3 * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, 2000,
4 * 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009 by Larry Wall
7 * You may distribute under the terms of either the GNU General Public
8 * License or the Artistic License, as specified in the README file.
13 * 'I wonder what the Entish is for "yes" and "no",' he thought.
16 * [p.480 of _The Lord of the Rings_, III/iv: "Treebeard"]
22 * This file contains the code that creates, manipulates and destroys
23 * scalar values (SVs). The other types (AV, HV, GV, etc.) reuse the
24 * structure of an SV, so their creation and destruction is handled
25 * here; higher-level functions are in av.c, hv.c, and so on. Opcode
26 * level functions (eg. substr, split, join) for each of the types are
36 # if __STDC_VERSION__ >= 199901L && !defined(VMS)
47 /* Missing proto on LynxOS */
48 char *gconvert(double, int, int, char *);
51 #ifdef PERL_UTF8_CACHE_ASSERT
52 /* if adding more checks watch out for the following tests:
53 * t/op/index.t t/op/length.t t/op/pat.t t/op/substr.t
54 * lib/utf8.t lib/Unicode/Collate/t/index.t
57 # define ASSERT_UTF8_CACHE(cache) \
58 STMT_START { if (cache) { assert((cache)[0] <= (cache)[1]); \
59 assert((cache)[2] <= (cache)[3]); \
60 assert((cache)[3] <= (cache)[1]);} \
63 # define ASSERT_UTF8_CACHE(cache) NOOP
66 #ifdef PERL_OLD_COPY_ON_WRITE
67 #define SV_COW_NEXT_SV(sv) INT2PTR(SV *,SvUVX(sv))
68 #define SV_COW_NEXT_SV_SET(current,next) SvUV_set(current, PTR2UV(next))
69 /* This is a pessimistic view. Scalar must be purely a read-write PV to copy-
73 /* ============================================================================
75 =head1 Allocation and deallocation of SVs.
77 An SV (or AV, HV, etc.) is allocated in two parts: the head (struct
78 sv, av, hv...) contains type and reference count information, and for
79 many types, a pointer to the body (struct xrv, xpv, xpviv...), which
80 contains fields specific to each type. Some types store all they need
81 in the head, so don't have a body.
83 In all but the most memory-paranoid configurations (ex: PURIFY), heads
84 and bodies are allocated out of arenas, which by default are
85 approximately 4K chunks of memory parcelled up into N heads or bodies.
86 Sv-bodies are allocated by their sv-type, guaranteeing size
87 consistency needed to allocate safely from arrays.
89 For SV-heads, the first slot in each arena is reserved, and holds a
90 link to the next arena, some flags, and a note of the number of slots.
91 Snaked through each arena chain is a linked list of free items; when
92 this becomes empty, an extra arena is allocated and divided up into N
93 items which are threaded into the free list.
95 SV-bodies are similar, but they use arena-sets by default, which
96 separate the link and info from the arena itself, and reclaim the 1st
97 slot in the arena. SV-bodies are further described later.
99 The following global variables are associated with arenas:
101 PL_sv_arenaroot pointer to list of SV arenas
102 PL_sv_root pointer to list of free SV structures
104 PL_body_arenas head of linked-list of body arenas
105 PL_body_roots[] array of pointers to list of free bodies of svtype
106 arrays are indexed by the svtype needed
108 A few special SV heads are not allocated from an arena, but are
109 instead directly created in the interpreter structure, eg PL_sv_undef.
110 The size of arenas can be changed from the default by setting
111 PERL_ARENA_SIZE appropriately at compile time.
113 The SV arena serves the secondary purpose of allowing still-live SVs
114 to be located and destroyed during final cleanup.
116 At the lowest level, the macros new_SV() and del_SV() grab and free
117 an SV head. (If debugging with -DD, del_SV() calls the function S_del_sv()
118 to return the SV to the free list with error checking.) new_SV() calls
119 more_sv() / sv_add_arena() to add an extra arena if the free list is empty.
120 SVs in the free list have their SvTYPE field set to all ones.
122 At the time of very final cleanup, sv_free_arenas() is called from
123 perl_destruct() to physically free all the arenas allocated since the
124 start of the interpreter.
126 The function visit() scans the SV arenas list, and calls a specified
127 function for each SV it finds which is still live - ie which has an SvTYPE
128 other than all 1's, and a non-zero SvREFCNT. visit() is used by the
129 following functions (specified as [function that calls visit()] / [function
130 called by visit() for each SV]):
132 sv_report_used() / do_report_used()
133 dump all remaining SVs (debugging aid)
135 sv_clean_objs() / do_clean_objs(),do_clean_named_objs(),
136 do_clean_named_io_objs()
137 Attempt to free all objects pointed to by RVs,
138 and try to do the same for all objects indirectly
139 referenced by typeglobs too. Called once from
140 perl_destruct(), prior to calling sv_clean_all()
143 sv_clean_all() / do_clean_all()
144 SvREFCNT_dec(sv) each remaining SV, possibly
145 triggering an sv_free(). It also sets the
146 SVf_BREAK flag on the SV to indicate that the
147 refcnt has been artificially lowered, and thus
148 stopping sv_free() from giving spurious warnings
149 about SVs which unexpectedly have a refcnt
150 of zero. called repeatedly from perl_destruct()
151 until there are no SVs left.
153 =head2 Arena allocator API Summary
155 Private API to rest of sv.c
159 new_XPVNV(), del_XPVGV(),
164 sv_report_used(), sv_clean_objs(), sv_clean_all(), sv_free_arenas()
168 * ========================================================================= */
171 * "A time to plant, and a time to uproot what was planted..."
175 # define MEM_LOG_NEW_SV(sv, file, line, func) \
176 Perl_mem_log_new_sv(sv, file, line, func)
177 # define MEM_LOG_DEL_SV(sv, file, line, func) \
178 Perl_mem_log_del_sv(sv, file, line, func)
180 # define MEM_LOG_NEW_SV(sv, file, line, func) NOOP
181 # define MEM_LOG_DEL_SV(sv, file, line, func) NOOP
184 #ifdef DEBUG_LEAKING_SCALARS
185 # define FREE_SV_DEBUG_FILE(sv) Safefree((sv)->sv_debug_file)
186 # define DEBUG_SV_SERIAL(sv) \
187 DEBUG_m(PerlIO_printf(Perl_debug_log, "0x%"UVxf": (%05ld) del_SV\n", \
188 PTR2UV(sv), (long)(sv)->sv_debug_serial))
190 # define FREE_SV_DEBUG_FILE(sv)
191 # define DEBUG_SV_SERIAL(sv) NOOP
195 # define SvARENA_CHAIN(sv) ((sv)->sv_u.svu_rv)
196 # define SvARENA_CHAIN_SET(sv,val) (sv)->sv_u.svu_rv = MUTABLE_SV((val))
197 /* Whilst I'd love to do this, it seems that things like to check on
199 # define POSION_SV_HEAD(sv) PoisonNew(sv, 1, struct STRUCT_SV)
201 # define POSION_SV_HEAD(sv) PoisonNew(&SvANY(sv), 1, void *), \
202 PoisonNew(&SvREFCNT(sv), 1, U32)
204 # define SvARENA_CHAIN(sv) SvANY(sv)
205 # define SvARENA_CHAIN_SET(sv,val) SvANY(sv) = (void *)(val)
206 # define POSION_SV_HEAD(sv)
209 /* Mark an SV head as unused, and add to free list.
211 * If SVf_BREAK is set, skip adding it to the free list, as this SV had
212 * its refcount artificially decremented during global destruction, so
213 * there may be dangling pointers to it. The last thing we want in that
214 * case is for it to be reused. */
216 #define plant_SV(p) \
218 const U32 old_flags = SvFLAGS(p); \
219 MEM_LOG_DEL_SV(p, __FILE__, __LINE__, FUNCTION__); \
220 DEBUG_SV_SERIAL(p); \
221 FREE_SV_DEBUG_FILE(p); \
223 SvFLAGS(p) = SVTYPEMASK; \
224 if (!(old_flags & SVf_BREAK)) { \
225 SvARENA_CHAIN_SET(p, PL_sv_root); \
231 #define uproot_SV(p) \
234 PL_sv_root = MUTABLE_SV(SvARENA_CHAIN(p)); \
239 /* make some more SVs by adding another arena */
246 char *chunk; /* must use New here to match call to */
247 Newx(chunk,PERL_ARENA_SIZE,char); /* Safefree() in sv_free_arenas() */
248 sv_add_arena(chunk, PERL_ARENA_SIZE, 0);
253 /* new_SV(): return a new, empty SV head */
255 #ifdef DEBUG_LEAKING_SCALARS
256 /* provide a real function for a debugger to play with */
258 S_new_SV(pTHX_ const char *file, int line, const char *func)
265 sv = S_more_sv(aTHX);
269 sv->sv_debug_optype = PL_op ? PL_op->op_type : 0;
270 sv->sv_debug_line = (U16) (PL_parser && PL_parser->copline != NOLINE
276 sv->sv_debug_inpad = 0;
277 sv->sv_debug_parent = NULL;
278 sv->sv_debug_file = PL_curcop ? savepv(CopFILE(PL_curcop)): NULL;
280 sv->sv_debug_serial = PL_sv_serial++;
282 MEM_LOG_NEW_SV(sv, file, line, func);
283 DEBUG_m(PerlIO_printf(Perl_debug_log, "0x%"UVxf": (%05ld) new_SV (from %s:%d [%s])\n",
284 PTR2UV(sv), (long)sv->sv_debug_serial, file, line, func));
288 # define new_SV(p) (p)=S_new_SV(aTHX_ __FILE__, __LINE__, FUNCTION__)
296 (p) = S_more_sv(aTHX); \
300 MEM_LOG_NEW_SV(p, __FILE__, __LINE__, FUNCTION__); \
305 /* del_SV(): return an empty SV head to the free list */
318 S_del_sv(pTHX_ SV *p)
322 PERL_ARGS_ASSERT_DEL_SV;
327 for (sva = PL_sv_arenaroot; sva; sva = MUTABLE_SV(SvANY(sva))) {
328 const SV * const sv = sva + 1;
329 const SV * const svend = &sva[SvREFCNT(sva)];
330 if (p >= sv && p < svend) {
336 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL),
337 "Attempt to free non-arena SV: 0x%"UVxf
338 pTHX__FORMAT, PTR2UV(p) pTHX__VALUE);
345 #else /* ! DEBUGGING */
347 #define del_SV(p) plant_SV(p)
349 #endif /* DEBUGGING */
353 =head1 SV Manipulation Functions
355 =for apidoc sv_add_arena
357 Given a chunk of memory, link it to the head of the list of arenas,
358 and split it into a list of free SVs.
364 S_sv_add_arena(pTHX_ char *const ptr, const U32 size, const U32 flags)
367 SV *const sva = MUTABLE_SV(ptr);
371 PERL_ARGS_ASSERT_SV_ADD_ARENA;
373 /* The first SV in an arena isn't an SV. */
374 SvANY(sva) = (void *) PL_sv_arenaroot; /* ptr to next arena */
375 SvREFCNT(sva) = size / sizeof(SV); /* number of SV slots */
376 SvFLAGS(sva) = flags; /* FAKE if not to be freed */
378 PL_sv_arenaroot = sva;
379 PL_sv_root = sva + 1;
381 svend = &sva[SvREFCNT(sva) - 1];
384 SvARENA_CHAIN_SET(sv, (sv + 1));
388 /* Must always set typemask because it's always checked in on cleanup
389 when the arenas are walked looking for objects. */
390 SvFLAGS(sv) = SVTYPEMASK;
393 SvARENA_CHAIN_SET(sv, 0);
397 SvFLAGS(sv) = SVTYPEMASK;
400 /* visit(): call the named function for each non-free SV in the arenas
401 * whose flags field matches the flags/mask args. */
404 S_visit(pTHX_ SVFUNC_t f, const U32 flags, const U32 mask)
410 PERL_ARGS_ASSERT_VISIT;
412 for (sva = PL_sv_arenaroot; sva; sva = MUTABLE_SV(SvANY(sva))) {
413 register const SV * const svend = &sva[SvREFCNT(sva)];
415 for (sv = sva + 1; sv < svend; ++sv) {
416 if (SvTYPE(sv) != (svtype)SVTYPEMASK
417 && (sv->sv_flags & mask) == flags
430 /* called by sv_report_used() for each live SV */
433 do_report_used(pTHX_ SV *const sv)
435 if (SvTYPE(sv) != (svtype)SVTYPEMASK) {
436 PerlIO_printf(Perl_debug_log, "****\n");
443 =for apidoc sv_report_used
445 Dump the contents of all SVs not yet freed (debugging aid).
451 Perl_sv_report_used(pTHX)
454 visit(do_report_used, 0, 0);
460 /* called by sv_clean_objs() for each live SV */
463 do_clean_objs(pTHX_ SV *const ref)
468 SV * const target = SvRV(ref);
469 if (SvOBJECT(target)) {
470 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning object ref:\n "), sv_dump(ref)));
471 if (SvWEAKREF(ref)) {
472 sv_del_backref(target, ref);
478 SvREFCNT_dec(target);
483 /* XXX Might want to check arrays, etc. */
487 /* clear any slots in a GV which hold objects - except IO;
488 * called by sv_clean_objs() for each live GV */
491 do_clean_named_objs(pTHX_ SV *const sv)
495 assert(SvTYPE(sv) == SVt_PVGV);
496 assert(isGV_with_GP(sv));
500 /* freeing GP entries may indirectly free the current GV;
501 * hold onto it while we mess with the GP slots */
504 if ( ((obj = GvSV(sv) )) && SvOBJECT(obj)) {
505 DEBUG_D((PerlIO_printf(Perl_debug_log,
506 "Cleaning named glob SV object:\n "), sv_dump(obj)));
510 if ( ((obj = MUTABLE_SV(GvAV(sv)) )) && SvOBJECT(obj)) {
511 DEBUG_D((PerlIO_printf(Perl_debug_log,
512 "Cleaning named glob AV object:\n "), sv_dump(obj)));
516 if ( ((obj = MUTABLE_SV(GvHV(sv)) )) && SvOBJECT(obj)) {
517 DEBUG_D((PerlIO_printf(Perl_debug_log,
518 "Cleaning named glob HV object:\n "), sv_dump(obj)));
522 if ( ((obj = MUTABLE_SV(GvCV(sv)) )) && SvOBJECT(obj)) {
523 DEBUG_D((PerlIO_printf(Perl_debug_log,
524 "Cleaning named glob CV object:\n "), sv_dump(obj)));
528 SvREFCNT_dec(sv); /* undo the inc above */
531 /* clear any IO slots in a GV which hold objects (except stderr, defout);
532 * called by sv_clean_objs() for each live GV */
535 do_clean_named_io_objs(pTHX_ SV *const sv)
539 assert(SvTYPE(sv) == SVt_PVGV);
540 assert(isGV_with_GP(sv));
541 if (!GvGP(sv) || sv == (SV*)PL_stderrgv || sv == (SV*)PL_defoutgv)
545 if ( ((obj = MUTABLE_SV(GvIO(sv)) )) && SvOBJECT(obj)) {
546 DEBUG_D((PerlIO_printf(Perl_debug_log,
547 "Cleaning named glob IO object:\n "), sv_dump(obj)));
551 SvREFCNT_dec(sv); /* undo the inc above */
554 /* Void wrapper to pass to visit() */
556 do_curse(pTHX_ SV * const sv) {
557 if ((PL_stderrgv && GvGP(PL_stderrgv) && (SV*)GvIO(PL_stderrgv) == sv)
558 || (PL_defoutgv && GvGP(PL_defoutgv) && (SV*)GvIO(PL_defoutgv) == sv))
564 =for apidoc sv_clean_objs
566 Attempt to destroy all objects not yet freed.
572 Perl_sv_clean_objs(pTHX)
576 PL_in_clean_objs = TRUE;
577 visit(do_clean_objs, SVf_ROK, SVf_ROK);
578 /* Some barnacles may yet remain, clinging to typeglobs.
579 * Run the non-IO destructors first: they may want to output
580 * error messages, close files etc */
581 visit(do_clean_named_objs, SVt_PVGV|SVpgv_GP, SVTYPEMASK|SVp_POK|SVpgv_GP);
582 visit(do_clean_named_io_objs, SVt_PVGV|SVpgv_GP, SVTYPEMASK|SVp_POK|SVpgv_GP);
583 /* And if there are some very tenacious barnacles clinging to arrays,
584 closures, or what have you.... */
585 visit(do_curse, SVs_OBJECT, SVs_OBJECT);
586 olddef = PL_defoutgv;
587 PL_defoutgv = NULL; /* disable skip of PL_defoutgv */
588 if (olddef && isGV_with_GP(olddef))
589 do_clean_named_io_objs(aTHX_ MUTABLE_SV(olddef));
590 olderr = PL_stderrgv;
591 PL_stderrgv = NULL; /* disable skip of PL_stderrgv */
592 if (olderr && isGV_with_GP(olderr))
593 do_clean_named_io_objs(aTHX_ MUTABLE_SV(olderr));
594 SvREFCNT_dec(olddef);
595 PL_in_clean_objs = FALSE;
598 /* called by sv_clean_all() for each live SV */
601 do_clean_all(pTHX_ SV *const sv)
604 if (sv == (const SV *) PL_fdpid || sv == (const SV *)PL_strtab) {
605 /* don't clean pid table and strtab */
608 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning loops: SV at 0x%"UVxf"\n", PTR2UV(sv)) ));
609 SvFLAGS(sv) |= SVf_BREAK;
614 =for apidoc sv_clean_all
616 Decrement the refcnt of each remaining SV, possibly triggering a
617 cleanup. This function may have to be called multiple times to free
618 SVs which are in complex self-referential hierarchies.
624 Perl_sv_clean_all(pTHX)
628 PL_in_clean_all = TRUE;
629 cleaned = visit(do_clean_all, 0,0);
634 ARENASETS: a meta-arena implementation which separates arena-info
635 into struct arena_set, which contains an array of struct
636 arena_descs, each holding info for a single arena. By separating
637 the meta-info from the arena, we recover the 1st slot, formerly
638 borrowed for list management. The arena_set is about the size of an
639 arena, avoiding the needless malloc overhead of a naive linked-list.
641 The cost is 1 arena-set malloc per ~320 arena-mallocs, + the unused
642 memory in the last arena-set (1/2 on average). In trade, we get
643 back the 1st slot in each arena (ie 1.7% of a CV-arena, less for
644 smaller types). The recovery of the wasted space allows use of
645 small arenas for large, rare body types, by changing array* fields
646 in body_details_by_type[] below.
649 char *arena; /* the raw storage, allocated aligned */
650 size_t size; /* its size ~4k typ */
651 svtype utype; /* bodytype stored in arena */
656 /* Get the maximum number of elements in set[] such that struct arena_set
657 will fit within PERL_ARENA_SIZE, which is probably just under 4K, and
658 therefore likely to be 1 aligned memory page. */
660 #define ARENAS_PER_SET ((PERL_ARENA_SIZE - sizeof(struct arena_set*) \
661 - 2 * sizeof(int)) / sizeof (struct arena_desc))
664 struct arena_set* next;
665 unsigned int set_size; /* ie ARENAS_PER_SET */
666 unsigned int curr; /* index of next available arena-desc */
667 struct arena_desc set[ARENAS_PER_SET];
671 =for apidoc sv_free_arenas
673 Deallocate the memory used by all arenas. Note that all the individual SV
674 heads and bodies within the arenas must already have been freed.
679 Perl_sv_free_arenas(pTHX)
686 /* Free arenas here, but be careful about fake ones. (We assume
687 contiguity of the fake ones with the corresponding real ones.) */
689 for (sva = PL_sv_arenaroot; sva; sva = svanext) {
690 svanext = MUTABLE_SV(SvANY(sva));
691 while (svanext && SvFAKE(svanext))
692 svanext = MUTABLE_SV(SvANY(svanext));
699 struct arena_set *aroot = (struct arena_set*) PL_body_arenas;
702 struct arena_set *current = aroot;
705 assert(aroot->set[i].arena);
706 Safefree(aroot->set[i].arena);
714 i = PERL_ARENA_ROOTS_SIZE;
716 PL_body_roots[i] = 0;
723 Here are mid-level routines that manage the allocation of bodies out
724 of the various arenas. There are 5 kinds of arenas:
726 1. SV-head arenas, which are discussed and handled above
727 2. regular body arenas
728 3. arenas for reduced-size bodies
731 Arena types 2 & 3 are chained by body-type off an array of
732 arena-root pointers, which is indexed by svtype. Some of the
733 larger/less used body types are malloced singly, since a large
734 unused block of them is wasteful. Also, several svtypes dont have
735 bodies; the data fits into the sv-head itself. The arena-root
736 pointer thus has a few unused root-pointers (which may be hijacked
737 later for arena types 4,5)
739 3 differs from 2 as an optimization; some body types have several
740 unused fields in the front of the structure (which are kept in-place
741 for consistency). These bodies can be allocated in smaller chunks,
742 because the leading fields arent accessed. Pointers to such bodies
743 are decremented to point at the unused 'ghost' memory, knowing that
744 the pointers are used with offsets to the real memory.
747 =head1 SV-Body Allocation
749 Allocation of SV-bodies is similar to SV-heads, differing as follows;
750 the allocation mechanism is used for many body types, so is somewhat
751 more complicated, it uses arena-sets, and has no need for still-live
754 At the outermost level, (new|del)_X*V macros return bodies of the
755 appropriate type. These macros call either (new|del)_body_type or
756 (new|del)_body_allocated macro pairs, depending on specifics of the
757 type. Most body types use the former pair, the latter pair is used to
758 allocate body types with "ghost fields".
760 "ghost fields" are fields that are unused in certain types, and
761 consequently don't need to actually exist. They are declared because
762 they're part of a "base type", which allows use of functions as
763 methods. The simplest examples are AVs and HVs, 2 aggregate types
764 which don't use the fields which support SCALAR semantics.
766 For these types, the arenas are carved up into appropriately sized
767 chunks, we thus avoid wasted memory for those unaccessed members.
768 When bodies are allocated, we adjust the pointer back in memory by the
769 size of the part not allocated, so it's as if we allocated the full
770 structure. (But things will all go boom if you write to the part that
771 is "not there", because you'll be overwriting the last members of the
772 preceding structure in memory.)
774 We calculate the correction using the STRUCT_OFFSET macro on the first
775 member present. If the allocated structure is smaller (no initial NV
776 actually allocated) then the net effect is to subtract the size of the NV
777 from the pointer, to return a new pointer as if an initial NV were actually
778 allocated. (We were using structures named *_allocated for this, but
779 this turned out to be a subtle bug, because a structure without an NV
780 could have a lower alignment constraint, but the compiler is allowed to
781 optimised accesses based on the alignment constraint of the actual pointer
782 to the full structure, for example, using a single 64 bit load instruction
783 because it "knows" that two adjacent 32 bit members will be 8-byte aligned.)
785 This is the same trick as was used for NV and IV bodies. Ironically it
786 doesn't need to be used for NV bodies any more, because NV is now at
787 the start of the structure. IV bodies don't need it either, because
788 they are no longer allocated.
790 In turn, the new_body_* allocators call S_new_body(), which invokes
791 new_body_inline macro, which takes a lock, and takes a body off the
792 linked list at PL_body_roots[sv_type], calling Perl_more_bodies() if
793 necessary to refresh an empty list. Then the lock is released, and
794 the body is returned.
796 Perl_more_bodies allocates a new arena, and carves it up into an array of N
797 bodies, which it strings into a linked list. It looks up arena-size
798 and body-size from the body_details table described below, thus
799 supporting the multiple body-types.
801 If PURIFY is defined, or PERL_ARENA_SIZE=0, arenas are not used, and
802 the (new|del)_X*V macros are mapped directly to malloc/free.
804 For each sv-type, struct body_details bodies_by_type[] carries
805 parameters which control these aspects of SV handling:
807 Arena_size determines whether arenas are used for this body type, and if
808 so, how big they are. PURIFY or PERL_ARENA_SIZE=0 set this field to
809 zero, forcing individual mallocs and frees.
811 Body_size determines how big a body is, and therefore how many fit into
812 each arena. Offset carries the body-pointer adjustment needed for
813 "ghost fields", and is used in *_allocated macros.
815 But its main purpose is to parameterize info needed in
816 Perl_sv_upgrade(). The info here dramatically simplifies the function
817 vs the implementation in 5.8.8, making it table-driven. All fields
818 are used for this, except for arena_size.
820 For the sv-types that have no bodies, arenas are not used, so those
821 PL_body_roots[sv_type] are unused, and can be overloaded. In
822 something of a special case, SVt_NULL is borrowed for HE arenas;
823 PL_body_roots[HE_SVSLOT=SVt_NULL] is filled by S_more_he, but the
824 bodies_by_type[SVt_NULL] slot is not used, as the table is not
829 struct body_details {
830 U8 body_size; /* Size to allocate */
831 U8 copy; /* Size of structure to copy (may be shorter) */
833 unsigned int type : 4; /* We have space for a sanity check. */
834 unsigned int cant_upgrade : 1; /* Cannot upgrade this type */
835 unsigned int zero_nv : 1; /* zero the NV when upgrading from this */
836 unsigned int arena : 1; /* Allocated from an arena */
837 size_t arena_size; /* Size of arena to allocate */
845 /* With -DPURFIY we allocate everything directly, and don't use arenas.
846 This seems a rather elegant way to simplify some of the code below. */
847 #define HASARENA FALSE
849 #define HASARENA TRUE
851 #define NOARENA FALSE
853 /* Size the arenas to exactly fit a given number of bodies. A count
854 of 0 fits the max number bodies into a PERL_ARENA_SIZE.block,
855 simplifying the default. If count > 0, the arena is sized to fit
856 only that many bodies, allowing arenas to be used for large, rare
857 bodies (XPVFM, XPVIO) without undue waste. The arena size is
858 limited by PERL_ARENA_SIZE, so we can safely oversize the
861 #define FIT_ARENA0(body_size) \
862 ((size_t)(PERL_ARENA_SIZE / body_size) * body_size)
863 #define FIT_ARENAn(count,body_size) \
864 ( count * body_size <= PERL_ARENA_SIZE) \
865 ? count * body_size \
866 : FIT_ARENA0 (body_size)
867 #define FIT_ARENA(count,body_size) \
869 ? FIT_ARENAn (count, body_size) \
870 : FIT_ARENA0 (body_size)
872 /* Calculate the length to copy. Specifically work out the length less any
873 final padding the compiler needed to add. See the comment in sv_upgrade
874 for why copying the padding proved to be a bug. */
876 #define copy_length(type, last_member) \
877 STRUCT_OFFSET(type, last_member) \
878 + sizeof (((type*)SvANY((const SV *)0))->last_member)
880 static const struct body_details bodies_by_type[] = {
881 /* HEs use this offset for their arena. */
882 { 0, 0, 0, SVt_NULL, FALSE, NONV, NOARENA, 0 },
884 /* The bind placeholder pretends to be an RV for now.
885 Also it's marked as "can't upgrade" to stop anyone using it before it's
887 { 0, 0, 0, SVt_BIND, TRUE, NONV, NOARENA, 0 },
889 /* IVs are in the head, so the allocation size is 0. */
891 sizeof(IV), /* This is used to copy out the IV body. */
892 STRUCT_OFFSET(XPVIV, xiv_iv), SVt_IV, FALSE, NONV,
893 NOARENA /* IVS don't need an arena */, 0
896 { sizeof(NV), sizeof(NV),
897 STRUCT_OFFSET(XPVNV, xnv_u),
898 SVt_NV, FALSE, HADNV, HASARENA, FIT_ARENA(0, sizeof(NV)) },
900 { sizeof(XPV) - STRUCT_OFFSET(XPV, xpv_cur),
901 copy_length(XPV, xpv_len) - STRUCT_OFFSET(XPV, xpv_cur),
902 + STRUCT_OFFSET(XPV, xpv_cur),
903 SVt_PV, FALSE, NONV, HASARENA,
904 FIT_ARENA(0, sizeof(XPV) - STRUCT_OFFSET(XPV, xpv_cur)) },
906 { sizeof(XPVIV) - STRUCT_OFFSET(XPV, xpv_cur),
907 copy_length(XPVIV, xiv_u) - STRUCT_OFFSET(XPV, xpv_cur),
908 + STRUCT_OFFSET(XPV, xpv_cur),
909 SVt_PVIV, FALSE, NONV, HASARENA,
910 FIT_ARENA(0, sizeof(XPVIV) - STRUCT_OFFSET(XPV, xpv_cur)) },
912 { sizeof(XPVNV) - STRUCT_OFFSET(XPV, xpv_cur),
913 copy_length(XPVNV, xnv_u) - STRUCT_OFFSET(XPV, xpv_cur),
914 + STRUCT_OFFSET(XPV, xpv_cur),
915 SVt_PVNV, FALSE, HADNV, HASARENA,
916 FIT_ARENA(0, sizeof(XPVNV) - STRUCT_OFFSET(XPV, xpv_cur)) },
918 { sizeof(XPVMG), copy_length(XPVMG, xnv_u), 0, SVt_PVMG, FALSE, HADNV,
919 HASARENA, FIT_ARENA(0, sizeof(XPVMG)) },
924 SVt_REGEXP, FALSE, NONV, HASARENA,
925 FIT_ARENA(0, sizeof(regexp))
928 { sizeof(XPVGV), sizeof(XPVGV), 0, SVt_PVGV, TRUE, HADNV,
929 HASARENA, FIT_ARENA(0, sizeof(XPVGV)) },
931 { sizeof(XPVLV), sizeof(XPVLV), 0, SVt_PVLV, TRUE, HADNV,
932 HASARENA, FIT_ARENA(0, sizeof(XPVLV)) },
935 copy_length(XPVAV, xav_alloc),
937 SVt_PVAV, TRUE, NONV, HASARENA,
938 FIT_ARENA(0, sizeof(XPVAV)) },
941 copy_length(XPVHV, xhv_max),
943 SVt_PVHV, TRUE, NONV, HASARENA,
944 FIT_ARENA(0, sizeof(XPVHV)) },
949 SVt_PVCV, TRUE, NONV, HASARENA,
950 FIT_ARENA(0, sizeof(XPVCV)) },
955 SVt_PVFM, TRUE, NONV, NOARENA,
956 FIT_ARENA(20, sizeof(XPVFM)) },
961 SVt_PVIO, TRUE, NONV, HASARENA,
962 FIT_ARENA(24, sizeof(XPVIO)) },
965 #define new_body_allocated(sv_type) \
966 (void *)((char *)S_new_body(aTHX_ sv_type) \
967 - bodies_by_type[sv_type].offset)
969 /* return a thing to the free list */
971 #define del_body(thing, root) \
973 void ** const thing_copy = (void **)thing; \
974 *thing_copy = *root; \
975 *root = (void*)thing_copy; \
980 #define new_XNV() safemalloc(sizeof(XPVNV))
981 #define new_XPVNV() safemalloc(sizeof(XPVNV))
982 #define new_XPVMG() safemalloc(sizeof(XPVMG))
984 #define del_XPVGV(p) safefree(p)
988 #define new_XNV() new_body_allocated(SVt_NV)
989 #define new_XPVNV() new_body_allocated(SVt_PVNV)
990 #define new_XPVMG() new_body_allocated(SVt_PVMG)
992 #define del_XPVGV(p) del_body(p + bodies_by_type[SVt_PVGV].offset, \
993 &PL_body_roots[SVt_PVGV])
997 /* no arena for you! */
999 #define new_NOARENA(details) \
1000 safemalloc((details)->body_size + (details)->offset)
1001 #define new_NOARENAZ(details) \
1002 safecalloc((details)->body_size + (details)->offset, 1)
1005 Perl_more_bodies (pTHX_ const svtype sv_type, const size_t body_size,
1006 const size_t arena_size)
1009 void ** const root = &PL_body_roots[sv_type];
1010 struct arena_desc *adesc;
1011 struct arena_set *aroot = (struct arena_set *) PL_body_arenas;
1015 const size_t good_arena_size = Perl_malloc_good_size(arena_size);
1016 #if defined(DEBUGGING) && !defined(PERL_GLOBAL_STRUCT_PRIVATE)
1017 static bool done_sanity_check;
1019 /* PERL_GLOBAL_STRUCT_PRIVATE cannot coexist with global
1020 * variables like done_sanity_check. */
1021 if (!done_sanity_check) {
1022 unsigned int i = SVt_LAST;
1024 done_sanity_check = TRUE;
1027 assert (bodies_by_type[i].type == i);
1033 /* may need new arena-set to hold new arena */
1034 if (!aroot || aroot->curr >= aroot->set_size) {
1035 struct arena_set *newroot;
1036 Newxz(newroot, 1, struct arena_set);
1037 newroot->set_size = ARENAS_PER_SET;
1038 newroot->next = aroot;
1040 PL_body_arenas = (void *) newroot;
1041 DEBUG_m(PerlIO_printf(Perl_debug_log, "new arenaset %p\n", (void*)aroot));
1044 /* ok, now have arena-set with at least 1 empty/available arena-desc */
1045 curr = aroot->curr++;
1046 adesc = &(aroot->set[curr]);
1047 assert(!adesc->arena);
1049 Newx(adesc->arena, good_arena_size, char);
1050 adesc->size = good_arena_size;
1051 adesc->utype = sv_type;
1052 DEBUG_m(PerlIO_printf(Perl_debug_log, "arena %d added: %p size %"UVuf"\n",
1053 curr, (void*)adesc->arena, (UV)good_arena_size));
1055 start = (char *) adesc->arena;
1057 /* Get the address of the byte after the end of the last body we can fit.
1058 Remember, this is integer division: */
1059 end = start + good_arena_size / body_size * body_size;
1061 /* computed count doesn't reflect the 1st slot reservation */
1062 #if defined(MYMALLOC) || defined(HAS_MALLOC_GOOD_SIZE)
1063 DEBUG_m(PerlIO_printf(Perl_debug_log,
1064 "arena %p end %p arena-size %d (from %d) type %d "
1066 (void*)start, (void*)end, (int)good_arena_size,
1067 (int)arena_size, sv_type, (int)body_size,
1068 (int)good_arena_size / (int)body_size));
1070 DEBUG_m(PerlIO_printf(Perl_debug_log,
1071 "arena %p end %p arena-size %d type %d size %d ct %d\n",
1072 (void*)start, (void*)end,
1073 (int)arena_size, sv_type, (int)body_size,
1074 (int)good_arena_size / (int)body_size));
1076 *root = (void *)start;
1079 /* Where the next body would start: */
1080 char * const next = start + body_size;
1083 /* This is the last body: */
1084 assert(next == end);
1086 *(void **)start = 0;
1090 *(void**) start = (void *)next;
1095 /* grab a new thing from the free list, allocating more if necessary.
1096 The inline version is used for speed in hot routines, and the
1097 function using it serves the rest (unless PURIFY).
1099 #define new_body_inline(xpv, sv_type) \
1101 void ** const r3wt = &PL_body_roots[sv_type]; \
1102 xpv = (PTR_TBL_ENT_t*) (*((void **)(r3wt)) \
1103 ? *((void **)(r3wt)) : Perl_more_bodies(aTHX_ sv_type, \
1104 bodies_by_type[sv_type].body_size,\
1105 bodies_by_type[sv_type].arena_size)); \
1106 *(r3wt) = *(void**)(xpv); \
1112 S_new_body(pTHX_ const svtype sv_type)
1116 new_body_inline(xpv, sv_type);
1122 static const struct body_details fake_rv =
1123 { 0, 0, 0, SVt_IV, FALSE, NONV, NOARENA, 0 };
1126 =for apidoc sv_upgrade
1128 Upgrade an SV to a more complex form. Generally adds a new body type to the
1129 SV, then copies across as much information as possible from the old body.
1130 It croaks if the SV is already in a more complex form than requested. You
1131 generally want to use the C<SvUPGRADE> macro wrapper, which checks the type
1132 before calling C<sv_upgrade>, and hence does not croak. See also
1139 Perl_sv_upgrade(pTHX_ register SV *const sv, svtype new_type)
1144 const svtype old_type = SvTYPE(sv);
1145 const struct body_details *new_type_details;
1146 const struct body_details *old_type_details
1147 = bodies_by_type + old_type;
1148 SV *referant = NULL;
1150 PERL_ARGS_ASSERT_SV_UPGRADE;
1152 if (old_type == new_type)
1155 /* This clause was purposefully added ahead of the early return above to
1156 the shared string hackery for (sort {$a <=> $b} keys %hash), with the
1157 inference by Nick I-S that it would fix other troublesome cases. See
1158 changes 7162, 7163 (f130fd4589cf5fbb24149cd4db4137c8326f49c1 and parent)
1160 Given that shared hash key scalars are no longer PVIV, but PV, there is
1161 no longer need to unshare so as to free up the IVX slot for its proper
1162 purpose. So it's safe to move the early return earlier. */
1164 if (new_type != SVt_PV && SvIsCOW(sv)) {
1165 sv_force_normal_flags(sv, 0);
1168 old_body = SvANY(sv);
1170 /* Copying structures onto other structures that have been neatly zeroed
1171 has a subtle gotcha. Consider XPVMG
1173 +------+------+------+------+------+-------+-------+
1174 | NV | CUR | LEN | IV | MAGIC | STASH |
1175 +------+------+------+------+------+-------+-------+
1176 0 4 8 12 16 20 24 28
1178 where NVs are aligned to 8 bytes, so that sizeof that structure is
1179 actually 32 bytes long, with 4 bytes of padding at the end:
1181 +------+------+------+------+------+-------+-------+------+
1182 | NV | CUR | LEN | IV | MAGIC | STASH | ??? |
1183 +------+------+------+------+------+-------+-------+------+
1184 0 4 8 12 16 20 24 28 32
1186 so what happens if you allocate memory for this structure:
1188 +------+------+------+------+------+-------+-------+------+------+...
1189 | NV | CUR | LEN | IV | MAGIC | STASH | GP | NAME |
1190 +------+------+------+------+------+-------+-------+------+------+...
1191 0 4 8 12 16 20 24 28 32 36
1193 zero it, then copy sizeof(XPVMG) bytes on top of it? Not quite what you
1194 expect, because you copy the area marked ??? onto GP. Now, ??? may have
1195 started out as zero once, but it's quite possible that it isn't. So now,
1196 rather than a nicely zeroed GP, you have it pointing somewhere random.
1199 (In fact, GP ends up pointing at a previous GP structure, because the
1200 principle cause of the padding in XPVMG getting garbage is a copy of
1201 sizeof(XPVMG) bytes from a XPVGV structure in sv_unglob. Right now
1202 this happens to be moot because XPVGV has been re-ordered, with GP
1203 no longer after STASH)
1205 So we are careful and work out the size of used parts of all the
1213 referant = SvRV(sv);
1214 old_type_details = &fake_rv;
1215 if (new_type == SVt_NV)
1216 new_type = SVt_PVNV;
1218 if (new_type < SVt_PVIV) {
1219 new_type = (new_type == SVt_NV)
1220 ? SVt_PVNV : SVt_PVIV;
1225 if (new_type < SVt_PVNV) {
1226 new_type = SVt_PVNV;
1230 assert(new_type > SVt_PV);
1231 assert(SVt_IV < SVt_PV);
1232 assert(SVt_NV < SVt_PV);
1239 /* Because the XPVMG of PL_mess_sv isn't allocated from the arena,
1240 there's no way that it can be safely upgraded, because perl.c
1241 expects to Safefree(SvANY(PL_mess_sv)) */
1242 assert(sv != PL_mess_sv);
1243 /* This flag bit is used to mean other things in other scalar types.
1244 Given that it only has meaning inside the pad, it shouldn't be set
1245 on anything that can get upgraded. */
1246 assert(!SvPAD_TYPED(sv));
1249 if (old_type_details->cant_upgrade)
1250 Perl_croak(aTHX_ "Can't upgrade %s (%" UVuf ") to %" UVuf,
1251 sv_reftype(sv, 0), (UV) old_type, (UV) new_type);
1254 if (old_type > new_type)
1255 Perl_croak(aTHX_ "sv_upgrade from type %d down to type %d",
1256 (int)old_type, (int)new_type);
1258 new_type_details = bodies_by_type + new_type;
1260 SvFLAGS(sv) &= ~SVTYPEMASK;
1261 SvFLAGS(sv) |= new_type;
1263 /* This can't happen, as SVt_NULL is <= all values of new_type, so one of
1264 the return statements above will have triggered. */
1265 assert (new_type != SVt_NULL);
1268 assert(old_type == SVt_NULL);
1269 SvANY(sv) = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
1273 assert(old_type == SVt_NULL);
1274 SvANY(sv) = new_XNV();
1279 assert(new_type_details->body_size);
1282 assert(new_type_details->arena);
1283 assert(new_type_details->arena_size);
1284 /* This points to the start of the allocated area. */
1285 new_body_inline(new_body, new_type);
1286 Zero(new_body, new_type_details->body_size, char);
1287 new_body = ((char *)new_body) - new_type_details->offset;
1289 /* We always allocated the full length item with PURIFY. To do this
1290 we fake things so that arena is false for all 16 types.. */
1291 new_body = new_NOARENAZ(new_type_details);
1293 SvANY(sv) = new_body;
1294 if (new_type == SVt_PVAV) {
1298 if (old_type_details->body_size) {
1301 /* It will have been zeroed when the new body was allocated.
1302 Lets not write to it, in case it confuses a write-back
1308 #ifndef NODEFAULT_SHAREKEYS
1309 HvSHAREKEYS_on(sv); /* key-sharing on by default */
1311 HvMAX(sv) = 7; /* (start with 8 buckets) */
1314 /* SVt_NULL isn't the only thing upgraded to AV or HV.
1315 The target created by newSVrv also is, and it can have magic.
1316 However, it never has SvPVX set.
1318 if (old_type == SVt_IV) {
1320 } else if (old_type >= SVt_PV) {
1321 assert(SvPVX_const(sv) == 0);
1324 if (old_type >= SVt_PVMG) {
1325 SvMAGIC_set(sv, ((XPVMG*)old_body)->xmg_u.xmg_magic);
1326 SvSTASH_set(sv, ((XPVMG*)old_body)->xmg_stash);
1328 sv->sv_u.svu_array = NULL; /* or svu_hash */
1334 /* This ensures that SvTHINKFIRST(sv) is true, and hence that
1335 sv_force_normal_flags(sv) is called. */
1338 /* XXX Is this still needed? Was it ever needed? Surely as there is
1339 no route from NV to PVIV, NOK can never be true */
1340 assert(!SvNOKp(sv));
1351 assert(new_type_details->body_size);
1352 /* We always allocated the full length item with PURIFY. To do this
1353 we fake things so that arena is false for all 16 types.. */
1354 if(new_type_details->arena) {
1355 /* This points to the start of the allocated area. */
1356 new_body_inline(new_body, new_type);
1357 Zero(new_body, new_type_details->body_size, char);
1358 new_body = ((char *)new_body) - new_type_details->offset;
1360 new_body = new_NOARENAZ(new_type_details);
1362 SvANY(sv) = new_body;
1364 if (old_type_details->copy) {
1365 /* There is now the potential for an upgrade from something without
1366 an offset (PVNV or PVMG) to something with one (PVCV, PVFM) */
1367 int offset = old_type_details->offset;
1368 int length = old_type_details->copy;
1370 if (new_type_details->offset > old_type_details->offset) {
1371 const int difference
1372 = new_type_details->offset - old_type_details->offset;
1373 offset += difference;
1374 length -= difference;
1376 assert (length >= 0);
1378 Copy((char *)old_body + offset, (char *)new_body + offset, length,
1382 #ifndef NV_ZERO_IS_ALLBITS_ZERO
1383 /* If NV 0.0 is stores as all bits 0 then Zero() already creates a
1384 * correct 0.0 for us. Otherwise, if the old body didn't have an
1385 * NV slot, but the new one does, then we need to initialise the
1386 * freshly created NV slot with whatever the correct bit pattern is
1388 if (old_type_details->zero_nv && !new_type_details->zero_nv
1389 && !isGV_with_GP(sv))
1393 if (new_type == SVt_PVIO) {
1394 IO * const io = MUTABLE_IO(sv);
1395 GV *iogv = gv_fetchpvs("IO::File::", GV_ADD, SVt_PVHV);
1398 /* Clear the stashcache because a new IO could overrule a package
1400 hv_clear(PL_stashcache);
1402 SvSTASH_set(io, MUTABLE_HV(SvREFCNT_inc(GvHV(iogv))));
1403 IoPAGE_LEN(sv) = 60;
1405 if (old_type < SVt_PV) {
1406 /* referant will be NULL unless the old type was SVt_IV emulating
1408 sv->sv_u.svu_rv = referant;
1412 Perl_croak(aTHX_ "panic: sv_upgrade to unknown type %lu",
1413 (unsigned long)new_type);
1416 if (old_type > SVt_IV) {
1420 /* Note that there is an assumption that all bodies of types that
1421 can be upgraded came from arenas. Only the more complex non-
1422 upgradable types are allowed to be directly malloc()ed. */
1423 assert(old_type_details->arena);
1424 del_body((void*)((char*)old_body + old_type_details->offset),
1425 &PL_body_roots[old_type]);
1431 =for apidoc sv_backoff
1433 Remove any string offset. You should normally use the C<SvOOK_off> macro
1440 Perl_sv_backoff(pTHX_ register SV *const sv)
1443 const char * const s = SvPVX_const(sv);
1445 PERL_ARGS_ASSERT_SV_BACKOFF;
1446 PERL_UNUSED_CONTEXT;
1449 assert(SvTYPE(sv) != SVt_PVHV);
1450 assert(SvTYPE(sv) != SVt_PVAV);
1452 SvOOK_offset(sv, delta);
1454 SvLEN_set(sv, SvLEN(sv) + delta);
1455 SvPV_set(sv, SvPVX(sv) - delta);
1456 Move(s, SvPVX(sv), SvCUR(sv)+1, char);
1457 SvFLAGS(sv) &= ~SVf_OOK;
1464 Expands the character buffer in the SV. If necessary, uses C<sv_unref> and
1465 upgrades the SV to C<SVt_PV>. Returns a pointer to the character buffer.
1466 Use the C<SvGROW> wrapper instead.
1472 Perl_sv_grow(pTHX_ register SV *const sv, register STRLEN newlen)
1476 PERL_ARGS_ASSERT_SV_GROW;
1478 if (PL_madskills && newlen >= 0x100000) {
1479 PerlIO_printf(Perl_debug_log,
1480 "Allocation too large: %"UVxf"\n", (UV)newlen);
1482 #ifdef HAS_64K_LIMIT
1483 if (newlen >= 0x10000) {
1484 PerlIO_printf(Perl_debug_log,
1485 "Allocation too large: %"UVxf"\n", (UV)newlen);
1488 #endif /* HAS_64K_LIMIT */
1491 if (SvTYPE(sv) < SVt_PV) {
1492 sv_upgrade(sv, SVt_PV);
1493 s = SvPVX_mutable(sv);
1495 else if (SvOOK(sv)) { /* pv is offset? */
1497 s = SvPVX_mutable(sv);
1498 if (newlen > SvLEN(sv))
1499 newlen += 10 * (newlen - SvCUR(sv)); /* avoid copy each time */
1500 #ifdef HAS_64K_LIMIT
1501 if (newlen >= 0x10000)
1506 s = SvPVX_mutable(sv);
1508 if (newlen > SvLEN(sv)) { /* need more room? */
1509 STRLEN minlen = SvCUR(sv);
1510 minlen += (minlen >> PERL_STRLEN_EXPAND_SHIFT) + 10;
1511 if (newlen < minlen)
1513 #ifndef Perl_safesysmalloc_size
1514 newlen = PERL_STRLEN_ROUNDUP(newlen);
1516 if (SvLEN(sv) && s) {
1517 s = (char*)saferealloc(s, newlen);
1520 s = (char*)safemalloc(newlen);
1521 if (SvPVX_const(sv) && SvCUR(sv)) {
1522 Move(SvPVX_const(sv), s, (newlen < SvCUR(sv)) ? newlen : SvCUR(sv), char);
1526 #ifdef Perl_safesysmalloc_size
1527 /* Do this here, do it once, do it right, and then we will never get
1528 called back into sv_grow() unless there really is some growing
1530 SvLEN_set(sv, Perl_safesysmalloc_size(s));
1532 SvLEN_set(sv, newlen);
1539 =for apidoc sv_setiv
1541 Copies an integer into the given SV, upgrading first if necessary.
1542 Does not handle 'set' magic. See also C<sv_setiv_mg>.
1548 Perl_sv_setiv(pTHX_ register SV *const sv, const IV i)
1552 PERL_ARGS_ASSERT_SV_SETIV;
1554 SV_CHECK_THINKFIRST_COW_DROP(sv);
1555 switch (SvTYPE(sv)) {
1558 sv_upgrade(sv, SVt_IV);
1561 sv_upgrade(sv, SVt_PVIV);
1565 if (!isGV_with_GP(sv))
1572 /* diag_listed_as: Can't coerce %s to %s in %s */
1573 Perl_croak(aTHX_ "Can't coerce %s to integer in %s", sv_reftype(sv,0),
1577 (void)SvIOK_only(sv); /* validate number */
1583 =for apidoc sv_setiv_mg
1585 Like C<sv_setiv>, but also handles 'set' magic.
1591 Perl_sv_setiv_mg(pTHX_ register SV *const sv, const IV i)
1593 PERL_ARGS_ASSERT_SV_SETIV_MG;
1600 =for apidoc sv_setuv
1602 Copies an unsigned integer into the given SV, upgrading first if necessary.
1603 Does not handle 'set' magic. See also C<sv_setuv_mg>.
1609 Perl_sv_setuv(pTHX_ register SV *const sv, const UV u)
1611 PERL_ARGS_ASSERT_SV_SETUV;
1613 /* With the if statement to ensure that integers are stored as IVs whenever
1615 u=1.49 s=0.52 cu=72.49 cs=10.64 scripts=270 tests=20865
1618 u=1.35 s=0.47 cu=73.45 cs=11.43 scripts=270 tests=20865
1620 If you wish to remove the following if statement, so that this routine
1621 (and its callers) always return UVs, please benchmark to see what the
1622 effect is. Modern CPUs may be different. Or may not :-)
1624 if (u <= (UV)IV_MAX) {
1625 sv_setiv(sv, (IV)u);
1634 =for apidoc sv_setuv_mg
1636 Like C<sv_setuv>, but also handles 'set' magic.
1642 Perl_sv_setuv_mg(pTHX_ register SV *const sv, const UV u)
1644 PERL_ARGS_ASSERT_SV_SETUV_MG;
1651 =for apidoc sv_setnv
1653 Copies a double into the given SV, upgrading first if necessary.
1654 Does not handle 'set' magic. See also C<sv_setnv_mg>.
1660 Perl_sv_setnv(pTHX_ register SV *const sv, const NV num)
1664 PERL_ARGS_ASSERT_SV_SETNV;
1666 SV_CHECK_THINKFIRST_COW_DROP(sv);
1667 switch (SvTYPE(sv)) {
1670 sv_upgrade(sv, SVt_NV);
1674 sv_upgrade(sv, SVt_PVNV);
1678 if (!isGV_with_GP(sv))
1685 /* diag_listed_as: Can't coerce %s to %s in %s */
1686 Perl_croak(aTHX_ "Can't coerce %s to number in %s", sv_reftype(sv,0),
1691 (void)SvNOK_only(sv); /* validate number */
1696 =for apidoc sv_setnv_mg
1698 Like C<sv_setnv>, but also handles 'set' magic.
1704 Perl_sv_setnv_mg(pTHX_ register SV *const sv, const NV num)
1706 PERL_ARGS_ASSERT_SV_SETNV_MG;
1712 /* Print an "isn't numeric" warning, using a cleaned-up,
1713 * printable version of the offending string
1717 S_not_a_number(pTHX_ SV *const sv)
1724 PERL_ARGS_ASSERT_NOT_A_NUMBER;
1727 dsv = newSVpvs_flags("", SVs_TEMP);
1728 pv = sv_uni_display(dsv, sv, 10, UNI_DISPLAY_ISPRINT);
1731 const char * const limit = tmpbuf + sizeof(tmpbuf) - 8;
1732 /* each *s can expand to 4 chars + "...\0",
1733 i.e. need room for 8 chars */
1735 const char *s = SvPVX_const(sv);
1736 const char * const end = s + SvCUR(sv);
1737 for ( ; s < end && d < limit; s++ ) {
1739 if (ch & 128 && !isPRINT_LC(ch)) {
1748 else if (ch == '\r') {
1752 else if (ch == '\f') {
1756 else if (ch == '\\') {
1760 else if (ch == '\0') {
1764 else if (isPRINT_LC(ch))
1781 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
1782 /* diag_listed_as: Argument "%s" isn't numeric%s */
1783 "Argument \"%s\" isn't numeric in %s", pv,
1786 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
1787 /* diag_listed_as: Argument "%s" isn't numeric%s */
1788 "Argument \"%s\" isn't numeric", pv);
1792 =for apidoc looks_like_number
1794 Test if the content of an SV looks like a number (or is a number).
1795 C<Inf> and C<Infinity> are treated as numbers (so will not issue a
1796 non-numeric warning), even if your atof() doesn't grok them. Get-magic is
1803 Perl_looks_like_number(pTHX_ SV *const sv)
1805 register const char *sbegin;
1808 PERL_ARGS_ASSERT_LOOKS_LIKE_NUMBER;
1810 if (SvPOK(sv) || SvPOKp(sv)) {
1811 sbegin = SvPV_nomg_const(sv, len);
1814 return SvFLAGS(sv) & (SVf_NOK|SVp_NOK|SVf_IOK|SVp_IOK);
1815 return grok_number(sbegin, len, NULL);
1819 S_glob_2number(pTHX_ GV * const gv)
1821 SV *const buffer = sv_newmortal();
1823 PERL_ARGS_ASSERT_GLOB_2NUMBER;
1825 gv_efullname3(buffer, gv, "*");
1827 /* We know that all GVs stringify to something that is not-a-number,
1828 so no need to test that. */
1829 if (ckWARN(WARN_NUMERIC))
1830 not_a_number(buffer);
1831 /* We just want something true to return, so that S_sv_2iuv_common
1832 can tail call us and return true. */
1836 /* Actually, ISO C leaves conversion of UV to IV undefined, but
1837 until proven guilty, assume that things are not that bad... */
1842 As 64 bit platforms often have an NV that doesn't preserve all bits of
1843 an IV (an assumption perl has been based on to date) it becomes necessary
1844 to remove the assumption that the NV always carries enough precision to
1845 recreate the IV whenever needed, and that the NV is the canonical form.
1846 Instead, IV/UV and NV need to be given equal rights. So as to not lose
1847 precision as a side effect of conversion (which would lead to insanity
1848 and the dragon(s) in t/op/numconvert.t getting very angry) the intent is
1849 1) to distinguish between IV/UV/NV slots that have cached a valid
1850 conversion where precision was lost and IV/UV/NV slots that have a
1851 valid conversion which has lost no precision
1852 2) to ensure that if a numeric conversion to one form is requested that
1853 would lose precision, the precise conversion (or differently
1854 imprecise conversion) is also performed and cached, to prevent
1855 requests for different numeric formats on the same SV causing
1856 lossy conversion chains. (lossless conversion chains are perfectly
1861 SvIOKp is true if the IV slot contains a valid value
1862 SvIOK is true only if the IV value is accurate (UV if SvIOK_UV true)
1863 SvNOKp is true if the NV slot contains a valid value
1864 SvNOK is true only if the NV value is accurate
1867 while converting from PV to NV, check to see if converting that NV to an
1868 IV(or UV) would lose accuracy over a direct conversion from PV to
1869 IV(or UV). If it would, cache both conversions, return NV, but mark
1870 SV as IOK NOKp (ie not NOK).
1872 While converting from PV to IV, check to see if converting that IV to an
1873 NV would lose accuracy over a direct conversion from PV to NV. If it
1874 would, cache both conversions, flag similarly.
1876 Before, the SV value "3.2" could become NV=3.2 IV=3 NOK, IOK quite
1877 correctly because if IV & NV were set NV *always* overruled.
1878 Now, "3.2" will become NV=3.2 IV=3 NOK, IOKp, because the flag's meaning
1879 changes - now IV and NV together means that the two are interchangeable:
1880 SvIVX == (IV) SvNVX && SvNVX == (NV) SvIVX;
1882 The benefit of this is that operations such as pp_add know that if
1883 SvIOK is true for both left and right operands, then integer addition
1884 can be used instead of floating point (for cases where the result won't
1885 overflow). Before, floating point was always used, which could lead to
1886 loss of precision compared with integer addition.
1888 * making IV and NV equal status should make maths accurate on 64 bit
1890 * may speed up maths somewhat if pp_add and friends start to use
1891 integers when possible instead of fp. (Hopefully the overhead in
1892 looking for SvIOK and checking for overflow will not outweigh the
1893 fp to integer speedup)
1894 * will slow down integer operations (callers of SvIV) on "inaccurate"
1895 values, as the change from SvIOK to SvIOKp will cause a call into
1896 sv_2iv each time rather than a macro access direct to the IV slot
1897 * should speed up number->string conversion on integers as IV is
1898 favoured when IV and NV are equally accurate
1900 ####################################################################
1901 You had better be using SvIOK_notUV if you want an IV for arithmetic:
1902 SvIOK is true if (IV or UV), so you might be getting (IV)SvUV.
1903 On the other hand, SvUOK is true iff UV.
1904 ####################################################################
1906 Your mileage will vary depending your CPU's relative fp to integer
1910 #ifndef NV_PRESERVES_UV
1911 # define IS_NUMBER_UNDERFLOW_IV 1
1912 # define IS_NUMBER_UNDERFLOW_UV 2
1913 # define IS_NUMBER_IV_AND_UV 2
1914 # define IS_NUMBER_OVERFLOW_IV 4
1915 # define IS_NUMBER_OVERFLOW_UV 5
1917 /* sv_2iuv_non_preserve(): private routine for use by sv_2iv() and sv_2uv() */
1919 /* For sv_2nv these three cases are "SvNOK and don't bother casting" */
1921 S_sv_2iuv_non_preserve(pTHX_ register SV *const sv
1929 PERL_ARGS_ASSERT_SV_2IUV_NON_PRESERVE;
1931 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_2iuv_non '%s', IV=0x%"UVxf" NV=%"NVgf" inttype=%"UVXf"\n", SvPVX_const(sv), SvIVX(sv), SvNVX(sv), (UV)numtype));
1932 if (SvNVX(sv) < (NV)IV_MIN) {
1933 (void)SvIOKp_on(sv);
1935 SvIV_set(sv, IV_MIN);
1936 return IS_NUMBER_UNDERFLOW_IV;
1938 if (SvNVX(sv) > (NV)UV_MAX) {
1939 (void)SvIOKp_on(sv);
1942 SvUV_set(sv, UV_MAX);
1943 return IS_NUMBER_OVERFLOW_UV;
1945 (void)SvIOKp_on(sv);
1947 /* Can't use strtol etc to convert this string. (See truth table in
1949 if (SvNVX(sv) <= (UV)IV_MAX) {
1950 SvIV_set(sv, I_V(SvNVX(sv)));
1951 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
1952 SvIOK_on(sv); /* Integer is precise. NOK, IOK */
1954 /* Integer is imprecise. NOK, IOKp */
1956 return SvNVX(sv) < 0 ? IS_NUMBER_UNDERFLOW_UV : IS_NUMBER_IV_AND_UV;
1959 SvUV_set(sv, U_V(SvNVX(sv)));
1960 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
1961 if (SvUVX(sv) == UV_MAX) {
1962 /* As we know that NVs don't preserve UVs, UV_MAX cannot
1963 possibly be preserved by NV. Hence, it must be overflow.
1965 return IS_NUMBER_OVERFLOW_UV;
1967 SvIOK_on(sv); /* Integer is precise. NOK, UOK */
1969 /* Integer is imprecise. NOK, IOKp */
1971 return IS_NUMBER_OVERFLOW_IV;
1973 #endif /* !NV_PRESERVES_UV*/
1976 S_sv_2iuv_common(pTHX_ SV *const sv)
1980 PERL_ARGS_ASSERT_SV_2IUV_COMMON;
1983 /* erm. not sure. *should* never get NOKp (without NOK) from sv_2nv
1984 * without also getting a cached IV/UV from it at the same time
1985 * (ie PV->NV conversion should detect loss of accuracy and cache
1986 * IV or UV at same time to avoid this. */
1987 /* IV-over-UV optimisation - choose to cache IV if possible */
1989 if (SvTYPE(sv) == SVt_NV)
1990 sv_upgrade(sv, SVt_PVNV);
1992 (void)SvIOKp_on(sv); /* Must do this first, to clear any SvOOK */
1993 /* < not <= as for NV doesn't preserve UV, ((NV)IV_MAX+1) will almost
1994 certainly cast into the IV range at IV_MAX, whereas the correct
1995 answer is the UV IV_MAX +1. Hence < ensures that dodgy boundary
1997 #if defined(NAN_COMPARE_BROKEN) && defined(Perl_isnan)
1998 if (Perl_isnan(SvNVX(sv))) {
2004 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2005 SvIV_set(sv, I_V(SvNVX(sv)));
2006 if (SvNVX(sv) == (NV) SvIVX(sv)
2007 #ifndef NV_PRESERVES_UV
2008 && (((UV)1 << NV_PRESERVES_UV_BITS) >
2009 (UV)(SvIVX(sv) > 0 ? SvIVX(sv) : -SvIVX(sv)))
2010 /* Don't flag it as "accurately an integer" if the number
2011 came from a (by definition imprecise) NV operation, and
2012 we're outside the range of NV integer precision */
2016 SvIOK_on(sv); /* Can this go wrong with rounding? NWC */
2018 /* scalar has trailing garbage, eg "42a" */
2020 DEBUG_c(PerlIO_printf(Perl_debug_log,
2021 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (precise)\n",
2027 /* IV not precise. No need to convert from PV, as NV
2028 conversion would already have cached IV if it detected
2029 that PV->IV would be better than PV->NV->IV
2030 flags already correct - don't set public IOK. */
2031 DEBUG_c(PerlIO_printf(Perl_debug_log,
2032 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (imprecise)\n",
2037 /* Can the above go wrong if SvIVX == IV_MIN and SvNVX < IV_MIN,
2038 but the cast (NV)IV_MIN rounds to a the value less (more
2039 negative) than IV_MIN which happens to be equal to SvNVX ??
2040 Analogous to 0xFFFFFFFFFFFFFFFF rounding up to NV (2**64) and
2041 NV rounding back to 0xFFFFFFFFFFFFFFFF, so UVX == UV(NVX) and
2042 (NV)UVX == NVX are both true, but the values differ. :-(
2043 Hopefully for 2s complement IV_MIN is something like
2044 0x8000000000000000 which will be exact. NWC */
2047 SvUV_set(sv, U_V(SvNVX(sv)));
2049 (SvNVX(sv) == (NV) SvUVX(sv))
2050 #ifndef NV_PRESERVES_UV
2051 /* Make sure it's not 0xFFFFFFFFFFFFFFFF */
2052 /*&& (SvUVX(sv) != UV_MAX) irrelevant with code below */
2053 && (((UV)1 << NV_PRESERVES_UV_BITS) > SvUVX(sv))
2054 /* Don't flag it as "accurately an integer" if the number
2055 came from a (by definition imprecise) NV operation, and
2056 we're outside the range of NV integer precision */
2062 DEBUG_c(PerlIO_printf(Perl_debug_log,
2063 "0x%"UVxf" 2iv(%"UVuf" => %"IVdf") (as unsigned)\n",
2069 else if (SvPOKp(sv) && SvLEN(sv)) {
2071 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2072 /* We want to avoid a possible problem when we cache an IV/ a UV which
2073 may be later translated to an NV, and the resulting NV is not
2074 the same as the direct translation of the initial string
2075 (eg 123.456 can shortcut to the IV 123 with atol(), but we must
2076 be careful to ensure that the value with the .456 is around if the
2077 NV value is requested in the future).
2079 This means that if we cache such an IV/a UV, we need to cache the
2080 NV as well. Moreover, we trade speed for space, and do not
2081 cache the NV if we are sure it's not needed.
2084 /* SVt_PVNV is one higher than SVt_PVIV, hence this order */
2085 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2086 == IS_NUMBER_IN_UV) {
2087 /* It's definitely an integer, only upgrade to PVIV */
2088 if (SvTYPE(sv) < SVt_PVIV)
2089 sv_upgrade(sv, SVt_PVIV);
2091 } else if (SvTYPE(sv) < SVt_PVNV)
2092 sv_upgrade(sv, SVt_PVNV);
2094 /* If NVs preserve UVs then we only use the UV value if we know that
2095 we aren't going to call atof() below. If NVs don't preserve UVs
2096 then the value returned may have more precision than atof() will
2097 return, even though value isn't perfectly accurate. */
2098 if ((numtype & (IS_NUMBER_IN_UV
2099 #ifdef NV_PRESERVES_UV
2102 )) == IS_NUMBER_IN_UV) {
2103 /* This won't turn off the public IOK flag if it was set above */
2104 (void)SvIOKp_on(sv);
2106 if (!(numtype & IS_NUMBER_NEG)) {
2108 if (value <= (UV)IV_MAX) {
2109 SvIV_set(sv, (IV)value);
2111 /* it didn't overflow, and it was positive. */
2112 SvUV_set(sv, value);
2116 /* 2s complement assumption */
2117 if (value <= (UV)IV_MIN) {
2118 SvIV_set(sv, -(IV)value);
2120 /* Too negative for an IV. This is a double upgrade, but
2121 I'm assuming it will be rare. */
2122 if (SvTYPE(sv) < SVt_PVNV)
2123 sv_upgrade(sv, SVt_PVNV);
2127 SvNV_set(sv, -(NV)value);
2128 SvIV_set(sv, IV_MIN);
2132 /* For !NV_PRESERVES_UV and IS_NUMBER_IN_UV and IS_NUMBER_NOT_INT we
2133 will be in the previous block to set the IV slot, and the next
2134 block to set the NV slot. So no else here. */
2136 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2137 != IS_NUMBER_IN_UV) {
2138 /* It wasn't an (integer that doesn't overflow the UV). */
2139 SvNV_set(sv, Atof(SvPVX_const(sv)));
2141 if (! numtype && ckWARN(WARN_NUMERIC))
2144 #if defined(USE_LONG_DOUBLE)
2145 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%" PERL_PRIgldbl ")\n",
2146 PTR2UV(sv), SvNVX(sv)));
2148 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"NVgf")\n",
2149 PTR2UV(sv), SvNVX(sv)));
2152 #ifdef NV_PRESERVES_UV
2153 (void)SvIOKp_on(sv);
2155 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2156 SvIV_set(sv, I_V(SvNVX(sv)));
2157 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
2160 NOOP; /* Integer is imprecise. NOK, IOKp */
2162 /* UV will not work better than IV */
2164 if (SvNVX(sv) > (NV)UV_MAX) {
2166 /* Integer is inaccurate. NOK, IOKp, is UV */
2167 SvUV_set(sv, UV_MAX);
2169 SvUV_set(sv, U_V(SvNVX(sv)));
2170 /* 0xFFFFFFFFFFFFFFFF not an issue in here, NVs
2171 NV preservse UV so can do correct comparison. */
2172 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
2175 NOOP; /* Integer is imprecise. NOK, IOKp, is UV */
2180 #else /* NV_PRESERVES_UV */
2181 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2182 == (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT)) {
2183 /* The IV/UV slot will have been set from value returned by
2184 grok_number above. The NV slot has just been set using
2187 assert (SvIOKp(sv));
2189 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2190 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2191 /* Small enough to preserve all bits. */
2192 (void)SvIOKp_on(sv);
2194 SvIV_set(sv, I_V(SvNVX(sv)));
2195 if ((NV)(SvIVX(sv)) == SvNVX(sv))
2197 /* Assumption: first non-preserved integer is < IV_MAX,
2198 this NV is in the preserved range, therefore: */
2199 if (!(U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))
2201 Perl_croak(aTHX_ "sv_2iv assumed (U_V(fabs((double)SvNVX(sv))) < (UV)IV_MAX) but SvNVX(sv)=%"NVgf" U_V is 0x%"UVxf", IV_MAX is 0x%"UVxf"\n", SvNVX(sv), U_V(SvNVX(sv)), (UV)IV_MAX);
2205 0 0 already failed to read UV.
2206 0 1 already failed to read UV.
2207 1 0 you won't get here in this case. IV/UV
2208 slot set, public IOK, Atof() unneeded.
2209 1 1 already read UV.
2210 so there's no point in sv_2iuv_non_preserve() attempting
2211 to use atol, strtol, strtoul etc. */
2213 sv_2iuv_non_preserve (sv, numtype);
2215 sv_2iuv_non_preserve (sv);
2219 #endif /* NV_PRESERVES_UV */
2220 /* It might be more code efficient to go through the entire logic above
2221 and conditionally set with SvIOKp_on() rather than SvIOK(), but it
2222 gets complex and potentially buggy, so more programmer efficient
2223 to do it this way, by turning off the public flags: */
2225 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK);
2229 if (isGV_with_GP(sv))
2230 return glob_2number(MUTABLE_GV(sv));
2232 if (!SvPADTMP(sv)) {
2233 if (!PL_localizing && ckWARN(WARN_UNINITIALIZED))
2236 if (SvTYPE(sv) < SVt_IV)
2237 /* Typically the caller expects that sv_any is not NULL now. */
2238 sv_upgrade(sv, SVt_IV);
2239 /* Return 0 from the caller. */
2246 =for apidoc sv_2iv_flags
2248 Return the integer value of an SV, doing any necessary string
2249 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2250 Normally used via the C<SvIV(sv)> and C<SvIVx(sv)> macros.
2256 Perl_sv_2iv_flags(pTHX_ register SV *const sv, const I32 flags)
2261 if (SvGMAGICAL(sv) || SvVALID(sv)) {
2262 /* FBMs use the space for SvIVX and SvNVX for other purposes, and use
2263 the same flag bit as SVf_IVisUV, so must not let them cache IVs.
2264 In practice they are extremely unlikely to actually get anywhere
2265 accessible by user Perl code - the only way that I'm aware of is when
2266 a constant subroutine which is used as the second argument to index.
2268 if (flags & SV_GMAGIC)
2273 return I_V(SvNVX(sv));
2275 if (SvPOKp(sv) && SvLEN(sv)) {
2278 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2280 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2281 == IS_NUMBER_IN_UV) {
2282 /* It's definitely an integer */
2283 if (numtype & IS_NUMBER_NEG) {
2284 if (value < (UV)IV_MIN)
2287 if (value < (UV)IV_MAX)
2292 if (ckWARN(WARN_NUMERIC))
2295 return I_V(Atof(SvPVX_const(sv)));
2300 assert(SvTYPE(sv) >= SVt_PVMG);
2301 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
2302 } else if (SvTHINKFIRST(sv)) {
2307 if (flags & SV_SKIP_OVERLOAD)
2309 tmpstr = AMG_CALLunary(sv, numer_amg);
2310 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2311 return SvIV(tmpstr);
2314 return PTR2IV(SvRV(sv));
2317 sv_force_normal_flags(sv, 0);
2319 if (SvREADONLY(sv) && !SvOK(sv)) {
2320 if (ckWARN(WARN_UNINITIALIZED))
2326 if (S_sv_2iuv_common(aTHX_ sv))
2329 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"IVdf")\n",
2330 PTR2UV(sv),SvIVX(sv)));
2331 return SvIsUV(sv) ? (IV)SvUVX(sv) : SvIVX(sv);
2335 =for apidoc sv_2uv_flags
2337 Return the unsigned integer value of an SV, doing any necessary string
2338 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2339 Normally used via the C<SvUV(sv)> and C<SvUVx(sv)> macros.
2345 Perl_sv_2uv_flags(pTHX_ register SV *const sv, const I32 flags)
2350 if (SvGMAGICAL(sv) || SvVALID(sv)) {
2351 /* FBMs use the space for SvIVX and SvNVX for other purposes, and use
2352 the same flag bit as SVf_IVisUV, so must not let them cache IVs. */
2353 if (flags & SV_GMAGIC)
2358 return U_V(SvNVX(sv));
2359 if (SvPOKp(sv) && SvLEN(sv)) {
2362 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2364 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2365 == IS_NUMBER_IN_UV) {
2366 /* It's definitely an integer */
2367 if (!(numtype & IS_NUMBER_NEG))
2371 if (ckWARN(WARN_NUMERIC))
2374 return U_V(Atof(SvPVX_const(sv)));
2379 assert(SvTYPE(sv) >= SVt_PVMG);
2380 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
2381 } else if (SvTHINKFIRST(sv)) {
2386 if (flags & SV_SKIP_OVERLOAD)
2388 tmpstr = AMG_CALLunary(sv, numer_amg);
2389 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2390 return SvUV(tmpstr);
2393 return PTR2UV(SvRV(sv));
2396 sv_force_normal_flags(sv, 0);
2398 if (SvREADONLY(sv) && !SvOK(sv)) {
2399 if (ckWARN(WARN_UNINITIALIZED))
2405 if (S_sv_2iuv_common(aTHX_ sv))
2409 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2uv(%"UVuf")\n",
2410 PTR2UV(sv),SvUVX(sv)));
2411 return SvIsUV(sv) ? SvUVX(sv) : (UV)SvIVX(sv);
2415 =for apidoc sv_2nv_flags
2417 Return the num value of an SV, doing any necessary string or integer
2418 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2419 Normally used via the C<SvNV(sv)> and C<SvNVx(sv)> macros.
2425 Perl_sv_2nv_flags(pTHX_ register SV *const sv, const I32 flags)
2430 if (SvGMAGICAL(sv) || SvVALID(sv)) {
2431 /* FBMs use the space for SvIVX and SvNVX for other purposes, and use
2432 the same flag bit as SVf_IVisUV, so must not let them cache NVs. */
2433 if (flags & SV_GMAGIC)
2437 if ((SvPOKp(sv) && SvLEN(sv)) && !SvIOKp(sv)) {
2438 if (!SvIOKp(sv) && ckWARN(WARN_NUMERIC) &&
2439 !grok_number(SvPVX_const(sv), SvCUR(sv), NULL))
2441 return Atof(SvPVX_const(sv));
2445 return (NV)SvUVX(sv);
2447 return (NV)SvIVX(sv);
2452 assert(SvTYPE(sv) >= SVt_PVMG);
2453 /* This falls through to the report_uninit near the end of the
2455 } else if (SvTHINKFIRST(sv)) {
2460 if (flags & SV_SKIP_OVERLOAD)
2462 tmpstr = AMG_CALLunary(sv, numer_amg);
2463 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2464 return SvNV(tmpstr);
2467 return PTR2NV(SvRV(sv));
2470 sv_force_normal_flags(sv, 0);
2472 if (SvREADONLY(sv) && !SvOK(sv)) {
2473 if (ckWARN(WARN_UNINITIALIZED))
2478 if (SvTYPE(sv) < SVt_NV) {
2479 /* The logic to use SVt_PVNV if necessary is in sv_upgrade. */
2480 sv_upgrade(sv, SVt_NV);
2481 #ifdef USE_LONG_DOUBLE
2483 STORE_NUMERIC_LOCAL_SET_STANDARD();
2484 PerlIO_printf(Perl_debug_log,
2485 "0x%"UVxf" num(%" PERL_PRIgldbl ")\n",
2486 PTR2UV(sv), SvNVX(sv));
2487 RESTORE_NUMERIC_LOCAL();
2491 STORE_NUMERIC_LOCAL_SET_STANDARD();
2492 PerlIO_printf(Perl_debug_log, "0x%"UVxf" num(%"NVgf")\n",
2493 PTR2UV(sv), SvNVX(sv));
2494 RESTORE_NUMERIC_LOCAL();
2498 else if (SvTYPE(sv) < SVt_PVNV)
2499 sv_upgrade(sv, SVt_PVNV);
2504 SvNV_set(sv, SvIsUV(sv) ? (NV)SvUVX(sv) : (NV)SvIVX(sv));
2505 #ifdef NV_PRESERVES_UV
2511 /* Only set the public NV OK flag if this NV preserves the IV */
2512 /* Check it's not 0xFFFFFFFFFFFFFFFF */
2514 SvIsUV(sv) ? ((SvUVX(sv) != UV_MAX)&&(SvUVX(sv) == U_V(SvNVX(sv))))
2515 : (SvIVX(sv) == I_V(SvNVX(sv))))
2521 else if (SvPOKp(sv) && SvLEN(sv)) {
2523 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2524 if (!SvIOKp(sv) && !numtype && ckWARN(WARN_NUMERIC))
2526 #ifdef NV_PRESERVES_UV
2527 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2528 == IS_NUMBER_IN_UV) {
2529 /* It's definitely an integer */
2530 SvNV_set(sv, (numtype & IS_NUMBER_NEG) ? -(NV)value : (NV)value);
2532 SvNV_set(sv, Atof(SvPVX_const(sv)));
2538 SvNV_set(sv, Atof(SvPVX_const(sv)));
2539 /* Only set the public NV OK flag if this NV preserves the value in
2540 the PV at least as well as an IV/UV would.
2541 Not sure how to do this 100% reliably. */
2542 /* if that shift count is out of range then Configure's test is
2543 wonky. We shouldn't be in here with NV_PRESERVES_UV_BITS ==
2545 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2546 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2547 SvNOK_on(sv); /* Definitely small enough to preserve all bits */
2548 } else if (!(numtype & IS_NUMBER_IN_UV)) {
2549 /* Can't use strtol etc to convert this string, so don't try.
2550 sv_2iv and sv_2uv will use the NV to convert, not the PV. */
2553 /* value has been set. It may not be precise. */
2554 if ((numtype & IS_NUMBER_NEG) && (value > (UV)IV_MIN)) {
2555 /* 2s complement assumption for (UV)IV_MIN */
2556 SvNOK_on(sv); /* Integer is too negative. */
2561 if (numtype & IS_NUMBER_NEG) {
2562 SvIV_set(sv, -(IV)value);
2563 } else if (value <= (UV)IV_MAX) {
2564 SvIV_set(sv, (IV)value);
2566 SvUV_set(sv, value);
2570 if (numtype & IS_NUMBER_NOT_INT) {
2571 /* I believe that even if the original PV had decimals,
2572 they are lost beyond the limit of the FP precision.
2573 However, neither is canonical, so both only get p
2574 flags. NWC, 2000/11/25 */
2575 /* Both already have p flags, so do nothing */
2577 const NV nv = SvNVX(sv);
2578 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2579 if (SvIVX(sv) == I_V(nv)) {
2582 /* It had no "." so it must be integer. */
2586 /* between IV_MAX and NV(UV_MAX).
2587 Could be slightly > UV_MAX */
2589 if (numtype & IS_NUMBER_NOT_INT) {
2590 /* UV and NV both imprecise. */
2592 const UV nv_as_uv = U_V(nv);
2594 if (value == nv_as_uv && SvUVX(sv) != UV_MAX) {
2603 /* It might be more code efficient to go through the entire logic above
2604 and conditionally set with SvNOKp_on() rather than SvNOK(), but it
2605 gets complex and potentially buggy, so more programmer efficient
2606 to do it this way, by turning off the public flags: */
2608 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK);
2609 #endif /* NV_PRESERVES_UV */
2612 if (isGV_with_GP(sv)) {
2613 glob_2number(MUTABLE_GV(sv));
2617 if (!PL_localizing && !SvPADTMP(sv) && ckWARN(WARN_UNINITIALIZED))
2619 assert (SvTYPE(sv) >= SVt_NV);
2620 /* Typically the caller expects that sv_any is not NULL now. */
2621 /* XXX Ilya implies that this is a bug in callers that assume this
2622 and ideally should be fixed. */
2625 #if defined(USE_LONG_DOUBLE)
2627 STORE_NUMERIC_LOCAL_SET_STANDARD();
2628 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2nv(%" PERL_PRIgldbl ")\n",
2629 PTR2UV(sv), SvNVX(sv));
2630 RESTORE_NUMERIC_LOCAL();
2634 STORE_NUMERIC_LOCAL_SET_STANDARD();
2635 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 1nv(%"NVgf")\n",
2636 PTR2UV(sv), SvNVX(sv));
2637 RESTORE_NUMERIC_LOCAL();
2646 Return an SV with the numeric value of the source SV, doing any necessary
2647 reference or overload conversion. You must use the C<SvNUM(sv)> macro to
2648 access this function.
2654 Perl_sv_2num(pTHX_ register SV *const sv)
2656 PERL_ARGS_ASSERT_SV_2NUM;
2661 SV * const tmpsv = AMG_CALLunary(sv, numer_amg);
2662 TAINT_IF(tmpsv && SvTAINTED(tmpsv));
2663 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
2664 return sv_2num(tmpsv);
2666 return sv_2mortal(newSVuv(PTR2UV(SvRV(sv))));
2669 /* uiv_2buf(): private routine for use by sv_2pv_flags(): print an IV or
2670 * UV as a string towards the end of buf, and return pointers to start and
2673 * We assume that buf is at least TYPE_CHARS(UV) long.
2677 S_uiv_2buf(char *const buf, const IV iv, UV uv, const int is_uv, char **const peob)
2679 char *ptr = buf + TYPE_CHARS(UV);
2680 char * const ebuf = ptr;
2683 PERL_ARGS_ASSERT_UIV_2BUF;
2695 *--ptr = '0' + (char)(uv % 10);
2704 =for apidoc sv_2pv_flags
2706 Returns a pointer to the string value of an SV, and sets *lp to its length.
2707 If flags includes SV_GMAGIC, does an mg_get() first. Coerces sv to a
2708 string if necessary. Normally invoked via the C<SvPV_flags> macro.
2709 C<sv_2pv()> and C<sv_2pv_nomg> usually end up here too.
2715 Perl_sv_2pv_flags(pTHX_ register SV *const sv, STRLEN *const lp, const I32 flags)
2725 if (SvGMAGICAL(sv)) {
2726 if (flags & SV_GMAGIC)
2731 if (flags & SV_MUTABLE_RETURN)
2732 return SvPVX_mutable(sv);
2733 if (flags & SV_CONST_RETURN)
2734 return (char *)SvPVX_const(sv);
2737 if (SvIOKp(sv) || SvNOKp(sv)) {
2738 char tbuf[64]; /* Must fit sprintf/Gconvert of longest IV/NV */
2743 ? my_snprintf(tbuf, sizeof(tbuf), "%"UVuf, (UV)SvUVX(sv))
2744 : my_snprintf(tbuf, sizeof(tbuf), "%"IVdf, (IV)SvIVX(sv));
2745 } else if(SvNVX(sv) == 0.0) {
2750 Gconvert(SvNVX(sv), NV_DIG, 0, tbuf);
2757 SvUPGRADE(sv, SVt_PV);
2760 s = SvGROW_mutable(sv, len + 1);
2763 return (char*)memcpy(s, tbuf, len + 1);
2769 assert(SvTYPE(sv) >= SVt_PVMG);
2770 /* This falls through to the report_uninit near the end of the
2772 } else if (SvTHINKFIRST(sv)) {
2777 if (flags & SV_SKIP_OVERLOAD)
2779 tmpstr = AMG_CALLunary(sv, string_amg);
2780 TAINT_IF(tmpstr && SvTAINTED(tmpstr));
2781 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2783 /* char *pv = lp ? SvPV(tmpstr, *lp) : SvPV_nolen(tmpstr);
2787 if ((SvFLAGS(tmpstr) & (SVf_POK)) == SVf_POK) {
2788 if (flags & SV_CONST_RETURN) {
2789 pv = (char *) SvPVX_const(tmpstr);
2791 pv = (flags & SV_MUTABLE_RETURN)
2792 ? SvPVX_mutable(tmpstr) : SvPVX(tmpstr);
2795 *lp = SvCUR(tmpstr);
2797 pv = sv_2pv_flags(tmpstr, lp, flags);
2810 SV *const referent = SvRV(sv);
2814 retval = buffer = savepvn("NULLREF", len);
2815 } else if (SvTYPE(referent) == SVt_REGEXP && (
2816 !(PL_curcop->cop_hints & HINT_NO_AMAGIC)
2817 || amagic_is_enabled(string_amg)
2819 REGEXP * const re = (REGEXP *)MUTABLE_PTR(referent);
2824 /* If the regex is UTF-8 we want the containing scalar to
2825 have an UTF-8 flag too */
2831 if ((seen_evals = RX_SEEN_EVALS(re)))
2832 PL_reginterp_cnt += seen_evals;
2835 *lp = RX_WRAPLEN(re);
2837 return RX_WRAPPED(re);
2839 const char *const typestr = sv_reftype(referent, 0);
2840 const STRLEN typelen = strlen(typestr);
2841 UV addr = PTR2UV(referent);
2842 const char *stashname = NULL;
2843 STRLEN stashnamelen = 0; /* hush, gcc */
2844 const char *buffer_end;
2846 if (SvOBJECT(referent)) {
2847 const HEK *const name = HvNAME_HEK(SvSTASH(referent));
2850 stashname = HEK_KEY(name);
2851 stashnamelen = HEK_LEN(name);
2853 if (HEK_UTF8(name)) {
2859 stashname = "__ANON__";
2862 len = stashnamelen + 1 /* = */ + typelen + 3 /* (0x */
2863 + 2 * sizeof(UV) + 2 /* )\0 */;
2865 len = typelen + 3 /* (0x */
2866 + 2 * sizeof(UV) + 2 /* )\0 */;
2869 Newx(buffer, len, char);
2870 buffer_end = retval = buffer + len;
2872 /* Working backwards */
2876 *--retval = PL_hexdigit[addr & 15];
2877 } while (addr >>= 4);
2883 memcpy(retval, typestr, typelen);
2887 retval -= stashnamelen;
2888 memcpy(retval, stashname, stashnamelen);
2890 /* retval may not necessarily have reached the start of the
2892 assert (retval >= buffer);
2894 len = buffer_end - retval - 1; /* -1 for that \0 */
2902 if (SvREADONLY(sv) && !SvOK(sv)) {
2905 if (flags & SV_UNDEF_RETURNS_NULL)
2907 if (ckWARN(WARN_UNINITIALIZED))
2912 if (SvIOK(sv) || ((SvIOKp(sv) && !SvNOKp(sv)))) {
2913 /* I'm assuming that if both IV and NV are equally valid then
2914 converting the IV is going to be more efficient */
2915 const U32 isUIOK = SvIsUV(sv);
2916 char buf[TYPE_CHARS(UV)];
2920 if (SvTYPE(sv) < SVt_PVIV)
2921 sv_upgrade(sv, SVt_PVIV);
2922 ptr = uiv_2buf(buf, SvIVX(sv), SvUVX(sv), isUIOK, &ebuf);
2924 /* inlined from sv_setpvn */
2925 s = SvGROW_mutable(sv, len + 1);
2926 Move(ptr, s, len, char);
2930 else if (SvNOKp(sv)) {
2931 if (SvTYPE(sv) < SVt_PVNV)
2932 sv_upgrade(sv, SVt_PVNV);
2933 if (SvNVX(sv) == 0.0) {
2934 s = SvGROW_mutable(sv, 2);
2939 /* The +20 is pure guesswork. Configure test needed. --jhi */
2940 s = SvGROW_mutable(sv, NV_DIG + 20);
2941 /* some Xenix systems wipe out errno here */
2942 Gconvert(SvNVX(sv), NV_DIG, 0, s);
2952 if (isGV_with_GP(sv)) {
2953 GV *const gv = MUTABLE_GV(sv);
2954 SV *const buffer = sv_newmortal();
2956 gv_efullname3(buffer, gv, "*");
2958 assert(SvPOK(buffer));
2960 *lp = SvCUR(buffer);
2962 if ( SvUTF8(buffer) ) SvUTF8_on(sv);
2963 return SvPVX(buffer);
2968 if (flags & SV_UNDEF_RETURNS_NULL)
2970 if (!PL_localizing && !SvPADTMP(sv) && ckWARN(WARN_UNINITIALIZED))
2972 if (SvTYPE(sv) < SVt_PV)
2973 /* Typically the caller expects that sv_any is not NULL now. */
2974 sv_upgrade(sv, SVt_PV);
2978 const STRLEN len = s - SvPVX_const(sv);
2984 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
2985 PTR2UV(sv),SvPVX_const(sv)));
2986 if (flags & SV_CONST_RETURN)
2987 return (char *)SvPVX_const(sv);
2988 if (flags & SV_MUTABLE_RETURN)
2989 return SvPVX_mutable(sv);
2994 =for apidoc sv_copypv
2996 Copies a stringified representation of the source SV into the
2997 destination SV. Automatically performs any necessary mg_get and
2998 coercion of numeric values into strings. Guaranteed to preserve
2999 UTF8 flag even from overloaded objects. Similar in nature to
3000 sv_2pv[_flags] but operates directly on an SV instead of just the
3001 string. Mostly uses sv_2pv_flags to do its work, except when that
3002 would lose the UTF-8'ness of the PV.
3008 Perl_sv_copypv(pTHX_ SV *const dsv, register SV *const ssv)
3011 const char * const s = SvPV_const(ssv,len);
3013 PERL_ARGS_ASSERT_SV_COPYPV;
3015 sv_setpvn(dsv,s,len);
3023 =for apidoc sv_2pvbyte
3025 Return a pointer to the byte-encoded representation of the SV, and set *lp
3026 to its length. May cause the SV to be downgraded from UTF-8 as a
3029 Usually accessed via the C<SvPVbyte> macro.
3035 Perl_sv_2pvbyte(pTHX_ register SV *sv, STRLEN *const lp)
3037 PERL_ARGS_ASSERT_SV_2PVBYTE;
3039 if ((SvTHINKFIRST(sv) && !SvIsCOW(sv)) || isGV_with_GP(sv)) {
3040 SV *sv2 = sv_newmortal();
3044 else SvGETMAGIC(sv);
3045 sv_utf8_downgrade(sv,0);
3046 return lp ? SvPV_nomg(sv,*lp) : SvPV_nomg_nolen(sv);
3050 =for apidoc sv_2pvutf8
3052 Return a pointer to the UTF-8-encoded representation of the SV, and set *lp
3053 to its length. May cause the SV to be upgraded to UTF-8 as a side-effect.
3055 Usually accessed via the C<SvPVutf8> macro.
3061 Perl_sv_2pvutf8(pTHX_ register SV *sv, STRLEN *const lp)
3063 PERL_ARGS_ASSERT_SV_2PVUTF8;
3065 if ((SvTHINKFIRST(sv) && !SvIsCOW(sv)) || isGV_with_GP(sv))
3066 sv = sv_mortalcopy(sv);
3067 sv_utf8_upgrade(sv);
3068 if (SvGMAGICAL(sv)) SvFLAGS(sv) &= ~SVf_POK;
3070 return lp ? SvPV_nomg(sv,*lp) : SvPV_nomg_nolen(sv);
3075 =for apidoc sv_2bool
3077 This macro is only used by sv_true() or its macro equivalent, and only if
3078 the latter's argument is neither SvPOK, SvIOK nor SvNOK.
3079 It calls sv_2bool_flags with the SV_GMAGIC flag.
3081 =for apidoc sv_2bool_flags
3083 This function is only used by sv_true() and friends, and only if
3084 the latter's argument is neither SvPOK, SvIOK nor SvNOK. If the flags
3085 contain SV_GMAGIC, then it does an mg_get() first.
3092 Perl_sv_2bool_flags(pTHX_ register SV *const sv, const I32 flags)
3096 PERL_ARGS_ASSERT_SV_2BOOL_FLAGS;
3098 if(flags & SV_GMAGIC) SvGETMAGIC(sv);
3104 SV * const tmpsv = AMG_CALLunary(sv, bool__amg);
3105 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
3106 return cBOOL(SvTRUE(tmpsv));
3108 return SvRV(sv) != 0;
3111 register XPV* const Xpvtmp = (XPV*)SvANY(sv);
3113 (*sv->sv_u.svu_pv > '0' ||
3114 Xpvtmp->xpv_cur > 1 ||
3115 (Xpvtmp->xpv_cur && *sv->sv_u.svu_pv != '0')))
3122 return SvIVX(sv) != 0;
3125 return SvNVX(sv) != 0.0;
3127 if (isGV_with_GP(sv))
3137 =for apidoc sv_utf8_upgrade
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 Will C<mg_get> on C<sv> if appropriate.
3142 Always sets the SvUTF8 flag to avoid future validity checks even
3143 if the whole string is the same in UTF-8 as not.
3144 Returns the number of bytes in the converted string
3146 This is not as a general purpose byte encoding to Unicode interface:
3147 use the Encode extension for that.
3149 =for apidoc sv_utf8_upgrade_nomg
3151 Like sv_utf8_upgrade, but doesn't do magic on C<sv>.
3153 =for apidoc sv_utf8_upgrade_flags
3155 Converts the PV of an SV to its UTF-8-encoded form.
3156 Forces the SV to string form if it is not already.
3157 Always sets the SvUTF8 flag to avoid future validity checks even
3158 if all the bytes are invariant in UTF-8.
3159 If C<flags> has C<SV_GMAGIC> bit set,
3160 will C<mg_get> on C<sv> if appropriate, else not.
3161 Returns the number of bytes in the converted string
3162 C<sv_utf8_upgrade> and
3163 C<sv_utf8_upgrade_nomg> are implemented in terms of this function.
3165 This is not as a general purpose byte encoding to Unicode interface:
3166 use the Encode extension for that.
3170 The grow version is currently not externally documented. It adds a parameter,
3171 extra, which is the number of unused bytes the string of 'sv' is guaranteed to
3172 have free after it upon return. This allows the caller to reserve extra space
3173 that it intends to fill, to avoid extra grows.
3175 Also externally undocumented for the moment is the flag SV_FORCE_UTF8_UPGRADE,
3176 which can be used to tell this function to not first check to see if there are
3177 any characters that are different in UTF-8 (variant characters) which would
3178 force it to allocate a new string to sv, but to assume there are. Typically
3179 this flag is used by a routine that has already parsed the string to find that
3180 there are such characters, and passes this information on so that the work
3181 doesn't have to be repeated.
3183 (One might think that the calling routine could pass in the position of the
3184 first such variant, so it wouldn't have to be found again. But that is not the
3185 case, because typically when the caller is likely to use this flag, it won't be
3186 calling this routine unless it finds something that won't fit into a byte.
3187 Otherwise it tries to not upgrade and just use bytes. But some things that
3188 do fit into a byte are variants in utf8, and the caller may not have been
3189 keeping track of these.)
3191 If the routine itself changes the string, it adds a trailing NUL. Such a NUL
3192 isn't guaranteed due to having other routines do the work in some input cases,
3193 or if the input is already flagged as being in utf8.
3195 The speed of this could perhaps be improved for many cases if someone wanted to
3196 write a fast function that counts the number of variant characters in a string,
3197 especially if it could return the position of the first one.
3202 Perl_sv_utf8_upgrade_flags_grow(pTHX_ register SV *const sv, const I32 flags, STRLEN extra)
3206 PERL_ARGS_ASSERT_SV_UTF8_UPGRADE_FLAGS_GROW;
3208 if (sv == &PL_sv_undef)
3212 if (SvREADONLY(sv) && (SvPOKp(sv) || SvIOKp(sv) || SvNOKp(sv))) {
3213 (void) sv_2pv_flags(sv,&len, flags);
3215 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3219 (void) SvPV_force_flags(sv,len,flags & SV_GMAGIC);
3224 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3229 sv_force_normal_flags(sv, 0);
3232 if (PL_encoding && !(flags & SV_UTF8_NO_ENCODING)) {
3233 sv_recode_to_utf8(sv, PL_encoding);
3234 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3238 if (SvCUR(sv) == 0) {
3239 if (extra) SvGROW(sv, extra);
3240 } else { /* Assume Latin-1/EBCDIC */
3241 /* This function could be much more efficient if we
3242 * had a FLAG in SVs to signal if there are any variant
3243 * chars in the PV. Given that there isn't such a flag
3244 * make the loop as fast as possible (although there are certainly ways
3245 * to speed this up, eg. through vectorization) */
3246 U8 * s = (U8 *) SvPVX_const(sv);
3247 U8 * e = (U8 *) SvEND(sv);
3249 STRLEN two_byte_count = 0;
3251 if (flags & SV_FORCE_UTF8_UPGRADE) goto must_be_utf8;
3253 /* See if really will need to convert to utf8. We mustn't rely on our
3254 * incoming SV being well formed and having a trailing '\0', as certain
3255 * code in pp_formline can send us partially built SVs. */
3259 if (NATIVE_IS_INVARIANT(ch)) continue;
3261 t--; /* t already incremented; re-point to first variant */
3266 /* utf8 conversion not needed because all are invariants. Mark as
3267 * UTF-8 even if no variant - saves scanning loop */
3269 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3274 /* Here, the string should be converted to utf8, either because of an
3275 * input flag (two_byte_count = 0), or because a character that
3276 * requires 2 bytes was found (two_byte_count = 1). t points either to
3277 * the beginning of the string (if we didn't examine anything), or to
3278 * the first variant. In either case, everything from s to t - 1 will
3279 * occupy only 1 byte each on output.
3281 * There are two main ways to convert. One is to create a new string
3282 * and go through the input starting from the beginning, appending each
3283 * converted value onto the new string as we go along. It's probably
3284 * best to allocate enough space in the string for the worst possible
3285 * case rather than possibly running out of space and having to
3286 * reallocate and then copy what we've done so far. Since everything
3287 * from s to t - 1 is invariant, the destination can be initialized
3288 * with these using a fast memory copy
3290 * The other way is to figure out exactly how big the string should be
3291 * by parsing the entire input. Then you don't have to make it big
3292 * enough to handle the worst possible case, and more importantly, if
3293 * the string you already have is large enough, you don't have to
3294 * allocate a new string, you can copy the last character in the input
3295 * string to the final position(s) that will be occupied by the
3296 * converted string and go backwards, stopping at t, since everything
3297 * before that is invariant.
3299 * There are advantages and disadvantages to each method.
3301 * In the first method, we can allocate a new string, do the memory
3302 * copy from the s to t - 1, and then proceed through the rest of the
3303 * string byte-by-byte.
3305 * In the second method, we proceed through the rest of the input
3306 * string just calculating how big the converted string will be. Then
3307 * there are two cases:
3308 * 1) if the string has enough extra space to handle the converted
3309 * value. We go backwards through the string, converting until we
3310 * get to the position we are at now, and then stop. If this
3311 * position is far enough along in the string, this method is
3312 * faster than the other method. If the memory copy were the same
3313 * speed as the byte-by-byte loop, that position would be about
3314 * half-way, as at the half-way mark, parsing to the end and back
3315 * is one complete string's parse, the same amount as starting
3316 * over and going all the way through. Actually, it would be
3317 * somewhat less than half-way, as it's faster to just count bytes
3318 * than to also copy, and we don't have the overhead of allocating
3319 * a new string, changing the scalar to use it, and freeing the
3320 * existing one. But if the memory copy is fast, the break-even
3321 * point is somewhere after half way. The counting loop could be
3322 * sped up by vectorization, etc, to move the break-even point
3323 * further towards the beginning.
3324 * 2) if the string doesn't have enough space to handle the converted
3325 * value. A new string will have to be allocated, and one might
3326 * as well, given that, start from the beginning doing the first
3327 * method. We've spent extra time parsing the string and in
3328 * exchange all we've gotten is that we know precisely how big to
3329 * make the new one. Perl is more optimized for time than space,
3330 * so this case is a loser.
3331 * So what I've decided to do is not use the 2nd method unless it is
3332 * guaranteed that a new string won't have to be allocated, assuming
3333 * the worst case. I also decided not to put any more conditions on it
3334 * than this, for now. It seems likely that, since the worst case is
3335 * twice as big as the unknown portion of the string (plus 1), we won't
3336 * be guaranteed enough space, causing us to go to the first method,
3337 * unless the string is short, or the first variant character is near
3338 * the end of it. In either of these cases, it seems best to use the
3339 * 2nd method. The only circumstance I can think of where this would
3340 * be really slower is if the string had once had much more data in it
3341 * than it does now, but there is still a substantial amount in it */
3344 STRLEN invariant_head = t - s;
3345 STRLEN size = invariant_head + (e - t) * 2 + 1 + extra;
3346 if (SvLEN(sv) < size) {
3348 /* Here, have decided to allocate a new string */
3353 Newx(dst, size, U8);
3355 /* If no known invariants at the beginning of the input string,
3356 * set so starts from there. Otherwise, can use memory copy to
3357 * get up to where we are now, and then start from here */
3359 if (invariant_head <= 0) {
3362 Copy(s, dst, invariant_head, char);
3363 d = dst + invariant_head;
3367 const UV uv = NATIVE8_TO_UNI(*t++);
3368 if (UNI_IS_INVARIANT(uv))
3369 *d++ = (U8)UNI_TO_NATIVE(uv);
3371 *d++ = (U8)UTF8_EIGHT_BIT_HI(uv);
3372 *d++ = (U8)UTF8_EIGHT_BIT_LO(uv);
3376 SvPV_free(sv); /* No longer using pre-existing string */
3377 SvPV_set(sv, (char*)dst);
3378 SvCUR_set(sv, d - dst);
3379 SvLEN_set(sv, size);
3382 /* Here, have decided to get the exact size of the string.
3383 * Currently this happens only when we know that there is
3384 * guaranteed enough space to fit the converted string, so
3385 * don't have to worry about growing. If two_byte_count is 0,
3386 * then t points to the first byte of the string which hasn't
3387 * been examined yet. Otherwise two_byte_count is 1, and t
3388 * points to the first byte in the string that will expand to
3389 * two. Depending on this, start examining at t or 1 after t.
3392 U8 *d = t + two_byte_count;
3395 /* Count up the remaining bytes that expand to two */
3398 const U8 chr = *d++;
3399 if (! NATIVE_IS_INVARIANT(chr)) two_byte_count++;
3402 /* The string will expand by just the number of bytes that
3403 * occupy two positions. But we are one afterwards because of
3404 * the increment just above. This is the place to put the
3405 * trailing NUL, and to set the length before we decrement */
3407 d += two_byte_count;
3408 SvCUR_set(sv, d - s);
3412 /* Having decremented d, it points to the position to put the
3413 * very last byte of the expanded string. Go backwards through
3414 * the string, copying and expanding as we go, stopping when we
3415 * get to the part that is invariant the rest of the way down */
3419 const U8 ch = NATIVE8_TO_UNI(*e--);
3420 if (UNI_IS_INVARIANT(ch)) {
3421 *d-- = UNI_TO_NATIVE(ch);
3423 *d-- = (U8)UTF8_EIGHT_BIT_LO(ch);
3424 *d-- = (U8)UTF8_EIGHT_BIT_HI(ch);
3429 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
3430 /* Update pos. We do it at the end rather than during
3431 * the upgrade, to avoid slowing down the common case
3432 * (upgrade without pos) */
3433 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
3435 I32 pos = mg->mg_len;
3436 if (pos > 0 && (U32)pos > invariant_head) {
3437 U8 *d = (U8*) SvPVX(sv) + invariant_head;
3438 STRLEN n = (U32)pos - invariant_head;
3440 if (UTF8_IS_START(*d))
3445 mg->mg_len = d - (U8*)SvPVX(sv);
3448 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
3449 magic_setutf8(sv,mg); /* clear UTF8 cache */
3454 /* Mark as UTF-8 even if no variant - saves scanning loop */
3460 =for apidoc sv_utf8_downgrade
3462 Attempts to convert the PV of an SV from characters to bytes.
3463 If the PV contains a character that cannot fit
3464 in a byte, this conversion will fail;
3465 in this case, either returns false or, if C<fail_ok> is not
3468 This is not as a general purpose Unicode to byte encoding interface:
3469 use the Encode extension for that.
3475 Perl_sv_utf8_downgrade(pTHX_ register SV *const sv, const bool fail_ok)
3479 PERL_ARGS_ASSERT_SV_UTF8_DOWNGRADE;
3481 if (SvPOKp(sv) && SvUTF8(sv)) {
3485 int mg_flags = SV_GMAGIC;
3488 sv_force_normal_flags(sv, 0);
3490 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
3492 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
3494 I32 pos = mg->mg_len;
3496 sv_pos_b2u(sv, &pos);
3497 mg_flags = 0; /* sv_pos_b2u does get magic */
3501 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
3502 magic_setutf8(sv,mg); /* clear UTF8 cache */
3505 s = (U8 *) SvPV_flags(sv, len, mg_flags);
3507 if (!utf8_to_bytes(s, &len)) {
3512 Perl_croak(aTHX_ "Wide character in %s",
3515 Perl_croak(aTHX_ "Wide character");
3526 =for apidoc sv_utf8_encode
3528 Converts the PV of an SV to UTF-8, but then turns the C<SvUTF8>
3529 flag off so that it looks like octets again.
3535 Perl_sv_utf8_encode(pTHX_ register SV *const sv)
3537 PERL_ARGS_ASSERT_SV_UTF8_ENCODE;
3539 if (SvREADONLY(sv)) {
3540 sv_force_normal_flags(sv, 0);
3542 (void) sv_utf8_upgrade(sv);
3547 =for apidoc sv_utf8_decode
3549 If the PV of the SV is an octet sequence in UTF-8
3550 and contains a multiple-byte character, the C<SvUTF8> flag is turned on
3551 so that it looks like a character. If the PV contains only single-byte
3552 characters, the C<SvUTF8> flag stays off.
3553 Scans PV for validity and returns false if the PV is invalid UTF-8.
3559 Perl_sv_utf8_decode(pTHX_ register SV *const sv)
3561 PERL_ARGS_ASSERT_SV_UTF8_DECODE;
3564 const U8 *start, *c;
3567 /* The octets may have got themselves encoded - get them back as
3570 if (!sv_utf8_downgrade(sv, TRUE))
3573 /* it is actually just a matter of turning the utf8 flag on, but
3574 * we want to make sure everything inside is valid utf8 first.
3576 c = start = (const U8 *) SvPVX_const(sv);
3577 if (!is_utf8_string(c, SvCUR(sv)))
3579 e = (const U8 *) SvEND(sv);
3582 if (!UTF8_IS_INVARIANT(ch)) {
3587 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
3588 /* adjust pos to the start of a UTF8 char sequence */
3589 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
3591 I32 pos = mg->mg_len;
3593 for (c = start + pos; c > start; c--) {
3594 if (UTF8_IS_START(*c))
3597 mg->mg_len = c - start;
3600 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
3601 magic_setutf8(sv,mg); /* clear UTF8 cache */
3608 =for apidoc sv_setsv
3610 Copies the contents of the source SV C<ssv> into the destination SV
3611 C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3612 function if the source SV needs to be reused. Does not handle 'set' magic.
3613 Loosely speaking, it performs a copy-by-value, obliterating any previous
3614 content of the destination.
3616 You probably want to use one of the assortment of wrappers, such as
3617 C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3618 C<SvSetMagicSV_nosteal>.
3620 =for apidoc sv_setsv_flags
3622 Copies the contents of the source SV C<ssv> into the destination SV
3623 C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3624 function if the source SV needs to be reused. Does not handle 'set' magic.
3625 Loosely speaking, it performs a copy-by-value, obliterating any previous
3626 content of the destination.
3627 If the C<flags> parameter has the C<SV_GMAGIC> bit set, will C<mg_get> on
3628 C<ssv> if appropriate, else not. If the C<flags>
3629 parameter has the C<NOSTEAL> bit set then the
3630 buffers of temps will not be stolen. <sv_setsv>
3631 and C<sv_setsv_nomg> are implemented in terms of this function.
3633 You probably want to use one of the assortment of wrappers, such as
3634 C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3635 C<SvSetMagicSV_nosteal>.
3637 This is the primary function for copying scalars, and most other
3638 copy-ish functions and macros use this underneath.
3644 S_glob_assign_glob(pTHX_ SV *const dstr, SV *const sstr, const int dtype)
3646 I32 mro_changes = 0; /* 1 = method, 2 = isa, 3 = recursive isa */
3647 HV *old_stash = NULL;
3649 PERL_ARGS_ASSERT_GLOB_ASSIGN_GLOB;
3651 if (dtype != SVt_PVGV && !isGV_with_GP(dstr)) {
3652 const char * const name = GvNAME(sstr);
3653 const STRLEN len = GvNAMELEN(sstr);
3655 if (dtype >= SVt_PV) {
3661 SvUPGRADE(dstr, SVt_PVGV);
3662 (void)SvOK_off(dstr);
3663 /* We have to turn this on here, even though we turn it off
3664 below, as GvSTASH will fail an assertion otherwise. */
3665 isGV_with_GP_on(dstr);
3667 GvSTASH(dstr) = GvSTASH(sstr);
3669 Perl_sv_add_backref(aTHX_ MUTABLE_SV(GvSTASH(dstr)), dstr);
3670 gv_name_set(MUTABLE_GV(dstr), name, len,
3671 GV_ADD | (GvNAMEUTF8(sstr) ? SVf_UTF8 : 0 ));
3672 SvFAKE_on(dstr); /* can coerce to non-glob */
3675 if(GvGP(MUTABLE_GV(sstr))) {
3676 /* If source has method cache entry, clear it */
3678 SvREFCNT_dec(GvCV(sstr));
3679 GvCV_set(sstr, NULL);
3682 /* If source has a real method, then a method is
3685 GvCV((const GV *)sstr) && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3691 /* If dest already had a real method, that's a change as well */
3693 !mro_changes && GvGP(MUTABLE_GV(dstr)) && GvCVu((const GV *)dstr)
3694 && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3699 /* We don't need to check the name of the destination if it was not a
3700 glob to begin with. */
3701 if(dtype == SVt_PVGV) {
3702 const char * const name = GvNAME((const GV *)dstr);
3705 /* The stash may have been detached from the symbol table, so
3707 && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3708 && GvAV((const GV *)sstr)
3712 const STRLEN len = GvNAMELEN(dstr);
3713 if ((len > 1 && name[len-2] == ':' && name[len-1] == ':')
3714 || (len == 1 && name[0] == ':')) {
3717 /* Set aside the old stash, so we can reset isa caches on
3719 if((old_stash = GvHV(dstr)))
3720 /* Make sure we do not lose it early. */
3721 SvREFCNT_inc_simple_void_NN(
3722 sv_2mortal((SV *)old_stash)
3728 gp_free(MUTABLE_GV(dstr));
3729 isGV_with_GP_off(dstr); /* SvOK_off does not like globs. */
3730 (void)SvOK_off(dstr);
3731 isGV_with_GP_on(dstr);
3732 GvINTRO_off(dstr); /* one-shot flag */
3733 GvGP_set(dstr, gp_ref(GvGP(sstr)));
3734 if (SvTAINTED(sstr))
3736 if (GvIMPORTED(dstr) != GVf_IMPORTED
3737 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3739 GvIMPORTED_on(dstr);
3742 if(mro_changes == 2) {
3744 SV * const sref = (SV *)GvAV((const GV *)dstr);
3745 if (SvSMAGICAL(sref) && (mg = mg_find(sref, PERL_MAGIC_isa))) {
3746 if (SvTYPE(mg->mg_obj) != SVt_PVAV) {
3747 AV * const ary = newAV();
3748 av_push(ary, mg->mg_obj); /* takes the refcount */
3749 mg->mg_obj = (SV *)ary;
3751 av_push((AV *)mg->mg_obj, SvREFCNT_inc_simple_NN(dstr));
3753 else sv_magic(sref, dstr, PERL_MAGIC_isa, NULL, 0);
3754 mro_isa_changed_in(GvSTASH(dstr));
3756 else if(mro_changes == 3) {
3757 HV * const stash = GvHV(dstr);
3758 if(old_stash ? (HV *)HvENAME_get(old_stash) : stash)
3764 else if(mro_changes) mro_method_changed_in(GvSTASH(dstr));
3769 S_glob_assign_ref(pTHX_ SV *const dstr, SV *const sstr)
3771 SV * const sref = SvREFCNT_inc(SvRV(sstr));
3773 const int intro = GvINTRO(dstr);
3776 const U32 stype = SvTYPE(sref);
3778 PERL_ARGS_ASSERT_GLOB_ASSIGN_REF;
3781 GvINTRO_off(dstr); /* one-shot flag */
3782 GvLINE(dstr) = CopLINE(PL_curcop);
3783 GvEGV(dstr) = MUTABLE_GV(dstr);
3788 location = (SV **) &(GvGP(dstr)->gp_cv); /* XXX bypassing GvCV_set */
3789 import_flag = GVf_IMPORTED_CV;
3792 location = (SV **) &GvHV(dstr);
3793 import_flag = GVf_IMPORTED_HV;
3796 location = (SV **) &GvAV(dstr);
3797 import_flag = GVf_IMPORTED_AV;
3800 location = (SV **) &GvIOp(dstr);
3803 location = (SV **) &GvFORM(dstr);
3806 location = &GvSV(dstr);
3807 import_flag = GVf_IMPORTED_SV;
3810 if (stype == SVt_PVCV) {
3811 /*if (GvCVGEN(dstr) && (GvCV(dstr) != (const CV *)sref || GvCVGEN(dstr))) {*/
3812 if (GvCVGEN(dstr)) {
3813 SvREFCNT_dec(GvCV(dstr));
3814 GvCV_set(dstr, NULL);
3815 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3818 SAVEGENERICSV(*location);
3822 if (stype == SVt_PVCV && (*location != sref || GvCVGEN(dstr))) {
3823 CV* const cv = MUTABLE_CV(*location);
3825 if (!GvCVGEN((const GV *)dstr) &&
3826 (CvROOT(cv) || CvXSUB(cv)) &&
3827 /* redundant check that avoids creating the extra SV
3828 most of the time: */
3829 (CvCONST(cv) || ckWARN(WARN_REDEFINE)))
3831 SV * const new_const_sv =
3832 CvCONST((const CV *)sref)
3833 ? cv_const_sv((const CV *)sref)
3835 report_redefined_cv(
3836 sv_2mortal(Perl_newSVpvf(aTHX_
3839 HvNAME_HEK(GvSTASH((const GV *)dstr))
3841 HEKfARG(GvENAME_HEK(MUTABLE_GV(dstr)))
3844 CvCONST((const CV *)sref) ? &new_const_sv : NULL
3848 cv_ckproto_len_flags(cv, (const GV *)dstr,
3849 SvPOK(sref) ? CvPROTO(sref) : NULL,
3850 SvPOK(sref) ? CvPROTOLEN(sref) : 0,
3851 SvPOK(sref) ? SvUTF8(sref) : 0);
3853 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3854 GvASSUMECV_on(dstr);
3855 if(GvSTASH(dstr)) mro_method_changed_in(GvSTASH(dstr)); /* sub foo { 1 } sub bar { 2 } *bar = \&foo */
3858 if (import_flag && !(GvFLAGS(dstr) & import_flag)
3859 && CopSTASH_ne(PL_curcop, GvSTASH(dstr))) {
3860 GvFLAGS(dstr) |= import_flag;
3862 if (stype == SVt_PVHV) {
3863 const char * const name = GvNAME((GV*)dstr);
3864 const STRLEN len = GvNAMELEN(dstr);
3867 (len > 1 && name[len-2] == ':' && name[len-1] == ':')
3868 || (len == 1 && name[0] == ':')
3870 && (!dref || HvENAME_get(dref))
3873 (HV *)sref, (HV *)dref,
3879 stype == SVt_PVAV && sref != dref
3880 && strEQ(GvNAME((GV*)dstr), "ISA")
3881 /* The stash may have been detached from the symbol table, so
3882 check its name before doing anything. */
3883 && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3886 MAGIC * const omg = dref && SvSMAGICAL(dref)
3887 ? mg_find(dref, PERL_MAGIC_isa)
3889 if (SvSMAGICAL(sref) && (mg = mg_find(sref, PERL_MAGIC_isa))) {
3890 if (SvTYPE(mg->mg_obj) != SVt_PVAV) {
3891 AV * const ary = newAV();
3892 av_push(ary, mg->mg_obj); /* takes the refcount */
3893 mg->mg_obj = (SV *)ary;
3896 if (SvTYPE(omg->mg_obj) == SVt_PVAV) {
3897 SV **svp = AvARRAY((AV *)omg->mg_obj);
3898 I32 items = AvFILLp((AV *)omg->mg_obj) + 1;
3902 SvREFCNT_inc_simple_NN(*svp++)
3908 SvREFCNT_inc_simple_NN(omg->mg_obj)
3912 av_push((AV *)mg->mg_obj,SvREFCNT_inc_simple_NN(dstr));
3917 sref, omg ? omg->mg_obj : dstr, PERL_MAGIC_isa, NULL, 0
3919 mg = mg_find(sref, PERL_MAGIC_isa);
3921 /* Since the *ISA assignment could have affected more than
3922 one stash, don't call mro_isa_changed_in directly, but let
3923 magic_clearisa do it for us, as it already has the logic for
3924 dealing with globs vs arrays of globs. */
3926 Perl_magic_clearisa(aTHX_ NULL, mg);
3931 if (SvTAINTED(sstr))
3937 Perl_sv_setsv_flags(pTHX_ SV *dstr, register SV* sstr, const I32 flags)
3940 register U32 sflags;
3942 register svtype stype;
3944 PERL_ARGS_ASSERT_SV_SETSV_FLAGS;
3949 if (SvIS_FREED(dstr)) {
3950 Perl_croak(aTHX_ "panic: attempt to copy value %" SVf
3951 " to a freed scalar %p", SVfARG(sstr), (void *)dstr);
3953 SV_CHECK_THINKFIRST_COW_DROP(dstr);
3955 sstr = &PL_sv_undef;
3956 if (SvIS_FREED(sstr)) {
3957 Perl_croak(aTHX_ "panic: attempt to copy freed scalar %p to %p",
3958 (void*)sstr, (void*)dstr);
3960 stype = SvTYPE(sstr);
3961 dtype = SvTYPE(dstr);
3963 (void)SvAMAGIC_off(dstr);
3966 /* need to nuke the magic */
3967 sv_unmagic(dstr, PERL_MAGIC_vstring);
3970 /* There's a lot of redundancy below but we're going for speed here */
3975 if (dtype != SVt_PVGV && dtype != SVt_PVLV) {
3976 (void)SvOK_off(dstr);
3984 sv_upgrade(dstr, SVt_IV);
3988 sv_upgrade(dstr, SVt_PVIV);
3992 goto end_of_first_switch;
3994 (void)SvIOK_only(dstr);
3995 SvIV_set(dstr, SvIVX(sstr));
3998 /* SvTAINTED can only be true if the SV has taint magic, which in
3999 turn means that the SV type is PVMG (or greater). This is the
4000 case statement for SVt_IV, so this cannot be true (whatever gcov
4002 assert(!SvTAINTED(sstr));
4007 if (dtype < SVt_PV && dtype != SVt_IV)
4008 sv_upgrade(dstr, SVt_IV);
4016 sv_upgrade(dstr, SVt_NV);
4020 sv_upgrade(dstr, SVt_PVNV);
4024 goto end_of_first_switch;
4026 SvNV_set(dstr, SvNVX(sstr));
4027 (void)SvNOK_only(dstr);
4028 /* SvTAINTED can only be true if the SV has taint magic, which in
4029 turn means that the SV type is PVMG (or greater). This is the
4030 case statement for SVt_NV, so this cannot be true (whatever gcov
4032 assert(!SvTAINTED(sstr));
4038 #ifdef PERL_OLD_COPY_ON_WRITE
4039 if ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS) {
4040 if (dtype < SVt_PVIV)
4041 sv_upgrade(dstr, SVt_PVIV);
4048 sv_upgrade(dstr, SVt_PV);
4051 if (dtype < SVt_PVIV)
4052 sv_upgrade(dstr, SVt_PVIV);
4055 if (dtype < SVt_PVNV)
4056 sv_upgrade(dstr, SVt_PVNV);
4060 const char * const type = sv_reftype(sstr,0);
4062 /* diag_listed_as: Bizarre copy of %s */
4063 Perl_croak(aTHX_ "Bizarre copy of %s in %s", type, OP_DESC(PL_op));
4065 Perl_croak(aTHX_ "Bizarre copy of %s", type);
4070 if (dtype < SVt_REGEXP)
4071 sv_upgrade(dstr, SVt_REGEXP);
4074 /* case SVt_BIND: */
4078 if (SvGMAGICAL(sstr) && (flags & SV_GMAGIC)) {
4080 if (SvTYPE(sstr) != stype)
4081 stype = SvTYPE(sstr);
4083 if (isGV_with_GP(sstr) && dtype <= SVt_PVLV) {
4084 glob_assign_glob(dstr, sstr, dtype);
4087 if (stype == SVt_PVLV)
4088 SvUPGRADE(dstr, SVt_PVNV);
4090 SvUPGRADE(dstr, (svtype)stype);
4092 end_of_first_switch:
4094 /* dstr may have been upgraded. */
4095 dtype = SvTYPE(dstr);
4096 sflags = SvFLAGS(sstr);
4098 if (dtype == SVt_PVCV || dtype == SVt_PVFM) {
4099 /* Assigning to a subroutine sets the prototype. */
4102 const char *const ptr = SvPV_const(sstr, len);
4104 SvGROW(dstr, len + 1);
4105 Copy(ptr, SvPVX(dstr), len + 1, char);
4106 SvCUR_set(dstr, len);
4108 SvFLAGS(dstr) |= sflags & SVf_UTF8;
4109 CvAUTOLOAD_off(dstr);
4113 } else if (dtype == SVt_PVAV || dtype == SVt_PVHV) {
4114 const char * const type = sv_reftype(dstr,0);
4116 /* diag_listed_as: Cannot copy to %s */
4117 Perl_croak(aTHX_ "Cannot copy to %s in %s", type, OP_DESC(PL_op));
4119 Perl_croak(aTHX_ "Cannot copy to %s", type);
4120 } else if (sflags & SVf_ROK) {
4121 if (isGV_with_GP(dstr)
4122 && SvTYPE(SvRV(sstr)) == SVt_PVGV && isGV_with_GP(SvRV(sstr))) {
4125 if (GvIMPORTED(dstr) != GVf_IMPORTED
4126 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
4128 GvIMPORTED_on(dstr);
4133 glob_assign_glob(dstr, sstr, dtype);
4137 if (dtype >= SVt_PV) {
4138 if (isGV_with_GP(dstr)) {
4139 glob_assign_ref(dstr, sstr);
4142 if (SvPVX_const(dstr)) {
4148 (void)SvOK_off(dstr);
4149 SvRV_set(dstr, SvREFCNT_inc(SvRV(sstr)));
4150 SvFLAGS(dstr) |= sflags & SVf_ROK;
4151 assert(!(sflags & SVp_NOK));
4152 assert(!(sflags & SVp_IOK));
4153 assert(!(sflags & SVf_NOK));
4154 assert(!(sflags & SVf_IOK));
4156 else if (isGV_with_GP(dstr)) {
4157 if (!(sflags & SVf_OK)) {
4158 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
4159 "Undefined value assigned to typeglob");
4162 GV *gv = gv_fetchsv_nomg(sstr, GV_ADD, SVt_PVGV);
4163 if (dstr != (const SV *)gv) {
4164 const char * const name = GvNAME((const GV *)dstr);
4165 const STRLEN len = GvNAMELEN(dstr);
4166 HV *old_stash = NULL;
4167 bool reset_isa = FALSE;
4168 if ((len > 1 && name[len-2] == ':' && name[len-1] == ':')
4169 || (len == 1 && name[0] == ':')) {
4170 /* Set aside the old stash, so we can reset isa caches
4171 on its subclasses. */
4172 if((old_stash = GvHV(dstr))) {
4173 /* Make sure we do not lose it early. */
4174 SvREFCNT_inc_simple_void_NN(
4175 sv_2mortal((SV *)old_stash)
4182 gp_free(MUTABLE_GV(dstr));
4183 GvGP_set(dstr, gp_ref(GvGP(gv)));
4186 HV * const stash = GvHV(dstr);
4188 old_stash ? (HV *)HvENAME_get(old_stash) : stash
4198 else if (dtype == SVt_REGEXP && stype == SVt_REGEXP) {
4199 reg_temp_copy((REGEXP*)dstr, (REGEXP*)sstr);
4201 else if (sflags & SVp_POK) {
4205 * Check to see if we can just swipe the string. If so, it's a
4206 * possible small lose on short strings, but a big win on long ones.
4207 * It might even be a win on short strings if SvPVX_const(dstr)
4208 * has to be allocated and SvPVX_const(sstr) has to be freed.
4209 * Likewise if we can set up COW rather than doing an actual copy, we
4210 * drop to the else clause, as the swipe code and the COW setup code
4211 * have much in common.
4214 /* Whichever path we take through the next code, we want this true,
4215 and doing it now facilitates the COW check. */
4216 (void)SvPOK_only(dstr);
4219 /* If we're already COW then this clause is not true, and if COW
4220 is allowed then we drop down to the else and make dest COW
4221 with us. If caller hasn't said that we're allowed to COW
4222 shared hash keys then we don't do the COW setup, even if the
4223 source scalar is a shared hash key scalar. */
4224 (((flags & SV_COW_SHARED_HASH_KEYS)
4225 ? (sflags & (SVf_FAKE|SVf_READONLY)) != (SVf_FAKE|SVf_READONLY)
4226 : 1 /* If making a COW copy is forbidden then the behaviour we
4227 desire is as if the source SV isn't actually already
4228 COW, even if it is. So we act as if the source flags
4229 are not COW, rather than actually testing them. */
4231 #ifndef PERL_OLD_COPY_ON_WRITE
4232 /* The change that added SV_COW_SHARED_HASH_KEYS makes the logic
4233 when PERL_OLD_COPY_ON_WRITE is defined a little wrong.
4234 Conceptually PERL_OLD_COPY_ON_WRITE being defined should
4235 override SV_COW_SHARED_HASH_KEYS, because it means "always COW"
4236 but in turn, it's somewhat dead code, never expected to go
4237 live, but more kept as a placeholder on how to do it better
4238 in a newer implementation. */
4239 /* If we are COW and dstr is a suitable target then we drop down
4240 into the else and make dest a COW of us. */
4241 || (SvFLAGS(dstr) & CAN_COW_MASK) != CAN_COW_FLAGS
4246 (sflags & SVs_TEMP) && /* slated for free anyway? */
4247 !(sflags & SVf_OOK) && /* and not involved in OOK hack? */
4248 (!(flags & SV_NOSTEAL)) &&
4249 /* and we're allowed to steal temps */
4250 SvREFCNT(sstr) == 1 && /* and no other references to it? */
4251 SvLEN(sstr)) /* and really is a string */
4252 #ifdef PERL_OLD_COPY_ON_WRITE
4253 && ((flags & SV_COW_SHARED_HASH_KEYS)
4254 ? (!((sflags & CAN_COW_MASK) == CAN_COW_FLAGS
4255 && (SvFLAGS(dstr) & CAN_COW_MASK) == CAN_COW_FLAGS
4256 && SvTYPE(sstr) >= SVt_PVIV && SvTYPE(sstr) != SVt_PVFM))
4260 /* Failed the swipe test, and it's not a shared hash key either.
4261 Have to copy the string. */
4262 STRLEN len = SvCUR(sstr);
4263 SvGROW(dstr, len + 1); /* inlined from sv_setpvn */
4264 Move(SvPVX_const(sstr),SvPVX(dstr),len,char);
4265 SvCUR_set(dstr, len);
4266 *SvEND(dstr) = '\0';
4268 /* If PERL_OLD_COPY_ON_WRITE is not defined, then isSwipe will always
4270 /* Either it's a shared hash key, or it's suitable for
4271 copy-on-write or we can swipe the string. */
4273 PerlIO_printf(Perl_debug_log, "Copy on write: sstr --> dstr\n");
4277 #ifdef PERL_OLD_COPY_ON_WRITE
4279 if ((sflags & (SVf_FAKE | SVf_READONLY))
4280 != (SVf_FAKE | SVf_READONLY)) {
4281 SvREADONLY_on(sstr);
4283 /* Make the source SV into a loop of 1.
4284 (about to become 2) */
4285 SV_COW_NEXT_SV_SET(sstr, sstr);
4289 /* Initial code is common. */
4290 if (SvPVX_const(dstr)) { /* we know that dtype >= SVt_PV */
4295 /* making another shared SV. */
4296 STRLEN cur = SvCUR(sstr);
4297 STRLEN len = SvLEN(sstr);
4298 #ifdef PERL_OLD_COPY_ON_WRITE
4300 assert (SvTYPE(dstr) >= SVt_PVIV);
4301 /* SvIsCOW_normal */
4302 /* splice us in between source and next-after-source. */
4303 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4304 SV_COW_NEXT_SV_SET(sstr, dstr);
4305 SvPV_set(dstr, SvPVX_mutable(sstr));
4309 /* SvIsCOW_shared_hash */
4310 DEBUG_C(PerlIO_printf(Perl_debug_log,
4311 "Copy on write: Sharing hash\n"));
4313 assert (SvTYPE(dstr) >= SVt_PV);
4315 HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)))));
4317 SvLEN_set(dstr, len);
4318 SvCUR_set(dstr, cur);
4319 SvREADONLY_on(dstr);
4323 { /* Passes the swipe test. */
4324 SvPV_set(dstr, SvPVX_mutable(sstr));
4325 SvLEN_set(dstr, SvLEN(sstr));
4326 SvCUR_set(dstr, SvCUR(sstr));
4329 (void)SvOK_off(sstr); /* NOTE: nukes most SvFLAGS on sstr */
4330 SvPV_set(sstr, NULL);
4336 if (sflags & SVp_NOK) {
4337 SvNV_set(dstr, SvNVX(sstr));
4339 if (sflags & SVp_IOK) {
4340 SvIV_set(dstr, SvIVX(sstr));
4341 /* Must do this otherwise some other overloaded use of 0x80000000
4342 gets confused. I guess SVpbm_VALID */
4343 if (sflags & SVf_IVisUV)
4346 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_NOK|SVp_NOK|SVf_UTF8);
4348 const MAGIC * const smg = SvVSTRING_mg(sstr);
4350 sv_magic(dstr, NULL, PERL_MAGIC_vstring,
4351 smg->mg_ptr, smg->mg_len);
4352 SvRMAGICAL_on(dstr);
4356 else if (sflags & (SVp_IOK|SVp_NOK)) {
4357 (void)SvOK_off(dstr);
4358 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_IVisUV|SVf_NOK|SVp_NOK);
4359 if (sflags & SVp_IOK) {
4360 /* XXXX Do we want to set IsUV for IV(ROK)? Be extra safe... */
4361 SvIV_set(dstr, SvIVX(sstr));
4363 if (sflags & SVp_NOK) {
4364 SvNV_set(dstr, SvNVX(sstr));
4368 if (isGV_with_GP(sstr)) {
4369 gv_efullname3(dstr, MUTABLE_GV(sstr), "*");
4372 (void)SvOK_off(dstr);
4374 if (SvTAINTED(sstr))
4379 =for apidoc sv_setsv_mg
4381 Like C<sv_setsv>, but also handles 'set' magic.
4387 Perl_sv_setsv_mg(pTHX_ SV *const dstr, register SV *const sstr)
4389 PERL_ARGS_ASSERT_SV_SETSV_MG;
4391 sv_setsv(dstr,sstr);
4395 #ifdef PERL_OLD_COPY_ON_WRITE
4397 Perl_sv_setsv_cow(pTHX_ SV *dstr, SV *sstr)
4399 STRLEN cur = SvCUR(sstr);
4400 STRLEN len = SvLEN(sstr);
4401 register char *new_pv;
4403 PERL_ARGS_ASSERT_SV_SETSV_COW;
4406 PerlIO_printf(Perl_debug_log, "Fast copy on write: %p -> %p\n",
4407 (void*)sstr, (void*)dstr);
4414 if (SvTHINKFIRST(dstr))
4415 sv_force_normal_flags(dstr, SV_COW_DROP_PV);
4416 else if (SvPVX_const(dstr))
4417 Safefree(SvPVX_const(dstr));
4421 SvUPGRADE(dstr, SVt_PVIV);
4423 assert (SvPOK(sstr));
4424 assert (SvPOKp(sstr));
4425 assert (!SvIOK(sstr));
4426 assert (!SvIOKp(sstr));
4427 assert (!SvNOK(sstr));
4428 assert (!SvNOKp(sstr));
4430 if (SvIsCOW(sstr)) {
4432 if (SvLEN(sstr) == 0) {
4433 /* source is a COW shared hash key. */
4434 DEBUG_C(PerlIO_printf(Perl_debug_log,
4435 "Fast copy on write: Sharing hash\n"));
4436 new_pv = HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr))));
4439 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4441 assert ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS);
4442 SvUPGRADE(sstr, SVt_PVIV);
4443 SvREADONLY_on(sstr);
4445 DEBUG_C(PerlIO_printf(Perl_debug_log,
4446 "Fast copy on write: Converting sstr to COW\n"));
4447 SV_COW_NEXT_SV_SET(dstr, sstr);
4449 SV_COW_NEXT_SV_SET(sstr, dstr);
4450 new_pv = SvPVX_mutable(sstr);
4453 SvPV_set(dstr, new_pv);
4454 SvFLAGS(dstr) = (SVt_PVIV|SVf_POK|SVp_POK|SVf_FAKE|SVf_READONLY);
4457 SvLEN_set(dstr, len);
4458 SvCUR_set(dstr, cur);
4467 =for apidoc sv_setpvn
4469 Copies a string into an SV. The C<len> parameter indicates the number of
4470 bytes to be copied. If the C<ptr> argument is NULL the SV will become
4471 undefined. Does not handle 'set' magic. See C<sv_setpvn_mg>.
4477 Perl_sv_setpvn(pTHX_ register SV *const sv, register const char *const ptr, register const STRLEN len)
4480 register char *dptr;
4482 PERL_ARGS_ASSERT_SV_SETPVN;
4484 SV_CHECK_THINKFIRST_COW_DROP(sv);
4490 /* len is STRLEN which is unsigned, need to copy to signed */
4493 Perl_croak(aTHX_ "panic: sv_setpvn called with negative strlen %"
4496 SvUPGRADE(sv, SVt_PV);
4498 dptr = SvGROW(sv, len + 1);
4499 Move(ptr,dptr,len,char);
4502 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4504 if (SvTYPE(sv) == SVt_PVCV) CvAUTOLOAD_off(sv);
4508 =for apidoc sv_setpvn_mg
4510 Like C<sv_setpvn>, but also handles 'set' magic.
4516 Perl_sv_setpvn_mg(pTHX_ register SV *const sv, register const char *const ptr, register const STRLEN len)
4518 PERL_ARGS_ASSERT_SV_SETPVN_MG;
4520 sv_setpvn(sv,ptr,len);
4525 =for apidoc sv_setpv
4527 Copies a string into an SV. The string must be null-terminated. Does not
4528 handle 'set' magic. See C<sv_setpv_mg>.
4534 Perl_sv_setpv(pTHX_ register SV *const sv, register const char *const ptr)
4537 register STRLEN len;
4539 PERL_ARGS_ASSERT_SV_SETPV;
4541 SV_CHECK_THINKFIRST_COW_DROP(sv);
4547 SvUPGRADE(sv, SVt_PV);
4549 SvGROW(sv, len + 1);
4550 Move(ptr,SvPVX(sv),len+1,char);
4552 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4554 if (SvTYPE(sv) == SVt_PVCV) CvAUTOLOAD_off(sv);
4558 =for apidoc sv_setpv_mg
4560 Like C<sv_setpv>, but also handles 'set' magic.
4566 Perl_sv_setpv_mg(pTHX_ register SV *const sv, register const char *const ptr)
4568 PERL_ARGS_ASSERT_SV_SETPV_MG;
4575 Perl_sv_sethek(pTHX_ register SV *const sv, const HEK *const hek)
4579 PERL_ARGS_ASSERT_SV_SETHEK;
4585 if (HEK_LEN(hek) == HEf_SVKEY) {
4586 sv_setsv(sv, *(SV**)HEK_KEY(hek));
4589 const int flags = HEK_FLAGS(hek);
4590 if (flags & HVhek_WASUTF8) {
4591 STRLEN utf8_len = HEK_LEN(hek);
4592 char *as_utf8 = (char *)bytes_to_utf8((U8*)HEK_KEY(hek), &utf8_len);
4593 sv_usepvn_flags(sv, as_utf8, utf8_len, SV_HAS_TRAILING_NUL);
4596 } else if (flags & (HVhek_REHASH|HVhek_UNSHARED)) {
4597 sv_setpvn(sv, HEK_KEY(hek), HEK_LEN(hek));
4600 else SvUTF8_off(sv);
4604 SV_CHECK_THINKFIRST_COW_DROP(sv);
4605 SvUPGRADE(sv, SVt_PV);
4606 SvPV_set(sv,(char *)HEK_KEY(share_hek_hek(hek)));
4607 SvCUR_set(sv, HEK_LEN(hek));
4614 else SvUTF8_off(sv);
4622 =for apidoc sv_usepvn_flags
4624 Tells an SV to use C<ptr> to find its string value. Normally the
4625 string is stored inside the SV but sv_usepvn allows the SV to use an
4626 outside string. The C<ptr> should point to memory that was allocated
4627 by C<malloc>. It must be the start of a mallocked block
4628 of memory, and not a pointer to the middle of it. The
4629 string length, C<len>, must be supplied. By default
4630 this function will realloc (i.e. move) the memory pointed to by C<ptr>,
4631 so that pointer should not be freed or used by the programmer after
4632 giving it to sv_usepvn, and neither should any pointers from "behind"
4633 that pointer (e.g. ptr + 1) be used.
4635 If C<flags> & SV_SMAGIC is true, will call SvSETMAGIC. If C<flags> &
4636 SV_HAS_TRAILING_NUL is true, then C<ptr[len]> must be NUL, and the realloc
4637 will be skipped (i.e. the buffer is actually at least 1 byte longer than
4638 C<len>, and already meets the requirements for storing in C<SvPVX>).
4644 Perl_sv_usepvn_flags(pTHX_ SV *const sv, char *ptr, const STRLEN len, const U32 flags)
4649 PERL_ARGS_ASSERT_SV_USEPVN_FLAGS;
4651 SV_CHECK_THINKFIRST_COW_DROP(sv);
4652 SvUPGRADE(sv, SVt_PV);
4655 if (flags & SV_SMAGIC)
4659 if (SvPVX_const(sv))
4663 if (flags & SV_HAS_TRAILING_NUL)
4664 assert(ptr[len] == '\0');
4667 allocate = (flags & SV_HAS_TRAILING_NUL)
4669 #ifdef Perl_safesysmalloc_size
4672 PERL_STRLEN_ROUNDUP(len + 1);
4674 if (flags & SV_HAS_TRAILING_NUL) {
4675 /* It's long enough - do nothing.
4676 Specifically Perl_newCONSTSUB is relying on this. */
4679 /* Force a move to shake out bugs in callers. */
4680 char *new_ptr = (char*)safemalloc(allocate);
4681 Copy(ptr, new_ptr, len, char);
4682 PoisonFree(ptr,len,char);
4686 ptr = (char*) saferealloc (ptr, allocate);
4689 #ifdef Perl_safesysmalloc_size
4690 SvLEN_set(sv, Perl_safesysmalloc_size(ptr));
4692 SvLEN_set(sv, allocate);
4696 if (!(flags & SV_HAS_TRAILING_NUL)) {
4699 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4701 if (flags & SV_SMAGIC)
4705 #ifdef PERL_OLD_COPY_ON_WRITE
4706 /* Need to do this *after* making the SV normal, as we need the buffer
4707 pointer to remain valid until after we've copied it. If we let go too early,
4708 another thread could invalidate it by unsharing last of the same hash key
4709 (which it can do by means other than releasing copy-on-write Svs)
4710 or by changing the other copy-on-write SVs in the loop. */
4712 S_sv_release_COW(pTHX_ register SV *sv, const char *pvx, SV *after)
4714 PERL_ARGS_ASSERT_SV_RELEASE_COW;
4716 { /* this SV was SvIsCOW_normal(sv) */
4717 /* we need to find the SV pointing to us. */
4718 SV *current = SV_COW_NEXT_SV(after);
4720 if (current == sv) {
4721 /* The SV we point to points back to us (there were only two of us
4723 Hence other SV is no longer copy on write either. */
4725 SvREADONLY_off(after);
4727 /* We need to follow the pointers around the loop. */
4729 while ((next = SV_COW_NEXT_SV(current)) != sv) {
4732 /* don't loop forever if the structure is bust, and we have
4733 a pointer into a closed loop. */
4734 assert (current != after);
4735 assert (SvPVX_const(current) == pvx);
4737 /* Make the SV before us point to the SV after us. */
4738 SV_COW_NEXT_SV_SET(current, after);
4744 =for apidoc sv_force_normal_flags
4746 Undo various types of fakery on an SV: if the PV is a shared string, make
4747 a private copy; if we're a ref, stop refing; if we're a glob, downgrade to
4748 an xpvmg; if we're a copy-on-write scalar, this is the on-write time when
4749 we do the copy, and is also used locally. If C<SV_COW_DROP_PV> is set
4750 then a copy-on-write scalar drops its PV buffer (if any) and becomes
4751 SvPOK_off rather than making a copy. (Used where this
4752 scalar is about to be set to some other value.) In addition,
4753 the C<flags> parameter gets passed to C<sv_unref_flags()>
4754 when unreffing. C<sv_force_normal> calls this function
4755 with flags set to 0.
4761 Perl_sv_force_normal_flags(pTHX_ register SV *const sv, const U32 flags)
4765 PERL_ARGS_ASSERT_SV_FORCE_NORMAL_FLAGS;
4767 #ifdef PERL_OLD_COPY_ON_WRITE
4768 if (SvREADONLY(sv)) {
4770 const char * const pvx = SvPVX_const(sv);
4771 const STRLEN len = SvLEN(sv);
4772 const STRLEN cur = SvCUR(sv);
4773 /* next COW sv in the loop. If len is 0 then this is a shared-hash
4774 key scalar, so we mustn't attempt to call SV_COW_NEXT_SV(), as
4775 we'll fail an assertion. */
4776 SV * const next = len ? SV_COW_NEXT_SV(sv) : 0;
4779 PerlIO_printf(Perl_debug_log,
4780 "Copy on write: Force normal %ld\n",
4786 /* This SV doesn't own the buffer, so need to Newx() a new one: */
4789 if (flags & SV_COW_DROP_PV) {
4790 /* OK, so we don't need to copy our buffer. */
4793 SvGROW(sv, cur + 1);
4794 Move(pvx,SvPVX(sv),cur,char);
4799 sv_release_COW(sv, pvx, next);
4801 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4807 else if (IN_PERL_RUNTIME)
4808 Perl_croak_no_modify(aTHX);
4811 if (SvREADONLY(sv)) {
4813 const char * const pvx = SvPVX_const(sv);
4814 const STRLEN len = SvCUR(sv);
4819 if (flags & SV_COW_DROP_PV) {
4820 /* OK, so we don't need to copy our buffer. */
4823 SvGROW(sv, len + 1);
4824 Move(pvx,SvPVX(sv),len,char);
4827 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4829 else if (IN_PERL_RUNTIME)
4830 Perl_croak_no_modify(aTHX);
4834 sv_unref_flags(sv, flags);
4835 else if (SvFAKE(sv) && isGV_with_GP(sv))
4836 sv_unglob(sv, flags);
4837 else if (SvFAKE(sv) && SvTYPE(sv) == SVt_REGEXP) {
4838 /* Need to downgrade the REGEXP to a simple(r) scalar. This is analogous
4839 to sv_unglob. We only need it here, so inline it. */
4840 const svtype new_type = SvMAGIC(sv) || SvSTASH(sv) ? SVt_PVMG : SVt_PV;
4841 SV *const temp = newSV_type(new_type);
4842 void *const temp_p = SvANY(sv);
4844 if (new_type == SVt_PVMG) {
4845 SvMAGIC_set(temp, SvMAGIC(sv));
4846 SvMAGIC_set(sv, NULL);
4847 SvSTASH_set(temp, SvSTASH(sv));
4848 SvSTASH_set(sv, NULL);
4850 SvCUR_set(temp, SvCUR(sv));
4851 /* Remember that SvPVX is in the head, not the body. */
4853 SvLEN_set(temp, SvLEN(sv));
4854 /* This signals "buffer is owned by someone else" in sv_clear,
4855 which is the least effort way to stop it freeing the buffer.
4857 SvLEN_set(sv, SvLEN(sv)+1);
4859 /* Their buffer is already owned by someone else. */
4860 SvPVX(sv) = savepvn(SvPVX(sv), SvCUR(sv));
4861 SvLEN_set(temp, SvCUR(sv)+1);
4864 /* Now swap the rest of the bodies. */
4866 SvFLAGS(sv) &= ~(SVf_FAKE|SVTYPEMASK);
4867 SvFLAGS(sv) |= new_type;
4868 SvANY(sv) = SvANY(temp);
4870 SvFLAGS(temp) &= ~(SVTYPEMASK);
4871 SvFLAGS(temp) |= SVt_REGEXP|SVf_FAKE;
4872 SvANY(temp) = temp_p;
4881 Efficient removal of characters from the beginning of the string buffer.
4882 SvPOK(sv) must be true and the C<ptr> must be a pointer to somewhere inside
4883 the string buffer. The C<ptr> becomes the first character of the adjusted
4884 string. Uses the "OOK hack".
4886 Beware: after this function returns, C<ptr> and SvPVX_const(sv) may no longer
4887 refer to the same chunk of data.
4889 The unfortunate similarity of this function's name to that of Perl's C<chop>
4890 operator is strictly coincidental. This function works from the left;
4891 C<chop> works from the right.
4897 Perl_sv_chop(pTHX_ register SV *const sv, register const char *const ptr)
4908 PERL_ARGS_ASSERT_SV_CHOP;
4910 if (!ptr || !SvPOKp(sv))
4912 delta = ptr - SvPVX_const(sv);
4914 /* Nothing to do. */
4917 max_delta = SvLEN(sv) ? SvLEN(sv) : SvCUR(sv);
4918 if (delta > max_delta)
4919 Perl_croak(aTHX_ "panic: sv_chop ptr=%p, start=%p, end=%p",
4920 ptr, SvPVX_const(sv), SvPVX_const(sv) + max_delta);
4921 /* SvPVX(sv) may move in SV_CHECK_THINKFIRST(sv), so don't use ptr any more */
4922 SV_CHECK_THINKFIRST(sv);
4925 if (!SvLEN(sv)) { /* make copy of shared string */
4926 const char *pvx = SvPVX_const(sv);
4927 const STRLEN len = SvCUR(sv);
4928 SvGROW(sv, len + 1);
4929 Move(pvx,SvPVX(sv),len,char);
4935 SvOOK_offset(sv, old_delta);
4937 SvLEN_set(sv, SvLEN(sv) - delta);
4938 SvCUR_set(sv, SvCUR(sv) - delta);
4939 SvPV_set(sv, SvPVX(sv) + delta);
4941 p = (U8 *)SvPVX_const(sv);
4944 /* how many bytes were evacuated? we will fill them with sentinel
4945 bytes, except for the part holding the new offset of course. */
4948 evacn += (old_delta < 0x100 ? 1 : 1 + sizeof(STRLEN));
4950 assert(evacn <= delta + old_delta);
4956 if (delta < 0x100) {
4960 p -= sizeof(STRLEN);
4961 Copy((U8*)&delta, p, sizeof(STRLEN), U8);
4965 /* Fill the preceding buffer with sentinals to verify that no-one is
4975 =for apidoc sv_catpvn
4977 Concatenates the string onto the end of the string which is in the SV. The
4978 C<len> indicates number of bytes to copy. If the SV has the UTF-8
4979 status set, then the bytes appended should be valid UTF-8.
4980 Handles 'get' magic, but not 'set' magic. See C<sv_catpvn_mg>.
4982 =for apidoc sv_catpvn_flags
4984 Concatenates the string onto the end of the string which is in the SV. The
4985 C<len> indicates number of bytes to copy. If the SV has the UTF-8
4986 status set, then the bytes appended should be valid UTF-8.
4987 If C<flags> has the C<SV_SMAGIC> bit set, will
4988 C<mg_set> on C<dsv> afterwards if appropriate.
4989 C<sv_catpvn> and C<sv_catpvn_nomg> are implemented
4990 in terms of this function.
4996 Perl_sv_catpvn_flags(pTHX_ register SV *const dsv, register const char *sstr, register const STRLEN slen, const I32 flags)
5000 const char * const dstr = SvPV_force_flags(dsv, dlen, flags);
5002 PERL_ARGS_ASSERT_SV_CATPVN_FLAGS;
5003 assert((flags & (SV_CATBYTES|SV_CATUTF8)) != (SV_CATBYTES|SV_CATUTF8));
5005 if (!(flags & SV_CATBYTES) || !SvUTF8(dsv)) {
5006 if (flags & SV_CATUTF8 && !SvUTF8(dsv)) {
5007 sv_utf8_upgrade_flags_grow(dsv, 0, slen + 1);
5010 else SvGROW(dsv, dlen + slen + 1);
5012 sstr = SvPVX_const(dsv);
5013 Move(sstr, SvPVX(dsv) + dlen, slen, char);
5014 SvCUR_set(dsv, SvCUR(dsv) + slen);
5017 /* We inline bytes_to_utf8, to avoid an extra malloc. */
5018 const char * const send = sstr + slen;
5021 /* Something this code does not account for, which I think is
5022 impossible; it would require the same pv to be treated as
5023 bytes *and* utf8, which would indicate a bug elsewhere. */
5024 assert(sstr != dstr);
5026 SvGROW(dsv, dlen + slen * 2 + 1);
5027 d = (U8 *)SvPVX(dsv) + dlen;
5029 while (sstr < send) {
5030 const UV uv = NATIVE_TO_ASCII((U8)*sstr++);
5031 if (UNI_IS_INVARIANT(uv))
5032 *d++ = (U8)UTF_TO_NATIVE(uv);
5034 *d++ = (U8)UTF8_EIGHT_BIT_HI(uv);
5035 *d++ = (U8)UTF8_EIGHT_BIT_LO(uv);
5038 SvCUR_set(dsv, d-(const U8 *)SvPVX(dsv));
5041 (void)SvPOK_only_UTF8(dsv); /* validate pointer */
5043 if (flags & SV_SMAGIC)
5048 =for apidoc sv_catsv
5050 Concatenates the string from SV C<ssv> onto the end of the string in
5051 SV C<dsv>. Modifies C<dsv> but not C<ssv>. Handles 'get' magic, but
5052 not 'set' magic. See C<sv_catsv_mg>.
5054 =for apidoc sv_catsv_flags
5056 Concatenates the string from SV C<ssv> onto the end of the string in
5057 SV C<dsv>. Modifies C<dsv> but not C<ssv>. If C<flags> has C<SV_GMAGIC>
5058 bit set, will C<mg_get> on the C<ssv>, if appropriate, before
5059 reading it. If the C<flags> contain C<SV_SMAGIC>, C<mg_set> will be
5060 called on the modified SV afterward, if appropriate. C<sv_catsv>
5061 and C<sv_catsv_nomg> are implemented in terms of this function.
5066 Perl_sv_catsv_flags(pTHX_ SV *const dsv, register SV *const ssv, const I32 flags)
5070 PERL_ARGS_ASSERT_SV_CATSV_FLAGS;
5074 const char *spv = SvPV_flags_const(ssv, slen, flags);
5076 if (SvGMAGICAL(dsv) && (flags & SV_GMAGIC))
5078 sv_catpvn_flags(dsv, spv, slen,
5079 DO_UTF8(ssv) ? SV_CATUTF8 : SV_CATBYTES);
5082 if (flags & SV_SMAGIC)
5087 =for apidoc sv_catpv
5089 Concatenates the string onto the end of the string which is in the SV.
5090 If the SV has the UTF-8 status set, then the bytes appended should be
5091 valid UTF-8. Handles 'get' magic, but not 'set' magic. See C<sv_catpv_mg>.
5096 Perl_sv_catpv(pTHX_ register SV *const sv, register const char *ptr)
5099 register STRLEN len;
5103 PERL_ARGS_ASSERT_SV_CATPV;
5107 junk = SvPV_force(sv, tlen);
5109 SvGROW(sv, tlen + len + 1);
5111 ptr = SvPVX_const(sv);
5112 Move(ptr,SvPVX(sv)+tlen,len+1,char);
5113 SvCUR_set(sv, SvCUR(sv) + len);
5114 (void)SvPOK_only_UTF8(sv); /* validate pointer */
5119 =for apidoc sv_catpv_flags
5121 Concatenates the string onto the end of the string which is in the SV.
5122 If the SV has the UTF-8 status set, then the bytes appended should
5123 be valid UTF-8. If C<flags> has the C<SV_SMAGIC> bit set, will C<mg_set>
5124 on the modified SV if appropriate.
5130 Perl_sv_catpv_flags(pTHX_ SV *dstr, const char *sstr, const I32 flags)
5132 PERL_ARGS_ASSERT_SV_CATPV_FLAGS;
5133 sv_catpvn_flags(dstr, sstr, strlen(sstr), flags);
5137 =for apidoc sv_catpv_mg
5139 Like C<sv_catpv>, but also handles 'set' magic.
5145 Perl_sv_catpv_mg(pTHX_ register SV *const sv, register const char *const ptr)
5147 PERL_ARGS_ASSERT_SV_CATPV_MG;
5156 Creates a new SV. A non-zero C<len> parameter indicates the number of
5157 bytes of preallocated string space the SV should have. An extra byte for a
5158 trailing NUL is also reserved. (SvPOK is not set for the SV even if string
5159 space is allocated.) The reference count for the new SV is set to 1.
5161 In 5.9.3, newSV() replaces the older NEWSV() API, and drops the first
5162 parameter, I<x>, a debug aid which allowed callers to identify themselves.
5163 This aid has been superseded by a new build option, PERL_MEM_LOG (see
5164 L<perlhacktips/PERL_MEM_LOG>). The older API is still there for use in XS
5165 modules supporting older perls.
5171 Perl_newSV(pTHX_ const STRLEN len)
5178 sv_upgrade(sv, SVt_PV);
5179 SvGROW(sv, len + 1);
5184 =for apidoc sv_magicext
5186 Adds magic to an SV, upgrading it if necessary. Applies the
5187 supplied vtable and returns a pointer to the magic added.
5189 Note that C<sv_magicext> will allow things that C<sv_magic> will not.
5190 In particular, you can add magic to SvREADONLY SVs, and add more than
5191 one instance of the same 'how'.
5193 If C<namlen> is greater than zero then a C<savepvn> I<copy> of C<name> is
5194 stored, if C<namlen> is zero then C<name> is stored as-is and - as another
5195 special case - if C<(name && namlen == HEf_SVKEY)> then C<name> is assumed
5196 to contain an C<SV*> and is stored as-is with its REFCNT incremented.
5198 (This is now used as a subroutine by C<sv_magic>.)
5203 Perl_sv_magicext(pTHX_ SV *const sv, SV *const obj, const int how,
5204 const MGVTBL *const vtable, const char *const name, const I32 namlen)
5209 PERL_ARGS_ASSERT_SV_MAGICEXT;
5211 SvUPGRADE(sv, SVt_PVMG);
5212 Newxz(mg, 1, MAGIC);
5213 mg->mg_moremagic = SvMAGIC(sv);
5214 SvMAGIC_set(sv, mg);
5216 /* Sometimes a magic contains a reference loop, where the sv and
5217 object refer to each other. To prevent a reference loop that
5218 would prevent such objects being freed, we look for such loops
5219 and if we find one we avoid incrementing the object refcount.
5221 Note we cannot do this to avoid self-tie loops as intervening RV must
5222 have its REFCNT incremented to keep it in existence.
5225 if (!obj || obj == sv ||
5226 how == PERL_MAGIC_arylen ||
5227 how == PERL_MAGIC_symtab ||
5228 (SvTYPE(obj) == SVt_PVGV &&
5229 (GvSV(obj) == sv || GvHV(obj) == (const HV *)sv
5230 || GvAV(obj) == (const AV *)sv || GvCV(obj) == (const CV *)sv
5231 || GvIOp(obj) == (const IO *)sv || GvFORM(obj) == (const CV *)sv)))
5236 mg->mg_obj = SvREFCNT_inc_simple(obj);
5237 mg->mg_flags |= MGf_REFCOUNTED;
5240 /* Normal self-ties simply pass a null object, and instead of
5241 using mg_obj directly, use the SvTIED_obj macro to produce a
5242 new RV as needed. For glob "self-ties", we are tieing the PVIO
5243 with an RV obj pointing to the glob containing the PVIO. In
5244 this case, to avoid a reference loop, we need to weaken the
5248 if (how == PERL_MAGIC_tiedscalar && SvTYPE(sv) == SVt_PVIO &&
5249 obj && SvROK(obj) && GvIO(SvRV(obj)) == (const IO *)sv)
5255 mg->mg_len = namlen;
5258 mg->mg_ptr = savepvn(name, namlen);
5259 else if (namlen == HEf_SVKEY) {
5260 /* Yes, this is casting away const. This is only for the case of
5261 HEf_SVKEY. I think we need to document this aberation of the
5262 constness of the API, rather than making name non-const, as
5263 that change propagating outwards a long way. */
5264 mg->mg_ptr = (char*)SvREFCNT_inc_simple_NN((SV *)name);
5266 mg->mg_ptr = (char *) name;
5268 mg->mg_virtual = (MGVTBL *) vtable;
5272 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
5277 =for apidoc sv_magic
5279 Adds magic to an SV. First upgrades C<sv> to type C<SVt_PVMG> if
5280 necessary, then adds a new magic item of type C<how> to the head of the
5283 See C<sv_magicext> (which C<sv_magic> now calls) for a description of the
5284 handling of the C<name> and C<namlen> arguments.
5286 You need to use C<sv_magicext> to add magic to SvREADONLY SVs and also
5287 to add more than one instance of the same 'how'.
5293 Perl_sv_magic(pTHX_ register SV *const sv, SV *const obj, const int how,
5294 const char *const name, const I32 namlen)
5297 const MGVTBL *vtable;
5300 unsigned int vtable_index;
5302 PERL_ARGS_ASSERT_SV_MAGIC;
5304 if (how < 0 || (unsigned)how > C_ARRAY_LENGTH(PL_magic_data)
5305 || ((flags = PL_magic_data[how]),
5306 (vtable_index = flags & PERL_MAGIC_VTABLE_MASK)
5307 > magic_vtable_max))
5308 Perl_croak(aTHX_ "Don't know how to handle magic of type \\%o", how);
5310 /* PERL_MAGIC_ext is reserved for use by extensions not perl internals.
5311 Useful for attaching extension internal data to perl vars.
5312 Note that multiple extensions may clash if magical scalars
5313 etc holding private data from one are passed to another. */
5315 vtable = (vtable_index == magic_vtable_max)
5316 ? NULL : PL_magic_vtables + vtable_index;
5318 #ifdef PERL_OLD_COPY_ON_WRITE
5320 sv_force_normal_flags(sv, 0);
5322 if (SvREADONLY(sv)) {
5324 /* its okay to attach magic to shared strings */
5328 && !PERL_MAGIC_TYPE_READONLY_ACCEPTABLE(how)
5331 Perl_croak_no_modify(aTHX);
5334 if (SvMAGICAL(sv) || (how == PERL_MAGIC_taint && SvTYPE(sv) >= SVt_PVMG)) {
5335 if (SvMAGIC(sv) && (mg = mg_find(sv, how))) {
5336 /* sv_magic() refuses to add a magic of the same 'how' as an
5339 if (how == PERL_MAGIC_taint) {
5341 /* Any scalar which already had taint magic on which someone
5342 (erroneously?) did SvIOK_on() or similar will now be
5343 incorrectly sporting public "OK" flags. */
5344 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
5350 /* Rest of work is done else where */
5351 mg = sv_magicext(sv,obj,how,vtable,name,namlen);
5354 case PERL_MAGIC_taint:
5357 case PERL_MAGIC_ext:
5358 case PERL_MAGIC_dbfile:
5365 S_sv_unmagicext_flags(pTHX_ SV *const sv, const int type, MGVTBL *vtbl, const U32 flags)
5372 if (SvTYPE(sv) < SVt_PVMG || !SvMAGIC(sv))
5374 mgp = &(((XPVMG*) SvANY(sv))->xmg_u.xmg_magic);
5375 for (mg = *mgp; mg; mg = *mgp) {
5376 const MGVTBL* const virt = mg->mg_virtual;
5377 if (mg->mg_type == type && (!flags || virt == vtbl)) {
5378 *mgp = mg->mg_moremagic;
5379 if (virt && virt->svt_free)
5380 virt->svt_free(aTHX_ sv, mg);
5381 if (mg->mg_ptr && mg->mg_type != PERL_MAGIC_regex_global) {
5383 Safefree(mg->mg_ptr);
5384 else if (mg->mg_len == HEf_SVKEY)
5385 SvREFCNT_dec(MUTABLE_SV(mg->mg_ptr));
5386 else if (mg->mg_type == PERL_MAGIC_utf8)
5387 Safefree(mg->mg_ptr);
5389 if (mg->mg_flags & MGf_REFCOUNTED)
5390 SvREFCNT_dec(mg->mg_obj);
5394 mgp = &mg->mg_moremagic;
5397 if (SvMAGICAL(sv)) /* if we're under save_magic, wait for restore_magic; */
5398 mg_magical(sv); /* else fix the flags now */
5402 SvFLAGS(sv) |= (SvFLAGS(sv) & (SVp_IOK|SVp_NOK|SVp_POK)) >> PRIVSHIFT;
5408 =for apidoc sv_unmagic
5410 Removes all magic of type C<type> from an SV.
5416 Perl_sv_unmagic(pTHX_ SV *const sv, const int type)
5418 PERL_ARGS_ASSERT_SV_UNMAGIC;
5419 return S_sv_unmagicext_flags(aTHX_ sv, type, NULL, 0);
5423 =for apidoc sv_unmagicext
5425 Removes all magic of type C<type> with the specified C<vtbl> from an SV.
5431 Perl_sv_unmagicext(pTHX_ SV *const sv, const int type, MGVTBL *vtbl)
5433 PERL_ARGS_ASSERT_SV_UNMAGICEXT;
5434 return S_sv_unmagicext_flags(aTHX_ sv, type, vtbl, 1);
5438 =for apidoc sv_rvweaken
5440 Weaken a reference: set the C<SvWEAKREF> flag on this RV; give the
5441 referred-to SV C<PERL_MAGIC_backref> magic if it hasn't already; and
5442 push a back-reference to this RV onto the array of backreferences
5443 associated with that magic. If the RV is magical, set magic will be
5444 called after the RV is cleared.
5450 Perl_sv_rvweaken(pTHX_ SV *const sv)
5454 PERL_ARGS_ASSERT_SV_RVWEAKEN;
5456 if (!SvOK(sv)) /* let undefs pass */
5459 Perl_croak(aTHX_ "Can't weaken a nonreference");
5460 else if (SvWEAKREF(sv)) {
5461 Perl_ck_warner(aTHX_ packWARN(WARN_MISC), "Reference is already weak");
5464 else if (SvREADONLY(sv)) croak_no_modify();
5466 Perl_sv_add_backref(aTHX_ tsv, sv);
5472 /* Give tsv backref magic if it hasn't already got it, then push a
5473 * back-reference to sv onto the array associated with the backref magic.
5475 * As an optimisation, if there's only one backref and it's not an AV,
5476 * store it directly in the HvAUX or mg_obj slot, avoiding the need to
5477 * allocate an AV. (Whether the slot holds an AV tells us whether this is
5481 /* A discussion about the backreferences array and its refcount:
5483 * The AV holding the backreferences is pointed to either as the mg_obj of
5484 * PERL_MAGIC_backref, or in the specific case of a HV, from the
5485 * xhv_backreferences field. The array is created with a refcount
5486 * of 2. This means that if during global destruction the array gets
5487 * picked on before its parent to have its refcount decremented by the
5488 * random zapper, it won't actually be freed, meaning it's still there for
5489 * when its parent gets freed.
5491 * When the parent SV is freed, the extra ref is killed by
5492 * Perl_sv_kill_backrefs. The other ref is killed, in the case of magic,
5493 * by mg_free() / MGf_REFCOUNTED, or for a hash, by Perl_hv_kill_backrefs.
5495 * When a single backref SV is stored directly, it is not reference
5500 Perl_sv_add_backref(pTHX_ SV *const tsv, SV *const sv)
5507 PERL_ARGS_ASSERT_SV_ADD_BACKREF;
5509 /* find slot to store array or singleton backref */
5511 if (SvTYPE(tsv) == SVt_PVHV) {
5512 svp = (SV**)Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5515 (SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL))))
5517 sv_magic(tsv, NULL, PERL_MAGIC_backref, NULL, 0);
5518 mg = mg_find(tsv, PERL_MAGIC_backref);
5520 svp = &(mg->mg_obj);
5523 /* create or retrieve the array */
5525 if ( (!*svp && SvTYPE(sv) == SVt_PVAV)
5526 || (*svp && SvTYPE(*svp) != SVt_PVAV)
5531 SvREFCNT_inc_simple_void(av);
5532 /* av now has a refcnt of 2; see discussion above */
5534 /* move single existing backref to the array */
5536 AvARRAY(av)[++AvFILLp(av)] = *svp; /* av_push() */
5540 mg->mg_flags |= MGf_REFCOUNTED;
5543 av = MUTABLE_AV(*svp);
5546 /* optimisation: store single backref directly in HvAUX or mg_obj */
5550 /* push new backref */
5551 assert(SvTYPE(av) == SVt_PVAV);
5552 if (AvFILLp(av) >= AvMAX(av)) {
5553 av_extend(av, AvFILLp(av)+1);
5555 AvARRAY(av)[++AvFILLp(av)] = sv; /* av_push() */
5558 /* delete a back-reference to ourselves from the backref magic associated
5559 * with the SV we point to.
5563 Perl_sv_del_backref(pTHX_ SV *const tsv, SV *const sv)
5568 PERL_ARGS_ASSERT_SV_DEL_BACKREF;
5570 if (SvTYPE(tsv) == SVt_PVHV) {
5572 svp = (SV**)Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5576 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
5577 svp = mg ? &(mg->mg_obj) : NULL;
5581 Perl_croak(aTHX_ "panic: del_backref, svp=0");
5583 /* It's possible that sv is being freed recursively part way through the
5584 freeing of tsv. If this happens, the backreferences array of tsv has
5585 already been freed, and so svp will be NULL. If this is the case,
5586 we should not panic. Instead, nothing needs doing, so return. */
5587 if (PL_phase == PERL_PHASE_DESTRUCT && SvREFCNT(tsv) == 0)
5589 Perl_croak(aTHX_ "panic: del_backref, *svp=%p phase=%s refcnt=%" UVuf,
5590 *svp, PL_phase_names[PL_phase], (UV)SvREFCNT(tsv));
5593 if (SvTYPE(*svp) == SVt_PVAV) {
5597 AV * const av = (AV*)*svp;
5599 assert(!SvIS_FREED(av));
5603 /* for an SV with N weak references to it, if all those
5604 * weak refs are deleted, then sv_del_backref will be called
5605 * N times and O(N^2) compares will be done within the backref
5606 * array. To ameliorate this potential slowness, we:
5607 * 1) make sure this code is as tight as possible;
5608 * 2) when looking for SV, look for it at both the head and tail of the
5609 * array first before searching the rest, since some create/destroy
5610 * patterns will cause the backrefs to be freed in order.
5617 SV **p = &svp[fill];
5618 SV *const topsv = *p;
5625 /* We weren't the last entry.
5626 An unordered list has this property that you
5627 can take the last element off the end to fill
5628 the hole, and it's still an unordered list :-)
5634 break; /* should only be one */
5641 AvFILLp(av) = fill-1;
5643 else if (SvIS_FREED(*svp) && PL_phase == PERL_PHASE_DESTRUCT) {
5644 /* freed AV; skip */
5647 /* optimisation: only a single backref, stored directly */
5649 Perl_croak(aTHX_ "panic: del_backref, *svp=%p, sv=%p", *svp, sv);
5656 Perl_sv_kill_backrefs(pTHX_ SV *const sv, AV *const av)
5662 PERL_ARGS_ASSERT_SV_KILL_BACKREFS;
5667 /* after multiple passes through Perl_sv_clean_all() for a thinngy
5668 * that has badly leaked, the backref array may have gotten freed,
5669 * since we only protect it against 1 round of cleanup */
5670 if (SvIS_FREED(av)) {
5671 if (PL_in_clean_all) /* All is fair */
5674 "panic: magic_killbackrefs (freed backref AV/SV)");
5678 is_array = (SvTYPE(av) == SVt_PVAV);
5680 assert(!SvIS_FREED(av));
5683 last = svp + AvFILLp(av);
5686 /* optimisation: only a single backref, stored directly */
5692 while (svp <= last) {
5694 SV *const referrer = *svp;
5695 if (SvWEAKREF(referrer)) {
5696 /* XXX Should we check that it hasn't changed? */
5697 assert(SvROK(referrer));
5698 SvRV_set(referrer, 0);
5700 SvWEAKREF_off(referrer);
5701 SvSETMAGIC(referrer);
5702 } else if (SvTYPE(referrer) == SVt_PVGV ||
5703 SvTYPE(referrer) == SVt_PVLV) {
5704 assert(SvTYPE(sv) == SVt_PVHV); /* stash backref */
5705 /* You lookin' at me? */
5706 assert(GvSTASH(referrer));
5707 assert(GvSTASH(referrer) == (const HV *)sv);
5708 GvSTASH(referrer) = 0;
5709 } else if (SvTYPE(referrer) == SVt_PVCV ||
5710 SvTYPE(referrer) == SVt_PVFM) {
5711 if (SvTYPE(sv) == SVt_PVHV) { /* stash backref */
5712 /* You lookin' at me? */
5713 assert(CvSTASH(referrer));
5714 assert(CvSTASH(referrer) == (const HV *)sv);
5715 SvANY(MUTABLE_CV(referrer))->xcv_stash = 0;
5718 assert(SvTYPE(sv) == SVt_PVGV);
5719 /* You lookin' at me? */
5720 assert(CvGV(referrer));
5721 assert(CvGV(referrer) == (const GV *)sv);
5722 anonymise_cv_maybe(MUTABLE_GV(sv),
5723 MUTABLE_CV(referrer));
5728 "panic: magic_killbackrefs (flags=%"UVxf")",
5729 (UV)SvFLAGS(referrer));
5740 SvREFCNT_dec(av); /* remove extra count added by sv_add_backref() */
5746 =for apidoc sv_insert
5748 Inserts a string at the specified offset/length within the SV. Similar to
5749 the Perl substr() function. Handles get magic.
5751 =for apidoc sv_insert_flags
5753 Same as C<sv_insert>, but the extra C<flags> are passed to the
5754 C<SvPV_force_flags> that applies to C<bigstr>.
5760 Perl_sv_insert_flags(pTHX_ SV *const bigstr, const STRLEN offset, const STRLEN len, const char *const little, const STRLEN littlelen, const U32 flags)
5765 register char *midend;
5766 register char *bigend;
5767 register SSize_t i; /* better be sizeof(STRLEN) or bad things happen */
5770 PERL_ARGS_ASSERT_SV_INSERT_FLAGS;
5773 Perl_croak(aTHX_ "Can't modify nonexistent substring");
5774 SvPV_force_flags(bigstr, curlen, flags);
5775 (void)SvPOK_only_UTF8(bigstr);
5776 if (offset + len > curlen) {
5777 SvGROW(bigstr, offset+len+1);
5778 Zero(SvPVX(bigstr)+curlen, offset+len-curlen, char);
5779 SvCUR_set(bigstr, offset+len);
5783 i = littlelen - len;
5784 if (i > 0) { /* string might grow */
5785 big = SvGROW(bigstr, SvCUR(bigstr) + i + 1);
5786 mid = big + offset + len;
5787 midend = bigend = big + SvCUR(bigstr);
5790 while (midend > mid) /* shove everything down */
5791 *--bigend = *--midend;
5792 Move(little,big+offset,littlelen,char);
5793 SvCUR_set(bigstr, SvCUR(bigstr) + i);
5798 Move(little,SvPVX(bigstr)+offset,len,char);
5803 big = SvPVX(bigstr);
5806 bigend = big + SvCUR(bigstr);
5808 if (midend > bigend)
5809 Perl_croak(aTHX_ "panic: sv_insert, midend=%p, bigend=%p",
5812 if (mid - big > bigend - midend) { /* faster to shorten from end */
5814 Move(little, mid, littlelen,char);
5817 i = bigend - midend;
5819 Move(midend, mid, i,char);
5823 SvCUR_set(bigstr, mid - big);
5825 else if ((i = mid - big)) { /* faster from front */
5826 midend -= littlelen;
5828 Move(big, midend - i, i, char);
5829 sv_chop(bigstr,midend-i);
5831 Move(little, mid, littlelen,char);
5833 else if (littlelen) {
5834 midend -= littlelen;
5835 sv_chop(bigstr,midend);
5836 Move(little,midend,littlelen,char);
5839 sv_chop(bigstr,midend);
5845 =for apidoc sv_replace
5847 Make the first argument a copy of the second, then delete the original.
5848 The target SV physically takes over ownership of the body of the source SV
5849 and inherits its flags; however, the target keeps any magic it owns,
5850 and any magic in the source is discarded.
5851 Note that this is a rather specialist SV copying operation; most of the
5852 time you'll want to use C<sv_setsv> or one of its many macro front-ends.
5858 Perl_sv_replace(pTHX_ register SV *const sv, register SV *const nsv)
5861 const U32 refcnt = SvREFCNT(sv);
5863 PERL_ARGS_ASSERT_SV_REPLACE;
5865 SV_CHECK_THINKFIRST_COW_DROP(sv);
5866 if (SvREFCNT(nsv) != 1) {
5867 Perl_croak(aTHX_ "panic: reference miscount on nsv in sv_replace()"
5868 " (%" UVuf " != 1)", (UV) SvREFCNT(nsv));
5870 if (SvMAGICAL(sv)) {
5874 sv_upgrade(nsv, SVt_PVMG);
5875 SvMAGIC_set(nsv, SvMAGIC(sv));
5876 SvFLAGS(nsv) |= SvMAGICAL(sv);
5878 SvMAGIC_set(sv, NULL);
5882 assert(!SvREFCNT(sv));
5883 #ifdef DEBUG_LEAKING_SCALARS
5884 sv->sv_flags = nsv->sv_flags;
5885 sv->sv_any = nsv->sv_any;
5886 sv->sv_refcnt = nsv->sv_refcnt;
5887 sv->sv_u = nsv->sv_u;
5889 StructCopy(nsv,sv,SV);
5891 if(SvTYPE(sv) == SVt_IV) {
5893 = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
5897 #ifdef PERL_OLD_COPY_ON_WRITE
5898 if (SvIsCOW_normal(nsv)) {
5899 /* We need to follow the pointers around the loop to make the
5900 previous SV point to sv, rather than nsv. */
5903 while ((next = SV_COW_NEXT_SV(current)) != nsv) {
5906 assert(SvPVX_const(current) == SvPVX_const(nsv));
5908 /* Make the SV before us point to the SV after us. */
5910 PerlIO_printf(Perl_debug_log, "previous is\n");
5912 PerlIO_printf(Perl_debug_log,
5913 "move it from 0x%"UVxf" to 0x%"UVxf"\n",
5914 (UV) SV_COW_NEXT_SV(current), (UV) sv);
5916 SV_COW_NEXT_SV_SET(current, sv);
5919 SvREFCNT(sv) = refcnt;
5920 SvFLAGS(nsv) |= SVTYPEMASK; /* Mark as freed */
5925 /* We're about to free a GV which has a CV that refers back to us.
5926 * If that CV will outlive us, make it anonymous (i.e. fix up its CvGV
5930 S_anonymise_cv_maybe(pTHX_ GV *gv, CV* cv)
5935 PERL_ARGS_ASSERT_ANONYMISE_CV_MAYBE;
5938 assert(SvREFCNT(gv) == 0);
5939 assert(isGV(gv) && isGV_with_GP(gv));
5941 assert(!CvANON(cv));
5942 assert(CvGV(cv) == gv);
5944 /* will the CV shortly be freed by gp_free() ? */
5945 if (GvCV(gv) == cv && GvGP(gv)->gp_refcnt < 2 && SvREFCNT(cv) < 2) {
5946 SvANY(cv)->xcv_gv = NULL;
5950 /* if not, anonymise: */
5951 gvname = (GvSTASH(gv) && HvNAME(GvSTASH(gv)) && HvENAME(GvSTASH(gv)))
5952 ? newSVhek(HvENAME_HEK(GvSTASH(gv)))
5953 : newSVpvn_flags( "__ANON__", 8, 0 );
5954 sv_catpvs(gvname, "::__ANON__");
5955 anongv = gv_fetchsv(gvname, GV_ADDMULTI, SVt_PVCV);
5956 SvREFCNT_dec(gvname);
5960 SvANY(cv)->xcv_gv = MUTABLE_GV(SvREFCNT_inc(anongv));
5965 =for apidoc sv_clear
5967 Clear an SV: call any destructors, free up any memory used by the body,
5968 and free the body itself. The SV's head is I<not> freed, although
5969 its type is set to all 1's so that it won't inadvertently be assumed
5970 to be live during global destruction etc.
5971 This function should only be called when REFCNT is zero. Most of the time
5972 you'll want to call C<sv_free()> (or its macro wrapper C<SvREFCNT_dec>)
5979 Perl_sv_clear(pTHX_ SV *const orig_sv)
5984 const struct body_details *sv_type_details;
5987 register SV *sv = orig_sv;
5990 PERL_ARGS_ASSERT_SV_CLEAR;
5992 /* within this loop, sv is the SV currently being freed, and
5993 * iter_sv is the most recent AV or whatever that's being iterated
5994 * over to provide more SVs */
6000 assert(SvREFCNT(sv) == 0);
6001 assert(SvTYPE(sv) != (svtype)SVTYPEMASK);
6003 if (type <= SVt_IV) {
6004 /* See the comment in sv.h about the collusion between this
6005 * early return and the overloading of the NULL slots in the
6009 SvFLAGS(sv) &= SVf_BREAK;
6010 SvFLAGS(sv) |= SVTYPEMASK;
6014 assert(!SvOBJECT(sv) || type >= SVt_PVMG); /* objs are always >= MG */
6016 if (type >= SVt_PVMG) {
6018 if (!curse(sv, 1)) goto get_next_sv;
6019 type = SvTYPE(sv); /* destructor may have changed it */
6021 /* Free back-references before magic, in case the magic calls
6022 * Perl code that has weak references to sv. */
6023 if (type == SVt_PVHV) {
6024 Perl_hv_kill_backrefs(aTHX_ MUTABLE_HV(sv));
6028 else if (type == SVt_PVMG && SvPAD_OUR(sv)) {
6029 SvREFCNT_dec(SvOURSTASH(sv));
6030 } else if (SvMAGIC(sv)) {
6031 /* Free back-references before other types of magic. */
6032 sv_unmagic(sv, PERL_MAGIC_backref);
6035 if (type == SVt_PVMG && SvPAD_TYPED(sv))
6036 SvREFCNT_dec(SvSTASH(sv));
6039 /* case SVt_BIND: */
6042 IoIFP(sv) != PerlIO_stdin() &&
6043 IoIFP(sv) != PerlIO_stdout() &&
6044 IoIFP(sv) != PerlIO_stderr() &&
6045 !(IoFLAGS(sv) & IOf_FAKE_DIRP))
6047 io_close(MUTABLE_IO(sv), FALSE);
6049 if (IoDIRP(sv) && !(IoFLAGS(sv) & IOf_FAKE_DIRP))
6050 PerlDir_close(IoDIRP(sv));
6051 IoDIRP(sv) = (DIR*)NULL;
6052 Safefree(IoTOP_NAME(sv));
6053 Safefree(IoFMT_NAME(sv));
6054 Safefree(IoBOTTOM_NAME(sv));
6055 if ((const GV *)sv == PL_statgv)
6059 /* FIXME for plugins */
6060 pregfree2((REGEXP*) sv);
6064 cv_undef(MUTABLE_CV(sv));
6065 /* If we're in a stash, we don't own a reference to it.
6066 * However it does have a back reference to us, which needs to
6068 if ((stash = CvSTASH(sv)))
6069 sv_del_backref(MUTABLE_SV(stash), sv);
6072 if (PL_last_swash_hv == (const HV *)sv) {
6073 PL_last_swash_hv = NULL;
6075 if (HvTOTALKEYS((HV*)sv) > 0) {
6077 /* this statement should match the one at the beginning of
6078 * hv_undef_flags() */
6079 if ( PL_phase != PERL_PHASE_DESTRUCT
6080 && (name = HvNAME((HV*)sv)))
6083 (void)hv_delete(PL_stashcache, name,
6084 HvNAMEUTF8((HV*)sv) ? -HvNAMELEN_get((HV*)sv) : HvNAMELEN_get((HV*)sv), G_DISCARD);
6085 hv_name_set((HV*)sv, NULL, 0, 0);
6088 /* save old iter_sv in unused SvSTASH field */
6089 assert(!SvOBJECT(sv));
6090 SvSTASH(sv) = (HV*)iter_sv;
6093 /* XXX ideally we should save the old value of hash_index
6094 * too, but I can't think of any place to hide it. The
6095 * effect of not saving it is that for freeing hashes of
6096 * hashes, we become quadratic in scanning the HvARRAY of
6097 * the top hash looking for new entries to free; but
6098 * hopefully this will be dwarfed by the freeing of all
6099 * the nested hashes. */
6101 next_sv = Perl_hfree_next_entry(aTHX_ (HV*)sv, &hash_index);
6102 goto get_next_sv; /* process this new sv */
6104 /* free empty hash */
6105 Perl_hv_undef_flags(aTHX_ MUTABLE_HV(sv), HV_NAME_SETALL);
6106 assert(!HvARRAY((HV*)sv));
6110 AV* av = MUTABLE_AV(sv);
6111 if (PL_comppad == av) {
6115 if (AvREAL(av) && AvFILLp(av) > -1) {
6116 next_sv = AvARRAY(av)[AvFILLp(av)--];
6117 /* save old iter_sv in top-most slot of AV,
6118 * and pray that it doesn't get wiped in the meantime */
6119 AvARRAY(av)[AvMAX(av)] = iter_sv;
6121 goto get_next_sv; /* process this new sv */
6123 Safefree(AvALLOC(av));
6128 if (LvTYPE(sv) == 'T') { /* for tie: return HE to pool */
6129 SvREFCNT_dec(HeKEY_sv((HE*)LvTARG(sv)));
6130 HeNEXT((HE*)LvTARG(sv)) = PL_hv_fetch_ent_mh;
6131 PL_hv_fetch_ent_mh = (HE*)LvTARG(sv);
6133 else if (LvTYPE(sv) != 't') /* unless tie: unrefcnted fake SV** */
6134 SvREFCNT_dec(LvTARG(sv));
6136 if (isGV_with_GP(sv)) {
6137 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
6138 && HvENAME_get(stash))
6139 mro_method_changed_in(stash);
6140 gp_free(MUTABLE_GV(sv));
6142 unshare_hek(GvNAME_HEK(sv));
6143 /* If we're in a stash, we don't own a reference to it.
6144 * However it does have a back reference to us, which
6145 * needs to be cleared. */
6146 if (!SvVALID(sv) && (stash = GvSTASH(sv)))
6147 sv_del_backref(MUTABLE_SV(stash), sv);
6149 /* FIXME. There are probably more unreferenced pointers to SVs
6150 * in the interpreter struct that we should check and tidy in
6151 * a similar fashion to this: */
6152 /* See also S_sv_unglob, which does the same thing. */
6153 if ((const GV *)sv == PL_last_in_gv)
6154 PL_last_in_gv = NULL;
6155 else if ((const GV *)sv == PL_statgv)
6162 /* Don't bother with SvOOK_off(sv); as we're only going to
6166 SvOOK_offset(sv, offset);
6167 SvPV_set(sv, SvPVX_mutable(sv) - offset);
6168 /* Don't even bother with turning off the OOK flag. */
6173 SV * const target = SvRV(sv);
6175 sv_del_backref(target, sv);
6180 #ifdef PERL_OLD_COPY_ON_WRITE
6181 else if (SvPVX_const(sv)
6182 && !(SvTYPE(sv) == SVt_PVIO
6183 && !(IoFLAGS(sv) & IOf_FAKE_DIRP)))
6187 PerlIO_printf(Perl_debug_log, "Copy on write: clear\n");
6191 sv_release_COW(sv, SvPVX_const(sv), SV_COW_NEXT_SV(sv));
6193 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
6197 } else if (SvLEN(sv)) {
6198 Safefree(SvPVX_const(sv));
6202 else if (SvPVX_const(sv) && SvLEN(sv)
6203 && !(SvTYPE(sv) == SVt_PVIO
6204 && !(IoFLAGS(sv) & IOf_FAKE_DIRP)))
6205 Safefree(SvPVX_mutable(sv));
6206 else if (SvPVX_const(sv) && SvIsCOW(sv)) {
6207 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
6218 SvFLAGS(sv) &= SVf_BREAK;
6219 SvFLAGS(sv) |= SVTYPEMASK;
6221 sv_type_details = bodies_by_type + type;
6222 if (sv_type_details->arena) {
6223 del_body(((char *)SvANY(sv) + sv_type_details->offset),
6224 &PL_body_roots[type]);
6226 else if (sv_type_details->body_size) {
6227 safefree(SvANY(sv));
6231 /* caller is responsible for freeing the head of the original sv */
6232 if (sv != orig_sv && !SvREFCNT(sv))
6235 /* grab and free next sv, if any */
6243 else if (!iter_sv) {
6245 } else if (SvTYPE(iter_sv) == SVt_PVAV) {
6246 AV *const av = (AV*)iter_sv;
6247 if (AvFILLp(av) > -1) {
6248 sv = AvARRAY(av)[AvFILLp(av)--];
6250 else { /* no more elements of current AV to free */
6253 /* restore previous value, squirrelled away */
6254 iter_sv = AvARRAY(av)[AvMAX(av)];
6255 Safefree(AvALLOC(av));
6258 } else if (SvTYPE(iter_sv) == SVt_PVHV) {
6259 sv = Perl_hfree_next_entry(aTHX_ (HV*)iter_sv, &hash_index);
6260 if (!sv && !HvTOTALKEYS((HV *)iter_sv)) {
6261 /* no more elements of current HV to free */
6264 /* Restore previous value of iter_sv, squirrelled away */
6265 assert(!SvOBJECT(sv));
6266 iter_sv = (SV*)SvSTASH(sv);
6268 /* ideally we should restore the old hash_index here,
6269 * but we don't currently save the old value */
6272 /* free any remaining detritus from the hash struct */
6273 Perl_hv_undef_flags(aTHX_ MUTABLE_HV(sv), HV_NAME_SETALL);
6274 assert(!HvARRAY((HV*)sv));
6279 /* unrolled SvREFCNT_dec and sv_free2 follows: */
6283 if (!SvREFCNT(sv)) {
6287 if (--(SvREFCNT(sv)))
6291 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEBUGGING),
6292 "Attempt to free temp prematurely: SV 0x%"UVxf
6293 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
6297 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
6298 /* make sure SvREFCNT(sv)==0 happens very seldom */
6299 SvREFCNT(sv) = (~(U32)0)/2;
6308 /* This routine curses the sv itself, not the object referenced by sv. So
6309 sv does not have to be ROK. */
6312 S_curse(pTHX_ SV * const sv, const bool check_refcnt) {
6315 PERL_ARGS_ASSERT_CURSE;
6316 assert(SvOBJECT(sv));
6318 if (PL_defstash && /* Still have a symbol table? */
6325 stash = SvSTASH(sv);
6326 destructor = StashHANDLER(stash,DESTROY);
6328 /* A constant subroutine can have no side effects, so
6329 don't bother calling it. */
6330 && !CvCONST(destructor)
6331 /* Don't bother calling an empty destructor or one that
6332 returns immediately. */
6333 && (CvISXSUB(destructor)
6334 || (CvSTART(destructor)
6335 && (CvSTART(destructor)->op_next->op_type
6337 && (CvSTART(destructor)->op_next->op_type
6339 || CvSTART(destructor)->op_next->op_next->op_type
6345 SV* const tmpref = newRV(sv);
6346 SvREADONLY_on(tmpref); /* DESTROY() could be naughty */
6348 PUSHSTACKi(PERLSI_DESTROY);
6353 call_sv(MUTABLE_SV(destructor),
6354 G_DISCARD|G_EVAL|G_KEEPERR|G_VOID);
6358 if(SvREFCNT(tmpref) < 2) {
6359 /* tmpref is not kept alive! */
6361 SvRV_set(tmpref, NULL);
6364 SvREFCNT_dec(tmpref);
6366 } while (SvOBJECT(sv) && SvSTASH(sv) != stash);
6369 if (check_refcnt && SvREFCNT(sv)) {
6370 if (PL_in_clean_objs)
6372 "DESTROY created new reference to dead object '%"HEKf"'",
6373 HEKfARG(HvNAME_HEK(stash)));
6374 /* DESTROY gave object new lease on life */
6380 SvREFCNT_dec(SvSTASH(sv)); /* possibly of changed persuasion */
6381 SvOBJECT_off(sv); /* Curse the object. */
6382 if (SvTYPE(sv) != SVt_PVIO)
6383 --PL_sv_objcount;/* XXX Might want something more general */
6389 =for apidoc sv_newref
6391 Increment an SV's reference count. Use the C<SvREFCNT_inc()> wrapper
6398 Perl_sv_newref(pTHX_ SV *const sv)
6400 PERL_UNUSED_CONTEXT;
6409 Decrement an SV's reference count, and if it drops to zero, call
6410 C<sv_clear> to invoke destructors and free up any memory used by
6411 the body; finally, deallocate the SV's head itself.
6412 Normally called via a wrapper macro C<SvREFCNT_dec>.
6418 Perl_sv_free(pTHX_ SV *const sv)
6423 if (SvREFCNT(sv) == 0) {
6424 if (SvFLAGS(sv) & SVf_BREAK)
6425 /* this SV's refcnt has been artificially decremented to
6426 * trigger cleanup */
6428 if (PL_in_clean_all) /* All is fair */
6430 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
6431 /* make sure SvREFCNT(sv)==0 happens very seldom */
6432 SvREFCNT(sv) = (~(U32)0)/2;
6435 if (ckWARN_d(WARN_INTERNAL)) {
6436 #ifdef DEBUG_LEAKING_SCALARS_FORK_DUMP
6437 Perl_dump_sv_child(aTHX_ sv);
6439 #ifdef DEBUG_LEAKING_SCALARS
6442 #ifdef DEBUG_LEAKING_SCALARS_ABORT
6443 if (PL_warnhook == PERL_WARNHOOK_FATAL
6444 || ckDEAD(packWARN(WARN_INTERNAL))) {
6445 /* Don't let Perl_warner cause us to escape our fate: */
6449 /* This may not return: */
6450 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
6451 "Attempt to free unreferenced scalar: SV 0x%"UVxf
6452 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
6455 #ifdef DEBUG_LEAKING_SCALARS_ABORT
6460 if (--(SvREFCNT(sv)) > 0)
6462 Perl_sv_free2(aTHX_ sv);
6466 Perl_sv_free2(pTHX_ SV *const sv)
6470 PERL_ARGS_ASSERT_SV_FREE2;
6474 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEBUGGING),
6475 "Attempt to free temp prematurely: SV 0x%"UVxf
6476 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
6480 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
6481 /* make sure SvREFCNT(sv)==0 happens very seldom */
6482 SvREFCNT(sv) = (~(U32)0)/2;
6493 Returns the length of the string in the SV. Handles magic and type
6494 coercion. See also C<SvCUR>, which gives raw access to the xpv_cur slot.
6500 Perl_sv_len(pTHX_ register SV *const sv)
6508 len = mg_length(sv);
6510 (void)SvPV_const(sv, len);
6515 =for apidoc sv_len_utf8
6517 Returns the number of characters in the string in an SV, counting wide
6518 UTF-8 bytes as a single character. Handles magic and type coercion.
6524 * The length is cached in PERL_MAGIC_utf8, in the mg_len field. Also the
6525 * mg_ptr is used, by sv_pos_u2b() and sv_pos_b2u() - see the comments below.
6526 * (Note that the mg_len is not the length of the mg_ptr field.
6527 * This allows the cache to store the character length of the string without
6528 * needing to malloc() extra storage to attach to the mg_ptr.)
6533 Perl_sv_len_utf8(pTHX_ register SV *const sv)
6539 return mg_length(sv);
6543 const U8 *s = (U8*)SvPV_const(sv, len);
6547 MAGIC *mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_utf8) : NULL;
6549 if (mg && (mg->mg_len != -1 || mg->mg_ptr)) {
6550 if (mg->mg_len != -1)
6553 /* We can use the offset cache for a headstart.
6554 The longer value is stored in the first pair. */
6555 STRLEN *cache = (STRLEN *) mg->mg_ptr;
6557 ulen = cache[0] + Perl_utf8_length(aTHX_ s + cache[1],
6561 if (PL_utf8cache < 0) {
6562 const STRLEN real = Perl_utf8_length(aTHX_ s, s + len);
6563 assert_uft8_cache_coherent("sv_len_utf8", ulen, real, sv);
6567 ulen = Perl_utf8_length(aTHX_ s, s + len);
6568 utf8_mg_len_cache_update(sv, &mg, ulen);
6572 return Perl_utf8_length(aTHX_ s, s + len);
6576 /* Walk forwards to find the byte corresponding to the passed in UTF-8
6579 S_sv_pos_u2b_forwards(const U8 *const start, const U8 *const send,
6580 STRLEN *const uoffset_p, bool *const at_end)
6582 const U8 *s = start;
6583 STRLEN uoffset = *uoffset_p;
6585 PERL_ARGS_ASSERT_SV_POS_U2B_FORWARDS;
6587 while (s < send && uoffset) {
6594 else if (s > send) {
6596 /* This is the existing behaviour. Possibly it should be a croak, as
6597 it's actually a bounds error */
6600 *uoffset_p -= uoffset;
6604 /* Given the length of the string in both bytes and UTF-8 characters, decide
6605 whether to walk forwards or backwards to find the byte corresponding to
6606 the passed in UTF-8 offset. */
6608 S_sv_pos_u2b_midway(const U8 *const start, const U8 *send,
6609 STRLEN uoffset, const STRLEN uend)
6611 STRLEN backw = uend - uoffset;
6613 PERL_ARGS_ASSERT_SV_POS_U2B_MIDWAY;
6615 if (uoffset < 2 * backw) {
6616 /* The assumption is that going forwards is twice the speed of going
6617 forward (that's where the 2 * backw comes from).
6618 (The real figure of course depends on the UTF-8 data.) */
6619 const U8 *s = start;
6621 while (s < send && uoffset--)
6631 while (UTF8_IS_CONTINUATION(*send))
6634 return send - start;
6637 /* For the string representation of the given scalar, find the byte
6638 corresponding to the passed in UTF-8 offset. uoffset0 and boffset0
6639 give another position in the string, *before* the sought offset, which
6640 (which is always true, as 0, 0 is a valid pair of positions), which should
6641 help reduce the amount of linear searching.
6642 If *mgp is non-NULL, it should point to the UTF-8 cache magic, which
6643 will be used to reduce the amount of linear searching. The cache will be
6644 created if necessary, and the found value offered to it for update. */
6646 S_sv_pos_u2b_cached(pTHX_ SV *const sv, MAGIC **const mgp, const U8 *const start,
6647 const U8 *const send, STRLEN uoffset,
6648 STRLEN uoffset0, STRLEN boffset0)
6650 STRLEN boffset = 0; /* Actually always set, but let's keep gcc happy. */
6652 bool at_end = FALSE;
6654 PERL_ARGS_ASSERT_SV_POS_U2B_CACHED;
6656 assert (uoffset >= uoffset0);
6663 && (*mgp || (SvTYPE(sv) >= SVt_PVMG &&
6664 (*mgp = mg_find(sv, PERL_MAGIC_utf8))))) {
6665 if ((*mgp)->mg_ptr) {
6666 STRLEN *cache = (STRLEN *) (*mgp)->mg_ptr;
6667 if (cache[0] == uoffset) {
6668 /* An exact match. */
6671 if (cache[2] == uoffset) {
6672 /* An exact match. */
6676 if (cache[0] < uoffset) {
6677 /* The cache already knows part of the way. */
6678 if (cache[0] > uoffset0) {
6679 /* The cache knows more than the passed in pair */
6680 uoffset0 = cache[0];
6681 boffset0 = cache[1];
6683 if ((*mgp)->mg_len != -1) {
6684 /* And we know the end too. */
6686 + sv_pos_u2b_midway(start + boffset0, send,
6688 (*mgp)->mg_len - uoffset0);
6690 uoffset -= uoffset0;
6692 + sv_pos_u2b_forwards(start + boffset0,
6693 send, &uoffset, &at_end);
6694 uoffset += uoffset0;
6697 else if (cache[2] < uoffset) {
6698 /* We're between the two cache entries. */
6699 if (cache[2] > uoffset0) {
6700 /* and the cache knows more than the passed in pair */
6701 uoffset0 = cache[2];
6702 boffset0 = cache[3];
6706 + sv_pos_u2b_midway(start + boffset0,
6709 cache[0] - uoffset0);
6712 + sv_pos_u2b_midway(start + boffset0,
6715 cache[2] - uoffset0);
6719 else if ((*mgp)->mg_len != -1) {
6720 /* If we can take advantage of a passed in offset, do so. */
6721 /* In fact, offset0 is either 0, or less than offset, so don't
6722 need to worry about the other possibility. */
6724 + sv_pos_u2b_midway(start + boffset0, send,
6726 (*mgp)->mg_len - uoffset0);
6731 if (!found || PL_utf8cache < 0) {
6732 STRLEN real_boffset;
6733 uoffset -= uoffset0;
6734 real_boffset = boffset0 + sv_pos_u2b_forwards(start + boffset0,
6735 send, &uoffset, &at_end);
6736 uoffset += uoffset0;
6738 if (found && PL_utf8cache < 0)
6739 assert_uft8_cache_coherent("sv_pos_u2b_cache", boffset,
6741 boffset = real_boffset;
6746 utf8_mg_len_cache_update(sv, mgp, uoffset);
6748 utf8_mg_pos_cache_update(sv, mgp, boffset, uoffset, send - start);
6755 =for apidoc sv_pos_u2b_flags
6757 Converts the value pointed to by offsetp from a count of UTF-8 chars from
6758 the start of the string, to a count of the equivalent number of bytes; if
6759 lenp is non-zero, it does the same to lenp, but this time starting from
6760 the offset, rather than from the start
6761 of the string. Handles type coercion.
6762 I<flags> is passed to C<SvPV_flags>, and usually should be
6763 C<SV_GMAGIC|SV_CONST_RETURN> to handle magic.
6769 * sv_pos_u2b_flags() uses, like sv_pos_b2u(), the mg_ptr of the potential
6770 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6771 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
6776 Perl_sv_pos_u2b_flags(pTHX_ SV *const sv, STRLEN uoffset, STRLEN *const lenp,
6783 PERL_ARGS_ASSERT_SV_POS_U2B_FLAGS;
6785 start = (U8*)SvPV_flags(sv, len, flags);
6787 const U8 * const send = start + len;
6789 boffset = sv_pos_u2b_cached(sv, &mg, start, send, uoffset, 0, 0);
6792 && *lenp /* don't bother doing work for 0, as its bytes equivalent
6793 is 0, and *lenp is already set to that. */) {
6794 /* Convert the relative offset to absolute. */
6795 const STRLEN uoffset2 = uoffset + *lenp;
6796 const STRLEN boffset2
6797 = sv_pos_u2b_cached(sv, &mg, start, send, uoffset2,
6798 uoffset, boffset) - boffset;
6812 =for apidoc sv_pos_u2b
6814 Converts the value pointed to by offsetp from a count of UTF-8 chars from
6815 the start of the string, to a count of the equivalent number of bytes; if
6816 lenp is non-zero, it does the same to lenp, but this time starting from
6817 the offset, rather than from the start of the string. Handles magic and
6820 Use C<sv_pos_u2b_flags> in preference, which correctly handles strings longer
6827 * sv_pos_u2b() uses, like sv_pos_b2u(), the mg_ptr of the potential
6828 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6829 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
6833 /* This function is subject to size and sign problems */
6836 Perl_sv_pos_u2b(pTHX_ register SV *const sv, I32 *const offsetp, I32 *const lenp)
6838 PERL_ARGS_ASSERT_SV_POS_U2B;
6841 STRLEN ulen = (STRLEN)*lenp;
6842 *offsetp = (I32)sv_pos_u2b_flags(sv, (STRLEN)*offsetp, &ulen,
6843 SV_GMAGIC|SV_CONST_RETURN);
6846 *offsetp = (I32)sv_pos_u2b_flags(sv, (STRLEN)*offsetp, NULL,
6847 SV_GMAGIC|SV_CONST_RETURN);
6852 S_utf8_mg_len_cache_update(pTHX_ SV *const sv, MAGIC **const mgp,
6855 PERL_ARGS_ASSERT_UTF8_MG_LEN_CACHE_UPDATE;
6859 if (!*mgp && (SvTYPE(sv) < SVt_PVMG ||
6860 !(*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
6861 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, &PL_vtbl_utf8, 0, 0);
6865 (*mgp)->mg_len = ulen;
6866 /* For now, treat "overflowed" as "still unknown". See RT #72924. */
6867 if (ulen != (STRLEN) (*mgp)->mg_len)
6868 (*mgp)->mg_len = -1;
6871 /* Create and update the UTF8 magic offset cache, with the proffered utf8/
6872 byte length pairing. The (byte) length of the total SV is passed in too,
6873 as blen, because for some (more esoteric) SVs, the call to SvPV_const()
6874 may not have updated SvCUR, so we can't rely on reading it directly.
6876 The proffered utf8/byte length pairing isn't used if the cache already has
6877 two pairs, and swapping either for the proffered pair would increase the
6878 RMS of the intervals between known byte offsets.
6880 The cache itself consists of 4 STRLEN values
6881 0: larger UTF-8 offset
6882 1: corresponding byte offset
6883 2: smaller UTF-8 offset
6884 3: corresponding byte offset
6886 Unused cache pairs have the value 0, 0.
6887 Keeping the cache "backwards" means that the invariant of
6888 cache[0] >= cache[2] is maintained even with empty slots, which means that
6889 the code that uses it doesn't need to worry if only 1 entry has actually
6890 been set to non-zero. It also makes the "position beyond the end of the
6891 cache" logic much simpler, as the first slot is always the one to start
6895 S_utf8_mg_pos_cache_update(pTHX_ SV *const sv, MAGIC **const mgp, const STRLEN byte,
6896 const STRLEN utf8, const STRLEN blen)
6900 PERL_ARGS_ASSERT_UTF8_MG_POS_CACHE_UPDATE;
6905 if (!*mgp && (SvTYPE(sv) < SVt_PVMG ||
6906 !(*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
6907 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, (MGVTBL*)&PL_vtbl_utf8, 0,
6909 (*mgp)->mg_len = -1;
6913 if (!(cache = (STRLEN *)(*mgp)->mg_ptr)) {
6914 Newxz(cache, PERL_MAGIC_UTF8_CACHESIZE * 2, STRLEN);
6915 (*mgp)->mg_ptr = (char *) cache;
6919 if (PL_utf8cache < 0 && SvPOKp(sv)) {
6920 /* SvPOKp() because it's possible that sv has string overloading, and
6921 therefore is a reference, hence SvPVX() is actually a pointer.
6922 This cures the (very real) symptoms of RT 69422, but I'm not actually
6923 sure whether we should even be caching the results of UTF-8
6924 operations on overloading, given that nothing stops overloading
6925 returning a different value every time it's called. */
6926 const U8 *start = (const U8 *) SvPVX_const(sv);
6927 const STRLEN realutf8 = utf8_length(start, start + byte);
6929 assert_uft8_cache_coherent("utf8_mg_pos_cache_update", utf8, realutf8,
6933 /* Cache is held with the later position first, to simplify the code
6934 that deals with unbounded ends. */
6936 ASSERT_UTF8_CACHE(cache);
6937 if (cache[1] == 0) {
6938 /* Cache is totally empty */
6941 } else if (cache[3] == 0) {
6942 if (byte > cache[1]) {
6943 /* New one is larger, so goes first. */
6944 cache[2] = cache[0];
6945 cache[3] = cache[1];
6953 #define THREEWAY_SQUARE(a,b,c,d) \
6954 ((float)((d) - (c))) * ((float)((d) - (c))) \
6955 + ((float)((c) - (b))) * ((float)((c) - (b))) \
6956 + ((float)((b) - (a))) * ((float)((b) - (a)))
6958 /* Cache has 2 slots in use, and we know three potential pairs.
6959 Keep the two that give the lowest RMS distance. Do the
6960 calculation in bytes simply because we always know the byte
6961 length. squareroot has the same ordering as the positive value,
6962 so don't bother with the actual square root. */
6963 const float existing = THREEWAY_SQUARE(0, cache[3], cache[1], blen);
6964 if (byte > cache[1]) {
6965 /* New position is after the existing pair of pairs. */
6966 const float keep_earlier
6967 = THREEWAY_SQUARE(0, cache[3], byte, blen);
6968 const float keep_later
6969 = THREEWAY_SQUARE(0, cache[1], byte, blen);
6971 if (keep_later < keep_earlier) {
6972 if (keep_later < existing) {
6973 cache[2] = cache[0];
6974 cache[3] = cache[1];
6980 if (keep_earlier < existing) {
6986 else if (byte > cache[3]) {
6987 /* New position is between the existing pair of pairs. */
6988 const float keep_earlier
6989 = THREEWAY_SQUARE(0, cache[3], byte, blen);
6990 const float keep_later
6991 = THREEWAY_SQUARE(0, byte, cache[1], blen);
6993 if (keep_later < keep_earlier) {
6994 if (keep_later < existing) {
7000 if (keep_earlier < existing) {
7007 /* New position is before the existing pair of pairs. */
7008 const float keep_earlier
7009 = THREEWAY_SQUARE(0, byte, cache[3], blen);
7010 const float keep_later
7011 = THREEWAY_SQUARE(0, byte, cache[1], blen);
7013 if (keep_later < keep_earlier) {
7014 if (keep_later < existing) {
7020 if (keep_earlier < existing) {
7021 cache[0] = cache[2];
7022 cache[1] = cache[3];
7029 ASSERT_UTF8_CACHE(cache);
7032 /* We already know all of the way, now we may be able to walk back. The same
7033 assumption is made as in S_sv_pos_u2b_midway(), namely that walking
7034 backward is half the speed of walking forward. */
7036 S_sv_pos_b2u_midway(pTHX_ const U8 *const s, const U8 *const target,
7037 const U8 *end, STRLEN endu)
7039 const STRLEN forw = target - s;
7040 STRLEN backw = end - target;
7042 PERL_ARGS_ASSERT_SV_POS_B2U_MIDWAY;
7044 if (forw < 2 * backw) {
7045 return utf8_length(s, target);
7048 while (end > target) {
7050 while (UTF8_IS_CONTINUATION(*end)) {
7059 =for apidoc sv_pos_b2u
7061 Converts the value pointed to by offsetp from a count of bytes from the
7062 start of the string, to a count of the equivalent number of UTF-8 chars.
7063 Handles magic and type coercion.
7069 * sv_pos_b2u() uses, like sv_pos_u2b(), the mg_ptr of the potential
7070 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
7075 Perl_sv_pos_b2u(pTHX_ register SV *const sv, I32 *const offsetp)
7078 const STRLEN byte = *offsetp;
7079 STRLEN len = 0; /* Actually always set, but let's keep gcc happy. */
7085 PERL_ARGS_ASSERT_SV_POS_B2U;
7090 s = (const U8*)SvPV_const(sv, blen);
7093 Perl_croak(aTHX_ "panic: sv_pos_b2u: bad byte offset, blen=%"UVuf
7094 ", byte=%"UVuf, (UV)blen, (UV)byte);
7100 && SvTYPE(sv) >= SVt_PVMG
7101 && (mg = mg_find(sv, PERL_MAGIC_utf8)))
7104 STRLEN * const cache = (STRLEN *) mg->mg_ptr;
7105 if (cache[1] == byte) {
7106 /* An exact match. */
7107 *offsetp = cache[0];
7110 if (cache[3] == byte) {
7111 /* An exact match. */
7112 *offsetp = cache[2];
7116 if (cache[1] < byte) {
7117 /* We already know part of the way. */
7118 if (mg->mg_len != -1) {
7119 /* Actually, we know the end too. */
7121 + S_sv_pos_b2u_midway(aTHX_ s + cache[1], send,
7122 s + blen, mg->mg_len - cache[0]);
7124 len = cache[0] + utf8_length(s + cache[1], send);
7127 else if (cache[3] < byte) {
7128 /* We're between the two cached pairs, so we do the calculation
7129 offset by the byte/utf-8 positions for the earlier pair,
7130 then add the utf-8 characters from the string start to
7132 len = S_sv_pos_b2u_midway(aTHX_ s + cache[3], send,
7133 s + cache[1], cache[0] - cache[2])
7137 else { /* cache[3] > byte */
7138 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + cache[3],
7142 ASSERT_UTF8_CACHE(cache);
7144 } else if (mg->mg_len != -1) {
7145 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + blen, mg->mg_len);
7149 if (!found || PL_utf8cache < 0) {
7150 const STRLEN real_len = utf8_length(s, send);
7152 if (found && PL_utf8cache < 0)
7153 assert_uft8_cache_coherent("sv_pos_b2u", len, real_len, sv);
7160 utf8_mg_len_cache_update(sv, &mg, len);
7162 utf8_mg_pos_cache_update(sv, &mg, byte, len, blen);
7167 S_assert_uft8_cache_coherent(pTHX_ const char *const func, STRLEN from_cache,
7168 STRLEN real, SV *const sv)
7170 PERL_ARGS_ASSERT_ASSERT_UFT8_CACHE_COHERENT;
7172 /* As this is debugging only code, save space by keeping this test here,
7173 rather than inlining it in all the callers. */
7174 if (from_cache == real)
7177 /* Need to turn the assertions off otherwise we may recurse infinitely
7178 while printing error messages. */
7179 SAVEI8(PL_utf8cache);
7181 Perl_croak(aTHX_ "panic: %s cache %"UVuf" real %"UVuf" for %"SVf,
7182 func, (UV) from_cache, (UV) real, SVfARG(sv));
7188 Returns a boolean indicating whether the strings in the two SVs are
7189 identical. Is UTF-8 and 'use bytes' aware, handles get magic, and will
7190 coerce its args to strings if necessary.
7192 =for apidoc sv_eq_flags
7194 Returns a boolean indicating whether the strings in the two SVs are
7195 identical. Is UTF-8 and 'use bytes' aware and coerces its args to strings
7196 if necessary. If the flags include SV_GMAGIC, it handles get-magic, too.
7202 Perl_sv_eq_flags(pTHX_ register SV *sv1, register SV *sv2, const U32 flags)
7210 SV* svrecode = NULL;
7217 /* if pv1 and pv2 are the same, second SvPV_const call may
7218 * invalidate pv1 (if we are handling magic), so we may need to
7220 if (sv1 == sv2 && flags & SV_GMAGIC
7221 && (SvTHINKFIRST(sv1) || SvGMAGICAL(sv1))) {
7222 pv1 = SvPV_const(sv1, cur1);
7223 sv1 = newSVpvn_flags(pv1, cur1, SVs_TEMP | SvUTF8(sv2));
7225 pv1 = SvPV_flags_const(sv1, cur1, flags);
7233 pv2 = SvPV_flags_const(sv2, cur2, flags);
7235 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
7236 /* Differing utf8ness.
7237 * Do not UTF8size the comparands as a side-effect. */
7240 svrecode = newSVpvn(pv2, cur2);
7241 sv_recode_to_utf8(svrecode, PL_encoding);
7242 pv2 = SvPV_const(svrecode, cur2);
7245 svrecode = newSVpvn(pv1, cur1);
7246 sv_recode_to_utf8(svrecode, PL_encoding);
7247 pv1 = SvPV_const(svrecode, cur1);
7249 /* Now both are in UTF-8. */
7251 SvREFCNT_dec(svrecode);
7257 /* sv1 is the UTF-8 one */
7258 return bytes_cmp_utf8((const U8*)pv2, cur2,
7259 (const U8*)pv1, cur1) == 0;
7262 /* sv2 is the UTF-8 one */
7263 return bytes_cmp_utf8((const U8*)pv1, cur1,
7264 (const U8*)pv2, cur2) == 0;
7270 eq = (pv1 == pv2) || memEQ(pv1, pv2, cur1);
7272 SvREFCNT_dec(svrecode);
7280 Compares the strings in two SVs. Returns -1, 0, or 1 indicating whether the
7281 string in C<sv1> is less than, equal to, or greater than the string in
7282 C<sv2>. Is UTF-8 and 'use bytes' aware, handles get magic, and will
7283 coerce its args to strings if necessary. See also C<sv_cmp_locale>.
7285 =for apidoc sv_cmp_flags
7287 Compares the strings in two SVs. Returns -1, 0, or 1 indicating whether the
7288 string in C<sv1> is less than, equal to, or greater than the string in
7289 C<sv2>. Is UTF-8 and 'use bytes' aware and will coerce its args to strings
7290 if necessary. If the flags include SV_GMAGIC, it handles get magic. See
7291 also C<sv_cmp_locale_flags>.
7297 Perl_sv_cmp(pTHX_ register SV *const sv1, register SV *const sv2)
7299 return sv_cmp_flags(sv1, sv2, SV_GMAGIC);
7303 Perl_sv_cmp_flags(pTHX_ register SV *const sv1, register SV *const sv2,
7308 const char *pv1, *pv2;
7311 SV *svrecode = NULL;
7318 pv1 = SvPV_flags_const(sv1, cur1, flags);
7325 pv2 = SvPV_flags_const(sv2, cur2, flags);
7327 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
7328 /* Differing utf8ness.
7329 * Do not UTF8size the comparands as a side-effect. */
7332 svrecode = newSVpvn(pv2, cur2);
7333 sv_recode_to_utf8(svrecode, PL_encoding);
7334 pv2 = SvPV_const(svrecode, cur2);
7337 const int retval = -bytes_cmp_utf8((const U8*)pv2, cur2,
7338 (const U8*)pv1, cur1);
7339 return retval ? retval < 0 ? -1 : +1 : 0;
7344 svrecode = newSVpvn(pv1, cur1);
7345 sv_recode_to_utf8(svrecode, PL_encoding);
7346 pv1 = SvPV_const(svrecode, cur1);
7349 const int retval = bytes_cmp_utf8((const U8*)pv1, cur1,
7350 (const U8*)pv2, cur2);
7351 return retval ? retval < 0 ? -1 : +1 : 0;
7357 cmp = cur2 ? -1 : 0;
7361 const I32 retval = memcmp((const void*)pv1, (const void*)pv2, cur1 < cur2 ? cur1 : cur2);
7364 cmp = retval < 0 ? -1 : 1;
7365 } else if (cur1 == cur2) {
7368 cmp = cur1 < cur2 ? -1 : 1;
7372 SvREFCNT_dec(svrecode);
7380 =for apidoc sv_cmp_locale
7382 Compares the strings in two SVs in a locale-aware manner. Is UTF-8 and
7383 'use bytes' aware, handles get magic, and will coerce its args to strings
7384 if necessary. See also C<sv_cmp>.
7386 =for apidoc sv_cmp_locale_flags
7388 Compares the strings in two SVs in a locale-aware manner. Is UTF-8 and
7389 'use bytes' aware and will coerce its args to strings if necessary. If the
7390 flags contain SV_GMAGIC, it handles get magic. See also C<sv_cmp_flags>.
7396 Perl_sv_cmp_locale(pTHX_ register SV *const sv1, register SV *const sv2)
7398 return sv_cmp_locale_flags(sv1, sv2, SV_GMAGIC);
7402 Perl_sv_cmp_locale_flags(pTHX_ register SV *const sv1, register SV *const sv2,
7406 #ifdef USE_LOCALE_COLLATE
7412 if (PL_collation_standard)
7416 pv1 = sv1 ? sv_collxfrm_flags(sv1, &len1, flags) : (char *) NULL;
7418 pv2 = sv2 ? sv_collxfrm_flags(sv2, &len2, flags) : (char *) NULL;
7420 if (!pv1 || !len1) {
7431 retval = memcmp((void*)pv1, (void*)pv2, len1 < len2 ? len1 : len2);
7434 return retval < 0 ? -1 : 1;
7437 * When the result of collation is equality, that doesn't mean
7438 * that there are no differences -- some locales exclude some
7439 * characters from consideration. So to avoid false equalities,
7440 * we use the raw string as a tiebreaker.
7446 #endif /* USE_LOCALE_COLLATE */
7448 return sv_cmp(sv1, sv2);
7452 #ifdef USE_LOCALE_COLLATE
7455 =for apidoc sv_collxfrm
7457 This calls C<sv_collxfrm_flags> with the SV_GMAGIC flag. See
7458 C<sv_collxfrm_flags>.
7460 =for apidoc sv_collxfrm_flags
7462 Add Collate Transform magic to an SV if it doesn't already have it. If the
7463 flags contain SV_GMAGIC, it handles get-magic.
7465 Any scalar variable may carry PERL_MAGIC_collxfrm magic that contains the
7466 scalar data of the variable, but transformed to such a format that a normal
7467 memory comparison can be used to compare the data according to the locale
7474 Perl_sv_collxfrm_flags(pTHX_ SV *const sv, STRLEN *const nxp, const I32 flags)
7479 PERL_ARGS_ASSERT_SV_COLLXFRM_FLAGS;
7481 mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_collxfrm) : (MAGIC *) NULL;
7482 if (!mg || !mg->mg_ptr || *(U32*)mg->mg_ptr != PL_collation_ix) {
7488 Safefree(mg->mg_ptr);
7489 s = SvPV_flags_const(sv, len, flags);
7490 if ((xf = mem_collxfrm(s, len, &xlen))) {
7492 #ifdef PERL_OLD_COPY_ON_WRITE
7494 sv_force_normal_flags(sv, 0);
7496 mg = sv_magicext(sv, 0, PERL_MAGIC_collxfrm, &PL_vtbl_collxfrm,
7510 if (mg && mg->mg_ptr) {
7512 return mg->mg_ptr + sizeof(PL_collation_ix);
7520 #endif /* USE_LOCALE_COLLATE */
7523 S_sv_gets_append_to_utf8(pTHX_ SV *const sv, PerlIO *const fp, I32 append)
7525 SV * const tsv = newSV(0);
7528 sv_gets(tsv, fp, 0);
7529 sv_utf8_upgrade_nomg(tsv);
7530 SvCUR_set(sv,append);
7533 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7537 S_sv_gets_read_record(pTHX_ SV *const sv, PerlIO *const fp, I32 append)
7540 const U32 recsize = SvUV(SvRV(PL_rs)); /* RsRECORD() guarantees > 0. */
7541 /* Grab the size of the record we're getting */
7542 char *const buffer = SvGROW(sv, (STRLEN)(recsize + append + 1)) + append;
7549 /* VMS wants read instead of fread, because fread doesn't respect */
7550 /* RMS record boundaries. This is not necessarily a good thing to be */
7551 /* doing, but we've got no other real choice - except avoid stdio
7552 as implementation - perhaps write a :vms layer ?
7554 fd = PerlIO_fileno(fp);
7556 bytesread = PerlLIO_read(fd, buffer, recsize);
7558 else /* in-memory file from PerlIO::Scalar */
7561 bytesread = PerlIO_read(fp, buffer, recsize);
7566 SvCUR_set(sv, bytesread + append);
7567 buffer[bytesread] = '\0';
7568 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7574 Get a line from the filehandle and store it into the SV, optionally
7575 appending to the currently-stored string.
7581 Perl_sv_gets(pTHX_ register SV *const sv, register PerlIO *const fp, I32 append)
7586 register STDCHAR rslast;
7587 register STDCHAR *bp;
7592 PERL_ARGS_ASSERT_SV_GETS;
7594 if (SvTHINKFIRST(sv))
7595 sv_force_normal_flags(sv, append ? 0 : SV_COW_DROP_PV);
7596 /* XXX. If you make this PVIV, then copy on write can copy scalars read
7598 However, perlbench says it's slower, because the existing swipe code
7599 is faster than copy on write.
7600 Swings and roundabouts. */
7601 SvUPGRADE(sv, SVt_PV);
7606 if (PerlIO_isutf8(fp)) {
7608 sv_utf8_upgrade_nomg(sv);
7609 sv_pos_u2b(sv,&append,0);
7611 } else if (SvUTF8(sv)) {
7612 return S_sv_gets_append_to_utf8(aTHX_ sv, fp, append);
7620 if (PerlIO_isutf8(fp))
7623 if (IN_PERL_COMPILETIME) {
7624 /* we always read code in line mode */
7628 else if (RsSNARF(PL_rs)) {
7629 /* If it is a regular disk file use size from stat() as estimate
7630 of amount we are going to read -- may result in mallocing
7631 more memory than we really need if the layers below reduce
7632 the size we read (e.g. CRLF or a gzip layer).
7635 if (!PerlLIO_fstat(PerlIO_fileno(fp), &st) && S_ISREG(st.st_mode)) {
7636 const Off_t offset = PerlIO_tell(fp);
7637 if (offset != (Off_t) -1 && st.st_size + append > offset) {
7638 (void) SvGROW(sv, (STRLEN)((st.st_size - offset) + append + 1));
7644 else if (RsRECORD(PL_rs)) {
7645 return S_sv_gets_read_record(aTHX_ sv, fp, append);
7647 else if (RsPARA(PL_rs)) {
7653 /* Get $/ i.e. PL_rs into same encoding as stream wants */
7654 if (PerlIO_isutf8(fp)) {
7655 rsptr = SvPVutf8(PL_rs, rslen);
7658 if (SvUTF8(PL_rs)) {
7659 if (!sv_utf8_downgrade(PL_rs, TRUE)) {
7660 Perl_croak(aTHX_ "Wide character in $/");
7663 rsptr = SvPV_const(PL_rs, rslen);
7667 rslast = rslen ? rsptr[rslen - 1] : '\0';
7669 if (rspara) { /* have to do this both before and after */
7670 do { /* to make sure file boundaries work right */
7673 i = PerlIO_getc(fp);
7677 PerlIO_ungetc(fp,i);
7683 /* See if we know enough about I/O mechanism to cheat it ! */
7685 /* This used to be #ifdef test - it is made run-time test for ease
7686 of abstracting out stdio interface. One call should be cheap
7687 enough here - and may even be a macro allowing compile
7691 if (PerlIO_fast_gets(fp)) {
7694 * We're going to steal some values from the stdio struct
7695 * and put EVERYTHING in the innermost loop into registers.
7697 register STDCHAR *ptr;
7701 #if defined(VMS) && defined(PERLIO_IS_STDIO)
7702 /* An ungetc()d char is handled separately from the regular
7703 * buffer, so we getc() it back out and stuff it in the buffer.
7705 i = PerlIO_getc(fp);
7706 if (i == EOF) return 0;
7707 *(--((*fp)->_ptr)) = (unsigned char) i;
7711 /* Here is some breathtakingly efficient cheating */
7713 cnt = PerlIO_get_cnt(fp); /* get count into register */
7714 /* make sure we have the room */
7715 if ((I32)(SvLEN(sv) - append) <= cnt + 1) {
7716 /* Not room for all of it
7717 if we are looking for a separator and room for some
7719 if (rslen && cnt > 80 && (I32)SvLEN(sv) > append) {
7720 /* just process what we have room for */
7721 shortbuffered = cnt - SvLEN(sv) + append + 1;
7722 cnt -= shortbuffered;
7726 /* remember that cnt can be negative */
7727 SvGROW(sv, (STRLEN)(append + (cnt <= 0 ? 2 : (cnt + 1))));
7732 bp = (STDCHAR*)SvPVX_const(sv) + append; /* move these two too to registers */
7733 ptr = (STDCHAR*)PerlIO_get_ptr(fp);
7734 DEBUG_P(PerlIO_printf(Perl_debug_log,
7735 "Screamer: entering, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7736 DEBUG_P(PerlIO_printf(Perl_debug_log,
7737 "Screamer: entering: PerlIO * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7738 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7739 PTR2UV(PerlIO_has_base(fp) ? PerlIO_get_base(fp) : 0)));
7744 while (cnt > 0) { /* this | eat */
7746 if ((*bp++ = *ptr++) == rslast) /* really | dust */
7747 goto thats_all_folks; /* screams | sed :-) */
7751 Copy(ptr, bp, cnt, char); /* this | eat */
7752 bp += cnt; /* screams | dust */
7753 ptr += cnt; /* louder | sed :-) */
7755 assert (!shortbuffered);
7756 goto cannot_be_shortbuffered;
7760 if (shortbuffered) { /* oh well, must extend */
7761 cnt = shortbuffered;
7763 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
7765 SvGROW(sv, SvLEN(sv) + append + cnt + 2);
7766 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
7770 cannot_be_shortbuffered:
7771 DEBUG_P(PerlIO_printf(Perl_debug_log,
7772 "Screamer: going to getc, ptr=%"UVuf", cnt=%ld\n",
7773 PTR2UV(ptr),(long)cnt));
7774 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* deregisterize cnt and ptr */
7776 DEBUG_Pv(PerlIO_printf(Perl_debug_log,
7777 "Screamer: pre: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7778 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7779 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7781 /* This used to call 'filbuf' in stdio form, but as that behaves like
7782 getc when cnt <= 0 we use PerlIO_getc here to avoid introducing
7783 another abstraction. */
7784 i = PerlIO_getc(fp); /* get more characters */
7786 DEBUG_Pv(PerlIO_printf(Perl_debug_log,
7787 "Screamer: post: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7788 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7789 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7791 cnt = PerlIO_get_cnt(fp);
7792 ptr = (STDCHAR*)PerlIO_get_ptr(fp); /* reregisterize cnt and ptr */
7793 DEBUG_P(PerlIO_printf(Perl_debug_log,
7794 "Screamer: after getc, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7796 if (i == EOF) /* all done for ever? */
7797 goto thats_really_all_folks;
7799 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
7801 SvGROW(sv, bpx + cnt + 2);
7802 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
7804 *bp++ = (STDCHAR)i; /* store character from PerlIO_getc */
7806 if (rslen && (STDCHAR)i == rslast) /* all done for now? */
7807 goto thats_all_folks;
7811 if ((rslen > 1 && (STRLEN)(bp - (STDCHAR*)SvPVX_const(sv)) < rslen) ||
7812 memNE((char*)bp - rslen, rsptr, rslen))
7813 goto screamer; /* go back to the fray */
7814 thats_really_all_folks:
7816 cnt += shortbuffered;
7817 DEBUG_P(PerlIO_printf(Perl_debug_log,
7818 "Screamer: quitting, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7819 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* put these back or we're in trouble */
7820 DEBUG_P(PerlIO_printf(Perl_debug_log,
7821 "Screamer: end: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7822 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7823 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7825 SvCUR_set(sv, bp - (STDCHAR*)SvPVX_const(sv)); /* set length */
7826 DEBUG_P(PerlIO_printf(Perl_debug_log,
7827 "Screamer: done, len=%ld, string=|%.*s|\n",
7828 (long)SvCUR(sv),(int)SvCUR(sv),SvPVX_const(sv)));
7832 /*The big, slow, and stupid way. */
7833 #ifdef USE_HEAP_INSTEAD_OF_STACK /* Even slower way. */
7834 STDCHAR *buf = NULL;
7835 Newx(buf, 8192, STDCHAR);
7843 register const STDCHAR * const bpe = buf + sizeof(buf);
7845 while ((i = PerlIO_getc(fp)) != EOF && (*bp++ = (STDCHAR)i) != rslast && bp < bpe)
7846 ; /* keep reading */
7850 cnt = PerlIO_read(fp,(char*)buf, sizeof(buf));
7851 /* Accommodate broken VAXC compiler, which applies U8 cast to
7852 * both args of ?: operator, causing EOF to change into 255
7855 i = (U8)buf[cnt - 1];
7861 cnt = 0; /* we do need to re-set the sv even when cnt <= 0 */
7863 sv_catpvn(sv, (char *) buf, cnt);
7865 sv_setpvn(sv, (char *) buf, cnt);
7867 if (i != EOF && /* joy */
7869 SvCUR(sv) < rslen ||
7870 memNE(SvPVX_const(sv) + SvCUR(sv) - rslen, rsptr, rslen)))
7874 * If we're reading from a TTY and we get a short read,
7875 * indicating that the user hit his EOF character, we need
7876 * to notice it now, because if we try to read from the TTY
7877 * again, the EOF condition will disappear.
7879 * The comparison of cnt to sizeof(buf) is an optimization
7880 * that prevents unnecessary calls to feof().
7884 if (!(cnt < (I32)sizeof(buf) && PerlIO_eof(fp)))
7888 #ifdef USE_HEAP_INSTEAD_OF_STACK
7893 if (rspara) { /* have to do this both before and after */
7894 while (i != EOF) { /* to make sure file boundaries work right */
7895 i = PerlIO_getc(fp);
7897 PerlIO_ungetc(fp,i);
7903 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7909 Auto-increment of the value in the SV, doing string to numeric conversion
7910 if necessary. Handles 'get' magic and operator overloading.
7916 Perl_sv_inc(pTHX_ register SV *const sv)
7925 =for apidoc sv_inc_nomg
7927 Auto-increment of the value in the SV, doing string to numeric conversion
7928 if necessary. Handles operator overloading. Skips handling 'get' magic.
7934 Perl_sv_inc_nomg(pTHX_ register SV *const sv)
7942 if (SvTHINKFIRST(sv)) {
7943 if (SvIsCOW(sv) || isGV_with_GP(sv))
7944 sv_force_normal_flags(sv, 0);
7945 if (SvREADONLY(sv)) {
7946 if (IN_PERL_RUNTIME)
7947 Perl_croak_no_modify(aTHX);
7951 if (SvAMAGIC(sv) && AMG_CALLunary(sv, inc_amg))
7953 i = PTR2IV(SvRV(sv));
7958 flags = SvFLAGS(sv);
7959 if ((flags & (SVp_NOK|SVp_IOK)) == SVp_NOK) {
7960 /* It's (privately or publicly) a float, but not tested as an
7961 integer, so test it to see. */
7963 flags = SvFLAGS(sv);
7965 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
7966 /* It's publicly an integer, or privately an integer-not-float */
7967 #ifdef PERL_PRESERVE_IVUV
7971 if (SvUVX(sv) == UV_MAX)
7972 sv_setnv(sv, UV_MAX_P1);
7974 (void)SvIOK_only_UV(sv);
7975 SvUV_set(sv, SvUVX(sv) + 1);
7977 if (SvIVX(sv) == IV_MAX)
7978 sv_setuv(sv, (UV)IV_MAX + 1);
7980 (void)SvIOK_only(sv);
7981 SvIV_set(sv, SvIVX(sv) + 1);
7986 if (flags & SVp_NOK) {
7987 const NV was = SvNVX(sv);
7988 if (NV_OVERFLOWS_INTEGERS_AT &&
7989 was >= NV_OVERFLOWS_INTEGERS_AT) {
7990 /* diag_listed_as: Lost precision when %s %f by 1 */
7991 Perl_ck_warner(aTHX_ packWARN(WARN_IMPRECISION),
7992 "Lost precision when incrementing %" NVff " by 1",
7995 (void)SvNOK_only(sv);
7996 SvNV_set(sv, was + 1.0);
8000 if (!(flags & SVp_POK) || !*SvPVX_const(sv)) {
8001 if ((flags & SVTYPEMASK) < SVt_PVIV)
8002 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV ? SVt_PVIV : SVt_IV));
8003 (void)SvIOK_only(sv);
8008 while (isALPHA(*d)) d++;
8009 while (isDIGIT(*d)) d++;
8010 if (d < SvEND(sv)) {
8011 #ifdef PERL_PRESERVE_IVUV
8012 /* Got to punt this as an integer if needs be, but we don't issue
8013 warnings. Probably ought to make the sv_iv_please() that does
8014 the conversion if possible, and silently. */
8015 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
8016 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
8017 /* Need to try really hard to see if it's an integer.
8018 9.22337203685478e+18 is an integer.
8019 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
8020 so $a="9.22337203685478e+18"; $a+0; $a++
8021 needs to be the same as $a="9.22337203685478e+18"; $a++
8028 /* sv_2iv *should* have made this an NV */
8029 if (flags & SVp_NOK) {
8030 (void)SvNOK_only(sv);
8031 SvNV_set(sv, SvNVX(sv) + 1.0);
8034 /* I don't think we can get here. Maybe I should assert this
8035 And if we do get here I suspect that sv_setnv will croak. NWC
8037 #if defined(USE_LONG_DOUBLE)
8038 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",
8039 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
8041 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_inc punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
8042 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
8045 #endif /* PERL_PRESERVE_IVUV */
8046 sv_setnv(sv,Atof(SvPVX_const(sv)) + 1.0);
8050 while (d >= SvPVX_const(sv)) {
8058 /* MKS: The original code here died if letters weren't consecutive.
8059 * at least it didn't have to worry about non-C locales. The
8060 * new code assumes that ('z'-'a')==('Z'-'A'), letters are
8061 * arranged in order (although not consecutively) and that only
8062 * [A-Za-z] are accepted by isALPHA in the C locale.
8064 if (*d != 'z' && *d != 'Z') {
8065 do { ++*d; } while (!isALPHA(*d));
8068 *(d--) -= 'z' - 'a';
8073 *(d--) -= 'z' - 'a' + 1;
8077 /* oh,oh, the number grew */
8078 SvGROW(sv, SvCUR(sv) + 2);
8079 SvCUR_set(sv, SvCUR(sv) + 1);
8080 for (d = SvPVX(sv) + SvCUR(sv); d > SvPVX_const(sv); d--)
8091 Auto-decrement of the value in the SV, doing string to numeric conversion
8092 if necessary. Handles 'get' magic and operator overloading.
8098 Perl_sv_dec(pTHX_ register SV *const sv)
8108 =for apidoc sv_dec_nomg
8110 Auto-decrement of the value in the SV, doing string to numeric conversion
8111 if necessary. Handles operator overloading. Skips handling 'get' magic.
8117 Perl_sv_dec_nomg(pTHX_ register SV *const sv)
8124 if (SvTHINKFIRST(sv)) {
8125 if (SvIsCOW(sv) || isGV_with_GP(sv))
8126 sv_force_normal_flags(sv, 0);
8127 if (SvREADONLY(sv)) {
8128 if (IN_PERL_RUNTIME)
8129 Perl_croak_no_modify(aTHX);
8133 if (SvAMAGIC(sv) && AMG_CALLunary(sv, dec_amg))
8135 i = PTR2IV(SvRV(sv));
8140 /* Unlike sv_inc we don't have to worry about string-never-numbers
8141 and keeping them magic. But we mustn't warn on punting */
8142 flags = SvFLAGS(sv);
8143 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
8144 /* It's publicly an integer, or privately an integer-not-float */
8145 #ifdef PERL_PRESERVE_IVUV
8149 if (SvUVX(sv) == 0) {
8150 (void)SvIOK_only(sv);
8154 (void)SvIOK_only_UV(sv);
8155 SvUV_set(sv, SvUVX(sv) - 1);
8158 if (SvIVX(sv) == IV_MIN) {
8159 sv_setnv(sv, (NV)IV_MIN);
8163 (void)SvIOK_only(sv);
8164 SvIV_set(sv, SvIVX(sv) - 1);
8169 if (flags & SVp_NOK) {
8172 const NV was = SvNVX(sv);
8173 if (NV_OVERFLOWS_INTEGERS_AT &&
8174 was <= -NV_OVERFLOWS_INTEGERS_AT) {
8175 /* diag_listed_as: Lost precision when %s %f by 1 */
8176 Perl_ck_warner(aTHX_ packWARN(WARN_IMPRECISION),
8177 "Lost precision when decrementing %" NVff " by 1",
8180 (void)SvNOK_only(sv);
8181 SvNV_set(sv, was - 1.0);
8185 if (!(flags & SVp_POK)) {
8186 if ((flags & SVTYPEMASK) < SVt_PVIV)
8187 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV) ? SVt_PVIV : SVt_IV);
8189 (void)SvIOK_only(sv);
8192 #ifdef PERL_PRESERVE_IVUV
8194 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
8195 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
8196 /* Need to try really hard to see if it's an integer.
8197 9.22337203685478e+18 is an integer.
8198 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
8199 so $a="9.22337203685478e+18"; $a+0; $a--
8200 needs to be the same as $a="9.22337203685478e+18"; $a--
8207 /* sv_2iv *should* have made this an NV */
8208 if (flags & SVp_NOK) {
8209 (void)SvNOK_only(sv);
8210 SvNV_set(sv, SvNVX(sv) - 1.0);
8213 /* I don't think we can get here. Maybe I should assert this
8214 And if we do get here I suspect that sv_setnv will croak. NWC
8216 #if defined(USE_LONG_DOUBLE)
8217 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",
8218 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
8220 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_dec punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
8221 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
8225 #endif /* PERL_PRESERVE_IVUV */
8226 sv_setnv(sv,Atof(SvPVX_const(sv)) - 1.0); /* punt */
8229 /* this define is used to eliminate a chunk of duplicated but shared logic
8230 * it has the suffix __SV_C to signal that it isnt API, and isnt meant to be
8231 * used anywhere but here - yves
8233 #define PUSH_EXTEND_MORTAL__SV_C(AnSv) \
8236 PL_tmps_stack[++PL_tmps_ix] = (AnSv); \
8240 =for apidoc sv_mortalcopy
8242 Creates a new SV which is a copy of the original SV (using C<sv_setsv>).
8243 The new SV is marked as mortal. It will be destroyed "soon", either by an
8244 explicit call to FREETMPS, or by an implicit call at places such as
8245 statement boundaries. See also C<sv_newmortal> and C<sv_2mortal>.
8250 /* Make a string that will exist for the duration of the expression
8251 * evaluation. Actually, it may have to last longer than that, but
8252 * hopefully we won't free it until it has been assigned to a
8253 * permanent location. */
8256 Perl_sv_mortalcopy(pTHX_ SV *const oldstr)
8262 sv_setsv(sv,oldstr);
8263 PUSH_EXTEND_MORTAL__SV_C(sv);
8269 =for apidoc sv_newmortal
8271 Creates a new null SV which is mortal. The reference count of the SV is
8272 set to 1. It will be destroyed "soon", either by an explicit call to
8273 FREETMPS, or by an implicit call at places such as statement boundaries.
8274 See also C<sv_mortalcopy> and C<sv_2mortal>.
8280 Perl_sv_newmortal(pTHX)
8286 SvFLAGS(sv) = SVs_TEMP;
8287 PUSH_EXTEND_MORTAL__SV_C(sv);
8293 =for apidoc newSVpvn_flags
8295 Creates a new SV and copies a string into it. The reference count for the
8296 SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
8297 string. You are responsible for ensuring that the source string is at least
8298 C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
8299 Currently the only flag bits accepted are C<SVf_UTF8> and C<SVs_TEMP>.
8300 If C<SVs_TEMP> is set, then C<sv_2mortal()> is called on the result before
8301 returning. If C<SVf_UTF8> is set, C<s>
8302 is considered to be in UTF-8 and the
8303 C<SVf_UTF8> flag will be set on the new SV.
8304 C<newSVpvn_utf8()> is a convenience wrapper for this function, defined as
8306 #define newSVpvn_utf8(s, len, u) \
8307 newSVpvn_flags((s), (len), (u) ? SVf_UTF8 : 0)
8313 Perl_newSVpvn_flags(pTHX_ const char *const s, const STRLEN len, const U32 flags)
8318 /* All the flags we don't support must be zero.
8319 And we're new code so I'm going to assert this from the start. */
8320 assert(!(flags & ~(SVf_UTF8|SVs_TEMP)));
8322 sv_setpvn(sv,s,len);
8324 /* This code used to a sv_2mortal(), however we now unroll the call to sv_2mortal()
8325 * and do what it does ourselves here.
8326 * Since we have asserted that flags can only have the SVf_UTF8 and/or SVs_TEMP flags
8327 * set above we can use it to enable the sv flags directly (bypassing SvTEMP_on), which
8328 * in turn means we dont need to mask out the SVf_UTF8 flag below, which means that we
8329 * eliminate quite a few steps than it looks - Yves (explaining patch by gfx)
8332 SvFLAGS(sv) |= flags;
8334 if(flags & SVs_TEMP){
8335 PUSH_EXTEND_MORTAL__SV_C(sv);
8342 =for apidoc sv_2mortal
8344 Marks an existing SV as mortal. The SV will be destroyed "soon", either
8345 by an explicit call to FREETMPS, or by an implicit call at places such as
8346 statement boundaries. SvTEMP() is turned on which means that the SV's
8347 string buffer can be "stolen" if this SV is copied. See also C<sv_newmortal>
8348 and C<sv_mortalcopy>.
8354 Perl_sv_2mortal(pTHX_ register SV *const sv)
8359 if (SvREADONLY(sv) && SvIMMORTAL(sv))
8361 PUSH_EXTEND_MORTAL__SV_C(sv);
8369 Creates a new SV and copies a string into it. The reference count for the
8370 SV is set to 1. If C<len> is zero, Perl will compute the length using
8371 strlen(). For efficiency, consider using C<newSVpvn> instead.
8377 Perl_newSVpv(pTHX_ const char *const s, const STRLEN len)
8383 sv_setpvn(sv, s, len || s == NULL ? len : strlen(s));
8388 =for apidoc newSVpvn
8390 Creates a new SV and copies a string into it. The reference count for the
8391 SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
8392 string. You are responsible for ensuring that the source string is at least
8393 C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
8399 Perl_newSVpvn(pTHX_ const char *const s, const STRLEN len)
8405 sv_setpvn(sv,s,len);
8410 =for apidoc newSVhek
8412 Creates a new SV from the hash key structure. It will generate scalars that
8413 point to the shared string table where possible. Returns a new (undefined)
8414 SV if the hek is NULL.
8420 Perl_newSVhek(pTHX_ const HEK *const hek)
8430 if (HEK_LEN(hek) == HEf_SVKEY) {
8431 return newSVsv(*(SV**)HEK_KEY(hek));
8433 const int flags = HEK_FLAGS(hek);
8434 if (flags & HVhek_WASUTF8) {
8436 Andreas would like keys he put in as utf8 to come back as utf8
8438 STRLEN utf8_len = HEK_LEN(hek);
8439 SV * const sv = newSV_type(SVt_PV);
8440 char *as_utf8 = (char *)bytes_to_utf8 ((U8*)HEK_KEY(hek), &utf8_len);
8441 /* bytes_to_utf8() allocates a new string, which we can repurpose: */
8442 sv_usepvn_flags(sv, as_utf8, utf8_len, SV_HAS_TRAILING_NUL);
8445 } else if (flags & (HVhek_REHASH|HVhek_UNSHARED)) {
8446 /* We don't have a pointer to the hv, so we have to replicate the
8447 flag into every HEK. This hv is using custom a hasing
8448 algorithm. Hence we can't return a shared string scalar, as
8449 that would contain the (wrong) hash value, and might get passed
8450 into an hv routine with a regular hash.
8451 Similarly, a hash that isn't using shared hash keys has to have
8452 the flag in every key so that we know not to try to call
8453 share_hek_hek on it. */
8455 SV * const sv = newSVpvn (HEK_KEY(hek), HEK_LEN(hek));
8460 /* This will be overwhelminly the most common case. */
8462 /* Inline most of newSVpvn_share(), because share_hek_hek() is far
8463 more efficient than sharepvn(). */
8467 sv_upgrade(sv, SVt_PV);
8468 SvPV_set(sv, (char *)HEK_KEY(share_hek_hek(hek)));
8469 SvCUR_set(sv, HEK_LEN(hek));
8482 =for apidoc newSVpvn_share
8484 Creates a new SV with its SvPVX_const pointing to a shared string in the string
8485 table. If the string does not already exist in the table, it is
8486 created first. Turns on READONLY and FAKE. If the C<hash> parameter
8487 is non-zero, that value is used; otherwise the hash is computed.
8488 The string's hash can later be retrieved from the SV
8489 with the C<SvSHARED_HASH()> macro. The idea here is
8490 that as the string table is used for shared hash keys these strings will have
8491 SvPVX_const == HeKEY and hash lookup will avoid string compare.
8497 Perl_newSVpvn_share(pTHX_ const char *src, I32 len, U32 hash)
8501 bool is_utf8 = FALSE;
8502 const char *const orig_src = src;
8505 STRLEN tmplen = -len;
8507 /* See the note in hv.c:hv_fetch() --jhi */
8508 src = (char*)bytes_from_utf8((const U8*)src, &tmplen, &is_utf8);
8512 PERL_HASH(hash, src, len);
8514 /* The logic for this is inlined in S_mro_get_linear_isa_dfs(), so if it
8515 changes here, update it there too. */
8516 sv_upgrade(sv, SVt_PV);
8517 SvPV_set(sv, sharepvn(src, is_utf8?-len:len, hash));
8525 if (src != orig_src)
8531 =for apidoc newSVpv_share
8533 Like C<newSVpvn_share>, but takes a nul-terminated string instead of a
8540 Perl_newSVpv_share(pTHX_ const char *src, U32 hash)
8542 return newSVpvn_share(src, strlen(src), hash);
8545 #if defined(PERL_IMPLICIT_CONTEXT)
8547 /* pTHX_ magic can't cope with varargs, so this is a no-context
8548 * version of the main function, (which may itself be aliased to us).
8549 * Don't access this version directly.
8553 Perl_newSVpvf_nocontext(const char *const pat, ...)
8559 PERL_ARGS_ASSERT_NEWSVPVF_NOCONTEXT;
8561 va_start(args, pat);
8562 sv = vnewSVpvf(pat, &args);
8569 =for apidoc newSVpvf
8571 Creates a new SV and initializes it with the string formatted like
8578 Perl_newSVpvf(pTHX_ const char *const pat, ...)
8583 PERL_ARGS_ASSERT_NEWSVPVF;
8585 va_start(args, pat);
8586 sv = vnewSVpvf(pat, &args);
8591 /* backend for newSVpvf() and newSVpvf_nocontext() */
8594 Perl_vnewSVpvf(pTHX_ const char *const pat, va_list *const args)
8599 PERL_ARGS_ASSERT_VNEWSVPVF;
8602 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
8609 Creates a new SV and copies a floating point value into it.
8610 The reference count for the SV is set to 1.
8616 Perl_newSVnv(pTHX_ const NV n)
8629 Creates a new SV and copies an integer into it. The reference count for the
8636 Perl_newSViv(pTHX_ const IV i)
8649 Creates a new SV and copies an unsigned integer into it.
8650 The reference count for the SV is set to 1.
8656 Perl_newSVuv(pTHX_ const UV u)
8667 =for apidoc newSV_type
8669 Creates a new SV, of the type specified. The reference count for the new SV
8676 Perl_newSV_type(pTHX_ const svtype type)
8681 sv_upgrade(sv, type);
8686 =for apidoc newRV_noinc
8688 Creates an RV wrapper for an SV. The reference count for the original
8689 SV is B<not> incremented.
8695 Perl_newRV_noinc(pTHX_ SV *const tmpRef)
8698 register SV *sv = newSV_type(SVt_IV);
8700 PERL_ARGS_ASSERT_NEWRV_NOINC;
8703 SvRV_set(sv, tmpRef);
8708 /* newRV_inc is the official function name to use now.
8709 * newRV_inc is in fact #defined to newRV in sv.h
8713 Perl_newRV(pTHX_ SV *const sv)
8717 PERL_ARGS_ASSERT_NEWRV;
8719 return newRV_noinc(SvREFCNT_inc_simple_NN(sv));
8725 Creates a new SV which is an exact duplicate of the original SV.
8732 Perl_newSVsv(pTHX_ register SV *const old)
8739 if (SvTYPE(old) == (svtype)SVTYPEMASK) {
8740 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL), "semi-panic: attempt to dup freed string");
8744 /* SV_GMAGIC is the default for sv_setv()
8745 SV_NOSTEAL prevents TEMP buffers being, well, stolen, and saves games
8746 with SvTEMP_off and SvTEMP_on round a call to sv_setsv. */
8747 sv_setsv_flags(sv, old, SV_GMAGIC | SV_NOSTEAL);
8752 =for apidoc sv_reset
8754 Underlying implementation for the C<reset> Perl function.
8755 Note that the perl-level function is vaguely deprecated.
8761 Perl_sv_reset(pTHX_ register const char *s, HV *const stash)
8764 char todo[PERL_UCHAR_MAX+1];
8766 PERL_ARGS_ASSERT_SV_RESET;
8771 if (!*s) { /* reset ?? searches */
8772 MAGIC * const mg = mg_find((const SV *)stash, PERL_MAGIC_symtab);
8774 const U32 count = mg->mg_len / sizeof(PMOP**);
8775 PMOP **pmp = (PMOP**) mg->mg_ptr;
8776 PMOP *const *const end = pmp + count;
8780 SvREADONLY_off(PL_regex_pad[(*pmp)->op_pmoffset]);
8782 (*pmp)->op_pmflags &= ~PMf_USED;
8790 /* reset variables */
8792 if (!HvARRAY(stash))
8795 Zero(todo, 256, char);
8798 I32 i = (unsigned char)*s;
8802 max = (unsigned char)*s++;
8803 for ( ; i <= max; i++) {
8806 for (i = 0; i <= (I32) HvMAX(stash); i++) {
8808 for (entry = HvARRAY(stash)[i];
8810 entry = HeNEXT(entry))
8815 if (!todo[(U8)*HeKEY(entry)])
8817 gv = MUTABLE_GV(HeVAL(entry));
8820 if (SvTHINKFIRST(sv)) {
8821 if (!SvREADONLY(sv) && SvROK(sv))
8823 /* XXX Is this continue a bug? Why should THINKFIRST
8824 exempt us from resetting arrays and hashes? */
8828 if (SvTYPE(sv) >= SVt_PV) {
8830 if (SvPVX_const(sv) != NULL)
8838 if (GvHV(gv) && !HvNAME_get(GvHV(gv))) {
8840 Perl_die(aTHX_ "Can't reset %%ENV on this system");
8843 # if defined(USE_ENVIRON_ARRAY)
8846 # endif /* USE_ENVIRON_ARRAY */
8857 Using various gambits, try to get an IO from an SV: the IO slot if its a
8858 GV; or the recursive result if we're an RV; or the IO slot of the symbol
8859 named after the PV if we're a string.
8861 'Get' magic is ignored on the sv passed in, but will be called on
8862 C<SvRV(sv)> if sv is an RV.
8868 Perl_sv_2io(pTHX_ SV *const sv)
8873 PERL_ARGS_ASSERT_SV_2IO;
8875 switch (SvTYPE(sv)) {
8877 io = MUTABLE_IO(sv);
8881 if (isGV_with_GP(sv)) {
8882 gv = MUTABLE_GV(sv);
8885 Perl_croak(aTHX_ "Bad filehandle: %"HEKf,
8886 HEKfARG(GvNAME_HEK(gv)));
8892 Perl_croak(aTHX_ PL_no_usym, "filehandle");
8894 SvGETMAGIC(SvRV(sv));
8895 return sv_2io(SvRV(sv));
8897 gv = gv_fetchsv_nomg(sv, 0, SVt_PVIO);
8904 if (SvGMAGICAL(sv)) {
8905 newsv = sv_newmortal();
8906 sv_setsv_nomg(newsv, sv);
8908 Perl_croak(aTHX_ "Bad filehandle: %"SVf, SVfARG(newsv));
8918 Using various gambits, try to get a CV from an SV; in addition, try if
8919 possible to set C<*st> and C<*gvp> to the stash and GV associated with it.
8920 The flags in C<lref> are passed to gv_fetchsv.
8926 Perl_sv_2cv(pTHX_ SV *sv, HV **const st, GV **const gvp, const I32 lref)
8932 PERL_ARGS_ASSERT_SV_2CV;
8939 switch (SvTYPE(sv)) {
8943 return MUTABLE_CV(sv);
8953 sv = amagic_deref_call(sv, to_cv_amg);
8956 if (SvTYPE(sv) == SVt_PVCV) {
8957 cv = MUTABLE_CV(sv);
8962 else if(SvGETMAGIC(sv), isGV_with_GP(sv))
8963 gv = MUTABLE_GV(sv);
8965 Perl_croak(aTHX_ "Not a subroutine reference");
8967 else if (isGV_with_GP(sv)) {
8968 gv = MUTABLE_GV(sv);
8971 gv = gv_fetchsv_nomg(sv, lref, SVt_PVCV);
8978 /* Some flags to gv_fetchsv mean don't really create the GV */
8979 if (!isGV_with_GP(gv)) {
8984 if (lref & ~GV_ADDMG && !GvCVu(gv)) {
8988 gv_efullname3(tmpsv, gv, NULL);
8989 /* XXX this is probably not what they think they're getting.
8990 * It has the same effect as "sub name;", i.e. just a forward
8992 newSUB(start_subparse(FALSE, 0),
8993 newSVOP(OP_CONST, 0, tmpsv),
8997 Perl_croak(aTHX_ "Unable to create sub named \"%"SVf"\"",
8998 SVfARG(SvOK(sv) ? sv : &PL_sv_no));
9007 Returns true if the SV has a true value by Perl's rules.
9008 Use the C<SvTRUE> macro instead, which may call C<sv_true()> or may
9009 instead use an in-line version.
9015 Perl_sv_true(pTHX_ register SV *const sv)
9020 register const XPV* const tXpv = (XPV*)SvANY(sv);
9022 (tXpv->xpv_cur > 1 ||
9023 (tXpv->xpv_cur && *sv->sv_u.svu_pv != '0')))
9030 return SvIVX(sv) != 0;
9033 return SvNVX(sv) != 0.0;
9035 return sv_2bool(sv);
9041 =for apidoc sv_pvn_force
9043 Get a sensible string out of the SV somehow.
9044 A private implementation of the C<SvPV_force> macro for compilers which
9045 can't cope with complex macro expressions. Always use the macro instead.
9047 =for apidoc sv_pvn_force_flags
9049 Get a sensible string out of the SV somehow.
9050 If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<sv> if
9051 appropriate, else not. C<sv_pvn_force> and C<sv_pvn_force_nomg> are
9052 implemented in terms of this function.
9053 You normally want to use the various wrapper macros instead: see
9054 C<SvPV_force> and C<SvPV_force_nomg>
9060 Perl_sv_pvn_force_flags(pTHX_ SV *const sv, STRLEN *const lp, const I32 flags)
9064 PERL_ARGS_ASSERT_SV_PVN_FORCE_FLAGS;
9066 if (flags & SV_GMAGIC) SvGETMAGIC(sv);
9067 if (SvTHINKFIRST(sv) && !SvROK(sv))
9068 sv_force_normal_flags(sv, 0);
9078 if (SvREADONLY(sv) && !(flags & SV_MUTABLE_RETURN)) {
9079 const char * const ref = sv_reftype(sv,0);
9081 Perl_croak(aTHX_ "Can't coerce readonly %s to string in %s",
9082 ref, OP_DESC(PL_op));
9084 Perl_croak(aTHX_ "Can't coerce readonly %s to string", ref);
9086 if ((SvTYPE(sv) > SVt_PVLV && SvTYPE(sv) != SVt_PVFM)
9087 || isGV_with_GP(sv))
9088 /* diag_listed_as: Can't coerce %s to %s in %s */
9089 Perl_croak(aTHX_ "Can't coerce %s to string in %s", sv_reftype(sv,0),
9091 s = sv_2pv_flags(sv, &len, flags &~ SV_GMAGIC);
9095 if (s != SvPVX_const(sv)) { /* Almost, but not quite, sv_setpvn() */
9098 SvUPGRADE(sv, SVt_PV); /* Never FALSE */
9099 SvGROW(sv, len + 1);
9100 Move(s,SvPVX(sv),len,char);
9102 SvPVX(sv)[len] = '\0';
9105 SvPOK_on(sv); /* validate pointer */
9107 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
9108 PTR2UV(sv),SvPVX_const(sv)));
9111 return SvPVX_mutable(sv);
9115 =for apidoc sv_pvbyten_force
9117 The backend for the C<SvPVbytex_force> macro. Always use the macro
9124 Perl_sv_pvbyten_force(pTHX_ SV *const sv, STRLEN *const lp)
9126 PERL_ARGS_ASSERT_SV_PVBYTEN_FORCE;
9128 sv_pvn_force(sv,lp);
9129 sv_utf8_downgrade(sv,0);
9135 =for apidoc sv_pvutf8n_force
9137 The backend for the C<SvPVutf8x_force> macro. Always use the macro
9144 Perl_sv_pvutf8n_force(pTHX_ SV *const sv, STRLEN *const lp)
9146 PERL_ARGS_ASSERT_SV_PVUTF8N_FORCE;
9148 sv_pvn_force(sv,lp);
9149 sv_utf8_upgrade(sv);
9155 =for apidoc sv_reftype
9157 Returns a string describing what the SV is a reference to.
9163 Perl_sv_reftype(pTHX_ const SV *const sv, const int ob)
9165 PERL_ARGS_ASSERT_SV_REFTYPE;
9166 if (ob && SvOBJECT(sv)) {
9167 return SvPV_nolen_const(sv_ref(NULL, sv, ob));
9170 switch (SvTYPE(sv)) {
9185 case SVt_PVLV: return (char *) (SvROK(sv) ? "REF"
9186 /* tied lvalues should appear to be
9187 * scalars for backwards compatibility */
9188 : (LvTYPE(sv) == 't' || LvTYPE(sv) == 'T')
9189 ? "SCALAR" : "LVALUE");
9190 case SVt_PVAV: return "ARRAY";
9191 case SVt_PVHV: return "HASH";
9192 case SVt_PVCV: return "CODE";
9193 case SVt_PVGV: return (char *) (isGV_with_GP(sv)
9194 ? "GLOB" : "SCALAR");
9195 case SVt_PVFM: return "FORMAT";
9196 case SVt_PVIO: return "IO";
9197 case SVt_BIND: return "BIND";
9198 case SVt_REGEXP: return "REGEXP";
9199 default: return "UNKNOWN";
9207 Returns a SV describing what the SV passed in is a reference to.
9213 Perl_sv_ref(pTHX_ register SV *dst, const SV *const sv, const int ob)
9215 PERL_ARGS_ASSERT_SV_REF;
9218 dst = sv_newmortal();
9220 if (ob && SvOBJECT(sv)) {
9221 HvNAME_get(SvSTASH(sv))
9222 ? sv_sethek(dst, HvNAME_HEK(SvSTASH(sv)))
9223 : sv_setpvn(dst, "__ANON__", 8);
9226 const char * reftype = sv_reftype(sv, 0);
9227 sv_setpv(dst, reftype);
9233 =for apidoc sv_isobject
9235 Returns a boolean indicating whether the SV is an RV pointing to a blessed
9236 object. If the SV is not an RV, or if the object is not blessed, then this
9243 Perl_sv_isobject(pTHX_ SV *sv)
9259 Returns a boolean indicating whether the SV is blessed into the specified
9260 class. This does not check for subtypes; use C<sv_derived_from> to verify
9261 an inheritance relationship.
9267 Perl_sv_isa(pTHX_ SV *sv, const char *const name)
9271 PERL_ARGS_ASSERT_SV_ISA;
9281 hvname = HvNAME_get(SvSTASH(sv));
9285 return strEQ(hvname, name);
9291 Creates a new SV for the RV, C<rv>, to point to. If C<rv> is not an RV then
9292 it will be upgraded to one. If C<classname> is non-null then the new SV will
9293 be blessed in the specified package. The new SV is returned and its
9294 reference count is 1.
9300 Perl_newSVrv(pTHX_ SV *const rv, const char *const classname)
9305 PERL_ARGS_ASSERT_NEWSVRV;
9309 SV_CHECK_THINKFIRST_COW_DROP(rv);
9310 (void)SvAMAGIC_off(rv);
9312 if (SvTYPE(rv) >= SVt_PVMG) {
9313 const U32 refcnt = SvREFCNT(rv);
9317 SvREFCNT(rv) = refcnt;
9319 sv_upgrade(rv, SVt_IV);
9320 } else if (SvROK(rv)) {
9321 SvREFCNT_dec(SvRV(rv));
9323 prepare_SV_for_RV(rv);
9331 HV* const stash = gv_stashpv(classname, GV_ADD);
9332 (void)sv_bless(rv, stash);
9338 =for apidoc sv_setref_pv
9340 Copies a pointer into a new SV, optionally blessing the SV. The C<rv>
9341 argument will be upgraded to an RV. That RV will be modified to point to
9342 the new SV. If the C<pv> argument is NULL then C<PL_sv_undef> will be placed
9343 into the SV. The C<classname> argument indicates the package for the
9344 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9345 will have a reference count of 1, and the RV will be returned.
9347 Do not use with other Perl types such as HV, AV, SV, CV, because those
9348 objects will become corrupted by the pointer copy process.
9350 Note that C<sv_setref_pvn> copies the string while this copies the pointer.
9356 Perl_sv_setref_pv(pTHX_ SV *const rv, const char *const classname, void *const pv)
9360 PERL_ARGS_ASSERT_SV_SETREF_PV;
9363 sv_setsv(rv, &PL_sv_undef);
9367 sv_setiv(newSVrv(rv,classname), PTR2IV(pv));
9372 =for apidoc sv_setref_iv
9374 Copies an integer into a new SV, optionally blessing the SV. The C<rv>
9375 argument will be upgraded to an RV. That RV will be modified to point to
9376 the new SV. The C<classname> argument indicates the package for the
9377 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9378 will have a reference count of 1, and the RV will be returned.
9384 Perl_sv_setref_iv(pTHX_ SV *const rv, const char *const classname, const IV iv)
9386 PERL_ARGS_ASSERT_SV_SETREF_IV;
9388 sv_setiv(newSVrv(rv,classname), iv);
9393 =for apidoc sv_setref_uv
9395 Copies an unsigned integer into a new SV, optionally blessing the SV. The C<rv>
9396 argument will be upgraded to an RV. That RV will be modified to point to
9397 the new SV. The C<classname> argument indicates the package for the
9398 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9399 will have a reference count of 1, and the RV will be returned.
9405 Perl_sv_setref_uv(pTHX_ SV *const rv, const char *const classname, const UV uv)
9407 PERL_ARGS_ASSERT_SV_SETREF_UV;
9409 sv_setuv(newSVrv(rv,classname), uv);
9414 =for apidoc sv_setref_nv
9416 Copies a double into a new SV, optionally blessing the SV. The C<rv>
9417 argument will be upgraded to an RV. That RV will be modified to point to
9418 the new SV. The C<classname> argument indicates the package for the
9419 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9420 will have a reference count of 1, and the RV will be returned.
9426 Perl_sv_setref_nv(pTHX_ SV *const rv, const char *const classname, const NV nv)
9428 PERL_ARGS_ASSERT_SV_SETREF_NV;
9430 sv_setnv(newSVrv(rv,classname), nv);
9435 =for apidoc sv_setref_pvn
9437 Copies a string into a new SV, optionally blessing the SV. The length of the
9438 string must be specified with C<n>. The C<rv> argument will be upgraded to
9439 an RV. That RV will be modified to point to the new SV. The C<classname>
9440 argument indicates the package for the blessing. Set C<classname> to
9441 C<NULL> to avoid the blessing. The new SV will have a reference count
9442 of 1, and the RV will be returned.
9444 Note that C<sv_setref_pv> copies the pointer while this copies the string.
9450 Perl_sv_setref_pvn(pTHX_ SV *const rv, const char *const classname,
9451 const char *const pv, const STRLEN n)
9453 PERL_ARGS_ASSERT_SV_SETREF_PVN;
9455 sv_setpvn(newSVrv(rv,classname), pv, n);
9460 =for apidoc sv_bless
9462 Blesses an SV into a specified package. The SV must be an RV. The package
9463 must be designated by its stash (see C<gv_stashpv()>). The reference count
9464 of the SV is unaffected.
9470 Perl_sv_bless(pTHX_ SV *const sv, HV *const stash)
9475 PERL_ARGS_ASSERT_SV_BLESS;
9478 Perl_croak(aTHX_ "Can't bless non-reference value");
9480 if (SvFLAGS(tmpRef) & (SVs_OBJECT|SVf_READONLY)) {
9481 if (SvIsCOW(tmpRef))
9482 sv_force_normal_flags(tmpRef, 0);
9483 if (SvREADONLY(tmpRef))
9484 Perl_croak_no_modify(aTHX);
9485 if (SvOBJECT(tmpRef)) {
9486 if (SvTYPE(tmpRef) != SVt_PVIO)
9488 SvREFCNT_dec(SvSTASH(tmpRef));
9491 SvOBJECT_on(tmpRef);
9492 if (SvTYPE(tmpRef) != SVt_PVIO)
9494 SvUPGRADE(tmpRef, SVt_PVMG);
9495 SvSTASH_set(tmpRef, MUTABLE_HV(SvREFCNT_inc_simple(stash)));
9500 (void)SvAMAGIC_off(sv);
9502 if(SvSMAGICAL(tmpRef))
9503 if(mg_find(tmpRef, PERL_MAGIC_ext) || mg_find(tmpRef, PERL_MAGIC_uvar))
9511 /* Downgrades a PVGV to a PVMG. If it's actually a PVLV, we leave the type
9512 * as it is after unglobbing it.
9515 PERL_STATIC_INLINE void
9516 S_sv_unglob(pTHX_ SV *const sv, U32 flags)
9521 SV * const temp = flags & SV_COW_DROP_PV ? NULL : sv_newmortal();
9523 PERL_ARGS_ASSERT_SV_UNGLOB;
9525 assert(SvTYPE(sv) == SVt_PVGV || SvTYPE(sv) == SVt_PVLV);
9527 if (!(flags & SV_COW_DROP_PV))
9528 gv_efullname3(temp, MUTABLE_GV(sv), "*");
9531 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
9532 && HvNAME_get(stash))
9533 mro_method_changed_in(stash);
9534 gp_free(MUTABLE_GV(sv));
9537 sv_del_backref(MUTABLE_SV(GvSTASH(sv)), sv);
9541 if (GvNAME_HEK(sv)) {
9542 unshare_hek(GvNAME_HEK(sv));
9544 isGV_with_GP_off(sv);
9546 if(SvTYPE(sv) == SVt_PVGV) {
9547 /* need to keep SvANY(sv) in the right arena */
9548 xpvmg = new_XPVMG();
9549 StructCopy(SvANY(sv), xpvmg, XPVMG);
9550 del_XPVGV(SvANY(sv));
9553 SvFLAGS(sv) &= ~SVTYPEMASK;
9554 SvFLAGS(sv) |= SVt_PVMG;
9557 /* Intentionally not calling any local SET magic, as this isn't so much a
9558 set operation as merely an internal storage change. */
9559 if (flags & SV_COW_DROP_PV) SvOK_off(sv);
9560 else sv_setsv_flags(sv, temp, 0);
9562 if ((const GV *)sv == PL_last_in_gv)
9563 PL_last_in_gv = NULL;
9564 else if ((const GV *)sv == PL_statgv)
9569 =for apidoc sv_unref_flags
9571 Unsets the RV status of the SV, and decrements the reference count of
9572 whatever was being referenced by the RV. This can almost be thought of
9573 as a reversal of C<newSVrv>. The C<cflags> argument can contain
9574 C<SV_IMMEDIATE_UNREF> to force the reference count to be decremented
9575 (otherwise the decrementing is conditional on the reference count being
9576 different from one or the reference being a readonly SV).
9583 Perl_sv_unref_flags(pTHX_ SV *const ref, const U32 flags)
9585 SV* const target = SvRV(ref);
9587 PERL_ARGS_ASSERT_SV_UNREF_FLAGS;
9589 if (SvWEAKREF(ref)) {
9590 sv_del_backref(target, ref);
9592 SvRV_set(ref, NULL);
9595 SvRV_set(ref, NULL);
9597 /* You can't have a || SvREADONLY(target) here, as $a = $$a, where $a was
9598 assigned to as BEGIN {$a = \"Foo"} will fail. */
9599 if (SvREFCNT(target) != 1 || (flags & SV_IMMEDIATE_UNREF))
9600 SvREFCNT_dec(target);
9601 else /* XXX Hack, but hard to make $a=$a->[1] work otherwise */
9602 sv_2mortal(target); /* Schedule for freeing later */
9606 =for apidoc sv_untaint
9608 Untaint an SV. Use C<SvTAINTED_off> instead.
9614 Perl_sv_untaint(pTHX_ SV *const sv)
9616 PERL_ARGS_ASSERT_SV_UNTAINT;
9618 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
9619 MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
9626 =for apidoc sv_tainted
9628 Test an SV for taintedness. Use C<SvTAINTED> instead.
9634 Perl_sv_tainted(pTHX_ SV *const sv)
9636 PERL_ARGS_ASSERT_SV_TAINTED;
9638 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
9639 const MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
9640 if (mg && (mg->mg_len & 1) )
9647 =for apidoc sv_setpviv
9649 Copies an integer into the given SV, also updating its string value.
9650 Does not handle 'set' magic. See C<sv_setpviv_mg>.
9656 Perl_sv_setpviv(pTHX_ SV *const sv, const IV iv)
9658 char buf[TYPE_CHARS(UV)];
9660 char * const ptr = uiv_2buf(buf, iv, 0, 0, &ebuf);
9662 PERL_ARGS_ASSERT_SV_SETPVIV;
9664 sv_setpvn(sv, ptr, ebuf - ptr);
9668 =for apidoc sv_setpviv_mg
9670 Like C<sv_setpviv>, but also handles 'set' magic.
9676 Perl_sv_setpviv_mg(pTHX_ SV *const sv, const IV iv)
9678 PERL_ARGS_ASSERT_SV_SETPVIV_MG;
9684 #if defined(PERL_IMPLICIT_CONTEXT)
9686 /* pTHX_ magic can't cope with varargs, so this is a no-context
9687 * version of the main function, (which may itself be aliased to us).
9688 * Don't access this version directly.
9692 Perl_sv_setpvf_nocontext(SV *const sv, const char *const pat, ...)
9697 PERL_ARGS_ASSERT_SV_SETPVF_NOCONTEXT;
9699 va_start(args, pat);
9700 sv_vsetpvf(sv, pat, &args);
9704 /* pTHX_ magic can't cope with varargs, so this is a no-context
9705 * version of the main function, (which may itself be aliased to us).
9706 * Don't access this version directly.
9710 Perl_sv_setpvf_mg_nocontext(SV *const sv, const char *const pat, ...)
9715 PERL_ARGS_ASSERT_SV_SETPVF_MG_NOCONTEXT;
9717 va_start(args, pat);
9718 sv_vsetpvf_mg(sv, pat, &args);
9724 =for apidoc sv_setpvf
9726 Works like C<sv_catpvf> but copies the text into the SV instead of
9727 appending it. Does not handle 'set' magic. See C<sv_setpvf_mg>.
9733 Perl_sv_setpvf(pTHX_ SV *const sv, const char *const pat, ...)
9737 PERL_ARGS_ASSERT_SV_SETPVF;
9739 va_start(args, pat);
9740 sv_vsetpvf(sv, pat, &args);
9745 =for apidoc sv_vsetpvf
9747 Works like C<sv_vcatpvf> but copies the text into the SV instead of
9748 appending it. Does not handle 'set' magic. See C<sv_vsetpvf_mg>.
9750 Usually used via its frontend C<sv_setpvf>.
9756 Perl_sv_vsetpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9758 PERL_ARGS_ASSERT_SV_VSETPVF;
9760 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9764 =for apidoc sv_setpvf_mg
9766 Like C<sv_setpvf>, but also handles 'set' magic.
9772 Perl_sv_setpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
9776 PERL_ARGS_ASSERT_SV_SETPVF_MG;
9778 va_start(args, pat);
9779 sv_vsetpvf_mg(sv, pat, &args);
9784 =for apidoc sv_vsetpvf_mg
9786 Like C<sv_vsetpvf>, but also handles 'set' magic.
9788 Usually used via its frontend C<sv_setpvf_mg>.
9794 Perl_sv_vsetpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9796 PERL_ARGS_ASSERT_SV_VSETPVF_MG;
9798 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9802 #if defined(PERL_IMPLICIT_CONTEXT)
9804 /* pTHX_ magic can't cope with varargs, so this is a no-context
9805 * version of the main function, (which may itself be aliased to us).
9806 * Don't access this version directly.
9810 Perl_sv_catpvf_nocontext(SV *const sv, const char *const pat, ...)
9815 PERL_ARGS_ASSERT_SV_CATPVF_NOCONTEXT;
9817 va_start(args, pat);
9818 sv_vcatpvf(sv, pat, &args);
9822 /* pTHX_ magic can't cope with varargs, so this is a no-context
9823 * version of the main function, (which may itself be aliased to us).
9824 * Don't access this version directly.
9828 Perl_sv_catpvf_mg_nocontext(SV *const sv, const char *const pat, ...)
9833 PERL_ARGS_ASSERT_SV_CATPVF_MG_NOCONTEXT;
9835 va_start(args, pat);
9836 sv_vcatpvf_mg(sv, pat, &args);
9842 =for apidoc sv_catpvf
9844 Processes its arguments like C<sprintf> and appends the formatted
9845 output to an SV. If the appended data contains "wide" characters
9846 (including, but not limited to, SVs with a UTF-8 PV formatted with %s,
9847 and characters >255 formatted with %c), the original SV might get
9848 upgraded to UTF-8. Handles 'get' magic, but not 'set' magic. See
9849 C<sv_catpvf_mg>. If the original SV was UTF-8, the pattern should be
9850 valid UTF-8; if the original SV was bytes, the pattern should be too.
9855 Perl_sv_catpvf(pTHX_ SV *const sv, const char *const pat, ...)
9859 PERL_ARGS_ASSERT_SV_CATPVF;
9861 va_start(args, pat);
9862 sv_vcatpvf(sv, pat, &args);
9867 =for apidoc sv_vcatpvf
9869 Processes its arguments like C<vsprintf> and appends the formatted output
9870 to an SV. Does not handle 'set' magic. See C<sv_vcatpvf_mg>.
9872 Usually used via its frontend C<sv_catpvf>.
9878 Perl_sv_vcatpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9880 PERL_ARGS_ASSERT_SV_VCATPVF;
9882 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9886 =for apidoc sv_catpvf_mg
9888 Like C<sv_catpvf>, but also handles 'set' magic.
9894 Perl_sv_catpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
9898 PERL_ARGS_ASSERT_SV_CATPVF_MG;
9900 va_start(args, pat);
9901 sv_vcatpvf_mg(sv, pat, &args);
9906 =for apidoc sv_vcatpvf_mg
9908 Like C<sv_vcatpvf>, but also handles 'set' magic.
9910 Usually used via its frontend C<sv_catpvf_mg>.
9916 Perl_sv_vcatpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9918 PERL_ARGS_ASSERT_SV_VCATPVF_MG;
9920 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9925 =for apidoc sv_vsetpvfn
9927 Works like C<sv_vcatpvfn> but copies the text into the SV instead of
9930 Usually used via one of its frontends C<sv_vsetpvf> and C<sv_vsetpvf_mg>.
9936 Perl_sv_vsetpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
9937 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
9939 PERL_ARGS_ASSERT_SV_VSETPVFN;
9942 sv_vcatpvfn(sv, pat, patlen, args, svargs, svmax, maybe_tainted);
9947 * Warn of missing argument to sprintf, and then return a defined value
9948 * to avoid inappropriate "use of uninit" warnings [perl #71000].
9950 #define WARN_MISSING WARN_UNINITIALIZED /* Not sure we want a new category */
9952 S_vcatpvfn_missing_argument(pTHX) {
9953 if (ckWARN(WARN_MISSING)) {
9954 Perl_warner(aTHX_ packWARN(WARN_MISSING), "Missing argument in %s",
9955 PL_op ? OP_DESC(PL_op) : "sv_vcatpvfn()");
9962 S_expect_number(pTHX_ char **const pattern)
9967 PERL_ARGS_ASSERT_EXPECT_NUMBER;
9969 switch (**pattern) {
9970 case '1': case '2': case '3':
9971 case '4': case '5': case '6':
9972 case '7': case '8': case '9':
9973 var = *(*pattern)++ - '0';
9974 while (isDIGIT(**pattern)) {
9975 const I32 tmp = var * 10 + (*(*pattern)++ - '0');
9977 Perl_croak(aTHX_ "Integer overflow in format string for %s", (PL_op ? OP_DESC(PL_op) : "sv_vcatpvfn"));
9985 S_F0convert(NV nv, char *const endbuf, STRLEN *const len)
9987 const int neg = nv < 0;
9990 PERL_ARGS_ASSERT_F0CONVERT;
9998 if (uv & 1 && uv == nv)
9999 uv--; /* Round to even */
10001 const unsigned dig = uv % 10;
10003 } while (uv /= 10);
10014 =for apidoc sv_vcatpvfn
10016 Processes its arguments like C<vsprintf> and appends the formatted output
10017 to an SV. Uses an array of SVs if the C style variable argument list is
10018 missing (NULL). When running with taint checks enabled, indicates via
10019 C<maybe_tainted> if results are untrustworthy (often due to the use of
10022 Usually used via one of its frontends C<sv_vcatpvf> and C<sv_vcatpvf_mg>.
10028 #define VECTORIZE_ARGS vecsv = va_arg(*args, SV*);\
10029 vecstr = (U8*)SvPV_const(vecsv,veclen);\
10030 vec_utf8 = DO_UTF8(vecsv);
10032 /* XXX maybe_tainted is never assigned to, so the doc above is lying. */
10035 Perl_sv_vcatpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
10036 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
10041 const char *patend;
10044 static const char nullstr[] = "(null)";
10046 bool has_utf8 = DO_UTF8(sv); /* has the result utf8? */
10047 const bool pat_utf8 = has_utf8; /* the pattern is in utf8? */
10049 /* Times 4: a decimal digit takes more than 3 binary digits.
10050 * NV_DIG: mantissa takes than many decimal digits.
10051 * Plus 32: Playing safe. */
10052 char ebuf[IV_DIG * 4 + NV_DIG + 32];
10053 /* large enough for "%#.#f" --chip */
10054 /* what about long double NVs? --jhi */
10056 PERL_ARGS_ASSERT_SV_VCATPVFN;
10057 PERL_UNUSED_ARG(maybe_tainted);
10059 /* no matter what, this is a string now */
10060 (void)SvPV_force(sv, origlen);
10062 /* special-case "", "%s", and "%-p" (SVf - see below) */
10065 if (patlen == 2 && pat[0] == '%' && pat[1] == 's') {
10067 const char * const s = va_arg(*args, char*);
10068 sv_catpv(sv, s ? s : nullstr);
10070 else if (svix < svmax) {
10071 sv_catsv(sv, *svargs);
10074 S_vcatpvfn_missing_argument(aTHX);
10077 if (args && patlen == 3 && pat[0] == '%' &&
10078 pat[1] == '-' && pat[2] == 'p') {
10079 argsv = MUTABLE_SV(va_arg(*args, void*));
10080 sv_catsv(sv, argsv);
10084 #ifndef USE_LONG_DOUBLE
10085 /* special-case "%.<number>[gf]" */
10086 if ( !args && patlen <= 5 && pat[0] == '%' && pat[1] == '.'
10087 && (pat[patlen-1] == 'g' || pat[patlen-1] == 'f') ) {
10088 unsigned digits = 0;
10092 while (*pp >= '0' && *pp <= '9')
10093 digits = 10 * digits + (*pp++ - '0');
10094 if (pp - pat == (int)patlen - 1 && svix < svmax) {
10095 const NV nv = SvNV(*svargs);
10097 /* Add check for digits != 0 because it seems that some
10098 gconverts are buggy in this case, and we don't yet have
10099 a Configure test for this. */
10100 if (digits && digits < sizeof(ebuf) - NV_DIG - 10) {
10101 /* 0, point, slack */
10102 Gconvert(nv, (int)digits, 0, ebuf);
10103 sv_catpv(sv, ebuf);
10104 if (*ebuf) /* May return an empty string for digits==0 */
10107 } else if (!digits) {
10110 if ((p = F0convert(nv, ebuf + sizeof ebuf, &l))) {
10111 sv_catpvn(sv, p, l);
10117 #endif /* !USE_LONG_DOUBLE */
10119 if (!args && svix < svmax && DO_UTF8(*svargs))
10122 patend = (char*)pat + patlen;
10123 for (p = (char*)pat; p < patend; p = q) {
10126 bool vectorize = FALSE;
10127 bool vectorarg = FALSE;
10128 bool vec_utf8 = FALSE;
10134 bool has_precis = FALSE;
10136 const I32 osvix = svix;
10137 bool is_utf8 = FALSE; /* is this item utf8? */
10138 #ifdef HAS_LDBL_SPRINTF_BUG
10139 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
10140 with sfio - Allen <allens@cpan.org> */
10141 bool fix_ldbl_sprintf_bug = FALSE;
10145 U8 utf8buf[UTF8_MAXBYTES+1];
10146 STRLEN esignlen = 0;
10148 const char *eptr = NULL;
10149 const char *fmtstart;
10152 const U8 *vecstr = NULL;
10159 /* we need a long double target in case HAS_LONG_DOUBLE but
10160 not USE_LONG_DOUBLE
10162 #if defined(HAS_LONG_DOUBLE) && LONG_DOUBLESIZE > DOUBLESIZE
10170 const char *dotstr = ".";
10171 STRLEN dotstrlen = 1;
10172 I32 efix = 0; /* explicit format parameter index */
10173 I32 ewix = 0; /* explicit width index */
10174 I32 epix = 0; /* explicit precision index */
10175 I32 evix = 0; /* explicit vector index */
10176 bool asterisk = FALSE;
10178 /* echo everything up to the next format specification */
10179 for (q = p; q < patend && *q != '%'; ++q) ;
10181 if (has_utf8 && !pat_utf8)
10182 sv_catpvn_utf8_upgrade(sv, p, q - p, nsv);
10184 sv_catpvn(sv, p, q - p);
10193 We allow format specification elements in this order:
10194 \d+\$ explicit format parameter index
10196 v|\*(\d+\$)?v vector with optional (optionally specified) arg
10197 0 flag (as above): repeated to allow "v02"
10198 \d+|\*(\d+\$)? width using optional (optionally specified) arg
10199 \.(\d*|\*(\d+\$)?) precision using optional (optionally specified) arg
10201 [%bcdefginopsuxDFOUX] format (mandatory)
10206 As of perl5.9.3, printf format checking is on by default.
10207 Internally, perl uses %p formats to provide an escape to
10208 some extended formatting. This block deals with those
10209 extensions: if it does not match, (char*)q is reset and
10210 the normal format processing code is used.
10212 Currently defined extensions are:
10213 %p include pointer address (standard)
10214 %-p (SVf) include an SV (previously %_)
10215 %-<num>p include an SV with precision <num>
10217 %3p include a HEK with precision of 256
10218 %<num>p (where num != 2 or 3) reserved for future
10221 Robin Barker 2005-07-14 (but modified since)
10223 %1p (VDf) removed. RMB 2007-10-19
10230 n = expect_number(&q);
10232 if (sv) { /* SVf */
10237 argsv = MUTABLE_SV(va_arg(*args, void*));
10238 eptr = SvPV_const(argsv, elen);
10239 if (DO_UTF8(argsv))
10243 else if (n==2 || n==3) { /* HEKf */
10244 HEK * const hek = va_arg(*args, HEK *);
10245 eptr = HEK_KEY(hek);
10246 elen = HEK_LEN(hek);
10247 if (HEK_UTF8(hek)) is_utf8 = TRUE;
10248 if (n==3) precis = 256, has_precis = TRUE;
10252 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL),
10253 "internal %%<num>p might conflict with future printf extensions");
10259 if ( (width = expect_number(&q)) ) {
10274 if (plus == '+' && *q == ' ') /* '+' over ' ' */
10303 if ( (ewix = expect_number(&q)) )
10312 if ((vectorarg = asterisk)) {
10325 width = expect_number(&q);
10328 if (vectorize && vectorarg) {
10329 /* vectorizing, but not with the default "." */
10331 vecsv = va_arg(*args, SV*);
10333 vecsv = (evix > 0 && evix <= svmax)
10334 ? svargs[evix-1] : S_vcatpvfn_missing_argument(aTHX);
10336 vecsv = svix < svmax
10337 ? svargs[svix++] : S_vcatpvfn_missing_argument(aTHX);
10339 dotstr = SvPV_const(vecsv, dotstrlen);
10340 /* Keep the DO_UTF8 test *after* the SvPV call, else things go
10341 bad with tied or overloaded values that return UTF8. */
10342 if (DO_UTF8(vecsv))
10344 else if (has_utf8) {
10345 vecsv = sv_mortalcopy(vecsv);
10346 sv_utf8_upgrade(vecsv);
10347 dotstr = SvPV_const(vecsv, dotstrlen);
10354 i = va_arg(*args, int);
10356 i = (ewix ? ewix <= svmax : svix < svmax) ?
10357 SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
10359 width = (i < 0) ? -i : i;
10369 if ( ((epix = expect_number(&q))) && (*q++ != '$') )
10371 /* XXX: todo, support specified precision parameter */
10375 i = va_arg(*args, int);
10377 i = (ewix ? ewix <= svmax : svix < svmax)
10378 ? SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
10380 has_precis = !(i < 0);
10384 while (isDIGIT(*q))
10385 precis = precis * 10 + (*q++ - '0');
10394 else if (efix ? (efix > 0 && efix <= svmax) : svix < svmax) {
10395 vecsv = svargs[efix ? efix-1 : svix++];
10396 vecstr = (U8*)SvPV_const(vecsv,veclen);
10397 vec_utf8 = DO_UTF8(vecsv);
10399 /* if this is a version object, we need to convert
10400 * back into v-string notation and then let the
10401 * vectorize happen normally
10403 if (sv_isobject(vecsv) && sv_derived_from(vecsv, "version")) {
10404 char *version = savesvpv(vecsv);
10405 if ( hv_exists(MUTABLE_HV(SvRV(vecsv)), "alpha", 5 ) ) {
10406 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
10407 "vector argument not supported with alpha versions");
10410 vecsv = sv_newmortal();
10411 scan_vstring(version, version + veclen, vecsv);
10412 vecstr = (U8*)SvPV_const(vecsv, veclen);
10413 vec_utf8 = DO_UTF8(vecsv);
10427 case 'I': /* Ix, I32x, and I64x */
10429 if (q[1] == '6' && q[2] == '4') {
10435 if (q[1] == '3' && q[2] == '2') {
10445 #if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
10457 #if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
10458 if (*q == 'l') { /* lld, llf */
10467 if (*++q == 'h') { /* hhd, hhu */
10496 if (!vectorize && !args) {
10498 const I32 i = efix-1;
10499 argsv = (i >= 0 && i < svmax)
10500 ? svargs[i] : S_vcatpvfn_missing_argument(aTHX);
10502 argsv = (svix >= 0 && svix < svmax)
10503 ? svargs[svix++] : S_vcatpvfn_missing_argument(aTHX);
10507 switch (c = *q++) {
10514 uv = (args) ? va_arg(*args, int) : SvIV(argsv);
10516 (!UNI_IS_INVARIANT(uv) && SvUTF8(sv)))
10518 eptr = (char*)utf8buf;
10519 elen = uvchr_to_utf8((U8*)eptr, uv) - utf8buf;
10533 eptr = va_arg(*args, char*);
10535 elen = strlen(eptr);
10537 eptr = (char *)nullstr;
10538 elen = sizeof nullstr - 1;
10542 eptr = SvPV_const(argsv, elen);
10543 if (DO_UTF8(argsv)) {
10544 STRLEN old_precis = precis;
10545 if (has_precis && precis < elen) {
10546 STRLEN ulen = sv_len_utf8(argsv);
10547 I32 p = precis > ulen ? ulen : precis;
10548 sv_pos_u2b(argsv, &p, 0); /* sticks at end */
10551 if (width) { /* fudge width (can't fudge elen) */
10552 if (has_precis && precis < elen)
10553 width += precis - old_precis;
10555 width += elen - sv_len_utf8(argsv);
10562 if (has_precis && precis < elen)
10569 if (alt || vectorize)
10571 uv = PTR2UV(args ? va_arg(*args, void*) : argsv);
10592 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
10601 esignbuf[esignlen++] = plus;
10605 case 'c': iv = (char)va_arg(*args, int); break;
10606 case 'h': iv = (short)va_arg(*args, int); break;
10607 case 'l': iv = va_arg(*args, long); break;
10608 case 'V': iv = va_arg(*args, IV); break;
10609 case 'z': iv = va_arg(*args, SSize_t); break;
10610 case 't': iv = va_arg(*args, ptrdiff_t); break;
10611 default: iv = va_arg(*args, int); break;
10613 case 'j': iv = va_arg(*args, intmax_t); break;
10617 iv = va_arg(*args, Quad_t); break;
10624 IV tiv = SvIV(argsv); /* work around GCC bug #13488 */
10626 case 'c': iv = (char)tiv; break;
10627 case 'h': iv = (short)tiv; break;
10628 case 'l': iv = (long)tiv; break;
10630 default: iv = tiv; break;
10633 iv = (Quad_t)tiv; break;
10639 if ( !vectorize ) /* we already set uv above */
10644 esignbuf[esignlen++] = plus;
10648 esignbuf[esignlen++] = '-';
10692 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
10703 case 'c': uv = (unsigned char)va_arg(*args, unsigned); break;
10704 case 'h': uv = (unsigned short)va_arg(*args, unsigned); break;
10705 case 'l': uv = va_arg(*args, unsigned long); break;
10706 case 'V': uv = va_arg(*args, UV); break;
10707 case 'z': uv = va_arg(*args, Size_t); break;
10708 case 't': uv = va_arg(*args, ptrdiff_t); break; /* will sign extend, but there is no uptrdiff_t, so oh well */
10710 case 'j': uv = va_arg(*args, uintmax_t); break;
10712 default: uv = va_arg(*args, unsigned); break;
10715 uv = va_arg(*args, Uquad_t); break;
10722 UV tuv = SvUV(argsv); /* work around GCC bug #13488 */
10724 case 'c': uv = (unsigned char)tuv; break;
10725 case 'h': uv = (unsigned short)tuv; break;
10726 case 'l': uv = (unsigned long)tuv; break;
10728 default: uv = tuv; break;
10731 uv = (Uquad_t)tuv; break;
10740 char *ptr = ebuf + sizeof ebuf;
10741 bool tempalt = uv ? alt : FALSE; /* Vectors can't change alt */
10747 p = (char *)((c == 'X') ? PL_hexdigit + 16 : PL_hexdigit);
10751 } while (uv >>= 4);
10753 esignbuf[esignlen++] = '0';
10754 esignbuf[esignlen++] = c; /* 'x' or 'X' */
10760 *--ptr = '0' + dig;
10761 } while (uv >>= 3);
10762 if (alt && *ptr != '0')
10768 *--ptr = '0' + dig;
10769 } while (uv >>= 1);
10771 esignbuf[esignlen++] = '0';
10772 esignbuf[esignlen++] = c;
10775 default: /* it had better be ten or less */
10778 *--ptr = '0' + dig;
10779 } while (uv /= base);
10782 elen = (ebuf + sizeof ebuf) - ptr;
10786 zeros = precis - elen;
10787 else if (precis == 0 && elen == 1 && *eptr == '0'
10788 && !(base == 8 && alt)) /* "%#.0o" prints "0" */
10791 /* a precision nullifies the 0 flag. */
10798 /* FLOATING POINT */
10801 c = 'f'; /* maybe %F isn't supported here */
10803 case 'e': case 'E':
10805 case 'g': case 'G':
10809 /* This is evil, but floating point is even more evil */
10811 /* for SV-style calling, we can only get NV
10812 for C-style calling, we assume %f is double;
10813 for simplicity we allow any of %Lf, %llf, %qf for long double
10817 #if defined(USE_LONG_DOUBLE)
10821 /* [perl #20339] - we should accept and ignore %lf rather than die */
10825 #if defined(USE_LONG_DOUBLE)
10826 intsize = args ? 0 : 'q';
10830 #if defined(HAS_LONG_DOUBLE)
10843 /* now we need (long double) if intsize == 'q', else (double) */
10845 #if LONG_DOUBLESIZE > DOUBLESIZE
10847 va_arg(*args, long double) :
10848 va_arg(*args, double)
10850 va_arg(*args, double)
10855 /* nv * 0 will be NaN for NaN, +Inf and -Inf, and 0 for anything
10856 else. frexp() has some unspecified behaviour for those three */
10857 if (c != 'e' && c != 'E' && (nv * 0) == 0) {
10859 /* FIXME: if HAS_LONG_DOUBLE but not USE_LONG_DOUBLE this
10860 will cast our (long double) to (double) */
10861 (void)Perl_frexp(nv, &i);
10862 if (i == PERL_INT_MIN)
10863 Perl_die(aTHX_ "panic: frexp");
10865 need = BIT_DIGITS(i);
10867 need += has_precis ? precis : 6; /* known default */
10872 #ifdef HAS_LDBL_SPRINTF_BUG
10873 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
10874 with sfio - Allen <allens@cpan.org> */
10877 # define MY_DBL_MAX DBL_MAX
10878 # else /* XXX guessing! HUGE_VAL may be defined as infinity, so not using */
10879 # if DOUBLESIZE >= 8
10880 # define MY_DBL_MAX 1.7976931348623157E+308L
10882 # define MY_DBL_MAX 3.40282347E+38L
10886 # ifdef HAS_LDBL_SPRINTF_BUG_LESS1 /* only between -1L & 1L - Allen */
10887 # define MY_DBL_MAX_BUG 1L
10889 # define MY_DBL_MAX_BUG MY_DBL_MAX
10893 # define MY_DBL_MIN DBL_MIN
10894 # else /* XXX guessing! -Allen */
10895 # if DOUBLESIZE >= 8
10896 # define MY_DBL_MIN 2.2250738585072014E-308L
10898 # define MY_DBL_MIN 1.17549435E-38L
10902 if ((intsize == 'q') && (c == 'f') &&
10903 ((nv < MY_DBL_MAX_BUG) && (nv > -MY_DBL_MAX_BUG)) &&
10904 (need < DBL_DIG)) {
10905 /* it's going to be short enough that
10906 * long double precision is not needed */
10908 if ((nv <= 0L) && (nv >= -0L))
10909 fix_ldbl_sprintf_bug = TRUE; /* 0 is 0 - easiest */
10911 /* would use Perl_fp_class as a double-check but not
10912 * functional on IRIX - see perl.h comments */
10914 if ((nv >= MY_DBL_MIN) || (nv <= -MY_DBL_MIN)) {
10915 /* It's within the range that a double can represent */
10916 #if defined(DBL_MAX) && !defined(DBL_MIN)
10917 if ((nv >= ((long double)1/DBL_MAX)) ||
10918 (nv <= (-(long double)1/DBL_MAX)))
10920 fix_ldbl_sprintf_bug = TRUE;
10923 if (fix_ldbl_sprintf_bug == TRUE) {
10933 # undef MY_DBL_MAX_BUG
10936 #endif /* HAS_LDBL_SPRINTF_BUG */
10938 need += 20; /* fudge factor */
10939 if (PL_efloatsize < need) {
10940 Safefree(PL_efloatbuf);
10941 PL_efloatsize = need + 20; /* more fudge */
10942 Newx(PL_efloatbuf, PL_efloatsize, char);
10943 PL_efloatbuf[0] = '\0';
10946 if ( !(width || left || plus || alt) && fill != '0'
10947 && has_precis && intsize != 'q' ) { /* Shortcuts */
10948 /* See earlier comment about buggy Gconvert when digits,
10950 if ( c == 'g' && precis) {
10951 Gconvert((NV)nv, (int)precis, 0, PL_efloatbuf);
10952 /* May return an empty string for digits==0 */
10953 if (*PL_efloatbuf) {
10954 elen = strlen(PL_efloatbuf);
10955 goto float_converted;
10957 } else if ( c == 'f' && !precis) {
10958 if ((eptr = F0convert(nv, ebuf + sizeof ebuf, &elen)))
10963 char *ptr = ebuf + sizeof ebuf;
10966 /* FIXME: what to do if HAS_LONG_DOUBLE but not PERL_PRIfldbl? */
10967 #if defined(HAS_LONG_DOUBLE) && defined(PERL_PRIfldbl)
10968 if (intsize == 'q') {
10969 /* Copy the one or more characters in a long double
10970 * format before the 'base' ([efgEFG]) character to
10971 * the format string. */
10972 static char const prifldbl[] = PERL_PRIfldbl;
10973 char const *p = prifldbl + sizeof(prifldbl) - 3;
10974 while (p >= prifldbl) { *--ptr = *p--; }
10979 do { *--ptr = '0' + (base % 10); } while (base /= 10);
10984 do { *--ptr = '0' + (base % 10); } while (base /= 10);
10996 /* No taint. Otherwise we are in the strange situation
10997 * where printf() taints but print($float) doesn't.
10999 #if defined(HAS_LONG_DOUBLE)
11000 elen = ((intsize == 'q')
11001 ? my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, nv)
11002 : my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, (double)nv));
11004 elen = my_sprintf(PL_efloatbuf, ptr, nv);
11008 eptr = PL_efloatbuf;
11016 i = SvCUR(sv) - origlen;
11019 case 'c': *(va_arg(*args, char*)) = i; break;
11020 case 'h': *(va_arg(*args, short*)) = i; break;
11021 default: *(va_arg(*args, int*)) = i; break;
11022 case 'l': *(va_arg(*args, long*)) = i; break;
11023 case 'V': *(va_arg(*args, IV*)) = i; break;
11024 case 'z': *(va_arg(*args, SSize_t*)) = i; break;
11025 case 't': *(va_arg(*args, ptrdiff_t*)) = i; break;
11027 case 'j': *(va_arg(*args, intmax_t*)) = i; break;
11031 *(va_arg(*args, Quad_t*)) = i; break;
11038 sv_setuv_mg(argsv, has_utf8 ? (UV)sv_len_utf8(sv) : (UV)i);
11039 continue; /* not "break" */
11046 && (PL_op->op_type == OP_PRTF || PL_op->op_type == OP_SPRINTF)
11047 && ckWARN(WARN_PRINTF))
11049 SV * const msg = sv_newmortal();
11050 Perl_sv_setpvf(aTHX_ msg, "Invalid conversion in %sprintf: ",
11051 (PL_op->op_type == OP_PRTF) ? "" : "s");
11052 if (fmtstart < patend) {
11053 const char * const fmtend = q < patend ? q : patend;
11055 sv_catpvs(msg, "\"%");
11056 for (f = fmtstart; f < fmtend; f++) {
11058 sv_catpvn(msg, f, 1);
11060 Perl_sv_catpvf(aTHX_ msg,
11061 "\\%03"UVof, (UV)*f & 0xFF);
11064 sv_catpvs(msg, "\"");
11066 sv_catpvs(msg, "end of string");
11068 Perl_warner(aTHX_ packWARN(WARN_PRINTF), "%"SVf, SVfARG(msg)); /* yes, this is reentrant */
11071 /* output mangled stuff ... */
11077 /* ... right here, because formatting flags should not apply */
11078 SvGROW(sv, SvCUR(sv) + elen + 1);
11080 Copy(eptr, p, elen, char);
11083 SvCUR_set(sv, p - SvPVX_const(sv));
11085 continue; /* not "break" */
11088 if (is_utf8 != has_utf8) {
11091 sv_utf8_upgrade(sv);
11094 const STRLEN old_elen = elen;
11095 SV * const nsv = newSVpvn_flags(eptr, elen, SVs_TEMP);
11096 sv_utf8_upgrade(nsv);
11097 eptr = SvPVX_const(nsv);
11100 if (width) { /* fudge width (can't fudge elen) */
11101 width += elen - old_elen;
11107 have = esignlen + zeros + elen;
11109 Perl_croak_nocontext("%s", PL_memory_wrap);
11111 need = (have > width ? have : width);
11114 if (need >= (((STRLEN)~0) - SvCUR(sv) - dotstrlen - 1))
11115 Perl_croak_nocontext("%s", PL_memory_wrap);
11116 SvGROW(sv, SvCUR(sv) + need + dotstrlen + 1);
11118 if (esignlen && fill == '0') {
11120 for (i = 0; i < (int)esignlen; i++)
11121 *p++ = esignbuf[i];
11123 if (gap && !left) {
11124 memset(p, fill, gap);
11127 if (esignlen && fill != '0') {
11129 for (i = 0; i < (int)esignlen; i++)
11130 *p++ = esignbuf[i];
11134 for (i = zeros; i; i--)
11138 Copy(eptr, p, elen, char);
11142 memset(p, ' ', gap);
11147 Copy(dotstr, p, dotstrlen, char);
11151 vectorize = FALSE; /* done iterating over vecstr */
11158 SvCUR_set(sv, p - SvPVX_const(sv));
11167 /* =========================================================================
11169 =head1 Cloning an interpreter
11171 All the macros and functions in this section are for the private use of
11172 the main function, perl_clone().
11174 The foo_dup() functions make an exact copy of an existing foo thingy.
11175 During the course of a cloning, a hash table is used to map old addresses
11176 to new addresses. The table is created and manipulated with the
11177 ptr_table_* functions.
11181 * =========================================================================*/
11184 #if defined(USE_ITHREADS)
11186 /* XXX Remove this so it doesn't have to go thru the macro and return for nothing */
11187 #ifndef GpREFCNT_inc
11188 # define GpREFCNT_inc(gp) ((gp) ? (++(gp)->gp_refcnt, (gp)) : (GP*)NULL)
11192 /* Certain cases in Perl_ss_dup have been merged, by relying on the fact
11193 that currently av_dup, gv_dup and hv_dup are the same as sv_dup.
11194 If this changes, please unmerge ss_dup.
11195 Likewise, sv_dup_inc_multiple() relies on this fact. */
11196 #define sv_dup_inc_NN(s,t) SvREFCNT_inc_NN(sv_dup_inc(s,t))
11197 #define av_dup(s,t) MUTABLE_AV(sv_dup((const SV *)s,t))
11198 #define av_dup_inc(s,t) MUTABLE_AV(sv_dup_inc((const SV *)s,t))
11199 #define hv_dup(s,t) MUTABLE_HV(sv_dup((const SV *)s,t))
11200 #define hv_dup_inc(s,t) MUTABLE_HV(sv_dup_inc((const SV *)s,t))
11201 #define cv_dup(s,t) MUTABLE_CV(sv_dup((const SV *)s,t))
11202 #define cv_dup_inc(s,t) MUTABLE_CV(sv_dup_inc((const SV *)s,t))
11203 #define io_dup(s,t) MUTABLE_IO(sv_dup((const SV *)s,t))
11204 #define io_dup_inc(s,t) MUTABLE_IO(sv_dup_inc((const SV *)s,t))
11205 #define gv_dup(s,t) MUTABLE_GV(sv_dup((const SV *)s,t))
11206 #define gv_dup_inc(s,t) MUTABLE_GV(sv_dup_inc((const SV *)s,t))
11207 #define SAVEPV(p) ((p) ? savepv(p) : NULL)
11208 #define SAVEPVN(p,n) ((p) ? savepvn(p,n) : NULL)
11210 /* clone a parser */
11213 Perl_parser_dup(pTHX_ const yy_parser *const proto, CLONE_PARAMS *const param)
11217 PERL_ARGS_ASSERT_PARSER_DUP;
11222 /* look for it in the table first */
11223 parser = (yy_parser *)ptr_table_fetch(PL_ptr_table, proto);
11227 /* create anew and remember what it is */
11228 Newxz(parser, 1, yy_parser);
11229 ptr_table_store(PL_ptr_table, proto, parser);
11231 /* XXX these not yet duped */
11232 parser->old_parser = NULL;
11233 parser->stack = NULL;
11235 parser->stack_size = 0;
11236 /* XXX parser->stack->state = 0; */
11238 /* XXX eventually, just Copy() most of the parser struct ? */
11240 parser->lex_brackets = proto->lex_brackets;
11241 parser->lex_casemods = proto->lex_casemods;
11242 parser->lex_brackstack = savepvn(proto->lex_brackstack,
11243 (proto->lex_brackets < 120 ? 120 : proto->lex_brackets));
11244 parser->lex_casestack = savepvn(proto->lex_casestack,
11245 (proto->lex_casemods < 12 ? 12 : proto->lex_casemods));
11246 parser->lex_defer = proto->lex_defer;
11247 parser->lex_dojoin = proto->lex_dojoin;
11248 parser->lex_expect = proto->lex_expect;
11249 parser->lex_formbrack = proto->lex_formbrack;
11250 parser->lex_inpat = proto->lex_inpat;
11251 parser->lex_inwhat = proto->lex_inwhat;
11252 parser->lex_op = proto->lex_op;
11253 parser->lex_repl = sv_dup_inc(proto->lex_repl, param);
11254 parser->lex_starts = proto->lex_starts;
11255 parser->lex_stuff = sv_dup_inc(proto->lex_stuff, param);
11256 parser->multi_close = proto->multi_close;
11257 parser->multi_open = proto->multi_open;
11258 parser->multi_start = proto->multi_start;
11259 parser->multi_end = proto->multi_end;
11260 parser->pending_ident = proto->pending_ident;
11261 parser->preambled = proto->preambled;
11262 parser->sublex_info = proto->sublex_info; /* XXX not quite right */
11263 parser->linestr = sv_dup_inc(proto->linestr, param);
11264 parser->expect = proto->expect;
11265 parser->copline = proto->copline;
11266 parser->last_lop_op = proto->last_lop_op;
11267 parser->lex_state = proto->lex_state;
11268 parser->rsfp = fp_dup(proto->rsfp, '<', param);
11269 /* rsfp_filters entries have fake IoDIRP() */
11270 parser->rsfp_filters= av_dup_inc(proto->rsfp_filters, param);
11271 parser->in_my = proto->in_my;
11272 parser->in_my_stash = hv_dup(proto->in_my_stash, param);
11273 parser->error_count = proto->error_count;
11276 parser->linestr = sv_dup_inc(proto->linestr, param);
11279 char * const ols = SvPVX(proto->linestr);
11280 char * const ls = SvPVX(parser->linestr);
11282 parser->bufptr = ls + (proto->bufptr >= ols ?
11283 proto->bufptr - ols : 0);
11284 parser->oldbufptr = ls + (proto->oldbufptr >= ols ?
11285 proto->oldbufptr - ols : 0);
11286 parser->oldoldbufptr= ls + (proto->oldoldbufptr >= ols ?
11287 proto->oldoldbufptr - ols : 0);
11288 parser->linestart = ls + (proto->linestart >= ols ?
11289 proto->linestart - ols : 0);
11290 parser->last_uni = ls + (proto->last_uni >= ols ?
11291 proto->last_uni - ols : 0);
11292 parser->last_lop = ls + (proto->last_lop >= ols ?
11293 proto->last_lop - ols : 0);
11295 parser->bufend = ls + SvCUR(parser->linestr);
11298 Copy(proto->tokenbuf, parser->tokenbuf, 256, char);
11302 parser->endwhite = proto->endwhite;
11303 parser->faketokens = proto->faketokens;
11304 parser->lasttoke = proto->lasttoke;
11305 parser->nextwhite = proto->nextwhite;
11306 parser->realtokenstart = proto->realtokenstart;
11307 parser->skipwhite = proto->skipwhite;
11308 parser->thisclose = proto->thisclose;
11309 parser->thismad = proto->thismad;
11310 parser->thisopen = proto->thisopen;
11311 parser->thisstuff = proto->thisstuff;
11312 parser->thistoken = proto->thistoken;
11313 parser->thiswhite = proto->thiswhite;
11315 Copy(proto->nexttoke, parser->nexttoke, 5, NEXTTOKE);
11316 parser->curforce = proto->curforce;
11318 Copy(proto->nextval, parser->nextval, 5, YYSTYPE);
11319 Copy(proto->nexttype, parser->nexttype, 5, I32);
11320 parser->nexttoke = proto->nexttoke;
11323 /* XXX should clone saved_curcop here, but we aren't passed
11324 * proto_perl; so do it in perl_clone_using instead */
11330 /* duplicate a file handle */
11333 Perl_fp_dup(pTHX_ PerlIO *const fp, const char type, CLONE_PARAMS *const param)
11337 PERL_ARGS_ASSERT_FP_DUP;
11338 PERL_UNUSED_ARG(type);
11341 return (PerlIO*)NULL;
11343 /* look for it in the table first */
11344 ret = (PerlIO*)ptr_table_fetch(PL_ptr_table, fp);
11348 /* create anew and remember what it is */
11349 ret = PerlIO_fdupopen(aTHX_ fp, param, PERLIO_DUP_CLONE);
11350 ptr_table_store(PL_ptr_table, fp, ret);
11354 /* duplicate a directory handle */
11357 Perl_dirp_dup(pTHX_ DIR *const dp, CLONE_PARAMS *const param)
11363 register const Direntry_t *dirent;
11364 char smallbuf[256];
11370 PERL_UNUSED_CONTEXT;
11371 PERL_ARGS_ASSERT_DIRP_DUP;
11376 /* look for it in the table first */
11377 ret = (DIR*)ptr_table_fetch(PL_ptr_table, dp);
11383 PERL_UNUSED_ARG(param);
11387 /* open the current directory (so we can switch back) */
11388 if (!(pwd = PerlDir_open("."))) return (DIR *)NULL;
11390 /* chdir to our dir handle and open the present working directory */
11391 if (fchdir(my_dirfd(dp)) < 0 || !(ret = PerlDir_open("."))) {
11392 PerlDir_close(pwd);
11393 return (DIR *)NULL;
11395 /* Now we should have two dir handles pointing to the same dir. */
11397 /* Be nice to the calling code and chdir back to where we were. */
11398 fchdir(my_dirfd(pwd)); /* If this fails, then what? */
11400 /* We have no need of the pwd handle any more. */
11401 PerlDir_close(pwd);
11404 # define d_namlen(d) (d)->d_namlen
11406 # define d_namlen(d) strlen((d)->d_name)
11408 /* Iterate once through dp, to get the file name at the current posi-
11409 tion. Then step back. */
11410 pos = PerlDir_tell(dp);
11411 if ((dirent = PerlDir_read(dp))) {
11412 len = d_namlen(dirent);
11413 if (len <= sizeof smallbuf) name = smallbuf;
11414 else Newx(name, len, char);
11415 Move(dirent->d_name, name, len, char);
11417 PerlDir_seek(dp, pos);
11419 /* Iterate through the new dir handle, till we find a file with the
11421 if (!dirent) /* just before the end */
11423 pos = PerlDir_tell(ret);
11424 if (PerlDir_read(ret)) continue; /* not there yet */
11425 PerlDir_seek(ret, pos); /* step back */
11429 const long pos0 = PerlDir_tell(ret);
11431 pos = PerlDir_tell(ret);
11432 if ((dirent = PerlDir_read(ret))) {
11433 if (len == d_namlen(dirent)
11434 && memEQ(name, dirent->d_name, len)) {
11436 PerlDir_seek(ret, pos); /* step back */
11439 /* else we are not there yet; keep iterating */
11441 else { /* This is not meant to happen. The best we can do is
11442 reset the iterator to the beginning. */
11443 PerlDir_seek(ret, pos0);
11450 if (name && name != smallbuf)
11455 ret = win32_dirp_dup(dp, param);
11458 /* pop it in the pointer table */
11460 ptr_table_store(PL_ptr_table, dp, ret);
11465 /* duplicate a typeglob */
11468 Perl_gp_dup(pTHX_ GP *const gp, CLONE_PARAMS *const param)
11472 PERL_ARGS_ASSERT_GP_DUP;
11476 /* look for it in the table first */
11477 ret = (GP*)ptr_table_fetch(PL_ptr_table, gp);
11481 /* create anew and remember what it is */
11483 ptr_table_store(PL_ptr_table, gp, ret);
11486 /* ret->gp_refcnt must be 0 before any other dups are called. We're relying
11487 on Newxz() to do this for us. */
11488 ret->gp_sv = sv_dup_inc(gp->gp_sv, param);
11489 ret->gp_io = io_dup_inc(gp->gp_io, param);
11490 ret->gp_form = cv_dup_inc(gp->gp_form, param);
11491 ret->gp_av = av_dup_inc(gp->gp_av, param);
11492 ret->gp_hv = hv_dup_inc(gp->gp_hv, param);
11493 ret->gp_egv = gv_dup(gp->gp_egv, param);/* GvEGV is not refcounted */
11494 ret->gp_cv = cv_dup_inc(gp->gp_cv, param);
11495 ret->gp_cvgen = gp->gp_cvgen;
11496 ret->gp_line = gp->gp_line;
11497 ret->gp_file_hek = hek_dup(gp->gp_file_hek, param);
11501 /* duplicate a chain of magic */
11504 Perl_mg_dup(pTHX_ MAGIC *mg, CLONE_PARAMS *const param)
11506 MAGIC *mgret = NULL;
11507 MAGIC **mgprev_p = &mgret;
11509 PERL_ARGS_ASSERT_MG_DUP;
11511 for (; mg; mg = mg->mg_moremagic) {
11514 if ((param->flags & CLONEf_JOIN_IN)
11515 && mg->mg_type == PERL_MAGIC_backref)
11516 /* when joining, we let the individual SVs add themselves to
11517 * backref as needed. */
11520 Newx(nmg, 1, MAGIC);
11522 mgprev_p = &(nmg->mg_moremagic);
11524 /* There was a comment "XXX copy dynamic vtable?" but as we don't have
11525 dynamic vtables, I'm not sure why Sarathy wrote it. The comment dates
11526 from the original commit adding Perl_mg_dup() - revision 4538.
11527 Similarly there is the annotation "XXX random ptr?" next to the
11528 assignment to nmg->mg_ptr. */
11531 /* FIXME for plugins
11532 if (nmg->mg_type == PERL_MAGIC_qr) {
11533 nmg->mg_obj = MUTABLE_SV(CALLREGDUPE((REGEXP*)nmg->mg_obj, param));
11537 nmg->mg_obj = (nmg->mg_flags & MGf_REFCOUNTED)
11538 ? nmg->mg_type == PERL_MAGIC_backref
11539 /* The backref AV has its reference
11540 * count deliberately bumped by 1 */
11541 ? SvREFCNT_inc(av_dup_inc((const AV *)
11542 nmg->mg_obj, param))
11543 : sv_dup_inc(nmg->mg_obj, param)
11544 : sv_dup(nmg->mg_obj, param);
11546 if (nmg->mg_ptr && nmg->mg_type != PERL_MAGIC_regex_global) {
11547 if (nmg->mg_len > 0) {
11548 nmg->mg_ptr = SAVEPVN(nmg->mg_ptr, nmg->mg_len);
11549 if (nmg->mg_type == PERL_MAGIC_overload_table &&
11550 AMT_AMAGIC((AMT*)nmg->mg_ptr))
11552 AMT * const namtp = (AMT*)nmg->mg_ptr;
11553 sv_dup_inc_multiple((SV**)(namtp->table),
11554 (SV**)(namtp->table), NofAMmeth, param);
11557 else if (nmg->mg_len == HEf_SVKEY)
11558 nmg->mg_ptr = (char*)sv_dup_inc((const SV *)nmg->mg_ptr, param);
11560 if ((nmg->mg_flags & MGf_DUP) && nmg->mg_virtual && nmg->mg_virtual->svt_dup) {
11561 nmg->mg_virtual->svt_dup(aTHX_ nmg, param);
11567 #endif /* USE_ITHREADS */
11569 struct ptr_tbl_arena {
11570 struct ptr_tbl_arena *next;
11571 struct ptr_tbl_ent array[1023/3]; /* as ptr_tbl_ent has 3 pointers. */
11574 /* create a new pointer-mapping table */
11577 Perl_ptr_table_new(pTHX)
11580 PERL_UNUSED_CONTEXT;
11582 Newx(tbl, 1, PTR_TBL_t);
11583 tbl->tbl_max = 511;
11584 tbl->tbl_items = 0;
11585 tbl->tbl_arena = NULL;
11586 tbl->tbl_arena_next = NULL;
11587 tbl->tbl_arena_end = NULL;
11588 Newxz(tbl->tbl_ary, tbl->tbl_max + 1, PTR_TBL_ENT_t*);
11592 #define PTR_TABLE_HASH(ptr) \
11593 ((PTR2UV(ptr) >> 3) ^ (PTR2UV(ptr) >> (3 + 7)) ^ (PTR2UV(ptr) >> (3 + 17)))
11595 /* map an existing pointer using a table */
11597 STATIC PTR_TBL_ENT_t *
11598 S_ptr_table_find(PTR_TBL_t *const tbl, const void *const sv)
11600 PTR_TBL_ENT_t *tblent;
11601 const UV hash = PTR_TABLE_HASH(sv);
11603 PERL_ARGS_ASSERT_PTR_TABLE_FIND;
11605 tblent = tbl->tbl_ary[hash & tbl->tbl_max];
11606 for (; tblent; tblent = tblent->next) {
11607 if (tblent->oldval == sv)
11614 Perl_ptr_table_fetch(pTHX_ PTR_TBL_t *const tbl, const void *const sv)
11616 PTR_TBL_ENT_t const *const tblent = ptr_table_find(tbl, sv);
11618 PERL_ARGS_ASSERT_PTR_TABLE_FETCH;
11619 PERL_UNUSED_CONTEXT;
11621 return tblent ? tblent->newval : NULL;
11624 /* add a new entry to a pointer-mapping table */
11627 Perl_ptr_table_store(pTHX_ PTR_TBL_t *const tbl, const void *const oldsv, void *const newsv)
11629 PTR_TBL_ENT_t *tblent = ptr_table_find(tbl, oldsv);
11631 PERL_ARGS_ASSERT_PTR_TABLE_STORE;
11632 PERL_UNUSED_CONTEXT;
11635 tblent->newval = newsv;
11637 const UV entry = PTR_TABLE_HASH(oldsv) & tbl->tbl_max;
11639 if (tbl->tbl_arena_next == tbl->tbl_arena_end) {
11640 struct ptr_tbl_arena *new_arena;
11642 Newx(new_arena, 1, struct ptr_tbl_arena);
11643 new_arena->next = tbl->tbl_arena;
11644 tbl->tbl_arena = new_arena;
11645 tbl->tbl_arena_next = new_arena->array;
11646 tbl->tbl_arena_end = new_arena->array
11647 + sizeof(new_arena->array) / sizeof(new_arena->array[0]);
11650 tblent = tbl->tbl_arena_next++;
11652 tblent->oldval = oldsv;
11653 tblent->newval = newsv;
11654 tblent->next = tbl->tbl_ary[entry];
11655 tbl->tbl_ary[entry] = tblent;
11657 if (tblent->next && tbl->tbl_items > tbl->tbl_max)
11658 ptr_table_split(tbl);
11662 /* double the hash bucket size of an existing ptr table */
11665 Perl_ptr_table_split(pTHX_ PTR_TBL_t *const tbl)
11667 PTR_TBL_ENT_t **ary = tbl->tbl_ary;
11668 const UV oldsize = tbl->tbl_max + 1;
11669 UV newsize = oldsize * 2;
11672 PERL_ARGS_ASSERT_PTR_TABLE_SPLIT;
11673 PERL_UNUSED_CONTEXT;
11675 Renew(ary, newsize, PTR_TBL_ENT_t*);
11676 Zero(&ary[oldsize], newsize-oldsize, PTR_TBL_ENT_t*);
11677 tbl->tbl_max = --newsize;
11678 tbl->tbl_ary = ary;
11679 for (i=0; i < oldsize; i++, ary++) {
11680 PTR_TBL_ENT_t **entp = ary;
11681 PTR_TBL_ENT_t *ent = *ary;
11682 PTR_TBL_ENT_t **curentp;
11685 curentp = ary + oldsize;
11687 if ((newsize & PTR_TABLE_HASH(ent->oldval)) != i) {
11689 ent->next = *curentp;
11699 /* remove all the entries from a ptr table */
11700 /* Deprecated - will be removed post 5.14 */
11703 Perl_ptr_table_clear(pTHX_ PTR_TBL_t *const tbl)
11705 if (tbl && tbl->tbl_items) {
11706 struct ptr_tbl_arena *arena = tbl->tbl_arena;
11708 Zero(tbl->tbl_ary, tbl->tbl_max + 1, struct ptr_tbl_ent **);
11711 struct ptr_tbl_arena *next = arena->next;
11717 tbl->tbl_items = 0;
11718 tbl->tbl_arena = NULL;
11719 tbl->tbl_arena_next = NULL;
11720 tbl->tbl_arena_end = NULL;
11724 /* clear and free a ptr table */
11727 Perl_ptr_table_free(pTHX_ PTR_TBL_t *const tbl)
11729 struct ptr_tbl_arena *arena;
11735 arena = tbl->tbl_arena;
11738 struct ptr_tbl_arena *next = arena->next;
11744 Safefree(tbl->tbl_ary);
11748 #if defined(USE_ITHREADS)
11751 Perl_rvpv_dup(pTHX_ SV *const dstr, const SV *const sstr, CLONE_PARAMS *const param)
11753 PERL_ARGS_ASSERT_RVPV_DUP;
11756 if (SvWEAKREF(sstr)) {
11757 SvRV_set(dstr, sv_dup(SvRV_const(sstr), param));
11758 if (param->flags & CLONEf_JOIN_IN) {
11759 /* if joining, we add any back references individually rather
11760 * than copying the whole backref array */
11761 Perl_sv_add_backref(aTHX_ SvRV(dstr), dstr);
11765 SvRV_set(dstr, sv_dup_inc(SvRV_const(sstr), param));
11767 else if (SvPVX_const(sstr)) {
11768 /* Has something there */
11770 /* Normal PV - clone whole allocated space */
11771 SvPV_set(dstr, SAVEPVN(SvPVX_const(sstr), SvLEN(sstr)-1));
11772 if (SvREADONLY(sstr) && SvFAKE(sstr)) {
11773 /* Not that normal - actually sstr is copy on write.
11774 But we are a true, independent SV, so: */
11775 SvREADONLY_off(dstr);
11780 /* Special case - not normally malloced for some reason */
11781 if (isGV_with_GP(sstr)) {
11782 /* Don't need to do anything here. */
11784 else if ((SvREADONLY(sstr) && SvFAKE(sstr))) {
11785 /* A "shared" PV - clone it as "shared" PV */
11787 HEK_KEY(hek_dup(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)),
11791 /* Some other special case - random pointer */
11792 SvPV_set(dstr, (char *) SvPVX_const(sstr));
11797 /* Copy the NULL */
11798 SvPV_set(dstr, NULL);
11802 /* duplicate a list of SVs. source and dest may point to the same memory. */
11804 S_sv_dup_inc_multiple(pTHX_ SV *const *source, SV **dest,
11805 SSize_t items, CLONE_PARAMS *const param)
11807 PERL_ARGS_ASSERT_SV_DUP_INC_MULTIPLE;
11809 while (items-- > 0) {
11810 *dest++ = sv_dup_inc(*source++, param);
11816 /* duplicate an SV of any type (including AV, HV etc) */
11819 S_sv_dup_common(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
11824 PERL_ARGS_ASSERT_SV_DUP_COMMON;
11826 if (SvTYPE(sstr) == (svtype)SVTYPEMASK) {
11827 #ifdef DEBUG_LEAKING_SCALARS_ABORT
11832 /* look for it in the table first */
11833 dstr = MUTABLE_SV(ptr_table_fetch(PL_ptr_table, sstr));
11837 if(param->flags & CLONEf_JOIN_IN) {
11838 /** We are joining here so we don't want do clone
11839 something that is bad **/
11840 if (SvTYPE(sstr) == SVt_PVHV) {
11841 const HEK * const hvname = HvNAME_HEK(sstr);
11843 /** don't clone stashes if they already exist **/
11844 dstr = MUTABLE_SV(gv_stashpvn(HEK_KEY(hvname), HEK_LEN(hvname),
11845 HEK_UTF8(hvname) ? SVf_UTF8 : 0));
11846 ptr_table_store(PL_ptr_table, sstr, dstr);
11850 else if (SvTYPE(sstr) == SVt_PVGV && !SvFAKE(sstr)) {
11851 HV *stash = GvSTASH(sstr);
11852 const HEK * hvname;
11853 if (stash && (hvname = HvNAME_HEK(stash))) {
11854 /** don't clone GVs if they already exist **/
11856 stash = gv_stashpvn(HEK_KEY(hvname), HEK_LEN(hvname),
11857 HEK_UTF8(hvname) ? SVf_UTF8 : 0);
11859 stash, GvNAME(sstr),
11865 if (svp && *svp && SvTYPE(*svp) == SVt_PVGV) {
11866 ptr_table_store(PL_ptr_table, sstr, *svp);
11873 /* create anew and remember what it is */
11876 #ifdef DEBUG_LEAKING_SCALARS
11877 dstr->sv_debug_optype = sstr->sv_debug_optype;
11878 dstr->sv_debug_line = sstr->sv_debug_line;
11879 dstr->sv_debug_inpad = sstr->sv_debug_inpad;
11880 dstr->sv_debug_parent = (SV*)sstr;
11881 FREE_SV_DEBUG_FILE(dstr);
11882 dstr->sv_debug_file = savepv(sstr->sv_debug_file);
11885 ptr_table_store(PL_ptr_table, sstr, dstr);
11888 SvFLAGS(dstr) = SvFLAGS(sstr);
11889 SvFLAGS(dstr) &= ~SVf_OOK; /* don't propagate OOK hack */
11890 SvREFCNT(dstr) = 0; /* must be before any other dups! */
11893 if (SvANY(sstr) && PL_watch_pvx && SvPVX_const(sstr) == PL_watch_pvx)
11894 PerlIO_printf(Perl_debug_log, "watch at %p hit, found string \"%s\"\n",
11895 (void*)PL_watch_pvx, SvPVX_const(sstr));
11898 /* don't clone objects whose class has asked us not to */
11899 if (SvOBJECT(sstr) && ! (SvFLAGS(SvSTASH(sstr)) & SVphv_CLONEABLE)) {
11904 switch (SvTYPE(sstr)) {
11906 SvANY(dstr) = NULL;
11909 SvANY(dstr) = (XPVIV*)((char*)&(dstr->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
11911 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
11913 SvIV_set(dstr, SvIVX(sstr));
11917 SvANY(dstr) = new_XNV();
11918 SvNV_set(dstr, SvNVX(sstr));
11920 /* case SVt_BIND: */
11923 /* These are all the types that need complex bodies allocating. */
11925 const svtype sv_type = SvTYPE(sstr);
11926 const struct body_details *const sv_type_details
11927 = bodies_by_type + sv_type;
11931 Perl_croak(aTHX_ "Bizarre SvTYPE [%" IVdf "]", (IV)SvTYPE(sstr));
11946 assert(sv_type_details->body_size);
11947 if (sv_type_details->arena) {
11948 new_body_inline(new_body, sv_type);
11950 = (void*)((char*)new_body - sv_type_details->offset);
11952 new_body = new_NOARENA(sv_type_details);
11956 SvANY(dstr) = new_body;
11959 Copy(((char*)SvANY(sstr)) + sv_type_details->offset,
11960 ((char*)SvANY(dstr)) + sv_type_details->offset,
11961 sv_type_details->copy, char);
11963 Copy(((char*)SvANY(sstr)),
11964 ((char*)SvANY(dstr)),
11965 sv_type_details->body_size + sv_type_details->offset, char);
11968 if (sv_type != SVt_PVAV && sv_type != SVt_PVHV
11969 && !isGV_with_GP(dstr)
11970 && !(sv_type == SVt_PVIO && !(IoFLAGS(dstr) & IOf_FAKE_DIRP)))
11971 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
11973 /* The Copy above means that all the source (unduplicated) pointers
11974 are now in the destination. We can check the flags and the
11975 pointers in either, but it's possible that there's less cache
11976 missing by always going for the destination.
11977 FIXME - instrument and check that assumption */
11978 if (sv_type >= SVt_PVMG) {
11979 if ((sv_type == SVt_PVMG) && SvPAD_OUR(dstr)) {
11980 SvOURSTASH_set(dstr, hv_dup_inc(SvOURSTASH(dstr), param));
11981 } else if (SvMAGIC(dstr))
11982 SvMAGIC_set(dstr, mg_dup(SvMAGIC(dstr), param));
11984 SvSTASH_set(dstr, hv_dup_inc(SvSTASH(dstr), param));
11987 /* The cast silences a GCC warning about unhandled types. */
11988 switch ((int)sv_type) {
11998 /* FIXME for plugins */
11999 re_dup_guts((REGEXP*) sstr, (REGEXP*) dstr, param);
12002 /* XXX LvTARGOFF sometimes holds PMOP* when DEBUGGING */
12003 if (LvTYPE(dstr) == 't') /* for tie: unrefcnted fake (SV**) */
12004 LvTARG(dstr) = dstr;
12005 else if (LvTYPE(dstr) == 'T') /* for tie: fake HE */
12006 LvTARG(dstr) = MUTABLE_SV(he_dup((HE*)LvTARG(dstr), 0, param));
12008 LvTARG(dstr) = sv_dup_inc(LvTARG(dstr), param);
12010 /* non-GP case already handled above */
12011 if(isGV_with_GP(sstr)) {
12012 GvNAME_HEK(dstr) = hek_dup(GvNAME_HEK(dstr), param);
12013 /* Don't call sv_add_backref here as it's going to be
12014 created as part of the magic cloning of the symbol
12015 table--unless this is during a join and the stash
12016 is not actually being cloned. */
12017 /* Danger Will Robinson - GvGP(dstr) isn't initialised
12018 at the point of this comment. */
12019 GvSTASH(dstr) = hv_dup(GvSTASH(dstr), param);
12020 if (param->flags & CLONEf_JOIN_IN)
12021 Perl_sv_add_backref(aTHX_ MUTABLE_SV(GvSTASH(dstr)), dstr);
12022 GvGP_set(dstr, gp_dup(GvGP(sstr), param));
12023 (void)GpREFCNT_inc(GvGP(dstr));
12027 /* PL_parser->rsfp_filters entries have fake IoDIRP() */
12028 if(IoFLAGS(dstr) & IOf_FAKE_DIRP) {
12029 /* I have no idea why fake dirp (rsfps)
12030 should be treated differently but otherwise
12031 we end up with leaks -- sky*/
12032 IoTOP_GV(dstr) = gv_dup_inc(IoTOP_GV(dstr), param);
12033 IoFMT_GV(dstr) = gv_dup_inc(IoFMT_GV(dstr), param);
12034 IoBOTTOM_GV(dstr) = gv_dup_inc(IoBOTTOM_GV(dstr), param);
12036 IoTOP_GV(dstr) = gv_dup(IoTOP_GV(dstr), param);
12037 IoFMT_GV(dstr) = gv_dup(IoFMT_GV(dstr), param);
12038 IoBOTTOM_GV(dstr) = gv_dup(IoBOTTOM_GV(dstr), param);
12039 if (IoDIRP(dstr)) {
12040 IoDIRP(dstr) = dirp_dup(IoDIRP(dstr), param);
12043 /* IoDIRP(dstr) is already a copy of IoDIRP(sstr) */
12045 IoIFP(dstr) = fp_dup(IoIFP(sstr), IoTYPE(dstr), param);
12047 if (IoOFP(dstr) == IoIFP(sstr))
12048 IoOFP(dstr) = IoIFP(dstr);
12050 IoOFP(dstr) = fp_dup(IoOFP(dstr), IoTYPE(dstr), param);
12051 IoTOP_NAME(dstr) = SAVEPV(IoTOP_NAME(dstr));
12052 IoFMT_NAME(dstr) = SAVEPV(IoFMT_NAME(dstr));
12053 IoBOTTOM_NAME(dstr) = SAVEPV(IoBOTTOM_NAME(dstr));
12056 /* avoid cloning an empty array */
12057 if (AvARRAY((const AV *)sstr) && AvFILLp((const AV *)sstr) >= 0) {
12058 SV **dst_ary, **src_ary;
12059 SSize_t items = AvFILLp((const AV *)sstr) + 1;
12061 src_ary = AvARRAY((const AV *)sstr);
12062 Newxz(dst_ary, AvMAX((const AV *)sstr)+1, SV*);
12063 ptr_table_store(PL_ptr_table, src_ary, dst_ary);
12064 AvARRAY(MUTABLE_AV(dstr)) = dst_ary;
12065 AvALLOC((const AV *)dstr) = dst_ary;
12066 if (AvREAL((const AV *)sstr)) {
12067 dst_ary = sv_dup_inc_multiple(src_ary, dst_ary, items,
12071 while (items-- > 0)
12072 *dst_ary++ = sv_dup(*src_ary++, param);
12074 items = AvMAX((const AV *)sstr) - AvFILLp((const AV *)sstr);
12075 while (items-- > 0) {
12076 *dst_ary++ = &PL_sv_undef;
12080 AvARRAY(MUTABLE_AV(dstr)) = NULL;
12081 AvALLOC((const AV *)dstr) = (SV**)NULL;
12082 AvMAX( (const AV *)dstr) = -1;
12083 AvFILLp((const AV *)dstr) = -1;
12087 if (HvARRAY((const HV *)sstr)) {
12089 const bool sharekeys = !!HvSHAREKEYS(sstr);
12090 XPVHV * const dxhv = (XPVHV*)SvANY(dstr);
12091 XPVHV * const sxhv = (XPVHV*)SvANY(sstr);
12093 Newx(darray, PERL_HV_ARRAY_ALLOC_BYTES(dxhv->xhv_max+1)
12094 + (SvOOK(sstr) ? sizeof(struct xpvhv_aux) : 0),
12096 HvARRAY(dstr) = (HE**)darray;
12097 while (i <= sxhv->xhv_max) {
12098 const HE * const source = HvARRAY(sstr)[i];
12099 HvARRAY(dstr)[i] = source
12100 ? he_dup(source, sharekeys, param) : 0;
12104 const struct xpvhv_aux * const saux = HvAUX(sstr);
12105 struct xpvhv_aux * const daux = HvAUX(dstr);
12106 /* This flag isn't copied. */
12109 if (saux->xhv_name_count) {
12110 HEK ** const sname = saux->xhv_name_u.xhvnameu_names;
12112 = saux->xhv_name_count < 0
12113 ? -saux->xhv_name_count
12114 : saux->xhv_name_count;
12115 HEK **shekp = sname + count;
12117 Newx(daux->xhv_name_u.xhvnameu_names, count, HEK *);
12118 dhekp = daux->xhv_name_u.xhvnameu_names + count;
12119 while (shekp-- > sname) {
12121 *dhekp = hek_dup(*shekp, param);
12125 daux->xhv_name_u.xhvnameu_name
12126 = hek_dup(saux->xhv_name_u.xhvnameu_name,
12129 daux->xhv_name_count = saux->xhv_name_count;
12131 daux->xhv_riter = saux->xhv_riter;
12132 daux->xhv_eiter = saux->xhv_eiter
12133 ? he_dup(saux->xhv_eiter,
12134 cBOOL(HvSHAREKEYS(sstr)), param) : 0;
12135 /* backref array needs refcnt=2; see sv_add_backref */
12136 daux->xhv_backreferences =
12137 (param->flags & CLONEf_JOIN_IN)
12138 /* when joining, we let the individual GVs and
12139 * CVs add themselves to backref as
12140 * needed. This avoids pulling in stuff
12141 * that isn't required, and simplifies the
12142 * case where stashes aren't cloned back
12143 * if they already exist in the parent
12146 : saux->xhv_backreferences
12147 ? (SvTYPE(saux->xhv_backreferences) == SVt_PVAV)
12148 ? MUTABLE_AV(SvREFCNT_inc(
12149 sv_dup_inc((const SV *)
12150 saux->xhv_backreferences, param)))
12151 : MUTABLE_AV(sv_dup((const SV *)
12152 saux->xhv_backreferences, param))
12155 daux->xhv_mro_meta = saux->xhv_mro_meta
12156 ? mro_meta_dup(saux->xhv_mro_meta, param)
12159 /* Record stashes for possible cloning in Perl_clone(). */
12161 av_push(param->stashes, dstr);
12165 HvARRAY(MUTABLE_HV(dstr)) = NULL;
12168 if (!(param->flags & CLONEf_COPY_STACKS)) {
12173 /* NOTE: not refcounted */
12174 SvANY(MUTABLE_CV(dstr))->xcv_stash =
12175 hv_dup(CvSTASH(dstr), param);
12176 if ((param->flags & CLONEf_JOIN_IN) && CvSTASH(dstr))
12177 Perl_sv_add_backref(aTHX_ MUTABLE_SV(CvSTASH(dstr)), dstr);
12178 if (!CvISXSUB(dstr)) {
12180 CvROOT(dstr) = OpREFCNT_inc(CvROOT(dstr));
12182 } else if (CvCONST(dstr)) {
12183 CvXSUBANY(dstr).any_ptr =
12184 sv_dup_inc((const SV *)CvXSUBANY(dstr).any_ptr, param);
12186 if (CvDYNFILE(dstr)) CvFILE(dstr) = SAVEPV(CvFILE(dstr));
12187 /* don't dup if copying back - CvGV isn't refcounted, so the
12188 * duped GV may never be freed. A bit of a hack! DAPM */
12189 SvANY(MUTABLE_CV(dstr))->xcv_gv =
12191 ? gv_dup_inc(CvGV(sstr), param)
12192 : (param->flags & CLONEf_JOIN_IN)
12194 : gv_dup(CvGV(sstr), param);
12196 CvPADLIST(dstr) = padlist_dup(CvPADLIST(sstr), param);
12198 CvWEAKOUTSIDE(sstr)
12199 ? cv_dup( CvOUTSIDE(dstr), param)
12200 : cv_dup_inc(CvOUTSIDE(dstr), param);
12206 if (SvOBJECT(dstr) && SvTYPE(dstr) != SVt_PVIO)
12213 Perl_sv_dup_inc(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
12215 PERL_ARGS_ASSERT_SV_DUP_INC;
12216 return sstr ? SvREFCNT_inc(sv_dup_common(sstr, param)) : NULL;
12220 Perl_sv_dup(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
12222 SV *dstr = sstr ? sv_dup_common(sstr, param) : NULL;
12223 PERL_ARGS_ASSERT_SV_DUP;
12225 /* Track every SV that (at least initially) had a reference count of 0.
12226 We need to do this by holding an actual reference to it in this array.
12227 If we attempt to cheat, turn AvREAL_off(), and store only pointers
12228 (akin to the stashes hash, and the perl stack), we come unstuck if
12229 a weak reference (or other SV legitimately SvREFCNT() == 0 for this
12230 thread) is manipulated in a CLONE method, because CLONE runs before the
12231 unreferenced array is walked to find SVs still with SvREFCNT() == 0
12232 (and fix things up by giving each a reference via the temps stack).
12233 Instead, during CLONE, if the 0-referenced SV has SvREFCNT_inc() and
12234 then SvREFCNT_dec(), it will be cleaned up (and added to the free list)
12235 before the walk of unreferenced happens and a reference to that is SV
12236 added to the temps stack. At which point we have the same SV considered
12237 to be in use, and free to be re-used. Not good.
12239 if (dstr && !(param->flags & CLONEf_COPY_STACKS) && !SvREFCNT(dstr)) {
12240 assert(param->unreferenced);
12241 av_push(param->unreferenced, SvREFCNT_inc(dstr));
12247 /* duplicate a context */
12250 Perl_cx_dup(pTHX_ PERL_CONTEXT *cxs, I32 ix, I32 max, CLONE_PARAMS* param)
12252 PERL_CONTEXT *ncxs;
12254 PERL_ARGS_ASSERT_CX_DUP;
12257 return (PERL_CONTEXT*)NULL;
12259 /* look for it in the table first */
12260 ncxs = (PERL_CONTEXT*)ptr_table_fetch(PL_ptr_table, cxs);
12264 /* create anew and remember what it is */
12265 Newx(ncxs, max + 1, PERL_CONTEXT);
12266 ptr_table_store(PL_ptr_table, cxs, ncxs);
12267 Copy(cxs, ncxs, max + 1, PERL_CONTEXT);
12270 PERL_CONTEXT * const ncx = &ncxs[ix];
12271 if (CxTYPE(ncx) == CXt_SUBST) {
12272 Perl_croak(aTHX_ "Cloning substitution context is unimplemented");
12275 switch (CxTYPE(ncx)) {
12277 ncx->blk_sub.cv = (ncx->blk_sub.olddepth == 0
12278 ? cv_dup_inc(ncx->blk_sub.cv, param)
12279 : cv_dup(ncx->blk_sub.cv,param));
12280 ncx->blk_sub.argarray = (CxHASARGS(ncx)
12281 ? av_dup_inc(ncx->blk_sub.argarray,
12284 ncx->blk_sub.savearray = av_dup_inc(ncx->blk_sub.savearray,
12286 ncx->blk_sub.oldcomppad = (PAD*)ptr_table_fetch(PL_ptr_table,
12287 ncx->blk_sub.oldcomppad);
12290 ncx->blk_eval.old_namesv = sv_dup_inc(ncx->blk_eval.old_namesv,
12292 ncx->blk_eval.cur_text = sv_dup(ncx->blk_eval.cur_text, param);
12294 case CXt_LOOP_LAZYSV:
12295 ncx->blk_loop.state_u.lazysv.end
12296 = sv_dup_inc(ncx->blk_loop.state_u.lazysv.end, param);
12297 /* We are taking advantage of av_dup_inc and sv_dup_inc
12298 actually being the same function, and order equivalence of
12300 We can assert the later [but only at run time :-(] */
12301 assert ((void *) &ncx->blk_loop.state_u.ary.ary ==
12302 (void *) &ncx->blk_loop.state_u.lazysv.cur);
12304 ncx->blk_loop.state_u.ary.ary
12305 = av_dup_inc(ncx->blk_loop.state_u.ary.ary, param);
12306 case CXt_LOOP_LAZYIV:
12307 case CXt_LOOP_PLAIN:
12308 if (CxPADLOOP(ncx)) {
12309 ncx->blk_loop.itervar_u.oldcomppad
12310 = (PAD*)ptr_table_fetch(PL_ptr_table,
12311 ncx->blk_loop.itervar_u.oldcomppad);
12313 ncx->blk_loop.itervar_u.gv
12314 = gv_dup((const GV *)ncx->blk_loop.itervar_u.gv,
12319 ncx->blk_format.cv = cv_dup(ncx->blk_format.cv, param);
12320 ncx->blk_format.gv = gv_dup(ncx->blk_format.gv, param);
12321 ncx->blk_format.dfoutgv = gv_dup_inc(ncx->blk_format.dfoutgv,
12334 /* duplicate a stack info structure */
12337 Perl_si_dup(pTHX_ PERL_SI *si, CLONE_PARAMS* param)
12341 PERL_ARGS_ASSERT_SI_DUP;
12344 return (PERL_SI*)NULL;
12346 /* look for it in the table first */
12347 nsi = (PERL_SI*)ptr_table_fetch(PL_ptr_table, si);
12351 /* create anew and remember what it is */
12352 Newxz(nsi, 1, PERL_SI);
12353 ptr_table_store(PL_ptr_table, si, nsi);
12355 nsi->si_stack = av_dup_inc(si->si_stack, param);
12356 nsi->si_cxix = si->si_cxix;
12357 nsi->si_cxmax = si->si_cxmax;
12358 nsi->si_cxstack = cx_dup(si->si_cxstack, si->si_cxix, si->si_cxmax, param);
12359 nsi->si_type = si->si_type;
12360 nsi->si_prev = si_dup(si->si_prev, param);
12361 nsi->si_next = si_dup(si->si_next, param);
12362 nsi->si_markoff = si->si_markoff;
12367 #define POPINT(ss,ix) ((ss)[--(ix)].any_i32)
12368 #define TOPINT(ss,ix) ((ss)[ix].any_i32)
12369 #define POPLONG(ss,ix) ((ss)[--(ix)].any_long)
12370 #define TOPLONG(ss,ix) ((ss)[ix].any_long)
12371 #define POPIV(ss,ix) ((ss)[--(ix)].any_iv)
12372 #define TOPIV(ss,ix) ((ss)[ix].any_iv)
12373 #define POPUV(ss,ix) ((ss)[--(ix)].any_uv)
12374 #define TOPUV(ss,ix) ((ss)[ix].any_uv)
12375 #define POPBOOL(ss,ix) ((ss)[--(ix)].any_bool)
12376 #define TOPBOOL(ss,ix) ((ss)[ix].any_bool)
12377 #define POPPTR(ss,ix) ((ss)[--(ix)].any_ptr)
12378 #define TOPPTR(ss,ix) ((ss)[ix].any_ptr)
12379 #define POPDPTR(ss,ix) ((ss)[--(ix)].any_dptr)
12380 #define TOPDPTR(ss,ix) ((ss)[ix].any_dptr)
12381 #define POPDXPTR(ss,ix) ((ss)[--(ix)].any_dxptr)
12382 #define TOPDXPTR(ss,ix) ((ss)[ix].any_dxptr)
12385 #define pv_dup_inc(p) SAVEPV(p)
12386 #define pv_dup(p) SAVEPV(p)
12387 #define svp_dup_inc(p,pp) any_dup(p,pp)
12389 /* map any object to the new equivent - either something in the
12390 * ptr table, or something in the interpreter structure
12394 Perl_any_dup(pTHX_ void *v, const PerlInterpreter *proto_perl)
12398 PERL_ARGS_ASSERT_ANY_DUP;
12401 return (void*)NULL;
12403 /* look for it in the table first */
12404 ret = ptr_table_fetch(PL_ptr_table, v);
12408 /* see if it is part of the interpreter structure */
12409 if (v >= (void*)proto_perl && v < (void*)(proto_perl+1))
12410 ret = (void*)(((char*)aTHX) + (((char*)v) - (char*)proto_perl));
12418 /* duplicate the save stack */
12421 Perl_ss_dup(pTHX_ PerlInterpreter *proto_perl, CLONE_PARAMS* param)
12424 ANY * const ss = proto_perl->Isavestack;
12425 const I32 max = proto_perl->Isavestack_max;
12426 I32 ix = proto_perl->Isavestack_ix;
12439 void (*dptr) (void*);
12440 void (*dxptr) (pTHX_ void*);
12442 PERL_ARGS_ASSERT_SS_DUP;
12444 Newxz(nss, max, ANY);
12447 const UV uv = POPUV(ss,ix);
12448 const U8 type = (U8)uv & SAVE_MASK;
12450 TOPUV(nss,ix) = uv;
12452 case SAVEt_CLEARSV:
12454 case SAVEt_HELEM: /* hash element */
12455 sv = (const SV *)POPPTR(ss,ix);
12456 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12458 case SAVEt_ITEM: /* normal string */
12459 case SAVEt_GVSV: /* scalar slot in GV */
12460 case SAVEt_SV: /* scalar reference */
12461 sv = (const SV *)POPPTR(ss,ix);
12462 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12465 case SAVEt_MORTALIZESV:
12466 sv = (const SV *)POPPTR(ss,ix);
12467 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12469 case SAVEt_SHARED_PVREF: /* char* in shared space */
12470 c = (char*)POPPTR(ss,ix);
12471 TOPPTR(nss,ix) = savesharedpv(c);
12472 ptr = POPPTR(ss,ix);
12473 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12475 case SAVEt_GENERIC_SVREF: /* generic sv */
12476 case SAVEt_SVREF: /* scalar reference */
12477 sv = (const SV *)POPPTR(ss,ix);
12478 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12479 ptr = POPPTR(ss,ix);
12480 TOPPTR(nss,ix) = svp_dup_inc((SV**)ptr, proto_perl);/* XXXXX */
12482 case SAVEt_HV: /* hash reference */
12483 case SAVEt_AV: /* array reference */
12484 sv = (const SV *) POPPTR(ss,ix);
12485 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12487 case SAVEt_COMPPAD:
12489 sv = (const SV *) POPPTR(ss,ix);
12490 TOPPTR(nss,ix) = sv_dup(sv, param);
12492 case SAVEt_INT: /* int reference */
12493 ptr = POPPTR(ss,ix);
12494 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12495 intval = (int)POPINT(ss,ix);
12496 TOPINT(nss,ix) = intval;
12498 case SAVEt_LONG: /* long reference */
12499 ptr = POPPTR(ss,ix);
12500 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12501 longval = (long)POPLONG(ss,ix);
12502 TOPLONG(nss,ix) = longval;
12504 case SAVEt_I32: /* I32 reference */
12505 ptr = POPPTR(ss,ix);
12506 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12508 TOPINT(nss,ix) = i;
12510 case SAVEt_IV: /* IV reference */
12511 ptr = POPPTR(ss,ix);
12512 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12514 TOPIV(nss,ix) = iv;
12516 case SAVEt_HPTR: /* HV* reference */
12517 case SAVEt_APTR: /* AV* reference */
12518 case SAVEt_SPTR: /* SV* reference */
12519 ptr = POPPTR(ss,ix);
12520 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12521 sv = (const SV *)POPPTR(ss,ix);
12522 TOPPTR(nss,ix) = sv_dup(sv, param);
12524 case SAVEt_VPTR: /* random* reference */
12525 ptr = POPPTR(ss,ix);
12526 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12528 case SAVEt_INT_SMALL:
12529 case SAVEt_I32_SMALL:
12530 case SAVEt_I16: /* I16 reference */
12531 case SAVEt_I8: /* I8 reference */
12533 ptr = POPPTR(ss,ix);
12534 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12536 case SAVEt_GENERIC_PVREF: /* generic char* */
12537 case SAVEt_PPTR: /* char* reference */
12538 ptr = POPPTR(ss,ix);
12539 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12540 c = (char*)POPPTR(ss,ix);
12541 TOPPTR(nss,ix) = pv_dup(c);
12543 case SAVEt_GP: /* scalar reference */
12544 gp = (GP*)POPPTR(ss,ix);
12545 TOPPTR(nss,ix) = gp = gp_dup(gp, param);
12546 (void)GpREFCNT_inc(gp);
12547 gv = (const GV *)POPPTR(ss,ix);
12548 TOPPTR(nss,ix) = gv_dup_inc(gv, param);
12551 ptr = POPPTR(ss,ix);
12552 if (ptr && (((OP*)ptr)->op_private & OPpREFCOUNTED)) {
12553 /* these are assumed to be refcounted properly */
12555 switch (((OP*)ptr)->op_type) {
12557 case OP_LEAVESUBLV:
12561 case OP_LEAVEWRITE:
12562 TOPPTR(nss,ix) = ptr;
12565 (void) OpREFCNT_inc(o);
12569 TOPPTR(nss,ix) = NULL;
12574 TOPPTR(nss,ix) = NULL;
12576 case SAVEt_FREECOPHH:
12577 ptr = POPPTR(ss,ix);
12578 TOPPTR(nss,ix) = cophh_copy((COPHH *)ptr);
12581 hv = (const HV *)POPPTR(ss,ix);
12582 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
12584 TOPINT(nss,ix) = i;
12587 c = (char*)POPPTR(ss,ix);
12588 TOPPTR(nss,ix) = pv_dup_inc(c);
12590 case SAVEt_STACK_POS: /* Position on Perl stack */
12592 TOPINT(nss,ix) = i;
12594 case SAVEt_DESTRUCTOR:
12595 ptr = POPPTR(ss,ix);
12596 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
12597 dptr = POPDPTR(ss,ix);
12598 TOPDPTR(nss,ix) = DPTR2FPTR(void (*)(void*),
12599 any_dup(FPTR2DPTR(void *, dptr),
12602 case SAVEt_DESTRUCTOR_X:
12603 ptr = POPPTR(ss,ix);
12604 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
12605 dxptr = POPDXPTR(ss,ix);
12606 TOPDXPTR(nss,ix) = DPTR2FPTR(void (*)(pTHX_ void*),
12607 any_dup(FPTR2DPTR(void *, dxptr),
12610 case SAVEt_REGCONTEXT:
12612 ix -= uv >> SAVE_TIGHT_SHIFT;
12614 case SAVEt_AELEM: /* array element */
12615 sv = (const SV *)POPPTR(ss,ix);
12616 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12618 TOPINT(nss,ix) = i;
12619 av = (const AV *)POPPTR(ss,ix);
12620 TOPPTR(nss,ix) = av_dup_inc(av, param);
12623 ptr = POPPTR(ss,ix);
12624 TOPPTR(nss,ix) = ptr;
12627 ptr = POPPTR(ss,ix);
12628 ptr = cophh_copy((COPHH*)ptr);
12629 TOPPTR(nss,ix) = ptr;
12631 TOPINT(nss,ix) = i;
12632 if (i & HINT_LOCALIZE_HH) {
12633 hv = (const HV *)POPPTR(ss,ix);
12634 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
12637 case SAVEt_PADSV_AND_MORTALIZE:
12638 longval = (long)POPLONG(ss,ix);
12639 TOPLONG(nss,ix) = longval;
12640 ptr = POPPTR(ss,ix);
12641 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12642 sv = (const SV *)POPPTR(ss,ix);
12643 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12645 case SAVEt_SET_SVFLAGS:
12647 TOPINT(nss,ix) = i;
12649 TOPINT(nss,ix) = i;
12650 sv = (const SV *)POPPTR(ss,ix);
12651 TOPPTR(nss,ix) = sv_dup(sv, param);
12653 case SAVEt_RE_STATE:
12655 const struct re_save_state *const old_state
12656 = (struct re_save_state *)
12657 (ss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
12658 struct re_save_state *const new_state
12659 = (struct re_save_state *)
12660 (nss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
12662 Copy(old_state, new_state, 1, struct re_save_state);
12663 ix -= SAVESTACK_ALLOC_FOR_RE_SAVE_STATE;
12665 new_state->re_state_bostr
12666 = pv_dup(old_state->re_state_bostr);
12667 new_state->re_state_reginput
12668 = pv_dup(old_state->re_state_reginput);
12669 new_state->re_state_regeol
12670 = pv_dup(old_state->re_state_regeol);
12671 new_state->re_state_regoffs
12672 = (regexp_paren_pair*)
12673 any_dup(old_state->re_state_regoffs, proto_perl);
12674 new_state->re_state_reglastparen
12675 = (U32*) any_dup(old_state->re_state_reglastparen,
12677 new_state->re_state_reglastcloseparen
12678 = (U32*)any_dup(old_state->re_state_reglastcloseparen,
12680 /* XXX This just has to be broken. The old save_re_context
12681 code did SAVEGENERICPV(PL_reg_start_tmp);
12682 PL_reg_start_tmp is char **.
12683 Look above to what the dup code does for
12684 SAVEt_GENERIC_PVREF
12685 It can never have worked.
12686 So this is merely a faithful copy of the exiting bug: */
12687 new_state->re_state_reg_start_tmp
12688 = (char **) pv_dup((char *)
12689 old_state->re_state_reg_start_tmp);
12690 /* I assume that it only ever "worked" because no-one called
12691 (pseudo)fork while the regexp engine had re-entered itself.
12693 #ifdef PERL_OLD_COPY_ON_WRITE
12694 new_state->re_state_nrs
12695 = sv_dup(old_state->re_state_nrs, param);
12697 new_state->re_state_reg_magic
12698 = (MAGIC*) any_dup(old_state->re_state_reg_magic,
12700 new_state->re_state_reg_oldcurpm
12701 = (PMOP*) any_dup(old_state->re_state_reg_oldcurpm,
12703 new_state->re_state_reg_curpm
12704 = (PMOP*) any_dup(old_state->re_state_reg_curpm,
12706 new_state->re_state_reg_oldsaved
12707 = pv_dup(old_state->re_state_reg_oldsaved);
12708 new_state->re_state_reg_poscache
12709 = pv_dup(old_state->re_state_reg_poscache);
12710 new_state->re_state_reg_starttry
12711 = pv_dup(old_state->re_state_reg_starttry);
12714 case SAVEt_COMPILE_WARNINGS:
12715 ptr = POPPTR(ss,ix);
12716 TOPPTR(nss,ix) = DUP_WARNINGS((STRLEN*)ptr);
12719 ptr = POPPTR(ss,ix);
12720 TOPPTR(nss,ix) = parser_dup((const yy_parser*)ptr, param);
12724 "panic: ss_dup inconsistency (%"IVdf")", (IV) type);
12732 /* if sv is a stash, call $class->CLONE_SKIP(), and set the SVphv_CLONEABLE
12733 * flag to the result. This is done for each stash before cloning starts,
12734 * so we know which stashes want their objects cloned */
12737 do_mark_cloneable_stash(pTHX_ SV *const sv)
12739 const HEK * const hvname = HvNAME_HEK((const HV *)sv);
12741 GV* const cloner = gv_fetchmethod_autoload(MUTABLE_HV(sv), "CLONE_SKIP", 0);
12742 SvFLAGS(sv) |= SVphv_CLONEABLE; /* clone objects by default */
12743 if (cloner && GvCV(cloner)) {
12750 mXPUSHs(newSVhek(hvname));
12752 call_sv(MUTABLE_SV(GvCV(cloner)), G_SCALAR);
12759 SvFLAGS(sv) &= ~SVphv_CLONEABLE;
12767 =for apidoc perl_clone
12769 Create and return a new interpreter by cloning the current one.
12771 perl_clone takes these flags as parameters:
12773 CLONEf_COPY_STACKS - is used to, well, copy the stacks also,
12774 without it we only clone the data and zero the stacks,
12775 with it we copy the stacks and the new perl interpreter is
12776 ready to run at the exact same point as the previous one.
12777 The pseudo-fork code uses COPY_STACKS while the
12778 threads->create doesn't.
12780 CLONEf_KEEP_PTR_TABLE -
12781 perl_clone keeps a ptr_table with the pointer of the old
12782 variable as a key and the new variable as a value,
12783 this allows it to check if something has been cloned and not
12784 clone it again but rather just use the value and increase the
12785 refcount. If KEEP_PTR_TABLE is not set then perl_clone will kill
12786 the ptr_table using the function
12787 C<ptr_table_free(PL_ptr_table); PL_ptr_table = NULL;>,
12788 reason to keep it around is if you want to dup some of your own
12789 variable who are outside the graph perl scans, example of this
12790 code is in threads.xs create.
12792 CLONEf_CLONE_HOST -
12793 This is a win32 thing, it is ignored on unix, it tells perls
12794 win32host code (which is c++) to clone itself, this is needed on
12795 win32 if you want to run two threads at the same time,
12796 if you just want to do some stuff in a separate perl interpreter
12797 and then throw it away and return to the original one,
12798 you don't need to do anything.
12803 /* XXX the above needs expanding by someone who actually understands it ! */
12804 EXTERN_C PerlInterpreter *
12805 perl_clone_host(PerlInterpreter* proto_perl, UV flags);
12808 perl_clone(PerlInterpreter *proto_perl, UV flags)
12811 #ifdef PERL_IMPLICIT_SYS
12813 PERL_ARGS_ASSERT_PERL_CLONE;
12815 /* perlhost.h so we need to call into it
12816 to clone the host, CPerlHost should have a c interface, sky */
12818 if (flags & CLONEf_CLONE_HOST) {
12819 return perl_clone_host(proto_perl,flags);
12821 return perl_clone_using(proto_perl, flags,
12823 proto_perl->IMemShared,
12824 proto_perl->IMemParse,
12826 proto_perl->IStdIO,
12830 proto_perl->IProc);
12834 perl_clone_using(PerlInterpreter *proto_perl, UV flags,
12835 struct IPerlMem* ipM, struct IPerlMem* ipMS,
12836 struct IPerlMem* ipMP, struct IPerlEnv* ipE,
12837 struct IPerlStdIO* ipStd, struct IPerlLIO* ipLIO,
12838 struct IPerlDir* ipD, struct IPerlSock* ipS,
12839 struct IPerlProc* ipP)
12841 /* XXX many of the string copies here can be optimized if they're
12842 * constants; they need to be allocated as common memory and just
12843 * their pointers copied. */
12846 CLONE_PARAMS clone_params;
12847 CLONE_PARAMS* const param = &clone_params;
12849 PerlInterpreter * const my_perl = (PerlInterpreter*)(*ipM->pMalloc)(ipM, sizeof(PerlInterpreter));
12851 PERL_ARGS_ASSERT_PERL_CLONE_USING;
12852 #else /* !PERL_IMPLICIT_SYS */
12854 CLONE_PARAMS clone_params;
12855 CLONE_PARAMS* param = &clone_params;
12856 PerlInterpreter * const my_perl = (PerlInterpreter*)PerlMem_malloc(sizeof(PerlInterpreter));
12858 PERL_ARGS_ASSERT_PERL_CLONE;
12859 #endif /* PERL_IMPLICIT_SYS */
12861 /* for each stash, determine whether its objects should be cloned */
12862 S_visit(proto_perl, do_mark_cloneable_stash, SVt_PVHV, SVTYPEMASK);
12863 PERL_SET_THX(my_perl);
12866 PoisonNew(my_perl, 1, PerlInterpreter);
12869 PL_defstash = NULL; /* may be used by perl malloc() */
12872 PL_scopestack_name = 0;
12874 PL_savestack_ix = 0;
12875 PL_savestack_max = -1;
12876 PL_sig_pending = 0;
12878 Zero(&PL_debug_pad, 1, struct perl_debug_pad);
12879 # ifdef DEBUG_LEAKING_SCALARS
12880 PL_sv_serial = (((UV)my_perl >> 2) & 0xfff) * 1000000;
12882 #else /* !DEBUGGING */
12883 Zero(my_perl, 1, PerlInterpreter);
12884 #endif /* DEBUGGING */
12886 #ifdef PERL_IMPLICIT_SYS
12887 /* host pointers */
12889 PL_MemShared = ipMS;
12890 PL_MemParse = ipMP;
12897 #endif /* PERL_IMPLICIT_SYS */
12899 param->flags = flags;
12900 /* Nothing in the core code uses this, but we make it available to
12901 extensions (using mg_dup). */
12902 param->proto_perl = proto_perl;
12903 /* Likely nothing will use this, but it is initialised to be consistent
12904 with Perl_clone_params_new(). */
12905 param->new_perl = my_perl;
12906 param->unreferenced = NULL;
12908 INIT_TRACK_MEMPOOL(my_perl->Imemory_debug_header, my_perl);
12910 PL_body_arenas = NULL;
12911 Zero(&PL_body_roots, 1, PL_body_roots);
12914 PL_sv_objcount = 0;
12916 PL_sv_arenaroot = NULL;
12918 PL_debug = proto_perl->Idebug;
12920 PL_hash_seed = proto_perl->Ihash_seed;
12921 PL_rehash_seed = proto_perl->Irehash_seed;
12923 SvANY(&PL_sv_undef) = NULL;
12924 SvREFCNT(&PL_sv_undef) = (~(U32)0)/2;
12925 SvFLAGS(&PL_sv_undef) = SVf_READONLY|SVt_NULL;
12926 SvREFCNT(&PL_sv_no) = (~(U32)0)/2;
12927 SvFLAGS(&PL_sv_no) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
12928 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
12930 SvANY(&PL_sv_yes) = new_XPVNV();
12931 SvREFCNT(&PL_sv_yes) = (~(U32)0)/2;
12932 SvFLAGS(&PL_sv_yes) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
12933 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
12935 /* dbargs array probably holds garbage */
12938 PL_compiling = proto_perl->Icompiling;
12940 #ifdef PERL_DEBUG_READONLY_OPS
12945 /* pseudo environmental stuff */
12946 PL_origargc = proto_perl->Iorigargc;
12947 PL_origargv = proto_perl->Iorigargv;
12949 /* Set tainting stuff before PerlIO_debug can possibly get called */
12950 PL_tainting = proto_perl->Itainting;
12951 PL_taint_warn = proto_perl->Itaint_warn;
12953 PL_minus_c = proto_perl->Iminus_c;
12955 PL_localpatches = proto_perl->Ilocalpatches;
12956 PL_splitstr = proto_perl->Isplitstr;
12957 PL_minus_n = proto_perl->Iminus_n;
12958 PL_minus_p = proto_perl->Iminus_p;
12959 PL_minus_l = proto_perl->Iminus_l;
12960 PL_minus_a = proto_perl->Iminus_a;
12961 PL_minus_E = proto_perl->Iminus_E;
12962 PL_minus_F = proto_perl->Iminus_F;
12963 PL_doswitches = proto_perl->Idoswitches;
12964 PL_dowarn = proto_perl->Idowarn;
12965 PL_sawampersand = proto_perl->Isawampersand;
12966 PL_unsafe = proto_perl->Iunsafe;
12967 PL_perldb = proto_perl->Iperldb;
12968 PL_perl_destruct_level = proto_perl->Iperl_destruct_level;
12969 PL_exit_flags = proto_perl->Iexit_flags;
12971 /* XXX time(&PL_basetime) when asked for? */
12972 PL_basetime = proto_perl->Ibasetime;
12974 PL_maxsysfd = proto_perl->Imaxsysfd;
12975 PL_statusvalue = proto_perl->Istatusvalue;
12977 PL_statusvalue_vms = proto_perl->Istatusvalue_vms;
12979 PL_statusvalue_posix = proto_perl->Istatusvalue_posix;
12982 /* RE engine related */
12983 Zero(&PL_reg_state, 1, struct re_save_state);
12984 PL_reginterp_cnt = 0;
12985 PL_regmatch_slab = NULL;
12987 PL_sub_generation = proto_perl->Isub_generation;
12989 /* funky return mechanisms */
12990 PL_forkprocess = proto_perl->Iforkprocess;
12992 /* internal state */
12993 PL_maxo = proto_perl->Imaxo;
12995 PL_main_start = proto_perl->Imain_start;
12996 PL_eval_root = proto_perl->Ieval_root;
12997 PL_eval_start = proto_perl->Ieval_start;
12999 PL_filemode = proto_perl->Ifilemode;
13000 PL_lastfd = proto_perl->Ilastfd;
13001 PL_oldname = proto_perl->Ioldname; /* XXX not quite right */
13004 PL_gensym = proto_perl->Igensym;
13006 PL_laststatval = proto_perl->Ilaststatval;
13007 PL_laststype = proto_perl->Ilaststype;
13010 PL_profiledata = NULL;
13012 PL_generation = proto_perl->Igeneration;
13014 PL_in_clean_objs = proto_perl->Iin_clean_objs;
13015 PL_in_clean_all = proto_perl->Iin_clean_all;
13017 PL_uid = proto_perl->Iuid;
13018 PL_euid = proto_perl->Ieuid;
13019 PL_gid = proto_perl->Igid;
13020 PL_egid = proto_perl->Iegid;
13021 PL_nomemok = proto_perl->Inomemok;
13022 PL_an = proto_perl->Ian;
13023 PL_evalseq = proto_perl->Ievalseq;
13024 PL_origenviron = proto_perl->Iorigenviron; /* XXX not quite right */
13025 PL_origalen = proto_perl->Iorigalen;
13027 PL_sighandlerp = proto_perl->Isighandlerp;
13029 PL_runops = proto_perl->Irunops;
13031 PL_subline = proto_perl->Isubline;
13034 PL_cryptseen = proto_perl->Icryptseen;
13037 PL_hints = proto_perl->Ihints;
13039 PL_amagic_generation = proto_perl->Iamagic_generation;
13041 #ifdef USE_LOCALE_COLLATE
13042 PL_collation_ix = proto_perl->Icollation_ix;
13043 PL_collation_standard = proto_perl->Icollation_standard;
13044 PL_collxfrm_base = proto_perl->Icollxfrm_base;
13045 PL_collxfrm_mult = proto_perl->Icollxfrm_mult;
13046 #endif /* USE_LOCALE_COLLATE */
13048 #ifdef USE_LOCALE_NUMERIC
13049 PL_numeric_standard = proto_perl->Inumeric_standard;
13050 PL_numeric_local = proto_perl->Inumeric_local;
13051 #endif /* !USE_LOCALE_NUMERIC */
13053 /* Did the locale setup indicate UTF-8? */
13054 PL_utf8locale = proto_perl->Iutf8locale;
13055 /* Unicode features (see perlrun/-C) */
13056 PL_unicode = proto_perl->Iunicode;
13058 /* Pre-5.8 signals control */
13059 PL_signals = proto_perl->Isignals;
13061 /* times() ticks per second */
13062 PL_clocktick = proto_perl->Iclocktick;
13064 /* Recursion stopper for PerlIO_find_layer */
13065 PL_in_load_module = proto_perl->Iin_load_module;
13067 /* sort() routine */
13068 PL_sort_RealCmp = proto_perl->Isort_RealCmp;
13070 /* Not really needed/useful since the reenrant_retint is "volatile",
13071 * but do it for consistency's sake. */
13072 PL_reentrant_retint = proto_perl->Ireentrant_retint;
13074 /* Hooks to shared SVs and locks. */
13075 PL_sharehook = proto_perl->Isharehook;
13076 PL_lockhook = proto_perl->Ilockhook;
13077 PL_unlockhook = proto_perl->Iunlockhook;
13078 PL_threadhook = proto_perl->Ithreadhook;
13079 PL_destroyhook = proto_perl->Idestroyhook;
13080 PL_signalhook = proto_perl->Isignalhook;
13082 PL_globhook = proto_perl->Iglobhook;
13084 #ifdef THREADS_HAVE_PIDS
13085 PL_ppid = proto_perl->Ippid;
13089 PL_last_swash_hv = NULL; /* reinits on demand */
13090 PL_last_swash_klen = 0;
13091 PL_last_swash_key[0]= '\0';
13092 PL_last_swash_tmps = (U8*)NULL;
13093 PL_last_swash_slen = 0;
13095 PL_glob_index = proto_perl->Iglob_index;
13096 PL_srand_called = proto_perl->Isrand_called;
13098 if (flags & CLONEf_COPY_STACKS) {
13099 /* next allocation will be PL_tmps_stack[PL_tmps_ix+1] */
13100 PL_tmps_ix = proto_perl->Itmps_ix;
13101 PL_tmps_max = proto_perl->Itmps_max;
13102 PL_tmps_floor = proto_perl->Itmps_floor;
13104 /* next push_scope()/ENTER sets PL_scopestack[PL_scopestack_ix]
13105 * NOTE: unlike the others! */
13106 PL_scopestack_ix = proto_perl->Iscopestack_ix;
13107 PL_scopestack_max = proto_perl->Iscopestack_max;
13109 /* next SSPUSHFOO() sets PL_savestack[PL_savestack_ix]
13110 * NOTE: unlike the others! */
13111 PL_savestack_ix = proto_perl->Isavestack_ix;
13112 PL_savestack_max = proto_perl->Isavestack_max;
13115 PL_start_env = proto_perl->Istart_env; /* XXXXXX */
13116 PL_top_env = &PL_start_env;
13118 PL_op = proto_perl->Iop;
13121 PL_Xpv = (XPV*)NULL;
13122 my_perl->Ina = proto_perl->Ina;
13124 PL_statbuf = proto_perl->Istatbuf;
13125 PL_statcache = proto_perl->Istatcache;
13128 PL_timesbuf = proto_perl->Itimesbuf;
13131 PL_tainted = proto_perl->Itainted;
13132 PL_curpm = proto_perl->Icurpm; /* XXX No PMOP ref count */
13134 PL_chopset = proto_perl->Ichopset; /* XXX never deallocated */
13136 PL_restartjmpenv = proto_perl->Irestartjmpenv;
13137 PL_restartop = proto_perl->Irestartop;
13138 PL_in_eval = proto_perl->Iin_eval;
13139 PL_delaymagic = proto_perl->Idelaymagic;
13140 PL_phase = proto_perl->Iphase;
13141 PL_localizing = proto_perl->Ilocalizing;
13143 PL_hv_fetch_ent_mh = NULL;
13144 PL_modcount = proto_perl->Imodcount;
13145 PL_lastgotoprobe = NULL;
13146 PL_dumpindent = proto_perl->Idumpindent;
13148 PL_efloatbuf = NULL; /* reinits on demand */
13149 PL_efloatsize = 0; /* reinits on demand */
13153 PL_regdummy = proto_perl->Iregdummy;
13154 PL_colorset = 0; /* reinits PL_colors[] */
13155 /*PL_colors[6] = {0,0,0,0,0,0};*/
13157 /* Pluggable optimizer */
13158 PL_peepp = proto_perl->Ipeepp;
13159 PL_rpeepp = proto_perl->Irpeepp;
13160 /* op_free() hook */
13161 PL_opfreehook = proto_perl->Iopfreehook;
13163 #ifdef USE_REENTRANT_API
13164 /* XXX: things like -Dm will segfault here in perlio, but doing
13165 * PERL_SET_CONTEXT(proto_perl);
13166 * breaks too many other things
13168 Perl_reentrant_init(aTHX);
13171 /* create SV map for pointer relocation */
13172 PL_ptr_table = ptr_table_new();
13174 /* initialize these special pointers as early as possible */
13175 ptr_table_store(PL_ptr_table, &proto_perl->Isv_undef, &PL_sv_undef);
13177 SvANY(&PL_sv_no) = new_XPVNV();
13178 SvPV_set(&PL_sv_no, savepvn(PL_No, 0));
13179 SvCUR_set(&PL_sv_no, 0);
13180 SvLEN_set(&PL_sv_no, 1);
13181 SvIV_set(&PL_sv_no, 0);
13182 SvNV_set(&PL_sv_no, 0);
13183 ptr_table_store(PL_ptr_table, &proto_perl->Isv_no, &PL_sv_no);
13185 SvPV_set(&PL_sv_yes, savepvn(PL_Yes, 1));
13186 SvCUR_set(&PL_sv_yes, 1);
13187 SvLEN_set(&PL_sv_yes, 2);
13188 SvIV_set(&PL_sv_yes, 1);
13189 SvNV_set(&PL_sv_yes, 1);
13190 ptr_table_store(PL_ptr_table, &proto_perl->Isv_yes, &PL_sv_yes);
13192 /* create (a non-shared!) shared string table */
13193 PL_strtab = newHV();
13194 HvSHAREKEYS_off(PL_strtab);
13195 hv_ksplit(PL_strtab, HvTOTALKEYS(proto_perl->Istrtab));
13196 ptr_table_store(PL_ptr_table, proto_perl->Istrtab, PL_strtab);
13198 /* These two PVs will be free'd special way so must set them same way op.c does */
13199 PL_compiling.cop_stashpv = savesharedpv(PL_compiling.cop_stashpv);
13200 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_stashpv, PL_compiling.cop_stashpv);
13202 PL_compiling.cop_file = savesharedpv(PL_compiling.cop_file);
13203 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_file, PL_compiling.cop_file);
13205 ptr_table_store(PL_ptr_table, &proto_perl->Icompiling, &PL_compiling);
13206 PL_compiling.cop_warnings = DUP_WARNINGS(PL_compiling.cop_warnings);
13207 CopHINTHASH_set(&PL_compiling, cophh_copy(CopHINTHASH_get(&PL_compiling)));
13208 PL_curcop = (COP*)any_dup(proto_perl->Icurcop, proto_perl);
13210 param->stashes = newAV(); /* Setup array of objects to call clone on */
13211 /* This makes no difference to the implementation, as it always pushes
13212 and shifts pointers to other SVs without changing their reference
13213 count, with the array becoming empty before it is freed. However, it
13214 makes it conceptually clear what is going on, and will avoid some
13215 work inside av.c, filling slots between AvFILL() and AvMAX() with
13216 &PL_sv_undef, and SvREFCNT_dec()ing those. */
13217 AvREAL_off(param->stashes);
13219 if (!(flags & CLONEf_COPY_STACKS)) {
13220 param->unreferenced = newAV();
13223 #ifdef PERLIO_LAYERS
13224 /* Clone PerlIO tables as soon as we can handle general xx_dup() */
13225 PerlIO_clone(aTHX_ proto_perl, param);
13228 PL_envgv = gv_dup(proto_perl->Ienvgv, param);
13229 PL_incgv = gv_dup(proto_perl->Iincgv, param);
13230 PL_hintgv = gv_dup(proto_perl->Ihintgv, param);
13231 PL_origfilename = SAVEPV(proto_perl->Iorigfilename);
13232 PL_diehook = sv_dup_inc(proto_perl->Idiehook, param);
13233 PL_warnhook = sv_dup_inc(proto_perl->Iwarnhook, param);
13236 PL_patchlevel = sv_dup_inc(proto_perl->Ipatchlevel, param);
13237 PL_apiversion = sv_dup_inc(proto_perl->Iapiversion, param);
13238 PL_inplace = SAVEPV(proto_perl->Iinplace);
13239 PL_e_script = sv_dup_inc(proto_perl->Ie_script, param);
13241 /* magical thingies */
13242 PL_formfeed = sv_dup(proto_perl->Iformfeed, param);
13244 PL_encoding = sv_dup(proto_perl->Iencoding, param);
13246 sv_setpvs(PERL_DEBUG_PAD(0), ""); /* For regex debugging. */
13247 sv_setpvs(PERL_DEBUG_PAD(1), ""); /* ext/re needs these */
13248 sv_setpvs(PERL_DEBUG_PAD(2), ""); /* even without DEBUGGING. */
13251 /* Clone the regex array */
13252 /* ORANGE FIXME for plugins, probably in the SV dup code.
13253 newSViv(PTR2IV(CALLREGDUPE(
13254 INT2PTR(REGEXP *, SvIVX(regex)), param))))
13256 PL_regex_padav = av_dup_inc(proto_perl->Iregex_padav, param);
13257 PL_regex_pad = AvARRAY(PL_regex_padav);
13259 /* shortcuts to various I/O objects */
13260 PL_ofsgv = gv_dup_inc(proto_perl->Iofsgv, param);
13261 PL_stdingv = gv_dup(proto_perl->Istdingv, param);
13262 PL_stderrgv = gv_dup(proto_perl->Istderrgv, param);
13263 PL_defgv = gv_dup(proto_perl->Idefgv, param);
13264 PL_argvgv = gv_dup(proto_perl->Iargvgv, param);
13265 PL_argvoutgv = gv_dup(proto_perl->Iargvoutgv, param);
13266 PL_argvout_stack = av_dup_inc(proto_perl->Iargvout_stack, param);
13268 /* shortcuts to regexp stuff */
13269 PL_replgv = gv_dup(proto_perl->Ireplgv, param);
13271 /* shortcuts to misc objects */
13272 PL_errgv = gv_dup(proto_perl->Ierrgv, param);
13274 /* shortcuts to debugging objects */
13275 PL_DBgv = gv_dup(proto_perl->IDBgv, param);
13276 PL_DBline = gv_dup(proto_perl->IDBline, param);
13277 PL_DBsub = gv_dup(proto_perl->IDBsub, param);
13278 PL_DBsingle = sv_dup(proto_perl->IDBsingle, param);
13279 PL_DBtrace = sv_dup(proto_perl->IDBtrace, param);
13280 PL_DBsignal = sv_dup(proto_perl->IDBsignal, param);
13282 /* symbol tables */
13283 PL_defstash = hv_dup_inc(proto_perl->Idefstash, param);
13284 PL_curstash = hv_dup_inc(proto_perl->Icurstash, param);
13285 PL_debstash = hv_dup(proto_perl->Idebstash, param);
13286 PL_globalstash = hv_dup(proto_perl->Iglobalstash, param);
13287 PL_curstname = sv_dup_inc(proto_perl->Icurstname, param);
13289 PL_beginav = av_dup_inc(proto_perl->Ibeginav, param);
13290 PL_beginav_save = av_dup_inc(proto_perl->Ibeginav_save, param);
13291 PL_checkav_save = av_dup_inc(proto_perl->Icheckav_save, param);
13292 PL_unitcheckav = av_dup_inc(proto_perl->Iunitcheckav, param);
13293 PL_unitcheckav_save = av_dup_inc(proto_perl->Iunitcheckav_save, param);
13294 PL_endav = av_dup_inc(proto_perl->Iendav, param);
13295 PL_checkav = av_dup_inc(proto_perl->Icheckav, param);
13296 PL_initav = av_dup_inc(proto_perl->Iinitav, param);
13298 PL_isarev = hv_dup_inc(proto_perl->Iisarev, param);
13300 /* subprocess state */
13301 PL_fdpid = av_dup_inc(proto_perl->Ifdpid, param);
13303 if (proto_perl->Iop_mask)
13304 PL_op_mask = SAVEPVN(proto_perl->Iop_mask, PL_maxo);
13307 /* PL_asserting = proto_perl->Iasserting; */
13309 /* current interpreter roots */
13310 PL_main_cv = cv_dup_inc(proto_perl->Imain_cv, param);
13312 PL_main_root = OpREFCNT_inc(proto_perl->Imain_root);
13315 /* runtime control stuff */
13316 PL_curcopdb = (COP*)any_dup(proto_perl->Icurcopdb, proto_perl);
13318 PL_preambleav = av_dup_inc(proto_perl->Ipreambleav, param);
13320 PL_ors_sv = sv_dup_inc(proto_perl->Iors_sv, param);
13322 /* interpreter atexit processing */
13323 PL_exitlistlen = proto_perl->Iexitlistlen;
13324 if (PL_exitlistlen) {
13325 Newx(PL_exitlist, PL_exitlistlen, PerlExitListEntry);
13326 Copy(proto_perl->Iexitlist, PL_exitlist, PL_exitlistlen, PerlExitListEntry);
13329 PL_exitlist = (PerlExitListEntry*)NULL;
13331 PL_my_cxt_size = proto_perl->Imy_cxt_size;
13332 if (PL_my_cxt_size) {
13333 Newx(PL_my_cxt_list, PL_my_cxt_size, void *);
13334 Copy(proto_perl->Imy_cxt_list, PL_my_cxt_list, PL_my_cxt_size, void *);
13335 #ifdef PERL_GLOBAL_STRUCT_PRIVATE
13336 Newx(PL_my_cxt_keys, PL_my_cxt_size, const char *);
13337 Copy(proto_perl->Imy_cxt_keys, PL_my_cxt_keys, PL_my_cxt_size, char *);
13341 PL_my_cxt_list = (void**)NULL;
13342 #ifdef PERL_GLOBAL_STRUCT_PRIVATE
13343 PL_my_cxt_keys = (const char**)NULL;
13346 PL_modglobal = hv_dup_inc(proto_perl->Imodglobal, param);
13347 PL_custom_op_names = hv_dup_inc(proto_perl->Icustom_op_names,param);
13348 PL_custom_op_descs = hv_dup_inc(proto_perl->Icustom_op_descs,param);
13349 PL_custom_ops = hv_dup_inc(proto_perl->Icustom_ops, param);
13351 PL_compcv = cv_dup(proto_perl->Icompcv, param);
13353 PAD_CLONE_VARS(proto_perl, param);
13355 #ifdef HAVE_INTERP_INTERN
13356 sys_intern_dup(&proto_perl->Isys_intern, &PL_sys_intern);
13359 PL_DBcv = cv_dup(proto_perl->IDBcv, param);
13361 #ifdef PERL_USES_PL_PIDSTATUS
13362 PL_pidstatus = newHV(); /* XXX flag for cloning? */
13364 PL_osname = SAVEPV(proto_perl->Iosname);
13365 PL_parser = parser_dup(proto_perl->Iparser, param);
13367 /* XXX this only works if the saved cop has already been cloned */
13368 if (proto_perl->Iparser) {
13369 PL_parser->saved_curcop = (COP*)any_dup(
13370 proto_perl->Iparser->saved_curcop,
13374 PL_subname = sv_dup_inc(proto_perl->Isubname, param);
13376 #ifdef USE_LOCALE_COLLATE
13377 PL_collation_name = SAVEPV(proto_perl->Icollation_name);
13378 #endif /* USE_LOCALE_COLLATE */
13380 #ifdef USE_LOCALE_NUMERIC
13381 PL_numeric_name = SAVEPV(proto_perl->Inumeric_name);
13382 PL_numeric_radix_sv = sv_dup_inc(proto_perl->Inumeric_radix_sv, param);
13383 #endif /* !USE_LOCALE_NUMERIC */
13385 /* Unicode inversion lists */
13386 PL_ASCII = sv_dup_inc(proto_perl->IASCII, param);
13387 PL_Latin1 = sv_dup_inc(proto_perl->ILatin1, param);
13389 PL_PerlSpace = sv_dup_inc(proto_perl->IPerlSpace, param);
13390 PL_XPerlSpace = sv_dup_inc(proto_perl->IXPerlSpace, param);
13392 PL_L1PosixAlnum = sv_dup_inc(proto_perl->IL1PosixAlnum, param);
13393 PL_PosixAlnum = sv_dup_inc(proto_perl->IPosixAlnum, param);
13395 PL_L1PosixAlpha = sv_dup_inc(proto_perl->IL1PosixAlpha, param);
13396 PL_PosixAlpha = sv_dup_inc(proto_perl->IPosixAlpha, param);
13398 PL_PosixBlank = sv_dup_inc(proto_perl->IPosixBlank, param);
13399 PL_XPosixBlank = sv_dup_inc(proto_perl->IXPosixBlank, param);
13401 PL_PosixCntrl = sv_dup_inc(proto_perl->IPosixCntrl, param);
13402 PL_XPosixCntrl = sv_dup_inc(proto_perl->IXPosixCntrl, param);
13404 PL_PosixDigit = sv_dup_inc(proto_perl->IPosixDigit, param);
13406 PL_L1PosixGraph = sv_dup_inc(proto_perl->IL1PosixGraph, param);
13407 PL_PosixGraph = sv_dup_inc(proto_perl->IPosixGraph, param);
13409 PL_HorizSpace = sv_dup_inc(proto_perl->IHorizSpace, param);
13411 PL_L1PosixLower = sv_dup_inc(proto_perl->IL1PosixLower, param);
13412 PL_PosixLower = sv_dup_inc(proto_perl->IPosixLower, param);
13414 PL_L1PosixPrint = sv_dup_inc(proto_perl->IL1PosixPrint, param);
13415 PL_PosixPrint = sv_dup_inc(proto_perl->IPosixPrint, param);
13417 PL_L1PosixPunct = sv_dup_inc(proto_perl->IL1PosixPunct, param);
13418 PL_PosixPunct = sv_dup_inc(proto_perl->IPosixPunct, param);
13420 PL_PosixSpace = sv_dup_inc(proto_perl->IPosixSpace, param);
13421 PL_XPosixSpace = sv_dup_inc(proto_perl->IXPosixSpace, param);
13423 PL_L1PosixUpper = sv_dup_inc(proto_perl->IL1PosixUpper, param);
13424 PL_PosixUpper = sv_dup_inc(proto_perl->IPosixUpper, param);
13426 PL_L1PosixWord = sv_dup_inc(proto_perl->IL1PosixWord, param);
13427 PL_PosixWord = sv_dup_inc(proto_perl->IPosixWord, param);
13429 PL_PosixXDigit = sv_dup_inc(proto_perl->IPosixXDigit, param);
13430 PL_XPosixXDigit = sv_dup_inc(proto_perl->IXPosixXDigit, param);
13432 PL_VertSpace = sv_dup_inc(proto_perl->IVertSpace, param);
13434 /* utf8 character class swashes */
13435 PL_utf8_alnum = sv_dup_inc(proto_perl->Iutf8_alnum, param);
13436 PL_utf8_alpha = sv_dup_inc(proto_perl->Iutf8_alpha, param);
13437 PL_utf8_space = sv_dup_inc(proto_perl->Iutf8_space, param);
13438 PL_utf8_graph = sv_dup_inc(proto_perl->Iutf8_graph, param);
13439 PL_utf8_digit = sv_dup_inc(proto_perl->Iutf8_digit, param);
13440 PL_utf8_upper = sv_dup_inc(proto_perl->Iutf8_upper, param);
13441 PL_utf8_lower = sv_dup_inc(proto_perl->Iutf8_lower, param);
13442 PL_utf8_print = sv_dup_inc(proto_perl->Iutf8_print, param);
13443 PL_utf8_punct = sv_dup_inc(proto_perl->Iutf8_punct, param);
13444 PL_utf8_xdigit = sv_dup_inc(proto_perl->Iutf8_xdigit, param);
13445 PL_utf8_mark = sv_dup_inc(proto_perl->Iutf8_mark, param);
13446 PL_utf8_X_begin = sv_dup_inc(proto_perl->Iutf8_X_begin, param);
13447 PL_utf8_X_extend = sv_dup_inc(proto_perl->Iutf8_X_extend, param);
13448 PL_utf8_X_prepend = sv_dup_inc(proto_perl->Iutf8_X_prepend, param);
13449 PL_utf8_X_non_hangul = sv_dup_inc(proto_perl->Iutf8_X_non_hangul, param);
13450 PL_utf8_X_L = sv_dup_inc(proto_perl->Iutf8_X_L, param);
13451 PL_utf8_X_LV = sv_dup_inc(proto_perl->Iutf8_X_LV, param);
13452 PL_utf8_X_LVT = sv_dup_inc(proto_perl->Iutf8_X_LVT, param);
13453 PL_utf8_X_T = sv_dup_inc(proto_perl->Iutf8_X_T, param);
13454 PL_utf8_X_V = sv_dup_inc(proto_perl->Iutf8_X_V, param);
13455 PL_utf8_X_LV_LVT_V = sv_dup_inc(proto_perl->Iutf8_X_LV_LVT_V, param);
13456 PL_utf8_toupper = sv_dup_inc(proto_perl->Iutf8_toupper, param);
13457 PL_utf8_totitle = sv_dup_inc(proto_perl->Iutf8_totitle, param);
13458 PL_utf8_tolower = sv_dup_inc(proto_perl->Iutf8_tolower, param);
13459 PL_utf8_tofold = sv_dup_inc(proto_perl->Iutf8_tofold, param);
13460 PL_utf8_idstart = sv_dup_inc(proto_perl->Iutf8_idstart, param);
13461 PL_utf8_xidstart = sv_dup_inc(proto_perl->Iutf8_xidstart, param);
13462 PL_utf8_perl_idstart = sv_dup_inc(proto_perl->Iutf8_perl_idstart, param);
13463 PL_utf8_idcont = sv_dup_inc(proto_perl->Iutf8_idcont, param);
13464 PL_utf8_xidcont = sv_dup_inc(proto_perl->Iutf8_xidcont, param);
13465 PL_utf8_foldable = sv_dup_inc(proto_perl->Iutf8_foldable, param);
13466 PL_ASCII = sv_dup_inc(proto_perl->IASCII, param);
13467 PL_AboveLatin1 = sv_dup_inc(proto_perl->IAboveLatin1, param);
13468 PL_Latin1 = sv_dup_inc(proto_perl->ILatin1, param);
13471 if (proto_perl->Ipsig_pend) {
13472 Newxz(PL_psig_pend, SIG_SIZE, int);
13475 PL_psig_pend = (int*)NULL;
13478 if (proto_perl->Ipsig_name) {
13479 Newx(PL_psig_name, 2 * SIG_SIZE, SV*);
13480 sv_dup_inc_multiple(proto_perl->Ipsig_name, PL_psig_name, 2 * SIG_SIZE,
13482 PL_psig_ptr = PL_psig_name + SIG_SIZE;
13485 PL_psig_ptr = (SV**)NULL;
13486 PL_psig_name = (SV**)NULL;
13489 if (flags & CLONEf_COPY_STACKS) {
13490 Newx(PL_tmps_stack, PL_tmps_max, SV*);
13491 sv_dup_inc_multiple(proto_perl->Itmps_stack, PL_tmps_stack,
13492 PL_tmps_ix+1, param);
13494 /* next PUSHMARK() sets *(PL_markstack_ptr+1) */
13495 i = proto_perl->Imarkstack_max - proto_perl->Imarkstack;
13496 Newxz(PL_markstack, i, I32);
13497 PL_markstack_max = PL_markstack + (proto_perl->Imarkstack_max
13498 - proto_perl->Imarkstack);
13499 PL_markstack_ptr = PL_markstack + (proto_perl->Imarkstack_ptr
13500 - proto_perl->Imarkstack);
13501 Copy(proto_perl->Imarkstack, PL_markstack,
13502 PL_markstack_ptr - PL_markstack + 1, I32);
13504 /* next push_scope()/ENTER sets PL_scopestack[PL_scopestack_ix]
13505 * NOTE: unlike the others! */
13506 Newxz(PL_scopestack, PL_scopestack_max, I32);
13507 Copy(proto_perl->Iscopestack, PL_scopestack, PL_scopestack_ix, I32);
13510 Newxz(PL_scopestack_name, PL_scopestack_max, const char *);
13511 Copy(proto_perl->Iscopestack_name, PL_scopestack_name, PL_scopestack_ix, const char *);
13513 /* NOTE: si_dup() looks at PL_markstack */
13514 PL_curstackinfo = si_dup(proto_perl->Icurstackinfo, param);
13516 /* PL_curstack = PL_curstackinfo->si_stack; */
13517 PL_curstack = av_dup(proto_perl->Icurstack, param);
13518 PL_mainstack = av_dup(proto_perl->Imainstack, param);
13520 /* next PUSHs() etc. set *(PL_stack_sp+1) */
13521 PL_stack_base = AvARRAY(PL_curstack);
13522 PL_stack_sp = PL_stack_base + (proto_perl->Istack_sp
13523 - proto_perl->Istack_base);
13524 PL_stack_max = PL_stack_base + AvMAX(PL_curstack);
13526 /*Newxz(PL_savestack, PL_savestack_max, ANY);*/
13527 PL_savestack = ss_dup(proto_perl, param);
13531 ENTER; /* perl_destruct() wants to LEAVE; */
13534 PL_statgv = gv_dup(proto_perl->Istatgv, param);
13535 PL_statname = sv_dup_inc(proto_perl->Istatname, param);
13537 PL_rs = sv_dup_inc(proto_perl->Irs, param);
13538 PL_last_in_gv = gv_dup(proto_perl->Ilast_in_gv, param);
13539 PL_defoutgv = gv_dup_inc(proto_perl->Idefoutgv, param);
13540 PL_toptarget = sv_dup_inc(proto_perl->Itoptarget, param);
13541 PL_bodytarget = sv_dup_inc(proto_perl->Ibodytarget, param);
13542 PL_formtarget = sv_dup(proto_perl->Iformtarget, param);
13544 PL_errors = sv_dup_inc(proto_perl->Ierrors, param);
13546 PL_sortcop = (OP*)any_dup(proto_perl->Isortcop, proto_perl);
13547 PL_sortstash = hv_dup(proto_perl->Isortstash, param);
13548 PL_firstgv = gv_dup(proto_perl->Ifirstgv, param);
13549 PL_secondgv = gv_dup(proto_perl->Isecondgv, param);
13551 PL_stashcache = newHV();
13553 PL_watchaddr = (char **) ptr_table_fetch(PL_ptr_table,
13554 proto_perl->Iwatchaddr);
13555 PL_watchok = PL_watchaddr ? * PL_watchaddr : NULL;
13556 if (PL_debug && PL_watchaddr) {
13557 PerlIO_printf(Perl_debug_log,
13558 "WATCHING: %"UVxf" cloned as %"UVxf" with value %"UVxf"\n",
13559 PTR2UV(proto_perl->Iwatchaddr), PTR2UV(PL_watchaddr),
13560 PTR2UV(PL_watchok));
13563 PL_registered_mros = hv_dup_inc(proto_perl->Iregistered_mros, param);
13564 PL_blockhooks = av_dup_inc(proto_perl->Iblockhooks, param);
13565 PL_utf8_foldclosures = hv_dup_inc(proto_perl->Iutf8_foldclosures, param);
13567 /* Call the ->CLONE method, if it exists, for each of the stashes
13568 identified by sv_dup() above.
13570 while(av_len(param->stashes) != -1) {
13571 HV* const stash = MUTABLE_HV(av_shift(param->stashes));
13572 GV* const cloner = gv_fetchmethod_autoload(stash, "CLONE", 0);
13573 if (cloner && GvCV(cloner)) {
13578 mXPUSHs(newSVhek(HvNAME_HEK(stash)));
13580 call_sv(MUTABLE_SV(GvCV(cloner)), G_DISCARD);
13586 if (!(flags & CLONEf_KEEP_PTR_TABLE)) {
13587 ptr_table_free(PL_ptr_table);
13588 PL_ptr_table = NULL;
13591 if (!(flags & CLONEf_COPY_STACKS)) {
13592 unreferenced_to_tmp_stack(param->unreferenced);
13595 SvREFCNT_dec(param->stashes);
13597 /* orphaned? eg threads->new inside BEGIN or use */
13598 if (PL_compcv && ! SvREFCNT(PL_compcv)) {
13599 SvREFCNT_inc_simple_void(PL_compcv);
13600 SAVEFREESV(PL_compcv);
13607 S_unreferenced_to_tmp_stack(pTHX_ AV *const unreferenced)
13609 PERL_ARGS_ASSERT_UNREFERENCED_TO_TMP_STACK;
13611 if (AvFILLp(unreferenced) > -1) {
13612 SV **svp = AvARRAY(unreferenced);
13613 SV **const last = svp + AvFILLp(unreferenced);
13617 if (SvREFCNT(*svp) == 1)
13619 } while (++svp <= last);
13621 EXTEND_MORTAL(count);
13622 svp = AvARRAY(unreferenced);
13625 if (SvREFCNT(*svp) == 1) {
13626 /* Our reference is the only one to this SV. This means that
13627 in this thread, the scalar effectively has a 0 reference.
13628 That doesn't work (cleanup never happens), so donate our
13629 reference to it onto the save stack. */
13630 PL_tmps_stack[++PL_tmps_ix] = *svp;
13632 /* As an optimisation, because we are already walking the
13633 entire array, instead of above doing either
13634 SvREFCNT_inc(*svp) or *svp = &PL_sv_undef, we can instead
13635 release our reference to the scalar, so that at the end of
13636 the array owns zero references to the scalars it happens to
13637 point to. We are effectively converting the array from
13638 AvREAL() on to AvREAL() off. This saves the av_clear()
13639 (triggered by the SvREFCNT_dec(unreferenced) below) from
13640 walking the array a second time. */
13641 SvREFCNT_dec(*svp);
13644 } while (++svp <= last);
13645 AvREAL_off(unreferenced);
13647 SvREFCNT_dec(unreferenced);
13651 Perl_clone_params_del(CLONE_PARAMS *param)
13653 /* This seemingly funky ordering keeps the build with PERL_GLOBAL_STRUCT
13655 PerlInterpreter *const to = param->new_perl;
13657 PerlInterpreter *const was = PERL_GET_THX;
13659 PERL_ARGS_ASSERT_CLONE_PARAMS_DEL;
13665 SvREFCNT_dec(param->stashes);
13666 if (param->unreferenced)
13667 unreferenced_to_tmp_stack(param->unreferenced);
13677 Perl_clone_params_new(PerlInterpreter *const from, PerlInterpreter *const to)
13680 /* Need to play this game, as newAV() can call safesysmalloc(), and that
13681 does a dTHX; to get the context from thread local storage.
13682 FIXME - under PERL_CORE Newx(), Safefree() and friends should expand to
13683 a version that passes in my_perl. */
13684 PerlInterpreter *const was = PERL_GET_THX;
13685 CLONE_PARAMS *param;
13687 PERL_ARGS_ASSERT_CLONE_PARAMS_NEW;
13693 /* Given that we've set the context, we can do this unshared. */
13694 Newx(param, 1, CLONE_PARAMS);
13697 param->proto_perl = from;
13698 param->new_perl = to;
13699 param->stashes = (AV *)Perl_newSV_type(to, SVt_PVAV);
13700 AvREAL_off(param->stashes);
13701 param->unreferenced = (AV *)Perl_newSV_type(to, SVt_PVAV);
13709 #endif /* USE_ITHREADS */
13712 =head1 Unicode Support
13714 =for apidoc sv_recode_to_utf8
13716 The encoding is assumed to be an Encode object, on entry the PV
13717 of the sv is assumed to be octets in that encoding, and the sv
13718 will be converted into Unicode (and UTF-8).
13720 If the sv already is UTF-8 (or if it is not POK), or if the encoding
13721 is not a reference, nothing is done to the sv. If the encoding is not
13722 an C<Encode::XS> Encoding object, bad things will happen.
13723 (See F<lib/encoding.pm> and L<Encode>.)
13725 The PV of the sv is returned.
13730 Perl_sv_recode_to_utf8(pTHX_ SV *sv, SV *encoding)
13734 PERL_ARGS_ASSERT_SV_RECODE_TO_UTF8;
13736 if (SvPOK(sv) && !SvUTF8(sv) && !IN_BYTES && SvROK(encoding)) {
13750 Passing sv_yes is wrong - it needs to be or'ed set of constants
13751 for Encode::XS, while UTf-8 decode (currently) assumes a true value means
13752 remove converted chars from source.
13754 Both will default the value - let them.
13756 XPUSHs(&PL_sv_yes);
13759 call_method("decode", G_SCALAR);
13763 s = SvPV_const(uni, len);
13764 if (s != SvPVX_const(sv)) {
13765 SvGROW(sv, len + 1);
13766 Move(s, SvPVX(sv), len + 1, char);
13767 SvCUR_set(sv, len);
13771 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
13772 /* clear pos and any utf8 cache */
13773 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
13776 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
13777 magic_setutf8(sv,mg); /* clear UTF8 cache */
13782 return SvPOKp(sv) ? SvPVX(sv) : NULL;
13786 =for apidoc sv_cat_decode
13788 The encoding is assumed to be an Encode object, the PV of the ssv is
13789 assumed to be octets in that encoding and decoding the input starts
13790 from the position which (PV + *offset) pointed to. The dsv will be
13791 concatenated the decoded UTF-8 string from ssv. Decoding will terminate
13792 when the string tstr appears in decoding output or the input ends on
13793 the PV of the ssv. The value which the offset points will be modified
13794 to the last input position on the ssv.
13796 Returns TRUE if the terminator was found, else returns FALSE.
13801 Perl_sv_cat_decode(pTHX_ SV *dsv, SV *encoding,
13802 SV *ssv, int *offset, char *tstr, int tlen)
13807 PERL_ARGS_ASSERT_SV_CAT_DECODE;
13809 if (SvPOK(ssv) && SvPOK(dsv) && SvROK(encoding) && offset) {
13820 offsv = newSViv(*offset);
13822 mXPUSHp(tstr, tlen);
13824 call_method("cat_decode", G_SCALAR);
13826 ret = SvTRUE(TOPs);
13827 *offset = SvIV(offsv);
13833 Perl_croak(aTHX_ "Invalid argument to sv_cat_decode");
13838 /* ---------------------------------------------------------------------
13840 * support functions for report_uninit()
13843 /* the maxiumum size of array or hash where we will scan looking
13844 * for the undefined element that triggered the warning */
13846 #define FUV_MAX_SEARCH_SIZE 1000
13848 /* Look for an entry in the hash whose value has the same SV as val;
13849 * If so, return a mortal copy of the key. */
13852 S_find_hash_subscript(pTHX_ const HV *const hv, const SV *const val)
13855 register HE **array;
13858 PERL_ARGS_ASSERT_FIND_HASH_SUBSCRIPT;
13860 if (!hv || SvMAGICAL(hv) || !HvARRAY(hv) ||
13861 (HvTOTALKEYS(hv) > FUV_MAX_SEARCH_SIZE))
13864 array = HvARRAY(hv);
13866 for (i=HvMAX(hv); i>0; i--) {
13867 register HE *entry;
13868 for (entry = array[i]; entry; entry = HeNEXT(entry)) {
13869 if (HeVAL(entry) != val)
13871 if ( HeVAL(entry) == &PL_sv_undef ||
13872 HeVAL(entry) == &PL_sv_placeholder)
13876 if (HeKLEN(entry) == HEf_SVKEY)
13877 return sv_mortalcopy(HeKEY_sv(entry));
13878 return sv_2mortal(newSVhek(HeKEY_hek(entry)));
13884 /* Look for an entry in the array whose value has the same SV as val;
13885 * If so, return the index, otherwise return -1. */
13888 S_find_array_subscript(pTHX_ const AV *const av, const SV *const val)
13892 PERL_ARGS_ASSERT_FIND_ARRAY_SUBSCRIPT;
13894 if (!av || SvMAGICAL(av) || !AvARRAY(av) ||
13895 (AvFILLp(av) > FUV_MAX_SEARCH_SIZE))
13898 if (val != &PL_sv_undef) {
13899 SV ** const svp = AvARRAY(av);
13902 for (i=AvFILLp(av); i>=0; i--)
13909 /* S_varname(): return the name of a variable, optionally with a subscript.
13910 * If gv is non-zero, use the name of that global, along with gvtype (one
13911 * of "$", "@", "%"); otherwise use the name of the lexical at pad offset
13912 * targ. Depending on the value of the subscript_type flag, return:
13915 #define FUV_SUBSCRIPT_NONE 1 /* "@foo" */
13916 #define FUV_SUBSCRIPT_ARRAY 2 /* "$foo[aindex]" */
13917 #define FUV_SUBSCRIPT_HASH 3 /* "$foo{keyname}" */
13918 #define FUV_SUBSCRIPT_WITHIN 4 /* "within @foo" */
13921 Perl_varname(pTHX_ const GV *const gv, const char gvtype, PADOFFSET targ,
13922 const SV *const keyname, I32 aindex, int subscript_type)
13925 SV * const name = sv_newmortal();
13926 if (gv && isGV(gv)) {
13928 buffer[0] = gvtype;
13931 /* as gv_fullname4(), but add literal '^' for $^FOO names */
13933 gv_fullname4(name, gv, buffer, 0);
13935 if ((unsigned int)SvPVX(name)[1] <= 26) {
13937 buffer[1] = SvPVX(name)[1] + 'A' - 1;
13939 /* Swap the 1 unprintable control character for the 2 byte pretty
13940 version - ie substr($name, 1, 1) = $buffer; */
13941 sv_insert(name, 1, 1, buffer, 2);
13945 CV * const cv = gv ? (CV *)gv : find_runcv(NULL);
13949 assert(!cv || SvTYPE(cv) == SVt_PVCV);
13951 if (!cv || !CvPADLIST(cv))
13953 av = MUTABLE_AV((*av_fetch(CvPADLIST(cv), 0, FALSE)));
13954 sv = *av_fetch(av, targ, FALSE);
13955 sv_setsv(name, sv);
13958 if (subscript_type == FUV_SUBSCRIPT_HASH) {
13959 SV * const sv = newSV(0);
13960 *SvPVX(name) = '$';
13961 Perl_sv_catpvf(aTHX_ name, "{%s}",
13962 pv_pretty(sv, SvPVX_const(keyname), SvCUR(keyname), 32, NULL, NULL,
13963 PERL_PV_PRETTY_DUMP | PERL_PV_ESCAPE_UNI_DETECT ));
13966 else if (subscript_type == FUV_SUBSCRIPT_ARRAY) {
13967 *SvPVX(name) = '$';
13968 Perl_sv_catpvf(aTHX_ name, "[%"IVdf"]", (IV)aindex);
13970 else if (subscript_type == FUV_SUBSCRIPT_WITHIN) {
13971 /* We know that name has no magic, so can use 0 instead of SV_GMAGIC */
13972 Perl_sv_insert_flags(aTHX_ name, 0, 0, STR_WITH_LEN("within "), 0);
13980 =for apidoc find_uninit_var
13982 Find the name of the undefined variable (if any) that caused the operator
13983 to issue a "Use of uninitialized value" warning.
13984 If match is true, only return a name if its value matches uninit_sv.
13985 So roughly speaking, if a unary operator (such as OP_COS) generates a
13986 warning, then following the direct child of the op may yield an
13987 OP_PADSV or OP_GV that gives the name of the undefined variable. On the
13988 other hand, with OP_ADD there are two branches to follow, so we only print
13989 the variable name if we get an exact match.
13991 The name is returned as a mortal SV.
13993 Assumes that PL_op is the op that originally triggered the error, and that
13994 PL_comppad/PL_curpad points to the currently executing pad.
14000 S_find_uninit_var(pTHX_ const OP *const obase, const SV *const uninit_sv,
14006 const OP *o, *o2, *kid;
14008 if (!obase || (match && (!uninit_sv || uninit_sv == &PL_sv_undef ||
14009 uninit_sv == &PL_sv_placeholder)))
14012 switch (obase->op_type) {
14019 const bool pad = (obase->op_type == OP_PADAV || obase->op_type == OP_PADHV);
14020 const bool hash = (obase->op_type == OP_PADHV || obase->op_type == OP_RV2HV);
14023 int subscript_type = FUV_SUBSCRIPT_WITHIN;
14025 if (pad) { /* @lex, %lex */
14026 sv = PAD_SVl(obase->op_targ);
14030 if (cUNOPx(obase)->op_first->op_type == OP_GV) {
14031 /* @global, %global */
14032 gv = cGVOPx_gv(cUNOPx(obase)->op_first);
14035 sv = hash ? MUTABLE_SV(GvHV(gv)): MUTABLE_SV(GvAV(gv));
14037 else if (obase == PL_op) /* @{expr}, %{expr} */
14038 return find_uninit_var(cUNOPx(obase)->op_first,
14040 else /* @{expr}, %{expr} as a sub-expression */
14044 /* attempt to find a match within the aggregate */
14046 keysv = find_hash_subscript((const HV*)sv, uninit_sv);
14048 subscript_type = FUV_SUBSCRIPT_HASH;
14051 index = find_array_subscript((const AV *)sv, uninit_sv);
14053 subscript_type = FUV_SUBSCRIPT_ARRAY;
14056 if (match && subscript_type == FUV_SUBSCRIPT_WITHIN)
14059 return varname(gv, hash ? '%' : '@', obase->op_targ,
14060 keysv, index, subscript_type);
14064 if (cUNOPx(obase)->op_first->op_type == OP_GV) {
14066 gv = cGVOPx_gv(cUNOPx(obase)->op_first);
14067 if (!gv || !GvSTASH(gv))
14069 if (match && (GvSV(gv) != uninit_sv))
14071 return varname(gv, '$', 0, NULL, 0, FUV_SUBSCRIPT_NONE);
14074 return find_uninit_var(cUNOPx(obase)->op_first, uninit_sv, 1);
14077 if (match && PAD_SVl(obase->op_targ) != uninit_sv)
14079 return varname(NULL, '$', obase->op_targ,
14080 NULL, 0, FUV_SUBSCRIPT_NONE);
14083 gv = cGVOPx_gv(obase);
14084 if (!gv || (match && GvSV(gv) != uninit_sv) || !GvSTASH(gv))
14086 return varname(gv, '$', 0, NULL, 0, FUV_SUBSCRIPT_NONE);
14088 case OP_AELEMFAST_LEX:
14091 AV *av = MUTABLE_AV(PAD_SV(obase->op_targ));
14092 if (!av || SvRMAGICAL(av))
14094 svp = av_fetch(av, (I32)obase->op_private, FALSE);
14095 if (!svp || *svp != uninit_sv)
14098 return varname(NULL, '$', obase->op_targ,
14099 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
14102 gv = cGVOPx_gv(obase);
14107 AV *const av = GvAV(gv);
14108 if (!av || SvRMAGICAL(av))
14110 svp = av_fetch(av, (I32)obase->op_private, FALSE);
14111 if (!svp || *svp != uninit_sv)
14114 return varname(gv, '$', 0,
14115 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
14120 o = cUNOPx(obase)->op_first;
14121 if (!o || o->op_type != OP_NULL ||
14122 ! (o->op_targ == OP_AELEM || o->op_targ == OP_HELEM))
14124 return find_uninit_var(cBINOPo->op_last, uninit_sv, match);
14129 bool negate = FALSE;
14131 if (PL_op == obase)
14132 /* $a[uninit_expr] or $h{uninit_expr} */
14133 return find_uninit_var(cBINOPx(obase)->op_last, uninit_sv, match);
14136 o = cBINOPx(obase)->op_first;
14137 kid = cBINOPx(obase)->op_last;
14139 /* get the av or hv, and optionally the gv */
14141 if (o->op_type == OP_PADAV || o->op_type == OP_PADHV) {
14142 sv = PAD_SV(o->op_targ);
14144 else if ((o->op_type == OP_RV2AV || o->op_type == OP_RV2HV)
14145 && cUNOPo->op_first->op_type == OP_GV)
14147 gv = cGVOPx_gv(cUNOPo->op_first);
14151 == OP_RV2HV ? MUTABLE_SV(GvHV(gv)) : MUTABLE_SV(GvAV(gv));
14156 if (kid && kid->op_type == OP_NEGATE) {
14158 kid = cUNOPx(kid)->op_first;
14161 if (kid && kid->op_type == OP_CONST && SvOK(cSVOPx_sv(kid))) {
14162 /* index is constant */
14165 kidsv = sv_2mortal(newSVpvs("-"));
14166 sv_catsv(kidsv, cSVOPx_sv(kid));
14169 kidsv = cSVOPx_sv(kid);
14173 if (obase->op_type == OP_HELEM) {
14174 HE* he = hv_fetch_ent(MUTABLE_HV(sv), kidsv, 0, 0);
14175 if (!he || HeVAL(he) != uninit_sv)
14179 SV * const * const svp = av_fetch(MUTABLE_AV(sv),
14180 negate ? - SvIV(cSVOPx_sv(kid)) : SvIV(cSVOPx_sv(kid)),
14182 if (!svp || *svp != uninit_sv)
14186 if (obase->op_type == OP_HELEM)
14187 return varname(gv, '%', o->op_targ,
14188 kidsv, 0, FUV_SUBSCRIPT_HASH);
14190 return varname(gv, '@', o->op_targ, NULL,
14191 negate ? - SvIV(cSVOPx_sv(kid)) : SvIV(cSVOPx_sv(kid)),
14192 FUV_SUBSCRIPT_ARRAY);
14195 /* index is an expression;
14196 * attempt to find a match within the aggregate */
14197 if (obase->op_type == OP_HELEM) {
14198 SV * const keysv = find_hash_subscript((const HV*)sv, uninit_sv);
14200 return varname(gv, '%', o->op_targ,
14201 keysv, 0, FUV_SUBSCRIPT_HASH);
14205 = find_array_subscript((const AV *)sv, uninit_sv);
14207 return varname(gv, '@', o->op_targ,
14208 NULL, index, FUV_SUBSCRIPT_ARRAY);
14213 (o->op_type == OP_PADAV || o->op_type == OP_RV2AV)
14215 o->op_targ, NULL, 0, FUV_SUBSCRIPT_WITHIN);
14221 /* only examine RHS */
14222 return find_uninit_var(cBINOPx(obase)->op_first, uninit_sv, match);
14225 o = cUNOPx(obase)->op_first;
14226 if (o->op_type == OP_PUSHMARK)
14229 if (!o->op_sibling) {
14230 /* one-arg version of open is highly magical */
14232 if (o->op_type == OP_GV) { /* open FOO; */
14234 if (match && GvSV(gv) != uninit_sv)
14236 return varname(gv, '$', 0,
14237 NULL, 0, FUV_SUBSCRIPT_NONE);
14239 /* other possibilities not handled are:
14240 * open $x; or open my $x; should return '${*$x}'
14241 * open expr; should return '$'.expr ideally
14247 /* ops where $_ may be an implicit arg */
14252 if ( !(obase->op_flags & OPf_STACKED)) {
14253 if (uninit_sv == ((obase->op_private & OPpTARGET_MY)
14254 ? PAD_SVl(obase->op_targ)
14257 sv = sv_newmortal();
14258 sv_setpvs(sv, "$_");
14267 match = 1; /* print etc can return undef on defined args */
14268 /* skip filehandle as it can't produce 'undef' warning */
14269 o = cUNOPx(obase)->op_first;
14270 if ((obase->op_flags & OPf_STACKED) && o->op_type == OP_PUSHMARK)
14271 o = o->op_sibling->op_sibling;
14275 case OP_ENTEREVAL: /* could be eval $undef or $x='$undef'; eval $x */
14276 case OP_CUSTOM: /* XS or custom code could trigger random warnings */
14278 /* the following ops are capable of returning PL_sv_undef even for
14279 * defined arg(s) */
14298 case OP_GETPEERNAME:
14346 case OP_SMARTMATCH:
14355 /* XXX tmp hack: these two may call an XS sub, and currently
14356 XS subs don't have a SUB entry on the context stack, so CV and
14357 pad determination goes wrong, and BAD things happen. So, just
14358 don't try to determine the value under those circumstances.
14359 Need a better fix at dome point. DAPM 11/2007 */
14365 GV * const gv = gv_fetchpvs(".", GV_NOTQUAL, SVt_PV);
14366 if (gv && GvSV(gv) == uninit_sv)
14367 return newSVpvs_flags("$.", SVs_TEMP);
14372 /* def-ness of rval pos() is independent of the def-ness of its arg */
14373 if ( !(obase->op_flags & OPf_MOD))
14378 if (SvROK(PL_rs) && uninit_sv == SvRV(PL_rs))
14379 return newSVpvs_flags("${$/}", SVs_TEMP);
14384 if (!(obase->op_flags & OPf_KIDS))
14386 o = cUNOPx(obase)->op_first;
14392 /* This loop checks all the kid ops, skipping any that cannot pos-
14393 * sibly be responsible for the uninitialized value; i.e., defined
14394 * constants and ops that return nothing. If there is only one op
14395 * left that is not skipped, then we *know* it is responsible for
14396 * the uninitialized value. If there is more than one op left, we
14397 * have to look for an exact match in the while() loop below.
14400 for (kid=o; kid; kid = kid->op_sibling) {
14402 const OPCODE type = kid->op_type;
14403 if ( (type == OP_CONST && SvOK(cSVOPx_sv(kid)))
14404 || (type == OP_NULL && ! (kid->op_flags & OPf_KIDS))
14405 || (type == OP_PUSHMARK)
14409 if (o2) { /* more than one found */
14416 return find_uninit_var(o2, uninit_sv, match);
14418 /* scan all args */
14420 sv = find_uninit_var(o, uninit_sv, 1);
14432 =for apidoc report_uninit
14434 Print appropriate "Use of uninitialized variable" warning.
14440 Perl_report_uninit(pTHX_ const SV *uninit_sv)
14444 SV* varname = NULL;
14445 if (uninit_sv && PL_curpad) {
14446 varname = find_uninit_var(PL_op, uninit_sv,0);
14448 sv_insert(varname, 0, 0, " ", 1);
14450 /* diag_listed_as: Use of uninitialized value%s */
14451 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit_sv,
14452 SVfARG(varname ? varname : &PL_sv_no),
14453 " in ", OP_DESC(PL_op));
14456 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
14462 * c-indentation-style: bsd
14463 * c-basic-offset: 4
14464 * indent-tabs-mode: t
14467 * ex: set ts=8 sts=4 sw=4 noet: