3 * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, 2000,
4 * 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009 by Larry Wall
7 * You may distribute under the terms of either the GNU General Public
8 * License or the Artistic License, as specified in the README file.
13 * 'I wonder what the Entish is for "yes" and "no",' he thought.
16 * [p.480 of _The Lord of the Rings_, III/iv: "Treebeard"]
22 * This file contains the code that creates, manipulates and destroys
23 * scalar values (SVs). The other types (AV, HV, GV, etc.) reuse the
24 * structure of an SV, so their creation and destruction is handled
25 * here; higher-level functions are in av.c, hv.c, and so on. Opcode
26 * level functions (eg. substr, split, join) for each of the types are
36 # if __STDC_VERSION__ >= 199901L && !defined(VMS)
47 /* Missing proto on LynxOS */
48 char *gconvert(double, int, int, char *);
51 #ifdef PERL_UTF8_CACHE_ASSERT
52 /* if adding more checks watch out for the following tests:
53 * t/op/index.t t/op/length.t t/op/pat.t t/op/substr.t
54 * lib/utf8.t lib/Unicode/Collate/t/index.t
57 # define ASSERT_UTF8_CACHE(cache) \
58 STMT_START { if (cache) { assert((cache)[0] <= (cache)[1]); \
59 assert((cache)[2] <= (cache)[3]); \
60 assert((cache)[3] <= (cache)[1]);} \
63 # define ASSERT_UTF8_CACHE(cache) NOOP
66 #ifdef PERL_OLD_COPY_ON_WRITE
67 #define SV_COW_NEXT_SV(sv) INT2PTR(SV *,SvUVX(sv))
68 #define SV_COW_NEXT_SV_SET(current,next) SvUV_set(current, PTR2UV(next))
69 /* This is a pessimistic view. Scalar must be purely a read-write PV to copy-
73 /* ============================================================================
75 =head1 Allocation and deallocation of SVs.
77 An SV (or AV, HV, etc.) is allocated in two parts: the head (struct
78 sv, av, hv...) contains type and reference count information, and for
79 many types, a pointer to the body (struct xrv, xpv, xpviv...), which
80 contains fields specific to each type. Some types store all they need
81 in the head, so don't have a body.
83 In all but the most memory-paranoid configurations (ex: PURIFY), heads
84 and bodies are allocated out of arenas, which by default are
85 approximately 4K chunks of memory parcelled up into N heads or bodies.
86 Sv-bodies are allocated by their sv-type, guaranteeing size
87 consistency needed to allocate safely from arrays.
89 For SV-heads, the first slot in each arena is reserved, and holds a
90 link to the next arena, some flags, and a note of the number of slots.
91 Snaked through each arena chain is a linked list of free items; when
92 this becomes empty, an extra arena is allocated and divided up into N
93 items which are threaded into the free list.
95 SV-bodies are similar, but they use arena-sets by default, which
96 separate the link and info from the arena itself, and reclaim the 1st
97 slot in the arena. SV-bodies are further described later.
99 The following global variables are associated with arenas:
101 PL_sv_arenaroot pointer to list of SV arenas
102 PL_sv_root pointer to list of free SV structures
104 PL_body_arenas head of linked-list of body arenas
105 PL_body_roots[] array of pointers to list of free bodies of svtype
106 arrays are indexed by the svtype needed
108 A few special SV heads are not allocated from an arena, but are
109 instead directly created in the interpreter structure, eg PL_sv_undef.
110 The size of arenas can be changed from the default by setting
111 PERL_ARENA_SIZE appropriately at compile time.
113 The SV arena serves the secondary purpose of allowing still-live SVs
114 to be located and destroyed during final cleanup.
116 At the lowest level, the macros new_SV() and del_SV() grab and free
117 an SV head. (If debugging with -DD, del_SV() calls the function S_del_sv()
118 to return the SV to the free list with error checking.) new_SV() calls
119 more_sv() / sv_add_arena() to add an extra arena if the free list is empty.
120 SVs in the free list have their SvTYPE field set to all ones.
122 At the time of very final cleanup, sv_free_arenas() is called from
123 perl_destruct() to physically free all the arenas allocated since the
124 start of the interpreter.
126 The function visit() scans the SV arenas list, and calls a specified
127 function for each SV it finds which is still live - ie which has an SvTYPE
128 other than all 1's, and a non-zero SvREFCNT. visit() is used by the
129 following functions (specified as [function that calls visit()] / [function
130 called by visit() for each SV]):
132 sv_report_used() / do_report_used()
133 dump all remaining SVs (debugging aid)
135 sv_clean_objs() / do_clean_objs(),do_clean_named_objs(),
136 do_clean_named_io_objs()
137 Attempt to free all objects pointed to by RVs,
138 and try to do the same for all objects indirectly
139 referenced by typeglobs too. Called once from
140 perl_destruct(), prior to calling sv_clean_all()
143 sv_clean_all() / do_clean_all()
144 SvREFCNT_dec(sv) each remaining SV, possibly
145 triggering an sv_free(). It also sets the
146 SVf_BREAK flag on the SV to indicate that the
147 refcnt has been artificially lowered, and thus
148 stopping sv_free() from giving spurious warnings
149 about SVs which unexpectedly have a refcnt
150 of zero. called repeatedly from perl_destruct()
151 until there are no SVs left.
153 =head2 Arena allocator API Summary
155 Private API to rest of sv.c
159 new_XPVNV(), del_XPVGV(),
164 sv_report_used(), sv_clean_objs(), sv_clean_all(), sv_free_arenas()
168 * ========================================================================= */
171 * "A time to plant, and a time to uproot what was planted..."
175 # define MEM_LOG_NEW_SV(sv, file, line, func) \
176 Perl_mem_log_new_sv(sv, file, line, func)
177 # define MEM_LOG_DEL_SV(sv, file, line, func) \
178 Perl_mem_log_del_sv(sv, file, line, func)
180 # define MEM_LOG_NEW_SV(sv, file, line, func) NOOP
181 # define MEM_LOG_DEL_SV(sv, file, line, func) NOOP
184 #ifdef DEBUG_LEAKING_SCALARS
185 # define FREE_SV_DEBUG_FILE(sv) Safefree((sv)->sv_debug_file)
186 # define DEBUG_SV_SERIAL(sv) \
187 DEBUG_m(PerlIO_printf(Perl_debug_log, "0x%"UVxf": (%05ld) del_SV\n", \
188 PTR2UV(sv), (long)(sv)->sv_debug_serial))
190 # define FREE_SV_DEBUG_FILE(sv)
191 # define DEBUG_SV_SERIAL(sv) NOOP
195 # define SvARENA_CHAIN(sv) ((sv)->sv_u.svu_rv)
196 # define SvARENA_CHAIN_SET(sv,val) (sv)->sv_u.svu_rv = MUTABLE_SV((val))
197 /* Whilst I'd love to do this, it seems that things like to check on
199 # define POSION_SV_HEAD(sv) PoisonNew(sv, 1, struct STRUCT_SV)
201 # define POSION_SV_HEAD(sv) PoisonNew(&SvANY(sv), 1, void *), \
202 PoisonNew(&SvREFCNT(sv), 1, U32)
204 # define SvARENA_CHAIN(sv) SvANY(sv)
205 # define SvARENA_CHAIN_SET(sv,val) SvANY(sv) = (void *)(val)
206 # define POSION_SV_HEAD(sv)
209 /* Mark an SV head as unused, and add to free list.
211 * If SVf_BREAK is set, skip adding it to the free list, as this SV had
212 * its refcount artificially decremented during global destruction, so
213 * there may be dangling pointers to it. The last thing we want in that
214 * case is for it to be reused. */
216 #define plant_SV(p) \
218 const U32 old_flags = SvFLAGS(p); \
219 MEM_LOG_DEL_SV(p, __FILE__, __LINE__, FUNCTION__); \
220 DEBUG_SV_SERIAL(p); \
221 FREE_SV_DEBUG_FILE(p); \
223 SvFLAGS(p) = SVTYPEMASK; \
224 if (!(old_flags & SVf_BREAK)) { \
225 SvARENA_CHAIN_SET(p, PL_sv_root); \
231 #define uproot_SV(p) \
234 PL_sv_root = MUTABLE_SV(SvARENA_CHAIN(p)); \
239 /* make some more SVs by adding another arena */
246 char *chunk; /* must use New here to match call to */
247 Newx(chunk,PERL_ARENA_SIZE,char); /* Safefree() in sv_free_arenas() */
248 sv_add_arena(chunk, PERL_ARENA_SIZE, 0);
253 /* new_SV(): return a new, empty SV head */
255 #ifdef DEBUG_LEAKING_SCALARS
256 /* provide a real function for a debugger to play with */
258 S_new_SV(pTHX_ const char *file, int line, const char *func)
265 sv = S_more_sv(aTHX);
269 sv->sv_debug_optype = PL_op ? PL_op->op_type : 0;
270 sv->sv_debug_line = (U16) (PL_parser && PL_parser->copline != NOLINE
276 sv->sv_debug_inpad = 0;
277 sv->sv_debug_parent = NULL;
278 sv->sv_debug_file = PL_curcop ? savepv(CopFILE(PL_curcop)): NULL;
280 sv->sv_debug_serial = PL_sv_serial++;
282 MEM_LOG_NEW_SV(sv, file, line, func);
283 DEBUG_m(PerlIO_printf(Perl_debug_log, "0x%"UVxf": (%05ld) new_SV (from %s:%d [%s])\n",
284 PTR2UV(sv), (long)sv->sv_debug_serial, file, line, func));
288 # define new_SV(p) (p)=S_new_SV(aTHX_ __FILE__, __LINE__, FUNCTION__)
296 (p) = S_more_sv(aTHX); \
300 MEM_LOG_NEW_SV(p, __FILE__, __LINE__, FUNCTION__); \
305 /* del_SV(): return an empty SV head to the free list */
318 S_del_sv(pTHX_ SV *p)
322 PERL_ARGS_ASSERT_DEL_SV;
327 for (sva = PL_sv_arenaroot; sva; sva = MUTABLE_SV(SvANY(sva))) {
328 const SV * const sv = sva + 1;
329 const SV * const svend = &sva[SvREFCNT(sva)];
330 if (p >= sv && p < svend) {
336 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL),
337 "Attempt to free non-arena SV: 0x%"UVxf
338 pTHX__FORMAT, PTR2UV(p) pTHX__VALUE);
345 #else /* ! DEBUGGING */
347 #define del_SV(p) plant_SV(p)
349 #endif /* DEBUGGING */
353 =head1 SV Manipulation Functions
355 =for apidoc sv_add_arena
357 Given a chunk of memory, link it to the head of the list of arenas,
358 and split it into a list of free SVs.
364 S_sv_add_arena(pTHX_ char *const ptr, const U32 size, const U32 flags)
367 SV *const sva = MUTABLE_SV(ptr);
371 PERL_ARGS_ASSERT_SV_ADD_ARENA;
373 /* The first SV in an arena isn't an SV. */
374 SvANY(sva) = (void *) PL_sv_arenaroot; /* ptr to next arena */
375 SvREFCNT(sva) = size / sizeof(SV); /* number of SV slots */
376 SvFLAGS(sva) = flags; /* FAKE if not to be freed */
378 PL_sv_arenaroot = sva;
379 PL_sv_root = sva + 1;
381 svend = &sva[SvREFCNT(sva) - 1];
384 SvARENA_CHAIN_SET(sv, (sv + 1));
388 /* Must always set typemask because it's always checked in on cleanup
389 when the arenas are walked looking for objects. */
390 SvFLAGS(sv) = SVTYPEMASK;
393 SvARENA_CHAIN_SET(sv, 0);
397 SvFLAGS(sv) = SVTYPEMASK;
400 /* visit(): call the named function for each non-free SV in the arenas
401 * whose flags field matches the flags/mask args. */
404 S_visit(pTHX_ SVFUNC_t f, const U32 flags, const U32 mask)
410 PERL_ARGS_ASSERT_VISIT;
412 for (sva = PL_sv_arenaroot; sva; sva = MUTABLE_SV(SvANY(sva))) {
413 register const SV * const svend = &sva[SvREFCNT(sva)];
415 for (sv = sva + 1; sv < svend; ++sv) {
416 if (SvTYPE(sv) != (svtype)SVTYPEMASK
417 && (sv->sv_flags & mask) == flags
430 /* called by sv_report_used() for each live SV */
433 do_report_used(pTHX_ SV *const sv)
435 if (SvTYPE(sv) != (svtype)SVTYPEMASK) {
436 PerlIO_printf(Perl_debug_log, "****\n");
443 =for apidoc sv_report_used
445 Dump the contents of all SVs not yet freed. (Debugging aid).
451 Perl_sv_report_used(pTHX)
454 visit(do_report_used, 0, 0);
460 /* called by sv_clean_objs() for each live SV */
463 do_clean_objs(pTHX_ SV *const ref)
468 SV * const target = SvRV(ref);
469 if (SvOBJECT(target)) {
470 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning object ref:\n "), sv_dump(ref)));
471 if (SvWEAKREF(ref)) {
472 sv_del_backref(target, ref);
478 SvREFCNT_dec(target);
483 /* XXX Might want to check arrays, etc. */
487 /* clear any slots in a GV which hold objects - except IO;
488 * called by sv_clean_objs() for each live GV */
491 do_clean_named_objs(pTHX_ SV *const sv)
495 assert(SvTYPE(sv) == SVt_PVGV);
496 assert(isGV_with_GP(sv));
500 /* freeing GP entries may indirectly free the current GV;
501 * hold onto it while we mess with the GP slots */
504 if ( ((obj = GvSV(sv) )) && SvOBJECT(obj)) {
505 DEBUG_D((PerlIO_printf(Perl_debug_log,
506 "Cleaning named glob SV object:\n "), sv_dump(obj)));
510 if ( ((obj = MUTABLE_SV(GvAV(sv)) )) && SvOBJECT(obj)) {
511 DEBUG_D((PerlIO_printf(Perl_debug_log,
512 "Cleaning named glob AV object:\n "), sv_dump(obj)));
516 if ( ((obj = MUTABLE_SV(GvHV(sv)) )) && SvOBJECT(obj)) {
517 DEBUG_D((PerlIO_printf(Perl_debug_log,
518 "Cleaning named glob HV object:\n "), sv_dump(obj)));
522 if ( ((obj = MUTABLE_SV(GvCV(sv)) )) && SvOBJECT(obj)) {
523 DEBUG_D((PerlIO_printf(Perl_debug_log,
524 "Cleaning named glob CV object:\n "), sv_dump(obj)));
528 SvREFCNT_dec(sv); /* undo the inc above */
531 /* clear any IO slots in a GV which hold objects (except stderr, defout);
532 * called by sv_clean_objs() for each live GV */
535 do_clean_named_io_objs(pTHX_ SV *const sv)
539 assert(SvTYPE(sv) == SVt_PVGV);
540 assert(isGV_with_GP(sv));
541 if (!GvGP(sv) || sv == (SV*)PL_stderrgv || sv == (SV*)PL_defoutgv)
545 if ( ((obj = MUTABLE_SV(GvIO(sv)) )) && SvOBJECT(obj)) {
546 DEBUG_D((PerlIO_printf(Perl_debug_log,
547 "Cleaning named glob IO object:\n "), sv_dump(obj)));
551 SvREFCNT_dec(sv); /* undo the inc above */
554 /* Void wrapper to pass to visit() */
556 do_curse(pTHX_ SV * const sv) {
557 if ((PL_stderrgv && GvGP(PL_stderrgv) && (SV*)GvIO(PL_stderrgv) == sv)
558 || (PL_defoutgv && GvGP(PL_defoutgv) && (SV*)GvIO(PL_defoutgv) == sv))
564 =for apidoc sv_clean_objs
566 Attempt to destroy all objects not yet freed
572 Perl_sv_clean_objs(pTHX)
576 PL_in_clean_objs = TRUE;
577 visit(do_clean_objs, SVf_ROK, SVf_ROK);
578 /* Some barnacles may yet remain, clinging to typeglobs.
579 * Run the non-IO destructors first: they may want to output
580 * error messages, close files etc */
581 visit(do_clean_named_objs, SVt_PVGV|SVpgv_GP, SVTYPEMASK|SVp_POK|SVpgv_GP);
582 visit(do_clean_named_io_objs, SVt_PVGV|SVpgv_GP, SVTYPEMASK|SVp_POK|SVpgv_GP);
583 /* And if there are some very tenacious barnacles clinging to arrays,
584 closures, or what have you.... */
585 visit(do_curse, SVs_OBJECT, SVs_OBJECT);
586 olddef = PL_defoutgv;
587 PL_defoutgv = NULL; /* disable skip of PL_defoutgv */
588 if (olddef && isGV_with_GP(olddef))
589 do_clean_named_io_objs(aTHX_ MUTABLE_SV(olddef));
590 olderr = PL_stderrgv;
591 PL_stderrgv = NULL; /* disable skip of PL_stderrgv */
592 if (olderr && isGV_with_GP(olderr))
593 do_clean_named_io_objs(aTHX_ MUTABLE_SV(olderr));
594 SvREFCNT_dec(olddef);
595 PL_in_clean_objs = FALSE;
598 /* called by sv_clean_all() for each live SV */
601 do_clean_all(pTHX_ SV *const sv)
604 if (sv == (const SV *) PL_fdpid || sv == (const SV *)PL_strtab) {
605 /* don't clean pid table and strtab */
608 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning loops: SV at 0x%"UVxf"\n", PTR2UV(sv)) ));
609 SvFLAGS(sv) |= SVf_BREAK;
614 =for apidoc sv_clean_all
616 Decrement the refcnt of each remaining SV, possibly triggering a
617 cleanup. This function may have to be called multiple times to free
618 SVs which are in complex self-referential hierarchies.
624 Perl_sv_clean_all(pTHX)
628 PL_in_clean_all = TRUE;
629 cleaned = visit(do_clean_all, 0,0);
634 ARENASETS: a meta-arena implementation which separates arena-info
635 into struct arena_set, which contains an array of struct
636 arena_descs, each holding info for a single arena. By separating
637 the meta-info from the arena, we recover the 1st slot, formerly
638 borrowed for list management. The arena_set is about the size of an
639 arena, avoiding the needless malloc overhead of a naive linked-list.
641 The cost is 1 arena-set malloc per ~320 arena-mallocs, + the unused
642 memory in the last arena-set (1/2 on average). In trade, we get
643 back the 1st slot in each arena (ie 1.7% of a CV-arena, less for
644 smaller types). The recovery of the wasted space allows use of
645 small arenas for large, rare body types, by changing array* fields
646 in body_details_by_type[] below.
649 char *arena; /* the raw storage, allocated aligned */
650 size_t size; /* its size ~4k typ */
651 svtype utype; /* bodytype stored in arena */
656 /* Get the maximum number of elements in set[] such that struct arena_set
657 will fit within PERL_ARENA_SIZE, which is probably just under 4K, and
658 therefore likely to be 1 aligned memory page. */
660 #define ARENAS_PER_SET ((PERL_ARENA_SIZE - sizeof(struct arena_set*) \
661 - 2 * sizeof(int)) / sizeof (struct arena_desc))
664 struct arena_set* next;
665 unsigned int set_size; /* ie ARENAS_PER_SET */
666 unsigned int curr; /* index of next available arena-desc */
667 struct arena_desc set[ARENAS_PER_SET];
671 =for apidoc sv_free_arenas
673 Deallocate the memory used by all arenas. Note that all the individual SV
674 heads and bodies within the arenas must already have been freed.
679 Perl_sv_free_arenas(pTHX)
686 /* Free arenas here, but be careful about fake ones. (We assume
687 contiguity of the fake ones with the corresponding real ones.) */
689 for (sva = PL_sv_arenaroot; sva; sva = svanext) {
690 svanext = MUTABLE_SV(SvANY(sva));
691 while (svanext && SvFAKE(svanext))
692 svanext = MUTABLE_SV(SvANY(svanext));
699 struct arena_set *aroot = (struct arena_set*) PL_body_arenas;
702 struct arena_set *current = aroot;
705 assert(aroot->set[i].arena);
706 Safefree(aroot->set[i].arena);
714 i = PERL_ARENA_ROOTS_SIZE;
716 PL_body_roots[i] = 0;
723 Here are mid-level routines that manage the allocation of bodies out
724 of the various arenas. There are 5 kinds of arenas:
726 1. SV-head arenas, which are discussed and handled above
727 2. regular body arenas
728 3. arenas for reduced-size bodies
731 Arena types 2 & 3 are chained by body-type off an array of
732 arena-root pointers, which is indexed by svtype. Some of the
733 larger/less used body types are malloced singly, since a large
734 unused block of them is wasteful. Also, several svtypes dont have
735 bodies; the data fits into the sv-head itself. The arena-root
736 pointer thus has a few unused root-pointers (which may be hijacked
737 later for arena types 4,5)
739 3 differs from 2 as an optimization; some body types have several
740 unused fields in the front of the structure (which are kept in-place
741 for consistency). These bodies can be allocated in smaller chunks,
742 because the leading fields arent accessed. Pointers to such bodies
743 are decremented to point at the unused 'ghost' memory, knowing that
744 the pointers are used with offsets to the real memory.
747 =head1 SV-Body Allocation
749 Allocation of SV-bodies is similar to SV-heads, differing as follows;
750 the allocation mechanism is used for many body types, so is somewhat
751 more complicated, it uses arena-sets, and has no need for still-live
754 At the outermost level, (new|del)_X*V macros return bodies of the
755 appropriate type. These macros call either (new|del)_body_type or
756 (new|del)_body_allocated macro pairs, depending on specifics of the
757 type. Most body types use the former pair, the latter pair is used to
758 allocate body types with "ghost fields".
760 "ghost fields" are fields that are unused in certain types, and
761 consequently don't need to actually exist. They are declared because
762 they're part of a "base type", which allows use of functions as
763 methods. The simplest examples are AVs and HVs, 2 aggregate types
764 which don't use the fields which support SCALAR semantics.
766 For these types, the arenas are carved up into appropriately sized
767 chunks, we thus avoid wasted memory for those unaccessed members.
768 When bodies are allocated, we adjust the pointer back in memory by the
769 size of the part not allocated, so it's as if we allocated the full
770 structure. (But things will all go boom if you write to the part that
771 is "not there", because you'll be overwriting the last members of the
772 preceding structure in memory.)
774 We calculate the correction using the STRUCT_OFFSET macro on the first
775 member present. If the allocated structure is smaller (no initial NV
776 actually allocated) then the net effect is to subtract the size of the NV
777 from the pointer, to return a new pointer as if an initial NV were actually
778 allocated. (We were using structures named *_allocated for this, but
779 this turned out to be a subtle bug, because a structure without an NV
780 could have a lower alignment constraint, but the compiler is allowed to
781 optimised accesses based on the alignment constraint of the actual pointer
782 to the full structure, for example, using a single 64 bit load instruction
783 because it "knows" that two adjacent 32 bit members will be 8-byte aligned.)
785 This is the same trick as was used for NV and IV bodies. Ironically it
786 doesn't need to be used for NV bodies any more, because NV is now at
787 the start of the structure. IV bodies don't need it either, because
788 they are no longer allocated.
790 In turn, the new_body_* allocators call S_new_body(), which invokes
791 new_body_inline macro, which takes a lock, and takes a body off the
792 linked list at PL_body_roots[sv_type], calling Perl_more_bodies() if
793 necessary to refresh an empty list. Then the lock is released, and
794 the body is returned.
796 Perl_more_bodies allocates a new arena, and carves it up into an array of N
797 bodies, which it strings into a linked list. It looks up arena-size
798 and body-size from the body_details table described below, thus
799 supporting the multiple body-types.
801 If PURIFY is defined, or PERL_ARENA_SIZE=0, arenas are not used, and
802 the (new|del)_X*V macros are mapped directly to malloc/free.
804 For each sv-type, struct body_details bodies_by_type[] carries
805 parameters which control these aspects of SV handling:
807 Arena_size determines whether arenas are used for this body type, and if
808 so, how big they are. PURIFY or PERL_ARENA_SIZE=0 set this field to
809 zero, forcing individual mallocs and frees.
811 Body_size determines how big a body is, and therefore how many fit into
812 each arena. Offset carries the body-pointer adjustment needed for
813 "ghost fields", and is used in *_allocated macros.
815 But its main purpose is to parameterize info needed in
816 Perl_sv_upgrade(). The info here dramatically simplifies the function
817 vs the implementation in 5.8.8, making it table-driven. All fields
818 are used for this, except for arena_size.
820 For the sv-types that have no bodies, arenas are not used, so those
821 PL_body_roots[sv_type] are unused, and can be overloaded. In
822 something of a special case, SVt_NULL is borrowed for HE arenas;
823 PL_body_roots[HE_SVSLOT=SVt_NULL] is filled by S_more_he, but the
824 bodies_by_type[SVt_NULL] slot is not used, as the table is not
829 struct body_details {
830 U8 body_size; /* Size to allocate */
831 U8 copy; /* Size of structure to copy (may be shorter) */
833 unsigned int type : 4; /* We have space for a sanity check. */
834 unsigned int cant_upgrade : 1; /* Cannot upgrade this type */
835 unsigned int zero_nv : 1; /* zero the NV when upgrading from this */
836 unsigned int arena : 1; /* Allocated from an arena */
837 size_t arena_size; /* Size of arena to allocate */
845 /* With -DPURFIY we allocate everything directly, and don't use arenas.
846 This seems a rather elegant way to simplify some of the code below. */
847 #define HASARENA FALSE
849 #define HASARENA TRUE
851 #define NOARENA FALSE
853 /* Size the arenas to exactly fit a given number of bodies. A count
854 of 0 fits the max number bodies into a PERL_ARENA_SIZE.block,
855 simplifying the default. If count > 0, the arena is sized to fit
856 only that many bodies, allowing arenas to be used for large, rare
857 bodies (XPVFM, XPVIO) without undue waste. The arena size is
858 limited by PERL_ARENA_SIZE, so we can safely oversize the
861 #define FIT_ARENA0(body_size) \
862 ((size_t)(PERL_ARENA_SIZE / body_size) * body_size)
863 #define FIT_ARENAn(count,body_size) \
864 ( count * body_size <= PERL_ARENA_SIZE) \
865 ? count * body_size \
866 : FIT_ARENA0 (body_size)
867 #define FIT_ARENA(count,body_size) \
869 ? FIT_ARENAn (count, body_size) \
870 : FIT_ARENA0 (body_size)
872 /* Calculate the length to copy. Specifically work out the length less any
873 final padding the compiler needed to add. See the comment in sv_upgrade
874 for why copying the padding proved to be a bug. */
876 #define copy_length(type, last_member) \
877 STRUCT_OFFSET(type, last_member) \
878 + sizeof (((type*)SvANY((const SV *)0))->last_member)
880 static const struct body_details bodies_by_type[] = {
881 /* HEs use this offset for their arena. */
882 { 0, 0, 0, SVt_NULL, FALSE, NONV, NOARENA, 0 },
884 /* The bind placeholder pretends to be an RV for now.
885 Also it's marked as "can't upgrade" to stop anyone using it before it's
887 { 0, 0, 0, SVt_BIND, TRUE, NONV, NOARENA, 0 },
889 /* IVs are in the head, so the allocation size is 0. */
891 sizeof(IV), /* This is used to copy out the IV body. */
892 STRUCT_OFFSET(XPVIV, xiv_iv), SVt_IV, FALSE, NONV,
893 NOARENA /* IVS don't need an arena */, 0
896 { sizeof(NV), sizeof(NV),
897 STRUCT_OFFSET(XPVNV, xnv_u),
898 SVt_NV, FALSE, HADNV, HASARENA, FIT_ARENA(0, sizeof(NV)) },
900 { sizeof(XPV) - STRUCT_OFFSET(XPV, xpv_cur),
901 copy_length(XPV, xpv_len) - STRUCT_OFFSET(XPV, xpv_cur),
902 + STRUCT_OFFSET(XPV, xpv_cur),
903 SVt_PV, FALSE, NONV, HASARENA,
904 FIT_ARENA(0, sizeof(XPV) - STRUCT_OFFSET(XPV, xpv_cur)) },
906 { sizeof(XPVIV) - STRUCT_OFFSET(XPV, xpv_cur),
907 copy_length(XPVIV, xiv_u) - STRUCT_OFFSET(XPV, xpv_cur),
908 + STRUCT_OFFSET(XPV, xpv_cur),
909 SVt_PVIV, FALSE, NONV, HASARENA,
910 FIT_ARENA(0, sizeof(XPVIV) - STRUCT_OFFSET(XPV, xpv_cur)) },
912 { sizeof(XPVNV) - STRUCT_OFFSET(XPV, xpv_cur),
913 copy_length(XPVNV, xnv_u) - STRUCT_OFFSET(XPV, xpv_cur),
914 + STRUCT_OFFSET(XPV, xpv_cur),
915 SVt_PVNV, FALSE, HADNV, HASARENA,
916 FIT_ARENA(0, sizeof(XPVNV) - STRUCT_OFFSET(XPV, xpv_cur)) },
918 { sizeof(XPVMG), copy_length(XPVMG, xnv_u), 0, SVt_PVMG, FALSE, HADNV,
919 HASARENA, FIT_ARENA(0, sizeof(XPVMG)) },
924 SVt_REGEXP, FALSE, NONV, HASARENA,
925 FIT_ARENA(0, sizeof(regexp))
928 { sizeof(XPVGV), sizeof(XPVGV), 0, SVt_PVGV, TRUE, HADNV,
929 HASARENA, FIT_ARENA(0, sizeof(XPVGV)) },
931 { sizeof(XPVLV), sizeof(XPVLV), 0, SVt_PVLV, TRUE, HADNV,
932 HASARENA, FIT_ARENA(0, sizeof(XPVLV)) },
935 copy_length(XPVAV, xav_alloc),
937 SVt_PVAV, TRUE, NONV, HASARENA,
938 FIT_ARENA(0, sizeof(XPVAV)) },
941 copy_length(XPVHV, xhv_max),
943 SVt_PVHV, TRUE, NONV, HASARENA,
944 FIT_ARENA(0, sizeof(XPVHV)) },
949 SVt_PVCV, TRUE, NONV, HASARENA,
950 FIT_ARENA(0, sizeof(XPVCV)) },
955 SVt_PVFM, TRUE, NONV, NOARENA,
956 FIT_ARENA(20, sizeof(XPVFM)) },
961 SVt_PVIO, TRUE, NONV, HASARENA,
962 FIT_ARENA(24, sizeof(XPVIO)) },
965 #define new_body_allocated(sv_type) \
966 (void *)((char *)S_new_body(aTHX_ sv_type) \
967 - bodies_by_type[sv_type].offset)
969 /* return a thing to the free list */
971 #define del_body(thing, root) \
973 void ** const thing_copy = (void **)thing; \
974 *thing_copy = *root; \
975 *root = (void*)thing_copy; \
980 #define new_XNV() safemalloc(sizeof(XPVNV))
981 #define new_XPVNV() safemalloc(sizeof(XPVNV))
982 #define new_XPVMG() safemalloc(sizeof(XPVMG))
984 #define del_XPVGV(p) safefree(p)
988 #define new_XNV() new_body_allocated(SVt_NV)
989 #define new_XPVNV() new_body_allocated(SVt_PVNV)
990 #define new_XPVMG() new_body_allocated(SVt_PVMG)
992 #define del_XPVGV(p) del_body(p + bodies_by_type[SVt_PVGV].offset, \
993 &PL_body_roots[SVt_PVGV])
997 /* no arena for you! */
999 #define new_NOARENA(details) \
1000 safemalloc((details)->body_size + (details)->offset)
1001 #define new_NOARENAZ(details) \
1002 safecalloc((details)->body_size + (details)->offset, 1)
1005 Perl_more_bodies (pTHX_ const svtype sv_type, const size_t body_size,
1006 const size_t arena_size)
1009 void ** const root = &PL_body_roots[sv_type];
1010 struct arena_desc *adesc;
1011 struct arena_set *aroot = (struct arena_set *) PL_body_arenas;
1015 const size_t good_arena_size = Perl_malloc_good_size(arena_size);
1016 #if defined(DEBUGGING) && !defined(PERL_GLOBAL_STRUCT_PRIVATE)
1017 static bool done_sanity_check;
1019 /* PERL_GLOBAL_STRUCT_PRIVATE cannot coexist with global
1020 * variables like done_sanity_check. */
1021 if (!done_sanity_check) {
1022 unsigned int i = SVt_LAST;
1024 done_sanity_check = TRUE;
1027 assert (bodies_by_type[i].type == i);
1033 /* may need new arena-set to hold new arena */
1034 if (!aroot || aroot->curr >= aroot->set_size) {
1035 struct arena_set *newroot;
1036 Newxz(newroot, 1, struct arena_set);
1037 newroot->set_size = ARENAS_PER_SET;
1038 newroot->next = aroot;
1040 PL_body_arenas = (void *) newroot;
1041 DEBUG_m(PerlIO_printf(Perl_debug_log, "new arenaset %p\n", (void*)aroot));
1044 /* ok, now have arena-set with at least 1 empty/available arena-desc */
1045 curr = aroot->curr++;
1046 adesc = &(aroot->set[curr]);
1047 assert(!adesc->arena);
1049 Newx(adesc->arena, good_arena_size, char);
1050 adesc->size = good_arena_size;
1051 adesc->utype = sv_type;
1052 DEBUG_m(PerlIO_printf(Perl_debug_log, "arena %d added: %p size %"UVuf"\n",
1053 curr, (void*)adesc->arena, (UV)good_arena_size));
1055 start = (char *) adesc->arena;
1057 /* Get the address of the byte after the end of the last body we can fit.
1058 Remember, this is integer division: */
1059 end = start + good_arena_size / body_size * body_size;
1061 /* computed count doesn't reflect the 1st slot reservation */
1062 #if defined(MYMALLOC) || defined(HAS_MALLOC_GOOD_SIZE)
1063 DEBUG_m(PerlIO_printf(Perl_debug_log,
1064 "arena %p end %p arena-size %d (from %d) type %d "
1066 (void*)start, (void*)end, (int)good_arena_size,
1067 (int)arena_size, sv_type, (int)body_size,
1068 (int)good_arena_size / (int)body_size));
1070 DEBUG_m(PerlIO_printf(Perl_debug_log,
1071 "arena %p end %p arena-size %d type %d size %d ct %d\n",
1072 (void*)start, (void*)end,
1073 (int)arena_size, sv_type, (int)body_size,
1074 (int)good_arena_size / (int)body_size));
1076 *root = (void *)start;
1079 /* Where the next body would start: */
1080 char * const next = start + body_size;
1083 /* This is the last body: */
1084 assert(next == end);
1086 *(void **)start = 0;
1090 *(void**) start = (void *)next;
1095 /* grab a new thing from the free list, allocating more if necessary.
1096 The inline version is used for speed in hot routines, and the
1097 function using it serves the rest (unless PURIFY).
1099 #define new_body_inline(xpv, sv_type) \
1101 void ** const r3wt = &PL_body_roots[sv_type]; \
1102 xpv = (PTR_TBL_ENT_t*) (*((void **)(r3wt)) \
1103 ? *((void **)(r3wt)) : Perl_more_bodies(aTHX_ sv_type, \
1104 bodies_by_type[sv_type].body_size,\
1105 bodies_by_type[sv_type].arena_size)); \
1106 *(r3wt) = *(void**)(xpv); \
1112 S_new_body(pTHX_ const svtype sv_type)
1116 new_body_inline(xpv, sv_type);
1122 static const struct body_details fake_rv =
1123 { 0, 0, 0, SVt_IV, FALSE, NONV, NOARENA, 0 };
1126 =for apidoc sv_upgrade
1128 Upgrade an SV to a more complex form. Generally adds a new body type to the
1129 SV, then copies across as much information as possible from the old body.
1130 You generally want to use the C<SvUPGRADE> macro wrapper. See also C<svtype>.
1136 Perl_sv_upgrade(pTHX_ register SV *const sv, svtype new_type)
1141 const svtype old_type = SvTYPE(sv);
1142 const struct body_details *new_type_details;
1143 const struct body_details *old_type_details
1144 = bodies_by_type + old_type;
1145 SV *referant = NULL;
1147 PERL_ARGS_ASSERT_SV_UPGRADE;
1149 if (old_type == new_type)
1152 /* This clause was purposefully added ahead of the early return above to
1153 the shared string hackery for (sort {$a <=> $b} keys %hash), with the
1154 inference by Nick I-S that it would fix other troublesome cases. See
1155 changes 7162, 7163 (f130fd4589cf5fbb24149cd4db4137c8326f49c1 and parent)
1157 Given that shared hash key scalars are no longer PVIV, but PV, there is
1158 no longer need to unshare so as to free up the IVX slot for its proper
1159 purpose. So it's safe to move the early return earlier. */
1161 if (new_type != SVt_PV && SvIsCOW(sv)) {
1162 sv_force_normal_flags(sv, 0);
1165 old_body = SvANY(sv);
1167 /* Copying structures onto other structures that have been neatly zeroed
1168 has a subtle gotcha. Consider XPVMG
1170 +------+------+------+------+------+-------+-------+
1171 | NV | CUR | LEN | IV | MAGIC | STASH |
1172 +------+------+------+------+------+-------+-------+
1173 0 4 8 12 16 20 24 28
1175 where NVs are aligned to 8 bytes, so that sizeof that structure is
1176 actually 32 bytes long, with 4 bytes of padding at the end:
1178 +------+------+------+------+------+-------+-------+------+
1179 | NV | CUR | LEN | IV | MAGIC | STASH | ??? |
1180 +------+------+------+------+------+-------+-------+------+
1181 0 4 8 12 16 20 24 28 32
1183 so what happens if you allocate memory for this structure:
1185 +------+------+------+------+------+-------+-------+------+------+...
1186 | NV | CUR | LEN | IV | MAGIC | STASH | GP | NAME |
1187 +------+------+------+------+------+-------+-------+------+------+...
1188 0 4 8 12 16 20 24 28 32 36
1190 zero it, then copy sizeof(XPVMG) bytes on top of it? Not quite what you
1191 expect, because you copy the area marked ??? onto GP. Now, ??? may have
1192 started out as zero once, but it's quite possible that it isn't. So now,
1193 rather than a nicely zeroed GP, you have it pointing somewhere random.
1196 (In fact, GP ends up pointing at a previous GP structure, because the
1197 principle cause of the padding in XPVMG getting garbage is a copy of
1198 sizeof(XPVMG) bytes from a XPVGV structure in sv_unglob. Right now
1199 this happens to be moot because XPVGV has been re-ordered, with GP
1200 no longer after STASH)
1202 So we are careful and work out the size of used parts of all the
1210 referant = SvRV(sv);
1211 old_type_details = &fake_rv;
1212 if (new_type == SVt_NV)
1213 new_type = SVt_PVNV;
1215 if (new_type < SVt_PVIV) {
1216 new_type = (new_type == SVt_NV)
1217 ? SVt_PVNV : SVt_PVIV;
1222 if (new_type < SVt_PVNV) {
1223 new_type = SVt_PVNV;
1227 assert(new_type > SVt_PV);
1228 assert(SVt_IV < SVt_PV);
1229 assert(SVt_NV < SVt_PV);
1236 /* Because the XPVMG of PL_mess_sv isn't allocated from the arena,
1237 there's no way that it can be safely upgraded, because perl.c
1238 expects to Safefree(SvANY(PL_mess_sv)) */
1239 assert(sv != PL_mess_sv);
1240 /* This flag bit is used to mean other things in other scalar types.
1241 Given that it only has meaning inside the pad, it shouldn't be set
1242 on anything that can get upgraded. */
1243 assert(!SvPAD_TYPED(sv));
1246 if (old_type_details->cant_upgrade)
1247 Perl_croak(aTHX_ "Can't upgrade %s (%" UVuf ") to %" UVuf,
1248 sv_reftype(sv, 0), (UV) old_type, (UV) new_type);
1251 if (old_type > new_type)
1252 Perl_croak(aTHX_ "sv_upgrade from type %d down to type %d",
1253 (int)old_type, (int)new_type);
1255 new_type_details = bodies_by_type + new_type;
1257 SvFLAGS(sv) &= ~SVTYPEMASK;
1258 SvFLAGS(sv) |= new_type;
1260 /* This can't happen, as SVt_NULL is <= all values of new_type, so one of
1261 the return statements above will have triggered. */
1262 assert (new_type != SVt_NULL);
1265 assert(old_type == SVt_NULL);
1266 SvANY(sv) = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
1270 assert(old_type == SVt_NULL);
1271 SvANY(sv) = new_XNV();
1276 assert(new_type_details->body_size);
1279 assert(new_type_details->arena);
1280 assert(new_type_details->arena_size);
1281 /* This points to the start of the allocated area. */
1282 new_body_inline(new_body, new_type);
1283 Zero(new_body, new_type_details->body_size, char);
1284 new_body = ((char *)new_body) - new_type_details->offset;
1286 /* We always allocated the full length item with PURIFY. To do this
1287 we fake things so that arena is false for all 16 types.. */
1288 new_body = new_NOARENAZ(new_type_details);
1290 SvANY(sv) = new_body;
1291 if (new_type == SVt_PVAV) {
1295 if (old_type_details->body_size) {
1298 /* It will have been zeroed when the new body was allocated.
1299 Lets not write to it, in case it confuses a write-back
1305 #ifndef NODEFAULT_SHAREKEYS
1306 HvSHAREKEYS_on(sv); /* key-sharing on by default */
1308 HvMAX(sv) = 7; /* (start with 8 buckets) */
1311 /* SVt_NULL isn't the only thing upgraded to AV or HV.
1312 The target created by newSVrv also is, and it can have magic.
1313 However, it never has SvPVX set.
1315 if (old_type == SVt_IV) {
1317 } else if (old_type >= SVt_PV) {
1318 assert(SvPVX_const(sv) == 0);
1321 if (old_type >= SVt_PVMG) {
1322 SvMAGIC_set(sv, ((XPVMG*)old_body)->xmg_u.xmg_magic);
1323 SvSTASH_set(sv, ((XPVMG*)old_body)->xmg_stash);
1325 sv->sv_u.svu_array = NULL; /* or svu_hash */
1331 /* This ensures that SvTHINKFIRST(sv) is true, and hence that
1332 sv_force_normal_flags(sv) is called. */
1335 /* XXX Is this still needed? Was it ever needed? Surely as there is
1336 no route from NV to PVIV, NOK can never be true */
1337 assert(!SvNOKp(sv));
1348 assert(new_type_details->body_size);
1349 /* We always allocated the full length item with PURIFY. To do this
1350 we fake things so that arena is false for all 16 types.. */
1351 if(new_type_details->arena) {
1352 /* This points to the start of the allocated area. */
1353 new_body_inline(new_body, new_type);
1354 Zero(new_body, new_type_details->body_size, char);
1355 new_body = ((char *)new_body) - new_type_details->offset;
1357 new_body = new_NOARENAZ(new_type_details);
1359 SvANY(sv) = new_body;
1361 if (old_type_details->copy) {
1362 /* There is now the potential for an upgrade from something without
1363 an offset (PVNV or PVMG) to something with one (PVCV, PVFM) */
1364 int offset = old_type_details->offset;
1365 int length = old_type_details->copy;
1367 if (new_type_details->offset > old_type_details->offset) {
1368 const int difference
1369 = new_type_details->offset - old_type_details->offset;
1370 offset += difference;
1371 length -= difference;
1373 assert (length >= 0);
1375 Copy((char *)old_body + offset, (char *)new_body + offset, length,
1379 #ifndef NV_ZERO_IS_ALLBITS_ZERO
1380 /* If NV 0.0 is stores as all bits 0 then Zero() already creates a
1381 * correct 0.0 for us. Otherwise, if the old body didn't have an
1382 * NV slot, but the new one does, then we need to initialise the
1383 * freshly created NV slot with whatever the correct bit pattern is
1385 if (old_type_details->zero_nv && !new_type_details->zero_nv
1386 && !isGV_with_GP(sv))
1390 if (new_type == SVt_PVIO) {
1391 IO * const io = MUTABLE_IO(sv);
1392 GV *iogv = gv_fetchpvs("IO::File::", GV_ADD, SVt_PVHV);
1395 /* Clear the stashcache because a new IO could overrule a package
1397 hv_clear(PL_stashcache);
1399 SvSTASH_set(io, MUTABLE_HV(SvREFCNT_inc(GvHV(iogv))));
1400 IoPAGE_LEN(sv) = 60;
1402 if (old_type < SVt_PV) {
1403 /* referant will be NULL unless the old type was SVt_IV emulating
1405 sv->sv_u.svu_rv = referant;
1409 Perl_croak(aTHX_ "panic: sv_upgrade to unknown type %lu",
1410 (unsigned long)new_type);
1413 if (old_type > SVt_IV) {
1417 /* Note that there is an assumption that all bodies of types that
1418 can be upgraded came from arenas. Only the more complex non-
1419 upgradable types are allowed to be directly malloc()ed. */
1420 assert(old_type_details->arena);
1421 del_body((void*)((char*)old_body + old_type_details->offset),
1422 &PL_body_roots[old_type]);
1428 =for apidoc sv_backoff
1430 Remove any string offset. You should normally use the C<SvOOK_off> macro
1437 Perl_sv_backoff(pTHX_ register SV *const sv)
1440 const char * const s = SvPVX_const(sv);
1442 PERL_ARGS_ASSERT_SV_BACKOFF;
1443 PERL_UNUSED_CONTEXT;
1446 assert(SvTYPE(sv) != SVt_PVHV);
1447 assert(SvTYPE(sv) != SVt_PVAV);
1449 SvOOK_offset(sv, delta);
1451 SvLEN_set(sv, SvLEN(sv) + delta);
1452 SvPV_set(sv, SvPVX(sv) - delta);
1453 Move(s, SvPVX(sv), SvCUR(sv)+1, char);
1454 SvFLAGS(sv) &= ~SVf_OOK;
1461 Expands the character buffer in the SV. If necessary, uses C<sv_unref> and
1462 upgrades the SV to C<SVt_PV>. Returns a pointer to the character buffer.
1463 Use the C<SvGROW> wrapper instead.
1469 Perl_sv_grow(pTHX_ register SV *const sv, register STRLEN newlen)
1473 PERL_ARGS_ASSERT_SV_GROW;
1475 if (PL_madskills && newlen >= 0x100000) {
1476 PerlIO_printf(Perl_debug_log,
1477 "Allocation too large: %"UVxf"\n", (UV)newlen);
1479 #ifdef HAS_64K_LIMIT
1480 if (newlen >= 0x10000) {
1481 PerlIO_printf(Perl_debug_log,
1482 "Allocation too large: %"UVxf"\n", (UV)newlen);
1485 #endif /* HAS_64K_LIMIT */
1488 if (SvTYPE(sv) < SVt_PV) {
1489 sv_upgrade(sv, SVt_PV);
1490 s = SvPVX_mutable(sv);
1492 else if (SvOOK(sv)) { /* pv is offset? */
1494 s = SvPVX_mutable(sv);
1495 if (newlen > SvLEN(sv))
1496 newlen += 10 * (newlen - SvCUR(sv)); /* avoid copy each time */
1497 #ifdef HAS_64K_LIMIT
1498 if (newlen >= 0x10000)
1503 s = SvPVX_mutable(sv);
1505 if (newlen > SvLEN(sv)) { /* need more room? */
1506 STRLEN minlen = SvCUR(sv);
1507 minlen += (minlen >> PERL_STRLEN_EXPAND_SHIFT) + 10;
1508 if (newlen < minlen)
1510 #ifndef Perl_safesysmalloc_size
1511 newlen = PERL_STRLEN_ROUNDUP(newlen);
1513 if (SvLEN(sv) && s) {
1514 s = (char*)saferealloc(s, newlen);
1517 s = (char*)safemalloc(newlen);
1518 if (SvPVX_const(sv) && SvCUR(sv)) {
1519 Move(SvPVX_const(sv), s, (newlen < SvCUR(sv)) ? newlen : SvCUR(sv), char);
1523 #ifdef Perl_safesysmalloc_size
1524 /* Do this here, do it once, do it right, and then we will never get
1525 called back into sv_grow() unless there really is some growing
1527 SvLEN_set(sv, Perl_safesysmalloc_size(s));
1529 SvLEN_set(sv, newlen);
1536 =for apidoc sv_setiv
1538 Copies an integer into the given SV, upgrading first if necessary.
1539 Does not handle 'set' magic. See also C<sv_setiv_mg>.
1545 Perl_sv_setiv(pTHX_ register SV *const sv, const IV i)
1549 PERL_ARGS_ASSERT_SV_SETIV;
1551 SV_CHECK_THINKFIRST_COW_DROP(sv);
1552 switch (SvTYPE(sv)) {
1555 sv_upgrade(sv, SVt_IV);
1558 sv_upgrade(sv, SVt_PVIV);
1562 if (!isGV_with_GP(sv))
1569 /* diag_listed_as: Can't coerce %s to %s in %s */
1570 Perl_croak(aTHX_ "Can't coerce %s to integer in %s", sv_reftype(sv,0),
1574 (void)SvIOK_only(sv); /* validate number */
1580 =for apidoc sv_setiv_mg
1582 Like C<sv_setiv>, but also handles 'set' magic.
1588 Perl_sv_setiv_mg(pTHX_ register SV *const sv, const IV i)
1590 PERL_ARGS_ASSERT_SV_SETIV_MG;
1597 =for apidoc sv_setuv
1599 Copies an unsigned integer into the given SV, upgrading first if necessary.
1600 Does not handle 'set' magic. See also C<sv_setuv_mg>.
1606 Perl_sv_setuv(pTHX_ register SV *const sv, const UV u)
1608 PERL_ARGS_ASSERT_SV_SETUV;
1610 /* With these two if statements:
1611 u=1.49 s=0.52 cu=72.49 cs=10.64 scripts=270 tests=20865
1614 u=1.35 s=0.47 cu=73.45 cs=11.43 scripts=270 tests=20865
1616 If you wish to remove them, please benchmark to see what the effect is
1618 if (u <= (UV)IV_MAX) {
1619 sv_setiv(sv, (IV)u);
1628 =for apidoc sv_setuv_mg
1630 Like C<sv_setuv>, but also handles 'set' magic.
1636 Perl_sv_setuv_mg(pTHX_ register SV *const sv, const UV u)
1638 PERL_ARGS_ASSERT_SV_SETUV_MG;
1645 =for apidoc sv_setnv
1647 Copies a double into the given SV, upgrading first if necessary.
1648 Does not handle 'set' magic. See also C<sv_setnv_mg>.
1654 Perl_sv_setnv(pTHX_ register SV *const sv, const NV num)
1658 PERL_ARGS_ASSERT_SV_SETNV;
1660 SV_CHECK_THINKFIRST_COW_DROP(sv);
1661 switch (SvTYPE(sv)) {
1664 sv_upgrade(sv, SVt_NV);
1668 sv_upgrade(sv, SVt_PVNV);
1672 if (!isGV_with_GP(sv))
1679 /* diag_listed_as: Can't coerce %s to %s in %s */
1680 Perl_croak(aTHX_ "Can't coerce %s to number in %s", sv_reftype(sv,0),
1685 (void)SvNOK_only(sv); /* validate number */
1690 =for apidoc sv_setnv_mg
1692 Like C<sv_setnv>, but also handles 'set' magic.
1698 Perl_sv_setnv_mg(pTHX_ register SV *const sv, const NV num)
1700 PERL_ARGS_ASSERT_SV_SETNV_MG;
1706 /* Print an "isn't numeric" warning, using a cleaned-up,
1707 * printable version of the offending string
1711 S_not_a_number(pTHX_ SV *const sv)
1718 PERL_ARGS_ASSERT_NOT_A_NUMBER;
1721 dsv = newSVpvs_flags("", SVs_TEMP);
1722 pv = sv_uni_display(dsv, sv, 10, UNI_DISPLAY_ISPRINT);
1725 const char * const limit = tmpbuf + sizeof(tmpbuf) - 8;
1726 /* each *s can expand to 4 chars + "...\0",
1727 i.e. need room for 8 chars */
1729 const char *s = SvPVX_const(sv);
1730 const char * const end = s + SvCUR(sv);
1731 for ( ; s < end && d < limit; s++ ) {
1733 if (ch & 128 && !isPRINT_LC(ch)) {
1742 else if (ch == '\r') {
1746 else if (ch == '\f') {
1750 else if (ch == '\\') {
1754 else if (ch == '\0') {
1758 else if (isPRINT_LC(ch))
1775 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
1776 "Argument \"%s\" isn't numeric in %s", pv,
1779 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
1780 "Argument \"%s\" isn't numeric", pv);
1784 =for apidoc looks_like_number
1786 Test if the content of an SV looks like a number (or is a number).
1787 C<Inf> and C<Infinity> are treated as numbers (so will not issue a
1788 non-numeric warning), even if your atof() doesn't grok them. Get-magic is
1795 Perl_looks_like_number(pTHX_ SV *const sv)
1797 register const char *sbegin;
1800 PERL_ARGS_ASSERT_LOOKS_LIKE_NUMBER;
1802 if (SvPOK(sv) || SvPOKp(sv)) {
1803 sbegin = SvPV_nomg_const(sv, len);
1806 return SvFLAGS(sv) & (SVf_NOK|SVp_NOK|SVf_IOK|SVp_IOK);
1807 return grok_number(sbegin, len, NULL);
1811 S_glob_2number(pTHX_ GV * const gv)
1813 const U32 wasfake = SvFLAGS(gv) & SVf_FAKE;
1814 SV *const buffer = sv_newmortal();
1816 PERL_ARGS_ASSERT_GLOB_2NUMBER;
1818 /* FAKE globs can get coerced, so need to turn this off temporarily if it
1821 gv_efullname3(buffer, gv, "*");
1822 SvFLAGS(gv) |= wasfake;
1824 /* We know that all GVs stringify to something that is not-a-number,
1825 so no need to test that. */
1826 if (ckWARN(WARN_NUMERIC))
1827 not_a_number(buffer);
1828 /* We just want something true to return, so that S_sv_2iuv_common
1829 can tail call us and return true. */
1833 /* Actually, ISO C leaves conversion of UV to IV undefined, but
1834 until proven guilty, assume that things are not that bad... */
1839 As 64 bit platforms often have an NV that doesn't preserve all bits of
1840 an IV (an assumption perl has been based on to date) it becomes necessary
1841 to remove the assumption that the NV always carries enough precision to
1842 recreate the IV whenever needed, and that the NV is the canonical form.
1843 Instead, IV/UV and NV need to be given equal rights. So as to not lose
1844 precision as a side effect of conversion (which would lead to insanity
1845 and the dragon(s) in t/op/numconvert.t getting very angry) the intent is
1846 1) to distinguish between IV/UV/NV slots that have cached a valid
1847 conversion where precision was lost and IV/UV/NV slots that have a
1848 valid conversion which has lost no precision
1849 2) to ensure that if a numeric conversion to one form is requested that
1850 would lose precision, the precise conversion (or differently
1851 imprecise conversion) is also performed and cached, to prevent
1852 requests for different numeric formats on the same SV causing
1853 lossy conversion chains. (lossless conversion chains are perfectly
1858 SvIOKp is true if the IV slot contains a valid value
1859 SvIOK is true only if the IV value is accurate (UV if SvIOK_UV true)
1860 SvNOKp is true if the NV slot contains a valid value
1861 SvNOK is true only if the NV value is accurate
1864 while converting from PV to NV, check to see if converting that NV to an
1865 IV(or UV) would lose accuracy over a direct conversion from PV to
1866 IV(or UV). If it would, cache both conversions, return NV, but mark
1867 SV as IOK NOKp (ie not NOK).
1869 While converting from PV to IV, check to see if converting that IV to an
1870 NV would lose accuracy over a direct conversion from PV to NV. If it
1871 would, cache both conversions, flag similarly.
1873 Before, the SV value "3.2" could become NV=3.2 IV=3 NOK, IOK quite
1874 correctly because if IV & NV were set NV *always* overruled.
1875 Now, "3.2" will become NV=3.2 IV=3 NOK, IOKp, because the flag's meaning
1876 changes - now IV and NV together means that the two are interchangeable:
1877 SvIVX == (IV) SvNVX && SvNVX == (NV) SvIVX;
1879 The benefit of this is that operations such as pp_add know that if
1880 SvIOK is true for both left and right operands, then integer addition
1881 can be used instead of floating point (for cases where the result won't
1882 overflow). Before, floating point was always used, which could lead to
1883 loss of precision compared with integer addition.
1885 * making IV and NV equal status should make maths accurate on 64 bit
1887 * may speed up maths somewhat if pp_add and friends start to use
1888 integers when possible instead of fp. (Hopefully the overhead in
1889 looking for SvIOK and checking for overflow will not outweigh the
1890 fp to integer speedup)
1891 * will slow down integer operations (callers of SvIV) on "inaccurate"
1892 values, as the change from SvIOK to SvIOKp will cause a call into
1893 sv_2iv each time rather than a macro access direct to the IV slot
1894 * should speed up number->string conversion on integers as IV is
1895 favoured when IV and NV are equally accurate
1897 ####################################################################
1898 You had better be using SvIOK_notUV if you want an IV for arithmetic:
1899 SvIOK is true if (IV or UV), so you might be getting (IV)SvUV.
1900 On the other hand, SvUOK is true iff UV.
1901 ####################################################################
1903 Your mileage will vary depending your CPU's relative fp to integer
1907 #ifndef NV_PRESERVES_UV
1908 # define IS_NUMBER_UNDERFLOW_IV 1
1909 # define IS_NUMBER_UNDERFLOW_UV 2
1910 # define IS_NUMBER_IV_AND_UV 2
1911 # define IS_NUMBER_OVERFLOW_IV 4
1912 # define IS_NUMBER_OVERFLOW_UV 5
1914 /* sv_2iuv_non_preserve(): private routine for use by sv_2iv() and sv_2uv() */
1916 /* For sv_2nv these three cases are "SvNOK and don't bother casting" */
1918 S_sv_2iuv_non_preserve(pTHX_ register SV *const sv
1926 PERL_ARGS_ASSERT_SV_2IUV_NON_PRESERVE;
1928 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_2iuv_non '%s', IV=0x%"UVxf" NV=%"NVgf" inttype=%"UVXf"\n", SvPVX_const(sv), SvIVX(sv), SvNVX(sv), (UV)numtype));
1929 if (SvNVX(sv) < (NV)IV_MIN) {
1930 (void)SvIOKp_on(sv);
1932 SvIV_set(sv, IV_MIN);
1933 return IS_NUMBER_UNDERFLOW_IV;
1935 if (SvNVX(sv) > (NV)UV_MAX) {
1936 (void)SvIOKp_on(sv);
1939 SvUV_set(sv, UV_MAX);
1940 return IS_NUMBER_OVERFLOW_UV;
1942 (void)SvIOKp_on(sv);
1944 /* Can't use strtol etc to convert this string. (See truth table in
1946 if (SvNVX(sv) <= (UV)IV_MAX) {
1947 SvIV_set(sv, I_V(SvNVX(sv)));
1948 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
1949 SvIOK_on(sv); /* Integer is precise. NOK, IOK */
1951 /* Integer is imprecise. NOK, IOKp */
1953 return SvNVX(sv) < 0 ? IS_NUMBER_UNDERFLOW_UV : IS_NUMBER_IV_AND_UV;
1956 SvUV_set(sv, U_V(SvNVX(sv)));
1957 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
1958 if (SvUVX(sv) == UV_MAX) {
1959 /* As we know that NVs don't preserve UVs, UV_MAX cannot
1960 possibly be preserved by NV. Hence, it must be overflow.
1962 return IS_NUMBER_OVERFLOW_UV;
1964 SvIOK_on(sv); /* Integer is precise. NOK, UOK */
1966 /* Integer is imprecise. NOK, IOKp */
1968 return IS_NUMBER_OVERFLOW_IV;
1970 #endif /* !NV_PRESERVES_UV*/
1973 S_sv_2iuv_common(pTHX_ SV *const sv)
1977 PERL_ARGS_ASSERT_SV_2IUV_COMMON;
1980 /* erm. not sure. *should* never get NOKp (without NOK) from sv_2nv
1981 * without also getting a cached IV/UV from it at the same time
1982 * (ie PV->NV conversion should detect loss of accuracy and cache
1983 * IV or UV at same time to avoid this. */
1984 /* IV-over-UV optimisation - choose to cache IV if possible */
1986 if (SvTYPE(sv) == SVt_NV)
1987 sv_upgrade(sv, SVt_PVNV);
1989 (void)SvIOKp_on(sv); /* Must do this first, to clear any SvOOK */
1990 /* < not <= as for NV doesn't preserve UV, ((NV)IV_MAX+1) will almost
1991 certainly cast into the IV range at IV_MAX, whereas the correct
1992 answer is the UV IV_MAX +1. Hence < ensures that dodgy boundary
1994 #if defined(NAN_COMPARE_BROKEN) && defined(Perl_isnan)
1995 if (Perl_isnan(SvNVX(sv))) {
2001 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2002 SvIV_set(sv, I_V(SvNVX(sv)));
2003 if (SvNVX(sv) == (NV) SvIVX(sv)
2004 #ifndef NV_PRESERVES_UV
2005 && (((UV)1 << NV_PRESERVES_UV_BITS) >
2006 (UV)(SvIVX(sv) > 0 ? SvIVX(sv) : -SvIVX(sv)))
2007 /* Don't flag it as "accurately an integer" if the number
2008 came from a (by definition imprecise) NV operation, and
2009 we're outside the range of NV integer precision */
2013 SvIOK_on(sv); /* Can this go wrong with rounding? NWC */
2015 /* scalar has trailing garbage, eg "42a" */
2017 DEBUG_c(PerlIO_printf(Perl_debug_log,
2018 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (precise)\n",
2024 /* IV not precise. No need to convert from PV, as NV
2025 conversion would already have cached IV if it detected
2026 that PV->IV would be better than PV->NV->IV
2027 flags already correct - don't set public IOK. */
2028 DEBUG_c(PerlIO_printf(Perl_debug_log,
2029 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (imprecise)\n",
2034 /* Can the above go wrong if SvIVX == IV_MIN and SvNVX < IV_MIN,
2035 but the cast (NV)IV_MIN rounds to a the value less (more
2036 negative) than IV_MIN which happens to be equal to SvNVX ??
2037 Analogous to 0xFFFFFFFFFFFFFFFF rounding up to NV (2**64) and
2038 NV rounding back to 0xFFFFFFFFFFFFFFFF, so UVX == UV(NVX) and
2039 (NV)UVX == NVX are both true, but the values differ. :-(
2040 Hopefully for 2s complement IV_MIN is something like
2041 0x8000000000000000 which will be exact. NWC */
2044 SvUV_set(sv, U_V(SvNVX(sv)));
2046 (SvNVX(sv) == (NV) SvUVX(sv))
2047 #ifndef NV_PRESERVES_UV
2048 /* Make sure it's not 0xFFFFFFFFFFFFFFFF */
2049 /*&& (SvUVX(sv) != UV_MAX) irrelevant with code below */
2050 && (((UV)1 << NV_PRESERVES_UV_BITS) > SvUVX(sv))
2051 /* Don't flag it as "accurately an integer" if the number
2052 came from a (by definition imprecise) NV operation, and
2053 we're outside the range of NV integer precision */
2059 DEBUG_c(PerlIO_printf(Perl_debug_log,
2060 "0x%"UVxf" 2iv(%"UVuf" => %"IVdf") (as unsigned)\n",
2066 else if (SvPOKp(sv) && SvLEN(sv)) {
2068 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2069 /* We want to avoid a possible problem when we cache an IV/ a UV which
2070 may be later translated to an NV, and the resulting NV is not
2071 the same as the direct translation of the initial string
2072 (eg 123.456 can shortcut to the IV 123 with atol(), but we must
2073 be careful to ensure that the value with the .456 is around if the
2074 NV value is requested in the future).
2076 This means that if we cache such an IV/a UV, we need to cache the
2077 NV as well. Moreover, we trade speed for space, and do not
2078 cache the NV if we are sure it's not needed.
2081 /* SVt_PVNV is one higher than SVt_PVIV, hence this order */
2082 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2083 == IS_NUMBER_IN_UV) {
2084 /* It's definitely an integer, only upgrade to PVIV */
2085 if (SvTYPE(sv) < SVt_PVIV)
2086 sv_upgrade(sv, SVt_PVIV);
2088 } else if (SvTYPE(sv) < SVt_PVNV)
2089 sv_upgrade(sv, SVt_PVNV);
2091 /* If NVs preserve UVs then we only use the UV value if we know that
2092 we aren't going to call atof() below. If NVs don't preserve UVs
2093 then the value returned may have more precision than atof() will
2094 return, even though value isn't perfectly accurate. */
2095 if ((numtype & (IS_NUMBER_IN_UV
2096 #ifdef NV_PRESERVES_UV
2099 )) == IS_NUMBER_IN_UV) {
2100 /* This won't turn off the public IOK flag if it was set above */
2101 (void)SvIOKp_on(sv);
2103 if (!(numtype & IS_NUMBER_NEG)) {
2105 if (value <= (UV)IV_MAX) {
2106 SvIV_set(sv, (IV)value);
2108 /* it didn't overflow, and it was positive. */
2109 SvUV_set(sv, value);
2113 /* 2s complement assumption */
2114 if (value <= (UV)IV_MIN) {
2115 SvIV_set(sv, -(IV)value);
2117 /* Too negative for an IV. This is a double upgrade, but
2118 I'm assuming it will be rare. */
2119 if (SvTYPE(sv) < SVt_PVNV)
2120 sv_upgrade(sv, SVt_PVNV);
2124 SvNV_set(sv, -(NV)value);
2125 SvIV_set(sv, IV_MIN);
2129 /* For !NV_PRESERVES_UV and IS_NUMBER_IN_UV and IS_NUMBER_NOT_INT we
2130 will be in the previous block to set the IV slot, and the next
2131 block to set the NV slot. So no else here. */
2133 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2134 != IS_NUMBER_IN_UV) {
2135 /* It wasn't an (integer that doesn't overflow the UV). */
2136 SvNV_set(sv, Atof(SvPVX_const(sv)));
2138 if (! numtype && ckWARN(WARN_NUMERIC))
2141 #if defined(USE_LONG_DOUBLE)
2142 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%" PERL_PRIgldbl ")\n",
2143 PTR2UV(sv), SvNVX(sv)));
2145 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"NVgf")\n",
2146 PTR2UV(sv), SvNVX(sv)));
2149 #ifdef NV_PRESERVES_UV
2150 (void)SvIOKp_on(sv);
2152 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2153 SvIV_set(sv, I_V(SvNVX(sv)));
2154 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
2157 NOOP; /* Integer is imprecise. NOK, IOKp */
2159 /* UV will not work better than IV */
2161 if (SvNVX(sv) > (NV)UV_MAX) {
2163 /* Integer is inaccurate. NOK, IOKp, is UV */
2164 SvUV_set(sv, UV_MAX);
2166 SvUV_set(sv, U_V(SvNVX(sv)));
2167 /* 0xFFFFFFFFFFFFFFFF not an issue in here, NVs
2168 NV preservse UV so can do correct comparison. */
2169 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
2172 NOOP; /* Integer is imprecise. NOK, IOKp, is UV */
2177 #else /* NV_PRESERVES_UV */
2178 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2179 == (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT)) {
2180 /* The IV/UV slot will have been set from value returned by
2181 grok_number above. The NV slot has just been set using
2184 assert (SvIOKp(sv));
2186 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2187 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2188 /* Small enough to preserve all bits. */
2189 (void)SvIOKp_on(sv);
2191 SvIV_set(sv, I_V(SvNVX(sv)));
2192 if ((NV)(SvIVX(sv)) == SvNVX(sv))
2194 /* Assumption: first non-preserved integer is < IV_MAX,
2195 this NV is in the preserved range, therefore: */
2196 if (!(U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))
2198 Perl_croak(aTHX_ "sv_2iv assumed (U_V(fabs((double)SvNVX(sv))) < (UV)IV_MAX) but SvNVX(sv)=%"NVgf" U_V is 0x%"UVxf", IV_MAX is 0x%"UVxf"\n", SvNVX(sv), U_V(SvNVX(sv)), (UV)IV_MAX);
2202 0 0 already failed to read UV.
2203 0 1 already failed to read UV.
2204 1 0 you won't get here in this case. IV/UV
2205 slot set, public IOK, Atof() unneeded.
2206 1 1 already read UV.
2207 so there's no point in sv_2iuv_non_preserve() attempting
2208 to use atol, strtol, strtoul etc. */
2210 sv_2iuv_non_preserve (sv, numtype);
2212 sv_2iuv_non_preserve (sv);
2216 #endif /* NV_PRESERVES_UV */
2217 /* It might be more code efficient to go through the entire logic above
2218 and conditionally set with SvIOKp_on() rather than SvIOK(), but it
2219 gets complex and potentially buggy, so more programmer efficient
2220 to do it this way, by turning off the public flags: */
2222 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK);
2226 if (isGV_with_GP(sv))
2227 return glob_2number(MUTABLE_GV(sv));
2229 if (!SvPADTMP(sv)) {
2230 if (!PL_localizing && ckWARN(WARN_UNINITIALIZED))
2233 if (SvTYPE(sv) < SVt_IV)
2234 /* Typically the caller expects that sv_any is not NULL now. */
2235 sv_upgrade(sv, SVt_IV);
2236 /* Return 0 from the caller. */
2243 =for apidoc sv_2iv_flags
2245 Return the integer value of an SV, doing any necessary string
2246 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2247 Normally used via the C<SvIV(sv)> and C<SvIVx(sv)> macros.
2253 Perl_sv_2iv_flags(pTHX_ register SV *const sv, const I32 flags)
2258 if (SvGMAGICAL(sv) || SvVALID(sv)) {
2259 /* FBMs use the space for SvIVX and SvNVX for other purposes, and use
2260 the same flag bit as SVf_IVisUV, so must not let them cache IVs.
2261 In practice they are extremely unlikely to actually get anywhere
2262 accessible by user Perl code - the only way that I'm aware of is when
2263 a constant subroutine which is used as the second argument to index.
2265 if (flags & SV_GMAGIC)
2270 return I_V(SvNVX(sv));
2272 if (SvPOKp(sv) && SvLEN(sv)) {
2275 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2277 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2278 == IS_NUMBER_IN_UV) {
2279 /* It's definitely an integer */
2280 if (numtype & IS_NUMBER_NEG) {
2281 if (value < (UV)IV_MIN)
2284 if (value < (UV)IV_MAX)
2289 if (ckWARN(WARN_NUMERIC))
2292 return I_V(Atof(SvPVX_const(sv)));
2297 assert(SvTYPE(sv) >= SVt_PVMG);
2298 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
2299 } else if (SvTHINKFIRST(sv)) {
2304 if (flags & SV_SKIP_OVERLOAD)
2306 tmpstr = AMG_CALLunary(sv, numer_amg);
2307 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2308 return SvIV(tmpstr);
2311 return PTR2IV(SvRV(sv));
2314 sv_force_normal_flags(sv, 0);
2316 if (SvREADONLY(sv) && !SvOK(sv)) {
2317 if (ckWARN(WARN_UNINITIALIZED))
2323 if (S_sv_2iuv_common(aTHX_ sv))
2326 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"IVdf")\n",
2327 PTR2UV(sv),SvIVX(sv)));
2328 return SvIsUV(sv) ? (IV)SvUVX(sv) : SvIVX(sv);
2332 =for apidoc sv_2uv_flags
2334 Return the unsigned integer value of an SV, doing any necessary string
2335 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2336 Normally used via the C<SvUV(sv)> and C<SvUVx(sv)> macros.
2342 Perl_sv_2uv_flags(pTHX_ register SV *const sv, const I32 flags)
2347 if (SvGMAGICAL(sv) || SvVALID(sv)) {
2348 /* FBMs use the space for SvIVX and SvNVX for other purposes, and use
2349 the same flag bit as SVf_IVisUV, so must not let them cache IVs. */
2350 if (flags & SV_GMAGIC)
2355 return U_V(SvNVX(sv));
2356 if (SvPOKp(sv) && SvLEN(sv)) {
2359 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2361 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2362 == IS_NUMBER_IN_UV) {
2363 /* It's definitely an integer */
2364 if (!(numtype & IS_NUMBER_NEG))
2368 if (ckWARN(WARN_NUMERIC))
2371 return U_V(Atof(SvPVX_const(sv)));
2376 assert(SvTYPE(sv) >= SVt_PVMG);
2377 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
2378 } else if (SvTHINKFIRST(sv)) {
2383 if (flags & SV_SKIP_OVERLOAD)
2385 tmpstr = AMG_CALLunary(sv, numer_amg);
2386 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2387 return SvUV(tmpstr);
2390 return PTR2UV(SvRV(sv));
2393 sv_force_normal_flags(sv, 0);
2395 if (SvREADONLY(sv) && !SvOK(sv)) {
2396 if (ckWARN(WARN_UNINITIALIZED))
2402 if (S_sv_2iuv_common(aTHX_ sv))
2406 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2uv(%"UVuf")\n",
2407 PTR2UV(sv),SvUVX(sv)));
2408 return SvIsUV(sv) ? SvUVX(sv) : (UV)SvIVX(sv);
2412 =for apidoc sv_2nv_flags
2414 Return the num value of an SV, doing any necessary string or integer
2415 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2416 Normally used via the C<SvNV(sv)> and C<SvNVx(sv)> macros.
2422 Perl_sv_2nv_flags(pTHX_ register SV *const sv, const I32 flags)
2427 if (SvGMAGICAL(sv) || SvVALID(sv)) {
2428 /* FBMs use the space for SvIVX and SvNVX for other purposes, and use
2429 the same flag bit as SVf_IVisUV, so must not let them cache NVs. */
2430 if (flags & SV_GMAGIC)
2434 if ((SvPOKp(sv) && SvLEN(sv)) && !SvIOKp(sv)) {
2435 if (!SvIOKp(sv) && ckWARN(WARN_NUMERIC) &&
2436 !grok_number(SvPVX_const(sv), SvCUR(sv), NULL))
2438 return Atof(SvPVX_const(sv));
2442 return (NV)SvUVX(sv);
2444 return (NV)SvIVX(sv);
2449 assert(SvTYPE(sv) >= SVt_PVMG);
2450 /* This falls through to the report_uninit near the end of the
2452 } else if (SvTHINKFIRST(sv)) {
2457 if (flags & SV_SKIP_OVERLOAD)
2459 tmpstr = AMG_CALLunary(sv, numer_amg);
2460 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2461 return SvNV(tmpstr);
2464 return PTR2NV(SvRV(sv));
2467 sv_force_normal_flags(sv, 0);
2469 if (SvREADONLY(sv) && !SvOK(sv)) {
2470 if (ckWARN(WARN_UNINITIALIZED))
2475 if (SvTYPE(sv) < SVt_NV) {
2476 /* The logic to use SVt_PVNV if necessary is in sv_upgrade. */
2477 sv_upgrade(sv, SVt_NV);
2478 #ifdef USE_LONG_DOUBLE
2480 STORE_NUMERIC_LOCAL_SET_STANDARD();
2481 PerlIO_printf(Perl_debug_log,
2482 "0x%"UVxf" num(%" PERL_PRIgldbl ")\n",
2483 PTR2UV(sv), SvNVX(sv));
2484 RESTORE_NUMERIC_LOCAL();
2488 STORE_NUMERIC_LOCAL_SET_STANDARD();
2489 PerlIO_printf(Perl_debug_log, "0x%"UVxf" num(%"NVgf")\n",
2490 PTR2UV(sv), SvNVX(sv));
2491 RESTORE_NUMERIC_LOCAL();
2495 else if (SvTYPE(sv) < SVt_PVNV)
2496 sv_upgrade(sv, SVt_PVNV);
2501 SvNV_set(sv, SvIsUV(sv) ? (NV)SvUVX(sv) : (NV)SvIVX(sv));
2502 #ifdef NV_PRESERVES_UV
2508 /* Only set the public NV OK flag if this NV preserves the IV */
2509 /* Check it's not 0xFFFFFFFFFFFFFFFF */
2511 SvIsUV(sv) ? ((SvUVX(sv) != UV_MAX)&&(SvUVX(sv) == U_V(SvNVX(sv))))
2512 : (SvIVX(sv) == I_V(SvNVX(sv))))
2518 else if (SvPOKp(sv) && SvLEN(sv)) {
2520 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2521 if (!SvIOKp(sv) && !numtype && ckWARN(WARN_NUMERIC))
2523 #ifdef NV_PRESERVES_UV
2524 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2525 == IS_NUMBER_IN_UV) {
2526 /* It's definitely an integer */
2527 SvNV_set(sv, (numtype & IS_NUMBER_NEG) ? -(NV)value : (NV)value);
2529 SvNV_set(sv, Atof(SvPVX_const(sv)));
2535 SvNV_set(sv, Atof(SvPVX_const(sv)));
2536 /* Only set the public NV OK flag if this NV preserves the value in
2537 the PV at least as well as an IV/UV would.
2538 Not sure how to do this 100% reliably. */
2539 /* if that shift count is out of range then Configure's test is
2540 wonky. We shouldn't be in here with NV_PRESERVES_UV_BITS ==
2542 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2543 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2544 SvNOK_on(sv); /* Definitely small enough to preserve all bits */
2545 } else if (!(numtype & IS_NUMBER_IN_UV)) {
2546 /* Can't use strtol etc to convert this string, so don't try.
2547 sv_2iv and sv_2uv will use the NV to convert, not the PV. */
2550 /* value has been set. It may not be precise. */
2551 if ((numtype & IS_NUMBER_NEG) && (value > (UV)IV_MIN)) {
2552 /* 2s complement assumption for (UV)IV_MIN */
2553 SvNOK_on(sv); /* Integer is too negative. */
2558 if (numtype & IS_NUMBER_NEG) {
2559 SvIV_set(sv, -(IV)value);
2560 } else if (value <= (UV)IV_MAX) {
2561 SvIV_set(sv, (IV)value);
2563 SvUV_set(sv, value);
2567 if (numtype & IS_NUMBER_NOT_INT) {
2568 /* I believe that even if the original PV had decimals,
2569 they are lost beyond the limit of the FP precision.
2570 However, neither is canonical, so both only get p
2571 flags. NWC, 2000/11/25 */
2572 /* Both already have p flags, so do nothing */
2574 const NV nv = SvNVX(sv);
2575 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2576 if (SvIVX(sv) == I_V(nv)) {
2579 /* It had no "." so it must be integer. */
2583 /* between IV_MAX and NV(UV_MAX).
2584 Could be slightly > UV_MAX */
2586 if (numtype & IS_NUMBER_NOT_INT) {
2587 /* UV and NV both imprecise. */
2589 const UV nv_as_uv = U_V(nv);
2591 if (value == nv_as_uv && SvUVX(sv) != UV_MAX) {
2600 /* It might be more code efficient to go through the entire logic above
2601 and conditionally set with SvNOKp_on() rather than SvNOK(), but it
2602 gets complex and potentially buggy, so more programmer efficient
2603 to do it this way, by turning off the public flags: */
2605 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK);
2606 #endif /* NV_PRESERVES_UV */
2609 if (isGV_with_GP(sv)) {
2610 glob_2number(MUTABLE_GV(sv));
2614 if (!PL_localizing && !SvPADTMP(sv) && ckWARN(WARN_UNINITIALIZED))
2616 assert (SvTYPE(sv) >= SVt_NV);
2617 /* Typically the caller expects that sv_any is not NULL now. */
2618 /* XXX Ilya implies that this is a bug in callers that assume this
2619 and ideally should be fixed. */
2622 #if defined(USE_LONG_DOUBLE)
2624 STORE_NUMERIC_LOCAL_SET_STANDARD();
2625 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2nv(%" PERL_PRIgldbl ")\n",
2626 PTR2UV(sv), SvNVX(sv));
2627 RESTORE_NUMERIC_LOCAL();
2631 STORE_NUMERIC_LOCAL_SET_STANDARD();
2632 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 1nv(%"NVgf")\n",
2633 PTR2UV(sv), SvNVX(sv));
2634 RESTORE_NUMERIC_LOCAL();
2643 Return an SV with the numeric value of the source SV, doing any necessary
2644 reference or overload conversion. You must use the C<SvNUM(sv)> macro to
2645 access this function.
2651 Perl_sv_2num(pTHX_ register SV *const sv)
2653 PERL_ARGS_ASSERT_SV_2NUM;
2658 SV * const tmpsv = AMG_CALLunary(sv, numer_amg);
2659 TAINT_IF(tmpsv && SvTAINTED(tmpsv));
2660 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
2661 return sv_2num(tmpsv);
2663 return sv_2mortal(newSVuv(PTR2UV(SvRV(sv))));
2666 /* uiv_2buf(): private routine for use by sv_2pv_flags(): print an IV or
2667 * UV as a string towards the end of buf, and return pointers to start and
2670 * We assume that buf is at least TYPE_CHARS(UV) long.
2674 S_uiv_2buf(char *const buf, const IV iv, UV uv, const int is_uv, char **const peob)
2676 char *ptr = buf + TYPE_CHARS(UV);
2677 char * const ebuf = ptr;
2680 PERL_ARGS_ASSERT_UIV_2BUF;
2692 *--ptr = '0' + (char)(uv % 10);
2701 =for apidoc sv_2pv_flags
2703 Returns a pointer to the string value of an SV, and sets *lp to its length.
2704 If flags includes SV_GMAGIC, does an mg_get() first. Coerces sv to a string
2706 Normally invoked via the C<SvPV_flags> macro. C<sv_2pv()> and C<sv_2pv_nomg>
2707 usually end up here too.
2713 Perl_sv_2pv_flags(pTHX_ register SV *const sv, STRLEN *const lp, const I32 flags)
2723 if (SvGMAGICAL(sv)) {
2724 if (flags & SV_GMAGIC)
2729 if (flags & SV_MUTABLE_RETURN)
2730 return SvPVX_mutable(sv);
2731 if (flags & SV_CONST_RETURN)
2732 return (char *)SvPVX_const(sv);
2735 if (SvIOKp(sv) || SvNOKp(sv)) {
2736 char tbuf[64]; /* Must fit sprintf/Gconvert of longest IV/NV */
2741 ? my_snprintf(tbuf, sizeof(tbuf), "%"UVuf, (UV)SvUVX(sv))
2742 : my_snprintf(tbuf, sizeof(tbuf), "%"IVdf, (IV)SvIVX(sv));
2743 } else if(SvNVX(sv) == 0.0) {
2748 Gconvert(SvNVX(sv), NV_DIG, 0, tbuf);
2755 SvUPGRADE(sv, SVt_PV);
2758 s = SvGROW_mutable(sv, len + 1);
2761 return (char*)memcpy(s, tbuf, len + 1);
2767 assert(SvTYPE(sv) >= SVt_PVMG);
2768 /* This falls through to the report_uninit near the end of the
2770 } else if (SvTHINKFIRST(sv)) {
2775 if (flags & SV_SKIP_OVERLOAD)
2777 tmpstr = AMG_CALLunary(sv, string_amg);
2778 TAINT_IF(tmpstr && SvTAINTED(tmpstr));
2779 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2781 /* char *pv = lp ? SvPV(tmpstr, *lp) : SvPV_nolen(tmpstr);
2785 if ((SvFLAGS(tmpstr) & (SVf_POK)) == SVf_POK) {
2786 if (flags & SV_CONST_RETURN) {
2787 pv = (char *) SvPVX_const(tmpstr);
2789 pv = (flags & SV_MUTABLE_RETURN)
2790 ? SvPVX_mutable(tmpstr) : SvPVX(tmpstr);
2793 *lp = SvCUR(tmpstr);
2795 pv = sv_2pv_flags(tmpstr, lp, flags);
2808 SV *const referent = SvRV(sv);
2812 retval = buffer = savepvn("NULLREF", len);
2813 } else if (SvTYPE(referent) == SVt_REGEXP) {
2814 REGEXP * const re = (REGEXP *)MUTABLE_PTR(referent);
2819 /* If the regex is UTF-8 we want the containing scalar to
2820 have an UTF-8 flag too */
2826 if ((seen_evals = RX_SEEN_EVALS(re)))
2827 PL_reginterp_cnt += seen_evals;
2830 *lp = RX_WRAPLEN(re);
2832 return RX_WRAPPED(re);
2834 const char *const typestr = sv_reftype(referent, 0);
2835 const STRLEN typelen = strlen(typestr);
2836 UV addr = PTR2UV(referent);
2837 const char *stashname = NULL;
2838 STRLEN stashnamelen = 0; /* hush, gcc */
2839 const char *buffer_end;
2841 if (SvOBJECT(referent)) {
2842 const HEK *const name = HvNAME_HEK(SvSTASH(referent));
2845 stashname = HEK_KEY(name);
2846 stashnamelen = HEK_LEN(name);
2848 if (HEK_UTF8(name)) {
2854 stashname = "__ANON__";
2857 len = stashnamelen + 1 /* = */ + typelen + 3 /* (0x */
2858 + 2 * sizeof(UV) + 2 /* )\0 */;
2860 len = typelen + 3 /* (0x */
2861 + 2 * sizeof(UV) + 2 /* )\0 */;
2864 Newx(buffer, len, char);
2865 buffer_end = retval = buffer + len;
2867 /* Working backwards */
2871 *--retval = PL_hexdigit[addr & 15];
2872 } while (addr >>= 4);
2878 memcpy(retval, typestr, typelen);
2882 retval -= stashnamelen;
2883 memcpy(retval, stashname, stashnamelen);
2885 /* retval may not necessarily have reached the start of the
2887 assert (retval >= buffer);
2889 len = buffer_end - retval - 1; /* -1 for that \0 */
2897 if (SvREADONLY(sv) && !SvOK(sv)) {
2900 if (flags & SV_UNDEF_RETURNS_NULL)
2902 if (ckWARN(WARN_UNINITIALIZED))
2907 if (SvIOK(sv) || ((SvIOKp(sv) && !SvNOKp(sv)))) {
2908 /* I'm assuming that if both IV and NV are equally valid then
2909 converting the IV is going to be more efficient */
2910 const U32 isUIOK = SvIsUV(sv);
2911 char buf[TYPE_CHARS(UV)];
2915 if (SvTYPE(sv) < SVt_PVIV)
2916 sv_upgrade(sv, SVt_PVIV);
2917 ptr = uiv_2buf(buf, SvIVX(sv), SvUVX(sv), isUIOK, &ebuf);
2919 /* inlined from sv_setpvn */
2920 s = SvGROW_mutable(sv, len + 1);
2921 Move(ptr, s, len, char);
2925 else if (SvNOKp(sv)) {
2926 if (SvTYPE(sv) < SVt_PVNV)
2927 sv_upgrade(sv, SVt_PVNV);
2928 if (SvNVX(sv) == 0.0) {
2929 s = SvGROW_mutable(sv, 2);
2934 /* The +20 is pure guesswork. Configure test needed. --jhi */
2935 s = SvGROW_mutable(sv, NV_DIG + 20);
2936 /* some Xenix systems wipe out errno here */
2937 Gconvert(SvNVX(sv), NV_DIG, 0, s);
2947 if (isGV_with_GP(sv)) {
2948 GV *const gv = MUTABLE_GV(sv);
2949 const U32 wasfake = SvFLAGS(gv) & SVf_FAKE;
2950 SV *const buffer = sv_newmortal();
2952 /* FAKE globs can get coerced, so need to turn this off temporarily
2955 gv_efullname3(buffer, gv, "*");
2956 SvFLAGS(gv) |= wasfake;
2958 if (SvPOK(buffer)) {
2960 *lp = SvCUR(buffer);
2962 if ( SvUTF8(buffer) ) SvUTF8_on(sv);
2963 return SvPVX(buffer);
2974 if (flags & SV_UNDEF_RETURNS_NULL)
2976 if (!PL_localizing && !SvPADTMP(sv) && ckWARN(WARN_UNINITIALIZED))
2978 if (SvTYPE(sv) < SVt_PV)
2979 /* Typically the caller expects that sv_any is not NULL now. */
2980 sv_upgrade(sv, SVt_PV);
2984 const STRLEN len = s - SvPVX_const(sv);
2990 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
2991 PTR2UV(sv),SvPVX_const(sv)));
2992 if (flags & SV_CONST_RETURN)
2993 return (char *)SvPVX_const(sv);
2994 if (flags & SV_MUTABLE_RETURN)
2995 return SvPVX_mutable(sv);
3000 =for apidoc sv_copypv
3002 Copies a stringified representation of the source SV into the
3003 destination SV. Automatically performs any necessary mg_get and
3004 coercion of numeric values into strings. Guaranteed to preserve
3005 UTF8 flag even from overloaded objects. Similar in nature to
3006 sv_2pv[_flags] but operates directly on an SV instead of just the
3007 string. Mostly uses sv_2pv_flags to do its work, except when that
3008 would lose the UTF-8'ness of the PV.
3014 Perl_sv_copypv(pTHX_ SV *const dsv, register SV *const ssv)
3017 const char * const s = SvPV_const(ssv,len);
3019 PERL_ARGS_ASSERT_SV_COPYPV;
3021 sv_setpvn(dsv,s,len);
3029 =for apidoc sv_2pvbyte
3031 Return a pointer to the byte-encoded representation of the SV, and set *lp
3032 to its length. May cause the SV to be downgraded from UTF-8 as a
3035 Usually accessed via the C<SvPVbyte> macro.
3041 Perl_sv_2pvbyte(pTHX_ register SV *const sv, STRLEN *const lp)
3043 PERL_ARGS_ASSERT_SV_2PVBYTE;
3046 sv_utf8_downgrade(sv,0);
3047 return lp ? SvPV_nomg(sv,*lp) : SvPV_nomg_nolen(sv);
3051 =for apidoc sv_2pvutf8
3053 Return a pointer to the UTF-8-encoded representation of the SV, and set *lp
3054 to its length. May cause the SV to be upgraded to UTF-8 as a side-effect.
3056 Usually accessed via the C<SvPVutf8> macro.
3062 Perl_sv_2pvutf8(pTHX_ register SV *const sv, STRLEN *const lp)
3064 PERL_ARGS_ASSERT_SV_2PVUTF8;
3066 sv_utf8_upgrade(sv);
3067 return lp ? SvPV(sv,*lp) : SvPV_nolen(sv);
3072 =for apidoc sv_2bool
3074 This macro is only used by sv_true() or its macro equivalent, and only if
3075 the latter's argument is neither SvPOK, SvIOK nor SvNOK.
3076 It calls sv_2bool_flags with the SV_GMAGIC flag.
3078 =for apidoc sv_2bool_flags
3080 This function is only used by sv_true() and friends, and only if
3081 the latter's argument is neither SvPOK, SvIOK nor SvNOK. If the flags
3082 contain SV_GMAGIC, then it does an mg_get() first.
3089 Perl_sv_2bool_flags(pTHX_ register SV *const sv, const I32 flags)
3093 PERL_ARGS_ASSERT_SV_2BOOL_FLAGS;
3095 if(flags & SV_GMAGIC) SvGETMAGIC(sv);
3101 SV * const tmpsv = AMG_CALLunary(sv, bool__amg);
3102 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
3103 return cBOOL(SvTRUE(tmpsv));
3105 return SvRV(sv) != 0;
3108 register XPV* const Xpvtmp = (XPV*)SvANY(sv);
3110 (*sv->sv_u.svu_pv > '0' ||
3111 Xpvtmp->xpv_cur > 1 ||
3112 (Xpvtmp->xpv_cur && *sv->sv_u.svu_pv != '0')))
3119 return SvIVX(sv) != 0;
3122 return SvNVX(sv) != 0.0;
3124 if (isGV_with_GP(sv))
3134 =for apidoc sv_utf8_upgrade
3136 Converts the PV of an SV to its UTF-8-encoded form.
3137 Forces the SV to string form if it is not already.
3138 Will C<mg_get> on C<sv> if appropriate.
3139 Always sets the SvUTF8 flag to avoid future validity checks even
3140 if the whole string is the same in UTF-8 as not.
3141 Returns the number of bytes in the converted string
3143 This is not as a general purpose byte encoding to Unicode interface:
3144 use the Encode extension for that.
3146 =for apidoc sv_utf8_upgrade_nomg
3148 Like sv_utf8_upgrade, but doesn't do magic on C<sv>
3150 =for apidoc sv_utf8_upgrade_flags
3152 Converts the PV of an SV to its UTF-8-encoded form.
3153 Forces the SV to string form if it is not already.
3154 Always sets the SvUTF8 flag to avoid future validity checks even
3155 if all the bytes are invariant in UTF-8. If C<flags> has C<SV_GMAGIC> bit set,
3156 will C<mg_get> on C<sv> if appropriate, else not.
3157 Returns the number of bytes in the converted string
3158 C<sv_utf8_upgrade> and
3159 C<sv_utf8_upgrade_nomg> are implemented in terms of this function.
3161 This is not as a general purpose byte encoding to Unicode interface:
3162 use the Encode extension for that.
3166 The grow version is currently not externally documented. It adds a parameter,
3167 extra, which is the number of unused bytes the string of 'sv' is guaranteed to
3168 have free after it upon return. This allows the caller to reserve extra space
3169 that it intends to fill, to avoid extra grows.
3171 Also externally undocumented for the moment is the flag SV_FORCE_UTF8_UPGRADE,
3172 which can be used to tell this function to not first check to see if there are
3173 any characters that are different in UTF-8 (variant characters) which would
3174 force it to allocate a new string to sv, but to assume there are. Typically
3175 this flag is used by a routine that has already parsed the string to find that
3176 there are such characters, and passes this information on so that the work
3177 doesn't have to be repeated.
3179 (One might think that the calling routine could pass in the position of the
3180 first such variant, so it wouldn't have to be found again. But that is not the
3181 case, because typically when the caller is likely to use this flag, it won't be
3182 calling this routine unless it finds something that won't fit into a byte.
3183 Otherwise it tries to not upgrade and just use bytes. But some things that
3184 do fit into a byte are variants in utf8, and the caller may not have been
3185 keeping track of these.)
3187 If the routine itself changes the string, it adds a trailing NUL. Such a NUL
3188 isn't guaranteed due to having other routines do the work in some input cases,
3189 or if the input is already flagged as being in utf8.
3191 The speed of this could perhaps be improved for many cases if someone wanted to
3192 write a fast function that counts the number of variant characters in a string,
3193 especially if it could return the position of the first one.
3198 Perl_sv_utf8_upgrade_flags_grow(pTHX_ register SV *const sv, const I32 flags, STRLEN extra)
3202 PERL_ARGS_ASSERT_SV_UTF8_UPGRADE_FLAGS_GROW;
3204 if (sv == &PL_sv_undef)
3208 if (SvREADONLY(sv) && (SvPOKp(sv) || SvIOKp(sv) || SvNOKp(sv))) {
3209 (void) sv_2pv_flags(sv,&len, flags);
3211 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3215 (void) SvPV_force_flags(sv,len,flags & SV_GMAGIC);
3220 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3225 sv_force_normal_flags(sv, 0);
3228 if (PL_encoding && !(flags & SV_UTF8_NO_ENCODING)) {
3229 sv_recode_to_utf8(sv, PL_encoding);
3230 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3234 if (SvCUR(sv) == 0) {
3235 if (extra) SvGROW(sv, extra);
3236 } else { /* Assume Latin-1/EBCDIC */
3237 /* This function could be much more efficient if we
3238 * had a FLAG in SVs to signal if there are any variant
3239 * chars in the PV. Given that there isn't such a flag
3240 * make the loop as fast as possible (although there are certainly ways
3241 * to speed this up, eg. through vectorization) */
3242 U8 * s = (U8 *) SvPVX_const(sv);
3243 U8 * e = (U8 *) SvEND(sv);
3245 STRLEN two_byte_count = 0;
3247 if (flags & SV_FORCE_UTF8_UPGRADE) goto must_be_utf8;
3249 /* See if really will need to convert to utf8. We mustn't rely on our
3250 * incoming SV being well formed and having a trailing '\0', as certain
3251 * code in pp_formline can send us partially built SVs. */
3255 if (NATIVE_IS_INVARIANT(ch)) continue;
3257 t--; /* t already incremented; re-point to first variant */
3262 /* utf8 conversion not needed because all are invariants. Mark as
3263 * UTF-8 even if no variant - saves scanning loop */
3265 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3270 /* Here, the string should be converted to utf8, either because of an
3271 * input flag (two_byte_count = 0), or because a character that
3272 * requires 2 bytes was found (two_byte_count = 1). t points either to
3273 * the beginning of the string (if we didn't examine anything), or to
3274 * the first variant. In either case, everything from s to t - 1 will
3275 * occupy only 1 byte each on output.
3277 * There are two main ways to convert. One is to create a new string
3278 * and go through the input starting from the beginning, appending each
3279 * converted value onto the new string as we go along. It's probably
3280 * best to allocate enough space in the string for the worst possible
3281 * case rather than possibly running out of space and having to
3282 * reallocate and then copy what we've done so far. Since everything
3283 * from s to t - 1 is invariant, the destination can be initialized
3284 * with these using a fast memory copy
3286 * The other way is to figure out exactly how big the string should be
3287 * by parsing the entire input. Then you don't have to make it big
3288 * enough to handle the worst possible case, and more importantly, if
3289 * the string you already have is large enough, you don't have to
3290 * allocate a new string, you can copy the last character in the input
3291 * string to the final position(s) that will be occupied by the
3292 * converted string and go backwards, stopping at t, since everything
3293 * before that is invariant.
3295 * There are advantages and disadvantages to each method.
3297 * In the first method, we can allocate a new string, do the memory
3298 * copy from the s to t - 1, and then proceed through the rest of the
3299 * string byte-by-byte.
3301 * In the second method, we proceed through the rest of the input
3302 * string just calculating how big the converted string will be. Then
3303 * there are two cases:
3304 * 1) if the string has enough extra space to handle the converted
3305 * value. We go backwards through the string, converting until we
3306 * get to the position we are at now, and then stop. If this
3307 * position is far enough along in the string, this method is
3308 * faster than the other method. If the memory copy were the same
3309 * speed as the byte-by-byte loop, that position would be about
3310 * half-way, as at the half-way mark, parsing to the end and back
3311 * is one complete string's parse, the same amount as starting
3312 * over and going all the way through. Actually, it would be
3313 * somewhat less than half-way, as it's faster to just count bytes
3314 * than to also copy, and we don't have the overhead of allocating
3315 * a new string, changing the scalar to use it, and freeing the
3316 * existing one. But if the memory copy is fast, the break-even
3317 * point is somewhere after half way. The counting loop could be
3318 * sped up by vectorization, etc, to move the break-even point
3319 * further towards the beginning.
3320 * 2) if the string doesn't have enough space to handle the converted
3321 * value. A new string will have to be allocated, and one might
3322 * as well, given that, start from the beginning doing the first
3323 * method. We've spent extra time parsing the string and in
3324 * exchange all we've gotten is that we know precisely how big to
3325 * make the new one. Perl is more optimized for time than space,
3326 * so this case is a loser.
3327 * So what I've decided to do is not use the 2nd method unless it is
3328 * guaranteed that a new string won't have to be allocated, assuming
3329 * the worst case. I also decided not to put any more conditions on it
3330 * than this, for now. It seems likely that, since the worst case is
3331 * twice as big as the unknown portion of the string (plus 1), we won't
3332 * be guaranteed enough space, causing us to go to the first method,
3333 * unless the string is short, or the first variant character is near
3334 * the end of it. In either of these cases, it seems best to use the
3335 * 2nd method. The only circumstance I can think of where this would
3336 * be really slower is if the string had once had much more data in it
3337 * than it does now, but there is still a substantial amount in it */
3340 STRLEN invariant_head = t - s;
3341 STRLEN size = invariant_head + (e - t) * 2 + 1 + extra;
3342 if (SvLEN(sv) < size) {
3344 /* Here, have decided to allocate a new string */
3349 Newx(dst, size, U8);
3351 /* If no known invariants at the beginning of the input string,
3352 * set so starts from there. Otherwise, can use memory copy to
3353 * get up to where we are now, and then start from here */
3355 if (invariant_head <= 0) {
3358 Copy(s, dst, invariant_head, char);
3359 d = dst + invariant_head;
3363 const UV uv = NATIVE8_TO_UNI(*t++);
3364 if (UNI_IS_INVARIANT(uv))
3365 *d++ = (U8)UNI_TO_NATIVE(uv);
3367 *d++ = (U8)UTF8_EIGHT_BIT_HI(uv);
3368 *d++ = (U8)UTF8_EIGHT_BIT_LO(uv);
3372 SvPV_free(sv); /* No longer using pre-existing string */
3373 SvPV_set(sv, (char*)dst);
3374 SvCUR_set(sv, d - dst);
3375 SvLEN_set(sv, size);
3378 /* Here, have decided to get the exact size of the string.
3379 * Currently this happens only when we know that there is
3380 * guaranteed enough space to fit the converted string, so
3381 * don't have to worry about growing. If two_byte_count is 0,
3382 * then t points to the first byte of the string which hasn't
3383 * been examined yet. Otherwise two_byte_count is 1, and t
3384 * points to the first byte in the string that will expand to
3385 * two. Depending on this, start examining at t or 1 after t.
3388 U8 *d = t + two_byte_count;
3391 /* Count up the remaining bytes that expand to two */
3394 const U8 chr = *d++;
3395 if (! NATIVE_IS_INVARIANT(chr)) two_byte_count++;
3398 /* The string will expand by just the number of bytes that
3399 * occupy two positions. But we are one afterwards because of
3400 * the increment just above. This is the place to put the
3401 * trailing NUL, and to set the length before we decrement */
3403 d += two_byte_count;
3404 SvCUR_set(sv, d - s);
3408 /* Having decremented d, it points to the position to put the
3409 * very last byte of the expanded string. Go backwards through
3410 * the string, copying and expanding as we go, stopping when we
3411 * get to the part that is invariant the rest of the way down */
3415 const U8 ch = NATIVE8_TO_UNI(*e--);
3416 if (UNI_IS_INVARIANT(ch)) {
3417 *d-- = UNI_TO_NATIVE(ch);
3419 *d-- = (U8)UTF8_EIGHT_BIT_LO(ch);
3420 *d-- = (U8)UTF8_EIGHT_BIT_HI(ch);
3425 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
3426 /* Update pos. We do it at the end rather than during
3427 * the upgrade, to avoid slowing down the common case
3428 * (upgrade without pos) */
3429 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
3431 I32 pos = mg->mg_len;
3432 if (pos > 0 && (U32)pos > invariant_head) {
3433 U8 *d = (U8*) SvPVX(sv) + invariant_head;
3434 STRLEN n = (U32)pos - invariant_head;
3436 if (UTF8_IS_START(*d))
3441 mg->mg_len = d - (U8*)SvPVX(sv);
3444 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
3445 magic_setutf8(sv,mg); /* clear UTF8 cache */
3450 /* Mark as UTF-8 even if no variant - saves scanning loop */
3456 =for apidoc sv_utf8_downgrade
3458 Attempts to convert the PV of an SV from characters to bytes.
3459 If the PV contains a character that cannot fit
3460 in a byte, this conversion will fail;
3461 in this case, either returns false or, if C<fail_ok> is not
3464 This is not as a general purpose Unicode to byte encoding interface:
3465 use the Encode extension for that.
3471 Perl_sv_utf8_downgrade(pTHX_ register SV *const sv, const bool fail_ok)
3475 PERL_ARGS_ASSERT_SV_UTF8_DOWNGRADE;
3477 if (SvPOKp(sv) && SvUTF8(sv)) {
3481 int mg_flags = SV_GMAGIC;
3484 sv_force_normal_flags(sv, 0);
3486 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
3488 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
3490 I32 pos = mg->mg_len;
3492 sv_pos_b2u(sv, &pos);
3493 mg_flags = 0; /* sv_pos_b2u does get magic */
3497 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
3498 magic_setutf8(sv,mg); /* clear UTF8 cache */
3501 s = (U8 *) SvPV_flags(sv, len, mg_flags);
3503 if (!utf8_to_bytes(s, &len)) {
3508 Perl_croak(aTHX_ "Wide character in %s",
3511 Perl_croak(aTHX_ "Wide character");
3522 =for apidoc sv_utf8_encode
3524 Converts the PV of an SV to UTF-8, but then turns the C<SvUTF8>
3525 flag off so that it looks like octets again.
3531 Perl_sv_utf8_encode(pTHX_ register SV *const sv)
3533 PERL_ARGS_ASSERT_SV_UTF8_ENCODE;
3536 sv_force_normal_flags(sv, 0);
3538 if (SvREADONLY(sv)) {
3539 Perl_croak_no_modify(aTHX);
3541 (void) sv_utf8_upgrade(sv);
3546 =for apidoc sv_utf8_decode
3548 If the PV of the SV is an octet sequence in UTF-8
3549 and contains a multiple-byte character, the C<SvUTF8> flag is turned on
3550 so that it looks like a character. If the PV contains only single-byte
3551 characters, the C<SvUTF8> flag stays off.
3552 Scans PV for validity and returns false if the PV is invalid UTF-8.
3558 Perl_sv_utf8_decode(pTHX_ register SV *const sv)
3560 PERL_ARGS_ASSERT_SV_UTF8_DECODE;
3563 const U8 *start, *c;
3566 /* The octets may have got themselves encoded - get them back as
3569 if (!sv_utf8_downgrade(sv, TRUE))
3572 /* it is actually just a matter of turning the utf8 flag on, but
3573 * we want to make sure everything inside is valid utf8 first.
3575 c = start = (const U8 *) SvPVX_const(sv);
3576 if (!is_utf8_string(c, SvCUR(sv)+1))
3578 e = (const U8 *) SvEND(sv);
3581 if (!UTF8_IS_INVARIANT(ch)) {
3586 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
3587 /* adjust pos to the start of a UTF8 char sequence */
3588 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
3590 I32 pos = mg->mg_len;
3592 for (c = start + pos; c > start; c--) {
3593 if (UTF8_IS_START(*c))
3596 mg->mg_len = c - start;
3599 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
3600 magic_setutf8(sv,mg); /* clear UTF8 cache */
3607 =for apidoc sv_setsv
3609 Copies the contents of the source SV C<ssv> into the destination SV
3610 C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3611 function if the source SV needs to be reused. Does not handle 'set' magic.
3612 Loosely speaking, it performs a copy-by-value, obliterating any previous
3613 content of the destination.
3615 You probably want to use one of the assortment of wrappers, such as
3616 C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3617 C<SvSetMagicSV_nosteal>.
3619 =for apidoc sv_setsv_flags
3621 Copies the contents of the source SV C<ssv> into the destination SV
3622 C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3623 function if the source SV needs to be reused. Does not handle 'set' magic.
3624 Loosely speaking, it performs a copy-by-value, obliterating any previous
3625 content of the destination.
3626 If the C<flags> parameter has the C<SV_GMAGIC> bit set, will C<mg_get> on
3627 C<ssv> if appropriate, else not. If the C<flags> parameter has the
3628 C<NOSTEAL> bit set then the buffers of temps will not be stolen. <sv_setsv>
3629 and C<sv_setsv_nomg> are implemented in terms of this function.
3631 You probably want to use one of the assortment of wrappers, such as
3632 C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3633 C<SvSetMagicSV_nosteal>.
3635 This is the primary function for copying scalars, and most other
3636 copy-ish functions and macros use this underneath.
3642 S_glob_assign_glob(pTHX_ SV *const dstr, SV *const sstr, const int dtype)
3644 I32 mro_changes = 0; /* 1 = method, 2 = isa, 3 = recursive isa */
3645 HV *old_stash = NULL;
3647 PERL_ARGS_ASSERT_GLOB_ASSIGN_GLOB;
3649 if (dtype != SVt_PVGV && !isGV_with_GP(dstr)) {
3650 const char * const name = GvNAME(sstr);
3651 const STRLEN len = GvNAMELEN(sstr);
3653 if (dtype >= SVt_PV) {
3659 SvUPGRADE(dstr, SVt_PVGV);
3660 (void)SvOK_off(dstr);
3661 /* FIXME - why are we doing this, then turning it off and on again
3663 isGV_with_GP_on(dstr);
3665 GvSTASH(dstr) = GvSTASH(sstr);
3667 Perl_sv_add_backref(aTHX_ MUTABLE_SV(GvSTASH(dstr)), dstr);
3668 gv_name_set(MUTABLE_GV(dstr), name, len,
3669 GV_ADD | (GvNAMEUTF8(sstr) ? SVf_UTF8 : 0 ));
3670 SvFAKE_on(dstr); /* can coerce to non-glob */
3673 if(GvGP(MUTABLE_GV(sstr))) {
3674 /* If source has method cache entry, clear it */
3676 SvREFCNT_dec(GvCV(sstr));
3677 GvCV_set(sstr, NULL);
3680 /* If source has a real method, then a method is
3683 GvCV((const GV *)sstr) && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3689 /* If dest already had a real method, that's a change as well */
3691 !mro_changes && GvGP(MUTABLE_GV(dstr)) && GvCVu((const GV *)dstr)
3692 && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3697 /* We don’t need to check the name of the destination if it was not a
3698 glob to begin with. */
3699 if(dtype == SVt_PVGV) {
3700 const char * const name = GvNAME((const GV *)dstr);
3703 /* The stash may have been detached from the symbol table, so
3705 && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3706 && GvAV((const GV *)sstr)
3710 const STRLEN len = GvNAMELEN(dstr);
3711 if ((len > 1 && name[len-2] == ':' && name[len-1] == ':')
3712 || (len == 1 && name[0] == ':')) {
3715 /* Set aside the old stash, so we can reset isa caches on
3717 if((old_stash = GvHV(dstr)))
3718 /* Make sure we do not lose it early. */
3719 SvREFCNT_inc_simple_void_NN(
3720 sv_2mortal((SV *)old_stash)
3726 gp_free(MUTABLE_GV(dstr));
3727 isGV_with_GP_off(dstr);
3728 (void)SvOK_off(dstr);
3729 isGV_with_GP_on(dstr);
3730 GvINTRO_off(dstr); /* one-shot flag */
3731 GvGP_set(dstr, gp_ref(GvGP(sstr)));
3732 if (SvTAINTED(sstr))
3734 if (GvIMPORTED(dstr) != GVf_IMPORTED
3735 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3737 GvIMPORTED_on(dstr);
3740 if(mro_changes == 2) {
3742 SV * const sref = (SV *)GvAV((const GV *)dstr);
3743 if (SvSMAGICAL(sref) && (mg = mg_find(sref, PERL_MAGIC_isa))) {
3744 if (SvTYPE(mg->mg_obj) != SVt_PVAV) {
3745 AV * const ary = newAV();
3746 av_push(ary, mg->mg_obj); /* takes the refcount */
3747 mg->mg_obj = (SV *)ary;
3749 av_push((AV *)mg->mg_obj, SvREFCNT_inc_simple_NN(dstr));
3751 else sv_magic(sref, dstr, PERL_MAGIC_isa, NULL, 0);
3752 mro_isa_changed_in(GvSTASH(dstr));
3754 else if(mro_changes == 3) {
3755 HV * const stash = GvHV(dstr);
3756 if(old_stash ? (HV *)HvENAME_get(old_stash) : stash)
3762 else if(mro_changes) mro_method_changed_in(GvSTASH(dstr));
3767 S_glob_assign_ref(pTHX_ SV *const dstr, SV *const sstr)
3769 SV * const sref = SvREFCNT_inc(SvRV(sstr));
3771 const int intro = GvINTRO(dstr);
3774 const U32 stype = SvTYPE(sref);
3776 PERL_ARGS_ASSERT_GLOB_ASSIGN_REF;
3779 GvINTRO_off(dstr); /* one-shot flag */
3780 GvLINE(dstr) = CopLINE(PL_curcop);
3781 GvEGV(dstr) = MUTABLE_GV(dstr);
3786 location = (SV **) &(GvGP(dstr)->gp_cv); /* XXX bypassing GvCV_set */
3787 import_flag = GVf_IMPORTED_CV;
3790 location = (SV **) &GvHV(dstr);
3791 import_flag = GVf_IMPORTED_HV;
3794 location = (SV **) &GvAV(dstr);
3795 import_flag = GVf_IMPORTED_AV;
3798 location = (SV **) &GvIOp(dstr);
3801 location = (SV **) &GvFORM(dstr);
3804 location = &GvSV(dstr);
3805 import_flag = GVf_IMPORTED_SV;
3808 if (stype == SVt_PVCV) {
3809 /*if (GvCVGEN(dstr) && (GvCV(dstr) != (const CV *)sref || GvCVGEN(dstr))) {*/
3810 if (GvCVGEN(dstr)) {
3811 SvREFCNT_dec(GvCV(dstr));
3812 GvCV_set(dstr, NULL);
3813 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3816 SAVEGENERICSV(*location);
3820 if (stype == SVt_PVCV && (*location != sref || GvCVGEN(dstr))) {
3821 CV* const cv = MUTABLE_CV(*location);
3823 if (!GvCVGEN((const GV *)dstr) &&
3824 (CvROOT(cv) || CvXSUB(cv)))
3826 /* Redefining a sub - warning is mandatory if
3827 it was a const and its value changed. */
3828 if (CvCONST(cv) && CvCONST((const CV *)sref)
3830 == cv_const_sv((const CV *)sref)) {
3832 /* They are 2 constant subroutines generated from
3833 the same constant. This probably means that
3834 they are really the "same" proxy subroutine
3835 instantiated in 2 places. Most likely this is
3836 when a constant is exported twice. Don't warn.
3839 else if (ckWARN(WARN_REDEFINE)
3841 && (!CvCONST((const CV *)sref)
3842 || sv_cmp(cv_const_sv(cv),
3843 cv_const_sv((const CV *)
3845 Perl_warner(aTHX_ packWARN(WARN_REDEFINE),
3848 ? "Constant subroutine %"HEKf
3849 "::%"HEKf" redefined"
3850 : "Subroutine %"HEKf"::%"HEKf
3853 HvNAME_HEK(GvSTASH((const GV *)dstr))
3855 HEKfARG(GvENAME_HEK(MUTABLE_GV(dstr))));
3859 cv_ckproto_len_flags(cv, (const GV *)dstr,
3860 SvPOK(sref) ? CvPROTO(sref) : NULL,
3861 SvPOK(sref) ? CvPROTOLEN(sref) : 0,
3862 SvPOK(sref) ? SvUTF8(sref) : 0);
3864 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3865 GvASSUMECV_on(dstr);
3866 if(GvSTASH(dstr)) mro_method_changed_in(GvSTASH(dstr)); /* sub foo { 1 } sub bar { 2 } *bar = \&foo */
3869 if (import_flag && !(GvFLAGS(dstr) & import_flag)
3870 && CopSTASH_ne(PL_curcop, GvSTASH(dstr))) {
3871 GvFLAGS(dstr) |= import_flag;
3873 if (stype == SVt_PVHV) {
3874 const char * const name = GvNAME((GV*)dstr);
3875 const STRLEN len = GvNAMELEN(dstr);
3878 (len > 1 && name[len-2] == ':' && name[len-1] == ':')
3879 || (len == 1 && name[0] == ':')
3881 && (!dref || HvENAME_get(dref))
3884 (HV *)sref, (HV *)dref,
3890 stype == SVt_PVAV && sref != dref
3891 && strEQ(GvNAME((GV*)dstr), "ISA")
3892 /* The stash may have been detached from the symbol table, so
3893 check its name before doing anything. */
3894 && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3897 MAGIC * const omg = dref && SvSMAGICAL(dref)
3898 ? mg_find(dref, PERL_MAGIC_isa)
3900 if (SvSMAGICAL(sref) && (mg = mg_find(sref, PERL_MAGIC_isa))) {
3901 if (SvTYPE(mg->mg_obj) != SVt_PVAV) {
3902 AV * const ary = newAV();
3903 av_push(ary, mg->mg_obj); /* takes the refcount */
3904 mg->mg_obj = (SV *)ary;
3907 if (SvTYPE(omg->mg_obj) == SVt_PVAV) {
3908 SV **svp = AvARRAY((AV *)omg->mg_obj);
3909 I32 items = AvFILLp((AV *)omg->mg_obj) + 1;
3913 SvREFCNT_inc_simple_NN(*svp++)
3919 SvREFCNT_inc_simple_NN(omg->mg_obj)
3923 av_push((AV *)mg->mg_obj,SvREFCNT_inc_simple_NN(dstr));
3928 sref, omg ? omg->mg_obj : dstr, PERL_MAGIC_isa, NULL, 0
3930 mg = mg_find(sref, PERL_MAGIC_isa);
3932 /* Since the *ISA assignment could have affected more than
3933 one stash, don’t call mro_isa_changed_in directly, but let
3934 magic_clearisa do it for us, as it already has the logic for
3935 dealing with globs vs arrays of globs. */
3937 Perl_magic_clearisa(aTHX_ NULL, mg);
3942 if (SvTAINTED(sstr))
3948 Perl_sv_setsv_flags(pTHX_ SV *dstr, register SV* sstr, const I32 flags)
3951 register U32 sflags;
3953 register svtype stype;
3955 PERL_ARGS_ASSERT_SV_SETSV_FLAGS;
3960 if (SvIS_FREED(dstr)) {
3961 Perl_croak(aTHX_ "panic: attempt to copy value %" SVf
3962 " to a freed scalar %p", SVfARG(sstr), (void *)dstr);
3964 SV_CHECK_THINKFIRST_COW_DROP(dstr);
3966 sstr = &PL_sv_undef;
3967 if (SvIS_FREED(sstr)) {
3968 Perl_croak(aTHX_ "panic: attempt to copy freed scalar %p to %p",
3969 (void*)sstr, (void*)dstr);
3971 stype = SvTYPE(sstr);
3972 dtype = SvTYPE(dstr);
3974 (void)SvAMAGIC_off(dstr);
3977 /* need to nuke the magic */
3981 /* There's a lot of redundancy below but we're going for speed here */
3986 if (dtype != SVt_PVGV && dtype != SVt_PVLV) {
3987 (void)SvOK_off(dstr);
3995 sv_upgrade(dstr, SVt_IV);
3999 sv_upgrade(dstr, SVt_PVIV);
4003 goto end_of_first_switch;
4005 (void)SvIOK_only(dstr);
4006 SvIV_set(dstr, SvIVX(sstr));
4009 /* SvTAINTED can only be true if the SV has taint magic, which in
4010 turn means that the SV type is PVMG (or greater). This is the
4011 case statement for SVt_IV, so this cannot be true (whatever gcov
4013 assert(!SvTAINTED(sstr));
4018 if (dtype < SVt_PV && dtype != SVt_IV)
4019 sv_upgrade(dstr, SVt_IV);
4027 sv_upgrade(dstr, SVt_NV);
4031 sv_upgrade(dstr, SVt_PVNV);
4035 goto end_of_first_switch;
4037 SvNV_set(dstr, SvNVX(sstr));
4038 (void)SvNOK_only(dstr);
4039 /* SvTAINTED can only be true if the SV has taint magic, which in
4040 turn means that the SV type is PVMG (or greater). This is the
4041 case statement for SVt_NV, so this cannot be true (whatever gcov
4043 assert(!SvTAINTED(sstr));
4049 #ifdef PERL_OLD_COPY_ON_WRITE
4050 if ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS) {
4051 if (dtype < SVt_PVIV)
4052 sv_upgrade(dstr, SVt_PVIV);
4059 sv_upgrade(dstr, SVt_PV);
4062 if (dtype < SVt_PVIV)
4063 sv_upgrade(dstr, SVt_PVIV);
4066 if (dtype < SVt_PVNV)
4067 sv_upgrade(dstr, SVt_PVNV);
4071 const char * const type = sv_reftype(sstr,0);
4073 Perl_croak(aTHX_ "Bizarre copy of %s in %s", type, OP_DESC(PL_op));
4075 Perl_croak(aTHX_ "Bizarre copy of %s", type);
4080 if (dtype < SVt_REGEXP)
4081 sv_upgrade(dstr, SVt_REGEXP);
4084 /* case SVt_BIND: */
4088 if (SvGMAGICAL(sstr) && (flags & SV_GMAGIC)) {
4090 if (SvTYPE(sstr) != stype)
4091 stype = SvTYPE(sstr);
4093 if (isGV_with_GP(sstr) && dtype <= SVt_PVLV) {
4094 glob_assign_glob(dstr, sstr, dtype);
4097 if (stype == SVt_PVLV)
4098 SvUPGRADE(dstr, SVt_PVNV);
4100 SvUPGRADE(dstr, (svtype)stype);
4102 end_of_first_switch:
4104 /* dstr may have been upgraded. */
4105 dtype = SvTYPE(dstr);
4106 sflags = SvFLAGS(sstr);
4108 if (dtype == SVt_PVCV || dtype == SVt_PVFM) {
4109 /* Assigning to a subroutine sets the prototype. */
4112 const char *const ptr = SvPV_const(sstr, len);
4114 SvGROW(dstr, len + 1);
4115 Copy(ptr, SvPVX(dstr), len + 1, char);
4116 SvCUR_set(dstr, len);
4118 SvFLAGS(dstr) |= sflags & SVf_UTF8;
4119 CvAUTOLOAD_off(dstr);
4123 } else if (dtype == SVt_PVAV || dtype == SVt_PVHV) {
4124 const char * const type = sv_reftype(dstr,0);
4126 Perl_croak(aTHX_ "Cannot copy to %s in %s", type, OP_DESC(PL_op));
4128 Perl_croak(aTHX_ "Cannot copy to %s", type);
4129 } else if (sflags & SVf_ROK) {
4130 if (isGV_with_GP(dstr)
4131 && SvTYPE(SvRV(sstr)) == SVt_PVGV && isGV_with_GP(SvRV(sstr))) {
4134 if (GvIMPORTED(dstr) != GVf_IMPORTED
4135 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
4137 GvIMPORTED_on(dstr);
4142 glob_assign_glob(dstr, sstr, dtype);
4146 if (dtype >= SVt_PV) {
4147 if (isGV_with_GP(dstr)) {
4148 glob_assign_ref(dstr, sstr);
4151 if (SvPVX_const(dstr)) {
4157 (void)SvOK_off(dstr);
4158 SvRV_set(dstr, SvREFCNT_inc(SvRV(sstr)));
4159 SvFLAGS(dstr) |= sflags & SVf_ROK;
4160 assert(!(sflags & SVp_NOK));
4161 assert(!(sflags & SVp_IOK));
4162 assert(!(sflags & SVf_NOK));
4163 assert(!(sflags & SVf_IOK));
4165 else if (isGV_with_GP(dstr)) {
4166 if (!(sflags & SVf_OK)) {
4167 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
4168 "Undefined value assigned to typeglob");
4171 GV *gv = gv_fetchsv_nomg(sstr, GV_ADD, SVt_PVGV);
4172 if (dstr != (const SV *)gv) {
4173 const char * const name = GvNAME((const GV *)dstr);
4174 const STRLEN len = GvNAMELEN(dstr);
4175 HV *old_stash = NULL;
4176 bool reset_isa = FALSE;
4177 if ((len > 1 && name[len-2] == ':' && name[len-1] == ':')
4178 || (len == 1 && name[0] == ':')) {
4179 /* Set aside the old stash, so we can reset isa caches
4180 on its subclasses. */
4181 if((old_stash = GvHV(dstr))) {
4182 /* Make sure we do not lose it early. */
4183 SvREFCNT_inc_simple_void_NN(
4184 sv_2mortal((SV *)old_stash)
4191 gp_free(MUTABLE_GV(dstr));
4192 GvGP_set(dstr, gp_ref(GvGP(gv)));
4195 HV * const stash = GvHV(dstr);
4197 old_stash ? (HV *)HvENAME_get(old_stash) : stash
4207 else if (dtype == SVt_REGEXP && stype == SVt_REGEXP) {
4208 reg_temp_copy((REGEXP*)dstr, (REGEXP*)sstr);
4210 else if (sflags & SVp_POK) {
4214 * Check to see if we can just swipe the string. If so, it's a
4215 * possible small lose on short strings, but a big win on long ones.
4216 * It might even be a win on short strings if SvPVX_const(dstr)
4217 * has to be allocated and SvPVX_const(sstr) has to be freed.
4218 * Likewise if we can set up COW rather than doing an actual copy, we
4219 * drop to the else clause, as the swipe code and the COW setup code
4220 * have much in common.
4223 /* Whichever path we take through the next code, we want this true,
4224 and doing it now facilitates the COW check. */
4225 (void)SvPOK_only(dstr);
4228 /* If we're already COW then this clause is not true, and if COW
4229 is allowed then we drop down to the else and make dest COW
4230 with us. If caller hasn't said that we're allowed to COW
4231 shared hash keys then we don't do the COW setup, even if the
4232 source scalar is a shared hash key scalar. */
4233 (((flags & SV_COW_SHARED_HASH_KEYS)
4234 ? (sflags & (SVf_FAKE|SVf_READONLY)) != (SVf_FAKE|SVf_READONLY)
4235 : 1 /* If making a COW copy is forbidden then the behaviour we
4236 desire is as if the source SV isn't actually already
4237 COW, even if it is. So we act as if the source flags
4238 are not COW, rather than actually testing them. */
4240 #ifndef PERL_OLD_COPY_ON_WRITE
4241 /* The change that added SV_COW_SHARED_HASH_KEYS makes the logic
4242 when PERL_OLD_COPY_ON_WRITE is defined a little wrong.
4243 Conceptually PERL_OLD_COPY_ON_WRITE being defined should
4244 override SV_COW_SHARED_HASH_KEYS, because it means "always COW"
4245 but in turn, it's somewhat dead code, never expected to go
4246 live, but more kept as a placeholder on how to do it better
4247 in a newer implementation. */
4248 /* If we are COW and dstr is a suitable target then we drop down
4249 into the else and make dest a COW of us. */
4250 || (SvFLAGS(dstr) & CAN_COW_MASK) != CAN_COW_FLAGS
4255 (sflags & SVs_TEMP) && /* slated for free anyway? */
4256 !(sflags & SVf_OOK) && /* and not involved in OOK hack? */
4257 (!(flags & SV_NOSTEAL)) &&
4258 /* and we're allowed to steal temps */
4259 SvREFCNT(sstr) == 1 && /* and no other references to it? */
4260 SvLEN(sstr)) /* and really is a string */
4261 #ifdef PERL_OLD_COPY_ON_WRITE
4262 && ((flags & SV_COW_SHARED_HASH_KEYS)
4263 ? (!((sflags & CAN_COW_MASK) == CAN_COW_FLAGS
4264 && (SvFLAGS(dstr) & CAN_COW_MASK) == CAN_COW_FLAGS
4265 && SvTYPE(sstr) >= SVt_PVIV && SvTYPE(sstr) != SVt_PVFM))
4269 /* Failed the swipe test, and it's not a shared hash key either.
4270 Have to copy the string. */
4271 STRLEN len = SvCUR(sstr);
4272 SvGROW(dstr, len + 1); /* inlined from sv_setpvn */
4273 Move(SvPVX_const(sstr),SvPVX(dstr),len,char);
4274 SvCUR_set(dstr, len);
4275 *SvEND(dstr) = '\0';
4277 /* If PERL_OLD_COPY_ON_WRITE is not defined, then isSwipe will always
4279 /* Either it's a shared hash key, or it's suitable for
4280 copy-on-write or we can swipe the string. */
4282 PerlIO_printf(Perl_debug_log, "Copy on write: sstr --> dstr\n");
4286 #ifdef PERL_OLD_COPY_ON_WRITE
4288 if ((sflags & (SVf_FAKE | SVf_READONLY))
4289 != (SVf_FAKE | SVf_READONLY)) {
4290 SvREADONLY_on(sstr);
4292 /* Make the source SV into a loop of 1.
4293 (about to become 2) */
4294 SV_COW_NEXT_SV_SET(sstr, sstr);
4298 /* Initial code is common. */
4299 if (SvPVX_const(dstr)) { /* we know that dtype >= SVt_PV */
4304 /* making another shared SV. */
4305 STRLEN cur = SvCUR(sstr);
4306 STRLEN len = SvLEN(sstr);
4307 #ifdef PERL_OLD_COPY_ON_WRITE
4309 assert (SvTYPE(dstr) >= SVt_PVIV);
4310 /* SvIsCOW_normal */
4311 /* splice us in between source and next-after-source. */
4312 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4313 SV_COW_NEXT_SV_SET(sstr, dstr);
4314 SvPV_set(dstr, SvPVX_mutable(sstr));
4318 /* SvIsCOW_shared_hash */
4319 DEBUG_C(PerlIO_printf(Perl_debug_log,
4320 "Copy on write: Sharing hash\n"));
4322 assert (SvTYPE(dstr) >= SVt_PV);
4324 HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)))));
4326 SvLEN_set(dstr, len);
4327 SvCUR_set(dstr, cur);
4328 SvREADONLY_on(dstr);
4332 { /* Passes the swipe test. */
4333 SvPV_set(dstr, SvPVX_mutable(sstr));
4334 SvLEN_set(dstr, SvLEN(sstr));
4335 SvCUR_set(dstr, SvCUR(sstr));
4338 (void)SvOK_off(sstr); /* NOTE: nukes most SvFLAGS on sstr */
4339 SvPV_set(sstr, NULL);
4345 if (sflags & SVp_NOK) {
4346 SvNV_set(dstr, SvNVX(sstr));
4348 if (sflags & SVp_IOK) {
4349 SvIV_set(dstr, SvIVX(sstr));
4350 /* Must do this otherwise some other overloaded use of 0x80000000
4351 gets confused. I guess SVpbm_VALID */
4352 if (sflags & SVf_IVisUV)
4355 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_NOK|SVp_NOK|SVf_UTF8);
4357 const MAGIC * const smg = SvVSTRING_mg(sstr);
4359 sv_magic(dstr, NULL, PERL_MAGIC_vstring,
4360 smg->mg_ptr, smg->mg_len);
4361 SvRMAGICAL_on(dstr);
4365 else if (sflags & (SVp_IOK|SVp_NOK)) {
4366 (void)SvOK_off(dstr);
4367 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_IVisUV|SVf_NOK|SVp_NOK);
4368 if (sflags & SVp_IOK) {
4369 /* XXXX Do we want to set IsUV for IV(ROK)? Be extra safe... */
4370 SvIV_set(dstr, SvIVX(sstr));
4372 if (sflags & SVp_NOK) {
4373 SvNV_set(dstr, SvNVX(sstr));
4377 if (isGV_with_GP(sstr)) {
4378 /* This stringification rule for globs is spread in 3 places.
4379 This feels bad. FIXME. */
4380 const U32 wasfake = sflags & SVf_FAKE;
4382 /* FAKE globs can get coerced, so need to turn this off
4383 temporarily if it is on. */
4385 gv_efullname3(dstr, MUTABLE_GV(sstr), "*");
4386 SvFLAGS(sstr) |= wasfake;
4389 (void)SvOK_off(dstr);
4391 if (SvTAINTED(sstr))
4396 =for apidoc sv_setsv_mg
4398 Like C<sv_setsv>, but also handles 'set' magic.
4404 Perl_sv_setsv_mg(pTHX_ SV *const dstr, register SV *const sstr)
4406 PERL_ARGS_ASSERT_SV_SETSV_MG;
4408 sv_setsv(dstr,sstr);
4412 #ifdef PERL_OLD_COPY_ON_WRITE
4414 Perl_sv_setsv_cow(pTHX_ SV *dstr, SV *sstr)
4416 STRLEN cur = SvCUR(sstr);
4417 STRLEN len = SvLEN(sstr);
4418 register char *new_pv;
4420 PERL_ARGS_ASSERT_SV_SETSV_COW;
4423 PerlIO_printf(Perl_debug_log, "Fast copy on write: %p -> %p\n",
4424 (void*)sstr, (void*)dstr);
4431 if (SvTHINKFIRST(dstr))
4432 sv_force_normal_flags(dstr, SV_COW_DROP_PV);
4433 else if (SvPVX_const(dstr))
4434 Safefree(SvPVX_const(dstr));
4438 SvUPGRADE(dstr, SVt_PVIV);
4440 assert (SvPOK(sstr));
4441 assert (SvPOKp(sstr));
4442 assert (!SvIOK(sstr));
4443 assert (!SvIOKp(sstr));
4444 assert (!SvNOK(sstr));
4445 assert (!SvNOKp(sstr));
4447 if (SvIsCOW(sstr)) {
4449 if (SvLEN(sstr) == 0) {
4450 /* source is a COW shared hash key. */
4451 DEBUG_C(PerlIO_printf(Perl_debug_log,
4452 "Fast copy on write: Sharing hash\n"));
4453 new_pv = HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr))));
4456 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4458 assert ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS);
4459 SvUPGRADE(sstr, SVt_PVIV);
4460 SvREADONLY_on(sstr);
4462 DEBUG_C(PerlIO_printf(Perl_debug_log,
4463 "Fast copy on write: Converting sstr to COW\n"));
4464 SV_COW_NEXT_SV_SET(dstr, sstr);
4466 SV_COW_NEXT_SV_SET(sstr, dstr);
4467 new_pv = SvPVX_mutable(sstr);
4470 SvPV_set(dstr, new_pv);
4471 SvFLAGS(dstr) = (SVt_PVIV|SVf_POK|SVp_POK|SVf_FAKE|SVf_READONLY);
4474 SvLEN_set(dstr, len);
4475 SvCUR_set(dstr, cur);
4484 =for apidoc sv_setpvn
4486 Copies a string into an SV. The C<len> parameter indicates the number of
4487 bytes to be copied. If the C<ptr> argument is NULL the SV will become
4488 undefined. Does not handle 'set' magic. See C<sv_setpvn_mg>.
4494 Perl_sv_setpvn(pTHX_ register SV *const sv, register const char *const ptr, register const STRLEN len)
4497 register char *dptr;
4499 PERL_ARGS_ASSERT_SV_SETPVN;
4501 SV_CHECK_THINKFIRST_COW_DROP(sv);
4507 /* len is STRLEN which is unsigned, need to copy to signed */
4510 Perl_croak(aTHX_ "panic: sv_setpvn called with negative strlen");
4512 SvUPGRADE(sv, SVt_PV);
4514 dptr = SvGROW(sv, len + 1);
4515 Move(ptr,dptr,len,char);
4518 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4520 if (SvTYPE(sv) == SVt_PVCV) CvAUTOLOAD_off(sv);
4524 =for apidoc sv_setpvn_mg
4526 Like C<sv_setpvn>, but also handles 'set' magic.
4532 Perl_sv_setpvn_mg(pTHX_ register SV *const sv, register const char *const ptr, register const STRLEN len)
4534 PERL_ARGS_ASSERT_SV_SETPVN_MG;
4536 sv_setpvn(sv,ptr,len);
4541 =for apidoc sv_setpv
4543 Copies a string into an SV. The string must be null-terminated. Does not
4544 handle 'set' magic. See C<sv_setpv_mg>.
4550 Perl_sv_setpv(pTHX_ register SV *const sv, register const char *const ptr)
4553 register STRLEN len;
4555 PERL_ARGS_ASSERT_SV_SETPV;
4557 SV_CHECK_THINKFIRST_COW_DROP(sv);
4563 SvUPGRADE(sv, SVt_PV);
4565 SvGROW(sv, len + 1);
4566 Move(ptr,SvPVX(sv),len+1,char);
4568 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4570 if (SvTYPE(sv) == SVt_PVCV) CvAUTOLOAD_off(sv);
4574 =for apidoc sv_setpv_mg
4576 Like C<sv_setpv>, but also handles 'set' magic.
4582 Perl_sv_setpv_mg(pTHX_ register SV *const sv, register const char *const ptr)
4584 PERL_ARGS_ASSERT_SV_SETPV_MG;
4591 Perl_sv_sethek(pTHX_ register SV *const sv, const HEK *const hek)
4595 PERL_ARGS_ASSERT_SV_SETHEK;
4601 if (HEK_LEN(hek) == HEf_SVKEY) {
4602 sv_setsv(sv, *(SV**)HEK_KEY(hek));
4605 const int flags = HEK_FLAGS(hek);
4606 if (flags & HVhek_WASUTF8) {
4607 STRLEN utf8_len = HEK_LEN(hek);
4608 char *as_utf8 = (char *)bytes_to_utf8((U8*)HEK_KEY(hek), &utf8_len);
4609 sv_usepvn_flags(sv, as_utf8, utf8_len, SV_HAS_TRAILING_NUL);
4612 } else if (flags & (HVhek_REHASH|HVhek_UNSHARED)) {
4613 sv_setpvn(sv, HEK_KEY(hek), HEK_LEN(hek));
4616 else SvUTF8_off(sv);
4620 sv_upgrade(sv, SVt_PV);
4621 sv_usepvn_flags(sv, (char *)HEK_KEY(share_hek_hek(hek)), HEK_LEN(hek), SV_HAS_TRAILING_NUL);
4628 else SvUTF8_off(sv);
4636 =for apidoc sv_usepvn_flags
4638 Tells an SV to use C<ptr> to find its string value. Normally the
4639 string is stored inside the SV but sv_usepvn allows the SV to use an
4640 outside string. The C<ptr> should point to memory that was allocated
4641 by C<malloc>. The string length, C<len>, must be supplied. By default
4642 this function will realloc (i.e. move) the memory pointed to by C<ptr>,
4643 so that pointer should not be freed or used by the programmer after
4644 giving it to sv_usepvn, and neither should any pointers from "behind"
4645 that pointer (e.g. ptr + 1) be used.
4647 If C<flags> & SV_SMAGIC is true, will call SvSETMAGIC. If C<flags> &
4648 SV_HAS_TRAILING_NUL is true, then C<ptr[len]> must be NUL, and the realloc
4649 will be skipped. (i.e. the buffer is actually at least 1 byte longer than
4650 C<len>, and already meets the requirements for storing in C<SvPVX>)
4656 Perl_sv_usepvn_flags(pTHX_ SV *const sv, char *ptr, const STRLEN len, const U32 flags)
4661 PERL_ARGS_ASSERT_SV_USEPVN_FLAGS;
4663 SV_CHECK_THINKFIRST_COW_DROP(sv);
4664 SvUPGRADE(sv, SVt_PV);
4667 if (flags & SV_SMAGIC)
4671 if (SvPVX_const(sv))
4675 if (flags & SV_HAS_TRAILING_NUL)
4676 assert(ptr[len] == '\0');
4679 allocate = (flags & SV_HAS_TRAILING_NUL)
4681 #ifdef Perl_safesysmalloc_size
4684 PERL_STRLEN_ROUNDUP(len + 1);
4686 if (flags & SV_HAS_TRAILING_NUL) {
4687 /* It's long enough - do nothing.
4688 Specifically Perl_newCONSTSUB is relying on this. */
4691 /* Force a move to shake out bugs in callers. */
4692 char *new_ptr = (char*)safemalloc(allocate);
4693 Copy(ptr, new_ptr, len, char);
4694 PoisonFree(ptr,len,char);
4698 ptr = (char*) saferealloc (ptr, allocate);
4701 #ifdef Perl_safesysmalloc_size
4702 SvLEN_set(sv, Perl_safesysmalloc_size(ptr));
4704 SvLEN_set(sv, allocate);
4708 if (!(flags & SV_HAS_TRAILING_NUL)) {
4711 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4713 if (flags & SV_SMAGIC)
4717 #ifdef PERL_OLD_COPY_ON_WRITE
4718 /* Need to do this *after* making the SV normal, as we need the buffer
4719 pointer to remain valid until after we've copied it. If we let go too early,
4720 another thread could invalidate it by unsharing last of the same hash key
4721 (which it can do by means other than releasing copy-on-write Svs)
4722 or by changing the other copy-on-write SVs in the loop. */
4724 S_sv_release_COW(pTHX_ register SV *sv, const char *pvx, SV *after)
4726 PERL_ARGS_ASSERT_SV_RELEASE_COW;
4728 { /* this SV was SvIsCOW_normal(sv) */
4729 /* we need to find the SV pointing to us. */
4730 SV *current = SV_COW_NEXT_SV(after);
4732 if (current == sv) {
4733 /* The SV we point to points back to us (there were only two of us
4735 Hence other SV is no longer copy on write either. */
4737 SvREADONLY_off(after);
4739 /* We need to follow the pointers around the loop. */
4741 while ((next = SV_COW_NEXT_SV(current)) != sv) {
4744 /* don't loop forever if the structure is bust, and we have
4745 a pointer into a closed loop. */
4746 assert (current != after);
4747 assert (SvPVX_const(current) == pvx);
4749 /* Make the SV before us point to the SV after us. */
4750 SV_COW_NEXT_SV_SET(current, after);
4756 =for apidoc sv_force_normal_flags
4758 Undo various types of fakery on an SV: if the PV is a shared string, make
4759 a private copy; if we're a ref, stop refing; if we're a glob, downgrade to
4760 an xpvmg; if we're a copy-on-write scalar, this is the on-write time when
4761 we do the copy, and is also used locally. If C<SV_COW_DROP_PV> is set
4762 then a copy-on-write scalar drops its PV buffer (if any) and becomes
4763 SvPOK_off rather than making a copy. (Used where this scalar is about to be
4764 set to some other value.) In addition, the C<flags> parameter gets passed to
4765 C<sv_unref_flags()> when unreffing. C<sv_force_normal> calls this function
4766 with flags set to 0.
4772 Perl_sv_force_normal_flags(pTHX_ register SV *const sv, const U32 flags)
4776 PERL_ARGS_ASSERT_SV_FORCE_NORMAL_FLAGS;
4778 #ifdef PERL_OLD_COPY_ON_WRITE
4779 if (SvREADONLY(sv)) {
4781 const char * const pvx = SvPVX_const(sv);
4782 const STRLEN len = SvLEN(sv);
4783 const STRLEN cur = SvCUR(sv);
4784 /* next COW sv in the loop. If len is 0 then this is a shared-hash
4785 key scalar, so we mustn't attempt to call SV_COW_NEXT_SV(), as
4786 we'll fail an assertion. */
4787 SV * const next = len ? SV_COW_NEXT_SV(sv) : 0;
4790 PerlIO_printf(Perl_debug_log,
4791 "Copy on write: Force normal %ld\n",
4797 /* This SV doesn't own the buffer, so need to Newx() a new one: */
4800 if (flags & SV_COW_DROP_PV) {
4801 /* OK, so we don't need to copy our buffer. */
4804 SvGROW(sv, cur + 1);
4805 Move(pvx,SvPVX(sv),cur,char);
4810 sv_release_COW(sv, pvx, next);
4812 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4818 else if (IN_PERL_RUNTIME)
4819 Perl_croak_no_modify(aTHX);
4822 if (SvREADONLY(sv)) {
4823 if (SvFAKE(sv) && !isGV_with_GP(sv)) {
4824 const char * const pvx = SvPVX_const(sv);
4825 const STRLEN len = SvCUR(sv);
4830 SvGROW(sv, len + 1);
4831 Move(pvx,SvPVX(sv),len,char);
4833 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4835 else if (IN_PERL_RUNTIME)
4836 Perl_croak_no_modify(aTHX);
4840 sv_unref_flags(sv, flags);
4841 else if (SvFAKE(sv) && isGV_with_GP(sv))
4843 else if (SvFAKE(sv) && SvTYPE(sv) == SVt_REGEXP) {
4844 /* Need to downgrade the REGEXP to a simple(r) scalar. This is analogous
4845 to sv_unglob. We only need it here, so inline it. */
4846 const svtype new_type = SvMAGIC(sv) || SvSTASH(sv) ? SVt_PVMG : SVt_PV;
4847 SV *const temp = newSV_type(new_type);
4848 void *const temp_p = SvANY(sv);
4850 if (new_type == SVt_PVMG) {
4851 SvMAGIC_set(temp, SvMAGIC(sv));
4852 SvMAGIC_set(sv, NULL);
4853 SvSTASH_set(temp, SvSTASH(sv));
4854 SvSTASH_set(sv, NULL);
4856 SvCUR_set(temp, SvCUR(sv));
4857 /* Remember that SvPVX is in the head, not the body. */
4859 SvLEN_set(temp, SvLEN(sv));
4860 /* This signals "buffer is owned by someone else" in sv_clear,
4861 which is the least effort way to stop it freeing the buffer.
4863 SvLEN_set(sv, SvLEN(sv)+1);
4865 /* Their buffer is already owned by someone else. */
4866 SvPVX(sv) = savepvn(SvPVX(sv), SvCUR(sv));
4867 SvLEN_set(temp, SvCUR(sv)+1);
4870 /* Now swap the rest of the bodies. */
4872 SvFLAGS(sv) &= ~(SVf_FAKE|SVTYPEMASK);
4873 SvFLAGS(sv) |= new_type;
4874 SvANY(sv) = SvANY(temp);
4876 SvFLAGS(temp) &= ~(SVTYPEMASK);
4877 SvFLAGS(temp) |= SVt_REGEXP|SVf_FAKE;
4878 SvANY(temp) = temp_p;
4887 Efficient removal of characters from the beginning of the string buffer.
4888 SvPOK(sv) must be true and the C<ptr> must be a pointer to somewhere inside
4889 the string buffer. The C<ptr> becomes the first character of the adjusted
4890 string. Uses the "OOK hack".
4892 Beware: after this function returns, C<ptr> and SvPVX_const(sv) may no longer
4893 refer to the same chunk of data.
4895 The unfortunate similarity of this function's name to that of Perl's C<chop>
4896 operator is strictly coincidental. This function works from the left;
4897 C<chop> works from the right.
4903 Perl_sv_chop(pTHX_ register SV *const sv, register const char *const ptr)
4914 PERL_ARGS_ASSERT_SV_CHOP;
4916 if (!ptr || !SvPOKp(sv))
4918 delta = ptr - SvPVX_const(sv);
4920 /* Nothing to do. */
4923 max_delta = SvLEN(sv) ? SvLEN(sv) : SvCUR(sv);
4924 if (delta > max_delta)
4925 Perl_croak(aTHX_ "panic: sv_chop ptr=%p, start=%p, end=%p",
4926 ptr, SvPVX_const(sv), SvPVX_const(sv) + max_delta);
4927 /* SvPVX(sv) may move in SV_CHECK_THINKFIRST(sv), so don't use ptr any more */
4928 SV_CHECK_THINKFIRST(sv);
4931 if (!SvLEN(sv)) { /* make copy of shared string */
4932 const char *pvx = SvPVX_const(sv);
4933 const STRLEN len = SvCUR(sv);
4934 SvGROW(sv, len + 1);
4935 Move(pvx,SvPVX(sv),len,char);
4938 SvFLAGS(sv) |= SVf_OOK;
4941 SvOOK_offset(sv, old_delta);
4943 SvLEN_set(sv, SvLEN(sv) - delta);
4944 SvCUR_set(sv, SvCUR(sv) - delta);
4945 SvPV_set(sv, SvPVX(sv) + delta);
4947 p = (U8 *)SvPVX_const(sv);
4950 /* how many bytes were evacuated? we will fill them with sentinel
4951 bytes, except for the part holding the new offset of course. */
4954 evacn += (old_delta < 0x100 ? 1 : 1 + sizeof(STRLEN));
4956 assert(evacn <= delta + old_delta);
4962 if (delta < 0x100) {
4966 p -= sizeof(STRLEN);
4967 Copy((U8*)&delta, p, sizeof(STRLEN), U8);
4971 /* Fill the preceding buffer with sentinals to verify that no-one is
4981 =for apidoc sv_catpvn
4983 Concatenates the string onto the end of the string which is in the SV. The
4984 C<len> indicates number of bytes to copy. If the SV has the UTF-8
4985 status set, then the bytes appended should be valid UTF-8.
4986 Handles 'get' magic, but not 'set' magic. See C<sv_catpvn_mg>.
4988 =for apidoc sv_catpvn_flags
4990 Concatenates the string onto the end of the string which is in the SV. The
4991 C<len> indicates number of bytes to copy. If the SV has the UTF-8
4992 status set, then the bytes appended should be valid UTF-8.
4993 If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<dsv> if
4994 appropriate, else not. C<sv_catpvn> and C<sv_catpvn_nomg> are implemented
4995 in terms of this function.
5001 Perl_sv_catpvn_flags(pTHX_ register SV *const dsv, register const char *sstr, register const STRLEN slen, const I32 flags)
5005 const char * const dstr = SvPV_force_flags(dsv, dlen, flags);
5007 PERL_ARGS_ASSERT_SV_CATPVN_FLAGS;
5008 assert((flags & (SV_CATBYTES|SV_CATUTF8)) != (SV_CATBYTES|SV_CATUTF8));
5010 if (!(flags & SV_CATBYTES) || !SvUTF8(dsv)) {
5011 if (flags & SV_CATUTF8 && !SvUTF8(dsv)) {
5012 sv_utf8_upgrade_flags_grow(dsv, 0, slen + 1);
5015 else SvGROW(dsv, dlen + slen + 1);
5017 sstr = SvPVX_const(dsv);
5018 Move(sstr, SvPVX(dsv) + dlen, slen, char);
5019 SvCUR_set(dsv, SvCUR(dsv) + slen);
5022 /* We inline bytes_to_utf8, to avoid an extra malloc. */
5023 const char * const send = sstr + slen;
5026 /* Something this code does not account for, which I think is
5027 impossible; it would require the same pv to be treated as
5028 bytes *and* utf8, which would indicate a bug elsewhere. */
5029 assert(sstr != dstr);
5031 SvGROW(dsv, dlen + slen * 2 + 1);
5032 d = (U8 *)SvPVX(dsv) + dlen;
5034 while (sstr < send) {
5035 const UV uv = NATIVE_TO_ASCII((U8)*sstr++);
5036 if (UNI_IS_INVARIANT(uv))
5037 *d++ = (U8)UTF_TO_NATIVE(uv);
5039 *d++ = (U8)UTF8_EIGHT_BIT_HI(uv);
5040 *d++ = (U8)UTF8_EIGHT_BIT_LO(uv);
5043 SvCUR_set(dsv, d-(const U8 *)SvPVX(dsv));
5046 (void)SvPOK_only_UTF8(dsv); /* validate pointer */
5048 if (flags & SV_SMAGIC)
5053 =for apidoc sv_catsv
5055 Concatenates the string from SV C<ssv> onto the end of the string in
5056 SV C<dsv>. Modifies C<dsv> but not C<ssv>. Handles 'get' magic, but
5057 not 'set' magic. See C<sv_catsv_mg>.
5059 =for apidoc sv_catsv_flags
5061 Concatenates the string from SV C<ssv> onto the end of the string in
5062 SV C<dsv>. Modifies C<dsv> but not C<ssv>. If C<flags> has C<SV_GMAGIC>
5063 bit set, will C<mg_get> on the SVs if appropriate, else not. C<sv_catsv>
5064 and C<sv_catsv_nomg> are implemented in terms of this function.
5069 Perl_sv_catsv_flags(pTHX_ SV *const dsv, register SV *const ssv, const I32 flags)
5073 PERL_ARGS_ASSERT_SV_CATSV_FLAGS;
5077 const char *spv = SvPV_flags_const(ssv, slen, flags);
5079 if (SvGMAGICAL(dsv) && (flags & SV_GMAGIC))
5081 sv_catpvn_flags(dsv, spv, slen,
5082 DO_UTF8(ssv) ? SV_CATUTF8 : SV_CATBYTES);
5085 if (flags & SV_SMAGIC)
5090 =for apidoc sv_catpv
5092 Concatenates the string onto the end of the string which is in the SV.
5093 If the SV has the UTF-8 status set, then the bytes appended should be
5094 valid UTF-8. Handles 'get' magic, but not 'set' magic. See C<sv_catpv_mg>.
5099 Perl_sv_catpv(pTHX_ register SV *const sv, register const char *ptr)
5102 register STRLEN len;
5106 PERL_ARGS_ASSERT_SV_CATPV;
5110 junk = SvPV_force(sv, tlen);
5112 SvGROW(sv, tlen + len + 1);
5114 ptr = SvPVX_const(sv);
5115 Move(ptr,SvPVX(sv)+tlen,len+1,char);
5116 SvCUR_set(sv, SvCUR(sv) + len);
5117 (void)SvPOK_only_UTF8(sv); /* validate pointer */
5122 =for apidoc sv_catpv_flags
5124 Concatenates the string onto the end of the string which is in the SV.
5125 If the SV has the UTF-8 status set, then the bytes appended should
5126 be valid UTF-8. If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get>
5127 on the SVs if appropriate, else not.
5133 Perl_sv_catpv_flags(pTHX_ SV *dstr, const char *sstr, const I32 flags)
5135 PERL_ARGS_ASSERT_SV_CATPV_FLAGS;
5136 sv_catpvn_flags(dstr, sstr, strlen(sstr), flags);
5140 =for apidoc sv_catpv_mg
5142 Like C<sv_catpv>, but also handles 'set' magic.
5148 Perl_sv_catpv_mg(pTHX_ register SV *const sv, register const char *const ptr)
5150 PERL_ARGS_ASSERT_SV_CATPV_MG;
5159 Creates a new SV. A non-zero C<len> parameter indicates the number of
5160 bytes of preallocated string space the SV should have. An extra byte for a
5161 trailing NUL is also reserved. (SvPOK is not set for the SV even if string
5162 space is allocated.) The reference count for the new SV is set to 1.
5164 In 5.9.3, newSV() replaces the older NEWSV() API, and drops the first
5165 parameter, I<x>, a debug aid which allowed callers to identify themselves.
5166 This aid has been superseded by a new build option, PERL_MEM_LOG (see
5167 L<perlhacktips/PERL_MEM_LOG>). The older API is still there for use in XS
5168 modules supporting older perls.
5174 Perl_newSV(pTHX_ const STRLEN len)
5181 sv_upgrade(sv, SVt_PV);
5182 SvGROW(sv, len + 1);
5187 =for apidoc sv_magicext
5189 Adds magic to an SV, upgrading it if necessary. Applies the
5190 supplied vtable and returns a pointer to the magic added.
5192 Note that C<sv_magicext> will allow things that C<sv_magic> will not.
5193 In particular, you can add magic to SvREADONLY SVs, and add more than
5194 one instance of the same 'how'.
5196 If C<namlen> is greater than zero then a C<savepvn> I<copy> of C<name> is
5197 stored, if C<namlen> is zero then C<name> is stored as-is and - as another
5198 special case - if C<(name && namlen == HEf_SVKEY)> then C<name> is assumed
5199 to contain an C<SV*> and is stored as-is with its REFCNT incremented.
5201 (This is now used as a subroutine by C<sv_magic>.)
5206 Perl_sv_magicext(pTHX_ SV *const sv, SV *const obj, const int how,
5207 const MGVTBL *const vtable, const char *const name, const I32 namlen)
5212 PERL_ARGS_ASSERT_SV_MAGICEXT;
5214 SvUPGRADE(sv, SVt_PVMG);
5215 Newxz(mg, 1, MAGIC);
5216 mg->mg_moremagic = SvMAGIC(sv);
5217 SvMAGIC_set(sv, mg);
5219 /* Sometimes a magic contains a reference loop, where the sv and
5220 object refer to each other. To prevent a reference loop that
5221 would prevent such objects being freed, we look for such loops
5222 and if we find one we avoid incrementing the object refcount.
5224 Note we cannot do this to avoid self-tie loops as intervening RV must
5225 have its REFCNT incremented to keep it in existence.
5228 if (!obj || obj == sv ||
5229 how == PERL_MAGIC_arylen ||
5230 how == PERL_MAGIC_symtab ||
5231 (SvTYPE(obj) == SVt_PVGV &&
5232 (GvSV(obj) == sv || GvHV(obj) == (const HV *)sv
5233 || GvAV(obj) == (const AV *)sv || GvCV(obj) == (const CV *)sv
5234 || GvIOp(obj) == (const IO *)sv || GvFORM(obj) == (const CV *)sv)))
5239 mg->mg_obj = SvREFCNT_inc_simple(obj);
5240 mg->mg_flags |= MGf_REFCOUNTED;
5243 /* Normal self-ties simply pass a null object, and instead of
5244 using mg_obj directly, use the SvTIED_obj macro to produce a
5245 new RV as needed. For glob "self-ties", we are tieing the PVIO
5246 with an RV obj pointing to the glob containing the PVIO. In
5247 this case, to avoid a reference loop, we need to weaken the
5251 if (how == PERL_MAGIC_tiedscalar && SvTYPE(sv) == SVt_PVIO &&
5252 obj && SvROK(obj) && GvIO(SvRV(obj)) == (const IO *)sv)
5258 mg->mg_len = namlen;
5261 mg->mg_ptr = savepvn(name, namlen);
5262 else if (namlen == HEf_SVKEY) {
5263 /* Yes, this is casting away const. This is only for the case of
5264 HEf_SVKEY. I think we need to document this aberation of the
5265 constness of the API, rather than making name non-const, as
5266 that change propagating outwards a long way. */
5267 mg->mg_ptr = (char*)SvREFCNT_inc_simple_NN((SV *)name);
5269 mg->mg_ptr = (char *) name;
5271 mg->mg_virtual = (MGVTBL *) vtable;
5275 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
5280 =for apidoc sv_magic
5282 Adds magic to an SV. First upgrades C<sv> to type C<SVt_PVMG> if necessary,
5283 then adds a new magic item of type C<how> to the head of the magic list.
5285 See C<sv_magicext> (which C<sv_magic> now calls) for a description of the
5286 handling of the C<name> and C<namlen> arguments.
5288 You need to use C<sv_magicext> to add magic to SvREADONLY SVs and also
5289 to add more than one instance of the same 'how'.
5295 Perl_sv_magic(pTHX_ register SV *const sv, SV *const obj, const int how,
5296 const char *const name, const I32 namlen)
5299 const MGVTBL *vtable;
5302 unsigned int vtable_index;
5304 PERL_ARGS_ASSERT_SV_MAGIC;
5306 if (how < 0 || (unsigned)how > C_ARRAY_LENGTH(PL_magic_data)
5307 || ((flags = PL_magic_data[how]),
5308 (vtable_index = flags & PERL_MAGIC_VTABLE_MASK)
5309 > magic_vtable_max))
5310 Perl_croak(aTHX_ "Don't know how to handle magic of type \\%o", how);
5312 /* PERL_MAGIC_ext is reserved for use by extensions not perl internals.
5313 Useful for attaching extension internal data to perl vars.
5314 Note that multiple extensions may clash if magical scalars
5315 etc holding private data from one are passed to another. */
5317 vtable = (vtable_index == magic_vtable_max)
5318 ? NULL : PL_magic_vtables + vtable_index;
5320 #ifdef PERL_OLD_COPY_ON_WRITE
5322 sv_force_normal_flags(sv, 0);
5324 if (SvREADONLY(sv)) {
5326 /* its okay to attach magic to shared strings; the subsequent
5327 * upgrade to PVMG will unshare the string */
5328 !(SvFAKE(sv) && SvTYPE(sv) < SVt_PVMG)
5331 && !PERL_MAGIC_TYPE_READONLY_ACCEPTABLE(how)
5334 Perl_croak_no_modify(aTHX);
5337 if (SvMAGICAL(sv) || (how == PERL_MAGIC_taint && SvTYPE(sv) >= SVt_PVMG)) {
5338 if (SvMAGIC(sv) && (mg = mg_find(sv, how))) {
5339 /* sv_magic() refuses to add a magic of the same 'how' as an
5342 if (how == PERL_MAGIC_taint) {
5344 /* Any scalar which already had taint magic on which someone
5345 (erroneously?) did SvIOK_on() or similar will now be
5346 incorrectly sporting public "OK" flags. */
5347 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
5353 /* Rest of work is done else where */
5354 mg = sv_magicext(sv,obj,how,vtable,name,namlen);
5357 case PERL_MAGIC_taint:
5360 case PERL_MAGIC_ext:
5361 case PERL_MAGIC_dbfile:
5368 S_sv_unmagicext_flags(pTHX_ SV *const sv, const int type, MGVTBL *vtbl, const U32 flags)
5375 if (SvTYPE(sv) < SVt_PVMG || !SvMAGIC(sv))
5377 mgp = &(((XPVMG*) SvANY(sv))->xmg_u.xmg_magic);
5378 for (mg = *mgp; mg; mg = *mgp) {
5379 const MGVTBL* const virt = mg->mg_virtual;
5380 if (mg->mg_type == type && (!flags || virt == vtbl)) {
5381 *mgp = mg->mg_moremagic;
5382 if (virt && virt->svt_free)
5383 virt->svt_free(aTHX_ sv, mg);
5384 if (mg->mg_ptr && mg->mg_type != PERL_MAGIC_regex_global) {
5386 Safefree(mg->mg_ptr);
5387 else if (mg->mg_len == HEf_SVKEY)
5388 SvREFCNT_dec(MUTABLE_SV(mg->mg_ptr));
5389 else if (mg->mg_type == PERL_MAGIC_utf8)
5390 Safefree(mg->mg_ptr);
5392 if (mg->mg_flags & MGf_REFCOUNTED)
5393 SvREFCNT_dec(mg->mg_obj);
5397 mgp = &mg->mg_moremagic;
5400 if (SvMAGICAL(sv)) /* if we're under save_magic, wait for restore_magic; */
5401 mg_magical(sv); /* else fix the flags now */
5405 SvFLAGS(sv) |= (SvFLAGS(sv) & (SVp_IOK|SVp_NOK|SVp_POK)) >> PRIVSHIFT;
5411 =for apidoc sv_unmagic
5413 Removes all magic of type C<type> from an SV.
5419 Perl_sv_unmagic(pTHX_ SV *const sv, const int type)
5421 PERL_ARGS_ASSERT_SV_UNMAGIC;
5422 return S_sv_unmagicext_flags(aTHX_ sv, type, NULL, 0);
5426 =for apidoc sv_unmagicext
5428 Removes all magic of type C<type> with the specified C<vtbl> from an SV.
5434 Perl_sv_unmagicext(pTHX_ SV *const sv, const int type, MGVTBL *vtbl)
5436 PERL_ARGS_ASSERT_SV_UNMAGICEXT;
5437 return S_sv_unmagicext_flags(aTHX_ sv, type, vtbl, 1);
5441 =for apidoc sv_rvweaken
5443 Weaken a reference: set the C<SvWEAKREF> flag on this RV; give the
5444 referred-to SV C<PERL_MAGIC_backref> magic if it hasn't already; and
5445 push a back-reference to this RV onto the array of backreferences
5446 associated with that magic. If the RV is magical, set magic will be
5447 called after the RV is cleared.
5453 Perl_sv_rvweaken(pTHX_ SV *const sv)
5457 PERL_ARGS_ASSERT_SV_RVWEAKEN;
5459 if (!SvOK(sv)) /* let undefs pass */
5462 Perl_croak(aTHX_ "Can't weaken a nonreference");
5463 else if (SvWEAKREF(sv)) {
5464 Perl_ck_warner(aTHX_ packWARN(WARN_MISC), "Reference is already weak");
5467 else if (SvREADONLY(sv)) croak_no_modify();
5469 Perl_sv_add_backref(aTHX_ tsv, sv);
5475 /* Give tsv backref magic if it hasn't already got it, then push a
5476 * back-reference to sv onto the array associated with the backref magic.
5478 * As an optimisation, if there's only one backref and it's not an AV,
5479 * store it directly in the HvAUX or mg_obj slot, avoiding the need to
5480 * allocate an AV. (Whether the slot holds an AV tells us whether this is
5484 /* A discussion about the backreferences array and its refcount:
5486 * The AV holding the backreferences is pointed to either as the mg_obj of
5487 * PERL_MAGIC_backref, or in the specific case of a HV, from the
5488 * xhv_backreferences field. The array is created with a refcount
5489 * of 2. This means that if during global destruction the array gets
5490 * picked on before its parent to have its refcount decremented by the
5491 * random zapper, it won't actually be freed, meaning it's still there for
5492 * when its parent gets freed.
5494 * When the parent SV is freed, the extra ref is killed by
5495 * Perl_sv_kill_backrefs. The other ref is killed, in the case of magic,
5496 * by mg_free() / MGf_REFCOUNTED, or for a hash, by Perl_hv_kill_backrefs.
5498 * When a single backref SV is stored directly, it is not reference
5503 Perl_sv_add_backref(pTHX_ SV *const tsv, SV *const sv)
5510 PERL_ARGS_ASSERT_SV_ADD_BACKREF;
5512 /* find slot to store array or singleton backref */
5514 if (SvTYPE(tsv) == SVt_PVHV) {
5515 svp = (SV**)Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5518 (SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL))))
5520 sv_magic(tsv, NULL, PERL_MAGIC_backref, NULL, 0);
5521 mg = mg_find(tsv, PERL_MAGIC_backref);
5523 svp = &(mg->mg_obj);
5526 /* create or retrieve the array */
5528 if ( (!*svp && SvTYPE(sv) == SVt_PVAV)
5529 || (*svp && SvTYPE(*svp) != SVt_PVAV)
5534 SvREFCNT_inc_simple_void(av);
5535 /* av now has a refcnt of 2; see discussion above */
5537 /* move single existing backref to the array */
5539 AvARRAY(av)[++AvFILLp(av)] = *svp; /* av_push() */
5543 mg->mg_flags |= MGf_REFCOUNTED;
5546 av = MUTABLE_AV(*svp);
5549 /* optimisation: store single backref directly in HvAUX or mg_obj */
5553 /* push new backref */
5554 assert(SvTYPE(av) == SVt_PVAV);
5555 if (AvFILLp(av) >= AvMAX(av)) {
5556 av_extend(av, AvFILLp(av)+1);
5558 AvARRAY(av)[++AvFILLp(av)] = sv; /* av_push() */
5561 /* delete a back-reference to ourselves from the backref magic associated
5562 * with the SV we point to.
5566 Perl_sv_del_backref(pTHX_ SV *const tsv, SV *const sv)
5571 PERL_ARGS_ASSERT_SV_DEL_BACKREF;
5573 if (SvTYPE(tsv) == SVt_PVHV) {
5575 svp = (SV**)Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5579 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
5580 svp = mg ? &(mg->mg_obj) : NULL;
5584 Perl_croak(aTHX_ "panic: del_backref");
5586 if (SvTYPE(*svp) == SVt_PVAV) {
5590 AV * const av = (AV*)*svp;
5592 assert(!SvIS_FREED(av));
5596 /* for an SV with N weak references to it, if all those
5597 * weak refs are deleted, then sv_del_backref will be called
5598 * N times and O(N^2) compares will be done within the backref
5599 * array. To ameliorate this potential slowness, we:
5600 * 1) make sure this code is as tight as possible;
5601 * 2) when looking for SV, look for it at both the head and tail of the
5602 * array first before searching the rest, since some create/destroy
5603 * patterns will cause the backrefs to be freed in order.
5610 SV **p = &svp[fill];
5611 SV *const topsv = *p;
5618 /* We weren't the last entry.
5619 An unordered list has this property that you
5620 can take the last element off the end to fill
5621 the hole, and it's still an unordered list :-)
5627 break; /* should only be one */
5634 AvFILLp(av) = fill-1;
5637 /* optimisation: only a single backref, stored directly */
5639 Perl_croak(aTHX_ "panic: del_backref");
5646 Perl_sv_kill_backrefs(pTHX_ SV *const sv, AV *const av)
5652 PERL_ARGS_ASSERT_SV_KILL_BACKREFS;
5657 /* after multiple passes through Perl_sv_clean_all() for a thinngy
5658 * that has badly leaked, the backref array may have gotten freed,
5659 * since we only protect it against 1 round of cleanup */
5660 if (SvIS_FREED(av)) {
5661 if (PL_in_clean_all) /* All is fair */
5664 "panic: magic_killbackrefs (freed backref AV/SV)");
5668 is_array = (SvTYPE(av) == SVt_PVAV);
5670 assert(!SvIS_FREED(av));
5673 last = svp + AvFILLp(av);
5676 /* optimisation: only a single backref, stored directly */
5682 while (svp <= last) {
5684 SV *const referrer = *svp;
5685 if (SvWEAKREF(referrer)) {
5686 /* XXX Should we check that it hasn't changed? */
5687 assert(SvROK(referrer));
5688 SvRV_set(referrer, 0);
5690 SvWEAKREF_off(referrer);
5691 SvSETMAGIC(referrer);
5692 } else if (SvTYPE(referrer) == SVt_PVGV ||
5693 SvTYPE(referrer) == SVt_PVLV) {
5694 assert(SvTYPE(sv) == SVt_PVHV); /* stash backref */
5695 /* You lookin' at me? */
5696 assert(GvSTASH(referrer));
5697 assert(GvSTASH(referrer) == (const HV *)sv);
5698 GvSTASH(referrer) = 0;
5699 } else if (SvTYPE(referrer) == SVt_PVCV ||
5700 SvTYPE(referrer) == SVt_PVFM) {
5701 if (SvTYPE(sv) == SVt_PVHV) { /* stash backref */
5702 /* You lookin' at me? */
5703 assert(CvSTASH(referrer));
5704 assert(CvSTASH(referrer) == (const HV *)sv);
5705 SvANY(MUTABLE_CV(referrer))->xcv_stash = 0;
5708 assert(SvTYPE(sv) == SVt_PVGV);
5709 /* You lookin' at me? */
5710 assert(CvGV(referrer));
5711 assert(CvGV(referrer) == (const GV *)sv);
5712 anonymise_cv_maybe(MUTABLE_GV(sv),
5713 MUTABLE_CV(referrer));
5718 "panic: magic_killbackrefs (flags=%"UVxf")",
5719 (UV)SvFLAGS(referrer));
5730 SvREFCNT_dec(av); /* remove extra count added by sv_add_backref() */
5736 =for apidoc sv_insert
5738 Inserts a string at the specified offset/length within the SV. Similar to
5739 the Perl substr() function. Handles get magic.
5741 =for apidoc sv_insert_flags
5743 Same as C<sv_insert>, but the extra C<flags> are passed the C<SvPV_force_flags> that applies to C<bigstr>.
5749 Perl_sv_insert_flags(pTHX_ SV *const bigstr, const STRLEN offset, const STRLEN len, const char *const little, const STRLEN littlelen, const U32 flags)
5754 register char *midend;
5755 register char *bigend;
5756 register SSize_t i; /* better be sizeof(STRLEN) or bad things happen */
5759 PERL_ARGS_ASSERT_SV_INSERT_FLAGS;
5762 Perl_croak(aTHX_ "Can't modify non-existent substring");
5763 SvPV_force_flags(bigstr, curlen, flags);
5764 (void)SvPOK_only_UTF8(bigstr);
5765 if (offset + len > curlen) {
5766 SvGROW(bigstr, offset+len+1);
5767 Zero(SvPVX(bigstr)+curlen, offset+len-curlen, char);
5768 SvCUR_set(bigstr, offset+len);
5772 i = littlelen - len;
5773 if (i > 0) { /* string might grow */
5774 big = SvGROW(bigstr, SvCUR(bigstr) + i + 1);
5775 mid = big + offset + len;
5776 midend = bigend = big + SvCUR(bigstr);
5779 while (midend > mid) /* shove everything down */
5780 *--bigend = *--midend;
5781 Move(little,big+offset,littlelen,char);
5782 SvCUR_set(bigstr, SvCUR(bigstr) + i);
5787 Move(little,SvPVX(bigstr)+offset,len,char);
5792 big = SvPVX(bigstr);
5795 bigend = big + SvCUR(bigstr);
5797 if (midend > bigend)
5798 Perl_croak(aTHX_ "panic: sv_insert");
5800 if (mid - big > bigend - midend) { /* faster to shorten from end */
5802 Move(little, mid, littlelen,char);
5805 i = bigend - midend;
5807 Move(midend, mid, i,char);
5811 SvCUR_set(bigstr, mid - big);
5813 else if ((i = mid - big)) { /* faster from front */
5814 midend -= littlelen;
5816 Move(big, midend - i, i, char);
5817 sv_chop(bigstr,midend-i);
5819 Move(little, mid, littlelen,char);
5821 else if (littlelen) {
5822 midend -= littlelen;
5823 sv_chop(bigstr,midend);
5824 Move(little,midend,littlelen,char);
5827 sv_chop(bigstr,midend);
5833 =for apidoc sv_replace
5835 Make the first argument a copy of the second, then delete the original.
5836 The target SV physically takes over ownership of the body of the source SV
5837 and inherits its flags; however, the target keeps any magic it owns,
5838 and any magic in the source is discarded.
5839 Note that this is a rather specialist SV copying operation; most of the
5840 time you'll want to use C<sv_setsv> or one of its many macro front-ends.
5846 Perl_sv_replace(pTHX_ register SV *const sv, register SV *const nsv)
5849 const U32 refcnt = SvREFCNT(sv);
5851 PERL_ARGS_ASSERT_SV_REPLACE;
5853 SV_CHECK_THINKFIRST_COW_DROP(sv);
5854 if (SvREFCNT(nsv) != 1) {
5855 Perl_croak(aTHX_ "panic: reference miscount on nsv in sv_replace()"
5856 " (%" UVuf " != 1)", (UV) SvREFCNT(nsv));
5858 if (SvMAGICAL(sv)) {
5862 sv_upgrade(nsv, SVt_PVMG);
5863 SvMAGIC_set(nsv, SvMAGIC(sv));
5864 SvFLAGS(nsv) |= SvMAGICAL(sv);
5866 SvMAGIC_set(sv, NULL);
5870 assert(!SvREFCNT(sv));
5871 #ifdef DEBUG_LEAKING_SCALARS
5872 sv->sv_flags = nsv->sv_flags;
5873 sv->sv_any = nsv->sv_any;
5874 sv->sv_refcnt = nsv->sv_refcnt;
5875 sv->sv_u = nsv->sv_u;
5877 StructCopy(nsv,sv,SV);
5879 if(SvTYPE(sv) == SVt_IV) {
5881 = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
5885 #ifdef PERL_OLD_COPY_ON_WRITE
5886 if (SvIsCOW_normal(nsv)) {
5887 /* We need to follow the pointers around the loop to make the
5888 previous SV point to sv, rather than nsv. */
5891 while ((next = SV_COW_NEXT_SV(current)) != nsv) {
5894 assert(SvPVX_const(current) == SvPVX_const(nsv));
5896 /* Make the SV before us point to the SV after us. */
5898 PerlIO_printf(Perl_debug_log, "previous is\n");
5900 PerlIO_printf(Perl_debug_log,
5901 "move it from 0x%"UVxf" to 0x%"UVxf"\n",
5902 (UV) SV_COW_NEXT_SV(current), (UV) sv);
5904 SV_COW_NEXT_SV_SET(current, sv);
5907 SvREFCNT(sv) = refcnt;
5908 SvFLAGS(nsv) |= SVTYPEMASK; /* Mark as freed */
5913 /* We're about to free a GV which has a CV that refers back to us.
5914 * If that CV will outlive us, make it anonymous (i.e. fix up its CvGV
5918 S_anonymise_cv_maybe(pTHX_ GV *gv, CV* cv)
5923 PERL_ARGS_ASSERT_ANONYMISE_CV_MAYBE;
5926 assert(SvREFCNT(gv) == 0);
5927 assert(isGV(gv) && isGV_with_GP(gv));
5929 assert(!CvANON(cv));
5930 assert(CvGV(cv) == gv);
5932 /* will the CV shortly be freed by gp_free() ? */
5933 if (GvCV(gv) == cv && GvGP(gv)->gp_refcnt < 2 && SvREFCNT(cv) < 2) {
5934 SvANY(cv)->xcv_gv = NULL;
5938 /* if not, anonymise: */
5939 gvname = (GvSTASH(gv) && HvNAME(GvSTASH(gv)) && HvENAME(GvSTASH(gv)))
5940 ? newSVhek(HvENAME_HEK(GvSTASH(gv)))
5941 : newSVpvn_flags( "__ANON__", 8, 0 );
5942 sv_catpvs(gvname, "::__ANON__");
5943 anongv = gv_fetchsv(gvname, GV_ADDMULTI, SVt_PVCV);
5944 SvREFCNT_dec(gvname);
5948 SvANY(cv)->xcv_gv = MUTABLE_GV(SvREFCNT_inc(anongv));
5953 =for apidoc sv_clear
5955 Clear an SV: call any destructors, free up any memory used by the body,
5956 and free the body itself. The SV's head is I<not> freed, although
5957 its type is set to all 1's so that it won't inadvertently be assumed
5958 to be live during global destruction etc.
5959 This function should only be called when REFCNT is zero. Most of the time
5960 you'll want to call C<sv_free()> (or its macro wrapper C<SvREFCNT_dec>)
5967 Perl_sv_clear(pTHX_ SV *const orig_sv)
5972 const struct body_details *sv_type_details;
5975 register SV *sv = orig_sv;
5978 PERL_ARGS_ASSERT_SV_CLEAR;
5980 /* within this loop, sv is the SV currently being freed, and
5981 * iter_sv is the most recent AV or whatever that's being iterated
5982 * over to provide more SVs */
5988 assert(SvREFCNT(sv) == 0);
5989 assert(SvTYPE(sv) != (svtype)SVTYPEMASK);
5991 if (type <= SVt_IV) {
5992 /* See the comment in sv.h about the collusion between this
5993 * early return and the overloading of the NULL slots in the
5997 SvFLAGS(sv) &= SVf_BREAK;
5998 SvFLAGS(sv) |= SVTYPEMASK;
6002 assert(!SvOBJECT(sv) || type >= SVt_PVMG); /* objs are always >= MG */
6004 if (type >= SVt_PVMG) {
6006 if (!curse(sv, 1)) goto get_next_sv;
6007 type = SvTYPE(sv); /* destructor may have changed it */
6009 /* Free back-references before magic, in case the magic calls
6010 * Perl code that has weak references to sv. */
6011 if (type == SVt_PVHV) {
6012 Perl_hv_kill_backrefs(aTHX_ MUTABLE_HV(sv));
6016 else if (type == SVt_PVMG && SvPAD_OUR(sv)) {
6017 SvREFCNT_dec(SvOURSTASH(sv));
6018 } else if (SvMAGIC(sv)) {
6019 /* Free back-references before other types of magic. */
6020 sv_unmagic(sv, PERL_MAGIC_backref);
6023 if (type == SVt_PVMG && SvPAD_TYPED(sv))
6024 SvREFCNT_dec(SvSTASH(sv));
6027 /* case SVt_BIND: */
6030 IoIFP(sv) != PerlIO_stdin() &&
6031 IoIFP(sv) != PerlIO_stdout() &&
6032 IoIFP(sv) != PerlIO_stderr() &&
6033 !(IoFLAGS(sv) & IOf_FAKE_DIRP))
6035 io_close(MUTABLE_IO(sv), FALSE);
6037 if (IoDIRP(sv) && !(IoFLAGS(sv) & IOf_FAKE_DIRP))
6038 PerlDir_close(IoDIRP(sv));
6039 IoDIRP(sv) = (DIR*)NULL;
6040 Safefree(IoTOP_NAME(sv));
6041 Safefree(IoFMT_NAME(sv));
6042 Safefree(IoBOTTOM_NAME(sv));
6045 /* FIXME for plugins */
6046 pregfree2((REGEXP*) sv);
6050 cv_undef(MUTABLE_CV(sv));
6051 /* If we're in a stash, we don't own a reference to it.
6052 * However it does have a back reference to us, which needs to
6054 if ((stash = CvSTASH(sv)))
6055 sv_del_backref(MUTABLE_SV(stash), sv);
6058 if (PL_last_swash_hv == (const HV *)sv) {
6059 PL_last_swash_hv = NULL;
6061 if (HvTOTALKEYS((HV*)sv) > 0) {
6063 /* this statement should match the one at the beginning of
6064 * hv_undef_flags() */
6065 if ( PL_phase != PERL_PHASE_DESTRUCT
6066 && (name = HvNAME((HV*)sv)))
6069 (void)hv_delete(PL_stashcache, name,
6070 HvNAMEUTF8((HV*)sv) ? -HvNAMELEN_get((HV*)sv) : HvNAMELEN_get((HV*)sv), G_DISCARD);
6071 hv_name_set((HV*)sv, NULL, 0, 0);
6074 /* save old iter_sv in unused SvSTASH field */
6075 assert(!SvOBJECT(sv));
6076 SvSTASH(sv) = (HV*)iter_sv;
6079 /* XXX ideally we should save the old value of hash_index
6080 * too, but I can't think of any place to hide it. The
6081 * effect of not saving it is that for freeing hashes of
6082 * hashes, we become quadratic in scanning the HvARRAY of
6083 * the top hash looking for new entries to free; but
6084 * hopefully this will be dwarfed by the freeing of all
6085 * the nested hashes. */
6087 next_sv = Perl_hfree_next_entry(aTHX_ (HV*)sv, &hash_index);
6088 goto get_next_sv; /* process this new sv */
6090 /* free empty hash */
6091 Perl_hv_undef_flags(aTHX_ MUTABLE_HV(sv), HV_NAME_SETALL);
6092 assert(!HvARRAY((HV*)sv));
6096 AV* av = MUTABLE_AV(sv);
6097 if (PL_comppad == av) {
6101 if (AvREAL(av) && AvFILLp(av) > -1) {
6102 next_sv = AvARRAY(av)[AvFILLp(av)--];
6103 /* save old iter_sv in top-most slot of AV,
6104 * and pray that it doesn't get wiped in the meantime */
6105 AvARRAY(av)[AvMAX(av)] = iter_sv;
6107 goto get_next_sv; /* process this new sv */
6109 Safefree(AvALLOC(av));
6114 if (LvTYPE(sv) == 'T') { /* for tie: return HE to pool */
6115 SvREFCNT_dec(HeKEY_sv((HE*)LvTARG(sv)));
6116 HeNEXT((HE*)LvTARG(sv)) = PL_hv_fetch_ent_mh;
6117 PL_hv_fetch_ent_mh = (HE*)LvTARG(sv);
6119 else if (LvTYPE(sv) != 't') /* unless tie: unrefcnted fake SV** */
6120 SvREFCNT_dec(LvTARG(sv));
6122 if (isGV_with_GP(sv)) {
6123 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
6124 && HvENAME_get(stash))
6125 mro_method_changed_in(stash);
6126 gp_free(MUTABLE_GV(sv));
6128 unshare_hek(GvNAME_HEK(sv));
6129 /* If we're in a stash, we don't own a reference to it.
6130 * However it does have a back reference to us, which
6131 * needs to be cleared. */
6132 if (!SvVALID(sv) && (stash = GvSTASH(sv)))
6133 sv_del_backref(MUTABLE_SV(stash), sv);
6135 /* FIXME. There are probably more unreferenced pointers to SVs
6136 * in the interpreter struct that we should check and tidy in
6137 * a similar fashion to this: */
6138 if ((const GV *)sv == PL_last_in_gv)
6139 PL_last_in_gv = NULL;
6145 /* Don't bother with SvOOK_off(sv); as we're only going to
6149 SvOOK_offset(sv, offset);
6150 SvPV_set(sv, SvPVX_mutable(sv) - offset);
6151 /* Don't even bother with turning off the OOK flag. */
6156 SV * const target = SvRV(sv);
6158 sv_del_backref(target, sv);
6163 #ifdef PERL_OLD_COPY_ON_WRITE
6164 else if (SvPVX_const(sv)
6165 && !(SvTYPE(sv) == SVt_PVIO
6166 && !(IoFLAGS(sv) & IOf_FAKE_DIRP)))
6170 PerlIO_printf(Perl_debug_log, "Copy on write: clear\n");
6174 sv_release_COW(sv, SvPVX_const(sv), SV_COW_NEXT_SV(sv));
6176 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
6180 } else if (SvLEN(sv)) {
6181 Safefree(SvPVX_const(sv));
6185 else if (SvPVX_const(sv) && SvLEN(sv)
6186 && !(SvTYPE(sv) == SVt_PVIO
6187 && !(IoFLAGS(sv) & IOf_FAKE_DIRP)))
6188 Safefree(SvPVX_mutable(sv));
6189 else if (SvPVX_const(sv) && SvREADONLY(sv) && SvFAKE(sv)) {
6190 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
6201 SvFLAGS(sv) &= SVf_BREAK;
6202 SvFLAGS(sv) |= SVTYPEMASK;
6204 sv_type_details = bodies_by_type + type;
6205 if (sv_type_details->arena) {
6206 del_body(((char *)SvANY(sv) + sv_type_details->offset),
6207 &PL_body_roots[type]);
6209 else if (sv_type_details->body_size) {
6210 safefree(SvANY(sv));
6214 /* caller is responsible for freeing the head of the original sv */
6215 if (sv != orig_sv && !SvREFCNT(sv))
6218 /* grab and free next sv, if any */
6226 else if (!iter_sv) {
6228 } else if (SvTYPE(iter_sv) == SVt_PVAV) {
6229 AV *const av = (AV*)iter_sv;
6230 if (AvFILLp(av) > -1) {
6231 sv = AvARRAY(av)[AvFILLp(av)--];
6233 else { /* no more elements of current AV to free */
6236 /* restore previous value, squirrelled away */
6237 iter_sv = AvARRAY(av)[AvMAX(av)];
6238 Safefree(AvALLOC(av));
6241 } else if (SvTYPE(iter_sv) == SVt_PVHV) {
6242 sv = Perl_hfree_next_entry(aTHX_ (HV*)iter_sv, &hash_index);
6243 if (!sv && !HvTOTALKEYS((HV *)iter_sv)) {
6244 /* no more elements of current HV to free */
6247 /* Restore previous value of iter_sv, squirrelled away */
6248 assert(!SvOBJECT(sv));
6249 iter_sv = (SV*)SvSTASH(sv);
6251 /* ideally we should restore the old hash_index here,
6252 * but we don't currently save the old value */
6255 /* free any remaining detritus from the hash struct */
6256 Perl_hv_undef_flags(aTHX_ MUTABLE_HV(sv), HV_NAME_SETALL);
6257 assert(!HvARRAY((HV*)sv));
6262 /* unrolled SvREFCNT_dec and sv_free2 follows: */
6266 if (!SvREFCNT(sv)) {
6270 if (--(SvREFCNT(sv)))
6274 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEBUGGING),
6275 "Attempt to free temp prematurely: SV 0x%"UVxf
6276 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
6280 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
6281 /* make sure SvREFCNT(sv)==0 happens very seldom */
6282 SvREFCNT(sv) = (~(U32)0)/2;
6291 /* This routine curses the sv itself, not the object referenced by sv. So
6292 sv does not have to be ROK. */
6295 S_curse(pTHX_ SV * const sv, const bool check_refcnt) {
6298 PERL_ARGS_ASSERT_CURSE;
6299 assert(SvOBJECT(sv));
6301 if (PL_defstash && /* Still have a symbol table? */
6308 stash = SvSTASH(sv);
6309 destructor = StashHANDLER(stash,DESTROY);
6311 /* A constant subroutine can have no side effects, so
6312 don't bother calling it. */
6313 && !CvCONST(destructor)
6314 /* Don't bother calling an empty destructor */
6315 && (CvISXSUB(destructor)
6316 || (CvSTART(destructor)
6317 && (CvSTART(destructor)->op_next->op_type
6320 SV* const tmpref = newRV(sv);
6321 SvREADONLY_on(tmpref); /* DESTROY() could be naughty */
6323 PUSHSTACKi(PERLSI_DESTROY);
6328 call_sv(MUTABLE_SV(destructor),
6329 G_DISCARD|G_EVAL|G_KEEPERR|G_VOID);
6333 if(SvREFCNT(tmpref) < 2) {
6334 /* tmpref is not kept alive! */
6336 SvRV_set(tmpref, NULL);
6339 SvREFCNT_dec(tmpref);
6341 } while (SvOBJECT(sv) && SvSTASH(sv) != stash);
6344 if (check_refcnt && SvREFCNT(sv)) {
6345 if (PL_in_clean_objs)
6347 "DESTROY created new reference to dead object '%"HEKf"'",
6348 HEKfARG(HvNAME_HEK(stash)));
6349 /* DESTROY gave object new lease on life */
6355 SvREFCNT_dec(SvSTASH(sv)); /* possibly of changed persuasion */
6356 SvOBJECT_off(sv); /* Curse the object. */
6357 if (SvTYPE(sv) != SVt_PVIO)
6358 --PL_sv_objcount;/* XXX Might want something more general */
6364 =for apidoc sv_newref
6366 Increment an SV's reference count. Use the C<SvREFCNT_inc()> wrapper
6373 Perl_sv_newref(pTHX_ SV *const sv)
6375 PERL_UNUSED_CONTEXT;
6384 Decrement an SV's reference count, and if it drops to zero, call
6385 C<sv_clear> to invoke destructors and free up any memory used by
6386 the body; finally, deallocate the SV's head itself.
6387 Normally called via a wrapper macro C<SvREFCNT_dec>.
6393 Perl_sv_free(pTHX_ SV *const sv)
6398 if (SvREFCNT(sv) == 0) {
6399 if (SvFLAGS(sv) & SVf_BREAK)
6400 /* this SV's refcnt has been artificially decremented to
6401 * trigger cleanup */
6403 if (PL_in_clean_all) /* All is fair */
6405 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
6406 /* make sure SvREFCNT(sv)==0 happens very seldom */
6407 SvREFCNT(sv) = (~(U32)0)/2;
6410 if (ckWARN_d(WARN_INTERNAL)) {
6411 #ifdef DEBUG_LEAKING_SCALARS_FORK_DUMP
6412 Perl_dump_sv_child(aTHX_ sv);
6414 #ifdef DEBUG_LEAKING_SCALARS
6417 #ifdef DEBUG_LEAKING_SCALARS_ABORT
6418 if (PL_warnhook == PERL_WARNHOOK_FATAL
6419 || ckDEAD(packWARN(WARN_INTERNAL))) {
6420 /* Don't let Perl_warner cause us to escape our fate: */
6424 /* This may not return: */
6425 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
6426 "Attempt to free unreferenced scalar: SV 0x%"UVxf
6427 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
6430 #ifdef DEBUG_LEAKING_SCALARS_ABORT
6435 if (--(SvREFCNT(sv)) > 0)
6437 Perl_sv_free2(aTHX_ sv);
6441 Perl_sv_free2(pTHX_ SV *const sv)
6445 PERL_ARGS_ASSERT_SV_FREE2;
6449 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEBUGGING),
6450 "Attempt to free temp prematurely: SV 0x%"UVxf
6451 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
6455 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
6456 /* make sure SvREFCNT(sv)==0 happens very seldom */
6457 SvREFCNT(sv) = (~(U32)0)/2;
6468 Returns the length of the string in the SV. Handles magic and type
6469 coercion. See also C<SvCUR>, which gives raw access to the xpv_cur slot.
6475 Perl_sv_len(pTHX_ register SV *const sv)
6483 len = mg_length(sv);
6485 (void)SvPV_const(sv, len);
6490 =for apidoc sv_len_utf8
6492 Returns the number of characters in the string in an SV, counting wide
6493 UTF-8 bytes as a single character. Handles magic and type coercion.
6499 * The length is cached in PERL_MAGIC_utf8, in the mg_len field. Also the
6500 * mg_ptr is used, by sv_pos_u2b() and sv_pos_b2u() - see the comments below.
6501 * (Note that the mg_len is not the length of the mg_ptr field.
6502 * This allows the cache to store the character length of the string without
6503 * needing to malloc() extra storage to attach to the mg_ptr.)
6508 Perl_sv_len_utf8(pTHX_ register SV *const sv)
6514 return mg_length(sv);
6518 const U8 *s = (U8*)SvPV_const(sv, len);
6522 MAGIC *mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_utf8) : NULL;
6524 if (mg && (mg->mg_len != -1 || mg->mg_ptr)) {
6525 if (mg->mg_len != -1)
6528 /* We can use the offset cache for a headstart.
6529 The longer value is stored in the first pair. */
6530 STRLEN *cache = (STRLEN *) mg->mg_ptr;
6532 ulen = cache[0] + Perl_utf8_length(aTHX_ s + cache[1],
6536 if (PL_utf8cache < 0) {
6537 const STRLEN real = Perl_utf8_length(aTHX_ s, s + len);
6538 assert_uft8_cache_coherent("sv_len_utf8", ulen, real, sv);
6542 ulen = Perl_utf8_length(aTHX_ s, s + len);
6543 utf8_mg_len_cache_update(sv, &mg, ulen);
6547 return Perl_utf8_length(aTHX_ s, s + len);
6551 /* Walk forwards to find the byte corresponding to the passed in UTF-8
6554 S_sv_pos_u2b_forwards(const U8 *const start, const U8 *const send,
6555 STRLEN *const uoffset_p, bool *const at_end)
6557 const U8 *s = start;
6558 STRLEN uoffset = *uoffset_p;
6560 PERL_ARGS_ASSERT_SV_POS_U2B_FORWARDS;
6562 while (s < send && uoffset) {
6569 else if (s > send) {
6571 /* This is the existing behaviour. Possibly it should be a croak, as
6572 it's actually a bounds error */
6575 *uoffset_p -= uoffset;
6579 /* Given the length of the string in both bytes and UTF-8 characters, decide
6580 whether to walk forwards or backwards to find the byte corresponding to
6581 the passed in UTF-8 offset. */
6583 S_sv_pos_u2b_midway(const U8 *const start, const U8 *send,
6584 STRLEN uoffset, const STRLEN uend)
6586 STRLEN backw = uend - uoffset;
6588 PERL_ARGS_ASSERT_SV_POS_U2B_MIDWAY;
6590 if (uoffset < 2 * backw) {
6591 /* The assumption is that going forwards is twice the speed of going
6592 forward (that's where the 2 * backw comes from).
6593 (The real figure of course depends on the UTF-8 data.) */
6594 const U8 *s = start;
6596 while (s < send && uoffset--)
6606 while (UTF8_IS_CONTINUATION(*send))
6609 return send - start;
6612 /* For the string representation of the given scalar, find the byte
6613 corresponding to the passed in UTF-8 offset. uoffset0 and boffset0
6614 give another position in the string, *before* the sought offset, which
6615 (which is always true, as 0, 0 is a valid pair of positions), which should
6616 help reduce the amount of linear searching.
6617 If *mgp is non-NULL, it should point to the UTF-8 cache magic, which
6618 will be used to reduce the amount of linear searching. The cache will be
6619 created if necessary, and the found value offered to it for update. */
6621 S_sv_pos_u2b_cached(pTHX_ SV *const sv, MAGIC **const mgp, const U8 *const start,
6622 const U8 *const send, STRLEN uoffset,
6623 STRLEN uoffset0, STRLEN boffset0)
6625 STRLEN boffset = 0; /* Actually always set, but let's keep gcc happy. */
6627 bool at_end = FALSE;
6629 PERL_ARGS_ASSERT_SV_POS_U2B_CACHED;
6631 assert (uoffset >= uoffset0);
6638 && (*mgp || (SvTYPE(sv) >= SVt_PVMG &&
6639 (*mgp = mg_find(sv, PERL_MAGIC_utf8))))) {
6640 if ((*mgp)->mg_ptr) {
6641 STRLEN *cache = (STRLEN *) (*mgp)->mg_ptr;
6642 if (cache[0] == uoffset) {
6643 /* An exact match. */
6646 if (cache[2] == uoffset) {
6647 /* An exact match. */
6651 if (cache[0] < uoffset) {
6652 /* The cache already knows part of the way. */
6653 if (cache[0] > uoffset0) {
6654 /* The cache knows more than the passed in pair */
6655 uoffset0 = cache[0];
6656 boffset0 = cache[1];
6658 if ((*mgp)->mg_len != -1) {
6659 /* And we know the end too. */
6661 + sv_pos_u2b_midway(start + boffset0, send,
6663 (*mgp)->mg_len - uoffset0);
6665 uoffset -= uoffset0;
6667 + sv_pos_u2b_forwards(start + boffset0,
6668 send, &uoffset, &at_end);
6669 uoffset += uoffset0;
6672 else if (cache[2] < uoffset) {
6673 /* We're between the two cache entries. */
6674 if (cache[2] > uoffset0) {
6675 /* and the cache knows more than the passed in pair */
6676 uoffset0 = cache[2];
6677 boffset0 = cache[3];
6681 + sv_pos_u2b_midway(start + boffset0,
6684 cache[0] - uoffset0);
6687 + sv_pos_u2b_midway(start + boffset0,
6690 cache[2] - uoffset0);
6694 else if ((*mgp)->mg_len != -1) {
6695 /* If we can take advantage of a passed in offset, do so. */
6696 /* In fact, offset0 is either 0, or less than offset, so don't
6697 need to worry about the other possibility. */
6699 + sv_pos_u2b_midway(start + boffset0, send,
6701 (*mgp)->mg_len - uoffset0);
6706 if (!found || PL_utf8cache < 0) {
6707 STRLEN real_boffset;
6708 uoffset -= uoffset0;
6709 real_boffset = boffset0 + sv_pos_u2b_forwards(start + boffset0,
6710 send, &uoffset, &at_end);
6711 uoffset += uoffset0;
6713 if (found && PL_utf8cache < 0)
6714 assert_uft8_cache_coherent("sv_pos_u2b_cache", boffset,
6716 boffset = real_boffset;
6721 utf8_mg_len_cache_update(sv, mgp, uoffset);
6723 utf8_mg_pos_cache_update(sv, mgp, boffset, uoffset, send - start);
6730 =for apidoc sv_pos_u2b_flags
6732 Converts the value pointed to by offsetp from a count of UTF-8 chars from
6733 the start of the string, to a count of the equivalent number of bytes; if
6734 lenp is non-zero, it does the same to lenp, but this time starting from
6735 the offset, rather than from the start of the string. Handles type coercion.
6736 I<flags> is passed to C<SvPV_flags>, and usually should be
6737 C<SV_GMAGIC|SV_CONST_RETURN> to handle magic.
6743 * sv_pos_u2b_flags() uses, like sv_pos_b2u(), the mg_ptr of the potential
6744 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6745 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
6750 Perl_sv_pos_u2b_flags(pTHX_ SV *const sv, STRLEN uoffset, STRLEN *const lenp,
6757 PERL_ARGS_ASSERT_SV_POS_U2B_FLAGS;
6759 start = (U8*)SvPV_flags(sv, len, flags);
6761 const U8 * const send = start + len;
6763 boffset = sv_pos_u2b_cached(sv, &mg, start, send, uoffset, 0, 0);
6766 && *lenp /* don't bother doing work for 0, as its bytes equivalent
6767 is 0, and *lenp is already set to that. */) {
6768 /* Convert the relative offset to absolute. */
6769 const STRLEN uoffset2 = uoffset + *lenp;
6770 const STRLEN boffset2
6771 = sv_pos_u2b_cached(sv, &mg, start, send, uoffset2,
6772 uoffset, boffset) - boffset;
6786 =for apidoc sv_pos_u2b
6788 Converts the value pointed to by offsetp from a count of UTF-8 chars from
6789 the start of the string, to a count of the equivalent number of bytes; if
6790 lenp is non-zero, it does the same to lenp, but this time starting from
6791 the offset, rather than from the start of the string. Handles magic and
6794 Use C<sv_pos_u2b_flags> in preference, which correctly handles strings longer
6801 * sv_pos_u2b() uses, like sv_pos_b2u(), the mg_ptr of the potential
6802 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6803 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
6807 /* This function is subject to size and sign problems */
6810 Perl_sv_pos_u2b(pTHX_ register SV *const sv, I32 *const offsetp, I32 *const lenp)
6812 PERL_ARGS_ASSERT_SV_POS_U2B;
6815 STRLEN ulen = (STRLEN)*lenp;
6816 *offsetp = (I32)sv_pos_u2b_flags(sv, (STRLEN)*offsetp, &ulen,
6817 SV_GMAGIC|SV_CONST_RETURN);
6820 *offsetp = (I32)sv_pos_u2b_flags(sv, (STRLEN)*offsetp, NULL,
6821 SV_GMAGIC|SV_CONST_RETURN);
6826 S_utf8_mg_len_cache_update(pTHX_ SV *const sv, MAGIC **const mgp,
6829 PERL_ARGS_ASSERT_UTF8_MG_LEN_CACHE_UPDATE;
6833 if (!*mgp && (SvTYPE(sv) < SVt_PVMG ||
6834 !(*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
6835 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, &PL_vtbl_utf8, 0, 0);
6839 (*mgp)->mg_len = ulen;
6840 /* For now, treat "overflowed" as "still unknown". See RT #72924. */
6841 if (ulen != (STRLEN) (*mgp)->mg_len)
6842 (*mgp)->mg_len = -1;
6845 /* Create and update the UTF8 magic offset cache, with the proffered utf8/
6846 byte length pairing. The (byte) length of the total SV is passed in too,
6847 as blen, because for some (more esoteric) SVs, the call to SvPV_const()
6848 may not have updated SvCUR, so we can't rely on reading it directly.
6850 The proffered utf8/byte length pairing isn't used if the cache already has
6851 two pairs, and swapping either for the proffered pair would increase the
6852 RMS of the intervals between known byte offsets.
6854 The cache itself consists of 4 STRLEN values
6855 0: larger UTF-8 offset
6856 1: corresponding byte offset
6857 2: smaller UTF-8 offset
6858 3: corresponding byte offset
6860 Unused cache pairs have the value 0, 0.
6861 Keeping the cache "backwards" means that the invariant of
6862 cache[0] >= cache[2] is maintained even with empty slots, which means that
6863 the code that uses it doesn't need to worry if only 1 entry has actually
6864 been set to non-zero. It also makes the "position beyond the end of the
6865 cache" logic much simpler, as the first slot is always the one to start
6869 S_utf8_mg_pos_cache_update(pTHX_ SV *const sv, MAGIC **const mgp, const STRLEN byte,
6870 const STRLEN utf8, const STRLEN blen)
6874 PERL_ARGS_ASSERT_UTF8_MG_POS_CACHE_UPDATE;
6879 if (!*mgp && (SvTYPE(sv) < SVt_PVMG ||
6880 !(*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
6881 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, (MGVTBL*)&PL_vtbl_utf8, 0,
6883 (*mgp)->mg_len = -1;
6887 if (!(cache = (STRLEN *)(*mgp)->mg_ptr)) {
6888 Newxz(cache, PERL_MAGIC_UTF8_CACHESIZE * 2, STRLEN);
6889 (*mgp)->mg_ptr = (char *) cache;
6893 if (PL_utf8cache < 0 && SvPOKp(sv)) {
6894 /* SvPOKp() because it's possible that sv has string overloading, and
6895 therefore is a reference, hence SvPVX() is actually a pointer.
6896 This cures the (very real) symptoms of RT 69422, but I'm not actually
6897 sure whether we should even be caching the results of UTF-8
6898 operations on overloading, given that nothing stops overloading
6899 returning a different value every time it's called. */
6900 const U8 *start = (const U8 *) SvPVX_const(sv);
6901 const STRLEN realutf8 = utf8_length(start, start + byte);
6903 assert_uft8_cache_coherent("utf8_mg_pos_cache_update", utf8, realutf8,
6907 /* Cache is held with the later position first, to simplify the code
6908 that deals with unbounded ends. */
6910 ASSERT_UTF8_CACHE(cache);
6911 if (cache[1] == 0) {
6912 /* Cache is totally empty */
6915 } else if (cache[3] == 0) {
6916 if (byte > cache[1]) {
6917 /* New one is larger, so goes first. */
6918 cache[2] = cache[0];
6919 cache[3] = cache[1];
6927 #define THREEWAY_SQUARE(a,b,c,d) \
6928 ((float)((d) - (c))) * ((float)((d) - (c))) \
6929 + ((float)((c) - (b))) * ((float)((c) - (b))) \
6930 + ((float)((b) - (a))) * ((float)((b) - (a)))
6932 /* Cache has 2 slots in use, and we know three potential pairs.
6933 Keep the two that give the lowest RMS distance. Do the
6934 calculation in bytes simply because we always know the byte
6935 length. squareroot has the same ordering as the positive value,
6936 so don't bother with the actual square root. */
6937 const float existing = THREEWAY_SQUARE(0, cache[3], cache[1], blen);
6938 if (byte > cache[1]) {
6939 /* New position is after the existing pair of pairs. */
6940 const float keep_earlier
6941 = THREEWAY_SQUARE(0, cache[3], byte, blen);
6942 const float keep_later
6943 = THREEWAY_SQUARE(0, cache[1], byte, blen);
6945 if (keep_later < keep_earlier) {
6946 if (keep_later < existing) {
6947 cache[2] = cache[0];
6948 cache[3] = cache[1];
6954 if (keep_earlier < existing) {
6960 else if (byte > cache[3]) {
6961 /* New position is between the existing pair of pairs. */
6962 const float keep_earlier
6963 = THREEWAY_SQUARE(0, cache[3], byte, blen);
6964 const float keep_later
6965 = THREEWAY_SQUARE(0, byte, cache[1], blen);
6967 if (keep_later < keep_earlier) {
6968 if (keep_later < existing) {
6974 if (keep_earlier < existing) {
6981 /* New position is before the existing pair of pairs. */
6982 const float keep_earlier
6983 = THREEWAY_SQUARE(0, byte, cache[3], blen);
6984 const float keep_later
6985 = THREEWAY_SQUARE(0, byte, cache[1], blen);
6987 if (keep_later < keep_earlier) {
6988 if (keep_later < existing) {
6994 if (keep_earlier < existing) {
6995 cache[0] = cache[2];
6996 cache[1] = cache[3];
7003 ASSERT_UTF8_CACHE(cache);
7006 /* We already know all of the way, now we may be able to walk back. The same
7007 assumption is made as in S_sv_pos_u2b_midway(), namely that walking
7008 backward is half the speed of walking forward. */
7010 S_sv_pos_b2u_midway(pTHX_ const U8 *const s, const U8 *const target,
7011 const U8 *end, STRLEN endu)
7013 const STRLEN forw = target - s;
7014 STRLEN backw = end - target;
7016 PERL_ARGS_ASSERT_SV_POS_B2U_MIDWAY;
7018 if (forw < 2 * backw) {
7019 return utf8_length(s, target);
7022 while (end > target) {
7024 while (UTF8_IS_CONTINUATION(*end)) {
7033 =for apidoc sv_pos_b2u
7035 Converts the value pointed to by offsetp from a count of bytes from the
7036 start of the string, to a count of the equivalent number of UTF-8 chars.
7037 Handles magic and type coercion.
7043 * sv_pos_b2u() uses, like sv_pos_u2b(), the mg_ptr of the potential
7044 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
7049 Perl_sv_pos_b2u(pTHX_ register SV *const sv, I32 *const offsetp)
7052 const STRLEN byte = *offsetp;
7053 STRLEN len = 0; /* Actually always set, but let's keep gcc happy. */
7059 PERL_ARGS_ASSERT_SV_POS_B2U;
7064 s = (const U8*)SvPV_const(sv, blen);
7067 Perl_croak(aTHX_ "panic: sv_pos_b2u: bad byte offset");
7073 && SvTYPE(sv) >= SVt_PVMG
7074 && (mg = mg_find(sv, PERL_MAGIC_utf8)))
7077 STRLEN * const cache = (STRLEN *) mg->mg_ptr;
7078 if (cache[1] == byte) {
7079 /* An exact match. */
7080 *offsetp = cache[0];
7083 if (cache[3] == byte) {
7084 /* An exact match. */
7085 *offsetp = cache[2];
7089 if (cache[1] < byte) {
7090 /* We already know part of the way. */
7091 if (mg->mg_len != -1) {
7092 /* Actually, we know the end too. */
7094 + S_sv_pos_b2u_midway(aTHX_ s + cache[1], send,
7095 s + blen, mg->mg_len - cache[0]);
7097 len = cache[0] + utf8_length(s + cache[1], send);
7100 else if (cache[3] < byte) {
7101 /* We're between the two cached pairs, so we do the calculation
7102 offset by the byte/utf-8 positions for the earlier pair,
7103 then add the utf-8 characters from the string start to
7105 len = S_sv_pos_b2u_midway(aTHX_ s + cache[3], send,
7106 s + cache[1], cache[0] - cache[2])
7110 else { /* cache[3] > byte */
7111 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + cache[3],
7115 ASSERT_UTF8_CACHE(cache);
7117 } else if (mg->mg_len != -1) {
7118 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + blen, mg->mg_len);
7122 if (!found || PL_utf8cache < 0) {
7123 const STRLEN real_len = utf8_length(s, send);
7125 if (found && PL_utf8cache < 0)
7126 assert_uft8_cache_coherent("sv_pos_b2u", len, real_len, sv);
7133 utf8_mg_len_cache_update(sv, &mg, len);
7135 utf8_mg_pos_cache_update(sv, &mg, byte, len, blen);
7140 S_assert_uft8_cache_coherent(pTHX_ const char *const func, STRLEN from_cache,
7141 STRLEN real, SV *const sv)
7143 PERL_ARGS_ASSERT_ASSERT_UFT8_CACHE_COHERENT;
7145 /* As this is debugging only code, save space by keeping this test here,
7146 rather than inlining it in all the callers. */
7147 if (from_cache == real)
7150 /* Need to turn the assertions off otherwise we may recurse infinitely
7151 while printing error messages. */
7152 SAVEI8(PL_utf8cache);
7154 Perl_croak(aTHX_ "panic: %s cache %"UVuf" real %"UVuf" for %"SVf,
7155 func, (UV) from_cache, (UV) real, SVfARG(sv));
7161 Returns a boolean indicating whether the strings in the two SVs are
7162 identical. Is UTF-8 and 'use bytes' aware, handles get magic, and will
7163 coerce its args to strings if necessary.
7165 =for apidoc sv_eq_flags
7167 Returns a boolean indicating whether the strings in the two SVs are
7168 identical. Is UTF-8 and 'use bytes' aware and coerces its args to strings
7169 if necessary. If the flags include SV_GMAGIC, it handles get-magic, too.
7175 Perl_sv_eq_flags(pTHX_ register SV *sv1, register SV *sv2, const U32 flags)
7184 SV* svrecode = NULL;
7191 /* if pv1 and pv2 are the same, second SvPV_const call may
7192 * invalidate pv1 (if we are handling magic), so we may need to
7194 if (sv1 == sv2 && flags & SV_GMAGIC
7195 && (SvTHINKFIRST(sv1) || SvGMAGICAL(sv1))) {
7196 pv1 = SvPV_const(sv1, cur1);
7197 sv1 = newSVpvn_flags(pv1, cur1, SVs_TEMP | SvUTF8(sv2));
7199 pv1 = SvPV_flags_const(sv1, cur1, flags);
7207 pv2 = SvPV_flags_const(sv2, cur2, flags);
7209 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
7210 /* Differing utf8ness.
7211 * Do not UTF8size the comparands as a side-effect. */
7214 svrecode = newSVpvn(pv2, cur2);
7215 sv_recode_to_utf8(svrecode, PL_encoding);
7216 pv2 = SvPV_const(svrecode, cur2);
7219 svrecode = newSVpvn(pv1, cur1);
7220 sv_recode_to_utf8(svrecode, PL_encoding);
7221 pv1 = SvPV_const(svrecode, cur1);
7223 /* Now both are in UTF-8. */
7225 SvREFCNT_dec(svrecode);
7231 /* sv1 is the UTF-8 one */
7232 return bytes_cmp_utf8((const U8*)pv2, cur2,
7233 (const U8*)pv1, cur1) == 0;
7236 /* sv2 is the UTF-8 one */
7237 return bytes_cmp_utf8((const U8*)pv1, cur1,
7238 (const U8*)pv2, cur2) == 0;
7244 eq = (pv1 == pv2) || memEQ(pv1, pv2, cur1);
7246 SvREFCNT_dec(svrecode);
7256 Compares the strings in two SVs. Returns -1, 0, or 1 indicating whether the
7257 string in C<sv1> is less than, equal to, or greater than the string in
7258 C<sv2>. Is UTF-8 and 'use bytes' aware, handles get magic, and will
7259 coerce its args to strings if necessary. See also C<sv_cmp_locale>.
7261 =for apidoc sv_cmp_flags
7263 Compares the strings in two SVs. Returns -1, 0, or 1 indicating whether the
7264 string in C<sv1> is less than, equal to, or greater than the string in
7265 C<sv2>. Is UTF-8 and 'use bytes' aware and will coerce its args to strings
7266 if necessary. If the flags include SV_GMAGIC, it handles get magic. See
7267 also C<sv_cmp_locale_flags>.
7273 Perl_sv_cmp(pTHX_ register SV *const sv1, register SV *const sv2)
7275 return sv_cmp_flags(sv1, sv2, SV_GMAGIC);
7279 Perl_sv_cmp_flags(pTHX_ register SV *const sv1, register SV *const sv2,
7284 const char *pv1, *pv2;
7287 SV *svrecode = NULL;
7294 pv1 = SvPV_flags_const(sv1, cur1, flags);
7301 pv2 = SvPV_flags_const(sv2, cur2, flags);
7303 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
7304 /* Differing utf8ness.
7305 * Do not UTF8size the comparands as a side-effect. */
7308 svrecode = newSVpvn(pv2, cur2);
7309 sv_recode_to_utf8(svrecode, PL_encoding);
7310 pv2 = SvPV_const(svrecode, cur2);
7313 const int retval = -bytes_cmp_utf8((const U8*)pv2, cur2,
7314 (const U8*)pv1, cur1);
7315 return retval ? retval < 0 ? -1 : +1 : 0;
7320 svrecode = newSVpvn(pv1, cur1);
7321 sv_recode_to_utf8(svrecode, PL_encoding);
7322 pv1 = SvPV_const(svrecode, cur1);
7325 const int retval = bytes_cmp_utf8((const U8*)pv1, cur1,
7326 (const U8*)pv2, cur2);
7327 return retval ? retval < 0 ? -1 : +1 : 0;
7333 cmp = cur2 ? -1 : 0;
7337 const I32 retval = memcmp((const void*)pv1, (const void*)pv2, cur1 < cur2 ? cur1 : cur2);
7340 cmp = retval < 0 ? -1 : 1;
7341 } else if (cur1 == cur2) {
7344 cmp = cur1 < cur2 ? -1 : 1;
7348 SvREFCNT_dec(svrecode);
7356 =for apidoc sv_cmp_locale
7358 Compares the strings in two SVs in a locale-aware manner. Is UTF-8 and
7359 'use bytes' aware, handles get magic, and will coerce its args to strings
7360 if necessary. See also C<sv_cmp>.
7362 =for apidoc sv_cmp_locale_flags
7364 Compares the strings in two SVs in a locale-aware manner. Is UTF-8 and
7365 'use bytes' aware and will coerce its args to strings if necessary. If the
7366 flags contain SV_GMAGIC, it handles get magic. See also C<sv_cmp_flags>.
7372 Perl_sv_cmp_locale(pTHX_ register SV *const sv1, register SV *const sv2)
7374 return sv_cmp_locale_flags(sv1, sv2, SV_GMAGIC);
7378 Perl_sv_cmp_locale_flags(pTHX_ register SV *const sv1, register SV *const sv2,
7382 #ifdef USE_LOCALE_COLLATE
7388 if (PL_collation_standard)
7392 pv1 = sv1 ? sv_collxfrm_flags(sv1, &len1, flags) : (char *) NULL;
7394 pv2 = sv2 ? sv_collxfrm_flags(sv2, &len2, flags) : (char *) NULL;
7396 if (!pv1 || !len1) {
7407 retval = memcmp((void*)pv1, (void*)pv2, len1 < len2 ? len1 : len2);
7410 return retval < 0 ? -1 : 1;
7413 * When the result of collation is equality, that doesn't mean
7414 * that there are no differences -- some locales exclude some
7415 * characters from consideration. So to avoid false equalities,
7416 * we use the raw string as a tiebreaker.
7422 #endif /* USE_LOCALE_COLLATE */
7424 return sv_cmp(sv1, sv2);
7428 #ifdef USE_LOCALE_COLLATE
7431 =for apidoc sv_collxfrm
7433 This calls C<sv_collxfrm_flags> with the SV_GMAGIC flag. See
7434 C<sv_collxfrm_flags>.
7436 =for apidoc sv_collxfrm_flags
7438 Add Collate Transform magic to an SV if it doesn't already have it. If the
7439 flags contain SV_GMAGIC, it handles get-magic.
7441 Any scalar variable may carry PERL_MAGIC_collxfrm magic that contains the
7442 scalar data of the variable, but transformed to such a format that a normal
7443 memory comparison can be used to compare the data according to the locale
7450 Perl_sv_collxfrm_flags(pTHX_ SV *const sv, STRLEN *const nxp, const I32 flags)
7455 PERL_ARGS_ASSERT_SV_COLLXFRM_FLAGS;
7457 mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_collxfrm) : (MAGIC *) NULL;
7458 if (!mg || !mg->mg_ptr || *(U32*)mg->mg_ptr != PL_collation_ix) {
7464 Safefree(mg->mg_ptr);
7465 s = SvPV_flags_const(sv, len, flags);
7466 if ((xf = mem_collxfrm(s, len, &xlen))) {
7468 #ifdef PERL_OLD_COPY_ON_WRITE
7470 sv_force_normal_flags(sv, 0);
7472 mg = sv_magicext(sv, 0, PERL_MAGIC_collxfrm, &PL_vtbl_collxfrm,
7486 if (mg && mg->mg_ptr) {
7488 return mg->mg_ptr + sizeof(PL_collation_ix);
7496 #endif /* USE_LOCALE_COLLATE */
7499 S_sv_gets_append_to_utf8(pTHX_ SV *const sv, PerlIO *const fp, I32 append)
7501 SV * const tsv = newSV(0);
7504 sv_gets(tsv, fp, 0);
7505 sv_utf8_upgrade_nomg(tsv);
7506 SvCUR_set(sv,append);
7509 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7513 S_sv_gets_read_record(pTHX_ SV *const sv, PerlIO *const fp, I32 append)
7516 const U32 recsize = SvUV(SvRV(PL_rs)); /* RsRECORD() guarantees > 0. */
7517 /* Grab the size of the record we're getting */
7518 char *const buffer = SvGROW(sv, (STRLEN)(recsize + append + 1)) + append;
7525 /* VMS wants read instead of fread, because fread doesn't respect */
7526 /* RMS record boundaries. This is not necessarily a good thing to be */
7527 /* doing, but we've got no other real choice - except avoid stdio
7528 as implementation - perhaps write a :vms layer ?
7530 fd = PerlIO_fileno(fp);
7532 bytesread = PerlLIO_read(fd, buffer, recsize);
7534 else /* in-memory file from PerlIO::Scalar */
7537 bytesread = PerlIO_read(fp, buffer, recsize);
7542 SvCUR_set(sv, bytesread + append);
7543 buffer[bytesread] = '\0';
7544 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7550 Get a line from the filehandle and store it into the SV, optionally
7551 appending to the currently-stored string.
7557 Perl_sv_gets(pTHX_ register SV *const sv, register PerlIO *const fp, I32 append)
7562 register STDCHAR rslast;
7563 register STDCHAR *bp;
7568 PERL_ARGS_ASSERT_SV_GETS;
7570 if (SvTHINKFIRST(sv))
7571 sv_force_normal_flags(sv, append ? 0 : SV_COW_DROP_PV);
7572 /* XXX. If you make this PVIV, then copy on write can copy scalars read
7574 However, perlbench says it's slower, because the existing swipe code
7575 is faster than copy on write.
7576 Swings and roundabouts. */
7577 SvUPGRADE(sv, SVt_PV);
7582 if (PerlIO_isutf8(fp)) {
7584 sv_utf8_upgrade_nomg(sv);
7585 sv_pos_u2b(sv,&append,0);
7587 } else if (SvUTF8(sv)) {
7588 return S_sv_gets_append_to_utf8(aTHX_ sv, fp, append);
7596 if (PerlIO_isutf8(fp))
7599 if (IN_PERL_COMPILETIME) {
7600 /* we always read code in line mode */
7604 else if (RsSNARF(PL_rs)) {
7605 /* If it is a regular disk file use size from stat() as estimate
7606 of amount we are going to read -- may result in mallocing
7607 more memory than we really need if the layers below reduce
7608 the size we read (e.g. CRLF or a gzip layer).
7611 if (!PerlLIO_fstat(PerlIO_fileno(fp), &st) && S_ISREG(st.st_mode)) {
7612 const Off_t offset = PerlIO_tell(fp);
7613 if (offset != (Off_t) -1 && st.st_size + append > offset) {
7614 (void) SvGROW(sv, (STRLEN)((st.st_size - offset) + append + 1));
7620 else if (RsRECORD(PL_rs)) {
7621 return S_sv_gets_read_record(aTHX_ sv, fp, append);
7623 else if (RsPARA(PL_rs)) {
7629 /* Get $/ i.e. PL_rs into same encoding as stream wants */
7630 if (PerlIO_isutf8(fp)) {
7631 rsptr = SvPVutf8(PL_rs, rslen);
7634 if (SvUTF8(PL_rs)) {
7635 if (!sv_utf8_downgrade(PL_rs, TRUE)) {
7636 Perl_croak(aTHX_ "Wide character in $/");
7639 rsptr = SvPV_const(PL_rs, rslen);
7643 rslast = rslen ? rsptr[rslen - 1] : '\0';
7645 if (rspara) { /* have to do this both before and after */
7646 do { /* to make sure file boundaries work right */
7649 i = PerlIO_getc(fp);
7653 PerlIO_ungetc(fp,i);
7659 /* See if we know enough about I/O mechanism to cheat it ! */
7661 /* This used to be #ifdef test - it is made run-time test for ease
7662 of abstracting out stdio interface. One call should be cheap
7663 enough here - and may even be a macro allowing compile
7667 if (PerlIO_fast_gets(fp)) {
7670 * We're going to steal some values from the stdio struct
7671 * and put EVERYTHING in the innermost loop into registers.
7673 register STDCHAR *ptr;
7677 #if defined(VMS) && defined(PERLIO_IS_STDIO)
7678 /* An ungetc()d char is handled separately from the regular
7679 * buffer, so we getc() it back out and stuff it in the buffer.
7681 i = PerlIO_getc(fp);
7682 if (i == EOF) return 0;
7683 *(--((*fp)->_ptr)) = (unsigned char) i;
7687 /* Here is some breathtakingly efficient cheating */
7689 cnt = PerlIO_get_cnt(fp); /* get count into register */
7690 /* make sure we have the room */
7691 if ((I32)(SvLEN(sv) - append) <= cnt + 1) {
7692 /* Not room for all of it
7693 if we are looking for a separator and room for some
7695 if (rslen && cnt > 80 && (I32)SvLEN(sv) > append) {
7696 /* just process what we have room for */
7697 shortbuffered = cnt - SvLEN(sv) + append + 1;
7698 cnt -= shortbuffered;
7702 /* remember that cnt can be negative */
7703 SvGROW(sv, (STRLEN)(append + (cnt <= 0 ? 2 : (cnt + 1))));
7708 bp = (STDCHAR*)SvPVX_const(sv) + append; /* move these two too to registers */
7709 ptr = (STDCHAR*)PerlIO_get_ptr(fp);
7710 DEBUG_P(PerlIO_printf(Perl_debug_log,
7711 "Screamer: entering, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7712 DEBUG_P(PerlIO_printf(Perl_debug_log,
7713 "Screamer: entering: PerlIO * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7714 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7715 PTR2UV(PerlIO_has_base(fp) ? PerlIO_get_base(fp) : 0)));
7720 while (cnt > 0) { /* this | eat */
7722 if ((*bp++ = *ptr++) == rslast) /* really | dust */
7723 goto thats_all_folks; /* screams | sed :-) */
7727 Copy(ptr, bp, cnt, char); /* this | eat */
7728 bp += cnt; /* screams | dust */
7729 ptr += cnt; /* louder | sed :-) */
7731 assert (!shortbuffered);
7732 goto cannot_be_shortbuffered;
7736 if (shortbuffered) { /* oh well, must extend */
7737 cnt = shortbuffered;
7739 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
7741 SvGROW(sv, SvLEN(sv) + append + cnt + 2);
7742 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
7746 cannot_be_shortbuffered:
7747 DEBUG_P(PerlIO_printf(Perl_debug_log,
7748 "Screamer: going to getc, ptr=%"UVuf", cnt=%ld\n",
7749 PTR2UV(ptr),(long)cnt));
7750 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* deregisterize cnt and ptr */
7752 DEBUG_Pv(PerlIO_printf(Perl_debug_log,
7753 "Screamer: pre: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7754 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7755 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7757 /* This used to call 'filbuf' in stdio form, but as that behaves like
7758 getc when cnt <= 0 we use PerlIO_getc here to avoid introducing
7759 another abstraction. */
7760 i = PerlIO_getc(fp); /* get more characters */
7762 DEBUG_Pv(PerlIO_printf(Perl_debug_log,
7763 "Screamer: post: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7764 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7765 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7767 cnt = PerlIO_get_cnt(fp);
7768 ptr = (STDCHAR*)PerlIO_get_ptr(fp); /* reregisterize cnt and ptr */
7769 DEBUG_P(PerlIO_printf(Perl_debug_log,
7770 "Screamer: after getc, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7772 if (i == EOF) /* all done for ever? */
7773 goto thats_really_all_folks;
7775 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
7777 SvGROW(sv, bpx + cnt + 2);
7778 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
7780 *bp++ = (STDCHAR)i; /* store character from PerlIO_getc */
7782 if (rslen && (STDCHAR)i == rslast) /* all done for now? */
7783 goto thats_all_folks;
7787 if ((rslen > 1 && (STRLEN)(bp - (STDCHAR*)SvPVX_const(sv)) < rslen) ||
7788 memNE((char*)bp - rslen, rsptr, rslen))
7789 goto screamer; /* go back to the fray */
7790 thats_really_all_folks:
7792 cnt += shortbuffered;
7793 DEBUG_P(PerlIO_printf(Perl_debug_log,
7794 "Screamer: quitting, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7795 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* put these back or we're in trouble */
7796 DEBUG_P(PerlIO_printf(Perl_debug_log,
7797 "Screamer: end: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7798 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7799 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7801 SvCUR_set(sv, bp - (STDCHAR*)SvPVX_const(sv)); /* set length */
7802 DEBUG_P(PerlIO_printf(Perl_debug_log,
7803 "Screamer: done, len=%ld, string=|%.*s|\n",
7804 (long)SvCUR(sv),(int)SvCUR(sv),SvPVX_const(sv)));
7808 /*The big, slow, and stupid way. */
7809 #ifdef USE_HEAP_INSTEAD_OF_STACK /* Even slower way. */
7810 STDCHAR *buf = NULL;
7811 Newx(buf, 8192, STDCHAR);
7819 register const STDCHAR * const bpe = buf + sizeof(buf);
7821 while ((i = PerlIO_getc(fp)) != EOF && (*bp++ = (STDCHAR)i) != rslast && bp < bpe)
7822 ; /* keep reading */
7826 cnt = PerlIO_read(fp,(char*)buf, sizeof(buf));
7827 /* Accommodate broken VAXC compiler, which applies U8 cast to
7828 * both args of ?: operator, causing EOF to change into 255
7831 i = (U8)buf[cnt - 1];
7837 cnt = 0; /* we do need to re-set the sv even when cnt <= 0 */
7839 sv_catpvn(sv, (char *) buf, cnt);
7841 sv_setpvn(sv, (char *) buf, cnt);
7843 if (i != EOF && /* joy */
7845 SvCUR(sv) < rslen ||
7846 memNE(SvPVX_const(sv) + SvCUR(sv) - rslen, rsptr, rslen)))
7850 * If we're reading from a TTY and we get a short read,
7851 * indicating that the user hit his EOF character, we need
7852 * to notice it now, because if we try to read from the TTY
7853 * again, the EOF condition will disappear.
7855 * The comparison of cnt to sizeof(buf) is an optimization
7856 * that prevents unnecessary calls to feof().
7860 if (!(cnt < (I32)sizeof(buf) && PerlIO_eof(fp)))
7864 #ifdef USE_HEAP_INSTEAD_OF_STACK
7869 if (rspara) { /* have to do this both before and after */
7870 while (i != EOF) { /* to make sure file boundaries work right */
7871 i = PerlIO_getc(fp);
7873 PerlIO_ungetc(fp,i);
7879 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7885 Auto-increment of the value in the SV, doing string to numeric conversion
7886 if necessary. Handles 'get' magic and operator overloading.
7892 Perl_sv_inc(pTHX_ register SV *const sv)
7901 =for apidoc sv_inc_nomg
7903 Auto-increment of the value in the SV, doing string to numeric conversion
7904 if necessary. Handles operator overloading. Skips handling 'get' magic.
7910 Perl_sv_inc_nomg(pTHX_ register SV *const sv)
7918 if (SvTHINKFIRST(sv)) {
7919 if (SvIsCOW(sv) || isGV_with_GP(sv))
7920 sv_force_normal_flags(sv, 0);
7921 if (SvREADONLY(sv)) {
7922 if (IN_PERL_RUNTIME)
7923 Perl_croak_no_modify(aTHX);
7927 if (SvAMAGIC(sv) && AMG_CALLunary(sv, inc_amg))
7929 i = PTR2IV(SvRV(sv));
7934 flags = SvFLAGS(sv);
7935 if ((flags & (SVp_NOK|SVp_IOK)) == SVp_NOK) {
7936 /* It's (privately or publicly) a float, but not tested as an
7937 integer, so test it to see. */
7939 flags = SvFLAGS(sv);
7941 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
7942 /* It's publicly an integer, or privately an integer-not-float */
7943 #ifdef PERL_PRESERVE_IVUV
7947 if (SvUVX(sv) == UV_MAX)
7948 sv_setnv(sv, UV_MAX_P1);
7950 (void)SvIOK_only_UV(sv);
7951 SvUV_set(sv, SvUVX(sv) + 1);
7953 if (SvIVX(sv) == IV_MAX)
7954 sv_setuv(sv, (UV)IV_MAX + 1);
7956 (void)SvIOK_only(sv);
7957 SvIV_set(sv, SvIVX(sv) + 1);
7962 if (flags & SVp_NOK) {
7963 const NV was = SvNVX(sv);
7964 if (NV_OVERFLOWS_INTEGERS_AT &&
7965 was >= NV_OVERFLOWS_INTEGERS_AT) {
7966 Perl_ck_warner(aTHX_ packWARN(WARN_IMPRECISION),
7967 "Lost precision when incrementing %" NVff " by 1",
7970 (void)SvNOK_only(sv);
7971 SvNV_set(sv, was + 1.0);
7975 if (!(flags & SVp_POK) || !*SvPVX_const(sv)) {
7976 if ((flags & SVTYPEMASK) < SVt_PVIV)
7977 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV ? SVt_PVIV : SVt_IV));
7978 (void)SvIOK_only(sv);
7983 while (isALPHA(*d)) d++;
7984 while (isDIGIT(*d)) d++;
7985 if (d < SvEND(sv)) {
7986 #ifdef PERL_PRESERVE_IVUV
7987 /* Got to punt this as an integer if needs be, but we don't issue
7988 warnings. Probably ought to make the sv_iv_please() that does
7989 the conversion if possible, and silently. */
7990 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
7991 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
7992 /* Need to try really hard to see if it's an integer.
7993 9.22337203685478e+18 is an integer.
7994 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
7995 so $a="9.22337203685478e+18"; $a+0; $a++
7996 needs to be the same as $a="9.22337203685478e+18"; $a++
8003 /* sv_2iv *should* have made this an NV */
8004 if (flags & SVp_NOK) {
8005 (void)SvNOK_only(sv);
8006 SvNV_set(sv, SvNVX(sv) + 1.0);
8009 /* I don't think we can get here. Maybe I should assert this
8010 And if we do get here I suspect that sv_setnv will croak. NWC
8012 #if defined(USE_LONG_DOUBLE)
8013 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",
8014 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
8016 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_inc punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
8017 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
8020 #endif /* PERL_PRESERVE_IVUV */
8021 sv_setnv(sv,Atof(SvPVX_const(sv)) + 1.0);
8025 while (d >= SvPVX_const(sv)) {
8033 /* MKS: The original code here died if letters weren't consecutive.
8034 * at least it didn't have to worry about non-C locales. The
8035 * new code assumes that ('z'-'a')==('Z'-'A'), letters are
8036 * arranged in order (although not consecutively) and that only
8037 * [A-Za-z] are accepted by isALPHA in the C locale.
8039 if (*d != 'z' && *d != 'Z') {
8040 do { ++*d; } while (!isALPHA(*d));
8043 *(d--) -= 'z' - 'a';
8048 *(d--) -= 'z' - 'a' + 1;
8052 /* oh,oh, the number grew */
8053 SvGROW(sv, SvCUR(sv) + 2);
8054 SvCUR_set(sv, SvCUR(sv) + 1);
8055 for (d = SvPVX(sv) + SvCUR(sv); d > SvPVX_const(sv); d--)
8066 Auto-decrement of the value in the SV, doing string to numeric conversion
8067 if necessary. Handles 'get' magic and operator overloading.
8073 Perl_sv_dec(pTHX_ register SV *const sv)
8083 =for apidoc sv_dec_nomg
8085 Auto-decrement of the value in the SV, doing string to numeric conversion
8086 if necessary. Handles operator overloading. Skips handling 'get' magic.
8092 Perl_sv_dec_nomg(pTHX_ register SV *const sv)
8099 if (SvTHINKFIRST(sv)) {
8100 if (SvIsCOW(sv) || isGV_with_GP(sv))
8101 sv_force_normal_flags(sv, 0);
8102 if (SvREADONLY(sv)) {
8103 if (IN_PERL_RUNTIME)
8104 Perl_croak_no_modify(aTHX);
8108 if (SvAMAGIC(sv) && AMG_CALLunary(sv, dec_amg))
8110 i = PTR2IV(SvRV(sv));
8115 /* Unlike sv_inc we don't have to worry about string-never-numbers
8116 and keeping them magic. But we mustn't warn on punting */
8117 flags = SvFLAGS(sv);
8118 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
8119 /* It's publicly an integer, or privately an integer-not-float */
8120 #ifdef PERL_PRESERVE_IVUV
8124 if (SvUVX(sv) == 0) {
8125 (void)SvIOK_only(sv);
8129 (void)SvIOK_only_UV(sv);
8130 SvUV_set(sv, SvUVX(sv) - 1);
8133 if (SvIVX(sv) == IV_MIN) {
8134 sv_setnv(sv, (NV)IV_MIN);
8138 (void)SvIOK_only(sv);
8139 SvIV_set(sv, SvIVX(sv) - 1);
8144 if (flags & SVp_NOK) {
8147 const NV was = SvNVX(sv);
8148 if (NV_OVERFLOWS_INTEGERS_AT &&
8149 was <= -NV_OVERFLOWS_INTEGERS_AT) {
8150 Perl_ck_warner(aTHX_ packWARN(WARN_IMPRECISION),
8151 "Lost precision when decrementing %" NVff " by 1",
8154 (void)SvNOK_only(sv);
8155 SvNV_set(sv, was - 1.0);
8159 if (!(flags & SVp_POK)) {
8160 if ((flags & SVTYPEMASK) < SVt_PVIV)
8161 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV) ? SVt_PVIV : SVt_IV);
8163 (void)SvIOK_only(sv);
8166 #ifdef PERL_PRESERVE_IVUV
8168 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
8169 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
8170 /* Need to try really hard to see if it's an integer.
8171 9.22337203685478e+18 is an integer.
8172 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
8173 so $a="9.22337203685478e+18"; $a+0; $a--
8174 needs to be the same as $a="9.22337203685478e+18"; $a--
8181 /* sv_2iv *should* have made this an NV */
8182 if (flags & SVp_NOK) {
8183 (void)SvNOK_only(sv);
8184 SvNV_set(sv, SvNVX(sv) - 1.0);
8187 /* I don't think we can get here. Maybe I should assert this
8188 And if we do get here I suspect that sv_setnv will croak. NWC
8190 #if defined(USE_LONG_DOUBLE)
8191 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",
8192 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
8194 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_dec punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
8195 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
8199 #endif /* PERL_PRESERVE_IVUV */
8200 sv_setnv(sv,Atof(SvPVX_const(sv)) - 1.0); /* punt */
8203 /* this define is used to eliminate a chunk of duplicated but shared logic
8204 * it has the suffix __SV_C to signal that it isnt API, and isnt meant to be
8205 * used anywhere but here - yves
8207 #define PUSH_EXTEND_MORTAL__SV_C(AnSv) \
8210 PL_tmps_stack[++PL_tmps_ix] = (AnSv); \
8214 =for apidoc sv_mortalcopy
8216 Creates a new SV which is a copy of the original SV (using C<sv_setsv>).
8217 The new SV is marked as mortal. It will be destroyed "soon", either by an
8218 explicit call to FREETMPS, or by an implicit call at places such as
8219 statement boundaries. See also C<sv_newmortal> and C<sv_2mortal>.
8224 /* Make a string that will exist for the duration of the expression
8225 * evaluation. Actually, it may have to last longer than that, but
8226 * hopefully we won't free it until it has been assigned to a
8227 * permanent location. */
8230 Perl_sv_mortalcopy(pTHX_ SV *const oldstr)
8236 sv_setsv(sv,oldstr);
8237 PUSH_EXTEND_MORTAL__SV_C(sv);
8243 =for apidoc sv_newmortal
8245 Creates a new null SV which is mortal. The reference count of the SV is
8246 set to 1. It will be destroyed "soon", either by an explicit call to
8247 FREETMPS, or by an implicit call at places such as statement boundaries.
8248 See also C<sv_mortalcopy> and C<sv_2mortal>.
8254 Perl_sv_newmortal(pTHX)
8260 SvFLAGS(sv) = SVs_TEMP;
8261 PUSH_EXTEND_MORTAL__SV_C(sv);
8267 =for apidoc newSVpvn_flags
8269 Creates a new SV and copies a string into it. The reference count for the
8270 SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
8271 string. You are responsible for ensuring that the source string is at least
8272 C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
8273 Currently the only flag bits accepted are C<SVf_UTF8> and C<SVs_TEMP>.
8274 If C<SVs_TEMP> is set, then C<sv_2mortal()> is called on the result before
8275 returning. If C<SVf_UTF8> is set, C<s> is considered to be in UTF-8 and the
8276 C<SVf_UTF8> flag will be set on the new SV.
8277 C<newSVpvn_utf8()> is a convenience wrapper for this function, defined as
8279 #define newSVpvn_utf8(s, len, u) \
8280 newSVpvn_flags((s), (len), (u) ? SVf_UTF8 : 0)
8286 Perl_newSVpvn_flags(pTHX_ const char *const s, const STRLEN len, const U32 flags)
8291 /* All the flags we don't support must be zero.
8292 And we're new code so I'm going to assert this from the start. */
8293 assert(!(flags & ~(SVf_UTF8|SVs_TEMP)));
8295 sv_setpvn(sv,s,len);
8297 /* This code used to a sv_2mortal(), however we now unroll the call to sv_2mortal()
8298 * and do what it does ourselves here.
8299 * Since we have asserted that flags can only have the SVf_UTF8 and/or SVs_TEMP flags
8300 * set above we can use it to enable the sv flags directly (bypassing SvTEMP_on), which
8301 * in turn means we dont need to mask out the SVf_UTF8 flag below, which means that we
8302 * eliminate quite a few steps than it looks - Yves (explaining patch by gfx)
8305 SvFLAGS(sv) |= flags;
8307 if(flags & SVs_TEMP){
8308 PUSH_EXTEND_MORTAL__SV_C(sv);
8315 =for apidoc sv_2mortal
8317 Marks an existing SV as mortal. The SV will be destroyed "soon", either
8318 by an explicit call to FREETMPS, or by an implicit call at places such as
8319 statement boundaries. SvTEMP() is turned on which means that the SV's
8320 string buffer can be "stolen" if this SV is copied. See also C<sv_newmortal>
8321 and C<sv_mortalcopy>.
8327 Perl_sv_2mortal(pTHX_ register SV *const sv)
8332 if (SvREADONLY(sv) && SvIMMORTAL(sv))
8334 PUSH_EXTEND_MORTAL__SV_C(sv);
8342 Creates a new SV and copies a string into it. The reference count for the
8343 SV is set to 1. If C<len> is zero, Perl will compute the length using
8344 strlen(). For efficiency, consider using C<newSVpvn> instead.
8350 Perl_newSVpv(pTHX_ const char *const s, const STRLEN len)
8356 sv_setpvn(sv, s, len || s == NULL ? len : strlen(s));
8361 =for apidoc newSVpvn
8363 Creates a new SV and copies a string into it. The reference count for the
8364 SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
8365 string. You are responsible for ensuring that the source string is at least
8366 C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
8372 Perl_newSVpvn(pTHX_ const char *const s, const STRLEN len)
8378 sv_setpvn(sv,s,len);
8383 =for apidoc newSVhek
8385 Creates a new SV from the hash key structure. It will generate scalars that
8386 point to the shared string table where possible. Returns a new (undefined)
8387 SV if the hek is NULL.
8393 Perl_newSVhek(pTHX_ const HEK *const hek)
8403 if (HEK_LEN(hek) == HEf_SVKEY) {
8404 return newSVsv(*(SV**)HEK_KEY(hek));
8406 const int flags = HEK_FLAGS(hek);
8407 if (flags & HVhek_WASUTF8) {
8409 Andreas would like keys he put in as utf8 to come back as utf8
8411 STRLEN utf8_len = HEK_LEN(hek);
8412 SV * const sv = newSV_type(SVt_PV);
8413 char *as_utf8 = (char *)bytes_to_utf8 ((U8*)HEK_KEY(hek), &utf8_len);
8414 /* bytes_to_utf8() allocates a new string, which we can repurpose: */
8415 sv_usepvn_flags(sv, as_utf8, utf8_len, SV_HAS_TRAILING_NUL);
8418 } else if (flags & (HVhek_REHASH|HVhek_UNSHARED)) {
8419 /* We don't have a pointer to the hv, so we have to replicate the
8420 flag into every HEK. This hv is using custom a hasing
8421 algorithm. Hence we can't return a shared string scalar, as
8422 that would contain the (wrong) hash value, and might get passed
8423 into an hv routine with a regular hash.
8424 Similarly, a hash that isn't using shared hash keys has to have
8425 the flag in every key so that we know not to try to call
8426 share_hek_hek on it. */
8428 SV * const sv = newSVpvn (HEK_KEY(hek), HEK_LEN(hek));
8433 /* This will be overwhelminly the most common case. */
8435 /* Inline most of newSVpvn_share(), because share_hek_hek() is far
8436 more efficient than sharepvn(). */
8440 sv_upgrade(sv, SVt_PV);
8441 SvPV_set(sv, (char *)HEK_KEY(share_hek_hek(hek)));
8442 SvCUR_set(sv, HEK_LEN(hek));
8455 =for apidoc newSVpvn_share
8457 Creates a new SV with its SvPVX_const pointing to a shared string in the string
8458 table. If the string does not already exist in the table, it is created
8459 first. Turns on READONLY and FAKE. If the C<hash> parameter is non-zero, that
8460 value is used; otherwise the hash is computed. The string's hash can be later
8461 be retrieved from the SV with the C<SvSHARED_HASH()> macro. The idea here is
8462 that as the string table is used for shared hash keys these strings will have
8463 SvPVX_const == HeKEY and hash lookup will avoid string compare.
8469 Perl_newSVpvn_share(pTHX_ const char *src, I32 len, U32 hash)
8473 bool is_utf8 = FALSE;
8474 const char *const orig_src = src;
8477 STRLEN tmplen = -len;
8479 /* See the note in hv.c:hv_fetch() --jhi */
8480 src = (char*)bytes_from_utf8((const U8*)src, &tmplen, &is_utf8);
8484 PERL_HASH(hash, src, len);
8486 /* The logic for this is inlined in S_mro_get_linear_isa_dfs(), so if it
8487 changes here, update it there too. */
8488 sv_upgrade(sv, SVt_PV);
8489 SvPV_set(sv, sharepvn(src, is_utf8?-len:len, hash));
8497 if (src != orig_src)
8503 =for apidoc newSVpv_share
8505 Like C<newSVpvn_share>, but takes a nul-terminated string instead of a
8512 Perl_newSVpv_share(pTHX_ const char *src, U32 hash)
8514 return newSVpvn_share(src, strlen(src), hash);
8517 #if defined(PERL_IMPLICIT_CONTEXT)
8519 /* pTHX_ magic can't cope with varargs, so this is a no-context
8520 * version of the main function, (which may itself be aliased to us).
8521 * Don't access this version directly.
8525 Perl_newSVpvf_nocontext(const char *const pat, ...)
8531 PERL_ARGS_ASSERT_NEWSVPVF_NOCONTEXT;
8533 va_start(args, pat);
8534 sv = vnewSVpvf(pat, &args);
8541 =for apidoc newSVpvf
8543 Creates a new SV and initializes it with the string formatted like
8550 Perl_newSVpvf(pTHX_ const char *const pat, ...)
8555 PERL_ARGS_ASSERT_NEWSVPVF;
8557 va_start(args, pat);
8558 sv = vnewSVpvf(pat, &args);
8563 /* backend for newSVpvf() and newSVpvf_nocontext() */
8566 Perl_vnewSVpvf(pTHX_ const char *const pat, va_list *const args)
8571 PERL_ARGS_ASSERT_VNEWSVPVF;
8574 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
8581 Creates a new SV and copies a floating point value into it.
8582 The reference count for the SV is set to 1.
8588 Perl_newSVnv(pTHX_ const NV n)
8601 Creates a new SV and copies an integer into it. The reference count for the
8608 Perl_newSViv(pTHX_ const IV i)
8621 Creates a new SV and copies an unsigned integer into it.
8622 The reference count for the SV is set to 1.
8628 Perl_newSVuv(pTHX_ const UV u)
8639 =for apidoc newSV_type
8641 Creates a new SV, of the type specified. The reference count for the new SV
8648 Perl_newSV_type(pTHX_ const svtype type)
8653 sv_upgrade(sv, type);
8658 =for apidoc newRV_noinc
8660 Creates an RV wrapper for an SV. The reference count for the original
8661 SV is B<not> incremented.
8667 Perl_newRV_noinc(pTHX_ SV *const tmpRef)
8670 register SV *sv = newSV_type(SVt_IV);
8672 PERL_ARGS_ASSERT_NEWRV_NOINC;
8675 SvRV_set(sv, tmpRef);
8680 /* newRV_inc is the official function name to use now.
8681 * newRV_inc is in fact #defined to newRV in sv.h
8685 Perl_newRV(pTHX_ SV *const sv)
8689 PERL_ARGS_ASSERT_NEWRV;
8691 return newRV_noinc(SvREFCNT_inc_simple_NN(sv));
8697 Creates a new SV which is an exact duplicate of the original SV.
8704 Perl_newSVsv(pTHX_ register SV *const old)
8711 if (SvTYPE(old) == (svtype)SVTYPEMASK) {
8712 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL), "semi-panic: attempt to dup freed string");
8716 /* SV_GMAGIC is the default for sv_setv()
8717 SV_NOSTEAL prevents TEMP buffers being, well, stolen, and saves games
8718 with SvTEMP_off and SvTEMP_on round a call to sv_setsv. */
8719 sv_setsv_flags(sv, old, SV_GMAGIC | SV_NOSTEAL);
8724 =for apidoc sv_reset
8726 Underlying implementation for the C<reset> Perl function.
8727 Note that the perl-level function is vaguely deprecated.
8733 Perl_sv_reset(pTHX_ register const char *s, HV *const stash)
8736 char todo[PERL_UCHAR_MAX+1];
8738 PERL_ARGS_ASSERT_SV_RESET;
8743 if (!*s) { /* reset ?? searches */
8744 MAGIC * const mg = mg_find((const SV *)stash, PERL_MAGIC_symtab);
8746 const U32 count = mg->mg_len / sizeof(PMOP**);
8747 PMOP **pmp = (PMOP**) mg->mg_ptr;
8748 PMOP *const *const end = pmp + count;
8752 SvREADONLY_off(PL_regex_pad[(*pmp)->op_pmoffset]);
8754 (*pmp)->op_pmflags &= ~PMf_USED;
8762 /* reset variables */
8764 if (!HvARRAY(stash))
8767 Zero(todo, 256, char);
8770 I32 i = (unsigned char)*s;
8774 max = (unsigned char)*s++;
8775 for ( ; i <= max; i++) {
8778 for (i = 0; i <= (I32) HvMAX(stash); i++) {
8780 for (entry = HvARRAY(stash)[i];
8782 entry = HeNEXT(entry))
8787 if (!todo[(U8)*HeKEY(entry)])
8789 gv = MUTABLE_GV(HeVAL(entry));
8792 if (SvTHINKFIRST(sv)) {
8793 if (!SvREADONLY(sv) && SvROK(sv))
8795 /* XXX Is this continue a bug? Why should THINKFIRST
8796 exempt us from resetting arrays and hashes? */
8800 if (SvTYPE(sv) >= SVt_PV) {
8802 if (SvPVX_const(sv) != NULL)
8810 if (GvHV(gv) && !HvNAME_get(GvHV(gv))) {
8812 Perl_die(aTHX_ "Can't reset %%ENV on this system");
8815 # if defined(USE_ENVIRON_ARRAY)
8818 # endif /* USE_ENVIRON_ARRAY */
8829 Using various gambits, try to get an IO from an SV: the IO slot if its a
8830 GV; or the recursive result if we're an RV; or the IO slot of the symbol
8831 named after the PV if we're a string.
8837 Perl_sv_2io(pTHX_ SV *const sv)
8842 PERL_ARGS_ASSERT_SV_2IO;
8844 switch (SvTYPE(sv)) {
8846 io = MUTABLE_IO(sv);
8850 if (isGV_with_GP(sv)) {
8851 gv = MUTABLE_GV(sv);
8854 Perl_croak(aTHX_ "Bad filehandle: %"HEKf,
8855 HEKfARG(GvNAME_HEK(gv)));
8861 Perl_croak(aTHX_ PL_no_usym, "filehandle");
8863 return sv_2io(SvRV(sv));
8864 gv = gv_fetchsv(sv, 0, SVt_PVIO);
8870 Perl_croak(aTHX_ "Bad filehandle: %"SVf, SVfARG(sv));
8879 Using various gambits, try to get a CV from an SV; in addition, try if
8880 possible to set C<*st> and C<*gvp> to the stash and GV associated with it.
8881 The flags in C<lref> are passed to gv_fetchsv.
8887 Perl_sv_2cv(pTHX_ SV *sv, HV **const st, GV **const gvp, const I32 lref)
8893 PERL_ARGS_ASSERT_SV_2CV;
8900 switch (SvTYPE(sv)) {
8904 return MUTABLE_CV(sv);
8914 sv = amagic_deref_call(sv, to_cv_amg);
8915 /* At this point I'd like to do SPAGAIN, but really I need to
8916 force it upon my callers. Hmmm. This is a mess... */
8919 if (SvTYPE(sv) == SVt_PVCV) {
8920 cv = MUTABLE_CV(sv);
8925 else if(isGV_with_GP(sv))
8926 gv = MUTABLE_GV(sv);
8928 Perl_croak(aTHX_ "Not a subroutine reference");
8930 else if (isGV_with_GP(sv)) {
8931 gv = MUTABLE_GV(sv);
8934 gv = gv_fetchsv_nomg(sv, lref, SVt_PVCV);
8941 /* Some flags to gv_fetchsv mean don't really create the GV */
8942 if (!isGV_with_GP(gv)) {
8947 if (lref & ~GV_ADDMG && !GvCVu(gv)) {
8951 gv_efullname3(tmpsv, gv, NULL);
8952 /* XXX this is probably not what they think they're getting.
8953 * It has the same effect as "sub name;", i.e. just a forward
8955 newSUB(start_subparse(FALSE, 0),
8956 newSVOP(OP_CONST, 0, tmpsv),
8960 Perl_croak(aTHX_ "Unable to create sub named \"%"SVf"\"",
8961 SVfARG(SvOK(sv) ? sv : &PL_sv_no));
8970 Returns true if the SV has a true value by Perl's rules.
8971 Use the C<SvTRUE> macro instead, which may call C<sv_true()> or may
8972 instead use an in-line version.
8978 Perl_sv_true(pTHX_ register SV *const sv)
8983 register const XPV* const tXpv = (XPV*)SvANY(sv);
8985 (tXpv->xpv_cur > 1 ||
8986 (tXpv->xpv_cur && *sv->sv_u.svu_pv != '0')))
8993 return SvIVX(sv) != 0;
8996 return SvNVX(sv) != 0.0;
8998 return sv_2bool(sv);
9004 =for apidoc sv_pvn_force
9006 Get a sensible string out of the SV somehow.
9007 A private implementation of the C<SvPV_force> macro for compilers which
9008 can't cope with complex macro expressions. Always use the macro instead.
9010 =for apidoc sv_pvn_force_flags
9012 Get a sensible string out of the SV somehow.
9013 If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<sv> if
9014 appropriate, else not. C<sv_pvn_force> and C<sv_pvn_force_nomg> are
9015 implemented in terms of this function.
9016 You normally want to use the various wrapper macros instead: see
9017 C<SvPV_force> and C<SvPV_force_nomg>
9023 Perl_sv_pvn_force_flags(pTHX_ SV *const sv, STRLEN *const lp, const I32 flags)
9027 PERL_ARGS_ASSERT_SV_PVN_FORCE_FLAGS;
9029 if (SvTHINKFIRST(sv) && !SvROK(sv))
9030 sv_force_normal_flags(sv, 0);
9040 if (SvREADONLY(sv) && !(flags & SV_MUTABLE_RETURN)) {
9041 const char * const ref = sv_reftype(sv,0);
9043 Perl_croak(aTHX_ "Can't coerce readonly %s to string in %s",
9044 ref, OP_DESC(PL_op));
9046 Perl_croak(aTHX_ "Can't coerce readonly %s to string", ref);
9048 if ((SvTYPE(sv) > SVt_PVLV && SvTYPE(sv) != SVt_PVFM)
9049 || isGV_with_GP(sv))
9050 /* diag_listed_as: Can't coerce %s to %s in %s */
9051 Perl_croak(aTHX_ "Can't coerce %s to string in %s", sv_reftype(sv,0),
9053 s = sv_2pv_flags(sv, &len, flags);
9057 if (s != SvPVX_const(sv)) { /* Almost, but not quite, sv_setpvn() */
9060 SvUPGRADE(sv, SVt_PV); /* Never FALSE */
9061 SvGROW(sv, len + 1);
9062 Move(s,SvPVX(sv),len,char);
9064 SvPVX(sv)[len] = '\0';
9067 SvPOK_on(sv); /* validate pointer */
9069 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
9070 PTR2UV(sv),SvPVX_const(sv)));
9073 return SvPVX_mutable(sv);
9077 =for apidoc sv_pvbyten_force
9079 The backend for the C<SvPVbytex_force> macro. Always use the macro instead.
9085 Perl_sv_pvbyten_force(pTHX_ SV *const sv, STRLEN *const lp)
9087 PERL_ARGS_ASSERT_SV_PVBYTEN_FORCE;
9089 sv_pvn_force(sv,lp);
9090 sv_utf8_downgrade(sv,0);
9096 =for apidoc sv_pvutf8n_force
9098 The backend for the C<SvPVutf8x_force> macro. Always use the macro instead.
9104 Perl_sv_pvutf8n_force(pTHX_ SV *const sv, STRLEN *const lp)
9106 PERL_ARGS_ASSERT_SV_PVUTF8N_FORCE;
9108 sv_pvn_force(sv,lp);
9109 sv_utf8_upgrade(sv);
9115 =for apidoc sv_reftype
9117 Returns a string describing what the SV is a reference to.
9123 Perl_sv_reftype(pTHX_ const SV *const sv, const int ob)
9125 PERL_ARGS_ASSERT_SV_REFTYPE;
9126 if (ob && SvOBJECT(sv)) {
9127 return SvPV_nolen_const(sv_ref(NULL, sv, ob));
9130 switch (SvTYPE(sv)) {
9145 case SVt_PVLV: return (char *) (SvROK(sv) ? "REF"
9146 /* tied lvalues should appear to be
9147 * scalars for backwards compatibility */
9148 : (LvTYPE(sv) == 't' || LvTYPE(sv) == 'T')
9149 ? "SCALAR" : "LVALUE");
9150 case SVt_PVAV: return "ARRAY";
9151 case SVt_PVHV: return "HASH";
9152 case SVt_PVCV: return "CODE";
9153 case SVt_PVGV: return (char *) (isGV_with_GP(sv)
9154 ? "GLOB" : "SCALAR");
9155 case SVt_PVFM: return "FORMAT";
9156 case SVt_PVIO: return "IO";
9157 case SVt_BIND: return "BIND";
9158 case SVt_REGEXP: return "REGEXP";
9159 default: return "UNKNOWN";
9167 Returns a SV describing what the SV passed in is a reference to.
9173 Perl_sv_ref(pTHX_ register SV *dst, const SV *const sv, const int ob)
9175 PERL_ARGS_ASSERT_SV_REF;
9178 dst = sv_newmortal();
9180 if (ob && SvOBJECT(sv)) {
9181 HvNAME_get(SvSTASH(sv))
9182 ? sv_sethek(dst, HvNAME_HEK(SvSTASH(sv)))
9183 : sv_setpvn(dst, "__ANON__", 8);
9186 const char * reftype = sv_reftype(sv, 0);
9187 sv_setpv(dst, reftype);
9193 =for apidoc sv_isobject
9195 Returns a boolean indicating whether the SV is an RV pointing to a blessed
9196 object. If the SV is not an RV, or if the object is not blessed, then this
9203 Perl_sv_isobject(pTHX_ SV *sv)
9219 Returns a boolean indicating whether the SV is blessed into the specified
9220 class. This does not check for subtypes; use C<sv_derived_from> to verify
9221 an inheritance relationship.
9227 Perl_sv_isa(pTHX_ SV *sv, const char *const name)
9231 PERL_ARGS_ASSERT_SV_ISA;
9241 hvname = HvNAME_get(SvSTASH(sv));
9245 return strEQ(hvname, name);
9251 Creates a new SV for the RV, C<rv>, to point to. If C<rv> is not an RV then
9252 it will be upgraded to one. If C<classname> is non-null then the new SV will
9253 be blessed in the specified package. The new SV is returned and its
9254 reference count is 1.
9260 Perl_newSVrv(pTHX_ SV *const rv, const char *const classname)
9265 PERL_ARGS_ASSERT_NEWSVRV;
9269 SV_CHECK_THINKFIRST_COW_DROP(rv);
9270 (void)SvAMAGIC_off(rv);
9272 if (SvTYPE(rv) >= SVt_PVMG) {
9273 const U32 refcnt = SvREFCNT(rv);
9277 SvREFCNT(rv) = refcnt;
9279 sv_upgrade(rv, SVt_IV);
9280 } else if (SvROK(rv)) {
9281 SvREFCNT_dec(SvRV(rv));
9283 prepare_SV_for_RV(rv);
9291 HV* const stash = gv_stashpv(classname, GV_ADD);
9292 (void)sv_bless(rv, stash);
9298 =for apidoc sv_setref_pv
9300 Copies a pointer into a new SV, optionally blessing the SV. The C<rv>
9301 argument will be upgraded to an RV. That RV will be modified to point to
9302 the new SV. If the C<pv> argument is NULL then C<PL_sv_undef> will be placed
9303 into the SV. The C<classname> argument indicates the package for the
9304 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9305 will have a reference count of 1, and the RV will be returned.
9307 Do not use with other Perl types such as HV, AV, SV, CV, because those
9308 objects will become corrupted by the pointer copy process.
9310 Note that C<sv_setref_pvn> copies the string while this copies the pointer.
9316 Perl_sv_setref_pv(pTHX_ SV *const rv, const char *const classname, void *const pv)
9320 PERL_ARGS_ASSERT_SV_SETREF_PV;
9323 sv_setsv(rv, &PL_sv_undef);
9327 sv_setiv(newSVrv(rv,classname), PTR2IV(pv));
9332 =for apidoc sv_setref_iv
9334 Copies an integer into a new SV, optionally blessing the SV. The C<rv>
9335 argument will be upgraded to an RV. That RV will be modified to point to
9336 the new SV. The C<classname> argument indicates the package for the
9337 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9338 will have a reference count of 1, and the RV will be returned.
9344 Perl_sv_setref_iv(pTHX_ SV *const rv, const char *const classname, const IV iv)
9346 PERL_ARGS_ASSERT_SV_SETREF_IV;
9348 sv_setiv(newSVrv(rv,classname), iv);
9353 =for apidoc sv_setref_uv
9355 Copies an unsigned integer into a new SV, optionally blessing the SV. The C<rv>
9356 argument will be upgraded to an RV. That RV will be modified to point to
9357 the new SV. The C<classname> argument indicates the package for the
9358 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9359 will have a reference count of 1, and the RV will be returned.
9365 Perl_sv_setref_uv(pTHX_ SV *const rv, const char *const classname, const UV uv)
9367 PERL_ARGS_ASSERT_SV_SETREF_UV;
9369 sv_setuv(newSVrv(rv,classname), uv);
9374 =for apidoc sv_setref_nv
9376 Copies a double into a new SV, optionally blessing the SV. The C<rv>
9377 argument will be upgraded to an RV. That RV will be modified to point to
9378 the new SV. The C<classname> argument indicates the package for the
9379 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9380 will have a reference count of 1, and the RV will be returned.
9386 Perl_sv_setref_nv(pTHX_ SV *const rv, const char *const classname, const NV nv)
9388 PERL_ARGS_ASSERT_SV_SETREF_NV;
9390 sv_setnv(newSVrv(rv,classname), nv);
9395 =for apidoc sv_setref_pvn
9397 Copies a string into a new SV, optionally blessing the SV. The length of the
9398 string must be specified with C<n>. The C<rv> argument will be upgraded to
9399 an RV. That RV will be modified to point to the new SV. The C<classname>
9400 argument indicates the package for the blessing. Set C<classname> to
9401 C<NULL> to avoid the blessing. The new SV will have a reference count
9402 of 1, and the RV will be returned.
9404 Note that C<sv_setref_pv> copies the pointer while this copies the string.
9410 Perl_sv_setref_pvn(pTHX_ SV *const rv, const char *const classname,
9411 const char *const pv, const STRLEN n)
9413 PERL_ARGS_ASSERT_SV_SETREF_PVN;
9415 sv_setpvn(newSVrv(rv,classname), pv, n);
9420 =for apidoc sv_bless
9422 Blesses an SV into a specified package. The SV must be an RV. The package
9423 must be designated by its stash (see C<gv_stashpv()>). The reference count
9424 of the SV is unaffected.
9430 Perl_sv_bless(pTHX_ SV *const sv, HV *const stash)
9435 PERL_ARGS_ASSERT_SV_BLESS;
9438 Perl_croak(aTHX_ "Can't bless non-reference value");
9440 if (SvFLAGS(tmpRef) & (SVs_OBJECT|SVf_READONLY)) {
9441 if (SvIsCOW(tmpRef))
9442 sv_force_normal_flags(tmpRef, 0);
9443 if (SvREADONLY(tmpRef))
9444 Perl_croak_no_modify(aTHX);
9445 if (SvOBJECT(tmpRef)) {
9446 if (SvTYPE(tmpRef) != SVt_PVIO)
9448 SvREFCNT_dec(SvSTASH(tmpRef));
9451 SvOBJECT_on(tmpRef);
9452 if (SvTYPE(tmpRef) != SVt_PVIO)
9454 SvUPGRADE(tmpRef, SVt_PVMG);
9455 SvSTASH_set(tmpRef, MUTABLE_HV(SvREFCNT_inc_simple(stash)));
9460 (void)SvAMAGIC_off(sv);
9462 if(SvSMAGICAL(tmpRef))
9463 if(mg_find(tmpRef, PERL_MAGIC_ext) || mg_find(tmpRef, PERL_MAGIC_uvar))
9471 /* Downgrades a PVGV to a PVMG. If it’s actually a PVLV, we leave the type
9472 * as it is after unglobbing it.
9476 S_sv_unglob(pTHX_ SV *const sv)
9481 SV * const temp = sv_newmortal();
9483 PERL_ARGS_ASSERT_SV_UNGLOB;
9485 assert(SvTYPE(sv) == SVt_PVGV || SvTYPE(sv) == SVt_PVLV);
9487 gv_efullname3(temp, MUTABLE_GV(sv), "*");
9490 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
9491 && HvNAME_get(stash))
9492 mro_method_changed_in(stash);
9493 gp_free(MUTABLE_GV(sv));
9496 sv_del_backref(MUTABLE_SV(GvSTASH(sv)), sv);
9500 if (GvNAME_HEK(sv)) {
9501 unshare_hek(GvNAME_HEK(sv));
9503 isGV_with_GP_off(sv);
9505 if(SvTYPE(sv) == SVt_PVGV) {
9506 /* need to keep SvANY(sv) in the right arena */
9507 xpvmg = new_XPVMG();
9508 StructCopy(SvANY(sv), xpvmg, XPVMG);
9509 del_XPVGV(SvANY(sv));
9512 SvFLAGS(sv) &= ~SVTYPEMASK;
9513 SvFLAGS(sv) |= SVt_PVMG;
9516 /* Intentionally not calling any local SET magic, as this isn't so much a
9517 set operation as merely an internal storage change. */
9518 sv_setsv_flags(sv, temp, 0);
9522 =for apidoc sv_unref_flags
9524 Unsets the RV status of the SV, and decrements the reference count of
9525 whatever was being referenced by the RV. This can almost be thought of
9526 as a reversal of C<newSVrv>. The C<cflags> argument can contain
9527 C<SV_IMMEDIATE_UNREF> to force the reference count to be decremented
9528 (otherwise the decrementing is conditional on the reference count being
9529 different from one or the reference being a readonly SV).
9536 Perl_sv_unref_flags(pTHX_ SV *const ref, const U32 flags)
9538 SV* const target = SvRV(ref);
9540 PERL_ARGS_ASSERT_SV_UNREF_FLAGS;
9542 if (SvWEAKREF(ref)) {
9543 sv_del_backref(target, ref);
9545 SvRV_set(ref, NULL);
9548 SvRV_set(ref, NULL);
9550 /* You can't have a || SvREADONLY(target) here, as $a = $$a, where $a was
9551 assigned to as BEGIN {$a = \"Foo"} will fail. */
9552 if (SvREFCNT(target) != 1 || (flags & SV_IMMEDIATE_UNREF))
9553 SvREFCNT_dec(target);
9554 else /* XXX Hack, but hard to make $a=$a->[1] work otherwise */
9555 sv_2mortal(target); /* Schedule for freeing later */
9559 =for apidoc sv_untaint
9561 Untaint an SV. Use C<SvTAINTED_off> instead.
9567 Perl_sv_untaint(pTHX_ SV *const sv)
9569 PERL_ARGS_ASSERT_SV_UNTAINT;
9571 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
9572 MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
9579 =for apidoc sv_tainted
9581 Test an SV for taintedness. Use C<SvTAINTED> instead.
9587 Perl_sv_tainted(pTHX_ SV *const sv)
9589 PERL_ARGS_ASSERT_SV_TAINTED;
9591 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
9592 const MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
9593 if (mg && (mg->mg_len & 1) )
9600 =for apidoc sv_setpviv
9602 Copies an integer into the given SV, also updating its string value.
9603 Does not handle 'set' magic. See C<sv_setpviv_mg>.
9609 Perl_sv_setpviv(pTHX_ SV *const sv, const IV iv)
9611 char buf[TYPE_CHARS(UV)];
9613 char * const ptr = uiv_2buf(buf, iv, 0, 0, &ebuf);
9615 PERL_ARGS_ASSERT_SV_SETPVIV;
9617 sv_setpvn(sv, ptr, ebuf - ptr);
9621 =for apidoc sv_setpviv_mg
9623 Like C<sv_setpviv>, but also handles 'set' magic.
9629 Perl_sv_setpviv_mg(pTHX_ SV *const sv, const IV iv)
9631 PERL_ARGS_ASSERT_SV_SETPVIV_MG;
9637 #if defined(PERL_IMPLICIT_CONTEXT)
9639 /* pTHX_ magic can't cope with varargs, so this is a no-context
9640 * version of the main function, (which may itself be aliased to us).
9641 * Don't access this version directly.
9645 Perl_sv_setpvf_nocontext(SV *const sv, const char *const pat, ...)
9650 PERL_ARGS_ASSERT_SV_SETPVF_NOCONTEXT;
9652 va_start(args, pat);
9653 sv_vsetpvf(sv, pat, &args);
9657 /* pTHX_ magic can't cope with varargs, so this is a no-context
9658 * version of the main function, (which may itself be aliased to us).
9659 * Don't access this version directly.
9663 Perl_sv_setpvf_mg_nocontext(SV *const sv, const char *const pat, ...)
9668 PERL_ARGS_ASSERT_SV_SETPVF_MG_NOCONTEXT;
9670 va_start(args, pat);
9671 sv_vsetpvf_mg(sv, pat, &args);
9677 =for apidoc sv_setpvf
9679 Works like C<sv_catpvf> but copies the text into the SV instead of
9680 appending it. Does not handle 'set' magic. See C<sv_setpvf_mg>.
9686 Perl_sv_setpvf(pTHX_ SV *const sv, const char *const pat, ...)
9690 PERL_ARGS_ASSERT_SV_SETPVF;
9692 va_start(args, pat);
9693 sv_vsetpvf(sv, pat, &args);
9698 =for apidoc sv_vsetpvf
9700 Works like C<sv_vcatpvf> but copies the text into the SV instead of
9701 appending it. Does not handle 'set' magic. See C<sv_vsetpvf_mg>.
9703 Usually used via its frontend C<sv_setpvf>.
9709 Perl_sv_vsetpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9711 PERL_ARGS_ASSERT_SV_VSETPVF;
9713 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9717 =for apidoc sv_setpvf_mg
9719 Like C<sv_setpvf>, but also handles 'set' magic.
9725 Perl_sv_setpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
9729 PERL_ARGS_ASSERT_SV_SETPVF_MG;
9731 va_start(args, pat);
9732 sv_vsetpvf_mg(sv, pat, &args);
9737 =for apidoc sv_vsetpvf_mg
9739 Like C<sv_vsetpvf>, but also handles 'set' magic.
9741 Usually used via its frontend C<sv_setpvf_mg>.
9747 Perl_sv_vsetpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9749 PERL_ARGS_ASSERT_SV_VSETPVF_MG;
9751 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9755 #if defined(PERL_IMPLICIT_CONTEXT)
9757 /* pTHX_ magic can't cope with varargs, so this is a no-context
9758 * version of the main function, (which may itself be aliased to us).
9759 * Don't access this version directly.
9763 Perl_sv_catpvf_nocontext(SV *const sv, const char *const pat, ...)
9768 PERL_ARGS_ASSERT_SV_CATPVF_NOCONTEXT;
9770 va_start(args, pat);
9771 sv_vcatpvf(sv, pat, &args);
9775 /* pTHX_ magic can't cope with varargs, so this is a no-context
9776 * version of the main function, (which may itself be aliased to us).
9777 * Don't access this version directly.
9781 Perl_sv_catpvf_mg_nocontext(SV *const sv, const char *const pat, ...)
9786 PERL_ARGS_ASSERT_SV_CATPVF_MG_NOCONTEXT;
9788 va_start(args, pat);
9789 sv_vcatpvf_mg(sv, pat, &args);
9795 =for apidoc sv_catpvf
9797 Processes its arguments like C<sprintf> and appends the formatted
9798 output to an SV. If the appended data contains "wide" characters
9799 (including, but not limited to, SVs with a UTF-8 PV formatted with %s,
9800 and characters >255 formatted with %c), the original SV might get
9801 upgraded to UTF-8. Handles 'get' magic, but not 'set' magic. See
9802 C<sv_catpvf_mg>. If the original SV was UTF-8, the pattern should be
9803 valid UTF-8; if the original SV was bytes, the pattern should be too.
9808 Perl_sv_catpvf(pTHX_ SV *const sv, const char *const pat, ...)
9812 PERL_ARGS_ASSERT_SV_CATPVF;
9814 va_start(args, pat);
9815 sv_vcatpvf(sv, pat, &args);
9820 =for apidoc sv_vcatpvf
9822 Processes its arguments like C<vsprintf> and appends the formatted output
9823 to an SV. Does not handle 'set' magic. See C<sv_vcatpvf_mg>.
9825 Usually used via its frontend C<sv_catpvf>.
9831 Perl_sv_vcatpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9833 PERL_ARGS_ASSERT_SV_VCATPVF;
9835 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9839 =for apidoc sv_catpvf_mg
9841 Like C<sv_catpvf>, but also handles 'set' magic.
9847 Perl_sv_catpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
9851 PERL_ARGS_ASSERT_SV_CATPVF_MG;
9853 va_start(args, pat);
9854 sv_vcatpvf_mg(sv, pat, &args);
9859 =for apidoc sv_vcatpvf_mg
9861 Like C<sv_vcatpvf>, but also handles 'set' magic.
9863 Usually used via its frontend C<sv_catpvf_mg>.
9869 Perl_sv_vcatpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9871 PERL_ARGS_ASSERT_SV_VCATPVF_MG;
9873 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9878 =for apidoc sv_vsetpvfn
9880 Works like C<sv_vcatpvfn> but copies the text into the SV instead of
9883 Usually used via one of its frontends C<sv_vsetpvf> and C<sv_vsetpvf_mg>.
9889 Perl_sv_vsetpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
9890 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
9892 PERL_ARGS_ASSERT_SV_VSETPVFN;
9895 sv_vcatpvfn(sv, pat, patlen, args, svargs, svmax, maybe_tainted);
9900 * Warn of missing argument to sprintf, and then return a defined value
9901 * to avoid inappropriate "use of uninit" warnings [perl #71000].
9903 #define WARN_MISSING WARN_UNINITIALIZED /* Not sure we want a new category */
9905 S_vcatpvfn_missing_argument(pTHX) {
9906 if (ckWARN(WARN_MISSING)) {
9907 Perl_warner(aTHX_ packWARN(WARN_MISSING), "Missing argument in %s",
9908 PL_op ? OP_DESC(PL_op) : "sv_vcatpvfn()");
9915 S_expect_number(pTHX_ char **const pattern)
9920 PERL_ARGS_ASSERT_EXPECT_NUMBER;
9922 switch (**pattern) {
9923 case '1': case '2': case '3':
9924 case '4': case '5': case '6':
9925 case '7': case '8': case '9':
9926 var = *(*pattern)++ - '0';
9927 while (isDIGIT(**pattern)) {
9928 const I32 tmp = var * 10 + (*(*pattern)++ - '0');
9930 Perl_croak(aTHX_ "Integer overflow in format string for %s", (PL_op ? OP_DESC(PL_op) : "sv_vcatpvfn"));
9938 S_F0convert(NV nv, char *const endbuf, STRLEN *const len)
9940 const int neg = nv < 0;
9943 PERL_ARGS_ASSERT_F0CONVERT;
9951 if (uv & 1 && uv == nv)
9952 uv--; /* Round to even */
9954 const unsigned dig = uv % 10;
9967 =for apidoc sv_vcatpvfn
9969 Processes its arguments like C<vsprintf> and appends the formatted output
9970 to an SV. Uses an array of SVs if the C style variable argument list is
9971 missing (NULL). When running with taint checks enabled, indicates via
9972 C<maybe_tainted> if results are untrustworthy (often due to the use of
9975 Usually used via one of its frontends C<sv_vcatpvf> and C<sv_vcatpvf_mg>.
9981 #define VECTORIZE_ARGS vecsv = va_arg(*args, SV*);\
9982 vecstr = (U8*)SvPV_const(vecsv,veclen);\
9983 vec_utf8 = DO_UTF8(vecsv);
9985 /* XXX maybe_tainted is never assigned to, so the doc above is lying. */
9988 Perl_sv_vcatpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
9989 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
9997 static const char nullstr[] = "(null)";
9999 bool has_utf8 = DO_UTF8(sv); /* has the result utf8? */
10000 const bool pat_utf8 = has_utf8; /* the pattern is in utf8? */
10002 /* Times 4: a decimal digit takes more than 3 binary digits.
10003 * NV_DIG: mantissa takes than many decimal digits.
10004 * Plus 32: Playing safe. */
10005 char ebuf[IV_DIG * 4 + NV_DIG + 32];
10006 /* large enough for "%#.#f" --chip */
10007 /* what about long double NVs? --jhi */
10009 PERL_ARGS_ASSERT_SV_VCATPVFN;
10010 PERL_UNUSED_ARG(maybe_tainted);
10012 /* no matter what, this is a string now */
10013 (void)SvPV_force(sv, origlen);
10015 /* special-case "", "%s", and "%-p" (SVf - see below) */
10018 if (patlen == 2 && pat[0] == '%' && pat[1] == 's') {
10020 const char * const s = va_arg(*args, char*);
10021 sv_catpv(sv, s ? s : nullstr);
10023 else if (svix < svmax) {
10024 sv_catsv(sv, *svargs);
10027 S_vcatpvfn_missing_argument(aTHX);
10030 if (args && patlen == 3 && pat[0] == '%' &&
10031 pat[1] == '-' && pat[2] == 'p') {
10032 argsv = MUTABLE_SV(va_arg(*args, void*));
10033 sv_catsv(sv, argsv);
10037 #ifndef USE_LONG_DOUBLE
10038 /* special-case "%.<number>[gf]" */
10039 if ( !args && patlen <= 5 && pat[0] == '%' && pat[1] == '.'
10040 && (pat[patlen-1] == 'g' || pat[patlen-1] == 'f') ) {
10041 unsigned digits = 0;
10045 while (*pp >= '0' && *pp <= '9')
10046 digits = 10 * digits + (*pp++ - '0');
10047 if (pp - pat == (int)patlen - 1 && svix < svmax) {
10048 const NV nv = SvNV(*svargs);
10050 /* Add check for digits != 0 because it seems that some
10051 gconverts are buggy in this case, and we don't yet have
10052 a Configure test for this. */
10053 if (digits && digits < sizeof(ebuf) - NV_DIG - 10) {
10054 /* 0, point, slack */
10055 Gconvert(nv, (int)digits, 0, ebuf);
10056 sv_catpv(sv, ebuf);
10057 if (*ebuf) /* May return an empty string for digits==0 */
10060 } else if (!digits) {
10063 if ((p = F0convert(nv, ebuf + sizeof ebuf, &l))) {
10064 sv_catpvn(sv, p, l);
10070 #endif /* !USE_LONG_DOUBLE */
10072 if (!args && svix < svmax && DO_UTF8(*svargs))
10075 patend = (char*)pat + patlen;
10076 for (p = (char*)pat; p < patend; p = q) {
10079 bool vectorize = FALSE;
10080 bool vectorarg = FALSE;
10081 bool vec_utf8 = FALSE;
10087 bool has_precis = FALSE;
10089 const I32 osvix = svix;
10090 bool is_utf8 = FALSE; /* is this item utf8? */
10091 #ifdef HAS_LDBL_SPRINTF_BUG
10092 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
10093 with sfio - Allen <allens@cpan.org> */
10094 bool fix_ldbl_sprintf_bug = FALSE;
10098 U8 utf8buf[UTF8_MAXBYTES+1];
10099 STRLEN esignlen = 0;
10101 const char *eptr = NULL;
10102 const char *fmtstart;
10105 const U8 *vecstr = NULL;
10112 /* we need a long double target in case HAS_LONG_DOUBLE but
10113 not USE_LONG_DOUBLE
10115 #if defined(HAS_LONG_DOUBLE) && LONG_DOUBLESIZE > DOUBLESIZE
10123 const char *dotstr = ".";
10124 STRLEN dotstrlen = 1;
10125 I32 efix = 0; /* explicit format parameter index */
10126 I32 ewix = 0; /* explicit width index */
10127 I32 epix = 0; /* explicit precision index */
10128 I32 evix = 0; /* explicit vector index */
10129 bool asterisk = FALSE;
10131 /* echo everything up to the next format specification */
10132 for (q = p; q < patend && *q != '%'; ++q) ;
10134 if (has_utf8 && !pat_utf8)
10135 sv_catpvn_utf8_upgrade(sv, p, q - p, nsv);
10137 sv_catpvn(sv, p, q - p);
10146 We allow format specification elements in this order:
10147 \d+\$ explicit format parameter index
10149 v|\*(\d+\$)?v vector with optional (optionally specified) arg
10150 0 flag (as above): repeated to allow "v02"
10151 \d+|\*(\d+\$)? width using optional (optionally specified) arg
10152 \.(\d*|\*(\d+\$)?) precision using optional (optionally specified) arg
10154 [%bcdefginopsuxDFOUX] format (mandatory)
10159 As of perl5.9.3, printf format checking is on by default.
10160 Internally, perl uses %p formats to provide an escape to
10161 some extended formatting. This block deals with those
10162 extensions: if it does not match, (char*)q is reset and
10163 the normal format processing code is used.
10165 Currently defined extensions are:
10166 %p include pointer address (standard)
10167 %-p (SVf) include an SV (previously %_)
10168 %-<num>p include an SV with precision <num>
10170 %3p include a HEK with precision of 256
10171 %<num>p (where num != 2 or 3) reserved for future
10174 Robin Barker 2005-07-14 (but modified since)
10176 %1p (VDf) removed. RMB 2007-10-19
10183 n = expect_number(&q);
10185 if (sv) { /* SVf */
10190 argsv = MUTABLE_SV(va_arg(*args, void*));
10191 eptr = SvPV_const(argsv, elen);
10192 if (DO_UTF8(argsv))
10196 else if (n==2 || n==3) { /* HEKf */
10197 HEK * const hek = va_arg(*args, HEK *);
10198 eptr = HEK_KEY(hek);
10199 elen = HEK_LEN(hek);
10200 if (HEK_UTF8(hek)) is_utf8 = TRUE;
10201 if (n==3) precis = 256, has_precis = TRUE;
10205 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL),
10206 "internal %%<num>p might conflict with future printf extensions");
10212 if ( (width = expect_number(&q)) ) {
10227 if (plus == '+' && *q == ' ') /* '+' over ' ' */
10256 if ( (ewix = expect_number(&q)) )
10265 if ((vectorarg = asterisk)) {
10278 width = expect_number(&q);
10281 if (vectorize && vectorarg) {
10282 /* vectorizing, but not with the default "." */
10284 vecsv = va_arg(*args, SV*);
10286 vecsv = (evix > 0 && evix <= svmax)
10287 ? svargs[evix-1] : S_vcatpvfn_missing_argument(aTHX);
10289 vecsv = svix < svmax
10290 ? svargs[svix++] : S_vcatpvfn_missing_argument(aTHX);
10292 dotstr = SvPV_const(vecsv, dotstrlen);
10293 /* Keep the DO_UTF8 test *after* the SvPV call, else things go
10294 bad with tied or overloaded values that return UTF8. */
10295 if (DO_UTF8(vecsv))
10297 else if (has_utf8) {
10298 vecsv = sv_mortalcopy(vecsv);
10299 sv_utf8_upgrade(vecsv);
10300 dotstr = SvPV_const(vecsv, dotstrlen);
10307 i = va_arg(*args, int);
10309 i = (ewix ? ewix <= svmax : svix < svmax) ?
10310 SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
10312 width = (i < 0) ? -i : i;
10322 if ( ((epix = expect_number(&q))) && (*q++ != '$') )
10324 /* XXX: todo, support specified precision parameter */
10328 i = va_arg(*args, int);
10330 i = (ewix ? ewix <= svmax : svix < svmax)
10331 ? SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
10333 has_precis = !(i < 0);
10337 while (isDIGIT(*q))
10338 precis = precis * 10 + (*q++ - '0');
10347 else if (efix ? (efix > 0 && efix <= svmax) : svix < svmax) {
10348 vecsv = svargs[efix ? efix-1 : svix++];
10349 vecstr = (U8*)SvPV_const(vecsv,veclen);
10350 vec_utf8 = DO_UTF8(vecsv);
10352 /* if this is a version object, we need to convert
10353 * back into v-string notation and then let the
10354 * vectorize happen normally
10356 if (sv_derived_from(vecsv, "version")) {
10357 char *version = savesvpv(vecsv);
10358 if ( hv_exists(MUTABLE_HV(SvRV(vecsv)), "alpha", 5 ) ) {
10359 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
10360 "vector argument not supported with alpha versions");
10363 vecsv = sv_newmortal();
10364 scan_vstring(version, version + veclen, vecsv);
10365 vecstr = (U8*)SvPV_const(vecsv, veclen);
10366 vec_utf8 = DO_UTF8(vecsv);
10380 case 'I': /* Ix, I32x, and I64x */
10382 if (q[1] == '6' && q[2] == '4') {
10388 if (q[1] == '3' && q[2] == '2') {
10398 #if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
10410 #if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
10411 if (*q == 'l') { /* lld, llf */
10420 if (*++q == 'h') { /* hhd, hhu */
10449 if (!vectorize && !args) {
10451 const I32 i = efix-1;
10452 argsv = (i >= 0 && i < svmax)
10453 ? svargs[i] : S_vcatpvfn_missing_argument(aTHX);
10455 argsv = (svix >= 0 && svix < svmax)
10456 ? svargs[svix++] : S_vcatpvfn_missing_argument(aTHX);
10460 switch (c = *q++) {
10467 uv = (args) ? va_arg(*args, int) : SvIV(argsv);
10469 (!UNI_IS_INVARIANT(uv) && SvUTF8(sv)))
10471 eptr = (char*)utf8buf;
10472 elen = uvchr_to_utf8((U8*)eptr, uv) - utf8buf;
10486 eptr = va_arg(*args, char*);
10488 elen = strlen(eptr);
10490 eptr = (char *)nullstr;
10491 elen = sizeof nullstr - 1;
10495 eptr = SvPV_const(argsv, elen);
10496 if (DO_UTF8(argsv)) {
10497 STRLEN old_precis = precis;
10498 if (has_precis && precis < elen) {
10499 STRLEN ulen = sv_len_utf8(argsv);
10500 I32 p = precis > ulen ? ulen : precis;
10501 sv_pos_u2b(argsv, &p, 0); /* sticks at end */
10504 if (width) { /* fudge width (can't fudge elen) */
10505 if (has_precis && precis < elen)
10506 width += precis - old_precis;
10508 width += elen - sv_len_utf8(argsv);
10515 if (has_precis && precis < elen)
10522 if (alt || vectorize)
10524 uv = PTR2UV(args ? va_arg(*args, void*) : argsv);
10545 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
10554 esignbuf[esignlen++] = plus;
10558 case 'c': iv = (char)va_arg(*args, int); break;
10559 case 'h': iv = (short)va_arg(*args, int); break;
10560 case 'l': iv = va_arg(*args, long); break;
10561 case 'V': iv = va_arg(*args, IV); break;
10562 case 'z': iv = va_arg(*args, SSize_t); break;
10563 case 't': iv = va_arg(*args, ptrdiff_t); break;
10564 default: iv = va_arg(*args, int); break;
10566 case 'j': iv = va_arg(*args, intmax_t); break;
10570 iv = va_arg(*args, Quad_t); break;
10577 IV tiv = SvIV(argsv); /* work around GCC bug #13488 */
10579 case 'c': iv = (char)tiv; break;
10580 case 'h': iv = (short)tiv; break;
10581 case 'l': iv = (long)tiv; break;
10583 default: iv = tiv; break;
10586 iv = (Quad_t)tiv; break;
10592 if ( !vectorize ) /* we already set uv above */
10597 esignbuf[esignlen++] = plus;
10601 esignbuf[esignlen++] = '-';
10645 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
10656 case 'c': uv = (unsigned char)va_arg(*args, unsigned); break;
10657 case 'h': uv = (unsigned short)va_arg(*args, unsigned); break;
10658 case 'l': uv = va_arg(*args, unsigned long); break;
10659 case 'V': uv = va_arg(*args, UV); break;
10660 case 'z': uv = va_arg(*args, Size_t); break;
10661 case 't': uv = va_arg(*args, ptrdiff_t); break; /* will sign extend, but there is no uptrdiff_t, so oh well */
10663 case 'j': uv = va_arg(*args, uintmax_t); break;
10665 default: uv = va_arg(*args, unsigned); break;
10668 uv = va_arg(*args, Uquad_t); break;
10675 UV tuv = SvUV(argsv); /* work around GCC bug #13488 */
10677 case 'c': uv = (unsigned char)tuv; break;
10678 case 'h': uv = (unsigned short)tuv; break;
10679 case 'l': uv = (unsigned long)tuv; break;
10681 default: uv = tuv; break;
10684 uv = (Uquad_t)tuv; break;
10693 char *ptr = ebuf + sizeof ebuf;
10694 bool tempalt = uv ? alt : FALSE; /* Vectors can't change alt */
10700 p = (char *)((c == 'X') ? PL_hexdigit + 16 : PL_hexdigit);
10704 } while (uv >>= 4);
10706 esignbuf[esignlen++] = '0';
10707 esignbuf[esignlen++] = c; /* 'x' or 'X' */
10713 *--ptr = '0' + dig;
10714 } while (uv >>= 3);
10715 if (alt && *ptr != '0')
10721 *--ptr = '0' + dig;
10722 } while (uv >>= 1);
10724 esignbuf[esignlen++] = '0';
10725 esignbuf[esignlen++] = c;
10728 default: /* it had better be ten or less */
10731 *--ptr = '0' + dig;
10732 } while (uv /= base);
10735 elen = (ebuf + sizeof ebuf) - ptr;
10739 zeros = precis - elen;
10740 else if (precis == 0 && elen == 1 && *eptr == '0'
10741 && !(base == 8 && alt)) /* "%#.0o" prints "0" */
10744 /* a precision nullifies the 0 flag. */
10751 /* FLOATING POINT */
10754 c = 'f'; /* maybe %F isn't supported here */
10756 case 'e': case 'E':
10758 case 'g': case 'G':
10762 /* This is evil, but floating point is even more evil */
10764 /* for SV-style calling, we can only get NV
10765 for C-style calling, we assume %f is double;
10766 for simplicity we allow any of %Lf, %llf, %qf for long double
10770 #if defined(USE_LONG_DOUBLE)
10774 /* [perl #20339] - we should accept and ignore %lf rather than die */
10778 #if defined(USE_LONG_DOUBLE)
10779 intsize = args ? 0 : 'q';
10783 #if defined(HAS_LONG_DOUBLE)
10796 /* now we need (long double) if intsize == 'q', else (double) */
10798 #if LONG_DOUBLESIZE > DOUBLESIZE
10800 va_arg(*args, long double) :
10801 va_arg(*args, double)
10803 va_arg(*args, double)
10808 /* nv * 0 will be NaN for NaN, +Inf and -Inf, and 0 for anything
10809 else. frexp() has some unspecified behaviour for those three */
10810 if (c != 'e' && c != 'E' && (nv * 0) == 0) {
10812 /* FIXME: if HAS_LONG_DOUBLE but not USE_LONG_DOUBLE this
10813 will cast our (long double) to (double) */
10814 (void)Perl_frexp(nv, &i);
10815 if (i == PERL_INT_MIN)
10816 Perl_die(aTHX_ "panic: frexp");
10818 need = BIT_DIGITS(i);
10820 need += has_precis ? precis : 6; /* known default */
10825 #ifdef HAS_LDBL_SPRINTF_BUG
10826 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
10827 with sfio - Allen <allens@cpan.org> */
10830 # define MY_DBL_MAX DBL_MAX
10831 # else /* XXX guessing! HUGE_VAL may be defined as infinity, so not using */
10832 # if DOUBLESIZE >= 8
10833 # define MY_DBL_MAX 1.7976931348623157E+308L
10835 # define MY_DBL_MAX 3.40282347E+38L
10839 # ifdef HAS_LDBL_SPRINTF_BUG_LESS1 /* only between -1L & 1L - Allen */
10840 # define MY_DBL_MAX_BUG 1L
10842 # define MY_DBL_MAX_BUG MY_DBL_MAX
10846 # define MY_DBL_MIN DBL_MIN
10847 # else /* XXX guessing! -Allen */
10848 # if DOUBLESIZE >= 8
10849 # define MY_DBL_MIN 2.2250738585072014E-308L
10851 # define MY_DBL_MIN 1.17549435E-38L
10855 if ((intsize == 'q') && (c == 'f') &&
10856 ((nv < MY_DBL_MAX_BUG) && (nv > -MY_DBL_MAX_BUG)) &&
10857 (need < DBL_DIG)) {
10858 /* it's going to be short enough that
10859 * long double precision is not needed */
10861 if ((nv <= 0L) && (nv >= -0L))
10862 fix_ldbl_sprintf_bug = TRUE; /* 0 is 0 - easiest */
10864 /* would use Perl_fp_class as a double-check but not
10865 * functional on IRIX - see perl.h comments */
10867 if ((nv >= MY_DBL_MIN) || (nv <= -MY_DBL_MIN)) {
10868 /* It's within the range that a double can represent */
10869 #if defined(DBL_MAX) && !defined(DBL_MIN)
10870 if ((nv >= ((long double)1/DBL_MAX)) ||
10871 (nv <= (-(long double)1/DBL_MAX)))
10873 fix_ldbl_sprintf_bug = TRUE;
10876 if (fix_ldbl_sprintf_bug == TRUE) {
10886 # undef MY_DBL_MAX_BUG
10889 #endif /* HAS_LDBL_SPRINTF_BUG */
10891 need += 20; /* fudge factor */
10892 if (PL_efloatsize < need) {
10893 Safefree(PL_efloatbuf);
10894 PL_efloatsize = need + 20; /* more fudge */
10895 Newx(PL_efloatbuf, PL_efloatsize, char);
10896 PL_efloatbuf[0] = '\0';
10899 if ( !(width || left || plus || alt) && fill != '0'
10900 && has_precis && intsize != 'q' ) { /* Shortcuts */
10901 /* See earlier comment about buggy Gconvert when digits,
10903 if ( c == 'g' && precis) {
10904 Gconvert((NV)nv, (int)precis, 0, PL_efloatbuf);
10905 /* May return an empty string for digits==0 */
10906 if (*PL_efloatbuf) {
10907 elen = strlen(PL_efloatbuf);
10908 goto float_converted;
10910 } else if ( c == 'f' && !precis) {
10911 if ((eptr = F0convert(nv, ebuf + sizeof ebuf, &elen)))
10916 char *ptr = ebuf + sizeof ebuf;
10919 /* FIXME: what to do if HAS_LONG_DOUBLE but not PERL_PRIfldbl? */
10920 #if defined(HAS_LONG_DOUBLE) && defined(PERL_PRIfldbl)
10921 if (intsize == 'q') {
10922 /* Copy the one or more characters in a long double
10923 * format before the 'base' ([efgEFG]) character to
10924 * the format string. */
10925 static char const prifldbl[] = PERL_PRIfldbl;
10926 char const *p = prifldbl + sizeof(prifldbl) - 3;
10927 while (p >= prifldbl) { *--ptr = *p--; }
10932 do { *--ptr = '0' + (base % 10); } while (base /= 10);
10937 do { *--ptr = '0' + (base % 10); } while (base /= 10);
10949 /* No taint. Otherwise we are in the strange situation
10950 * where printf() taints but print($float) doesn't.
10952 #if defined(HAS_LONG_DOUBLE)
10953 elen = ((intsize == 'q')
10954 ? my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, nv)
10955 : my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, (double)nv));
10957 elen = my_sprintf(PL_efloatbuf, ptr, nv);
10961 eptr = PL_efloatbuf;
10969 i = SvCUR(sv) - origlen;
10972 case 'c': *(va_arg(*args, char*)) = i; break;
10973 case 'h': *(va_arg(*args, short*)) = i; break;
10974 default: *(va_arg(*args, int*)) = i; break;
10975 case 'l': *(va_arg(*args, long*)) = i; break;
10976 case 'V': *(va_arg(*args, IV*)) = i; break;
10977 case 'z': *(va_arg(*args, SSize_t*)) = i; break;
10978 case 't': *(va_arg(*args, ptrdiff_t*)) = i; break;
10980 case 'j': *(va_arg(*args, intmax_t*)) = i; break;
10984 *(va_arg(*args, Quad_t*)) = i; break;
10991 sv_setuv_mg(argsv, (UV)i);
10992 continue; /* not "break" */
10999 && (PL_op->op_type == OP_PRTF || PL_op->op_type == OP_SPRINTF)
11000 && ckWARN(WARN_PRINTF))
11002 SV * const msg = sv_newmortal();
11003 Perl_sv_setpvf(aTHX_ msg, "Invalid conversion in %sprintf: ",
11004 (PL_op->op_type == OP_PRTF) ? "" : "s");
11005 if (fmtstart < patend) {
11006 const char * const fmtend = q < patend ? q : patend;
11008 sv_catpvs(msg, "\"%");
11009 for (f = fmtstart; f < fmtend; f++) {
11011 sv_catpvn(msg, f, 1);
11013 Perl_sv_catpvf(aTHX_ msg,
11014 "\\%03"UVof, (UV)*f & 0xFF);
11017 sv_catpvs(msg, "\"");
11019 sv_catpvs(msg, "end of string");
11021 Perl_warner(aTHX_ packWARN(WARN_PRINTF), "%"SVf, SVfARG(msg)); /* yes, this is reentrant */
11024 /* output mangled stuff ... */
11030 /* ... right here, because formatting flags should not apply */
11031 SvGROW(sv, SvCUR(sv) + elen + 1);
11033 Copy(eptr, p, elen, char);
11036 SvCUR_set(sv, p - SvPVX_const(sv));
11038 continue; /* not "break" */
11041 if (is_utf8 != has_utf8) {
11044 sv_utf8_upgrade(sv);
11047 const STRLEN old_elen = elen;
11048 SV * const nsv = newSVpvn_flags(eptr, elen, SVs_TEMP);
11049 sv_utf8_upgrade(nsv);
11050 eptr = SvPVX_const(nsv);
11053 if (width) { /* fudge width (can't fudge elen) */
11054 width += elen - old_elen;
11060 have = esignlen + zeros + elen;
11062 Perl_croak_nocontext("%s", PL_memory_wrap);
11064 need = (have > width ? have : width);
11067 if (need >= (((STRLEN)~0) - SvCUR(sv) - dotstrlen - 1))
11068 Perl_croak_nocontext("%s", PL_memory_wrap);
11069 SvGROW(sv, SvCUR(sv) + need + dotstrlen + 1);
11071 if (esignlen && fill == '0') {
11073 for (i = 0; i < (int)esignlen; i++)
11074 *p++ = esignbuf[i];
11076 if (gap && !left) {
11077 memset(p, fill, gap);
11080 if (esignlen && fill != '0') {
11082 for (i = 0; i < (int)esignlen; i++)
11083 *p++ = esignbuf[i];
11087 for (i = zeros; i; i--)
11091 Copy(eptr, p, elen, char);
11095 memset(p, ' ', gap);
11100 Copy(dotstr, p, dotstrlen, char);
11104 vectorize = FALSE; /* done iterating over vecstr */
11111 SvCUR_set(sv, p - SvPVX_const(sv));
11120 /* =========================================================================
11122 =head1 Cloning an interpreter
11124 All the macros and functions in this section are for the private use of
11125 the main function, perl_clone().
11127 The foo_dup() functions make an exact copy of an existing foo thingy.
11128 During the course of a cloning, a hash table is used to map old addresses
11129 to new addresses. The table is created and manipulated with the
11130 ptr_table_* functions.
11134 * =========================================================================*/
11137 #if defined(USE_ITHREADS)
11139 /* XXX Remove this so it doesn't have to go thru the macro and return for nothing */
11140 #ifndef GpREFCNT_inc
11141 # define GpREFCNT_inc(gp) ((gp) ? (++(gp)->gp_refcnt, (gp)) : (GP*)NULL)
11145 /* Certain cases in Perl_ss_dup have been merged, by relying on the fact
11146 that currently av_dup, gv_dup and hv_dup are the same as sv_dup.
11147 If this changes, please unmerge ss_dup.
11148 Likewise, sv_dup_inc_multiple() relies on this fact. */
11149 #define sv_dup_inc_NN(s,t) SvREFCNT_inc_NN(sv_dup_inc(s,t))
11150 #define av_dup(s,t) MUTABLE_AV(sv_dup((const SV *)s,t))
11151 #define av_dup_inc(s,t) MUTABLE_AV(sv_dup_inc((const SV *)s,t))
11152 #define hv_dup(s,t) MUTABLE_HV(sv_dup((const SV *)s,t))
11153 #define hv_dup_inc(s,t) MUTABLE_HV(sv_dup_inc((const SV *)s,t))
11154 #define cv_dup(s,t) MUTABLE_CV(sv_dup((const SV *)s,t))
11155 #define cv_dup_inc(s,t) MUTABLE_CV(sv_dup_inc((const SV *)s,t))
11156 #define io_dup(s,t) MUTABLE_IO(sv_dup((const SV *)s,t))
11157 #define io_dup_inc(s,t) MUTABLE_IO(sv_dup_inc((const SV *)s,t))
11158 #define gv_dup(s,t) MUTABLE_GV(sv_dup((const SV *)s,t))
11159 #define gv_dup_inc(s,t) MUTABLE_GV(sv_dup_inc((const SV *)s,t))
11160 #define SAVEPV(p) ((p) ? savepv(p) : NULL)
11161 #define SAVEPVN(p,n) ((p) ? savepvn(p,n) : NULL)
11163 /* clone a parser */
11166 Perl_parser_dup(pTHX_ const yy_parser *const proto, CLONE_PARAMS *const param)
11170 PERL_ARGS_ASSERT_PARSER_DUP;
11175 /* look for it in the table first */
11176 parser = (yy_parser *)ptr_table_fetch(PL_ptr_table, proto);
11180 /* create anew and remember what it is */
11181 Newxz(parser, 1, yy_parser);
11182 ptr_table_store(PL_ptr_table, proto, parser);
11184 /* XXX these not yet duped */
11185 parser->old_parser = NULL;
11186 parser->stack = NULL;
11188 parser->stack_size = 0;
11189 /* XXX parser->stack->state = 0; */
11191 /* XXX eventually, just Copy() most of the parser struct ? */
11193 parser->lex_brackets = proto->lex_brackets;
11194 parser->lex_casemods = proto->lex_casemods;
11195 parser->lex_brackstack = savepvn(proto->lex_brackstack,
11196 (proto->lex_brackets < 120 ? 120 : proto->lex_brackets));
11197 parser->lex_casestack = savepvn(proto->lex_casestack,
11198 (proto->lex_casemods < 12 ? 12 : proto->lex_casemods));
11199 parser->lex_defer = proto->lex_defer;
11200 parser->lex_dojoin = proto->lex_dojoin;
11201 parser->lex_expect = proto->lex_expect;
11202 parser->lex_formbrack = proto->lex_formbrack;
11203 parser->lex_inpat = proto->lex_inpat;
11204 parser->lex_inwhat = proto->lex_inwhat;
11205 parser->lex_op = proto->lex_op;
11206 parser->lex_repl = sv_dup_inc(proto->lex_repl, param);
11207 parser->lex_starts = proto->lex_starts;
11208 parser->lex_stuff = sv_dup_inc(proto->lex_stuff, param);
11209 parser->multi_close = proto->multi_close;
11210 parser->multi_open = proto->multi_open;
11211 parser->multi_start = proto->multi_start;
11212 parser->multi_end = proto->multi_end;
11213 parser->pending_ident = proto->pending_ident;
11214 parser->preambled = proto->preambled;
11215 parser->sublex_info = proto->sublex_info; /* XXX not quite right */
11216 parser->linestr = sv_dup_inc(proto->linestr, param);
11217 parser->expect = proto->expect;
11218 parser->copline = proto->copline;
11219 parser->last_lop_op = proto->last_lop_op;
11220 parser->lex_state = proto->lex_state;
11221 parser->rsfp = fp_dup(proto->rsfp, '<', param);
11222 /* rsfp_filters entries have fake IoDIRP() */
11223 parser->rsfp_filters= av_dup_inc(proto->rsfp_filters, param);
11224 parser->in_my = proto->in_my;
11225 parser->in_my_stash = hv_dup(proto->in_my_stash, param);
11226 parser->error_count = proto->error_count;
11229 parser->linestr = sv_dup_inc(proto->linestr, param);
11232 char * const ols = SvPVX(proto->linestr);
11233 char * const ls = SvPVX(parser->linestr);
11235 parser->bufptr = ls + (proto->bufptr >= ols ?
11236 proto->bufptr - ols : 0);
11237 parser->oldbufptr = ls + (proto->oldbufptr >= ols ?
11238 proto->oldbufptr - ols : 0);
11239 parser->oldoldbufptr= ls + (proto->oldoldbufptr >= ols ?
11240 proto->oldoldbufptr - ols : 0);
11241 parser->linestart = ls + (proto->linestart >= ols ?
11242 proto->linestart - ols : 0);
11243 parser->last_uni = ls + (proto->last_uni >= ols ?
11244 proto->last_uni - ols : 0);
11245 parser->last_lop = ls + (proto->last_lop >= ols ?
11246 proto->last_lop - ols : 0);
11248 parser->bufend = ls + SvCUR(parser->linestr);
11251 Copy(proto->tokenbuf, parser->tokenbuf, 256, char);
11255 parser->endwhite = proto->endwhite;
11256 parser->faketokens = proto->faketokens;
11257 parser->lasttoke = proto->lasttoke;
11258 parser->nextwhite = proto->nextwhite;
11259 parser->realtokenstart = proto->realtokenstart;
11260 parser->skipwhite = proto->skipwhite;
11261 parser->thisclose = proto->thisclose;
11262 parser->thismad = proto->thismad;
11263 parser->thisopen = proto->thisopen;
11264 parser->thisstuff = proto->thisstuff;
11265 parser->thistoken = proto->thistoken;
11266 parser->thiswhite = proto->thiswhite;
11268 Copy(proto->nexttoke, parser->nexttoke, 5, NEXTTOKE);
11269 parser->curforce = proto->curforce;
11271 Copy(proto->nextval, parser->nextval, 5, YYSTYPE);
11272 Copy(proto->nexttype, parser->nexttype, 5, I32);
11273 parser->nexttoke = proto->nexttoke;
11276 /* XXX should clone saved_curcop here, but we aren't passed
11277 * proto_perl; so do it in perl_clone_using instead */
11283 /* duplicate a file handle */
11286 Perl_fp_dup(pTHX_ PerlIO *const fp, const char type, CLONE_PARAMS *const param)
11290 PERL_ARGS_ASSERT_FP_DUP;
11291 PERL_UNUSED_ARG(type);
11294 return (PerlIO*)NULL;
11296 /* look for it in the table first */
11297 ret = (PerlIO*)ptr_table_fetch(PL_ptr_table, fp);
11301 /* create anew and remember what it is */
11302 ret = PerlIO_fdupopen(aTHX_ fp, param, PERLIO_DUP_CLONE);
11303 ptr_table_store(PL_ptr_table, fp, ret);
11307 /* duplicate a directory handle */
11310 Perl_dirp_dup(pTHX_ DIR *const dp, CLONE_PARAMS *const param)
11316 register const Direntry_t *dirent;
11317 char smallbuf[256];
11323 PERL_UNUSED_CONTEXT;
11324 PERL_ARGS_ASSERT_DIRP_DUP;
11329 /* look for it in the table first */
11330 ret = (DIR*)ptr_table_fetch(PL_ptr_table, dp);
11336 PERL_UNUSED_ARG(param);
11340 /* open the current directory (so we can switch back) */
11341 if (!(pwd = PerlDir_open("."))) return (DIR *)NULL;
11343 /* chdir to our dir handle and open the present working directory */
11344 if (fchdir(my_dirfd(dp)) < 0 || !(ret = PerlDir_open("."))) {
11345 PerlDir_close(pwd);
11346 return (DIR *)NULL;
11348 /* Now we should have two dir handles pointing to the same dir. */
11350 /* Be nice to the calling code and chdir back to where we were. */
11351 fchdir(my_dirfd(pwd)); /* If this fails, then what? */
11353 /* We have no need of the pwd handle any more. */
11354 PerlDir_close(pwd);
11357 # define d_namlen(d) (d)->d_namlen
11359 # define d_namlen(d) strlen((d)->d_name)
11361 /* Iterate once through dp, to get the file name at the current posi-
11362 tion. Then step back. */
11363 pos = PerlDir_tell(dp);
11364 if ((dirent = PerlDir_read(dp))) {
11365 len = d_namlen(dirent);
11366 if (len <= sizeof smallbuf) name = smallbuf;
11367 else Newx(name, len, char);
11368 Move(dirent->d_name, name, len, char);
11370 PerlDir_seek(dp, pos);
11372 /* Iterate through the new dir handle, till we find a file with the
11374 if (!dirent) /* just before the end */
11376 pos = PerlDir_tell(ret);
11377 if (PerlDir_read(ret)) continue; /* not there yet */
11378 PerlDir_seek(ret, pos); /* step back */
11382 const long pos0 = PerlDir_tell(ret);
11384 pos = PerlDir_tell(ret);
11385 if ((dirent = PerlDir_read(ret))) {
11386 if (len == d_namlen(dirent)
11387 && memEQ(name, dirent->d_name, len)) {
11389 PerlDir_seek(ret, pos); /* step back */
11392 /* else we are not there yet; keep iterating */
11394 else { /* This is not meant to happen. The best we can do is
11395 reset the iterator to the beginning. */
11396 PerlDir_seek(ret, pos0);
11403 if (name && name != smallbuf)
11408 ret = win32_dirp_dup(dp, param);
11411 /* pop it in the pointer table */
11413 ptr_table_store(PL_ptr_table, dp, ret);
11418 /* duplicate a typeglob */
11421 Perl_gp_dup(pTHX_ GP *const gp, CLONE_PARAMS *const param)
11425 PERL_ARGS_ASSERT_GP_DUP;
11429 /* look for it in the table first */
11430 ret = (GP*)ptr_table_fetch(PL_ptr_table, gp);
11434 /* create anew and remember what it is */
11436 ptr_table_store(PL_ptr_table, gp, ret);
11439 /* ret->gp_refcnt must be 0 before any other dups are called. We're relying
11440 on Newxz() to do this for us. */
11441 ret->gp_sv = sv_dup_inc(gp->gp_sv, param);
11442 ret->gp_io = io_dup_inc(gp->gp_io, param);
11443 ret->gp_form = cv_dup_inc(gp->gp_form, param);
11444 ret->gp_av = av_dup_inc(gp->gp_av, param);
11445 ret->gp_hv = hv_dup_inc(gp->gp_hv, param);
11446 ret->gp_egv = gv_dup(gp->gp_egv, param);/* GvEGV is not refcounted */
11447 ret->gp_cv = cv_dup_inc(gp->gp_cv, param);
11448 ret->gp_cvgen = gp->gp_cvgen;
11449 ret->gp_line = gp->gp_line;
11450 ret->gp_file_hek = hek_dup(gp->gp_file_hek, param);
11454 /* duplicate a chain of magic */
11457 Perl_mg_dup(pTHX_ MAGIC *mg, CLONE_PARAMS *const param)
11459 MAGIC *mgret = NULL;
11460 MAGIC **mgprev_p = &mgret;
11462 PERL_ARGS_ASSERT_MG_DUP;
11464 for (; mg; mg = mg->mg_moremagic) {
11467 if ((param->flags & CLONEf_JOIN_IN)
11468 && mg->mg_type == PERL_MAGIC_backref)
11469 /* when joining, we let the individual SVs add themselves to
11470 * backref as needed. */
11473 Newx(nmg, 1, MAGIC);
11475 mgprev_p = &(nmg->mg_moremagic);
11477 /* There was a comment "XXX copy dynamic vtable?" but as we don't have
11478 dynamic vtables, I'm not sure why Sarathy wrote it. The comment dates
11479 from the original commit adding Perl_mg_dup() - revision 4538.
11480 Similarly there is the annotation "XXX random ptr?" next to the
11481 assignment to nmg->mg_ptr. */
11484 /* FIXME for plugins
11485 if (nmg->mg_type == PERL_MAGIC_qr) {
11486 nmg->mg_obj = MUTABLE_SV(CALLREGDUPE((REGEXP*)nmg->mg_obj, param));
11490 nmg->mg_obj = (nmg->mg_flags & MGf_REFCOUNTED)
11491 ? nmg->mg_type == PERL_MAGIC_backref
11492 /* The backref AV has its reference
11493 * count deliberately bumped by 1 */
11494 ? SvREFCNT_inc(av_dup_inc((const AV *)
11495 nmg->mg_obj, param))
11496 : sv_dup_inc(nmg->mg_obj, param)
11497 : sv_dup(nmg->mg_obj, param);
11499 if (nmg->mg_ptr && nmg->mg_type != PERL_MAGIC_regex_global) {
11500 if (nmg->mg_len > 0) {
11501 nmg->mg_ptr = SAVEPVN(nmg->mg_ptr, nmg->mg_len);
11502 if (nmg->mg_type == PERL_MAGIC_overload_table &&
11503 AMT_AMAGIC((AMT*)nmg->mg_ptr))
11505 AMT * const namtp = (AMT*)nmg->mg_ptr;
11506 sv_dup_inc_multiple((SV**)(namtp->table),
11507 (SV**)(namtp->table), NofAMmeth, param);
11510 else if (nmg->mg_len == HEf_SVKEY)
11511 nmg->mg_ptr = (char*)sv_dup_inc((const SV *)nmg->mg_ptr, param);
11513 if ((nmg->mg_flags & MGf_DUP) && nmg->mg_virtual && nmg->mg_virtual->svt_dup) {
11514 nmg->mg_virtual->svt_dup(aTHX_ nmg, param);
11520 #endif /* USE_ITHREADS */
11522 struct ptr_tbl_arena {
11523 struct ptr_tbl_arena *next;
11524 struct ptr_tbl_ent array[1023/3]; /* as ptr_tbl_ent has 3 pointers. */
11527 /* create a new pointer-mapping table */
11530 Perl_ptr_table_new(pTHX)
11533 PERL_UNUSED_CONTEXT;
11535 Newx(tbl, 1, PTR_TBL_t);
11536 tbl->tbl_max = 511;
11537 tbl->tbl_items = 0;
11538 tbl->tbl_arena = NULL;
11539 tbl->tbl_arena_next = NULL;
11540 tbl->tbl_arena_end = NULL;
11541 Newxz(tbl->tbl_ary, tbl->tbl_max + 1, PTR_TBL_ENT_t*);
11545 #define PTR_TABLE_HASH(ptr) \
11546 ((PTR2UV(ptr) >> 3) ^ (PTR2UV(ptr) >> (3 + 7)) ^ (PTR2UV(ptr) >> (3 + 17)))
11548 /* map an existing pointer using a table */
11550 STATIC PTR_TBL_ENT_t *
11551 S_ptr_table_find(PTR_TBL_t *const tbl, const void *const sv)
11553 PTR_TBL_ENT_t *tblent;
11554 const UV hash = PTR_TABLE_HASH(sv);
11556 PERL_ARGS_ASSERT_PTR_TABLE_FIND;
11558 tblent = tbl->tbl_ary[hash & tbl->tbl_max];
11559 for (; tblent; tblent = tblent->next) {
11560 if (tblent->oldval == sv)
11567 Perl_ptr_table_fetch(pTHX_ PTR_TBL_t *const tbl, const void *const sv)
11569 PTR_TBL_ENT_t const *const tblent = ptr_table_find(tbl, sv);
11571 PERL_ARGS_ASSERT_PTR_TABLE_FETCH;
11572 PERL_UNUSED_CONTEXT;
11574 return tblent ? tblent->newval : NULL;
11577 /* add a new entry to a pointer-mapping table */
11580 Perl_ptr_table_store(pTHX_ PTR_TBL_t *const tbl, const void *const oldsv, void *const newsv)
11582 PTR_TBL_ENT_t *tblent = ptr_table_find(tbl, oldsv);
11584 PERL_ARGS_ASSERT_PTR_TABLE_STORE;
11585 PERL_UNUSED_CONTEXT;
11588 tblent->newval = newsv;
11590 const UV entry = PTR_TABLE_HASH(oldsv) & tbl->tbl_max;
11592 if (tbl->tbl_arena_next == tbl->tbl_arena_end) {
11593 struct ptr_tbl_arena *new_arena;
11595 Newx(new_arena, 1, struct ptr_tbl_arena);
11596 new_arena->next = tbl->tbl_arena;
11597 tbl->tbl_arena = new_arena;
11598 tbl->tbl_arena_next = new_arena->array;
11599 tbl->tbl_arena_end = new_arena->array
11600 + sizeof(new_arena->array) / sizeof(new_arena->array[0]);
11603 tblent = tbl->tbl_arena_next++;
11605 tblent->oldval = oldsv;
11606 tblent->newval = newsv;
11607 tblent->next = tbl->tbl_ary[entry];
11608 tbl->tbl_ary[entry] = tblent;
11610 if (tblent->next && tbl->tbl_items > tbl->tbl_max)
11611 ptr_table_split(tbl);
11615 /* double the hash bucket size of an existing ptr table */
11618 Perl_ptr_table_split(pTHX_ PTR_TBL_t *const tbl)
11620 PTR_TBL_ENT_t **ary = tbl->tbl_ary;
11621 const UV oldsize = tbl->tbl_max + 1;
11622 UV newsize = oldsize * 2;
11625 PERL_ARGS_ASSERT_PTR_TABLE_SPLIT;
11626 PERL_UNUSED_CONTEXT;
11628 Renew(ary, newsize, PTR_TBL_ENT_t*);
11629 Zero(&ary[oldsize], newsize-oldsize, PTR_TBL_ENT_t*);
11630 tbl->tbl_max = --newsize;
11631 tbl->tbl_ary = ary;
11632 for (i=0; i < oldsize; i++, ary++) {
11633 PTR_TBL_ENT_t **entp = ary;
11634 PTR_TBL_ENT_t *ent = *ary;
11635 PTR_TBL_ENT_t **curentp;
11638 curentp = ary + oldsize;
11640 if ((newsize & PTR_TABLE_HASH(ent->oldval)) != i) {
11642 ent->next = *curentp;
11652 /* remove all the entries from a ptr table */
11653 /* Deprecated - will be removed post 5.14 */
11656 Perl_ptr_table_clear(pTHX_ PTR_TBL_t *const tbl)
11658 if (tbl && tbl->tbl_items) {
11659 struct ptr_tbl_arena *arena = tbl->tbl_arena;
11661 Zero(tbl->tbl_ary, tbl->tbl_max + 1, struct ptr_tbl_ent **);
11664 struct ptr_tbl_arena *next = arena->next;
11670 tbl->tbl_items = 0;
11671 tbl->tbl_arena = NULL;
11672 tbl->tbl_arena_next = NULL;
11673 tbl->tbl_arena_end = NULL;
11677 /* clear and free a ptr table */
11680 Perl_ptr_table_free(pTHX_ PTR_TBL_t *const tbl)
11682 struct ptr_tbl_arena *arena;
11688 arena = tbl->tbl_arena;
11691 struct ptr_tbl_arena *next = arena->next;
11697 Safefree(tbl->tbl_ary);
11701 #if defined(USE_ITHREADS)
11704 Perl_rvpv_dup(pTHX_ SV *const dstr, const SV *const sstr, CLONE_PARAMS *const param)
11706 PERL_ARGS_ASSERT_RVPV_DUP;
11709 if (SvWEAKREF(sstr)) {
11710 SvRV_set(dstr, sv_dup(SvRV_const(sstr), param));
11711 if (param->flags & CLONEf_JOIN_IN) {
11712 /* if joining, we add any back references individually rather
11713 * than copying the whole backref array */
11714 Perl_sv_add_backref(aTHX_ SvRV(dstr), dstr);
11718 SvRV_set(dstr, sv_dup_inc(SvRV_const(sstr), param));
11720 else if (SvPVX_const(sstr)) {
11721 /* Has something there */
11723 /* Normal PV - clone whole allocated space */
11724 SvPV_set(dstr, SAVEPVN(SvPVX_const(sstr), SvLEN(sstr)-1));
11725 if (SvREADONLY(sstr) && SvFAKE(sstr)) {
11726 /* Not that normal - actually sstr is copy on write.
11727 But we are a true, independent SV, so: */
11728 SvREADONLY_off(dstr);
11733 /* Special case - not normally malloced for some reason */
11734 if (isGV_with_GP(sstr)) {
11735 /* Don't need to do anything here. */
11737 else if ((SvREADONLY(sstr) && SvFAKE(sstr))) {
11738 /* A "shared" PV - clone it as "shared" PV */
11740 HEK_KEY(hek_dup(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)),
11744 /* Some other special case - random pointer */
11745 SvPV_set(dstr, (char *) SvPVX_const(sstr));
11750 /* Copy the NULL */
11751 SvPV_set(dstr, NULL);
11755 /* duplicate a list of SVs. source and dest may point to the same memory. */
11757 S_sv_dup_inc_multiple(pTHX_ SV *const *source, SV **dest,
11758 SSize_t items, CLONE_PARAMS *const param)
11760 PERL_ARGS_ASSERT_SV_DUP_INC_MULTIPLE;
11762 while (items-- > 0) {
11763 *dest++ = sv_dup_inc(*source++, param);
11769 /* duplicate an SV of any type (including AV, HV etc) */
11772 S_sv_dup_common(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
11777 PERL_ARGS_ASSERT_SV_DUP_COMMON;
11779 if (SvTYPE(sstr) == (svtype)SVTYPEMASK) {
11780 #ifdef DEBUG_LEAKING_SCALARS_ABORT
11785 /* look for it in the table first */
11786 dstr = MUTABLE_SV(ptr_table_fetch(PL_ptr_table, sstr));
11790 if(param->flags & CLONEf_JOIN_IN) {
11791 /** We are joining here so we don't want do clone
11792 something that is bad **/
11793 if (SvTYPE(sstr) == SVt_PVHV) {
11794 const HEK * const hvname = HvNAME_HEK(sstr);
11796 /** don't clone stashes if they already exist **/
11797 dstr = MUTABLE_SV(gv_stashpvn(HEK_KEY(hvname), HEK_LEN(hvname),
11798 HEK_UTF8(hvname) ? SVf_UTF8 : 0));
11799 ptr_table_store(PL_ptr_table, sstr, dstr);
11805 /* create anew and remember what it is */
11808 #ifdef DEBUG_LEAKING_SCALARS
11809 dstr->sv_debug_optype = sstr->sv_debug_optype;
11810 dstr->sv_debug_line = sstr->sv_debug_line;
11811 dstr->sv_debug_inpad = sstr->sv_debug_inpad;
11812 dstr->sv_debug_parent = (SV*)sstr;
11813 FREE_SV_DEBUG_FILE(dstr);
11814 dstr->sv_debug_file = savepv(sstr->sv_debug_file);
11817 ptr_table_store(PL_ptr_table, sstr, dstr);
11820 SvFLAGS(dstr) = SvFLAGS(sstr);
11821 SvFLAGS(dstr) &= ~SVf_OOK; /* don't propagate OOK hack */
11822 SvREFCNT(dstr) = 0; /* must be before any other dups! */
11825 if (SvANY(sstr) && PL_watch_pvx && SvPVX_const(sstr) == PL_watch_pvx)
11826 PerlIO_printf(Perl_debug_log, "watch at %p hit, found string \"%s\"\n",
11827 (void*)PL_watch_pvx, SvPVX_const(sstr));
11830 /* don't clone objects whose class has asked us not to */
11831 if (SvOBJECT(sstr) && ! (SvFLAGS(SvSTASH(sstr)) & SVphv_CLONEABLE)) {
11836 switch (SvTYPE(sstr)) {
11838 SvANY(dstr) = NULL;
11841 SvANY(dstr) = (XPVIV*)((char*)&(dstr->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
11843 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
11845 SvIV_set(dstr, SvIVX(sstr));
11849 SvANY(dstr) = new_XNV();
11850 SvNV_set(dstr, SvNVX(sstr));
11852 /* case SVt_BIND: */
11855 /* These are all the types that need complex bodies allocating. */
11857 const svtype sv_type = SvTYPE(sstr);
11858 const struct body_details *const sv_type_details
11859 = bodies_by_type + sv_type;
11863 Perl_croak(aTHX_ "Bizarre SvTYPE [%" IVdf "]", (IV)SvTYPE(sstr));
11878 assert(sv_type_details->body_size);
11879 if (sv_type_details->arena) {
11880 new_body_inline(new_body, sv_type);
11882 = (void*)((char*)new_body - sv_type_details->offset);
11884 new_body = new_NOARENA(sv_type_details);
11888 SvANY(dstr) = new_body;
11891 Copy(((char*)SvANY(sstr)) + sv_type_details->offset,
11892 ((char*)SvANY(dstr)) + sv_type_details->offset,
11893 sv_type_details->copy, char);
11895 Copy(((char*)SvANY(sstr)),
11896 ((char*)SvANY(dstr)),
11897 sv_type_details->body_size + sv_type_details->offset, char);
11900 if (sv_type != SVt_PVAV && sv_type != SVt_PVHV
11901 && !isGV_with_GP(dstr)
11902 && !(sv_type == SVt_PVIO && !(IoFLAGS(dstr) & IOf_FAKE_DIRP)))
11903 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
11905 /* The Copy above means that all the source (unduplicated) pointers
11906 are now in the destination. We can check the flags and the
11907 pointers in either, but it's possible that there's less cache
11908 missing by always going for the destination.
11909 FIXME - instrument and check that assumption */
11910 if (sv_type >= SVt_PVMG) {
11911 if ((sv_type == SVt_PVMG) && SvPAD_OUR(dstr)) {
11912 SvOURSTASH_set(dstr, hv_dup_inc(SvOURSTASH(dstr), param));
11913 } else if (SvMAGIC(dstr))
11914 SvMAGIC_set(dstr, mg_dup(SvMAGIC(dstr), param));
11916 SvSTASH_set(dstr, hv_dup_inc(SvSTASH(dstr), param));
11919 /* The cast silences a GCC warning about unhandled types. */
11920 switch ((int)sv_type) {
11930 /* FIXME for plugins */
11931 re_dup_guts((REGEXP*) sstr, (REGEXP*) dstr, param);
11934 /* XXX LvTARGOFF sometimes holds PMOP* when DEBUGGING */
11935 if (LvTYPE(dstr) == 't') /* for tie: unrefcnted fake (SV**) */
11936 LvTARG(dstr) = dstr;
11937 else if (LvTYPE(dstr) == 'T') /* for tie: fake HE */
11938 LvTARG(dstr) = MUTABLE_SV(he_dup((HE*)LvTARG(dstr), 0, param));
11940 LvTARG(dstr) = sv_dup_inc(LvTARG(dstr), param);
11942 /* non-GP case already handled above */
11943 if(isGV_with_GP(sstr)) {
11944 GvNAME_HEK(dstr) = hek_dup(GvNAME_HEK(dstr), param);
11945 /* Don't call sv_add_backref here as it's going to be
11946 created as part of the magic cloning of the symbol
11947 table--unless this is during a join and the stash
11948 is not actually being cloned. */
11949 /* Danger Will Robinson - GvGP(dstr) isn't initialised
11950 at the point of this comment. */
11951 GvSTASH(dstr) = hv_dup(GvSTASH(dstr), param);
11952 if (param->flags & CLONEf_JOIN_IN)
11953 Perl_sv_add_backref(aTHX_ MUTABLE_SV(GvSTASH(dstr)), dstr);
11954 GvGP_set(dstr, gp_dup(GvGP(sstr), param));
11955 (void)GpREFCNT_inc(GvGP(dstr));
11959 /* PL_parser->rsfp_filters entries have fake IoDIRP() */
11960 if(IoFLAGS(dstr) & IOf_FAKE_DIRP) {
11961 /* I have no idea why fake dirp (rsfps)
11962 should be treated differently but otherwise
11963 we end up with leaks -- sky*/
11964 IoTOP_GV(dstr) = gv_dup_inc(IoTOP_GV(dstr), param);
11965 IoFMT_GV(dstr) = gv_dup_inc(IoFMT_GV(dstr), param);
11966 IoBOTTOM_GV(dstr) = gv_dup_inc(IoBOTTOM_GV(dstr), param);
11968 IoTOP_GV(dstr) = gv_dup(IoTOP_GV(dstr), param);
11969 IoFMT_GV(dstr) = gv_dup(IoFMT_GV(dstr), param);
11970 IoBOTTOM_GV(dstr) = gv_dup(IoBOTTOM_GV(dstr), param);
11971 if (IoDIRP(dstr)) {
11972 IoDIRP(dstr) = dirp_dup(IoDIRP(dstr), param);
11975 /* IoDIRP(dstr) is already a copy of IoDIRP(sstr) */
11977 IoIFP(dstr) = fp_dup(IoIFP(sstr), IoTYPE(dstr), param);
11979 if (IoOFP(dstr) == IoIFP(sstr))
11980 IoOFP(dstr) = IoIFP(dstr);
11982 IoOFP(dstr) = fp_dup(IoOFP(dstr), IoTYPE(dstr), param);
11983 IoTOP_NAME(dstr) = SAVEPV(IoTOP_NAME(dstr));
11984 IoFMT_NAME(dstr) = SAVEPV(IoFMT_NAME(dstr));
11985 IoBOTTOM_NAME(dstr) = SAVEPV(IoBOTTOM_NAME(dstr));
11988 /* avoid cloning an empty array */
11989 if (AvARRAY((const AV *)sstr) && AvFILLp((const AV *)sstr) >= 0) {
11990 SV **dst_ary, **src_ary;
11991 SSize_t items = AvFILLp((const AV *)sstr) + 1;
11993 src_ary = AvARRAY((const AV *)sstr);
11994 Newxz(dst_ary, AvMAX((const AV *)sstr)+1, SV*);
11995 ptr_table_store(PL_ptr_table, src_ary, dst_ary);
11996 AvARRAY(MUTABLE_AV(dstr)) = dst_ary;
11997 AvALLOC((const AV *)dstr) = dst_ary;
11998 if (AvREAL((const AV *)sstr)) {
11999 dst_ary = sv_dup_inc_multiple(src_ary, dst_ary, items,
12003 while (items-- > 0)
12004 *dst_ary++ = sv_dup(*src_ary++, param);
12006 items = AvMAX((const AV *)sstr) - AvFILLp((const AV *)sstr);
12007 while (items-- > 0) {
12008 *dst_ary++ = &PL_sv_undef;
12012 AvARRAY(MUTABLE_AV(dstr)) = NULL;
12013 AvALLOC((const AV *)dstr) = (SV**)NULL;
12014 AvMAX( (const AV *)dstr) = -1;
12015 AvFILLp((const AV *)dstr) = -1;
12019 if (HvARRAY((const HV *)sstr)) {
12021 const bool sharekeys = !!HvSHAREKEYS(sstr);
12022 XPVHV * const dxhv = (XPVHV*)SvANY(dstr);
12023 XPVHV * const sxhv = (XPVHV*)SvANY(sstr);
12025 Newx(darray, PERL_HV_ARRAY_ALLOC_BYTES(dxhv->xhv_max+1)
12026 + (SvOOK(sstr) ? sizeof(struct xpvhv_aux) : 0),
12028 HvARRAY(dstr) = (HE**)darray;
12029 while (i <= sxhv->xhv_max) {
12030 const HE * const source = HvARRAY(sstr)[i];
12031 HvARRAY(dstr)[i] = source
12032 ? he_dup(source, sharekeys, param) : 0;
12036 const struct xpvhv_aux * const saux = HvAUX(sstr);
12037 struct xpvhv_aux * const daux = HvAUX(dstr);
12038 /* This flag isn't copied. */
12039 /* SvOOK_on(hv) attacks the IV flags. */
12040 SvFLAGS(dstr) |= SVf_OOK;
12042 if (saux->xhv_name_count) {
12043 HEK ** const sname = saux->xhv_name_u.xhvnameu_names;
12045 = saux->xhv_name_count < 0
12046 ? -saux->xhv_name_count
12047 : saux->xhv_name_count;
12048 HEK **shekp = sname + count;
12050 Newx(daux->xhv_name_u.xhvnameu_names, count, HEK *);
12051 dhekp = daux->xhv_name_u.xhvnameu_names + count;
12052 while (shekp-- > sname) {
12054 *dhekp = hek_dup(*shekp, param);
12058 daux->xhv_name_u.xhvnameu_name
12059 = hek_dup(saux->xhv_name_u.xhvnameu_name,
12062 daux->xhv_name_count = saux->xhv_name_count;
12064 daux->xhv_riter = saux->xhv_riter;
12065 daux->xhv_eiter = saux->xhv_eiter
12066 ? he_dup(saux->xhv_eiter,
12067 cBOOL(HvSHAREKEYS(sstr)), param) : 0;
12068 /* backref array needs refcnt=2; see sv_add_backref */
12069 daux->xhv_backreferences =
12070 (param->flags & CLONEf_JOIN_IN)
12071 /* when joining, we let the individual GVs and
12072 * CVs add themselves to backref as
12073 * needed. This avoids pulling in stuff
12074 * that isn't required, and simplifies the
12075 * case where stashes aren't cloned back
12076 * if they already exist in the parent
12079 : saux->xhv_backreferences
12080 ? (SvTYPE(saux->xhv_backreferences) == SVt_PVAV)
12081 ? MUTABLE_AV(SvREFCNT_inc(
12082 sv_dup_inc((const SV *)
12083 saux->xhv_backreferences, param)))
12084 : MUTABLE_AV(sv_dup((const SV *)
12085 saux->xhv_backreferences, param))
12088 daux->xhv_mro_meta = saux->xhv_mro_meta
12089 ? mro_meta_dup(saux->xhv_mro_meta, param)
12092 /* Record stashes for possible cloning in Perl_clone(). */
12094 av_push(param->stashes, dstr);
12098 HvARRAY(MUTABLE_HV(dstr)) = NULL;
12101 if (!(param->flags & CLONEf_COPY_STACKS)) {
12106 /* NOTE: not refcounted */
12107 SvANY(MUTABLE_CV(dstr))->xcv_stash =
12108 hv_dup(CvSTASH(dstr), param);
12109 if ((param->flags & CLONEf_JOIN_IN) && CvSTASH(dstr))
12110 Perl_sv_add_backref(aTHX_ MUTABLE_SV(CvSTASH(dstr)), dstr);
12111 if (!CvISXSUB(dstr)) {
12113 CvROOT(dstr) = OpREFCNT_inc(CvROOT(dstr));
12115 } else if (CvCONST(dstr)) {
12116 CvXSUBANY(dstr).any_ptr =
12117 sv_dup_inc((const SV *)CvXSUBANY(dstr).any_ptr, param);
12119 if (CvDYNFILE(dstr)) CvFILE(dstr) = SAVEPV(CvFILE(dstr));
12120 /* don't dup if copying back - CvGV isn't refcounted, so the
12121 * duped GV may never be freed. A bit of a hack! DAPM */
12122 SvANY(MUTABLE_CV(dstr))->xcv_gv =
12124 ? gv_dup_inc(CvGV(sstr), param)
12125 : (param->flags & CLONEf_JOIN_IN)
12127 : gv_dup(CvGV(sstr), param);
12129 CvPADLIST(dstr) = padlist_dup(CvPADLIST(sstr), param);
12131 CvWEAKOUTSIDE(sstr)
12132 ? cv_dup( CvOUTSIDE(dstr), param)
12133 : cv_dup_inc(CvOUTSIDE(dstr), param);
12139 if (SvOBJECT(dstr) && SvTYPE(dstr) != SVt_PVIO)
12146 Perl_sv_dup_inc(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
12148 PERL_ARGS_ASSERT_SV_DUP_INC;
12149 return sstr ? SvREFCNT_inc(sv_dup_common(sstr, param)) : NULL;
12153 Perl_sv_dup(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
12155 SV *dstr = sstr ? sv_dup_common(sstr, param) : NULL;
12156 PERL_ARGS_ASSERT_SV_DUP;
12158 /* Track every SV that (at least initially) had a reference count of 0.
12159 We need to do this by holding an actual reference to it in this array.
12160 If we attempt to cheat, turn AvREAL_off(), and store only pointers
12161 (akin to the stashes hash, and the perl stack), we come unstuck if
12162 a weak reference (or other SV legitimately SvREFCNT() == 0 for this
12163 thread) is manipulated in a CLONE method, because CLONE runs before the
12164 unreferenced array is walked to find SVs still with SvREFCNT() == 0
12165 (and fix things up by giving each a reference via the temps stack).
12166 Instead, during CLONE, if the 0-referenced SV has SvREFCNT_inc() and
12167 then SvREFCNT_dec(), it will be cleaned up (and added to the free list)
12168 before the walk of unreferenced happens and a reference to that is SV
12169 added to the temps stack. At which point we have the same SV considered
12170 to be in use, and free to be re-used. Not good.
12172 if (dstr && !(param->flags & CLONEf_COPY_STACKS) && !SvREFCNT(dstr)) {
12173 assert(param->unreferenced);
12174 av_push(param->unreferenced, SvREFCNT_inc(dstr));
12180 /* duplicate a context */
12183 Perl_cx_dup(pTHX_ PERL_CONTEXT *cxs, I32 ix, I32 max, CLONE_PARAMS* param)
12185 PERL_CONTEXT *ncxs;
12187 PERL_ARGS_ASSERT_CX_DUP;
12190 return (PERL_CONTEXT*)NULL;
12192 /* look for it in the table first */
12193 ncxs = (PERL_CONTEXT*)ptr_table_fetch(PL_ptr_table, cxs);
12197 /* create anew and remember what it is */
12198 Newx(ncxs, max + 1, PERL_CONTEXT);
12199 ptr_table_store(PL_ptr_table, cxs, ncxs);
12200 Copy(cxs, ncxs, max + 1, PERL_CONTEXT);
12203 PERL_CONTEXT * const ncx = &ncxs[ix];
12204 if (CxTYPE(ncx) == CXt_SUBST) {
12205 Perl_croak(aTHX_ "Cloning substitution context is unimplemented");
12208 switch (CxTYPE(ncx)) {
12210 ncx->blk_sub.cv = (ncx->blk_sub.olddepth == 0
12211 ? cv_dup_inc(ncx->blk_sub.cv, param)
12212 : cv_dup(ncx->blk_sub.cv,param));
12213 ncx->blk_sub.argarray = (CxHASARGS(ncx)
12214 ? av_dup_inc(ncx->blk_sub.argarray,
12217 ncx->blk_sub.savearray = av_dup_inc(ncx->blk_sub.savearray,
12219 ncx->blk_sub.oldcomppad = (PAD*)ptr_table_fetch(PL_ptr_table,
12220 ncx->blk_sub.oldcomppad);
12223 ncx->blk_eval.old_namesv = sv_dup_inc(ncx->blk_eval.old_namesv,
12225 ncx->blk_eval.cur_text = sv_dup(ncx->blk_eval.cur_text, param);
12227 case CXt_LOOP_LAZYSV:
12228 ncx->blk_loop.state_u.lazysv.end
12229 = sv_dup_inc(ncx->blk_loop.state_u.lazysv.end, param);
12230 /* We are taking advantage of av_dup_inc and sv_dup_inc
12231 actually being the same function, and order equivalence of
12233 We can assert the later [but only at run time :-(] */
12234 assert ((void *) &ncx->blk_loop.state_u.ary.ary ==
12235 (void *) &ncx->blk_loop.state_u.lazysv.cur);
12237 ncx->blk_loop.state_u.ary.ary
12238 = av_dup_inc(ncx->blk_loop.state_u.ary.ary, param);
12239 case CXt_LOOP_LAZYIV:
12240 case CXt_LOOP_PLAIN:
12241 if (CxPADLOOP(ncx)) {
12242 ncx->blk_loop.itervar_u.oldcomppad
12243 = (PAD*)ptr_table_fetch(PL_ptr_table,
12244 ncx->blk_loop.itervar_u.oldcomppad);
12246 ncx->blk_loop.itervar_u.gv
12247 = gv_dup((const GV *)ncx->blk_loop.itervar_u.gv,
12252 ncx->blk_format.cv = cv_dup(ncx->blk_format.cv, param);
12253 ncx->blk_format.gv = gv_dup(ncx->blk_format.gv, param);
12254 ncx->blk_format.dfoutgv = gv_dup_inc(ncx->blk_format.dfoutgv,
12267 /* duplicate a stack info structure */
12270 Perl_si_dup(pTHX_ PERL_SI *si, CLONE_PARAMS* param)
12274 PERL_ARGS_ASSERT_SI_DUP;
12277 return (PERL_SI*)NULL;
12279 /* look for it in the table first */
12280 nsi = (PERL_SI*)ptr_table_fetch(PL_ptr_table, si);
12284 /* create anew and remember what it is */
12285 Newxz(nsi, 1, PERL_SI);
12286 ptr_table_store(PL_ptr_table, si, nsi);
12288 nsi->si_stack = av_dup_inc(si->si_stack, param);
12289 nsi->si_cxix = si->si_cxix;
12290 nsi->si_cxmax = si->si_cxmax;
12291 nsi->si_cxstack = cx_dup(si->si_cxstack, si->si_cxix, si->si_cxmax, param);
12292 nsi->si_type = si->si_type;
12293 nsi->si_prev = si_dup(si->si_prev, param);
12294 nsi->si_next = si_dup(si->si_next, param);
12295 nsi->si_markoff = si->si_markoff;
12300 #define POPINT(ss,ix) ((ss)[--(ix)].any_i32)
12301 #define TOPINT(ss,ix) ((ss)[ix].any_i32)
12302 #define POPLONG(ss,ix) ((ss)[--(ix)].any_long)
12303 #define TOPLONG(ss,ix) ((ss)[ix].any_long)
12304 #define POPIV(ss,ix) ((ss)[--(ix)].any_iv)
12305 #define TOPIV(ss,ix) ((ss)[ix].any_iv)
12306 #define POPUV(ss,ix) ((ss)[--(ix)].any_uv)
12307 #define TOPUV(ss,ix) ((ss)[ix].any_uv)
12308 #define POPBOOL(ss,ix) ((ss)[--(ix)].any_bool)
12309 #define TOPBOOL(ss,ix) ((ss)[ix].any_bool)
12310 #define POPPTR(ss,ix) ((ss)[--(ix)].any_ptr)
12311 #define TOPPTR(ss,ix) ((ss)[ix].any_ptr)
12312 #define POPDPTR(ss,ix) ((ss)[--(ix)].any_dptr)
12313 #define TOPDPTR(ss,ix) ((ss)[ix].any_dptr)
12314 #define POPDXPTR(ss,ix) ((ss)[--(ix)].any_dxptr)
12315 #define TOPDXPTR(ss,ix) ((ss)[ix].any_dxptr)
12318 #define pv_dup_inc(p) SAVEPV(p)
12319 #define pv_dup(p) SAVEPV(p)
12320 #define svp_dup_inc(p,pp) any_dup(p,pp)
12322 /* map any object to the new equivent - either something in the
12323 * ptr table, or something in the interpreter structure
12327 Perl_any_dup(pTHX_ void *v, const PerlInterpreter *proto_perl)
12331 PERL_ARGS_ASSERT_ANY_DUP;
12334 return (void*)NULL;
12336 /* look for it in the table first */
12337 ret = ptr_table_fetch(PL_ptr_table, v);
12341 /* see if it is part of the interpreter structure */
12342 if (v >= (void*)proto_perl && v < (void*)(proto_perl+1))
12343 ret = (void*)(((char*)aTHX) + (((char*)v) - (char*)proto_perl));
12351 /* duplicate the save stack */
12354 Perl_ss_dup(pTHX_ PerlInterpreter *proto_perl, CLONE_PARAMS* param)
12357 ANY * const ss = proto_perl->Isavestack;
12358 const I32 max = proto_perl->Isavestack_max;
12359 I32 ix = proto_perl->Isavestack_ix;
12372 void (*dptr) (void*);
12373 void (*dxptr) (pTHX_ void*);
12375 PERL_ARGS_ASSERT_SS_DUP;
12377 Newxz(nss, max, ANY);
12380 const UV uv = POPUV(ss,ix);
12381 const U8 type = (U8)uv & SAVE_MASK;
12383 TOPUV(nss,ix) = uv;
12385 case SAVEt_CLEARSV:
12387 case SAVEt_HELEM: /* hash element */
12388 sv = (const SV *)POPPTR(ss,ix);
12389 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12391 case SAVEt_ITEM: /* normal string */
12392 case SAVEt_GVSV: /* scalar slot in GV */
12393 case SAVEt_SV: /* scalar reference */
12394 sv = (const SV *)POPPTR(ss,ix);
12395 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12398 case SAVEt_MORTALIZESV:
12399 sv = (const SV *)POPPTR(ss,ix);
12400 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12402 case SAVEt_SHARED_PVREF: /* char* in shared space */
12403 c = (char*)POPPTR(ss,ix);
12404 TOPPTR(nss,ix) = savesharedpv(c);
12405 ptr = POPPTR(ss,ix);
12406 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12408 case SAVEt_GENERIC_SVREF: /* generic sv */
12409 case SAVEt_SVREF: /* scalar reference */
12410 sv = (const SV *)POPPTR(ss,ix);
12411 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12412 ptr = POPPTR(ss,ix);
12413 TOPPTR(nss,ix) = svp_dup_inc((SV**)ptr, proto_perl);/* XXXXX */
12415 case SAVEt_HV: /* hash reference */
12416 case SAVEt_AV: /* array reference */
12417 sv = (const SV *) POPPTR(ss,ix);
12418 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12420 case SAVEt_COMPPAD:
12422 sv = (const SV *) POPPTR(ss,ix);
12423 TOPPTR(nss,ix) = sv_dup(sv, param);
12425 case SAVEt_INT: /* int reference */
12426 ptr = POPPTR(ss,ix);
12427 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12428 intval = (int)POPINT(ss,ix);
12429 TOPINT(nss,ix) = intval;
12431 case SAVEt_LONG: /* long reference */
12432 ptr = POPPTR(ss,ix);
12433 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12434 longval = (long)POPLONG(ss,ix);
12435 TOPLONG(nss,ix) = longval;
12437 case SAVEt_I32: /* I32 reference */
12438 ptr = POPPTR(ss,ix);
12439 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12441 TOPINT(nss,ix) = i;
12443 case SAVEt_IV: /* IV reference */
12444 ptr = POPPTR(ss,ix);
12445 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12447 TOPIV(nss,ix) = iv;
12449 case SAVEt_HPTR: /* HV* reference */
12450 case SAVEt_APTR: /* AV* reference */
12451 case SAVEt_SPTR: /* SV* reference */
12452 ptr = POPPTR(ss,ix);
12453 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12454 sv = (const SV *)POPPTR(ss,ix);
12455 TOPPTR(nss,ix) = sv_dup(sv, param);
12457 case SAVEt_VPTR: /* random* reference */
12458 ptr = POPPTR(ss,ix);
12459 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12461 case SAVEt_INT_SMALL:
12462 case SAVEt_I32_SMALL:
12463 case SAVEt_I16: /* I16 reference */
12464 case SAVEt_I8: /* I8 reference */
12466 ptr = POPPTR(ss,ix);
12467 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12469 case SAVEt_GENERIC_PVREF: /* generic char* */
12470 case SAVEt_PPTR: /* char* reference */
12471 ptr = POPPTR(ss,ix);
12472 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12473 c = (char*)POPPTR(ss,ix);
12474 TOPPTR(nss,ix) = pv_dup(c);
12476 case SAVEt_GP: /* scalar reference */
12477 gp = (GP*)POPPTR(ss,ix);
12478 TOPPTR(nss,ix) = gp = gp_dup(gp, param);
12479 (void)GpREFCNT_inc(gp);
12480 gv = (const GV *)POPPTR(ss,ix);
12481 TOPPTR(nss,ix) = gv_dup_inc(gv, param);
12484 ptr = POPPTR(ss,ix);
12485 if (ptr && (((OP*)ptr)->op_private & OPpREFCOUNTED)) {
12486 /* these are assumed to be refcounted properly */
12488 switch (((OP*)ptr)->op_type) {
12490 case OP_LEAVESUBLV:
12494 case OP_LEAVEWRITE:
12495 TOPPTR(nss,ix) = ptr;
12498 (void) OpREFCNT_inc(o);
12502 TOPPTR(nss,ix) = NULL;
12507 TOPPTR(nss,ix) = NULL;
12509 case SAVEt_FREECOPHH:
12510 ptr = POPPTR(ss,ix);
12511 TOPPTR(nss,ix) = cophh_copy((COPHH *)ptr);
12514 hv = (const HV *)POPPTR(ss,ix);
12515 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
12517 TOPINT(nss,ix) = i;
12520 c = (char*)POPPTR(ss,ix);
12521 TOPPTR(nss,ix) = pv_dup_inc(c);
12523 case SAVEt_STACK_POS: /* Position on Perl stack */
12525 TOPINT(nss,ix) = i;
12527 case SAVEt_DESTRUCTOR:
12528 ptr = POPPTR(ss,ix);
12529 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
12530 dptr = POPDPTR(ss,ix);
12531 TOPDPTR(nss,ix) = DPTR2FPTR(void (*)(void*),
12532 any_dup(FPTR2DPTR(void *, dptr),
12535 case SAVEt_DESTRUCTOR_X:
12536 ptr = POPPTR(ss,ix);
12537 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
12538 dxptr = POPDXPTR(ss,ix);
12539 TOPDXPTR(nss,ix) = DPTR2FPTR(void (*)(pTHX_ void*),
12540 any_dup(FPTR2DPTR(void *, dxptr),
12543 case SAVEt_REGCONTEXT:
12545 ix -= uv >> SAVE_TIGHT_SHIFT;
12547 case SAVEt_AELEM: /* array element */
12548 sv = (const SV *)POPPTR(ss,ix);
12549 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12551 TOPINT(nss,ix) = i;
12552 av = (const AV *)POPPTR(ss,ix);
12553 TOPPTR(nss,ix) = av_dup_inc(av, param);
12556 ptr = POPPTR(ss,ix);
12557 TOPPTR(nss,ix) = ptr;
12560 ptr = POPPTR(ss,ix);
12561 ptr = cophh_copy((COPHH*)ptr);
12562 TOPPTR(nss,ix) = ptr;
12564 TOPINT(nss,ix) = i;
12565 if (i & HINT_LOCALIZE_HH) {
12566 hv = (const HV *)POPPTR(ss,ix);
12567 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
12570 case SAVEt_PADSV_AND_MORTALIZE:
12571 longval = (long)POPLONG(ss,ix);
12572 TOPLONG(nss,ix) = longval;
12573 ptr = POPPTR(ss,ix);
12574 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12575 sv = (const SV *)POPPTR(ss,ix);
12576 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12578 case SAVEt_SET_SVFLAGS:
12580 TOPINT(nss,ix) = i;
12582 TOPINT(nss,ix) = i;
12583 sv = (const SV *)POPPTR(ss,ix);
12584 TOPPTR(nss,ix) = sv_dup(sv, param);
12586 case SAVEt_RE_STATE:
12588 const struct re_save_state *const old_state
12589 = (struct re_save_state *)
12590 (ss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
12591 struct re_save_state *const new_state
12592 = (struct re_save_state *)
12593 (nss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
12595 Copy(old_state, new_state, 1, struct re_save_state);
12596 ix -= SAVESTACK_ALLOC_FOR_RE_SAVE_STATE;
12598 new_state->re_state_bostr
12599 = pv_dup(old_state->re_state_bostr);
12600 new_state->re_state_reginput
12601 = pv_dup(old_state->re_state_reginput);
12602 new_state->re_state_regeol
12603 = pv_dup(old_state->re_state_regeol);
12604 new_state->re_state_regoffs
12605 = (regexp_paren_pair*)
12606 any_dup(old_state->re_state_regoffs, proto_perl);
12607 new_state->re_state_reglastparen
12608 = (U32*) any_dup(old_state->re_state_reglastparen,
12610 new_state->re_state_reglastcloseparen
12611 = (U32*)any_dup(old_state->re_state_reglastcloseparen,
12613 /* XXX This just has to be broken. The old save_re_context
12614 code did SAVEGENERICPV(PL_reg_start_tmp);
12615 PL_reg_start_tmp is char **.
12616 Look above to what the dup code does for
12617 SAVEt_GENERIC_PVREF
12618 It can never have worked.
12619 So this is merely a faithful copy of the exiting bug: */
12620 new_state->re_state_reg_start_tmp
12621 = (char **) pv_dup((char *)
12622 old_state->re_state_reg_start_tmp);
12623 /* I assume that it only ever "worked" because no-one called
12624 (pseudo)fork while the regexp engine had re-entered itself.
12626 #ifdef PERL_OLD_COPY_ON_WRITE
12627 new_state->re_state_nrs
12628 = sv_dup(old_state->re_state_nrs, param);
12630 new_state->re_state_reg_magic
12631 = (MAGIC*) any_dup(old_state->re_state_reg_magic,
12633 new_state->re_state_reg_oldcurpm
12634 = (PMOP*) any_dup(old_state->re_state_reg_oldcurpm,
12636 new_state->re_state_reg_curpm
12637 = (PMOP*) any_dup(old_state->re_state_reg_curpm,
12639 new_state->re_state_reg_oldsaved
12640 = pv_dup(old_state->re_state_reg_oldsaved);
12641 new_state->re_state_reg_poscache
12642 = pv_dup(old_state->re_state_reg_poscache);
12643 new_state->re_state_reg_starttry
12644 = pv_dup(old_state->re_state_reg_starttry);
12647 case SAVEt_COMPILE_WARNINGS:
12648 ptr = POPPTR(ss,ix);
12649 TOPPTR(nss,ix) = DUP_WARNINGS((STRLEN*)ptr);
12652 ptr = POPPTR(ss,ix);
12653 TOPPTR(nss,ix) = parser_dup((const yy_parser*)ptr, param);
12657 "panic: ss_dup inconsistency (%"IVdf")", (IV) type);
12665 /* if sv is a stash, call $class->CLONE_SKIP(), and set the SVphv_CLONEABLE
12666 * flag to the result. This is done for each stash before cloning starts,
12667 * so we know which stashes want their objects cloned */
12670 do_mark_cloneable_stash(pTHX_ SV *const sv)
12672 const HEK * const hvname = HvNAME_HEK((const HV *)sv);
12674 GV* const cloner = gv_fetchmethod_autoload(MUTABLE_HV(sv), "CLONE_SKIP", 0);
12675 SvFLAGS(sv) |= SVphv_CLONEABLE; /* clone objects by default */
12676 if (cloner && GvCV(cloner)) {
12683 mXPUSHs(newSVhek(hvname));
12685 call_sv(MUTABLE_SV(GvCV(cloner)), G_SCALAR);
12692 SvFLAGS(sv) &= ~SVphv_CLONEABLE;
12700 =for apidoc perl_clone
12702 Create and return a new interpreter by cloning the current one.
12704 perl_clone takes these flags as parameters:
12706 CLONEf_COPY_STACKS - is used to, well, copy the stacks also,
12707 without it we only clone the data and zero the stacks,
12708 with it we copy the stacks and the new perl interpreter is
12709 ready to run at the exact same point as the previous one.
12710 The pseudo-fork code uses COPY_STACKS while the
12711 threads->create doesn't.
12713 CLONEf_KEEP_PTR_TABLE
12714 perl_clone keeps a ptr_table with the pointer of the old
12715 variable as a key and the new variable as a value,
12716 this allows it to check if something has been cloned and not
12717 clone it again but rather just use the value and increase the
12718 refcount. If KEEP_PTR_TABLE is not set then perl_clone will kill
12719 the ptr_table using the function
12720 C<ptr_table_free(PL_ptr_table); PL_ptr_table = NULL;>,
12721 reason to keep it around is if you want to dup some of your own
12722 variable who are outside the graph perl scans, example of this
12723 code is in threads.xs create
12726 This is a win32 thing, it is ignored on unix, it tells perls
12727 win32host code (which is c++) to clone itself, this is needed on
12728 win32 if you want to run two threads at the same time,
12729 if you just want to do some stuff in a separate perl interpreter
12730 and then throw it away and return to the original one,
12731 you don't need to do anything.
12736 /* XXX the above needs expanding by someone who actually understands it ! */
12737 EXTERN_C PerlInterpreter *
12738 perl_clone_host(PerlInterpreter* proto_perl, UV flags);
12741 perl_clone(PerlInterpreter *proto_perl, UV flags)
12744 #ifdef PERL_IMPLICIT_SYS
12746 PERL_ARGS_ASSERT_PERL_CLONE;
12748 /* perlhost.h so we need to call into it
12749 to clone the host, CPerlHost should have a c interface, sky */
12751 if (flags & CLONEf_CLONE_HOST) {
12752 return perl_clone_host(proto_perl,flags);
12754 return perl_clone_using(proto_perl, flags,
12756 proto_perl->IMemShared,
12757 proto_perl->IMemParse,
12759 proto_perl->IStdIO,
12763 proto_perl->IProc);
12767 perl_clone_using(PerlInterpreter *proto_perl, UV flags,
12768 struct IPerlMem* ipM, struct IPerlMem* ipMS,
12769 struct IPerlMem* ipMP, struct IPerlEnv* ipE,
12770 struct IPerlStdIO* ipStd, struct IPerlLIO* ipLIO,
12771 struct IPerlDir* ipD, struct IPerlSock* ipS,
12772 struct IPerlProc* ipP)
12774 /* XXX many of the string copies here can be optimized if they're
12775 * constants; they need to be allocated as common memory and just
12776 * their pointers copied. */
12779 CLONE_PARAMS clone_params;
12780 CLONE_PARAMS* const param = &clone_params;
12782 PerlInterpreter * const my_perl = (PerlInterpreter*)(*ipM->pMalloc)(ipM, sizeof(PerlInterpreter));
12784 PERL_ARGS_ASSERT_PERL_CLONE_USING;
12785 #else /* !PERL_IMPLICIT_SYS */
12787 CLONE_PARAMS clone_params;
12788 CLONE_PARAMS* param = &clone_params;
12789 PerlInterpreter * const my_perl = (PerlInterpreter*)PerlMem_malloc(sizeof(PerlInterpreter));
12791 PERL_ARGS_ASSERT_PERL_CLONE;
12792 #endif /* PERL_IMPLICIT_SYS */
12794 /* for each stash, determine whether its objects should be cloned */
12795 S_visit(proto_perl, do_mark_cloneable_stash, SVt_PVHV, SVTYPEMASK);
12796 PERL_SET_THX(my_perl);
12799 PoisonNew(my_perl, 1, PerlInterpreter);
12802 PL_defstash = NULL; /* may be used by perl malloc() */
12805 PL_scopestack_name = 0;
12807 PL_savestack_ix = 0;
12808 PL_savestack_max = -1;
12809 PL_sig_pending = 0;
12811 Zero(&PL_debug_pad, 1, struct perl_debug_pad);
12812 # ifdef DEBUG_LEAKING_SCALARS
12813 PL_sv_serial = (((UV)my_perl >> 2) & 0xfff) * 1000000;
12815 #else /* !DEBUGGING */
12816 Zero(my_perl, 1, PerlInterpreter);
12817 #endif /* DEBUGGING */
12819 #ifdef PERL_IMPLICIT_SYS
12820 /* host pointers */
12822 PL_MemShared = ipMS;
12823 PL_MemParse = ipMP;
12830 #endif /* PERL_IMPLICIT_SYS */
12832 param->flags = flags;
12833 /* Nothing in the core code uses this, but we make it available to
12834 extensions (using mg_dup). */
12835 param->proto_perl = proto_perl;
12836 /* Likely nothing will use this, but it is initialised to be consistent
12837 with Perl_clone_params_new(). */
12838 param->new_perl = my_perl;
12839 param->unreferenced = NULL;
12841 INIT_TRACK_MEMPOOL(my_perl->Imemory_debug_header, my_perl);
12843 PL_body_arenas = NULL;
12844 Zero(&PL_body_roots, 1, PL_body_roots);
12847 PL_sv_objcount = 0;
12849 PL_sv_arenaroot = NULL;
12851 PL_debug = proto_perl->Idebug;
12853 PL_hash_seed = proto_perl->Ihash_seed;
12854 PL_rehash_seed = proto_perl->Irehash_seed;
12856 SvANY(&PL_sv_undef) = NULL;
12857 SvREFCNT(&PL_sv_undef) = (~(U32)0)/2;
12858 SvFLAGS(&PL_sv_undef) = SVf_READONLY|SVt_NULL;
12859 SvREFCNT(&PL_sv_no) = (~(U32)0)/2;
12860 SvFLAGS(&PL_sv_no) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
12861 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
12863 SvANY(&PL_sv_yes) = new_XPVNV();
12864 SvREFCNT(&PL_sv_yes) = (~(U32)0)/2;
12865 SvFLAGS(&PL_sv_yes) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
12866 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
12868 /* dbargs array probably holds garbage */
12871 PL_compiling = proto_perl->Icompiling;
12873 #ifdef PERL_DEBUG_READONLY_OPS
12878 /* pseudo environmental stuff */
12879 PL_origargc = proto_perl->Iorigargc;
12880 PL_origargv = proto_perl->Iorigargv;
12882 /* Set tainting stuff before PerlIO_debug can possibly get called */
12883 PL_tainting = proto_perl->Itainting;
12884 PL_taint_warn = proto_perl->Itaint_warn;
12886 PL_minus_c = proto_perl->Iminus_c;
12888 PL_localpatches = proto_perl->Ilocalpatches;
12889 PL_splitstr = proto_perl->Isplitstr;
12890 PL_minus_n = proto_perl->Iminus_n;
12891 PL_minus_p = proto_perl->Iminus_p;
12892 PL_minus_l = proto_perl->Iminus_l;
12893 PL_minus_a = proto_perl->Iminus_a;
12894 PL_minus_E = proto_perl->Iminus_E;
12895 PL_minus_F = proto_perl->Iminus_F;
12896 PL_doswitches = proto_perl->Idoswitches;
12897 PL_dowarn = proto_perl->Idowarn;
12898 PL_sawampersand = proto_perl->Isawampersand;
12899 PL_unsafe = proto_perl->Iunsafe;
12900 PL_perldb = proto_perl->Iperldb;
12901 PL_perl_destruct_level = proto_perl->Iperl_destruct_level;
12902 PL_exit_flags = proto_perl->Iexit_flags;
12904 /* XXX time(&PL_basetime) when asked for? */
12905 PL_basetime = proto_perl->Ibasetime;
12907 PL_maxsysfd = proto_perl->Imaxsysfd;
12908 PL_statusvalue = proto_perl->Istatusvalue;
12910 PL_statusvalue_vms = proto_perl->Istatusvalue_vms;
12912 PL_statusvalue_posix = proto_perl->Istatusvalue_posix;
12915 /* RE engine related */
12916 Zero(&PL_reg_state, 1, struct re_save_state);
12917 PL_reginterp_cnt = 0;
12918 PL_regmatch_slab = NULL;
12920 PL_sub_generation = proto_perl->Isub_generation;
12922 /* funky return mechanisms */
12923 PL_forkprocess = proto_perl->Iforkprocess;
12925 /* internal state */
12926 PL_maxo = proto_perl->Imaxo;
12928 PL_main_start = proto_perl->Imain_start;
12929 PL_eval_root = proto_perl->Ieval_root;
12930 PL_eval_start = proto_perl->Ieval_start;
12932 PL_filemode = proto_perl->Ifilemode;
12933 PL_lastfd = proto_perl->Ilastfd;
12934 PL_oldname = proto_perl->Ioldname; /* XXX not quite right */
12937 PL_gensym = proto_perl->Igensym;
12939 PL_laststatval = proto_perl->Ilaststatval;
12940 PL_laststype = proto_perl->Ilaststype;
12943 PL_profiledata = NULL;
12945 PL_generation = proto_perl->Igeneration;
12947 PL_in_clean_objs = proto_perl->Iin_clean_objs;
12948 PL_in_clean_all = proto_perl->Iin_clean_all;
12950 PL_uid = proto_perl->Iuid;
12951 PL_euid = proto_perl->Ieuid;
12952 PL_gid = proto_perl->Igid;
12953 PL_egid = proto_perl->Iegid;
12954 PL_nomemok = proto_perl->Inomemok;
12955 PL_an = proto_perl->Ian;
12956 PL_evalseq = proto_perl->Ievalseq;
12957 PL_origenviron = proto_perl->Iorigenviron; /* XXX not quite right */
12958 PL_origalen = proto_perl->Iorigalen;
12960 PL_sighandlerp = proto_perl->Isighandlerp;
12962 PL_runops = proto_perl->Irunops;
12964 PL_subline = proto_perl->Isubline;
12967 PL_cryptseen = proto_perl->Icryptseen;
12970 PL_hints = proto_perl->Ihints;
12972 PL_amagic_generation = proto_perl->Iamagic_generation;
12974 #ifdef USE_LOCALE_COLLATE
12975 PL_collation_ix = proto_perl->Icollation_ix;
12976 PL_collation_standard = proto_perl->Icollation_standard;
12977 PL_collxfrm_base = proto_perl->Icollxfrm_base;
12978 PL_collxfrm_mult = proto_perl->Icollxfrm_mult;
12979 #endif /* USE_LOCALE_COLLATE */
12981 #ifdef USE_LOCALE_NUMERIC
12982 PL_numeric_standard = proto_perl->Inumeric_standard;
12983 PL_numeric_local = proto_perl->Inumeric_local;
12984 #endif /* !USE_LOCALE_NUMERIC */
12986 /* Did the locale setup indicate UTF-8? */
12987 PL_utf8locale = proto_perl->Iutf8locale;
12988 /* Unicode features (see perlrun/-C) */
12989 PL_unicode = proto_perl->Iunicode;
12991 /* Pre-5.8 signals control */
12992 PL_signals = proto_perl->Isignals;
12994 /* times() ticks per second */
12995 PL_clocktick = proto_perl->Iclocktick;
12997 /* Recursion stopper for PerlIO_find_layer */
12998 PL_in_load_module = proto_perl->Iin_load_module;
13000 /* sort() routine */
13001 PL_sort_RealCmp = proto_perl->Isort_RealCmp;
13003 /* Not really needed/useful since the reenrant_retint is "volatile",
13004 * but do it for consistency's sake. */
13005 PL_reentrant_retint = proto_perl->Ireentrant_retint;
13007 /* Hooks to shared SVs and locks. */
13008 PL_sharehook = proto_perl->Isharehook;
13009 PL_lockhook = proto_perl->Ilockhook;
13010 PL_unlockhook = proto_perl->Iunlockhook;
13011 PL_threadhook = proto_perl->Ithreadhook;
13012 PL_destroyhook = proto_perl->Idestroyhook;
13013 PL_signalhook = proto_perl->Isignalhook;
13015 PL_globhook = proto_perl->Iglobhook;
13017 #ifdef THREADS_HAVE_PIDS
13018 PL_ppid = proto_perl->Ippid;
13022 PL_last_swash_hv = NULL; /* reinits on demand */
13023 PL_last_swash_klen = 0;
13024 PL_last_swash_key[0]= '\0';
13025 PL_last_swash_tmps = (U8*)NULL;
13026 PL_last_swash_slen = 0;
13028 PL_glob_index = proto_perl->Iglob_index;
13029 PL_srand_called = proto_perl->Isrand_called;
13031 if (flags & CLONEf_COPY_STACKS) {
13032 /* next allocation will be PL_tmps_stack[PL_tmps_ix+1] */
13033 PL_tmps_ix = proto_perl->Itmps_ix;
13034 PL_tmps_max = proto_perl->Itmps_max;
13035 PL_tmps_floor = proto_perl->Itmps_floor;
13037 /* next push_scope()/ENTER sets PL_scopestack[PL_scopestack_ix]
13038 * NOTE: unlike the others! */
13039 PL_scopestack_ix = proto_perl->Iscopestack_ix;
13040 PL_scopestack_max = proto_perl->Iscopestack_max;
13042 /* next SSPUSHFOO() sets PL_savestack[PL_savestack_ix]
13043 * NOTE: unlike the others! */
13044 PL_savestack_ix = proto_perl->Isavestack_ix;
13045 PL_savestack_max = proto_perl->Isavestack_max;
13048 PL_start_env = proto_perl->Istart_env; /* XXXXXX */
13049 PL_top_env = &PL_start_env;
13051 PL_op = proto_perl->Iop;
13054 PL_Xpv = (XPV*)NULL;
13055 my_perl->Ina = proto_perl->Ina;
13057 PL_statbuf = proto_perl->Istatbuf;
13058 PL_statcache = proto_perl->Istatcache;
13061 PL_timesbuf = proto_perl->Itimesbuf;
13064 PL_tainted = proto_perl->Itainted;
13065 PL_curpm = proto_perl->Icurpm; /* XXX No PMOP ref count */
13067 PL_chopset = proto_perl->Ichopset; /* XXX never deallocated */
13069 PL_restartjmpenv = proto_perl->Irestartjmpenv;
13070 PL_restartop = proto_perl->Irestartop;
13071 PL_in_eval = proto_perl->Iin_eval;
13072 PL_delaymagic = proto_perl->Idelaymagic;
13073 PL_phase = proto_perl->Iphase;
13074 PL_localizing = proto_perl->Ilocalizing;
13076 PL_hv_fetch_ent_mh = NULL;
13077 PL_modcount = proto_perl->Imodcount;
13078 PL_lastgotoprobe = NULL;
13079 PL_dumpindent = proto_perl->Idumpindent;
13081 PL_efloatbuf = NULL; /* reinits on demand */
13082 PL_efloatsize = 0; /* reinits on demand */
13086 PL_regdummy = proto_perl->Iregdummy;
13087 PL_colorset = 0; /* reinits PL_colors[] */
13088 /*PL_colors[6] = {0,0,0,0,0,0};*/
13090 /* Pluggable optimizer */
13091 PL_peepp = proto_perl->Ipeepp;
13092 PL_rpeepp = proto_perl->Irpeepp;
13093 /* op_free() hook */
13094 PL_opfreehook = proto_perl->Iopfreehook;
13096 #ifdef USE_REENTRANT_API
13097 /* XXX: things like -Dm will segfault here in perlio, but doing
13098 * PERL_SET_CONTEXT(proto_perl);
13099 * breaks too many other things
13101 Perl_reentrant_init(aTHX);
13104 /* create SV map for pointer relocation */
13105 PL_ptr_table = ptr_table_new();
13107 /* initialize these special pointers as early as possible */
13108 ptr_table_store(PL_ptr_table, &proto_perl->Isv_undef, &PL_sv_undef);
13110 SvANY(&PL_sv_no) = new_XPVNV();
13111 SvPV_set(&PL_sv_no, savepvn(PL_No, 0));
13112 SvCUR_set(&PL_sv_no, 0);
13113 SvLEN_set(&PL_sv_no, 1);
13114 SvIV_set(&PL_sv_no, 0);
13115 SvNV_set(&PL_sv_no, 0);
13116 ptr_table_store(PL_ptr_table, &proto_perl->Isv_no, &PL_sv_no);
13118 SvPV_set(&PL_sv_yes, savepvn(PL_Yes, 1));
13119 SvCUR_set(&PL_sv_yes, 1);
13120 SvLEN_set(&PL_sv_yes, 2);
13121 SvIV_set(&PL_sv_yes, 1);
13122 SvNV_set(&PL_sv_yes, 1);
13123 ptr_table_store(PL_ptr_table, &proto_perl->Isv_yes, &PL_sv_yes);
13125 /* create (a non-shared!) shared string table */
13126 PL_strtab = newHV();
13127 HvSHAREKEYS_off(PL_strtab);
13128 hv_ksplit(PL_strtab, HvTOTALKEYS(proto_perl->Istrtab));
13129 ptr_table_store(PL_ptr_table, proto_perl->Istrtab, PL_strtab);
13131 /* These two PVs will be free'd special way so must set them same way op.c does */
13132 PL_compiling.cop_stashpv = savesharedpv(PL_compiling.cop_stashpv);
13133 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_stashpv, PL_compiling.cop_stashpv);
13135 PL_compiling.cop_file = savesharedpv(PL_compiling.cop_file);
13136 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_file, PL_compiling.cop_file);
13138 ptr_table_store(PL_ptr_table, &proto_perl->Icompiling, &PL_compiling);
13139 PL_compiling.cop_warnings = DUP_WARNINGS(PL_compiling.cop_warnings);
13140 CopHINTHASH_set(&PL_compiling, cophh_copy(CopHINTHASH_get(&PL_compiling)));
13141 PL_curcop = (COP*)any_dup(proto_perl->Icurcop, proto_perl);
13143 param->stashes = newAV(); /* Setup array of objects to call clone on */
13144 /* This makes no difference to the implementation, as it always pushes
13145 and shifts pointers to other SVs without changing their reference
13146 count, with the array becoming empty before it is freed. However, it
13147 makes it conceptually clear what is going on, and will avoid some
13148 work inside av.c, filling slots between AvFILL() and AvMAX() with
13149 &PL_sv_undef, and SvREFCNT_dec()ing those. */
13150 AvREAL_off(param->stashes);
13152 if (!(flags & CLONEf_COPY_STACKS)) {
13153 param->unreferenced = newAV();
13156 #ifdef PERLIO_LAYERS
13157 /* Clone PerlIO tables as soon as we can handle general xx_dup() */
13158 PerlIO_clone(aTHX_ proto_perl, param);
13161 PL_envgv = gv_dup(proto_perl->Ienvgv, param);
13162 PL_incgv = gv_dup(proto_perl->Iincgv, param);
13163 PL_hintgv = gv_dup(proto_perl->Ihintgv, param);
13164 PL_origfilename = SAVEPV(proto_perl->Iorigfilename);
13165 PL_diehook = sv_dup_inc(proto_perl->Idiehook, param);
13166 PL_warnhook = sv_dup_inc(proto_perl->Iwarnhook, param);
13169 PL_patchlevel = sv_dup_inc(proto_perl->Ipatchlevel, param);
13170 PL_apiversion = sv_dup_inc(proto_perl->Iapiversion, param);
13171 PL_inplace = SAVEPV(proto_perl->Iinplace);
13172 PL_e_script = sv_dup_inc(proto_perl->Ie_script, param);
13174 /* magical thingies */
13175 PL_formfeed = sv_dup(proto_perl->Iformfeed, param);
13177 PL_encoding = sv_dup(proto_perl->Iencoding, param);
13179 sv_setpvs(PERL_DEBUG_PAD(0), ""); /* For regex debugging. */
13180 sv_setpvs(PERL_DEBUG_PAD(1), ""); /* ext/re needs these */
13181 sv_setpvs(PERL_DEBUG_PAD(2), ""); /* even without DEBUGGING. */
13184 /* Clone the regex array */
13185 /* ORANGE FIXME for plugins, probably in the SV dup code.
13186 newSViv(PTR2IV(CALLREGDUPE(
13187 INT2PTR(REGEXP *, SvIVX(regex)), param))))
13189 PL_regex_padav = av_dup_inc(proto_perl->Iregex_padav, param);
13190 PL_regex_pad = AvARRAY(PL_regex_padav);
13192 /* shortcuts to various I/O objects */
13193 PL_ofsgv = gv_dup_inc(proto_perl->Iofsgv, param);
13194 PL_stdingv = gv_dup(proto_perl->Istdingv, param);
13195 PL_stderrgv = gv_dup(proto_perl->Istderrgv, param);
13196 PL_defgv = gv_dup(proto_perl->Idefgv, param);
13197 PL_argvgv = gv_dup(proto_perl->Iargvgv, param);
13198 PL_argvoutgv = gv_dup(proto_perl->Iargvoutgv, param);
13199 PL_argvout_stack = av_dup_inc(proto_perl->Iargvout_stack, param);
13201 /* shortcuts to regexp stuff */
13202 PL_replgv = gv_dup(proto_perl->Ireplgv, param);
13204 /* shortcuts to misc objects */
13205 PL_errgv = gv_dup(proto_perl->Ierrgv, param);
13207 /* shortcuts to debugging objects */
13208 PL_DBgv = gv_dup(proto_perl->IDBgv, param);
13209 PL_DBline = gv_dup(proto_perl->IDBline, param);
13210 PL_DBsub = gv_dup(proto_perl->IDBsub, param);
13211 PL_DBsingle = sv_dup(proto_perl->IDBsingle, param);
13212 PL_DBtrace = sv_dup(proto_perl->IDBtrace, param);
13213 PL_DBsignal = sv_dup(proto_perl->IDBsignal, param);
13215 /* symbol tables */
13216 PL_defstash = hv_dup_inc(proto_perl->Idefstash, param);
13217 PL_curstash = hv_dup_inc(proto_perl->Icurstash, param);
13218 PL_debstash = hv_dup(proto_perl->Idebstash, param);
13219 PL_globalstash = hv_dup(proto_perl->Iglobalstash, param);
13220 PL_curstname = sv_dup_inc(proto_perl->Icurstname, param);
13222 PL_beginav = av_dup_inc(proto_perl->Ibeginav, param);
13223 PL_beginav_save = av_dup_inc(proto_perl->Ibeginav_save, param);
13224 PL_checkav_save = av_dup_inc(proto_perl->Icheckav_save, param);
13225 PL_unitcheckav = av_dup_inc(proto_perl->Iunitcheckav, param);
13226 PL_unitcheckav_save = av_dup_inc(proto_perl->Iunitcheckav_save, param);
13227 PL_endav = av_dup_inc(proto_perl->Iendav, param);
13228 PL_checkav = av_dup_inc(proto_perl->Icheckav, param);
13229 PL_initav = av_dup_inc(proto_perl->Iinitav, param);
13231 PL_isarev = hv_dup_inc(proto_perl->Iisarev, param);
13233 /* subprocess state */
13234 PL_fdpid = av_dup_inc(proto_perl->Ifdpid, param);
13236 if (proto_perl->Iop_mask)
13237 PL_op_mask = SAVEPVN(proto_perl->Iop_mask, PL_maxo);
13240 /* PL_asserting = proto_perl->Iasserting; */
13242 /* current interpreter roots */
13243 PL_main_cv = cv_dup_inc(proto_perl->Imain_cv, param);
13245 PL_main_root = OpREFCNT_inc(proto_perl->Imain_root);
13248 /* runtime control stuff */
13249 PL_curcopdb = (COP*)any_dup(proto_perl->Icurcopdb, proto_perl);
13251 PL_preambleav = av_dup_inc(proto_perl->Ipreambleav, param);
13253 PL_ors_sv = sv_dup_inc(proto_perl->Iors_sv, param);
13255 /* interpreter atexit processing */
13256 PL_exitlistlen = proto_perl->Iexitlistlen;
13257 if (PL_exitlistlen) {
13258 Newx(PL_exitlist, PL_exitlistlen, PerlExitListEntry);
13259 Copy(proto_perl->Iexitlist, PL_exitlist, PL_exitlistlen, PerlExitListEntry);
13262 PL_exitlist = (PerlExitListEntry*)NULL;
13264 PL_my_cxt_size = proto_perl->Imy_cxt_size;
13265 if (PL_my_cxt_size) {
13266 Newx(PL_my_cxt_list, PL_my_cxt_size, void *);
13267 Copy(proto_perl->Imy_cxt_list, PL_my_cxt_list, PL_my_cxt_size, void *);
13268 #ifdef PERL_GLOBAL_STRUCT_PRIVATE
13269 Newx(PL_my_cxt_keys, PL_my_cxt_size, const char *);
13270 Copy(proto_perl->Imy_cxt_keys, PL_my_cxt_keys, PL_my_cxt_size, char *);
13274 PL_my_cxt_list = (void**)NULL;
13275 #ifdef PERL_GLOBAL_STRUCT_PRIVATE
13276 PL_my_cxt_keys = (const char**)NULL;
13279 PL_modglobal = hv_dup_inc(proto_perl->Imodglobal, param);
13280 PL_custom_op_names = hv_dup_inc(proto_perl->Icustom_op_names,param);
13281 PL_custom_op_descs = hv_dup_inc(proto_perl->Icustom_op_descs,param);
13282 PL_custom_ops = hv_dup_inc(proto_perl->Icustom_ops, param);
13284 PL_compcv = cv_dup(proto_perl->Icompcv, param);
13286 PAD_CLONE_VARS(proto_perl, param);
13288 #ifdef HAVE_INTERP_INTERN
13289 sys_intern_dup(&proto_perl->Isys_intern, &PL_sys_intern);
13292 PL_DBcv = cv_dup(proto_perl->IDBcv, param);
13294 #ifdef PERL_USES_PL_PIDSTATUS
13295 PL_pidstatus = newHV(); /* XXX flag for cloning? */
13297 PL_osname = SAVEPV(proto_perl->Iosname);
13298 PL_parser = parser_dup(proto_perl->Iparser, param);
13300 /* XXX this only works if the saved cop has already been cloned */
13301 if (proto_perl->Iparser) {
13302 PL_parser->saved_curcop = (COP*)any_dup(
13303 proto_perl->Iparser->saved_curcop,
13307 PL_subname = sv_dup_inc(proto_perl->Isubname, param);
13309 #ifdef USE_LOCALE_COLLATE
13310 PL_collation_name = SAVEPV(proto_perl->Icollation_name);
13311 #endif /* USE_LOCALE_COLLATE */
13313 #ifdef USE_LOCALE_NUMERIC
13314 PL_numeric_name = SAVEPV(proto_perl->Inumeric_name);
13315 PL_numeric_radix_sv = sv_dup_inc(proto_perl->Inumeric_radix_sv, param);
13316 #endif /* !USE_LOCALE_NUMERIC */
13318 /* utf8 character classes */
13319 PL_utf8_alnum = sv_dup_inc(proto_perl->Iutf8_alnum, param);
13320 PL_utf8_alpha = sv_dup_inc(proto_perl->Iutf8_alpha, param);
13321 PL_utf8_space = sv_dup_inc(proto_perl->Iutf8_space, param);
13322 PL_utf8_graph = sv_dup_inc(proto_perl->Iutf8_graph, param);
13323 PL_utf8_digit = sv_dup_inc(proto_perl->Iutf8_digit, param);
13324 PL_utf8_upper = sv_dup_inc(proto_perl->Iutf8_upper, param);
13325 PL_utf8_lower = sv_dup_inc(proto_perl->Iutf8_lower, param);
13326 PL_utf8_print = sv_dup_inc(proto_perl->Iutf8_print, param);
13327 PL_utf8_punct = sv_dup_inc(proto_perl->Iutf8_punct, param);
13328 PL_utf8_xdigit = sv_dup_inc(proto_perl->Iutf8_xdigit, param);
13329 PL_utf8_mark = sv_dup_inc(proto_perl->Iutf8_mark, param);
13330 PL_utf8_X_begin = sv_dup_inc(proto_perl->Iutf8_X_begin, param);
13331 PL_utf8_X_extend = sv_dup_inc(proto_perl->Iutf8_X_extend, param);
13332 PL_utf8_X_prepend = sv_dup_inc(proto_perl->Iutf8_X_prepend, param);
13333 PL_utf8_X_non_hangul = sv_dup_inc(proto_perl->Iutf8_X_non_hangul, param);
13334 PL_utf8_X_L = sv_dup_inc(proto_perl->Iutf8_X_L, param);
13335 PL_utf8_X_LV = sv_dup_inc(proto_perl->Iutf8_X_LV, param);
13336 PL_utf8_X_LVT = sv_dup_inc(proto_perl->Iutf8_X_LVT, param);
13337 PL_utf8_X_T = sv_dup_inc(proto_perl->Iutf8_X_T, param);
13338 PL_utf8_X_V = sv_dup_inc(proto_perl->Iutf8_X_V, param);
13339 PL_utf8_X_LV_LVT_V = sv_dup_inc(proto_perl->Iutf8_X_LV_LVT_V, param);
13340 PL_utf8_toupper = sv_dup_inc(proto_perl->Iutf8_toupper, param);
13341 PL_utf8_totitle = sv_dup_inc(proto_perl->Iutf8_totitle, param);
13342 PL_utf8_tolower = sv_dup_inc(proto_perl->Iutf8_tolower, param);
13343 PL_utf8_tofold = sv_dup_inc(proto_perl->Iutf8_tofold, param);
13344 PL_utf8_idstart = sv_dup_inc(proto_perl->Iutf8_idstart, param);
13345 PL_utf8_xidstart = sv_dup_inc(proto_perl->Iutf8_xidstart, param);
13346 PL_utf8_perl_idstart = sv_dup_inc(proto_perl->Iutf8_perl_idstart, param);
13347 PL_utf8_idcont = sv_dup_inc(proto_perl->Iutf8_idcont, param);
13348 PL_utf8_xidcont = sv_dup_inc(proto_perl->Iutf8_xidcont, param);
13349 PL_utf8_foldable = sv_dup_inc(proto_perl->Iutf8_foldable, param);
13352 if (proto_perl->Ipsig_pend) {
13353 Newxz(PL_psig_pend, SIG_SIZE, int);
13356 PL_psig_pend = (int*)NULL;
13359 if (proto_perl->Ipsig_name) {
13360 Newx(PL_psig_name, 2 * SIG_SIZE, SV*);
13361 sv_dup_inc_multiple(proto_perl->Ipsig_name, PL_psig_name, 2 * SIG_SIZE,
13363 PL_psig_ptr = PL_psig_name + SIG_SIZE;
13366 PL_psig_ptr = (SV**)NULL;
13367 PL_psig_name = (SV**)NULL;
13370 if (flags & CLONEf_COPY_STACKS) {
13371 Newx(PL_tmps_stack, PL_tmps_max, SV*);
13372 sv_dup_inc_multiple(proto_perl->Itmps_stack, PL_tmps_stack,
13373 PL_tmps_ix+1, param);
13375 /* next PUSHMARK() sets *(PL_markstack_ptr+1) */
13376 i = proto_perl->Imarkstack_max - proto_perl->Imarkstack;
13377 Newxz(PL_markstack, i, I32);
13378 PL_markstack_max = PL_markstack + (proto_perl->Imarkstack_max
13379 - proto_perl->Imarkstack);
13380 PL_markstack_ptr = PL_markstack + (proto_perl->Imarkstack_ptr
13381 - proto_perl->Imarkstack);
13382 Copy(proto_perl->Imarkstack, PL_markstack,
13383 PL_markstack_ptr - PL_markstack + 1, I32);
13385 /* next push_scope()/ENTER sets PL_scopestack[PL_scopestack_ix]
13386 * NOTE: unlike the others! */
13387 Newxz(PL_scopestack, PL_scopestack_max, I32);
13388 Copy(proto_perl->Iscopestack, PL_scopestack, PL_scopestack_ix, I32);
13391 Newxz(PL_scopestack_name, PL_scopestack_max, const char *);
13392 Copy(proto_perl->Iscopestack_name, PL_scopestack_name, PL_scopestack_ix, const char *);
13394 /* NOTE: si_dup() looks at PL_markstack */
13395 PL_curstackinfo = si_dup(proto_perl->Icurstackinfo, param);
13397 /* PL_curstack = PL_curstackinfo->si_stack; */
13398 PL_curstack = av_dup(proto_perl->Icurstack, param);
13399 PL_mainstack = av_dup(proto_perl->Imainstack, param);
13401 /* next PUSHs() etc. set *(PL_stack_sp+1) */
13402 PL_stack_base = AvARRAY(PL_curstack);
13403 PL_stack_sp = PL_stack_base + (proto_perl->Istack_sp
13404 - proto_perl->Istack_base);
13405 PL_stack_max = PL_stack_base + AvMAX(PL_curstack);
13407 /*Newxz(PL_savestack, PL_savestack_max, ANY);*/
13408 PL_savestack = ss_dup(proto_perl, param);
13412 ENTER; /* perl_destruct() wants to LEAVE; */
13415 PL_statgv = gv_dup(proto_perl->Istatgv, param);
13416 PL_statname = sv_dup_inc(proto_perl->Istatname, param);
13418 PL_rs = sv_dup_inc(proto_perl->Irs, param);
13419 PL_last_in_gv = gv_dup(proto_perl->Ilast_in_gv, param);
13420 PL_defoutgv = gv_dup_inc(proto_perl->Idefoutgv, param);
13421 PL_toptarget = sv_dup_inc(proto_perl->Itoptarget, param);
13422 PL_bodytarget = sv_dup_inc(proto_perl->Ibodytarget, param);
13423 PL_formtarget = sv_dup(proto_perl->Iformtarget, param);
13425 PL_errors = sv_dup_inc(proto_perl->Ierrors, param);
13427 PL_sortcop = (OP*)any_dup(proto_perl->Isortcop, proto_perl);
13428 PL_sortstash = hv_dup(proto_perl->Isortstash, param);
13429 PL_firstgv = gv_dup(proto_perl->Ifirstgv, param);
13430 PL_secondgv = gv_dup(proto_perl->Isecondgv, param);
13432 PL_stashcache = newHV();
13434 PL_watchaddr = (char **) ptr_table_fetch(PL_ptr_table,
13435 proto_perl->Iwatchaddr);
13436 PL_watchok = PL_watchaddr ? * PL_watchaddr : NULL;
13437 if (PL_debug && PL_watchaddr) {
13438 PerlIO_printf(Perl_debug_log,
13439 "WATCHING: %"UVxf" cloned as %"UVxf" with value %"UVxf"\n",
13440 PTR2UV(proto_perl->Iwatchaddr), PTR2UV(PL_watchaddr),
13441 PTR2UV(PL_watchok));
13444 PL_registered_mros = hv_dup_inc(proto_perl->Iregistered_mros, param);
13445 PL_blockhooks = av_dup_inc(proto_perl->Iblockhooks, param);
13446 PL_utf8_foldclosures = hv_dup_inc(proto_perl->Iutf8_foldclosures, param);
13448 /* Call the ->CLONE method, if it exists, for each of the stashes
13449 identified by sv_dup() above.
13451 while(av_len(param->stashes) != -1) {
13452 HV* const stash = MUTABLE_HV(av_shift(param->stashes));
13453 GV* const cloner = gv_fetchmethod_autoload(stash, "CLONE", 0);
13454 if (cloner && GvCV(cloner)) {
13459 mXPUSHs(newSVhek(HvNAME_HEK(stash)));
13461 call_sv(MUTABLE_SV(GvCV(cloner)), G_DISCARD);
13467 if (!(flags & CLONEf_KEEP_PTR_TABLE)) {
13468 ptr_table_free(PL_ptr_table);
13469 PL_ptr_table = NULL;
13472 if (!(flags & CLONEf_COPY_STACKS)) {
13473 unreferenced_to_tmp_stack(param->unreferenced);
13476 SvREFCNT_dec(param->stashes);
13478 /* orphaned? eg threads->new inside BEGIN or use */
13479 if (PL_compcv && ! SvREFCNT(PL_compcv)) {
13480 SvREFCNT_inc_simple_void(PL_compcv);
13481 SAVEFREESV(PL_compcv);
13488 S_unreferenced_to_tmp_stack(pTHX_ AV *const unreferenced)
13490 PERL_ARGS_ASSERT_UNREFERENCED_TO_TMP_STACK;
13492 if (AvFILLp(unreferenced) > -1) {
13493 SV **svp = AvARRAY(unreferenced);
13494 SV **const last = svp + AvFILLp(unreferenced);
13498 if (SvREFCNT(*svp) == 1)
13500 } while (++svp <= last);
13502 EXTEND_MORTAL(count);
13503 svp = AvARRAY(unreferenced);
13506 if (SvREFCNT(*svp) == 1) {
13507 /* Our reference is the only one to this SV. This means that
13508 in this thread, the scalar effectively has a 0 reference.
13509 That doesn't work (cleanup never happens), so donate our
13510 reference to it onto the save stack. */
13511 PL_tmps_stack[++PL_tmps_ix] = *svp;
13513 /* As an optimisation, because we are already walking the
13514 entire array, instead of above doing either
13515 SvREFCNT_inc(*svp) or *svp = &PL_sv_undef, we can instead
13516 release our reference to the scalar, so that at the end of
13517 the array owns zero references to the scalars it happens to
13518 point to. We are effectively converting the array from
13519 AvREAL() on to AvREAL() off. This saves the av_clear()
13520 (triggered by the SvREFCNT_dec(unreferenced) below) from
13521 walking the array a second time. */
13522 SvREFCNT_dec(*svp);
13525 } while (++svp <= last);
13526 AvREAL_off(unreferenced);
13528 SvREFCNT_dec(unreferenced);
13532 Perl_clone_params_del(CLONE_PARAMS *param)
13534 /* This seemingly funky ordering keeps the build with PERL_GLOBAL_STRUCT
13536 PerlInterpreter *const to = param->new_perl;
13538 PerlInterpreter *const was = PERL_GET_THX;
13540 PERL_ARGS_ASSERT_CLONE_PARAMS_DEL;
13546 SvREFCNT_dec(param->stashes);
13547 if (param->unreferenced)
13548 unreferenced_to_tmp_stack(param->unreferenced);
13558 Perl_clone_params_new(PerlInterpreter *const from, PerlInterpreter *const to)
13561 /* Need to play this game, as newAV() can call safesysmalloc(), and that
13562 does a dTHX; to get the context from thread local storage.
13563 FIXME - under PERL_CORE Newx(), Safefree() and friends should expand to
13564 a version that passes in my_perl. */
13565 PerlInterpreter *const was = PERL_GET_THX;
13566 CLONE_PARAMS *param;
13568 PERL_ARGS_ASSERT_CLONE_PARAMS_NEW;
13574 /* Given that we've set the context, we can do this unshared. */
13575 Newx(param, 1, CLONE_PARAMS);
13578 param->proto_perl = from;
13579 param->new_perl = to;
13580 param->stashes = (AV *)Perl_newSV_type(to, SVt_PVAV);
13581 AvREAL_off(param->stashes);
13582 param->unreferenced = (AV *)Perl_newSV_type(to, SVt_PVAV);
13590 #endif /* USE_ITHREADS */
13593 =head1 Unicode Support
13595 =for apidoc sv_recode_to_utf8
13597 The encoding is assumed to be an Encode object, on entry the PV
13598 of the sv is assumed to be octets in that encoding, and the sv
13599 will be converted into Unicode (and UTF-8).
13601 If the sv already is UTF-8 (or if it is not POK), or if the encoding
13602 is not a reference, nothing is done to the sv. If the encoding is not
13603 an C<Encode::XS> Encoding object, bad things will happen.
13604 (See F<lib/encoding.pm> and L<Encode>).
13606 The PV of the sv is returned.
13611 Perl_sv_recode_to_utf8(pTHX_ SV *sv, SV *encoding)
13615 PERL_ARGS_ASSERT_SV_RECODE_TO_UTF8;
13617 if (SvPOK(sv) && !SvUTF8(sv) && !IN_BYTES && SvROK(encoding)) {
13631 Passing sv_yes is wrong - it needs to be or'ed set of constants
13632 for Encode::XS, while UTf-8 decode (currently) assumes a true value means
13633 remove converted chars from source.
13635 Both will default the value - let them.
13637 XPUSHs(&PL_sv_yes);
13640 call_method("decode", G_SCALAR);
13644 s = SvPV_const(uni, len);
13645 if (s != SvPVX_const(sv)) {
13646 SvGROW(sv, len + 1);
13647 Move(s, SvPVX(sv), len + 1, char);
13648 SvCUR_set(sv, len);
13652 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
13653 /* clear pos and any utf8 cache */
13654 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
13657 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
13658 magic_setutf8(sv,mg); /* clear UTF8 cache */
13663 return SvPOKp(sv) ? SvPVX(sv) : NULL;
13667 =for apidoc sv_cat_decode
13669 The encoding is assumed to be an Encode object, the PV of the ssv is
13670 assumed to be octets in that encoding and decoding the input starts
13671 from the position which (PV + *offset) pointed to. The dsv will be
13672 concatenated the decoded UTF-8 string from ssv. Decoding will terminate
13673 when the string tstr appears in decoding output or the input ends on
13674 the PV of the ssv. The value which the offset points will be modified
13675 to the last input position on the ssv.
13677 Returns TRUE if the terminator was found, else returns FALSE.
13682 Perl_sv_cat_decode(pTHX_ SV *dsv, SV *encoding,
13683 SV *ssv, int *offset, char *tstr, int tlen)
13688 PERL_ARGS_ASSERT_SV_CAT_DECODE;
13690 if (SvPOK(ssv) && SvPOK(dsv) && SvROK(encoding) && offset) {
13701 offsv = newSViv(*offset);
13703 mXPUSHp(tstr, tlen);
13705 call_method("cat_decode", G_SCALAR);
13707 ret = SvTRUE(TOPs);
13708 *offset = SvIV(offsv);
13714 Perl_croak(aTHX_ "Invalid argument to sv_cat_decode");
13719 /* ---------------------------------------------------------------------
13721 * support functions for report_uninit()
13724 /* the maxiumum size of array or hash where we will scan looking
13725 * for the undefined element that triggered the warning */
13727 #define FUV_MAX_SEARCH_SIZE 1000
13729 /* Look for an entry in the hash whose value has the same SV as val;
13730 * If so, return a mortal copy of the key. */
13733 S_find_hash_subscript(pTHX_ const HV *const hv, const SV *const val)
13736 register HE **array;
13739 PERL_ARGS_ASSERT_FIND_HASH_SUBSCRIPT;
13741 if (!hv || SvMAGICAL(hv) || !HvARRAY(hv) ||
13742 (HvTOTALKEYS(hv) > FUV_MAX_SEARCH_SIZE))
13745 array = HvARRAY(hv);
13747 for (i=HvMAX(hv); i>0; i--) {
13748 register HE *entry;
13749 for (entry = array[i]; entry; entry = HeNEXT(entry)) {
13750 if (HeVAL(entry) != val)
13752 if ( HeVAL(entry) == &PL_sv_undef ||
13753 HeVAL(entry) == &PL_sv_placeholder)
13757 if (HeKLEN(entry) == HEf_SVKEY)
13758 return sv_mortalcopy(HeKEY_sv(entry));
13759 return sv_2mortal(newSVhek(HeKEY_hek(entry)));
13765 /* Look for an entry in the array whose value has the same SV as val;
13766 * If so, return the index, otherwise return -1. */
13769 S_find_array_subscript(pTHX_ const AV *const av, const SV *const val)
13773 PERL_ARGS_ASSERT_FIND_ARRAY_SUBSCRIPT;
13775 if (!av || SvMAGICAL(av) || !AvARRAY(av) ||
13776 (AvFILLp(av) > FUV_MAX_SEARCH_SIZE))
13779 if (val != &PL_sv_undef) {
13780 SV ** const svp = AvARRAY(av);
13783 for (i=AvFILLp(av); i>=0; i--)
13790 /* S_varname(): return the name of a variable, optionally with a subscript.
13791 * If gv is non-zero, use the name of that global, along with gvtype (one
13792 * of "$", "@", "%"); otherwise use the name of the lexical at pad offset
13793 * targ. Depending on the value of the subscript_type flag, return:
13796 #define FUV_SUBSCRIPT_NONE 1 /* "@foo" */
13797 #define FUV_SUBSCRIPT_ARRAY 2 /* "$foo[aindex]" */
13798 #define FUV_SUBSCRIPT_HASH 3 /* "$foo{keyname}" */
13799 #define FUV_SUBSCRIPT_WITHIN 4 /* "within @foo" */
13802 S_varname(pTHX_ const GV *const gv, const char gvtype, PADOFFSET targ,
13803 const SV *const keyname, I32 aindex, int subscript_type)
13806 SV * const name = sv_newmortal();
13809 buffer[0] = gvtype;
13812 /* as gv_fullname4(), but add literal '^' for $^FOO names */
13814 gv_fullname4(name, gv, buffer, 0);
13816 if ((unsigned int)SvPVX(name)[1] <= 26) {
13818 buffer[1] = SvPVX(name)[1] + 'A' - 1;
13820 /* Swap the 1 unprintable control character for the 2 byte pretty
13821 version - ie substr($name, 1, 1) = $buffer; */
13822 sv_insert(name, 1, 1, buffer, 2);
13826 CV * const cv = find_runcv(NULL);
13830 if (!cv || !CvPADLIST(cv))
13832 av = MUTABLE_AV((*av_fetch(CvPADLIST(cv), 0, FALSE)));
13833 sv = *av_fetch(av, targ, FALSE);
13834 sv_setsv(name, sv);
13837 if (subscript_type == FUV_SUBSCRIPT_HASH) {
13838 SV * const sv = newSV(0);
13839 *SvPVX(name) = '$';
13840 Perl_sv_catpvf(aTHX_ name, "{%s}",
13841 pv_display(sv,SvPVX_const(keyname), SvCUR(keyname), 0, 32));
13844 else if (subscript_type == FUV_SUBSCRIPT_ARRAY) {
13845 *SvPVX(name) = '$';
13846 Perl_sv_catpvf(aTHX_ name, "[%"IVdf"]", (IV)aindex);
13848 else if (subscript_type == FUV_SUBSCRIPT_WITHIN) {
13849 /* We know that name has no magic, so can use 0 instead of SV_GMAGIC */
13850 Perl_sv_insert_flags(aTHX_ name, 0, 0, STR_WITH_LEN("within "), 0);
13858 =for apidoc find_uninit_var
13860 Find the name of the undefined variable (if any) that caused the operator o
13861 to issue a "Use of uninitialized value" warning.
13862 If match is true, only return a name if it's value matches uninit_sv.
13863 So roughly speaking, if a unary operator (such as OP_COS) generates a
13864 warning, then following the direct child of the op may yield an
13865 OP_PADSV or OP_GV that gives the name of the undefined variable. On the
13866 other hand, with OP_ADD there are two branches to follow, so we only print
13867 the variable name if we get an exact match.
13869 The name is returned as a mortal SV.
13871 Assumes that PL_op is the op that originally triggered the error, and that
13872 PL_comppad/PL_curpad points to the currently executing pad.
13878 S_find_uninit_var(pTHX_ const OP *const obase, const SV *const uninit_sv,
13884 const OP *o, *o2, *kid;
13886 if (!obase || (match && (!uninit_sv || uninit_sv == &PL_sv_undef ||
13887 uninit_sv == &PL_sv_placeholder)))
13890 switch (obase->op_type) {
13897 const bool pad = (obase->op_type == OP_PADAV || obase->op_type == OP_PADHV);
13898 const bool hash = (obase->op_type == OP_PADHV || obase->op_type == OP_RV2HV);
13901 int subscript_type = FUV_SUBSCRIPT_WITHIN;
13903 if (pad) { /* @lex, %lex */
13904 sv = PAD_SVl(obase->op_targ);
13908 if (cUNOPx(obase)->op_first->op_type == OP_GV) {
13909 /* @global, %global */
13910 gv = cGVOPx_gv(cUNOPx(obase)->op_first);
13913 sv = hash ? MUTABLE_SV(GvHV(gv)): MUTABLE_SV(GvAV(gv));
13915 else /* @{expr}, %{expr} */
13916 return find_uninit_var(cUNOPx(obase)->op_first,
13920 /* attempt to find a match within the aggregate */
13922 keysv = find_hash_subscript((const HV*)sv, uninit_sv);
13924 subscript_type = FUV_SUBSCRIPT_HASH;
13927 index = find_array_subscript((const AV *)sv, uninit_sv);
13929 subscript_type = FUV_SUBSCRIPT_ARRAY;
13932 if (match && subscript_type == FUV_SUBSCRIPT_WITHIN)
13935 return varname(gv, hash ? '%' : '@', obase->op_targ,
13936 keysv, index, subscript_type);
13940 if (cUNOPx(obase)->op_first->op_type == OP_GV) {
13942 gv = cGVOPx_gv(cUNOPx(obase)->op_first);
13943 if (!gv || !GvSTASH(gv))
13945 if (match && (GvSV(gv) != uninit_sv))
13947 return varname(gv, '$', 0, NULL, 0, FUV_SUBSCRIPT_NONE);
13950 return find_uninit_var(cUNOPx(obase)->op_first, uninit_sv, 1);
13953 if (match && PAD_SVl(obase->op_targ) != uninit_sv)
13955 return varname(NULL, '$', obase->op_targ,
13956 NULL, 0, FUV_SUBSCRIPT_NONE);
13959 gv = cGVOPx_gv(obase);
13960 if (!gv || (match && GvSV(gv) != uninit_sv) || !GvSTASH(gv))
13962 return varname(gv, '$', 0, NULL, 0, FUV_SUBSCRIPT_NONE);
13964 case OP_AELEMFAST_LEX:
13967 AV *av = MUTABLE_AV(PAD_SV(obase->op_targ));
13968 if (!av || SvRMAGICAL(av))
13970 svp = av_fetch(av, (I32)obase->op_private, FALSE);
13971 if (!svp || *svp != uninit_sv)
13974 return varname(NULL, '$', obase->op_targ,
13975 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
13978 gv = cGVOPx_gv(obase);
13983 AV *const av = GvAV(gv);
13984 if (!av || SvRMAGICAL(av))
13986 svp = av_fetch(av, (I32)obase->op_private, FALSE);
13987 if (!svp || *svp != uninit_sv)
13990 return varname(gv, '$', 0,
13991 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
13996 o = cUNOPx(obase)->op_first;
13997 if (!o || o->op_type != OP_NULL ||
13998 ! (o->op_targ == OP_AELEM || o->op_targ == OP_HELEM))
14000 return find_uninit_var(cBINOPo->op_last, uninit_sv, match);
14005 bool negate = FALSE;
14007 if (PL_op == obase)
14008 /* $a[uninit_expr] or $h{uninit_expr} */
14009 return find_uninit_var(cBINOPx(obase)->op_last, uninit_sv, match);
14012 o = cBINOPx(obase)->op_first;
14013 kid = cBINOPx(obase)->op_last;
14015 /* get the av or hv, and optionally the gv */
14017 if (o->op_type == OP_PADAV || o->op_type == OP_PADHV) {
14018 sv = PAD_SV(o->op_targ);
14020 else if ((o->op_type == OP_RV2AV || o->op_type == OP_RV2HV)
14021 && cUNOPo->op_first->op_type == OP_GV)
14023 gv = cGVOPx_gv(cUNOPo->op_first);
14027 == OP_RV2HV ? MUTABLE_SV(GvHV(gv)) : MUTABLE_SV(GvAV(gv));
14032 if (kid && kid->op_type == OP_NEGATE) {
14034 kid = cUNOPx(kid)->op_first;
14037 if (kid && kid->op_type == OP_CONST && SvOK(cSVOPx_sv(kid))) {
14038 /* index is constant */
14041 kidsv = sv_2mortal(newSVpvs("-"));
14042 sv_catsv(kidsv, cSVOPx_sv(kid));
14045 kidsv = cSVOPx_sv(kid);
14049 if (obase->op_type == OP_HELEM) {
14050 HE* he = hv_fetch_ent(MUTABLE_HV(sv), kidsv, 0, 0);
14051 if (!he || HeVAL(he) != uninit_sv)
14055 SV * const * const svp = av_fetch(MUTABLE_AV(sv),
14056 negate ? - SvIV(cSVOPx_sv(kid)) : SvIV(cSVOPx_sv(kid)),
14058 if (!svp || *svp != uninit_sv)
14062 if (obase->op_type == OP_HELEM)
14063 return varname(gv, '%', o->op_targ,
14064 kidsv, 0, FUV_SUBSCRIPT_HASH);
14066 return varname(gv, '@', o->op_targ, NULL,
14067 negate ? - SvIV(cSVOPx_sv(kid)) : SvIV(cSVOPx_sv(kid)),
14068 FUV_SUBSCRIPT_ARRAY);
14071 /* index is an expression;
14072 * attempt to find a match within the aggregate */
14073 if (obase->op_type == OP_HELEM) {
14074 SV * const keysv = find_hash_subscript((const HV*)sv, uninit_sv);
14076 return varname(gv, '%', o->op_targ,
14077 keysv, 0, FUV_SUBSCRIPT_HASH);
14081 = find_array_subscript((const AV *)sv, uninit_sv);
14083 return varname(gv, '@', o->op_targ,
14084 NULL, index, FUV_SUBSCRIPT_ARRAY);
14089 (o->op_type == OP_PADAV || o->op_type == OP_RV2AV)
14091 o->op_targ, NULL, 0, FUV_SUBSCRIPT_WITHIN);
14097 /* only examine RHS */
14098 return find_uninit_var(cBINOPx(obase)->op_first, uninit_sv, match);
14101 o = cUNOPx(obase)->op_first;
14102 if (o->op_type == OP_PUSHMARK)
14105 if (!o->op_sibling) {
14106 /* one-arg version of open is highly magical */
14108 if (o->op_type == OP_GV) { /* open FOO; */
14110 if (match && GvSV(gv) != uninit_sv)
14112 return varname(gv, '$', 0,
14113 NULL, 0, FUV_SUBSCRIPT_NONE);
14115 /* other possibilities not handled are:
14116 * open $x; or open my $x; should return '${*$x}'
14117 * open expr; should return '$'.expr ideally
14123 /* ops where $_ may be an implicit arg */
14127 if ( !(obase->op_flags & OPf_STACKED)) {
14128 if (uninit_sv == ((obase->op_private & OPpTARGET_MY)
14129 ? PAD_SVl(obase->op_targ)
14132 sv = sv_newmortal();
14133 sv_setpvs(sv, "$_");
14142 match = 1; /* print etc can return undef on defined args */
14143 /* skip filehandle as it can't produce 'undef' warning */
14144 o = cUNOPx(obase)->op_first;
14145 if ((obase->op_flags & OPf_STACKED) && o->op_type == OP_PUSHMARK)
14146 o = o->op_sibling->op_sibling;
14150 case OP_ENTEREVAL: /* could be eval $undef or $x='$undef'; eval $x */
14151 case OP_CUSTOM: /* XS or custom code could trigger random warnings */
14153 /* the following ops are capable of returning PL_sv_undef even for
14154 * defined arg(s) */
14173 case OP_GETPEERNAME:
14221 case OP_SMARTMATCH:
14230 /* XXX tmp hack: these two may call an XS sub, and currently
14231 XS subs don't have a SUB entry on the context stack, so CV and
14232 pad determination goes wrong, and BAD things happen. So, just
14233 don't try to determine the value under those circumstances.
14234 Need a better fix at dome point. DAPM 11/2007 */
14240 GV * const gv = gv_fetchpvs(".", GV_NOTQUAL, SVt_PV);
14241 if (gv && GvSV(gv) == uninit_sv)
14242 return newSVpvs_flags("$.", SVs_TEMP);
14247 /* def-ness of rval pos() is independent of the def-ness of its arg */
14248 if ( !(obase->op_flags & OPf_MOD))
14253 if (SvROK(PL_rs) && uninit_sv == SvRV(PL_rs))
14254 return newSVpvs_flags("${$/}", SVs_TEMP);
14259 if (!(obase->op_flags & OPf_KIDS))
14261 o = cUNOPx(obase)->op_first;
14267 /* if all except one arg are constant, or have no side-effects,
14268 * or are optimized away, then it's unambiguous */
14270 for (kid=o; kid; kid = kid->op_sibling) {
14272 const OPCODE type = kid->op_type;
14273 if ( (type == OP_CONST && SvOK(cSVOPx_sv(kid)))
14274 || (type == OP_NULL && ! (kid->op_flags & OPf_KIDS))
14275 || (type == OP_PUSHMARK)
14277 /* @$a and %$a, but not @a or %a */
14278 (type == OP_RV2AV || type == OP_RV2HV)
14279 && cUNOPx(kid)->op_first
14280 && cUNOPx(kid)->op_first->op_type != OP_GV
14285 if (o2) { /* more than one found */
14292 return find_uninit_var(o2, uninit_sv, match);
14294 /* scan all args */
14296 sv = find_uninit_var(o, uninit_sv, 1);
14308 =for apidoc report_uninit
14310 Print appropriate "Use of uninitialized variable" warning
14316 Perl_report_uninit(pTHX_ const SV *uninit_sv)
14320 SV* varname = NULL;
14321 if (uninit_sv && PL_curpad) {
14322 varname = find_uninit_var(PL_op, uninit_sv,0);
14324 sv_insert(varname, 0, 0, " ", 1);
14326 /* diag_listed_as: Use of uninitialized value%s */
14327 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit_sv,
14328 SVfARG(varname ? varname : &PL_sv_no),
14329 " in ", OP_DESC(PL_op));
14332 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
14338 * c-indentation-style: bsd
14339 * c-basic-offset: 4
14340 * indent-tabs-mode: t
14343 * ex: set ts=8 sts=4 sw=4 noet: