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) != 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) != 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 /* 8 bytes on most ILP32 with IEEE doubles */
897 { sizeof(NV), sizeof(NV),
898 STRUCT_OFFSET(XPVNV, xnv_u),
899 SVt_NV, FALSE, HADNV, HASARENA, FIT_ARENA(0, sizeof(NV)) },
901 /* 8 bytes on most ILP32 with IEEE doubles */
902 { sizeof(XPV) - STRUCT_OFFSET(XPV, xpv_cur),
903 copy_length(XPV, xpv_len) - STRUCT_OFFSET(XPV, xpv_cur),
904 + STRUCT_OFFSET(XPV, xpv_cur),
905 SVt_PV, FALSE, NONV, HASARENA,
906 FIT_ARENA(0, sizeof(XPV) - STRUCT_OFFSET(XPV, xpv_cur)) },
909 { sizeof(XPVIV) - STRUCT_OFFSET(XPV, xpv_cur),
910 copy_length(XPVIV, xiv_u) - STRUCT_OFFSET(XPV, xpv_cur),
911 + STRUCT_OFFSET(XPV, xpv_cur),
912 SVt_PVIV, FALSE, NONV, HASARENA,
913 FIT_ARENA(0, sizeof(XPVIV) - STRUCT_OFFSET(XPV, xpv_cur)) },
916 { sizeof(XPVNV) - STRUCT_OFFSET(XPV, xpv_cur),
917 copy_length(XPVNV, xnv_u) - STRUCT_OFFSET(XPV, xpv_cur),
918 + STRUCT_OFFSET(XPV, xpv_cur),
919 SVt_PVNV, FALSE, HADNV, HASARENA,
920 FIT_ARENA(0, sizeof(XPVNV) - STRUCT_OFFSET(XPV, xpv_cur)) },
923 { sizeof(XPVMG), copy_length(XPVMG, xnv_u), 0, SVt_PVMG, FALSE, HADNV,
924 HASARENA, FIT_ARENA(0, sizeof(XPVMG)) },
930 SVt_REGEXP, FALSE, NONV, HASARENA,
931 FIT_ARENA(0, sizeof(regexp))
935 { sizeof(XPVGV), sizeof(XPVGV), 0, SVt_PVGV, TRUE, HADNV,
936 HASARENA, FIT_ARENA(0, sizeof(XPVGV)) },
939 { sizeof(XPVLV), sizeof(XPVLV), 0, SVt_PVLV, TRUE, HADNV,
940 HASARENA, FIT_ARENA(0, sizeof(XPVLV)) },
943 copy_length(XPVAV, xav_alloc),
945 SVt_PVAV, TRUE, NONV, HASARENA,
946 FIT_ARENA(0, sizeof(XPVAV)) },
949 copy_length(XPVHV, xhv_max),
951 SVt_PVHV, TRUE, NONV, HASARENA,
952 FIT_ARENA(0, sizeof(XPVHV)) },
958 SVt_PVCV, TRUE, NONV, HASARENA,
959 FIT_ARENA(0, sizeof(XPVCV)) },
964 SVt_PVFM, TRUE, NONV, NOARENA,
965 FIT_ARENA(20, sizeof(XPVFM)) },
967 /* XPVIO is 84 bytes, fits 48x */
971 SVt_PVIO, TRUE, NONV, HASARENA,
972 FIT_ARENA(24, sizeof(XPVIO)) },
975 #define new_body_allocated(sv_type) \
976 (void *)((char *)S_new_body(aTHX_ sv_type) \
977 - bodies_by_type[sv_type].offset)
979 /* return a thing to the free list */
981 #define del_body(thing, root) \
983 void ** const thing_copy = (void **)thing; \
984 *thing_copy = *root; \
985 *root = (void*)thing_copy; \
990 #define new_XNV() safemalloc(sizeof(XPVNV))
991 #define new_XPVNV() safemalloc(sizeof(XPVNV))
992 #define new_XPVMG() safemalloc(sizeof(XPVMG))
994 #define del_XPVGV(p) safefree(p)
998 #define new_XNV() new_body_allocated(SVt_NV)
999 #define new_XPVNV() new_body_allocated(SVt_PVNV)
1000 #define new_XPVMG() new_body_allocated(SVt_PVMG)
1002 #define del_XPVGV(p) del_body(p + bodies_by_type[SVt_PVGV].offset, \
1003 &PL_body_roots[SVt_PVGV])
1007 /* no arena for you! */
1009 #define new_NOARENA(details) \
1010 safemalloc((details)->body_size + (details)->offset)
1011 #define new_NOARENAZ(details) \
1012 safecalloc((details)->body_size + (details)->offset, 1)
1015 Perl_more_bodies (pTHX_ const svtype sv_type, const size_t body_size,
1016 const size_t arena_size)
1019 void ** const root = &PL_body_roots[sv_type];
1020 struct arena_desc *adesc;
1021 struct arena_set *aroot = (struct arena_set *) PL_body_arenas;
1025 const size_t good_arena_size = Perl_malloc_good_size(arena_size);
1026 #if defined(DEBUGGING) && !defined(PERL_GLOBAL_STRUCT_PRIVATE)
1027 static bool done_sanity_check;
1029 /* PERL_GLOBAL_STRUCT_PRIVATE cannot coexist with global
1030 * variables like done_sanity_check. */
1031 if (!done_sanity_check) {
1032 unsigned int i = SVt_LAST;
1034 done_sanity_check = TRUE;
1037 assert (bodies_by_type[i].type == i);
1043 /* may need new arena-set to hold new arena */
1044 if (!aroot || aroot->curr >= aroot->set_size) {
1045 struct arena_set *newroot;
1046 Newxz(newroot, 1, struct arena_set);
1047 newroot->set_size = ARENAS_PER_SET;
1048 newroot->next = aroot;
1050 PL_body_arenas = (void *) newroot;
1051 DEBUG_m(PerlIO_printf(Perl_debug_log, "new arenaset %p\n", (void*)aroot));
1054 /* ok, now have arena-set with at least 1 empty/available arena-desc */
1055 curr = aroot->curr++;
1056 adesc = &(aroot->set[curr]);
1057 assert(!adesc->arena);
1059 Newx(adesc->arena, good_arena_size, char);
1060 adesc->size = good_arena_size;
1061 adesc->utype = sv_type;
1062 DEBUG_m(PerlIO_printf(Perl_debug_log, "arena %d added: %p size %"UVuf"\n",
1063 curr, (void*)adesc->arena, (UV)good_arena_size));
1065 start = (char *) adesc->arena;
1067 /* Get the address of the byte after the end of the last body we can fit.
1068 Remember, this is integer division: */
1069 end = start + good_arena_size / body_size * body_size;
1071 /* computed count doesn't reflect the 1st slot reservation */
1072 #if defined(MYMALLOC) || defined(HAS_MALLOC_GOOD_SIZE)
1073 DEBUG_m(PerlIO_printf(Perl_debug_log,
1074 "arena %p end %p arena-size %d (from %d) type %d "
1076 (void*)start, (void*)end, (int)good_arena_size,
1077 (int)arena_size, sv_type, (int)body_size,
1078 (int)good_arena_size / (int)body_size));
1080 DEBUG_m(PerlIO_printf(Perl_debug_log,
1081 "arena %p end %p arena-size %d type %d size %d ct %d\n",
1082 (void*)start, (void*)end,
1083 (int)arena_size, sv_type, (int)body_size,
1084 (int)good_arena_size / (int)body_size));
1086 *root = (void *)start;
1089 /* Where the next body would start: */
1090 char * const next = start + body_size;
1093 /* This is the last body: */
1094 assert(next == end);
1096 *(void **)start = 0;
1100 *(void**) start = (void *)next;
1105 /* grab a new thing from the free list, allocating more if necessary.
1106 The inline version is used for speed in hot routines, and the
1107 function using it serves the rest (unless PURIFY).
1109 #define new_body_inline(xpv, sv_type) \
1111 void ** const r3wt = &PL_body_roots[sv_type]; \
1112 xpv = (PTR_TBL_ENT_t*) (*((void **)(r3wt)) \
1113 ? *((void **)(r3wt)) : Perl_more_bodies(aTHX_ sv_type, \
1114 bodies_by_type[sv_type].body_size,\
1115 bodies_by_type[sv_type].arena_size)); \
1116 *(r3wt) = *(void**)(xpv); \
1122 S_new_body(pTHX_ const svtype sv_type)
1126 new_body_inline(xpv, sv_type);
1132 static const struct body_details fake_rv =
1133 { 0, 0, 0, SVt_IV, FALSE, NONV, NOARENA, 0 };
1136 =for apidoc sv_upgrade
1138 Upgrade an SV to a more complex form. Generally adds a new body type to the
1139 SV, then copies across as much information as possible from the old body.
1140 You generally want to use the C<SvUPGRADE> macro wrapper. See also C<svtype>.
1146 Perl_sv_upgrade(pTHX_ register SV *const sv, svtype new_type)
1151 const svtype old_type = SvTYPE(sv);
1152 const struct body_details *new_type_details;
1153 const struct body_details *old_type_details
1154 = bodies_by_type + old_type;
1155 SV *referant = NULL;
1157 PERL_ARGS_ASSERT_SV_UPGRADE;
1159 if (old_type == new_type)
1162 /* This clause was purposefully added ahead of the early return above to
1163 the shared string hackery for (sort {$a <=> $b} keys %hash), with the
1164 inference by Nick I-S that it would fix other troublesome cases. See
1165 changes 7162, 7163 (f130fd4589cf5fbb24149cd4db4137c8326f49c1 and parent)
1167 Given that shared hash key scalars are no longer PVIV, but PV, there is
1168 no longer need to unshare so as to free up the IVX slot for its proper
1169 purpose. So it's safe to move the early return earlier. */
1171 if (new_type != SVt_PV && SvIsCOW(sv)) {
1172 sv_force_normal_flags(sv, 0);
1175 old_body = SvANY(sv);
1177 /* Copying structures onto other structures that have been neatly zeroed
1178 has a subtle gotcha. Consider XPVMG
1180 +------+------+------+------+------+-------+-------+
1181 | NV | CUR | LEN | IV | MAGIC | STASH |
1182 +------+------+------+------+------+-------+-------+
1183 0 4 8 12 16 20 24 28
1185 where NVs are aligned to 8 bytes, so that sizeof that structure is
1186 actually 32 bytes long, with 4 bytes of padding at the end:
1188 +------+------+------+------+------+-------+-------+------+
1189 | NV | CUR | LEN | IV | MAGIC | STASH | ??? |
1190 +------+------+------+------+------+-------+-------+------+
1191 0 4 8 12 16 20 24 28 32
1193 so what happens if you allocate memory for this structure:
1195 +------+------+------+------+------+-------+-------+------+------+...
1196 | NV | CUR | LEN | IV | MAGIC | STASH | GP | NAME |
1197 +------+------+------+------+------+-------+-------+------+------+...
1198 0 4 8 12 16 20 24 28 32 36
1200 zero it, then copy sizeof(XPVMG) bytes on top of it? Not quite what you
1201 expect, because you copy the area marked ??? onto GP. Now, ??? may have
1202 started out as zero once, but it's quite possible that it isn't. So now,
1203 rather than a nicely zeroed GP, you have it pointing somewhere random.
1206 (In fact, GP ends up pointing at a previous GP structure, because the
1207 principle cause of the padding in XPVMG getting garbage is a copy of
1208 sizeof(XPVMG) bytes from a XPVGV structure in sv_unglob. Right now
1209 this happens to be moot because XPVGV has been re-ordered, with GP
1210 no longer after STASH)
1212 So we are careful and work out the size of used parts of all the
1220 referant = SvRV(sv);
1221 old_type_details = &fake_rv;
1222 if (new_type == SVt_NV)
1223 new_type = SVt_PVNV;
1225 if (new_type < SVt_PVIV) {
1226 new_type = (new_type == SVt_NV)
1227 ? SVt_PVNV : SVt_PVIV;
1232 if (new_type < SVt_PVNV) {
1233 new_type = SVt_PVNV;
1237 assert(new_type > SVt_PV);
1238 assert(SVt_IV < SVt_PV);
1239 assert(SVt_NV < SVt_PV);
1246 /* Because the XPVMG of PL_mess_sv isn't allocated from the arena,
1247 there's no way that it can be safely upgraded, because perl.c
1248 expects to Safefree(SvANY(PL_mess_sv)) */
1249 assert(sv != PL_mess_sv);
1250 /* This flag bit is used to mean other things in other scalar types.
1251 Given that it only has meaning inside the pad, it shouldn't be set
1252 on anything that can get upgraded. */
1253 assert(!SvPAD_TYPED(sv));
1256 if (old_type_details->cant_upgrade)
1257 Perl_croak(aTHX_ "Can't upgrade %s (%" UVuf ") to %" UVuf,
1258 sv_reftype(sv, 0), (UV) old_type, (UV) new_type);
1261 if (old_type > new_type)
1262 Perl_croak(aTHX_ "sv_upgrade from type %d down to type %d",
1263 (int)old_type, (int)new_type);
1265 new_type_details = bodies_by_type + new_type;
1267 SvFLAGS(sv) &= ~SVTYPEMASK;
1268 SvFLAGS(sv) |= new_type;
1270 /* This can't happen, as SVt_NULL is <= all values of new_type, so one of
1271 the return statements above will have triggered. */
1272 assert (new_type != SVt_NULL);
1275 assert(old_type == SVt_NULL);
1276 SvANY(sv) = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
1280 assert(old_type == SVt_NULL);
1281 SvANY(sv) = new_XNV();
1286 assert(new_type_details->body_size);
1289 assert(new_type_details->arena);
1290 assert(new_type_details->arena_size);
1291 /* This points to the start of the allocated area. */
1292 new_body_inline(new_body, new_type);
1293 Zero(new_body, new_type_details->body_size, char);
1294 new_body = ((char *)new_body) - new_type_details->offset;
1296 /* We always allocated the full length item with PURIFY. To do this
1297 we fake things so that arena is false for all 16 types.. */
1298 new_body = new_NOARENAZ(new_type_details);
1300 SvANY(sv) = new_body;
1301 if (new_type == SVt_PVAV) {
1305 if (old_type_details->body_size) {
1308 /* It will have been zeroed when the new body was allocated.
1309 Lets not write to it, in case it confuses a write-back
1315 #ifndef NODEFAULT_SHAREKEYS
1316 HvSHAREKEYS_on(sv); /* key-sharing on by default */
1318 HvMAX(sv) = 7; /* (start with 8 buckets) */
1321 /* SVt_NULL isn't the only thing upgraded to AV or HV.
1322 The target created by newSVrv also is, and it can have magic.
1323 However, it never has SvPVX set.
1325 if (old_type == SVt_IV) {
1327 } else if (old_type >= SVt_PV) {
1328 assert(SvPVX_const(sv) == 0);
1331 if (old_type >= SVt_PVMG) {
1332 SvMAGIC_set(sv, ((XPVMG*)old_body)->xmg_u.xmg_magic);
1333 SvSTASH_set(sv, ((XPVMG*)old_body)->xmg_stash);
1335 sv->sv_u.svu_array = NULL; /* or svu_hash */
1341 /* This ensures that SvTHINKFIRST(sv) is true, and hence that
1342 sv_force_normal_flags(sv) is called. */
1345 /* XXX Is this still needed? Was it ever needed? Surely as there is
1346 no route from NV to PVIV, NOK can never be true */
1347 assert(!SvNOKp(sv));
1358 assert(new_type_details->body_size);
1359 /* We always allocated the full length item with PURIFY. To do this
1360 we fake things so that arena is false for all 16 types.. */
1361 if(new_type_details->arena) {
1362 /* This points to the start of the allocated area. */
1363 new_body_inline(new_body, new_type);
1364 Zero(new_body, new_type_details->body_size, char);
1365 new_body = ((char *)new_body) - new_type_details->offset;
1367 new_body = new_NOARENAZ(new_type_details);
1369 SvANY(sv) = new_body;
1371 if (old_type_details->copy) {
1372 /* There is now the potential for an upgrade from something without
1373 an offset (PVNV or PVMG) to something with one (PVCV, PVFM) */
1374 int offset = old_type_details->offset;
1375 int length = old_type_details->copy;
1377 if (new_type_details->offset > old_type_details->offset) {
1378 const int difference
1379 = new_type_details->offset - old_type_details->offset;
1380 offset += difference;
1381 length -= difference;
1383 assert (length >= 0);
1385 Copy((char *)old_body + offset, (char *)new_body + offset, length,
1389 #ifndef NV_ZERO_IS_ALLBITS_ZERO
1390 /* If NV 0.0 is stores as all bits 0 then Zero() already creates a
1391 * correct 0.0 for us. Otherwise, if the old body didn't have an
1392 * NV slot, but the new one does, then we need to initialise the
1393 * freshly created NV slot with whatever the correct bit pattern is
1395 if (old_type_details->zero_nv && !new_type_details->zero_nv
1396 && !isGV_with_GP(sv))
1400 if (new_type == SVt_PVIO) {
1401 IO * const io = MUTABLE_IO(sv);
1402 GV *iogv = gv_fetchpvs("IO::File::", GV_ADD, SVt_PVHV);
1405 /* Clear the stashcache because a new IO could overrule a package
1407 hv_clear(PL_stashcache);
1409 SvSTASH_set(io, MUTABLE_HV(SvREFCNT_inc(GvHV(iogv))));
1410 IoPAGE_LEN(sv) = 60;
1412 if (old_type < SVt_PV) {
1413 /* referant will be NULL unless the old type was SVt_IV emulating
1415 sv->sv_u.svu_rv = referant;
1419 Perl_croak(aTHX_ "panic: sv_upgrade to unknown type %lu",
1420 (unsigned long)new_type);
1423 if (old_type > SVt_IV) {
1427 /* Note that there is an assumption that all bodies of types that
1428 can be upgraded came from arenas. Only the more complex non-
1429 upgradable types are allowed to be directly malloc()ed. */
1430 assert(old_type_details->arena);
1431 del_body((void*)((char*)old_body + old_type_details->offset),
1432 &PL_body_roots[old_type]);
1438 =for apidoc sv_backoff
1440 Remove any string offset. You should normally use the C<SvOOK_off> macro
1447 Perl_sv_backoff(pTHX_ register SV *const sv)
1450 const char * const s = SvPVX_const(sv);
1452 PERL_ARGS_ASSERT_SV_BACKOFF;
1453 PERL_UNUSED_CONTEXT;
1456 assert(SvTYPE(sv) != SVt_PVHV);
1457 assert(SvTYPE(sv) != SVt_PVAV);
1459 SvOOK_offset(sv, delta);
1461 SvLEN_set(sv, SvLEN(sv) + delta);
1462 SvPV_set(sv, SvPVX(sv) - delta);
1463 Move(s, SvPVX(sv), SvCUR(sv)+1, char);
1464 SvFLAGS(sv) &= ~SVf_OOK;
1471 Expands the character buffer in the SV. If necessary, uses C<sv_unref> and
1472 upgrades the SV to C<SVt_PV>. Returns a pointer to the character buffer.
1473 Use the C<SvGROW> wrapper instead.
1479 Perl_sv_grow(pTHX_ register SV *const sv, register STRLEN newlen)
1483 PERL_ARGS_ASSERT_SV_GROW;
1485 if (PL_madskills && newlen >= 0x100000) {
1486 PerlIO_printf(Perl_debug_log,
1487 "Allocation too large: %"UVxf"\n", (UV)newlen);
1489 #ifdef HAS_64K_LIMIT
1490 if (newlen >= 0x10000) {
1491 PerlIO_printf(Perl_debug_log,
1492 "Allocation too large: %"UVxf"\n", (UV)newlen);
1495 #endif /* HAS_64K_LIMIT */
1498 if (SvTYPE(sv) < SVt_PV) {
1499 sv_upgrade(sv, SVt_PV);
1500 s = SvPVX_mutable(sv);
1502 else if (SvOOK(sv)) { /* pv is offset? */
1504 s = SvPVX_mutable(sv);
1505 if (newlen > SvLEN(sv))
1506 newlen += 10 * (newlen - SvCUR(sv)); /* avoid copy each time */
1507 #ifdef HAS_64K_LIMIT
1508 if (newlen >= 0x10000)
1513 s = SvPVX_mutable(sv);
1515 if (newlen > SvLEN(sv)) { /* need more room? */
1516 STRLEN minlen = SvCUR(sv);
1517 minlen += (minlen >> PERL_STRLEN_EXPAND_SHIFT) + 10;
1518 if (newlen < minlen)
1520 #ifndef Perl_safesysmalloc_size
1521 newlen = PERL_STRLEN_ROUNDUP(newlen);
1523 if (SvLEN(sv) && s) {
1524 s = (char*)saferealloc(s, newlen);
1527 s = (char*)safemalloc(newlen);
1528 if (SvPVX_const(sv) && SvCUR(sv)) {
1529 Move(SvPVX_const(sv), s, (newlen < SvCUR(sv)) ? newlen : SvCUR(sv), char);
1533 #ifdef Perl_safesysmalloc_size
1534 /* Do this here, do it once, do it right, and then we will never get
1535 called back into sv_grow() unless there really is some growing
1537 SvLEN_set(sv, Perl_safesysmalloc_size(s));
1539 SvLEN_set(sv, newlen);
1546 =for apidoc sv_setiv
1548 Copies an integer into the given SV, upgrading first if necessary.
1549 Does not handle 'set' magic. See also C<sv_setiv_mg>.
1555 Perl_sv_setiv(pTHX_ register SV *const sv, const IV i)
1559 PERL_ARGS_ASSERT_SV_SETIV;
1561 SV_CHECK_THINKFIRST_COW_DROP(sv);
1562 switch (SvTYPE(sv)) {
1565 sv_upgrade(sv, SVt_IV);
1568 sv_upgrade(sv, SVt_PVIV);
1572 if (!isGV_with_GP(sv))
1579 /* diag_listed_as: Can't coerce %s to %s in %s */
1580 Perl_croak(aTHX_ "Can't coerce %s to integer in %s", sv_reftype(sv,0),
1584 (void)SvIOK_only(sv); /* validate number */
1590 =for apidoc sv_setiv_mg
1592 Like C<sv_setiv>, but also handles 'set' magic.
1598 Perl_sv_setiv_mg(pTHX_ register SV *const sv, const IV i)
1600 PERL_ARGS_ASSERT_SV_SETIV_MG;
1607 =for apidoc sv_setuv
1609 Copies an unsigned integer into the given SV, upgrading first if necessary.
1610 Does not handle 'set' magic. See also C<sv_setuv_mg>.
1616 Perl_sv_setuv(pTHX_ register SV *const sv, const UV u)
1618 PERL_ARGS_ASSERT_SV_SETUV;
1620 /* With these two if statements:
1621 u=1.49 s=0.52 cu=72.49 cs=10.64 scripts=270 tests=20865
1624 u=1.35 s=0.47 cu=73.45 cs=11.43 scripts=270 tests=20865
1626 If you wish to remove them, please benchmark to see what the effect is
1628 if (u <= (UV)IV_MAX) {
1629 sv_setiv(sv, (IV)u);
1638 =for apidoc sv_setuv_mg
1640 Like C<sv_setuv>, but also handles 'set' magic.
1646 Perl_sv_setuv_mg(pTHX_ register SV *const sv, const UV u)
1648 PERL_ARGS_ASSERT_SV_SETUV_MG;
1655 =for apidoc sv_setnv
1657 Copies a double into the given SV, upgrading first if necessary.
1658 Does not handle 'set' magic. See also C<sv_setnv_mg>.
1664 Perl_sv_setnv(pTHX_ register SV *const sv, const NV num)
1668 PERL_ARGS_ASSERT_SV_SETNV;
1670 SV_CHECK_THINKFIRST_COW_DROP(sv);
1671 switch (SvTYPE(sv)) {
1674 sv_upgrade(sv, SVt_NV);
1678 sv_upgrade(sv, SVt_PVNV);
1682 if (!isGV_with_GP(sv))
1689 /* diag_listed_as: Can't coerce %s to %s in %s */
1690 Perl_croak(aTHX_ "Can't coerce %s to number in %s", sv_reftype(sv,0),
1695 (void)SvNOK_only(sv); /* validate number */
1700 =for apidoc sv_setnv_mg
1702 Like C<sv_setnv>, but also handles 'set' magic.
1708 Perl_sv_setnv_mg(pTHX_ register SV *const sv, const NV num)
1710 PERL_ARGS_ASSERT_SV_SETNV_MG;
1716 /* Print an "isn't numeric" warning, using a cleaned-up,
1717 * printable version of the offending string
1721 S_not_a_number(pTHX_ SV *const sv)
1728 PERL_ARGS_ASSERT_NOT_A_NUMBER;
1731 dsv = newSVpvs_flags("", SVs_TEMP);
1732 pv = sv_uni_display(dsv, sv, 10, 0);
1735 const char * const limit = tmpbuf + sizeof(tmpbuf) - 8;
1736 /* each *s can expand to 4 chars + "...\0",
1737 i.e. need room for 8 chars */
1739 const char *s = SvPVX_const(sv);
1740 const char * const end = s + SvCUR(sv);
1741 for ( ; s < end && d < limit; s++ ) {
1743 if (ch & 128 && !isPRINT_LC(ch)) {
1752 else if (ch == '\r') {
1756 else if (ch == '\f') {
1760 else if (ch == '\\') {
1764 else if (ch == '\0') {
1768 else if (isPRINT_LC(ch))
1785 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
1786 "Argument \"%s\" isn't numeric in %s", pv,
1789 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
1790 "Argument \"%s\" isn't numeric", pv);
1794 =for apidoc looks_like_number
1796 Test if the content of an SV looks like a number (or is a number).
1797 C<Inf> and C<Infinity> are treated as numbers (so will not issue a
1798 non-numeric warning), even if your atof() doesn't grok them.
1804 Perl_looks_like_number(pTHX_ SV *const sv)
1806 register const char *sbegin;
1809 PERL_ARGS_ASSERT_LOOKS_LIKE_NUMBER;
1812 sbegin = SvPVX_const(sv);
1815 else if (SvPOKp(sv))
1816 sbegin = SvPV_const(sv, len);
1818 return SvFLAGS(sv) & (SVf_NOK|SVp_NOK|SVf_IOK|SVp_IOK);
1819 return grok_number(sbegin, len, NULL);
1823 S_glob_2number(pTHX_ GV * const gv)
1825 const U32 wasfake = SvFLAGS(gv) & SVf_FAKE;
1826 SV *const buffer = sv_newmortal();
1828 PERL_ARGS_ASSERT_GLOB_2NUMBER;
1830 /* FAKE globs can get coerced, so need to turn this off temporarily if it
1833 gv_efullname3(buffer, gv, "*");
1834 SvFLAGS(gv) |= wasfake;
1836 /* We know that all GVs stringify to something that is not-a-number,
1837 so no need to test that. */
1838 if (ckWARN(WARN_NUMERIC))
1839 not_a_number(buffer);
1840 /* We just want something true to return, so that S_sv_2iuv_common
1841 can tail call us and return true. */
1845 /* Actually, ISO C leaves conversion of UV to IV undefined, but
1846 until proven guilty, assume that things are not that bad... */
1851 As 64 bit platforms often have an NV that doesn't preserve all bits of
1852 an IV (an assumption perl has been based on to date) it becomes necessary
1853 to remove the assumption that the NV always carries enough precision to
1854 recreate the IV whenever needed, and that the NV is the canonical form.
1855 Instead, IV/UV and NV need to be given equal rights. So as to not lose
1856 precision as a side effect of conversion (which would lead to insanity
1857 and the dragon(s) in t/op/numconvert.t getting very angry) the intent is
1858 1) to distinguish between IV/UV/NV slots that have cached a valid
1859 conversion where precision was lost and IV/UV/NV slots that have a
1860 valid conversion which has lost no precision
1861 2) to ensure that if a numeric conversion to one form is requested that
1862 would lose precision, the precise conversion (or differently
1863 imprecise conversion) is also performed and cached, to prevent
1864 requests for different numeric formats on the same SV causing
1865 lossy conversion chains. (lossless conversion chains are perfectly
1870 SvIOKp is true if the IV slot contains a valid value
1871 SvIOK is true only if the IV value is accurate (UV if SvIOK_UV true)
1872 SvNOKp is true if the NV slot contains a valid value
1873 SvNOK is true only if the NV value is accurate
1876 while converting from PV to NV, check to see if converting that NV to an
1877 IV(or UV) would lose accuracy over a direct conversion from PV to
1878 IV(or UV). If it would, cache both conversions, return NV, but mark
1879 SV as IOK NOKp (ie not NOK).
1881 While converting from PV to IV, check to see if converting that IV to an
1882 NV would lose accuracy over a direct conversion from PV to NV. If it
1883 would, cache both conversions, flag similarly.
1885 Before, the SV value "3.2" could become NV=3.2 IV=3 NOK, IOK quite
1886 correctly because if IV & NV were set NV *always* overruled.
1887 Now, "3.2" will become NV=3.2 IV=3 NOK, IOKp, because the flag's meaning
1888 changes - now IV and NV together means that the two are interchangeable:
1889 SvIVX == (IV) SvNVX && SvNVX == (NV) SvIVX;
1891 The benefit of this is that operations such as pp_add know that if
1892 SvIOK is true for both left and right operands, then integer addition
1893 can be used instead of floating point (for cases where the result won't
1894 overflow). Before, floating point was always used, which could lead to
1895 loss of precision compared with integer addition.
1897 * making IV and NV equal status should make maths accurate on 64 bit
1899 * may speed up maths somewhat if pp_add and friends start to use
1900 integers when possible instead of fp. (Hopefully the overhead in
1901 looking for SvIOK and checking for overflow will not outweigh the
1902 fp to integer speedup)
1903 * will slow down integer operations (callers of SvIV) on "inaccurate"
1904 values, as the change from SvIOK to SvIOKp will cause a call into
1905 sv_2iv each time rather than a macro access direct to the IV slot
1906 * should speed up number->string conversion on integers as IV is
1907 favoured when IV and NV are equally accurate
1909 ####################################################################
1910 You had better be using SvIOK_notUV if you want an IV for arithmetic:
1911 SvIOK is true if (IV or UV), so you might be getting (IV)SvUV.
1912 On the other hand, SvUOK is true iff UV.
1913 ####################################################################
1915 Your mileage will vary depending your CPU's relative fp to integer
1919 #ifndef NV_PRESERVES_UV
1920 # define IS_NUMBER_UNDERFLOW_IV 1
1921 # define IS_NUMBER_UNDERFLOW_UV 2
1922 # define IS_NUMBER_IV_AND_UV 2
1923 # define IS_NUMBER_OVERFLOW_IV 4
1924 # define IS_NUMBER_OVERFLOW_UV 5
1926 /* sv_2iuv_non_preserve(): private routine for use by sv_2iv() and sv_2uv() */
1928 /* For sv_2nv these three cases are "SvNOK and don't bother casting" */
1930 S_sv_2iuv_non_preserve(pTHX_ register SV *const sv
1938 PERL_ARGS_ASSERT_SV_2IUV_NON_PRESERVE;
1940 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));
1941 if (SvNVX(sv) < (NV)IV_MIN) {
1942 (void)SvIOKp_on(sv);
1944 SvIV_set(sv, IV_MIN);
1945 return IS_NUMBER_UNDERFLOW_IV;
1947 if (SvNVX(sv) > (NV)UV_MAX) {
1948 (void)SvIOKp_on(sv);
1951 SvUV_set(sv, UV_MAX);
1952 return IS_NUMBER_OVERFLOW_UV;
1954 (void)SvIOKp_on(sv);
1956 /* Can't use strtol etc to convert this string. (See truth table in
1958 if (SvNVX(sv) <= (UV)IV_MAX) {
1959 SvIV_set(sv, I_V(SvNVX(sv)));
1960 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
1961 SvIOK_on(sv); /* Integer is precise. NOK, IOK */
1963 /* Integer is imprecise. NOK, IOKp */
1965 return SvNVX(sv) < 0 ? IS_NUMBER_UNDERFLOW_UV : IS_NUMBER_IV_AND_UV;
1968 SvUV_set(sv, U_V(SvNVX(sv)));
1969 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
1970 if (SvUVX(sv) == UV_MAX) {
1971 /* As we know that NVs don't preserve UVs, UV_MAX cannot
1972 possibly be preserved by NV. Hence, it must be overflow.
1974 return IS_NUMBER_OVERFLOW_UV;
1976 SvIOK_on(sv); /* Integer is precise. NOK, UOK */
1978 /* Integer is imprecise. NOK, IOKp */
1980 return IS_NUMBER_OVERFLOW_IV;
1982 #endif /* !NV_PRESERVES_UV*/
1985 S_sv_2iuv_common(pTHX_ SV *const sv)
1989 PERL_ARGS_ASSERT_SV_2IUV_COMMON;
1992 /* erm. not sure. *should* never get NOKp (without NOK) from sv_2nv
1993 * without also getting a cached IV/UV from it at the same time
1994 * (ie PV->NV conversion should detect loss of accuracy and cache
1995 * IV or UV at same time to avoid this. */
1996 /* IV-over-UV optimisation - choose to cache IV if possible */
1998 if (SvTYPE(sv) == SVt_NV)
1999 sv_upgrade(sv, SVt_PVNV);
2001 (void)SvIOKp_on(sv); /* Must do this first, to clear any SvOOK */
2002 /* < not <= as for NV doesn't preserve UV, ((NV)IV_MAX+1) will almost
2003 certainly cast into the IV range at IV_MAX, whereas the correct
2004 answer is the UV IV_MAX +1. Hence < ensures that dodgy boundary
2006 #if defined(NAN_COMPARE_BROKEN) && defined(Perl_isnan)
2007 if (Perl_isnan(SvNVX(sv))) {
2013 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2014 SvIV_set(sv, I_V(SvNVX(sv)));
2015 if (SvNVX(sv) == (NV) SvIVX(sv)
2016 #ifndef NV_PRESERVES_UV
2017 && (((UV)1 << NV_PRESERVES_UV_BITS) >
2018 (UV)(SvIVX(sv) > 0 ? SvIVX(sv) : -SvIVX(sv)))
2019 /* Don't flag it as "accurately an integer" if the number
2020 came from a (by definition imprecise) NV operation, and
2021 we're outside the range of NV integer precision */
2025 SvIOK_on(sv); /* Can this go wrong with rounding? NWC */
2027 /* scalar has trailing garbage, eg "42a" */
2029 DEBUG_c(PerlIO_printf(Perl_debug_log,
2030 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (precise)\n",
2036 /* IV not precise. No need to convert from PV, as NV
2037 conversion would already have cached IV if it detected
2038 that PV->IV would be better than PV->NV->IV
2039 flags already correct - don't set public IOK. */
2040 DEBUG_c(PerlIO_printf(Perl_debug_log,
2041 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (imprecise)\n",
2046 /* Can the above go wrong if SvIVX == IV_MIN and SvNVX < IV_MIN,
2047 but the cast (NV)IV_MIN rounds to a the value less (more
2048 negative) than IV_MIN which happens to be equal to SvNVX ??
2049 Analogous to 0xFFFFFFFFFFFFFFFF rounding up to NV (2**64) and
2050 NV rounding back to 0xFFFFFFFFFFFFFFFF, so UVX == UV(NVX) and
2051 (NV)UVX == NVX are both true, but the values differ. :-(
2052 Hopefully for 2s complement IV_MIN is something like
2053 0x8000000000000000 which will be exact. NWC */
2056 SvUV_set(sv, U_V(SvNVX(sv)));
2058 (SvNVX(sv) == (NV) SvUVX(sv))
2059 #ifndef NV_PRESERVES_UV
2060 /* Make sure it's not 0xFFFFFFFFFFFFFFFF */
2061 /*&& (SvUVX(sv) != UV_MAX) irrelevant with code below */
2062 && (((UV)1 << NV_PRESERVES_UV_BITS) > SvUVX(sv))
2063 /* Don't flag it as "accurately an integer" if the number
2064 came from a (by definition imprecise) NV operation, and
2065 we're outside the range of NV integer precision */
2071 DEBUG_c(PerlIO_printf(Perl_debug_log,
2072 "0x%"UVxf" 2iv(%"UVuf" => %"IVdf") (as unsigned)\n",
2078 else if (SvPOKp(sv) && SvLEN(sv)) {
2080 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2081 /* We want to avoid a possible problem when we cache an IV/ a UV which
2082 may be later translated to an NV, and the resulting NV is not
2083 the same as the direct translation of the initial string
2084 (eg 123.456 can shortcut to the IV 123 with atol(), but we must
2085 be careful to ensure that the value with the .456 is around if the
2086 NV value is requested in the future).
2088 This means that if we cache such an IV/a UV, we need to cache the
2089 NV as well. Moreover, we trade speed for space, and do not
2090 cache the NV if we are sure it's not needed.
2093 /* SVt_PVNV is one higher than SVt_PVIV, hence this order */
2094 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2095 == IS_NUMBER_IN_UV) {
2096 /* It's definitely an integer, only upgrade to PVIV */
2097 if (SvTYPE(sv) < SVt_PVIV)
2098 sv_upgrade(sv, SVt_PVIV);
2100 } else if (SvTYPE(sv) < SVt_PVNV)
2101 sv_upgrade(sv, SVt_PVNV);
2103 /* If NVs preserve UVs then we only use the UV value if we know that
2104 we aren't going to call atof() below. If NVs don't preserve UVs
2105 then the value returned may have more precision than atof() will
2106 return, even though value isn't perfectly accurate. */
2107 if ((numtype & (IS_NUMBER_IN_UV
2108 #ifdef NV_PRESERVES_UV
2111 )) == IS_NUMBER_IN_UV) {
2112 /* This won't turn off the public IOK flag if it was set above */
2113 (void)SvIOKp_on(sv);
2115 if (!(numtype & IS_NUMBER_NEG)) {
2117 if (value <= (UV)IV_MAX) {
2118 SvIV_set(sv, (IV)value);
2120 /* it didn't overflow, and it was positive. */
2121 SvUV_set(sv, value);
2125 /* 2s complement assumption */
2126 if (value <= (UV)IV_MIN) {
2127 SvIV_set(sv, -(IV)value);
2129 /* Too negative for an IV. This is a double upgrade, but
2130 I'm assuming it will be rare. */
2131 if (SvTYPE(sv) < SVt_PVNV)
2132 sv_upgrade(sv, SVt_PVNV);
2136 SvNV_set(sv, -(NV)value);
2137 SvIV_set(sv, IV_MIN);
2141 /* For !NV_PRESERVES_UV and IS_NUMBER_IN_UV and IS_NUMBER_NOT_INT we
2142 will be in the previous block to set the IV slot, and the next
2143 block to set the NV slot. So no else here. */
2145 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2146 != IS_NUMBER_IN_UV) {
2147 /* It wasn't an (integer that doesn't overflow the UV). */
2148 SvNV_set(sv, Atof(SvPVX_const(sv)));
2150 if (! numtype && ckWARN(WARN_NUMERIC))
2153 #if defined(USE_LONG_DOUBLE)
2154 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%" PERL_PRIgldbl ")\n",
2155 PTR2UV(sv), SvNVX(sv)));
2157 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"NVgf")\n",
2158 PTR2UV(sv), SvNVX(sv)));
2161 #ifdef NV_PRESERVES_UV
2162 (void)SvIOKp_on(sv);
2164 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2165 SvIV_set(sv, I_V(SvNVX(sv)));
2166 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
2169 NOOP; /* Integer is imprecise. NOK, IOKp */
2171 /* UV will not work better than IV */
2173 if (SvNVX(sv) > (NV)UV_MAX) {
2175 /* Integer is inaccurate. NOK, IOKp, is UV */
2176 SvUV_set(sv, UV_MAX);
2178 SvUV_set(sv, U_V(SvNVX(sv)));
2179 /* 0xFFFFFFFFFFFFFFFF not an issue in here, NVs
2180 NV preservse UV so can do correct comparison. */
2181 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
2184 NOOP; /* Integer is imprecise. NOK, IOKp, is UV */
2189 #else /* NV_PRESERVES_UV */
2190 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2191 == (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT)) {
2192 /* The IV/UV slot will have been set from value returned by
2193 grok_number above. The NV slot has just been set using
2196 assert (SvIOKp(sv));
2198 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2199 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2200 /* Small enough to preserve all bits. */
2201 (void)SvIOKp_on(sv);
2203 SvIV_set(sv, I_V(SvNVX(sv)));
2204 if ((NV)(SvIVX(sv)) == SvNVX(sv))
2206 /* Assumption: first non-preserved integer is < IV_MAX,
2207 this NV is in the preserved range, therefore: */
2208 if (!(U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))
2210 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);
2214 0 0 already failed to read UV.
2215 0 1 already failed to read UV.
2216 1 0 you won't get here in this case. IV/UV
2217 slot set, public IOK, Atof() unneeded.
2218 1 1 already read UV.
2219 so there's no point in sv_2iuv_non_preserve() attempting
2220 to use atol, strtol, strtoul etc. */
2222 sv_2iuv_non_preserve (sv, numtype);
2224 sv_2iuv_non_preserve (sv);
2228 #endif /* NV_PRESERVES_UV */
2229 /* It might be more code efficient to go through the entire logic above
2230 and conditionally set with SvIOKp_on() rather than SvIOK(), but it
2231 gets complex and potentially buggy, so more programmer efficient
2232 to do it this way, by turning off the public flags: */
2234 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK);
2238 if (isGV_with_GP(sv))
2239 return glob_2number(MUTABLE_GV(sv));
2241 if (!(SvFLAGS(sv) & SVs_PADTMP)) {
2242 if (!PL_localizing && ckWARN(WARN_UNINITIALIZED))
2245 if (SvTYPE(sv) < SVt_IV)
2246 /* Typically the caller expects that sv_any is not NULL now. */
2247 sv_upgrade(sv, SVt_IV);
2248 /* Return 0 from the caller. */
2255 =for apidoc sv_2iv_flags
2257 Return the integer value of an SV, doing any necessary string
2258 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2259 Normally used via the C<SvIV(sv)> and C<SvIVx(sv)> macros.
2265 Perl_sv_2iv_flags(pTHX_ register SV *const sv, const I32 flags)
2270 if (SvGMAGICAL(sv) || SvVALID(sv)) {
2271 /* FBMs use the space for SvIVX and SvNVX for other purposes, and use
2272 the same flag bit as SVf_IVisUV, so must not let them cache IVs.
2273 In practice they are extremely unlikely to actually get anywhere
2274 accessible by user Perl code - the only way that I'm aware of is when
2275 a constant subroutine which is used as the second argument to index.
2277 if (flags & SV_GMAGIC)
2282 return I_V(SvNVX(sv));
2284 if (SvPOKp(sv) && SvLEN(sv)) {
2287 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2289 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2290 == IS_NUMBER_IN_UV) {
2291 /* It's definitely an integer */
2292 if (numtype & IS_NUMBER_NEG) {
2293 if (value < (UV)IV_MIN)
2296 if (value < (UV)IV_MAX)
2301 if (ckWARN(WARN_NUMERIC))
2304 return I_V(Atof(SvPVX_const(sv)));
2309 assert(SvTYPE(sv) >= SVt_PVMG);
2310 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
2311 } else if (SvTHINKFIRST(sv)) {
2316 if (flags & SV_SKIP_OVERLOAD)
2318 tmpstr = AMG_CALLunary(sv, numer_amg);
2319 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2320 return SvIV(tmpstr);
2323 return PTR2IV(SvRV(sv));
2326 sv_force_normal_flags(sv, 0);
2328 if (SvREADONLY(sv) && !SvOK(sv)) {
2329 if (ckWARN(WARN_UNINITIALIZED))
2335 if (S_sv_2iuv_common(aTHX_ sv))
2338 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"IVdf")\n",
2339 PTR2UV(sv),SvIVX(sv)));
2340 return SvIsUV(sv) ? (IV)SvUVX(sv) : SvIVX(sv);
2344 =for apidoc sv_2uv_flags
2346 Return the unsigned integer value of an SV, doing any necessary string
2347 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2348 Normally used via the C<SvUV(sv)> and C<SvUVx(sv)> macros.
2354 Perl_sv_2uv_flags(pTHX_ register SV *const sv, const I32 flags)
2359 if (SvGMAGICAL(sv) || SvVALID(sv)) {
2360 /* FBMs use the space for SvIVX and SvNVX for other purposes, and use
2361 the same flag bit as SVf_IVisUV, so must not let them cache IVs. */
2362 if (flags & SV_GMAGIC)
2367 return U_V(SvNVX(sv));
2368 if (SvPOKp(sv) && SvLEN(sv)) {
2371 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2373 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2374 == IS_NUMBER_IN_UV) {
2375 /* It's definitely an integer */
2376 if (!(numtype & IS_NUMBER_NEG))
2380 if (ckWARN(WARN_NUMERIC))
2383 return U_V(Atof(SvPVX_const(sv)));
2388 assert(SvTYPE(sv) >= SVt_PVMG);
2389 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
2390 } else if (SvTHINKFIRST(sv)) {
2395 if (flags & SV_SKIP_OVERLOAD)
2397 tmpstr = AMG_CALLunary(sv, numer_amg);
2398 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2399 return SvUV(tmpstr);
2402 return PTR2UV(SvRV(sv));
2405 sv_force_normal_flags(sv, 0);
2407 if (SvREADONLY(sv) && !SvOK(sv)) {
2408 if (ckWARN(WARN_UNINITIALIZED))
2414 if (S_sv_2iuv_common(aTHX_ sv))
2418 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2uv(%"UVuf")\n",
2419 PTR2UV(sv),SvUVX(sv)));
2420 return SvIsUV(sv) ? SvUVX(sv) : (UV)SvIVX(sv);
2424 =for apidoc sv_2nv_flags
2426 Return the num value of an SV, doing any necessary string or integer
2427 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2428 Normally used via the C<SvNV(sv)> and C<SvNVx(sv)> macros.
2434 Perl_sv_2nv_flags(pTHX_ register SV *const sv, const I32 flags)
2439 if (SvGMAGICAL(sv) || SvVALID(sv)) {
2440 /* FBMs use the space for SvIVX and SvNVX for other purposes, and use
2441 the same flag bit as SVf_IVisUV, so must not let them cache NVs. */
2442 if (flags & SV_GMAGIC)
2446 if ((SvPOKp(sv) && SvLEN(sv)) && !SvIOKp(sv)) {
2447 if (!SvIOKp(sv) && ckWARN(WARN_NUMERIC) &&
2448 !grok_number(SvPVX_const(sv), SvCUR(sv), NULL))
2450 return Atof(SvPVX_const(sv));
2454 return (NV)SvUVX(sv);
2456 return (NV)SvIVX(sv);
2461 assert(SvTYPE(sv) >= SVt_PVMG);
2462 /* This falls through to the report_uninit near the end of the
2464 } else if (SvTHINKFIRST(sv)) {
2469 if (flags & SV_SKIP_OVERLOAD)
2471 tmpstr = AMG_CALLunary(sv, numer_amg);
2472 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2473 return SvNV(tmpstr);
2476 return PTR2NV(SvRV(sv));
2479 sv_force_normal_flags(sv, 0);
2481 if (SvREADONLY(sv) && !SvOK(sv)) {
2482 if (ckWARN(WARN_UNINITIALIZED))
2487 if (SvTYPE(sv) < SVt_NV) {
2488 /* The logic to use SVt_PVNV if necessary is in sv_upgrade. */
2489 sv_upgrade(sv, SVt_NV);
2490 #ifdef USE_LONG_DOUBLE
2492 STORE_NUMERIC_LOCAL_SET_STANDARD();
2493 PerlIO_printf(Perl_debug_log,
2494 "0x%"UVxf" num(%" PERL_PRIgldbl ")\n",
2495 PTR2UV(sv), SvNVX(sv));
2496 RESTORE_NUMERIC_LOCAL();
2500 STORE_NUMERIC_LOCAL_SET_STANDARD();
2501 PerlIO_printf(Perl_debug_log, "0x%"UVxf" num(%"NVgf")\n",
2502 PTR2UV(sv), SvNVX(sv));
2503 RESTORE_NUMERIC_LOCAL();
2507 else if (SvTYPE(sv) < SVt_PVNV)
2508 sv_upgrade(sv, SVt_PVNV);
2513 SvNV_set(sv, SvIsUV(sv) ? (NV)SvUVX(sv) : (NV)SvIVX(sv));
2514 #ifdef NV_PRESERVES_UV
2520 /* Only set the public NV OK flag if this NV preserves the IV */
2521 /* Check it's not 0xFFFFFFFFFFFFFFFF */
2523 SvIsUV(sv) ? ((SvUVX(sv) != UV_MAX)&&(SvUVX(sv) == U_V(SvNVX(sv))))
2524 : (SvIVX(sv) == I_V(SvNVX(sv))))
2530 else if (SvPOKp(sv) && SvLEN(sv)) {
2532 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2533 if (!SvIOKp(sv) && !numtype && ckWARN(WARN_NUMERIC))
2535 #ifdef NV_PRESERVES_UV
2536 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2537 == IS_NUMBER_IN_UV) {
2538 /* It's definitely an integer */
2539 SvNV_set(sv, (numtype & IS_NUMBER_NEG) ? -(NV)value : (NV)value);
2541 SvNV_set(sv, Atof(SvPVX_const(sv)));
2547 SvNV_set(sv, Atof(SvPVX_const(sv)));
2548 /* Only set the public NV OK flag if this NV preserves the value in
2549 the PV at least as well as an IV/UV would.
2550 Not sure how to do this 100% reliably. */
2551 /* if that shift count is out of range then Configure's test is
2552 wonky. We shouldn't be in here with NV_PRESERVES_UV_BITS ==
2554 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2555 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2556 SvNOK_on(sv); /* Definitely small enough to preserve all bits */
2557 } else if (!(numtype & IS_NUMBER_IN_UV)) {
2558 /* Can't use strtol etc to convert this string, so don't try.
2559 sv_2iv and sv_2uv will use the NV to convert, not the PV. */
2562 /* value has been set. It may not be precise. */
2563 if ((numtype & IS_NUMBER_NEG) && (value > (UV)IV_MIN)) {
2564 /* 2s complement assumption for (UV)IV_MIN */
2565 SvNOK_on(sv); /* Integer is too negative. */
2570 if (numtype & IS_NUMBER_NEG) {
2571 SvIV_set(sv, -(IV)value);
2572 } else if (value <= (UV)IV_MAX) {
2573 SvIV_set(sv, (IV)value);
2575 SvUV_set(sv, value);
2579 if (numtype & IS_NUMBER_NOT_INT) {
2580 /* I believe that even if the original PV had decimals,
2581 they are lost beyond the limit of the FP precision.
2582 However, neither is canonical, so both only get p
2583 flags. NWC, 2000/11/25 */
2584 /* Both already have p flags, so do nothing */
2586 const NV nv = SvNVX(sv);
2587 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2588 if (SvIVX(sv) == I_V(nv)) {
2591 /* It had no "." so it must be integer. */
2595 /* between IV_MAX and NV(UV_MAX).
2596 Could be slightly > UV_MAX */
2598 if (numtype & IS_NUMBER_NOT_INT) {
2599 /* UV and NV both imprecise. */
2601 const UV nv_as_uv = U_V(nv);
2603 if (value == nv_as_uv && SvUVX(sv) != UV_MAX) {
2612 /* It might be more code efficient to go through the entire logic above
2613 and conditionally set with SvNOKp_on() rather than SvNOK(), but it
2614 gets complex and potentially buggy, so more programmer efficient
2615 to do it this way, by turning off the public flags: */
2617 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK);
2618 #endif /* NV_PRESERVES_UV */
2621 if (isGV_with_GP(sv)) {
2622 glob_2number(MUTABLE_GV(sv));
2626 if (!PL_localizing && !(SvFLAGS(sv) & SVs_PADTMP) && ckWARN(WARN_UNINITIALIZED))
2628 assert (SvTYPE(sv) >= SVt_NV);
2629 /* Typically the caller expects that sv_any is not NULL now. */
2630 /* XXX Ilya implies that this is a bug in callers that assume this
2631 and ideally should be fixed. */
2634 #if defined(USE_LONG_DOUBLE)
2636 STORE_NUMERIC_LOCAL_SET_STANDARD();
2637 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2nv(%" PERL_PRIgldbl ")\n",
2638 PTR2UV(sv), SvNVX(sv));
2639 RESTORE_NUMERIC_LOCAL();
2643 STORE_NUMERIC_LOCAL_SET_STANDARD();
2644 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 1nv(%"NVgf")\n",
2645 PTR2UV(sv), SvNVX(sv));
2646 RESTORE_NUMERIC_LOCAL();
2655 Return an SV with the numeric value of the source SV, doing any necessary
2656 reference or overload conversion. You must use the C<SvNUM(sv)> macro to
2657 access this function.
2663 Perl_sv_2num(pTHX_ register SV *const sv)
2665 PERL_ARGS_ASSERT_SV_2NUM;
2670 SV * const tmpsv = AMG_CALLunary(sv, numer_amg);
2671 TAINT_IF(tmpsv && SvTAINTED(tmpsv));
2672 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
2673 return sv_2num(tmpsv);
2675 return sv_2mortal(newSVuv(PTR2UV(SvRV(sv))));
2678 /* uiv_2buf(): private routine for use by sv_2pv_flags(): print an IV or
2679 * UV as a string towards the end of buf, and return pointers to start and
2682 * We assume that buf is at least TYPE_CHARS(UV) long.
2686 S_uiv_2buf(char *const buf, const IV iv, UV uv, const int is_uv, char **const peob)
2688 char *ptr = buf + TYPE_CHARS(UV);
2689 char * const ebuf = ptr;
2692 PERL_ARGS_ASSERT_UIV_2BUF;
2704 *--ptr = '0' + (char)(uv % 10);
2713 =for apidoc sv_2pv_flags
2715 Returns a pointer to the string value of an SV, and sets *lp to its length.
2716 If flags includes SV_GMAGIC, does an mg_get() first. Coerces sv to a string
2718 Normally invoked via the C<SvPV_flags> macro. C<sv_2pv()> and C<sv_2pv_nomg>
2719 usually end up here too.
2725 Perl_sv_2pv_flags(pTHX_ register SV *const sv, STRLEN *const lp, const I32 flags)
2735 if (SvGMAGICAL(sv)) {
2736 if (flags & SV_GMAGIC)
2741 if (flags & SV_MUTABLE_RETURN)
2742 return SvPVX_mutable(sv);
2743 if (flags & SV_CONST_RETURN)
2744 return (char *)SvPVX_const(sv);
2747 if (SvIOKp(sv) || SvNOKp(sv)) {
2748 char tbuf[64]; /* Must fit sprintf/Gconvert of longest IV/NV */
2753 ? my_snprintf(tbuf, sizeof(tbuf), "%"UVuf, (UV)SvUVX(sv))
2754 : my_snprintf(tbuf, sizeof(tbuf), "%"IVdf, (IV)SvIVX(sv));
2755 } else if(SvNVX(sv) == 0.0) {
2760 Gconvert(SvNVX(sv), NV_DIG, 0, tbuf);
2767 SvUPGRADE(sv, SVt_PV);
2770 s = SvGROW_mutable(sv, len + 1);
2773 return (char*)memcpy(s, tbuf, len + 1);
2779 assert(SvTYPE(sv) >= SVt_PVMG);
2780 /* This falls through to the report_uninit near the end of the
2782 } else if (SvTHINKFIRST(sv)) {
2787 if (flags & SV_SKIP_OVERLOAD)
2789 tmpstr = AMG_CALLunary(sv, string_amg);
2790 TAINT_IF(tmpstr && SvTAINTED(tmpstr));
2791 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2793 /* char *pv = lp ? SvPV(tmpstr, *lp) : SvPV_nolen(tmpstr);
2797 if ((SvFLAGS(tmpstr) & (SVf_POK)) == SVf_POK) {
2798 if (flags & SV_CONST_RETURN) {
2799 pv = (char *) SvPVX_const(tmpstr);
2801 pv = (flags & SV_MUTABLE_RETURN)
2802 ? SvPVX_mutable(tmpstr) : SvPVX(tmpstr);
2805 *lp = SvCUR(tmpstr);
2807 pv = sv_2pv_flags(tmpstr, lp, flags);
2820 SV *const referent = SvRV(sv);
2824 retval = buffer = savepvn("NULLREF", len);
2825 } else if (SvTYPE(referent) == SVt_REGEXP) {
2826 REGEXP * const re = (REGEXP *)MUTABLE_PTR(referent);
2831 /* If the regex is UTF-8 we want the containing scalar to
2832 have an UTF-8 flag too */
2838 if ((seen_evals = RX_SEEN_EVALS(re)))
2839 PL_reginterp_cnt += seen_evals;
2842 *lp = RX_WRAPLEN(re);
2844 return RX_WRAPPED(re);
2846 const char *const typestr = sv_reftype(referent, 0);
2847 const STRLEN typelen = strlen(typestr);
2848 UV addr = PTR2UV(referent);
2849 const char *stashname = NULL;
2850 STRLEN stashnamelen = 0; /* hush, gcc */
2851 const char *buffer_end;
2853 if (SvOBJECT(referent)) {
2854 const HEK *const name = HvNAME_HEK(SvSTASH(referent));
2857 stashname = HEK_KEY(name);
2858 stashnamelen = HEK_LEN(name);
2860 if (HEK_UTF8(name)) {
2866 stashname = "__ANON__";
2869 len = stashnamelen + 1 /* = */ + typelen + 3 /* (0x */
2870 + 2 * sizeof(UV) + 2 /* )\0 */;
2872 len = typelen + 3 /* (0x */
2873 + 2 * sizeof(UV) + 2 /* )\0 */;
2876 Newx(buffer, len, char);
2877 buffer_end = retval = buffer + len;
2879 /* Working backwards */
2883 *--retval = PL_hexdigit[addr & 15];
2884 } while (addr >>= 4);
2890 memcpy(retval, typestr, typelen);
2894 retval -= stashnamelen;
2895 memcpy(retval, stashname, stashnamelen);
2897 /* retval may not necessarily have reached the start of the
2899 assert (retval >= buffer);
2901 len = buffer_end - retval - 1; /* -1 for that \0 */
2909 if (SvREADONLY(sv) && !SvOK(sv)) {
2912 if (flags & SV_UNDEF_RETURNS_NULL)
2914 if (ckWARN(WARN_UNINITIALIZED))
2919 if (SvIOK(sv) || ((SvIOKp(sv) && !SvNOKp(sv)))) {
2920 /* I'm assuming that if both IV and NV are equally valid then
2921 converting the IV is going to be more efficient */
2922 const U32 isUIOK = SvIsUV(sv);
2923 char buf[TYPE_CHARS(UV)];
2927 if (SvTYPE(sv) < SVt_PVIV)
2928 sv_upgrade(sv, SVt_PVIV);
2929 ptr = uiv_2buf(buf, SvIVX(sv), SvUVX(sv), isUIOK, &ebuf);
2931 /* inlined from sv_setpvn */
2932 s = SvGROW_mutable(sv, len + 1);
2933 Move(ptr, s, len, char);
2937 else if (SvNOKp(sv)) {
2938 if (SvTYPE(sv) < SVt_PVNV)
2939 sv_upgrade(sv, SVt_PVNV);
2940 if (SvNVX(sv) == 0.0) {
2941 s = SvGROW_mutable(sv, 2);
2946 /* The +20 is pure guesswork. Configure test needed. --jhi */
2947 s = SvGROW_mutable(sv, NV_DIG + 20);
2948 /* some Xenix systems wipe out errno here */
2949 Gconvert(SvNVX(sv), NV_DIG, 0, s);
2959 if (isGV_with_GP(sv)) {
2960 GV *const gv = MUTABLE_GV(sv);
2961 const U32 wasfake = SvFLAGS(gv) & SVf_FAKE;
2962 SV *const buffer = sv_newmortal();
2964 /* FAKE globs can get coerced, so need to turn this off temporarily
2967 gv_efullname3(buffer, gv, "*");
2968 SvFLAGS(gv) |= wasfake;
2970 if (SvPOK(buffer)) {
2972 *lp = SvCUR(buffer);
2974 return SvPVX(buffer);
2985 if (flags & SV_UNDEF_RETURNS_NULL)
2987 if (!PL_localizing && !(SvFLAGS(sv) & SVs_PADTMP) && ckWARN(WARN_UNINITIALIZED))
2989 if (SvTYPE(sv) < SVt_PV)
2990 /* Typically the caller expects that sv_any is not NULL now. */
2991 sv_upgrade(sv, SVt_PV);
2995 const STRLEN len = s - SvPVX_const(sv);
3001 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
3002 PTR2UV(sv),SvPVX_const(sv)));
3003 if (flags & SV_CONST_RETURN)
3004 return (char *)SvPVX_const(sv);
3005 if (flags & SV_MUTABLE_RETURN)
3006 return SvPVX_mutable(sv);
3011 =for apidoc sv_copypv
3013 Copies a stringified representation of the source SV into the
3014 destination SV. Automatically performs any necessary mg_get and
3015 coercion of numeric values into strings. Guaranteed to preserve
3016 UTF8 flag even from overloaded objects. Similar in nature to
3017 sv_2pv[_flags] but operates directly on an SV instead of just the
3018 string. Mostly uses sv_2pv_flags to do its work, except when that
3019 would lose the UTF-8'ness of the PV.
3025 Perl_sv_copypv(pTHX_ SV *const dsv, register SV *const ssv)
3028 const char * const s = SvPV_const(ssv,len);
3030 PERL_ARGS_ASSERT_SV_COPYPV;
3032 sv_setpvn(dsv,s,len);
3040 =for apidoc sv_2pvbyte
3042 Return a pointer to the byte-encoded representation of the SV, and set *lp
3043 to its length. May cause the SV to be downgraded from UTF-8 as a
3046 Usually accessed via the C<SvPVbyte> macro.
3052 Perl_sv_2pvbyte(pTHX_ register SV *const sv, STRLEN *const lp)
3054 PERL_ARGS_ASSERT_SV_2PVBYTE;
3057 sv_utf8_downgrade(sv,0);
3058 return lp ? SvPV_nomg(sv,*lp) : SvPV_nomg_nolen(sv);
3062 =for apidoc sv_2pvutf8
3064 Return a pointer to the UTF-8-encoded representation of the SV, and set *lp
3065 to its length. May cause the SV to be upgraded to UTF-8 as a side-effect.
3067 Usually accessed via the C<SvPVutf8> macro.
3073 Perl_sv_2pvutf8(pTHX_ register SV *const sv, STRLEN *const lp)
3075 PERL_ARGS_ASSERT_SV_2PVUTF8;
3077 sv_utf8_upgrade(sv);
3078 return lp ? SvPV(sv,*lp) : SvPV_nolen(sv);
3083 =for apidoc sv_2bool
3085 This macro is only used by sv_true() or its macro equivalent, and only if
3086 the latter's argument is neither SvPOK, SvIOK nor SvNOK.
3087 It calls sv_2bool_flags with the SV_GMAGIC flag.
3089 =for apidoc sv_2bool_flags
3091 This function is only used by sv_true() and friends, and only if
3092 the latter's argument is neither SvPOK, SvIOK nor SvNOK. If the flags
3093 contain SV_GMAGIC, then it does an mg_get() first.
3100 Perl_sv_2bool_flags(pTHX_ register SV *const sv, const I32 flags)
3104 PERL_ARGS_ASSERT_SV_2BOOL_FLAGS;
3106 if(flags & SV_GMAGIC) SvGETMAGIC(sv);
3112 SV * const tmpsv = AMG_CALLunary(sv, bool__amg);
3113 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
3114 return cBOOL(SvTRUE(tmpsv));
3116 return SvRV(sv) != 0;
3119 register XPV* const Xpvtmp = (XPV*)SvANY(sv);
3121 (*sv->sv_u.svu_pv > '0' ||
3122 Xpvtmp->xpv_cur > 1 ||
3123 (Xpvtmp->xpv_cur && *sv->sv_u.svu_pv != '0')))
3130 return SvIVX(sv) != 0;
3133 return SvNVX(sv) != 0.0;
3135 if (isGV_with_GP(sv))
3145 =for apidoc sv_utf8_upgrade
3147 Converts the PV of an SV to its UTF-8-encoded form.
3148 Forces the SV to string form if it is not already.
3149 Will C<mg_get> on C<sv> if appropriate.
3150 Always sets the SvUTF8 flag to avoid future validity checks even
3151 if the whole string is the same in UTF-8 as not.
3152 Returns the number of bytes in the converted string
3154 This is not as a general purpose byte encoding to Unicode interface:
3155 use the Encode extension for that.
3157 =for apidoc sv_utf8_upgrade_nomg
3159 Like sv_utf8_upgrade, but doesn't do magic on C<sv>
3161 =for apidoc sv_utf8_upgrade_flags
3163 Converts the PV of an SV to its UTF-8-encoded form.
3164 Forces the SV to string form if it is not already.
3165 Always sets the SvUTF8 flag to avoid future validity checks even
3166 if all the bytes are invariant in UTF-8. If C<flags> has C<SV_GMAGIC> bit set,
3167 will C<mg_get> on C<sv> if appropriate, else not.
3168 Returns the number of bytes in the converted string
3169 C<sv_utf8_upgrade> and
3170 C<sv_utf8_upgrade_nomg> are implemented in terms of this function.
3172 This is not as a general purpose byte encoding to Unicode interface:
3173 use the Encode extension for that.
3177 The grow version is currently not externally documented. It adds a parameter,
3178 extra, which is the number of unused bytes the string of 'sv' is guaranteed to
3179 have free after it upon return. This allows the caller to reserve extra space
3180 that it intends to fill, to avoid extra grows.
3182 Also externally undocumented for the moment is the flag SV_FORCE_UTF8_UPGRADE,
3183 which can be used to tell this function to not first check to see if there are
3184 any characters that are different in UTF-8 (variant characters) which would
3185 force it to allocate a new string to sv, but to assume there are. Typically
3186 this flag is used by a routine that has already parsed the string to find that
3187 there are such characters, and passes this information on so that the work
3188 doesn't have to be repeated.
3190 (One might think that the calling routine could pass in the position of the
3191 first such variant, so it wouldn't have to be found again. But that is not the
3192 case, because typically when the caller is likely to use this flag, it won't be
3193 calling this routine unless it finds something that won't fit into a byte.
3194 Otherwise it tries to not upgrade and just use bytes. But some things that
3195 do fit into a byte are variants in utf8, and the caller may not have been
3196 keeping track of these.)
3198 If the routine itself changes the string, it adds a trailing NUL. Such a NUL
3199 isn't guaranteed due to having other routines do the work in some input cases,
3200 or if the input is already flagged as being in utf8.
3202 The speed of this could perhaps be improved for many cases if someone wanted to
3203 write a fast function that counts the number of variant characters in a string,
3204 especially if it could return the position of the first one.
3209 Perl_sv_utf8_upgrade_flags_grow(pTHX_ register SV *const sv, const I32 flags, STRLEN extra)
3213 PERL_ARGS_ASSERT_SV_UTF8_UPGRADE_FLAGS_GROW;
3215 if (sv == &PL_sv_undef)
3219 if (SvREADONLY(sv) && (SvPOKp(sv) || SvIOKp(sv) || SvNOKp(sv))) {
3220 (void) sv_2pv_flags(sv,&len, flags);
3222 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3226 (void) SvPV_force_flags(sv,len,flags & SV_GMAGIC);
3231 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3236 sv_force_normal_flags(sv, 0);
3239 if (PL_encoding && !(flags & SV_UTF8_NO_ENCODING)) {
3240 sv_recode_to_utf8(sv, PL_encoding);
3241 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3245 if (SvCUR(sv) == 0) {
3246 if (extra) SvGROW(sv, extra);
3247 } else { /* Assume Latin-1/EBCDIC */
3248 /* This function could be much more efficient if we
3249 * had a FLAG in SVs to signal if there are any variant
3250 * chars in the PV. Given that there isn't such a flag
3251 * make the loop as fast as possible (although there are certainly ways
3252 * to speed this up, eg. through vectorization) */
3253 U8 * s = (U8 *) SvPVX_const(sv);
3254 U8 * e = (U8 *) SvEND(sv);
3256 STRLEN two_byte_count = 0;
3258 if (flags & SV_FORCE_UTF8_UPGRADE) goto must_be_utf8;
3260 /* See if really will need to convert to utf8. We mustn't rely on our
3261 * incoming SV being well formed and having a trailing '\0', as certain
3262 * code in pp_formline can send us partially built SVs. */
3266 if (NATIVE_IS_INVARIANT(ch)) continue;
3268 t--; /* t already incremented; re-point to first variant */
3273 /* utf8 conversion not needed because all are invariants. Mark as
3274 * UTF-8 even if no variant - saves scanning loop */
3280 /* Here, the string should be converted to utf8, either because of an
3281 * input flag (two_byte_count = 0), or because a character that
3282 * requires 2 bytes was found (two_byte_count = 1). t points either to
3283 * the beginning of the string (if we didn't examine anything), or to
3284 * the first variant. In either case, everything from s to t - 1 will
3285 * occupy only 1 byte each on output.
3287 * There are two main ways to convert. One is to create a new string
3288 * and go through the input starting from the beginning, appending each
3289 * converted value onto the new string as we go along. It's probably
3290 * best to allocate enough space in the string for the worst possible
3291 * case rather than possibly running out of space and having to
3292 * reallocate and then copy what we've done so far. Since everything
3293 * from s to t - 1 is invariant, the destination can be initialized
3294 * with these using a fast memory copy
3296 * The other way is to figure out exactly how big the string should be
3297 * by parsing the entire input. Then you don't have to make it big
3298 * enough to handle the worst possible case, and more importantly, if
3299 * the string you already have is large enough, you don't have to
3300 * allocate a new string, you can copy the last character in the input
3301 * string to the final position(s) that will be occupied by the
3302 * converted string and go backwards, stopping at t, since everything
3303 * before that is invariant.
3305 * There are advantages and disadvantages to each method.
3307 * In the first method, we can allocate a new string, do the memory
3308 * copy from the s to t - 1, and then proceed through the rest of the
3309 * string byte-by-byte.
3311 * In the second method, we proceed through the rest of the input
3312 * string just calculating how big the converted string will be. Then
3313 * there are two cases:
3314 * 1) if the string has enough extra space to handle the converted
3315 * value. We go backwards through the string, converting until we
3316 * get to the position we are at now, and then stop. If this
3317 * position is far enough along in the string, this method is
3318 * faster than the other method. If the memory copy were the same
3319 * speed as the byte-by-byte loop, that position would be about
3320 * half-way, as at the half-way mark, parsing to the end and back
3321 * is one complete string's parse, the same amount as starting
3322 * over and going all the way through. Actually, it would be
3323 * somewhat less than half-way, as it's faster to just count bytes
3324 * than to also copy, and we don't have the overhead of allocating
3325 * a new string, changing the scalar to use it, and freeing the
3326 * existing one. But if the memory copy is fast, the break-even
3327 * point is somewhere after half way. The counting loop could be
3328 * sped up by vectorization, etc, to move the break-even point
3329 * further towards the beginning.
3330 * 2) if the string doesn't have enough space to handle the converted
3331 * value. A new string will have to be allocated, and one might
3332 * as well, given that, start from the beginning doing the first
3333 * method. We've spent extra time parsing the string and in
3334 * exchange all we've gotten is that we know precisely how big to
3335 * make the new one. Perl is more optimized for time than space,
3336 * so this case is a loser.
3337 * So what I've decided to do is not use the 2nd method unless it is
3338 * guaranteed that a new string won't have to be allocated, assuming
3339 * the worst case. I also decided not to put any more conditions on it
3340 * than this, for now. It seems likely that, since the worst case is
3341 * twice as big as the unknown portion of the string (plus 1), we won't
3342 * be guaranteed enough space, causing us to go to the first method,
3343 * unless the string is short, or the first variant character is near
3344 * the end of it. In either of these cases, it seems best to use the
3345 * 2nd method. The only circumstance I can think of where this would
3346 * be really slower is if the string had once had much more data in it
3347 * than it does now, but there is still a substantial amount in it */
3350 STRLEN invariant_head = t - s;
3351 STRLEN size = invariant_head + (e - t) * 2 + 1 + extra;
3352 if (SvLEN(sv) < size) {
3354 /* Here, have decided to allocate a new string */
3359 Newx(dst, size, U8);
3361 /* If no known invariants at the beginning of the input string,
3362 * set so starts from there. Otherwise, can use memory copy to
3363 * get up to where we are now, and then start from here */
3365 if (invariant_head <= 0) {
3368 Copy(s, dst, invariant_head, char);
3369 d = dst + invariant_head;
3373 const UV uv = NATIVE8_TO_UNI(*t++);
3374 if (UNI_IS_INVARIANT(uv))
3375 *d++ = (U8)UNI_TO_NATIVE(uv);
3377 *d++ = (U8)UTF8_EIGHT_BIT_HI(uv);
3378 *d++ = (U8)UTF8_EIGHT_BIT_LO(uv);
3382 SvPV_free(sv); /* No longer using pre-existing string */
3383 SvPV_set(sv, (char*)dst);
3384 SvCUR_set(sv, d - dst);
3385 SvLEN_set(sv, size);
3388 /* Here, have decided to get the exact size of the string.
3389 * Currently this happens only when we know that there is
3390 * guaranteed enough space to fit the converted string, so
3391 * don't have to worry about growing. If two_byte_count is 0,
3392 * then t points to the first byte of the string which hasn't
3393 * been examined yet. Otherwise two_byte_count is 1, and t
3394 * points to the first byte in the string that will expand to
3395 * two. Depending on this, start examining at t or 1 after t.
3398 U8 *d = t + two_byte_count;
3401 /* Count up the remaining bytes that expand to two */
3404 const U8 chr = *d++;
3405 if (! NATIVE_IS_INVARIANT(chr)) two_byte_count++;
3408 /* The string will expand by just the number of bytes that
3409 * occupy two positions. But we are one afterwards because of
3410 * the increment just above. This is the place to put the
3411 * trailing NUL, and to set the length before we decrement */
3413 d += two_byte_count;
3414 SvCUR_set(sv, d - s);
3418 /* Having decremented d, it points to the position to put the
3419 * very last byte of the expanded string. Go backwards through
3420 * the string, copying and expanding as we go, stopping when we
3421 * get to the part that is invariant the rest of the way down */
3425 const U8 ch = NATIVE8_TO_UNI(*e--);
3426 if (UNI_IS_INVARIANT(ch)) {
3427 *d-- = UNI_TO_NATIVE(ch);
3429 *d-- = (U8)UTF8_EIGHT_BIT_LO(ch);
3430 *d-- = (U8)UTF8_EIGHT_BIT_HI(ch);
3435 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
3436 /* Update pos. We do it at the end rather than during
3437 * the upgrade, to avoid slowing down the common case
3438 * (upgrade without pos) */
3439 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
3441 I32 pos = mg->mg_len;
3442 if (pos > 0 && (U32)pos > invariant_head) {
3443 U8 *d = (U8*) SvPVX(sv) + invariant_head;
3444 STRLEN n = (U32)pos - invariant_head;
3446 if (UTF8_IS_START(*d))
3451 mg->mg_len = d - (U8*)SvPVX(sv);
3454 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
3455 magic_setutf8(sv,mg); /* clear UTF8 cache */
3460 /* Mark as UTF-8 even if no variant - saves scanning loop */
3466 =for apidoc sv_utf8_downgrade
3468 Attempts to convert the PV of an SV from characters to bytes.
3469 If the PV contains a character that cannot fit
3470 in a byte, this conversion will fail;
3471 in this case, either returns false or, if C<fail_ok> is not
3474 This is not as a general purpose Unicode to byte encoding interface:
3475 use the Encode extension for that.
3481 Perl_sv_utf8_downgrade(pTHX_ register SV *const sv, const bool fail_ok)
3485 PERL_ARGS_ASSERT_SV_UTF8_DOWNGRADE;
3487 if (SvPOKp(sv) && SvUTF8(sv)) {
3491 int mg_flags = SV_GMAGIC;
3494 sv_force_normal_flags(sv, 0);
3496 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
3498 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
3500 I32 pos = mg->mg_len;
3502 sv_pos_b2u(sv, &pos);
3503 mg_flags = 0; /* sv_pos_b2u does get magic */
3507 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
3508 magic_setutf8(sv,mg); /* clear UTF8 cache */
3511 s = (U8 *) SvPV_flags(sv, len, mg_flags);
3513 if (!utf8_to_bytes(s, &len)) {
3518 Perl_croak(aTHX_ "Wide character in %s",
3521 Perl_croak(aTHX_ "Wide character");
3532 =for apidoc sv_utf8_encode
3534 Converts the PV of an SV to UTF-8, but then turns the C<SvUTF8>
3535 flag off so that it looks like octets again.
3541 Perl_sv_utf8_encode(pTHX_ register SV *const sv)
3543 PERL_ARGS_ASSERT_SV_UTF8_ENCODE;
3546 sv_force_normal_flags(sv, 0);
3548 if (SvREADONLY(sv)) {
3549 Perl_croak_no_modify(aTHX);
3551 (void) sv_utf8_upgrade(sv);
3556 =for apidoc sv_utf8_decode
3558 If the PV of the SV is an octet sequence in UTF-8
3559 and contains a multiple-byte character, the C<SvUTF8> flag is turned on
3560 so that it looks like a character. If the PV contains only single-byte
3561 characters, the C<SvUTF8> flag stays off.
3562 Scans PV for validity and returns false if the PV is invalid UTF-8.
3568 Perl_sv_utf8_decode(pTHX_ register SV *const sv)
3570 PERL_ARGS_ASSERT_SV_UTF8_DECODE;
3573 const U8 *start, *c;
3576 /* The octets may have got themselves encoded - get them back as
3579 if (!sv_utf8_downgrade(sv, TRUE))
3582 /* it is actually just a matter of turning the utf8 flag on, but
3583 * we want to make sure everything inside is valid utf8 first.
3585 c = start = (const U8 *) SvPVX_const(sv);
3586 if (!is_utf8_string(c, SvCUR(sv)+1))
3588 e = (const U8 *) SvEND(sv);
3591 if (!UTF8_IS_INVARIANT(ch)) {
3596 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
3597 /* adjust pos to the start of a UTF8 char sequence */
3598 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
3600 I32 pos = mg->mg_len;
3602 for (c = start + pos; c > start; c--) {
3603 if (UTF8_IS_START(*c))
3606 mg->mg_len = c - start;
3609 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
3610 magic_setutf8(sv,mg); /* clear UTF8 cache */
3617 =for apidoc sv_setsv
3619 Copies the contents of the source SV C<ssv> into the destination SV
3620 C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3621 function if the source SV needs to be reused. Does not handle 'set' magic.
3622 Loosely speaking, it performs a copy-by-value, obliterating any previous
3623 content of the destination.
3625 You probably want to use one of the assortment of wrappers, such as
3626 C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3627 C<SvSetMagicSV_nosteal>.
3629 =for apidoc sv_setsv_flags
3631 Copies the contents of the source SV C<ssv> into the destination SV
3632 C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3633 function if the source SV needs to be reused. Does not handle 'set' magic.
3634 Loosely speaking, it performs a copy-by-value, obliterating any previous
3635 content of the destination.
3636 If the C<flags> parameter has the C<SV_GMAGIC> bit set, will C<mg_get> on
3637 C<ssv> if appropriate, else not. If the C<flags> parameter has the
3638 C<NOSTEAL> bit set then the buffers of temps will not be stolen. <sv_setsv>
3639 and C<sv_setsv_nomg> are implemented in terms of this function.
3641 You probably want to use one of the assortment of wrappers, such as
3642 C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3643 C<SvSetMagicSV_nosteal>.
3645 This is the primary function for copying scalars, and most other
3646 copy-ish functions and macros use this underneath.
3652 S_glob_assign_glob(pTHX_ SV *const dstr, SV *const sstr, const int dtype)
3654 I32 mro_changes = 0; /* 1 = method, 2 = isa, 3 = recursive isa */
3655 HV *old_stash = NULL;
3657 PERL_ARGS_ASSERT_GLOB_ASSIGN_GLOB;
3659 if (dtype != SVt_PVGV && !isGV_with_GP(dstr)) {
3660 const char * const name = GvNAME(sstr);
3661 const STRLEN len = GvNAMELEN(sstr);
3663 if (dtype >= SVt_PV) {
3669 SvUPGRADE(dstr, SVt_PVGV);
3670 (void)SvOK_off(dstr);
3671 /* FIXME - why are we doing this, then turning it off and on again
3673 isGV_with_GP_on(dstr);
3675 GvSTASH(dstr) = GvSTASH(sstr);
3677 Perl_sv_add_backref(aTHX_ MUTABLE_SV(GvSTASH(dstr)), dstr);
3678 gv_name_set(MUTABLE_GV(dstr), name, len, GV_ADD);
3679 SvFAKE_on(dstr); /* can coerce to non-glob */
3682 if(GvGP(MUTABLE_GV(sstr))) {
3683 /* If source has method cache entry, clear it */
3685 SvREFCNT_dec(GvCV(sstr));
3686 GvCV_set(sstr, NULL);
3689 /* If source has a real method, then a method is
3692 GvCV((const GV *)sstr) && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3698 /* If dest already had a real method, that's a change as well */
3700 !mro_changes && GvGP(MUTABLE_GV(dstr)) && GvCVu((const GV *)dstr)
3701 && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3706 /* We don’t need to check the name of the destination if it was not a
3707 glob to begin with. */
3708 if(dtype == SVt_PVGV) {
3709 const char * const name = GvNAME((const GV *)dstr);
3712 /* The stash may have been detached from the symbol table, so
3714 && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3715 && GvAV((const GV *)sstr)
3719 const STRLEN len = GvNAMELEN(dstr);
3720 if ((len > 1 && name[len-2] == ':' && name[len-1] == ':')
3721 || (len == 1 && name[0] == ':')) {
3724 /* Set aside the old stash, so we can reset isa caches on
3726 if((old_stash = GvHV(dstr)))
3727 /* Make sure we do not lose it early. */
3728 SvREFCNT_inc_simple_void_NN(
3729 sv_2mortal((SV *)old_stash)
3735 gp_free(MUTABLE_GV(dstr));
3736 isGV_with_GP_off(dstr);
3737 (void)SvOK_off(dstr);
3738 isGV_with_GP_on(dstr);
3739 GvINTRO_off(dstr); /* one-shot flag */
3740 GvGP_set(dstr, gp_ref(GvGP(sstr)));
3741 if (SvTAINTED(sstr))
3743 if (GvIMPORTED(dstr) != GVf_IMPORTED
3744 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3746 GvIMPORTED_on(dstr);
3749 if(mro_changes == 2) {
3751 SV * const sref = (SV *)GvAV((const GV *)dstr);
3752 if (SvSMAGICAL(sref) && (mg = mg_find(sref, PERL_MAGIC_isa))) {
3753 if (SvTYPE(mg->mg_obj) != SVt_PVAV) {
3754 AV * const ary = newAV();
3755 av_push(ary, mg->mg_obj); /* takes the refcount */
3756 mg->mg_obj = (SV *)ary;
3758 av_push((AV *)mg->mg_obj, SvREFCNT_inc_simple_NN(dstr));
3760 else sv_magic(sref, dstr, PERL_MAGIC_isa, NULL, 0);
3761 mro_isa_changed_in(GvSTASH(dstr));
3763 else if(mro_changes == 3) {
3764 HV * const stash = GvHV(dstr);
3765 if(old_stash ? (HV *)HvENAME_get(old_stash) : stash)
3771 else if(mro_changes) mro_method_changed_in(GvSTASH(dstr));
3776 S_glob_assign_ref(pTHX_ SV *const dstr, SV *const sstr)
3778 SV * const sref = SvREFCNT_inc(SvRV(sstr));
3780 const int intro = GvINTRO(dstr);
3783 const U32 stype = SvTYPE(sref);
3785 PERL_ARGS_ASSERT_GLOB_ASSIGN_REF;
3788 GvINTRO_off(dstr); /* one-shot flag */
3789 GvLINE(dstr) = CopLINE(PL_curcop);
3790 GvEGV(dstr) = MUTABLE_GV(dstr);
3795 location = (SV **) &(GvGP(dstr)->gp_cv); /* XXX bypassing GvCV_set */
3796 import_flag = GVf_IMPORTED_CV;
3799 location = (SV **) &GvHV(dstr);
3800 import_flag = GVf_IMPORTED_HV;
3803 location = (SV **) &GvAV(dstr);
3804 import_flag = GVf_IMPORTED_AV;
3807 location = (SV **) &GvIOp(dstr);
3810 location = (SV **) &GvFORM(dstr);
3813 location = &GvSV(dstr);
3814 import_flag = GVf_IMPORTED_SV;
3817 if (stype == SVt_PVCV) {
3818 /*if (GvCVGEN(dstr) && (GvCV(dstr) != (const CV *)sref || GvCVGEN(dstr))) {*/
3819 if (GvCVGEN(dstr)) {
3820 SvREFCNT_dec(GvCV(dstr));
3821 GvCV_set(dstr, NULL);
3822 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3825 SAVEGENERICSV(*location);
3829 if (stype == SVt_PVCV && (*location != sref || GvCVGEN(dstr))) {
3830 CV* const cv = MUTABLE_CV(*location);
3832 if (!GvCVGEN((const GV *)dstr) &&
3833 (CvROOT(cv) || CvXSUB(cv)))
3835 /* Redefining a sub - warning is mandatory if
3836 it was a const and its value changed. */
3837 if (CvCONST(cv) && CvCONST((const CV *)sref)
3839 == cv_const_sv((const CV *)sref)) {
3841 /* They are 2 constant subroutines generated from
3842 the same constant. This probably means that
3843 they are really the "same" proxy subroutine
3844 instantiated in 2 places. Most likely this is
3845 when a constant is exported twice. Don't warn.
3848 else if (ckWARN(WARN_REDEFINE)
3850 && (!CvCONST((const CV *)sref)
3851 || sv_cmp(cv_const_sv(cv),
3852 cv_const_sv((const CV *)
3854 Perl_warner(aTHX_ packWARN(WARN_REDEFINE),
3857 ? "Constant subroutine %s::%s redefined"
3858 : "Subroutine %s::%s redefined"),
3859 HvNAME_get(GvSTASH((const GV *)dstr)),
3860 GvENAME(MUTABLE_GV(dstr)));
3864 cv_ckproto_len(cv, (const GV *)dstr,
3865 SvPOK(sref) ? SvPVX_const(sref) : NULL,
3866 SvPOK(sref) ? SvCUR(sref) : 0);
3868 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3869 GvASSUMECV_on(dstr);
3870 if(GvSTASH(dstr)) mro_method_changed_in(GvSTASH(dstr)); /* sub foo { 1 } sub bar { 2 } *bar = \&foo */
3873 if (import_flag && !(GvFLAGS(dstr) & import_flag)
3874 && CopSTASH_ne(PL_curcop, GvSTASH(dstr))) {
3875 GvFLAGS(dstr) |= import_flag;
3877 if (stype == SVt_PVHV) {
3878 const char * const name = GvNAME((GV*)dstr);
3879 const STRLEN len = GvNAMELEN(dstr);
3882 (len > 1 && name[len-2] == ':' && name[len-1] == ':')
3883 || (len == 1 && name[0] == ':')
3885 && (!dref || HvENAME_get(dref))
3888 (HV *)sref, (HV *)dref,
3894 stype == SVt_PVAV && sref != dref
3895 && strEQ(GvNAME((GV*)dstr), "ISA")
3896 /* The stash may have been detached from the symbol table, so
3897 check its name before doing anything. */
3898 && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3901 MAGIC * const omg = dref && SvSMAGICAL(dref)
3902 ? mg_find(dref, PERL_MAGIC_isa)
3904 if (SvSMAGICAL(sref) && (mg = mg_find(sref, PERL_MAGIC_isa))) {
3905 if (SvTYPE(mg->mg_obj) != SVt_PVAV) {
3906 AV * const ary = newAV();
3907 av_push(ary, mg->mg_obj); /* takes the refcount */
3908 mg->mg_obj = (SV *)ary;
3911 if (SvTYPE(omg->mg_obj) == SVt_PVAV) {
3912 SV **svp = AvARRAY((AV *)omg->mg_obj);
3913 I32 items = AvFILLp((AV *)omg->mg_obj) + 1;
3917 SvREFCNT_inc_simple_NN(*svp++)
3923 SvREFCNT_inc_simple_NN(omg->mg_obj)
3927 av_push((AV *)mg->mg_obj,SvREFCNT_inc_simple_NN(dstr));
3932 sref, omg ? omg->mg_obj : dstr, PERL_MAGIC_isa, NULL, 0
3934 mg = mg_find(sref, PERL_MAGIC_isa);
3936 /* Since the *ISA assignment could have affected more than
3937 one stash, don’t call mro_isa_changed_in directly, but let
3938 magic_clearisa do it for us, as it already has the logic for
3939 dealing with globs vs arrays of globs. */
3941 Perl_magic_clearisa(aTHX_ NULL, mg);
3946 if (SvTAINTED(sstr))
3952 Perl_sv_setsv_flags(pTHX_ SV *dstr, register SV* sstr, const I32 flags)
3955 register U32 sflags;
3957 register svtype stype;
3959 PERL_ARGS_ASSERT_SV_SETSV_FLAGS;
3964 if (SvIS_FREED(dstr)) {
3965 Perl_croak(aTHX_ "panic: attempt to copy value %" SVf
3966 " to a freed scalar %p", SVfARG(sstr), (void *)dstr);
3968 SV_CHECK_THINKFIRST_COW_DROP(dstr);
3970 sstr = &PL_sv_undef;
3971 if (SvIS_FREED(sstr)) {
3972 Perl_croak(aTHX_ "panic: attempt to copy freed scalar %p to %p",
3973 (void*)sstr, (void*)dstr);
3975 stype = SvTYPE(sstr);
3976 dtype = SvTYPE(dstr);
3978 (void)SvAMAGIC_off(dstr);
3981 /* need to nuke the magic */
3985 /* There's a lot of redundancy below but we're going for speed here */
3990 if (dtype != SVt_PVGV && dtype != SVt_PVLV) {
3991 (void)SvOK_off(dstr);
3999 sv_upgrade(dstr, SVt_IV);
4003 sv_upgrade(dstr, SVt_PVIV);
4007 goto end_of_first_switch;
4009 (void)SvIOK_only(dstr);
4010 SvIV_set(dstr, SvIVX(sstr));
4013 /* SvTAINTED can only be true if the SV has taint magic, which in
4014 turn means that the SV type is PVMG (or greater). This is the
4015 case statement for SVt_IV, so this cannot be true (whatever gcov
4017 assert(!SvTAINTED(sstr));
4022 if (dtype < SVt_PV && dtype != SVt_IV)
4023 sv_upgrade(dstr, SVt_IV);
4031 sv_upgrade(dstr, SVt_NV);
4035 sv_upgrade(dstr, SVt_PVNV);
4039 goto end_of_first_switch;
4041 SvNV_set(dstr, SvNVX(sstr));
4042 (void)SvNOK_only(dstr);
4043 /* SvTAINTED can only be true if the SV has taint magic, which in
4044 turn means that the SV type is PVMG (or greater). This is the
4045 case statement for SVt_NV, so this cannot be true (whatever gcov
4047 assert(!SvTAINTED(sstr));
4053 #ifdef PERL_OLD_COPY_ON_WRITE
4054 if ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS) {
4055 if (dtype < SVt_PVIV)
4056 sv_upgrade(dstr, SVt_PVIV);
4063 sv_upgrade(dstr, SVt_PV);
4066 if (dtype < SVt_PVIV)
4067 sv_upgrade(dstr, SVt_PVIV);
4070 if (dtype < SVt_PVNV)
4071 sv_upgrade(dstr, SVt_PVNV);
4075 const char * const type = sv_reftype(sstr,0);
4077 Perl_croak(aTHX_ "Bizarre copy of %s in %s", type, OP_DESC(PL_op));
4079 Perl_croak(aTHX_ "Bizarre copy of %s", type);
4084 if (dtype < SVt_REGEXP)
4085 sv_upgrade(dstr, SVt_REGEXP);
4088 /* case SVt_BIND: */
4092 if (SvGMAGICAL(sstr) && (flags & SV_GMAGIC)) {
4094 if (SvTYPE(sstr) != stype)
4095 stype = SvTYPE(sstr);
4097 if (isGV_with_GP(sstr) && dtype <= SVt_PVLV) {
4098 glob_assign_glob(dstr, sstr, dtype);
4101 if (stype == SVt_PVLV)
4102 SvUPGRADE(dstr, SVt_PVNV);
4104 SvUPGRADE(dstr, (svtype)stype);
4106 end_of_first_switch:
4108 /* dstr may have been upgraded. */
4109 dtype = SvTYPE(dstr);
4110 sflags = SvFLAGS(sstr);
4112 if (dtype == SVt_PVCV || dtype == SVt_PVFM) {
4113 /* Assigning to a subroutine sets the prototype. */
4116 const char *const ptr = SvPV_const(sstr, len);
4118 SvGROW(dstr, len + 1);
4119 Copy(ptr, SvPVX(dstr), len + 1, char);
4120 SvCUR_set(dstr, len);
4122 SvFLAGS(dstr) |= sflags & SVf_UTF8;
4126 } else if (dtype == SVt_PVAV || dtype == SVt_PVHV) {
4127 const char * const type = sv_reftype(dstr,0);
4129 Perl_croak(aTHX_ "Cannot copy to %s in %s", type, OP_DESC(PL_op));
4131 Perl_croak(aTHX_ "Cannot copy to %s", type);
4132 } else if (sflags & SVf_ROK) {
4133 if (isGV_with_GP(dstr)
4134 && SvTYPE(SvRV(sstr)) == SVt_PVGV && isGV_with_GP(SvRV(sstr))) {
4137 if (GvIMPORTED(dstr) != GVf_IMPORTED
4138 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
4140 GvIMPORTED_on(dstr);
4145 glob_assign_glob(dstr, sstr, dtype);
4149 if (dtype >= SVt_PV) {
4150 if (isGV_with_GP(dstr)) {
4151 glob_assign_ref(dstr, sstr);
4154 if (SvPVX_const(dstr)) {
4160 (void)SvOK_off(dstr);
4161 SvRV_set(dstr, SvREFCNT_inc(SvRV(sstr)));
4162 SvFLAGS(dstr) |= sflags & SVf_ROK;
4163 assert(!(sflags & SVp_NOK));
4164 assert(!(sflags & SVp_IOK));
4165 assert(!(sflags & SVf_NOK));
4166 assert(!(sflags & SVf_IOK));
4168 else if (isGV_with_GP(dstr)) {
4169 if (!(sflags & SVf_OK)) {
4170 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
4171 "Undefined value assigned to typeglob");
4174 GV *gv = gv_fetchsv(sstr, GV_ADD, SVt_PVGV);
4175 if (dstr != (const SV *)gv) {
4176 const char * const name = GvNAME((const GV *)dstr);
4177 const STRLEN len = GvNAMELEN(dstr);
4178 HV *old_stash = NULL;
4179 bool reset_isa = FALSE;
4180 if ((len > 1 && name[len-2] == ':' && name[len-1] == ':')
4181 || (len == 1 && name[0] == ':')) {
4182 /* Set aside the old stash, so we can reset isa caches
4183 on its subclasses. */
4184 if((old_stash = GvHV(dstr))) {
4185 /* Make sure we do not lose it early. */
4186 SvREFCNT_inc_simple_void_NN(
4187 sv_2mortal((SV *)old_stash)
4194 gp_free(MUTABLE_GV(dstr));
4195 GvGP_set(dstr, gp_ref(GvGP(gv)));
4198 HV * const stash = GvHV(dstr);
4200 old_stash ? (HV *)HvENAME_get(old_stash) : stash
4210 else if (dtype == SVt_REGEXP && stype == SVt_REGEXP) {
4211 reg_temp_copy((REGEXP*)dstr, (REGEXP*)sstr);
4213 else if (sflags & SVp_POK) {
4217 * Check to see if we can just swipe the string. If so, it's a
4218 * possible small lose on short strings, but a big win on long ones.
4219 * It might even be a win on short strings if SvPVX_const(dstr)
4220 * has to be allocated and SvPVX_const(sstr) has to be freed.
4221 * Likewise if we can set up COW rather than doing an actual copy, we
4222 * drop to the else clause, as the swipe code and the COW setup code
4223 * have much in common.
4226 /* Whichever path we take through the next code, we want this true,
4227 and doing it now facilitates the COW check. */
4228 (void)SvPOK_only(dstr);
4231 /* If we're already COW then this clause is not true, and if COW
4232 is allowed then we drop down to the else and make dest COW
4233 with us. If caller hasn't said that we're allowed to COW
4234 shared hash keys then we don't do the COW setup, even if the
4235 source scalar is a shared hash key scalar. */
4236 (((flags & SV_COW_SHARED_HASH_KEYS)
4237 ? (sflags & (SVf_FAKE|SVf_READONLY)) != (SVf_FAKE|SVf_READONLY)
4238 : 1 /* If making a COW copy is forbidden then the behaviour we
4239 desire is as if the source SV isn't actually already
4240 COW, even if it is. So we act as if the source flags
4241 are not COW, rather than actually testing them. */
4243 #ifndef PERL_OLD_COPY_ON_WRITE
4244 /* The change that added SV_COW_SHARED_HASH_KEYS makes the logic
4245 when PERL_OLD_COPY_ON_WRITE is defined a little wrong.
4246 Conceptually PERL_OLD_COPY_ON_WRITE being defined should
4247 override SV_COW_SHARED_HASH_KEYS, because it means "always COW"
4248 but in turn, it's somewhat dead code, never expected to go
4249 live, but more kept as a placeholder on how to do it better
4250 in a newer implementation. */
4251 /* If we are COW and dstr is a suitable target then we drop down
4252 into the else and make dest a COW of us. */
4253 || (SvFLAGS(dstr) & CAN_COW_MASK) != CAN_COW_FLAGS
4258 (sflags & SVs_TEMP) && /* slated for free anyway? */
4259 !(sflags & SVf_OOK) && /* and not involved in OOK hack? */
4260 (!(flags & SV_NOSTEAL)) &&
4261 /* and we're allowed to steal temps */
4262 SvREFCNT(sstr) == 1 && /* and no other references to it? */
4263 SvLEN(sstr)) /* and really is a string */
4264 #ifdef PERL_OLD_COPY_ON_WRITE
4265 && ((flags & SV_COW_SHARED_HASH_KEYS)
4266 ? (!((sflags & CAN_COW_MASK) == CAN_COW_FLAGS
4267 && (SvFLAGS(dstr) & CAN_COW_MASK) == CAN_COW_FLAGS
4268 && SvTYPE(sstr) >= SVt_PVIV && SvTYPE(sstr) != SVt_PVFM))
4272 /* Failed the swipe test, and it's not a shared hash key either.
4273 Have to copy the string. */
4274 STRLEN len = SvCUR(sstr);
4275 SvGROW(dstr, len + 1); /* inlined from sv_setpvn */
4276 Move(SvPVX_const(sstr),SvPVX(dstr),len,char);
4277 SvCUR_set(dstr, len);
4278 *SvEND(dstr) = '\0';
4280 /* If PERL_OLD_COPY_ON_WRITE is not defined, then isSwipe will always
4282 /* Either it's a shared hash key, or it's suitable for
4283 copy-on-write or we can swipe the string. */
4285 PerlIO_printf(Perl_debug_log, "Copy on write: sstr --> dstr\n");
4289 #ifdef PERL_OLD_COPY_ON_WRITE
4291 if ((sflags & (SVf_FAKE | SVf_READONLY))
4292 != (SVf_FAKE | SVf_READONLY)) {
4293 SvREADONLY_on(sstr);
4295 /* Make the source SV into a loop of 1.
4296 (about to become 2) */
4297 SV_COW_NEXT_SV_SET(sstr, sstr);
4301 /* Initial code is common. */
4302 if (SvPVX_const(dstr)) { /* we know that dtype >= SVt_PV */
4307 /* making another shared SV. */
4308 STRLEN cur = SvCUR(sstr);
4309 STRLEN len = SvLEN(sstr);
4310 #ifdef PERL_OLD_COPY_ON_WRITE
4312 assert (SvTYPE(dstr) >= SVt_PVIV);
4313 /* SvIsCOW_normal */
4314 /* splice us in between source and next-after-source. */
4315 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4316 SV_COW_NEXT_SV_SET(sstr, dstr);
4317 SvPV_set(dstr, SvPVX_mutable(sstr));
4321 /* SvIsCOW_shared_hash */
4322 DEBUG_C(PerlIO_printf(Perl_debug_log,
4323 "Copy on write: Sharing hash\n"));
4325 assert (SvTYPE(dstr) >= SVt_PV);
4327 HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)))));
4329 SvLEN_set(dstr, len);
4330 SvCUR_set(dstr, cur);
4331 SvREADONLY_on(dstr);
4335 { /* Passes the swipe test. */
4336 SvPV_set(dstr, SvPVX_mutable(sstr));
4337 SvLEN_set(dstr, SvLEN(sstr));
4338 SvCUR_set(dstr, SvCUR(sstr));
4341 (void)SvOK_off(sstr); /* NOTE: nukes most SvFLAGS on sstr */
4342 SvPV_set(sstr, NULL);
4348 if (sflags & SVp_NOK) {
4349 SvNV_set(dstr, SvNVX(sstr));
4351 if (sflags & SVp_IOK) {
4352 SvIV_set(dstr, SvIVX(sstr));
4353 /* Must do this otherwise some other overloaded use of 0x80000000
4354 gets confused. I guess SVpbm_VALID */
4355 if (sflags & SVf_IVisUV)
4358 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_NOK|SVp_NOK|SVf_UTF8);
4360 const MAGIC * const smg = SvVSTRING_mg(sstr);
4362 sv_magic(dstr, NULL, PERL_MAGIC_vstring,
4363 smg->mg_ptr, smg->mg_len);
4364 SvRMAGICAL_on(dstr);
4368 else if (sflags & (SVp_IOK|SVp_NOK)) {
4369 (void)SvOK_off(dstr);
4370 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_IVisUV|SVf_NOK|SVp_NOK);
4371 if (sflags & SVp_IOK) {
4372 /* XXXX Do we want to set IsUV for IV(ROK)? Be extra safe... */
4373 SvIV_set(dstr, SvIVX(sstr));
4375 if (sflags & SVp_NOK) {
4376 SvNV_set(dstr, SvNVX(sstr));
4380 if (isGV_with_GP(sstr)) {
4381 /* This stringification rule for globs is spread in 3 places.
4382 This feels bad. FIXME. */
4383 const U32 wasfake = sflags & SVf_FAKE;
4385 /* FAKE globs can get coerced, so need to turn this off
4386 temporarily if it is on. */
4388 gv_efullname3(dstr, MUTABLE_GV(sstr), "*");
4389 SvFLAGS(sstr) |= wasfake;
4392 (void)SvOK_off(dstr);
4394 if (SvTAINTED(sstr))
4399 =for apidoc sv_setsv_mg
4401 Like C<sv_setsv>, but also handles 'set' magic.
4407 Perl_sv_setsv_mg(pTHX_ SV *const dstr, register SV *const sstr)
4409 PERL_ARGS_ASSERT_SV_SETSV_MG;
4411 sv_setsv(dstr,sstr);
4415 #ifdef PERL_OLD_COPY_ON_WRITE
4417 Perl_sv_setsv_cow(pTHX_ SV *dstr, SV *sstr)
4419 STRLEN cur = SvCUR(sstr);
4420 STRLEN len = SvLEN(sstr);
4421 register char *new_pv;
4423 PERL_ARGS_ASSERT_SV_SETSV_COW;
4426 PerlIO_printf(Perl_debug_log, "Fast copy on write: %p -> %p\n",
4427 (void*)sstr, (void*)dstr);
4434 if (SvTHINKFIRST(dstr))
4435 sv_force_normal_flags(dstr, SV_COW_DROP_PV);
4436 else if (SvPVX_const(dstr))
4437 Safefree(SvPVX_const(dstr));
4441 SvUPGRADE(dstr, SVt_PVIV);
4443 assert (SvPOK(sstr));
4444 assert (SvPOKp(sstr));
4445 assert (!SvIOK(sstr));
4446 assert (!SvIOKp(sstr));
4447 assert (!SvNOK(sstr));
4448 assert (!SvNOKp(sstr));
4450 if (SvIsCOW(sstr)) {
4452 if (SvLEN(sstr) == 0) {
4453 /* source is a COW shared hash key. */
4454 DEBUG_C(PerlIO_printf(Perl_debug_log,
4455 "Fast copy on write: Sharing hash\n"));
4456 new_pv = HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr))));
4459 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4461 assert ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS);
4462 SvUPGRADE(sstr, SVt_PVIV);
4463 SvREADONLY_on(sstr);
4465 DEBUG_C(PerlIO_printf(Perl_debug_log,
4466 "Fast copy on write: Converting sstr to COW\n"));
4467 SV_COW_NEXT_SV_SET(dstr, sstr);
4469 SV_COW_NEXT_SV_SET(sstr, dstr);
4470 new_pv = SvPVX_mutable(sstr);
4473 SvPV_set(dstr, new_pv);
4474 SvFLAGS(dstr) = (SVt_PVIV|SVf_POK|SVp_POK|SVf_FAKE|SVf_READONLY);
4477 SvLEN_set(dstr, len);
4478 SvCUR_set(dstr, cur);
4487 =for apidoc sv_setpvn
4489 Copies a string into an SV. The C<len> parameter indicates the number of
4490 bytes to be copied. If the C<ptr> argument is NULL the SV will become
4491 undefined. Does not handle 'set' magic. See C<sv_setpvn_mg>.
4497 Perl_sv_setpvn(pTHX_ register SV *const sv, register const char *const ptr, register const STRLEN len)
4500 register char *dptr;
4502 PERL_ARGS_ASSERT_SV_SETPVN;
4504 SV_CHECK_THINKFIRST_COW_DROP(sv);
4510 /* len is STRLEN which is unsigned, need to copy to signed */
4513 Perl_croak(aTHX_ "panic: sv_setpvn called with negative strlen");
4515 SvUPGRADE(sv, SVt_PV);
4517 dptr = SvGROW(sv, len + 1);
4518 Move(ptr,dptr,len,char);
4521 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4526 =for apidoc sv_setpvn_mg
4528 Like C<sv_setpvn>, but also handles 'set' magic.
4534 Perl_sv_setpvn_mg(pTHX_ register SV *const sv, register const char *const ptr, register const STRLEN len)
4536 PERL_ARGS_ASSERT_SV_SETPVN_MG;
4538 sv_setpvn(sv,ptr,len);
4543 =for apidoc sv_setpv
4545 Copies a string into an SV. The string must be null-terminated. Does not
4546 handle 'set' magic. See C<sv_setpv_mg>.
4552 Perl_sv_setpv(pTHX_ register SV *const sv, register const char *const ptr)
4555 register STRLEN len;
4557 PERL_ARGS_ASSERT_SV_SETPV;
4559 SV_CHECK_THINKFIRST_COW_DROP(sv);
4565 SvUPGRADE(sv, SVt_PV);
4567 SvGROW(sv, len + 1);
4568 Move(ptr,SvPVX(sv),len+1,char);
4570 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4575 =for apidoc sv_setpv_mg
4577 Like C<sv_setpv>, but also handles 'set' magic.
4583 Perl_sv_setpv_mg(pTHX_ register SV *const sv, register const char *const ptr)
4585 PERL_ARGS_ASSERT_SV_SETPV_MG;
4592 =for apidoc sv_usepvn_flags
4594 Tells an SV to use C<ptr> to find its string value. Normally the
4595 string is stored inside the SV but sv_usepvn allows the SV to use an
4596 outside string. The C<ptr> should point to memory that was allocated
4597 by C<malloc>. The string length, C<len>, must be supplied. By default
4598 this function will realloc (i.e. move) the memory pointed to by C<ptr>,
4599 so that pointer should not be freed or used by the programmer after
4600 giving it to sv_usepvn, and neither should any pointers from "behind"
4601 that pointer (e.g. ptr + 1) be used.
4603 If C<flags> & SV_SMAGIC is true, will call SvSETMAGIC. If C<flags> &
4604 SV_HAS_TRAILING_NUL is true, then C<ptr[len]> must be NUL, and the realloc
4605 will be skipped. (i.e. the buffer is actually at least 1 byte longer than
4606 C<len>, and already meets the requirements for storing in C<SvPVX>)
4612 Perl_sv_usepvn_flags(pTHX_ SV *const sv, char *ptr, const STRLEN len, const U32 flags)
4617 PERL_ARGS_ASSERT_SV_USEPVN_FLAGS;
4619 SV_CHECK_THINKFIRST_COW_DROP(sv);
4620 SvUPGRADE(sv, SVt_PV);
4623 if (flags & SV_SMAGIC)
4627 if (SvPVX_const(sv))
4631 if (flags & SV_HAS_TRAILING_NUL)
4632 assert(ptr[len] == '\0');
4635 allocate = (flags & SV_HAS_TRAILING_NUL)
4637 #ifdef Perl_safesysmalloc_size
4640 PERL_STRLEN_ROUNDUP(len + 1);
4642 if (flags & SV_HAS_TRAILING_NUL) {
4643 /* It's long enough - do nothing.
4644 Specifically Perl_newCONSTSUB is relying on this. */
4647 /* Force a move to shake out bugs in callers. */
4648 char *new_ptr = (char*)safemalloc(allocate);
4649 Copy(ptr, new_ptr, len, char);
4650 PoisonFree(ptr,len,char);
4654 ptr = (char*) saferealloc (ptr, allocate);
4657 #ifdef Perl_safesysmalloc_size
4658 SvLEN_set(sv, Perl_safesysmalloc_size(ptr));
4660 SvLEN_set(sv, allocate);
4664 if (!(flags & SV_HAS_TRAILING_NUL)) {
4667 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4669 if (flags & SV_SMAGIC)
4673 #ifdef PERL_OLD_COPY_ON_WRITE
4674 /* Need to do this *after* making the SV normal, as we need the buffer
4675 pointer to remain valid until after we've copied it. If we let go too early,
4676 another thread could invalidate it by unsharing last of the same hash key
4677 (which it can do by means other than releasing copy-on-write Svs)
4678 or by changing the other copy-on-write SVs in the loop. */
4680 S_sv_release_COW(pTHX_ register SV *sv, const char *pvx, SV *after)
4682 PERL_ARGS_ASSERT_SV_RELEASE_COW;
4684 { /* this SV was SvIsCOW_normal(sv) */
4685 /* we need to find the SV pointing to us. */
4686 SV *current = SV_COW_NEXT_SV(after);
4688 if (current == sv) {
4689 /* The SV we point to points back to us (there were only two of us
4691 Hence other SV is no longer copy on write either. */
4693 SvREADONLY_off(after);
4695 /* We need to follow the pointers around the loop. */
4697 while ((next = SV_COW_NEXT_SV(current)) != sv) {
4700 /* don't loop forever if the structure is bust, and we have
4701 a pointer into a closed loop. */
4702 assert (current != after);
4703 assert (SvPVX_const(current) == pvx);
4705 /* Make the SV before us point to the SV after us. */
4706 SV_COW_NEXT_SV_SET(current, after);
4712 =for apidoc sv_force_normal_flags
4714 Undo various types of fakery on an SV: if the PV is a shared string, make
4715 a private copy; if we're a ref, stop refing; if we're a glob, downgrade to
4716 an xpvmg; if we're a copy-on-write scalar, this is the on-write time when
4717 we do the copy, and is also used locally. If C<SV_COW_DROP_PV> is set
4718 then a copy-on-write scalar drops its PV buffer (if any) and becomes
4719 SvPOK_off rather than making a copy. (Used where this scalar is about to be
4720 set to some other value.) In addition, the C<flags> parameter gets passed to
4721 C<sv_unref_flags()> when unreffing. C<sv_force_normal> calls this function
4722 with flags set to 0.
4728 Perl_sv_force_normal_flags(pTHX_ register SV *const sv, const U32 flags)
4732 PERL_ARGS_ASSERT_SV_FORCE_NORMAL_FLAGS;
4734 #ifdef PERL_OLD_COPY_ON_WRITE
4735 if (SvREADONLY(sv)) {
4737 const char * const pvx = SvPVX_const(sv);
4738 const STRLEN len = SvLEN(sv);
4739 const STRLEN cur = SvCUR(sv);
4740 /* next COW sv in the loop. If len is 0 then this is a shared-hash
4741 key scalar, so we mustn't attempt to call SV_COW_NEXT_SV(), as
4742 we'll fail an assertion. */
4743 SV * const next = len ? SV_COW_NEXT_SV(sv) : 0;
4746 PerlIO_printf(Perl_debug_log,
4747 "Copy on write: Force normal %ld\n",
4753 /* This SV doesn't own the buffer, so need to Newx() a new one: */
4756 if (flags & SV_COW_DROP_PV) {
4757 /* OK, so we don't need to copy our buffer. */
4760 SvGROW(sv, cur + 1);
4761 Move(pvx,SvPVX(sv),cur,char);
4766 sv_release_COW(sv, pvx, next);
4768 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4774 else if (IN_PERL_RUNTIME)
4775 Perl_croak_no_modify(aTHX);
4778 if (SvREADONLY(sv)) {
4780 const char * const pvx = SvPVX_const(sv);
4781 const STRLEN len = SvCUR(sv);
4786 SvGROW(sv, len + 1);
4787 Move(pvx,SvPVX(sv),len,char);
4789 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4791 else if (IN_PERL_RUNTIME)
4792 Perl_croak_no_modify(aTHX);
4796 sv_unref_flags(sv, flags);
4797 else if (SvFAKE(sv) && isGV_with_GP(sv))
4799 else if (SvFAKE(sv) && SvTYPE(sv) == SVt_REGEXP) {
4800 /* Need to downgrade the REGEXP to a simple(r) scalar. This is analogous
4801 to sv_unglob. We only need it here, so inline it. */
4802 const svtype new_type = SvMAGIC(sv) || SvSTASH(sv) ? SVt_PVMG : SVt_PV;
4803 SV *const temp = newSV_type(new_type);
4804 void *const temp_p = SvANY(sv);
4806 if (new_type == SVt_PVMG) {
4807 SvMAGIC_set(temp, SvMAGIC(sv));
4808 SvMAGIC_set(sv, NULL);
4809 SvSTASH_set(temp, SvSTASH(sv));
4810 SvSTASH_set(sv, NULL);
4812 SvCUR_set(temp, SvCUR(sv));
4813 /* Remember that SvPVX is in the head, not the body. */
4815 SvLEN_set(temp, SvLEN(sv));
4816 /* This signals "buffer is owned by someone else" in sv_clear,
4817 which is the least effort way to stop it freeing the buffer.
4819 SvLEN_set(sv, SvLEN(sv)+1);
4821 /* Their buffer is already owned by someone else. */
4822 SvPVX(sv) = savepvn(SvPVX(sv), SvCUR(sv));
4823 SvLEN_set(temp, SvCUR(sv)+1);
4826 /* Now swap the rest of the bodies. */
4828 SvFLAGS(sv) &= ~(SVf_FAKE|SVTYPEMASK);
4829 SvFLAGS(sv) |= new_type;
4830 SvANY(sv) = SvANY(temp);
4832 SvFLAGS(temp) &= ~(SVTYPEMASK);
4833 SvFLAGS(temp) |= SVt_REGEXP|SVf_FAKE;
4834 SvANY(temp) = temp_p;
4843 Efficient removal of characters from the beginning of the string buffer.
4844 SvPOK(sv) must be true and the C<ptr> must be a pointer to somewhere inside
4845 the string buffer. The C<ptr> becomes the first character of the adjusted
4846 string. Uses the "OOK hack".
4847 Beware: after this function returns, C<ptr> and SvPVX_const(sv) may no longer
4848 refer to the same chunk of data.
4854 Perl_sv_chop(pTHX_ register SV *const sv, register const char *const ptr)
4860 const U8 *real_start;
4864 PERL_ARGS_ASSERT_SV_CHOP;
4866 if (!ptr || !SvPOKp(sv))
4868 delta = ptr - SvPVX_const(sv);
4870 /* Nothing to do. */
4873 /* SvPVX(sv) may move in SV_CHECK_THINKFIRST(sv), but after this line,
4874 nothing uses the value of ptr any more. */
4875 max_delta = SvLEN(sv) ? SvLEN(sv) : SvCUR(sv);
4876 if (ptr <= SvPVX_const(sv))
4877 Perl_croak(aTHX_ "panic: sv_chop ptr=%p, start=%p, end=%p",
4878 ptr, SvPVX_const(sv), SvPVX_const(sv) + max_delta);
4879 SV_CHECK_THINKFIRST(sv);
4880 if (delta > max_delta)
4881 Perl_croak(aTHX_ "panic: sv_chop ptr=%p (was %p), start=%p, end=%p",
4882 SvPVX_const(sv) + delta, ptr, SvPVX_const(sv),
4883 SvPVX_const(sv) + max_delta);
4886 if (!SvLEN(sv)) { /* make copy of shared string */
4887 const char *pvx = SvPVX_const(sv);
4888 const STRLEN len = SvCUR(sv);
4889 SvGROW(sv, len + 1);
4890 Move(pvx,SvPVX(sv),len,char);
4893 SvFLAGS(sv) |= SVf_OOK;
4896 SvOOK_offset(sv, old_delta);
4898 SvLEN_set(sv, SvLEN(sv) - delta);
4899 SvCUR_set(sv, SvCUR(sv) - delta);
4900 SvPV_set(sv, SvPVX(sv) + delta);
4902 p = (U8 *)SvPVX_const(sv);
4907 real_start = p - delta;
4911 if (delta < 0x100) {
4915 p -= sizeof(STRLEN);
4916 Copy((U8*)&delta, p, sizeof(STRLEN), U8);
4920 /* Fill the preceding buffer with sentinals to verify that no-one is
4922 while (p > real_start) {
4930 =for apidoc sv_catpvn
4932 Concatenates the string onto the end of the string which is in the SV. The
4933 C<len> indicates number of bytes to copy. If the SV has the UTF-8
4934 status set, then the bytes appended should be valid UTF-8.
4935 Handles 'get' magic, but not 'set' magic. See C<sv_catpvn_mg>.
4937 =for apidoc sv_catpvn_flags
4939 Concatenates the string onto the end of the string which is in the SV. The
4940 C<len> indicates number of bytes to copy. If the SV has the UTF-8
4941 status set, then the bytes appended should be valid UTF-8.
4942 If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<dsv> if
4943 appropriate, else not. C<sv_catpvn> and C<sv_catpvn_nomg> are implemented
4944 in terms of this function.
4950 Perl_sv_catpvn_flags(pTHX_ register SV *const dsv, register const char *sstr, register const STRLEN slen, const I32 flags)
4954 const char * const dstr = SvPV_force_flags(dsv, dlen, flags);
4956 PERL_ARGS_ASSERT_SV_CATPVN_FLAGS;
4958 SvGROW(dsv, dlen + slen + 1);
4960 sstr = SvPVX_const(dsv);
4961 Move(sstr, SvPVX(dsv) + dlen, slen, char);
4962 SvCUR_set(dsv, SvCUR(dsv) + slen);
4964 (void)SvPOK_only_UTF8(dsv); /* validate pointer */
4966 if (flags & SV_SMAGIC)
4971 =for apidoc sv_catsv
4973 Concatenates the string from SV C<ssv> onto the end of the string in
4974 SV C<dsv>. Modifies C<dsv> but not C<ssv>. Handles 'get' magic, but
4975 not 'set' magic. See C<sv_catsv_mg>.
4977 =for apidoc sv_catsv_flags
4979 Concatenates the string from SV C<ssv> onto the end of the string in
4980 SV C<dsv>. Modifies C<dsv> but not C<ssv>. If C<flags> has C<SV_GMAGIC>
4981 bit set, will C<mg_get> on the SVs if appropriate, else not. C<sv_catsv>
4982 and C<sv_catsv_nomg> are implemented in terms of this function.
4987 Perl_sv_catsv_flags(pTHX_ SV *const dsv, register SV *const ssv, const I32 flags)
4991 PERL_ARGS_ASSERT_SV_CATSV_FLAGS;
4995 const char *spv = SvPV_flags_const(ssv, slen, flags);
4997 /* sutf8 and dutf8 were type bool, but under USE_ITHREADS,
4998 gcc version 2.95.2 20000220 (Debian GNU/Linux) for
4999 Linux xxx 2.2.17 on sparc64 with gcc -O2, we erroneously
5000 get dutf8 = 0x20000000, (i.e. SVf_UTF8) even though
5001 dsv->sv_flags doesn't have that bit set.
5002 Andy Dougherty 12 Oct 2001
5004 const I32 sutf8 = DO_UTF8(ssv);
5007 if (SvGMAGICAL(dsv) && (flags & SV_GMAGIC))
5009 dutf8 = DO_UTF8(dsv);
5011 if (dutf8 != sutf8) {
5013 /* Not modifying source SV, so taking a temporary copy. */
5014 SV* const csv = newSVpvn_flags(spv, slen, SVs_TEMP);
5016 sv_utf8_upgrade(csv);
5017 spv = SvPV_const(csv, slen);
5020 /* Leave enough space for the cat that's about to happen */
5021 sv_utf8_upgrade_flags_grow(dsv, 0, slen);
5023 sv_catpvn_nomg(dsv, spv, slen);
5026 if (flags & SV_SMAGIC)
5031 =for apidoc sv_catpv
5033 Concatenates the string onto the end of the string which is in the SV.
5034 If the SV has the UTF-8 status set, then the bytes appended should be
5035 valid UTF-8. Handles 'get' magic, but not 'set' magic. See C<sv_catpv_mg>.
5040 Perl_sv_catpv(pTHX_ register SV *const sv, register const char *ptr)
5043 register STRLEN len;
5047 PERL_ARGS_ASSERT_SV_CATPV;
5051 junk = SvPV_force(sv, tlen);
5053 SvGROW(sv, tlen + len + 1);
5055 ptr = SvPVX_const(sv);
5056 Move(ptr,SvPVX(sv)+tlen,len+1,char);
5057 SvCUR_set(sv, SvCUR(sv) + len);
5058 (void)SvPOK_only_UTF8(sv); /* validate pointer */
5063 =for apidoc sv_catpv_flags
5065 Concatenates the string onto the end of the string which is in the SV.
5066 If the SV has the UTF-8 status set, then the bytes appended should
5067 be valid UTF-8. If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get>
5068 on the SVs if appropriate, else not.
5074 Perl_sv_catpv_flags(pTHX_ SV *dstr, const char *sstr, const I32 flags)
5076 PERL_ARGS_ASSERT_SV_CATPV_FLAGS;
5077 sv_catpvn_flags(dstr, sstr, strlen(sstr), flags);
5081 =for apidoc sv_catpv_mg
5083 Like C<sv_catpv>, but also handles 'set' magic.
5089 Perl_sv_catpv_mg(pTHX_ register SV *const sv, register const char *const ptr)
5091 PERL_ARGS_ASSERT_SV_CATPV_MG;
5100 Creates a new SV. A non-zero C<len> parameter indicates the number of
5101 bytes of preallocated string space the SV should have. An extra byte for a
5102 trailing NUL is also reserved. (SvPOK is not set for the SV even if string
5103 space is allocated.) The reference count for the new SV is set to 1.
5105 In 5.9.3, newSV() replaces the older NEWSV() API, and drops the first
5106 parameter, I<x>, a debug aid which allowed callers to identify themselves.
5107 This aid has been superseded by a new build option, PERL_MEM_LOG (see
5108 L<perlhacktips/PERL_MEM_LOG>). The older API is still there for use in XS
5109 modules supporting older perls.
5115 Perl_newSV(pTHX_ const STRLEN len)
5122 sv_upgrade(sv, SVt_PV);
5123 SvGROW(sv, len + 1);
5128 =for apidoc sv_magicext
5130 Adds magic to an SV, upgrading it if necessary. Applies the
5131 supplied vtable and returns a pointer to the magic added.
5133 Note that C<sv_magicext> will allow things that C<sv_magic> will not.
5134 In particular, you can add magic to SvREADONLY SVs, and add more than
5135 one instance of the same 'how'.
5137 If C<namlen> is greater than zero then a C<savepvn> I<copy> of C<name> is
5138 stored, if C<namlen> is zero then C<name> is stored as-is and - as another
5139 special case - if C<(name && namlen == HEf_SVKEY)> then C<name> is assumed
5140 to contain an C<SV*> and is stored as-is with its REFCNT incremented.
5142 (This is now used as a subroutine by C<sv_magic>.)
5147 Perl_sv_magicext(pTHX_ SV *const sv, SV *const obj, const int how,
5148 const MGVTBL *const vtable, const char *const name, const I32 namlen)
5153 PERL_ARGS_ASSERT_SV_MAGICEXT;
5155 SvUPGRADE(sv, SVt_PVMG);
5156 Newxz(mg, 1, MAGIC);
5157 mg->mg_moremagic = SvMAGIC(sv);
5158 SvMAGIC_set(sv, mg);
5160 /* Sometimes a magic contains a reference loop, where the sv and
5161 object refer to each other. To prevent a reference loop that
5162 would prevent such objects being freed, we look for such loops
5163 and if we find one we avoid incrementing the object refcount.
5165 Note we cannot do this to avoid self-tie loops as intervening RV must
5166 have its REFCNT incremented to keep it in existence.
5169 if (!obj || obj == sv ||
5170 how == PERL_MAGIC_arylen ||
5171 how == PERL_MAGIC_symtab ||
5172 (SvTYPE(obj) == SVt_PVGV &&
5173 (GvSV(obj) == sv || GvHV(obj) == (const HV *)sv
5174 || GvAV(obj) == (const AV *)sv || GvCV(obj) == (const CV *)sv
5175 || GvIOp(obj) == (const IO *)sv || GvFORM(obj) == (const CV *)sv)))
5180 mg->mg_obj = SvREFCNT_inc_simple(obj);
5181 mg->mg_flags |= MGf_REFCOUNTED;
5184 /* Normal self-ties simply pass a null object, and instead of
5185 using mg_obj directly, use the SvTIED_obj macro to produce a
5186 new RV as needed. For glob "self-ties", we are tieing the PVIO
5187 with an RV obj pointing to the glob containing the PVIO. In
5188 this case, to avoid a reference loop, we need to weaken the
5192 if (how == PERL_MAGIC_tiedscalar && SvTYPE(sv) == SVt_PVIO &&
5193 obj && SvROK(obj) && GvIO(SvRV(obj)) == (const IO *)sv)
5199 mg->mg_len = namlen;
5202 mg->mg_ptr = savepvn(name, namlen);
5203 else if (namlen == HEf_SVKEY) {
5204 /* Yes, this is casting away const. This is only for the case of
5205 HEf_SVKEY. I think we need to document this aberation of the
5206 constness of the API, rather than making name non-const, as
5207 that change propagating outwards a long way. */
5208 mg->mg_ptr = (char*)SvREFCNT_inc_simple_NN((SV *)name);
5210 mg->mg_ptr = (char *) name;
5212 mg->mg_virtual = (MGVTBL *) vtable;
5216 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
5221 =for apidoc sv_magic
5223 Adds magic to an SV. First upgrades C<sv> to type C<SVt_PVMG> if necessary,
5224 then adds a new magic item of type C<how> to the head of the magic list.
5226 See C<sv_magicext> (which C<sv_magic> now calls) for a description of the
5227 handling of the C<name> and C<namlen> arguments.
5229 You need to use C<sv_magicext> to add magic to SvREADONLY SVs and also
5230 to add more than one instance of the same 'how'.
5236 Perl_sv_magic(pTHX_ register SV *const sv, SV *const obj, const int how,
5237 const char *const name, const I32 namlen)
5240 const MGVTBL *vtable;
5243 unsigned int vtable_index;
5245 PERL_ARGS_ASSERT_SV_MAGIC;
5247 if (how < 0 || how > C_ARRAY_LENGTH(PL_magic_data)
5248 || ((flags = PL_magic_data[how]),
5249 (vtable_index = flags & PERL_MAGIC_VTABLE_MASK)
5250 > magic_vtable_max))
5251 Perl_croak(aTHX_ "Don't know how to handle magic of type \\%o", how);
5253 /* PERL_MAGIC_ext is reserved for use by extensions not perl internals.
5254 Useful for attaching extension internal data to perl vars.
5255 Note that multiple extensions may clash if magical scalars
5256 etc holding private data from one are passed to another. */
5258 vtable = (vtable_index == magic_vtable_max)
5259 ? NULL : PL_magic_vtables + vtable_index;
5261 #ifdef PERL_OLD_COPY_ON_WRITE
5263 sv_force_normal_flags(sv, 0);
5265 if (SvREADONLY(sv)) {
5267 /* its okay to attach magic to shared strings; the subsequent
5268 * upgrade to PVMG will unshare the string */
5269 !(SvFAKE(sv) && SvTYPE(sv) < SVt_PVMG)
5272 && !PERL_MAGIC_TYPE_READONLY_ACCEPTABLE(how)
5275 Perl_croak_no_modify(aTHX);
5278 if (SvMAGICAL(sv) || (how == PERL_MAGIC_taint && SvTYPE(sv) >= SVt_PVMG)) {
5279 if (SvMAGIC(sv) && (mg = mg_find(sv, how))) {
5280 /* sv_magic() refuses to add a magic of the same 'how' as an
5283 if (how == PERL_MAGIC_taint) {
5285 /* Any scalar which already had taint magic on which someone
5286 (erroneously?) did SvIOK_on() or similar will now be
5287 incorrectly sporting public "OK" flags. */
5288 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
5294 /* Rest of work is done else where */
5295 mg = sv_magicext(sv,obj,how,vtable,name,namlen);
5298 case PERL_MAGIC_taint:
5301 case PERL_MAGIC_ext:
5302 case PERL_MAGIC_dbfile:
5309 S_sv_unmagicext_flags(pTHX_ SV *const sv, const int type, MGVTBL *vtbl, const U32 flags)
5316 if (SvTYPE(sv) < SVt_PVMG || !SvMAGIC(sv))
5318 mgp = &(((XPVMG*) SvANY(sv))->xmg_u.xmg_magic);
5319 for (mg = *mgp; mg; mg = *mgp) {
5320 const MGVTBL* const virt = mg->mg_virtual;
5321 if (mg->mg_type == type && (!flags || virt == vtbl)) {
5322 *mgp = mg->mg_moremagic;
5323 if (virt && virt->svt_free)
5324 virt->svt_free(aTHX_ sv, mg);
5325 if (mg->mg_ptr && mg->mg_type != PERL_MAGIC_regex_global) {
5327 Safefree(mg->mg_ptr);
5328 else if (mg->mg_len == HEf_SVKEY)
5329 SvREFCNT_dec(MUTABLE_SV(mg->mg_ptr));
5330 else if (mg->mg_type == PERL_MAGIC_utf8)
5331 Safefree(mg->mg_ptr);
5333 if (mg->mg_flags & MGf_REFCOUNTED)
5334 SvREFCNT_dec(mg->mg_obj);
5338 mgp = &mg->mg_moremagic;
5341 if (SvMAGICAL(sv)) /* if we're under save_magic, wait for restore_magic; */
5342 mg_magical(sv); /* else fix the flags now */
5346 SvFLAGS(sv) |= (SvFLAGS(sv) & (SVp_IOK|SVp_NOK|SVp_POK)) >> PRIVSHIFT;
5352 =for apidoc sv_unmagic
5354 Removes all magic of type C<type> from an SV.
5360 Perl_sv_unmagic(pTHX_ SV *const sv, const int type)
5362 PERL_ARGS_ASSERT_SV_UNMAGIC;
5363 return S_sv_unmagicext_flags(aTHX_ sv, type, NULL, 0);
5367 =for apidoc sv_unmagicext
5369 Removes all magic of type C<type> with the specified C<vtbl> from an SV.
5375 Perl_sv_unmagicext(pTHX_ SV *const sv, const int type, MGVTBL *vtbl)
5377 PERL_ARGS_ASSERT_SV_UNMAGICEXT;
5378 return S_sv_unmagicext_flags(aTHX_ sv, type, vtbl, 1);
5382 =for apidoc sv_rvweaken
5384 Weaken a reference: set the C<SvWEAKREF> flag on this RV; give the
5385 referred-to SV C<PERL_MAGIC_backref> magic if it hasn't already; and
5386 push a back-reference to this RV onto the array of backreferences
5387 associated with that magic. If the RV is magical, set magic will be
5388 called after the RV is cleared.
5394 Perl_sv_rvweaken(pTHX_ SV *const sv)
5398 PERL_ARGS_ASSERT_SV_RVWEAKEN;
5400 if (!SvOK(sv)) /* let undefs pass */
5403 Perl_croak(aTHX_ "Can't weaken a nonreference");
5404 else if (SvWEAKREF(sv)) {
5405 Perl_ck_warner(aTHX_ packWARN(WARN_MISC), "Reference is already weak");
5409 Perl_sv_add_backref(aTHX_ tsv, sv);
5415 /* Give tsv backref magic if it hasn't already got it, then push a
5416 * back-reference to sv onto the array associated with the backref magic.
5418 * As an optimisation, if there's only one backref and it's not an AV,
5419 * store it directly in the HvAUX or mg_obj slot, avoiding the need to
5420 * allocate an AV. (Whether the slot holds an AV tells us whether this is
5424 /* A discussion about the backreferences array and its refcount:
5426 * The AV holding the backreferences is pointed to either as the mg_obj of
5427 * PERL_MAGIC_backref, or in the specific case of a HV, from the
5428 * xhv_backreferences field. The array is created with a refcount
5429 * of 2. This means that if during global destruction the array gets
5430 * picked on before its parent to have its refcount decremented by the
5431 * random zapper, it won't actually be freed, meaning it's still there for
5432 * when its parent gets freed.
5434 * When the parent SV is freed, the extra ref is killed by
5435 * Perl_sv_kill_backrefs. The other ref is killed, in the case of magic,
5436 * by mg_free() / MGf_REFCOUNTED, or for a hash, by Perl_hv_kill_backrefs.
5438 * When a single backref SV is stored directly, it is not reference
5443 Perl_sv_add_backref(pTHX_ SV *const tsv, SV *const sv)
5450 PERL_ARGS_ASSERT_SV_ADD_BACKREF;
5452 /* find slot to store array or singleton backref */
5454 if (SvTYPE(tsv) == SVt_PVHV) {
5455 svp = (SV**)Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5458 (SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL))))
5460 sv_magic(tsv, NULL, PERL_MAGIC_backref, NULL, 0);
5461 mg = mg_find(tsv, PERL_MAGIC_backref);
5463 svp = &(mg->mg_obj);
5466 /* create or retrieve the array */
5468 if ( (!*svp && SvTYPE(sv) == SVt_PVAV)
5469 || (*svp && SvTYPE(*svp) != SVt_PVAV)
5474 SvREFCNT_inc_simple_void(av);
5475 /* av now has a refcnt of 2; see discussion above */
5477 /* move single existing backref to the array */
5479 AvARRAY(av)[++AvFILLp(av)] = *svp; /* av_push() */
5483 mg->mg_flags |= MGf_REFCOUNTED;
5486 av = MUTABLE_AV(*svp);
5489 /* optimisation: store single backref directly in HvAUX or mg_obj */
5493 /* push new backref */
5494 assert(SvTYPE(av) == SVt_PVAV);
5495 if (AvFILLp(av) >= AvMAX(av)) {
5496 av_extend(av, AvFILLp(av)+1);
5498 AvARRAY(av)[++AvFILLp(av)] = sv; /* av_push() */
5501 /* delete a back-reference to ourselves from the backref magic associated
5502 * with the SV we point to.
5506 Perl_sv_del_backref(pTHX_ SV *const tsv, SV *const sv)
5511 PERL_ARGS_ASSERT_SV_DEL_BACKREF;
5513 if (SvTYPE(tsv) == SVt_PVHV) {
5515 svp = (SV**)Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5519 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
5520 svp = mg ? &(mg->mg_obj) : NULL;
5524 Perl_croak(aTHX_ "panic: del_backref");
5526 if (SvTYPE(*svp) == SVt_PVAV) {
5530 AV * const av = (AV*)*svp;
5532 assert(!SvIS_FREED(av));
5536 /* for an SV with N weak references to it, if all those
5537 * weak refs are deleted, then sv_del_backref will be called
5538 * N times and O(N^2) compares will be done within the backref
5539 * array. To ameliorate this potential slowness, we:
5540 * 1) make sure this code is as tight as possible;
5541 * 2) when looking for SV, look for it at both the head and tail of the
5542 * array first before searching the rest, since some create/destroy
5543 * patterns will cause the backrefs to be freed in order.
5550 SV **p = &svp[fill];
5551 SV *const topsv = *p;
5558 /* We weren't the last entry.
5559 An unordered list has this property that you
5560 can take the last element off the end to fill
5561 the hole, and it's still an unordered list :-)
5567 break; /* should only be one */
5574 AvFILLp(av) = fill-1;
5577 /* optimisation: only a single backref, stored directly */
5579 Perl_croak(aTHX_ "panic: del_backref");
5586 Perl_sv_kill_backrefs(pTHX_ SV *const sv, AV *const av)
5592 PERL_ARGS_ASSERT_SV_KILL_BACKREFS;
5597 /* after multiple passes through Perl_sv_clean_all() for a thinngy
5598 * that has badly leaked, the backref array may have gotten freed,
5599 * since we only protect it against 1 round of cleanup */
5600 if (SvIS_FREED(av)) {
5601 if (PL_in_clean_all) /* All is fair */
5604 "panic: magic_killbackrefs (freed backref AV/SV)");
5608 is_array = (SvTYPE(av) == SVt_PVAV);
5610 assert(!SvIS_FREED(av));
5613 last = svp + AvFILLp(av);
5616 /* optimisation: only a single backref, stored directly */
5622 while (svp <= last) {
5624 SV *const referrer = *svp;
5625 if (SvWEAKREF(referrer)) {
5626 /* XXX Should we check that it hasn't changed? */
5627 assert(SvROK(referrer));
5628 SvRV_set(referrer, 0);
5630 SvWEAKREF_off(referrer);
5631 SvSETMAGIC(referrer);
5632 } else if (SvTYPE(referrer) == SVt_PVGV ||
5633 SvTYPE(referrer) == SVt_PVLV) {
5634 assert(SvTYPE(sv) == SVt_PVHV); /* stash backref */
5635 /* You lookin' at me? */
5636 assert(GvSTASH(referrer));
5637 assert(GvSTASH(referrer) == (const HV *)sv);
5638 GvSTASH(referrer) = 0;
5639 } else if (SvTYPE(referrer) == SVt_PVCV ||
5640 SvTYPE(referrer) == SVt_PVFM) {
5641 if (SvTYPE(sv) == SVt_PVHV) { /* stash backref */
5642 /* You lookin' at me? */
5643 assert(CvSTASH(referrer));
5644 assert(CvSTASH(referrer) == (const HV *)sv);
5645 SvANY(MUTABLE_CV(referrer))->xcv_stash = 0;
5648 assert(SvTYPE(sv) == SVt_PVGV);
5649 /* You lookin' at me? */
5650 assert(CvGV(referrer));
5651 assert(CvGV(referrer) == (const GV *)sv);
5652 anonymise_cv_maybe(MUTABLE_GV(sv),
5653 MUTABLE_CV(referrer));
5658 "panic: magic_killbackrefs (flags=%"UVxf")",
5659 (UV)SvFLAGS(referrer));
5670 SvREFCNT_dec(av); /* remove extra count added by sv_add_backref() */
5676 =for apidoc sv_insert
5678 Inserts a string at the specified offset/length within the SV. Similar to
5679 the Perl substr() function. Handles get magic.
5681 =for apidoc sv_insert_flags
5683 Same as C<sv_insert>, but the extra C<flags> are passed the C<SvPV_force_flags> that applies to C<bigstr>.
5689 Perl_sv_insert_flags(pTHX_ SV *const bigstr, const STRLEN offset, const STRLEN len, const char *const little, const STRLEN littlelen, const U32 flags)
5694 register char *midend;
5695 register char *bigend;
5699 PERL_ARGS_ASSERT_SV_INSERT_FLAGS;
5702 Perl_croak(aTHX_ "Can't modify non-existent substring");
5703 SvPV_force_flags(bigstr, curlen, flags);
5704 (void)SvPOK_only_UTF8(bigstr);
5705 if (offset + len > curlen) {
5706 SvGROW(bigstr, offset+len+1);
5707 Zero(SvPVX(bigstr)+curlen, offset+len-curlen, char);
5708 SvCUR_set(bigstr, offset+len);
5712 i = littlelen - len;
5713 if (i > 0) { /* string might grow */
5714 big = SvGROW(bigstr, SvCUR(bigstr) + i + 1);
5715 mid = big + offset + len;
5716 midend = bigend = big + SvCUR(bigstr);
5719 while (midend > mid) /* shove everything down */
5720 *--bigend = *--midend;
5721 Move(little,big+offset,littlelen,char);
5722 SvCUR_set(bigstr, SvCUR(bigstr) + i);
5727 Move(little,SvPVX(bigstr)+offset,len,char);
5732 big = SvPVX(bigstr);
5735 bigend = big + SvCUR(bigstr);
5737 if (midend > bigend)
5738 Perl_croak(aTHX_ "panic: sv_insert");
5740 if (mid - big > bigend - midend) { /* faster to shorten from end */
5742 Move(little, mid, littlelen,char);
5745 i = bigend - midend;
5747 Move(midend, mid, i,char);
5751 SvCUR_set(bigstr, mid - big);
5753 else if ((i = mid - big)) { /* faster from front */
5754 midend -= littlelen;
5756 Move(big, midend - i, i, char);
5757 sv_chop(bigstr,midend-i);
5759 Move(little, mid, littlelen,char);
5761 else if (littlelen) {
5762 midend -= littlelen;
5763 sv_chop(bigstr,midend);
5764 Move(little,midend,littlelen,char);
5767 sv_chop(bigstr,midend);
5773 =for apidoc sv_replace
5775 Make the first argument a copy of the second, then delete the original.
5776 The target SV physically takes over ownership of the body of the source SV
5777 and inherits its flags; however, the target keeps any magic it owns,
5778 and any magic in the source is discarded.
5779 Note that this is a rather specialist SV copying operation; most of the
5780 time you'll want to use C<sv_setsv> or one of its many macro front-ends.
5786 Perl_sv_replace(pTHX_ register SV *const sv, register SV *const nsv)
5789 const U32 refcnt = SvREFCNT(sv);
5791 PERL_ARGS_ASSERT_SV_REPLACE;
5793 SV_CHECK_THINKFIRST_COW_DROP(sv);
5794 if (SvREFCNT(nsv) != 1) {
5795 Perl_croak(aTHX_ "panic: reference miscount on nsv in sv_replace()"
5796 " (%" UVuf " != 1)", (UV) SvREFCNT(nsv));
5798 if (SvMAGICAL(sv)) {
5802 sv_upgrade(nsv, SVt_PVMG);
5803 SvMAGIC_set(nsv, SvMAGIC(sv));
5804 SvFLAGS(nsv) |= SvMAGICAL(sv);
5806 SvMAGIC_set(sv, NULL);
5810 assert(!SvREFCNT(sv));
5811 #ifdef DEBUG_LEAKING_SCALARS
5812 sv->sv_flags = nsv->sv_flags;
5813 sv->sv_any = nsv->sv_any;
5814 sv->sv_refcnt = nsv->sv_refcnt;
5815 sv->sv_u = nsv->sv_u;
5817 StructCopy(nsv,sv,SV);
5819 if(SvTYPE(sv) == SVt_IV) {
5821 = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
5825 #ifdef PERL_OLD_COPY_ON_WRITE
5826 if (SvIsCOW_normal(nsv)) {
5827 /* We need to follow the pointers around the loop to make the
5828 previous SV point to sv, rather than nsv. */
5831 while ((next = SV_COW_NEXT_SV(current)) != nsv) {
5834 assert(SvPVX_const(current) == SvPVX_const(nsv));
5836 /* Make the SV before us point to the SV after us. */
5838 PerlIO_printf(Perl_debug_log, "previous is\n");
5840 PerlIO_printf(Perl_debug_log,
5841 "move it from 0x%"UVxf" to 0x%"UVxf"\n",
5842 (UV) SV_COW_NEXT_SV(current), (UV) sv);
5844 SV_COW_NEXT_SV_SET(current, sv);
5847 SvREFCNT(sv) = refcnt;
5848 SvFLAGS(nsv) |= SVTYPEMASK; /* Mark as freed */
5853 /* We're about to free a GV which has a CV that refers back to us.
5854 * If that CV will outlive us, make it anonymous (i.e. fix up its CvGV
5858 S_anonymise_cv_maybe(pTHX_ GV *gv, CV* cv)
5864 PERL_ARGS_ASSERT_ANONYMISE_CV_MAYBE;
5867 assert(SvREFCNT(gv) == 0);
5868 assert(isGV(gv) && isGV_with_GP(gv));
5870 assert(!CvANON(cv));
5871 assert(CvGV(cv) == gv);
5873 /* will the CV shortly be freed by gp_free() ? */
5874 if (GvCV(gv) == cv && GvGP(gv)->gp_refcnt < 2 && SvREFCNT(cv) < 2) {
5875 SvANY(cv)->xcv_gv = NULL;
5879 /* if not, anonymise: */
5880 stash = GvSTASH(gv) && HvNAME(GvSTASH(gv))
5881 ? HvENAME(GvSTASH(gv)) : NULL;
5882 gvname = Perl_newSVpvf(aTHX_ "%s::__ANON__",
5883 stash ? stash : "__ANON__");
5884 anongv = gv_fetchsv(gvname, GV_ADDMULTI, SVt_PVCV);
5885 SvREFCNT_dec(gvname);
5889 SvANY(cv)->xcv_gv = MUTABLE_GV(SvREFCNT_inc(anongv));
5894 =for apidoc sv_clear
5896 Clear an SV: call any destructors, free up any memory used by the body,
5897 and free the body itself. The SV's head is I<not> freed, although
5898 its type is set to all 1's so that it won't inadvertently be assumed
5899 to be live during global destruction etc.
5900 This function should only be called when REFCNT is zero. Most of the time
5901 you'll want to call C<sv_free()> (or its macro wrapper C<SvREFCNT_dec>)
5908 Perl_sv_clear(pTHX_ SV *const orig_sv)
5913 const struct body_details *sv_type_details;
5916 register SV *sv = orig_sv;
5919 PERL_ARGS_ASSERT_SV_CLEAR;
5921 /* within this loop, sv is the SV currently being freed, and
5922 * iter_sv is the most recent AV or whatever that's being iterated
5923 * over to provide more SVs */
5929 assert(SvREFCNT(sv) == 0);
5930 assert(SvTYPE(sv) != SVTYPEMASK);
5932 if (type <= SVt_IV) {
5933 /* See the comment in sv.h about the collusion between this
5934 * early return and the overloading of the NULL slots in the
5938 SvFLAGS(sv) &= SVf_BREAK;
5939 SvFLAGS(sv) |= SVTYPEMASK;
5943 assert(!SvOBJECT(sv) || type >= SVt_PVMG); /* objs are always >= MG */
5945 if (type >= SVt_PVMG) {
5947 if (!curse(sv, 1)) goto get_next_sv;
5948 type = SvTYPE(sv); /* destructor may have changed it */
5950 /* Free back-references before magic, in case the magic calls
5951 * Perl code that has weak references to sv. */
5952 if (type == SVt_PVHV) {
5953 Perl_hv_kill_backrefs(aTHX_ MUTABLE_HV(sv));
5957 else if (type == SVt_PVMG && SvPAD_OUR(sv)) {
5958 SvREFCNT_dec(SvOURSTASH(sv));
5959 } else if (SvMAGIC(sv)) {
5960 /* Free back-references before other types of magic. */
5961 sv_unmagic(sv, PERL_MAGIC_backref);
5964 if (type == SVt_PVMG && SvPAD_TYPED(sv))
5965 SvREFCNT_dec(SvSTASH(sv));
5968 /* case SVt_BIND: */
5971 IoIFP(sv) != PerlIO_stdin() &&
5972 IoIFP(sv) != PerlIO_stdout() &&
5973 IoIFP(sv) != PerlIO_stderr() &&
5974 !(IoFLAGS(sv) & IOf_FAKE_DIRP))
5976 io_close(MUTABLE_IO(sv), FALSE);
5978 if (IoDIRP(sv) && !(IoFLAGS(sv) & IOf_FAKE_DIRP))
5979 PerlDir_close(IoDIRP(sv));
5980 IoDIRP(sv) = (DIR*)NULL;
5981 Safefree(IoTOP_NAME(sv));
5982 Safefree(IoFMT_NAME(sv));
5983 Safefree(IoBOTTOM_NAME(sv));
5986 /* FIXME for plugins */
5987 pregfree2((REGEXP*) sv);
5991 cv_undef(MUTABLE_CV(sv));
5992 /* If we're in a stash, we don't own a reference to it.
5993 * However it does have a back reference to us, which needs to
5995 if ((stash = CvSTASH(sv)))
5996 sv_del_backref(MUTABLE_SV(stash), sv);
5999 if (PL_last_swash_hv == (const HV *)sv) {
6000 PL_last_swash_hv = NULL;
6002 if (HvTOTALKEYS((HV*)sv) > 0) {
6004 /* this statement should match the one at the beginning of
6005 * hv_undef_flags() */
6006 if ( PL_phase != PERL_PHASE_DESTRUCT
6007 && (name = HvNAME((HV*)sv)))
6010 (void)hv_delete(PL_stashcache, name,
6011 HvNAMELEN_get((HV*)sv), G_DISCARD);
6012 hv_name_set((HV*)sv, NULL, 0, 0);
6015 /* save old iter_sv in unused SvSTASH field */
6016 assert(!SvOBJECT(sv));
6017 SvSTASH(sv) = (HV*)iter_sv;
6020 /* XXX ideally we should save the old value of hash_index
6021 * too, but I can't think of any place to hide it. The
6022 * effect of not saving it is that for freeing hashes of
6023 * hashes, we become quadratic in scanning the HvARRAY of
6024 * the top hash looking for new entries to free; but
6025 * hopefully this will be dwarfed by the freeing of all
6026 * the nested hashes. */
6028 next_sv = Perl_hfree_next_entry(aTHX_ (HV*)sv, &hash_index);
6029 goto get_next_sv; /* process this new sv */
6031 /* free empty hash */
6032 Perl_hv_undef_flags(aTHX_ MUTABLE_HV(sv), HV_NAME_SETALL);
6033 assert(!HvARRAY((HV*)sv));
6037 AV* av = MUTABLE_AV(sv);
6038 if (PL_comppad == av) {
6042 if (AvREAL(av) && AvFILLp(av) > -1) {
6043 next_sv = AvARRAY(av)[AvFILLp(av)--];
6044 /* save old iter_sv in top-most slot of AV,
6045 * and pray that it doesn't get wiped in the meantime */
6046 AvARRAY(av)[AvMAX(av)] = iter_sv;
6048 goto get_next_sv; /* process this new sv */
6050 Safefree(AvALLOC(av));
6055 if (LvTYPE(sv) == 'T') { /* for tie: return HE to pool */
6056 SvREFCNT_dec(HeKEY_sv((HE*)LvTARG(sv)));
6057 HeNEXT((HE*)LvTARG(sv)) = PL_hv_fetch_ent_mh;
6058 PL_hv_fetch_ent_mh = (HE*)LvTARG(sv);
6060 else if (LvTYPE(sv) != 't') /* unless tie: unrefcnted fake SV** */
6061 SvREFCNT_dec(LvTARG(sv));
6063 if (isGV_with_GP(sv)) {
6064 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
6065 && HvENAME_get(stash))
6066 mro_method_changed_in(stash);
6067 gp_free(MUTABLE_GV(sv));
6069 unshare_hek(GvNAME_HEK(sv));
6070 /* If we're in a stash, we don't own a reference to it.
6071 * However it does have a back reference to us, which
6072 * needs to be cleared. */
6073 if (!SvVALID(sv) && (stash = GvSTASH(sv)))
6074 sv_del_backref(MUTABLE_SV(stash), sv);
6076 /* FIXME. There are probably more unreferenced pointers to SVs
6077 * in the interpreter struct that we should check and tidy in
6078 * a similar fashion to this: */
6079 if ((const GV *)sv == PL_last_in_gv)
6080 PL_last_in_gv = NULL;
6086 /* Don't bother with SvOOK_off(sv); as we're only going to
6090 SvOOK_offset(sv, offset);
6091 SvPV_set(sv, SvPVX_mutable(sv) - offset);
6092 /* Don't even bother with turning off the OOK flag. */
6097 SV * const target = SvRV(sv);
6099 sv_del_backref(target, sv);
6104 #ifdef PERL_OLD_COPY_ON_WRITE
6105 else if (SvPVX_const(sv)
6106 && !(SvTYPE(sv) == SVt_PVIO
6107 && !(IoFLAGS(sv) & IOf_FAKE_DIRP)))
6111 PerlIO_printf(Perl_debug_log, "Copy on write: clear\n");
6115 sv_release_COW(sv, SvPVX_const(sv), SV_COW_NEXT_SV(sv));
6117 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
6121 } else if (SvLEN(sv)) {
6122 Safefree(SvPVX_const(sv));
6126 else if (SvPVX_const(sv) && SvLEN(sv)
6127 && !(SvTYPE(sv) == SVt_PVIO
6128 && !(IoFLAGS(sv) & IOf_FAKE_DIRP)))
6129 Safefree(SvPVX_mutable(sv));
6130 else if (SvPVX_const(sv) && SvREADONLY(sv) && SvFAKE(sv)) {
6131 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
6142 SvFLAGS(sv) &= SVf_BREAK;
6143 SvFLAGS(sv) |= SVTYPEMASK;
6145 sv_type_details = bodies_by_type + type;
6146 if (sv_type_details->arena) {
6147 del_body(((char *)SvANY(sv) + sv_type_details->offset),
6148 &PL_body_roots[type]);
6150 else if (sv_type_details->body_size) {
6151 safefree(SvANY(sv));
6155 /* caller is responsible for freeing the head of the original sv */
6156 if (sv != orig_sv && !SvREFCNT(sv))
6159 /* grab and free next sv, if any */
6167 else if (!iter_sv) {
6169 } else if (SvTYPE(iter_sv) == SVt_PVAV) {
6170 AV *const av = (AV*)iter_sv;
6171 if (AvFILLp(av) > -1) {
6172 sv = AvARRAY(av)[AvFILLp(av)--];
6174 else { /* no more elements of current AV to free */
6177 /* restore previous value, squirrelled away */
6178 iter_sv = AvARRAY(av)[AvMAX(av)];
6179 Safefree(AvALLOC(av));
6182 } else if (SvTYPE(iter_sv) == SVt_PVHV) {
6183 if (!HvTOTALKEYS((HV *)iter_sv)) {
6184 /* no more elements of current HV to free */
6187 /* Restore previous value of iter_sv, squirrelled away */
6188 assert(!SvOBJECT(sv));
6189 iter_sv = (SV*)SvSTASH(sv);
6191 /* ideally we should restore the old hash_index here,
6192 * but we don't currently save the old value */
6195 /* free any remaining detritus from the hash struct */
6196 Perl_hv_undef_flags(aTHX_ MUTABLE_HV(sv), HV_NAME_SETALL);
6197 assert(!HvARRAY((HV*)sv));
6200 sv = Perl_hfree_next_entry(aTHX_ (HV*)iter_sv, &hash_index);
6203 /* unrolled SvREFCNT_dec and sv_free2 follows: */
6207 if (!SvREFCNT(sv)) {
6211 if (--(SvREFCNT(sv)))
6215 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEBUGGING),
6216 "Attempt to free temp prematurely: SV 0x%"UVxf
6217 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
6221 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
6222 /* make sure SvREFCNT(sv)==0 happens very seldom */
6223 SvREFCNT(sv) = (~(U32)0)/2;
6232 /* This routine curses the sv itself, not the object referenced by sv. So
6233 sv does not have to be ROK. */
6236 S_curse(pTHX_ SV * const sv, const bool check_refcnt) {
6239 PERL_ARGS_ASSERT_CURSE;
6240 assert(SvOBJECT(sv));
6242 if (PL_defstash && /* Still have a symbol table? */
6249 stash = SvSTASH(sv);
6250 destructor = StashHANDLER(stash,DESTROY);
6252 /* A constant subroutine can have no side effects, so
6253 don't bother calling it. */
6254 && !CvCONST(destructor)
6255 /* Don't bother calling an empty destructor */
6256 && (CvISXSUB(destructor)
6257 || (CvSTART(destructor)
6258 && (CvSTART(destructor)->op_next->op_type
6261 SV* const tmpref = newRV(sv);
6262 SvREADONLY_on(tmpref); /* DESTROY() could be naughty */
6264 PUSHSTACKi(PERLSI_DESTROY);
6269 call_sv(MUTABLE_SV(destructor),
6270 G_DISCARD|G_EVAL|G_KEEPERR|G_VOID);
6274 if(SvREFCNT(tmpref) < 2) {
6275 /* tmpref is not kept alive! */
6277 SvRV_set(tmpref, NULL);
6280 SvREFCNT_dec(tmpref);
6282 } while (SvOBJECT(sv) && SvSTASH(sv) != stash);
6285 if (check_refcnt && SvREFCNT(sv)) {
6286 if (PL_in_clean_objs)
6288 "DESTROY created new reference to dead object '%s'",
6290 /* DESTROY gave object new lease on life */
6296 SvREFCNT_dec(SvSTASH(sv)); /* possibly of changed persuasion */
6297 SvOBJECT_off(sv); /* Curse the object. */
6298 if (SvTYPE(sv) != SVt_PVIO)
6299 --PL_sv_objcount;/* XXX Might want something more general */
6305 =for apidoc sv_newref
6307 Increment an SV's reference count. Use the C<SvREFCNT_inc()> wrapper
6314 Perl_sv_newref(pTHX_ SV *const sv)
6316 PERL_UNUSED_CONTEXT;
6325 Decrement an SV's reference count, and if it drops to zero, call
6326 C<sv_clear> to invoke destructors and free up any memory used by
6327 the body; finally, deallocate the SV's head itself.
6328 Normally called via a wrapper macro C<SvREFCNT_dec>.
6334 Perl_sv_free(pTHX_ SV *const sv)
6339 if (SvREFCNT(sv) == 0) {
6340 if (SvFLAGS(sv) & SVf_BREAK)
6341 /* this SV's refcnt has been artificially decremented to
6342 * trigger cleanup */
6344 if (PL_in_clean_all) /* All is fair */
6346 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
6347 /* make sure SvREFCNT(sv)==0 happens very seldom */
6348 SvREFCNT(sv) = (~(U32)0)/2;
6351 if (ckWARN_d(WARN_INTERNAL)) {
6352 #ifdef DEBUG_LEAKING_SCALARS_FORK_DUMP
6353 Perl_dump_sv_child(aTHX_ sv);
6355 #ifdef DEBUG_LEAKING_SCALARS
6358 #ifdef DEBUG_LEAKING_SCALARS_ABORT
6359 if (PL_warnhook == PERL_WARNHOOK_FATAL
6360 || ckDEAD(packWARN(WARN_INTERNAL))) {
6361 /* Don't let Perl_warner cause us to escape our fate: */
6365 /* This may not return: */
6366 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
6367 "Attempt to free unreferenced scalar: SV 0x%"UVxf
6368 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
6371 #ifdef DEBUG_LEAKING_SCALARS_ABORT
6376 if (--(SvREFCNT(sv)) > 0)
6378 Perl_sv_free2(aTHX_ sv);
6382 Perl_sv_free2(pTHX_ SV *const sv)
6386 PERL_ARGS_ASSERT_SV_FREE2;
6390 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEBUGGING),
6391 "Attempt to free temp prematurely: SV 0x%"UVxf
6392 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
6396 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
6397 /* make sure SvREFCNT(sv)==0 happens very seldom */
6398 SvREFCNT(sv) = (~(U32)0)/2;
6409 Returns the length of the string in the SV. Handles magic and type
6410 coercion. See also C<SvCUR>, which gives raw access to the xpv_cur slot.
6416 Perl_sv_len(pTHX_ register SV *const sv)
6424 len = mg_length(sv);
6426 (void)SvPV_const(sv, len);
6431 =for apidoc sv_len_utf8
6433 Returns the number of characters in the string in an SV, counting wide
6434 UTF-8 bytes as a single character. Handles magic and type coercion.
6440 * The length is cached in PERL_MAGIC_utf8, in the mg_len field. Also the
6441 * mg_ptr is used, by sv_pos_u2b() and sv_pos_b2u() - see the comments below.
6442 * (Note that the mg_len is not the length of the mg_ptr field.
6443 * This allows the cache to store the character length of the string without
6444 * needing to malloc() extra storage to attach to the mg_ptr.)
6449 Perl_sv_len_utf8(pTHX_ register SV *const sv)
6455 return mg_length(sv);
6459 const U8 *s = (U8*)SvPV_const(sv, len);
6463 MAGIC *mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_utf8) : NULL;
6465 if (mg && (mg->mg_len != -1 || mg->mg_ptr)) {
6466 if (mg->mg_len != -1)
6469 /* We can use the offset cache for a headstart.
6470 The longer value is stored in the first pair. */
6471 STRLEN *cache = (STRLEN *) mg->mg_ptr;
6473 ulen = cache[0] + Perl_utf8_length(aTHX_ s + cache[1],
6477 if (PL_utf8cache < 0) {
6478 const STRLEN real = Perl_utf8_length(aTHX_ s, s + len);
6479 assert_uft8_cache_coherent("sv_len_utf8", ulen, real, sv);
6483 ulen = Perl_utf8_length(aTHX_ s, s + len);
6484 utf8_mg_len_cache_update(sv, &mg, ulen);
6488 return Perl_utf8_length(aTHX_ s, s + len);
6492 /* Walk forwards to find the byte corresponding to the passed in UTF-8
6495 S_sv_pos_u2b_forwards(const U8 *const start, const U8 *const send,
6496 STRLEN *const uoffset_p, bool *const at_end)
6498 const U8 *s = start;
6499 STRLEN uoffset = *uoffset_p;
6501 PERL_ARGS_ASSERT_SV_POS_U2B_FORWARDS;
6503 while (s < send && uoffset) {
6510 else if (s > send) {
6512 /* This is the existing behaviour. Possibly it should be a croak, as
6513 it's actually a bounds error */
6516 *uoffset_p -= uoffset;
6520 /* Given the length of the string in both bytes and UTF-8 characters, decide
6521 whether to walk forwards or backwards to find the byte corresponding to
6522 the passed in UTF-8 offset. */
6524 S_sv_pos_u2b_midway(const U8 *const start, const U8 *send,
6525 STRLEN uoffset, const STRLEN uend)
6527 STRLEN backw = uend - uoffset;
6529 PERL_ARGS_ASSERT_SV_POS_U2B_MIDWAY;
6531 if (uoffset < 2 * backw) {
6532 /* The assumption is that going forwards is twice the speed of going
6533 forward (that's where the 2 * backw comes from).
6534 (The real figure of course depends on the UTF-8 data.) */
6535 const U8 *s = start;
6537 while (s < send && uoffset--)
6547 while (UTF8_IS_CONTINUATION(*send))
6550 return send - start;
6553 /* For the string representation of the given scalar, find the byte
6554 corresponding to the passed in UTF-8 offset. uoffset0 and boffset0
6555 give another position in the string, *before* the sought offset, which
6556 (which is always true, as 0, 0 is a valid pair of positions), which should
6557 help reduce the amount of linear searching.
6558 If *mgp is non-NULL, it should point to the UTF-8 cache magic, which
6559 will be used to reduce the amount of linear searching. The cache will be
6560 created if necessary, and the found value offered to it for update. */
6562 S_sv_pos_u2b_cached(pTHX_ SV *const sv, MAGIC **const mgp, const U8 *const start,
6563 const U8 *const send, STRLEN uoffset,
6564 STRLEN uoffset0, STRLEN boffset0)
6566 STRLEN boffset = 0; /* Actually always set, but let's keep gcc happy. */
6568 bool at_end = FALSE;
6570 PERL_ARGS_ASSERT_SV_POS_U2B_CACHED;
6572 assert (uoffset >= uoffset0);
6579 && (*mgp || (SvTYPE(sv) >= SVt_PVMG &&
6580 (*mgp = mg_find(sv, PERL_MAGIC_utf8))))) {
6581 if ((*mgp)->mg_ptr) {
6582 STRLEN *cache = (STRLEN *) (*mgp)->mg_ptr;
6583 if (cache[0] == uoffset) {
6584 /* An exact match. */
6587 if (cache[2] == uoffset) {
6588 /* An exact match. */
6592 if (cache[0] < uoffset) {
6593 /* The cache already knows part of the way. */
6594 if (cache[0] > uoffset0) {
6595 /* The cache knows more than the passed in pair */
6596 uoffset0 = cache[0];
6597 boffset0 = cache[1];
6599 if ((*mgp)->mg_len != -1) {
6600 /* And we know the end too. */
6602 + sv_pos_u2b_midway(start + boffset0, send,
6604 (*mgp)->mg_len - uoffset0);
6606 uoffset -= uoffset0;
6608 + sv_pos_u2b_forwards(start + boffset0,
6609 send, &uoffset, &at_end);
6610 uoffset += uoffset0;
6613 else if (cache[2] < uoffset) {
6614 /* We're between the two cache entries. */
6615 if (cache[2] > uoffset0) {
6616 /* and the cache knows more than the passed in pair */
6617 uoffset0 = cache[2];
6618 boffset0 = cache[3];
6622 + sv_pos_u2b_midway(start + boffset0,
6625 cache[0] - uoffset0);
6628 + sv_pos_u2b_midway(start + boffset0,
6631 cache[2] - uoffset0);
6635 else if ((*mgp)->mg_len != -1) {
6636 /* If we can take advantage of a passed in offset, do so. */
6637 /* In fact, offset0 is either 0, or less than offset, so don't
6638 need to worry about the other possibility. */
6640 + sv_pos_u2b_midway(start + boffset0, send,
6642 (*mgp)->mg_len - uoffset0);
6647 if (!found || PL_utf8cache < 0) {
6648 STRLEN real_boffset;
6649 uoffset -= uoffset0;
6650 real_boffset = boffset0 + sv_pos_u2b_forwards(start + boffset0,
6651 send, &uoffset, &at_end);
6652 uoffset += uoffset0;
6654 if (found && PL_utf8cache < 0)
6655 assert_uft8_cache_coherent("sv_pos_u2b_cache", boffset,
6657 boffset = real_boffset;
6662 utf8_mg_len_cache_update(sv, mgp, uoffset);
6664 utf8_mg_pos_cache_update(sv, mgp, boffset, uoffset, send - start);
6671 =for apidoc sv_pos_u2b_flags
6673 Converts the value pointed to by offsetp from a count of UTF-8 chars from
6674 the start of the string, to a count of the equivalent number of bytes; if
6675 lenp is non-zero, it does the same to lenp, but this time starting from
6676 the offset, rather than from the start of the string. Handles type coercion.
6677 I<flags> is passed to C<SvPV_flags>, and usually should be
6678 C<SV_GMAGIC|SV_CONST_RETURN> to handle magic.
6684 * sv_pos_u2b_flags() uses, like sv_pos_b2u(), the mg_ptr of the potential
6685 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6686 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
6691 Perl_sv_pos_u2b_flags(pTHX_ SV *const sv, STRLEN uoffset, STRLEN *const lenp,
6698 PERL_ARGS_ASSERT_SV_POS_U2B_FLAGS;
6700 start = (U8*)SvPV_flags(sv, len, flags);
6702 const U8 * const send = start + len;
6704 boffset = sv_pos_u2b_cached(sv, &mg, start, send, uoffset, 0, 0);
6707 && *lenp /* don't bother doing work for 0, as its bytes equivalent
6708 is 0, and *lenp is already set to that. */) {
6709 /* Convert the relative offset to absolute. */
6710 const STRLEN uoffset2 = uoffset + *lenp;
6711 const STRLEN boffset2
6712 = sv_pos_u2b_cached(sv, &mg, start, send, uoffset2,
6713 uoffset, boffset) - boffset;
6727 =for apidoc sv_pos_u2b
6729 Converts the value pointed to by offsetp from a count of UTF-8 chars from
6730 the start of the string, to a count of the equivalent number of bytes; if
6731 lenp is non-zero, it does the same to lenp, but this time starting from
6732 the offset, rather than from the start of the string. Handles magic and
6735 Use C<sv_pos_u2b_flags> in preference, which correctly handles strings longer
6742 * sv_pos_u2b() uses, like sv_pos_b2u(), the mg_ptr of the potential
6743 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6744 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
6748 /* This function is subject to size and sign problems */
6751 Perl_sv_pos_u2b(pTHX_ register SV *const sv, I32 *const offsetp, I32 *const lenp)
6753 PERL_ARGS_ASSERT_SV_POS_U2B;
6756 STRLEN ulen = (STRLEN)*lenp;
6757 *offsetp = (I32)sv_pos_u2b_flags(sv, (STRLEN)*offsetp, &ulen,
6758 SV_GMAGIC|SV_CONST_RETURN);
6761 *offsetp = (I32)sv_pos_u2b_flags(sv, (STRLEN)*offsetp, NULL,
6762 SV_GMAGIC|SV_CONST_RETURN);
6767 S_utf8_mg_len_cache_update(pTHX_ SV *const sv, MAGIC **const mgp,
6770 PERL_ARGS_ASSERT_UTF8_MG_LEN_CACHE_UPDATE;
6774 if (!*mgp && (SvTYPE(sv) < SVt_PVMG ||
6775 !(*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
6776 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, &PL_vtbl_utf8, 0, 0);
6780 (*mgp)->mg_len = ulen;
6781 /* For now, treat "overflowed" as "still unknown". See RT #72924. */
6782 if (ulen != (STRLEN) (*mgp)->mg_len)
6783 (*mgp)->mg_len = -1;
6786 /* Create and update the UTF8 magic offset cache, with the proffered utf8/
6787 byte length pairing. The (byte) length of the total SV is passed in too,
6788 as blen, because for some (more esoteric) SVs, the call to SvPV_const()
6789 may not have updated SvCUR, so we can't rely on reading it directly.
6791 The proffered utf8/byte length pairing isn't used if the cache already has
6792 two pairs, and swapping either for the proffered pair would increase the
6793 RMS of the intervals between known byte offsets.
6795 The cache itself consists of 4 STRLEN values
6796 0: larger UTF-8 offset
6797 1: corresponding byte offset
6798 2: smaller UTF-8 offset
6799 3: corresponding byte offset
6801 Unused cache pairs have the value 0, 0.
6802 Keeping the cache "backwards" means that the invariant of
6803 cache[0] >= cache[2] is maintained even with empty slots, which means that
6804 the code that uses it doesn't need to worry if only 1 entry has actually
6805 been set to non-zero. It also makes the "position beyond the end of the
6806 cache" logic much simpler, as the first slot is always the one to start
6810 S_utf8_mg_pos_cache_update(pTHX_ SV *const sv, MAGIC **const mgp, const STRLEN byte,
6811 const STRLEN utf8, const STRLEN blen)
6815 PERL_ARGS_ASSERT_UTF8_MG_POS_CACHE_UPDATE;
6820 if (!*mgp && (SvTYPE(sv) < SVt_PVMG ||
6821 !(*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
6822 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, (MGVTBL*)&PL_vtbl_utf8, 0,
6824 (*mgp)->mg_len = -1;
6828 if (!(cache = (STRLEN *)(*mgp)->mg_ptr)) {
6829 Newxz(cache, PERL_MAGIC_UTF8_CACHESIZE * 2, STRLEN);
6830 (*mgp)->mg_ptr = (char *) cache;
6834 if (PL_utf8cache < 0 && SvPOKp(sv)) {
6835 /* SvPOKp() because it's possible that sv has string overloading, and
6836 therefore is a reference, hence SvPVX() is actually a pointer.
6837 This cures the (very real) symptoms of RT 69422, but I'm not actually
6838 sure whether we should even be caching the results of UTF-8
6839 operations on overloading, given that nothing stops overloading
6840 returning a different value every time it's called. */
6841 const U8 *start = (const U8 *) SvPVX_const(sv);
6842 const STRLEN realutf8 = utf8_length(start, start + byte);
6844 assert_uft8_cache_coherent("utf8_mg_pos_cache_update", utf8, realutf8,
6848 /* Cache is held with the later position first, to simplify the code
6849 that deals with unbounded ends. */
6851 ASSERT_UTF8_CACHE(cache);
6852 if (cache[1] == 0) {
6853 /* Cache is totally empty */
6856 } else if (cache[3] == 0) {
6857 if (byte > cache[1]) {
6858 /* New one is larger, so goes first. */
6859 cache[2] = cache[0];
6860 cache[3] = cache[1];
6868 #define THREEWAY_SQUARE(a,b,c,d) \
6869 ((float)((d) - (c))) * ((float)((d) - (c))) \
6870 + ((float)((c) - (b))) * ((float)((c) - (b))) \
6871 + ((float)((b) - (a))) * ((float)((b) - (a)))
6873 /* Cache has 2 slots in use, and we know three potential pairs.
6874 Keep the two that give the lowest RMS distance. Do the
6875 calculation in bytes simply because we always know the byte
6876 length. squareroot has the same ordering as the positive value,
6877 so don't bother with the actual square root. */
6878 const float existing = THREEWAY_SQUARE(0, cache[3], cache[1], blen);
6879 if (byte > cache[1]) {
6880 /* New position is after the existing pair of pairs. */
6881 const float keep_earlier
6882 = THREEWAY_SQUARE(0, cache[3], byte, blen);
6883 const float keep_later
6884 = THREEWAY_SQUARE(0, cache[1], byte, blen);
6886 if (keep_later < keep_earlier) {
6887 if (keep_later < existing) {
6888 cache[2] = cache[0];
6889 cache[3] = cache[1];
6895 if (keep_earlier < existing) {
6901 else if (byte > cache[3]) {
6902 /* New position is between the existing pair of pairs. */
6903 const float keep_earlier
6904 = THREEWAY_SQUARE(0, cache[3], byte, blen);
6905 const float keep_later
6906 = THREEWAY_SQUARE(0, byte, cache[1], blen);
6908 if (keep_later < keep_earlier) {
6909 if (keep_later < existing) {
6915 if (keep_earlier < existing) {
6922 /* New position is before the existing pair of pairs. */
6923 const float keep_earlier
6924 = THREEWAY_SQUARE(0, byte, cache[3], blen);
6925 const float keep_later
6926 = THREEWAY_SQUARE(0, byte, cache[1], blen);
6928 if (keep_later < keep_earlier) {
6929 if (keep_later < existing) {
6935 if (keep_earlier < existing) {
6936 cache[0] = cache[2];
6937 cache[1] = cache[3];
6944 ASSERT_UTF8_CACHE(cache);
6947 /* We already know all of the way, now we may be able to walk back. The same
6948 assumption is made as in S_sv_pos_u2b_midway(), namely that walking
6949 backward is half the speed of walking forward. */
6951 S_sv_pos_b2u_midway(pTHX_ const U8 *const s, const U8 *const target,
6952 const U8 *end, STRLEN endu)
6954 const STRLEN forw = target - s;
6955 STRLEN backw = end - target;
6957 PERL_ARGS_ASSERT_SV_POS_B2U_MIDWAY;
6959 if (forw < 2 * backw) {
6960 return utf8_length(s, target);
6963 while (end > target) {
6965 while (UTF8_IS_CONTINUATION(*end)) {
6974 =for apidoc sv_pos_b2u
6976 Converts the value pointed to by offsetp from a count of bytes from the
6977 start of the string, to a count of the equivalent number of UTF-8 chars.
6978 Handles magic and type coercion.
6984 * sv_pos_b2u() uses, like sv_pos_u2b(), the mg_ptr of the potential
6985 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6990 Perl_sv_pos_b2u(pTHX_ register SV *const sv, I32 *const offsetp)
6993 const STRLEN byte = *offsetp;
6994 STRLEN len = 0; /* Actually always set, but let's keep gcc happy. */
7000 PERL_ARGS_ASSERT_SV_POS_B2U;
7005 s = (const U8*)SvPV_const(sv, blen);
7008 Perl_croak(aTHX_ "panic: sv_pos_b2u: bad byte offset");
7014 && SvTYPE(sv) >= SVt_PVMG
7015 && (mg = mg_find(sv, PERL_MAGIC_utf8)))
7018 STRLEN * const cache = (STRLEN *) mg->mg_ptr;
7019 if (cache[1] == byte) {
7020 /* An exact match. */
7021 *offsetp = cache[0];
7024 if (cache[3] == byte) {
7025 /* An exact match. */
7026 *offsetp = cache[2];
7030 if (cache[1] < byte) {
7031 /* We already know part of the way. */
7032 if (mg->mg_len != -1) {
7033 /* Actually, we know the end too. */
7035 + S_sv_pos_b2u_midway(aTHX_ s + cache[1], send,
7036 s + blen, mg->mg_len - cache[0]);
7038 len = cache[0] + utf8_length(s + cache[1], send);
7041 else if (cache[3] < byte) {
7042 /* We're between the two cached pairs, so we do the calculation
7043 offset by the byte/utf-8 positions for the earlier pair,
7044 then add the utf-8 characters from the string start to
7046 len = S_sv_pos_b2u_midway(aTHX_ s + cache[3], send,
7047 s + cache[1], cache[0] - cache[2])
7051 else { /* cache[3] > byte */
7052 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + cache[3],
7056 ASSERT_UTF8_CACHE(cache);
7058 } else if (mg->mg_len != -1) {
7059 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + blen, mg->mg_len);
7063 if (!found || PL_utf8cache < 0) {
7064 const STRLEN real_len = utf8_length(s, send);
7066 if (found && PL_utf8cache < 0)
7067 assert_uft8_cache_coherent("sv_pos_b2u", len, real_len, sv);
7074 utf8_mg_len_cache_update(sv, &mg, len);
7076 utf8_mg_pos_cache_update(sv, &mg, byte, len, blen);
7081 S_assert_uft8_cache_coherent(pTHX_ const char *const func, STRLEN from_cache,
7082 STRLEN real, SV *const sv)
7084 PERL_ARGS_ASSERT_ASSERT_UFT8_CACHE_COHERENT;
7086 /* As this is debugging only code, save space by keeping this test here,
7087 rather than inlining it in all the callers. */
7088 if (from_cache == real)
7091 /* Need to turn the assertions off otherwise we may recurse infinitely
7092 while printing error messages. */
7093 SAVEI8(PL_utf8cache);
7095 Perl_croak(aTHX_ "panic: %s cache %"UVuf" real %"UVuf" for %"SVf,
7096 func, (UV) from_cache, (UV) real, SVfARG(sv));
7102 Returns a boolean indicating whether the strings in the two SVs are
7103 identical. Is UTF-8 and 'use bytes' aware, handles get magic, and will
7104 coerce its args to strings if necessary.
7106 =for apidoc sv_eq_flags
7108 Returns a boolean indicating whether the strings in the two SVs are
7109 identical. Is UTF-8 and 'use bytes' aware and coerces its args to strings
7110 if necessary. If the flags include SV_GMAGIC, it handles get-magic, too.
7116 Perl_sv_eq_flags(pTHX_ register SV *sv1, register SV *sv2, const U32 flags)
7125 SV* svrecode = NULL;
7132 /* if pv1 and pv2 are the same, second SvPV_const call may
7133 * invalidate pv1 (if we are handling magic), so we may need to
7135 if (sv1 == sv2 && flags & SV_GMAGIC
7136 && (SvTHINKFIRST(sv1) || SvGMAGICAL(sv1))) {
7137 pv1 = SvPV_const(sv1, cur1);
7138 sv1 = newSVpvn_flags(pv1, cur1, SVs_TEMP | SvUTF8(sv2));
7140 pv1 = SvPV_flags_const(sv1, cur1, flags);
7148 pv2 = SvPV_flags_const(sv2, cur2, flags);
7150 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
7151 /* Differing utf8ness.
7152 * Do not UTF8size the comparands as a side-effect. */
7155 svrecode = newSVpvn(pv2, cur2);
7156 sv_recode_to_utf8(svrecode, PL_encoding);
7157 pv2 = SvPV_const(svrecode, cur2);
7160 svrecode = newSVpvn(pv1, cur1);
7161 sv_recode_to_utf8(svrecode, PL_encoding);
7162 pv1 = SvPV_const(svrecode, cur1);
7164 /* Now both are in UTF-8. */
7166 SvREFCNT_dec(svrecode);
7172 /* sv1 is the UTF-8 one */
7173 return bytes_cmp_utf8((const U8*)pv2, cur2,
7174 (const U8*)pv1, cur1) == 0;
7177 /* sv2 is the UTF-8 one */
7178 return bytes_cmp_utf8((const U8*)pv1, cur1,
7179 (const U8*)pv2, cur2) == 0;
7185 eq = (pv1 == pv2) || memEQ(pv1, pv2, cur1);
7187 SvREFCNT_dec(svrecode);
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, handles get magic, and will
7200 coerce its args to strings if necessary. See also C<sv_cmp_locale>.
7202 =for apidoc sv_cmp_flags
7204 Compares the strings in two SVs. Returns -1, 0, or 1 indicating whether the
7205 string in C<sv1> is less than, equal to, or greater than the string in
7206 C<sv2>. Is UTF-8 and 'use bytes' aware and will coerce its args to strings
7207 if necessary. If the flags include SV_GMAGIC, it handles get magic. See
7208 also C<sv_cmp_locale_flags>.
7214 Perl_sv_cmp(pTHX_ register SV *const sv1, register SV *const sv2)
7216 return sv_cmp_flags(sv1, sv2, SV_GMAGIC);
7220 Perl_sv_cmp_flags(pTHX_ register SV *const sv1, register SV *const sv2,
7225 const char *pv1, *pv2;
7228 SV *svrecode = NULL;
7235 pv1 = SvPV_flags_const(sv1, cur1, flags);
7242 pv2 = SvPV_flags_const(sv2, cur2, flags);
7244 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
7245 /* Differing utf8ness.
7246 * Do not UTF8size the comparands as a side-effect. */
7249 svrecode = newSVpvn(pv2, cur2);
7250 sv_recode_to_utf8(svrecode, PL_encoding);
7251 pv2 = SvPV_const(svrecode, cur2);
7254 const int retval = -bytes_cmp_utf8((const U8*)pv2, cur2,
7255 (const U8*)pv1, cur1);
7256 return retval ? retval < 0 ? -1 : +1 : 0;
7261 svrecode = newSVpvn(pv1, cur1);
7262 sv_recode_to_utf8(svrecode, PL_encoding);
7263 pv1 = SvPV_const(svrecode, cur1);
7266 const int retval = bytes_cmp_utf8((const U8*)pv1, cur1,
7267 (const U8*)pv2, cur2);
7268 return retval ? retval < 0 ? -1 : +1 : 0;
7274 cmp = cur2 ? -1 : 0;
7278 const I32 retval = memcmp((const void*)pv1, (const void*)pv2, cur1 < cur2 ? cur1 : cur2);
7281 cmp = retval < 0 ? -1 : 1;
7282 } else if (cur1 == cur2) {
7285 cmp = cur1 < cur2 ? -1 : 1;
7289 SvREFCNT_dec(svrecode);
7297 =for apidoc sv_cmp_locale
7299 Compares the strings in two SVs in a locale-aware manner. Is UTF-8 and
7300 'use bytes' aware, handles get magic, and will coerce its args to strings
7301 if necessary. See also C<sv_cmp>.
7303 =for apidoc sv_cmp_locale_flags
7305 Compares the strings in two SVs in a locale-aware manner. Is UTF-8 and
7306 'use bytes' aware and will coerce its args to strings if necessary. If the
7307 flags contain SV_GMAGIC, it handles get magic. See also C<sv_cmp_flags>.
7313 Perl_sv_cmp_locale(pTHX_ register SV *const sv1, register SV *const sv2)
7315 return sv_cmp_locale_flags(sv1, sv2, SV_GMAGIC);
7319 Perl_sv_cmp_locale_flags(pTHX_ register SV *const sv1, register SV *const sv2,
7323 #ifdef USE_LOCALE_COLLATE
7329 if (PL_collation_standard)
7333 pv1 = sv1 ? sv_collxfrm_flags(sv1, &len1, flags) : (char *) NULL;
7335 pv2 = sv2 ? sv_collxfrm_flags(sv2, &len2, flags) : (char *) NULL;
7337 if (!pv1 || !len1) {
7348 retval = memcmp((void*)pv1, (void*)pv2, len1 < len2 ? len1 : len2);
7351 return retval < 0 ? -1 : 1;
7354 * When the result of collation is equality, that doesn't mean
7355 * that there are no differences -- some locales exclude some
7356 * characters from consideration. So to avoid false equalities,
7357 * we use the raw string as a tiebreaker.
7363 #endif /* USE_LOCALE_COLLATE */
7365 return sv_cmp(sv1, sv2);
7369 #ifdef USE_LOCALE_COLLATE
7372 =for apidoc sv_collxfrm
7374 This calls C<sv_collxfrm_flags> with the SV_GMAGIC flag. See
7375 C<sv_collxfrm_flags>.
7377 =for apidoc sv_collxfrm_flags
7379 Add Collate Transform magic to an SV if it doesn't already have it. If the
7380 flags contain SV_GMAGIC, it handles get-magic.
7382 Any scalar variable may carry PERL_MAGIC_collxfrm magic that contains the
7383 scalar data of the variable, but transformed to such a format that a normal
7384 memory comparison can be used to compare the data according to the locale
7391 Perl_sv_collxfrm_flags(pTHX_ SV *const sv, STRLEN *const nxp, const I32 flags)
7396 PERL_ARGS_ASSERT_SV_COLLXFRM_FLAGS;
7398 mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_collxfrm) : (MAGIC *) NULL;
7399 if (!mg || !mg->mg_ptr || *(U32*)mg->mg_ptr != PL_collation_ix) {
7405 Safefree(mg->mg_ptr);
7406 s = SvPV_flags_const(sv, len, flags);
7407 if ((xf = mem_collxfrm(s, len, &xlen))) {
7409 #ifdef PERL_OLD_COPY_ON_WRITE
7411 sv_force_normal_flags(sv, 0);
7413 mg = sv_magicext(sv, 0, PERL_MAGIC_collxfrm, &PL_vtbl_collxfrm,
7427 if (mg && mg->mg_ptr) {
7429 return mg->mg_ptr + sizeof(PL_collation_ix);
7437 #endif /* USE_LOCALE_COLLATE */
7440 S_sv_gets_append_to_utf8(pTHX_ SV *const sv, PerlIO *const fp, I32 append)
7442 SV * const tsv = newSV(0);
7445 sv_gets(tsv, fp, 0);
7446 sv_utf8_upgrade_nomg(tsv);
7447 SvCUR_set(sv,append);
7450 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7454 S_sv_gets_read_record(pTHX_ SV *const sv, PerlIO *const fp, I32 append)
7457 const U32 recsize = SvUV(SvRV(PL_rs)); /* RsRECORD() guarantees > 0. */
7458 /* Grab the size of the record we're getting */
7459 char *const buffer = SvGROW(sv, (STRLEN)(recsize + append + 1)) + append;
7466 /* VMS wants read instead of fread, because fread doesn't respect */
7467 /* RMS record boundaries. This is not necessarily a good thing to be */
7468 /* doing, but we've got no other real choice - except avoid stdio
7469 as implementation - perhaps write a :vms layer ?
7471 fd = PerlIO_fileno(fp);
7473 bytesread = PerlLIO_read(fd, buffer, recsize);
7475 else /* in-memory file from PerlIO::Scalar */
7478 bytesread = PerlIO_read(fp, buffer, recsize);
7483 SvCUR_set(sv, bytesread + append);
7484 buffer[bytesread] = '\0';
7485 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7491 Get a line from the filehandle and store it into the SV, optionally
7492 appending to the currently-stored string.
7498 Perl_sv_gets(pTHX_ register SV *const sv, register PerlIO *const fp, I32 append)
7503 register STDCHAR rslast;
7504 register STDCHAR *bp;
7509 PERL_ARGS_ASSERT_SV_GETS;
7511 if (SvTHINKFIRST(sv))
7512 sv_force_normal_flags(sv, append ? 0 : SV_COW_DROP_PV);
7513 /* XXX. If you make this PVIV, then copy on write can copy scalars read
7515 However, perlbench says it's slower, because the existing swipe code
7516 is faster than copy on write.
7517 Swings and roundabouts. */
7518 SvUPGRADE(sv, SVt_PV);
7523 if (PerlIO_isutf8(fp)) {
7525 sv_utf8_upgrade_nomg(sv);
7526 sv_pos_u2b(sv,&append,0);
7528 } else if (SvUTF8(sv)) {
7529 return S_sv_gets_append_to_utf8(aTHX_ sv, fp, append);
7537 if (PerlIO_isutf8(fp))
7540 if (IN_PERL_COMPILETIME) {
7541 /* we always read code in line mode */
7545 else if (RsSNARF(PL_rs)) {
7546 /* If it is a regular disk file use size from stat() as estimate
7547 of amount we are going to read -- may result in mallocing
7548 more memory than we really need if the layers below reduce
7549 the size we read (e.g. CRLF or a gzip layer).
7552 if (!PerlLIO_fstat(PerlIO_fileno(fp), &st) && S_ISREG(st.st_mode)) {
7553 const Off_t offset = PerlIO_tell(fp);
7554 if (offset != (Off_t) -1 && st.st_size + append > offset) {
7555 (void) SvGROW(sv, (STRLEN)((st.st_size - offset) + append + 1));
7561 else if (RsRECORD(PL_rs)) {
7562 return S_sv_gets_read_record(aTHX_ sv, fp, append);
7564 else if (RsPARA(PL_rs)) {
7570 /* Get $/ i.e. PL_rs into same encoding as stream wants */
7571 if (PerlIO_isutf8(fp)) {
7572 rsptr = SvPVutf8(PL_rs, rslen);
7575 if (SvUTF8(PL_rs)) {
7576 if (!sv_utf8_downgrade(PL_rs, TRUE)) {
7577 Perl_croak(aTHX_ "Wide character in $/");
7580 rsptr = SvPV_const(PL_rs, rslen);
7584 rslast = rslen ? rsptr[rslen - 1] : '\0';
7586 if (rspara) { /* have to do this both before and after */
7587 do { /* to make sure file boundaries work right */
7590 i = PerlIO_getc(fp);
7594 PerlIO_ungetc(fp,i);
7600 /* See if we know enough about I/O mechanism to cheat it ! */
7602 /* This used to be #ifdef test - it is made run-time test for ease
7603 of abstracting out stdio interface. One call should be cheap
7604 enough here - and may even be a macro allowing compile
7608 if (PerlIO_fast_gets(fp)) {
7611 * We're going to steal some values from the stdio struct
7612 * and put EVERYTHING in the innermost loop into registers.
7614 register STDCHAR *ptr;
7618 #if defined(VMS) && defined(PERLIO_IS_STDIO)
7619 /* An ungetc()d char is handled separately from the regular
7620 * buffer, so we getc() it back out and stuff it in the buffer.
7622 i = PerlIO_getc(fp);
7623 if (i == EOF) return 0;
7624 *(--((*fp)->_ptr)) = (unsigned char) i;
7628 /* Here is some breathtakingly efficient cheating */
7630 cnt = PerlIO_get_cnt(fp); /* get count into register */
7631 /* make sure we have the room */
7632 if ((I32)(SvLEN(sv) - append) <= cnt + 1) {
7633 /* Not room for all of it
7634 if we are looking for a separator and room for some
7636 if (rslen && cnt > 80 && (I32)SvLEN(sv) > append) {
7637 /* just process what we have room for */
7638 shortbuffered = cnt - SvLEN(sv) + append + 1;
7639 cnt -= shortbuffered;
7643 /* remember that cnt can be negative */
7644 SvGROW(sv, (STRLEN)(append + (cnt <= 0 ? 2 : (cnt + 1))));
7649 bp = (STDCHAR*)SvPVX_const(sv) + append; /* move these two too to registers */
7650 ptr = (STDCHAR*)PerlIO_get_ptr(fp);
7651 DEBUG_P(PerlIO_printf(Perl_debug_log,
7652 "Screamer: entering, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7653 DEBUG_P(PerlIO_printf(Perl_debug_log,
7654 "Screamer: entering: PerlIO * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7655 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7656 PTR2UV(PerlIO_has_base(fp) ? PerlIO_get_base(fp) : 0)));
7661 while (cnt > 0) { /* this | eat */
7663 if ((*bp++ = *ptr++) == rslast) /* really | dust */
7664 goto thats_all_folks; /* screams | sed :-) */
7668 Copy(ptr, bp, cnt, char); /* this | eat */
7669 bp += cnt; /* screams | dust */
7670 ptr += cnt; /* louder | sed :-) */
7672 assert (!shortbuffered);
7673 goto cannot_be_shortbuffered;
7677 if (shortbuffered) { /* oh well, must extend */
7678 cnt = shortbuffered;
7680 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
7682 SvGROW(sv, SvLEN(sv) + append + cnt + 2);
7683 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
7687 cannot_be_shortbuffered:
7688 DEBUG_P(PerlIO_printf(Perl_debug_log,
7689 "Screamer: going to getc, ptr=%"UVuf", cnt=%ld\n",
7690 PTR2UV(ptr),(long)cnt));
7691 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* deregisterize cnt and ptr */
7693 DEBUG_Pv(PerlIO_printf(Perl_debug_log,
7694 "Screamer: pre: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7695 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7696 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7698 /* This used to call 'filbuf' in stdio form, but as that behaves like
7699 getc when cnt <= 0 we use PerlIO_getc here to avoid introducing
7700 another abstraction. */
7701 i = PerlIO_getc(fp); /* get more characters */
7703 DEBUG_Pv(PerlIO_printf(Perl_debug_log,
7704 "Screamer: post: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7705 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7706 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7708 cnt = PerlIO_get_cnt(fp);
7709 ptr = (STDCHAR*)PerlIO_get_ptr(fp); /* reregisterize cnt and ptr */
7710 DEBUG_P(PerlIO_printf(Perl_debug_log,
7711 "Screamer: after getc, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7713 if (i == EOF) /* all done for ever? */
7714 goto thats_really_all_folks;
7716 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
7718 SvGROW(sv, bpx + cnt + 2);
7719 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
7721 *bp++ = (STDCHAR)i; /* store character from PerlIO_getc */
7723 if (rslen && (STDCHAR)i == rslast) /* all done for now? */
7724 goto thats_all_folks;
7728 if ((rslen > 1 && (STRLEN)(bp - (STDCHAR*)SvPVX_const(sv)) < rslen) ||
7729 memNE((char*)bp - rslen, rsptr, rslen))
7730 goto screamer; /* go back to the fray */
7731 thats_really_all_folks:
7733 cnt += shortbuffered;
7734 DEBUG_P(PerlIO_printf(Perl_debug_log,
7735 "Screamer: quitting, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7736 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* put these back or we're in trouble */
7737 DEBUG_P(PerlIO_printf(Perl_debug_log,
7738 "Screamer: end: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7739 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7740 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7742 SvCUR_set(sv, bp - (STDCHAR*)SvPVX_const(sv)); /* set length */
7743 DEBUG_P(PerlIO_printf(Perl_debug_log,
7744 "Screamer: done, len=%ld, string=|%.*s|\n",
7745 (long)SvCUR(sv),(int)SvCUR(sv),SvPVX_const(sv)));
7749 /*The big, slow, and stupid way. */
7750 #ifdef USE_HEAP_INSTEAD_OF_STACK /* Even slower way. */
7751 STDCHAR *buf = NULL;
7752 Newx(buf, 8192, STDCHAR);
7760 register const STDCHAR * const bpe = buf + sizeof(buf);
7762 while ((i = PerlIO_getc(fp)) != EOF && (*bp++ = (STDCHAR)i) != rslast && bp < bpe)
7763 ; /* keep reading */
7767 cnt = PerlIO_read(fp,(char*)buf, sizeof(buf));
7768 /* Accommodate broken VAXC compiler, which applies U8 cast to
7769 * both args of ?: operator, causing EOF to change into 255
7772 i = (U8)buf[cnt - 1];
7778 cnt = 0; /* we do need to re-set the sv even when cnt <= 0 */
7780 sv_catpvn(sv, (char *) buf, cnt);
7782 sv_setpvn(sv, (char *) buf, cnt);
7784 if (i != EOF && /* joy */
7786 SvCUR(sv) < rslen ||
7787 memNE(SvPVX_const(sv) + SvCUR(sv) - rslen, rsptr, rslen)))
7791 * If we're reading from a TTY and we get a short read,
7792 * indicating that the user hit his EOF character, we need
7793 * to notice it now, because if we try to read from the TTY
7794 * again, the EOF condition will disappear.
7796 * The comparison of cnt to sizeof(buf) is an optimization
7797 * that prevents unnecessary calls to feof().
7801 if (!(cnt < (I32)sizeof(buf) && PerlIO_eof(fp)))
7805 #ifdef USE_HEAP_INSTEAD_OF_STACK
7810 if (rspara) { /* have to do this both before and after */
7811 while (i != EOF) { /* to make sure file boundaries work right */
7812 i = PerlIO_getc(fp);
7814 PerlIO_ungetc(fp,i);
7820 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7826 Auto-increment of the value in the SV, doing string to numeric conversion
7827 if necessary. Handles 'get' magic and operator overloading.
7833 Perl_sv_inc(pTHX_ register SV *const sv)
7842 =for apidoc sv_inc_nomg
7844 Auto-increment of the value in the SV, doing string to numeric conversion
7845 if necessary. Handles operator overloading. Skips handling 'get' magic.
7851 Perl_sv_inc_nomg(pTHX_ register SV *const sv)
7859 if (SvTHINKFIRST(sv)) {
7861 sv_force_normal_flags(sv, 0);
7862 if (SvREADONLY(sv)) {
7863 if (IN_PERL_RUNTIME)
7864 Perl_croak_no_modify(aTHX);
7868 if (SvAMAGIC(sv) && AMG_CALLunary(sv, inc_amg))
7870 i = PTR2IV(SvRV(sv));
7875 flags = SvFLAGS(sv);
7876 if ((flags & (SVp_NOK|SVp_IOK)) == SVp_NOK) {
7877 /* It's (privately or publicly) a float, but not tested as an
7878 integer, so test it to see. */
7880 flags = SvFLAGS(sv);
7882 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
7883 /* It's publicly an integer, or privately an integer-not-float */
7884 #ifdef PERL_PRESERVE_IVUV
7888 if (SvUVX(sv) == UV_MAX)
7889 sv_setnv(sv, UV_MAX_P1);
7891 (void)SvIOK_only_UV(sv);
7892 SvUV_set(sv, SvUVX(sv) + 1);
7894 if (SvIVX(sv) == IV_MAX)
7895 sv_setuv(sv, (UV)IV_MAX + 1);
7897 (void)SvIOK_only(sv);
7898 SvIV_set(sv, SvIVX(sv) + 1);
7903 if (flags & SVp_NOK) {
7904 const NV was = SvNVX(sv);
7905 if (NV_OVERFLOWS_INTEGERS_AT &&
7906 was >= NV_OVERFLOWS_INTEGERS_AT) {
7907 Perl_ck_warner(aTHX_ packWARN(WARN_IMPRECISION),
7908 "Lost precision when incrementing %" NVff " by 1",
7911 (void)SvNOK_only(sv);
7912 SvNV_set(sv, was + 1.0);
7916 if (!(flags & SVp_POK) || !*SvPVX_const(sv)) {
7917 if ((flags & SVTYPEMASK) < SVt_PVIV)
7918 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV ? SVt_PVIV : SVt_IV));
7919 (void)SvIOK_only(sv);
7924 while (isALPHA(*d)) d++;
7925 while (isDIGIT(*d)) d++;
7926 if (d < SvEND(sv)) {
7927 #ifdef PERL_PRESERVE_IVUV
7928 /* Got to punt this as an integer if needs be, but we don't issue
7929 warnings. Probably ought to make the sv_iv_please() that does
7930 the conversion if possible, and silently. */
7931 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
7932 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
7933 /* Need to try really hard to see if it's an integer.
7934 9.22337203685478e+18 is an integer.
7935 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
7936 so $a="9.22337203685478e+18"; $a+0; $a++
7937 needs to be the same as $a="9.22337203685478e+18"; $a++
7944 /* sv_2iv *should* have made this an NV */
7945 if (flags & SVp_NOK) {
7946 (void)SvNOK_only(sv);
7947 SvNV_set(sv, SvNVX(sv) + 1.0);
7950 /* I don't think we can get here. Maybe I should assert this
7951 And if we do get here I suspect that sv_setnv will croak. NWC
7953 #if defined(USE_LONG_DOUBLE)
7954 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",
7955 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7957 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_inc punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
7958 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7961 #endif /* PERL_PRESERVE_IVUV */
7962 sv_setnv(sv,Atof(SvPVX_const(sv)) + 1.0);
7966 while (d >= SvPVX_const(sv)) {
7974 /* MKS: The original code here died if letters weren't consecutive.
7975 * at least it didn't have to worry about non-C locales. The
7976 * new code assumes that ('z'-'a')==('Z'-'A'), letters are
7977 * arranged in order (although not consecutively) and that only
7978 * [A-Za-z] are accepted by isALPHA in the C locale.
7980 if (*d != 'z' && *d != 'Z') {
7981 do { ++*d; } while (!isALPHA(*d));
7984 *(d--) -= 'z' - 'a';
7989 *(d--) -= 'z' - 'a' + 1;
7993 /* oh,oh, the number grew */
7994 SvGROW(sv, SvCUR(sv) + 2);
7995 SvCUR_set(sv, SvCUR(sv) + 1);
7996 for (d = SvPVX(sv) + SvCUR(sv); d > SvPVX_const(sv); d--)
8007 Auto-decrement of the value in the SV, doing string to numeric conversion
8008 if necessary. Handles 'get' magic and operator overloading.
8014 Perl_sv_dec(pTHX_ register SV *const sv)
8024 =for apidoc sv_dec_nomg
8026 Auto-decrement of the value in the SV, doing string to numeric conversion
8027 if necessary. Handles operator overloading. Skips handling 'get' magic.
8033 Perl_sv_dec_nomg(pTHX_ register SV *const sv)
8040 if (SvTHINKFIRST(sv)) {
8042 sv_force_normal_flags(sv, 0);
8043 if (SvREADONLY(sv)) {
8044 if (IN_PERL_RUNTIME)
8045 Perl_croak_no_modify(aTHX);
8049 if (SvAMAGIC(sv) && AMG_CALLunary(sv, dec_amg))
8051 i = PTR2IV(SvRV(sv));
8056 /* Unlike sv_inc we don't have to worry about string-never-numbers
8057 and keeping them magic. But we mustn't warn on punting */
8058 flags = SvFLAGS(sv);
8059 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
8060 /* It's publicly an integer, or privately an integer-not-float */
8061 #ifdef PERL_PRESERVE_IVUV
8065 if (SvUVX(sv) == 0) {
8066 (void)SvIOK_only(sv);
8070 (void)SvIOK_only_UV(sv);
8071 SvUV_set(sv, SvUVX(sv) - 1);
8074 if (SvIVX(sv) == IV_MIN) {
8075 sv_setnv(sv, (NV)IV_MIN);
8079 (void)SvIOK_only(sv);
8080 SvIV_set(sv, SvIVX(sv) - 1);
8085 if (flags & SVp_NOK) {
8088 const NV was = SvNVX(sv);
8089 if (NV_OVERFLOWS_INTEGERS_AT &&
8090 was <= -NV_OVERFLOWS_INTEGERS_AT) {
8091 Perl_ck_warner(aTHX_ packWARN(WARN_IMPRECISION),
8092 "Lost precision when decrementing %" NVff " by 1",
8095 (void)SvNOK_only(sv);
8096 SvNV_set(sv, was - 1.0);
8100 if (!(flags & SVp_POK)) {
8101 if ((flags & SVTYPEMASK) < SVt_PVIV)
8102 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV) ? SVt_PVIV : SVt_IV);
8104 (void)SvIOK_only(sv);
8107 #ifdef PERL_PRESERVE_IVUV
8109 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
8110 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
8111 /* Need to try really hard to see if it's an integer.
8112 9.22337203685478e+18 is an integer.
8113 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
8114 so $a="9.22337203685478e+18"; $a+0; $a--
8115 needs to be the same as $a="9.22337203685478e+18"; $a--
8122 /* sv_2iv *should* have made this an NV */
8123 if (flags & SVp_NOK) {
8124 (void)SvNOK_only(sv);
8125 SvNV_set(sv, SvNVX(sv) - 1.0);
8128 /* I don't think we can get here. Maybe I should assert this
8129 And if we do get here I suspect that sv_setnv will croak. NWC
8131 #if defined(USE_LONG_DOUBLE)
8132 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",
8133 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
8135 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_dec punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
8136 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
8140 #endif /* PERL_PRESERVE_IVUV */
8141 sv_setnv(sv,Atof(SvPVX_const(sv)) - 1.0); /* punt */
8144 /* this define is used to eliminate a chunk of duplicated but shared logic
8145 * it has the suffix __SV_C to signal that it isnt API, and isnt meant to be
8146 * used anywhere but here - yves
8148 #define PUSH_EXTEND_MORTAL__SV_C(AnSv) \
8151 PL_tmps_stack[++PL_tmps_ix] = (AnSv); \
8155 =for apidoc sv_mortalcopy
8157 Creates a new SV which is a copy of the original SV (using C<sv_setsv>).
8158 The new SV is marked as mortal. It will be destroyed "soon", either by an
8159 explicit call to FREETMPS, or by an implicit call at places such as
8160 statement boundaries. See also C<sv_newmortal> and C<sv_2mortal>.
8165 /* Make a string that will exist for the duration of the expression
8166 * evaluation. Actually, it may have to last longer than that, but
8167 * hopefully we won't free it until it has been assigned to a
8168 * permanent location. */
8171 Perl_sv_mortalcopy(pTHX_ SV *const oldstr)
8177 sv_setsv(sv,oldstr);
8178 PUSH_EXTEND_MORTAL__SV_C(sv);
8184 =for apidoc sv_newmortal
8186 Creates a new null SV which is mortal. The reference count of the SV is
8187 set to 1. It will be destroyed "soon", either by an explicit call to
8188 FREETMPS, or by an implicit call at places such as statement boundaries.
8189 See also C<sv_mortalcopy> and C<sv_2mortal>.
8195 Perl_sv_newmortal(pTHX)
8201 SvFLAGS(sv) = SVs_TEMP;
8202 PUSH_EXTEND_MORTAL__SV_C(sv);
8208 =for apidoc newSVpvn_flags
8210 Creates a new SV and copies a string into it. The reference count for the
8211 SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
8212 string. You are responsible for ensuring that the source string is at least
8213 C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
8214 Currently the only flag bits accepted are C<SVf_UTF8> and C<SVs_TEMP>.
8215 If C<SVs_TEMP> is set, then C<sv_2mortal()> is called on the result before
8216 returning. If C<SVf_UTF8> is set, C<s> is considered to be in UTF-8 and the
8217 C<SVf_UTF8> flag will be set on the new SV.
8218 C<newSVpvn_utf8()> is a convenience wrapper for this function, defined as
8220 #define newSVpvn_utf8(s, len, u) \
8221 newSVpvn_flags((s), (len), (u) ? SVf_UTF8 : 0)
8227 Perl_newSVpvn_flags(pTHX_ const char *const s, const STRLEN len, const U32 flags)
8232 /* All the flags we don't support must be zero.
8233 And we're new code so I'm going to assert this from the start. */
8234 assert(!(flags & ~(SVf_UTF8|SVs_TEMP)));
8236 sv_setpvn(sv,s,len);
8238 /* This code used to a sv_2mortal(), however we now unroll the call to sv_2mortal()
8239 * and do what it does ourselves here.
8240 * Since we have asserted that flags can only have the SVf_UTF8 and/or SVs_TEMP flags
8241 * set above we can use it to enable the sv flags directly (bypassing SvTEMP_on), which
8242 * in turn means we dont need to mask out the SVf_UTF8 flag below, which means that we
8243 * eliminate quite a few steps than it looks - Yves (explaining patch by gfx)
8246 SvFLAGS(sv) |= flags;
8248 if(flags & SVs_TEMP){
8249 PUSH_EXTEND_MORTAL__SV_C(sv);
8256 =for apidoc sv_2mortal
8258 Marks an existing SV as mortal. The SV will be destroyed "soon", either
8259 by an explicit call to FREETMPS, or by an implicit call at places such as
8260 statement boundaries. SvTEMP() is turned on which means that the SV's
8261 string buffer can be "stolen" if this SV is copied. See also C<sv_newmortal>
8262 and C<sv_mortalcopy>.
8268 Perl_sv_2mortal(pTHX_ register SV *const sv)
8273 if (SvREADONLY(sv) && SvIMMORTAL(sv))
8275 PUSH_EXTEND_MORTAL__SV_C(sv);
8283 Creates a new SV and copies a string into it. The reference count for the
8284 SV is set to 1. If C<len> is zero, Perl will compute the length using
8285 strlen(). For efficiency, consider using C<newSVpvn> instead.
8291 Perl_newSVpv(pTHX_ const char *const s, const STRLEN len)
8297 sv_setpvn(sv, s, len || s == NULL ? len : strlen(s));
8302 =for apidoc newSVpvn
8304 Creates a new SV and copies a string into it. The reference count for the
8305 SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
8306 string. You are responsible for ensuring that the source string is at least
8307 C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
8313 Perl_newSVpvn(pTHX_ const char *const s, const STRLEN len)
8319 sv_setpvn(sv,s,len);
8324 =for apidoc newSVhek
8326 Creates a new SV from the hash key structure. It will generate scalars that
8327 point to the shared string table where possible. Returns a new (undefined)
8328 SV if the hek is NULL.
8334 Perl_newSVhek(pTHX_ const HEK *const hek)
8344 if (HEK_LEN(hek) == HEf_SVKEY) {
8345 return newSVsv(*(SV**)HEK_KEY(hek));
8347 const int flags = HEK_FLAGS(hek);
8348 if (flags & HVhek_WASUTF8) {
8350 Andreas would like keys he put in as utf8 to come back as utf8
8352 STRLEN utf8_len = HEK_LEN(hek);
8353 SV * const sv = newSV_type(SVt_PV);
8354 char *as_utf8 = (char *)bytes_to_utf8 ((U8*)HEK_KEY(hek), &utf8_len);
8355 /* bytes_to_utf8() allocates a new string, which we can repurpose: */
8356 sv_usepvn_flags(sv, as_utf8, utf8_len, SV_HAS_TRAILING_NUL);
8359 } else if (flags & (HVhek_REHASH|HVhek_UNSHARED)) {
8360 /* We don't have a pointer to the hv, so we have to replicate the
8361 flag into every HEK. This hv is using custom a hasing
8362 algorithm. Hence we can't return a shared string scalar, as
8363 that would contain the (wrong) hash value, and might get passed
8364 into an hv routine with a regular hash.
8365 Similarly, a hash that isn't using shared hash keys has to have
8366 the flag in every key so that we know not to try to call
8367 share_hek_kek on it. */
8369 SV * const sv = newSVpvn (HEK_KEY(hek), HEK_LEN(hek));
8374 /* This will be overwhelminly the most common case. */
8376 /* Inline most of newSVpvn_share(), because share_hek_hek() is far
8377 more efficient than sharepvn(). */
8381 sv_upgrade(sv, SVt_PV);
8382 SvPV_set(sv, (char *)HEK_KEY(share_hek_hek(hek)));
8383 SvCUR_set(sv, HEK_LEN(hek));
8396 =for apidoc newSVpvn_share
8398 Creates a new SV with its SvPVX_const pointing to a shared string in the string
8399 table. If the string does not already exist in the table, it is created
8400 first. Turns on READONLY and FAKE. If the C<hash> parameter is non-zero, that
8401 value is used; otherwise the hash is computed. The string's hash can be later
8402 be retrieved from the SV with the C<SvSHARED_HASH()> macro. The idea here is
8403 that as the string table is used for shared hash keys these strings will have
8404 SvPVX_const == HeKEY and hash lookup will avoid string compare.
8410 Perl_newSVpvn_share(pTHX_ const char *src, I32 len, U32 hash)
8414 bool is_utf8 = FALSE;
8415 const char *const orig_src = src;
8418 STRLEN tmplen = -len;
8420 /* See the note in hv.c:hv_fetch() --jhi */
8421 src = (char*)bytes_from_utf8((const U8*)src, &tmplen, &is_utf8);
8425 PERL_HASH(hash, src, len);
8427 /* The logic for this is inlined in S_mro_get_linear_isa_dfs(), so if it
8428 changes here, update it there too. */
8429 sv_upgrade(sv, SVt_PV);
8430 SvPV_set(sv, sharepvn(src, is_utf8?-len:len, hash));
8438 if (src != orig_src)
8444 =for apidoc newSVpv_share
8446 Like C<newSVpvn_share>, but takes a nul-terminated string instead of a
8453 Perl_newSVpv_share(pTHX_ const char *src, U32 hash)
8455 return newSVpvn_share(src, strlen(src), hash);
8458 #if defined(PERL_IMPLICIT_CONTEXT)
8460 /* pTHX_ magic can't cope with varargs, so this is a no-context
8461 * version of the main function, (which may itself be aliased to us).
8462 * Don't access this version directly.
8466 Perl_newSVpvf_nocontext(const char *const pat, ...)
8472 PERL_ARGS_ASSERT_NEWSVPVF_NOCONTEXT;
8474 va_start(args, pat);
8475 sv = vnewSVpvf(pat, &args);
8482 =for apidoc newSVpvf
8484 Creates a new SV and initializes it with the string formatted like
8491 Perl_newSVpvf(pTHX_ const char *const pat, ...)
8496 PERL_ARGS_ASSERT_NEWSVPVF;
8498 va_start(args, pat);
8499 sv = vnewSVpvf(pat, &args);
8504 /* backend for newSVpvf() and newSVpvf_nocontext() */
8507 Perl_vnewSVpvf(pTHX_ const char *const pat, va_list *const args)
8512 PERL_ARGS_ASSERT_VNEWSVPVF;
8515 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
8522 Creates a new SV and copies a floating point value into it.
8523 The reference count for the SV is set to 1.
8529 Perl_newSVnv(pTHX_ const NV n)
8542 Creates a new SV and copies an integer into it. The reference count for the
8549 Perl_newSViv(pTHX_ const IV i)
8562 Creates a new SV and copies an unsigned integer into it.
8563 The reference count for the SV is set to 1.
8569 Perl_newSVuv(pTHX_ const UV u)
8580 =for apidoc newSV_type
8582 Creates a new SV, of the type specified. The reference count for the new SV
8589 Perl_newSV_type(pTHX_ const svtype type)
8594 sv_upgrade(sv, type);
8599 =for apidoc newRV_noinc
8601 Creates an RV wrapper for an SV. The reference count for the original
8602 SV is B<not> incremented.
8608 Perl_newRV_noinc(pTHX_ SV *const tmpRef)
8611 register SV *sv = newSV_type(SVt_IV);
8613 PERL_ARGS_ASSERT_NEWRV_NOINC;
8616 SvRV_set(sv, tmpRef);
8621 /* newRV_inc is the official function name to use now.
8622 * newRV_inc is in fact #defined to newRV in sv.h
8626 Perl_newRV(pTHX_ SV *const sv)
8630 PERL_ARGS_ASSERT_NEWRV;
8632 return newRV_noinc(SvREFCNT_inc_simple_NN(sv));
8638 Creates a new SV which is an exact duplicate of the original SV.
8645 Perl_newSVsv(pTHX_ register SV *const old)
8652 if (SvTYPE(old) == SVTYPEMASK) {
8653 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL), "semi-panic: attempt to dup freed string");
8657 /* SV_GMAGIC is the default for sv_setv()
8658 SV_NOSTEAL prevents TEMP buffers being, well, stolen, and saves games
8659 with SvTEMP_off and SvTEMP_on round a call to sv_setsv. */
8660 sv_setsv_flags(sv, old, SV_GMAGIC | SV_NOSTEAL);
8665 =for apidoc sv_reset
8667 Underlying implementation for the C<reset> Perl function.
8668 Note that the perl-level function is vaguely deprecated.
8674 Perl_sv_reset(pTHX_ register const char *s, HV *const stash)
8677 char todo[PERL_UCHAR_MAX+1];
8679 PERL_ARGS_ASSERT_SV_RESET;
8684 if (!*s) { /* reset ?? searches */
8685 MAGIC * const mg = mg_find((const SV *)stash, PERL_MAGIC_symtab);
8687 const U32 count = mg->mg_len / sizeof(PMOP**);
8688 PMOP **pmp = (PMOP**) mg->mg_ptr;
8689 PMOP *const *const end = pmp + count;
8693 SvREADONLY_off(PL_regex_pad[(*pmp)->op_pmoffset]);
8695 (*pmp)->op_pmflags &= ~PMf_USED;
8703 /* reset variables */
8705 if (!HvARRAY(stash))
8708 Zero(todo, 256, char);
8711 I32 i = (unsigned char)*s;
8715 max = (unsigned char)*s++;
8716 for ( ; i <= max; i++) {
8719 for (i = 0; i <= (I32) HvMAX(stash); i++) {
8721 for (entry = HvARRAY(stash)[i];
8723 entry = HeNEXT(entry))
8728 if (!todo[(U8)*HeKEY(entry)])
8730 gv = MUTABLE_GV(HeVAL(entry));
8733 if (SvTHINKFIRST(sv)) {
8734 if (!SvREADONLY(sv) && SvROK(sv))
8736 /* XXX Is this continue a bug? Why should THINKFIRST
8737 exempt us from resetting arrays and hashes? */
8741 if (SvTYPE(sv) >= SVt_PV) {
8743 if (SvPVX_const(sv) != NULL)
8751 if (GvHV(gv) && !HvNAME_get(GvHV(gv))) {
8753 Perl_die(aTHX_ "Can't reset %%ENV on this system");
8756 # if defined(USE_ENVIRON_ARRAY)
8759 # endif /* USE_ENVIRON_ARRAY */
8770 Using various gambits, try to get an IO from an SV: the IO slot if its a
8771 GV; or the recursive result if we're an RV; or the IO slot of the symbol
8772 named after the PV if we're a string.
8778 Perl_sv_2io(pTHX_ SV *const sv)
8783 PERL_ARGS_ASSERT_SV_2IO;
8785 switch (SvTYPE(sv)) {
8787 io = MUTABLE_IO(sv);
8791 if (isGV_with_GP(sv)) {
8792 gv = MUTABLE_GV(sv);
8795 Perl_croak(aTHX_ "Bad filehandle: %s", GvNAME(gv));
8801 Perl_croak(aTHX_ PL_no_usym, "filehandle");
8803 return sv_2io(SvRV(sv));
8804 gv = gv_fetchsv(sv, 0, SVt_PVIO);
8810 Perl_croak(aTHX_ "Bad filehandle: %"SVf, SVfARG(sv));
8819 Using various gambits, try to get a CV from an SV; in addition, try if
8820 possible to set C<*st> and C<*gvp> to the stash and GV associated with it.
8821 The flags in C<lref> are passed to gv_fetchsv.
8827 Perl_sv_2cv(pTHX_ SV *sv, HV **const st, GV **const gvp, const I32 lref)
8833 PERL_ARGS_ASSERT_SV_2CV;
8840 switch (SvTYPE(sv)) {
8844 return MUTABLE_CV(sv);
8851 if (isGV_with_GP(sv)) {
8852 gv = MUTABLE_GV(sv);
8863 sv = amagic_deref_call(sv, to_cv_amg);
8864 /* At this point I'd like to do SPAGAIN, but really I need to
8865 force it upon my callers. Hmmm. This is a mess... */
8868 if (SvTYPE(sv) == SVt_PVCV) {
8869 cv = MUTABLE_CV(sv);
8874 else if(isGV_with_GP(sv))
8875 gv = MUTABLE_GV(sv);
8877 Perl_croak(aTHX_ "Not a subroutine reference");
8879 else if (isGV_with_GP(sv)) {
8881 gv = MUTABLE_GV(sv);
8884 gv = gv_fetchsv(sv, lref, SVt_PVCV); /* Calls get magic */
8890 /* Some flags to gv_fetchsv mean don't really create the GV */
8891 if (!isGV_with_GP(gv)) {
8897 if (lref && !GvCVu(gv)) {
8901 gv_efullname3(tmpsv, gv, NULL);
8902 /* XXX this is probably not what they think they're getting.
8903 * It has the same effect as "sub name;", i.e. just a forward
8905 newSUB(start_subparse(FALSE, 0),
8906 newSVOP(OP_CONST, 0, tmpsv),
8910 Perl_croak(aTHX_ "Unable to create sub named \"%"SVf"\"",
8911 SVfARG(SvOK(sv) ? sv : &PL_sv_no));
8920 Returns true if the SV has a true value by Perl's rules.
8921 Use the C<SvTRUE> macro instead, which may call C<sv_true()> or may
8922 instead use an in-line version.
8928 Perl_sv_true(pTHX_ register SV *const sv)
8933 register const XPV* const tXpv = (XPV*)SvANY(sv);
8935 (tXpv->xpv_cur > 1 ||
8936 (tXpv->xpv_cur && *sv->sv_u.svu_pv != '0')))
8943 return SvIVX(sv) != 0;
8946 return SvNVX(sv) != 0.0;
8948 return sv_2bool(sv);
8954 =for apidoc sv_pvn_force
8956 Get a sensible string out of the SV somehow.
8957 A private implementation of the C<SvPV_force> macro for compilers which
8958 can't cope with complex macro expressions. Always use the macro instead.
8960 =for apidoc sv_pvn_force_flags
8962 Get a sensible string out of the SV somehow.
8963 If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<sv> if
8964 appropriate, else not. C<sv_pvn_force> and C<sv_pvn_force_nomg> are
8965 implemented in terms of this function.
8966 You normally want to use the various wrapper macros instead: see
8967 C<SvPV_force> and C<SvPV_force_nomg>
8973 Perl_sv_pvn_force_flags(pTHX_ SV *const sv, STRLEN *const lp, const I32 flags)
8977 PERL_ARGS_ASSERT_SV_PVN_FORCE_FLAGS;
8979 if (SvTHINKFIRST(sv) && !SvROK(sv))
8980 sv_force_normal_flags(sv, 0);
8990 if (SvREADONLY(sv) && !(flags & SV_MUTABLE_RETURN)) {
8991 const char * const ref = sv_reftype(sv,0);
8993 Perl_croak(aTHX_ "Can't coerce readonly %s to string in %s",
8994 ref, OP_DESC(PL_op));
8996 Perl_croak(aTHX_ "Can't coerce readonly %s to string", ref);
8998 if ((SvTYPE(sv) > SVt_PVLV && SvTYPE(sv) != SVt_PVFM)
8999 || isGV_with_GP(sv))
9000 /* diag_listed_as: Can't coerce %s to %s in %s */
9001 Perl_croak(aTHX_ "Can't coerce %s to string in %s", sv_reftype(sv,0),
9003 s = sv_2pv_flags(sv, &len, flags);
9007 if (s != SvPVX_const(sv)) { /* Almost, but not quite, sv_setpvn() */
9010 SvUPGRADE(sv, SVt_PV); /* Never FALSE */
9011 SvGROW(sv, len + 1);
9012 Move(s,SvPVX(sv),len,char);
9014 SvPVX(sv)[len] = '\0';
9017 SvPOK_on(sv); /* validate pointer */
9019 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
9020 PTR2UV(sv),SvPVX_const(sv)));
9023 return SvPVX_mutable(sv);
9027 =for apidoc sv_pvbyten_force
9029 The backend for the C<SvPVbytex_force> macro. Always use the macro instead.
9035 Perl_sv_pvbyten_force(pTHX_ SV *const sv, STRLEN *const lp)
9037 PERL_ARGS_ASSERT_SV_PVBYTEN_FORCE;
9039 sv_pvn_force(sv,lp);
9040 sv_utf8_downgrade(sv,0);
9046 =for apidoc sv_pvutf8n_force
9048 The backend for the C<SvPVutf8x_force> macro. Always use the macro instead.
9054 Perl_sv_pvutf8n_force(pTHX_ SV *const sv, STRLEN *const lp)
9056 PERL_ARGS_ASSERT_SV_PVUTF8N_FORCE;
9058 sv_pvn_force(sv,lp);
9059 sv_utf8_upgrade(sv);
9065 =for apidoc sv_reftype
9067 Returns a string describing what the SV is a reference to.
9073 Perl_sv_reftype(pTHX_ const SV *const sv, const int ob)
9075 PERL_ARGS_ASSERT_SV_REFTYPE;
9077 /* The fact that I don't need to downcast to char * everywhere, only in ?:
9078 inside return suggests a const propagation bug in g++. */
9079 if (ob && SvOBJECT(sv)) {
9080 char * const name = HvNAME_get(SvSTASH(sv));
9081 return name ? name : (char *) "__ANON__";
9084 switch (SvTYPE(sv)) {
9099 case SVt_PVLV: return (char *) (SvROK(sv) ? "REF"
9100 /* tied lvalues should appear to be
9101 * scalars for backwards compatibility */
9102 : (LvTYPE(sv) == 't' || LvTYPE(sv) == 'T')
9103 ? "SCALAR" : "LVALUE");
9104 case SVt_PVAV: return "ARRAY";
9105 case SVt_PVHV: return "HASH";
9106 case SVt_PVCV: return "CODE";
9107 case SVt_PVGV: return (char *) (isGV_with_GP(sv)
9108 ? "GLOB" : "SCALAR");
9109 case SVt_PVFM: return "FORMAT";
9110 case SVt_PVIO: return "IO";
9111 case SVt_BIND: return "BIND";
9112 case SVt_REGEXP: return "REGEXP";
9113 default: return "UNKNOWN";
9119 =for apidoc sv_isobject
9121 Returns a boolean indicating whether the SV is an RV pointing to a blessed
9122 object. If the SV is not an RV, or if the object is not blessed, then this
9129 Perl_sv_isobject(pTHX_ SV *sv)
9145 Returns a boolean indicating whether the SV is blessed into the specified
9146 class. This does not check for subtypes; use C<sv_derived_from> to verify
9147 an inheritance relationship.
9153 Perl_sv_isa(pTHX_ SV *sv, const char *const name)
9157 PERL_ARGS_ASSERT_SV_ISA;
9167 hvname = HvNAME_get(SvSTASH(sv));
9171 return strEQ(hvname, name);
9177 Creates a new SV for the RV, C<rv>, to point to. If C<rv> is not an RV then
9178 it will be upgraded to one. If C<classname> is non-null then the new SV will
9179 be blessed in the specified package. The new SV is returned and its
9180 reference count is 1.
9186 Perl_newSVrv(pTHX_ SV *const rv, const char *const classname)
9191 PERL_ARGS_ASSERT_NEWSVRV;
9195 SV_CHECK_THINKFIRST_COW_DROP(rv);
9196 (void)SvAMAGIC_off(rv);
9198 if (SvTYPE(rv) >= SVt_PVMG) {
9199 const U32 refcnt = SvREFCNT(rv);
9203 SvREFCNT(rv) = refcnt;
9205 sv_upgrade(rv, SVt_IV);
9206 } else if (SvROK(rv)) {
9207 SvREFCNT_dec(SvRV(rv));
9209 prepare_SV_for_RV(rv);
9217 HV* const stash = gv_stashpv(classname, GV_ADD);
9218 (void)sv_bless(rv, stash);
9224 =for apidoc sv_setref_pv
9226 Copies a pointer into a new SV, optionally blessing the SV. The C<rv>
9227 argument will be upgraded to an RV. That RV will be modified to point to
9228 the new SV. If the C<pv> argument is NULL then C<PL_sv_undef> will be placed
9229 into the SV. The C<classname> argument indicates the package for the
9230 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9231 will have a reference count of 1, and the RV will be returned.
9233 Do not use with other Perl types such as HV, AV, SV, CV, because those
9234 objects will become corrupted by the pointer copy process.
9236 Note that C<sv_setref_pvn> copies the string while this copies the pointer.
9242 Perl_sv_setref_pv(pTHX_ SV *const rv, const char *const classname, void *const pv)
9246 PERL_ARGS_ASSERT_SV_SETREF_PV;
9249 sv_setsv(rv, &PL_sv_undef);
9253 sv_setiv(newSVrv(rv,classname), PTR2IV(pv));
9258 =for apidoc sv_setref_iv
9260 Copies an integer into a new SV, optionally blessing the SV. The C<rv>
9261 argument will be upgraded to an RV. That RV will be modified to point to
9262 the new SV. The C<classname> argument indicates the package for the
9263 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9264 will have a reference count of 1, and the RV will be returned.
9270 Perl_sv_setref_iv(pTHX_ SV *const rv, const char *const classname, const IV iv)
9272 PERL_ARGS_ASSERT_SV_SETREF_IV;
9274 sv_setiv(newSVrv(rv,classname), iv);
9279 =for apidoc sv_setref_uv
9281 Copies an unsigned integer into a new SV, optionally blessing the SV. The C<rv>
9282 argument will be upgraded to an RV. That RV will be modified to point to
9283 the new SV. The C<classname> argument indicates the package for the
9284 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9285 will have a reference count of 1, and the RV will be returned.
9291 Perl_sv_setref_uv(pTHX_ SV *const rv, const char *const classname, const UV uv)
9293 PERL_ARGS_ASSERT_SV_SETREF_UV;
9295 sv_setuv(newSVrv(rv,classname), uv);
9300 =for apidoc sv_setref_nv
9302 Copies a double into a new SV, optionally blessing the SV. The C<rv>
9303 argument will be upgraded to an RV. That RV will be modified to point to
9304 the new SV. The C<classname> argument indicates the package for the
9305 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9306 will have a reference count of 1, and the RV will be returned.
9312 Perl_sv_setref_nv(pTHX_ SV *const rv, const char *const classname, const NV nv)
9314 PERL_ARGS_ASSERT_SV_SETREF_NV;
9316 sv_setnv(newSVrv(rv,classname), nv);
9321 =for apidoc sv_setref_pvn
9323 Copies a string into a new SV, optionally blessing the SV. The length of the
9324 string must be specified with C<n>. The C<rv> argument will be upgraded to
9325 an RV. That RV will be modified to point to the new SV. The C<classname>
9326 argument indicates the package for the blessing. Set C<classname> to
9327 C<NULL> to avoid the blessing. The new SV will have a reference count
9328 of 1, and the RV will be returned.
9330 Note that C<sv_setref_pv> copies the pointer while this copies the string.
9336 Perl_sv_setref_pvn(pTHX_ SV *const rv, const char *const classname,
9337 const char *const pv, const STRLEN n)
9339 PERL_ARGS_ASSERT_SV_SETREF_PVN;
9341 sv_setpvn(newSVrv(rv,classname), pv, n);
9346 =for apidoc sv_bless
9348 Blesses an SV into a specified package. The SV must be an RV. The package
9349 must be designated by its stash (see C<gv_stashpv()>). The reference count
9350 of the SV is unaffected.
9356 Perl_sv_bless(pTHX_ SV *const sv, HV *const stash)
9361 PERL_ARGS_ASSERT_SV_BLESS;
9364 Perl_croak(aTHX_ "Can't bless non-reference value");
9366 if (SvFLAGS(tmpRef) & (SVs_OBJECT|SVf_READONLY)) {
9367 if (SvIsCOW(tmpRef))
9368 sv_force_normal_flags(tmpRef, 0);
9369 if (SvREADONLY(tmpRef))
9370 Perl_croak_no_modify(aTHX);
9371 if (SvOBJECT(tmpRef)) {
9372 if (SvTYPE(tmpRef) != SVt_PVIO)
9374 SvREFCNT_dec(SvSTASH(tmpRef));
9377 SvOBJECT_on(tmpRef);
9378 if (SvTYPE(tmpRef) != SVt_PVIO)
9380 SvUPGRADE(tmpRef, SVt_PVMG);
9381 SvSTASH_set(tmpRef, MUTABLE_HV(SvREFCNT_inc_simple(stash)));
9386 (void)SvAMAGIC_off(sv);
9388 if(SvSMAGICAL(tmpRef))
9389 if(mg_find(tmpRef, PERL_MAGIC_ext) || mg_find(tmpRef, PERL_MAGIC_uvar))
9397 /* Downgrades a PVGV to a PVMG. If it’s actually a PVLV, we leave the type
9398 * as it is after unglobbing it.
9402 S_sv_unglob(pTHX_ SV *const sv)
9407 SV * const temp = sv_newmortal();
9409 PERL_ARGS_ASSERT_SV_UNGLOB;
9411 assert(SvTYPE(sv) == SVt_PVGV || SvTYPE(sv) == SVt_PVLV);
9413 gv_efullname3(temp, MUTABLE_GV(sv), "*");
9416 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
9417 && HvNAME_get(stash))
9418 mro_method_changed_in(stash);
9419 gp_free(MUTABLE_GV(sv));
9422 sv_del_backref(MUTABLE_SV(GvSTASH(sv)), sv);
9426 if (GvNAME_HEK(sv)) {
9427 unshare_hek(GvNAME_HEK(sv));
9429 isGV_with_GP_off(sv);
9431 if(SvTYPE(sv) == SVt_PVGV) {
9432 /* need to keep SvANY(sv) in the right arena */
9433 xpvmg = new_XPVMG();
9434 StructCopy(SvANY(sv), xpvmg, XPVMG);
9435 del_XPVGV(SvANY(sv));
9438 SvFLAGS(sv) &= ~SVTYPEMASK;
9439 SvFLAGS(sv) |= SVt_PVMG;
9442 /* Intentionally not calling any local SET magic, as this isn't so much a
9443 set operation as merely an internal storage change. */
9444 sv_setsv_flags(sv, temp, 0);
9448 =for apidoc sv_unref_flags
9450 Unsets the RV status of the SV, and decrements the reference count of
9451 whatever was being referenced by the RV. This can almost be thought of
9452 as a reversal of C<newSVrv>. The C<cflags> argument can contain
9453 C<SV_IMMEDIATE_UNREF> to force the reference count to be decremented
9454 (otherwise the decrementing is conditional on the reference count being
9455 different from one or the reference being a readonly SV).
9462 Perl_sv_unref_flags(pTHX_ SV *const ref, const U32 flags)
9464 SV* const target = SvRV(ref);
9466 PERL_ARGS_ASSERT_SV_UNREF_FLAGS;
9468 if (SvWEAKREF(ref)) {
9469 sv_del_backref(target, ref);
9471 SvRV_set(ref, NULL);
9474 SvRV_set(ref, NULL);
9476 /* You can't have a || SvREADONLY(target) here, as $a = $$a, where $a was
9477 assigned to as BEGIN {$a = \"Foo"} will fail. */
9478 if (SvREFCNT(target) != 1 || (flags & SV_IMMEDIATE_UNREF))
9479 SvREFCNT_dec(target);
9480 else /* XXX Hack, but hard to make $a=$a->[1] work otherwise */
9481 sv_2mortal(target); /* Schedule for freeing later */
9485 =for apidoc sv_untaint
9487 Untaint an SV. Use C<SvTAINTED_off> instead.
9492 Perl_sv_untaint(pTHX_ SV *const sv)
9494 PERL_ARGS_ASSERT_SV_UNTAINT;
9496 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
9497 MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
9504 =for apidoc sv_tainted
9506 Test an SV for taintedness. Use C<SvTAINTED> instead.
9511 Perl_sv_tainted(pTHX_ SV *const sv)
9513 PERL_ARGS_ASSERT_SV_TAINTED;
9515 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
9516 const MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
9517 if (mg && (mg->mg_len & 1) )
9524 =for apidoc sv_setpviv
9526 Copies an integer into the given SV, also updating its string value.
9527 Does not handle 'set' magic. See C<sv_setpviv_mg>.
9533 Perl_sv_setpviv(pTHX_ SV *const sv, const IV iv)
9535 char buf[TYPE_CHARS(UV)];
9537 char * const ptr = uiv_2buf(buf, iv, 0, 0, &ebuf);
9539 PERL_ARGS_ASSERT_SV_SETPVIV;
9541 sv_setpvn(sv, ptr, ebuf - ptr);
9545 =for apidoc sv_setpviv_mg
9547 Like C<sv_setpviv>, but also handles 'set' magic.
9553 Perl_sv_setpviv_mg(pTHX_ SV *const sv, const IV iv)
9555 PERL_ARGS_ASSERT_SV_SETPVIV_MG;
9561 #if defined(PERL_IMPLICIT_CONTEXT)
9563 /* pTHX_ magic can't cope with varargs, so this is a no-context
9564 * version of the main function, (which may itself be aliased to us).
9565 * Don't access this version directly.
9569 Perl_sv_setpvf_nocontext(SV *const sv, const char *const pat, ...)
9574 PERL_ARGS_ASSERT_SV_SETPVF_NOCONTEXT;
9576 va_start(args, pat);
9577 sv_vsetpvf(sv, pat, &args);
9581 /* pTHX_ magic can't cope with varargs, so this is a no-context
9582 * version of the main function, (which may itself be aliased to us).
9583 * Don't access this version directly.
9587 Perl_sv_setpvf_mg_nocontext(SV *const sv, const char *const pat, ...)
9592 PERL_ARGS_ASSERT_SV_SETPVF_MG_NOCONTEXT;
9594 va_start(args, pat);
9595 sv_vsetpvf_mg(sv, pat, &args);
9601 =for apidoc sv_setpvf
9603 Works like C<sv_catpvf> but copies the text into the SV instead of
9604 appending it. Does not handle 'set' magic. See C<sv_setpvf_mg>.
9610 Perl_sv_setpvf(pTHX_ SV *const sv, const char *const pat, ...)
9614 PERL_ARGS_ASSERT_SV_SETPVF;
9616 va_start(args, pat);
9617 sv_vsetpvf(sv, pat, &args);
9622 =for apidoc sv_vsetpvf
9624 Works like C<sv_vcatpvf> but copies the text into the SV instead of
9625 appending it. Does not handle 'set' magic. See C<sv_vsetpvf_mg>.
9627 Usually used via its frontend C<sv_setpvf>.
9633 Perl_sv_vsetpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9635 PERL_ARGS_ASSERT_SV_VSETPVF;
9637 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9641 =for apidoc sv_setpvf_mg
9643 Like C<sv_setpvf>, but also handles 'set' magic.
9649 Perl_sv_setpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
9653 PERL_ARGS_ASSERT_SV_SETPVF_MG;
9655 va_start(args, pat);
9656 sv_vsetpvf_mg(sv, pat, &args);
9661 =for apidoc sv_vsetpvf_mg
9663 Like C<sv_vsetpvf>, but also handles 'set' magic.
9665 Usually used via its frontend C<sv_setpvf_mg>.
9671 Perl_sv_vsetpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9673 PERL_ARGS_ASSERT_SV_VSETPVF_MG;
9675 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9679 #if defined(PERL_IMPLICIT_CONTEXT)
9681 /* pTHX_ magic can't cope with varargs, so this is a no-context
9682 * version of the main function, (which may itself be aliased to us).
9683 * Don't access this version directly.
9687 Perl_sv_catpvf_nocontext(SV *const sv, const char *const pat, ...)
9692 PERL_ARGS_ASSERT_SV_CATPVF_NOCONTEXT;
9694 va_start(args, pat);
9695 sv_vcatpvf(sv, pat, &args);
9699 /* pTHX_ magic can't cope with varargs, so this is a no-context
9700 * version of the main function, (which may itself be aliased to us).
9701 * Don't access this version directly.
9705 Perl_sv_catpvf_mg_nocontext(SV *const sv, const char *const pat, ...)
9710 PERL_ARGS_ASSERT_SV_CATPVF_MG_NOCONTEXT;
9712 va_start(args, pat);
9713 sv_vcatpvf_mg(sv, pat, &args);
9719 =for apidoc sv_catpvf
9721 Processes its arguments like C<sprintf> and appends the formatted
9722 output to an SV. If the appended data contains "wide" characters
9723 (including, but not limited to, SVs with a UTF-8 PV formatted with %s,
9724 and characters >255 formatted with %c), the original SV might get
9725 upgraded to UTF-8. Handles 'get' magic, but not 'set' magic. See
9726 C<sv_catpvf_mg>. If the original SV was UTF-8, the pattern should be
9727 valid UTF-8; if the original SV was bytes, the pattern should be too.
9732 Perl_sv_catpvf(pTHX_ SV *const sv, const char *const pat, ...)
9736 PERL_ARGS_ASSERT_SV_CATPVF;
9738 va_start(args, pat);
9739 sv_vcatpvf(sv, pat, &args);
9744 =for apidoc sv_vcatpvf
9746 Processes its arguments like C<vsprintf> and appends the formatted output
9747 to an SV. Does not handle 'set' magic. See C<sv_vcatpvf_mg>.
9749 Usually used via its frontend C<sv_catpvf>.
9755 Perl_sv_vcatpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9757 PERL_ARGS_ASSERT_SV_VCATPVF;
9759 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9763 =for apidoc sv_catpvf_mg
9765 Like C<sv_catpvf>, but also handles 'set' magic.
9771 Perl_sv_catpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
9775 PERL_ARGS_ASSERT_SV_CATPVF_MG;
9777 va_start(args, pat);
9778 sv_vcatpvf_mg(sv, pat, &args);
9783 =for apidoc sv_vcatpvf_mg
9785 Like C<sv_vcatpvf>, but also handles 'set' magic.
9787 Usually used via its frontend C<sv_catpvf_mg>.
9793 Perl_sv_vcatpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9795 PERL_ARGS_ASSERT_SV_VCATPVF_MG;
9797 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9802 =for apidoc sv_vsetpvfn
9804 Works like C<sv_vcatpvfn> but copies the text into the SV instead of
9807 Usually used via one of its frontends C<sv_vsetpvf> and C<sv_vsetpvf_mg>.
9813 Perl_sv_vsetpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
9814 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
9816 PERL_ARGS_ASSERT_SV_VSETPVFN;
9819 sv_vcatpvfn(sv, pat, patlen, args, svargs, svmax, maybe_tainted);
9824 * Warn of missing argument to sprintf, and then return a defined value
9825 * to avoid inappropriate "use of uninit" warnings [perl #71000].
9827 #define WARN_MISSING WARN_UNINITIALIZED /* Not sure we want a new category */
9829 S_vcatpvfn_missing_argument(pTHX) {
9830 if (ckWARN(WARN_MISSING)) {
9831 Perl_warner(aTHX_ packWARN(WARN_MISSING), "Missing argument in %s",
9832 PL_op ? OP_DESC(PL_op) : "sv_vcatpvfn()");
9839 S_expect_number(pTHX_ char **const pattern)
9844 PERL_ARGS_ASSERT_EXPECT_NUMBER;
9846 switch (**pattern) {
9847 case '1': case '2': case '3':
9848 case '4': case '5': case '6':
9849 case '7': case '8': case '9':
9850 var = *(*pattern)++ - '0';
9851 while (isDIGIT(**pattern)) {
9852 const I32 tmp = var * 10 + (*(*pattern)++ - '0');
9854 Perl_croak(aTHX_ "Integer overflow in format string for %s", (PL_op ? OP_DESC(PL_op) : "sv_vcatpvfn"));
9862 S_F0convert(NV nv, char *const endbuf, STRLEN *const len)
9864 const int neg = nv < 0;
9867 PERL_ARGS_ASSERT_F0CONVERT;
9875 if (uv & 1 && uv == nv)
9876 uv--; /* Round to even */
9878 const unsigned dig = uv % 10;
9891 =for apidoc sv_vcatpvfn
9893 Processes its arguments like C<vsprintf> and appends the formatted output
9894 to an SV. Uses an array of SVs if the C style variable argument list is
9895 missing (NULL). When running with taint checks enabled, indicates via
9896 C<maybe_tainted> if results are untrustworthy (often due to the use of
9899 Usually used via one of its frontends C<sv_vcatpvf> and C<sv_vcatpvf_mg>.
9905 #define VECTORIZE_ARGS vecsv = va_arg(*args, SV*);\
9906 vecstr = (U8*)SvPV_const(vecsv,veclen);\
9907 vec_utf8 = DO_UTF8(vecsv);
9909 /* XXX maybe_tainted is never assigned to, so the doc above is lying. */
9912 Perl_sv_vcatpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
9913 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
9921 static const char nullstr[] = "(null)";
9923 bool has_utf8 = DO_UTF8(sv); /* has the result utf8? */
9924 const bool pat_utf8 = has_utf8; /* the pattern is in utf8? */
9926 /* Times 4: a decimal digit takes more than 3 binary digits.
9927 * NV_DIG: mantissa takes than many decimal digits.
9928 * Plus 32: Playing safe. */
9929 char ebuf[IV_DIG * 4 + NV_DIG + 32];
9930 /* large enough for "%#.#f" --chip */
9931 /* what about long double NVs? --jhi */
9933 PERL_ARGS_ASSERT_SV_VCATPVFN;
9934 PERL_UNUSED_ARG(maybe_tainted);
9936 /* no matter what, this is a string now */
9937 (void)SvPV_force(sv, origlen);
9939 /* special-case "", "%s", and "%-p" (SVf - see below) */
9942 if (patlen == 2 && pat[0] == '%' && pat[1] == 's') {
9944 const char * const s = va_arg(*args, char*);
9945 sv_catpv(sv, s ? s : nullstr);
9947 else if (svix < svmax) {
9948 sv_catsv(sv, *svargs);
9951 S_vcatpvfn_missing_argument(aTHX);
9954 if (args && patlen == 3 && pat[0] == '%' &&
9955 pat[1] == '-' && pat[2] == 'p') {
9956 argsv = MUTABLE_SV(va_arg(*args, void*));
9957 sv_catsv(sv, argsv);
9961 #ifndef USE_LONG_DOUBLE
9962 /* special-case "%.<number>[gf]" */
9963 if ( !args && patlen <= 5 && pat[0] == '%' && pat[1] == '.'
9964 && (pat[patlen-1] == 'g' || pat[patlen-1] == 'f') ) {
9965 unsigned digits = 0;
9969 while (*pp >= '0' && *pp <= '9')
9970 digits = 10 * digits + (*pp++ - '0');
9971 if (pp - pat == (int)patlen - 1 && svix < svmax) {
9972 const NV nv = SvNV(*svargs);
9974 /* Add check for digits != 0 because it seems that some
9975 gconverts are buggy in this case, and we don't yet have
9976 a Configure test for this. */
9977 if (digits && digits < sizeof(ebuf) - NV_DIG - 10) {
9978 /* 0, point, slack */
9979 Gconvert(nv, (int)digits, 0, ebuf);
9981 if (*ebuf) /* May return an empty string for digits==0 */
9984 } else if (!digits) {
9987 if ((p = F0convert(nv, ebuf + sizeof ebuf, &l))) {
9988 sv_catpvn(sv, p, l);
9994 #endif /* !USE_LONG_DOUBLE */
9996 if (!args && svix < svmax && DO_UTF8(*svargs))
9999 patend = (char*)pat + patlen;
10000 for (p = (char*)pat; p < patend; p = q) {
10003 bool vectorize = FALSE;
10004 bool vectorarg = FALSE;
10005 bool vec_utf8 = FALSE;
10011 bool has_precis = FALSE;
10013 const I32 osvix = svix;
10014 bool is_utf8 = FALSE; /* is this item utf8? */
10015 #ifdef HAS_LDBL_SPRINTF_BUG
10016 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
10017 with sfio - Allen <allens@cpan.org> */
10018 bool fix_ldbl_sprintf_bug = FALSE;
10022 U8 utf8buf[UTF8_MAXBYTES+1];
10023 STRLEN esignlen = 0;
10025 const char *eptr = NULL;
10026 const char *fmtstart;
10029 const U8 *vecstr = NULL;
10036 /* we need a long double target in case HAS_LONG_DOUBLE but
10037 not USE_LONG_DOUBLE
10039 #if defined(HAS_LONG_DOUBLE) && LONG_DOUBLESIZE > DOUBLESIZE
10047 const char *dotstr = ".";
10048 STRLEN dotstrlen = 1;
10049 I32 efix = 0; /* explicit format parameter index */
10050 I32 ewix = 0; /* explicit width index */
10051 I32 epix = 0; /* explicit precision index */
10052 I32 evix = 0; /* explicit vector index */
10053 bool asterisk = FALSE;
10055 /* echo everything up to the next format specification */
10056 for (q = p; q < patend && *q != '%'; ++q) ;
10058 if (has_utf8 && !pat_utf8)
10059 sv_catpvn_utf8_upgrade(sv, p, q - p, nsv);
10061 sv_catpvn(sv, p, q - p);
10070 We allow format specification elements in this order:
10071 \d+\$ explicit format parameter index
10073 v|\*(\d+\$)?v vector with optional (optionally specified) arg
10074 0 flag (as above): repeated to allow "v02"
10075 \d+|\*(\d+\$)? width using optional (optionally specified) arg
10076 \.(\d*|\*(\d+\$)?) precision using optional (optionally specified) arg
10078 [%bcdefginopsuxDFOUX] format (mandatory)
10083 As of perl5.9.3, printf format checking is on by default.
10084 Internally, perl uses %p formats to provide an escape to
10085 some extended formatting. This block deals with those
10086 extensions: if it does not match, (char*)q is reset and
10087 the normal format processing code is used.
10089 Currently defined extensions are:
10090 %p include pointer address (standard)
10091 %-p (SVf) include an SV (previously %_)
10092 %-<num>p include an SV with precision <num>
10093 %<num>p reserved for future extensions
10095 Robin Barker 2005-07-14
10097 %1p (VDf) removed. RMB 2007-10-19
10104 n = expect_number(&q);
10106 if (sv) { /* SVf */
10111 argsv = MUTABLE_SV(va_arg(*args, void*));
10112 eptr = SvPV_const(argsv, elen);
10113 if (DO_UTF8(argsv))
10118 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL),
10119 "internal %%<num>p might conflict with future printf extensions");
10125 if ( (width = expect_number(&q)) ) {
10140 if (plus == '+' && *q == ' ') /* '+' over ' ' */
10169 if ( (ewix = expect_number(&q)) )
10178 if ((vectorarg = asterisk)) {
10191 width = expect_number(&q);
10194 if (vectorize && vectorarg) {
10195 /* vectorizing, but not with the default "." */
10197 vecsv = va_arg(*args, SV*);
10199 vecsv = (evix > 0 && evix <= svmax)
10200 ? svargs[evix-1] : S_vcatpvfn_missing_argument(aTHX);
10202 vecsv = svix < svmax
10203 ? svargs[svix++] : S_vcatpvfn_missing_argument(aTHX);
10205 dotstr = SvPV_const(vecsv, dotstrlen);
10206 /* Keep the DO_UTF8 test *after* the SvPV call, else things go
10207 bad with tied or overloaded values that return UTF8. */
10208 if (DO_UTF8(vecsv))
10210 else if (has_utf8) {
10211 vecsv = sv_mortalcopy(vecsv);
10212 sv_utf8_upgrade(vecsv);
10213 dotstr = SvPV_const(vecsv, dotstrlen);
10220 i = va_arg(*args, int);
10222 i = (ewix ? ewix <= svmax : svix < svmax) ?
10223 SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
10225 width = (i < 0) ? -i : i;
10235 if ( ((epix = expect_number(&q))) && (*q++ != '$') )
10237 /* XXX: todo, support specified precision parameter */
10241 i = va_arg(*args, int);
10243 i = (ewix ? ewix <= svmax : svix < svmax)
10244 ? SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
10246 has_precis = !(i < 0);
10250 while (isDIGIT(*q))
10251 precis = precis * 10 + (*q++ - '0');
10260 else if (efix ? (efix > 0 && efix <= svmax) : svix < svmax) {
10261 vecsv = svargs[efix ? efix-1 : svix++];
10262 vecstr = (U8*)SvPV_const(vecsv,veclen);
10263 vec_utf8 = DO_UTF8(vecsv);
10265 /* if this is a version object, we need to convert
10266 * back into v-string notation and then let the
10267 * vectorize happen normally
10269 if (sv_derived_from(vecsv, "version")) {
10270 char *version = savesvpv(vecsv);
10271 if ( hv_exists(MUTABLE_HV(SvRV(vecsv)), "alpha", 5 ) ) {
10272 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
10273 "vector argument not supported with alpha versions");
10276 vecsv = sv_newmortal();
10277 scan_vstring(version, version + veclen, vecsv);
10278 vecstr = (U8*)SvPV_const(vecsv, veclen);
10279 vec_utf8 = DO_UTF8(vecsv);
10293 case 'I': /* Ix, I32x, and I64x */
10295 if (q[1] == '6' && q[2] == '4') {
10301 if (q[1] == '3' && q[2] == '2') {
10311 #if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
10323 #if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
10324 if (*q == 'l') { /* lld, llf */
10333 if (*++q == 'h') { /* hhd, hhu */
10362 if (!vectorize && !args) {
10364 const I32 i = efix-1;
10365 argsv = (i >= 0 && i < svmax)
10366 ? svargs[i] : S_vcatpvfn_missing_argument(aTHX);
10368 argsv = (svix >= 0 && svix < svmax)
10369 ? svargs[svix++] : S_vcatpvfn_missing_argument(aTHX);
10373 switch (c = *q++) {
10380 uv = (args) ? va_arg(*args, int) : SvIV(argsv);
10382 (!UNI_IS_INVARIANT(uv) && SvUTF8(sv)))
10384 eptr = (char*)utf8buf;
10385 elen = uvchr_to_utf8((U8*)eptr, uv) - utf8buf;
10399 eptr = va_arg(*args, char*);
10401 elen = strlen(eptr);
10403 eptr = (char *)nullstr;
10404 elen = sizeof nullstr - 1;
10408 eptr = SvPV_const(argsv, elen);
10409 if (DO_UTF8(argsv)) {
10410 STRLEN old_precis = precis;
10411 if (has_precis && precis < elen) {
10412 STRLEN ulen = sv_len_utf8(argsv);
10413 I32 p = precis > ulen ? ulen : precis;
10414 sv_pos_u2b(argsv, &p, 0); /* sticks at end */
10417 if (width) { /* fudge width (can't fudge elen) */
10418 if (has_precis && precis < elen)
10419 width += precis - old_precis;
10421 width += elen - sv_len_utf8(argsv);
10428 if (has_precis && precis < elen)
10435 if (alt || vectorize)
10437 uv = PTR2UV(args ? va_arg(*args, void*) : argsv);
10458 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
10467 esignbuf[esignlen++] = plus;
10471 case 'c': iv = (char)va_arg(*args, int); break;
10472 case 'h': iv = (short)va_arg(*args, int); break;
10473 case 'l': iv = va_arg(*args, long); break;
10474 case 'V': iv = va_arg(*args, IV); break;
10475 case 'z': iv = va_arg(*args, SSize_t); break;
10476 case 't': iv = va_arg(*args, ptrdiff_t); break;
10477 default: iv = va_arg(*args, int); break;
10479 case 'j': iv = va_arg(*args, intmax_t); break;
10483 iv = va_arg(*args, Quad_t); break;
10490 IV tiv = SvIV(argsv); /* work around GCC bug #13488 */
10492 case 'c': iv = (char)tiv; break;
10493 case 'h': iv = (short)tiv; break;
10494 case 'l': iv = (long)tiv; break;
10496 default: iv = tiv; break;
10499 iv = (Quad_t)tiv; break;
10505 if ( !vectorize ) /* we already set uv above */
10510 esignbuf[esignlen++] = plus;
10514 esignbuf[esignlen++] = '-';
10558 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
10569 case 'c': uv = (unsigned char)va_arg(*args, unsigned); break;
10570 case 'h': uv = (unsigned short)va_arg(*args, unsigned); break;
10571 case 'l': uv = va_arg(*args, unsigned long); break;
10572 case 'V': uv = va_arg(*args, UV); break;
10573 case 'z': uv = va_arg(*args, Size_t); break;
10574 case 't': uv = va_arg(*args, ptrdiff_t); break; /* will sign extend, but there is no uptrdiff_t, so oh well */
10576 case 'j': uv = va_arg(*args, uintmax_t); break;
10578 default: uv = va_arg(*args, unsigned); break;
10581 uv = va_arg(*args, Uquad_t); break;
10588 UV tuv = SvUV(argsv); /* work around GCC bug #13488 */
10590 case 'c': uv = (unsigned char)tuv; break;
10591 case 'h': uv = (unsigned short)tuv; break;
10592 case 'l': uv = (unsigned long)tuv; break;
10594 default: uv = tuv; break;
10597 uv = (Uquad_t)tuv; break;
10606 char *ptr = ebuf + sizeof ebuf;
10607 bool tempalt = uv ? alt : FALSE; /* Vectors can't change alt */
10613 p = (char *)((c == 'X') ? PL_hexdigit + 16 : PL_hexdigit);
10617 } while (uv >>= 4);
10619 esignbuf[esignlen++] = '0';
10620 esignbuf[esignlen++] = c; /* 'x' or 'X' */
10626 *--ptr = '0' + dig;
10627 } while (uv >>= 3);
10628 if (alt && *ptr != '0')
10634 *--ptr = '0' + dig;
10635 } while (uv >>= 1);
10637 esignbuf[esignlen++] = '0';
10638 esignbuf[esignlen++] = c;
10641 default: /* it had better be ten or less */
10644 *--ptr = '0' + dig;
10645 } while (uv /= base);
10648 elen = (ebuf + sizeof ebuf) - ptr;
10652 zeros = precis - elen;
10653 else if (precis == 0 && elen == 1 && *eptr == '0'
10654 && !(base == 8 && alt)) /* "%#.0o" prints "0" */
10657 /* a precision nullifies the 0 flag. */
10664 /* FLOATING POINT */
10667 c = 'f'; /* maybe %F isn't supported here */
10669 case 'e': case 'E':
10671 case 'g': case 'G':
10675 /* This is evil, but floating point is even more evil */
10677 /* for SV-style calling, we can only get NV
10678 for C-style calling, we assume %f is double;
10679 for simplicity we allow any of %Lf, %llf, %qf for long double
10683 #if defined(USE_LONG_DOUBLE)
10687 /* [perl #20339] - we should accept and ignore %lf rather than die */
10691 #if defined(USE_LONG_DOUBLE)
10692 intsize = args ? 0 : 'q';
10696 #if defined(HAS_LONG_DOUBLE)
10709 /* now we need (long double) if intsize == 'q', else (double) */
10711 #if LONG_DOUBLESIZE > DOUBLESIZE
10713 va_arg(*args, long double) :
10714 va_arg(*args, double)
10716 va_arg(*args, double)
10721 /* nv * 0 will be NaN for NaN, +Inf and -Inf, and 0 for anything
10722 else. frexp() has some unspecified behaviour for those three */
10723 if (c != 'e' && c != 'E' && (nv * 0) == 0) {
10725 /* FIXME: if HAS_LONG_DOUBLE but not USE_LONG_DOUBLE this
10726 will cast our (long double) to (double) */
10727 (void)Perl_frexp(nv, &i);
10728 if (i == PERL_INT_MIN)
10729 Perl_die(aTHX_ "panic: frexp");
10731 need = BIT_DIGITS(i);
10733 need += has_precis ? precis : 6; /* known default */
10738 #ifdef HAS_LDBL_SPRINTF_BUG
10739 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
10740 with sfio - Allen <allens@cpan.org> */
10743 # define MY_DBL_MAX DBL_MAX
10744 # else /* XXX guessing! HUGE_VAL may be defined as infinity, so not using */
10745 # if DOUBLESIZE >= 8
10746 # define MY_DBL_MAX 1.7976931348623157E+308L
10748 # define MY_DBL_MAX 3.40282347E+38L
10752 # ifdef HAS_LDBL_SPRINTF_BUG_LESS1 /* only between -1L & 1L - Allen */
10753 # define MY_DBL_MAX_BUG 1L
10755 # define MY_DBL_MAX_BUG MY_DBL_MAX
10759 # define MY_DBL_MIN DBL_MIN
10760 # else /* XXX guessing! -Allen */
10761 # if DOUBLESIZE >= 8
10762 # define MY_DBL_MIN 2.2250738585072014E-308L
10764 # define MY_DBL_MIN 1.17549435E-38L
10768 if ((intsize == 'q') && (c == 'f') &&
10769 ((nv < MY_DBL_MAX_BUG) && (nv > -MY_DBL_MAX_BUG)) &&
10770 (need < DBL_DIG)) {
10771 /* it's going to be short enough that
10772 * long double precision is not needed */
10774 if ((nv <= 0L) && (nv >= -0L))
10775 fix_ldbl_sprintf_bug = TRUE; /* 0 is 0 - easiest */
10777 /* would use Perl_fp_class as a double-check but not
10778 * functional on IRIX - see perl.h comments */
10780 if ((nv >= MY_DBL_MIN) || (nv <= -MY_DBL_MIN)) {
10781 /* It's within the range that a double can represent */
10782 #if defined(DBL_MAX) && !defined(DBL_MIN)
10783 if ((nv >= ((long double)1/DBL_MAX)) ||
10784 (nv <= (-(long double)1/DBL_MAX)))
10786 fix_ldbl_sprintf_bug = TRUE;
10789 if (fix_ldbl_sprintf_bug == TRUE) {
10799 # undef MY_DBL_MAX_BUG
10802 #endif /* HAS_LDBL_SPRINTF_BUG */
10804 need += 20; /* fudge factor */
10805 if (PL_efloatsize < need) {
10806 Safefree(PL_efloatbuf);
10807 PL_efloatsize = need + 20; /* more fudge */
10808 Newx(PL_efloatbuf, PL_efloatsize, char);
10809 PL_efloatbuf[0] = '\0';
10812 if ( !(width || left || plus || alt) && fill != '0'
10813 && has_precis && intsize != 'q' ) { /* Shortcuts */
10814 /* See earlier comment about buggy Gconvert when digits,
10816 if ( c == 'g' && precis) {
10817 Gconvert((NV)nv, (int)precis, 0, PL_efloatbuf);
10818 /* May return an empty string for digits==0 */
10819 if (*PL_efloatbuf) {
10820 elen = strlen(PL_efloatbuf);
10821 goto float_converted;
10823 } else if ( c == 'f' && !precis) {
10824 if ((eptr = F0convert(nv, ebuf + sizeof ebuf, &elen)))
10829 char *ptr = ebuf + sizeof ebuf;
10832 /* FIXME: what to do if HAS_LONG_DOUBLE but not PERL_PRIfldbl? */
10833 #if defined(HAS_LONG_DOUBLE) && defined(PERL_PRIfldbl)
10834 if (intsize == 'q') {
10835 /* Copy the one or more characters in a long double
10836 * format before the 'base' ([efgEFG]) character to
10837 * the format string. */
10838 static char const prifldbl[] = PERL_PRIfldbl;
10839 char const *p = prifldbl + sizeof(prifldbl) - 3;
10840 while (p >= prifldbl) { *--ptr = *p--; }
10845 do { *--ptr = '0' + (base % 10); } while (base /= 10);
10850 do { *--ptr = '0' + (base % 10); } while (base /= 10);
10862 /* No taint. Otherwise we are in the strange situation
10863 * where printf() taints but print($float) doesn't.
10865 #if defined(HAS_LONG_DOUBLE)
10866 elen = ((intsize == 'q')
10867 ? my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, nv)
10868 : my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, (double)nv));
10870 elen = my_sprintf(PL_efloatbuf, ptr, nv);
10874 eptr = PL_efloatbuf;
10882 i = SvCUR(sv) - origlen;
10885 case 'c': *(va_arg(*args, char*)) = i; break;
10886 case 'h': *(va_arg(*args, short*)) = i; break;
10887 default: *(va_arg(*args, int*)) = i; break;
10888 case 'l': *(va_arg(*args, long*)) = i; break;
10889 case 'V': *(va_arg(*args, IV*)) = i; break;
10890 case 'z': *(va_arg(*args, SSize_t*)) = i; break;
10891 case 't': *(va_arg(*args, ptrdiff_t*)) = i; break;
10893 case 'j': *(va_arg(*args, intmax_t*)) = i; break;
10897 *(va_arg(*args, Quad_t*)) = i; break;
10904 sv_setuv_mg(argsv, (UV)i);
10905 continue; /* not "break" */
10912 && (PL_op->op_type == OP_PRTF || PL_op->op_type == OP_SPRINTF)
10913 && ckWARN(WARN_PRINTF))
10915 SV * const msg = sv_newmortal();
10916 Perl_sv_setpvf(aTHX_ msg, "Invalid conversion in %sprintf: ",
10917 (PL_op->op_type == OP_PRTF) ? "" : "s");
10918 if (fmtstart < patend) {
10919 const char * const fmtend = q < patend ? q : patend;
10921 sv_catpvs(msg, "\"%");
10922 for (f = fmtstart; f < fmtend; f++) {
10924 sv_catpvn(msg, f, 1);
10926 Perl_sv_catpvf(aTHX_ msg,
10927 "\\%03"UVof, (UV)*f & 0xFF);
10930 sv_catpvs(msg, "\"");
10932 sv_catpvs(msg, "end of string");
10934 Perl_warner(aTHX_ packWARN(WARN_PRINTF), "%"SVf, SVfARG(msg)); /* yes, this is reentrant */
10937 /* output mangled stuff ... */
10943 /* ... right here, because formatting flags should not apply */
10944 SvGROW(sv, SvCUR(sv) + elen + 1);
10946 Copy(eptr, p, elen, char);
10949 SvCUR_set(sv, p - SvPVX_const(sv));
10951 continue; /* not "break" */
10954 if (is_utf8 != has_utf8) {
10957 sv_utf8_upgrade(sv);
10960 const STRLEN old_elen = elen;
10961 SV * const nsv = newSVpvn_flags(eptr, elen, SVs_TEMP);
10962 sv_utf8_upgrade(nsv);
10963 eptr = SvPVX_const(nsv);
10966 if (width) { /* fudge width (can't fudge elen) */
10967 width += elen - old_elen;
10973 have = esignlen + zeros + elen;
10975 Perl_croak_nocontext("%s", PL_memory_wrap);
10977 need = (have > width ? have : width);
10980 if (need >= (((STRLEN)~0) - SvCUR(sv) - dotstrlen - 1))
10981 Perl_croak_nocontext("%s", PL_memory_wrap);
10982 SvGROW(sv, SvCUR(sv) + need + dotstrlen + 1);
10984 if (esignlen && fill == '0') {
10986 for (i = 0; i < (int)esignlen; i++)
10987 *p++ = esignbuf[i];
10989 if (gap && !left) {
10990 memset(p, fill, gap);
10993 if (esignlen && fill != '0') {
10995 for (i = 0; i < (int)esignlen; i++)
10996 *p++ = esignbuf[i];
11000 for (i = zeros; i; i--)
11004 Copy(eptr, p, elen, char);
11008 memset(p, ' ', gap);
11013 Copy(dotstr, p, dotstrlen, char);
11017 vectorize = FALSE; /* done iterating over vecstr */
11024 SvCUR_set(sv, p - SvPVX_const(sv));
11033 /* =========================================================================
11035 =head1 Cloning an interpreter
11037 All the macros and functions in this section are for the private use of
11038 the main function, perl_clone().
11040 The foo_dup() functions make an exact copy of an existing foo thingy.
11041 During the course of a cloning, a hash table is used to map old addresses
11042 to new addresses. The table is created and manipulated with the
11043 ptr_table_* functions.
11047 * =========================================================================*/
11050 #if defined(USE_ITHREADS)
11052 /* XXX Remove this so it doesn't have to go thru the macro and return for nothing */
11053 #ifndef GpREFCNT_inc
11054 # define GpREFCNT_inc(gp) ((gp) ? (++(gp)->gp_refcnt, (gp)) : (GP*)NULL)
11058 /* Certain cases in Perl_ss_dup have been merged, by relying on the fact
11059 that currently av_dup, gv_dup and hv_dup are the same as sv_dup.
11060 If this changes, please unmerge ss_dup.
11061 Likewise, sv_dup_inc_multiple() relies on this fact. */
11062 #define sv_dup_inc_NN(s,t) SvREFCNT_inc_NN(sv_dup_inc(s,t))
11063 #define av_dup(s,t) MUTABLE_AV(sv_dup((const SV *)s,t))
11064 #define av_dup_inc(s,t) MUTABLE_AV(sv_dup_inc((const SV *)s,t))
11065 #define hv_dup(s,t) MUTABLE_HV(sv_dup((const SV *)s,t))
11066 #define hv_dup_inc(s,t) MUTABLE_HV(sv_dup_inc((const SV *)s,t))
11067 #define cv_dup(s,t) MUTABLE_CV(sv_dup((const SV *)s,t))
11068 #define cv_dup_inc(s,t) MUTABLE_CV(sv_dup_inc((const SV *)s,t))
11069 #define io_dup(s,t) MUTABLE_IO(sv_dup((const SV *)s,t))
11070 #define io_dup_inc(s,t) MUTABLE_IO(sv_dup_inc((const SV *)s,t))
11071 #define gv_dup(s,t) MUTABLE_GV(sv_dup((const SV *)s,t))
11072 #define gv_dup_inc(s,t) MUTABLE_GV(sv_dup_inc((const SV *)s,t))
11073 #define SAVEPV(p) ((p) ? savepv(p) : NULL)
11074 #define SAVEPVN(p,n) ((p) ? savepvn(p,n) : NULL)
11076 /* clone a parser */
11079 Perl_parser_dup(pTHX_ const yy_parser *const proto, CLONE_PARAMS *const param)
11083 PERL_ARGS_ASSERT_PARSER_DUP;
11088 /* look for it in the table first */
11089 parser = (yy_parser *)ptr_table_fetch(PL_ptr_table, proto);
11093 /* create anew and remember what it is */
11094 Newxz(parser, 1, yy_parser);
11095 ptr_table_store(PL_ptr_table, proto, parser);
11097 /* XXX these not yet duped */
11098 parser->old_parser = NULL;
11099 parser->stack = NULL;
11101 parser->stack_size = 0;
11102 /* XXX parser->stack->state = 0; */
11104 /* XXX eventually, just Copy() most of the parser struct ? */
11106 parser->lex_brackets = proto->lex_brackets;
11107 parser->lex_casemods = proto->lex_casemods;
11108 parser->lex_brackstack = savepvn(proto->lex_brackstack,
11109 (proto->lex_brackets < 120 ? 120 : proto->lex_brackets));
11110 parser->lex_casestack = savepvn(proto->lex_casestack,
11111 (proto->lex_casemods < 12 ? 12 : proto->lex_casemods));
11112 parser->lex_defer = proto->lex_defer;
11113 parser->lex_dojoin = proto->lex_dojoin;
11114 parser->lex_expect = proto->lex_expect;
11115 parser->lex_formbrack = proto->lex_formbrack;
11116 parser->lex_inpat = proto->lex_inpat;
11117 parser->lex_inwhat = proto->lex_inwhat;
11118 parser->lex_op = proto->lex_op;
11119 parser->lex_repl = sv_dup_inc(proto->lex_repl, param);
11120 parser->lex_starts = proto->lex_starts;
11121 parser->lex_stuff = sv_dup_inc(proto->lex_stuff, param);
11122 parser->multi_close = proto->multi_close;
11123 parser->multi_open = proto->multi_open;
11124 parser->multi_start = proto->multi_start;
11125 parser->multi_end = proto->multi_end;
11126 parser->pending_ident = proto->pending_ident;
11127 parser->preambled = proto->preambled;
11128 parser->sublex_info = proto->sublex_info; /* XXX not quite right */
11129 parser->linestr = sv_dup_inc(proto->linestr, param);
11130 parser->expect = proto->expect;
11131 parser->copline = proto->copline;
11132 parser->last_lop_op = proto->last_lop_op;
11133 parser->lex_state = proto->lex_state;
11134 parser->rsfp = fp_dup(proto->rsfp, '<', param);
11135 /* rsfp_filters entries have fake IoDIRP() */
11136 parser->rsfp_filters= av_dup_inc(proto->rsfp_filters, param);
11137 parser->in_my = proto->in_my;
11138 parser->in_my_stash = hv_dup(proto->in_my_stash, param);
11139 parser->error_count = proto->error_count;
11142 parser->linestr = sv_dup_inc(proto->linestr, param);
11145 char * const ols = SvPVX(proto->linestr);
11146 char * const ls = SvPVX(parser->linestr);
11148 parser->bufptr = ls + (proto->bufptr >= ols ?
11149 proto->bufptr - ols : 0);
11150 parser->oldbufptr = ls + (proto->oldbufptr >= ols ?
11151 proto->oldbufptr - ols : 0);
11152 parser->oldoldbufptr= ls + (proto->oldoldbufptr >= ols ?
11153 proto->oldoldbufptr - ols : 0);
11154 parser->linestart = ls + (proto->linestart >= ols ?
11155 proto->linestart - ols : 0);
11156 parser->last_uni = ls + (proto->last_uni >= ols ?
11157 proto->last_uni - ols : 0);
11158 parser->last_lop = ls + (proto->last_lop >= ols ?
11159 proto->last_lop - ols : 0);
11161 parser->bufend = ls + SvCUR(parser->linestr);
11164 Copy(proto->tokenbuf, parser->tokenbuf, 256, char);
11168 parser->endwhite = proto->endwhite;
11169 parser->faketokens = proto->faketokens;
11170 parser->lasttoke = proto->lasttoke;
11171 parser->nextwhite = proto->nextwhite;
11172 parser->realtokenstart = proto->realtokenstart;
11173 parser->skipwhite = proto->skipwhite;
11174 parser->thisclose = proto->thisclose;
11175 parser->thismad = proto->thismad;
11176 parser->thisopen = proto->thisopen;
11177 parser->thisstuff = proto->thisstuff;
11178 parser->thistoken = proto->thistoken;
11179 parser->thiswhite = proto->thiswhite;
11181 Copy(proto->nexttoke, parser->nexttoke, 5, NEXTTOKE);
11182 parser->curforce = proto->curforce;
11184 Copy(proto->nextval, parser->nextval, 5, YYSTYPE);
11185 Copy(proto->nexttype, parser->nexttype, 5, I32);
11186 parser->nexttoke = proto->nexttoke;
11189 /* XXX should clone saved_curcop here, but we aren't passed
11190 * proto_perl; so do it in perl_clone_using instead */
11196 /* duplicate a file handle */
11199 Perl_fp_dup(pTHX_ PerlIO *const fp, const char type, CLONE_PARAMS *const param)
11203 PERL_ARGS_ASSERT_FP_DUP;
11204 PERL_UNUSED_ARG(type);
11207 return (PerlIO*)NULL;
11209 /* look for it in the table first */
11210 ret = (PerlIO*)ptr_table_fetch(PL_ptr_table, fp);
11214 /* create anew and remember what it is */
11215 ret = PerlIO_fdupopen(aTHX_ fp, param, PERLIO_DUP_CLONE);
11216 ptr_table_store(PL_ptr_table, fp, ret);
11220 /* duplicate a directory handle */
11223 Perl_dirp_dup(pTHX_ DIR *const dp, CLONE_PARAMS *const param)
11229 register const Direntry_t *dirent;
11230 char smallbuf[256];
11236 PERL_UNUSED_CONTEXT;
11237 PERL_ARGS_ASSERT_DIRP_DUP;
11242 /* look for it in the table first */
11243 ret = (DIR*)ptr_table_fetch(PL_ptr_table, dp);
11249 PERL_UNUSED_ARG(param);
11253 /* open the current directory (so we can switch back) */
11254 if (!(pwd = PerlDir_open("."))) return (DIR *)NULL;
11256 /* chdir to our dir handle and open the present working directory */
11257 if (fchdir(my_dirfd(dp)) < 0 || !(ret = PerlDir_open("."))) {
11258 PerlDir_close(pwd);
11259 return (DIR *)NULL;
11261 /* Now we should have two dir handles pointing to the same dir. */
11263 /* Be nice to the calling code and chdir back to where we were. */
11264 fchdir(my_dirfd(pwd)); /* If this fails, then what? */
11266 /* We have no need of the pwd handle any more. */
11267 PerlDir_close(pwd);
11270 # define d_namlen(d) (d)->d_namlen
11272 # define d_namlen(d) strlen((d)->d_name)
11274 /* Iterate once through dp, to get the file name at the current posi-
11275 tion. Then step back. */
11276 pos = PerlDir_tell(dp);
11277 if ((dirent = PerlDir_read(dp))) {
11278 len = d_namlen(dirent);
11279 if (len <= sizeof smallbuf) name = smallbuf;
11280 else Newx(name, len, char);
11281 Move(dirent->d_name, name, len, char);
11283 PerlDir_seek(dp, pos);
11285 /* Iterate through the new dir handle, till we find a file with the
11287 if (!dirent) /* just before the end */
11289 pos = PerlDir_tell(ret);
11290 if (PerlDir_read(ret)) continue; /* not there yet */
11291 PerlDir_seek(ret, pos); /* step back */
11295 const long pos0 = PerlDir_tell(ret);
11297 pos = PerlDir_tell(ret);
11298 if ((dirent = PerlDir_read(ret))) {
11299 if (len == d_namlen(dirent)
11300 && memEQ(name, dirent->d_name, len)) {
11302 PerlDir_seek(ret, pos); /* step back */
11305 /* else we are not there yet; keep iterating */
11307 else { /* This is not meant to happen. The best we can do is
11308 reset the iterator to the beginning. */
11309 PerlDir_seek(ret, pos0);
11316 if (name && name != smallbuf)
11321 ret = win32_dirp_dup(dp, param);
11324 /* pop it in the pointer table */
11326 ptr_table_store(PL_ptr_table, dp, ret);
11331 /* duplicate a typeglob */
11334 Perl_gp_dup(pTHX_ GP *const gp, CLONE_PARAMS *const param)
11338 PERL_ARGS_ASSERT_GP_DUP;
11342 /* look for it in the table first */
11343 ret = (GP*)ptr_table_fetch(PL_ptr_table, gp);
11347 /* create anew and remember what it is */
11349 ptr_table_store(PL_ptr_table, gp, ret);
11352 /* ret->gp_refcnt must be 0 before any other dups are called. We're relying
11353 on Newxz() to do this for us. */
11354 ret->gp_sv = sv_dup_inc(gp->gp_sv, param);
11355 ret->gp_io = io_dup_inc(gp->gp_io, param);
11356 ret->gp_form = cv_dup_inc(gp->gp_form, param);
11357 ret->gp_av = av_dup_inc(gp->gp_av, param);
11358 ret->gp_hv = hv_dup_inc(gp->gp_hv, param);
11359 ret->gp_egv = gv_dup(gp->gp_egv, param);/* GvEGV is not refcounted */
11360 ret->gp_cv = cv_dup_inc(gp->gp_cv, param);
11361 ret->gp_cvgen = gp->gp_cvgen;
11362 ret->gp_line = gp->gp_line;
11363 ret->gp_file_hek = hek_dup(gp->gp_file_hek, param);
11367 /* duplicate a chain of magic */
11370 Perl_mg_dup(pTHX_ MAGIC *mg, CLONE_PARAMS *const param)
11372 MAGIC *mgret = NULL;
11373 MAGIC **mgprev_p = &mgret;
11375 PERL_ARGS_ASSERT_MG_DUP;
11377 for (; mg; mg = mg->mg_moremagic) {
11380 if ((param->flags & CLONEf_JOIN_IN)
11381 && mg->mg_type == PERL_MAGIC_backref)
11382 /* when joining, we let the individual SVs add themselves to
11383 * backref as needed. */
11386 Newx(nmg, 1, MAGIC);
11388 mgprev_p = &(nmg->mg_moremagic);
11390 /* There was a comment "XXX copy dynamic vtable?" but as we don't have
11391 dynamic vtables, I'm not sure why Sarathy wrote it. The comment dates
11392 from the original commit adding Perl_mg_dup() - revision 4538.
11393 Similarly there is the annotation "XXX random ptr?" next to the
11394 assignment to nmg->mg_ptr. */
11397 /* FIXME for plugins
11398 if (nmg->mg_type == PERL_MAGIC_qr) {
11399 nmg->mg_obj = MUTABLE_SV(CALLREGDUPE((REGEXP*)nmg->mg_obj, param));
11403 nmg->mg_obj = (nmg->mg_flags & MGf_REFCOUNTED)
11404 ? nmg->mg_type == PERL_MAGIC_backref
11405 /* The backref AV has its reference
11406 * count deliberately bumped by 1 */
11407 ? SvREFCNT_inc(av_dup_inc((const AV *)
11408 nmg->mg_obj, param))
11409 : sv_dup_inc(nmg->mg_obj, param)
11410 : sv_dup(nmg->mg_obj, param);
11412 if (nmg->mg_ptr && nmg->mg_type != PERL_MAGIC_regex_global) {
11413 if (nmg->mg_len > 0) {
11414 nmg->mg_ptr = SAVEPVN(nmg->mg_ptr, nmg->mg_len);
11415 if (nmg->mg_type == PERL_MAGIC_overload_table &&
11416 AMT_AMAGIC((AMT*)nmg->mg_ptr))
11418 AMT * const namtp = (AMT*)nmg->mg_ptr;
11419 sv_dup_inc_multiple((SV**)(namtp->table),
11420 (SV**)(namtp->table), NofAMmeth, param);
11423 else if (nmg->mg_len == HEf_SVKEY)
11424 nmg->mg_ptr = (char*)sv_dup_inc((const SV *)nmg->mg_ptr, param);
11426 if ((nmg->mg_flags & MGf_DUP) && nmg->mg_virtual && nmg->mg_virtual->svt_dup) {
11427 nmg->mg_virtual->svt_dup(aTHX_ nmg, param);
11433 #endif /* USE_ITHREADS */
11435 struct ptr_tbl_arena {
11436 struct ptr_tbl_arena *next;
11437 struct ptr_tbl_ent array[1023/3]; /* as ptr_tbl_ent has 3 pointers. */
11440 /* create a new pointer-mapping table */
11443 Perl_ptr_table_new(pTHX)
11446 PERL_UNUSED_CONTEXT;
11448 Newx(tbl, 1, PTR_TBL_t);
11449 tbl->tbl_max = 511;
11450 tbl->tbl_items = 0;
11451 tbl->tbl_arena = NULL;
11452 tbl->tbl_arena_next = NULL;
11453 tbl->tbl_arena_end = NULL;
11454 Newxz(tbl->tbl_ary, tbl->tbl_max + 1, PTR_TBL_ENT_t*);
11458 #define PTR_TABLE_HASH(ptr) \
11459 ((PTR2UV(ptr) >> 3) ^ (PTR2UV(ptr) >> (3 + 7)) ^ (PTR2UV(ptr) >> (3 + 17)))
11461 /* map an existing pointer using a table */
11463 STATIC PTR_TBL_ENT_t *
11464 S_ptr_table_find(PTR_TBL_t *const tbl, const void *const sv)
11466 PTR_TBL_ENT_t *tblent;
11467 const UV hash = PTR_TABLE_HASH(sv);
11469 PERL_ARGS_ASSERT_PTR_TABLE_FIND;
11471 tblent = tbl->tbl_ary[hash & tbl->tbl_max];
11472 for (; tblent; tblent = tblent->next) {
11473 if (tblent->oldval == sv)
11480 Perl_ptr_table_fetch(pTHX_ PTR_TBL_t *const tbl, const void *const sv)
11482 PTR_TBL_ENT_t const *const tblent = ptr_table_find(tbl, sv);
11484 PERL_ARGS_ASSERT_PTR_TABLE_FETCH;
11485 PERL_UNUSED_CONTEXT;
11487 return tblent ? tblent->newval : NULL;
11490 /* add a new entry to a pointer-mapping table */
11493 Perl_ptr_table_store(pTHX_ PTR_TBL_t *const tbl, const void *const oldsv, void *const newsv)
11495 PTR_TBL_ENT_t *tblent = ptr_table_find(tbl, oldsv);
11497 PERL_ARGS_ASSERT_PTR_TABLE_STORE;
11498 PERL_UNUSED_CONTEXT;
11501 tblent->newval = newsv;
11503 const UV entry = PTR_TABLE_HASH(oldsv) & tbl->tbl_max;
11505 if (tbl->tbl_arena_next == tbl->tbl_arena_end) {
11506 struct ptr_tbl_arena *new_arena;
11508 Newx(new_arena, 1, struct ptr_tbl_arena);
11509 new_arena->next = tbl->tbl_arena;
11510 tbl->tbl_arena = new_arena;
11511 tbl->tbl_arena_next = new_arena->array;
11512 tbl->tbl_arena_end = new_arena->array
11513 + sizeof(new_arena->array) / sizeof(new_arena->array[0]);
11516 tblent = tbl->tbl_arena_next++;
11518 tblent->oldval = oldsv;
11519 tblent->newval = newsv;
11520 tblent->next = tbl->tbl_ary[entry];
11521 tbl->tbl_ary[entry] = tblent;
11523 if (tblent->next && tbl->tbl_items > tbl->tbl_max)
11524 ptr_table_split(tbl);
11528 /* double the hash bucket size of an existing ptr table */
11531 Perl_ptr_table_split(pTHX_ PTR_TBL_t *const tbl)
11533 PTR_TBL_ENT_t **ary = tbl->tbl_ary;
11534 const UV oldsize = tbl->tbl_max + 1;
11535 UV newsize = oldsize * 2;
11538 PERL_ARGS_ASSERT_PTR_TABLE_SPLIT;
11539 PERL_UNUSED_CONTEXT;
11541 Renew(ary, newsize, PTR_TBL_ENT_t*);
11542 Zero(&ary[oldsize], newsize-oldsize, PTR_TBL_ENT_t*);
11543 tbl->tbl_max = --newsize;
11544 tbl->tbl_ary = ary;
11545 for (i=0; i < oldsize; i++, ary++) {
11546 PTR_TBL_ENT_t **entp = ary;
11547 PTR_TBL_ENT_t *ent = *ary;
11548 PTR_TBL_ENT_t **curentp;
11551 curentp = ary + oldsize;
11553 if ((newsize & PTR_TABLE_HASH(ent->oldval)) != i) {
11555 ent->next = *curentp;
11565 /* remove all the entries from a ptr table */
11566 /* Deprecated - will be removed post 5.14 */
11569 Perl_ptr_table_clear(pTHX_ PTR_TBL_t *const tbl)
11571 if (tbl && tbl->tbl_items) {
11572 struct ptr_tbl_arena *arena = tbl->tbl_arena;
11574 Zero(tbl->tbl_ary, tbl->tbl_max + 1, struct ptr_tbl_ent **);
11577 struct ptr_tbl_arena *next = arena->next;
11583 tbl->tbl_items = 0;
11584 tbl->tbl_arena = NULL;
11585 tbl->tbl_arena_next = NULL;
11586 tbl->tbl_arena_end = NULL;
11590 /* clear and free a ptr table */
11593 Perl_ptr_table_free(pTHX_ PTR_TBL_t *const tbl)
11595 struct ptr_tbl_arena *arena;
11601 arena = tbl->tbl_arena;
11604 struct ptr_tbl_arena *next = arena->next;
11610 Safefree(tbl->tbl_ary);
11614 #if defined(USE_ITHREADS)
11617 Perl_rvpv_dup(pTHX_ SV *const dstr, const SV *const sstr, CLONE_PARAMS *const param)
11619 PERL_ARGS_ASSERT_RVPV_DUP;
11622 if (SvWEAKREF(sstr)) {
11623 SvRV_set(dstr, sv_dup(SvRV_const(sstr), param));
11624 if (param->flags & CLONEf_JOIN_IN) {
11625 /* if joining, we add any back references individually rather
11626 * than copying the whole backref array */
11627 Perl_sv_add_backref(aTHX_ SvRV(dstr), dstr);
11631 SvRV_set(dstr, sv_dup_inc(SvRV_const(sstr), param));
11633 else if (SvPVX_const(sstr)) {
11634 /* Has something there */
11636 /* Normal PV - clone whole allocated space */
11637 SvPV_set(dstr, SAVEPVN(SvPVX_const(sstr), SvLEN(sstr)-1));
11638 if (SvREADONLY(sstr) && SvFAKE(sstr)) {
11639 /* Not that normal - actually sstr is copy on write.
11640 But we are a true, independent SV, so: */
11641 SvREADONLY_off(dstr);
11646 /* Special case - not normally malloced for some reason */
11647 if (isGV_with_GP(sstr)) {
11648 /* Don't need to do anything here. */
11650 else if ((SvREADONLY(sstr) && SvFAKE(sstr))) {
11651 /* A "shared" PV - clone it as "shared" PV */
11653 HEK_KEY(hek_dup(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)),
11657 /* Some other special case - random pointer */
11658 SvPV_set(dstr, (char *) SvPVX_const(sstr));
11663 /* Copy the NULL */
11664 SvPV_set(dstr, NULL);
11668 /* duplicate a list of SVs. source and dest may point to the same memory. */
11670 S_sv_dup_inc_multiple(pTHX_ SV *const *source, SV **dest,
11671 SSize_t items, CLONE_PARAMS *const param)
11673 PERL_ARGS_ASSERT_SV_DUP_INC_MULTIPLE;
11675 while (items-- > 0) {
11676 *dest++ = sv_dup_inc(*source++, param);
11682 /* duplicate an SV of any type (including AV, HV etc) */
11685 S_sv_dup_common(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
11690 PERL_ARGS_ASSERT_SV_DUP_COMMON;
11692 if (SvTYPE(sstr) == SVTYPEMASK) {
11693 #ifdef DEBUG_LEAKING_SCALARS_ABORT
11698 /* look for it in the table first */
11699 dstr = MUTABLE_SV(ptr_table_fetch(PL_ptr_table, sstr));
11703 if(param->flags & CLONEf_JOIN_IN) {
11704 /** We are joining here so we don't want do clone
11705 something that is bad **/
11706 if (SvTYPE(sstr) == SVt_PVHV) {
11707 const HEK * const hvname = HvNAME_HEK(sstr);
11709 /** don't clone stashes if they already exist **/
11710 dstr = MUTABLE_SV(gv_stashpvn(HEK_KEY(hvname), HEK_LEN(hvname), 0));
11711 ptr_table_store(PL_ptr_table, sstr, dstr);
11717 /* create anew and remember what it is */
11720 #ifdef DEBUG_LEAKING_SCALARS
11721 dstr->sv_debug_optype = sstr->sv_debug_optype;
11722 dstr->sv_debug_line = sstr->sv_debug_line;
11723 dstr->sv_debug_inpad = sstr->sv_debug_inpad;
11724 dstr->sv_debug_parent = (SV*)sstr;
11725 FREE_SV_DEBUG_FILE(dstr);
11726 dstr->sv_debug_file = savepv(sstr->sv_debug_file);
11729 ptr_table_store(PL_ptr_table, sstr, dstr);
11732 SvFLAGS(dstr) = SvFLAGS(sstr);
11733 SvFLAGS(dstr) &= ~SVf_OOK; /* don't propagate OOK hack */
11734 SvREFCNT(dstr) = 0; /* must be before any other dups! */
11737 if (SvANY(sstr) && PL_watch_pvx && SvPVX_const(sstr) == PL_watch_pvx)
11738 PerlIO_printf(Perl_debug_log, "watch at %p hit, found string \"%s\"\n",
11739 (void*)PL_watch_pvx, SvPVX_const(sstr));
11742 /* don't clone objects whose class has asked us not to */
11743 if (SvOBJECT(sstr) && ! (SvFLAGS(SvSTASH(sstr)) & SVphv_CLONEABLE)) {
11748 switch (SvTYPE(sstr)) {
11750 SvANY(dstr) = NULL;
11753 SvANY(dstr) = (XPVIV*)((char*)&(dstr->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
11755 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
11757 SvIV_set(dstr, SvIVX(sstr));
11761 SvANY(dstr) = new_XNV();
11762 SvNV_set(dstr, SvNVX(sstr));
11764 /* case SVt_BIND: */
11767 /* These are all the types that need complex bodies allocating. */
11769 const svtype sv_type = SvTYPE(sstr);
11770 const struct body_details *const sv_type_details
11771 = bodies_by_type + sv_type;
11775 Perl_croak(aTHX_ "Bizarre SvTYPE [%" IVdf "]", (IV)SvTYPE(sstr));
11790 assert(sv_type_details->body_size);
11791 if (sv_type_details->arena) {
11792 new_body_inline(new_body, sv_type);
11794 = (void*)((char*)new_body - sv_type_details->offset);
11796 new_body = new_NOARENA(sv_type_details);
11800 SvANY(dstr) = new_body;
11803 Copy(((char*)SvANY(sstr)) + sv_type_details->offset,
11804 ((char*)SvANY(dstr)) + sv_type_details->offset,
11805 sv_type_details->copy, char);
11807 Copy(((char*)SvANY(sstr)),
11808 ((char*)SvANY(dstr)),
11809 sv_type_details->body_size + sv_type_details->offset, char);
11812 if (sv_type != SVt_PVAV && sv_type != SVt_PVHV
11813 && !isGV_with_GP(dstr)
11814 && !(sv_type == SVt_PVIO && !(IoFLAGS(dstr) & IOf_FAKE_DIRP)))
11815 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
11817 /* The Copy above means that all the source (unduplicated) pointers
11818 are now in the destination. We can check the flags and the
11819 pointers in either, but it's possible that there's less cache
11820 missing by always going for the destination.
11821 FIXME - instrument and check that assumption */
11822 if (sv_type >= SVt_PVMG) {
11823 if ((sv_type == SVt_PVMG) && SvPAD_OUR(dstr)) {
11824 SvOURSTASH_set(dstr, hv_dup_inc(SvOURSTASH(dstr), param));
11825 } else if (SvMAGIC(dstr))
11826 SvMAGIC_set(dstr, mg_dup(SvMAGIC(dstr), param));
11828 SvSTASH_set(dstr, hv_dup_inc(SvSTASH(dstr), param));
11831 /* The cast silences a GCC warning about unhandled types. */
11832 switch ((int)sv_type) {
11842 /* FIXME for plugins */
11843 re_dup_guts((REGEXP*) sstr, (REGEXP*) dstr, param);
11846 /* XXX LvTARGOFF sometimes holds PMOP* when DEBUGGING */
11847 if (LvTYPE(dstr) == 't') /* for tie: unrefcnted fake (SV**) */
11848 LvTARG(dstr) = dstr;
11849 else if (LvTYPE(dstr) == 'T') /* for tie: fake HE */
11850 LvTARG(dstr) = MUTABLE_SV(he_dup((HE*)LvTARG(dstr), 0, param));
11852 LvTARG(dstr) = sv_dup_inc(LvTARG(dstr), param);
11854 /* non-GP case already handled above */
11855 if(isGV_with_GP(sstr)) {
11856 GvNAME_HEK(dstr) = hek_dup(GvNAME_HEK(dstr), param);
11857 /* Don't call sv_add_backref here as it's going to be
11858 created as part of the magic cloning of the symbol
11859 table--unless this is during a join and the stash
11860 is not actually being cloned. */
11861 /* Danger Will Robinson - GvGP(dstr) isn't initialised
11862 at the point of this comment. */
11863 GvSTASH(dstr) = hv_dup(GvSTASH(dstr), param);
11864 if (param->flags & CLONEf_JOIN_IN)
11865 Perl_sv_add_backref(aTHX_ MUTABLE_SV(GvSTASH(dstr)), dstr);
11866 GvGP_set(dstr, gp_dup(GvGP(sstr), param));
11867 (void)GpREFCNT_inc(GvGP(dstr));
11871 /* PL_parser->rsfp_filters entries have fake IoDIRP() */
11872 if(IoFLAGS(dstr) & IOf_FAKE_DIRP) {
11873 /* I have no idea why fake dirp (rsfps)
11874 should be treated differently but otherwise
11875 we end up with leaks -- sky*/
11876 IoTOP_GV(dstr) = gv_dup_inc(IoTOP_GV(dstr), param);
11877 IoFMT_GV(dstr) = gv_dup_inc(IoFMT_GV(dstr), param);
11878 IoBOTTOM_GV(dstr) = gv_dup_inc(IoBOTTOM_GV(dstr), param);
11880 IoTOP_GV(dstr) = gv_dup(IoTOP_GV(dstr), param);
11881 IoFMT_GV(dstr) = gv_dup(IoFMT_GV(dstr), param);
11882 IoBOTTOM_GV(dstr) = gv_dup(IoBOTTOM_GV(dstr), param);
11883 if (IoDIRP(dstr)) {
11884 IoDIRP(dstr) = dirp_dup(IoDIRP(dstr), param);
11887 /* IoDIRP(dstr) is already a copy of IoDIRP(sstr) */
11889 IoIFP(dstr) = fp_dup(IoIFP(sstr), IoTYPE(dstr), param);
11891 if (IoOFP(dstr) == IoIFP(sstr))
11892 IoOFP(dstr) = IoIFP(dstr);
11894 IoOFP(dstr) = fp_dup(IoOFP(dstr), IoTYPE(dstr), param);
11895 IoTOP_NAME(dstr) = SAVEPV(IoTOP_NAME(dstr));
11896 IoFMT_NAME(dstr) = SAVEPV(IoFMT_NAME(dstr));
11897 IoBOTTOM_NAME(dstr) = SAVEPV(IoBOTTOM_NAME(dstr));
11900 /* avoid cloning an empty array */
11901 if (AvARRAY((const AV *)sstr) && AvFILLp((const AV *)sstr) >= 0) {
11902 SV **dst_ary, **src_ary;
11903 SSize_t items = AvFILLp((const AV *)sstr) + 1;
11905 src_ary = AvARRAY((const AV *)sstr);
11906 Newxz(dst_ary, AvMAX((const AV *)sstr)+1, SV*);
11907 ptr_table_store(PL_ptr_table, src_ary, dst_ary);
11908 AvARRAY(MUTABLE_AV(dstr)) = dst_ary;
11909 AvALLOC((const AV *)dstr) = dst_ary;
11910 if (AvREAL((const AV *)sstr)) {
11911 dst_ary = sv_dup_inc_multiple(src_ary, dst_ary, items,
11915 while (items-- > 0)
11916 *dst_ary++ = sv_dup(*src_ary++, param);
11918 items = AvMAX((const AV *)sstr) - AvFILLp((const AV *)sstr);
11919 while (items-- > 0) {
11920 *dst_ary++ = &PL_sv_undef;
11924 AvARRAY(MUTABLE_AV(dstr)) = NULL;
11925 AvALLOC((const AV *)dstr) = (SV**)NULL;
11926 AvMAX( (const AV *)dstr) = -1;
11927 AvFILLp((const AV *)dstr) = -1;
11931 if (HvARRAY((const HV *)sstr)) {
11933 const bool sharekeys = !!HvSHAREKEYS(sstr);
11934 XPVHV * const dxhv = (XPVHV*)SvANY(dstr);
11935 XPVHV * const sxhv = (XPVHV*)SvANY(sstr);
11937 Newx(darray, PERL_HV_ARRAY_ALLOC_BYTES(dxhv->xhv_max+1)
11938 + (SvOOK(sstr) ? sizeof(struct xpvhv_aux) : 0),
11940 HvARRAY(dstr) = (HE**)darray;
11941 while (i <= sxhv->xhv_max) {
11942 const HE * const source = HvARRAY(sstr)[i];
11943 HvARRAY(dstr)[i] = source
11944 ? he_dup(source, sharekeys, param) : 0;
11948 const struct xpvhv_aux * const saux = HvAUX(sstr);
11949 struct xpvhv_aux * const daux = HvAUX(dstr);
11950 /* This flag isn't copied. */
11951 /* SvOOK_on(hv) attacks the IV flags. */
11952 SvFLAGS(dstr) |= SVf_OOK;
11954 if (saux->xhv_name_count) {
11955 HEK ** const sname = saux->xhv_name_u.xhvnameu_names;
11957 = saux->xhv_name_count < 0
11958 ? -saux->xhv_name_count
11959 : saux->xhv_name_count;
11960 HEK **shekp = sname + count;
11962 Newx(daux->xhv_name_u.xhvnameu_names, count, HEK *);
11963 dhekp = daux->xhv_name_u.xhvnameu_names + count;
11964 while (shekp-- > sname) {
11966 *dhekp = hek_dup(*shekp, param);
11970 daux->xhv_name_u.xhvnameu_name
11971 = hek_dup(saux->xhv_name_u.xhvnameu_name,
11974 daux->xhv_name_count = saux->xhv_name_count;
11976 daux->xhv_riter = saux->xhv_riter;
11977 daux->xhv_eiter = saux->xhv_eiter
11978 ? he_dup(saux->xhv_eiter,
11979 cBOOL(HvSHAREKEYS(sstr)), param) : 0;
11980 /* backref array needs refcnt=2; see sv_add_backref */
11981 daux->xhv_backreferences =
11982 (param->flags & CLONEf_JOIN_IN)
11983 /* when joining, we let the individual GVs and
11984 * CVs add themselves to backref as
11985 * needed. This avoids pulling in stuff
11986 * that isn't required, and simplifies the
11987 * case where stashes aren't cloned back
11988 * if they already exist in the parent
11991 : saux->xhv_backreferences
11992 ? (SvTYPE(saux->xhv_backreferences) == SVt_PVAV)
11993 ? MUTABLE_AV(SvREFCNT_inc(
11994 sv_dup_inc((const SV *)
11995 saux->xhv_backreferences, param)))
11996 : MUTABLE_AV(sv_dup((const SV *)
11997 saux->xhv_backreferences, param))
12000 daux->xhv_mro_meta = saux->xhv_mro_meta
12001 ? mro_meta_dup(saux->xhv_mro_meta, param)
12004 /* Record stashes for possible cloning in Perl_clone(). */
12006 av_push(param->stashes, dstr);
12010 HvARRAY(MUTABLE_HV(dstr)) = NULL;
12013 if (!(param->flags & CLONEf_COPY_STACKS)) {
12018 /* NOTE: not refcounted */
12019 SvANY(MUTABLE_CV(dstr))->xcv_stash =
12020 hv_dup(CvSTASH(dstr), param);
12021 if ((param->flags & CLONEf_JOIN_IN) && CvSTASH(dstr))
12022 Perl_sv_add_backref(aTHX_ MUTABLE_SV(CvSTASH(dstr)), dstr);
12023 if (!CvISXSUB(dstr)) {
12025 CvROOT(dstr) = OpREFCNT_inc(CvROOT(dstr));
12027 CvFILE(dstr) = SAVEPV(CvFILE(dstr));
12028 } else if (CvCONST(dstr)) {
12029 CvXSUBANY(dstr).any_ptr =
12030 sv_dup_inc((const SV *)CvXSUBANY(dstr).any_ptr, param);
12032 /* don't dup if copying back - CvGV isn't refcounted, so the
12033 * duped GV may never be freed. A bit of a hack! DAPM */
12034 SvANY(MUTABLE_CV(dstr))->xcv_gv =
12036 ? gv_dup_inc(CvGV(sstr), param)
12037 : (param->flags & CLONEf_JOIN_IN)
12039 : gv_dup(CvGV(sstr), param);
12041 CvPADLIST(dstr) = padlist_dup(CvPADLIST(sstr), param);
12043 CvWEAKOUTSIDE(sstr)
12044 ? cv_dup( CvOUTSIDE(dstr), param)
12045 : cv_dup_inc(CvOUTSIDE(dstr), param);
12051 if (SvOBJECT(dstr) && SvTYPE(dstr) != SVt_PVIO)
12058 Perl_sv_dup_inc(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
12060 PERL_ARGS_ASSERT_SV_DUP_INC;
12061 return sstr ? SvREFCNT_inc(sv_dup_common(sstr, param)) : NULL;
12065 Perl_sv_dup(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
12067 SV *dstr = sstr ? sv_dup_common(sstr, param) : NULL;
12068 PERL_ARGS_ASSERT_SV_DUP;
12070 /* Track every SV that (at least initially) had a reference count of 0.
12071 We need to do this by holding an actual reference to it in this array.
12072 If we attempt to cheat, turn AvREAL_off(), and store only pointers
12073 (akin to the stashes hash, and the perl stack), we come unstuck if
12074 a weak reference (or other SV legitimately SvREFCNT() == 0 for this
12075 thread) is manipulated in a CLONE method, because CLONE runs before the
12076 unreferenced array is walked to find SVs still with SvREFCNT() == 0
12077 (and fix things up by giving each a reference via the temps stack).
12078 Instead, during CLONE, if the 0-referenced SV has SvREFCNT_inc() and
12079 then SvREFCNT_dec(), it will be cleaned up (and added to the free list)
12080 before the walk of unreferenced happens and a reference to that is SV
12081 added to the temps stack. At which point we have the same SV considered
12082 to be in use, and free to be re-used. Not good.
12084 if (dstr && !(param->flags & CLONEf_COPY_STACKS) && !SvREFCNT(dstr)) {
12085 assert(param->unreferenced);
12086 av_push(param->unreferenced, SvREFCNT_inc(dstr));
12092 /* duplicate a context */
12095 Perl_cx_dup(pTHX_ PERL_CONTEXT *cxs, I32 ix, I32 max, CLONE_PARAMS* param)
12097 PERL_CONTEXT *ncxs;
12099 PERL_ARGS_ASSERT_CX_DUP;
12102 return (PERL_CONTEXT*)NULL;
12104 /* look for it in the table first */
12105 ncxs = (PERL_CONTEXT*)ptr_table_fetch(PL_ptr_table, cxs);
12109 /* create anew and remember what it is */
12110 Newx(ncxs, max + 1, PERL_CONTEXT);
12111 ptr_table_store(PL_ptr_table, cxs, ncxs);
12112 Copy(cxs, ncxs, max + 1, PERL_CONTEXT);
12115 PERL_CONTEXT * const ncx = &ncxs[ix];
12116 if (CxTYPE(ncx) == CXt_SUBST) {
12117 Perl_croak(aTHX_ "Cloning substitution context is unimplemented");
12120 switch (CxTYPE(ncx)) {
12122 ncx->blk_sub.cv = (ncx->blk_sub.olddepth == 0
12123 ? cv_dup_inc(ncx->blk_sub.cv, param)
12124 : cv_dup(ncx->blk_sub.cv,param));
12125 ncx->blk_sub.argarray = (CxHASARGS(ncx)
12126 ? av_dup_inc(ncx->blk_sub.argarray,
12129 ncx->blk_sub.savearray = av_dup_inc(ncx->blk_sub.savearray,
12131 ncx->blk_sub.oldcomppad = (PAD*)ptr_table_fetch(PL_ptr_table,
12132 ncx->blk_sub.oldcomppad);
12135 ncx->blk_eval.old_namesv = sv_dup_inc(ncx->blk_eval.old_namesv,
12137 ncx->blk_eval.cur_text = sv_dup(ncx->blk_eval.cur_text, param);
12139 case CXt_LOOP_LAZYSV:
12140 ncx->blk_loop.state_u.lazysv.end
12141 = sv_dup_inc(ncx->blk_loop.state_u.lazysv.end, param);
12142 /* We are taking advantage of av_dup_inc and sv_dup_inc
12143 actually being the same function, and order equivalence of
12145 We can assert the later [but only at run time :-(] */
12146 assert ((void *) &ncx->blk_loop.state_u.ary.ary ==
12147 (void *) &ncx->blk_loop.state_u.lazysv.cur);
12149 ncx->blk_loop.state_u.ary.ary
12150 = av_dup_inc(ncx->blk_loop.state_u.ary.ary, param);
12151 case CXt_LOOP_LAZYIV:
12152 case CXt_LOOP_PLAIN:
12153 if (CxPADLOOP(ncx)) {
12154 ncx->blk_loop.itervar_u.oldcomppad
12155 = (PAD*)ptr_table_fetch(PL_ptr_table,
12156 ncx->blk_loop.itervar_u.oldcomppad);
12158 ncx->blk_loop.itervar_u.gv
12159 = gv_dup((const GV *)ncx->blk_loop.itervar_u.gv,
12164 ncx->blk_format.cv = cv_dup(ncx->blk_format.cv, param);
12165 ncx->blk_format.gv = gv_dup(ncx->blk_format.gv, param);
12166 ncx->blk_format.dfoutgv = gv_dup_inc(ncx->blk_format.dfoutgv,
12179 /* duplicate a stack info structure */
12182 Perl_si_dup(pTHX_ PERL_SI *si, CLONE_PARAMS* param)
12186 PERL_ARGS_ASSERT_SI_DUP;
12189 return (PERL_SI*)NULL;
12191 /* look for it in the table first */
12192 nsi = (PERL_SI*)ptr_table_fetch(PL_ptr_table, si);
12196 /* create anew and remember what it is */
12197 Newxz(nsi, 1, PERL_SI);
12198 ptr_table_store(PL_ptr_table, si, nsi);
12200 nsi->si_stack = av_dup_inc(si->si_stack, param);
12201 nsi->si_cxix = si->si_cxix;
12202 nsi->si_cxmax = si->si_cxmax;
12203 nsi->si_cxstack = cx_dup(si->si_cxstack, si->si_cxix, si->si_cxmax, param);
12204 nsi->si_type = si->si_type;
12205 nsi->si_prev = si_dup(si->si_prev, param);
12206 nsi->si_next = si_dup(si->si_next, param);
12207 nsi->si_markoff = si->si_markoff;
12212 #define POPINT(ss,ix) ((ss)[--(ix)].any_i32)
12213 #define TOPINT(ss,ix) ((ss)[ix].any_i32)
12214 #define POPLONG(ss,ix) ((ss)[--(ix)].any_long)
12215 #define TOPLONG(ss,ix) ((ss)[ix].any_long)
12216 #define POPIV(ss,ix) ((ss)[--(ix)].any_iv)
12217 #define TOPIV(ss,ix) ((ss)[ix].any_iv)
12218 #define POPUV(ss,ix) ((ss)[--(ix)].any_uv)
12219 #define TOPUV(ss,ix) ((ss)[ix].any_uv)
12220 #define POPBOOL(ss,ix) ((ss)[--(ix)].any_bool)
12221 #define TOPBOOL(ss,ix) ((ss)[ix].any_bool)
12222 #define POPPTR(ss,ix) ((ss)[--(ix)].any_ptr)
12223 #define TOPPTR(ss,ix) ((ss)[ix].any_ptr)
12224 #define POPDPTR(ss,ix) ((ss)[--(ix)].any_dptr)
12225 #define TOPDPTR(ss,ix) ((ss)[ix].any_dptr)
12226 #define POPDXPTR(ss,ix) ((ss)[--(ix)].any_dxptr)
12227 #define TOPDXPTR(ss,ix) ((ss)[ix].any_dxptr)
12230 #define pv_dup_inc(p) SAVEPV(p)
12231 #define pv_dup(p) SAVEPV(p)
12232 #define svp_dup_inc(p,pp) any_dup(p,pp)
12234 /* map any object to the new equivent - either something in the
12235 * ptr table, or something in the interpreter structure
12239 Perl_any_dup(pTHX_ void *v, const PerlInterpreter *proto_perl)
12243 PERL_ARGS_ASSERT_ANY_DUP;
12246 return (void*)NULL;
12248 /* look for it in the table first */
12249 ret = ptr_table_fetch(PL_ptr_table, v);
12253 /* see if it is part of the interpreter structure */
12254 if (v >= (void*)proto_perl && v < (void*)(proto_perl+1))
12255 ret = (void*)(((char*)aTHX) + (((char*)v) - (char*)proto_perl));
12263 /* duplicate the save stack */
12266 Perl_ss_dup(pTHX_ PerlInterpreter *proto_perl, CLONE_PARAMS* param)
12269 ANY * const ss = proto_perl->Isavestack;
12270 const I32 max = proto_perl->Isavestack_max;
12271 I32 ix = proto_perl->Isavestack_ix;
12284 void (*dptr) (void*);
12285 void (*dxptr) (pTHX_ void*);
12287 PERL_ARGS_ASSERT_SS_DUP;
12289 Newxz(nss, max, ANY);
12292 const UV uv = POPUV(ss,ix);
12293 const U8 type = (U8)uv & SAVE_MASK;
12295 TOPUV(nss,ix) = uv;
12297 case SAVEt_CLEARSV:
12299 case SAVEt_HELEM: /* hash element */
12300 sv = (const SV *)POPPTR(ss,ix);
12301 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12303 case SAVEt_ITEM: /* normal string */
12304 case SAVEt_GVSV: /* scalar slot in GV */
12305 case SAVEt_SV: /* scalar reference */
12306 sv = (const SV *)POPPTR(ss,ix);
12307 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12310 case SAVEt_MORTALIZESV:
12311 sv = (const SV *)POPPTR(ss,ix);
12312 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12314 case SAVEt_SHARED_PVREF: /* char* in shared space */
12315 c = (char*)POPPTR(ss,ix);
12316 TOPPTR(nss,ix) = savesharedpv(c);
12317 ptr = POPPTR(ss,ix);
12318 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12320 case SAVEt_GENERIC_SVREF: /* generic sv */
12321 case SAVEt_SVREF: /* scalar reference */
12322 sv = (const SV *)POPPTR(ss,ix);
12323 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12324 ptr = POPPTR(ss,ix);
12325 TOPPTR(nss,ix) = svp_dup_inc((SV**)ptr, proto_perl);/* XXXXX */
12327 case SAVEt_HV: /* hash reference */
12328 case SAVEt_AV: /* array reference */
12329 sv = (const SV *) POPPTR(ss,ix);
12330 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12332 case SAVEt_COMPPAD:
12334 sv = (const SV *) POPPTR(ss,ix);
12335 TOPPTR(nss,ix) = sv_dup(sv, param);
12337 case SAVEt_INT: /* int reference */
12338 ptr = POPPTR(ss,ix);
12339 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12340 intval = (int)POPINT(ss,ix);
12341 TOPINT(nss,ix) = intval;
12343 case SAVEt_LONG: /* long reference */
12344 ptr = POPPTR(ss,ix);
12345 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12346 longval = (long)POPLONG(ss,ix);
12347 TOPLONG(nss,ix) = longval;
12349 case SAVEt_I32: /* I32 reference */
12350 case SAVEt_COP_ARYBASE: /* call CopARYBASE_set */
12351 ptr = POPPTR(ss,ix);
12352 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12354 TOPINT(nss,ix) = i;
12356 case SAVEt_IV: /* IV reference */
12357 ptr = POPPTR(ss,ix);
12358 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12360 TOPIV(nss,ix) = iv;
12362 case SAVEt_HPTR: /* HV* reference */
12363 case SAVEt_APTR: /* AV* reference */
12364 case SAVEt_SPTR: /* SV* reference */
12365 ptr = POPPTR(ss,ix);
12366 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12367 sv = (const SV *)POPPTR(ss,ix);
12368 TOPPTR(nss,ix) = sv_dup(sv, param);
12370 case SAVEt_VPTR: /* random* reference */
12371 ptr = POPPTR(ss,ix);
12372 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12374 case SAVEt_INT_SMALL:
12375 case SAVEt_I32_SMALL:
12376 case SAVEt_I16: /* I16 reference */
12377 case SAVEt_I8: /* I8 reference */
12379 ptr = POPPTR(ss,ix);
12380 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12382 case SAVEt_GENERIC_PVREF: /* generic char* */
12383 case SAVEt_PPTR: /* char* reference */
12384 ptr = POPPTR(ss,ix);
12385 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12386 c = (char*)POPPTR(ss,ix);
12387 TOPPTR(nss,ix) = pv_dup(c);
12389 case SAVEt_GP: /* scalar reference */
12390 gp = (GP*)POPPTR(ss,ix);
12391 TOPPTR(nss,ix) = gp = gp_dup(gp, param);
12392 (void)GpREFCNT_inc(gp);
12393 gv = (const GV *)POPPTR(ss,ix);
12394 TOPPTR(nss,ix) = gv_dup_inc(gv, param);
12397 ptr = POPPTR(ss,ix);
12398 if (ptr && (((OP*)ptr)->op_private & OPpREFCOUNTED)) {
12399 /* these are assumed to be refcounted properly */
12401 switch (((OP*)ptr)->op_type) {
12403 case OP_LEAVESUBLV:
12407 case OP_LEAVEWRITE:
12408 TOPPTR(nss,ix) = ptr;
12411 (void) OpREFCNT_inc(o);
12415 TOPPTR(nss,ix) = NULL;
12420 TOPPTR(nss,ix) = NULL;
12422 case SAVEt_FREECOPHH:
12423 ptr = POPPTR(ss,ix);
12424 TOPPTR(nss,ix) = cophh_copy((COPHH *)ptr);
12427 hv = (const HV *)POPPTR(ss,ix);
12428 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
12430 TOPINT(nss,ix) = i;
12433 c = (char*)POPPTR(ss,ix);
12434 TOPPTR(nss,ix) = pv_dup_inc(c);
12436 case SAVEt_STACK_POS: /* Position on Perl stack */
12438 TOPINT(nss,ix) = i;
12440 case SAVEt_DESTRUCTOR:
12441 ptr = POPPTR(ss,ix);
12442 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
12443 dptr = POPDPTR(ss,ix);
12444 TOPDPTR(nss,ix) = DPTR2FPTR(void (*)(void*),
12445 any_dup(FPTR2DPTR(void *, dptr),
12448 case SAVEt_DESTRUCTOR_X:
12449 ptr = POPPTR(ss,ix);
12450 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
12451 dxptr = POPDXPTR(ss,ix);
12452 TOPDXPTR(nss,ix) = DPTR2FPTR(void (*)(pTHX_ void*),
12453 any_dup(FPTR2DPTR(void *, dxptr),
12456 case SAVEt_REGCONTEXT:
12458 ix -= uv >> SAVE_TIGHT_SHIFT;
12460 case SAVEt_AELEM: /* array element */
12461 sv = (const SV *)POPPTR(ss,ix);
12462 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12464 TOPINT(nss,ix) = i;
12465 av = (const AV *)POPPTR(ss,ix);
12466 TOPPTR(nss,ix) = av_dup_inc(av, param);
12469 ptr = POPPTR(ss,ix);
12470 TOPPTR(nss,ix) = ptr;
12473 ptr = POPPTR(ss,ix);
12474 ptr = cophh_copy((COPHH*)ptr);
12475 TOPPTR(nss,ix) = ptr;
12477 TOPINT(nss,ix) = i;
12478 if (i & HINT_LOCALIZE_HH) {
12479 hv = (const HV *)POPPTR(ss,ix);
12480 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
12483 case SAVEt_PADSV_AND_MORTALIZE:
12484 longval = (long)POPLONG(ss,ix);
12485 TOPLONG(nss,ix) = longval;
12486 ptr = POPPTR(ss,ix);
12487 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12488 sv = (const SV *)POPPTR(ss,ix);
12489 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12491 case SAVEt_SET_SVFLAGS:
12493 TOPINT(nss,ix) = i;
12495 TOPINT(nss,ix) = i;
12496 sv = (const SV *)POPPTR(ss,ix);
12497 TOPPTR(nss,ix) = sv_dup(sv, param);
12499 case SAVEt_RE_STATE:
12501 const struct re_save_state *const old_state
12502 = (struct re_save_state *)
12503 (ss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
12504 struct re_save_state *const new_state
12505 = (struct re_save_state *)
12506 (nss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
12508 Copy(old_state, new_state, 1, struct re_save_state);
12509 ix -= SAVESTACK_ALLOC_FOR_RE_SAVE_STATE;
12511 new_state->re_state_bostr
12512 = pv_dup(old_state->re_state_bostr);
12513 new_state->re_state_reginput
12514 = pv_dup(old_state->re_state_reginput);
12515 new_state->re_state_regeol
12516 = pv_dup(old_state->re_state_regeol);
12517 new_state->re_state_regoffs
12518 = (regexp_paren_pair*)
12519 any_dup(old_state->re_state_regoffs, proto_perl);
12520 new_state->re_state_reglastparen
12521 = (U32*) any_dup(old_state->re_state_reglastparen,
12523 new_state->re_state_reglastcloseparen
12524 = (U32*)any_dup(old_state->re_state_reglastcloseparen,
12526 /* XXX This just has to be broken. The old save_re_context
12527 code did SAVEGENERICPV(PL_reg_start_tmp);
12528 PL_reg_start_tmp is char **.
12529 Look above to what the dup code does for
12530 SAVEt_GENERIC_PVREF
12531 It can never have worked.
12532 So this is merely a faithful copy of the exiting bug: */
12533 new_state->re_state_reg_start_tmp
12534 = (char **) pv_dup((char *)
12535 old_state->re_state_reg_start_tmp);
12536 /* I assume that it only ever "worked" because no-one called
12537 (pseudo)fork while the regexp engine had re-entered itself.
12539 #ifdef PERL_OLD_COPY_ON_WRITE
12540 new_state->re_state_nrs
12541 = sv_dup(old_state->re_state_nrs, param);
12543 new_state->re_state_reg_magic
12544 = (MAGIC*) any_dup(old_state->re_state_reg_magic,
12546 new_state->re_state_reg_oldcurpm
12547 = (PMOP*) any_dup(old_state->re_state_reg_oldcurpm,
12549 new_state->re_state_reg_curpm
12550 = (PMOP*) any_dup(old_state->re_state_reg_curpm,
12552 new_state->re_state_reg_oldsaved
12553 = pv_dup(old_state->re_state_reg_oldsaved);
12554 new_state->re_state_reg_poscache
12555 = pv_dup(old_state->re_state_reg_poscache);
12556 new_state->re_state_reg_starttry
12557 = pv_dup(old_state->re_state_reg_starttry);
12560 case SAVEt_COMPILE_WARNINGS:
12561 ptr = POPPTR(ss,ix);
12562 TOPPTR(nss,ix) = DUP_WARNINGS((STRLEN*)ptr);
12565 ptr = POPPTR(ss,ix);
12566 TOPPTR(nss,ix) = parser_dup((const yy_parser*)ptr, param);
12570 "panic: ss_dup inconsistency (%"IVdf")", (IV) type);
12578 /* if sv is a stash, call $class->CLONE_SKIP(), and set the SVphv_CLONEABLE
12579 * flag to the result. This is done for each stash before cloning starts,
12580 * so we know which stashes want their objects cloned */
12583 do_mark_cloneable_stash(pTHX_ SV *const sv)
12585 const HEK * const hvname = HvNAME_HEK((const HV *)sv);
12587 GV* const cloner = gv_fetchmethod_autoload(MUTABLE_HV(sv), "CLONE_SKIP", 0);
12588 SvFLAGS(sv) |= SVphv_CLONEABLE; /* clone objects by default */
12589 if (cloner && GvCV(cloner)) {
12596 mXPUSHs(newSVhek(hvname));
12598 call_sv(MUTABLE_SV(GvCV(cloner)), G_SCALAR);
12605 SvFLAGS(sv) &= ~SVphv_CLONEABLE;
12613 =for apidoc perl_clone
12615 Create and return a new interpreter by cloning the current one.
12617 perl_clone takes these flags as parameters:
12619 CLONEf_COPY_STACKS - is used to, well, copy the stacks also,
12620 without it we only clone the data and zero the stacks,
12621 with it we copy the stacks and the new perl interpreter is
12622 ready to run at the exact same point as the previous one.
12623 The pseudo-fork code uses COPY_STACKS while the
12624 threads->create doesn't.
12626 CLONEf_KEEP_PTR_TABLE
12627 perl_clone keeps a ptr_table with the pointer of the old
12628 variable as a key and the new variable as a value,
12629 this allows it to check if something has been cloned and not
12630 clone it again but rather just use the value and increase the
12631 refcount. If KEEP_PTR_TABLE is not set then perl_clone will kill
12632 the ptr_table using the function
12633 C<ptr_table_free(PL_ptr_table); PL_ptr_table = NULL;>,
12634 reason to keep it around is if you want to dup some of your own
12635 variable who are outside the graph perl scans, example of this
12636 code is in threads.xs create
12639 This is a win32 thing, it is ignored on unix, it tells perls
12640 win32host code (which is c++) to clone itself, this is needed on
12641 win32 if you want to run two threads at the same time,
12642 if you just want to do some stuff in a separate perl interpreter
12643 and then throw it away and return to the original one,
12644 you don't need to do anything.
12649 /* XXX the above needs expanding by someone who actually understands it ! */
12650 EXTERN_C PerlInterpreter *
12651 perl_clone_host(PerlInterpreter* proto_perl, UV flags);
12654 perl_clone(PerlInterpreter *proto_perl, UV flags)
12657 #ifdef PERL_IMPLICIT_SYS
12659 PERL_ARGS_ASSERT_PERL_CLONE;
12661 /* perlhost.h so we need to call into it
12662 to clone the host, CPerlHost should have a c interface, sky */
12664 if (flags & CLONEf_CLONE_HOST) {
12665 return perl_clone_host(proto_perl,flags);
12667 return perl_clone_using(proto_perl, flags,
12669 proto_perl->IMemShared,
12670 proto_perl->IMemParse,
12672 proto_perl->IStdIO,
12676 proto_perl->IProc);
12680 perl_clone_using(PerlInterpreter *proto_perl, UV flags,
12681 struct IPerlMem* ipM, struct IPerlMem* ipMS,
12682 struct IPerlMem* ipMP, struct IPerlEnv* ipE,
12683 struct IPerlStdIO* ipStd, struct IPerlLIO* ipLIO,
12684 struct IPerlDir* ipD, struct IPerlSock* ipS,
12685 struct IPerlProc* ipP)
12687 /* XXX many of the string copies here can be optimized if they're
12688 * constants; they need to be allocated as common memory and just
12689 * their pointers copied. */
12692 CLONE_PARAMS clone_params;
12693 CLONE_PARAMS* const param = &clone_params;
12695 PerlInterpreter * const my_perl = (PerlInterpreter*)(*ipM->pMalloc)(ipM, sizeof(PerlInterpreter));
12697 PERL_ARGS_ASSERT_PERL_CLONE_USING;
12698 #else /* !PERL_IMPLICIT_SYS */
12700 CLONE_PARAMS clone_params;
12701 CLONE_PARAMS* param = &clone_params;
12702 PerlInterpreter * const my_perl = (PerlInterpreter*)PerlMem_malloc(sizeof(PerlInterpreter));
12704 PERL_ARGS_ASSERT_PERL_CLONE;
12705 #endif /* PERL_IMPLICIT_SYS */
12707 /* for each stash, determine whether its objects should be cloned */
12708 S_visit(proto_perl, do_mark_cloneable_stash, SVt_PVHV, SVTYPEMASK);
12709 PERL_SET_THX(my_perl);
12712 PoisonNew(my_perl, 1, PerlInterpreter);
12715 PL_defstash = NULL; /* may be used by perl malloc() */
12718 PL_scopestack_name = 0;
12720 PL_savestack_ix = 0;
12721 PL_savestack_max = -1;
12722 PL_sig_pending = 0;
12724 Zero(&PL_debug_pad, 1, struct perl_debug_pad);
12725 # ifdef DEBUG_LEAKING_SCALARS
12726 PL_sv_serial = (((UV)my_perl >> 2) & 0xfff) * 1000000;
12728 #else /* !DEBUGGING */
12729 Zero(my_perl, 1, PerlInterpreter);
12730 #endif /* DEBUGGING */
12732 #ifdef PERL_IMPLICIT_SYS
12733 /* host pointers */
12735 PL_MemShared = ipMS;
12736 PL_MemParse = ipMP;
12743 #endif /* PERL_IMPLICIT_SYS */
12745 param->flags = flags;
12746 /* Nothing in the core code uses this, but we make it available to
12747 extensions (using mg_dup). */
12748 param->proto_perl = proto_perl;
12749 /* Likely nothing will use this, but it is initialised to be consistent
12750 with Perl_clone_params_new(). */
12751 param->new_perl = my_perl;
12752 param->unreferenced = NULL;
12754 INIT_TRACK_MEMPOOL(my_perl->Imemory_debug_header, my_perl);
12756 PL_body_arenas = NULL;
12757 Zero(&PL_body_roots, 1, PL_body_roots);
12760 PL_sv_objcount = 0;
12762 PL_sv_arenaroot = NULL;
12764 PL_debug = proto_perl->Idebug;
12766 PL_hash_seed = proto_perl->Ihash_seed;
12767 PL_rehash_seed = proto_perl->Irehash_seed;
12769 SvANY(&PL_sv_undef) = NULL;
12770 SvREFCNT(&PL_sv_undef) = (~(U32)0)/2;
12771 SvFLAGS(&PL_sv_undef) = SVf_READONLY|SVt_NULL;
12772 SvREFCNT(&PL_sv_no) = (~(U32)0)/2;
12773 SvFLAGS(&PL_sv_no) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
12774 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
12776 SvANY(&PL_sv_yes) = new_XPVNV();
12777 SvREFCNT(&PL_sv_yes) = (~(U32)0)/2;
12778 SvFLAGS(&PL_sv_yes) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
12779 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
12781 /* dbargs array probably holds garbage */
12784 PL_compiling = proto_perl->Icompiling;
12786 #ifdef PERL_DEBUG_READONLY_OPS
12791 /* pseudo environmental stuff */
12792 PL_origargc = proto_perl->Iorigargc;
12793 PL_origargv = proto_perl->Iorigargv;
12795 /* Set tainting stuff before PerlIO_debug can possibly get called */
12796 PL_tainting = proto_perl->Itainting;
12797 PL_taint_warn = proto_perl->Itaint_warn;
12799 PL_minus_c = proto_perl->Iminus_c;
12801 PL_localpatches = proto_perl->Ilocalpatches;
12802 PL_splitstr = proto_perl->Isplitstr;
12803 PL_minus_n = proto_perl->Iminus_n;
12804 PL_minus_p = proto_perl->Iminus_p;
12805 PL_minus_l = proto_perl->Iminus_l;
12806 PL_minus_a = proto_perl->Iminus_a;
12807 PL_minus_E = proto_perl->Iminus_E;
12808 PL_minus_F = proto_perl->Iminus_F;
12809 PL_doswitches = proto_perl->Idoswitches;
12810 PL_dowarn = proto_perl->Idowarn;
12811 PL_sawampersand = proto_perl->Isawampersand;
12812 PL_unsafe = proto_perl->Iunsafe;
12813 PL_perldb = proto_perl->Iperldb;
12814 PL_perl_destruct_level = proto_perl->Iperl_destruct_level;
12815 PL_exit_flags = proto_perl->Iexit_flags;
12817 /* XXX time(&PL_basetime) when asked for? */
12818 PL_basetime = proto_perl->Ibasetime;
12820 PL_maxsysfd = proto_perl->Imaxsysfd;
12821 PL_statusvalue = proto_perl->Istatusvalue;
12823 PL_statusvalue_vms = proto_perl->Istatusvalue_vms;
12825 PL_statusvalue_posix = proto_perl->Istatusvalue_posix;
12828 /* RE engine related */
12829 Zero(&PL_reg_state, 1, struct re_save_state);
12830 PL_reginterp_cnt = 0;
12831 PL_regmatch_slab = NULL;
12833 PL_sub_generation = proto_perl->Isub_generation;
12835 /* funky return mechanisms */
12836 PL_forkprocess = proto_perl->Iforkprocess;
12838 /* internal state */
12839 PL_maxo = proto_perl->Imaxo;
12841 PL_main_start = proto_perl->Imain_start;
12842 PL_eval_root = proto_perl->Ieval_root;
12843 PL_eval_start = proto_perl->Ieval_start;
12845 PL_filemode = proto_perl->Ifilemode;
12846 PL_lastfd = proto_perl->Ilastfd;
12847 PL_oldname = proto_perl->Ioldname; /* XXX not quite right */
12850 PL_gensym = proto_perl->Igensym;
12852 PL_laststatval = proto_perl->Ilaststatval;
12853 PL_laststype = proto_perl->Ilaststype;
12856 PL_profiledata = NULL;
12858 PL_generation = proto_perl->Igeneration;
12860 PL_in_clean_objs = proto_perl->Iin_clean_objs;
12861 PL_in_clean_all = proto_perl->Iin_clean_all;
12863 PL_uid = proto_perl->Iuid;
12864 PL_euid = proto_perl->Ieuid;
12865 PL_gid = proto_perl->Igid;
12866 PL_egid = proto_perl->Iegid;
12867 PL_nomemok = proto_perl->Inomemok;
12868 PL_an = proto_perl->Ian;
12869 PL_evalseq = proto_perl->Ievalseq;
12870 PL_origenviron = proto_perl->Iorigenviron; /* XXX not quite right */
12871 PL_origalen = proto_perl->Iorigalen;
12873 PL_sighandlerp = proto_perl->Isighandlerp;
12875 PL_runops = proto_perl->Irunops;
12877 PL_subline = proto_perl->Isubline;
12880 PL_cryptseen = proto_perl->Icryptseen;
12883 PL_hints = proto_perl->Ihints;
12885 PL_amagic_generation = proto_perl->Iamagic_generation;
12887 #ifdef USE_LOCALE_COLLATE
12888 PL_collation_ix = proto_perl->Icollation_ix;
12889 PL_collation_standard = proto_perl->Icollation_standard;
12890 PL_collxfrm_base = proto_perl->Icollxfrm_base;
12891 PL_collxfrm_mult = proto_perl->Icollxfrm_mult;
12892 #endif /* USE_LOCALE_COLLATE */
12894 #ifdef USE_LOCALE_NUMERIC
12895 PL_numeric_standard = proto_perl->Inumeric_standard;
12896 PL_numeric_local = proto_perl->Inumeric_local;
12897 #endif /* !USE_LOCALE_NUMERIC */
12899 /* Did the locale setup indicate UTF-8? */
12900 PL_utf8locale = proto_perl->Iutf8locale;
12901 /* Unicode features (see perlrun/-C) */
12902 PL_unicode = proto_perl->Iunicode;
12904 /* Pre-5.8 signals control */
12905 PL_signals = proto_perl->Isignals;
12907 /* times() ticks per second */
12908 PL_clocktick = proto_perl->Iclocktick;
12910 /* Recursion stopper for PerlIO_find_layer */
12911 PL_in_load_module = proto_perl->Iin_load_module;
12913 /* sort() routine */
12914 PL_sort_RealCmp = proto_perl->Isort_RealCmp;
12916 /* Not really needed/useful since the reenrant_retint is "volatile",
12917 * but do it for consistency's sake. */
12918 PL_reentrant_retint = proto_perl->Ireentrant_retint;
12920 /* Hooks to shared SVs and locks. */
12921 PL_sharehook = proto_perl->Isharehook;
12922 PL_lockhook = proto_perl->Ilockhook;
12923 PL_unlockhook = proto_perl->Iunlockhook;
12924 PL_threadhook = proto_perl->Ithreadhook;
12925 PL_destroyhook = proto_perl->Idestroyhook;
12926 PL_signalhook = proto_perl->Isignalhook;
12928 #ifdef THREADS_HAVE_PIDS
12929 PL_ppid = proto_perl->Ippid;
12933 PL_last_swash_hv = NULL; /* reinits on demand */
12934 PL_last_swash_klen = 0;
12935 PL_last_swash_key[0]= '\0';
12936 PL_last_swash_tmps = (U8*)NULL;
12937 PL_last_swash_slen = 0;
12939 PL_glob_index = proto_perl->Iglob_index;
12940 PL_srand_called = proto_perl->Isrand_called;
12942 if (flags & CLONEf_COPY_STACKS) {
12943 /* next allocation will be PL_tmps_stack[PL_tmps_ix+1] */
12944 PL_tmps_ix = proto_perl->Itmps_ix;
12945 PL_tmps_max = proto_perl->Itmps_max;
12946 PL_tmps_floor = proto_perl->Itmps_floor;
12948 /* next push_scope()/ENTER sets PL_scopestack[PL_scopestack_ix]
12949 * NOTE: unlike the others! */
12950 PL_scopestack_ix = proto_perl->Iscopestack_ix;
12951 PL_scopestack_max = proto_perl->Iscopestack_max;
12953 /* next SSPUSHFOO() sets PL_savestack[PL_savestack_ix]
12954 * NOTE: unlike the others! */
12955 PL_savestack_ix = proto_perl->Isavestack_ix;
12956 PL_savestack_max = proto_perl->Isavestack_max;
12959 PL_start_env = proto_perl->Istart_env; /* XXXXXX */
12960 PL_top_env = &PL_start_env;
12962 PL_op = proto_perl->Iop;
12965 PL_Xpv = (XPV*)NULL;
12966 my_perl->Ina = proto_perl->Ina;
12968 PL_statbuf = proto_perl->Istatbuf;
12969 PL_statcache = proto_perl->Istatcache;
12972 PL_timesbuf = proto_perl->Itimesbuf;
12975 PL_tainted = proto_perl->Itainted;
12976 PL_curpm = proto_perl->Icurpm; /* XXX No PMOP ref count */
12978 PL_chopset = proto_perl->Ichopset; /* XXX never deallocated */
12980 PL_restartjmpenv = proto_perl->Irestartjmpenv;
12981 PL_restartop = proto_perl->Irestartop;
12982 PL_in_eval = proto_perl->Iin_eval;
12983 PL_delaymagic = proto_perl->Idelaymagic;
12984 PL_phase = proto_perl->Iphase;
12985 PL_localizing = proto_perl->Ilocalizing;
12987 PL_hv_fetch_ent_mh = NULL;
12988 PL_modcount = proto_perl->Imodcount;
12989 PL_lastgotoprobe = NULL;
12990 PL_dumpindent = proto_perl->Idumpindent;
12992 PL_efloatbuf = NULL; /* reinits on demand */
12993 PL_efloatsize = 0; /* reinits on demand */
12997 PL_screamfirst = NULL;
12998 PL_screamnext = NULL;
12999 PL_maxscream = -1; /* reinits on demand */
13000 PL_lastscream = NULL;
13003 PL_regdummy = proto_perl->Iregdummy;
13004 PL_colorset = 0; /* reinits PL_colors[] */
13005 /*PL_colors[6] = {0,0,0,0,0,0};*/
13007 /* Pluggable optimizer */
13008 PL_peepp = proto_perl->Ipeepp;
13009 PL_rpeepp = proto_perl->Irpeepp;
13010 /* op_free() hook */
13011 PL_opfreehook = proto_perl->Iopfreehook;
13013 #ifdef USE_REENTRANT_API
13014 /* XXX: things like -Dm will segfault here in perlio, but doing
13015 * PERL_SET_CONTEXT(proto_perl);
13016 * breaks too many other things
13018 Perl_reentrant_init(aTHX);
13021 /* create SV map for pointer relocation */
13022 PL_ptr_table = ptr_table_new();
13024 /* initialize these special pointers as early as possible */
13025 ptr_table_store(PL_ptr_table, &proto_perl->Isv_undef, &PL_sv_undef);
13027 SvANY(&PL_sv_no) = new_XPVNV();
13028 SvPV_set(&PL_sv_no, savepvn(PL_No, 0));
13029 SvCUR_set(&PL_sv_no, 0);
13030 SvLEN_set(&PL_sv_no, 1);
13031 SvIV_set(&PL_sv_no, 0);
13032 SvNV_set(&PL_sv_no, 0);
13033 ptr_table_store(PL_ptr_table, &proto_perl->Isv_no, &PL_sv_no);
13035 SvPV_set(&PL_sv_yes, savepvn(PL_Yes, 1));
13036 SvCUR_set(&PL_sv_yes, 1);
13037 SvLEN_set(&PL_sv_yes, 2);
13038 SvIV_set(&PL_sv_yes, 1);
13039 SvNV_set(&PL_sv_yes, 1);
13040 ptr_table_store(PL_ptr_table, &proto_perl->Isv_yes, &PL_sv_yes);
13042 /* create (a non-shared!) shared string table */
13043 PL_strtab = newHV();
13044 HvSHAREKEYS_off(PL_strtab);
13045 hv_ksplit(PL_strtab, HvTOTALKEYS(proto_perl->Istrtab));
13046 ptr_table_store(PL_ptr_table, proto_perl->Istrtab, PL_strtab);
13048 /* These two PVs will be free'd special way so must set them same way op.c does */
13049 PL_compiling.cop_stashpv = savesharedpv(PL_compiling.cop_stashpv);
13050 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_stashpv, PL_compiling.cop_stashpv);
13052 PL_compiling.cop_file = savesharedpv(PL_compiling.cop_file);
13053 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_file, PL_compiling.cop_file);
13055 ptr_table_store(PL_ptr_table, &proto_perl->Icompiling, &PL_compiling);
13056 PL_compiling.cop_warnings = DUP_WARNINGS(PL_compiling.cop_warnings);
13057 CopHINTHASH_set(&PL_compiling, cophh_copy(CopHINTHASH_get(&PL_compiling)));
13058 PL_curcop = (COP*)any_dup(proto_perl->Icurcop, proto_perl);
13060 param->stashes = newAV(); /* Setup array of objects to call clone on */
13061 /* This makes no difference to the implementation, as it always pushes
13062 and shifts pointers to other SVs without changing their reference
13063 count, with the array becoming empty before it is freed. However, it
13064 makes it conceptually clear what is going on, and will avoid some
13065 work inside av.c, filling slots between AvFILL() and AvMAX() with
13066 &PL_sv_undef, and SvREFCNT_dec()ing those. */
13067 AvREAL_off(param->stashes);
13069 if (!(flags & CLONEf_COPY_STACKS)) {
13070 param->unreferenced = newAV();
13073 #ifdef PERLIO_LAYERS
13074 /* Clone PerlIO tables as soon as we can handle general xx_dup() */
13075 PerlIO_clone(aTHX_ proto_perl, param);
13078 PL_envgv = gv_dup(proto_perl->Ienvgv, param);
13079 PL_incgv = gv_dup(proto_perl->Iincgv, param);
13080 PL_hintgv = gv_dup(proto_perl->Ihintgv, param);
13081 PL_origfilename = SAVEPV(proto_perl->Iorigfilename);
13082 PL_diehook = sv_dup_inc(proto_perl->Idiehook, param);
13083 PL_warnhook = sv_dup_inc(proto_perl->Iwarnhook, param);
13086 PL_patchlevel = sv_dup_inc(proto_perl->Ipatchlevel, param);
13087 PL_apiversion = sv_dup_inc(proto_perl->Iapiversion, param);
13088 PL_inplace = SAVEPV(proto_perl->Iinplace);
13089 PL_e_script = sv_dup_inc(proto_perl->Ie_script, param);
13091 /* magical thingies */
13092 PL_formfeed = sv_dup(proto_perl->Iformfeed, param);
13094 PL_encoding = sv_dup(proto_perl->Iencoding, param);
13096 sv_setpvs(PERL_DEBUG_PAD(0), ""); /* For regex debugging. */
13097 sv_setpvs(PERL_DEBUG_PAD(1), ""); /* ext/re needs these */
13098 sv_setpvs(PERL_DEBUG_PAD(2), ""); /* even without DEBUGGING. */
13101 /* Clone the regex array */
13102 /* ORANGE FIXME for plugins, probably in the SV dup code.
13103 newSViv(PTR2IV(CALLREGDUPE(
13104 INT2PTR(REGEXP *, SvIVX(regex)), param))))
13106 PL_regex_padav = av_dup_inc(proto_perl->Iregex_padav, param);
13107 PL_regex_pad = AvARRAY(PL_regex_padav);
13109 /* shortcuts to various I/O objects */
13110 PL_ofsgv = gv_dup_inc(proto_perl->Iofsgv, param);
13111 PL_stdingv = gv_dup(proto_perl->Istdingv, param);
13112 PL_stderrgv = gv_dup(proto_perl->Istderrgv, param);
13113 PL_defgv = gv_dup(proto_perl->Idefgv, param);
13114 PL_argvgv = gv_dup(proto_perl->Iargvgv, param);
13115 PL_argvoutgv = gv_dup(proto_perl->Iargvoutgv, param);
13116 PL_argvout_stack = av_dup_inc(proto_perl->Iargvout_stack, param);
13118 /* shortcuts to regexp stuff */
13119 PL_replgv = gv_dup(proto_perl->Ireplgv, param);
13121 /* shortcuts to misc objects */
13122 PL_errgv = gv_dup(proto_perl->Ierrgv, param);
13124 /* shortcuts to debugging objects */
13125 PL_DBgv = gv_dup(proto_perl->IDBgv, param);
13126 PL_DBline = gv_dup(proto_perl->IDBline, param);
13127 PL_DBsub = gv_dup(proto_perl->IDBsub, param);
13128 PL_DBsingle = sv_dup(proto_perl->IDBsingle, param);
13129 PL_DBtrace = sv_dup(proto_perl->IDBtrace, param);
13130 PL_DBsignal = sv_dup(proto_perl->IDBsignal, param);
13132 /* symbol tables */
13133 PL_defstash = hv_dup_inc(proto_perl->Idefstash, param);
13134 PL_curstash = hv_dup(proto_perl->Icurstash, param);
13135 PL_debstash = hv_dup(proto_perl->Idebstash, param);
13136 PL_globalstash = hv_dup(proto_perl->Iglobalstash, param);
13137 PL_curstname = sv_dup_inc(proto_perl->Icurstname, param);
13139 PL_beginav = av_dup_inc(proto_perl->Ibeginav, param);
13140 PL_beginav_save = av_dup_inc(proto_perl->Ibeginav_save, param);
13141 PL_checkav_save = av_dup_inc(proto_perl->Icheckav_save, param);
13142 PL_unitcheckav = av_dup_inc(proto_perl->Iunitcheckav, param);
13143 PL_unitcheckav_save = av_dup_inc(proto_perl->Iunitcheckav_save, param);
13144 PL_endav = av_dup_inc(proto_perl->Iendav, param);
13145 PL_checkav = av_dup_inc(proto_perl->Icheckav, param);
13146 PL_initav = av_dup_inc(proto_perl->Iinitav, param);
13148 PL_isarev = hv_dup_inc(proto_perl->Iisarev, param);
13150 /* subprocess state */
13151 PL_fdpid = av_dup_inc(proto_perl->Ifdpid, param);
13153 if (proto_perl->Iop_mask)
13154 PL_op_mask = SAVEPVN(proto_perl->Iop_mask, PL_maxo);
13157 /* PL_asserting = proto_perl->Iasserting; */
13159 /* current interpreter roots */
13160 PL_main_cv = cv_dup_inc(proto_perl->Imain_cv, param);
13162 PL_main_root = OpREFCNT_inc(proto_perl->Imain_root);
13165 /* runtime control stuff */
13166 PL_curcopdb = (COP*)any_dup(proto_perl->Icurcopdb, proto_perl);
13168 PL_preambleav = av_dup_inc(proto_perl->Ipreambleav, param);
13170 PL_ors_sv = sv_dup_inc(proto_perl->Iors_sv, param);
13172 /* interpreter atexit processing */
13173 PL_exitlistlen = proto_perl->Iexitlistlen;
13174 if (PL_exitlistlen) {
13175 Newx(PL_exitlist, PL_exitlistlen, PerlExitListEntry);
13176 Copy(proto_perl->Iexitlist, PL_exitlist, PL_exitlistlen, PerlExitListEntry);
13179 PL_exitlist = (PerlExitListEntry*)NULL;
13181 PL_my_cxt_size = proto_perl->Imy_cxt_size;
13182 if (PL_my_cxt_size) {
13183 Newx(PL_my_cxt_list, PL_my_cxt_size, void *);
13184 Copy(proto_perl->Imy_cxt_list, PL_my_cxt_list, PL_my_cxt_size, void *);
13185 #ifdef PERL_GLOBAL_STRUCT_PRIVATE
13186 Newx(PL_my_cxt_keys, PL_my_cxt_size, const char *);
13187 Copy(proto_perl->Imy_cxt_keys, PL_my_cxt_keys, PL_my_cxt_size, char *);
13191 PL_my_cxt_list = (void**)NULL;
13192 #ifdef PERL_GLOBAL_STRUCT_PRIVATE
13193 PL_my_cxt_keys = (const char**)NULL;
13196 PL_modglobal = hv_dup_inc(proto_perl->Imodglobal, param);
13197 PL_custom_op_names = hv_dup_inc(proto_perl->Icustom_op_names,param);
13198 PL_custom_op_descs = hv_dup_inc(proto_perl->Icustom_op_descs,param);
13199 PL_custom_ops = hv_dup_inc(proto_perl->Icustom_ops, param);
13201 PL_compcv = cv_dup(proto_perl->Icompcv, param);
13203 PAD_CLONE_VARS(proto_perl, param);
13205 #ifdef HAVE_INTERP_INTERN
13206 sys_intern_dup(&proto_perl->Isys_intern, &PL_sys_intern);
13209 PL_DBcv = cv_dup(proto_perl->IDBcv, param);
13211 #ifdef PERL_USES_PL_PIDSTATUS
13212 PL_pidstatus = newHV(); /* XXX flag for cloning? */
13214 PL_osname = SAVEPV(proto_perl->Iosname);
13215 PL_parser = parser_dup(proto_perl->Iparser, param);
13217 /* XXX this only works if the saved cop has already been cloned */
13218 if (proto_perl->Iparser) {
13219 PL_parser->saved_curcop = (COP*)any_dup(
13220 proto_perl->Iparser->saved_curcop,
13224 PL_subname = sv_dup_inc(proto_perl->Isubname, param);
13226 #ifdef USE_LOCALE_COLLATE
13227 PL_collation_name = SAVEPV(proto_perl->Icollation_name);
13228 #endif /* USE_LOCALE_COLLATE */
13230 #ifdef USE_LOCALE_NUMERIC
13231 PL_numeric_name = SAVEPV(proto_perl->Inumeric_name);
13232 PL_numeric_radix_sv = sv_dup_inc(proto_perl->Inumeric_radix_sv, param);
13233 #endif /* !USE_LOCALE_NUMERIC */
13235 /* utf8 character classes */
13236 PL_utf8_alnum = sv_dup_inc(proto_perl->Iutf8_alnum, param);
13237 PL_utf8_ascii = sv_dup_inc(proto_perl->Iutf8_ascii, param);
13238 PL_utf8_alpha = sv_dup_inc(proto_perl->Iutf8_alpha, param);
13239 PL_utf8_space = sv_dup_inc(proto_perl->Iutf8_space, param);
13240 PL_utf8_cntrl = sv_dup_inc(proto_perl->Iutf8_cntrl, param);
13241 PL_utf8_graph = sv_dup_inc(proto_perl->Iutf8_graph, param);
13242 PL_utf8_digit = sv_dup_inc(proto_perl->Iutf8_digit, param);
13243 PL_utf8_upper = sv_dup_inc(proto_perl->Iutf8_upper, param);
13244 PL_utf8_lower = sv_dup_inc(proto_perl->Iutf8_lower, param);
13245 PL_utf8_print = sv_dup_inc(proto_perl->Iutf8_print, param);
13246 PL_utf8_punct = sv_dup_inc(proto_perl->Iutf8_punct, param);
13247 PL_utf8_xdigit = sv_dup_inc(proto_perl->Iutf8_xdigit, param);
13248 PL_utf8_mark = sv_dup_inc(proto_perl->Iutf8_mark, param);
13249 PL_utf8_X_begin = sv_dup_inc(proto_perl->Iutf8_X_begin, param);
13250 PL_utf8_X_extend = sv_dup_inc(proto_perl->Iutf8_X_extend, param);
13251 PL_utf8_X_prepend = sv_dup_inc(proto_perl->Iutf8_X_prepend, param);
13252 PL_utf8_X_non_hangul = sv_dup_inc(proto_perl->Iutf8_X_non_hangul, param);
13253 PL_utf8_X_L = sv_dup_inc(proto_perl->Iutf8_X_L, param);
13254 PL_utf8_X_LV = sv_dup_inc(proto_perl->Iutf8_X_LV, param);
13255 PL_utf8_X_LVT = sv_dup_inc(proto_perl->Iutf8_X_LVT, param);
13256 PL_utf8_X_T = sv_dup_inc(proto_perl->Iutf8_X_T, param);
13257 PL_utf8_X_V = sv_dup_inc(proto_perl->Iutf8_X_V, param);
13258 PL_utf8_X_LV_LVT_V = sv_dup_inc(proto_perl->Iutf8_X_LV_LVT_V, param);
13259 PL_utf8_toupper = sv_dup_inc(proto_perl->Iutf8_toupper, param);
13260 PL_utf8_totitle = sv_dup_inc(proto_perl->Iutf8_totitle, param);
13261 PL_utf8_tolower = sv_dup_inc(proto_perl->Iutf8_tolower, param);
13262 PL_utf8_tofold = sv_dup_inc(proto_perl->Iutf8_tofold, param);
13263 PL_utf8_idstart = sv_dup_inc(proto_perl->Iutf8_idstart, param);
13264 PL_utf8_xidstart = sv_dup_inc(proto_perl->Iutf8_xidstart, param);
13265 PL_utf8_idcont = sv_dup_inc(proto_perl->Iutf8_idcont, param);
13266 PL_utf8_xidcont = sv_dup_inc(proto_perl->Iutf8_xidcont, param);
13267 PL_utf8_foldable = hv_dup_inc(proto_perl->Iutf8_foldable, param);
13270 if (proto_perl->Ipsig_pend) {
13271 Newxz(PL_psig_pend, SIG_SIZE, int);
13274 PL_psig_pend = (int*)NULL;
13277 if (proto_perl->Ipsig_name) {
13278 Newx(PL_psig_name, 2 * SIG_SIZE, SV*);
13279 sv_dup_inc_multiple(proto_perl->Ipsig_name, PL_psig_name, 2 * SIG_SIZE,
13281 PL_psig_ptr = PL_psig_name + SIG_SIZE;
13284 PL_psig_ptr = (SV**)NULL;
13285 PL_psig_name = (SV**)NULL;
13288 if (flags & CLONEf_COPY_STACKS) {
13289 Newx(PL_tmps_stack, PL_tmps_max, SV*);
13290 sv_dup_inc_multiple(proto_perl->Itmps_stack, PL_tmps_stack,
13291 PL_tmps_ix+1, param);
13293 /* next PUSHMARK() sets *(PL_markstack_ptr+1) */
13294 i = proto_perl->Imarkstack_max - proto_perl->Imarkstack;
13295 Newxz(PL_markstack, i, I32);
13296 PL_markstack_max = PL_markstack + (proto_perl->Imarkstack_max
13297 - proto_perl->Imarkstack);
13298 PL_markstack_ptr = PL_markstack + (proto_perl->Imarkstack_ptr
13299 - proto_perl->Imarkstack);
13300 Copy(proto_perl->Imarkstack, PL_markstack,
13301 PL_markstack_ptr - PL_markstack + 1, I32);
13303 /* next push_scope()/ENTER sets PL_scopestack[PL_scopestack_ix]
13304 * NOTE: unlike the others! */
13305 Newxz(PL_scopestack, PL_scopestack_max, I32);
13306 Copy(proto_perl->Iscopestack, PL_scopestack, PL_scopestack_ix, I32);
13309 Newxz(PL_scopestack_name, PL_scopestack_max, const char *);
13310 Copy(proto_perl->Iscopestack_name, PL_scopestack_name, PL_scopestack_ix, const char *);
13312 /* NOTE: si_dup() looks at PL_markstack */
13313 PL_curstackinfo = si_dup(proto_perl->Icurstackinfo, param);
13315 /* PL_curstack = PL_curstackinfo->si_stack; */
13316 PL_curstack = av_dup(proto_perl->Icurstack, param);
13317 PL_mainstack = av_dup(proto_perl->Imainstack, param);
13319 /* next PUSHs() etc. set *(PL_stack_sp+1) */
13320 PL_stack_base = AvARRAY(PL_curstack);
13321 PL_stack_sp = PL_stack_base + (proto_perl->Istack_sp
13322 - proto_perl->Istack_base);
13323 PL_stack_max = PL_stack_base + AvMAX(PL_curstack);
13325 /*Newxz(PL_savestack, PL_savestack_max, ANY);*/
13326 PL_savestack = ss_dup(proto_perl, param);
13330 ENTER; /* perl_destruct() wants to LEAVE; */
13333 PL_statgv = gv_dup(proto_perl->Istatgv, param);
13334 PL_statname = sv_dup_inc(proto_perl->Istatname, param);
13336 PL_rs = sv_dup_inc(proto_perl->Irs, param);
13337 PL_last_in_gv = gv_dup(proto_perl->Ilast_in_gv, param);
13338 PL_defoutgv = gv_dup_inc(proto_perl->Idefoutgv, param);
13339 PL_toptarget = sv_dup_inc(proto_perl->Itoptarget, param);
13340 PL_bodytarget = sv_dup_inc(proto_perl->Ibodytarget, param);
13341 PL_formtarget = sv_dup(proto_perl->Iformtarget, param);
13343 PL_errors = sv_dup_inc(proto_perl->Ierrors, param);
13345 PL_sortcop = (OP*)any_dup(proto_perl->Isortcop, proto_perl);
13346 PL_sortstash = hv_dup(proto_perl->Isortstash, param);
13347 PL_firstgv = gv_dup(proto_perl->Ifirstgv, param);
13348 PL_secondgv = gv_dup(proto_perl->Isecondgv, param);
13350 PL_stashcache = newHV();
13352 PL_watchaddr = (char **) ptr_table_fetch(PL_ptr_table,
13353 proto_perl->Iwatchaddr);
13354 PL_watchok = PL_watchaddr ? * PL_watchaddr : NULL;
13355 if (PL_debug && PL_watchaddr) {
13356 PerlIO_printf(Perl_debug_log,
13357 "WATCHING: %"UVxf" cloned as %"UVxf" with value %"UVxf"\n",
13358 PTR2UV(proto_perl->Iwatchaddr), PTR2UV(PL_watchaddr),
13359 PTR2UV(PL_watchok));
13362 PL_registered_mros = hv_dup_inc(proto_perl->Iregistered_mros, param);
13363 PL_blockhooks = av_dup_inc(proto_perl->Iblockhooks, param);
13364 PL_utf8_foldclosures = hv_dup_inc(proto_perl->Iutf8_foldclosures, param);
13366 /* Call the ->CLONE method, if it exists, for each of the stashes
13367 identified by sv_dup() above.
13369 while(av_len(param->stashes) != -1) {
13370 HV* const stash = MUTABLE_HV(av_shift(param->stashes));
13371 GV* const cloner = gv_fetchmethod_autoload(stash, "CLONE", 0);
13372 if (cloner && GvCV(cloner)) {
13377 mXPUSHs(newSVhek(HvNAME_HEK(stash)));
13379 call_sv(MUTABLE_SV(GvCV(cloner)), G_DISCARD);
13385 if (!(flags & CLONEf_KEEP_PTR_TABLE)) {
13386 ptr_table_free(PL_ptr_table);
13387 PL_ptr_table = NULL;
13390 if (!(flags & CLONEf_COPY_STACKS)) {
13391 unreferenced_to_tmp_stack(param->unreferenced);
13394 SvREFCNT_dec(param->stashes);
13396 /* orphaned? eg threads->new inside BEGIN or use */
13397 if (PL_compcv && ! SvREFCNT(PL_compcv)) {
13398 SvREFCNT_inc_simple_void(PL_compcv);
13399 SAVEFREESV(PL_compcv);
13406 S_unreferenced_to_tmp_stack(pTHX_ AV *const unreferenced)
13408 PERL_ARGS_ASSERT_UNREFERENCED_TO_TMP_STACK;
13410 if (AvFILLp(unreferenced) > -1) {
13411 SV **svp = AvARRAY(unreferenced);
13412 SV **const last = svp + AvFILLp(unreferenced);
13416 if (SvREFCNT(*svp) == 1)
13418 } while (++svp <= last);
13420 EXTEND_MORTAL(count);
13421 svp = AvARRAY(unreferenced);
13424 if (SvREFCNT(*svp) == 1) {
13425 /* Our reference is the only one to this SV. This means that
13426 in this thread, the scalar effectively has a 0 reference.
13427 That doesn't work (cleanup never happens), so donate our
13428 reference to it onto the save stack. */
13429 PL_tmps_stack[++PL_tmps_ix] = *svp;
13431 /* As an optimisation, because we are already walking the
13432 entire array, instead of above doing either
13433 SvREFCNT_inc(*svp) or *svp = &PL_sv_undef, we can instead
13434 release our reference to the scalar, so that at the end of
13435 the array owns zero references to the scalars it happens to
13436 point to. We are effectively converting the array from
13437 AvREAL() on to AvREAL() off. This saves the av_clear()
13438 (triggered by the SvREFCNT_dec(unreferenced) below) from
13439 walking the array a second time. */
13440 SvREFCNT_dec(*svp);
13443 } while (++svp <= last);
13444 AvREAL_off(unreferenced);
13446 SvREFCNT_dec(unreferenced);
13450 Perl_clone_params_del(CLONE_PARAMS *param)
13452 /* This seemingly funky ordering keeps the build with PERL_GLOBAL_STRUCT
13454 PerlInterpreter *const to = param->new_perl;
13456 PerlInterpreter *const was = PERL_GET_THX;
13458 PERL_ARGS_ASSERT_CLONE_PARAMS_DEL;
13464 SvREFCNT_dec(param->stashes);
13465 if (param->unreferenced)
13466 unreferenced_to_tmp_stack(param->unreferenced);
13476 Perl_clone_params_new(PerlInterpreter *const from, PerlInterpreter *const to)
13479 /* Need to play this game, as newAV() can call safesysmalloc(), and that
13480 does a dTHX; to get the context from thread local storage.
13481 FIXME - under PERL_CORE Newx(), Safefree() and friends should expand to
13482 a version that passes in my_perl. */
13483 PerlInterpreter *const was = PERL_GET_THX;
13484 CLONE_PARAMS *param;
13486 PERL_ARGS_ASSERT_CLONE_PARAMS_NEW;
13492 /* Given that we've set the context, we can do this unshared. */
13493 Newx(param, 1, CLONE_PARAMS);
13496 param->proto_perl = from;
13497 param->new_perl = to;
13498 param->stashes = (AV *)Perl_newSV_type(to, SVt_PVAV);
13499 AvREAL_off(param->stashes);
13500 param->unreferenced = (AV *)Perl_newSV_type(to, SVt_PVAV);
13508 #endif /* USE_ITHREADS */
13511 =head1 Unicode Support
13513 =for apidoc sv_recode_to_utf8
13515 The encoding is assumed to be an Encode object, on entry the PV
13516 of the sv is assumed to be octets in that encoding, and the sv
13517 will be converted into Unicode (and UTF-8).
13519 If the sv already is UTF-8 (or if it is not POK), or if the encoding
13520 is not a reference, nothing is done to the sv. If the encoding is not
13521 an C<Encode::XS> Encoding object, bad things will happen.
13522 (See F<lib/encoding.pm> and L<Encode>).
13524 The PV of the sv is returned.
13529 Perl_sv_recode_to_utf8(pTHX_ SV *sv, SV *encoding)
13533 PERL_ARGS_ASSERT_SV_RECODE_TO_UTF8;
13535 if (SvPOK(sv) && !SvUTF8(sv) && !IN_BYTES && SvROK(encoding)) {
13549 Passing sv_yes is wrong - it needs to be or'ed set of constants
13550 for Encode::XS, while UTf-8 decode (currently) assumes a true value means
13551 remove converted chars from source.
13553 Both will default the value - let them.
13555 XPUSHs(&PL_sv_yes);
13558 call_method("decode", G_SCALAR);
13562 s = SvPV_const(uni, len);
13563 if (s != SvPVX_const(sv)) {
13564 SvGROW(sv, len + 1);
13565 Move(s, SvPVX(sv), len + 1, char);
13566 SvCUR_set(sv, len);
13570 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
13571 /* clear pos and any utf8 cache */
13572 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
13575 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
13576 magic_setutf8(sv,mg); /* clear UTF8 cache */
13581 return SvPOKp(sv) ? SvPVX(sv) : NULL;
13585 =for apidoc sv_cat_decode
13587 The encoding is assumed to be an Encode object, the PV of the ssv is
13588 assumed to be octets in that encoding and decoding the input starts
13589 from the position which (PV + *offset) pointed to. The dsv will be
13590 concatenated the decoded UTF-8 string from ssv. Decoding will terminate
13591 when the string tstr appears in decoding output or the input ends on
13592 the PV of the ssv. The value which the offset points will be modified
13593 to the last input position on the ssv.
13595 Returns TRUE if the terminator was found, else returns FALSE.
13600 Perl_sv_cat_decode(pTHX_ SV *dsv, SV *encoding,
13601 SV *ssv, int *offset, char *tstr, int tlen)
13606 PERL_ARGS_ASSERT_SV_CAT_DECODE;
13608 if (SvPOK(ssv) && SvPOK(dsv) && SvROK(encoding) && offset) {
13619 offsv = newSViv(*offset);
13621 mXPUSHp(tstr, tlen);
13623 call_method("cat_decode", G_SCALAR);
13625 ret = SvTRUE(TOPs);
13626 *offset = SvIV(offsv);
13632 Perl_croak(aTHX_ "Invalid argument to sv_cat_decode");
13637 /* ---------------------------------------------------------------------
13639 * support functions for report_uninit()
13642 /* the maxiumum size of array or hash where we will scan looking
13643 * for the undefined element that triggered the warning */
13645 #define FUV_MAX_SEARCH_SIZE 1000
13647 /* Look for an entry in the hash whose value has the same SV as val;
13648 * If so, return a mortal copy of the key. */
13651 S_find_hash_subscript(pTHX_ const HV *const hv, const SV *const val)
13654 register HE **array;
13657 PERL_ARGS_ASSERT_FIND_HASH_SUBSCRIPT;
13659 if (!hv || SvMAGICAL(hv) || !HvARRAY(hv) ||
13660 (HvTOTALKEYS(hv) > FUV_MAX_SEARCH_SIZE))
13663 array = HvARRAY(hv);
13665 for (i=HvMAX(hv); i>0; i--) {
13666 register HE *entry;
13667 for (entry = array[i]; entry; entry = HeNEXT(entry)) {
13668 if (HeVAL(entry) != val)
13670 if ( HeVAL(entry) == &PL_sv_undef ||
13671 HeVAL(entry) == &PL_sv_placeholder)
13675 if (HeKLEN(entry) == HEf_SVKEY)
13676 return sv_mortalcopy(HeKEY_sv(entry));
13677 return sv_2mortal(newSVhek(HeKEY_hek(entry)));
13683 /* Look for an entry in the array whose value has the same SV as val;
13684 * If so, return the index, otherwise return -1. */
13687 S_find_array_subscript(pTHX_ const AV *const av, const SV *const val)
13691 PERL_ARGS_ASSERT_FIND_ARRAY_SUBSCRIPT;
13693 if (!av || SvMAGICAL(av) || !AvARRAY(av) ||
13694 (AvFILLp(av) > FUV_MAX_SEARCH_SIZE))
13697 if (val != &PL_sv_undef) {
13698 SV ** const svp = AvARRAY(av);
13701 for (i=AvFILLp(av); i>=0; i--)
13708 /* S_varname(): return the name of a variable, optionally with a subscript.
13709 * If gv is non-zero, use the name of that global, along with gvtype (one
13710 * of "$", "@", "%"); otherwise use the name of the lexical at pad offset
13711 * targ. Depending on the value of the subscript_type flag, return:
13714 #define FUV_SUBSCRIPT_NONE 1 /* "@foo" */
13715 #define FUV_SUBSCRIPT_ARRAY 2 /* "$foo[aindex]" */
13716 #define FUV_SUBSCRIPT_HASH 3 /* "$foo{keyname}" */
13717 #define FUV_SUBSCRIPT_WITHIN 4 /* "within @foo" */
13720 S_varname(pTHX_ const GV *const gv, const char gvtype, PADOFFSET targ,
13721 const SV *const keyname, I32 aindex, int subscript_type)
13724 SV * const name = sv_newmortal();
13727 buffer[0] = gvtype;
13730 /* as gv_fullname4(), but add literal '^' for $^FOO names */
13732 gv_fullname4(name, gv, buffer, 0);
13734 if ((unsigned int)SvPVX(name)[1] <= 26) {
13736 buffer[1] = SvPVX(name)[1] + 'A' - 1;
13738 /* Swap the 1 unprintable control character for the 2 byte pretty
13739 version - ie substr($name, 1, 1) = $buffer; */
13740 sv_insert(name, 1, 1, buffer, 2);
13744 CV * const cv = find_runcv(NULL);
13748 if (!cv || !CvPADLIST(cv))
13750 av = MUTABLE_AV((*av_fetch(CvPADLIST(cv), 0, FALSE)));
13751 sv = *av_fetch(av, targ, FALSE);
13752 sv_setpvn(name, SvPV_nolen_const(sv), SvCUR(sv));
13755 if (subscript_type == FUV_SUBSCRIPT_HASH) {
13756 SV * const sv = newSV(0);
13757 *SvPVX(name) = '$';
13758 Perl_sv_catpvf(aTHX_ name, "{%s}",
13759 pv_display(sv,SvPVX_const(keyname), SvCUR(keyname), 0, 32));
13762 else if (subscript_type == FUV_SUBSCRIPT_ARRAY) {
13763 *SvPVX(name) = '$';
13764 Perl_sv_catpvf(aTHX_ name, "[%"IVdf"]", (IV)aindex);
13766 else if (subscript_type == FUV_SUBSCRIPT_WITHIN) {
13767 /* We know that name has no magic, so can use 0 instead of SV_GMAGIC */
13768 Perl_sv_insert_flags(aTHX_ name, 0, 0, STR_WITH_LEN("within "), 0);
13776 =for apidoc find_uninit_var
13778 Find the name of the undefined variable (if any) that caused the operator o
13779 to issue a "Use of uninitialized value" warning.
13780 If match is true, only return a name if it's value matches uninit_sv.
13781 So roughly speaking, if a unary operator (such as OP_COS) generates a
13782 warning, then following the direct child of the op may yield an
13783 OP_PADSV or OP_GV that gives the name of the undefined variable. On the
13784 other hand, with OP_ADD there are two branches to follow, so we only print
13785 the variable name if we get an exact match.
13787 The name is returned as a mortal SV.
13789 Assumes that PL_op is the op that originally triggered the error, and that
13790 PL_comppad/PL_curpad points to the currently executing pad.
13796 S_find_uninit_var(pTHX_ const OP *const obase, const SV *const uninit_sv,
13802 const OP *o, *o2, *kid;
13804 if (!obase || (match && (!uninit_sv || uninit_sv == &PL_sv_undef ||
13805 uninit_sv == &PL_sv_placeholder)))
13808 switch (obase->op_type) {
13815 const bool pad = (obase->op_type == OP_PADAV || obase->op_type == OP_PADHV);
13816 const bool hash = (obase->op_type == OP_PADHV || obase->op_type == OP_RV2HV);
13819 int subscript_type = FUV_SUBSCRIPT_WITHIN;
13821 if (pad) { /* @lex, %lex */
13822 sv = PAD_SVl(obase->op_targ);
13826 if (cUNOPx(obase)->op_first->op_type == OP_GV) {
13827 /* @global, %global */
13828 gv = cGVOPx_gv(cUNOPx(obase)->op_first);
13831 sv = hash ? MUTABLE_SV(GvHV(gv)): MUTABLE_SV(GvAV(gv));
13833 else /* @{expr}, %{expr} */
13834 return find_uninit_var(cUNOPx(obase)->op_first,
13838 /* attempt to find a match within the aggregate */
13840 keysv = find_hash_subscript((const HV*)sv, uninit_sv);
13842 subscript_type = FUV_SUBSCRIPT_HASH;
13845 index = find_array_subscript((const AV *)sv, uninit_sv);
13847 subscript_type = FUV_SUBSCRIPT_ARRAY;
13850 if (match && subscript_type == FUV_SUBSCRIPT_WITHIN)
13853 return varname(gv, hash ? '%' : '@', obase->op_targ,
13854 keysv, index, subscript_type);
13858 if (match && PAD_SVl(obase->op_targ) != uninit_sv)
13860 return varname(NULL, '$', obase->op_targ,
13861 NULL, 0, FUV_SUBSCRIPT_NONE);
13864 gv = cGVOPx_gv(obase);
13865 if (!gv || (match && GvSV(gv) != uninit_sv) || !GvSTASH(gv))
13867 return varname(gv, '$', 0, NULL, 0, FUV_SUBSCRIPT_NONE);
13869 case OP_AELEMFAST_LEX:
13872 AV *av = MUTABLE_AV(PAD_SV(obase->op_targ));
13873 if (!av || SvRMAGICAL(av))
13875 svp = av_fetch(av, (I32)obase->op_private, FALSE);
13876 if (!svp || *svp != uninit_sv)
13879 return varname(NULL, '$', obase->op_targ,
13880 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
13883 gv = cGVOPx_gv(obase);
13888 AV *const av = GvAV(gv);
13889 if (!av || SvRMAGICAL(av))
13891 svp = av_fetch(av, (I32)obase->op_private, FALSE);
13892 if (!svp || *svp != uninit_sv)
13895 return varname(gv, '$', 0,
13896 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
13901 o = cUNOPx(obase)->op_first;
13902 if (!o || o->op_type != OP_NULL ||
13903 ! (o->op_targ == OP_AELEM || o->op_targ == OP_HELEM))
13905 return find_uninit_var(cBINOPo->op_last, uninit_sv, match);
13909 if (PL_op == obase)
13910 /* $a[uninit_expr] or $h{uninit_expr} */
13911 return find_uninit_var(cBINOPx(obase)->op_last, uninit_sv, match);
13914 o = cBINOPx(obase)->op_first;
13915 kid = cBINOPx(obase)->op_last;
13917 /* get the av or hv, and optionally the gv */
13919 if (o->op_type == OP_PADAV || o->op_type == OP_PADHV) {
13920 sv = PAD_SV(o->op_targ);
13922 else if ((o->op_type == OP_RV2AV || o->op_type == OP_RV2HV)
13923 && cUNOPo->op_first->op_type == OP_GV)
13925 gv = cGVOPx_gv(cUNOPo->op_first);
13929 == OP_RV2HV ? MUTABLE_SV(GvHV(gv)) : MUTABLE_SV(GvAV(gv));
13934 if (kid && kid->op_type == OP_CONST && SvOK(cSVOPx_sv(kid))) {
13935 /* index is constant */
13939 if (obase->op_type == OP_HELEM) {
13940 HE* he = hv_fetch_ent(MUTABLE_HV(sv), cSVOPx_sv(kid), 0, 0);
13941 if (!he || HeVAL(he) != uninit_sv)
13945 SV * const * const svp = av_fetch(MUTABLE_AV(sv), SvIV(cSVOPx_sv(kid)), FALSE);
13946 if (!svp || *svp != uninit_sv)
13950 if (obase->op_type == OP_HELEM)
13951 return varname(gv, '%', o->op_targ,
13952 cSVOPx_sv(kid), 0, FUV_SUBSCRIPT_HASH);
13954 return varname(gv, '@', o->op_targ, NULL,
13955 SvIV(cSVOPx_sv(kid)), FUV_SUBSCRIPT_ARRAY);
13958 /* index is an expression;
13959 * attempt to find a match within the aggregate */
13960 if (obase->op_type == OP_HELEM) {
13961 SV * const keysv = find_hash_subscript((const HV*)sv, uninit_sv);
13963 return varname(gv, '%', o->op_targ,
13964 keysv, 0, FUV_SUBSCRIPT_HASH);
13968 = find_array_subscript((const AV *)sv, uninit_sv);
13970 return varname(gv, '@', o->op_targ,
13971 NULL, index, FUV_SUBSCRIPT_ARRAY);
13976 (o->op_type == OP_PADAV || o->op_type == OP_RV2AV)
13978 o->op_targ, NULL, 0, FUV_SUBSCRIPT_WITHIN);
13983 /* only examine RHS */
13984 return find_uninit_var(cBINOPx(obase)->op_first, uninit_sv, match);
13987 o = cUNOPx(obase)->op_first;
13988 if (o->op_type == OP_PUSHMARK)
13991 if (!o->op_sibling) {
13992 /* one-arg version of open is highly magical */
13994 if (o->op_type == OP_GV) { /* open FOO; */
13996 if (match && GvSV(gv) != uninit_sv)
13998 return varname(gv, '$', 0,
13999 NULL, 0, FUV_SUBSCRIPT_NONE);
14001 /* other possibilities not handled are:
14002 * open $x; or open my $x; should return '${*$x}'
14003 * open expr; should return '$'.expr ideally
14009 /* ops where $_ may be an implicit arg */
14013 if ( !(obase->op_flags & OPf_STACKED)) {
14014 if (uninit_sv == ((obase->op_private & OPpTARGET_MY)
14015 ? PAD_SVl(obase->op_targ)
14018 sv = sv_newmortal();
14019 sv_setpvs(sv, "$_");
14028 match = 1; /* print etc can return undef on defined args */
14029 /* skip filehandle as it can't produce 'undef' warning */
14030 o = cUNOPx(obase)->op_first;
14031 if ((obase->op_flags & OPf_STACKED) && o->op_type == OP_PUSHMARK)
14032 o = o->op_sibling->op_sibling;
14036 case OP_ENTEREVAL: /* could be eval $undef or $x='$undef'; eval $x */
14038 case OP_CUSTOM: /* XS or custom code could trigger random warnings */
14040 /* the following ops are capable of returning PL_sv_undef even for
14041 * defined arg(s) */
14060 case OP_GETPEERNAME:
14108 case OP_SMARTMATCH:
14117 /* XXX tmp hack: these two may call an XS sub, and currently
14118 XS subs don't have a SUB entry on the context stack, so CV and
14119 pad determination goes wrong, and BAD things happen. So, just
14120 don't try to determine the value under those circumstances.
14121 Need a better fix at dome point. DAPM 11/2007 */
14127 GV * const gv = gv_fetchpvs(".", GV_NOTQUAL, SVt_PV);
14128 if (gv && GvSV(gv) == uninit_sv)
14129 return newSVpvs_flags("$.", SVs_TEMP);
14134 /* def-ness of rval pos() is independent of the def-ness of its arg */
14135 if ( !(obase->op_flags & OPf_MOD))
14140 if (SvROK(PL_rs) && uninit_sv == SvRV(PL_rs))
14141 return newSVpvs_flags("${$/}", SVs_TEMP);
14146 if (!(obase->op_flags & OPf_KIDS))
14148 o = cUNOPx(obase)->op_first;
14154 /* if all except one arg are constant, or have no side-effects,
14155 * or are optimized away, then it's unambiguous */
14157 for (kid=o; kid; kid = kid->op_sibling) {
14159 const OPCODE type = kid->op_type;
14160 if ( (type == OP_CONST && SvOK(cSVOPx_sv(kid)))
14161 || (type == OP_NULL && ! (kid->op_flags & OPf_KIDS))
14162 || (type == OP_PUSHMARK)
14164 /* @$a and %$a, but not @a or %a */
14165 (type == OP_RV2AV || type == OP_RV2HV)
14166 && cUNOPx(kid)->op_first
14167 && cUNOPx(kid)->op_first->op_type != OP_GV
14172 if (o2) { /* more than one found */
14179 return find_uninit_var(o2, uninit_sv, match);
14181 /* scan all args */
14183 sv = find_uninit_var(o, uninit_sv, 1);
14195 =for apidoc report_uninit
14197 Print appropriate "Use of uninitialized variable" warning
14203 Perl_report_uninit(pTHX_ const SV *uninit_sv)
14207 SV* varname = NULL;
14209 varname = find_uninit_var(PL_op, uninit_sv,0);
14211 sv_insert(varname, 0, 0, " ", 1);
14213 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
14214 varname ? SvPV_nolen_const(varname) : "",
14215 " in ", OP_DESC(PL_op));
14218 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
14224 * c-indentation-style: bsd
14225 * c-basic-offset: 4
14226 * indent-tabs-mode: t
14229 * ex: set ts=8 sts=4 sw=4 noet: