint_warn;
12865 PL_minus_c = proto_perl->Iminus_c;
12867 PL_localpatches = proto_perl->Ilocalpatches;
12868 PL_splitstr = proto_perl->Isplitstr;
12869 PL_minus_n = proto_perl->Iminus_n;
12870 PL_minus_p = proto_perl->Iminus_p;
12871 PL_minus_l = proto_perl->Iminus_l;
12872 PL_minus_a = proto_perl->Iminus_a;
12873 PL_minus_E = proto_perl->Iminus_E;
12874 PL_minus_F = proto_perl->Iminus_F;
12875 PL_doswitches = proto_perl->Idoswitches;
12876 PL_dowarn = proto_perl->Idowarn;
12877 PL_sawampersand = proto_perl->Isawampersand;
12878 PL_unsafe = proto_perl->Iunsafe;
12879 PL_perldb = proto_perl->Iperldb;
12880 PL_perl_destruct_level = proto_perl->Iperl_destruct_level;
12881 PL_exit_flags = proto_perl->Iexit_flags;
12883 /* XXX time(&PL_basetime) when asked for? */
12884 PL_basetime = proto_perl->Ibasetime;
12886 PL_maxsysfd = proto_perl->Imaxsysfd;
12887 PL_statusvalue = proto_perl->Istatusvalue;
12889 PL_statusvalue_vms = proto_perl->Istatusvalue_vms;
12891 PL_statusvalue_posix = proto_perl->Istatusvalue_posix;
12894 /* RE engine related */
12895 Zero(&PL_reg_state, 1, struct re_save_state);
12896 PL_reginterp_cnt = 0;
12897 PL_regmatch_slab = NULL;
12899 PL_sub_generation = proto_perl->Isub_generation;
12901 /* funky return mechanisms */
12902 PL_forkprocess = proto_perl->Iforkprocess;
12904 /* internal state */
12905 PL_maxo = proto_perl->Imaxo;
12907 PL_main_start = proto_perl->Imain_start;
12908 PL_eval_root = proto_perl->Ieval_root;
12909 PL_eval_start = proto_perl->Ieval_start;
12911 PL_filemode = proto_perl->Ifilemode;
12912 PL_lastfd = proto_perl->Ilastfd;
12913 PL_oldname = proto_perl->Ioldname; /* XXX not quite right */
12916 PL_gensym = proto_perl->Igensym;
12918 PL_laststatval = proto_perl->Ilaststatval;
12919 PL_laststype = proto_perl->Ilaststype;
12922 PL_profiledata = NULL;
12924 PL_generation = proto_perl->Igeneration;
12926 PL_in_clean_objs = proto_perl->Iin_clean_objs;
12927 PL_in_clean_all = proto_perl->Iin_clean_all;
12929 PL_uid = proto_perl->Iuid;
12930 PL_euid = proto_perl->Ieuid;
12931 PL_gid = proto_perl->Igid;
12932 PL_egid = proto_perl->Iegid;
12933 PL_nomemok = proto_perl->Inomemok;
12934 PL_an = proto_perl->Ian;
12935 PL_evalseq = proto_perl->Ievalseq;
12936 PL_origenviron = proto_perl->Iorigenviron; /* XXX not quite right */
12937 PL_origalen = proto_perl->Iorigalen;
12939 PL_sighandlerp = proto_perl->Isighandlerp;
12941 PL_runops = proto_perl->Irunops;
12943 PL_subline = proto_perl->Isubline;
12946 PL_cryptseen = proto_perl->Icryptseen;
12949 PL_hints = proto_perl->Ihints;
12951 PL_amagic_generation = proto_perl->Iamagic_generation;
12953 #ifdef USE_LOCALE_COLLATE
12954 PL_collation_ix = proto_perl->Icollation_ix;
12955 PL_collation_standard = proto_perl->Icollation_standard;
12956 PL_collxfrm_base = proto_perl->Icollxfrm_base;
12957 PL_collxfrm_mult = proto_perl->Icollxfrm_mult;
12958 #endif /* USE_LOCALE_COLLATE */
12960 #ifdef USE_LOCALE_NUMERIC
12961 PL_numeric_standard = proto_perl->Inumeric_standard;
12962 PL_numeric_local = proto_perl->Inumeric_local;
12963 #endif /* !USE_LOCALE_NUMERIC */
12965 /* Did the locale setup indicate UTF-8? */
12966 PL_utf8locale = proto_perl->Iutf8locale;
12967 /* Unicode features (see perlrun/-C) */
12968 PL_unicode = proto_perl->Iunicode;
12970 /* Pre-5.8 signals control */
12971 PL_signals = proto_perl->Isignals;
12973 /* times() ticks per second */
12974 PL_clocktick = proto_perl->Iclocktick;
12976 /* Recursion stopper for PerlIO_find_layer */
12977 PL_in_load_module = proto_perl->Iin_load_module;
12979 /* sort() routine */
12980 PL_sort_RealCmp = proto_perl->Isort_RealCmp;
12982 /* Not really needed/useful since the reenrant_retint is "volatile",
12983 * but do it for consistency's sake. */
12984 PL_reentrant_retint = proto_perl->Ireentrant_retint;
12986 /* Hooks to shared SVs and locks. */
12987 PL_sharehook = proto_perl->Isharehook;
12988 PL_lockhook = proto_perl->Ilockhook;
12989 PL_unlockhook = proto_perl->Iunlockhook;
12990 PL_threadhook = proto_perl->Ithreadhook;
12991 PL_destroyhook = proto_perl->Idestroyhook;
12992 PL_signalhook = proto_perl->Isignalhook;
12994 PL_globhook = proto_perl->Iglobhook;
12996 #ifdef THREADS_HAVE_PIDS
12997 PL_ppid = proto_perl->Ippid;
13001 PL_last_swash_hv = NULL; /* reinits on demand */
13002 PL_last_swash_klen = 0;
13003 PL_last_swash_key[0]= '\0';
13004 PL_last_swash_tmps = (U8*)NULL;
13005 PL_last_swash_slen = 0;
13007 PL_glob_index = proto_perl->Iglob_index;
13008 PL_srand_called = proto_perl->Isrand_called;
13010 if (flags & CLONEf_COPY_STACKS) {
13011 /* next allocation will be PL_tmps_stack[PL_tmps_ix+1] */
13012 PL_tmps_ix = proto_perl->Itmps_ix;
13013 PL_tmps_max = proto_perl->Itmps_max;
13014 PL_tmps_floor = proto_perl->Itmps_floor;
13016 /* next push_scope()/ENTER sets PL_scopestack[PL_scopestack_ix]
13017 * NOTE: unlike the others! */
13018 PL_scopestack_ix = proto_perl->Iscopestack_ix;
13019 PL_scopestack_max = proto_perl->Iscopestack_max;
13021 /* next SSPUSHFOO() sets PL_savestack[PL_savestack_ix]
13022 * NOTE: unlike the others! */
13023 PL_savestack_ix = proto_perl->Isavestack_ix;
13024 PL_savestack_max = proto_perl->Isavestack_max;
13027 PL_start_env = proto_perl->Istart_env; /* XXXXXX */
13028 PL_top_env = &PL_start_env;
13030 PL_op = proto_perl->Iop;
13033 PL_Xpv = (XPV*)NULL;
13034 my_perl->Ina = proto_perl->Ina;
13036 PL_statbuf = proto_perl->Istatbuf;
13037 PL_statcache = proto_perl->Istatcache;
13040 PL_timesbuf = proto_perl->Itimesbuf;
13043 PL_tainted = proto_perl->Itainted;
13044 PL_curpm = proto_perl->Icurpm; /* XXX No PMOP ref count */
13046 PL_chopset = proto_perl->Ichopset; /* XXX never deallocated */
13048 PL_restartjmpenv = proto_perl->Irestartjmpenv;
13049 PL_restartop = proto_perl->Irestartop;
13050 PL_in_eval = proto_perl->Iin_eval;
13051 PL_delaymagic = proto_perl->Idelaymagic;
13052 PL_phase = proto_perl->Iphase;
13053 PL_localizing = proto_perl->Ilocalizing;
13055 PL_hv_fetch_ent_mh = NULL;
13056 PL_modcount = proto_perl->Imodcount;
13057 PL_lastgotoprobe = NULL;
13058 PL_dumpindent = proto_perl->Idumpindent;
13060 PL_efloatbuf = NULL; /* reinits on demand */
13061 PL_efloatsize = 0; /* reinits on demand */
13065 PL_regdummy = proto_perl->Iregdummy;
13066 PL_colorset = 0; /* reinits PL_colors[] */
13067 /*PL_colors[6] = {0,0,0,0,0,0};*/
13069 /* Pluggable optimizer */
13070 PL_peepp = proto_perl->Ipeepp;
13071 PL_rpeepp = proto_perl->Irpeepp;
13072 /* op_free() hook */
13073 PL_opfreehook = proto_perl->Iopfreehook;
13075 #ifdef USE_REENTRANT_API
13076 /* XXX: things like -Dm will segfault here in perlio, but doing
13077 * PERL_SET_CONTEXT(proto_perl);
13078 * breaks too many other things
13080 Perl_reentrant_init(aTHX);
13083 /* create SV map for pointer relocation */
13084 PL_ptr_table = ptr_table_new();
13086 /* initialize these special pointers as early as possible */
13087 ptr_table_store(PL_ptr_table, &proto_perl->Isv_undef, &PL_sv_undef);
13089 SvANY(&PL_sv_no) = new_XPVNV();
13090 SvPV_set(&PL_sv_no, savepvn(PL_No, 0));
13091 SvCUR_set(&PL_sv_no, 0);
13092 SvLEN_set(&PL_sv_no, 1);
13093 SvIV_set(&PL_sv_no, 0);
13094 SvNV_set(&PL_sv_no, 0);
13095 ptr_table_store(PL_ptr_table, &proto_perl->Isv_no, &PL_sv_no);
13097 SvPV_set(&PL_sv_yes, savepvn(PL_Yes, 1));
13098 SvCUR_set(&PL_sv_yes, 1);
13099 SvLEN_set(&PL_sv_yes, 2);
13100 SvIV_set(&PL_sv_yes, 1);
13101 SvNV_set(&PL_sv_yes, 1);
13102 ptr_table_store(PL_ptr_table, &proto_perl->Isv_yes, &PL_sv_yes);
13104 /* create (a non-shared!) shared string table */
13105 PL_strtab = newHV();
13106 HvSHAREKEYS_off(PL_strtab);
13107 hv_ksplit(PL_strtab, HvTOTALKEYS(proto_perl->Istrtab));
13108 ptr_table_store(PL_ptr_table, proto_perl->Istrtab, PL_strtab);
13110 /* These two PVs will be free'd special way so must set them same way op.c does */
13111 PL_compiling.cop_stashpv = savesharedpv(PL_compiling.cop_stashpv);
13112 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_stashpv, PL_compiling.cop_stashpv);
13114 PL_compiling.cop_file = savesharedpv(PL_compiling.cop_file);
13115 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_file, PL_compiling.cop_file);
13117 ptr_table_store(PL_ptr_table, &proto_perl->Icompiling, &PL_compiling);
13118 PL_compiling.cop_warnings = DUP_WARNINGS(PL_compiling.cop_warnings);
13119 CopHINTHASH_set(&PL_compiling, cophh_copy(CopHINTHASH_get(&PL_compiling)));
13120 PL_curcop = (COP*)any_dup(proto_perl->Icurcop, proto_perl);
13122 param->stashes = newAV(); /* Setup array of objects to call clone on */
13123 /* This makes no difference to the implementation, as it always pushes
13124 and shifts pointers to other SVs without changing their reference
13125 count, with the array becoming empty before it is freed. However, it
13126 makes it conceptually clear what is going on, and will avoid some
13127 work inside av.c, filling slots between AvFILL() and AvMAX() with
13128 &PL_sv_undef, and SvREFCNT_dec()ing those. */
13129 AvREAL_off(param->stashes);
13131 if (!(flags & CLONEf_COPY_STACKS)) {
13132 param->unreferenced = newAV();
13135 #ifdef PERLIO_LAYERS
13136 /* Clone PerlIO tables as soon as we can handle general xx_dup() */
13137 PerlIO_clone(aTHX_ proto_perl, param);
13140 PL_envgv = gv_dup(proto_perl->Ienvgv, param);
13141 PL_incgv = gv_dup(proto_perl->Iincgv, param);
13142 PL_hintgv = gv_dup(proto_perl->Ihintgv, param);
13143 PL_origfilename = SAVEPV(proto_perl->Iorigfilename);
13144 PL_diehook = sv_dup_inc(proto_perl->Idiehook, param);
13145 PL_warnhook = sv_dup_inc(proto_perl->Iwarnhook, param);
13148 PL_patchlevel = sv_dup_inc(proto_perl->Ipatchlevel, param);
13149 PL_apiversion = sv_dup_inc(proto_perl->Iapiversion, param);
13150 PL_inplace = SAVEPV(proto_perl->Iinplace);
13151 PL_e_script = sv_dup_inc(proto_perl->Ie_script, param);
13153 /* magical thingies */
13154 PL_formfeed = sv_dup(proto_perl->Iformfeed, param);
13156 PL_encoding = sv_dup(proto_perl->Iencoding, param);
13158 sv_setpvs(PERL_DEBUG_PAD(0), ""); /* For regex debugging. */
13159 sv_setpvs(PERL_DEBUG_PAD(1), ""); /* ext/re needs these */
13160 sv_setpvs(PERL_DEBUG_PAD(2), ""); /* even without DEBUGGING. */
13163 /* Clone the regex array */
13164 /* ORANGE FIXME for plugins, probably in the SV dup code.
13165 newSViv(PTR2IV(CALLREGDUPE(
13166 INT2PTR(REGEXP *, SvIVX(regex)), param))))
13168 PL_regex_padav = av_dup_inc(proto_perl->Iregex_padav, param);
13169 PL_regex_pad = AvARRAY(PL_regex_padav);
13171 /* shortcuts to various I/O objects */
13172 PL_ofsgv = gv_dup_inc(proto_perl->Iofsgv, param);
13173 PL_stdingv = gv_dup(proto_perl->Istdingv, param);
13174 PL_stderrgv = gv_dup(proto_perl->Istderrgv, param);
13175 PL_defgv = gv_dup(proto_perl->Idefgv, param);
13176 PL_argvgv = gv_dup(proto_perl->Iargvgv, param);
13177 PL_argvoutgv = gv_dup(proto_perl->Iargvoutgv, param);
13178 PL_argvout_stack = av_dup_inc(proto_perl->Iargvout_stack, param);
13180 /* shortcuts to regexp stuff */
13181 PL_replgv = gv_dup(proto_perl->Ireplgv, param);
13183 /* shortcuts to misc objects */
13184 PL_errgv = gv_dup(proto_perl->Ierrgv, param);
13186 /* shortcuts to debugging objects */
13187 PL_DBgv = gv_dup(proto_perl->IDBgv, param);
13188 PL_DBline = gv_dup(proto_perl->IDBline, param);
13189 PL_DBsub = gv_dup(proto_perl->IDBsub, param);
13190 PL_DBsingle = sv_dup(proto_perl->IDBsingle, param);
13191 PL_DBtrace = sv_dup(proto_perl->IDBtrace, param);
13192 PL_DBsignal = sv_dup(proto_perl->IDBsignal, param);
13194 /* symbol tables */
13195 PL_defstash = hv_dup_inc(proto_perl->Idefstash, param);
13196 PL_curstash = hv_dup_inc(proto_perl->Icurstash, param);
13197 PL_debstash = hv_dup(proto_perl->Idebstash, param);
13198 PL_globalstash = hv_dup(proto_perl->Iglobalstash, param);
13199 PL_curstname = sv_dup_inc(proto_perl->Icurstname, param);
13201 PL_beginav = av_dup_inc(proto_perl->Ibeginav, param);
13202 PL_beginav_save = av_dup_inc(proto_perl->Ibeginav_save, param);
13203 PL_checkav_save = av_dup_inc(proto_perl->Icheckav_save, param);
13204 PL_unitcheckav = av_dup_inc(proto_perl->Iunitcheckav, param);
13205 PL_unitcheckav_save = av_dup_inc(proto_perl->Iunitcheckav_save, param);
13206 PL_endav = av_dup_inc(proto_perl->Iendav, param);
13207 PL_checkav = av_dup_inc(proto_perl->Icheckav, param);
13208 PL_initav = av_dup_inc(proto_perl->Iinitav, param);
13210 PL_isarev = hv_dup_inc(proto_perl->Iisarev, param);
13212 /* subprocess state */
13213 PL_fdpid = av_dup_inc(proto_perl->Ifdpid, param);
13215 if (proto_perl->Iop_mask)
13216 PL_op_mask = SAVEPVN(proto_perl->Iop_mask, PL_maxo);
13219 /* PL_asserting = proto_perl->Iasserting; */
13221 /* current interpreter roots */
13222 PL_main_cv = cv_dup_inc(proto_perl->Imain_cv, param);
13224 PL_main_root = OpREFCNT_inc(proto_perl->Imain_root);
13227 /* runtime control stuff */
13228 PL_curcopdb = (COP*)any_dup(proto_perl->Icurcopdb, proto_perl);
13230 PL_preambleav = av_dup_inc(proto_perl->Ipreambleav, param);
13232 PL_ors_sv = sv_dup_inc(proto_perl->Iors_sv, param);
13234 /* interpreter atexit processing */
13235 PL_exitlistlen = proto_perl->Iexitlistlen;
13236 if (PL_exitlistlen) {
13237 Newx(PL_exitlist, PL_exitlistlen, PerlExitListEntry);
13238 Copy(proto_perl->Iexitlist, PL_exitlist, PL_exitlistlen, PerlExitListEntry);
13241 PL_exitlist = (PerlExitListEntry*)NULL;
13243 PL_my_cxt_size = proto_perl->Imy_cxt_size;
13244 if (PL_my_cxt_size) {
13245 Newx(PL_my_cxt_list, PL_my_cxt_size, void *);
13246 Copy(proto_perl->Imy_cxt_list, PL_my_cxt_list, PL_my_cxt_size, void *);
13247 #ifdef PERL_GLOBAL_STRUCT_PRIVATE
13248 Newx(PL_my_cxt_keys, PL_my_cxt_size, const char *);
13249 Copy(proto_perl->Imy_cxt_keys, PL_my_cxt_keys, PL_my_cxt_size, char *);
13253 PL_my_cxt_list = (void**)NULL;
13254 #ifdef PERL_GLOBAL_STRUCT_PRIVATE
13255 PL_my_cxt_keys = (const char**)NULL;
13258 PL_modglobal = hv_dup_inc(proto_perl->Imodglobal, param);
13259 PL_custom_op_names = hv_dup_inc(proto_perl->Icustom_op_names,param);
13260 PL_custom_op_descs = hv_dup_inc(proto_perl->Icustom_op_descs,param);
13261 PL_custom_ops = hv_dup_inc(proto_perl->Icustom_ops, param);
13263 PL_compcv = cv_dup(proto_perl->Icompcv, param);
13265 PAD_CLONE_VARS(proto_perl, param);
13267 #ifdef HAVE_INTERP_INTERN
13268 sys_intern_dup(&proto_perl->Isys_intern, &PL_sys_intern);
13271 PL_DBcv = cv_dup(proto_perl->IDBcv, param);
13273 #ifdef PERL_USES_PL_PIDSTATUS
13274 PL_pidstatus = newHV(); /* XXX flag for cloning? */
13276 PL_osname = SAVEPV(proto_perl->Iosname);
13277 PL_parser = parser_dup(proto_perl->Iparser, param);
13279 /* XXX this only works if the saved cop has already been cloned */
13280 if (proto_perl->Iparser) {
13281 PL_parser->saved_curcop = (COP*)any_dup(
13282 proto_perl->Iparser->saved_curcop,
13286 PL_subname = sv_dup_inc(proto_perl->Isubname, param);
13288 #ifdef USE_LOCALE_COLLATE
13289 PL_collation_name = SAVEPV(proto_perl->Icollation_name);
13290 #endif /* USE_LOCALE_COLLATE */
13292 #ifdef USE_LOCALE_NUMERIC
13293 PL_numeric_name = SAVEPV(proto_perl->Inumeric_name);
13294 PL_numeric_radix_sv = sv_dup_inc(proto_perl->Inumeric_radix_sv, param);
13295 #endif /* !USE_LOCALE_NUMERIC */
13297 /* utf8 character classes */
13298 PL_utf8_alnum = sv_dup_inc(proto_perl->Iutf8_alnum, param);
13299 PL_utf8_alpha = sv_dup_inc(proto_perl->Iutf8_alpha, param);
13300 PL_utf8_space = sv_dup_inc(proto_perl->Iutf8_space, param);
13301 PL_utf8_graph = sv_dup_inc(proto_perl->Iutf8_graph, param);
13302 PL_utf8_digit = sv_dup_inc(proto_perl->Iutf8_digit, param);
13303 PL_utf8_upper = sv_dup_inc(proto_perl->Iutf8_upper, param);
13304 PL_utf8_lower = sv_dup_inc(proto_perl->Iutf8_lower, param);
13305 PL_utf8_print = sv_dup_inc(proto_perl->Iutf8_print, param);
13306 PL_utf8_punct = sv_dup_inc(proto_perl->Iutf8_punct, param);
13307 PL_utf8_xdigit = sv_dup_inc(proto_perl->Iutf8_xdigit, param);
13308 PL_utf8_mark = sv_dup_inc(proto_perl->Iutf8_mark, param);
13309 PL_utf8_X_begin = sv_dup_inc(proto_perl->Iutf8_X_begin, param);
13310 PL_utf8_X_extend = sv_dup_inc(proto_perl->Iutf8_X_extend, param);
13311 PL_utf8_X_prepend = sv_dup_inc(proto_perl->Iutf8_X_prepend, param);
13312 PL_utf8_X_non_hangul = sv_dup_inc(proto_perl->Iutf8_X_non_hangul, param);
13313 PL_utf8_X_L = sv_dup_inc(proto_perl->Iutf8_X_L, param);
13314 PL_utf8_X_LV = sv_dup_inc(proto_perl->Iutf8_X_LV, param);
13315 PL_utf8_X_LVT = sv_dup_inc(proto_perl->Iutf8_X_LVT, param);
13316 PL_utf8_X_T = sv_dup_inc(proto_perl->Iutf8_X_T, param);
13317 PL_utf8_X_V = sv_dup_inc(proto_perl->Iutf8_X_V, param);
13318 PL_utf8_X_LV_LVT_V = sv_dup_inc(proto_perl->Iutf8_X_LV_LVT_V, param);
13319 PL_utf8_toupper = sv_dup_inc(proto_perl->Iutf8_toupper, param);
13320 PL_utf8_totitle = sv_dup_inc(proto_perl->Iutf8_totitle, param);
13321 PL_utf8_tolower = sv_dup_inc(proto_perl->Iutf8_tolower, param);
13322 PL_utf8_tofold = sv_dup_inc(proto_perl->Iutf8_tofold, param);
13323 PL_utf8_idstart = sv_dup_inc(proto_perl->Iutf8_idstart, param);
13324 PL_utf8_xidstart = sv_dup_inc(proto_perl->Iutf8_xidstart, param);
13325 PL_utf8_perl_idstart = sv_dup_inc(proto_perl->Iutf8_perl_idstart, param);
13326 PL_utf8_idcont = sv_dup_inc(proto_perl->Iutf8_idcont, param);
13327 PL_utf8_xidcont = sv_dup_inc(proto_perl->Iutf8_xidcont, param);
13328 PL_utf8_foldable = sv_dup_inc(proto_perl->Iutf8_foldable, param);
13331 if (proto_perl->Ipsig_pend) {
13332 Newxz(PL_psig_pend, SIG_SIZE, int);
13335 PL_psig_pend = (int*)NULL;
13338 if (proto_perl->Ipsig_name) {
13339 Newx(PL_psig_name, 2 * SIG_SIZE, SV*);
13340 sv_dup_inc_multiple(proto_perl->Ipsig_name, PL_psig_name, 2 * SIG_SIZE,
13342 PL_psig_ptr = PL_psig_name + SIG_SIZE;
13345 PL_psig_ptr = (SV**)NULL;
13346 PL_psig_name = (SV**)NULL;
13349 if (flags & CLONEf_COPY_STACKS) {
13350 Newx(PL_tmps_stack, PL_tmps_max, SV*);
13351 sv_dup_inc_multiple(proto_perl->Itmps_stack, PL_tmps_stack,
13352 PL_tmps_ix+1, param);
13354 /* next PUSHMARK() sets *(PL_markstack_ptr+1) */
13355 i = proto_perl->Imarkstack_max - proto_perl->Imarkstack;
13356 Newxz(PL_markstack, i, I32);
13357 PL_markstack_max = PL_markstack + (proto_perl->Imarkstack_max
13358 - proto_perl->Imarkstack);
13359 PL_markstack_ptr = PL_markstack + (proto_perl->Imarkstack_ptr
13360 - proto_perl->Imarkstack);
13361 Copy(proto_perl->Imarkstack, PL_markstack,
13362 PL_markstack_ptr - PL_markstack + 1, I32);
13364 /* next push_scope()/ENTER sets PL_scopestack[PL_scopestack_ix]
13365 * NOTE: unlike the others! */
13366 Newxz(PL_scopestack, PL_scopestack_max, I32);
13367 Copy(proto_perl->Iscopestack, PL_scopestack, PL_scopestack_ix, I32);
13370 Newxz(PL_scopestack_name, PL_scopestack_max, const char *);
13371 Copy(proto_perl->Iscopestack_name, PL_scopestack_name, PL_scopestack_ix, const char *);
13373 /* NOTE: si_dup() looks at PL_markstack */
13374 PL_curstackinfo = si_dup(proto_perl->Icurstackinfo, param);
13376 /* PL_curstack = PL_curstackinfo->si_stack; */
13377 PL_curstack = av_dup(proto_perl->Icurstack, param);
13378 PL_mainstack = av_dup(proto_perl->Imainstack, param);
13380 /* next PUSHs() etc. set *(PL_stack_sp+1) */
13381 PL_stack_base = AvARRAY(PL_curstack);
13382 PL_stack_sp = PL_stack_base + (proto_perl->Istack_sp
13383 - proto_perl->Istack_base);
13384 PL_stack_max = PL_stack_base + AvMAX(PL_curstack);
13386 /*Newxz(PL_savestack, PL_savestack_max, ANY);*/
13387 PL_savestack = ss_dup(proto_perl, param);
13391 ENTER; /* perl_destruct() wants to LEAVE; */
13394 PL_statgv = gv_dup(proto_perl->Istatgv, param);
13395 PL_statname = sv_dup_inc(proto_perl->Istatname, param);
13397 PL_rs = sv_dup_inc(proto_perl->Irs, param);
13398 PL_last_in_gv = gv_dup(proto_perl->Ilast_in_gv, param);
13399 PL_defoutgv = gv_dup_inc(proto_perl->Idefoutgv, param);
13400 PL_toptarget = sv_dup_inc(proto_perl->Itoptarget, param);
13401 PL_bodytarget = sv_dup_inc(proto_perl->Ibodytarget, param);
13402 PL_formtarget = sv_dup(proto_perl->Iformtarget, param);
13404 PL_errors = sv_dup_inc(proto_perl->Ierrors, param);
13406 PL_sortcop = (OP*)any_dup(proto_perl->Isortcop, proto_perl);
13407 PL_sortstash = hv_dup(proto_perl->Isortstash, param);
13408 PL_firstgv = gv_dup(proto_perl->Ifirstgv, param);
13409 PL_secondgv = gv_dup(proto_perl->Isecondgv, param);
13411 PL_stashcache = newHV();
13413 PL_watchaddr = (char **) ptr_table_fetch(PL_ptr_table,
13414 proto_perl->Iwatchaddr);
13415 PL_watchok = PL_watchaddr ? * PL_watchaddr : NULL;
13416 if (PL_debug && PL_watchaddr) {
13417 PerlIO_printf(Perl_debug_log,
13418 "WATCHING: %"UVxf" cloned as %"UVxf" with value %"UVxf"\n",
13419 PTR2UV(proto_perl->Iwatchaddr), PTR2UV(PL_watchaddr),
13420 PTR2UV(PL_watchok));
13423 PL_registered_mros = hv_dup_inc(proto_perl->Iregistered_mros, param);
13424 PL_blockhooks = av_dup_inc(proto_perl->Iblockhooks, param);
13425 PL_utf8_foldclosures = hv_dup_inc(proto_perl->Iutf8_foldclosures, param);
13427 /* Call the ->CLONE method, if it exists, for each of the stashes
13428 identified by sv_dup() above.
13430 while(av_len(param->stashes) != -1) {
13431 HV* const stash = MUTABLE_HV(av_shift(param->stashes));
13432 GV* const cloner = gv_fetchmethod_autoload(stash, "CLONE", 0);
13433 if (cloner && GvCV(cloner)) {
13438 mXPUSHs(newSVhek(HvNAME_HEK(stash)));
13440 call_sv(MUTABLE_SV(GvCV(cloner)), G_DISCARD);
13446 if (!(flags & CLONEf_KEEP_PTR_TABLE)) {
13447 ptr_table_free(PL_ptr_table);
13448 PL_ptr_table = NULL;
13451 if (!(flags & CLONEf_COPY_STACKS)) {
13452 unreferenced_to_tmp_stack(param->unreferenced);
13455 SvREFCNT_dec(param->stashes);
13457 /* orphaned? eg threads->new inside BEGIN or use */
13458 if (PL_compcv && ! SvREFCNT(PL_compcv)) {
13459 SvREFCNT_inc_simple_void(PL_compcv);
13460 SAVEFREESV(PL_compcv);
13467 S_unreferenced_to_tmp_stack(pTHX_ AV *const unreferenced)
13469 PERL_ARGS_ASSERT_UNREFERENCED_TO_TMP_STACK;
13471 if (AvFILLp(unreferenced) > -1) {
13472 SV **svp = AvARRAY(unreferenced);
13473 SV **const last = svp + AvFILLp(unreferenced);
13477 if (SvREFCNT(*svp) == 1)
13479 } while (++svp <= last);
13481 EXTEND_MORTAL(count);
13482 svp = AvARRAY(unreferenced);
13485 if (SvREFCNT(*svp) == 1) {
13486 /* Our reference is the only one to this SV. This means that
13487 in this thread, the scalar effectively has a 0 reference.
13488 That doesn't work (cleanup never happens), so donate our
13489 reference to it onto the save stack. */
13490 PL_tmps_stack[++PL_tmps_ix] = *svp;
13492 /* As an optimisation, because we are already walking the
13493 entire array, instead of above doing either
13494 SvREFCNT_inc(*svp) or *svp = &PL_sv_undef, we can instead
13495 release our reference to the scalar, so that at the end of
13496 the array owns zero references to the scalars it happens to
13497 point to. We are effectively converting the array from
13498 AvREAL() on to AvREAL() off. This saves the av_clear()
13499 (triggered by the SvREFCNT_dec(unreferenced) below) from
13500 walking the array a second time. */
13501 SvREFCNT_dec(*svp);
13504 } while (++svp <= last);
13505 AvREAL_off(unreferenced);
13507 SvREFCNT_dec(unreferenced);
13511 Perl_clone_params_del(CLONE_PARAMS *param)
13513 /* This seemingly funky ordering keeps the build with PERL_GLOBAL_STRUCT
13515 PerlInterpreter *const to = param->new_perl;
13517 PerlInterpreter *const was = PERL_GET_THX;
13519 PERL_ARGS_ASSERT_CLONE_PARAMS_DEL;
13525 SvREFCNT_dec(param->stashes);
13526 if (param->unreferenced)
13527 unreferenced_to_tmp_stack(param->unreferenced);
13537 Perl_clone_params_new(PerlInterpreter *const from, PerlInterpreter *const to)
13540 /* Need to play this game, as newAV() can call safesysmalloc(), and that
13541 does a dTHX; to get the context from thread local storage.
13542 FIXME - under PERL_CORE Newx(), Safefree() and friends should expand to
13543 a version that passes in my_perl. */
13544 PerlInterpreter *const was = PERL_GET_THX;
13545 CLONE_PARAMS *param;
13547 PERL_ARGS_ASSERT_CLONE_PARAMS_NEW;
13553 /* Given that we've set the context, we can do this unshared. */
13554 Newx(param, 1, CLONE_PARAMS);
13557 param->proto_perl = from;
13558 param->new_perl = to;
13559 param->stashes = (AV *)Perl_newSV_type(to, SVt_PVAV);
13560 AvREAL_off(param->stashes);
13561 param->unreferenced = (AV *)Perl_newSV_type(to, SVt_PVAV);
13569 #endif /* USE_ITHREADS */
13572 =head1 Unicode Support
13574 =for apidoc sv_recode_to_utf8
13576 The encoding is assumed to be an Encode object, on entry the PV
13577 of the sv is assumed to be octets in that encoding, and the sv
13578 will be converted into Unicode (and UTF-8).
13580 If the sv already is UTF-8 (or if it is not POK), or if the encoding
13581 is not a reference, nothing is done to the sv. If the encoding is not
13582 an C<Encode::XS> Encoding object, bad things will happen.
13583 (See F<lib/encoding.pm> and L<Encode>).
13585 The PV of the sv is returned.
13590 Perl_sv_recode_to_utf8(pTHX_ SV *sv, SV *encoding)
13594 PERL_ARGS_ASSERT_SV_RECODE_TO_UTF8;
13596 if (SvPOK(sv) && !SvUTF8(sv) && !IN_BYTES && SvROK(encoding)) {
13610 Passing sv_yes is wrong - it needs to be or'ed set of constants
13611 for Encode::XS, while UTf-8 decode (currently) assumes a true value means
13612 remove converted chars from source.
13614 Both will default the value - let them.
13616 XPUSHs(&PL_sv_yes);
13619 call_method("decode", G_SCALAR);
13623 s = SvPV_const(uni, len);
13624 if (s != SvPVX_const(sv)) {
13625 SvGROW(sv, len + 1);
13626 Move(s, SvPVX(sv), len + 1, char);
13627 SvCUR_set(sv, len);
13631 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
13632 /* clear pos and any utf8 cache */
13633 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
13636 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
13637 magic_setutf8(sv,mg); /* clear UTF8 cache */
13642 return SvPOKp(sv) ? SvPVX(sv) : NULL;
13646 =for apidoc sv_cat_decode
13648 The encoding is assumed to be an Encode object, the PV of the ssv is
13649 assumed to be octets in that encoding and decoding the input starts
13650 from the position which (PV + *offset) pointed to. The dsv will be
13651 concatenated the decoded UTF-8 string from ssv. Decoding will terminate
13652 when the string tstr appears in decoding output or the input ends on
13653 the PV of the ssv. The value which the offset points will be modified
13654 to the last input position on the ssv.
13656 Returns TRUE if the terminator was found, else returns FALSE.
13661 Perl_sv_cat_decode(pTHX_ SV *dsv, SV *encoding,
13662 SV *ssv, int *offset, char *tstr, int tlen)
13667 PERL_ARGS_ASSERT_SV_CAT_DECODE;
13669 if (SvPOK(ssv) && SvPOK(dsv) && SvROK(encoding) && offset) {
13680 offsv = newSViv(*offset);
13682 mXPUSHp(tstr, tlen);
13684 call_method("cat_decode", G_SCALAR);
13686 ret = SvTRUE(TOPs);
13687 *offset = SvIV(offsv);
13693 Perl_croak(aTHX_ "Invalid argument to sv_cat_decode");
13698 /* ---------------------------------------------------------------------
13700 * support functions for report_uninit()
13703 /* the maxiumum size of array or hash where we will scan looking
13704 * for the undefined element that triggered the warning */
13706 #define FUV_MAX_SEARCH_SIZE 1000
13708 /* Look for an entry in the hash whose value has the same SV as val;
13709 * If so, return a mortal copy of the key. */
13712 S_find_hash_subscript(pTHX_ const HV *const hv, const SV *const val)
13715 register HE **array;
13718 PERL_ARGS_ASSERT_FIND_HASH_SUBSCRIPT;
13720 if (!hv || SvMAGICAL(hv) || !HvARRAY(hv) ||
13721 (HvTOTALKEYS(hv) > FUV_MAX_SEARCH_SIZE))
13724 array = HvARRAY(hv);
13726 for (i=HvMAX(hv); i>0; i--) {
13727 register HE *entry;
13728 for (entry = array[i]; entry; entry = HeNEXT(entry)) {
13729 if (HeVAL(entry) != val)
13731 if ( HeVAL(entry) == &PL_sv_undef ||
13732 HeVAL(entry) == &PL_sv_placeholder)
13736 if (HeKLEN(entry) == HEf_SVKEY)
13737 return sv_mortalcopy(HeKEY_sv(entry));
13738 return sv_2mortal(newSVhek(HeKEY_hek(entry)));
13744 /* Look for an entry in the array whose value has the same SV as val;
13745 * If so, return the index, otherwise return -1. */
13748 S_find_array_subscript(pTHX_ const AV *const av, const SV *const val)
13752 PERL_ARGS_ASSERT_FIND_ARRAY_SUBSCRIPT;
13754 if (!av || SvMAGICAL(av) || !AvARRAY(av) ||
13755 (AvFILLp(av) > FUV_MAX_SEARCH_SIZE))
13758 if (val != &PL_sv_undef) {
13759 SV ** const svp = AvARRAY(av);
13762 for (i=AvFILLp(av); i>=0; i--)
13769 /* S_varname(): return the name of a variable, optionally with a subscript.
13770 * If gv is non-zero, use the name of that global, along with gvtype (one
13771 * of "$", "@", "%"); otherwise use the name of the lexical at pad offset
13772 * targ. Depending on the value of the subscript_type flag, return:
13775 #define FUV_SUBSCRIPT_NONE 1 /* "@foo" */
13776 #define FUV_SUBSCRIPT_ARRAY 2 /* "$foo[aindex]" */
13777 #define FUV_SUBSCRIPT_HASH 3 /* "$foo{keyname}" */
13778 #define FUV_SUBSCRIPT_WITHIN 4 /* "within @foo" */
13781 S_varname(pTHX_ const GV *const gv, const char gvtype, PADOFFSET targ,
13782 const SV *const keyname, I32 aindex, int subscript_type)
13785 SV * const name = sv_newmortal();
13788 buffer[0] = gvtype;
13791 /* as gv_fullname4(), but add literal '^' for $^FOO names */
13793 gv_fullname4(name, gv, buffer, 0);
13795 if ((unsigned int)SvPVX(name)[1] <= 26) {
13797 buffer[1] = SvPVX(name)[1] + 'A' - 1;
13799 /* Swap the 1 unprintable control character for the 2 byte pretty
13800 version - ie substr($name, 1, 1) = $buffer; */
13801 sv_insert(name, 1, 1, buffer, 2);
13805 CV * const cv = find_runcv(NULL);
13809 if (!cv || !CvPADLIST(cv))
13811 av = MUTABLE_AV((*av_fetch(CvPADLIST(cv), 0, FALSE)));
13812 sv = *av_fetch(av, targ, FALSE);
13813 sv_setsv(name, sv);
13816 if (subscript_type == FUV_SUBSCRIPT_HASH) {
13817 SV * const sv = newSV(0);
13818 *SvPVX(name) = '$';
13819 Perl_sv_catpvf(aTHX_ name, "{%s}",
13820 pv_display(sv,SvPVX_const(keyname), SvCUR(keyname), 0, 32));
13823 else if (subscript_type == FUV_SUBSCRIPT_ARRAY) {
13824 *SvPVX(name) = '$';
13825 Perl_sv_catpvf(aTHX_ name, "[%"IVdf"]", (IV)aindex);
13827 else if (subscript_type == FUV_SUBSCRIPT_WITHIN) {
13828 /* We know that name has no magic, so can use 0 instead of SV_GMAGIC */
13829 Perl_sv_insert_flags(aTHX_ name, 0, 0, STR_WITH_LEN("within "), 0);
13837 =for apidoc find_uninit_var
13839 Find the name of the undefined variable (if any) that caused the operator o
13840 to issue a "Use of uninitialized value" warning.
13841 If match is true, only return a name if it's value matches uninit_sv.
13842 So roughly speaking, if a unary operator (such as OP_COS) generates a
13843 warning, then following the direct child of the op may yield an
13844 OP_PADSV or OP_GV that gives the name of the undefined variable. On the
13845 other hand, with OP_ADD there are two branches to follow, so we only print
13846 the variable name if we get an exact match.
13848 The name is returned as a mortal SV.
13850 Assumes that PL_op is the op that originally triggered the error, and that
13851 PL_comppad/PL_curpad points to the currently executing pad.
13857 S_find_uninit_var(pTHX_ const OP *const obase, const SV *const uninit_sv,
13863 const OP *o, *o2, *kid;
13865 if (!obase || (match && (!uninit_sv || uninit_sv == &PL_sv_undef ||
13866 uninit_sv == &PL_sv_placeholder)))
13869 switch (obase->op_type) {
13876 const bool pad = (obase->op_type == OP_PADAV || obase->op_type == OP_PADHV);
13877 const bool hash = (obase->op_type == OP_PADHV || obase->op_type == OP_RV2HV);
13880 int subscript_type = FUV_SUBSCRIPT_WITHIN;
13882 if (pad) { /* @lex, %lex */
13883 sv = PAD_SVl(obase->op_targ);
13887 if (cUNOPx(obase)->op_first->op_type == OP_GV) {
13888 /* @global, %global */
13889 gv = cGVOPx_gv(cUNOPx(obase)->op_first);
13892 sv = hash ? MUTABLE_SV(GvHV(gv)): MUTABLE_SV(GvAV(gv));
13894 else /* @{expr}, %{expr} */
13895 return find_uninit_var(cUNOPx(obase)->op_first,
13899 /* attempt to find a match within the aggregate */
13901 keysv = find_hash_subscript((const HV*)sv, uninit_sv);
13903 subscript_type = FUV_SUBSCRIPT_HASH;
13906 index = find_array_subscript((const AV *)sv, uninit_sv);
13908 subscript_type = FUV_SUBSCRIPT_ARRAY;
13911 if (match && subscript_type == FUV_SUBSCRIPT_WITHIN)
13914 return varname(gv, hash ? '%' : '@', obase->op_targ,
13915 keysv, index, subscript_type);
13919 if (cUNOPx(obase)->op_first->op_type == OP_GV) {
13921 gv = cGVOPx_gv(cUNOPx(obase)->op_first);
13922 if (!gv || !GvSTASH(gv))
13924 if (match && (GvSV(gv) != uninit_sv))
13926 return varname(gv, '$', 0, NULL, 0, FUV_SUBSCRIPT_NONE);
13929 return find_uninit_var(cUNOPx(obase)->op_first, uninit_sv, 1);
13932 if (match && PAD_SVl(obase->op_targ) != uninit_sv)
13934 return varname(NULL, '$', obase->op_targ,
13935 NULL, 0, FUV_SUBSCRIPT_NONE);
13938 gv = cGVOPx_gv(obase);
13939 if (!gv || (match && GvSV(gv) != uninit_sv) || !GvSTASH(gv))
13941 return varname(gv, '$', 0, NULL, 0, FUV_SUBSCRIPT_NONE);
13943 case OP_AELEMFAST_LEX:
13946 AV *av = MUTABLE_AV(PAD_SV(obase->op_targ));
13947 if (!av || SvRMAGICAL(av))
13949 svp = av_fetch(av, (I32)obase->op_private, FALSE);
13950 if (!svp || *svp != uninit_sv)
13953 return varname(NULL, '$', obase->op_targ,
13954 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
13957 gv = cGVOPx_gv(obase);
13962 AV *const av = GvAV(gv);
13963 if (!av || SvRMAGICAL(av))
13965 svp = av_fetch(av, (I32)obase->op_private, FALSE);
13966 if (!svp || *svp != uninit_sv)
13969 return varname(gv, '$', 0,
13970 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
13975 o = cUNOPx(obase)->op_first;
13976 if (!o || o->op_type != OP_NULL ||
13977 ! (o->op_targ == OP_AELEM || o->op_targ == OP_HELEM))
13979 return find_uninit_var(cBINOPo->op_last, uninit_sv, match);
13984 bool negate = FALSE;
13986 if (PL_op == obase)
13987 /* $a[uninit_expr] or $h{uninit_expr} */
13988 return find_uninit_var(cBINOPx(obase)->op_last, uninit_sv, match);
13991 o = cBINOPx(obase)->op_first;
13992 kid = cBINOPx(obase)->op_last;
13994 /* get the av or hv, and optionally the gv */
13996 if (o->op_type == OP_PADAV || o->op_type == OP_PADHV) {
13997 sv = PAD_SV(o->op_targ);
13999 else if ((o->op_type == OP_RV2AV || o->op_type == OP_RV2HV)
14000 && cUNOPo->op_first->op_type == OP_GV)
14002 gv = cGVOPx_gv(cUNOPo->op_first);
14006 == OP_RV2HV ? MUTABLE_SV(GvHV(gv)) : MUTABLE_SV(GvAV(gv));
14011 if (kid && kid->op_type == OP_NEGATE) {
14013 kid = cUNOPx(kid)->op_first;
14016 if (kid && kid->op_type == OP_CONST && SvOK(cSVOPx_sv(kid))) {
14017 /* index is constant */
14020 kidsv = sv_2mortal(newSVpvs("-"));
14021 sv_catsv(kidsv, cSVOPx_sv(kid));
14024 kidsv = cSVOPx_sv(kid);
14028 if (obase->op_type == OP_HELEM) {
14029 HE* he = hv_fetch_ent(MUTABLE_HV(sv), kidsv, 0, 0);
14030 if (!he || HeVAL(he) != uninit_sv)
14034 SV * const * const svp = av_fetch(MUTABLE_AV(sv),
14035 negate ? - SvIV(cSVOPx_sv(kid)) : SvIV(cSVOPx_sv(kid)),
14037 if (!svp || *svp != uninit_sv)
14041 if (obase->op_type == OP_HELEM)
14042 return varname(gv, '%', o->op_targ,
14043 kidsv, 0, FUV_SUBSCRIPT_HASH);
14045 return varname(gv, '@', o->op_targ, NULL,
14046 negate ? - SvIV(cSVOPx_sv(kid)) : SvIV(cSVOPx_sv(kid)),
14047 FUV_SUBSCRIPT_ARRAY);
14050 /* index is an expression;
14051 * attempt to find a match within the aggregate */
14052 if (obase->op_type == OP_HELEM) {
14053 SV * const keysv = find_hash_subscript((const HV*)sv, uninit_sv);
14055 return varname(gv, '%', o->op_targ,
14056 keysv, 0, FUV_SUBSCRIPT_HASH);
14060 = find_array_subscript((const AV *)sv, uninit_sv);
14062 return varname(gv, '@', o->op_targ,
14063 NULL, index, FUV_SUBSCRIPT_ARRAY);
14068 (o->op_type == OP_PADAV || o->op_type == OP_RV2AV)
14070 o->op_targ, NULL, 0, FUV_SUBSCRIPT_WITHIN);
14076 /* only examine RHS */
14077 return find_uninit_var(cBINOPx(obase)->op_first, uninit_sv, match);
14080 o = cUNOPx(obase)->op_first;
14081 if (o->op_type == OP_PUSHMARK)
14084 if (!o->op_sibling) {
14085 /* one-arg version of open is highly magical */
14087 if (o->op_type == OP_GV) { /* open FOO; */
14089 if (match && GvSV(gv) != uninit_sv)
14091 return varname(gv, '$', 0,
14092 NULL, 0, FUV_SUBSCRIPT_NONE);
14094 /* other possibilities not handled are:
14095 * open $x; or open my $x; should return '${*$x}'
14096 * open expr; should return '$'.expr ideally
14102 /* ops where $_ may be an implicit arg */
14106 if ( !(obase->op_flags & OPf_STACKED)) {
14107 if (uninit_sv == ((obase->op_private & OPpTARGET_MY)
14108 ? PAD_SVl(obase->op_targ)
14111 sv = sv_newmortal();
14112 sv_setpvs(sv, "$_");
14121 match = 1; /* print etc can return undef on defined args */
14122 /* skip filehandle as it can't produce 'undef' warning */
14123 o = cUNOPx(obase)->op_first;
14124 if ((obase->op_flags & OPf_STACKED) && o->op_type == OP_PUSHMARK)
14125 o = o->op_sibling->op_sibling;
14129 case OP_ENTEREVAL: /* could be eval $undef or $x='$undef'; eval $x */
14130 case OP_CUSTOM: /* XS or custom code could trigger random warnings */
14132 /* the following ops are capable of returning PL_sv_undef even for
14133 * defined arg(s) */
14152 case OP_GETPEERNAME:
14200 case OP_SMARTMATCH:
14209 /* XXX tmp hack: these two may call an XS sub, and currently
14210 XS subs don't have a SUB entry on the context stack, so CV and
14211 pad determination goes wrong, and BAD things happen. So, just
14212 don't try to determine the value under those circumstances.
14213 Need a better fix at dome point. DAPM 11/2007 */
14219 GV * const gv = gv_fetchpvs(".", GV_NOTQUAL, SVt_PV);
14220 if (gv && GvSV(gv) == uninit_sv)
14221 return newSVpvs_flags("$.", SVs_TEMP);
14226 /* def-ness of rval pos() is independent of the def-ness of its arg */
14227 if ( !(obase->op_flags & OPf_MOD))
14232 if (SvROK(PL_rs) && uninit_sv == SvRV(PL_rs))
14233 return newSVpvs_flags("${$/}", SVs_TEMP);
14238 if (!(obase->op_flags & OPf_KIDS))
14240 o = cUNOPx(obase)->op_first;
14246 /* if all except one arg are constant, or have no side-effects,
14247 * or are optimized away, then it's unambiguous */
14249 for (kid=o; kid; kid = kid->op_sibling) {
14251 const OPCODE type = kid->op_type;
14252 if ( (type == OP_CONST && SvOK(cSVOPx_sv(kid)))
14253 || (type == OP_NULL && ! (kid->op_flags & OPf_KIDS))
14254 || (type == OP_PUSHMARK)
14256 /* @$a and %$a, but not @a or %a */
14257 (type == OP_RV2AV || type == OP_RV2HV)
14258 && cUNOPx(kid)->op_first
14259 && cUNOPx(kid)->op_first->op_type != OP_GV
14264 if (o2) { /* more than one found */
14271 return find_uninit_var(o2, uninit_sv, match);
14273 /* scan all args */
14275 sv = find_uninit_var(o, uninit_sv, 1);
14287 =for apidoc report_uninit
14289 Print appropriate "Use of uninitialized variable" warning
14295 Perl_report_uninit(pTHX_ const SV *uninit_sv)
14299 SV* varname = NULL;
14300 if (uninit_sv && PL_curpad) {
14301 varname = find_uninit_var(PL_op, uninit_sv,0);
14303 sv_insert(varname, 0, 0, " ", 1);
14305 /* diag_listed_as: Use of uninitialized value%s */
14306 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit_sv,
14307 SVfARG(varname ? varname : &PL_sv_no),
14308 " in ", OP_DESC(PL_op));
14311 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
14317 * c-indentation-style: bsd
14318 * c-basic-offset: 4
14319 * indent-tabs-mode: t
14322 * ex: set ts=8 sts=4 sw=4 noet: