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, 0);
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.
1794 Perl_looks_like_number(pTHX_ SV *const sv)
1796 register const char *sbegin;
1799 PERL_ARGS_ASSERT_LOOKS_LIKE_NUMBER;
1802 sbegin = SvPVX_const(sv);
1805 else if (SvPOKp(sv))
1806 sbegin = SvPV_const(sv, len);
1808 return SvFLAGS(sv) & (SVf_NOK|SVp_NOK|SVf_IOK|SVp_IOK);
1809 return grok_number(sbegin, len, NULL);
1813 S_glob_2number(pTHX_ GV * const gv)
1815 const U32 wasfake = SvFLAGS(gv) & SVf_FAKE;
1816 SV *const buffer = sv_newmortal();
1818 PERL_ARGS_ASSERT_GLOB_2NUMBER;
1820 /* FAKE globs can get coerced, so need to turn this off temporarily if it
1823 gv_efullname3(buffer, gv, "*");
1824 SvFLAGS(gv) |= wasfake;
1826 /* We know that all GVs stringify to something that is not-a-number,
1827 so no need to test that. */
1828 if (ckWARN(WARN_NUMERIC))
1829 not_a_number(buffer);
1830 /* We just want something true to return, so that S_sv_2iuv_common
1831 can tail call us and return true. */
1835 /* Actually, ISO C leaves conversion of UV to IV undefined, but
1836 until proven guilty, assume that things are not that bad... */
1841 As 64 bit platforms often have an NV that doesn't preserve all bits of
1842 an IV (an assumption perl has been based on to date) it becomes necessary
1843 to remove the assumption that the NV always carries enough precision to
1844 recreate the IV whenever needed, and that the NV is the canonical form.
1845 Instead, IV/UV and NV need to be given equal rights. So as to not lose
1846 precision as a side effect of conversion (which would lead to insanity
1847 and the dragon(s) in t/op/numconvert.t getting very angry) the intent is
1848 1) to distinguish between IV/UV/NV slots that have cached a valid
1849 conversion where precision was lost and IV/UV/NV slots that have a
1850 valid conversion which has lost no precision
1851 2) to ensure that if a numeric conversion to one form is requested that
1852 would lose precision, the precise conversion (or differently
1853 imprecise conversion) is also performed and cached, to prevent
1854 requests for different numeric formats on the same SV causing
1855 lossy conversion chains. (lossless conversion chains are perfectly
1860 SvIOKp is true if the IV slot contains a valid value
1861 SvIOK is true only if the IV value is accurate (UV if SvIOK_UV true)
1862 SvNOKp is true if the NV slot contains a valid value
1863 SvNOK is true only if the NV value is accurate
1866 while converting from PV to NV, check to see if converting that NV to an
1867 IV(or UV) would lose accuracy over a direct conversion from PV to
1868 IV(or UV). If it would, cache both conversions, return NV, but mark
1869 SV as IOK NOKp (ie not NOK).
1871 While converting from PV to IV, check to see if converting that IV to an
1872 NV would lose accuracy over a direct conversion from PV to NV. If it
1873 would, cache both conversions, flag similarly.
1875 Before, the SV value "3.2" could become NV=3.2 IV=3 NOK, IOK quite
1876 correctly because if IV & NV were set NV *always* overruled.
1877 Now, "3.2" will become NV=3.2 IV=3 NOK, IOKp, because the flag's meaning
1878 changes - now IV and NV together means that the two are interchangeable:
1879 SvIVX == (IV) SvNVX && SvNVX == (NV) SvIVX;
1881 The benefit of this is that operations such as pp_add know that if
1882 SvIOK is true for both left and right operands, then integer addition
1883 can be used instead of floating point (for cases where the result won't
1884 overflow). Before, floating point was always used, which could lead to
1885 loss of precision compared with integer addition.
1887 * making IV and NV equal status should make maths accurate on 64 bit
1889 * may speed up maths somewhat if pp_add and friends start to use
1890 integers when possible instead of fp. (Hopefully the overhead in
1891 looking for SvIOK and checking for overflow will not outweigh the
1892 fp to integer speedup)
1893 * will slow down integer operations (callers of SvIV) on "inaccurate"
1894 values, as the change from SvIOK to SvIOKp will cause a call into
1895 sv_2iv each time rather than a macro access direct to the IV slot
1896 * should speed up number->string conversion on integers as IV is
1897 favoured when IV and NV are equally accurate
1899 ####################################################################
1900 You had better be using SvIOK_notUV if you want an IV for arithmetic:
1901 SvIOK is true if (IV or UV), so you might be getting (IV)SvUV.
1902 On the other hand, SvUOK is true iff UV.
1903 ####################################################################
1905 Your mileage will vary depending your CPU's relative fp to integer
1909 #ifndef NV_PRESERVES_UV
1910 # define IS_NUMBER_UNDERFLOW_IV 1
1911 # define IS_NUMBER_UNDERFLOW_UV 2
1912 # define IS_NUMBER_IV_AND_UV 2
1913 # define IS_NUMBER_OVERFLOW_IV 4
1914 # define IS_NUMBER_OVERFLOW_UV 5
1916 /* sv_2iuv_non_preserve(): private routine for use by sv_2iv() and sv_2uv() */
1918 /* For sv_2nv these three cases are "SvNOK and don't bother casting" */
1920 S_sv_2iuv_non_preserve(pTHX_ register SV *const sv
1928 PERL_ARGS_ASSERT_SV_2IUV_NON_PRESERVE;
1930 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));
1931 if (SvNVX(sv) < (NV)IV_MIN) {
1932 (void)SvIOKp_on(sv);
1934 SvIV_set(sv, IV_MIN);
1935 return IS_NUMBER_UNDERFLOW_IV;
1937 if (SvNVX(sv) > (NV)UV_MAX) {
1938 (void)SvIOKp_on(sv);
1941 SvUV_set(sv, UV_MAX);
1942 return IS_NUMBER_OVERFLOW_UV;
1944 (void)SvIOKp_on(sv);
1946 /* Can't use strtol etc to convert this string. (See truth table in
1948 if (SvNVX(sv) <= (UV)IV_MAX) {
1949 SvIV_set(sv, I_V(SvNVX(sv)));
1950 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
1951 SvIOK_on(sv); /* Integer is precise. NOK, IOK */
1953 /* Integer is imprecise. NOK, IOKp */
1955 return SvNVX(sv) < 0 ? IS_NUMBER_UNDERFLOW_UV : IS_NUMBER_IV_AND_UV;
1958 SvUV_set(sv, U_V(SvNVX(sv)));
1959 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
1960 if (SvUVX(sv) == UV_MAX) {
1961 /* As we know that NVs don't preserve UVs, UV_MAX cannot
1962 possibly be preserved by NV. Hence, it must be overflow.
1964 return IS_NUMBER_OVERFLOW_UV;
1966 SvIOK_on(sv); /* Integer is precise. NOK, UOK */
1968 /* Integer is imprecise. NOK, IOKp */
1970 return IS_NUMBER_OVERFLOW_IV;
1972 #endif /* !NV_PRESERVES_UV*/
1975 S_sv_2iuv_common(pTHX_ SV *const sv)
1979 PERL_ARGS_ASSERT_SV_2IUV_COMMON;
1982 /* erm. not sure. *should* never get NOKp (without NOK) from sv_2nv
1983 * without also getting a cached IV/UV from it at the same time
1984 * (ie PV->NV conversion should detect loss of accuracy and cache
1985 * IV or UV at same time to avoid this. */
1986 /* IV-over-UV optimisation - choose to cache IV if possible */
1988 if (SvTYPE(sv) == SVt_NV)
1989 sv_upgrade(sv, SVt_PVNV);
1991 (void)SvIOKp_on(sv); /* Must do this first, to clear any SvOOK */
1992 /* < not <= as for NV doesn't preserve UV, ((NV)IV_MAX+1) will almost
1993 certainly cast into the IV range at IV_MAX, whereas the correct
1994 answer is the UV IV_MAX +1. Hence < ensures that dodgy boundary
1996 #if defined(NAN_COMPARE_BROKEN) && defined(Perl_isnan)
1997 if (Perl_isnan(SvNVX(sv))) {
2003 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2004 SvIV_set(sv, I_V(SvNVX(sv)));
2005 if (SvNVX(sv) == (NV) SvIVX(sv)
2006 #ifndef NV_PRESERVES_UV
2007 && (((UV)1 << NV_PRESERVES_UV_BITS) >
2008 (UV)(SvIVX(sv) > 0 ? SvIVX(sv) : -SvIVX(sv)))
2009 /* Don't flag it as "accurately an integer" if the number
2010 came from a (by definition imprecise) NV operation, and
2011 we're outside the range of NV integer precision */
2015 SvIOK_on(sv); /* Can this go wrong with rounding? NWC */
2017 /* scalar has trailing garbage, eg "42a" */
2019 DEBUG_c(PerlIO_printf(Perl_debug_log,
2020 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (precise)\n",
2026 /* IV not precise. No need to convert from PV, as NV
2027 conversion would already have cached IV if it detected
2028 that PV->IV would be better than PV->NV->IV
2029 flags already correct - don't set public IOK. */
2030 DEBUG_c(PerlIO_printf(Perl_debug_log,
2031 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (imprecise)\n",
2036 /* Can the above go wrong if SvIVX == IV_MIN and SvNVX < IV_MIN,
2037 but the cast (NV)IV_MIN rounds to a the value less (more
2038 negative) than IV_MIN which happens to be equal to SvNVX ??
2039 Analogous to 0xFFFFFFFFFFFFFFFF rounding up to NV (2**64) and
2040 NV rounding back to 0xFFFFFFFFFFFFFFFF, so UVX == UV(NVX) and
2041 (NV)UVX == NVX are both true, but the values differ. :-(
2042 Hopefully for 2s complement IV_MIN is something like
2043 0x8000000000000000 which will be exact. NWC */
2046 SvUV_set(sv, U_V(SvNVX(sv)));
2048 (SvNVX(sv) == (NV) SvUVX(sv))
2049 #ifndef NV_PRESERVES_UV
2050 /* Make sure it's not 0xFFFFFFFFFFFFFFFF */
2051 /*&& (SvUVX(sv) != UV_MAX) irrelevant with code below */
2052 && (((UV)1 << NV_PRESERVES_UV_BITS) > SvUVX(sv))
2053 /* Don't flag it as "accurately an integer" if the number
2054 came from a (by definition imprecise) NV operation, and
2055 we're outside the range of NV integer precision */
2061 DEBUG_c(PerlIO_printf(Perl_debug_log,
2062 "0x%"UVxf" 2iv(%"UVuf" => %"IVdf") (as unsigned)\n",
2068 else if (SvPOKp(sv) && SvLEN(sv)) {
2070 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2071 /* We want to avoid a possible problem when we cache an IV/ a UV which
2072 may be later translated to an NV, and the resulting NV is not
2073 the same as the direct translation of the initial string
2074 (eg 123.456 can shortcut to the IV 123 with atol(), but we must
2075 be careful to ensure that the value with the .456 is around if the
2076 NV value is requested in the future).
2078 This means that if we cache such an IV/a UV, we need to cache the
2079 NV as well. Moreover, we trade speed for space, and do not
2080 cache the NV if we are sure it's not needed.
2083 /* SVt_PVNV is one higher than SVt_PVIV, hence this order */
2084 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2085 == IS_NUMBER_IN_UV) {
2086 /* It's definitely an integer, only upgrade to PVIV */
2087 if (SvTYPE(sv) < SVt_PVIV)
2088 sv_upgrade(sv, SVt_PVIV);
2090 } else if (SvTYPE(sv) < SVt_PVNV)
2091 sv_upgrade(sv, SVt_PVNV);
2093 /* If NVs preserve UVs then we only use the UV value if we know that
2094 we aren't going to call atof() below. If NVs don't preserve UVs
2095 then the value returned may have more precision than atof() will
2096 return, even though value isn't perfectly accurate. */
2097 if ((numtype & (IS_NUMBER_IN_UV
2098 #ifdef NV_PRESERVES_UV
2101 )) == IS_NUMBER_IN_UV) {
2102 /* This won't turn off the public IOK flag if it was set above */
2103 (void)SvIOKp_on(sv);
2105 if (!(numtype & IS_NUMBER_NEG)) {
2107 if (value <= (UV)IV_MAX) {
2108 SvIV_set(sv, (IV)value);
2110 /* it didn't overflow, and it was positive. */
2111 SvUV_set(sv, value);
2115 /* 2s complement assumption */
2116 if (value <= (UV)IV_MIN) {
2117 SvIV_set(sv, -(IV)value);
2119 /* Too negative for an IV. This is a double upgrade, but
2120 I'm assuming it will be rare. */
2121 if (SvTYPE(sv) < SVt_PVNV)
2122 sv_upgrade(sv, SVt_PVNV);
2126 SvNV_set(sv, -(NV)value);
2127 SvIV_set(sv, IV_MIN);
2131 /* For !NV_PRESERVES_UV and IS_NUMBER_IN_UV and IS_NUMBER_NOT_INT we
2132 will be in the previous block to set the IV slot, and the next
2133 block to set the NV slot. So no else here. */
2135 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2136 != IS_NUMBER_IN_UV) {
2137 /* It wasn't an (integer that doesn't overflow the UV). */
2138 SvNV_set(sv, Atof(SvPVX_const(sv)));
2140 if (! numtype && ckWARN(WARN_NUMERIC))
2143 #if defined(USE_LONG_DOUBLE)
2144 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%" PERL_PRIgldbl ")\n",
2145 PTR2UV(sv), SvNVX(sv)));
2147 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"NVgf")\n",
2148 PTR2UV(sv), SvNVX(sv)));
2151 #ifdef NV_PRESERVES_UV
2152 (void)SvIOKp_on(sv);
2154 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2155 SvIV_set(sv, I_V(SvNVX(sv)));
2156 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
2159 NOOP; /* Integer is imprecise. NOK, IOKp */
2161 /* UV will not work better than IV */
2163 if (SvNVX(sv) > (NV)UV_MAX) {
2165 /* Integer is inaccurate. NOK, IOKp, is UV */
2166 SvUV_set(sv, UV_MAX);
2168 SvUV_set(sv, U_V(SvNVX(sv)));
2169 /* 0xFFFFFFFFFFFFFFFF not an issue in here, NVs
2170 NV preservse UV so can do correct comparison. */
2171 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
2174 NOOP; /* Integer is imprecise. NOK, IOKp, is UV */
2179 #else /* NV_PRESERVES_UV */
2180 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2181 == (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT)) {
2182 /* The IV/UV slot will have been set from value returned by
2183 grok_number above. The NV slot has just been set using
2186 assert (SvIOKp(sv));
2188 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2189 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2190 /* Small enough to preserve all bits. */
2191 (void)SvIOKp_on(sv);
2193 SvIV_set(sv, I_V(SvNVX(sv)));
2194 if ((NV)(SvIVX(sv)) == SvNVX(sv))
2196 /* Assumption: first non-preserved integer is < IV_MAX,
2197 this NV is in the preserved range, therefore: */
2198 if (!(U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))
2200 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);
2204 0 0 already failed to read UV.
2205 0 1 already failed to read UV.
2206 1 0 you won't get here in this case. IV/UV
2207 slot set, public IOK, Atof() unneeded.
2208 1 1 already read UV.
2209 so there's no point in sv_2iuv_non_preserve() attempting
2210 to use atol, strtol, strtoul etc. */
2212 sv_2iuv_non_preserve (sv, numtype);
2214 sv_2iuv_non_preserve (sv);
2218 #endif /* NV_PRESERVES_UV */
2219 /* It might be more code efficient to go through the entire logic above
2220 and conditionally set with SvIOKp_on() rather than SvIOK(), but it
2221 gets complex and potentially buggy, so more programmer efficient
2222 to do it this way, by turning off the public flags: */
2224 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK);
2228 if (isGV_with_GP(sv))
2229 return glob_2number(MUTABLE_GV(sv));
2231 if (!(SvFLAGS(sv) & SVs_PADTMP)) {
2232 if (!PL_localizing && ckWARN(WARN_UNINITIALIZED))
2235 if (SvTYPE(sv) < SVt_IV)
2236 /* Typically the caller expects that sv_any is not NULL now. */
2237 sv_upgrade(sv, SVt_IV);
2238 /* Return 0 from the caller. */
2245 =for apidoc sv_2iv_flags
2247 Return the integer value of an SV, doing any necessary string
2248 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2249 Normally used via the C<SvIV(sv)> and C<SvIVx(sv)> macros.
2255 Perl_sv_2iv_flags(pTHX_ register SV *const sv, const I32 flags)
2260 if (SvGMAGICAL(sv) || SvVALID(sv)) {
2261 /* FBMs use the space for SvIVX and SvNVX for other purposes, and use
2262 the same flag bit as SVf_IVisUV, so must not let them cache IVs.
2263 In practice they are extremely unlikely to actually get anywhere
2264 accessible by user Perl code - the only way that I'm aware of is when
2265 a constant subroutine which is used as the second argument to index.
2267 if (flags & SV_GMAGIC)
2272 return I_V(SvNVX(sv));
2274 if (SvPOKp(sv) && SvLEN(sv)) {
2277 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2279 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2280 == IS_NUMBER_IN_UV) {
2281 /* It's definitely an integer */
2282 if (numtype & IS_NUMBER_NEG) {
2283 if (value < (UV)IV_MIN)
2286 if (value < (UV)IV_MAX)
2291 if (ckWARN(WARN_NUMERIC))
2294 return I_V(Atof(SvPVX_const(sv)));
2299 assert(SvTYPE(sv) >= SVt_PVMG);
2300 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
2301 } else if (SvTHINKFIRST(sv)) {
2306 if (flags & SV_SKIP_OVERLOAD)
2308 tmpstr = AMG_CALLunary(sv, numer_amg);
2309 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2310 return SvIV(tmpstr);
2313 return PTR2IV(SvRV(sv));
2316 sv_force_normal_flags(sv, 0);
2318 if (SvREADONLY(sv) && !SvOK(sv)) {
2319 if (ckWARN(WARN_UNINITIALIZED))
2325 if (S_sv_2iuv_common(aTHX_ sv))
2328 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"IVdf")\n",
2329 PTR2UV(sv),SvIVX(sv)));
2330 return SvIsUV(sv) ? (IV)SvUVX(sv) : SvIVX(sv);
2334 =for apidoc sv_2uv_flags
2336 Return the unsigned integer value of an SV, doing any necessary string
2337 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2338 Normally used via the C<SvUV(sv)> and C<SvUVx(sv)> macros.
2344 Perl_sv_2uv_flags(pTHX_ register SV *const sv, const I32 flags)
2349 if (SvGMAGICAL(sv) || SvVALID(sv)) {
2350 /* FBMs use the space for SvIVX and SvNVX for other purposes, and use
2351 the same flag bit as SVf_IVisUV, so must not let them cache IVs. */
2352 if (flags & SV_GMAGIC)
2357 return U_V(SvNVX(sv));
2358 if (SvPOKp(sv) && SvLEN(sv)) {
2361 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2363 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2364 == IS_NUMBER_IN_UV) {
2365 /* It's definitely an integer */
2366 if (!(numtype & IS_NUMBER_NEG))
2370 if (ckWARN(WARN_NUMERIC))
2373 return U_V(Atof(SvPVX_const(sv)));
2378 assert(SvTYPE(sv) >= SVt_PVMG);
2379 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
2380 } else if (SvTHINKFIRST(sv)) {
2385 if (flags & SV_SKIP_OVERLOAD)
2387 tmpstr = AMG_CALLunary(sv, numer_amg);
2388 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2389 return SvUV(tmpstr);
2392 return PTR2UV(SvRV(sv));
2395 sv_force_normal_flags(sv, 0);
2397 if (SvREADONLY(sv) && !SvOK(sv)) {
2398 if (ckWARN(WARN_UNINITIALIZED))
2404 if (S_sv_2iuv_common(aTHX_ sv))
2408 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2uv(%"UVuf")\n",
2409 PTR2UV(sv),SvUVX(sv)));
2410 return SvIsUV(sv) ? SvUVX(sv) : (UV)SvIVX(sv);
2414 =for apidoc sv_2nv_flags
2416 Return the num value of an SV, doing any necessary string or integer
2417 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2418 Normally used via the C<SvNV(sv)> and C<SvNVx(sv)> macros.
2424 Perl_sv_2nv_flags(pTHX_ register SV *const sv, const I32 flags)
2429 if (SvGMAGICAL(sv) || SvVALID(sv)) {
2430 /* FBMs use the space for SvIVX and SvNVX for other purposes, and use
2431 the same flag bit as SVf_IVisUV, so must not let them cache NVs. */
2432 if (flags & SV_GMAGIC)
2436 if ((SvPOKp(sv) && SvLEN(sv)) && !SvIOKp(sv)) {
2437 if (!SvIOKp(sv) && ckWARN(WARN_NUMERIC) &&
2438 !grok_number(SvPVX_const(sv), SvCUR(sv), NULL))
2440 return Atof(SvPVX_const(sv));
2444 return (NV)SvUVX(sv);
2446 return (NV)SvIVX(sv);
2451 assert(SvTYPE(sv) >= SVt_PVMG);
2452 /* This falls through to the report_uninit near the end of the
2454 } else if (SvTHINKFIRST(sv)) {
2459 if (flags & SV_SKIP_OVERLOAD)
2461 tmpstr = AMG_CALLunary(sv, numer_amg);
2462 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2463 return SvNV(tmpstr);
2466 return PTR2NV(SvRV(sv));
2469 sv_force_normal_flags(sv, 0);
2471 if (SvREADONLY(sv) && !SvOK(sv)) {
2472 if (ckWARN(WARN_UNINITIALIZED))
2477 if (SvTYPE(sv) < SVt_NV) {
2478 /* The logic to use SVt_PVNV if necessary is in sv_upgrade. */
2479 sv_upgrade(sv, SVt_NV);
2480 #ifdef USE_LONG_DOUBLE
2482 STORE_NUMERIC_LOCAL_SET_STANDARD();
2483 PerlIO_printf(Perl_debug_log,
2484 "0x%"UVxf" num(%" PERL_PRIgldbl ")\n",
2485 PTR2UV(sv), SvNVX(sv));
2486 RESTORE_NUMERIC_LOCAL();
2490 STORE_NUMERIC_LOCAL_SET_STANDARD();
2491 PerlIO_printf(Perl_debug_log, "0x%"UVxf" num(%"NVgf")\n",
2492 PTR2UV(sv), SvNVX(sv));
2493 RESTORE_NUMERIC_LOCAL();
2497 else if (SvTYPE(sv) < SVt_PVNV)
2498 sv_upgrade(sv, SVt_PVNV);
2503 SvNV_set(sv, SvIsUV(sv) ? (NV)SvUVX(sv) : (NV)SvIVX(sv));
2504 #ifdef NV_PRESERVES_UV
2510 /* Only set the public NV OK flag if this NV preserves the IV */
2511 /* Check it's not 0xFFFFFFFFFFFFFFFF */
2513 SvIsUV(sv) ? ((SvUVX(sv) != UV_MAX)&&(SvUVX(sv) == U_V(SvNVX(sv))))
2514 : (SvIVX(sv) == I_V(SvNVX(sv))))
2520 else if (SvPOKp(sv) && SvLEN(sv)) {
2522 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2523 if (!SvIOKp(sv) && !numtype && ckWARN(WARN_NUMERIC))
2525 #ifdef NV_PRESERVES_UV
2526 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2527 == IS_NUMBER_IN_UV) {
2528 /* It's definitely an integer */
2529 SvNV_set(sv, (numtype & IS_NUMBER_NEG) ? -(NV)value : (NV)value);
2531 SvNV_set(sv, Atof(SvPVX_const(sv)));
2537 SvNV_set(sv, Atof(SvPVX_const(sv)));
2538 /* Only set the public NV OK flag if this NV preserves the value in
2539 the PV at least as well as an IV/UV would.
2540 Not sure how to do this 100% reliably. */
2541 /* if that shift count is out of range then Configure's test is
2542 wonky. We shouldn't be in here with NV_PRESERVES_UV_BITS ==
2544 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2545 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2546 SvNOK_on(sv); /* Definitely small enough to preserve all bits */
2547 } else if (!(numtype & IS_NUMBER_IN_UV)) {
2548 /* Can't use strtol etc to convert this string, so don't try.
2549 sv_2iv and sv_2uv will use the NV to convert, not the PV. */
2552 /* value has been set. It may not be precise. */
2553 if ((numtype & IS_NUMBER_NEG) && (value > (UV)IV_MIN)) {
2554 /* 2s complement assumption for (UV)IV_MIN */
2555 SvNOK_on(sv); /* Integer is too negative. */
2560 if (numtype & IS_NUMBER_NEG) {
2561 SvIV_set(sv, -(IV)value);
2562 } else if (value <= (UV)IV_MAX) {
2563 SvIV_set(sv, (IV)value);
2565 SvUV_set(sv, value);
2569 if (numtype & IS_NUMBER_NOT_INT) {
2570 /* I believe that even if the original PV had decimals,
2571 they are lost beyond the limit of the FP precision.
2572 However, neither is canonical, so both only get p
2573 flags. NWC, 2000/11/25 */
2574 /* Both already have p flags, so do nothing */
2576 const NV nv = SvNVX(sv);
2577 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2578 if (SvIVX(sv) == I_V(nv)) {
2581 /* It had no "." so it must be integer. */
2585 /* between IV_MAX and NV(UV_MAX).
2586 Could be slightly > UV_MAX */
2588 if (numtype & IS_NUMBER_NOT_INT) {
2589 /* UV and NV both imprecise. */
2591 const UV nv_as_uv = U_V(nv);
2593 if (value == nv_as_uv && SvUVX(sv) != UV_MAX) {
2602 /* It might be more code efficient to go through the entire logic above
2603 and conditionally set with SvNOKp_on() rather than SvNOK(), but it
2604 gets complex and potentially buggy, so more programmer efficient
2605 to do it this way, by turning off the public flags: */
2607 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK);
2608 #endif /* NV_PRESERVES_UV */
2611 if (isGV_with_GP(sv)) {
2612 glob_2number(MUTABLE_GV(sv));
2616 if (!PL_localizing && !(SvFLAGS(sv) & SVs_PADTMP) && ckWARN(WARN_UNINITIALIZED))
2618 assert (SvTYPE(sv) >= SVt_NV);
2619 /* Typically the caller expects that sv_any is not NULL now. */
2620 /* XXX Ilya implies that this is a bug in callers that assume this
2621 and ideally should be fixed. */
2624 #if defined(USE_LONG_DOUBLE)
2626 STORE_NUMERIC_LOCAL_SET_STANDARD();
2627 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2nv(%" PERL_PRIgldbl ")\n",
2628 PTR2UV(sv), SvNVX(sv));
2629 RESTORE_NUMERIC_LOCAL();
2633 STORE_NUMERIC_LOCAL_SET_STANDARD();
2634 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 1nv(%"NVgf")\n",
2635 PTR2UV(sv), SvNVX(sv));
2636 RESTORE_NUMERIC_LOCAL();
2645 Return an SV with the numeric value of the source SV, doing any necessary
2646 reference or overload conversion. You must use the C<SvNUM(sv)> macro to
2647 access this function.
2653 Perl_sv_2num(pTHX_ register SV *const sv)
2655 PERL_ARGS_ASSERT_SV_2NUM;
2660 SV * const tmpsv = AMG_CALLunary(sv, numer_amg);
2661 TAINT_IF(tmpsv && SvTAINTED(tmpsv));
2662 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
2663 return sv_2num(tmpsv);
2665 return sv_2mortal(newSVuv(PTR2UV(SvRV(sv))));
2668 /* uiv_2buf(): private routine for use by sv_2pv_flags(): print an IV or
2669 * UV as a string towards the end of buf, and return pointers to start and
2672 * We assume that buf is at least TYPE_CHARS(UV) long.
2676 S_uiv_2buf(char *const buf, const IV iv, UV uv, const int is_uv, char **const peob)
2678 char *ptr = buf + TYPE_CHARS(UV);
2679 char * const ebuf = ptr;
2682 PERL_ARGS_ASSERT_UIV_2BUF;
2694 *--ptr = '0' + (char)(uv % 10);
2703 =for apidoc sv_2pv_flags
2705 Returns a pointer to the string value of an SV, and sets *lp to its length.
2706 If flags includes SV_GMAGIC, does an mg_get() first. Coerces sv to a string
2708 Normally invoked via the C<SvPV_flags> macro. C<sv_2pv()> and C<sv_2pv_nomg>
2709 usually end up here too.
2715 Perl_sv_2pv_flags(pTHX_ register SV *const sv, STRLEN *const lp, const I32 flags)
2725 if (SvGMAGICAL(sv)) {
2726 if (flags & SV_GMAGIC)
2731 if (flags & SV_MUTABLE_RETURN)
2732 return SvPVX_mutable(sv);
2733 if (flags & SV_CONST_RETURN)
2734 return (char *)SvPVX_const(sv);
2737 if (SvIOKp(sv) || SvNOKp(sv)) {
2738 char tbuf[64]; /* Must fit sprintf/Gconvert of longest IV/NV */
2743 ? my_snprintf(tbuf, sizeof(tbuf), "%"UVuf, (UV)SvUVX(sv))
2744 : my_snprintf(tbuf, sizeof(tbuf), "%"IVdf, (IV)SvIVX(sv));
2745 } else if(SvNVX(sv) == 0.0) {
2750 Gconvert(SvNVX(sv), NV_DIG, 0, tbuf);
2757 SvUPGRADE(sv, SVt_PV);
2760 s = SvGROW_mutable(sv, len + 1);
2763 return (char*)memcpy(s, tbuf, len + 1);
2769 assert(SvTYPE(sv) >= SVt_PVMG);
2770 /* This falls through to the report_uninit near the end of the
2772 } else if (SvTHINKFIRST(sv)) {
2777 if (flags & SV_SKIP_OVERLOAD)
2779 tmpstr = AMG_CALLunary(sv, string_amg);
2780 TAINT_IF(tmpstr && SvTAINTED(tmpstr));
2781 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2783 /* char *pv = lp ? SvPV(tmpstr, *lp) : SvPV_nolen(tmpstr);
2787 if ((SvFLAGS(tmpstr) & (SVf_POK)) == SVf_POK) {
2788 if (flags & SV_CONST_RETURN) {
2789 pv = (char *) SvPVX_const(tmpstr);
2791 pv = (flags & SV_MUTABLE_RETURN)
2792 ? SvPVX_mutable(tmpstr) : SvPVX(tmpstr);
2795 *lp = SvCUR(tmpstr);
2797 pv = sv_2pv_flags(tmpstr, lp, flags);
2810 SV *const referent = SvRV(sv);
2814 retval = buffer = savepvn("NULLREF", len);
2815 } else if (SvTYPE(referent) == SVt_REGEXP) {
2816 REGEXP * const re = (REGEXP *)MUTABLE_PTR(referent);
2821 /* If the regex is UTF-8 we want the containing scalar to
2822 have an UTF-8 flag too */
2828 if ((seen_evals = RX_SEEN_EVALS(re)))
2829 PL_reginterp_cnt += seen_evals;
2832 *lp = RX_WRAPLEN(re);
2834 return RX_WRAPPED(re);
2836 const char *const typestr = sv_reftype(referent, 0);
2837 const STRLEN typelen = strlen(typestr);
2838 UV addr = PTR2UV(referent);
2839 const char *stashname = NULL;
2840 STRLEN stashnamelen = 0; /* hush, gcc */
2841 const char *buffer_end;
2843 if (SvOBJECT(referent)) {
2844 const HEK *const name = HvNAME_HEK(SvSTASH(referent));
2847 stashname = HEK_KEY(name);
2848 stashnamelen = HEK_LEN(name);
2850 if (HEK_UTF8(name)) {
2856 stashname = "__ANON__";
2859 len = stashnamelen + 1 /* = */ + typelen + 3 /* (0x */
2860 + 2 * sizeof(UV) + 2 /* )\0 */;
2862 len = typelen + 3 /* (0x */
2863 + 2 * sizeof(UV) + 2 /* )\0 */;
2866 Newx(buffer, len, char);
2867 buffer_end = retval = buffer + len;
2869 /* Working backwards */
2873 *--retval = PL_hexdigit[addr & 15];
2874 } while (addr >>= 4);
2880 memcpy(retval, typestr, typelen);
2884 retval -= stashnamelen;
2885 memcpy(retval, stashname, stashnamelen);
2887 /* retval may not necessarily have reached the start of the
2889 assert (retval >= buffer);
2891 len = buffer_end - retval - 1; /* -1 for that \0 */
2899 if (SvREADONLY(sv) && !SvOK(sv)) {
2902 if (flags & SV_UNDEF_RETURNS_NULL)
2904 if (ckWARN(WARN_UNINITIALIZED))
2909 if (SvIOK(sv) || ((SvIOKp(sv) && !SvNOKp(sv)))) {
2910 /* I'm assuming that if both IV and NV are equally valid then
2911 converting the IV is going to be more efficient */
2912 const U32 isUIOK = SvIsUV(sv);
2913 char buf[TYPE_CHARS(UV)];
2917 if (SvTYPE(sv) < SVt_PVIV)
2918 sv_upgrade(sv, SVt_PVIV);
2919 ptr = uiv_2buf(buf, SvIVX(sv), SvUVX(sv), isUIOK, &ebuf);
2921 /* inlined from sv_setpvn */
2922 s = SvGROW_mutable(sv, len + 1);
2923 Move(ptr, s, len, char);
2927 else if (SvNOKp(sv)) {
2928 if (SvTYPE(sv) < SVt_PVNV)
2929 sv_upgrade(sv, SVt_PVNV);
2930 if (SvNVX(sv) == 0.0) {
2931 s = SvGROW_mutable(sv, 2);
2936 /* The +20 is pure guesswork. Configure test needed. --jhi */
2937 s = SvGROW_mutable(sv, NV_DIG + 20);
2938 /* some Xenix systems wipe out errno here */
2939 Gconvert(SvNVX(sv), NV_DIG, 0, s);
2949 if (isGV_with_GP(sv)) {
2950 GV *const gv = MUTABLE_GV(sv);
2951 const U32 wasfake = SvFLAGS(gv) & SVf_FAKE;
2952 SV *const buffer = sv_newmortal();
2954 /* FAKE globs can get coerced, so need to turn this off temporarily
2957 gv_efullname3(buffer, gv, "*");
2958 SvFLAGS(gv) |= wasfake;
2960 if (SvPOK(buffer)) {
2962 *lp = SvCUR(buffer);
2964 if ( SvUTF8(buffer) ) SvUTF8_on(sv);
2965 return SvPVX(buffer);
2976 if (flags & SV_UNDEF_RETURNS_NULL)
2978 if (!PL_localizing && !(SvFLAGS(sv) & SVs_PADTMP) && ckWARN(WARN_UNINITIALIZED))
2980 if (SvTYPE(sv) < SVt_PV)
2981 /* Typically the caller expects that sv_any is not NULL now. */
2982 sv_upgrade(sv, SVt_PV);
2986 const STRLEN len = s - SvPVX_const(sv);
2992 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
2993 PTR2UV(sv),SvPVX_const(sv)));
2994 if (flags & SV_CONST_RETURN)
2995 return (char *)SvPVX_const(sv);
2996 if (flags & SV_MUTABLE_RETURN)
2997 return SvPVX_mutable(sv);
3002 =for apidoc sv_copypv
3004 Copies a stringified representation of the source SV into the
3005 destination SV. Automatically performs any necessary mg_get and
3006 coercion of numeric values into strings. Guaranteed to preserve
3007 UTF8 flag even from overloaded objects. Similar in nature to
3008 sv_2pv[_flags] but operates directly on an SV instead of just the
3009 string. Mostly uses sv_2pv_flags to do its work, except when that
3010 would lose the UTF-8'ness of the PV.
3016 Perl_sv_copypv(pTHX_ SV *const dsv, register SV *const ssv)
3019 const char * const s = SvPV_const(ssv,len);
3021 PERL_ARGS_ASSERT_SV_COPYPV;
3023 sv_setpvn(dsv,s,len);
3031 =for apidoc sv_2pvbyte
3033 Return a pointer to the byte-encoded representation of the SV, and set *lp
3034 to its length. May cause the SV to be downgraded from UTF-8 as a
3037 Usually accessed via the C<SvPVbyte> macro.
3043 Perl_sv_2pvbyte(pTHX_ register SV *const sv, STRLEN *const lp)
3045 PERL_ARGS_ASSERT_SV_2PVBYTE;
3048 sv_utf8_downgrade(sv,0);
3049 return lp ? SvPV_nomg(sv,*lp) : SvPV_nomg_nolen(sv);
3053 =for apidoc sv_2pvutf8
3055 Return a pointer to the UTF-8-encoded representation of the SV, and set *lp
3056 to its length. May cause the SV to be upgraded to UTF-8 as a side-effect.
3058 Usually accessed via the C<SvPVutf8> macro.
3064 Perl_sv_2pvutf8(pTHX_ register SV *const sv, STRLEN *const lp)
3066 PERL_ARGS_ASSERT_SV_2PVUTF8;
3068 sv_utf8_upgrade(sv);
3069 return lp ? SvPV(sv,*lp) : SvPV_nolen(sv);
3074 =for apidoc sv_2bool
3076 This macro is only used by sv_true() or its macro equivalent, and only if
3077 the latter's argument is neither SvPOK, SvIOK nor SvNOK.
3078 It calls sv_2bool_flags with the SV_GMAGIC flag.
3080 =for apidoc sv_2bool_flags
3082 This function is only used by sv_true() and friends, and only if
3083 the latter's argument is neither SvPOK, SvIOK nor SvNOK. If the flags
3084 contain SV_GMAGIC, then it does an mg_get() first.
3091 Perl_sv_2bool_flags(pTHX_ register SV *const sv, const I32 flags)
3095 PERL_ARGS_ASSERT_SV_2BOOL_FLAGS;
3097 if(flags & SV_GMAGIC) SvGETMAGIC(sv);
3103 SV * const tmpsv = AMG_CALLunary(sv, bool__amg);
3104 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
3105 return cBOOL(SvTRUE(tmpsv));
3107 return SvRV(sv) != 0;
3110 register XPV* const Xpvtmp = (XPV*)SvANY(sv);
3112 (*sv->sv_u.svu_pv > '0' ||
3113 Xpvtmp->xpv_cur > 1 ||
3114 (Xpvtmp->xpv_cur && *sv->sv_u.svu_pv != '0')))
3121 return SvIVX(sv) != 0;
3124 return SvNVX(sv) != 0.0;
3126 if (isGV_with_GP(sv))
3136 =for apidoc sv_utf8_upgrade
3138 Converts the PV of an SV to its UTF-8-encoded form.
3139 Forces the SV to string form if it is not already.
3140 Will C<mg_get> on C<sv> if appropriate.
3141 Always sets the SvUTF8 flag to avoid future validity checks even
3142 if the whole string is the same in UTF-8 as not.
3143 Returns the number of bytes in the converted string
3145 This is not as a general purpose byte encoding to Unicode interface:
3146 use the Encode extension for that.
3148 =for apidoc sv_utf8_upgrade_nomg
3150 Like sv_utf8_upgrade, but doesn't do magic on C<sv>
3152 =for apidoc sv_utf8_upgrade_flags
3154 Converts the PV of an SV to its UTF-8-encoded form.
3155 Forces the SV to string form if it is not already.
3156 Always sets the SvUTF8 flag to avoid future validity checks even
3157 if all the bytes are invariant in UTF-8. If C<flags> has C<SV_GMAGIC> bit set,
3158 will C<mg_get> on C<sv> if appropriate, else not.
3159 Returns the number of bytes in the converted string
3160 C<sv_utf8_upgrade> and
3161 C<sv_utf8_upgrade_nomg> are implemented in terms of this function.
3163 This is not as a general purpose byte encoding to Unicode interface:
3164 use the Encode extension for that.
3168 The grow version is currently not externally documented. It adds a parameter,
3169 extra, which is the number of unused bytes the string of 'sv' is guaranteed to
3170 have free after it upon return. This allows the caller to reserve extra space
3171 that it intends to fill, to avoid extra grows.
3173 Also externally undocumented for the moment is the flag SV_FORCE_UTF8_UPGRADE,
3174 which can be used to tell this function to not first check to see if there are
3175 any characters that are different in UTF-8 (variant characters) which would
3176 force it to allocate a new string to sv, but to assume there are. Typically
3177 this flag is used by a routine that has already parsed the string to find that
3178 there are such characters, and passes this information on so that the work
3179 doesn't have to be repeated.
3181 (One might think that the calling routine could pass in the position of the
3182 first such variant, so it wouldn't have to be found again. But that is not the
3183 case, because typically when the caller is likely to use this flag, it won't be
3184 calling this routine unless it finds something that won't fit into a byte.
3185 Otherwise it tries to not upgrade and just use bytes. But some things that
3186 do fit into a byte are variants in utf8, and the caller may not have been
3187 keeping track of these.)
3189 If the routine itself changes the string, it adds a trailing NUL. Such a NUL
3190 isn't guaranteed due to having other routines do the work in some input cases,
3191 or if the input is already flagged as being in utf8.
3193 The speed of this could perhaps be improved for many cases if someone wanted to
3194 write a fast function that counts the number of variant characters in a string,
3195 especially if it could return the position of the first one.
3200 Perl_sv_utf8_upgrade_flags_grow(pTHX_ register SV *const sv, const I32 flags, STRLEN extra)
3204 PERL_ARGS_ASSERT_SV_UTF8_UPGRADE_FLAGS_GROW;
3206 if (sv == &PL_sv_undef)
3210 if (SvREADONLY(sv) && (SvPOKp(sv) || SvIOKp(sv) || SvNOKp(sv))) {
3211 (void) sv_2pv_flags(sv,&len, flags);
3213 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3217 (void) SvPV_force_flags(sv,len,flags & SV_GMAGIC);
3222 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3227 sv_force_normal_flags(sv, 0);
3230 if (PL_encoding && !(flags & SV_UTF8_NO_ENCODING)) {
3231 sv_recode_to_utf8(sv, PL_encoding);
3232 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3236 if (SvCUR(sv) == 0) {
3237 if (extra) SvGROW(sv, extra);
3238 } else { /* Assume Latin-1/EBCDIC */
3239 /* This function could be much more efficient if we
3240 * had a FLAG in SVs to signal if there are any variant
3241 * chars in the PV. Given that there isn't such a flag
3242 * make the loop as fast as possible (although there are certainly ways
3243 * to speed this up, eg. through vectorization) */
3244 U8 * s = (U8 *) SvPVX_const(sv);
3245 U8 * e = (U8 *) SvEND(sv);
3247 STRLEN two_byte_count = 0;
3249 if (flags & SV_FORCE_UTF8_UPGRADE) goto must_be_utf8;
3251 /* See if really will need to convert to utf8. We mustn't rely on our
3252 * incoming SV being well formed and having a trailing '\0', as certain
3253 * code in pp_formline can send us partially built SVs. */
3257 if (NATIVE_IS_INVARIANT(ch)) continue;
3259 t--; /* t already incremented; re-point to first variant */
3264 /* utf8 conversion not needed because all are invariants. Mark as
3265 * UTF-8 even if no variant - saves scanning loop */
3271 /* Here, the string should be converted to utf8, either because of an
3272 * input flag (two_byte_count = 0), or because a character that
3273 * requires 2 bytes was found (two_byte_count = 1). t points either to
3274 * the beginning of the string (if we didn't examine anything), or to
3275 * the first variant. In either case, everything from s to t - 1 will
3276 * occupy only 1 byte each on output.
3278 * There are two main ways to convert. One is to create a new string
3279 * and go through the input starting from the beginning, appending each
3280 * converted value onto the new string as we go along. It's probably
3281 * best to allocate enough space in the string for the worst possible
3282 * case rather than possibly running out of space and having to
3283 * reallocate and then copy what we've done so far. Since everything
3284 * from s to t - 1 is invariant, the destination can be initialized
3285 * with these using a fast memory copy
3287 * The other way is to figure out exactly how big the string should be
3288 * by parsing the entire input. Then you don't have to make it big
3289 * enough to handle the worst possible case, and more importantly, if
3290 * the string you already have is large enough, you don't have to
3291 * allocate a new string, you can copy the last character in the input
3292 * string to the final position(s) that will be occupied by the
3293 * converted string and go backwards, stopping at t, since everything
3294 * before that is invariant.
3296 * There are advantages and disadvantages to each method.
3298 * In the first method, we can allocate a new string, do the memory
3299 * copy from the s to t - 1, and then proceed through the rest of the
3300 * string byte-by-byte.
3302 * In the second method, we proceed through the rest of the input
3303 * string just calculating how big the converted string will be. Then
3304 * there are two cases:
3305 * 1) if the string has enough extra space to handle the converted
3306 * value. We go backwards through the string, converting until we
3307 * get to the position we are at now, and then stop. If this
3308 * position is far enough along in the string, this method is
3309 * faster than the other method. If the memory copy were the same
3310 * speed as the byte-by-byte loop, that position would be about
3311 * half-way, as at the half-way mark, parsing to the end and back
3312 * is one complete string's parse, the same amount as starting
3313 * over and going all the way through. Actually, it would be
3314 * somewhat less than half-way, as it's faster to just count bytes
3315 * than to also copy, and we don't have the overhead of allocating
3316 * a new string, changing the scalar to use it, and freeing the
3317 * existing one. But if the memory copy is fast, the break-even
3318 * point is somewhere after half way. The counting loop could be
3319 * sped up by vectorization, etc, to move the break-even point
3320 * further towards the beginning.
3321 * 2) if the string doesn't have enough space to handle the converted
3322 * value. A new string will have to be allocated, and one might
3323 * as well, given that, start from the beginning doing the first
3324 * method. We've spent extra time parsing the string and in
3325 * exchange all we've gotten is that we know precisely how big to
3326 * make the new one. Perl is more optimized for time than space,
3327 * so this case is a loser.
3328 * So what I've decided to do is not use the 2nd method unless it is
3329 * guaranteed that a new string won't have to be allocated, assuming
3330 * the worst case. I also decided not to put any more conditions on it
3331 * than this, for now. It seems likely that, since the worst case is
3332 * twice as big as the unknown portion of the string (plus 1), we won't
3333 * be guaranteed enough space, causing us to go to the first method,
3334 * unless the string is short, or the first variant character is near
3335 * the end of it. In either of these cases, it seems best to use the
3336 * 2nd method. The only circumstance I can think of where this would
3337 * be really slower is if the string had once had much more data in it
3338 * than it does now, but there is still a substantial amount in it */
3341 STRLEN invariant_head = t - s;
3342 STRLEN size = invariant_head + (e - t) * 2 + 1 + extra;
3343 if (SvLEN(sv) < size) {
3345 /* Here, have decided to allocate a new string */
3350 Newx(dst, size, U8);
3352 /* If no known invariants at the beginning of the input string,
3353 * set so starts from there. Otherwise, can use memory copy to
3354 * get up to where we are now, and then start from here */
3356 if (invariant_head <= 0) {
3359 Copy(s, dst, invariant_head, char);
3360 d = dst + invariant_head;
3364 const UV uv = NATIVE8_TO_UNI(*t++);
3365 if (UNI_IS_INVARIANT(uv))
3366 *d++ = (U8)UNI_TO_NATIVE(uv);
3368 *d++ = (U8)UTF8_EIGHT_BIT_HI(uv);
3369 *d++ = (U8)UTF8_EIGHT_BIT_LO(uv);
3373 SvPV_free(sv); /* No longer using pre-existing string */
3374 SvPV_set(sv, (char*)dst);
3375 SvCUR_set(sv, d - dst);
3376 SvLEN_set(sv, size);
3379 /* Here, have decided to get the exact size of the string.
3380 * Currently this happens only when we know that there is
3381 * guaranteed enough space to fit the converted string, so
3382 * don't have to worry about growing. If two_byte_count is 0,
3383 * then t points to the first byte of the string which hasn't
3384 * been examined yet. Otherwise two_byte_count is 1, and t
3385 * points to the first byte in the string that will expand to
3386 * two. Depending on this, start examining at t or 1 after t.
3389 U8 *d = t + two_byte_count;
3392 /* Count up the remaining bytes that expand to two */
3395 const U8 chr = *d++;
3396 if (! NATIVE_IS_INVARIANT(chr)) two_byte_count++;
3399 /* The string will expand by just the number of bytes that
3400 * occupy two positions. But we are one afterwards because of
3401 * the increment just above. This is the place to put the
3402 * trailing NUL, and to set the length before we decrement */
3404 d += two_byte_count;
3405 SvCUR_set(sv, d - s);
3409 /* Having decremented d, it points to the position to put the
3410 * very last byte of the expanded string. Go backwards through
3411 * the string, copying and expanding as we go, stopping when we
3412 * get to the part that is invariant the rest of the way down */
3416 const U8 ch = NATIVE8_TO_UNI(*e--);
3417 if (UNI_IS_INVARIANT(ch)) {
3418 *d-- = UNI_TO_NATIVE(ch);
3420 *d-- = (U8)UTF8_EIGHT_BIT_LO(ch);
3421 *d-- = (U8)UTF8_EIGHT_BIT_HI(ch);
3426 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
3427 /* Update pos. We do it at the end rather than during
3428 * the upgrade, to avoid slowing down the common case
3429 * (upgrade without pos) */
3430 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
3432 I32 pos = mg->mg_len;
3433 if (pos > 0 && (U32)pos > invariant_head) {
3434 U8 *d = (U8*) SvPVX(sv) + invariant_head;
3435 STRLEN n = (U32)pos - invariant_head;
3437 if (UTF8_IS_START(*d))
3442 mg->mg_len = d - (U8*)SvPVX(sv);
3445 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
3446 magic_setutf8(sv,mg); /* clear UTF8 cache */
3451 /* Mark as UTF-8 even if no variant - saves scanning loop */
3457 =for apidoc sv_utf8_downgrade
3459 Attempts to convert the PV of an SV from characters to bytes.
3460 If the PV contains a character that cannot fit
3461 in a byte, this conversion will fail;
3462 in this case, either returns false or, if C<fail_ok> is not
3465 This is not as a general purpose Unicode to byte encoding interface:
3466 use the Encode extension for that.
3472 Perl_sv_utf8_downgrade(pTHX_ register SV *const sv, const bool fail_ok)
3476 PERL_ARGS_ASSERT_SV_UTF8_DOWNGRADE;
3478 if (SvPOKp(sv) && SvUTF8(sv)) {
3482 int mg_flags = SV_GMAGIC;
3485 sv_force_normal_flags(sv, 0);
3487 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
3489 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
3491 I32 pos = mg->mg_len;
3493 sv_pos_b2u(sv, &pos);
3494 mg_flags = 0; /* sv_pos_b2u does get magic */
3498 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
3499 magic_setutf8(sv,mg); /* clear UTF8 cache */
3502 s = (U8 *) SvPV_flags(sv, len, mg_flags);
3504 if (!utf8_to_bytes(s, &len)) {
3509 Perl_croak(aTHX_ "Wide character in %s",
3512 Perl_croak(aTHX_ "Wide character");
3523 =for apidoc sv_utf8_encode
3525 Converts the PV of an SV to UTF-8, but then turns the C<SvUTF8>
3526 flag off so that it looks like octets again.
3532 Perl_sv_utf8_encode(pTHX_ register SV *const sv)
3534 PERL_ARGS_ASSERT_SV_UTF8_ENCODE;
3537 sv_force_normal_flags(sv, 0);
3539 if (SvREADONLY(sv)) {
3540 Perl_croak_no_modify(aTHX);
3542 (void) sv_utf8_upgrade(sv);
3547 =for apidoc sv_utf8_decode
3549 If the PV of the SV is an octet sequence in UTF-8
3550 and contains a multiple-byte character, the C<SvUTF8> flag is turned on
3551 so that it looks like a character. If the PV contains only single-byte
3552 characters, the C<SvUTF8> flag stays off.
3553 Scans PV for validity and returns false if the PV is invalid UTF-8.
3559 Perl_sv_utf8_decode(pTHX_ register SV *const sv)
3561 PERL_ARGS_ASSERT_SV_UTF8_DECODE;
3564 const U8 *start, *c;
3567 /* The octets may have got themselves encoded - get them back as
3570 if (!sv_utf8_downgrade(sv, TRUE))
3573 /* it is actually just a matter of turning the utf8 flag on, but
3574 * we want to make sure everything inside is valid utf8 first.
3576 c = start = (const U8 *) SvPVX_const(sv);
3577 if (!is_utf8_string(c, SvCUR(sv)+1))
3579 e = (const U8 *) SvEND(sv);
3582 if (!UTF8_IS_INVARIANT(ch)) {
3587 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
3588 /* adjust pos to the start of a UTF8 char sequence */
3589 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
3591 I32 pos = mg->mg_len;
3593 for (c = start + pos; c > start; c--) {
3594 if (UTF8_IS_START(*c))
3597 mg->mg_len = c - start;
3600 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
3601 magic_setutf8(sv,mg); /* clear UTF8 cache */
3608 =for apidoc sv_setsv
3610 Copies the contents of the source SV C<ssv> into the destination SV
3611 C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3612 function if the source SV needs to be reused. Does not handle 'set' magic.
3613 Loosely speaking, it performs a copy-by-value, obliterating any previous
3614 content of the destination.
3616 You probably want to use one of the assortment of wrappers, such as
3617 C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3618 C<SvSetMagicSV_nosteal>.
3620 =for apidoc sv_setsv_flags
3622 Copies the contents of the source SV C<ssv> into the destination SV
3623 C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3624 function if the source SV needs to be reused. Does not handle 'set' magic.
3625 Loosely speaking, it performs a copy-by-value, obliterating any previous
3626 content of the destination.
3627 If the C<flags> parameter has the C<SV_GMAGIC> bit set, will C<mg_get> on
3628 C<ssv> if appropriate, else not. If the C<flags> parameter has the
3629 C<NOSTEAL> bit set then the buffers of temps will not be stolen. <sv_setsv>
3630 and C<sv_setsv_nomg> are implemented in terms of this function.
3632 You probably want to use one of the assortment of wrappers, such as
3633 C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3634 C<SvSetMagicSV_nosteal>.
3636 This is the primary function for copying scalars, and most other
3637 copy-ish functions and macros use this underneath.
3643 S_glob_assign_glob(pTHX_ SV *const dstr, SV *const sstr, const int dtype)
3645 I32 mro_changes = 0; /* 1 = method, 2 = isa, 3 = recursive isa */
3646 HV *old_stash = NULL;
3648 PERL_ARGS_ASSERT_GLOB_ASSIGN_GLOB;
3650 if (dtype != SVt_PVGV && !isGV_with_GP(dstr)) {
3651 const char * const name = GvNAME(sstr);
3652 const STRLEN len = GvNAMELEN(sstr);
3654 if (dtype >= SVt_PV) {
3660 SvUPGRADE(dstr, SVt_PVGV);
3661 (void)SvOK_off(dstr);
3662 /* FIXME - why are we doing this, then turning it off and on again
3664 isGV_with_GP_on(dstr);
3666 GvSTASH(dstr) = GvSTASH(sstr);
3668 Perl_sv_add_backref(aTHX_ MUTABLE_SV(GvSTASH(dstr)), dstr);
3669 gv_name_set(MUTABLE_GV(dstr), name, len,
3670 GV_ADD | (GvNAMEUTF8(sstr) ? SVf_UTF8 : 0 ));
3671 SvFAKE_on(dstr); /* can coerce to non-glob */
3674 if(GvGP(MUTABLE_GV(sstr))) {
3675 /* If source has method cache entry, clear it */
3677 SvREFCNT_dec(GvCV(sstr));
3678 GvCV_set(sstr, NULL);
3681 /* If source has a real method, then a method is
3684 GvCV((const GV *)sstr) && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3690 /* If dest already had a real method, that's a change as well */
3692 !mro_changes && GvGP(MUTABLE_GV(dstr)) && GvCVu((const GV *)dstr)
3693 && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3698 /* We don’t need to check the name of the destination if it was not a
3699 glob to begin with. */
3700 if(dtype == SVt_PVGV) {
3701 const char * const name = GvNAME((const GV *)dstr);
3704 /* The stash may have been detached from the symbol table, so
3706 && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3707 && GvAV((const GV *)sstr)
3711 const STRLEN len = GvNAMELEN(dstr);
3712 if ((len > 1 && name[len-2] == ':' && name[len-1] == ':')
3713 || (len == 1 && name[0] == ':')) {
3716 /* Set aside the old stash, so we can reset isa caches on
3718 if((old_stash = GvHV(dstr)))
3719 /* Make sure we do not lose it early. */
3720 SvREFCNT_inc_simple_void_NN(
3721 sv_2mortal((SV *)old_stash)
3727 gp_free(MUTABLE_GV(dstr));
3728 isGV_with_GP_off(dstr);
3729 (void)SvOK_off(dstr);
3730 isGV_with_GP_on(dstr);
3731 GvINTRO_off(dstr); /* one-shot flag */
3732 GvGP_set(dstr, gp_ref(GvGP(sstr)));
3733 if (SvTAINTED(sstr))
3735 if (GvIMPORTED(dstr) != GVf_IMPORTED
3736 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3738 GvIMPORTED_on(dstr);
3741 if(mro_changes == 2) {
3743 SV * const sref = (SV *)GvAV((const GV *)dstr);
3744 if (SvSMAGICAL(sref) && (mg = mg_find(sref, PERL_MAGIC_isa))) {
3745 if (SvTYPE(mg->mg_obj) != SVt_PVAV) {
3746 AV * const ary = newAV();
3747 av_push(ary, mg->mg_obj); /* takes the refcount */
3748 mg->mg_obj = (SV *)ary;
3750 av_push((AV *)mg->mg_obj, SvREFCNT_inc_simple_NN(dstr));
3752 else sv_magic(sref, dstr, PERL_MAGIC_isa, NULL, 0);
3753 mro_isa_changed_in(GvSTASH(dstr));
3755 else if(mro_changes == 3) {
3756 HV * const stash = GvHV(dstr);
3757 if(old_stash ? (HV *)HvENAME_get(old_stash) : stash)
3763 else if(mro_changes) mro_method_changed_in(GvSTASH(dstr));
3768 S_glob_assign_ref(pTHX_ SV *const dstr, SV *const sstr)
3770 SV * const sref = SvREFCNT_inc(SvRV(sstr));
3772 const int intro = GvINTRO(dstr);
3775 const U32 stype = SvTYPE(sref);
3777 PERL_ARGS_ASSERT_GLOB_ASSIGN_REF;
3780 GvINTRO_off(dstr); /* one-shot flag */
3781 GvLINE(dstr) = CopLINE(PL_curcop);
3782 GvEGV(dstr) = MUTABLE_GV(dstr);
3787 location = (SV **) &(GvGP(dstr)->gp_cv); /* XXX bypassing GvCV_set */
3788 import_flag = GVf_IMPORTED_CV;
3791 location = (SV **) &GvHV(dstr);
3792 import_flag = GVf_IMPORTED_HV;
3795 location = (SV **) &GvAV(dstr);
3796 import_flag = GVf_IMPORTED_AV;
3799 location = (SV **) &GvIOp(dstr);
3802 location = (SV **) &GvFORM(dstr);
3805 location = &GvSV(dstr);
3806 import_flag = GVf_IMPORTED_SV;
3809 if (stype == SVt_PVCV) {
3810 /*if (GvCVGEN(dstr) && (GvCV(dstr) != (const CV *)sref || GvCVGEN(dstr))) {*/
3811 if (GvCVGEN(dstr)) {
3812 SvREFCNT_dec(GvCV(dstr));
3813 GvCV_set(dstr, NULL);
3814 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3817 SAVEGENERICSV(*location);
3821 if (stype == SVt_PVCV && (*location != sref || GvCVGEN(dstr))) {
3822 CV* const cv = MUTABLE_CV(*location);
3824 if (!GvCVGEN((const GV *)dstr) &&
3825 (CvROOT(cv) || CvXSUB(cv)))
3827 /* Redefining a sub - warning is mandatory if
3828 it was a const and its value changed. */
3829 if (CvCONST(cv) && CvCONST((const CV *)sref)
3831 == cv_const_sv((const CV *)sref)) {
3833 /* They are 2 constant subroutines generated from
3834 the same constant. This probably means that
3835 they are really the "same" proxy subroutine
3836 instantiated in 2 places. Most likely this is
3837 when a constant is exported twice. Don't warn.
3840 else if (ckWARN(WARN_REDEFINE)
3842 && (!CvCONST((const CV *)sref)
3843 || sv_cmp(cv_const_sv(cv),
3844 cv_const_sv((const CV *)
3846 Perl_warner(aTHX_ packWARN(WARN_REDEFINE),
3849 ? "Constant subroutine %s::%s redefined"
3850 : "Subroutine %s::%s redefined"),
3851 HvNAME_get(GvSTASH((const GV *)dstr)),
3852 GvENAME(MUTABLE_GV(dstr)));
3856 cv_ckproto_len(cv, (const GV *)dstr,
3857 SvPOK(sref) ? SvPVX_const(sref) : NULL,
3858 SvPOK(sref) ? SvCUR(sref) : 0);
3860 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3861 GvASSUMECV_on(dstr);
3862 if(GvSTASH(dstr)) mro_method_changed_in(GvSTASH(dstr)); /* sub foo { 1 } sub bar { 2 } *bar = \&foo */
3865 if (import_flag && !(GvFLAGS(dstr) & import_flag)
3866 && CopSTASH_ne(PL_curcop, GvSTASH(dstr))) {
3867 GvFLAGS(dstr) |= import_flag;
3869 if (stype == SVt_PVHV) {
3870 const char * const name = GvNAME((GV*)dstr);
3871 const STRLEN len = GvNAMELEN(dstr);
3874 (len > 1 && name[len-2] == ':' && name[len-1] == ':')
3875 || (len == 1 && name[0] == ':')
3877 && (!dref || HvENAME_get(dref))
3880 (HV *)sref, (HV *)dref,
3886 stype == SVt_PVAV && sref != dref
3887 && strEQ(GvNAME((GV*)dstr), "ISA")
3888 /* The stash may have been detached from the symbol table, so
3889 check its name before doing anything. */
3890 && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3893 MAGIC * const omg = dref && SvSMAGICAL(dref)
3894 ? mg_find(dref, PERL_MAGIC_isa)
3896 if (SvSMAGICAL(sref) && (mg = mg_find(sref, PERL_MAGIC_isa))) {
3897 if (SvTYPE(mg->mg_obj) != SVt_PVAV) {
3898 AV * const ary = newAV();
3899 av_push(ary, mg->mg_obj); /* takes the refcount */
3900 mg->mg_obj = (SV *)ary;
3903 if (SvTYPE(omg->mg_obj) == SVt_PVAV) {
3904 SV **svp = AvARRAY((AV *)omg->mg_obj);
3905 I32 items = AvFILLp((AV *)omg->mg_obj) + 1;
3909 SvREFCNT_inc_simple_NN(*svp++)
3915 SvREFCNT_inc_simple_NN(omg->mg_obj)
3919 av_push((AV *)mg->mg_obj,SvREFCNT_inc_simple_NN(dstr));
3924 sref, omg ? omg->mg_obj : dstr, PERL_MAGIC_isa, NULL, 0
3926 mg = mg_find(sref, PERL_MAGIC_isa);
3928 /* Since the *ISA assignment could have affected more than
3929 one stash, don’t call mro_isa_changed_in directly, but let
3930 magic_clearisa do it for us, as it already has the logic for
3931 dealing with globs vs arrays of globs. */
3933 Perl_magic_clearisa(aTHX_ NULL, mg);
3938 if (SvTAINTED(sstr))
3944 Perl_sv_setsv_flags(pTHX_ SV *dstr, register SV* sstr, const I32 flags)
3947 register U32 sflags;
3949 register svtype stype;
3951 PERL_ARGS_ASSERT_SV_SETSV_FLAGS;
3956 if (SvIS_FREED(dstr)) {
3957 Perl_croak(aTHX_ "panic: attempt to copy value %" SVf
3958 " to a freed scalar %p", SVfARG(sstr), (void *)dstr);
3960 SV_CHECK_THINKFIRST_COW_DROP(dstr);
3962 sstr = &PL_sv_undef;
3963 if (SvIS_FREED(sstr)) {
3964 Perl_croak(aTHX_ "panic: attempt to copy freed scalar %p to %p",
3965 (void*)sstr, (void*)dstr);
3967 stype = SvTYPE(sstr);
3968 dtype = SvTYPE(dstr);
3970 (void)SvAMAGIC_off(dstr);
3973 /* need to nuke the magic */
3977 /* There's a lot of redundancy below but we're going for speed here */
3982 if (dtype != SVt_PVGV && dtype != SVt_PVLV) {
3983 (void)SvOK_off(dstr);
3991 sv_upgrade(dstr, SVt_IV);
3995 sv_upgrade(dstr, SVt_PVIV);
3999 goto end_of_first_switch;
4001 (void)SvIOK_only(dstr);
4002 SvIV_set(dstr, SvIVX(sstr));
4005 /* SvTAINTED can only be true if the SV has taint magic, which in
4006 turn means that the SV type is PVMG (or greater). This is the
4007 case statement for SVt_IV, so this cannot be true (whatever gcov
4009 assert(!SvTAINTED(sstr));
4014 if (dtype < SVt_PV && dtype != SVt_IV)
4015 sv_upgrade(dstr, SVt_IV);
4023 sv_upgrade(dstr, SVt_NV);
4027 sv_upgrade(dstr, SVt_PVNV);
4031 goto end_of_first_switch;
4033 SvNV_set(dstr, SvNVX(sstr));
4034 (void)SvNOK_only(dstr);
4035 /* SvTAINTED can only be true if the SV has taint magic, which in
4036 turn means that the SV type is PVMG (or greater). This is the
4037 case statement for SVt_NV, so this cannot be true (whatever gcov
4039 assert(!SvTAINTED(sstr));
4045 #ifdef PERL_OLD_COPY_ON_WRITE
4046 if ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS) {
4047 if (dtype < SVt_PVIV)
4048 sv_upgrade(dstr, SVt_PVIV);
4055 sv_upgrade(dstr, SVt_PV);
4058 if (dtype < SVt_PVIV)
4059 sv_upgrade(dstr, SVt_PVIV);
4062 if (dtype < SVt_PVNV)
4063 sv_upgrade(dstr, SVt_PVNV);
4067 const char * const type = sv_reftype(sstr,0);
4069 Perl_croak(aTHX_ "Bizarre copy of %s in %s", type, OP_DESC(PL_op));
4071 Perl_croak(aTHX_ "Bizarre copy of %s", type);
4076 if (dtype < SVt_REGEXP)
4077 sv_upgrade(dstr, SVt_REGEXP);
4080 /* case SVt_BIND: */
4084 if (SvGMAGICAL(sstr) && (flags & SV_GMAGIC)) {
4086 if (SvTYPE(sstr) != stype)
4087 stype = SvTYPE(sstr);
4089 if (isGV_with_GP(sstr) && dtype <= SVt_PVLV) {
4090 glob_assign_glob(dstr, sstr, dtype);
4093 if (stype == SVt_PVLV)
4094 SvUPGRADE(dstr, SVt_PVNV);
4096 SvUPGRADE(dstr, (svtype)stype);
4098 end_of_first_switch:
4100 /* dstr may have been upgraded. */
4101 dtype = SvTYPE(dstr);
4102 sflags = SvFLAGS(sstr);
4104 if (dtype == SVt_PVCV || dtype == SVt_PVFM) {
4105 /* Assigning to a subroutine sets the prototype. */
4108 const char *const ptr = SvPV_const(sstr, len);
4110 SvGROW(dstr, len + 1);
4111 Copy(ptr, SvPVX(dstr), len + 1, char);
4112 SvCUR_set(dstr, len);
4114 SvFLAGS(dstr) |= sflags & SVf_UTF8;
4118 } else if (dtype == SVt_PVAV || dtype == SVt_PVHV) {
4119 const char * const type = sv_reftype(dstr,0);
4121 Perl_croak(aTHX_ "Cannot copy to %s in %s", type, OP_DESC(PL_op));
4123 Perl_croak(aTHX_ "Cannot copy to %s", type);
4124 } else if (sflags & SVf_ROK) {
4125 if (isGV_with_GP(dstr)
4126 && SvTYPE(SvRV(sstr)) == SVt_PVGV && isGV_with_GP(SvRV(sstr))) {
4129 if (GvIMPORTED(dstr) != GVf_IMPORTED
4130 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
4132 GvIMPORTED_on(dstr);
4137 glob_assign_glob(dstr, sstr, dtype);
4141 if (dtype >= SVt_PV) {
4142 if (isGV_with_GP(dstr)) {
4143 glob_assign_ref(dstr, sstr);
4146 if (SvPVX_const(dstr)) {
4152 (void)SvOK_off(dstr);
4153 SvRV_set(dstr, SvREFCNT_inc(SvRV(sstr)));
4154 SvFLAGS(dstr) |= sflags & SVf_ROK;
4155 assert(!(sflags & SVp_NOK));
4156 assert(!(sflags & SVp_IOK));
4157 assert(!(sflags & SVf_NOK));
4158 assert(!(sflags & SVf_IOK));
4160 else if (isGV_with_GP(dstr)) {
4161 if (!(sflags & SVf_OK)) {
4162 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
4163 "Undefined value assigned to typeglob");
4166 GV *gv = gv_fetchsv_nomg(sstr, GV_ADD, SVt_PVGV);
4167 if (dstr != (const SV *)gv) {
4168 const char * const name = GvNAME((const GV *)dstr);
4169 const STRLEN len = GvNAMELEN(dstr);
4170 HV *old_stash = NULL;
4171 bool reset_isa = FALSE;
4172 if ((len > 1 && name[len-2] == ':' && name[len-1] == ':')
4173 || (len == 1 && name[0] == ':')) {
4174 /* Set aside the old stash, so we can reset isa caches
4175 on its subclasses. */
4176 if((old_stash = GvHV(dstr))) {
4177 /* Make sure we do not lose it early. */
4178 SvREFCNT_inc_simple_void_NN(
4179 sv_2mortal((SV *)old_stash)
4186 gp_free(MUTABLE_GV(dstr));
4187 GvGP_set(dstr, gp_ref(GvGP(gv)));
4190 HV * const stash = GvHV(dstr);
4192 old_stash ? (HV *)HvENAME_get(old_stash) : stash
4202 else if (dtype == SVt_REGEXP && stype == SVt_REGEXP) {
4203 reg_temp_copy((REGEXP*)dstr, (REGEXP*)sstr);
4205 else if (sflags & SVp_POK) {
4209 * Check to see if we can just swipe the string. If so, it's a
4210 * possible small lose on short strings, but a big win on long ones.
4211 * It might even be a win on short strings if SvPVX_const(dstr)
4212 * has to be allocated and SvPVX_const(sstr) has to be freed.
4213 * Likewise if we can set up COW rather than doing an actual copy, we
4214 * drop to the else clause, as the swipe code and the COW setup code
4215 * have much in common.
4218 /* Whichever path we take through the next code, we want this true,
4219 and doing it now facilitates the COW check. */
4220 (void)SvPOK_only(dstr);
4223 /* If we're already COW then this clause is not true, and if COW
4224 is allowed then we drop down to the else and make dest COW
4225 with us. If caller hasn't said that we're allowed to COW
4226 shared hash keys then we don't do the COW setup, even if the
4227 source scalar is a shared hash key scalar. */
4228 (((flags & SV_COW_SHARED_HASH_KEYS)
4229 ? (sflags & (SVf_FAKE|SVf_READONLY)) != (SVf_FAKE|SVf_READONLY)
4230 : 1 /* If making a COW copy is forbidden then the behaviour we
4231 desire is as if the source SV isn't actually already
4232 COW, even if it is. So we act as if the source flags
4233 are not COW, rather than actually testing them. */
4235 #ifndef PERL_OLD_COPY_ON_WRITE
4236 /* The change that added SV_COW_SHARED_HASH_KEYS makes the logic
4237 when PERL_OLD_COPY_ON_WRITE is defined a little wrong.
4238 Conceptually PERL_OLD_COPY_ON_WRITE being defined should
4239 override SV_COW_SHARED_HASH_KEYS, because it means "always COW"
4240 but in turn, it's somewhat dead code, never expected to go
4241 live, but more kept as a placeholder on how to do it better
4242 in a newer implementation. */
4243 /* If we are COW and dstr is a suitable target then we drop down
4244 into the else and make dest a COW of us. */
4245 || (SvFLAGS(dstr) & CAN_COW_MASK) != CAN_COW_FLAGS
4250 (sflags & SVs_TEMP) && /* slated for free anyway? */
4251 !(sflags & SVf_OOK) && /* and not involved in OOK hack? */
4252 (!(flags & SV_NOSTEAL)) &&
4253 /* and we're allowed to steal temps */
4254 SvREFCNT(sstr) == 1 && /* and no other references to it? */
4255 SvLEN(sstr)) /* and really is a string */
4256 #ifdef PERL_OLD_COPY_ON_WRITE
4257 && ((flags & SV_COW_SHARED_HASH_KEYS)
4258 ? (!((sflags & CAN_COW_MASK) == CAN_COW_FLAGS
4259 && (SvFLAGS(dstr) & CAN_COW_MASK) == CAN_COW_FLAGS
4260 && SvTYPE(sstr) >= SVt_PVIV && SvTYPE(sstr) != SVt_PVFM))
4264 /* Failed the swipe test, and it's not a shared hash key either.
4265 Have to copy the string. */
4266 STRLEN len = SvCUR(sstr);
4267 SvGROW(dstr, len + 1); /* inlined from sv_setpvn */
4268 Move(SvPVX_const(sstr),SvPVX(dstr),len,char);
4269 SvCUR_set(dstr, len);
4270 *SvEND(dstr) = '\0';
4272 /* If PERL_OLD_COPY_ON_WRITE is not defined, then isSwipe will always
4274 /* Either it's a shared hash key, or it's suitable for
4275 copy-on-write or we can swipe the string. */
4277 PerlIO_printf(Perl_debug_log, "Copy on write: sstr --> dstr\n");
4281 #ifdef PERL_OLD_COPY_ON_WRITE
4283 if ((sflags & (SVf_FAKE | SVf_READONLY))
4284 != (SVf_FAKE | SVf_READONLY)) {
4285 SvREADONLY_on(sstr);
4287 /* Make the source SV into a loop of 1.
4288 (about to become 2) */
4289 SV_COW_NEXT_SV_SET(sstr, sstr);
4293 /* Initial code is common. */
4294 if (SvPVX_const(dstr)) { /* we know that dtype >= SVt_PV */
4299 /* making another shared SV. */
4300 STRLEN cur = SvCUR(sstr);
4301 STRLEN len = SvLEN(sstr);
4302 #ifdef PERL_OLD_COPY_ON_WRITE
4304 assert (SvTYPE(dstr) >= SVt_PVIV);
4305 /* SvIsCOW_normal */
4306 /* splice us in between source and next-after-source. */
4307 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4308 SV_COW_NEXT_SV_SET(sstr, dstr);
4309 SvPV_set(dstr, SvPVX_mutable(sstr));
4313 /* SvIsCOW_shared_hash */
4314 DEBUG_C(PerlIO_printf(Perl_debug_log,
4315 "Copy on write: Sharing hash\n"));
4317 assert (SvTYPE(dstr) >= SVt_PV);
4319 HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)))));
4321 SvLEN_set(dstr, len);
4322 SvCUR_set(dstr, cur);
4323 SvREADONLY_on(dstr);
4327 { /* Passes the swipe test. */
4328 SvPV_set(dstr, SvPVX_mutable(sstr));
4329 SvLEN_set(dstr, SvLEN(sstr));
4330 SvCUR_set(dstr, SvCUR(sstr));
4333 (void)SvOK_off(sstr); /* NOTE: nukes most SvFLAGS on sstr */
4334 SvPV_set(sstr, NULL);
4340 if (sflags & SVp_NOK) {
4341 SvNV_set(dstr, SvNVX(sstr));
4343 if (sflags & SVp_IOK) {
4344 SvIV_set(dstr, SvIVX(sstr));
4345 /* Must do this otherwise some other overloaded use of 0x80000000
4346 gets confused. I guess SVpbm_VALID */
4347 if (sflags & SVf_IVisUV)
4350 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_NOK|SVp_NOK|SVf_UTF8);
4352 const MAGIC * const smg = SvVSTRING_mg(sstr);
4354 sv_magic(dstr, NULL, PERL_MAGIC_vstring,
4355 smg->mg_ptr, smg->mg_len);
4356 SvRMAGICAL_on(dstr);
4360 else if (sflags & (SVp_IOK|SVp_NOK)) {
4361 (void)SvOK_off(dstr);
4362 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_IVisUV|SVf_NOK|SVp_NOK);
4363 if (sflags & SVp_IOK) {
4364 /* XXXX Do we want to set IsUV for IV(ROK)? Be extra safe... */
4365 SvIV_set(dstr, SvIVX(sstr));
4367 if (sflags & SVp_NOK) {
4368 SvNV_set(dstr, SvNVX(sstr));
4372 if (isGV_with_GP(sstr)) {
4373 /* This stringification rule for globs is spread in 3 places.
4374 This feels bad. FIXME. */
4375 const U32 wasfake = sflags & SVf_FAKE;
4377 /* FAKE globs can get coerced, so need to turn this off
4378 temporarily if it is on. */
4380 gv_efullname3(dstr, MUTABLE_GV(sstr), "*");
4381 SvFLAGS(sstr) |= wasfake;
4384 (void)SvOK_off(dstr);
4386 if (SvTAINTED(sstr))
4391 =for apidoc sv_setsv_mg
4393 Like C<sv_setsv>, but also handles 'set' magic.
4399 Perl_sv_setsv_mg(pTHX_ SV *const dstr, register SV *const sstr)
4401 PERL_ARGS_ASSERT_SV_SETSV_MG;
4403 sv_setsv(dstr,sstr);
4407 #ifdef PERL_OLD_COPY_ON_WRITE
4409 Perl_sv_setsv_cow(pTHX_ SV *dstr, SV *sstr)
4411 STRLEN cur = SvCUR(sstr);
4412 STRLEN len = SvLEN(sstr);
4413 register char *new_pv;
4415 PERL_ARGS_ASSERT_SV_SETSV_COW;
4418 PerlIO_printf(Perl_debug_log, "Fast copy on write: %p -> %p\n",
4419 (void*)sstr, (void*)dstr);
4426 if (SvTHINKFIRST(dstr))
4427 sv_force_normal_flags(dstr, SV_COW_DROP_PV);
4428 else if (SvPVX_const(dstr))
4429 Safefree(SvPVX_const(dstr));
4433 SvUPGRADE(dstr, SVt_PVIV);
4435 assert (SvPOK(sstr));
4436 assert (SvPOKp(sstr));
4437 assert (!SvIOK(sstr));
4438 assert (!SvIOKp(sstr));
4439 assert (!SvNOK(sstr));
4440 assert (!SvNOKp(sstr));
4442 if (SvIsCOW(sstr)) {
4444 if (SvLEN(sstr) == 0) {
4445 /* source is a COW shared hash key. */
4446 DEBUG_C(PerlIO_printf(Perl_debug_log,
4447 "Fast copy on write: Sharing hash\n"));
4448 new_pv = HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr))));
4451 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4453 assert ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS);
4454 SvUPGRADE(sstr, SVt_PVIV);
4455 SvREADONLY_on(sstr);
4457 DEBUG_C(PerlIO_printf(Perl_debug_log,
4458 "Fast copy on write: Converting sstr to COW\n"));
4459 SV_COW_NEXT_SV_SET(dstr, sstr);
4461 SV_COW_NEXT_SV_SET(sstr, dstr);
4462 new_pv = SvPVX_mutable(sstr);
4465 SvPV_set(dstr, new_pv);
4466 SvFLAGS(dstr) = (SVt_PVIV|SVf_POK|SVp_POK|SVf_FAKE|SVf_READONLY);
4469 SvLEN_set(dstr, len);
4470 SvCUR_set(dstr, cur);
4479 =for apidoc sv_setpvn
4481 Copies a string into an SV. The C<len> parameter indicates the number of
4482 bytes to be copied. If the C<ptr> argument is NULL the SV will become
4483 undefined. Does not handle 'set' magic. See C<sv_setpvn_mg>.
4489 Perl_sv_setpvn(pTHX_ register SV *const sv, register const char *const ptr, register const STRLEN len)
4492 register char *dptr;
4494 PERL_ARGS_ASSERT_SV_SETPVN;
4496 SV_CHECK_THINKFIRST_COW_DROP(sv);
4502 /* len is STRLEN which is unsigned, need to copy to signed */
4505 Perl_croak(aTHX_ "panic: sv_setpvn called with negative strlen");
4507 SvUPGRADE(sv, SVt_PV);
4509 dptr = SvGROW(sv, len + 1);
4510 Move(ptr,dptr,len,char);
4513 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4518 =for apidoc sv_setpvn_mg
4520 Like C<sv_setpvn>, but also handles 'set' magic.
4526 Perl_sv_setpvn_mg(pTHX_ register SV *const sv, register const char *const ptr, register const STRLEN len)
4528 PERL_ARGS_ASSERT_SV_SETPVN_MG;
4530 sv_setpvn(sv,ptr,len);
4535 =for apidoc sv_setpv
4537 Copies a string into an SV. The string must be null-terminated. Does not
4538 handle 'set' magic. See C<sv_setpv_mg>.
4544 Perl_sv_setpv(pTHX_ register SV *const sv, register const char *const ptr)
4547 register STRLEN len;
4549 PERL_ARGS_ASSERT_SV_SETPV;
4551 SV_CHECK_THINKFIRST_COW_DROP(sv);
4557 SvUPGRADE(sv, SVt_PV);
4559 SvGROW(sv, len + 1);
4560 Move(ptr,SvPVX(sv),len+1,char);
4562 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4567 =for apidoc sv_setpv_mg
4569 Like C<sv_setpv>, but also handles 'set' magic.
4575 Perl_sv_setpv_mg(pTHX_ register SV *const sv, register const char *const ptr)
4577 PERL_ARGS_ASSERT_SV_SETPV_MG;
4584 =for apidoc sv_usepvn_flags
4586 Tells an SV to use C<ptr> to find its string value. Normally the
4587 string is stored inside the SV but sv_usepvn allows the SV to use an
4588 outside string. The C<ptr> should point to memory that was allocated
4589 by C<malloc>. The string length, C<len>, must be supplied. By default
4590 this function will realloc (i.e. move) the memory pointed to by C<ptr>,
4591 so that pointer should not be freed or used by the programmer after
4592 giving it to sv_usepvn, and neither should any pointers from "behind"
4593 that pointer (e.g. ptr + 1) be used.
4595 If C<flags> & SV_SMAGIC is true, will call SvSETMAGIC. If C<flags> &
4596 SV_HAS_TRAILING_NUL is true, then C<ptr[len]> must be NUL, and the realloc
4597 will be skipped. (i.e. the buffer is actually at least 1 byte longer than
4598 C<len>, and already meets the requirements for storing in C<SvPVX>)
4604 Perl_sv_usepvn_flags(pTHX_ SV *const sv, char *ptr, const STRLEN len, const U32 flags)
4609 PERL_ARGS_ASSERT_SV_USEPVN_FLAGS;
4611 SV_CHECK_THINKFIRST_COW_DROP(sv);
4612 SvUPGRADE(sv, SVt_PV);
4615 if (flags & SV_SMAGIC)
4619 if (SvPVX_const(sv))
4623 if (flags & SV_HAS_TRAILING_NUL)
4624 assert(ptr[len] == '\0');
4627 allocate = (flags & SV_HAS_TRAILING_NUL)
4629 #ifdef Perl_safesysmalloc_size
4632 PERL_STRLEN_ROUNDUP(len + 1);
4634 if (flags & SV_HAS_TRAILING_NUL) {
4635 /* It's long enough - do nothing.
4636 Specifically Perl_newCONSTSUB is relying on this. */
4639 /* Force a move to shake out bugs in callers. */
4640 char *new_ptr = (char*)safemalloc(allocate);
4641 Copy(ptr, new_ptr, len, char);
4642 PoisonFree(ptr,len,char);
4646 ptr = (char*) saferealloc (ptr, allocate);
4649 #ifdef Perl_safesysmalloc_size
4650 SvLEN_set(sv, Perl_safesysmalloc_size(ptr));
4652 SvLEN_set(sv, allocate);
4656 if (!(flags & SV_HAS_TRAILING_NUL)) {
4659 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4661 if (flags & SV_SMAGIC)
4665 #ifdef PERL_OLD_COPY_ON_WRITE
4666 /* Need to do this *after* making the SV normal, as we need the buffer
4667 pointer to remain valid until after we've copied it. If we let go too early,
4668 another thread could invalidate it by unsharing last of the same hash key
4669 (which it can do by means other than releasing copy-on-write Svs)
4670 or by changing the other copy-on-write SVs in the loop. */
4672 S_sv_release_COW(pTHX_ register SV *sv, const char *pvx, SV *after)
4674 PERL_ARGS_ASSERT_SV_RELEASE_COW;
4676 { /* this SV was SvIsCOW_normal(sv) */
4677 /* we need to find the SV pointing to us. */
4678 SV *current = SV_COW_NEXT_SV(after);
4680 if (current == sv) {
4681 /* The SV we point to points back to us (there were only two of us
4683 Hence other SV is no longer copy on write either. */
4685 SvREADONLY_off(after);
4687 /* We need to follow the pointers around the loop. */
4689 while ((next = SV_COW_NEXT_SV(current)) != sv) {
4692 /* don't loop forever if the structure is bust, and we have
4693 a pointer into a closed loop. */
4694 assert (current != after);
4695 assert (SvPVX_const(current) == pvx);
4697 /* Make the SV before us point to the SV after us. */
4698 SV_COW_NEXT_SV_SET(current, after);
4704 =for apidoc sv_force_normal_flags
4706 Undo various types of fakery on an SV: if the PV is a shared string, make
4707 a private copy; if we're a ref, stop refing; if we're a glob, downgrade to
4708 an xpvmg; if we're a copy-on-write scalar, this is the on-write time when
4709 we do the copy, and is also used locally. If C<SV_COW_DROP_PV> is set
4710 then a copy-on-write scalar drops its PV buffer (if any) and becomes
4711 SvPOK_off rather than making a copy. (Used where this scalar is about to be
4712 set to some other value.) In addition, the C<flags> parameter gets passed to
4713 C<sv_unref_flags()> when unreffing. C<sv_force_normal> calls this function
4714 with flags set to 0.
4720 Perl_sv_force_normal_flags(pTHX_ register SV *const sv, const U32 flags)
4724 PERL_ARGS_ASSERT_SV_FORCE_NORMAL_FLAGS;
4726 #ifdef PERL_OLD_COPY_ON_WRITE
4727 if (SvREADONLY(sv)) {
4729 const char * const pvx = SvPVX_const(sv);
4730 const STRLEN len = SvLEN(sv);
4731 const STRLEN cur = SvCUR(sv);
4732 /* next COW sv in the loop. If len is 0 then this is a shared-hash
4733 key scalar, so we mustn't attempt to call SV_COW_NEXT_SV(), as
4734 we'll fail an assertion. */
4735 SV * const next = len ? SV_COW_NEXT_SV(sv) : 0;
4738 PerlIO_printf(Perl_debug_log,
4739 "Copy on write: Force normal %ld\n",
4745 /* This SV doesn't own the buffer, so need to Newx() a new one: */
4748 if (flags & SV_COW_DROP_PV) {
4749 /* OK, so we don't need to copy our buffer. */
4752 SvGROW(sv, cur + 1);
4753 Move(pvx,SvPVX(sv),cur,char);
4758 sv_release_COW(sv, pvx, next);
4760 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4766 else if (IN_PERL_RUNTIME)
4767 Perl_croak_no_modify(aTHX);
4770 if (SvREADONLY(sv)) {
4771 if (SvFAKE(sv) && !isGV_with_GP(sv)) {
4772 const char * const pvx = SvPVX_const(sv);
4773 const STRLEN len = SvCUR(sv);
4778 SvGROW(sv, len + 1);
4779 Move(pvx,SvPVX(sv),len,char);
4781 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4783 else if (IN_PERL_RUNTIME)
4784 Perl_croak_no_modify(aTHX);
4788 sv_unref_flags(sv, flags);
4789 else if (SvFAKE(sv) && isGV_with_GP(sv))
4791 else if (SvFAKE(sv) && SvTYPE(sv) == SVt_REGEXP) {
4792 /* Need to downgrade the REGEXP to a simple(r) scalar. This is analogous
4793 to sv_unglob. We only need it here, so inline it. */
4794 const svtype new_type = SvMAGIC(sv) || SvSTASH(sv) ? SVt_PVMG : SVt_PV;
4795 SV *const temp = newSV_type(new_type);
4796 void *const temp_p = SvANY(sv);
4798 if (new_type == SVt_PVMG) {
4799 SvMAGIC_set(temp, SvMAGIC(sv));
4800 SvMAGIC_set(sv, NULL);
4801 SvSTASH_set(temp, SvSTASH(sv));
4802 SvSTASH_set(sv, NULL);
4804 SvCUR_set(temp, SvCUR(sv));
4805 /* Remember that SvPVX is in the head, not the body. */
4807 SvLEN_set(temp, SvLEN(sv));
4808 /* This signals "buffer is owned by someone else" in sv_clear,
4809 which is the least effort way to stop it freeing the buffer.
4811 SvLEN_set(sv, SvLEN(sv)+1);
4813 /* Their buffer is already owned by someone else. */
4814 SvPVX(sv) = savepvn(SvPVX(sv), SvCUR(sv));
4815 SvLEN_set(temp, SvCUR(sv)+1);
4818 /* Now swap the rest of the bodies. */
4820 SvFLAGS(sv) &= ~(SVf_FAKE|SVTYPEMASK);
4821 SvFLAGS(sv) |= new_type;
4822 SvANY(sv) = SvANY(temp);
4824 SvFLAGS(temp) &= ~(SVTYPEMASK);
4825 SvFLAGS(temp) |= SVt_REGEXP|SVf_FAKE;
4826 SvANY(temp) = temp_p;
4835 Efficient removal of characters from the beginning of the string buffer.
4836 SvPOK(sv) must be true and the C<ptr> must be a pointer to somewhere inside
4837 the string buffer. The C<ptr> becomes the first character of the adjusted
4838 string. Uses the "OOK hack".
4839 Beware: after this function returns, C<ptr> and SvPVX_const(sv) may no longer
4840 refer to the same chunk of data.
4846 Perl_sv_chop(pTHX_ register SV *const sv, register const char *const ptr)
4852 const U8 *real_start;
4856 PERL_ARGS_ASSERT_SV_CHOP;
4858 if (!ptr || !SvPOKp(sv))
4860 delta = ptr - SvPVX_const(sv);
4862 /* Nothing to do. */
4865 /* SvPVX(sv) may move in SV_CHECK_THINKFIRST(sv), but after this line,
4866 nothing uses the value of ptr any more. */
4867 max_delta = SvLEN(sv) ? SvLEN(sv) : SvCUR(sv);
4868 if (ptr <= SvPVX_const(sv))
4869 Perl_croak(aTHX_ "panic: sv_chop ptr=%p, start=%p, end=%p",
4870 ptr, SvPVX_const(sv), SvPVX_const(sv) + max_delta);
4871 SV_CHECK_THINKFIRST(sv);
4872 if (delta > max_delta)
4873 Perl_croak(aTHX_ "panic: sv_chop ptr=%p (was %p), start=%p, end=%p",
4874 SvPVX_const(sv) + delta, ptr, SvPVX_const(sv),
4875 SvPVX_const(sv) + max_delta);
4878 if (!SvLEN(sv)) { /* make copy of shared string */
4879 const char *pvx = SvPVX_const(sv);
4880 const STRLEN len = SvCUR(sv);
4881 SvGROW(sv, len + 1);
4882 Move(pvx,SvPVX(sv),len,char);
4885 SvFLAGS(sv) |= SVf_OOK;
4888 SvOOK_offset(sv, old_delta);
4890 SvLEN_set(sv, SvLEN(sv) - delta);
4891 SvCUR_set(sv, SvCUR(sv) - delta);
4892 SvPV_set(sv, SvPVX(sv) + delta);
4894 p = (U8 *)SvPVX_const(sv);
4899 real_start = p - delta;
4903 if (delta < 0x100) {
4907 p -= sizeof(STRLEN);
4908 Copy((U8*)&delta, p, sizeof(STRLEN), U8);
4912 /* Fill the preceding buffer with sentinals to verify that no-one is
4914 while (p > real_start) {
4922 =for apidoc sv_catpvn
4924 Concatenates the string onto the end of the string which is in the SV. The
4925 C<len> indicates number of bytes to copy. If the SV has the UTF-8
4926 status set, then the bytes appended should be valid UTF-8.
4927 Handles 'get' magic, but not 'set' magic. See C<sv_catpvn_mg>.
4929 =for apidoc sv_catpvn_flags
4931 Concatenates the string onto the end of the string which is in the SV. The
4932 C<len> indicates number of bytes to copy. If the SV has the UTF-8
4933 status set, then the bytes appended should be valid UTF-8.
4934 If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<dsv> if
4935 appropriate, else not. C<sv_catpvn> and C<sv_catpvn_nomg> are implemented
4936 in terms of this function.
4942 Perl_sv_catpvn_flags(pTHX_ register SV *const dsv, register const char *sstr, register const STRLEN slen, const I32 flags)
4946 const char * const dstr = SvPV_force_flags(dsv, dlen, flags);
4948 PERL_ARGS_ASSERT_SV_CATPVN_FLAGS;
4950 SvGROW(dsv, dlen + slen + 1);
4952 sstr = SvPVX_const(dsv);
4953 Move(sstr, SvPVX(dsv) + dlen, slen, char);
4954 SvCUR_set(dsv, SvCUR(dsv) + slen);
4956 (void)SvPOK_only_UTF8(dsv); /* validate pointer */
4958 if (flags & SV_SMAGIC)
4963 =for apidoc sv_catsv
4965 Concatenates the string from SV C<ssv> onto the end of the string in
4966 SV C<dsv>. Modifies C<dsv> but not C<ssv>. Handles 'get' magic, but
4967 not 'set' magic. See C<sv_catsv_mg>.
4969 =for apidoc sv_catsv_flags
4971 Concatenates the string from SV C<ssv> onto the end of the string in
4972 SV C<dsv>. Modifies C<dsv> but not C<ssv>. If C<flags> has C<SV_GMAGIC>
4973 bit set, will C<mg_get> on the SVs if appropriate, else not. C<sv_catsv>
4974 and C<sv_catsv_nomg> are implemented in terms of this function.
4979 Perl_sv_catsv_flags(pTHX_ SV *const dsv, register SV *const ssv, const I32 flags)
4983 PERL_ARGS_ASSERT_SV_CATSV_FLAGS;
4987 const char *spv = SvPV_flags_const(ssv, slen, flags);
4989 /* sutf8 and dutf8 were type bool, but under USE_ITHREADS,
4990 gcc version 2.95.2 20000220 (Debian GNU/Linux) for
4991 Linux xxx 2.2.17 on sparc64 with gcc -O2, we erroneously
4992 get dutf8 = 0x20000000, (i.e. SVf_UTF8) even though
4993 dsv->sv_flags doesn't have that bit set.
4994 Andy Dougherty 12 Oct 2001
4996 const I32 sutf8 = DO_UTF8(ssv);
4999 if (SvGMAGICAL(dsv) && (flags & SV_GMAGIC))
5001 dutf8 = DO_UTF8(dsv);
5003 if (dutf8 != sutf8) {
5005 /* Not modifying source SV, so taking a temporary copy. */
5006 SV* const csv = newSVpvn_flags(spv, slen, SVs_TEMP);
5008 sv_utf8_upgrade(csv);
5009 spv = SvPV_const(csv, slen);
5012 /* Leave enough space for the cat that's about to happen */
5013 sv_utf8_upgrade_flags_grow(dsv, 0, slen);
5015 sv_catpvn_nomg(dsv, spv, slen);
5018 if (flags & SV_SMAGIC)
5023 =for apidoc sv_catpv
5025 Concatenates the string onto the end of the string which is in the SV.
5026 If the SV has the UTF-8 status set, then the bytes appended should be
5027 valid UTF-8. Handles 'get' magic, but not 'set' magic. See C<sv_catpv_mg>.
5032 Perl_sv_catpv(pTHX_ register SV *const sv, register const char *ptr)
5035 register STRLEN len;
5039 PERL_ARGS_ASSERT_SV_CATPV;
5043 junk = SvPV_force(sv, tlen);
5045 SvGROW(sv, tlen + len + 1);
5047 ptr = SvPVX_const(sv);
5048 Move(ptr,SvPVX(sv)+tlen,len+1,char);
5049 SvCUR_set(sv, SvCUR(sv) + len);
5050 (void)SvPOK_only_UTF8(sv); /* validate pointer */
5055 =for apidoc sv_catpv_flags
5057 Concatenates the string onto the end of the string which is in the SV.
5058 If the SV has the UTF-8 status set, then the bytes appended should
5059 be valid UTF-8. If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get>
5060 on the SVs if appropriate, else not.
5066 Perl_sv_catpv_flags(pTHX_ SV *dstr, const char *sstr, const I32 flags)
5068 PERL_ARGS_ASSERT_SV_CATPV_FLAGS;
5069 sv_catpvn_flags(dstr, sstr, strlen(sstr), flags);
5073 =for apidoc sv_catpv_mg
5075 Like C<sv_catpv>, but also handles 'set' magic.
5081 Perl_sv_catpv_mg(pTHX_ register SV *const sv, register const char *const ptr)
5083 PERL_ARGS_ASSERT_SV_CATPV_MG;
5092 Creates a new SV. A non-zero C<len> parameter indicates the number of
5093 bytes of preallocated string space the SV should have. An extra byte for a
5094 trailing NUL is also reserved. (SvPOK is not set for the SV even if string
5095 space is allocated.) The reference count for the new SV is set to 1.
5097 In 5.9.3, newSV() replaces the older NEWSV() API, and drops the first
5098 parameter, I<x>, a debug aid which allowed callers to identify themselves.
5099 This aid has been superseded by a new build option, PERL_MEM_LOG (see
5100 L<perlhacktips/PERL_MEM_LOG>). The older API is still there for use in XS
5101 modules supporting older perls.
5107 Perl_newSV(pTHX_ const STRLEN len)
5114 sv_upgrade(sv, SVt_PV);
5115 SvGROW(sv, len + 1);
5120 =for apidoc sv_magicext
5122 Adds magic to an SV, upgrading it if necessary. Applies the
5123 supplied vtable and returns a pointer to the magic added.
5125 Note that C<sv_magicext> will allow things that C<sv_magic> will not.
5126 In particular, you can add magic to SvREADONLY SVs, and add more than
5127 one instance of the same 'how'.
5129 If C<namlen> is greater than zero then a C<savepvn> I<copy> of C<name> is
5130 stored, if C<namlen> is zero then C<name> is stored as-is and - as another
5131 special case - if C<(name && namlen == HEf_SVKEY)> then C<name> is assumed
5132 to contain an C<SV*> and is stored as-is with its REFCNT incremented.
5134 (This is now used as a subroutine by C<sv_magic>.)
5139 Perl_sv_magicext(pTHX_ SV *const sv, SV *const obj, const int how,
5140 const MGVTBL *const vtable, const char *const name, const I32 namlen)
5145 PERL_ARGS_ASSERT_SV_MAGICEXT;
5147 SvUPGRADE(sv, SVt_PVMG);
5148 Newxz(mg, 1, MAGIC);
5149 mg->mg_moremagic = SvMAGIC(sv);
5150 SvMAGIC_set(sv, mg);
5152 /* Sometimes a magic contains a reference loop, where the sv and
5153 object refer to each other. To prevent a reference loop that
5154 would prevent such objects being freed, we look for such loops
5155 and if we find one we avoid incrementing the object refcount.
5157 Note we cannot do this to avoid self-tie loops as intervening RV must
5158 have its REFCNT incremented to keep it in existence.
5161 if (!obj || obj == sv ||
5162 how == PERL_MAGIC_arylen ||
5163 how == PERL_MAGIC_symtab ||
5164 (SvTYPE(obj) == SVt_PVGV &&
5165 (GvSV(obj) == sv || GvHV(obj) == (const HV *)sv
5166 || GvAV(obj) == (const AV *)sv || GvCV(obj) == (const CV *)sv
5167 || GvIOp(obj) == (const IO *)sv || GvFORM(obj) == (const CV *)sv)))
5172 mg->mg_obj = SvREFCNT_inc_simple(obj);
5173 mg->mg_flags |= MGf_REFCOUNTED;
5176 /* Normal self-ties simply pass a null object, and instead of
5177 using mg_obj directly, use the SvTIED_obj macro to produce a
5178 new RV as needed. For glob "self-ties", we are tieing the PVIO
5179 with an RV obj pointing to the glob containing the PVIO. In
5180 this case, to avoid a reference loop, we need to weaken the
5184 if (how == PERL_MAGIC_tiedscalar && SvTYPE(sv) == SVt_PVIO &&
5185 obj && SvROK(obj) && GvIO(SvRV(obj)) == (const IO *)sv)
5191 mg->mg_len = namlen;
5194 mg->mg_ptr = savepvn(name, namlen);
5195 else if (namlen == HEf_SVKEY) {
5196 /* Yes, this is casting away const. This is only for the case of
5197 HEf_SVKEY. I think we need to document this aberation of the
5198 constness of the API, rather than making name non-const, as
5199 that change propagating outwards a long way. */
5200 mg->mg_ptr = (char*)SvREFCNT_inc_simple_NN((SV *)name);
5202 mg->mg_ptr = (char *) name;
5204 mg->mg_virtual = (MGVTBL *) vtable;
5208 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
5213 =for apidoc sv_magic
5215 Adds magic to an SV. First upgrades C<sv> to type C<SVt_PVMG> if necessary,
5216 then adds a new magic item of type C<how> to the head of the magic list.
5218 See C<sv_magicext> (which C<sv_magic> now calls) for a description of the
5219 handling of the C<name> and C<namlen> arguments.
5221 You need to use C<sv_magicext> to add magic to SvREADONLY SVs and also
5222 to add more than one instance of the same 'how'.
5228 Perl_sv_magic(pTHX_ register SV *const sv, SV *const obj, const int how,
5229 const char *const name, const I32 namlen)
5232 const MGVTBL *vtable;
5235 unsigned int vtable_index;
5237 PERL_ARGS_ASSERT_SV_MAGIC;
5239 if (how < 0 || (unsigned)how > C_ARRAY_LENGTH(PL_magic_data)
5240 || ((flags = PL_magic_data[how]),
5241 (vtable_index = flags & PERL_MAGIC_VTABLE_MASK)
5242 > magic_vtable_max))
5243 Perl_croak(aTHX_ "Don't know how to handle magic of type \\%o", how);
5245 /* PERL_MAGIC_ext is reserved for use by extensions not perl internals.
5246 Useful for attaching extension internal data to perl vars.
5247 Note that multiple extensions may clash if magical scalars
5248 etc holding private data from one are passed to another. */
5250 vtable = (vtable_index == magic_vtable_max)
5251 ? NULL : PL_magic_vtables + vtable_index;
5253 #ifdef PERL_OLD_COPY_ON_WRITE
5255 sv_force_normal_flags(sv, 0);
5257 if (SvREADONLY(sv)) {
5259 /* its okay to attach magic to shared strings; the subsequent
5260 * upgrade to PVMG will unshare the string */
5261 !(SvFAKE(sv) && SvTYPE(sv) < SVt_PVMG)
5264 && !PERL_MAGIC_TYPE_READONLY_ACCEPTABLE(how)
5267 Perl_croak_no_modify(aTHX);
5270 if (SvMAGICAL(sv) || (how == PERL_MAGIC_taint && SvTYPE(sv) >= SVt_PVMG)) {
5271 if (SvMAGIC(sv) && (mg = mg_find(sv, how))) {
5272 /* sv_magic() refuses to add a magic of the same 'how' as an
5275 if (how == PERL_MAGIC_taint) {
5277 /* Any scalar which already had taint magic on which someone
5278 (erroneously?) did SvIOK_on() or similar will now be
5279 incorrectly sporting public "OK" flags. */
5280 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
5286 /* Rest of work is done else where */
5287 mg = sv_magicext(sv,obj,how,vtable,name,namlen);
5290 case PERL_MAGIC_taint:
5293 case PERL_MAGIC_ext:
5294 case PERL_MAGIC_dbfile:
5301 S_sv_unmagicext_flags(pTHX_ SV *const sv, const int type, MGVTBL *vtbl, const U32 flags)
5308 if (SvTYPE(sv) < SVt_PVMG || !SvMAGIC(sv))
5310 mgp = &(((XPVMG*) SvANY(sv))->xmg_u.xmg_magic);
5311 for (mg = *mgp; mg; mg = *mgp) {
5312 const MGVTBL* const virt = mg->mg_virtual;
5313 if (mg->mg_type == type && (!flags || virt == vtbl)) {
5314 *mgp = mg->mg_moremagic;
5315 if (virt && virt->svt_free)
5316 virt->svt_free(aTHX_ sv, mg);
5317 if (mg->mg_ptr && mg->mg_type != PERL_MAGIC_regex_global) {
5319 Safefree(mg->mg_ptr);
5320 else if (mg->mg_len == HEf_SVKEY)
5321 SvREFCNT_dec(MUTABLE_SV(mg->mg_ptr));
5322 else if (mg->mg_type == PERL_MAGIC_utf8)
5323 Safefree(mg->mg_ptr);
5325 if (mg->mg_flags & MGf_REFCOUNTED)
5326 SvREFCNT_dec(mg->mg_obj);
5330 mgp = &mg->mg_moremagic;
5333 if (SvMAGICAL(sv)) /* if we're under save_magic, wait for restore_magic; */
5334 mg_magical(sv); /* else fix the flags now */
5338 SvFLAGS(sv) |= (SvFLAGS(sv) & (SVp_IOK|SVp_NOK|SVp_POK)) >> PRIVSHIFT;
5344 =for apidoc sv_unmagic
5346 Removes all magic of type C<type> from an SV.
5352 Perl_sv_unmagic(pTHX_ SV *const sv, const int type)
5354 PERL_ARGS_ASSERT_SV_UNMAGIC;
5355 return S_sv_unmagicext_flags(aTHX_ sv, type, NULL, 0);
5359 =for apidoc sv_unmagicext
5361 Removes all magic of type C<type> with the specified C<vtbl> from an SV.
5367 Perl_sv_unmagicext(pTHX_ SV *const sv, const int type, MGVTBL *vtbl)
5369 PERL_ARGS_ASSERT_SV_UNMAGICEXT;
5370 return S_sv_unmagicext_flags(aTHX_ sv, type, vtbl, 1);
5374 =for apidoc sv_rvweaken
5376 Weaken a reference: set the C<SvWEAKREF> flag on this RV; give the
5377 referred-to SV C<PERL_MAGIC_backref> magic if it hasn't already; and
5378 push a back-reference to this RV onto the array of backreferences
5379 associated with that magic. If the RV is magical, set magic will be
5380 called after the RV is cleared.
5386 Perl_sv_rvweaken(pTHX_ SV *const sv)
5390 PERL_ARGS_ASSERT_SV_RVWEAKEN;
5392 if (!SvOK(sv)) /* let undefs pass */
5395 Perl_croak(aTHX_ "Can't weaken a nonreference");
5396 else if (SvWEAKREF(sv)) {
5397 Perl_ck_warner(aTHX_ packWARN(WARN_MISC), "Reference is already weak");
5400 else if (SvREADONLY(sv)) croak_no_modify();
5402 Perl_sv_add_backref(aTHX_ tsv, sv);
5408 /* Give tsv backref magic if it hasn't already got it, then push a
5409 * back-reference to sv onto the array associated with the backref magic.
5411 * As an optimisation, if there's only one backref and it's not an AV,
5412 * store it directly in the HvAUX or mg_obj slot, avoiding the need to
5413 * allocate an AV. (Whether the slot holds an AV tells us whether this is
5417 /* A discussion about the backreferences array and its refcount:
5419 * The AV holding the backreferences is pointed to either as the mg_obj of
5420 * PERL_MAGIC_backref, or in the specific case of a HV, from the
5421 * xhv_backreferences field. The array is created with a refcount
5422 * of 2. This means that if during global destruction the array gets
5423 * picked on before its parent to have its refcount decremented by the
5424 * random zapper, it won't actually be freed, meaning it's still there for
5425 * when its parent gets freed.
5427 * When the parent SV is freed, the extra ref is killed by
5428 * Perl_sv_kill_backrefs. The other ref is killed, in the case of magic,
5429 * by mg_free() / MGf_REFCOUNTED, or for a hash, by Perl_hv_kill_backrefs.
5431 * When a single backref SV is stored directly, it is not reference
5436 Perl_sv_add_backref(pTHX_ SV *const tsv, SV *const sv)
5443 PERL_ARGS_ASSERT_SV_ADD_BACKREF;
5445 /* find slot to store array or singleton backref */
5447 if (SvTYPE(tsv) == SVt_PVHV) {
5448 svp = (SV**)Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5451 (SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL))))
5453 sv_magic(tsv, NULL, PERL_MAGIC_backref, NULL, 0);
5454 mg = mg_find(tsv, PERL_MAGIC_backref);
5456 svp = &(mg->mg_obj);
5459 /* create or retrieve the array */
5461 if ( (!*svp && SvTYPE(sv) == SVt_PVAV)
5462 || (*svp && SvTYPE(*svp) != SVt_PVAV)
5467 SvREFCNT_inc_simple_void(av);
5468 /* av now has a refcnt of 2; see discussion above */
5470 /* move single existing backref to the array */
5472 AvARRAY(av)[++AvFILLp(av)] = *svp; /* av_push() */
5476 mg->mg_flags |= MGf_REFCOUNTED;
5479 av = MUTABLE_AV(*svp);
5482 /* optimisation: store single backref directly in HvAUX or mg_obj */
5486 /* push new backref */
5487 assert(SvTYPE(av) == SVt_PVAV);
5488 if (AvFILLp(av) >= AvMAX(av)) {
5489 av_extend(av, AvFILLp(av)+1);
5491 AvARRAY(av)[++AvFILLp(av)] = sv; /* av_push() */
5494 /* delete a back-reference to ourselves from the backref magic associated
5495 * with the SV we point to.
5499 Perl_sv_del_backref(pTHX_ SV *const tsv, SV *const sv)
5504 PERL_ARGS_ASSERT_SV_DEL_BACKREF;
5506 if (SvTYPE(tsv) == SVt_PVHV) {
5508 svp = (SV**)Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5512 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
5513 svp = mg ? &(mg->mg_obj) : NULL;
5517 Perl_croak(aTHX_ "panic: del_backref");
5519 if (SvTYPE(*svp) == SVt_PVAV) {
5523 AV * const av = (AV*)*svp;
5525 assert(!SvIS_FREED(av));
5529 /* for an SV with N weak references to it, if all those
5530 * weak refs are deleted, then sv_del_backref will be called
5531 * N times and O(N^2) compares will be done within the backref
5532 * array. To ameliorate this potential slowness, we:
5533 * 1) make sure this code is as tight as possible;
5534 * 2) when looking for SV, look for it at both the head and tail of the
5535 * array first before searching the rest, since some create/destroy
5536 * patterns will cause the backrefs to be freed in order.
5543 SV **p = &svp[fill];
5544 SV *const topsv = *p;
5551 /* We weren't the last entry.
5552 An unordered list has this property that you
5553 can take the last element off the end to fill
5554 the hole, and it's still an unordered list :-)
5560 break; /* should only be one */
5567 AvFILLp(av) = fill-1;
5570 /* optimisation: only a single backref, stored directly */
5572 Perl_croak(aTHX_ "panic: del_backref");
5579 Perl_sv_kill_backrefs(pTHX_ SV *const sv, AV *const av)
5585 PERL_ARGS_ASSERT_SV_KILL_BACKREFS;
5590 /* after multiple passes through Perl_sv_clean_all() for a thinngy
5591 * that has badly leaked, the backref array may have gotten freed,
5592 * since we only protect it against 1 round of cleanup */
5593 if (SvIS_FREED(av)) {
5594 if (PL_in_clean_all) /* All is fair */
5597 "panic: magic_killbackrefs (freed backref AV/SV)");
5601 is_array = (SvTYPE(av) == SVt_PVAV);
5603 assert(!SvIS_FREED(av));
5606 last = svp + AvFILLp(av);
5609 /* optimisation: only a single backref, stored directly */
5615 while (svp <= last) {
5617 SV *const referrer = *svp;
5618 if (SvWEAKREF(referrer)) {
5619 /* XXX Should we check that it hasn't changed? */
5620 assert(SvROK(referrer));
5621 SvRV_set(referrer, 0);
5623 SvWEAKREF_off(referrer);
5624 SvSETMAGIC(referrer);
5625 } else if (SvTYPE(referrer) == SVt_PVGV ||
5626 SvTYPE(referrer) == SVt_PVLV) {
5627 assert(SvTYPE(sv) == SVt_PVHV); /* stash backref */
5628 /* You lookin' at me? */
5629 assert(GvSTASH(referrer));
5630 assert(GvSTASH(referrer) == (const HV *)sv);
5631 GvSTASH(referrer) = 0;
5632 } else if (SvTYPE(referrer) == SVt_PVCV ||
5633 SvTYPE(referrer) == SVt_PVFM) {
5634 if (SvTYPE(sv) == SVt_PVHV) { /* stash backref */
5635 /* You lookin' at me? */
5636 assert(CvSTASH(referrer));
5637 assert(CvSTASH(referrer) == (const HV *)sv);
5638 SvANY(MUTABLE_CV(referrer))->xcv_stash = 0;
5641 assert(SvTYPE(sv) == SVt_PVGV);
5642 /* You lookin' at me? */
5643 assert(CvGV(referrer));
5644 assert(CvGV(referrer) == (const GV *)sv);
5645 anonymise_cv_maybe(MUTABLE_GV(sv),
5646 MUTABLE_CV(referrer));
5651 "panic: magic_killbackrefs (flags=%"UVxf")",
5652 (UV)SvFLAGS(referrer));
5663 SvREFCNT_dec(av); /* remove extra count added by sv_add_backref() */
5669 =for apidoc sv_insert
5671 Inserts a string at the specified offset/length within the SV. Similar to
5672 the Perl substr() function. Handles get magic.
5674 =for apidoc sv_insert_flags
5676 Same as C<sv_insert>, but the extra C<flags> are passed the C<SvPV_force_flags> that applies to C<bigstr>.
5682 Perl_sv_insert_flags(pTHX_ SV *const bigstr, const STRLEN offset, const STRLEN len, const char *const little, const STRLEN littlelen, const U32 flags)
5687 register char *midend;
5688 register char *bigend;
5692 PERL_ARGS_ASSERT_SV_INSERT_FLAGS;
5695 Perl_croak(aTHX_ "Can't modify non-existent substring");
5696 SvPV_force_flags(bigstr, curlen, flags);
5697 (void)SvPOK_only_UTF8(bigstr);
5698 if (offset + len > curlen) {
5699 SvGROW(bigstr, offset+len+1);
5700 Zero(SvPVX(bigstr)+curlen, offset+len-curlen, char);
5701 SvCUR_set(bigstr, offset+len);
5705 i = littlelen - len;
5706 if (i > 0) { /* string might grow */
5707 big = SvGROW(bigstr, SvCUR(bigstr) + i + 1);
5708 mid = big + offset + len;
5709 midend = bigend = big + SvCUR(bigstr);
5712 while (midend > mid) /* shove everything down */
5713 *--bigend = *--midend;
5714 Move(little,big+offset,littlelen,char);
5715 SvCUR_set(bigstr, SvCUR(bigstr) + i);
5720 Move(little,SvPVX(bigstr)+offset,len,char);
5725 big = SvPVX(bigstr);
5728 bigend = big + SvCUR(bigstr);
5730 if (midend > bigend)
5731 Perl_croak(aTHX_ "panic: sv_insert");
5733 if (mid - big > bigend - midend) { /* faster to shorten from end */
5735 Move(little, mid, littlelen,char);
5738 i = bigend - midend;
5740 Move(midend, mid, i,char);
5744 SvCUR_set(bigstr, mid - big);
5746 else if ((i = mid - big)) { /* faster from front */
5747 midend -= littlelen;
5749 Move(big, midend - i, i, char);
5750 sv_chop(bigstr,midend-i);
5752 Move(little, mid, littlelen,char);
5754 else if (littlelen) {
5755 midend -= littlelen;
5756 sv_chop(bigstr,midend);
5757 Move(little,midend,littlelen,char);
5760 sv_chop(bigstr,midend);
5766 =for apidoc sv_replace
5768 Make the first argument a copy of the second, then delete the original.
5769 The target SV physically takes over ownership of the body of the source SV
5770 and inherits its flags; however, the target keeps any magic it owns,
5771 and any magic in the source is discarded.
5772 Note that this is a rather specialist SV copying operation; most of the
5773 time you'll want to use C<sv_setsv> or one of its many macro front-ends.
5779 Perl_sv_replace(pTHX_ register SV *const sv, register SV *const nsv)
5782 const U32 refcnt = SvREFCNT(sv);
5784 PERL_ARGS_ASSERT_SV_REPLACE;
5786 SV_CHECK_THINKFIRST_COW_DROP(sv);
5787 if (SvREFCNT(nsv) != 1) {
5788 Perl_croak(aTHX_ "panic: reference miscount on nsv in sv_replace()"
5789 " (%" UVuf " != 1)", (UV) SvREFCNT(nsv));
5791 if (SvMAGICAL(sv)) {
5795 sv_upgrade(nsv, SVt_PVMG);
5796 SvMAGIC_set(nsv, SvMAGIC(sv));
5797 SvFLAGS(nsv) |= SvMAGICAL(sv);
5799 SvMAGIC_set(sv, NULL);
5803 assert(!SvREFCNT(sv));
5804 #ifdef DEBUG_LEAKING_SCALARS
5805 sv->sv_flags = nsv->sv_flags;
5806 sv->sv_any = nsv->sv_any;
5807 sv->sv_refcnt = nsv->sv_refcnt;
5808 sv->sv_u = nsv->sv_u;
5810 StructCopy(nsv,sv,SV);
5812 if(SvTYPE(sv) == SVt_IV) {
5814 = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
5818 #ifdef PERL_OLD_COPY_ON_WRITE
5819 if (SvIsCOW_normal(nsv)) {
5820 /* We need to follow the pointers around the loop to make the
5821 previous SV point to sv, rather than nsv. */
5824 while ((next = SV_COW_NEXT_SV(current)) != nsv) {
5827 assert(SvPVX_const(current) == SvPVX_const(nsv));
5829 /* Make the SV before us point to the SV after us. */
5831 PerlIO_printf(Perl_debug_log, "previous is\n");
5833 PerlIO_printf(Perl_debug_log,
5834 "move it from 0x%"UVxf" to 0x%"UVxf"\n",
5835 (UV) SV_COW_NEXT_SV(current), (UV) sv);
5837 SV_COW_NEXT_SV_SET(current, sv);
5840 SvREFCNT(sv) = refcnt;
5841 SvFLAGS(nsv) |= SVTYPEMASK; /* Mark as freed */
5846 /* We're about to free a GV which has a CV that refers back to us.
5847 * If that CV will outlive us, make it anonymous (i.e. fix up its CvGV
5851 S_anonymise_cv_maybe(pTHX_ GV *gv, CV* cv)
5857 PERL_ARGS_ASSERT_ANONYMISE_CV_MAYBE;
5860 assert(SvREFCNT(gv) == 0);
5861 assert(isGV(gv) && isGV_with_GP(gv));
5863 assert(!CvANON(cv));
5864 assert(CvGV(cv) == gv);
5866 /* will the CV shortly be freed by gp_free() ? */
5867 if (GvCV(gv) == cv && GvGP(gv)->gp_refcnt < 2 && SvREFCNT(cv) < 2) {
5868 SvANY(cv)->xcv_gv = NULL;
5872 /* if not, anonymise: */
5873 stash = GvSTASH(gv) && HvNAME(GvSTASH(gv))
5874 ? HvENAME(GvSTASH(gv)) : NULL;
5875 gvname = Perl_newSVpvf(aTHX_ "%s::__ANON__",
5876 stash ? stash : "__ANON__");
5877 anongv = gv_fetchsv(gvname, GV_ADDMULTI, SVt_PVCV);
5878 SvREFCNT_dec(gvname);
5882 SvANY(cv)->xcv_gv = MUTABLE_GV(SvREFCNT_inc(anongv));
5887 =for apidoc sv_clear
5889 Clear an SV: call any destructors, free up any memory used by the body,
5890 and free the body itself. The SV's head is I<not> freed, although
5891 its type is set to all 1's so that it won't inadvertently be assumed
5892 to be live during global destruction etc.
5893 This function should only be called when REFCNT is zero. Most of the time
5894 you'll want to call C<sv_free()> (or its macro wrapper C<SvREFCNT_dec>)
5901 Perl_sv_clear(pTHX_ SV *const orig_sv)
5906 const struct body_details *sv_type_details;
5909 register SV *sv = orig_sv;
5912 PERL_ARGS_ASSERT_SV_CLEAR;
5914 /* within this loop, sv is the SV currently being freed, and
5915 * iter_sv is the most recent AV or whatever that's being iterated
5916 * over to provide more SVs */
5922 assert(SvREFCNT(sv) == 0);
5923 assert(SvTYPE(sv) != (svtype)SVTYPEMASK);
5925 if (type <= SVt_IV) {
5926 /* See the comment in sv.h about the collusion between this
5927 * early return and the overloading of the NULL slots in the
5931 SvFLAGS(sv) &= SVf_BREAK;
5932 SvFLAGS(sv) |= SVTYPEMASK;
5936 assert(!SvOBJECT(sv) || type >= SVt_PVMG); /* objs are always >= MG */
5938 if (type >= SVt_PVMG) {
5940 if (!curse(sv, 1)) goto get_next_sv;
5941 type = SvTYPE(sv); /* destructor may have changed it */
5943 /* Free back-references before magic, in case the magic calls
5944 * Perl code that has weak references to sv. */
5945 if (type == SVt_PVHV) {
5946 Perl_hv_kill_backrefs(aTHX_ MUTABLE_HV(sv));
5950 else if (type == SVt_PVMG && SvPAD_OUR(sv)) {
5951 SvREFCNT_dec(SvOURSTASH(sv));
5952 } else if (SvMAGIC(sv)) {
5953 /* Free back-references before other types of magic. */
5954 sv_unmagic(sv, PERL_MAGIC_backref);
5957 if (type == SVt_PVMG && SvPAD_TYPED(sv))
5958 SvREFCNT_dec(SvSTASH(sv));
5961 /* case SVt_BIND: */
5964 IoIFP(sv) != PerlIO_stdin() &&
5965 IoIFP(sv) != PerlIO_stdout() &&
5966 IoIFP(sv) != PerlIO_stderr() &&
5967 !(IoFLAGS(sv) & IOf_FAKE_DIRP))
5969 io_close(MUTABLE_IO(sv), FALSE);
5971 if (IoDIRP(sv) && !(IoFLAGS(sv) & IOf_FAKE_DIRP))
5972 PerlDir_close(IoDIRP(sv));
5973 IoDIRP(sv) = (DIR*)NULL;
5974 Safefree(IoTOP_NAME(sv));
5975 Safefree(IoFMT_NAME(sv));
5976 Safefree(IoBOTTOM_NAME(sv));
5979 /* FIXME for plugins */
5980 pregfree2((REGEXP*) sv);
5984 cv_undef(MUTABLE_CV(sv));
5985 /* If we're in a stash, we don't own a reference to it.
5986 * However it does have a back reference to us, which needs to
5988 if ((stash = CvSTASH(sv)))
5989 sv_del_backref(MUTABLE_SV(stash), sv);
5992 if (PL_last_swash_hv == (const HV *)sv) {
5993 PL_last_swash_hv = NULL;
5995 if (HvTOTALKEYS((HV*)sv) > 0) {
5997 /* this statement should match the one at the beginning of
5998 * hv_undef_flags() */
5999 if ( PL_phase != PERL_PHASE_DESTRUCT
6000 && (name = HvNAME((HV*)sv)))
6003 (void)hv_delete(PL_stashcache, name,
6004 HvNAMELEN_get((HV*)sv), G_DISCARD);
6005 hv_name_set((HV*)sv, NULL, 0, 0);
6008 /* save old iter_sv in unused SvSTASH field */
6009 assert(!SvOBJECT(sv));
6010 SvSTASH(sv) = (HV*)iter_sv;
6013 /* XXX ideally we should save the old value of hash_index
6014 * too, but I can't think of any place to hide it. The
6015 * effect of not saving it is that for freeing hashes of
6016 * hashes, we become quadratic in scanning the HvARRAY of
6017 * the top hash looking for new entries to free; but
6018 * hopefully this will be dwarfed by the freeing of all
6019 * the nested hashes. */
6021 next_sv = Perl_hfree_next_entry(aTHX_ (HV*)sv, &hash_index);
6022 goto get_next_sv; /* process this new sv */
6024 /* free empty hash */
6025 Perl_hv_undef_flags(aTHX_ MUTABLE_HV(sv), HV_NAME_SETALL);
6026 assert(!HvARRAY((HV*)sv));
6030 AV* av = MUTABLE_AV(sv);
6031 if (PL_comppad == av) {
6035 if (AvREAL(av) && AvFILLp(av) > -1) {
6036 next_sv = AvARRAY(av)[AvFILLp(av)--];
6037 /* save old iter_sv in top-most slot of AV,
6038 * and pray that it doesn't get wiped in the meantime */
6039 AvARRAY(av)[AvMAX(av)] = iter_sv;
6041 goto get_next_sv; /* process this new sv */
6043 Safefree(AvALLOC(av));
6048 if (LvTYPE(sv) == 'T') { /* for tie: return HE to pool */
6049 SvREFCNT_dec(HeKEY_sv((HE*)LvTARG(sv)));
6050 HeNEXT((HE*)LvTARG(sv)) = PL_hv_fetch_ent_mh;
6051 PL_hv_fetch_ent_mh = (HE*)LvTARG(sv);
6053 else if (LvTYPE(sv) != 't') /* unless tie: unrefcnted fake SV** */
6054 SvREFCNT_dec(LvTARG(sv));
6056 if (isGV_with_GP(sv)) {
6057 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
6058 && HvENAME_get(stash))
6059 mro_method_changed_in(stash);
6060 gp_free(MUTABLE_GV(sv));
6062 unshare_hek(GvNAME_HEK(sv));
6063 /* If we're in a stash, we don't own a reference to it.
6064 * However it does have a back reference to us, which
6065 * needs to be cleared. */
6066 if (!SvVALID(sv) && (stash = GvSTASH(sv)))
6067 sv_del_backref(MUTABLE_SV(stash), sv);
6069 /* FIXME. There are probably more unreferenced pointers to SVs
6070 * in the interpreter struct that we should check and tidy in
6071 * a similar fashion to this: */
6072 if ((const GV *)sv == PL_last_in_gv)
6073 PL_last_in_gv = NULL;
6079 /* Don't bother with SvOOK_off(sv); as we're only going to
6083 SvOOK_offset(sv, offset);
6084 SvPV_set(sv, SvPVX_mutable(sv) - offset);
6085 /* Don't even bother with turning off the OOK flag. */
6090 SV * const target = SvRV(sv);
6092 sv_del_backref(target, sv);
6097 #ifdef PERL_OLD_COPY_ON_WRITE
6098 else if (SvPVX_const(sv)
6099 && !(SvTYPE(sv) == SVt_PVIO
6100 && !(IoFLAGS(sv) & IOf_FAKE_DIRP)))
6104 PerlIO_printf(Perl_debug_log, "Copy on write: clear\n");
6108 sv_release_COW(sv, SvPVX_const(sv), SV_COW_NEXT_SV(sv));
6110 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
6114 } else if (SvLEN(sv)) {
6115 Safefree(SvPVX_const(sv));
6119 else if (SvPVX_const(sv) && SvLEN(sv)
6120 && !(SvTYPE(sv) == SVt_PVIO
6121 && !(IoFLAGS(sv) & IOf_FAKE_DIRP)))
6122 Safefree(SvPVX_mutable(sv));
6123 else if (SvPVX_const(sv) && SvREADONLY(sv) && SvFAKE(sv)) {
6124 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
6135 SvFLAGS(sv) &= SVf_BREAK;
6136 SvFLAGS(sv) |= SVTYPEMASK;
6138 sv_type_details = bodies_by_type + type;
6139 if (sv_type_details->arena) {
6140 del_body(((char *)SvANY(sv) + sv_type_details->offset),
6141 &PL_body_roots[type]);
6143 else if (sv_type_details->body_size) {
6144 safefree(SvANY(sv));
6148 /* caller is responsible for freeing the head of the original sv */
6149 if (sv != orig_sv && !SvREFCNT(sv))
6152 /* grab and free next sv, if any */
6160 else if (!iter_sv) {
6162 } else if (SvTYPE(iter_sv) == SVt_PVAV) {
6163 AV *const av = (AV*)iter_sv;
6164 if (AvFILLp(av) > -1) {
6165 sv = AvARRAY(av)[AvFILLp(av)--];
6167 else { /* no more elements of current AV to free */
6170 /* restore previous value, squirrelled away */
6171 iter_sv = AvARRAY(av)[AvMAX(av)];
6172 Safefree(AvALLOC(av));
6175 } else if (SvTYPE(iter_sv) == SVt_PVHV) {
6176 sv = Perl_hfree_next_entry(aTHX_ (HV*)iter_sv, &hash_index);
6177 if (!sv && !HvTOTALKEYS((HV *)iter_sv)) {
6178 /* no more elements of current HV to free */
6181 /* Restore previous value of iter_sv, squirrelled away */
6182 assert(!SvOBJECT(sv));
6183 iter_sv = (SV*)SvSTASH(sv);
6185 /* ideally we should restore the old hash_index here,
6186 * but we don't currently save the old value */
6189 /* free any remaining detritus from the hash struct */
6190 Perl_hv_undef_flags(aTHX_ MUTABLE_HV(sv), HV_NAME_SETALL);
6191 assert(!HvARRAY((HV*)sv));
6196 /* unrolled SvREFCNT_dec and sv_free2 follows: */
6200 if (!SvREFCNT(sv)) {
6204 if (--(SvREFCNT(sv)))
6208 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEBUGGING),
6209 "Attempt to free temp prematurely: SV 0x%"UVxf
6210 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
6214 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
6215 /* make sure SvREFCNT(sv)==0 happens very seldom */
6216 SvREFCNT(sv) = (~(U32)0)/2;
6225 /* This routine curses the sv itself, not the object referenced by sv. So
6226 sv does not have to be ROK. */
6229 S_curse(pTHX_ SV * const sv, const bool check_refcnt) {
6232 PERL_ARGS_ASSERT_CURSE;
6233 assert(SvOBJECT(sv));
6235 if (PL_defstash && /* Still have a symbol table? */
6242 stash = SvSTASH(sv);
6243 destructor = StashHANDLER(stash,DESTROY);
6245 /* A constant subroutine can have no side effects, so
6246 don't bother calling it. */
6247 && !CvCONST(destructor)
6248 /* Don't bother calling an empty destructor */
6249 && (CvISXSUB(destructor)
6250 || (CvSTART(destructor)
6251 && (CvSTART(destructor)->op_next->op_type
6254 SV* const tmpref = newRV(sv);
6255 SvREADONLY_on(tmpref); /* DESTROY() could be naughty */
6257 PUSHSTACKi(PERLSI_DESTROY);
6262 call_sv(MUTABLE_SV(destructor),
6263 G_DISCARD|G_EVAL|G_KEEPERR|G_VOID);
6267 if(SvREFCNT(tmpref) < 2) {
6268 /* tmpref is not kept alive! */
6270 SvRV_set(tmpref, NULL);
6273 SvREFCNT_dec(tmpref);
6275 } while (SvOBJECT(sv) && SvSTASH(sv) != stash);
6278 if (check_refcnt && SvREFCNT(sv)) {
6279 if (PL_in_clean_objs)
6281 "DESTROY created new reference to dead object '%s'",
6283 /* DESTROY gave object new lease on life */
6289 SvREFCNT_dec(SvSTASH(sv)); /* possibly of changed persuasion */
6290 SvOBJECT_off(sv); /* Curse the object. */
6291 if (SvTYPE(sv) != SVt_PVIO)
6292 --PL_sv_objcount;/* XXX Might want something more general */
6298 =for apidoc sv_newref
6300 Increment an SV's reference count. Use the C<SvREFCNT_inc()> wrapper
6307 Perl_sv_newref(pTHX_ SV *const sv)
6309 PERL_UNUSED_CONTEXT;
6318 Decrement an SV's reference count, and if it drops to zero, call
6319 C<sv_clear> to invoke destructors and free up any memory used by
6320 the body; finally, deallocate the SV's head itself.
6321 Normally called via a wrapper macro C<SvREFCNT_dec>.
6327 Perl_sv_free(pTHX_ SV *const sv)
6332 if (SvREFCNT(sv) == 0) {
6333 if (SvFLAGS(sv) & SVf_BREAK)
6334 /* this SV's refcnt has been artificially decremented to
6335 * trigger cleanup */
6337 if (PL_in_clean_all) /* All is fair */
6339 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
6340 /* make sure SvREFCNT(sv)==0 happens very seldom */
6341 SvREFCNT(sv) = (~(U32)0)/2;
6344 if (ckWARN_d(WARN_INTERNAL)) {
6345 #ifdef DEBUG_LEAKING_SCALARS_FORK_DUMP
6346 Perl_dump_sv_child(aTHX_ sv);
6348 #ifdef DEBUG_LEAKING_SCALARS
6351 #ifdef DEBUG_LEAKING_SCALARS_ABORT
6352 if (PL_warnhook == PERL_WARNHOOK_FATAL
6353 || ckDEAD(packWARN(WARN_INTERNAL))) {
6354 /* Don't let Perl_warner cause us to escape our fate: */
6358 /* This may not return: */
6359 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
6360 "Attempt to free unreferenced scalar: SV 0x%"UVxf
6361 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
6364 #ifdef DEBUG_LEAKING_SCALARS_ABORT
6369 if (--(SvREFCNT(sv)) > 0)
6371 Perl_sv_free2(aTHX_ sv);
6375 Perl_sv_free2(pTHX_ SV *const sv)
6379 PERL_ARGS_ASSERT_SV_FREE2;
6383 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEBUGGING),
6384 "Attempt to free temp prematurely: SV 0x%"UVxf
6385 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
6389 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
6390 /* make sure SvREFCNT(sv)==0 happens very seldom */
6391 SvREFCNT(sv) = (~(U32)0)/2;
6402 Returns the length of the string in the SV. Handles magic and type
6403 coercion. See also C<SvCUR>, which gives raw access to the xpv_cur slot.
6409 Perl_sv_len(pTHX_ register SV *const sv)
6417 len = mg_length(sv);
6419 (void)SvPV_const(sv, len);
6424 =for apidoc sv_len_utf8
6426 Returns the number of characters in the string in an SV, counting wide
6427 UTF-8 bytes as a single character. Handles magic and type coercion.
6433 * The length is cached in PERL_MAGIC_utf8, in the mg_len field. Also the
6434 * mg_ptr is used, by sv_pos_u2b() and sv_pos_b2u() - see the comments below.
6435 * (Note that the mg_len is not the length of the mg_ptr field.
6436 * This allows the cache to store the character length of the string without
6437 * needing to malloc() extra storage to attach to the mg_ptr.)
6442 Perl_sv_len_utf8(pTHX_ register SV *const sv)
6448 return mg_length(sv);
6452 const U8 *s = (U8*)SvPV_const(sv, len);
6456 MAGIC *mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_utf8) : NULL;
6458 if (mg && (mg->mg_len != -1 || mg->mg_ptr)) {
6459 if (mg->mg_len != -1)
6462 /* We can use the offset cache for a headstart.
6463 The longer value is stored in the first pair. */
6464 STRLEN *cache = (STRLEN *) mg->mg_ptr;
6466 ulen = cache[0] + Perl_utf8_length(aTHX_ s + cache[1],
6470 if (PL_utf8cache < 0) {
6471 const STRLEN real = Perl_utf8_length(aTHX_ s, s + len);
6472 assert_uft8_cache_coherent("sv_len_utf8", ulen, real, sv);
6476 ulen = Perl_utf8_length(aTHX_ s, s + len);
6477 utf8_mg_len_cache_update(sv, &mg, ulen);
6481 return Perl_utf8_length(aTHX_ s, s + len);
6485 /* Walk forwards to find the byte corresponding to the passed in UTF-8
6488 S_sv_pos_u2b_forwards(const U8 *const start, const U8 *const send,
6489 STRLEN *const uoffset_p, bool *const at_end)
6491 const U8 *s = start;
6492 STRLEN uoffset = *uoffset_p;
6494 PERL_ARGS_ASSERT_SV_POS_U2B_FORWARDS;
6496 while (s < send && uoffset) {
6503 else if (s > send) {
6505 /* This is the existing behaviour. Possibly it should be a croak, as
6506 it's actually a bounds error */
6509 *uoffset_p -= uoffset;
6513 /* Given the length of the string in both bytes and UTF-8 characters, decide
6514 whether to walk forwards or backwards to find the byte corresponding to
6515 the passed in UTF-8 offset. */
6517 S_sv_pos_u2b_midway(const U8 *const start, const U8 *send,
6518 STRLEN uoffset, const STRLEN uend)
6520 STRLEN backw = uend - uoffset;
6522 PERL_ARGS_ASSERT_SV_POS_U2B_MIDWAY;
6524 if (uoffset < 2 * backw) {
6525 /* The assumption is that going forwards is twice the speed of going
6526 forward (that's where the 2 * backw comes from).
6527 (The real figure of course depends on the UTF-8 data.) */
6528 const U8 *s = start;
6530 while (s < send && uoffset--)
6540 while (UTF8_IS_CONTINUATION(*send))
6543 return send - start;
6546 /* For the string representation of the given scalar, find the byte
6547 corresponding to the passed in UTF-8 offset. uoffset0 and boffset0
6548 give another position in the string, *before* the sought offset, which
6549 (which is always true, as 0, 0 is a valid pair of positions), which should
6550 help reduce the amount of linear searching.
6551 If *mgp is non-NULL, it should point to the UTF-8 cache magic, which
6552 will be used to reduce the amount of linear searching. The cache will be
6553 created if necessary, and the found value offered to it for update. */
6555 S_sv_pos_u2b_cached(pTHX_ SV *const sv, MAGIC **const mgp, const U8 *const start,
6556 const U8 *const send, STRLEN uoffset,
6557 STRLEN uoffset0, STRLEN boffset0)
6559 STRLEN boffset = 0; /* Actually always set, but let's keep gcc happy. */
6561 bool at_end = FALSE;
6563 PERL_ARGS_ASSERT_SV_POS_U2B_CACHED;
6565 assert (uoffset >= uoffset0);
6572 && (*mgp || (SvTYPE(sv) >= SVt_PVMG &&
6573 (*mgp = mg_find(sv, PERL_MAGIC_utf8))))) {
6574 if ((*mgp)->mg_ptr) {
6575 STRLEN *cache = (STRLEN *) (*mgp)->mg_ptr;
6576 if (cache[0] == uoffset) {
6577 /* An exact match. */
6580 if (cache[2] == uoffset) {
6581 /* An exact match. */
6585 if (cache[0] < uoffset) {
6586 /* The cache already knows part of the way. */
6587 if (cache[0] > uoffset0) {
6588 /* The cache knows more than the passed in pair */
6589 uoffset0 = cache[0];
6590 boffset0 = cache[1];
6592 if ((*mgp)->mg_len != -1) {
6593 /* And we know the end too. */
6595 + sv_pos_u2b_midway(start + boffset0, send,
6597 (*mgp)->mg_len - uoffset0);
6599 uoffset -= uoffset0;
6601 + sv_pos_u2b_forwards(start + boffset0,
6602 send, &uoffset, &at_end);
6603 uoffset += uoffset0;
6606 else if (cache[2] < uoffset) {
6607 /* We're between the two cache entries. */
6608 if (cache[2] > uoffset0) {
6609 /* and the cache knows more than the passed in pair */
6610 uoffset0 = cache[2];
6611 boffset0 = cache[3];
6615 + sv_pos_u2b_midway(start + boffset0,
6618 cache[0] - uoffset0);
6621 + sv_pos_u2b_midway(start + boffset0,
6624 cache[2] - uoffset0);
6628 else if ((*mgp)->mg_len != -1) {
6629 /* If we can take advantage of a passed in offset, do so. */
6630 /* In fact, offset0 is either 0, or less than offset, so don't
6631 need to worry about the other possibility. */
6633 + sv_pos_u2b_midway(start + boffset0, send,
6635 (*mgp)->mg_len - uoffset0);
6640 if (!found || PL_utf8cache < 0) {
6641 STRLEN real_boffset;
6642 uoffset -= uoffset0;
6643 real_boffset = boffset0 + sv_pos_u2b_forwards(start + boffset0,
6644 send, &uoffset, &at_end);
6645 uoffset += uoffset0;
6647 if (found && PL_utf8cache < 0)
6648 assert_uft8_cache_coherent("sv_pos_u2b_cache", boffset,
6650 boffset = real_boffset;
6655 utf8_mg_len_cache_update(sv, mgp, uoffset);
6657 utf8_mg_pos_cache_update(sv, mgp, boffset, uoffset, send - start);
6664 =for apidoc sv_pos_u2b_flags
6666 Converts the value pointed to by offsetp from a count of UTF-8 chars from
6667 the start of the string, to a count of the equivalent number of bytes; if
6668 lenp is non-zero, it does the same to lenp, but this time starting from
6669 the offset, rather than from the start of the string. Handles type coercion.
6670 I<flags> is passed to C<SvPV_flags>, and usually should be
6671 C<SV_GMAGIC|SV_CONST_RETURN> to handle magic.
6677 * sv_pos_u2b_flags() uses, like sv_pos_b2u(), the mg_ptr of the potential
6678 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6679 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
6684 Perl_sv_pos_u2b_flags(pTHX_ SV *const sv, STRLEN uoffset, STRLEN *const lenp,
6691 PERL_ARGS_ASSERT_SV_POS_U2B_FLAGS;
6693 start = (U8*)SvPV_flags(sv, len, flags);
6695 const U8 * const send = start + len;
6697 boffset = sv_pos_u2b_cached(sv, &mg, start, send, uoffset, 0, 0);
6700 && *lenp /* don't bother doing work for 0, as its bytes equivalent
6701 is 0, and *lenp is already set to that. */) {
6702 /* Convert the relative offset to absolute. */
6703 const STRLEN uoffset2 = uoffset + *lenp;
6704 const STRLEN boffset2
6705 = sv_pos_u2b_cached(sv, &mg, start, send, uoffset2,
6706 uoffset, boffset) - boffset;
6720 =for apidoc sv_pos_u2b
6722 Converts the value pointed to by offsetp from a count of UTF-8 chars from
6723 the start of the string, to a count of the equivalent number of bytes; if
6724 lenp is non-zero, it does the same to lenp, but this time starting from
6725 the offset, rather than from the start of the string. Handles magic and
6728 Use C<sv_pos_u2b_flags> in preference, which correctly handles strings longer
6735 * sv_pos_u2b() uses, like sv_pos_b2u(), the mg_ptr of the potential
6736 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6737 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
6741 /* This function is subject to size and sign problems */
6744 Perl_sv_pos_u2b(pTHX_ register SV *const sv, I32 *const offsetp, I32 *const lenp)
6746 PERL_ARGS_ASSERT_SV_POS_U2B;
6749 STRLEN ulen = (STRLEN)*lenp;
6750 *offsetp = (I32)sv_pos_u2b_flags(sv, (STRLEN)*offsetp, &ulen,
6751 SV_GMAGIC|SV_CONST_RETURN);
6754 *offsetp = (I32)sv_pos_u2b_flags(sv, (STRLEN)*offsetp, NULL,
6755 SV_GMAGIC|SV_CONST_RETURN);
6760 S_utf8_mg_len_cache_update(pTHX_ SV *const sv, MAGIC **const mgp,
6763 PERL_ARGS_ASSERT_UTF8_MG_LEN_CACHE_UPDATE;
6767 if (!*mgp && (SvTYPE(sv) < SVt_PVMG ||
6768 !(*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
6769 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, &PL_vtbl_utf8, 0, 0);
6773 (*mgp)->mg_len = ulen;
6774 /* For now, treat "overflowed" as "still unknown". See RT #72924. */
6775 if (ulen != (STRLEN) (*mgp)->mg_len)
6776 (*mgp)->mg_len = -1;
6779 /* Create and update the UTF8 magic offset cache, with the proffered utf8/
6780 byte length pairing. The (byte) length of the total SV is passed in too,
6781 as blen, because for some (more esoteric) SVs, the call to SvPV_const()
6782 may not have updated SvCUR, so we can't rely on reading it directly.
6784 The proffered utf8/byte length pairing isn't used if the cache already has
6785 two pairs, and swapping either for the proffered pair would increase the
6786 RMS of the intervals between known byte offsets.
6788 The cache itself consists of 4 STRLEN values
6789 0: larger UTF-8 offset
6790 1: corresponding byte offset
6791 2: smaller UTF-8 offset
6792 3: corresponding byte offset
6794 Unused cache pairs have the value 0, 0.
6795 Keeping the cache "backwards" means that the invariant of
6796 cache[0] >= cache[2] is maintained even with empty slots, which means that
6797 the code that uses it doesn't need to worry if only 1 entry has actually
6798 been set to non-zero. It also makes the "position beyond the end of the
6799 cache" logic much simpler, as the first slot is always the one to start
6803 S_utf8_mg_pos_cache_update(pTHX_ SV *const sv, MAGIC **const mgp, const STRLEN byte,
6804 const STRLEN utf8, const STRLEN blen)
6808 PERL_ARGS_ASSERT_UTF8_MG_POS_CACHE_UPDATE;
6813 if (!*mgp && (SvTYPE(sv) < SVt_PVMG ||
6814 !(*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
6815 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, (MGVTBL*)&PL_vtbl_utf8, 0,
6817 (*mgp)->mg_len = -1;
6821 if (!(cache = (STRLEN *)(*mgp)->mg_ptr)) {
6822 Newxz(cache, PERL_MAGIC_UTF8_CACHESIZE * 2, STRLEN);
6823 (*mgp)->mg_ptr = (char *) cache;
6827 if (PL_utf8cache < 0 && SvPOKp(sv)) {
6828 /* SvPOKp() because it's possible that sv has string overloading, and
6829 therefore is a reference, hence SvPVX() is actually a pointer.
6830 This cures the (very real) symptoms of RT 69422, but I'm not actually
6831 sure whether we should even be caching the results of UTF-8
6832 operations on overloading, given that nothing stops overloading
6833 returning a different value every time it's called. */
6834 const U8 *start = (const U8 *) SvPVX_const(sv);
6835 const STRLEN realutf8 = utf8_length(start, start + byte);
6837 assert_uft8_cache_coherent("utf8_mg_pos_cache_update", utf8, realutf8,
6841 /* Cache is held with the later position first, to simplify the code
6842 that deals with unbounded ends. */
6844 ASSERT_UTF8_CACHE(cache);
6845 if (cache[1] == 0) {
6846 /* Cache is totally empty */
6849 } else if (cache[3] == 0) {
6850 if (byte > cache[1]) {
6851 /* New one is larger, so goes first. */
6852 cache[2] = cache[0];
6853 cache[3] = cache[1];
6861 #define THREEWAY_SQUARE(a,b,c,d) \
6862 ((float)((d) - (c))) * ((float)((d) - (c))) \
6863 + ((float)((c) - (b))) * ((float)((c) - (b))) \
6864 + ((float)((b) - (a))) * ((float)((b) - (a)))
6866 /* Cache has 2 slots in use, and we know three potential pairs.
6867 Keep the two that give the lowest RMS distance. Do the
6868 calculation in bytes simply because we always know the byte
6869 length. squareroot has the same ordering as the positive value,
6870 so don't bother with the actual square root. */
6871 const float existing = THREEWAY_SQUARE(0, cache[3], cache[1], blen);
6872 if (byte > cache[1]) {
6873 /* New position is after the existing pair of pairs. */
6874 const float keep_earlier
6875 = THREEWAY_SQUARE(0, cache[3], byte, blen);
6876 const float keep_later
6877 = THREEWAY_SQUARE(0, cache[1], byte, blen);
6879 if (keep_later < keep_earlier) {
6880 if (keep_later < existing) {
6881 cache[2] = cache[0];
6882 cache[3] = cache[1];
6888 if (keep_earlier < existing) {
6894 else if (byte > cache[3]) {
6895 /* New position is between the existing pair of pairs. */
6896 const float keep_earlier
6897 = THREEWAY_SQUARE(0, cache[3], byte, blen);
6898 const float keep_later
6899 = THREEWAY_SQUARE(0, byte, cache[1], blen);
6901 if (keep_later < keep_earlier) {
6902 if (keep_later < existing) {
6908 if (keep_earlier < existing) {
6915 /* New position is before the existing pair of pairs. */
6916 const float keep_earlier
6917 = THREEWAY_SQUARE(0, byte, cache[3], blen);
6918 const float keep_later
6919 = THREEWAY_SQUARE(0, byte, cache[1], blen);
6921 if (keep_later < keep_earlier) {
6922 if (keep_later < existing) {
6928 if (keep_earlier < existing) {
6929 cache[0] = cache[2];
6930 cache[1] = cache[3];
6937 ASSERT_UTF8_CACHE(cache);
6940 /* We already know all of the way, now we may be able to walk back. The same
6941 assumption is made as in S_sv_pos_u2b_midway(), namely that walking
6942 backward is half the speed of walking forward. */
6944 S_sv_pos_b2u_midway(pTHX_ const U8 *const s, const U8 *const target,
6945 const U8 *end, STRLEN endu)
6947 const STRLEN forw = target - s;
6948 STRLEN backw = end - target;
6950 PERL_ARGS_ASSERT_SV_POS_B2U_MIDWAY;
6952 if (forw < 2 * backw) {
6953 return utf8_length(s, target);
6956 while (end > target) {
6958 while (UTF8_IS_CONTINUATION(*end)) {
6967 =for apidoc sv_pos_b2u
6969 Converts the value pointed to by offsetp from a count of bytes from the
6970 start of the string, to a count of the equivalent number of UTF-8 chars.
6971 Handles magic and type coercion.
6977 * sv_pos_b2u() uses, like sv_pos_u2b(), the mg_ptr of the potential
6978 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6983 Perl_sv_pos_b2u(pTHX_ register SV *const sv, I32 *const offsetp)
6986 const STRLEN byte = *offsetp;
6987 STRLEN len = 0; /* Actually always set, but let's keep gcc happy. */
6993 PERL_ARGS_ASSERT_SV_POS_B2U;
6998 s = (const U8*)SvPV_const(sv, blen);
7001 Perl_croak(aTHX_ "panic: sv_pos_b2u: bad byte offset");
7007 && SvTYPE(sv) >= SVt_PVMG
7008 && (mg = mg_find(sv, PERL_MAGIC_utf8)))
7011 STRLEN * const cache = (STRLEN *) mg->mg_ptr;
7012 if (cache[1] == byte) {
7013 /* An exact match. */
7014 *offsetp = cache[0];
7017 if (cache[3] == byte) {
7018 /* An exact match. */
7019 *offsetp = cache[2];
7023 if (cache[1] < byte) {
7024 /* We already know part of the way. */
7025 if (mg->mg_len != -1) {
7026 /* Actually, we know the end too. */
7028 + S_sv_pos_b2u_midway(aTHX_ s + cache[1], send,
7029 s + blen, mg->mg_len - cache[0]);
7031 len = cache[0] + utf8_length(s + cache[1], send);
7034 else if (cache[3] < byte) {
7035 /* We're between the two cached pairs, so we do the calculation
7036 offset by the byte/utf-8 positions for the earlier pair,
7037 then add the utf-8 characters from the string start to
7039 len = S_sv_pos_b2u_midway(aTHX_ s + cache[3], send,
7040 s + cache[1], cache[0] - cache[2])
7044 else { /* cache[3] > byte */
7045 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + cache[3],
7049 ASSERT_UTF8_CACHE(cache);
7051 } else if (mg->mg_len != -1) {
7052 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + blen, mg->mg_len);
7056 if (!found || PL_utf8cache < 0) {
7057 const STRLEN real_len = utf8_length(s, send);
7059 if (found && PL_utf8cache < 0)
7060 assert_uft8_cache_coherent("sv_pos_b2u", len, real_len, sv);
7067 utf8_mg_len_cache_update(sv, &mg, len);
7069 utf8_mg_pos_cache_update(sv, &mg, byte, len, blen);
7074 S_assert_uft8_cache_coherent(pTHX_ const char *const func, STRLEN from_cache,
7075 STRLEN real, SV *const sv)
7077 PERL_ARGS_ASSERT_ASSERT_UFT8_CACHE_COHERENT;
7079 /* As this is debugging only code, save space by keeping this test here,
7080 rather than inlining it in all the callers. */
7081 if (from_cache == real)
7084 /* Need to turn the assertions off otherwise we may recurse infinitely
7085 while printing error messages. */
7086 SAVEI8(PL_utf8cache);
7088 Perl_croak(aTHX_ "panic: %s cache %"UVuf" real %"UVuf" for %"SVf,
7089 func, (UV) from_cache, (UV) real, SVfARG(sv));
7095 Returns a boolean indicating whether the strings in the two SVs are
7096 identical. Is UTF-8 and 'use bytes' aware, handles get magic, and will
7097 coerce its args to strings if necessary.
7099 =for apidoc sv_eq_flags
7101 Returns a boolean indicating whether the strings in the two SVs are
7102 identical. Is UTF-8 and 'use bytes' aware and coerces its args to strings
7103 if necessary. If the flags include SV_GMAGIC, it handles get-magic, too.
7109 Perl_sv_eq_flags(pTHX_ register SV *sv1, register SV *sv2, const U32 flags)
7118 SV* svrecode = NULL;
7125 /* if pv1 and pv2 are the same, second SvPV_const call may
7126 * invalidate pv1 (if we are handling magic), so we may need to
7128 if (sv1 == sv2 && flags & SV_GMAGIC
7129 && (SvTHINKFIRST(sv1) || SvGMAGICAL(sv1))) {
7130 pv1 = SvPV_const(sv1, cur1);
7131 sv1 = newSVpvn_flags(pv1, cur1, SVs_TEMP | SvUTF8(sv2));
7133 pv1 = SvPV_flags_const(sv1, cur1, flags);
7141 pv2 = SvPV_flags_const(sv2, cur2, flags);
7143 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
7144 /* Differing utf8ness.
7145 * Do not UTF8size the comparands as a side-effect. */
7148 svrecode = newSVpvn(pv2, cur2);
7149 sv_recode_to_utf8(svrecode, PL_encoding);
7150 pv2 = SvPV_const(svrecode, cur2);
7153 svrecode = newSVpvn(pv1, cur1);
7154 sv_recode_to_utf8(svrecode, PL_encoding);
7155 pv1 = SvPV_const(svrecode, cur1);
7157 /* Now both are in UTF-8. */
7159 SvREFCNT_dec(svrecode);
7165 /* sv1 is the UTF-8 one */
7166 return bytes_cmp_utf8((const U8*)pv2, cur2,
7167 (const U8*)pv1, cur1) == 0;
7170 /* sv2 is the UTF-8 one */
7171 return bytes_cmp_utf8((const U8*)pv1, cur1,
7172 (const U8*)pv2, cur2) == 0;
7178 eq = (pv1 == pv2) || memEQ(pv1, pv2, cur1);
7180 SvREFCNT_dec(svrecode);
7190 Compares the strings in two SVs. Returns -1, 0, or 1 indicating whether the
7191 string in C<sv1> is less than, equal to, or greater than the string in
7192 C<sv2>. Is UTF-8 and 'use bytes' aware, handles get magic, and will
7193 coerce its args to strings if necessary. See also C<sv_cmp_locale>.
7195 =for apidoc sv_cmp_flags
7197 Compares the strings in two SVs. Returns -1, 0, or 1 indicating whether the
7198 string in C<sv1> is less than, equal to, or greater than the string in
7199 C<sv2>. Is UTF-8 and 'use bytes' aware and will coerce its args to strings
7200 if necessary. If the flags include SV_GMAGIC, it handles get magic. See
7201 also C<sv_cmp_locale_flags>.
7207 Perl_sv_cmp(pTHX_ register SV *const sv1, register SV *const sv2)
7209 return sv_cmp_flags(sv1, sv2, SV_GMAGIC);
7213 Perl_sv_cmp_flags(pTHX_ register SV *const sv1, register SV *const sv2,
7218 const char *pv1, *pv2;
7221 SV *svrecode = NULL;
7228 pv1 = SvPV_flags_const(sv1, cur1, flags);
7235 pv2 = SvPV_flags_const(sv2, cur2, flags);
7237 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
7238 /* Differing utf8ness.
7239 * Do not UTF8size the comparands as a side-effect. */
7242 svrecode = newSVpvn(pv2, cur2);
7243 sv_recode_to_utf8(svrecode, PL_encoding);
7244 pv2 = SvPV_const(svrecode, cur2);
7247 const int retval = -bytes_cmp_utf8((const U8*)pv2, cur2,
7248 (const U8*)pv1, cur1);
7249 return retval ? retval < 0 ? -1 : +1 : 0;
7254 svrecode = newSVpvn(pv1, cur1);
7255 sv_recode_to_utf8(svrecode, PL_encoding);
7256 pv1 = SvPV_const(svrecode, cur1);
7259 const int retval = bytes_cmp_utf8((const U8*)pv1, cur1,
7260 (const U8*)pv2, cur2);
7261 return retval ? retval < 0 ? -1 : +1 : 0;
7267 cmp = cur2 ? -1 : 0;
7271 const I32 retval = memcmp((const void*)pv1, (const void*)pv2, cur1 < cur2 ? cur1 : cur2);
7274 cmp = retval < 0 ? -1 : 1;
7275 } else if (cur1 == cur2) {
7278 cmp = cur1 < cur2 ? -1 : 1;
7282 SvREFCNT_dec(svrecode);
7290 =for apidoc sv_cmp_locale
7292 Compares the strings in two SVs in a locale-aware manner. Is UTF-8 and
7293 'use bytes' aware, handles get magic, and will coerce its args to strings
7294 if necessary. See also C<sv_cmp>.
7296 =for apidoc sv_cmp_locale_flags
7298 Compares the strings in two SVs in a locale-aware manner. Is UTF-8 and
7299 'use bytes' aware and will coerce its args to strings if necessary. If the
7300 flags contain SV_GMAGIC, it handles get magic. See also C<sv_cmp_flags>.
7306 Perl_sv_cmp_locale(pTHX_ register SV *const sv1, register SV *const sv2)
7308 return sv_cmp_locale_flags(sv1, sv2, SV_GMAGIC);
7312 Perl_sv_cmp_locale_flags(pTHX_ register SV *const sv1, register SV *const sv2,
7316 #ifdef USE_LOCALE_COLLATE
7322 if (PL_collation_standard)
7326 pv1 = sv1 ? sv_collxfrm_flags(sv1, &len1, flags) : (char *) NULL;
7328 pv2 = sv2 ? sv_collxfrm_flags(sv2, &len2, flags) : (char *) NULL;
7330 if (!pv1 || !len1) {
7341 retval = memcmp((void*)pv1, (void*)pv2, len1 < len2 ? len1 : len2);
7344 return retval < 0 ? -1 : 1;
7347 * When the result of collation is equality, that doesn't mean
7348 * that there are no differences -- some locales exclude some
7349 * characters from consideration. So to avoid false equalities,
7350 * we use the raw string as a tiebreaker.
7356 #endif /* USE_LOCALE_COLLATE */
7358 return sv_cmp(sv1, sv2);
7362 #ifdef USE_LOCALE_COLLATE
7365 =for apidoc sv_collxfrm
7367 This calls C<sv_collxfrm_flags> with the SV_GMAGIC flag. See
7368 C<sv_collxfrm_flags>.
7370 =for apidoc sv_collxfrm_flags
7372 Add Collate Transform magic to an SV if it doesn't already have it. If the
7373 flags contain SV_GMAGIC, it handles get-magic.
7375 Any scalar variable may carry PERL_MAGIC_collxfrm magic that contains the
7376 scalar data of the variable, but transformed to such a format that a normal
7377 memory comparison can be used to compare the data according to the locale
7384 Perl_sv_collxfrm_flags(pTHX_ SV *const sv, STRLEN *const nxp, const I32 flags)
7389 PERL_ARGS_ASSERT_SV_COLLXFRM_FLAGS;
7391 mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_collxfrm) : (MAGIC *) NULL;
7392 if (!mg || !mg->mg_ptr || *(U32*)mg->mg_ptr != PL_collation_ix) {
7398 Safefree(mg->mg_ptr);
7399 s = SvPV_flags_const(sv, len, flags);
7400 if ((xf = mem_collxfrm(s, len, &xlen))) {
7402 #ifdef PERL_OLD_COPY_ON_WRITE
7404 sv_force_normal_flags(sv, 0);
7406 mg = sv_magicext(sv, 0, PERL_MAGIC_collxfrm, &PL_vtbl_collxfrm,
7420 if (mg && mg->mg_ptr) {
7422 return mg->mg_ptr + sizeof(PL_collation_ix);
7430 #endif /* USE_LOCALE_COLLATE */
7433 S_sv_gets_append_to_utf8(pTHX_ SV *const sv, PerlIO *const fp, I32 append)
7435 SV * const tsv = newSV(0);
7438 sv_gets(tsv, fp, 0);
7439 sv_utf8_upgrade_nomg(tsv);
7440 SvCUR_set(sv,append);
7443 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7447 S_sv_gets_read_record(pTHX_ SV *const sv, PerlIO *const fp, I32 append)
7450 const U32 recsize = SvUV(SvRV(PL_rs)); /* RsRECORD() guarantees > 0. */
7451 /* Grab the size of the record we're getting */
7452 char *const buffer = SvGROW(sv, (STRLEN)(recsize + append + 1)) + append;
7459 /* VMS wants read instead of fread, because fread doesn't respect */
7460 /* RMS record boundaries. This is not necessarily a good thing to be */
7461 /* doing, but we've got no other real choice - except avoid stdio
7462 as implementation - perhaps write a :vms layer ?
7464 fd = PerlIO_fileno(fp);
7466 bytesread = PerlLIO_read(fd, buffer, recsize);
7468 else /* in-memory file from PerlIO::Scalar */
7471 bytesread = PerlIO_read(fp, buffer, recsize);
7476 SvCUR_set(sv, bytesread + append);
7477 buffer[bytesread] = '\0';
7478 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7484 Get a line from the filehandle and store it into the SV, optionally
7485 appending to the currently-stored string.
7491 Perl_sv_gets(pTHX_ register SV *const sv, register PerlIO *const fp, I32 append)
7496 register STDCHAR rslast;
7497 register STDCHAR *bp;
7502 PERL_ARGS_ASSERT_SV_GETS;
7504 if (SvTHINKFIRST(sv))
7505 sv_force_normal_flags(sv, append ? 0 : SV_COW_DROP_PV);
7506 /* XXX. If you make this PVIV, then copy on write can copy scalars read
7508 However, perlbench says it's slower, because the existing swipe code
7509 is faster than copy on write.
7510 Swings and roundabouts. */
7511 SvUPGRADE(sv, SVt_PV);
7516 if (PerlIO_isutf8(fp)) {
7518 sv_utf8_upgrade_nomg(sv);
7519 sv_pos_u2b(sv,&append,0);
7521 } else if (SvUTF8(sv)) {
7522 return S_sv_gets_append_to_utf8(aTHX_ sv, fp, append);
7530 if (PerlIO_isutf8(fp))
7533 if (IN_PERL_COMPILETIME) {
7534 /* we always read code in line mode */
7538 else if (RsSNARF(PL_rs)) {
7539 /* If it is a regular disk file use size from stat() as estimate
7540 of amount we are going to read -- may result in mallocing
7541 more memory than we really need if the layers below reduce
7542 the size we read (e.g. CRLF or a gzip layer).
7545 if (!PerlLIO_fstat(PerlIO_fileno(fp), &st) && S_ISREG(st.st_mode)) {
7546 const Off_t offset = PerlIO_tell(fp);
7547 if (offset != (Off_t) -1 && st.st_size + append > offset) {
7548 (void) SvGROW(sv, (STRLEN)((st.st_size - offset) + append + 1));
7554 else if (RsRECORD(PL_rs)) {
7555 return S_sv_gets_read_record(aTHX_ sv, fp, append);
7557 else if (RsPARA(PL_rs)) {
7563 /* Get $/ i.e. PL_rs into same encoding as stream wants */
7564 if (PerlIO_isutf8(fp)) {
7565 rsptr = SvPVutf8(PL_rs, rslen);
7568 if (SvUTF8(PL_rs)) {
7569 if (!sv_utf8_downgrade(PL_rs, TRUE)) {
7570 Perl_croak(aTHX_ "Wide character in $/");
7573 rsptr = SvPV_const(PL_rs, rslen);
7577 rslast = rslen ? rsptr[rslen - 1] : '\0';
7579 if (rspara) { /* have to do this both before and after */
7580 do { /* to make sure file boundaries work right */
7583 i = PerlIO_getc(fp);
7587 PerlIO_ungetc(fp,i);
7593 /* See if we know enough about I/O mechanism to cheat it ! */
7595 /* This used to be #ifdef test - it is made run-time test for ease
7596 of abstracting out stdio interface. One call should be cheap
7597 enough here - and may even be a macro allowing compile
7601 if (PerlIO_fast_gets(fp)) {
7604 * We're going to steal some values from the stdio struct
7605 * and put EVERYTHING in the innermost loop into registers.
7607 register STDCHAR *ptr;
7611 #if defined(VMS) && defined(PERLIO_IS_STDIO)
7612 /* An ungetc()d char is handled separately from the regular
7613 * buffer, so we getc() it back out and stuff it in the buffer.
7615 i = PerlIO_getc(fp);
7616 if (i == EOF) return 0;
7617 *(--((*fp)->_ptr)) = (unsigned char) i;
7621 /* Here is some breathtakingly efficient cheating */
7623 cnt = PerlIO_get_cnt(fp); /* get count into register */
7624 /* make sure we have the room */
7625 if ((I32)(SvLEN(sv) - append) <= cnt + 1) {
7626 /* Not room for all of it
7627 if we are looking for a separator and room for some
7629 if (rslen && cnt > 80 && (I32)SvLEN(sv) > append) {
7630 /* just process what we have room for */
7631 shortbuffered = cnt - SvLEN(sv) + append + 1;
7632 cnt -= shortbuffered;
7636 /* remember that cnt can be negative */
7637 SvGROW(sv, (STRLEN)(append + (cnt <= 0 ? 2 : (cnt + 1))));
7642 bp = (STDCHAR*)SvPVX_const(sv) + append; /* move these two too to registers */
7643 ptr = (STDCHAR*)PerlIO_get_ptr(fp);
7644 DEBUG_P(PerlIO_printf(Perl_debug_log,
7645 "Screamer: entering, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7646 DEBUG_P(PerlIO_printf(Perl_debug_log,
7647 "Screamer: entering: PerlIO * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7648 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7649 PTR2UV(PerlIO_has_base(fp) ? PerlIO_get_base(fp) : 0)));
7654 while (cnt > 0) { /* this | eat */
7656 if ((*bp++ = *ptr++) == rslast) /* really | dust */
7657 goto thats_all_folks; /* screams | sed :-) */
7661 Copy(ptr, bp, cnt, char); /* this | eat */
7662 bp += cnt; /* screams | dust */
7663 ptr += cnt; /* louder | sed :-) */
7665 assert (!shortbuffered);
7666 goto cannot_be_shortbuffered;
7670 if (shortbuffered) { /* oh well, must extend */
7671 cnt = shortbuffered;
7673 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
7675 SvGROW(sv, SvLEN(sv) + append + cnt + 2);
7676 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
7680 cannot_be_shortbuffered:
7681 DEBUG_P(PerlIO_printf(Perl_debug_log,
7682 "Screamer: going to getc, ptr=%"UVuf", cnt=%ld\n",
7683 PTR2UV(ptr),(long)cnt));
7684 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* deregisterize cnt and ptr */
7686 DEBUG_Pv(PerlIO_printf(Perl_debug_log,
7687 "Screamer: pre: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7688 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7689 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7691 /* This used to call 'filbuf' in stdio form, but as that behaves like
7692 getc when cnt <= 0 we use PerlIO_getc here to avoid introducing
7693 another abstraction. */
7694 i = PerlIO_getc(fp); /* get more characters */
7696 DEBUG_Pv(PerlIO_printf(Perl_debug_log,
7697 "Screamer: post: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7698 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7699 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7701 cnt = PerlIO_get_cnt(fp);
7702 ptr = (STDCHAR*)PerlIO_get_ptr(fp); /* reregisterize cnt and ptr */
7703 DEBUG_P(PerlIO_printf(Perl_debug_log,
7704 "Screamer: after getc, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7706 if (i == EOF) /* all done for ever? */
7707 goto thats_really_all_folks;
7709 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
7711 SvGROW(sv, bpx + cnt + 2);
7712 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
7714 *bp++ = (STDCHAR)i; /* store character from PerlIO_getc */
7716 if (rslen && (STDCHAR)i == rslast) /* all done for now? */
7717 goto thats_all_folks;
7721 if ((rslen > 1 && (STRLEN)(bp - (STDCHAR*)SvPVX_const(sv)) < rslen) ||
7722 memNE((char*)bp - rslen, rsptr, rslen))
7723 goto screamer; /* go back to the fray */
7724 thats_really_all_folks:
7726 cnt += shortbuffered;
7727 DEBUG_P(PerlIO_printf(Perl_debug_log,
7728 "Screamer: quitting, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7729 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* put these back or we're in trouble */
7730 DEBUG_P(PerlIO_printf(Perl_debug_log,
7731 "Screamer: end: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7732 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7733 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7735 SvCUR_set(sv, bp - (STDCHAR*)SvPVX_const(sv)); /* set length */
7736 DEBUG_P(PerlIO_printf(Perl_debug_log,
7737 "Screamer: done, len=%ld, string=|%.*s|\n",
7738 (long)SvCUR(sv),(int)SvCUR(sv),SvPVX_const(sv)));
7742 /*The big, slow, and stupid way. */
7743 #ifdef USE_HEAP_INSTEAD_OF_STACK /* Even slower way. */
7744 STDCHAR *buf = NULL;
7745 Newx(buf, 8192, STDCHAR);
7753 register const STDCHAR * const bpe = buf + sizeof(buf);
7755 while ((i = PerlIO_getc(fp)) != EOF && (*bp++ = (STDCHAR)i) != rslast && bp < bpe)
7756 ; /* keep reading */
7760 cnt = PerlIO_read(fp,(char*)buf, sizeof(buf));
7761 /* Accommodate broken VAXC compiler, which applies U8 cast to
7762 * both args of ?: operator, causing EOF to change into 255
7765 i = (U8)buf[cnt - 1];
7771 cnt = 0; /* we do need to re-set the sv even when cnt <= 0 */
7773 sv_catpvn(sv, (char *) buf, cnt);
7775 sv_setpvn(sv, (char *) buf, cnt);
7777 if (i != EOF && /* joy */
7779 SvCUR(sv) < rslen ||
7780 memNE(SvPVX_const(sv) + SvCUR(sv) - rslen, rsptr, rslen)))
7784 * If we're reading from a TTY and we get a short read,
7785 * indicating that the user hit his EOF character, we need
7786 * to notice it now, because if we try to read from the TTY
7787 * again, the EOF condition will disappear.
7789 * The comparison of cnt to sizeof(buf) is an optimization
7790 * that prevents unnecessary calls to feof().
7794 if (!(cnt < (I32)sizeof(buf) && PerlIO_eof(fp)))
7798 #ifdef USE_HEAP_INSTEAD_OF_STACK
7803 if (rspara) { /* have to do this both before and after */
7804 while (i != EOF) { /* to make sure file boundaries work right */
7805 i = PerlIO_getc(fp);
7807 PerlIO_ungetc(fp,i);
7813 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7819 Auto-increment of the value in the SV, doing string to numeric conversion
7820 if necessary. Handles 'get' magic and operator overloading.
7826 Perl_sv_inc(pTHX_ register SV *const sv)
7835 =for apidoc sv_inc_nomg
7837 Auto-increment of the value in the SV, doing string to numeric conversion
7838 if necessary. Handles operator overloading. Skips handling 'get' magic.
7844 Perl_sv_inc_nomg(pTHX_ register SV *const sv)
7852 if (SvTHINKFIRST(sv)) {
7853 if (SvIsCOW(sv) || isGV_with_GP(sv))
7854 sv_force_normal_flags(sv, 0);
7855 if (SvREADONLY(sv)) {
7856 if (IN_PERL_RUNTIME)
7857 Perl_croak_no_modify(aTHX);
7861 if (SvAMAGIC(sv) && AMG_CALLunary(sv, inc_amg))
7863 i = PTR2IV(SvRV(sv));
7868 flags = SvFLAGS(sv);
7869 if ((flags & (SVp_NOK|SVp_IOK)) == SVp_NOK) {
7870 /* It's (privately or publicly) a float, but not tested as an
7871 integer, so test it to see. */
7873 flags = SvFLAGS(sv);
7875 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
7876 /* It's publicly an integer, or privately an integer-not-float */
7877 #ifdef PERL_PRESERVE_IVUV
7881 if (SvUVX(sv) == UV_MAX)
7882 sv_setnv(sv, UV_MAX_P1);
7884 (void)SvIOK_only_UV(sv);
7885 SvUV_set(sv, SvUVX(sv) + 1);
7887 if (SvIVX(sv) == IV_MAX)
7888 sv_setuv(sv, (UV)IV_MAX + 1);
7890 (void)SvIOK_only(sv);
7891 SvIV_set(sv, SvIVX(sv) + 1);
7896 if (flags & SVp_NOK) {
7897 const NV was = SvNVX(sv);
7898 if (NV_OVERFLOWS_INTEGERS_AT &&
7899 was >= NV_OVERFLOWS_INTEGERS_AT) {
7900 Perl_ck_warner(aTHX_ packWARN(WARN_IMPRECISION),
7901 "Lost precision when incrementing %" NVff " by 1",
7904 (void)SvNOK_only(sv);
7905 SvNV_set(sv, was + 1.0);
7909 if (!(flags & SVp_POK) || !*SvPVX_const(sv)) {
7910 if ((flags & SVTYPEMASK) < SVt_PVIV)
7911 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV ? SVt_PVIV : SVt_IV));
7912 (void)SvIOK_only(sv);
7917 while (isALPHA(*d)) d++;
7918 while (isDIGIT(*d)) d++;
7919 if (d < SvEND(sv)) {
7920 #ifdef PERL_PRESERVE_IVUV
7921 /* Got to punt this as an integer if needs be, but we don't issue
7922 warnings. Probably ought to make the sv_iv_please() that does
7923 the conversion if possible, and silently. */
7924 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
7925 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
7926 /* Need to try really hard to see if it's an integer.
7927 9.22337203685478e+18 is an integer.
7928 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
7929 so $a="9.22337203685478e+18"; $a+0; $a++
7930 needs to be the same as $a="9.22337203685478e+18"; $a++
7937 /* sv_2iv *should* have made this an NV */
7938 if (flags & SVp_NOK) {
7939 (void)SvNOK_only(sv);
7940 SvNV_set(sv, SvNVX(sv) + 1.0);
7943 /* I don't think we can get here. Maybe I should assert this
7944 And if we do get here I suspect that sv_setnv will croak. NWC
7946 #if defined(USE_LONG_DOUBLE)
7947 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",
7948 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7950 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_inc punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
7951 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7954 #endif /* PERL_PRESERVE_IVUV */
7955 sv_setnv(sv,Atof(SvPVX_const(sv)) + 1.0);
7959 while (d >= SvPVX_const(sv)) {
7967 /* MKS: The original code here died if letters weren't consecutive.
7968 * at least it didn't have to worry about non-C locales. The
7969 * new code assumes that ('z'-'a')==('Z'-'A'), letters are
7970 * arranged in order (although not consecutively) and that only
7971 * [A-Za-z] are accepted by isALPHA in the C locale.
7973 if (*d != 'z' && *d != 'Z') {
7974 do { ++*d; } while (!isALPHA(*d));
7977 *(d--) -= 'z' - 'a';
7982 *(d--) -= 'z' - 'a' + 1;
7986 /* oh,oh, the number grew */
7987 SvGROW(sv, SvCUR(sv) + 2);
7988 SvCUR_set(sv, SvCUR(sv) + 1);
7989 for (d = SvPVX(sv) + SvCUR(sv); d > SvPVX_const(sv); d--)
8000 Auto-decrement of the value in the SV, doing string to numeric conversion
8001 if necessary. Handles 'get' magic and operator overloading.
8007 Perl_sv_dec(pTHX_ register SV *const sv)
8017 =for apidoc sv_dec_nomg
8019 Auto-decrement of the value in the SV, doing string to numeric conversion
8020 if necessary. Handles operator overloading. Skips handling 'get' magic.
8026 Perl_sv_dec_nomg(pTHX_ register SV *const sv)
8033 if (SvTHINKFIRST(sv)) {
8034 if (SvIsCOW(sv) || isGV_with_GP(sv))
8035 sv_force_normal_flags(sv, 0);
8036 if (SvREADONLY(sv)) {
8037 if (IN_PERL_RUNTIME)
8038 Perl_croak_no_modify(aTHX);
8042 if (SvAMAGIC(sv) && AMG_CALLunary(sv, dec_amg))
8044 i = PTR2IV(SvRV(sv));
8049 /* Unlike sv_inc we don't have to worry about string-never-numbers
8050 and keeping them magic. But we mustn't warn on punting */
8051 flags = SvFLAGS(sv);
8052 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
8053 /* It's publicly an integer, or privately an integer-not-float */
8054 #ifdef PERL_PRESERVE_IVUV
8058 if (SvUVX(sv) == 0) {
8059 (void)SvIOK_only(sv);
8063 (void)SvIOK_only_UV(sv);
8064 SvUV_set(sv, SvUVX(sv) - 1);
8067 if (SvIVX(sv) == IV_MIN) {
8068 sv_setnv(sv, (NV)IV_MIN);
8072 (void)SvIOK_only(sv);
8073 SvIV_set(sv, SvIVX(sv) - 1);
8078 if (flags & SVp_NOK) {
8081 const NV was = SvNVX(sv);
8082 if (NV_OVERFLOWS_INTEGERS_AT &&
8083 was <= -NV_OVERFLOWS_INTEGERS_AT) {
8084 Perl_ck_warner(aTHX_ packWARN(WARN_IMPRECISION),
8085 "Lost precision when decrementing %" NVff " by 1",
8088 (void)SvNOK_only(sv);
8089 SvNV_set(sv, was - 1.0);
8093 if (!(flags & SVp_POK)) {
8094 if ((flags & SVTYPEMASK) < SVt_PVIV)
8095 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV) ? SVt_PVIV : SVt_IV);
8097 (void)SvIOK_only(sv);
8100 #ifdef PERL_PRESERVE_IVUV
8102 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
8103 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
8104 /* Need to try really hard to see if it's an integer.
8105 9.22337203685478e+18 is an integer.
8106 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
8107 so $a="9.22337203685478e+18"; $a+0; $a--
8108 needs to be the same as $a="9.22337203685478e+18"; $a--
8115 /* sv_2iv *should* have made this an NV */
8116 if (flags & SVp_NOK) {
8117 (void)SvNOK_only(sv);
8118 SvNV_set(sv, SvNVX(sv) - 1.0);
8121 /* I don't think we can get here. Maybe I should assert this
8122 And if we do get here I suspect that sv_setnv will croak. NWC
8124 #if defined(USE_LONG_DOUBLE)
8125 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",
8126 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
8128 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_dec punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
8129 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
8133 #endif /* PERL_PRESERVE_IVUV */
8134 sv_setnv(sv,Atof(SvPVX_const(sv)) - 1.0); /* punt */
8137 /* this define is used to eliminate a chunk of duplicated but shared logic
8138 * it has the suffix __SV_C to signal that it isnt API, and isnt meant to be
8139 * used anywhere but here - yves
8141 #define PUSH_EXTEND_MORTAL__SV_C(AnSv) \
8144 PL_tmps_stack[++PL_tmps_ix] = (AnSv); \
8148 =for apidoc sv_mortalcopy
8150 Creates a new SV which is a copy of the original SV (using C<sv_setsv>).
8151 The new SV is marked as mortal. It will be destroyed "soon", either by an
8152 explicit call to FREETMPS, or by an implicit call at places such as
8153 statement boundaries. See also C<sv_newmortal> and C<sv_2mortal>.
8158 /* Make a string that will exist for the duration of the expression
8159 * evaluation. Actually, it may have to last longer than that, but
8160 * hopefully we won't free it until it has been assigned to a
8161 * permanent location. */
8164 Perl_sv_mortalcopy(pTHX_ SV *const oldstr)
8170 sv_setsv(sv,oldstr);
8171 PUSH_EXTEND_MORTAL__SV_C(sv);
8177 =for apidoc sv_newmortal
8179 Creates a new null SV which is mortal. The reference count of the SV is
8180 set to 1. It will be destroyed "soon", either by an explicit call to
8181 FREETMPS, or by an implicit call at places such as statement boundaries.
8182 See also C<sv_mortalcopy> and C<sv_2mortal>.
8188 Perl_sv_newmortal(pTHX)
8194 SvFLAGS(sv) = SVs_TEMP;
8195 PUSH_EXTEND_MORTAL__SV_C(sv);
8201 =for apidoc newSVpvn_flags
8203 Creates a new SV and copies a string into it. The reference count for the
8204 SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
8205 string. You are responsible for ensuring that the source string is at least
8206 C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
8207 Currently the only flag bits accepted are C<SVf_UTF8> and C<SVs_TEMP>.
8208 If C<SVs_TEMP> is set, then C<sv_2mortal()> is called on the result before
8209 returning. If C<SVf_UTF8> is set, C<s> is considered to be in UTF-8 and the
8210 C<SVf_UTF8> flag will be set on the new SV.
8211 C<newSVpvn_utf8()> is a convenience wrapper for this function, defined as
8213 #define newSVpvn_utf8(s, len, u) \
8214 newSVpvn_flags((s), (len), (u) ? SVf_UTF8 : 0)
8220 Perl_newSVpvn_flags(pTHX_ const char *const s, const STRLEN len, const U32 flags)
8225 /* All the flags we don't support must be zero.
8226 And we're new code so I'm going to assert this from the start. */
8227 assert(!(flags & ~(SVf_UTF8|SVs_TEMP)));
8229 sv_setpvn(sv,s,len);
8231 /* This code used to a sv_2mortal(), however we now unroll the call to sv_2mortal()
8232 * and do what it does ourselves here.
8233 * Since we have asserted that flags can only have the SVf_UTF8 and/or SVs_TEMP flags
8234 * set above we can use it to enable the sv flags directly (bypassing SvTEMP_on), which
8235 * in turn means we dont need to mask out the SVf_UTF8 flag below, which means that we
8236 * eliminate quite a few steps than it looks - Yves (explaining patch by gfx)
8239 SvFLAGS(sv) |= flags;
8241 if(flags & SVs_TEMP){
8242 PUSH_EXTEND_MORTAL__SV_C(sv);
8249 =for apidoc sv_2mortal
8251 Marks an existing SV as mortal. The SV will be destroyed "soon", either
8252 by an explicit call to FREETMPS, or by an implicit call at places such as
8253 statement boundaries. SvTEMP() is turned on which means that the SV's
8254 string buffer can be "stolen" if this SV is copied. See also C<sv_newmortal>
8255 and C<sv_mortalcopy>.
8261 Perl_sv_2mortal(pTHX_ register SV *const sv)
8266 if (SvREADONLY(sv) && SvIMMORTAL(sv))
8268 PUSH_EXTEND_MORTAL__SV_C(sv);
8276 Creates a new SV and copies a string into it. The reference count for the
8277 SV is set to 1. If C<len> is zero, Perl will compute the length using
8278 strlen(). For efficiency, consider using C<newSVpvn> instead.
8284 Perl_newSVpv(pTHX_ const char *const s, const STRLEN len)
8290 sv_setpvn(sv, s, len || s == NULL ? len : strlen(s));
8295 =for apidoc newSVpvn
8297 Creates a new SV and copies a string into it. The reference count for the
8298 SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
8299 string. You are responsible for ensuring that the source string is at least
8300 C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
8306 Perl_newSVpvn(pTHX_ const char *const s, const STRLEN len)
8312 sv_setpvn(sv,s,len);
8317 =for apidoc newSVhek
8319 Creates a new SV from the hash key structure. It will generate scalars that
8320 point to the shared string table where possible. Returns a new (undefined)
8321 SV if the hek is NULL.
8327 Perl_newSVhek(pTHX_ const HEK *const hek)
8337 if (HEK_LEN(hek) == HEf_SVKEY) {
8338 return newSVsv(*(SV**)HEK_KEY(hek));
8340 const int flags = HEK_FLAGS(hek);
8341 if (flags & HVhek_WASUTF8) {
8343 Andreas would like keys he put in as utf8 to come back as utf8
8345 STRLEN utf8_len = HEK_LEN(hek);
8346 SV * const sv = newSV_type(SVt_PV);
8347 char *as_utf8 = (char *)bytes_to_utf8 ((U8*)HEK_KEY(hek), &utf8_len);
8348 /* bytes_to_utf8() allocates a new string, which we can repurpose: */
8349 sv_usepvn_flags(sv, as_utf8, utf8_len, SV_HAS_TRAILING_NUL);
8352 } else if (flags & (HVhek_REHASH|HVhek_UNSHARED)) {
8353 /* We don't have a pointer to the hv, so we have to replicate the
8354 flag into every HEK. This hv is using custom a hasing
8355 algorithm. Hence we can't return a shared string scalar, as
8356 that would contain the (wrong) hash value, and might get passed
8357 into an hv routine with a regular hash.
8358 Similarly, a hash that isn't using shared hash keys has to have
8359 the flag in every key so that we know not to try to call
8360 share_hek_hek on it. */
8362 SV * const sv = newSVpvn (HEK_KEY(hek), HEK_LEN(hek));
8367 /* This will be overwhelminly the most common case. */
8369 /* Inline most of newSVpvn_share(), because share_hek_hek() is far
8370 more efficient than sharepvn(). */
8374 sv_upgrade(sv, SVt_PV);
8375 SvPV_set(sv, (char *)HEK_KEY(share_hek_hek(hek)));
8376 SvCUR_set(sv, HEK_LEN(hek));
8389 =for apidoc newSVpvn_share
8391 Creates a new SV with its SvPVX_const pointing to a shared string in the string
8392 table. If the string does not already exist in the table, it is created
8393 first. Turns on READONLY and FAKE. If the C<hash> parameter is non-zero, that
8394 value is used; otherwise the hash is computed. The string's hash can be later
8395 be retrieved from the SV with the C<SvSHARED_HASH()> macro. The idea here is
8396 that as the string table is used for shared hash keys these strings will have
8397 SvPVX_const == HeKEY and hash lookup will avoid string compare.
8403 Perl_newSVpvn_share(pTHX_ const char *src, I32 len, U32 hash)
8407 bool is_utf8 = FALSE;
8408 const char *const orig_src = src;
8411 STRLEN tmplen = -len;
8413 /* See the note in hv.c:hv_fetch() --jhi */
8414 src = (char*)bytes_from_utf8((const U8*)src, &tmplen, &is_utf8);
8418 PERL_HASH(hash, src, len);
8420 /* The logic for this is inlined in S_mro_get_linear_isa_dfs(), so if it
8421 changes here, update it there too. */
8422 sv_upgrade(sv, SVt_PV);
8423 SvPV_set(sv, sharepvn(src, is_utf8?-len:len, hash));
8431 if (src != orig_src)
8437 =for apidoc newSVpv_share
8439 Like C<newSVpvn_share>, but takes a nul-terminated string instead of a
8446 Perl_newSVpv_share(pTHX_ const char *src, U32 hash)
8448 return newSVpvn_share(src, strlen(src), hash);
8451 #if defined(PERL_IMPLICIT_CONTEXT)
8453 /* pTHX_ magic can't cope with varargs, so this is a no-context
8454 * version of the main function, (which may itself be aliased to us).
8455 * Don't access this version directly.
8459 Perl_newSVpvf_nocontext(const char *const pat, ...)
8465 PERL_ARGS_ASSERT_NEWSVPVF_NOCONTEXT;
8467 va_start(args, pat);
8468 sv = vnewSVpvf(pat, &args);
8475 =for apidoc newSVpvf
8477 Creates a new SV and initializes it with the string formatted like
8484 Perl_newSVpvf(pTHX_ const char *const pat, ...)
8489 PERL_ARGS_ASSERT_NEWSVPVF;
8491 va_start(args, pat);
8492 sv = vnewSVpvf(pat, &args);
8497 /* backend for newSVpvf() and newSVpvf_nocontext() */
8500 Perl_vnewSVpvf(pTHX_ const char *const pat, va_list *const args)
8505 PERL_ARGS_ASSERT_VNEWSVPVF;
8508 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
8515 Creates a new SV and copies a floating point value into it.
8516 The reference count for the SV is set to 1.
8522 Perl_newSVnv(pTHX_ const NV n)
8535 Creates a new SV and copies an integer into it. The reference count for the
8542 Perl_newSViv(pTHX_ const IV i)
8555 Creates a new SV and copies an unsigned integer into it.
8556 The reference count for the SV is set to 1.
8562 Perl_newSVuv(pTHX_ const UV u)
8573 =for apidoc newSV_type
8575 Creates a new SV, of the type specified. The reference count for the new SV
8582 Perl_newSV_type(pTHX_ const svtype type)
8587 sv_upgrade(sv, type);
8592 =for apidoc newRV_noinc
8594 Creates an RV wrapper for an SV. The reference count for the original
8595 SV is B<not> incremented.
8601 Perl_newRV_noinc(pTHX_ SV *const tmpRef)
8604 register SV *sv = newSV_type(SVt_IV);
8606 PERL_ARGS_ASSERT_NEWRV_NOINC;
8609 SvRV_set(sv, tmpRef);
8614 /* newRV_inc is the official function name to use now.
8615 * newRV_inc is in fact #defined to newRV in sv.h
8619 Perl_newRV(pTHX_ SV *const sv)
8623 PERL_ARGS_ASSERT_NEWRV;
8625 return newRV_noinc(SvREFCNT_inc_simple_NN(sv));
8631 Creates a new SV which is an exact duplicate of the original SV.
8638 Perl_newSVsv(pTHX_ register SV *const old)
8645 if (SvTYPE(old) == (svtype)SVTYPEMASK) {
8646 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL), "semi-panic: attempt to dup freed string");
8650 /* SV_GMAGIC is the default for sv_setv()
8651 SV_NOSTEAL prevents TEMP buffers being, well, stolen, and saves games
8652 with SvTEMP_off and SvTEMP_on round a call to sv_setsv. */
8653 sv_setsv_flags(sv, old, SV_GMAGIC | SV_NOSTEAL);
8658 =for apidoc sv_reset
8660 Underlying implementation for the C<reset> Perl function.
8661 Note that the perl-level function is vaguely deprecated.
8667 Perl_sv_reset(pTHX_ register const char *s, HV *const stash)
8670 char todo[PERL_UCHAR_MAX+1];
8672 PERL_ARGS_ASSERT_SV_RESET;
8677 if (!*s) { /* reset ?? searches */
8678 MAGIC * const mg = mg_find((const SV *)stash, PERL_MAGIC_symtab);
8680 const U32 count = mg->mg_len / sizeof(PMOP**);
8681 PMOP **pmp = (PMOP**) mg->mg_ptr;
8682 PMOP *const *const end = pmp + count;
8686 SvREADONLY_off(PL_regex_pad[(*pmp)->op_pmoffset]);
8688 (*pmp)->op_pmflags &= ~PMf_USED;
8696 /* reset variables */
8698 if (!HvARRAY(stash))
8701 Zero(todo, 256, char);
8704 I32 i = (unsigned char)*s;
8708 max = (unsigned char)*s++;
8709 for ( ; i <= max; i++) {
8712 for (i = 0; i <= (I32) HvMAX(stash); i++) {
8714 for (entry = HvARRAY(stash)[i];
8716 entry = HeNEXT(entry))
8721 if (!todo[(U8)*HeKEY(entry)])
8723 gv = MUTABLE_GV(HeVAL(entry));
8726 if (SvTHINKFIRST(sv)) {
8727 if (!SvREADONLY(sv) && SvROK(sv))
8729 /* XXX Is this continue a bug? Why should THINKFIRST
8730 exempt us from resetting arrays and hashes? */
8734 if (SvTYPE(sv) >= SVt_PV) {
8736 if (SvPVX_const(sv) != NULL)
8744 if (GvHV(gv) && !HvNAME_get(GvHV(gv))) {
8746 Perl_die(aTHX_ "Can't reset %%ENV on this system");
8749 # if defined(USE_ENVIRON_ARRAY)
8752 # endif /* USE_ENVIRON_ARRAY */
8763 Using various gambits, try to get an IO from an SV: the IO slot if its a
8764 GV; or the recursive result if we're an RV; or the IO slot of the symbol
8765 named after the PV if we're a string.
8771 Perl_sv_2io(pTHX_ SV *const sv)
8776 PERL_ARGS_ASSERT_SV_2IO;
8778 switch (SvTYPE(sv)) {
8780 io = MUTABLE_IO(sv);
8784 if (isGV_with_GP(sv)) {
8785 gv = MUTABLE_GV(sv);
8788 Perl_croak(aTHX_ "Bad filehandle: %s", GvNAME(gv));
8794 Perl_croak(aTHX_ PL_no_usym, "filehandle");
8796 return sv_2io(SvRV(sv));
8797 gv = gv_fetchsv(sv, 0, SVt_PVIO);
8803 Perl_croak(aTHX_ "Bad filehandle: %"SVf, SVfARG(sv));
8812 Using various gambits, try to get a CV from an SV; in addition, try if
8813 possible to set C<*st> and C<*gvp> to the stash and GV associated with it.
8814 The flags in C<lref> are passed to gv_fetchsv.
8820 Perl_sv_2cv(pTHX_ SV *sv, HV **const st, GV **const gvp, const I32 lref)
8826 PERL_ARGS_ASSERT_SV_2CV;
8833 switch (SvTYPE(sv)) {
8837 return MUTABLE_CV(sv);
8847 sv = amagic_deref_call(sv, to_cv_amg);
8848 /* At this point I'd like to do SPAGAIN, but really I need to
8849 force it upon my callers. Hmmm. This is a mess... */
8852 if (SvTYPE(sv) == SVt_PVCV) {
8853 cv = MUTABLE_CV(sv);
8858 else if(isGV_with_GP(sv))
8859 gv = MUTABLE_GV(sv);
8861 Perl_croak(aTHX_ "Not a subroutine reference");
8863 else if (isGV_with_GP(sv)) {
8864 gv = MUTABLE_GV(sv);
8867 gv = gv_fetchsv_nomg(sv, lref, SVt_PVCV);
8874 /* Some flags to gv_fetchsv mean don't really create the GV */
8875 if (!isGV_with_GP(gv)) {
8880 if (lref & ~GV_ADDMG && !GvCVu(gv)) {
8884 gv_efullname3(tmpsv, gv, NULL);
8885 /* XXX this is probably not what they think they're getting.
8886 * It has the same effect as "sub name;", i.e. just a forward
8888 newSUB(start_subparse(FALSE, 0),
8889 newSVOP(OP_CONST, 0, tmpsv),
8893 Perl_croak(aTHX_ "Unable to create sub named \"%"SVf"\"",
8894 SVfARG(SvOK(sv) ? sv : &PL_sv_no));
8903 Returns true if the SV has a true value by Perl's rules.
8904 Use the C<SvTRUE> macro instead, which may call C<sv_true()> or may
8905 instead use an in-line version.
8911 Perl_sv_true(pTHX_ register SV *const sv)
8916 register const XPV* const tXpv = (XPV*)SvANY(sv);
8918 (tXpv->xpv_cur > 1 ||
8919 (tXpv->xpv_cur && *sv->sv_u.svu_pv != '0')))
8926 return SvIVX(sv) != 0;
8929 return SvNVX(sv) != 0.0;
8931 return sv_2bool(sv);
8937 =for apidoc sv_pvn_force
8939 Get a sensible string out of the SV somehow.
8940 A private implementation of the C<SvPV_force> macro for compilers which
8941 can't cope with complex macro expressions. Always use the macro instead.
8943 =for apidoc sv_pvn_force_flags
8945 Get a sensible string out of the SV somehow.
8946 If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<sv> if
8947 appropriate, else not. C<sv_pvn_force> and C<sv_pvn_force_nomg> are
8948 implemented in terms of this function.
8949 You normally want to use the various wrapper macros instead: see
8950 C<SvPV_force> and C<SvPV_force_nomg>
8956 Perl_sv_pvn_force_flags(pTHX_ SV *const sv, STRLEN *const lp, const I32 flags)
8960 PERL_ARGS_ASSERT_SV_PVN_FORCE_FLAGS;
8962 if (SvTHINKFIRST(sv) && !SvROK(sv))
8963 sv_force_normal_flags(sv, 0);
8973 if (SvREADONLY(sv) && !(flags & SV_MUTABLE_RETURN)) {
8974 const char * const ref = sv_reftype(sv,0);
8976 Perl_croak(aTHX_ "Can't coerce readonly %s to string in %s",
8977 ref, OP_DESC(PL_op));
8979 Perl_croak(aTHX_ "Can't coerce readonly %s to string", ref);
8981 if ((SvTYPE(sv) > SVt_PVLV && SvTYPE(sv) != SVt_PVFM)
8982 || isGV_with_GP(sv))
8983 /* diag_listed_as: Can't coerce %s to %s in %s */
8984 Perl_croak(aTHX_ "Can't coerce %s to string in %s", sv_reftype(sv,0),
8986 s = sv_2pv_flags(sv, &len, flags);
8990 if (s != SvPVX_const(sv)) { /* Almost, but not quite, sv_setpvn() */
8993 SvUPGRADE(sv, SVt_PV); /* Never FALSE */
8994 SvGROW(sv, len + 1);
8995 Move(s,SvPVX(sv),len,char);
8997 SvPVX(sv)[len] = '\0';
9000 SvPOK_on(sv); /* validate pointer */
9002 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
9003 PTR2UV(sv),SvPVX_const(sv)));
9006 return SvPVX_mutable(sv);
9010 =for apidoc sv_pvbyten_force
9012 The backend for the C<SvPVbytex_force> macro. Always use the macro instead.
9018 Perl_sv_pvbyten_force(pTHX_ SV *const sv, STRLEN *const lp)
9020 PERL_ARGS_ASSERT_SV_PVBYTEN_FORCE;
9022 sv_pvn_force(sv,lp);
9023 sv_utf8_downgrade(sv,0);
9029 =for apidoc sv_pvutf8n_force
9031 The backend for the C<SvPVutf8x_force> macro. Always use the macro instead.
9037 Perl_sv_pvutf8n_force(pTHX_ SV *const sv, STRLEN *const lp)
9039 PERL_ARGS_ASSERT_SV_PVUTF8N_FORCE;
9041 sv_pvn_force(sv,lp);
9042 sv_utf8_upgrade(sv);
9048 =for apidoc sv_reftype
9050 Returns a string describing what the SV is a reference to.
9056 Perl_sv_reftype(pTHX_ const SV *const sv, const int ob)
9058 PERL_ARGS_ASSERT_SV_REFTYPE;
9060 /* The fact that I don't need to downcast to char * everywhere, only in ?:
9061 inside return suggests a const propagation bug in g++. */
9062 if (ob && SvOBJECT(sv)) {
9063 char * const name = HvNAME_get(SvSTASH(sv));
9064 return name ? name : (char *) "__ANON__";
9067 switch (SvTYPE(sv)) {
9082 case SVt_PVLV: return (char *) (SvROK(sv) ? "REF"
9083 /* tied lvalues should appear to be
9084 * scalars for backwards compatibility */
9085 : (LvTYPE(sv) == 't' || LvTYPE(sv) == 'T')
9086 ? "SCALAR" : "LVALUE");
9087 case SVt_PVAV: return "ARRAY";
9088 case SVt_PVHV: return "HASH";
9089 case SVt_PVCV: return "CODE";
9090 case SVt_PVGV: return (char *) (isGV_with_GP(sv)
9091 ? "GLOB" : "SCALAR");
9092 case SVt_PVFM: return "FORMAT";
9093 case SVt_PVIO: return "IO";
9094 case SVt_BIND: return "BIND";
9095 case SVt_REGEXP: return "REGEXP";
9096 default: return "UNKNOWN";
9102 =for apidoc sv_isobject
9104 Returns a boolean indicating whether the SV is an RV pointing to a blessed
9105 object. If the SV is not an RV, or if the object is not blessed, then this
9112 Perl_sv_isobject(pTHX_ SV *sv)
9128 Returns a boolean indicating whether the SV is blessed into the specified
9129 class. This does not check for subtypes; use C<sv_derived_from> to verify
9130 an inheritance relationship.
9136 Perl_sv_isa(pTHX_ SV *sv, const char *const name)
9140 PERL_ARGS_ASSERT_SV_ISA;
9150 hvname = HvNAME_get(SvSTASH(sv));
9154 return strEQ(hvname, name);
9160 Creates a new SV for the RV, C<rv>, to point to. If C<rv> is not an RV then
9161 it will be upgraded to one. If C<classname> is non-null then the new SV will
9162 be blessed in the specified package. The new SV is returned and its
9163 reference count is 1.
9169 Perl_newSVrv(pTHX_ SV *const rv, const char *const classname)
9174 PERL_ARGS_ASSERT_NEWSVRV;
9178 SV_CHECK_THINKFIRST_COW_DROP(rv);
9179 (void)SvAMAGIC_off(rv);
9181 if (SvTYPE(rv) >= SVt_PVMG) {
9182 const U32 refcnt = SvREFCNT(rv);
9186 SvREFCNT(rv) = refcnt;
9188 sv_upgrade(rv, SVt_IV);
9189 } else if (SvROK(rv)) {
9190 SvREFCNT_dec(SvRV(rv));
9192 prepare_SV_for_RV(rv);
9200 HV* const stash = gv_stashpv(classname, GV_ADD);
9201 (void)sv_bless(rv, stash);
9207 =for apidoc sv_setref_pv
9209 Copies a pointer into a new SV, optionally blessing the SV. The C<rv>
9210 argument will be upgraded to an RV. That RV will be modified to point to
9211 the new SV. If the C<pv> argument is NULL then C<PL_sv_undef> will be placed
9212 into the SV. The C<classname> argument indicates the package for the
9213 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9214 will have a reference count of 1, and the RV will be returned.
9216 Do not use with other Perl types such as HV, AV, SV, CV, because those
9217 objects will become corrupted by the pointer copy process.
9219 Note that C<sv_setref_pvn> copies the string while this copies the pointer.
9225 Perl_sv_setref_pv(pTHX_ SV *const rv, const char *const classname, void *const pv)
9229 PERL_ARGS_ASSERT_SV_SETREF_PV;
9232 sv_setsv(rv, &PL_sv_undef);
9236 sv_setiv(newSVrv(rv,classname), PTR2IV(pv));
9241 =for apidoc sv_setref_iv
9243 Copies an integer into a new SV, optionally blessing the SV. The C<rv>
9244 argument will be upgraded to an RV. That RV will be modified to point to
9245 the new SV. The C<classname> argument indicates the package for the
9246 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9247 will have a reference count of 1, and the RV will be returned.
9253 Perl_sv_setref_iv(pTHX_ SV *const rv, const char *const classname, const IV iv)
9255 PERL_ARGS_ASSERT_SV_SETREF_IV;
9257 sv_setiv(newSVrv(rv,classname), iv);
9262 =for apidoc sv_setref_uv
9264 Copies an unsigned integer into a new SV, optionally blessing the SV. The C<rv>
9265 argument will be upgraded to an RV. That RV will be modified to point to
9266 the new SV. The C<classname> argument indicates the package for the
9267 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9268 will have a reference count of 1, and the RV will be returned.
9274 Perl_sv_setref_uv(pTHX_ SV *const rv, const char *const classname, const UV uv)
9276 PERL_ARGS_ASSERT_SV_SETREF_UV;
9278 sv_setuv(newSVrv(rv,classname), uv);
9283 =for apidoc sv_setref_nv
9285 Copies a double into a new SV, optionally blessing the SV. The C<rv>
9286 argument will be upgraded to an RV. That RV will be modified to point to
9287 the new SV. The C<classname> argument indicates the package for the
9288 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9289 will have a reference count of 1, and the RV will be returned.
9295 Perl_sv_setref_nv(pTHX_ SV *const rv, const char *const classname, const NV nv)
9297 PERL_ARGS_ASSERT_SV_SETREF_NV;
9299 sv_setnv(newSVrv(rv,classname), nv);
9304 =for apidoc sv_setref_pvn
9306 Copies a string into a new SV, optionally blessing the SV. The length of the
9307 string must be specified with C<n>. The C<rv> argument will be upgraded to
9308 an RV. That RV will be modified to point to the new SV. The C<classname>
9309 argument indicates the package for the blessing. Set C<classname> to
9310 C<NULL> to avoid the blessing. The new SV will have a reference count
9311 of 1, and the RV will be returned.
9313 Note that C<sv_setref_pv> copies the pointer while this copies the string.
9319 Perl_sv_setref_pvn(pTHX_ SV *const rv, const char *const classname,
9320 const char *const pv, const STRLEN n)
9322 PERL_ARGS_ASSERT_SV_SETREF_PVN;
9324 sv_setpvn(newSVrv(rv,classname), pv, n);
9329 =for apidoc sv_bless
9331 Blesses an SV into a specified package. The SV must be an RV. The package
9332 must be designated by its stash (see C<gv_stashpv()>). The reference count
9333 of the SV is unaffected.
9339 Perl_sv_bless(pTHX_ SV *const sv, HV *const stash)
9344 PERL_ARGS_ASSERT_SV_BLESS;
9347 Perl_croak(aTHX_ "Can't bless non-reference value");
9349 if (SvFLAGS(tmpRef) & (SVs_OBJECT|SVf_READONLY)) {
9350 if (SvIsCOW(tmpRef))
9351 sv_force_normal_flags(tmpRef, 0);
9352 if (SvREADONLY(tmpRef))
9353 Perl_croak_no_modify(aTHX);
9354 if (SvOBJECT(tmpRef)) {
9355 if (SvTYPE(tmpRef) != SVt_PVIO)
9357 SvREFCNT_dec(SvSTASH(tmpRef));
9360 SvOBJECT_on(tmpRef);
9361 if (SvTYPE(tmpRef) != SVt_PVIO)
9363 SvUPGRADE(tmpRef, SVt_PVMG);
9364 SvSTASH_set(tmpRef, MUTABLE_HV(SvREFCNT_inc_simple(stash)));
9369 (void)SvAMAGIC_off(sv);
9371 if(SvSMAGICAL(tmpRef))
9372 if(mg_find(tmpRef, PERL_MAGIC_ext) || mg_find(tmpRef, PERL_MAGIC_uvar))
9380 /* Downgrades a PVGV to a PVMG. If it’s actually a PVLV, we leave the type
9381 * as it is after unglobbing it.
9385 S_sv_unglob(pTHX_ SV *const sv)
9390 SV * const temp = sv_newmortal();
9392 PERL_ARGS_ASSERT_SV_UNGLOB;
9394 assert(SvTYPE(sv) == SVt_PVGV || SvTYPE(sv) == SVt_PVLV);
9396 gv_efullname3(temp, MUTABLE_GV(sv), "*");
9399 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
9400 && HvNAME_get(stash))
9401 mro_method_changed_in(stash);
9402 gp_free(MUTABLE_GV(sv));
9405 sv_del_backref(MUTABLE_SV(GvSTASH(sv)), sv);
9409 if (GvNAME_HEK(sv)) {
9410 unshare_hek(GvNAME_HEK(sv));
9412 isGV_with_GP_off(sv);
9414 if(SvTYPE(sv) == SVt_PVGV) {
9415 /* need to keep SvANY(sv) in the right arena */
9416 xpvmg = new_XPVMG();
9417 StructCopy(SvANY(sv), xpvmg, XPVMG);
9418 del_XPVGV(SvANY(sv));
9421 SvFLAGS(sv) &= ~SVTYPEMASK;
9422 SvFLAGS(sv) |= SVt_PVMG;
9425 /* Intentionally not calling any local SET magic, as this isn't so much a
9426 set operation as merely an internal storage change. */
9427 sv_setsv_flags(sv, temp, 0);
9431 =for apidoc sv_unref_flags
9433 Unsets the RV status of the SV, and decrements the reference count of
9434 whatever was being referenced by the RV. This can almost be thought of
9435 as a reversal of C<newSVrv>. The C<cflags> argument can contain
9436 C<SV_IMMEDIATE_UNREF> to force the reference count to be decremented
9437 (otherwise the decrementing is conditional on the reference count being
9438 different from one or the reference being a readonly SV).
9445 Perl_sv_unref_flags(pTHX_ SV *const ref, const U32 flags)
9447 SV* const target = SvRV(ref);
9449 PERL_ARGS_ASSERT_SV_UNREF_FLAGS;
9451 if (SvWEAKREF(ref)) {
9452 sv_del_backref(target, ref);
9454 SvRV_set(ref, NULL);
9457 SvRV_set(ref, NULL);
9459 /* You can't have a || SvREADONLY(target) here, as $a = $$a, where $a was
9460 assigned to as BEGIN {$a = \"Foo"} will fail. */
9461 if (SvREFCNT(target) != 1 || (flags & SV_IMMEDIATE_UNREF))
9462 SvREFCNT_dec(target);
9463 else /* XXX Hack, but hard to make $a=$a->[1] work otherwise */
9464 sv_2mortal(target); /* Schedule for freeing later */
9468 =for apidoc sv_untaint
9470 Untaint an SV. Use C<SvTAINTED_off> instead.
9476 Perl_sv_untaint(pTHX_ SV *const sv)
9478 PERL_ARGS_ASSERT_SV_UNTAINT;
9480 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
9481 MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
9488 =for apidoc sv_tainted
9490 Test an SV for taintedness. Use C<SvTAINTED> instead.
9496 Perl_sv_tainted(pTHX_ SV *const sv)
9498 PERL_ARGS_ASSERT_SV_TAINTED;
9500 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
9501 const MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
9502 if (mg && (mg->mg_len & 1) )
9509 =for apidoc sv_setpviv
9511 Copies an integer into the given SV, also updating its string value.
9512 Does not handle 'set' magic. See C<sv_setpviv_mg>.
9518 Perl_sv_setpviv(pTHX_ SV *const sv, const IV iv)
9520 char buf[TYPE_CHARS(UV)];
9522 char * const ptr = uiv_2buf(buf, iv, 0, 0, &ebuf);
9524 PERL_ARGS_ASSERT_SV_SETPVIV;
9526 sv_setpvn(sv, ptr, ebuf - ptr);
9530 =for apidoc sv_setpviv_mg
9532 Like C<sv_setpviv>, but also handles 'set' magic.
9538 Perl_sv_setpviv_mg(pTHX_ SV *const sv, const IV iv)
9540 PERL_ARGS_ASSERT_SV_SETPVIV_MG;
9546 #if defined(PERL_IMPLICIT_CONTEXT)
9548 /* pTHX_ magic can't cope with varargs, so this is a no-context
9549 * version of the main function, (which may itself be aliased to us).
9550 * Don't access this version directly.
9554 Perl_sv_setpvf_nocontext(SV *const sv, const char *const pat, ...)
9559 PERL_ARGS_ASSERT_SV_SETPVF_NOCONTEXT;
9561 va_start(args, pat);
9562 sv_vsetpvf(sv, pat, &args);
9566 /* pTHX_ magic can't cope with varargs, so this is a no-context
9567 * version of the main function, (which may itself be aliased to us).
9568 * Don't access this version directly.
9572 Perl_sv_setpvf_mg_nocontext(SV *const sv, const char *const pat, ...)
9577 PERL_ARGS_ASSERT_SV_SETPVF_MG_NOCONTEXT;
9579 va_start(args, pat);
9580 sv_vsetpvf_mg(sv, pat, &args);
9586 =for apidoc sv_setpvf
9588 Works like C<sv_catpvf> but copies the text into the SV instead of
9589 appending it. Does not handle 'set' magic. See C<sv_setpvf_mg>.
9595 Perl_sv_setpvf(pTHX_ SV *const sv, const char *const pat, ...)
9599 PERL_ARGS_ASSERT_SV_SETPVF;
9601 va_start(args, pat);
9602 sv_vsetpvf(sv, pat, &args);
9607 =for apidoc sv_vsetpvf
9609 Works like C<sv_vcatpvf> but copies the text into the SV instead of
9610 appending it. Does not handle 'set' magic. See C<sv_vsetpvf_mg>.
9612 Usually used via its frontend C<sv_setpvf>.
9618 Perl_sv_vsetpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9620 PERL_ARGS_ASSERT_SV_VSETPVF;
9622 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9626 =for apidoc sv_setpvf_mg
9628 Like C<sv_setpvf>, but also handles 'set' magic.
9634 Perl_sv_setpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
9638 PERL_ARGS_ASSERT_SV_SETPVF_MG;
9640 va_start(args, pat);
9641 sv_vsetpvf_mg(sv, pat, &args);
9646 =for apidoc sv_vsetpvf_mg
9648 Like C<sv_vsetpvf>, but also handles 'set' magic.
9650 Usually used via its frontend C<sv_setpvf_mg>.
9656 Perl_sv_vsetpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9658 PERL_ARGS_ASSERT_SV_VSETPVF_MG;
9660 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9664 #if defined(PERL_IMPLICIT_CONTEXT)
9666 /* pTHX_ magic can't cope with varargs, so this is a no-context
9667 * version of the main function, (which may itself be aliased to us).
9668 * Don't access this version directly.
9672 Perl_sv_catpvf_nocontext(SV *const sv, const char *const pat, ...)
9677 PERL_ARGS_ASSERT_SV_CATPVF_NOCONTEXT;
9679 va_start(args, pat);
9680 sv_vcatpvf(sv, pat, &args);
9684 /* pTHX_ magic can't cope with varargs, so this is a no-context
9685 * version of the main function, (which may itself be aliased to us).
9686 * Don't access this version directly.
9690 Perl_sv_catpvf_mg_nocontext(SV *const sv, const char *const pat, ...)
9695 PERL_ARGS_ASSERT_SV_CATPVF_MG_NOCONTEXT;
9697 va_start(args, pat);
9698 sv_vcatpvf_mg(sv, pat, &args);
9704 =for apidoc sv_catpvf
9706 Processes its arguments like C<sprintf> and appends the formatted
9707 output to an SV. If the appended data contains "wide" characters
9708 (including, but not limited to, SVs with a UTF-8 PV formatted with %s,
9709 and characters >255 formatted with %c), the original SV might get
9710 upgraded to UTF-8. Handles 'get' magic, but not 'set' magic. See
9711 C<sv_catpvf_mg>. If the original SV was UTF-8, the pattern should be
9712 valid UTF-8; if the original SV was bytes, the pattern should be too.
9717 Perl_sv_catpvf(pTHX_ SV *const sv, const char *const pat, ...)
9721 PERL_ARGS_ASSERT_SV_CATPVF;
9723 va_start(args, pat);
9724 sv_vcatpvf(sv, pat, &args);
9729 =for apidoc sv_vcatpvf
9731 Processes its arguments like C<vsprintf> and appends the formatted output
9732 to an SV. Does not handle 'set' magic. See C<sv_vcatpvf_mg>.
9734 Usually used via its frontend C<sv_catpvf>.
9740 Perl_sv_vcatpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9742 PERL_ARGS_ASSERT_SV_VCATPVF;
9744 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9748 =for apidoc sv_catpvf_mg
9750 Like C<sv_catpvf>, but also handles 'set' magic.
9756 Perl_sv_catpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
9760 PERL_ARGS_ASSERT_SV_CATPVF_MG;
9762 va_start(args, pat);
9763 sv_vcatpvf_mg(sv, pat, &args);
9768 =for apidoc sv_vcatpvf_mg
9770 Like C<sv_vcatpvf>, but also handles 'set' magic.
9772 Usually used via its frontend C<sv_catpvf_mg>.
9778 Perl_sv_vcatpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9780 PERL_ARGS_ASSERT_SV_VCATPVF_MG;
9782 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9787 =for apidoc sv_vsetpvfn
9789 Works like C<sv_vcatpvfn> but copies the text into the SV instead of
9792 Usually used via one of its frontends C<sv_vsetpvf> and C<sv_vsetpvf_mg>.
9798 Perl_sv_vsetpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
9799 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
9801 PERL_ARGS_ASSERT_SV_VSETPVFN;
9804 sv_vcatpvfn(sv, pat, patlen, args, svargs, svmax, maybe_tainted);
9809 * Warn of missing argument to sprintf, and then return a defined value
9810 * to avoid inappropriate "use of uninit" warnings [perl #71000].
9812 #define WARN_MISSING WARN_UNINITIALIZED /* Not sure we want a new category */
9814 S_vcatpvfn_missing_argument(pTHX) {
9815 if (ckWARN(WARN_MISSING)) {
9816 Perl_warner(aTHX_ packWARN(WARN_MISSING), "Missing argument in %s",
9817 PL_op ? OP_DESC(PL_op) : "sv_vcatpvfn()");
9824 S_expect_number(pTHX_ char **const pattern)
9829 PERL_ARGS_ASSERT_EXPECT_NUMBER;
9831 switch (**pattern) {
9832 case '1': case '2': case '3':
9833 case '4': case '5': case '6':
9834 case '7': case '8': case '9':
9835 var = *(*pattern)++ - '0';
9836 while (isDIGIT(**pattern)) {
9837 const I32 tmp = var * 10 + (*(*pattern)++ - '0');
9839 Perl_croak(aTHX_ "Integer overflow in format string for %s", (PL_op ? OP_DESC(PL_op) : "sv_vcatpvfn"));
9847 S_F0convert(NV nv, char *const endbuf, STRLEN *const len)
9849 const int neg = nv < 0;
9852 PERL_ARGS_ASSERT_F0CONVERT;
9860 if (uv & 1 && uv == nv)
9861 uv--; /* Round to even */
9863 const unsigned dig = uv % 10;
9876 =for apidoc sv_vcatpvfn
9878 Processes its arguments like C<vsprintf> and appends the formatted output
9879 to an SV. Uses an array of SVs if the C style variable argument list is
9880 missing (NULL). When running with taint checks enabled, indicates via
9881 C<maybe_tainted> if results are untrustworthy (often due to the use of
9884 Usually used via one of its frontends C<sv_vcatpvf> and C<sv_vcatpvf_mg>.
9890 #define VECTORIZE_ARGS vecsv = va_arg(*args, SV*);\
9891 vecstr = (U8*)SvPV_const(vecsv,veclen);\
9892 vec_utf8 = DO_UTF8(vecsv);
9894 /* XXX maybe_tainted is never assigned to, so the doc above is lying. */
9897 Perl_sv_vcatpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
9898 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
9906 static const char nullstr[] = "(null)";
9908 bool has_utf8 = DO_UTF8(sv); /* has the result utf8? */
9909 const bool pat_utf8 = has_utf8; /* the pattern is in utf8? */
9911 /* Times 4: a decimal digit takes more than 3 binary digits.
9912 * NV_DIG: mantissa takes than many decimal digits.
9913 * Plus 32: Playing safe. */
9914 char ebuf[IV_DIG * 4 + NV_DIG + 32];
9915 /* large enough for "%#.#f" --chip */
9916 /* what about long double NVs? --jhi */
9918 PERL_ARGS_ASSERT_SV_VCATPVFN;
9919 PERL_UNUSED_ARG(maybe_tainted);
9921 /* no matter what, this is a string now */
9922 (void)SvPV_force(sv, origlen);
9924 /* special-case "", "%s", and "%-p" (SVf - see below) */
9927 if (patlen == 2 && pat[0] == '%' && pat[1] == 's') {
9929 const char * const s = va_arg(*args, char*);
9930 sv_catpv(sv, s ? s : nullstr);
9932 else if (svix < svmax) {
9933 sv_catsv(sv, *svargs);
9936 S_vcatpvfn_missing_argument(aTHX);
9939 if (args && patlen == 3 && pat[0] == '%' &&
9940 pat[1] == '-' && pat[2] == 'p') {
9941 argsv = MUTABLE_SV(va_arg(*args, void*));
9942 sv_catsv(sv, argsv);
9946 #ifndef USE_LONG_DOUBLE
9947 /* special-case "%.<number>[gf]" */
9948 if ( !args && patlen <= 5 && pat[0] == '%' && pat[1] == '.'
9949 && (pat[patlen-1] == 'g' || pat[patlen-1] == 'f') ) {
9950 unsigned digits = 0;
9954 while (*pp >= '0' && *pp <= '9')
9955 digits = 10 * digits + (*pp++ - '0');
9956 if (pp - pat == (int)patlen - 1 && svix < svmax) {
9957 const NV nv = SvNV(*svargs);
9959 /* Add check for digits != 0 because it seems that some
9960 gconverts are buggy in this case, and we don't yet have
9961 a Configure test for this. */
9962 if (digits && digits < sizeof(ebuf) - NV_DIG - 10) {
9963 /* 0, point, slack */
9964 Gconvert(nv, (int)digits, 0, ebuf);
9966 if (*ebuf) /* May return an empty string for digits==0 */
9969 } else if (!digits) {
9972 if ((p = F0convert(nv, ebuf + sizeof ebuf, &l))) {
9973 sv_catpvn(sv, p, l);
9979 #endif /* !USE_LONG_DOUBLE */
9981 if (!args && svix < svmax && DO_UTF8(*svargs))
9984 patend = (char*)pat + patlen;
9985 for (p = (char*)pat; p < patend; p = q) {
9988 bool vectorize = FALSE;
9989 bool vectorarg = FALSE;
9990 bool vec_utf8 = FALSE;
9996 bool has_precis = FALSE;
9998 const I32 osvix = svix;
9999 bool is_utf8 = FALSE; /* is this item utf8? */
10000 #ifdef HAS_LDBL_SPRINTF_BUG
10001 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
10002 with sfio - Allen <allens@cpan.org> */
10003 bool fix_ldbl_sprintf_bug = FALSE;
10007 U8 utf8buf[UTF8_MAXBYTES+1];
10008 STRLEN esignlen = 0;
10010 const char *eptr = NULL;
10011 const char *fmtstart;
10014 const U8 *vecstr = NULL;
10021 /* we need a long double target in case HAS_LONG_DOUBLE but
10022 not USE_LONG_DOUBLE
10024 #if defined(HAS_LONG_DOUBLE) && LONG_DOUBLESIZE > DOUBLESIZE
10032 const char *dotstr = ".";
10033 STRLEN dotstrlen = 1;
10034 I32 efix = 0; /* explicit format parameter index */
10035 I32 ewix = 0; /* explicit width index */
10036 I32 epix = 0; /* explicit precision index */
10037 I32 evix = 0; /* explicit vector index */
10038 bool asterisk = FALSE;
10040 /* echo everything up to the next format specification */
10041 for (q = p; q < patend && *q != '%'; ++q) ;
10043 if (has_utf8 && !pat_utf8)
10044 sv_catpvn_utf8_upgrade(sv, p, q - p, nsv);
10046 sv_catpvn(sv, p, q - p);
10055 We allow format specification elements in this order:
10056 \d+\$ explicit format parameter index
10058 v|\*(\d+\$)?v vector with optional (optionally specified) arg
10059 0 flag (as above): repeated to allow "v02"
10060 \d+|\*(\d+\$)? width using optional (optionally specified) arg
10061 \.(\d*|\*(\d+\$)?) precision using optional (optionally specified) arg
10063 [%bcdefginopsuxDFOUX] format (mandatory)
10068 As of perl5.9.3, printf format checking is on by default.
10069 Internally, perl uses %p formats to provide an escape to
10070 some extended formatting. This block deals with those
10071 extensions: if it does not match, (char*)q is reset and
10072 the normal format processing code is used.
10074 Currently defined extensions are:
10075 %p include pointer address (standard)
10076 %-p (SVf) include an SV (previously %_)
10077 %-<num>p include an SV with precision <num>
10078 %<num>p reserved for future extensions
10080 Robin Barker 2005-07-14
10082 %1p (VDf) removed. RMB 2007-10-19
10089 n = expect_number(&q);
10091 if (sv) { /* SVf */
10096 argsv = MUTABLE_SV(va_arg(*args, void*));
10097 eptr = SvPV_const(argsv, elen);
10098 if (DO_UTF8(argsv))
10103 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL),
10104 "internal %%<num>p might conflict with future printf extensions");
10110 if ( (width = expect_number(&q)) ) {
10125 if (plus == '+' && *q == ' ') /* '+' over ' ' */
10154 if ( (ewix = expect_number(&q)) )
10163 if ((vectorarg = asterisk)) {
10176 width = expect_number(&q);
10179 if (vectorize && vectorarg) {
10180 /* vectorizing, but not with the default "." */
10182 vecsv = va_arg(*args, SV*);
10184 vecsv = (evix > 0 && evix <= svmax)
10185 ? svargs[evix-1] : S_vcatpvfn_missing_argument(aTHX);
10187 vecsv = svix < svmax
10188 ? svargs[svix++] : S_vcatpvfn_missing_argument(aTHX);
10190 dotstr = SvPV_const(vecsv, dotstrlen);
10191 /* Keep the DO_UTF8 test *after* the SvPV call, else things go
10192 bad with tied or overloaded values that return UTF8. */
10193 if (DO_UTF8(vecsv))
10195 else if (has_utf8) {
10196 vecsv = sv_mortalcopy(vecsv);
10197 sv_utf8_upgrade(vecsv);
10198 dotstr = SvPV_const(vecsv, dotstrlen);
10205 i = va_arg(*args, int);
10207 i = (ewix ? ewix <= svmax : svix < svmax) ?
10208 SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
10210 width = (i < 0) ? -i : i;
10220 if ( ((epix = expect_number(&q))) && (*q++ != '$') )
10222 /* XXX: todo, support specified precision parameter */
10226 i = va_arg(*args, int);
10228 i = (ewix ? ewix <= svmax : svix < svmax)
10229 ? SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
10231 has_precis = !(i < 0);
10235 while (isDIGIT(*q))
10236 precis = precis * 10 + (*q++ - '0');
10245 else if (efix ? (efix > 0 && efix <= svmax) : svix < svmax) {
10246 vecsv = svargs[efix ? efix-1 : svix++];
10247 vecstr = (U8*)SvPV_const(vecsv,veclen);
10248 vec_utf8 = DO_UTF8(vecsv);
10250 /* if this is a version object, we need to convert
10251 * back into v-string notation and then let the
10252 * vectorize happen normally
10254 if (sv_derived_from(vecsv, "version")) {
10255 char *version = savesvpv(vecsv);
10256 if ( hv_exists(MUTABLE_HV(SvRV(vecsv)), "alpha", 5 ) ) {
10257 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
10258 "vector argument not supported with alpha versions");
10261 vecsv = sv_newmortal();
10262 scan_vstring(version, version + veclen, vecsv);
10263 vecstr = (U8*)SvPV_const(vecsv, veclen);
10264 vec_utf8 = DO_UTF8(vecsv);
10278 case 'I': /* Ix, I32x, and I64x */
10280 if (q[1] == '6' && q[2] == '4') {
10286 if (q[1] == '3' && q[2] == '2') {
10296 #if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
10308 #if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
10309 if (*q == 'l') { /* lld, llf */
10318 if (*++q == 'h') { /* hhd, hhu */
10347 if (!vectorize && !args) {
10349 const I32 i = efix-1;
10350 argsv = (i >= 0 && i < svmax)
10351 ? svargs[i] : S_vcatpvfn_missing_argument(aTHX);
10353 argsv = (svix >= 0 && svix < svmax)
10354 ? svargs[svix++] : S_vcatpvfn_missing_argument(aTHX);
10358 switch (c = *q++) {
10365 uv = (args) ? va_arg(*args, int) : SvIV(argsv);
10367 (!UNI_IS_INVARIANT(uv) && SvUTF8(sv)))
10369 eptr = (char*)utf8buf;
10370 elen = uvchr_to_utf8((U8*)eptr, uv) - utf8buf;
10384 eptr = va_arg(*args, char*);
10386 elen = strlen(eptr);
10388 eptr = (char *)nullstr;
10389 elen = sizeof nullstr - 1;
10393 eptr = SvPV_const(argsv, elen);
10394 if (DO_UTF8(argsv)) {
10395 STRLEN old_precis = precis;
10396 if (has_precis && precis < elen) {
10397 STRLEN ulen = sv_len_utf8(argsv);
10398 I32 p = precis > ulen ? ulen : precis;
10399 sv_pos_u2b(argsv, &p, 0); /* sticks at end */
10402 if (width) { /* fudge width (can't fudge elen) */
10403 if (has_precis && precis < elen)
10404 width += precis - old_precis;
10406 width += elen - sv_len_utf8(argsv);
10413 if (has_precis && precis < elen)
10420 if (alt || vectorize)
10422 uv = PTR2UV(args ? va_arg(*args, void*) : argsv);
10443 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
10452 esignbuf[esignlen++] = plus;
10456 case 'c': iv = (char)va_arg(*args, int); break;
10457 case 'h': iv = (short)va_arg(*args, int); break;
10458 case 'l': iv = va_arg(*args, long); break;
10459 case 'V': iv = va_arg(*args, IV); break;
10460 case 'z': iv = va_arg(*args, SSize_t); break;
10461 case 't': iv = va_arg(*args, ptrdiff_t); break;
10462 default: iv = va_arg(*args, int); break;
10464 case 'j': iv = va_arg(*args, intmax_t); break;
10468 iv = va_arg(*args, Quad_t); break;
10475 IV tiv = SvIV(argsv); /* work around GCC bug #13488 */
10477 case 'c': iv = (char)tiv; break;
10478 case 'h': iv = (short)tiv; break;
10479 case 'l': iv = (long)tiv; break;
10481 default: iv = tiv; break;
10484 iv = (Quad_t)tiv; break;
10490 if ( !vectorize ) /* we already set uv above */
10495 esignbuf[esignlen++] = plus;
10499 esignbuf[esignlen++] = '-';
10543 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
10554 case 'c': uv = (unsigned char)va_arg(*args, unsigned); break;
10555 case 'h': uv = (unsigned short)va_arg(*args, unsigned); break;
10556 case 'l': uv = va_arg(*args, unsigned long); break;
10557 case 'V': uv = va_arg(*args, UV); break;
10558 case 'z': uv = va_arg(*args, Size_t); break;
10559 case 't': uv = va_arg(*args, ptrdiff_t); break; /* will sign extend, but there is no uptrdiff_t, so oh well */
10561 case 'j': uv = va_arg(*args, uintmax_t); break;
10563 default: uv = va_arg(*args, unsigned); break;
10566 uv = va_arg(*args, Uquad_t); break;
10573 UV tuv = SvUV(argsv); /* work around GCC bug #13488 */
10575 case 'c': uv = (unsigned char)tuv; break;
10576 case 'h': uv = (unsigned short)tuv; break;
10577 case 'l': uv = (unsigned long)tuv; break;
10579 default: uv = tuv; break;
10582 uv = (Uquad_t)tuv; break;
10591 char *ptr = ebuf + sizeof ebuf;
10592 bool tempalt = uv ? alt : FALSE; /* Vectors can't change alt */
10598 p = (char *)((c == 'X') ? PL_hexdigit + 16 : PL_hexdigit);
10602 } while (uv >>= 4);
10604 esignbuf[esignlen++] = '0';
10605 esignbuf[esignlen++] = c; /* 'x' or 'X' */
10611 *--ptr = '0' + dig;
10612 } while (uv >>= 3);
10613 if (alt && *ptr != '0')
10619 *--ptr = '0' + dig;
10620 } while (uv >>= 1);
10622 esignbuf[esignlen++] = '0';
10623 esignbuf[esignlen++] = c;
10626 default: /* it had better be ten or less */
10629 *--ptr = '0' + dig;
10630 } while (uv /= base);
10633 elen = (ebuf + sizeof ebuf) - ptr;
10637 zeros = precis - elen;
10638 else if (precis == 0 && elen == 1 && *eptr == '0'
10639 && !(base == 8 && alt)) /* "%#.0o" prints "0" */
10642 /* a precision nullifies the 0 flag. */
10649 /* FLOATING POINT */
10652 c = 'f'; /* maybe %F isn't supported here */
10654 case 'e': case 'E':
10656 case 'g': case 'G':
10660 /* This is evil, but floating point is even more evil */
10662 /* for SV-style calling, we can only get NV
10663 for C-style calling, we assume %f is double;
10664 for simplicity we allow any of %Lf, %llf, %qf for long double
10668 #if defined(USE_LONG_DOUBLE)
10672 /* [perl #20339] - we should accept and ignore %lf rather than die */
10676 #if defined(USE_LONG_DOUBLE)
10677 intsize = args ? 0 : 'q';
10681 #if defined(HAS_LONG_DOUBLE)
10694 /* now we need (long double) if intsize == 'q', else (double) */
10696 #if LONG_DOUBLESIZE > DOUBLESIZE
10698 va_arg(*args, long double) :
10699 va_arg(*args, double)
10701 va_arg(*args, double)
10706 /* nv * 0 will be NaN for NaN, +Inf and -Inf, and 0 for anything
10707 else. frexp() has some unspecified behaviour for those three */
10708 if (c != 'e' && c != 'E' && (nv * 0) == 0) {
10710 /* FIXME: if HAS_LONG_DOUBLE but not USE_LONG_DOUBLE this
10711 will cast our (long double) to (double) */
10712 (void)Perl_frexp(nv, &i);
10713 if (i == PERL_INT_MIN)
10714 Perl_die(aTHX_ "panic: frexp");
10716 need = BIT_DIGITS(i);
10718 need += has_precis ? precis : 6; /* known default */
10723 #ifdef HAS_LDBL_SPRINTF_BUG
10724 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
10725 with sfio - Allen <allens@cpan.org> */
10728 # define MY_DBL_MAX DBL_MAX
10729 # else /* XXX guessing! HUGE_VAL may be defined as infinity, so not using */
10730 # if DOUBLESIZE >= 8
10731 # define MY_DBL_MAX 1.7976931348623157E+308L
10733 # define MY_DBL_MAX 3.40282347E+38L
10737 # ifdef HAS_LDBL_SPRINTF_BUG_LESS1 /* only between -1L & 1L - Allen */
10738 # define MY_DBL_MAX_BUG 1L
10740 # define MY_DBL_MAX_BUG MY_DBL_MAX
10744 # define MY_DBL_MIN DBL_MIN
10745 # else /* XXX guessing! -Allen */
10746 # if DOUBLESIZE >= 8
10747 # define MY_DBL_MIN 2.2250738585072014E-308L
10749 # define MY_DBL_MIN 1.17549435E-38L
10753 if ((intsize == 'q') && (c == 'f') &&
10754 ((nv < MY_DBL_MAX_BUG) && (nv > -MY_DBL_MAX_BUG)) &&
10755 (need < DBL_DIG)) {
10756 /* it's going to be short enough that
10757 * long double precision is not needed */
10759 if ((nv <= 0L) && (nv >= -0L))
10760 fix_ldbl_sprintf_bug = TRUE; /* 0 is 0 - easiest */
10762 /* would use Perl_fp_class as a double-check but not
10763 * functional on IRIX - see perl.h comments */
10765 if ((nv >= MY_DBL_MIN) || (nv <= -MY_DBL_MIN)) {
10766 /* It's within the range that a double can represent */
10767 #if defined(DBL_MAX) && !defined(DBL_MIN)
10768 if ((nv >= ((long double)1/DBL_MAX)) ||
10769 (nv <= (-(long double)1/DBL_MAX)))
10771 fix_ldbl_sprintf_bug = TRUE;
10774 if (fix_ldbl_sprintf_bug == TRUE) {
10784 # undef MY_DBL_MAX_BUG
10787 #endif /* HAS_LDBL_SPRINTF_BUG */
10789 need += 20; /* fudge factor */
10790 if (PL_efloatsize < need) {
10791 Safefree(PL_efloatbuf);
10792 PL_efloatsize = need + 20; /* more fudge */
10793 Newx(PL_efloatbuf, PL_efloatsize, char);
10794 PL_efloatbuf[0] = '\0';
10797 if ( !(width || left || plus || alt) && fill != '0'
10798 && has_precis && intsize != 'q' ) { /* Shortcuts */
10799 /* See earlier comment about buggy Gconvert when digits,
10801 if ( c == 'g' && precis) {
10802 Gconvert((NV)nv, (int)precis, 0, PL_efloatbuf);
10803 /* May return an empty string for digits==0 */
10804 if (*PL_efloatbuf) {
10805 elen = strlen(PL_efloatbuf);
10806 goto float_converted;
10808 } else if ( c == 'f' && !precis) {
10809 if ((eptr = F0convert(nv, ebuf + sizeof ebuf, &elen)))
10814 char *ptr = ebuf + sizeof ebuf;
10817 /* FIXME: what to do if HAS_LONG_DOUBLE but not PERL_PRIfldbl? */
10818 #if defined(HAS_LONG_DOUBLE) && defined(PERL_PRIfldbl)
10819 if (intsize == 'q') {
10820 /* Copy the one or more characters in a long double
10821 * format before the 'base' ([efgEFG]) character to
10822 * the format string. */
10823 static char const prifldbl[] = PERL_PRIfldbl;
10824 char const *p = prifldbl + sizeof(prifldbl) - 3;
10825 while (p >= prifldbl) { *--ptr = *p--; }
10830 do { *--ptr = '0' + (base % 10); } while (base /= 10);
10835 do { *--ptr = '0' + (base % 10); } while (base /= 10);
10847 /* No taint. Otherwise we are in the strange situation
10848 * where printf() taints but print($float) doesn't.
10850 #if defined(HAS_LONG_DOUBLE)
10851 elen = ((intsize == 'q')
10852 ? my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, nv)
10853 : my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, (double)nv));
10855 elen = my_sprintf(PL_efloatbuf, ptr, nv);
10859 eptr = PL_efloatbuf;
10867 i = SvCUR(sv) - origlen;
10870 case 'c': *(va_arg(*args, char*)) = i; break;
10871 case 'h': *(va_arg(*args, short*)) = i; break;
10872 default: *(va_arg(*args, int*)) = i; break;
10873 case 'l': *(va_arg(*args, long*)) = i; break;
10874 case 'V': *(va_arg(*args, IV*)) = i; break;
10875 case 'z': *(va_arg(*args, SSize_t*)) = i; break;
10876 case 't': *(va_arg(*args, ptrdiff_t*)) = i; break;
10878 case 'j': *(va_arg(*args, intmax_t*)) = i; break;
10882 *(va_arg(*args, Quad_t*)) = i; break;
10889 sv_setuv_mg(argsv, (UV)i);
10890 continue; /* not "break" */
10897 && (PL_op->op_type == OP_PRTF || PL_op->op_type == OP_SPRINTF)
10898 && ckWARN(WARN_PRINTF))
10900 SV * const msg = sv_newmortal();
10901 Perl_sv_setpvf(aTHX_ msg, "Invalid conversion in %sprintf: ",
10902 (PL_op->op_type == OP_PRTF) ? "" : "s");
10903 if (fmtstart < patend) {
10904 const char * const fmtend = q < patend ? q : patend;
10906 sv_catpvs(msg, "\"%");
10907 for (f = fmtstart; f < fmtend; f++) {
10909 sv_catpvn(msg, f, 1);
10911 Perl_sv_catpvf(aTHX_ msg,
10912 "\\%03"UVof, (UV)*f & 0xFF);
10915 sv_catpvs(msg, "\"");
10917 sv_catpvs(msg, "end of string");
10919 Perl_warner(aTHX_ packWARN(WARN_PRINTF), "%"SVf, SVfARG(msg)); /* yes, this is reentrant */
10922 /* output mangled stuff ... */
10928 /* ... right here, because formatting flags should not apply */
10929 SvGROW(sv, SvCUR(sv) + elen + 1);
10931 Copy(eptr, p, elen, char);
10934 SvCUR_set(sv, p - SvPVX_const(sv));
10936 continue; /* not "break" */
10939 if (is_utf8 != has_utf8) {
10942 sv_utf8_upgrade(sv);
10945 const STRLEN old_elen = elen;
10946 SV * const nsv = newSVpvn_flags(eptr, elen, SVs_TEMP);
10947 sv_utf8_upgrade(nsv);
10948 eptr = SvPVX_const(nsv);
10951 if (width) { /* fudge width (can't fudge elen) */
10952 width += elen - old_elen;
10958 have = esignlen + zeros + elen;
10960 Perl_croak_nocontext("%s", PL_memory_wrap);
10962 need = (have > width ? have : width);
10965 if (need >= (((STRLEN)~0) - SvCUR(sv) - dotstrlen - 1))
10966 Perl_croak_nocontext("%s", PL_memory_wrap);
10967 SvGROW(sv, SvCUR(sv) + need + dotstrlen + 1);
10969 if (esignlen && fill == '0') {
10971 for (i = 0; i < (int)esignlen; i++)
10972 *p++ = esignbuf[i];
10974 if (gap && !left) {
10975 memset(p, fill, gap);
10978 if (esignlen && fill != '0') {
10980 for (i = 0; i < (int)esignlen; i++)
10981 *p++ = esignbuf[i];
10985 for (i = zeros; i; i--)
10989 Copy(eptr, p, elen, char);
10993 memset(p, ' ', gap);
10998 Copy(dotstr, p, dotstrlen, char);
11002 vectorize = FALSE; /* done iterating over vecstr */
11009 SvCUR_set(sv, p - SvPVX_const(sv));
11018 /* =========================================================================
11020 =head1 Cloning an interpreter
11022 All the macros and functions in this section are for the private use of
11023 the main function, perl_clone().
11025 The foo_dup() functions make an exact copy of an existing foo thingy.
11026 During the course of a cloning, a hash table is used to map old addresses
11027 to new addresses. The table is created and manipulated with the
11028 ptr_table_* functions.
11032 * =========================================================================*/
11035 #if defined(USE_ITHREADS)
11037 /* XXX Remove this so it doesn't have to go thru the macro and return for nothing */
11038 #ifndef GpREFCNT_inc
11039 # define GpREFCNT_inc(gp) ((gp) ? (++(gp)->gp_refcnt, (gp)) : (GP*)NULL)
11043 /* Certain cases in Perl_ss_dup have been merged, by relying on the fact
11044 that currently av_dup, gv_dup and hv_dup are the same as sv_dup.
11045 If this changes, please unmerge ss_dup.
11046 Likewise, sv_dup_inc_multiple() relies on this fact. */
11047 #define sv_dup_inc_NN(s,t) SvREFCNT_inc_NN(sv_dup_inc(s,t))
11048 #define av_dup(s,t) MUTABLE_AV(sv_dup((const SV *)s,t))
11049 #define av_dup_inc(s,t) MUTABLE_AV(sv_dup_inc((const SV *)s,t))
11050 #define hv_dup(s,t) MUTABLE_HV(sv_dup((const SV *)s,t))
11051 #define hv_dup_inc(s,t) MUTABLE_HV(sv_dup_inc((const SV *)s,t))
11052 #define cv_dup(s,t) MUTABLE_CV(sv_dup((const SV *)s,t))
11053 #define cv_dup_inc(s,t) MUTABLE_CV(sv_dup_inc((const SV *)s,t))
11054 #define io_dup(s,t) MUTABLE_IO(sv_dup((const SV *)s,t))
11055 #define io_dup_inc(s,t) MUTABLE_IO(sv_dup_inc((const SV *)s,t))
11056 #define gv_dup(s,t) MUTABLE_GV(sv_dup((const SV *)s,t))
11057 #define gv_dup_inc(s,t) MUTABLE_GV(sv_dup_inc((const SV *)s,t))
11058 #define SAVEPV(p) ((p) ? savepv(p) : NULL)
11059 #define SAVEPVN(p,n) ((p) ? savepvn(p,n) : NULL)
11061 /* clone a parser */
11064 Perl_parser_dup(pTHX_ const yy_parser *const proto, CLONE_PARAMS *const param)
11068 PERL_ARGS_ASSERT_PARSER_DUP;
11073 /* look for it in the table first */
11074 parser = (yy_parser *)ptr_table_fetch(PL_ptr_table, proto);
11078 /* create anew and remember what it is */
11079 Newxz(parser, 1, yy_parser);
11080 ptr_table_store(PL_ptr_table, proto, parser);
11082 /* XXX these not yet duped */
11083 parser->old_parser = NULL;
11084 parser->stack = NULL;
11086 parser->stack_size = 0;
11087 /* XXX parser->stack->state = 0; */
11089 /* XXX eventually, just Copy() most of the parser struct ? */
11091 parser->lex_brackets = proto->lex_brackets;
11092 parser->lex_casemods = proto->lex_casemods;
11093 parser->lex_brackstack = savepvn(proto->lex_brackstack,
11094 (proto->lex_brackets < 120 ? 120 : proto->lex_brackets));
11095 parser->lex_casestack = savepvn(proto->lex_casestack,
11096 (proto->lex_casemods < 12 ? 12 : proto->lex_casemods));
11097 parser->lex_defer = proto->lex_defer;
11098 parser->lex_dojoin = proto->lex_dojoin;
11099 parser->lex_expect = proto->lex_expect;
11100 parser->lex_formbrack = proto->lex_formbrack;
11101 parser->lex_inpat = proto->lex_inpat;
11102 parser->lex_inwhat = proto->lex_inwhat;
11103 parser->lex_op = proto->lex_op;
11104 parser->lex_repl = sv_dup_inc(proto->lex_repl, param);
11105 parser->lex_starts = proto->lex_starts;
11106 parser->lex_stuff = sv_dup_inc(proto->lex_stuff, param);
11107 parser->multi_close = proto->multi_close;
11108 parser->multi_open = proto->multi_open;
11109 parser->multi_start = proto->multi_start;
11110 parser->multi_end = proto->multi_end;
11111 parser->pending_ident = proto->pending_ident;
11112 parser->preambled = proto->preambled;
11113 parser->sublex_info = proto->sublex_info; /* XXX not quite right */
11114 parser->linestr = sv_dup_inc(proto->linestr, param);
11115 parser->expect = proto->expect;
11116 parser->copline = proto->copline;
11117 parser->last_lop_op = proto->last_lop_op;
11118 parser->lex_state = proto->lex_state;
11119 parser->rsfp = fp_dup(proto->rsfp, '<', param);
11120 /* rsfp_filters entries have fake IoDIRP() */
11121 parser->rsfp_filters= av_dup_inc(proto->rsfp_filters, param);
11122 parser->in_my = proto->in_my;
11123 parser->in_my_stash = hv_dup(proto->in_my_stash, param);
11124 parser->error_count = proto->error_count;
11127 parser->linestr = sv_dup_inc(proto->linestr, param);
11130 char * const ols = SvPVX(proto->linestr);
11131 char * const ls = SvPVX(parser->linestr);
11133 parser->bufptr = ls + (proto->bufptr >= ols ?
11134 proto->bufptr - ols : 0);
11135 parser->oldbufptr = ls + (proto->oldbufptr >= ols ?
11136 proto->oldbufptr - ols : 0);
11137 parser->oldoldbufptr= ls + (proto->oldoldbufptr >= ols ?
11138 proto->oldoldbufptr - ols : 0);
11139 parser->linestart = ls + (proto->linestart >= ols ?
11140 proto->linestart - ols : 0);
11141 parser->last_uni = ls + (proto->last_uni >= ols ?
11142 proto->last_uni - ols : 0);
11143 parser->last_lop = ls + (proto->last_lop >= ols ?
11144 proto->last_lop - ols : 0);
11146 parser->bufend = ls + SvCUR(parser->linestr);
11149 Copy(proto->tokenbuf, parser->tokenbuf, 256, char);
11153 parser->endwhite = proto->endwhite;
11154 parser->faketokens = proto->faketokens;
11155 parser->lasttoke = proto->lasttoke;
11156 parser->nextwhite = proto->nextwhite;
11157 parser->realtokenstart = proto->realtokenstart;
11158 parser->skipwhite = proto->skipwhite;
11159 parser->thisclose = proto->thisclose;
11160 parser->thismad = proto->thismad;
11161 parser->thisopen = proto->thisopen;
11162 parser->thisstuff = proto->thisstuff;
11163 parser->thistoken = proto->thistoken;
11164 parser->thiswhite = proto->thiswhite;
11166 Copy(proto->nexttoke, parser->nexttoke, 5, NEXTTOKE);
11167 parser->curforce = proto->curforce;
11169 Copy(proto->nextval, parser->nextval, 5, YYSTYPE);
11170 Copy(proto->nexttype, parser->nexttype, 5, I32);
11171 parser->nexttoke = proto->nexttoke;
11174 /* XXX should clone saved_curcop here, but we aren't passed
11175 * proto_perl; so do it in perl_clone_using instead */
11181 /* duplicate a file handle */
11184 Perl_fp_dup(pTHX_ PerlIO *const fp, const char type, CLONE_PARAMS *const param)
11188 PERL_ARGS_ASSERT_FP_DUP;
11189 PERL_UNUSED_ARG(type);
11192 return (PerlIO*)NULL;
11194 /* look for it in the table first */
11195 ret = (PerlIO*)ptr_table_fetch(PL_ptr_table, fp);
11199 /* create anew and remember what it is */
11200 ret = PerlIO_fdupopen(aTHX_ fp, param, PERLIO_DUP_CLONE);
11201 ptr_table_store(PL_ptr_table, fp, ret);
11205 /* duplicate a directory handle */
11208 Perl_dirp_dup(pTHX_ DIR *const dp, CLONE_PARAMS *const param)
11214 register const Direntry_t *dirent;
11215 char smallbuf[256];
11221 PERL_UNUSED_CONTEXT;
11222 PERL_ARGS_ASSERT_DIRP_DUP;
11227 /* look for it in the table first */
11228 ret = (DIR*)ptr_table_fetch(PL_ptr_table, dp);
11234 PERL_UNUSED_ARG(param);
11238 /* open the current directory (so we can switch back) */
11239 if (!(pwd = PerlDir_open("."))) return (DIR *)NULL;
11241 /* chdir to our dir handle and open the present working directory */
11242 if (fchdir(my_dirfd(dp)) < 0 || !(ret = PerlDir_open("."))) {
11243 PerlDir_close(pwd);
11244 return (DIR *)NULL;
11246 /* Now we should have two dir handles pointing to the same dir. */
11248 /* Be nice to the calling code and chdir back to where we were. */
11249 fchdir(my_dirfd(pwd)); /* If this fails, then what? */
11251 /* We have no need of the pwd handle any more. */
11252 PerlDir_close(pwd);
11255 # define d_namlen(d) (d)->d_namlen
11257 # define d_namlen(d) strlen((d)->d_name)
11259 /* Iterate once through dp, to get the file name at the current posi-
11260 tion. Then step back. */
11261 pos = PerlDir_tell(dp);
11262 if ((dirent = PerlDir_read(dp))) {
11263 len = d_namlen(dirent);
11264 if (len <= sizeof smallbuf) name = smallbuf;
11265 else Newx(name, len, char);
11266 Move(dirent->d_name, name, len, char);
11268 PerlDir_seek(dp, pos);
11270 /* Iterate through the new dir handle, till we find a file with the
11272 if (!dirent) /* just before the end */
11274 pos = PerlDir_tell(ret);
11275 if (PerlDir_read(ret)) continue; /* not there yet */
11276 PerlDir_seek(ret, pos); /* step back */
11280 const long pos0 = PerlDir_tell(ret);
11282 pos = PerlDir_tell(ret);
11283 if ((dirent = PerlDir_read(ret))) {
11284 if (len == d_namlen(dirent)
11285 && memEQ(name, dirent->d_name, len)) {
11287 PerlDir_seek(ret, pos); /* step back */
11290 /* else we are not there yet; keep iterating */
11292 else { /* This is not meant to happen. The best we can do is
11293 reset the iterator to the beginning. */
11294 PerlDir_seek(ret, pos0);
11301 if (name && name != smallbuf)
11306 ret = win32_dirp_dup(dp, param);
11309 /* pop it in the pointer table */
11311 ptr_table_store(PL_ptr_table, dp, ret);
11316 /* duplicate a typeglob */
11319 Perl_gp_dup(pTHX_ GP *const gp, CLONE_PARAMS *const param)
11323 PERL_ARGS_ASSERT_GP_DUP;
11327 /* look for it in the table first */
11328 ret = (GP*)ptr_table_fetch(PL_ptr_table, gp);
11332 /* create anew and remember what it is */
11334 ptr_table_store(PL_ptr_table, gp, ret);
11337 /* ret->gp_refcnt must be 0 before any other dups are called. We're relying
11338 on Newxz() to do this for us. */
11339 ret->gp_sv = sv_dup_inc(gp->gp_sv, param);
11340 ret->gp_io = io_dup_inc(gp->gp_io, param);
11341 ret->gp_form = cv_dup_inc(gp->gp_form, param);
11342 ret->gp_av = av_dup_inc(gp->gp_av, param);
11343 ret->gp_hv = hv_dup_inc(gp->gp_hv, param);
11344 ret->gp_egv = gv_dup(gp->gp_egv, param);/* GvEGV is not refcounted */
11345 ret->gp_cv = cv_dup_inc(gp->gp_cv, param);
11346 ret->gp_cvgen = gp->gp_cvgen;
11347 ret->gp_line = gp->gp_line;
11348 ret->gp_file_hek = hek_dup(gp->gp_file_hek, param);
11352 /* duplicate a chain of magic */
11355 Perl_mg_dup(pTHX_ MAGIC *mg, CLONE_PARAMS *const param)
11357 MAGIC *mgret = NULL;
11358 MAGIC **mgprev_p = &mgret;
11360 PERL_ARGS_ASSERT_MG_DUP;
11362 for (; mg; mg = mg->mg_moremagic) {
11365 if ((param->flags & CLONEf_JOIN_IN)
11366 && mg->mg_type == PERL_MAGIC_backref)
11367 /* when joining, we let the individual SVs add themselves to
11368 * backref as needed. */
11371 Newx(nmg, 1, MAGIC);
11373 mgprev_p = &(nmg->mg_moremagic);
11375 /* There was a comment "XXX copy dynamic vtable?" but as we don't have
11376 dynamic vtables, I'm not sure why Sarathy wrote it. The comment dates
11377 from the original commit adding Perl_mg_dup() - revision 4538.
11378 Similarly there is the annotation "XXX random ptr?" next to the
11379 assignment to nmg->mg_ptr. */
11382 /* FIXME for plugins
11383 if (nmg->mg_type == PERL_MAGIC_qr) {
11384 nmg->mg_obj = MUTABLE_SV(CALLREGDUPE((REGEXP*)nmg->mg_obj, param));
11388 nmg->mg_obj = (nmg->mg_flags & MGf_REFCOUNTED)
11389 ? nmg->mg_type == PERL_MAGIC_backref
11390 /* The backref AV has its reference
11391 * count deliberately bumped by 1 */
11392 ? SvREFCNT_inc(av_dup_inc((const AV *)
11393 nmg->mg_obj, param))
11394 : sv_dup_inc(nmg->mg_obj, param)
11395 : sv_dup(nmg->mg_obj, param);
11397 if (nmg->mg_ptr && nmg->mg_type != PERL_MAGIC_regex_global) {
11398 if (nmg->mg_len > 0) {
11399 nmg->mg_ptr = SAVEPVN(nmg->mg_ptr, nmg->mg_len);
11400 if (nmg->mg_type == PERL_MAGIC_overload_table &&
11401 AMT_AMAGIC((AMT*)nmg->mg_ptr))
11403 AMT * const namtp = (AMT*)nmg->mg_ptr;
11404 sv_dup_inc_multiple((SV**)(namtp->table),
11405 (SV**)(namtp->table), NofAMmeth, param);
11408 else if (nmg->mg_len == HEf_SVKEY)
11409 nmg->mg_ptr = (char*)sv_dup_inc((const SV *)nmg->mg_ptr, param);
11411 if ((nmg->mg_flags & MGf_DUP) && nmg->mg_virtual && nmg->mg_virtual->svt_dup) {
11412 nmg->mg_virtual->svt_dup(aTHX_ nmg, param);
11418 #endif /* USE_ITHREADS */
11420 struct ptr_tbl_arena {
11421 struct ptr_tbl_arena *next;
11422 struct ptr_tbl_ent array[1023/3]; /* as ptr_tbl_ent has 3 pointers. */
11425 /* create a new pointer-mapping table */
11428 Perl_ptr_table_new(pTHX)
11431 PERL_UNUSED_CONTEXT;
11433 Newx(tbl, 1, PTR_TBL_t);
11434 tbl->tbl_max = 511;
11435 tbl->tbl_items = 0;
11436 tbl->tbl_arena = NULL;
11437 tbl->tbl_arena_next = NULL;
11438 tbl->tbl_arena_end = NULL;
11439 Newxz(tbl->tbl_ary, tbl->tbl_max + 1, PTR_TBL_ENT_t*);
11443 #define PTR_TABLE_HASH(ptr) \
11444 ((PTR2UV(ptr) >> 3) ^ (PTR2UV(ptr) >> (3 + 7)) ^ (PTR2UV(ptr) >> (3 + 17)))
11446 /* map an existing pointer using a table */
11448 STATIC PTR_TBL_ENT_t *
11449 S_ptr_table_find(PTR_TBL_t *const tbl, const void *const sv)
11451 PTR_TBL_ENT_t *tblent;
11452 const UV hash = PTR_TABLE_HASH(sv);
11454 PERL_ARGS_ASSERT_PTR_TABLE_FIND;
11456 tblent = tbl->tbl_ary[hash & tbl->tbl_max];
11457 for (; tblent; tblent = tblent->next) {
11458 if (tblent->oldval == sv)
11465 Perl_ptr_table_fetch(pTHX_ PTR_TBL_t *const tbl, const void *const sv)
11467 PTR_TBL_ENT_t const *const tblent = ptr_table_find(tbl, sv);
11469 PERL_ARGS_ASSERT_PTR_TABLE_FETCH;
11470 PERL_UNUSED_CONTEXT;
11472 return tblent ? tblent->newval : NULL;
11475 /* add a new entry to a pointer-mapping table */
11478 Perl_ptr_table_store(pTHX_ PTR_TBL_t *const tbl, const void *const oldsv, void *const newsv)
11480 PTR_TBL_ENT_t *tblent = ptr_table_find(tbl, oldsv);
11482 PERL_ARGS_ASSERT_PTR_TABLE_STORE;
11483 PERL_UNUSED_CONTEXT;
11486 tblent->newval = newsv;
11488 const UV entry = PTR_TABLE_HASH(oldsv) & tbl->tbl_max;
11490 if (tbl->tbl_arena_next == tbl->tbl_arena_end) {
11491 struct ptr_tbl_arena *new_arena;
11493 Newx(new_arena, 1, struct ptr_tbl_arena);
11494 new_arena->next = tbl->tbl_arena;
11495 tbl->tbl_arena = new_arena;
11496 tbl->tbl_arena_next = new_arena->array;
11497 tbl->tbl_arena_end = new_arena->array
11498 + sizeof(new_arena->array) / sizeof(new_arena->array[0]);
11501 tblent = tbl->tbl_arena_next++;
11503 tblent->oldval = oldsv;
11504 tblent->newval = newsv;
11505 tblent->next = tbl->tbl_ary[entry];
11506 tbl->tbl_ary[entry] = tblent;
11508 if (tblent->next && tbl->tbl_items > tbl->tbl_max)
11509 ptr_table_split(tbl);
11513 /* double the hash bucket size of an existing ptr table */
11516 Perl_ptr_table_split(pTHX_ PTR_TBL_t *const tbl)
11518 PTR_TBL_ENT_t **ary = tbl->tbl_ary;
11519 const UV oldsize = tbl->tbl_max + 1;
11520 UV newsize = oldsize * 2;
11523 PERL_ARGS_ASSERT_PTR_TABLE_SPLIT;
11524 PERL_UNUSED_CONTEXT;
11526 Renew(ary, newsize, PTR_TBL_ENT_t*);
11527 Zero(&ary[oldsize], newsize-oldsize, PTR_TBL_ENT_t*);
11528 tbl->tbl_max = --newsize;
11529 tbl->tbl_ary = ary;
11530 for (i=0; i < oldsize; i++, ary++) {
11531 PTR_TBL_ENT_t **entp = ary;
11532 PTR_TBL_ENT_t *ent = *ary;
11533 PTR_TBL_ENT_t **curentp;
11536 curentp = ary + oldsize;
11538 if ((newsize & PTR_TABLE_HASH(ent->oldval)) != i) {
11540 ent->next = *curentp;
11550 /* remove all the entries from a ptr table */
11551 /* Deprecated - will be removed post 5.14 */
11554 Perl_ptr_table_clear(pTHX_ PTR_TBL_t *const tbl)
11556 if (tbl && tbl->tbl_items) {
11557 struct ptr_tbl_arena *arena = tbl->tbl_arena;
11559 Zero(tbl->tbl_ary, tbl->tbl_max + 1, struct ptr_tbl_ent **);
11562 struct ptr_tbl_arena *next = arena->next;
11568 tbl->tbl_items = 0;
11569 tbl->tbl_arena = NULL;
11570 tbl->tbl_arena_next = NULL;
11571 tbl->tbl_arena_end = NULL;
11575 /* clear and free a ptr table */
11578 Perl_ptr_table_free(pTHX_ PTR_TBL_t *const tbl)
11580 struct ptr_tbl_arena *arena;
11586 arena = tbl->tbl_arena;
11589 struct ptr_tbl_arena *next = arena->next;
11595 Safefree(tbl->tbl_ary);
11599 #if defined(USE_ITHREADS)
11602 Perl_rvpv_dup(pTHX_ SV *const dstr, const SV *const sstr, CLONE_PARAMS *const param)
11604 PERL_ARGS_ASSERT_RVPV_DUP;
11607 if (SvWEAKREF(sstr)) {
11608 SvRV_set(dstr, sv_dup(SvRV_const(sstr), param));
11609 if (param->flags & CLONEf_JOIN_IN) {
11610 /* if joining, we add any back references individually rather
11611 * than copying the whole backref array */
11612 Perl_sv_add_backref(aTHX_ SvRV(dstr), dstr);
11616 SvRV_set(dstr, sv_dup_inc(SvRV_const(sstr), param));
11618 else if (SvPVX_const(sstr)) {
11619 /* Has something there */
11621 /* Normal PV - clone whole allocated space */
11622 SvPV_set(dstr, SAVEPVN(SvPVX_const(sstr), SvLEN(sstr)-1));
11623 if (SvREADONLY(sstr) && SvFAKE(sstr)) {
11624 /* Not that normal - actually sstr is copy on write.
11625 But we are a true, independent SV, so: */
11626 SvREADONLY_off(dstr);
11631 /* Special case - not normally malloced for some reason */
11632 if (isGV_with_GP(sstr)) {
11633 /* Don't need to do anything here. */
11635 else if ((SvREADONLY(sstr) && SvFAKE(sstr))) {
11636 /* A "shared" PV - clone it as "shared" PV */
11638 HEK_KEY(hek_dup(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)),
11642 /* Some other special case - random pointer */
11643 SvPV_set(dstr, (char *) SvPVX_const(sstr));
11648 /* Copy the NULL */
11649 SvPV_set(dstr, NULL);
11653 /* duplicate a list of SVs. source and dest may point to the same memory. */
11655 S_sv_dup_inc_multiple(pTHX_ SV *const *source, SV **dest,
11656 SSize_t items, CLONE_PARAMS *const param)
11658 PERL_ARGS_ASSERT_SV_DUP_INC_MULTIPLE;
11660 while (items-- > 0) {
11661 *dest++ = sv_dup_inc(*source++, param);
11667 /* duplicate an SV of any type (including AV, HV etc) */
11670 S_sv_dup_common(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
11675 PERL_ARGS_ASSERT_SV_DUP_COMMON;
11677 if (SvTYPE(sstr) == (svtype)SVTYPEMASK) {
11678 #ifdef DEBUG_LEAKING_SCALARS_ABORT
11683 /* look for it in the table first */
11684 dstr = MUTABLE_SV(ptr_table_fetch(PL_ptr_table, sstr));
11688 if(param->flags & CLONEf_JOIN_IN) {
11689 /** We are joining here so we don't want do clone
11690 something that is bad **/
11691 if (SvTYPE(sstr) == SVt_PVHV) {
11692 const HEK * const hvname = HvNAME_HEK(sstr);
11694 /** don't clone stashes if they already exist **/
11695 dstr = MUTABLE_SV(gv_stashpvn(HEK_KEY(hvname), HEK_LEN(hvname), 0));
11696 ptr_table_store(PL_ptr_table, sstr, dstr);
11702 /* create anew and remember what it is */
11705 #ifdef DEBUG_LEAKING_SCALARS
11706 dstr->sv_debug_optype = sstr->sv_debug_optype;
11707 dstr->sv_debug_line = sstr->sv_debug_line;
11708 dstr->sv_debug_inpad = sstr->sv_debug_inpad;
11709 dstr->sv_debug_parent = (SV*)sstr;
11710 FREE_SV_DEBUG_FILE(dstr);
11711 dstr->sv_debug_file = savepv(sstr->sv_debug_file);
11714 ptr_table_store(PL_ptr_table, sstr, dstr);
11717 SvFLAGS(dstr) = SvFLAGS(sstr);
11718 SvFLAGS(dstr) &= ~SVf_OOK; /* don't propagate OOK hack */
11719 SvREFCNT(dstr) = 0; /* must be before any other dups! */
11722 if (SvANY(sstr) && PL_watch_pvx && SvPVX_const(sstr) == PL_watch_pvx)
11723 PerlIO_printf(Perl_debug_log, "watch at %p hit, found string \"%s\"\n",
11724 (void*)PL_watch_pvx, SvPVX_const(sstr));
11727 /* don't clone objects whose class has asked us not to */
11728 if (SvOBJECT(sstr) && ! (SvFLAGS(SvSTASH(sstr)) & SVphv_CLONEABLE)) {
11733 switch (SvTYPE(sstr)) {
11735 SvANY(dstr) = NULL;
11738 SvANY(dstr) = (XPVIV*)((char*)&(dstr->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
11740 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
11742 SvIV_set(dstr, SvIVX(sstr));
11746 SvANY(dstr) = new_XNV();
11747 SvNV_set(dstr, SvNVX(sstr));
11749 /* case SVt_BIND: */
11752 /* These are all the types that need complex bodies allocating. */
11754 const svtype sv_type = SvTYPE(sstr);
11755 const struct body_details *const sv_type_details
11756 = bodies_by_type + sv_type;
11760 Perl_croak(aTHX_ "Bizarre SvTYPE [%" IVdf "]", (IV)SvTYPE(sstr));
11775 assert(sv_type_details->body_size);
11776 if (sv_type_details->arena) {
11777 new_body_inline(new_body, sv_type);
11779 = (void*)((char*)new_body - sv_type_details->offset);
11781 new_body = new_NOARENA(sv_type_details);
11785 SvANY(dstr) = new_body;
11788 Copy(((char*)SvANY(sstr)) + sv_type_details->offset,
11789 ((char*)SvANY(dstr)) + sv_type_details->offset,
11790 sv_type_details->copy, char);
11792 Copy(((char*)SvANY(sstr)),
11793 ((char*)SvANY(dstr)),
11794 sv_type_details->body_size + sv_type_details->offset, char);
11797 if (sv_type != SVt_PVAV && sv_type != SVt_PVHV
11798 && !isGV_with_GP(dstr)
11799 && !(sv_type == SVt_PVIO && !(IoFLAGS(dstr) & IOf_FAKE_DIRP)))
11800 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
11802 /* The Copy above means that all the source (unduplicated) pointers
11803 are now in the destination. We can check the flags and the
11804 pointers in either, but it's possible that there's less cache
11805 missing by always going for the destination.
11806 FIXME - instrument and check that assumption */
11807 if (sv_type >= SVt_PVMG) {
11808 if ((sv_type == SVt_PVMG) && SvPAD_OUR(dstr)) {
11809 SvOURSTASH_set(dstr, hv_dup_inc(SvOURSTASH(dstr), param));
11810 } else if (SvMAGIC(dstr))
11811 SvMAGIC_set(dstr, mg_dup(SvMAGIC(dstr), param));
11813 SvSTASH_set(dstr, hv_dup_inc(SvSTASH(dstr), param));
11816 /* The cast silences a GCC warning about unhandled types. */
11817 switch ((int)sv_type) {
11827 /* FIXME for plugins */
11828 re_dup_guts((REGEXP*) sstr, (REGEXP*) dstr, param);
11831 /* XXX LvTARGOFF sometimes holds PMOP* when DEBUGGING */
11832 if (LvTYPE(dstr) == 't') /* for tie: unrefcnted fake (SV**) */
11833 LvTARG(dstr) = dstr;
11834 else if (LvTYPE(dstr) == 'T') /* for tie: fake HE */
11835 LvTARG(dstr) = MUTABLE_SV(he_dup((HE*)LvTARG(dstr), 0, param));
11837 LvTARG(dstr) = sv_dup_inc(LvTARG(dstr), param);
11839 /* non-GP case already handled above */
11840 if(isGV_with_GP(sstr)) {
11841 GvNAME_HEK(dstr) = hek_dup(GvNAME_HEK(dstr), param);
11842 /* Don't call sv_add_backref here as it's going to be
11843 created as part of the magic cloning of the symbol
11844 table--unless this is during a join and the stash
11845 is not actually being cloned. */
11846 /* Danger Will Robinson - GvGP(dstr) isn't initialised
11847 at the point of this comment. */
11848 GvSTASH(dstr) = hv_dup(GvSTASH(dstr), param);
11849 if (param->flags & CLONEf_JOIN_IN)
11850 Perl_sv_add_backref(aTHX_ MUTABLE_SV(GvSTASH(dstr)), dstr);
11851 GvGP_set(dstr, gp_dup(GvGP(sstr), param));
11852 (void)GpREFCNT_inc(GvGP(dstr));
11856 /* PL_parser->rsfp_filters entries have fake IoDIRP() */
11857 if(IoFLAGS(dstr) & IOf_FAKE_DIRP) {
11858 /* I have no idea why fake dirp (rsfps)
11859 should be treated differently but otherwise
11860 we end up with leaks -- sky*/
11861 IoTOP_GV(dstr) = gv_dup_inc(IoTOP_GV(dstr), param);
11862 IoFMT_GV(dstr) = gv_dup_inc(IoFMT_GV(dstr), param);
11863 IoBOTTOM_GV(dstr) = gv_dup_inc(IoBOTTOM_GV(dstr), param);
11865 IoTOP_GV(dstr) = gv_dup(IoTOP_GV(dstr), param);
11866 IoFMT_GV(dstr) = gv_dup(IoFMT_GV(dstr), param);
11867 IoBOTTOM_GV(dstr) = gv_dup(IoBOTTOM_GV(dstr), param);
11868 if (IoDIRP(dstr)) {
11869 IoDIRP(dstr) = dirp_dup(IoDIRP(dstr), param);
11872 /* IoDIRP(dstr) is already a copy of IoDIRP(sstr) */
11874 IoIFP(dstr) = fp_dup(IoIFP(sstr), IoTYPE(dstr), param);
11876 if (IoOFP(dstr) == IoIFP(sstr))
11877 IoOFP(dstr) = IoIFP(dstr);
11879 IoOFP(dstr) = fp_dup(IoOFP(dstr), IoTYPE(dstr), param);
11880 IoTOP_NAME(dstr) = SAVEPV(IoTOP_NAME(dstr));
11881 IoFMT_NAME(dstr) = SAVEPV(IoFMT_NAME(dstr));
11882 IoBOTTOM_NAME(dstr) = SAVEPV(IoBOTTOM_NAME(dstr));
11885 /* avoid cloning an empty array */
11886 if (AvARRAY((const AV *)sstr) && AvFILLp((const AV *)sstr) >= 0) {
11887 SV **dst_ary, **src_ary;
11888 SSize_t items = AvFILLp((const AV *)sstr) + 1;
11890 src_ary = AvARRAY((const AV *)sstr);
11891 Newxz(dst_ary, AvMAX((const AV *)sstr)+1, SV*);
11892 ptr_table_store(PL_ptr_table, src_ary, dst_ary);
11893 AvARRAY(MUTABLE_AV(dstr)) = dst_ary;
11894 AvALLOC((const AV *)dstr) = dst_ary;
11895 if (AvREAL((const AV *)sstr)) {
11896 dst_ary = sv_dup_inc_multiple(src_ary, dst_ary, items,
11900 while (items-- > 0)
11901 *dst_ary++ = sv_dup(*src_ary++, param);
11903 items = AvMAX((const AV *)sstr) - AvFILLp((const AV *)sstr);
11904 while (items-- > 0) {
11905 *dst_ary++ = &PL_sv_undef;
11909 AvARRAY(MUTABLE_AV(dstr)) = NULL;
11910 AvALLOC((const AV *)dstr) = (SV**)NULL;
11911 AvMAX( (const AV *)dstr) = -1;
11912 AvFILLp((const AV *)dstr) = -1;
11916 if (HvARRAY((const HV *)sstr)) {
11918 const bool sharekeys = !!HvSHAREKEYS(sstr);
11919 XPVHV * const dxhv = (XPVHV*)SvANY(dstr);
11920 XPVHV * const sxhv = (XPVHV*)SvANY(sstr);
11922 Newx(darray, PERL_HV_ARRAY_ALLOC_BYTES(dxhv->xhv_max+1)
11923 + (SvOOK(sstr) ? sizeof(struct xpvhv_aux) : 0),
11925 HvARRAY(dstr) = (HE**)darray;
11926 while (i <= sxhv->xhv_max) {
11927 const HE * const source = HvARRAY(sstr)[i];
11928 HvARRAY(dstr)[i] = source
11929 ? he_dup(source, sharekeys, param) : 0;
11933 const struct xpvhv_aux * const saux = HvAUX(sstr);
11934 struct xpvhv_aux * const daux = HvAUX(dstr);
11935 /* This flag isn't copied. */
11936 /* SvOOK_on(hv) attacks the IV flags. */
11937 SvFLAGS(dstr) |= SVf_OOK;
11939 if (saux->xhv_name_count) {
11940 HEK ** const sname = saux->xhv_name_u.xhvnameu_names;
11942 = saux->xhv_name_count < 0
11943 ? -saux->xhv_name_count
11944 : saux->xhv_name_count;
11945 HEK **shekp = sname + count;
11947 Newx(daux->xhv_name_u.xhvnameu_names, count, HEK *);
11948 dhekp = daux->xhv_name_u.xhvnameu_names + count;
11949 while (shekp-- > sname) {
11951 *dhekp = hek_dup(*shekp, param);
11955 daux->xhv_name_u.xhvnameu_name
11956 = hek_dup(saux->xhv_name_u.xhvnameu_name,
11959 daux->xhv_name_count = saux->xhv_name_count;
11961 daux->xhv_riter = saux->xhv_riter;
11962 daux->xhv_eiter = saux->xhv_eiter
11963 ? he_dup(saux->xhv_eiter,
11964 cBOOL(HvSHAREKEYS(sstr)), param) : 0;
11965 /* backref array needs refcnt=2; see sv_add_backref */
11966 daux->xhv_backreferences =
11967 (param->flags & CLONEf_JOIN_IN)
11968 /* when joining, we let the individual GVs and
11969 * CVs add themselves to backref as
11970 * needed. This avoids pulling in stuff
11971 * that isn't required, and simplifies the
11972 * case where stashes aren't cloned back
11973 * if they already exist in the parent
11976 : saux->xhv_backreferences
11977 ? (SvTYPE(saux->xhv_backreferences) == SVt_PVAV)
11978 ? MUTABLE_AV(SvREFCNT_inc(
11979 sv_dup_inc((const SV *)
11980 saux->xhv_backreferences, param)))
11981 : MUTABLE_AV(sv_dup((const SV *)
11982 saux->xhv_backreferences, param))
11985 daux->xhv_mro_meta = saux->xhv_mro_meta
11986 ? mro_meta_dup(saux->xhv_mro_meta, param)
11989 /* Record stashes for possible cloning in Perl_clone(). */
11991 av_push(param->stashes, dstr);
11995 HvARRAY(MUTABLE_HV(dstr)) = NULL;
11998 if (!(param->flags & CLONEf_COPY_STACKS)) {
12003 /* NOTE: not refcounted */
12004 SvANY(MUTABLE_CV(dstr))->xcv_stash =
12005 hv_dup(CvSTASH(dstr), param);
12006 if ((param->flags & CLONEf_JOIN_IN) && CvSTASH(dstr))
12007 Perl_sv_add_backref(aTHX_ MUTABLE_SV(CvSTASH(dstr)), dstr);
12008 if (!CvISXSUB(dstr)) {
12010 CvROOT(dstr) = OpREFCNT_inc(CvROOT(dstr));
12012 } else if (CvCONST(dstr)) {
12013 CvXSUBANY(dstr).any_ptr =
12014 sv_dup_inc((const SV *)CvXSUBANY(dstr).any_ptr, param);
12016 if (CvDYNFILE(dstr)) CvFILE(dstr) = SAVEPV(CvFILE(dstr));
12017 /* don't dup if copying back - CvGV isn't refcounted, so the
12018 * duped GV may never be freed. A bit of a hack! DAPM */
12019 SvANY(MUTABLE_CV(dstr))->xcv_gv =
12021 ? gv_dup_inc(CvGV(sstr), param)
12022 : (param->flags & CLONEf_JOIN_IN)
12024 : gv_dup(CvGV(sstr), param);
12026 CvPADLIST(dstr) = padlist_dup(CvPADLIST(sstr), param);
12028 CvWEAKOUTSIDE(sstr)
12029 ? cv_dup( CvOUTSIDE(dstr), param)
12030 : cv_dup_inc(CvOUTSIDE(dstr), param);
12036 if (SvOBJECT(dstr) && SvTYPE(dstr) != SVt_PVIO)
12043 Perl_sv_dup_inc(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
12045 PERL_ARGS_ASSERT_SV_DUP_INC;
12046 return sstr ? SvREFCNT_inc(sv_dup_common(sstr, param)) : NULL;
12050 Perl_sv_dup(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
12052 SV *dstr = sstr ? sv_dup_common(sstr, param) : NULL;
12053 PERL_ARGS_ASSERT_SV_DUP;
12055 /* Track every SV that (at least initially) had a reference count of 0.
12056 We need to do this by holding an actual reference to it in this array.
12057 If we attempt to cheat, turn AvREAL_off(), and store only pointers
12058 (akin to the stashes hash, and the perl stack), we come unstuck if
12059 a weak reference (or other SV legitimately SvREFCNT() == 0 for this
12060 thread) is manipulated in a CLONE method, because CLONE runs before the
12061 unreferenced array is walked to find SVs still with SvREFCNT() == 0
12062 (and fix things up by giving each a reference via the temps stack).
12063 Instead, during CLONE, if the 0-referenced SV has SvREFCNT_inc() and
12064 then SvREFCNT_dec(), it will be cleaned up (and added to the free list)
12065 before the walk of unreferenced happens and a reference to that is SV
12066 added to the temps stack. At which point we have the same SV considered
12067 to be in use, and free to be re-used. Not good.
12069 if (dstr && !(param->flags & CLONEf_COPY_STACKS) && !SvREFCNT(dstr)) {
12070 assert(param->unreferenced);
12071 av_push(param->unreferenced, SvREFCNT_inc(dstr));
12077 /* duplicate a context */
12080 Perl_cx_dup(pTHX_ PERL_CONTEXT *cxs, I32 ix, I32 max, CLONE_PARAMS* param)
12082 PERL_CONTEXT *ncxs;
12084 PERL_ARGS_ASSERT_CX_DUP;
12087 return (PERL_CONTEXT*)NULL;
12089 /* look for it in the table first */
12090 ncxs = (PERL_CONTEXT*)ptr_table_fetch(PL_ptr_table, cxs);
12094 /* create anew and remember what it is */
12095 Newx(ncxs, max + 1, PERL_CONTEXT);
12096 ptr_table_store(PL_ptr_table, cxs, ncxs);
12097 Copy(cxs, ncxs, max + 1, PERL_CONTEXT);
12100 PERL_CONTEXT * const ncx = &ncxs[ix];
12101 if (CxTYPE(ncx) == CXt_SUBST) {
12102 Perl_croak(aTHX_ "Cloning substitution context is unimplemented");
12105 switch (CxTYPE(ncx)) {
12107 ncx->blk_sub.cv = (ncx->blk_sub.olddepth == 0
12108 ? cv_dup_inc(ncx->blk_sub.cv, param)
12109 : cv_dup(ncx->blk_sub.cv,param));
12110 ncx->blk_sub.argarray = (CxHASARGS(ncx)
12111 ? av_dup_inc(ncx->blk_sub.argarray,
12114 ncx->blk_sub.savearray = av_dup_inc(ncx->blk_sub.savearray,
12116 ncx->blk_sub.oldcomppad = (PAD*)ptr_table_fetch(PL_ptr_table,
12117 ncx->blk_sub.oldcomppad);
12120 ncx->blk_eval.old_namesv = sv_dup_inc(ncx->blk_eval.old_namesv,
12122 ncx->blk_eval.cur_text = sv_dup(ncx->blk_eval.cur_text, param);
12124 case CXt_LOOP_LAZYSV:
12125 ncx->blk_loop.state_u.lazysv.end
12126 = sv_dup_inc(ncx->blk_loop.state_u.lazysv.end, param);
12127 /* We are taking advantage of av_dup_inc and sv_dup_inc
12128 actually being the same function, and order equivalence of
12130 We can assert the later [but only at run time :-(] */
12131 assert ((void *) &ncx->blk_loop.state_u.ary.ary ==
12132 (void *) &ncx->blk_loop.state_u.lazysv.cur);
12134 ncx->blk_loop.state_u.ary.ary
12135 = av_dup_inc(ncx->blk_loop.state_u.ary.ary, param);
12136 case CXt_LOOP_LAZYIV:
12137 case CXt_LOOP_PLAIN:
12138 if (CxPADLOOP(ncx)) {
12139 ncx->blk_loop.itervar_u.oldcomppad
12140 = (PAD*)ptr_table_fetch(PL_ptr_table,
12141 ncx->blk_loop.itervar_u.oldcomppad);
12143 ncx->blk_loop.itervar_u.gv
12144 = gv_dup((const GV *)ncx->blk_loop.itervar_u.gv,
12149 ncx->blk_format.cv = cv_dup(ncx->blk_format.cv, param);
12150 ncx->blk_format.gv = gv_dup(ncx->blk_format.gv, param);
12151 ncx->blk_format.dfoutgv = gv_dup_inc(ncx->blk_format.dfoutgv,
12164 /* duplicate a stack info structure */
12167 Perl_si_dup(pTHX_ PERL_SI *si, CLONE_PARAMS* param)
12171 PERL_ARGS_ASSERT_SI_DUP;
12174 return (PERL_SI*)NULL;
12176 /* look for it in the table first */
12177 nsi = (PERL_SI*)ptr_table_fetch(PL_ptr_table, si);
12181 /* create anew and remember what it is */
12182 Newxz(nsi, 1, PERL_SI);
12183 ptr_table_store(PL_ptr_table, si, nsi);
12185 nsi->si_stack = av_dup_inc(si->si_stack, param);
12186 nsi->si_cxix = si->si_cxix;
12187 nsi->si_cxmax = si->si_cxmax;
12188 nsi->si_cxstack = cx_dup(si->si_cxstack, si->si_cxix, si->si_cxmax, param);
12189 nsi->si_type = si->si_type;
12190 nsi->si_prev = si_dup(si->si_prev, param);
12191 nsi->si_next = si_dup(si->si_next, param);
12192 nsi->si_markoff = si->si_markoff;
12197 #define POPINT(ss,ix) ((ss)[--(ix)].any_i32)
12198 #define TOPINT(ss,ix) ((ss)[ix].any_i32)
12199 #define POPLONG(ss,ix) ((ss)[--(ix)].any_long)
12200 #define TOPLONG(ss,ix) ((ss)[ix].any_long)
12201 #define POPIV(ss,ix) ((ss)[--(ix)].any_iv)
12202 #define TOPIV(ss,ix) ((ss)[ix].any_iv)
12203 #define POPUV(ss,ix) ((ss)[--(ix)].any_uv)
12204 #define TOPUV(ss,ix) ((ss)[ix].any_uv)
12205 #define POPBOOL(ss,ix) ((ss)[--(ix)].any_bool)
12206 #define TOPBOOL(ss,ix) ((ss)[ix].any_bool)
12207 #define POPPTR(ss,ix) ((ss)[--(ix)].any_ptr)
12208 #define TOPPTR(ss,ix) ((ss)[ix].any_ptr)
12209 #define POPDPTR(ss,ix) ((ss)[--(ix)].any_dptr)
12210 #define TOPDPTR(ss,ix) ((ss)[ix].any_dptr)
12211 #define POPDXPTR(ss,ix) ((ss)[--(ix)].any_dxptr)
12212 #define TOPDXPTR(ss,ix) ((ss)[ix].any_dxptr)
12215 #define pv_dup_inc(p) SAVEPV(p)
12216 #define pv_dup(p) SAVEPV(p)
12217 #define svp_dup_inc(p,pp) any_dup(p,pp)
12219 /* map any object to the new equivent - either something in the
12220 * ptr table, or something in the interpreter structure
12224 Perl_any_dup(pTHX_ void *v, const PerlInterpreter *proto_perl)
12228 PERL_ARGS_ASSERT_ANY_DUP;
12231 return (void*)NULL;
12233 /* look for it in the table first */
12234 ret = ptr_table_fetch(PL_ptr_table, v);
12238 /* see if it is part of the interpreter structure */
12239 if (v >= (void*)proto_perl && v < (void*)(proto_perl+1))
12240 ret = (void*)(((char*)aTHX) + (((char*)v) - (char*)proto_perl));
12248 /* duplicate the save stack */
12251 Perl_ss_dup(pTHX_ PerlInterpreter *proto_perl, CLONE_PARAMS* param)
12254 ANY * const ss = proto_perl->Isavestack;
12255 const I32 max = proto_perl->Isavestack_max;
12256 I32 ix = proto_perl->Isavestack_ix;
12269 void (*dptr) (void*);
12270 void (*dxptr) (pTHX_ void*);
12272 PERL_ARGS_ASSERT_SS_DUP;
12274 Newxz(nss, max, ANY);
12277 const UV uv = POPUV(ss,ix);
12278 const U8 type = (U8)uv & SAVE_MASK;
12280 TOPUV(nss,ix) = uv;
12282 case SAVEt_CLEARSV:
12284 case SAVEt_HELEM: /* hash element */
12285 sv = (const SV *)POPPTR(ss,ix);
12286 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12288 case SAVEt_ITEM: /* normal string */
12289 case SAVEt_GVSV: /* scalar slot in GV */
12290 case SAVEt_SV: /* scalar reference */
12291 sv = (const SV *)POPPTR(ss,ix);
12292 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12295 case SAVEt_MORTALIZESV:
12296 sv = (const SV *)POPPTR(ss,ix);
12297 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12299 case SAVEt_SHARED_PVREF: /* char* in shared space */
12300 c = (char*)POPPTR(ss,ix);
12301 TOPPTR(nss,ix) = savesharedpv(c);
12302 ptr = POPPTR(ss,ix);
12303 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12305 case SAVEt_GENERIC_SVREF: /* generic sv */
12306 case SAVEt_SVREF: /* scalar reference */
12307 sv = (const SV *)POPPTR(ss,ix);
12308 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12309 ptr = POPPTR(ss,ix);
12310 TOPPTR(nss,ix) = svp_dup_inc((SV**)ptr, proto_perl);/* XXXXX */
12312 case SAVEt_HV: /* hash reference */
12313 case SAVEt_AV: /* array reference */
12314 sv = (const SV *) POPPTR(ss,ix);
12315 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12317 case SAVEt_COMPPAD:
12319 sv = (const SV *) POPPTR(ss,ix);
12320 TOPPTR(nss,ix) = sv_dup(sv, param);
12322 case SAVEt_INT: /* int reference */
12323 ptr = POPPTR(ss,ix);
12324 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12325 intval = (int)POPINT(ss,ix);
12326 TOPINT(nss,ix) = intval;
12328 case SAVEt_LONG: /* long reference */
12329 ptr = POPPTR(ss,ix);
12330 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12331 longval = (long)POPLONG(ss,ix);
12332 TOPLONG(nss,ix) = longval;
12334 case SAVEt_I32: /* I32 reference */
12335 ptr = POPPTR(ss,ix);
12336 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12338 TOPINT(nss,ix) = i;
12340 case SAVEt_IV: /* IV reference */
12341 ptr = POPPTR(ss,ix);
12342 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12344 TOPIV(nss,ix) = iv;
12346 case SAVEt_HPTR: /* HV* reference */
12347 case SAVEt_APTR: /* AV* reference */
12348 case SAVEt_SPTR: /* SV* reference */
12349 ptr = POPPTR(ss,ix);
12350 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12351 sv = (const SV *)POPPTR(ss,ix);
12352 TOPPTR(nss,ix) = sv_dup(sv, param);
12354 case SAVEt_VPTR: /* random* reference */
12355 ptr = POPPTR(ss,ix);
12356 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12358 case SAVEt_INT_SMALL:
12359 case SAVEt_I32_SMALL:
12360 case SAVEt_I16: /* I16 reference */
12361 case SAVEt_I8: /* I8 reference */
12363 ptr = POPPTR(ss,ix);
12364 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12366 case SAVEt_GENERIC_PVREF: /* generic char* */
12367 case SAVEt_PPTR: /* char* reference */
12368 ptr = POPPTR(ss,ix);
12369 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12370 c = (char*)POPPTR(ss,ix);
12371 TOPPTR(nss,ix) = pv_dup(c);
12373 case SAVEt_GP: /* scalar reference */
12374 gp = (GP*)POPPTR(ss,ix);
12375 TOPPTR(nss,ix) = gp = gp_dup(gp, param);
12376 (void)GpREFCNT_inc(gp);
12377 gv = (const GV *)POPPTR(ss,ix);
12378 TOPPTR(nss,ix) = gv_dup_inc(gv, param);
12381 ptr = POPPTR(ss,ix);
12382 if (ptr && (((OP*)ptr)->op_private & OPpREFCOUNTED)) {
12383 /* these are assumed to be refcounted properly */
12385 switch (((OP*)ptr)->op_type) {
12387 case OP_LEAVESUBLV:
12391 case OP_LEAVEWRITE:
12392 TOPPTR(nss,ix) = ptr;
12395 (void) OpREFCNT_inc(o);
12399 TOPPTR(nss,ix) = NULL;
12404 TOPPTR(nss,ix) = NULL;
12406 case SAVEt_FREECOPHH:
12407 ptr = POPPTR(ss,ix);
12408 TOPPTR(nss,ix) = cophh_copy((COPHH *)ptr);
12411 hv = (const HV *)POPPTR(ss,ix);
12412 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
12414 TOPINT(nss,ix) = i;
12417 c = (char*)POPPTR(ss,ix);
12418 TOPPTR(nss,ix) = pv_dup_inc(c);
12420 case SAVEt_STACK_POS: /* Position on Perl stack */
12422 TOPINT(nss,ix) = i;
12424 case SAVEt_DESTRUCTOR:
12425 ptr = POPPTR(ss,ix);
12426 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
12427 dptr = POPDPTR(ss,ix);
12428 TOPDPTR(nss,ix) = DPTR2FPTR(void (*)(void*),
12429 any_dup(FPTR2DPTR(void *, dptr),
12432 case SAVEt_DESTRUCTOR_X:
12433 ptr = POPPTR(ss,ix);
12434 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
12435 dxptr = POPDXPTR(ss,ix);
12436 TOPDXPTR(nss,ix) = DPTR2FPTR(void (*)(pTHX_ void*),
12437 any_dup(FPTR2DPTR(void *, dxptr),
12440 case SAVEt_REGCONTEXT:
12442 ix -= uv >> SAVE_TIGHT_SHIFT;
12444 case SAVEt_AELEM: /* array element */
12445 sv = (const SV *)POPPTR(ss,ix);
12446 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12448 TOPINT(nss,ix) = i;
12449 av = (const AV *)POPPTR(ss,ix);
12450 TOPPTR(nss,ix) = av_dup_inc(av, param);
12453 ptr = POPPTR(ss,ix);
12454 TOPPTR(nss,ix) = ptr;
12457 ptr = POPPTR(ss,ix);
12458 ptr = cophh_copy((COPHH*)ptr);
12459 TOPPTR(nss,ix) = ptr;
12461 TOPINT(nss,ix) = i;
12462 if (i & HINT_LOCALIZE_HH) {
12463 hv = (const HV *)POPPTR(ss,ix);
12464 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
12467 case SAVEt_PADSV_AND_MORTALIZE:
12468 longval = (long)POPLONG(ss,ix);
12469 TOPLONG(nss,ix) = longval;
12470 ptr = POPPTR(ss,ix);
12471 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12472 sv = (const SV *)POPPTR(ss,ix);
12473 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12475 case SAVEt_SET_SVFLAGS:
12477 TOPINT(nss,ix) = i;
12479 TOPINT(nss,ix) = i;
12480 sv = (const SV *)POPPTR(ss,ix);
12481 TOPPTR(nss,ix) = sv_dup(sv, param);
12483 case SAVEt_RE_STATE:
12485 const struct re_save_state *const old_state
12486 = (struct re_save_state *)
12487 (ss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
12488 struct re_save_state *const new_state
12489 = (struct re_save_state *)
12490 (nss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
12492 Copy(old_state, new_state, 1, struct re_save_state);
12493 ix -= SAVESTACK_ALLOC_FOR_RE_SAVE_STATE;
12495 new_state->re_state_bostr
12496 = pv_dup(old_state->re_state_bostr);
12497 new_state->re_state_reginput
12498 = pv_dup(old_state->re_state_reginput);
12499 new_state->re_state_regeol
12500 = pv_dup(old_state->re_state_regeol);
12501 new_state->re_state_regoffs
12502 = (regexp_paren_pair*)
12503 any_dup(old_state->re_state_regoffs, proto_perl);
12504 new_state->re_state_reglastparen
12505 = (U32*) any_dup(old_state->re_state_reglastparen,
12507 new_state->re_state_reglastcloseparen
12508 = (U32*)any_dup(old_state->re_state_reglastcloseparen,
12510 /* XXX This just has to be broken. The old save_re_context
12511 code did SAVEGENERICPV(PL_reg_start_tmp);
12512 PL_reg_start_tmp is char **.
12513 Look above to what the dup code does for
12514 SAVEt_GENERIC_PVREF
12515 It can never have worked.
12516 So this is merely a faithful copy of the exiting bug: */
12517 new_state->re_state_reg_start_tmp
12518 = (char **) pv_dup((char *)
12519 old_state->re_state_reg_start_tmp);
12520 /* I assume that it only ever "worked" because no-one called
12521 (pseudo)fork while the regexp engine had re-entered itself.
12523 #ifdef PERL_OLD_COPY_ON_WRITE
12524 new_state->re_state_nrs
12525 = sv_dup(old_state->re_state_nrs, param);
12527 new_state->re_state_reg_magic
12528 = (MAGIC*) any_dup(old_state->re_state_reg_magic,
12530 new_state->re_state_reg_oldcurpm
12531 = (PMOP*) any_dup(old_state->re_state_reg_oldcurpm,
12533 new_state->re_state_reg_curpm
12534 = (PMOP*) any_dup(old_state->re_state_reg_curpm,
12536 new_state->re_state_reg_oldsaved
12537 = pv_dup(old_state->re_state_reg_oldsaved);
12538 new_state->re_state_reg_poscache
12539 = pv_dup(old_state->re_state_reg_poscache);
12540 new_state->re_state_reg_starttry
12541 = pv_dup(old_state->re_state_reg_starttry);
12544 case SAVEt_COMPILE_WARNINGS:
12545 ptr = POPPTR(ss,ix);
12546 TOPPTR(nss,ix) = DUP_WARNINGS((STRLEN*)ptr);
12549 ptr = POPPTR(ss,ix);
12550 TOPPTR(nss,ix) = parser_dup((const yy_parser*)ptr, param);
12554 "panic: ss_dup inconsistency (%"IVdf")", (IV) type);
12562 /* if sv is a stash, call $class->CLONE_SKIP(), and set the SVphv_CLONEABLE
12563 * flag to the result. This is done for each stash before cloning starts,
12564 * so we know which stashes want their objects cloned */
12567 do_mark_cloneable_stash(pTHX_ SV *const sv)
12569 const HEK * const hvname = HvNAME_HEK((const HV *)sv);
12571 GV* const cloner = gv_fetchmethod_autoload(MUTABLE_HV(sv), "CLONE_SKIP", 0);
12572 SvFLAGS(sv) |= SVphv_CLONEABLE; /* clone objects by default */
12573 if (cloner && GvCV(cloner)) {
12580 mXPUSHs(newSVhek(hvname));
12582 call_sv(MUTABLE_SV(GvCV(cloner)), G_SCALAR);
12589 SvFLAGS(sv) &= ~SVphv_CLONEABLE;
12597 =for apidoc perl_clone
12599 Create and return a new interpreter by cloning the current one.
12601 perl_clone takes these flags as parameters:
12603 CLONEf_COPY_STACKS - is used to, well, copy the stacks also,
12604 without it we only clone the data and zero the stacks,
12605 with it we copy the stacks and the new perl interpreter is
12606 ready to run at the exact same point as the previous one.
12607 The pseudo-fork code uses COPY_STACKS while the
12608 threads->create doesn't.
12610 CLONEf_KEEP_PTR_TABLE
12611 perl_clone keeps a ptr_table with the pointer of the old
12612 variable as a key and the new variable as a value,
12613 this allows it to check if something has been cloned and not
12614 clone it again but rather just use the value and increase the
12615 refcount. If KEEP_PTR_TABLE is not set then perl_clone will kill
12616 the ptr_table using the function
12617 C<ptr_table_free(PL_ptr_table); PL_ptr_table = NULL;>,
12618 reason to keep it around is if you want to dup some of your own
12619 variable who are outside the graph perl scans, example of this
12620 code is in threads.xs create
12623 This is a win32 thing, it is ignored on unix, it tells perls
12624 win32host code (which is c++) to clone itself, this is needed on
12625 win32 if you want to run two threads at the same time,
12626 if you just want to do some stuff in a separate perl interpreter
12627 and then throw it away and return to the original one,
12628 you don't need to do anything.
12633 /* XXX the above needs expanding by someone who actually understands it ! */
12634 EXTERN_C PerlInterpreter *
12635 perl_clone_host(PerlInterpreter* proto_perl, UV flags);
12638 perl_clone(PerlInterpreter *proto_perl, UV flags)
12641 #ifdef PERL_IMPLICIT_SYS
12643 PERL_ARGS_ASSERT_PERL_CLONE;
12645 /* perlhost.h so we need to call into it
12646 to clone the host, CPerlHost should have a c interface, sky */
12648 if (flags & CLONEf_CLONE_HOST) {
12649 return perl_clone_host(proto_perl,flags);
12651 return perl_clone_using(proto_perl, flags,
12653 proto_perl->IMemShared,
12654 proto_perl->IMemParse,
12656 proto_perl->IStdIO,
12660 proto_perl->IProc);
12664 perl_clone_using(PerlInterpreter *proto_perl, UV flags,
12665 struct IPerlMem* ipM, struct IPerlMem* ipMS,
12666 struct IPerlMem* ipMP, struct IPerlEnv* ipE,
12667 struct IPerlStdIO* ipStd, struct IPerlLIO* ipLIO,
12668 struct IPerlDir* ipD, struct IPerlSock* ipS,
12669 struct IPerlProc* ipP)
12671 /* XXX many of the string copies here can be optimized if they're
12672 * constants; they need to be allocated as common memory and just
12673 * their pointers copied. */
12676 CLONE_PARAMS clone_params;
12677 CLONE_PARAMS* const param = &clone_params;
12679 PerlInterpreter * const my_perl = (PerlInterpreter*)(*ipM->pMalloc)(ipM, sizeof(PerlInterpreter));
12681 PERL_ARGS_ASSERT_PERL_CLONE_USING;
12682 #else /* !PERL_IMPLICIT_SYS */
12684 CLONE_PARAMS clone_params;
12685 CLONE_PARAMS* param = &clone_params;
12686 PerlInterpreter * const my_perl = (PerlInterpreter*)PerlMem_malloc(sizeof(PerlInterpreter));
12688 PERL_ARGS_ASSERT_PERL_CLONE;
12689 #endif /* PERL_IMPLICIT_SYS */
12691 /* for each stash, determine whether its objects should be cloned */
12692 S_visit(proto_perl, do_mark_cloneable_stash, SVt_PVHV, SVTYPEMASK);
12693 PERL_SET_THX(my_perl);
12696 PoisonNew(my_perl, 1, PerlInterpreter);
12699 PL_defstash = NULL; /* may be used by perl malloc() */
12702 PL_scopestack_name = 0;
12704 PL_savestack_ix = 0;
12705 PL_savestack_max = -1;
12706 PL_sig_pending = 0;
12708 Zero(&PL_debug_pad, 1, struct perl_debug_pad);
12709 # ifdef DEBUG_LEAKING_SCALARS
12710 PL_sv_serial = (((UV)my_perl >> 2) & 0xfff) * 1000000;
12712 #else /* !DEBUGGING */
12713 Zero(my_perl, 1, PerlInterpreter);
12714 #endif /* DEBUGGING */
12716 #ifdef PERL_IMPLICIT_SYS
12717 /* host pointers */
12719 PL_MemShared = ipMS;
12720 PL_MemParse = ipMP;
12727 #endif /* PERL_IMPLICIT_SYS */
12729 param->flags = flags;
12730 /* Nothing in the core code uses this, but we make it available to
12731 extensions (using mg_dup). */
12732 param->proto_perl = proto_perl;
12733 /* Likely nothing will use this, but it is initialised to be consistent
12734 with Perl_clone_params_new(). */
12735 param->new_perl = my_perl;
12736 param->unreferenced = NULL;
12738 INIT_TRACK_MEMPOOL(my_perl->Imemory_debug_header, my_perl);
12740 PL_body_arenas = NULL;
12741 Zero(&PL_body_roots, 1, PL_body_roots);
12744 PL_sv_objcount = 0;
12746 PL_sv_arenaroot = NULL;
12748 PL_debug = proto_perl->Idebug;
12750 PL_hash_seed = proto_perl->Ihash_seed;
12751 PL_rehash_seed = proto_perl->Irehash_seed;
12753 SvANY(&PL_sv_undef) = NULL;
12754 SvREFCNT(&PL_sv_undef) = (~(U32)0)/2;
12755 SvFLAGS(&PL_sv_undef) = SVf_READONLY|SVt_NULL;
12756 SvREFCNT(&PL_sv_no) = (~(U32)0)/2;
12757 SvFLAGS(&PL_sv_no) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
12758 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
12760 SvANY(&PL_sv_yes) = new_XPVNV();
12761 SvREFCNT(&PL_sv_yes) = (~(U32)0)/2;
12762 SvFLAGS(&PL_sv_yes) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
12763 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
12765 /* dbargs array probably holds garbage */
12768 PL_compiling = proto_perl->Icompiling;
12770 #ifdef PERL_DEBUG_READONLY_OPS
12775 /* pseudo environmental stuff */
12776 PL_origargc = proto_perl->Iorigargc;
12777 PL_origargv = proto_perl->Iorigargv;
12779 /* Set tainting stuff before PerlIO_debug can possibly get called */
12780 PL_tainting = proto_perl->Itainting;
12781 PL_taint_warn = proto_perl->Itaint_warn;
12783 PL_minus_c = proto_perl->Iminus_c;
12785 PL_localpatches = proto_perl->Ilocalpatches;
12786 PL_splitstr = proto_perl->Isplitstr;
12787 PL_minus_n = proto_perl->Iminus_n;
12788 PL_minus_p = proto_perl->Iminus_p;
12789 PL_minus_l = proto_perl->Iminus_l;
12790 PL_minus_a = proto_perl->Iminus_a;
12791 PL_minus_E = proto_perl->Iminus_E;
12792 PL_minus_F = proto_perl->Iminus_F;
12793 PL_doswitches = proto_perl->Idoswitches;
12794 PL_dowarn = proto_perl->Idowarn;
12795 PL_sawampersand = proto_perl->Isawampersand;
12796 PL_unsafe = proto_perl->Iunsafe;
12797 PL_perldb = proto_perl->Iperldb;
12798 PL_perl_destruct_level = proto_perl->Iperl_destruct_level;
12799 PL_exit_flags = proto_perl->Iexit_flags;
12801 /* XXX time(&PL_basetime) when asked for? */
12802 PL_basetime = proto_perl->Ibasetime;
12804 PL_maxsysfd = proto_perl->Imaxsysfd;
12805 PL_statusvalue = proto_perl->Istatusvalue;
12807 PL_statusvalue_vms = proto_perl->Istatusvalue_vms;
12809 PL_statusvalue_posix = proto_perl->Istatusvalue_posix;
12812 /* RE engine related */
12813 Zero(&PL_reg_state, 1, struct re_save_state);
12814 PL_reginterp_cnt = 0;
12815 PL_regmatch_slab = NULL;
12817 PL_sub_generation = proto_perl->Isub_generation;
12819 /* funky return mechanisms */
12820 PL_forkprocess = proto_perl->Iforkprocess;
12822 /* internal state */
12823 PL_maxo = proto_perl->Imaxo;
12825 PL_main_start = proto_perl->Imain_start;
12826 PL_eval_root = proto_perl->Ieval_root;
12827 PL_eval_start = proto_perl->Ieval_start;
12829 PL_filemode = proto_perl->Ifilemode;
12830 PL_lastfd = proto_perl->Ilastfd;
12831 PL_oldname = proto_perl->Ioldname; /* XXX not quite right */
12834 PL_gensym = proto_perl->Igensym;
12836 PL_laststatval = proto_perl->Ilaststatval;
12837 PL_laststype = proto_perl->Ilaststype;
12840 PL_profiledata = NULL;
12842 PL_generation = proto_perl->Igeneration;
12844 PL_in_clean_objs = proto_perl->Iin_clean_objs;
12845 PL_in_clean_all = proto_perl->Iin_clean_all;
12847 PL_uid = proto_perl->Iuid;
12848 PL_euid = proto_perl->Ieuid;
12849 PL_gid = proto_perl->Igid;
12850 PL_egid = proto_perl->Iegid;
12851 PL_nomemok = proto_perl->Inomemok;
12852 PL_an = proto_perl->Ian;
12853 PL_evalseq = proto_perl->Ievalseq;
12854 PL_origenviron = proto_perl->Iorigenviron; /* XXX not quite right */
12855 PL_origalen = proto_perl->Iorigalen;
12857 PL_sighandlerp = proto_perl->Isighandlerp;
12859 PL_runops = proto_perl->Irunops;
12861 PL_subline = proto_perl->Isubline;
12864 PL_cryptseen = proto_perl->Icryptseen;
12867 PL_hints = proto_perl->Ihints;
12869 PL_amagic_generation = proto_perl->Iamagic_generation;
12871 #ifdef USE_LOCALE_COLLATE
12872 PL_collation_ix = proto_perl->Icollation_ix;
12873 PL_collation_standard = proto_perl->Icollation_standard;
12874 PL_collxfrm_base = proto_perl->Icollxfrm_base;
12875 PL_collxfrm_mult = proto_perl->Icollxfrm_mult;
12876 #endif /* USE_LOCALE_COLLATE */
12878 #ifdef USE_LOCALE_NUMERIC
12879 PL_numeric_standard = proto_perl->Inumeric_standard;
12880 PL_numeric_local = proto_perl->Inumeric_local;
12881 #endif /* !USE_LOCALE_NUMERIC */
12883 /* Did the locale setup indicate UTF-8? */
12884 PL_utf8locale = proto_perl->Iutf8locale;
12885 /* Unicode features (see perlrun/-C) */
12886 PL_unicode = proto_perl->Iunicode;
12888 /* Pre-5.8 signals control */
12889 PL_signals = proto_perl->Isignals;
12891 /* times() ticks per second */
12892 PL_clocktick = proto_perl->Iclocktick;
12894 /* Recursion stopper for PerlIO_find_layer */
12895 PL_in_load_module = proto_perl->Iin_load_module;
12897 /* sort() routine */
12898 PL_sort_RealCmp = proto_perl->Isort_RealCmp;
12900 /* Not really needed/useful since the reenrant_retint is "volatile",
12901 * but do it for consistency's sake. */
12902 PL_reentrant_retint = proto_perl->Ireentrant_retint;
12904 /* Hooks to shared SVs and locks. */
12905 PL_sharehook = proto_perl->Isharehook;
12906 PL_lockhook = proto_perl->Ilockhook;
12907 PL_unlockhook = proto_perl->Iunlockhook;
12908 PL_threadhook = proto_perl->Ithreadhook;
12909 PL_destroyhook = proto_perl->Idestroyhook;
12910 PL_signalhook = proto_perl->Isignalhook;
12912 #ifdef THREADS_HAVE_PIDS
12913 PL_ppid = proto_perl->Ippid;
12917 PL_last_swash_hv = NULL; /* reinits on demand */
12918 PL_last_swash_klen = 0;
12919 PL_last_swash_key[0]= '\0';
12920 PL_last_swash_tmps = (U8*)NULL;
12921 PL_last_swash_slen = 0;
12923 PL_glob_index = proto_perl->Iglob_index;
12924 PL_srand_called = proto_perl->Isrand_called;
12926 if (flags & CLONEf_COPY_STACKS) {
12927 /* next allocation will be PL_tmps_stack[PL_tmps_ix+1] */
12928 PL_tmps_ix = proto_perl->Itmps_ix;
12929 PL_tmps_max = proto_perl->Itmps_max;
12930 PL_tmps_floor = proto_perl->Itmps_floor;
12932 /* next push_scope()/ENTER sets PL_scopestack[PL_scopestack_ix]
12933 * NOTE: unlike the others! */
12934 PL_scopestack_ix = proto_perl->Iscopestack_ix;
12935 PL_scopestack_max = proto_perl->Iscopestack_max;
12937 /* next SSPUSHFOO() sets PL_savestack[PL_savestack_ix]
12938 * NOTE: unlike the others! */
12939 PL_savestack_ix = proto_perl->Isavestack_ix;
12940 PL_savestack_max = proto_perl->Isavestack_max;
12943 PL_start_env = proto_perl->Istart_env; /* XXXXXX */
12944 PL_top_env = &PL_start_env;
12946 PL_op = proto_perl->Iop;
12949 PL_Xpv = (XPV*)NULL;
12950 my_perl->Ina = proto_perl->Ina;
12952 PL_statbuf = proto_perl->Istatbuf;
12953 PL_statcache = proto_perl->Istatcache;
12956 PL_timesbuf = proto_perl->Itimesbuf;
12959 PL_tainted = proto_perl->Itainted;
12960 PL_curpm = proto_perl->Icurpm; /* XXX No PMOP ref count */
12962 PL_chopset = proto_perl->Ichopset; /* XXX never deallocated */
12964 PL_restartjmpenv = proto_perl->Irestartjmpenv;
12965 PL_restartop = proto_perl->Irestartop;
12966 PL_in_eval = proto_perl->Iin_eval;
12967 PL_delaymagic = proto_perl->Idelaymagic;
12968 PL_phase = proto_perl->Iphase;
12969 PL_localizing = proto_perl->Ilocalizing;
12971 PL_hv_fetch_ent_mh = NULL;
12972 PL_modcount = proto_perl->Imodcount;
12973 PL_lastgotoprobe = NULL;
12974 PL_dumpindent = proto_perl->Idumpindent;
12976 PL_efloatbuf = NULL; /* reinits on demand */
12977 PL_efloatsize = 0; /* reinits on demand */
12981 PL_regdummy = proto_perl->Iregdummy;
12982 PL_colorset = 0; /* reinits PL_colors[] */
12983 /*PL_colors[6] = {0,0,0,0,0,0};*/
12985 /* Pluggable optimizer */
12986 PL_peepp = proto_perl->Ipeepp;
12987 PL_rpeepp = proto_perl->Irpeepp;
12988 /* op_free() hook */
12989 PL_opfreehook = proto_perl->Iopfreehook;
12991 #ifdef USE_REENTRANT_API
12992 /* XXX: things like -Dm will segfault here in perlio, but doing
12993 * PERL_SET_CONTEXT(proto_perl);
12994 * breaks too many other things
12996 Perl_reentrant_init(aTHX);
12999 /* create SV map for pointer relocation */
13000 PL_ptr_table = ptr_table_new();
13002 /* initialize these special pointers as early as possible */
13003 ptr_table_store(PL_ptr_table, &proto_perl->Isv_undef, &PL_sv_undef);
13005 SvANY(&PL_sv_no) = new_XPVNV();
13006 SvPV_set(&PL_sv_no, savepvn(PL_No, 0));
13007 SvCUR_set(&PL_sv_no, 0);
13008 SvLEN_set(&PL_sv_no, 1);
13009 SvIV_set(&PL_sv_no, 0);
13010 SvNV_set(&PL_sv_no, 0);
13011 ptr_table_store(PL_ptr_table, &proto_perl->Isv_no, &PL_sv_no);
13013 SvPV_set(&PL_sv_yes, savepvn(PL_Yes, 1));
13014 SvCUR_set(&PL_sv_yes, 1);
13015 SvLEN_set(&PL_sv_yes, 2);
13016 SvIV_set(&PL_sv_yes, 1);
13017 SvNV_set(&PL_sv_yes, 1);
13018 ptr_table_store(PL_ptr_table, &proto_perl->Isv_yes, &PL_sv_yes);
13020 /* create (a non-shared!) shared string table */
13021 PL_strtab = newHV();
13022 HvSHAREKEYS_off(PL_strtab);
13023 hv_ksplit(PL_strtab, HvTOTALKEYS(proto_perl->Istrtab));
13024 ptr_table_store(PL_ptr_table, proto_perl->Istrtab, PL_strtab);
13026 /* These two PVs will be free'd special way so must set them same way op.c does */
13027 PL_compiling.cop_stashpv = savesharedpv(PL_compiling.cop_stashpv);
13028 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_stashpv, PL_compiling.cop_stashpv);
13030 PL_compiling.cop_file = savesharedpv(PL_compiling.cop_file);
13031 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_file, PL_compiling.cop_file);
13033 ptr_table_store(PL_ptr_table, &proto_perl->Icompiling, &PL_compiling);
13034 PL_compiling.cop_warnings = DUP_WARNINGS(PL_compiling.cop_warnings);
13035 CopHINTHASH_set(&PL_compiling, cophh_copy(CopHINTHASH_get(&PL_compiling)));
13036 PL_curcop = (COP*)any_dup(proto_perl->Icurcop, proto_perl);
13038 param->stashes = newAV(); /* Setup array of objects to call clone on */
13039 /* This makes no difference to the implementation, as it always pushes
13040 and shifts pointers to other SVs without changing their reference
13041 count, with the array becoming empty before it is freed. However, it
13042 makes it conceptually clear what is going on, and will avoid some
13043 work inside av.c, filling slots between AvFILL() and AvMAX() with
13044 &PL_sv_undef, and SvREFCNT_dec()ing those. */
13045 AvREAL_off(param->stashes);
13047 if (!(flags & CLONEf_COPY_STACKS)) {
13048 param->unreferenced = newAV();
13051 #ifdef PERLIO_LAYERS
13052 /* Clone PerlIO tables as soon as we can handle general xx_dup() */
13053 PerlIO_clone(aTHX_ proto_perl, param);
13056 PL_envgv = gv_dup(proto_perl->Ienvgv, param);
13057 PL_incgv = gv_dup(proto_perl->Iincgv, param);
13058 PL_hintgv = gv_dup(proto_perl->Ihintgv, param);
13059 PL_origfilename = SAVEPV(proto_perl->Iorigfilename);
13060 PL_diehook = sv_dup_inc(proto_perl->Idiehook, param);
13061 PL_warnhook = sv_dup_inc(proto_perl->Iwarnhook, param);
13064 PL_patchlevel = sv_dup_inc(proto_perl->Ipatchlevel, param);
13065 PL_apiversion = sv_dup_inc(proto_perl->Iapiversion, param);
13066 PL_inplace = SAVEPV(proto_perl->Iinplace);
13067 PL_e_script = sv_dup_inc(proto_perl->Ie_script, param);
13069 /* magical thingies */
13070 PL_formfeed = sv_dup(proto_perl->Iformfeed, param);
13072 PL_encoding = sv_dup(proto_perl->Iencoding, param);
13074 sv_setpvs(PERL_DEBUG_PAD(0), ""); /* For regex debugging. */
13075 sv_setpvs(PERL_DEBUG_PAD(1), ""); /* ext/re needs these */
13076 sv_setpvs(PERL_DEBUG_PAD(2), ""); /* even without DEBUGGING. */
13079 /* Clone the regex array */
13080 /* ORANGE FIXME for plugins, probably in the SV dup code.
13081 newSViv(PTR2IV(CALLREGDUPE(
13082 INT2PTR(REGEXP *, SvIVX(regex)), param))))
13084 PL_regex_padav = av_dup_inc(proto_perl->Iregex_padav, param);
13085 PL_regex_pad = AvARRAY(PL_regex_padav);
13087 /* shortcuts to various I/O objects */
13088 PL_ofsgv = gv_dup_inc(proto_perl->Iofsgv, param);
13089 PL_stdingv = gv_dup(proto_perl->Istdingv, param);
13090 PL_stderrgv = gv_dup(proto_perl->Istderrgv, param);
13091 PL_defgv = gv_dup(proto_perl->Idefgv, param);
13092 PL_argvgv = gv_dup(proto_perl->Iargvgv, param);
13093 PL_argvoutgv = gv_dup(proto_perl->Iargvoutgv, param);
13094 PL_argvout_stack = av_dup_inc(proto_perl->Iargvout_stack, param);
13096 /* shortcuts to regexp stuff */
13097 PL_replgv = gv_dup(proto_perl->Ireplgv, param);
13099 /* shortcuts to misc objects */
13100 PL_errgv = gv_dup(proto_perl->Ierrgv, param);
13102 /* shortcuts to debugging objects */
13103 PL_DBgv = gv_dup(proto_perl->IDBgv, param);
13104 PL_DBline = gv_dup(proto_perl->IDBline, param);
13105 PL_DBsub = gv_dup(proto_perl->IDBsub, param);
13106 PL_DBsingle = sv_dup(proto_perl->IDBsingle, param);
13107 PL_DBtrace = sv_dup(proto_perl->IDBtrace, param);
13108 PL_DBsignal = sv_dup(proto_perl->IDBsignal, param);
13110 /* symbol tables */
13111 PL_defstash = hv_dup_inc(proto_perl->Idefstash, param);
13112 PL_curstash = hv_dup(proto_perl->Icurstash, param);
13113 PL_debstash = hv_dup(proto_perl->Idebstash, param);
13114 PL_globalstash = hv_dup(proto_perl->Iglobalstash, param);
13115 PL_curstname = sv_dup_inc(proto_perl->Icurstname, param);
13117 PL_beginav = av_dup_inc(proto_perl->Ibeginav, param);
13118 PL_beginav_save = av_dup_inc(proto_perl->Ibeginav_save, param);
13119 PL_checkav_save = av_dup_inc(proto_perl->Icheckav_save, param);
13120 PL_unitcheckav = av_dup_inc(proto_perl->Iunitcheckav, param);
13121 PL_unitcheckav_save = av_dup_inc(proto_perl->Iunitcheckav_save, param);
13122 PL_endav = av_dup_inc(proto_perl->Iendav, param);
13123 PL_checkav = av_dup_inc(proto_perl->Icheckav, param);
13124 PL_initav = av_dup_inc(proto_perl->Iinitav, param);
13126 PL_isarev = hv_dup_inc(proto_perl->Iisarev, param);
13128 /* subprocess state */
13129 PL_fdpid = av_dup_inc(proto_perl->Ifdpid, param);
13131 if (proto_perl->Iop_mask)
13132 PL_op_mask = SAVEPVN(proto_perl->Iop_mask, PL_maxo);
13135 /* PL_asserting = proto_perl->Iasserting; */
13137 /* current interpreter roots */
13138 PL_main_cv = cv_dup_inc(proto_perl->Imain_cv, param);
13140 PL_main_root = OpREFCNT_inc(proto_perl->Imain_root);
13143 /* runtime control stuff */
13144 PL_curcopdb = (COP*)any_dup(proto_perl->Icurcopdb, proto_perl);
13146 PL_preambleav = av_dup_inc(proto_perl->Ipreambleav, param);
13148 PL_ors_sv = sv_dup_inc(proto_perl->Iors_sv, param);
13150 /* interpreter atexit processing */
13151 PL_exitlistlen = proto_perl->Iexitlistlen;
13152 if (PL_exitlistlen) {
13153 Newx(PL_exitlist, PL_exitlistlen, PerlExitListEntry);
13154 Copy(proto_perl->Iexitlist, PL_exitlist, PL_exitlistlen, PerlExitListEntry);
13157 PL_exitlist = (PerlExitListEntry*)NULL;
13159 PL_my_cxt_size = proto_perl->Imy_cxt_size;
13160 if (PL_my_cxt_size) {
13161 Newx(PL_my_cxt_list, PL_my_cxt_size, void *);
13162 Copy(proto_perl->Imy_cxt_list, PL_my_cxt_list, PL_my_cxt_size, void *);
13163 #ifdef PERL_GLOBAL_STRUCT_PRIVATE
13164 Newx(PL_my_cxt_keys, PL_my_cxt_size, const char *);
13165 Copy(proto_perl->Imy_cxt_keys, PL_my_cxt_keys, PL_my_cxt_size, char *);
13169 PL_my_cxt_list = (void**)NULL;
13170 #ifdef PERL_GLOBAL_STRUCT_PRIVATE
13171 PL_my_cxt_keys = (const char**)NULL;
13174 PL_modglobal = hv_dup_inc(proto_perl->Imodglobal, param);
13175 PL_custom_op_names = hv_dup_inc(proto_perl->Icustom_op_names,param);
13176 PL_custom_op_descs = hv_dup_inc(proto_perl->Icustom_op_descs,param);
13177 PL_custom_ops = hv_dup_inc(proto_perl->Icustom_ops, param);
13179 PL_compcv = cv_dup(proto_perl->Icompcv, param);
13181 PAD_CLONE_VARS(proto_perl, param);
13183 #ifdef HAVE_INTERP_INTERN
13184 sys_intern_dup(&proto_perl->Isys_intern, &PL_sys_intern);
13187 PL_DBcv = cv_dup(proto_perl->IDBcv, param);
13189 #ifdef PERL_USES_PL_PIDSTATUS
13190 PL_pidstatus = newHV(); /* XXX flag for cloning? */
13192 PL_osname = SAVEPV(proto_perl->Iosname);
13193 PL_parser = parser_dup(proto_perl->Iparser, param);
13195 /* XXX this only works if the saved cop has already been cloned */
13196 if (proto_perl->Iparser) {
13197 PL_parser->saved_curcop = (COP*)any_dup(
13198 proto_perl->Iparser->saved_curcop,
13202 PL_subname = sv_dup_inc(proto_perl->Isubname, param);
13204 #ifdef USE_LOCALE_COLLATE
13205 PL_collation_name = SAVEPV(proto_perl->Icollation_name);
13206 #endif /* USE_LOCALE_COLLATE */
13208 #ifdef USE_LOCALE_NUMERIC
13209 PL_numeric_name = SAVEPV(proto_perl->Inumeric_name);
13210 PL_numeric_radix_sv = sv_dup_inc(proto_perl->Inumeric_radix_sv, param);
13211 #endif /* !USE_LOCALE_NUMERIC */
13213 /* utf8 character classes */
13214 PL_utf8_alnum = sv_dup_inc(proto_perl->Iutf8_alnum, param);
13215 PL_utf8_alpha = sv_dup_inc(proto_perl->Iutf8_alpha, param);
13216 PL_utf8_space = sv_dup_inc(proto_perl->Iutf8_space, param);
13217 PL_utf8_graph = sv_dup_inc(proto_perl->Iutf8_graph, param);
13218 PL_utf8_digit = sv_dup_inc(proto_perl->Iutf8_digit, param);
13219 PL_utf8_upper = sv_dup_inc(proto_perl->Iutf8_upper, param);
13220 PL_utf8_lower = sv_dup_inc(proto_perl->Iutf8_lower, param);
13221 PL_utf8_print = sv_dup_inc(proto_perl->Iutf8_print, param);
13222 PL_utf8_punct = sv_dup_inc(proto_perl->Iutf8_punct, param);
13223 PL_utf8_xdigit = sv_dup_inc(proto_perl->Iutf8_xdigit, param);
13224 PL_utf8_mark = sv_dup_inc(proto_perl->Iutf8_mark, param);
13225 PL_utf8_X_begin = sv_dup_inc(proto_perl->Iutf8_X_begin, param);
13226 PL_utf8_X_extend = sv_dup_inc(proto_perl->Iutf8_X_extend, param);
13227 PL_utf8_X_prepend = sv_dup_inc(proto_perl->Iutf8_X_prepend, param);
13228 PL_utf8_X_non_hangul = sv_dup_inc(proto_perl->Iutf8_X_non_hangul, param);
13229 PL_utf8_X_L = sv_dup_inc(proto_perl->Iutf8_X_L, param);
13230 PL_utf8_X_LV = sv_dup_inc(proto_perl->Iutf8_X_LV, param);
13231 PL_utf8_X_LVT = sv_dup_inc(proto_perl->Iutf8_X_LVT, param);
13232 PL_utf8_X_T = sv_dup_inc(proto_perl->Iutf8_X_T, param);
13233 PL_utf8_X_V = sv_dup_inc(proto_perl->Iutf8_X_V, param);
13234 PL_utf8_X_LV_LVT_V = sv_dup_inc(proto_perl->Iutf8_X_LV_LVT_V, param);
13235 PL_utf8_toupper = sv_dup_inc(proto_perl->Iutf8_toupper, param);
13236 PL_utf8_totitle = sv_dup_inc(proto_perl->Iutf8_totitle, param);
13237 PL_utf8_tolower = sv_dup_inc(proto_perl->Iutf8_tolower, param);
13238 PL_utf8_tofold = sv_dup_inc(proto_perl->Iutf8_tofold, param);
13239 PL_utf8_idstart = sv_dup_inc(proto_perl->Iutf8_idstart, param);
13240 PL_utf8_xidstart = sv_dup_inc(proto_perl->Iutf8_xidstart, param);
13241 PL_utf8_perl_idstart = sv_dup_inc(proto_perl->Iutf8_perl_idstart, param);
13242 PL_utf8_idcont = sv_dup_inc(proto_perl->Iutf8_idcont, param);
13243 PL_utf8_xidcont = sv_dup_inc(proto_perl->Iutf8_xidcont, param);
13244 PL_utf8_foldable = sv_dup_inc(proto_perl->Iutf8_foldable, param);
13247 if (proto_perl->Ipsig_pend) {
13248 Newxz(PL_psig_pend, SIG_SIZE, int);
13251 PL_psig_pend = (int*)NULL;
13254 if (proto_perl->Ipsig_name) {
13255 Newx(PL_psig_name, 2 * SIG_SIZE, SV*);
13256 sv_dup_inc_multiple(proto_perl->Ipsig_name, PL_psig_name, 2 * SIG_SIZE,
13258 PL_psig_ptr = PL_psig_name + SIG_SIZE;
13261 PL_psig_ptr = (SV**)NULL;
13262 PL_psig_name = (SV**)NULL;
13265 if (flags & CLONEf_COPY_STACKS) {
13266 Newx(PL_tmps_stack, PL_tmps_max, SV*);
13267 sv_dup_inc_multiple(proto_perl->Itmps_stack, PL_tmps_stack,
13268 PL_tmps_ix+1, param);
13270 /* next PUSHMARK() sets *(PL_markstack_ptr+1) */
13271 i = proto_perl->Imarkstack_max - proto_perl->Imarkstack;
13272 Newxz(PL_markstack, i, I32);
13273 PL_markstack_max = PL_markstack + (proto_perl->Imarkstack_max
13274 - proto_perl->Imarkstack);
13275 PL_markstack_ptr = PL_markstack + (proto_perl->Imarkstack_ptr
13276 - proto_perl->Imarkstack);
13277 Copy(proto_perl->Imarkstack, PL_markstack,
13278 PL_markstack_ptr - PL_markstack + 1, I32);
13280 /* next push_scope()/ENTER sets PL_scopestack[PL_scopestack_ix]
13281 * NOTE: unlike the others! */
13282 Newxz(PL_scopestack, PL_scopestack_max, I32);
13283 Copy(proto_perl->Iscopestack, PL_scopestack, PL_scopestack_ix, I32);
13286 Newxz(PL_scopestack_name, PL_scopestack_max, const char *);
13287 Copy(proto_perl->Iscopestack_name, PL_scopestack_name, PL_scopestack_ix, const char *);
13289 /* NOTE: si_dup() looks at PL_markstack */
13290 PL_curstackinfo = si_dup(proto_perl->Icurstackinfo, param);
13292 /* PL_curstack = PL_curstackinfo->si_stack; */
13293 PL_curstack = av_dup(proto_perl->Icurstack, param);
13294 PL_mainstack = av_dup(proto_perl->Imainstack, param);
13296 /* next PUSHs() etc. set *(PL_stack_sp+1) */
13297 PL_stack_base = AvARRAY(PL_curstack);
13298 PL_stack_sp = PL_stack_base + (proto_perl->Istack_sp
13299 - proto_perl->Istack_base);
13300 PL_stack_max = PL_stack_base + AvMAX(PL_curstack);
13302 /*Newxz(PL_savestack, PL_savestack_max, ANY);*/
13303 PL_savestack = ss_dup(proto_perl, param);
13307 ENTER; /* perl_destruct() wants to LEAVE; */
13310 PL_statgv = gv_dup(proto_perl->Istatgv, param);
13311 PL_statname = sv_dup_inc(proto_perl->Istatname, param);
13313 PL_rs = sv_dup_inc(proto_perl->Irs, param);
13314 PL_last_in_gv = gv_dup(proto_perl->Ilast_in_gv, param);
13315 PL_defoutgv = gv_dup_inc(proto_perl->Idefoutgv, param);
13316 PL_toptarget = sv_dup_inc(proto_perl->Itoptarget, param);
13317 PL_bodytarget = sv_dup_inc(proto_perl->Ibodytarget, param);
13318 PL_formtarget = sv_dup(proto_perl->Iformtarget, param);
13320 PL_errors = sv_dup_inc(proto_perl->Ierrors, param);
13322 PL_sortcop = (OP*)any_dup(proto_perl->Isortcop, proto_perl);
13323 PL_sortstash = hv_dup(proto_perl->Isortstash, param);
13324 PL_firstgv = gv_dup(proto_perl->Ifirstgv, param);
13325 PL_secondgv = gv_dup(proto_perl->Isecondgv, param);
13327 PL_stashcache = newHV();
13329 PL_watchaddr = (char **) ptr_table_fetch(PL_ptr_table,
13330 proto_perl->Iwatchaddr);
13331 PL_watchok = PL_watchaddr ? * PL_watchaddr : NULL;
13332 if (PL_debug && PL_watchaddr) {
13333 PerlIO_printf(Perl_debug_log,
13334 "WATCHING: %"UVxf" cloned as %"UVxf" with value %"UVxf"\n",
13335 PTR2UV(proto_perl->Iwatchaddr), PTR2UV(PL_watchaddr),
13336 PTR2UV(PL_watchok));
13339 PL_registered_mros = hv_dup_inc(proto_perl->Iregistered_mros, param);
13340 PL_blockhooks = av_dup_inc(proto_perl->Iblockhooks, param);
13341 PL_utf8_foldclosures = hv_dup_inc(proto_perl->Iutf8_foldclosures, param);
13343 /* Call the ->CLONE method, if it exists, for each of the stashes
13344 identified by sv_dup() above.
13346 while(av_len(param->stashes) != -1) {
13347 HV* const stash = MUTABLE_HV(av_shift(param->stashes));
13348 GV* const cloner = gv_fetchmethod_autoload(stash, "CLONE", 0);
13349 if (cloner && GvCV(cloner)) {
13354 mXPUSHs(newSVhek(HvNAME_HEK(stash)));
13356 call_sv(MUTABLE_SV(GvCV(cloner)), G_DISCARD);
13362 if (!(flags & CLONEf_KEEP_PTR_TABLE)) {
13363 ptr_table_free(PL_ptr_table);
13364 PL_ptr_table = NULL;
13367 if (!(flags & CLONEf_COPY_STACKS)) {
13368 unreferenced_to_tmp_stack(param->unreferenced);
13371 SvREFCNT_dec(param->stashes);
13373 /* orphaned? eg threads->new inside BEGIN or use */
13374 if (PL_compcv && ! SvREFCNT(PL_compcv)) {
13375 SvREFCNT_inc_simple_void(PL_compcv);
13376 SAVEFREESV(PL_compcv);
13383 S_unreferenced_to_tmp_stack(pTHX_ AV *const unreferenced)
13385 PERL_ARGS_ASSERT_UNREFERENCED_TO_TMP_STACK;
13387 if (AvFILLp(unreferenced) > -1) {
13388 SV **svp = AvARRAY(unreferenced);
13389 SV **const last = svp + AvFILLp(unreferenced);
13393 if (SvREFCNT(*svp) == 1)
13395 } while (++svp <= last);
13397 EXTEND_MORTAL(count);
13398 svp = AvARRAY(unreferenced);
13401 if (SvREFCNT(*svp) == 1) {
13402 /* Our reference is the only one to this SV. This means that
13403 in this thread, the scalar effectively has a 0 reference.
13404 That doesn't work (cleanup never happens), so donate our
13405 reference to it onto the save stack. */
13406 PL_tmps_stack[++PL_tmps_ix] = *svp;
13408 /* As an optimisation, because we are already walking the
13409 entire array, instead of above doing either
13410 SvREFCNT_inc(*svp) or *svp = &PL_sv_undef, we can instead
13411 release our reference to the scalar, so that at the end of
13412 the array owns zero references to the scalars it happens to
13413 point to. We are effectively converting the array from
13414 AvREAL() on to AvREAL() off. This saves the av_clear()
13415 (triggered by the SvREFCNT_dec(unreferenced) below) from
13416 walking the array a second time. */
13417 SvREFCNT_dec(*svp);
13420 } while (++svp <= last);
13421 AvREAL_off(unreferenced);
13423 SvREFCNT_dec(unreferenced);
13427 Perl_clone_params_del(CLONE_PARAMS *param)
13429 /* This seemingly funky ordering keeps the build with PERL_GLOBAL_STRUCT
13431 PerlInterpreter *const to = param->new_perl;
13433 PerlInterpreter *const was = PERL_GET_THX;
13435 PERL_ARGS_ASSERT_CLONE_PARAMS_DEL;
13441 SvREFCNT_dec(param->stashes);
13442 if (param->unreferenced)
13443 unreferenced_to_tmp_stack(param->unreferenced);
13453 Perl_clone_params_new(PerlInterpreter *const from, PerlInterpreter *const to)
13456 /* Need to play this game, as newAV() can call safesysmalloc(), and that
13457 does a dTHX; to get the context from thread local storage.
13458 FIXME - under PERL_CORE Newx(), Safefree() and friends should expand to
13459 a version that passes in my_perl. */
13460 PerlInterpreter *const was = PERL_GET_THX;
13461 CLONE_PARAMS *param;
13463 PERL_ARGS_ASSERT_CLONE_PARAMS_NEW;
13469 /* Given that we've set the context, we can do this unshared. */
13470 Newx(param, 1, CLONE_PARAMS);
13473 param->proto_perl = from;
13474 param->new_perl = to;
13475 param->stashes = (AV *)Perl_newSV_type(to, SVt_PVAV);
13476 AvREAL_off(param->stashes);
13477 param->unreferenced = (AV *)Perl_newSV_type(to, SVt_PVAV);
13485 #endif /* USE_ITHREADS */
13488 =head1 Unicode Support
13490 =for apidoc sv_recode_to_utf8
13492 The encoding is assumed to be an Encode object, on entry the PV
13493 of the sv is assumed to be octets in that encoding, and the sv
13494 will be converted into Unicode (and UTF-8).
13496 If the sv already is UTF-8 (or if it is not POK), or if the encoding
13497 is not a reference, nothing is done to the sv. If the encoding is not
13498 an C<Encode::XS> Encoding object, bad things will happen.
13499 (See F<lib/encoding.pm> and L<Encode>).
13501 The PV of the sv is returned.
13506 Perl_sv_recode_to_utf8(pTHX_ SV *sv, SV *encoding)
13510 PERL_ARGS_ASSERT_SV_RECODE_TO_UTF8;
13512 if (SvPOK(sv) && !SvUTF8(sv) && !IN_BYTES && SvROK(encoding)) {
13526 Passing sv_yes is wrong - it needs to be or'ed set of constants
13527 for Encode::XS, while UTf-8 decode (currently) assumes a true value means
13528 remove converted chars from source.
13530 Both will default the value - let them.
13532 XPUSHs(&PL_sv_yes);
13535 call_method("decode", G_SCALAR);
13539 s = SvPV_const(uni, len);
13540 if (s != SvPVX_const(sv)) {
13541 SvGROW(sv, len + 1);
13542 Move(s, SvPVX(sv), len + 1, char);
13543 SvCUR_set(sv, len);
13547 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
13548 /* clear pos and any utf8 cache */
13549 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
13552 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
13553 magic_setutf8(sv,mg); /* clear UTF8 cache */
13558 return SvPOKp(sv) ? SvPVX(sv) : NULL;
13562 =for apidoc sv_cat_decode
13564 The encoding is assumed to be an Encode object, the PV of the ssv is
13565 assumed to be octets in that encoding and decoding the input starts
13566 from the position which (PV + *offset) pointed to. The dsv will be
13567 concatenated the decoded UTF-8 string from ssv. Decoding will terminate
13568 when the string tstr appears in decoding output or the input ends on
13569 the PV of the ssv. The value which the offset points will be modified
13570 to the last input position on the ssv.
13572 Returns TRUE if the terminator was found, else returns FALSE.
13577 Perl_sv_cat_decode(pTHX_ SV *dsv, SV *encoding,
13578 SV *ssv, int *offset, char *tstr, int tlen)
13583 PERL_ARGS_ASSERT_SV_CAT_DECODE;
13585 if (SvPOK(ssv) && SvPOK(dsv) && SvROK(encoding) && offset) {
13596 offsv = newSViv(*offset);
13598 mXPUSHp(tstr, tlen);
13600 call_method("cat_decode", G_SCALAR);
13602 ret = SvTRUE(TOPs);
13603 *offset = SvIV(offsv);
13609 Perl_croak(aTHX_ "Invalid argument to sv_cat_decode");
13614 /* ---------------------------------------------------------------------
13616 * support functions for report_uninit()
13619 /* the maxiumum size of array or hash where we will scan looking
13620 * for the undefined element that triggered the warning */
13622 #define FUV_MAX_SEARCH_SIZE 1000
13624 /* Look for an entry in the hash whose value has the same SV as val;
13625 * If so, return a mortal copy of the key. */
13628 S_find_hash_subscript(pTHX_ const HV *const hv, const SV *const val)
13631 register HE **array;
13634 PERL_ARGS_ASSERT_FIND_HASH_SUBSCRIPT;
13636 if (!hv || SvMAGICAL(hv) || !HvARRAY(hv) ||
13637 (HvTOTALKEYS(hv) > FUV_MAX_SEARCH_SIZE))
13640 array = HvARRAY(hv);
13642 for (i=HvMAX(hv); i>0; i--) {
13643 register HE *entry;
13644 for (entry = array[i]; entry; entry = HeNEXT(entry)) {
13645 if (HeVAL(entry) != val)
13647 if ( HeVAL(entry) == &PL_sv_undef ||
13648 HeVAL(entry) == &PL_sv_placeholder)
13652 if (HeKLEN(entry) == HEf_SVKEY)
13653 return sv_mortalcopy(HeKEY_sv(entry));
13654 return sv_2mortal(newSVhek(HeKEY_hek(entry)));
13660 /* Look for an entry in the array whose value has the same SV as val;
13661 * If so, return the index, otherwise return -1. */
13664 S_find_array_subscript(pTHX_ const AV *const av, const SV *const val)
13668 PERL_ARGS_ASSERT_FIND_ARRAY_SUBSCRIPT;
13670 if (!av || SvMAGICAL(av) || !AvARRAY(av) ||
13671 (AvFILLp(av) > FUV_MAX_SEARCH_SIZE))
13674 if (val != &PL_sv_undef) {
13675 SV ** const svp = AvARRAY(av);
13678 for (i=AvFILLp(av); i>=0; i--)
13685 /* S_varname(): return the name of a variable, optionally with a subscript.
13686 * If gv is non-zero, use the name of that global, along with gvtype (one
13687 * of "$", "@", "%"); otherwise use the name of the lexical at pad offset
13688 * targ. Depending on the value of the subscript_type flag, return:
13691 #define FUV_SUBSCRIPT_NONE 1 /* "@foo" */
13692 #define FUV_SUBSCRIPT_ARRAY 2 /* "$foo[aindex]" */
13693 #define FUV_SUBSCRIPT_HASH 3 /* "$foo{keyname}" */
13694 #define FUV_SUBSCRIPT_WITHIN 4 /* "within @foo" */
13697 S_varname(pTHX_ const GV *const gv, const char gvtype, PADOFFSET targ,
13698 const SV *const keyname, I32 aindex, int subscript_type)
13701 SV * const name = sv_newmortal();
13704 buffer[0] = gvtype;
13707 /* as gv_fullname4(), but add literal '^' for $^FOO names */
13709 gv_fullname4(name, gv, buffer, 0);
13711 if ((unsigned int)SvPVX(name)[1] <= 26) {
13713 buffer[1] = SvPVX(name)[1] + 'A' - 1;
13715 /* Swap the 1 unprintable control character for the 2 byte pretty
13716 version - ie substr($name, 1, 1) = $buffer; */
13717 sv_insert(name, 1, 1, buffer, 2);
13721 CV * const cv = find_runcv(NULL);
13725 if (!cv || !CvPADLIST(cv))
13727 av = MUTABLE_AV((*av_fetch(CvPADLIST(cv), 0, FALSE)));
13728 sv = *av_fetch(av, targ, FALSE);
13729 sv_setpvn(name, SvPV_nolen_const(sv), SvCUR(sv));
13732 if (subscript_type == FUV_SUBSCRIPT_HASH) {
13733 SV * const sv = newSV(0);
13734 *SvPVX(name) = '$';
13735 Perl_sv_catpvf(aTHX_ name, "{%s}",
13736 pv_display(sv,SvPVX_const(keyname), SvCUR(keyname), 0, 32));
13739 else if (subscript_type == FUV_SUBSCRIPT_ARRAY) {
13740 *SvPVX(name) = '$';
13741 Perl_sv_catpvf(aTHX_ name, "[%"IVdf"]", (IV)aindex);
13743 else if (subscript_type == FUV_SUBSCRIPT_WITHIN) {
13744 /* We know that name has no magic, so can use 0 instead of SV_GMAGIC */
13745 Perl_sv_insert_flags(aTHX_ name, 0, 0, STR_WITH_LEN("within "), 0);
13753 =for apidoc find_uninit_var
13755 Find the name of the undefined variable (if any) that caused the operator o
13756 to issue a "Use of uninitialized value" warning.
13757 If match is true, only return a name if it's value matches uninit_sv.
13758 So roughly speaking, if a unary operator (such as OP_COS) generates a
13759 warning, then following the direct child of the op may yield an
13760 OP_PADSV or OP_GV that gives the name of the undefined variable. On the
13761 other hand, with OP_ADD there are two branches to follow, so we only print
13762 the variable name if we get an exact match.
13764 The name is returned as a mortal SV.
13766 Assumes that PL_op is the op that originally triggered the error, and that
13767 PL_comppad/PL_curpad points to the currently executing pad.
13773 S_find_uninit_var(pTHX_ const OP *const obase, const SV *const uninit_sv,
13779 const OP *o, *o2, *kid;
13781 if (!obase || (match && (!uninit_sv || uninit_sv == &PL_sv_undef ||
13782 uninit_sv == &PL_sv_placeholder)))
13785 switch (obase->op_type) {
13792 const bool pad = (obase->op_type == OP_PADAV || obase->op_type == OP_PADHV);
13793 const bool hash = (obase->op_type == OP_PADHV || obase->op_type == OP_RV2HV);
13796 int subscript_type = FUV_SUBSCRIPT_WITHIN;
13798 if (pad) { /* @lex, %lex */
13799 sv = PAD_SVl(obase->op_targ);
13803 if (cUNOPx(obase)->op_first->op_type == OP_GV) {
13804 /* @global, %global */
13805 gv = cGVOPx_gv(cUNOPx(obase)->op_first);
13808 sv = hash ? MUTABLE_SV(GvHV(gv)): MUTABLE_SV(GvAV(gv));
13810 else /* @{expr}, %{expr} */
13811 return find_uninit_var(cUNOPx(obase)->op_first,
13815 /* attempt to find a match within the aggregate */
13817 keysv = find_hash_subscript((const HV*)sv, uninit_sv);
13819 subscript_type = FUV_SUBSCRIPT_HASH;
13822 index = find_array_subscript((const AV *)sv, uninit_sv);
13824 subscript_type = FUV_SUBSCRIPT_ARRAY;
13827 if (match && subscript_type == FUV_SUBSCRIPT_WITHIN)
13830 return varname(gv, hash ? '%' : '@', obase->op_targ,
13831 keysv, index, subscript_type);
13835 if (cUNOPx(obase)->op_first->op_type == OP_GV) {
13837 gv = cGVOPx_gv(cUNOPx(obase)->op_first);
13838 if (!gv || !GvSTASH(gv))
13840 if (match && (GvSV(gv) != uninit_sv))
13842 return varname(gv, '$', 0, NULL, 0, FUV_SUBSCRIPT_NONE);
13845 return find_uninit_var(cUNOPx(obase)->op_first, uninit_sv, 1);
13848 if (match && PAD_SVl(obase->op_targ) != uninit_sv)
13850 return varname(NULL, '$', obase->op_targ,
13851 NULL, 0, FUV_SUBSCRIPT_NONE);
13854 gv = cGVOPx_gv(obase);
13855 if (!gv || (match && GvSV(gv) != uninit_sv) || !GvSTASH(gv))
13857 return varname(gv, '$', 0, NULL, 0, FUV_SUBSCRIPT_NONE);
13859 case OP_AELEMFAST_LEX:
13862 AV *av = MUTABLE_AV(PAD_SV(obase->op_targ));
13863 if (!av || SvRMAGICAL(av))
13865 svp = av_fetch(av, (I32)obase->op_private, FALSE);
13866 if (!svp || *svp != uninit_sv)
13869 return varname(NULL, '$', obase->op_targ,
13870 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
13873 gv = cGVOPx_gv(obase);
13878 AV *const av = GvAV(gv);
13879 if (!av || SvRMAGICAL(av))
13881 svp = av_fetch(av, (I32)obase->op_private, FALSE);
13882 if (!svp || *svp != uninit_sv)
13885 return varname(gv, '$', 0,
13886 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
13891 o = cUNOPx(obase)->op_first;
13892 if (!o || o->op_type != OP_NULL ||
13893 ! (o->op_targ == OP_AELEM || o->op_targ == OP_HELEM))
13895 return find_uninit_var(cBINOPo->op_last, uninit_sv, match);
13900 bool negate = FALSE;
13902 if (PL_op == obase)
13903 /* $a[uninit_expr] or $h{uninit_expr} */
13904 return find_uninit_var(cBINOPx(obase)->op_last, uninit_sv, match);
13907 o = cBINOPx(obase)->op_first;
13908 kid = cBINOPx(obase)->op_last;
13910 /* get the av or hv, and optionally the gv */
13912 if (o->op_type == OP_PADAV || o->op_type == OP_PADHV) {
13913 sv = PAD_SV(o->op_targ);
13915 else if ((o->op_type == OP_RV2AV || o->op_type == OP_RV2HV)
13916 && cUNOPo->op_first->op_type == OP_GV)
13918 gv = cGVOPx_gv(cUNOPo->op_first);
13922 == OP_RV2HV ? MUTABLE_SV(GvHV(gv)) : MUTABLE_SV(GvAV(gv));
13927 if (kid && kid->op_type == OP_NEGATE) {
13929 kid = cUNOPx(kid)->op_first;
13932 if (kid && kid->op_type == OP_CONST && SvOK(cSVOPx_sv(kid))) {
13933 /* index is constant */
13936 kidsv = sv_2mortal(newSVpvs("-"));
13937 sv_catsv(kidsv, cSVOPx_sv(kid));
13940 kidsv = cSVOPx_sv(kid);
13944 if (obase->op_type == OP_HELEM) {
13945 HE* he = hv_fetch_ent(MUTABLE_HV(sv), kidsv, 0, 0);
13946 if (!he || HeVAL(he) != uninit_sv)
13950 SV * const * const svp = av_fetch(MUTABLE_AV(sv),
13951 negate ? - SvIV(cSVOPx_sv(kid)) : SvIV(cSVOPx_sv(kid)),
13953 if (!svp || *svp != uninit_sv)
13957 if (obase->op_type == OP_HELEM)
13958 return varname(gv, '%', o->op_targ,
13959 kidsv, 0, FUV_SUBSCRIPT_HASH);
13961 return varname(gv, '@', o->op_targ, NULL,
13962 negate ? - SvIV(cSVOPx_sv(kid)) : SvIV(cSVOPx_sv(kid)),
13963 FUV_SUBSCRIPT_ARRAY);
13966 /* index is an expression;
13967 * attempt to find a match within the aggregate */
13968 if (obase->op_type == OP_HELEM) {
13969 SV * const keysv = find_hash_subscript((const HV*)sv, uninit_sv);
13971 return varname(gv, '%', o->op_targ,
13972 keysv, 0, FUV_SUBSCRIPT_HASH);
13976 = find_array_subscript((const AV *)sv, uninit_sv);
13978 return varname(gv, '@', o->op_targ,
13979 NULL, index, FUV_SUBSCRIPT_ARRAY);
13984 (o->op_type == OP_PADAV || o->op_type == OP_RV2AV)
13986 o->op_targ, NULL, 0, FUV_SUBSCRIPT_WITHIN);
13992 /* only examine RHS */
13993 return find_uninit_var(cBINOPx(obase)->op_first, uninit_sv, match);
13996 o = cUNOPx(obase)->op_first;
13997 if (o->op_type == OP_PUSHMARK)
14000 if (!o->op_sibling) {
14001 /* one-arg version of open is highly magical */
14003 if (o->op_type == OP_GV) { /* open FOO; */
14005 if (match && GvSV(gv) != uninit_sv)
14007 return varname(gv, '$', 0,
14008 NULL, 0, FUV_SUBSCRIPT_NONE);
14010 /* other possibilities not handled are:
14011 * open $x; or open my $x; should return '${*$x}'
14012 * open expr; should return '$'.expr ideally
14018 /* ops where $_ may be an implicit arg */
14022 if ( !(obase->op_flags & OPf_STACKED)) {
14023 if (uninit_sv == ((obase->op_private & OPpTARGET_MY)
14024 ? PAD_SVl(obase->op_targ)
14027 sv = sv_newmortal();
14028 sv_setpvs(sv, "$_");
14037 match = 1; /* print etc can return undef on defined args */
14038 /* skip filehandle as it can't produce 'undef' warning */
14039 o = cUNOPx(obase)->op_first;
14040 if ((obase->op_flags & OPf_STACKED) && o->op_type == OP_PUSHMARK)
14041 o = o->op_sibling->op_sibling;
14045 case OP_ENTEREVAL: /* could be eval $undef or $x='$undef'; eval $x */
14046 case OP_CUSTOM: /* XS or custom code could trigger random warnings */
14048 /* the following ops are capable of returning PL_sv_undef even for
14049 * defined arg(s) */
14068 case OP_GETPEERNAME:
14116 case OP_SMARTMATCH:
14125 /* XXX tmp hack: these two may call an XS sub, and currently
14126 XS subs don't have a SUB entry on the context stack, so CV and
14127 pad determination goes wrong, and BAD things happen. So, just
14128 don't try to determine the value under those circumstances.
14129 Need a better fix at dome point. DAPM 11/2007 */
14135 GV * const gv = gv_fetchpvs(".", GV_NOTQUAL, SVt_PV);
14136 if (gv && GvSV(gv) == uninit_sv)
14137 return newSVpvs_flags("$.", SVs_TEMP);
14142 /* def-ness of rval pos() is independent of the def-ness of its arg */
14143 if ( !(obase->op_flags & OPf_MOD))
14148 if (SvROK(PL_rs) && uninit_sv == SvRV(PL_rs))
14149 return newSVpvs_flags("${$/}", SVs_TEMP);
14154 if (!(obase->op_flags & OPf_KIDS))
14156 o = cUNOPx(obase)->op_first;
14162 /* if all except one arg are constant, or have no side-effects,
14163 * or are optimized away, then it's unambiguous */
14165 for (kid=o; kid; kid = kid->op_sibling) {
14167 const OPCODE type = kid->op_type;
14168 if ( (type == OP_CONST && SvOK(cSVOPx_sv(kid)))
14169 || (type == OP_NULL && ! (kid->op_flags & OPf_KIDS))
14170 || (type == OP_PUSHMARK)
14172 /* @$a and %$a, but not @a or %a */
14173 (type == OP_RV2AV || type == OP_RV2HV)
14174 && cUNOPx(kid)->op_first
14175 && cUNOPx(kid)->op_first->op_type != OP_GV
14180 if (o2) { /* more than one found */
14187 return find_uninit_var(o2, uninit_sv, match);
14189 /* scan all args */
14191 sv = find_uninit_var(o, uninit_sv, 1);
14203 =for apidoc report_uninit
14205 Print appropriate "Use of uninitialized variable" warning
14211 Perl_report_uninit(pTHX_ const SV *uninit_sv)
14215 SV* varname = NULL;
14217 varname = find_uninit_var(PL_op, uninit_sv,0);
14219 sv_insert(varname, 0, 0, " ", 1);
14221 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
14222 varname ? SvPV_nolen_const(varname) : "",
14223 " in ", OP_DESC(PL_op));
14226 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
14232 * c-indentation-style: bsd
14233 * c-basic-offset: 4
14234 * indent-tabs-mode: t
14237 * ex: set ts=8 sts=4 sw=4 noet: