3 * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, 2000,
4 * 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009 by Larry Wall
7 * You may distribute under the terms of either the GNU General Public
8 * License or the Artistic License, as specified in the README file.
13 * 'I wonder what the Entish is for "yes" and "no",' he thought.
16 * [p.480 of _The Lord of the Rings_, III/iv: "Treebeard"]
22 * This file contains the code that creates, manipulates and destroys
23 * scalar values (SVs). The other types (AV, HV, GV, etc.) reuse the
24 * structure of an SV, so their creation and destruction is handled
25 * here; higher-level functions are in av.c, hv.c, and so on. Opcode
26 * level functions (eg. substr, split, join) for each of the types are
36 # if __STDC_VERSION__ >= 199901L && !defined(VMS)
47 /* Missing proto on LynxOS */
48 char *gconvert(double, int, int, char *);
51 #ifdef PERL_UTF8_CACHE_ASSERT
52 /* if adding more checks watch out for the following tests:
53 * t/op/index.t t/op/length.t t/op/pat.t t/op/substr.t
54 * lib/utf8.t lib/Unicode/Collate/t/index.t
57 # define ASSERT_UTF8_CACHE(cache) \
58 STMT_START { if (cache) { assert((cache)[0] <= (cache)[1]); \
59 assert((cache)[2] <= (cache)[3]); \
60 assert((cache)[3] <= (cache)[1]);} \
63 # define ASSERT_UTF8_CACHE(cache) NOOP
66 #ifdef PERL_OLD_COPY_ON_WRITE
67 #define SV_COW_NEXT_SV(sv) INT2PTR(SV *,SvUVX(sv))
68 #define SV_COW_NEXT_SV_SET(current,next) SvUV_set(current, PTR2UV(next))
69 /* This is a pessimistic view. Scalar must be purely a read-write PV to copy-
73 /* ============================================================================
75 =head1 Allocation and deallocation of SVs.
77 An SV (or AV, HV, etc.) is allocated in two parts: the head (struct
78 sv, av, hv...) contains type and reference count information, and for
79 many types, a pointer to the body (struct xrv, xpv, xpviv...), which
80 contains fields specific to each type. Some types store all they need
81 in the head, so don't have a body.
83 In all but the most memory-paranoid configurations (ex: PURIFY), heads
84 and bodies are allocated out of arenas, which by default are
85 approximately 4K chunks of memory parcelled up into N heads or bodies.
86 Sv-bodies are allocated by their sv-type, guaranteeing size
87 consistency needed to allocate safely from arrays.
89 For SV-heads, the first slot in each arena is reserved, and holds a
90 link to the next arena, some flags, and a note of the number of slots.
91 Snaked through each arena chain is a linked list of free items; when
92 this becomes empty, an extra arena is allocated and divided up into N
93 items which are threaded into the free list.
95 SV-bodies are similar, but they use arena-sets by default, which
96 separate the link and info from the arena itself, and reclaim the 1st
97 slot in the arena. SV-bodies are further described later.
99 The following global variables are associated with arenas:
101 PL_sv_arenaroot pointer to list of SV arenas
102 PL_sv_root pointer to list of free SV structures
104 PL_body_arenas head of linked-list of body arenas
105 PL_body_roots[] array of pointers to list of free bodies of svtype
106 arrays are indexed by the svtype needed
108 A few special SV heads are not allocated from an arena, but are
109 instead directly created in the interpreter structure, eg PL_sv_undef.
110 The size of arenas can be changed from the default by setting
111 PERL_ARENA_SIZE appropriately at compile time.
113 The SV arena serves the secondary purpose of allowing still-live SVs
114 to be located and destroyed during final cleanup.
116 At the lowest level, the macros new_SV() and del_SV() grab and free
117 an SV head. (If debugging with -DD, del_SV() calls the function S_del_sv()
118 to return the SV to the free list with error checking.) new_SV() calls
119 more_sv() / sv_add_arena() to add an extra arena if the free list is empty.
120 SVs in the free list have their SvTYPE field set to all ones.
122 At the time of very final cleanup, sv_free_arenas() is called from
123 perl_destruct() to physically free all the arenas allocated since the
124 start of the interpreter.
126 The function visit() scans the SV arenas list, and calls a specified
127 function for each SV it finds which is still live - ie which has an SvTYPE
128 other than all 1's, and a non-zero SvREFCNT. visit() is used by the
129 following functions (specified as [function that calls visit()] / [function
130 called by visit() for each SV]):
132 sv_report_used() / do_report_used()
133 dump all remaining SVs (debugging aid)
135 sv_clean_objs() / do_clean_objs(),do_clean_named_objs(),
136 do_clean_named_io_objs()
137 Attempt to free all objects pointed to by RVs,
138 and try to do the same for all objects indirectly
139 referenced by typeglobs too. Called once from
140 perl_destruct(), prior to calling sv_clean_all()
143 sv_clean_all() / do_clean_all()
144 SvREFCNT_dec(sv) each remaining SV, possibly
145 triggering an sv_free(). It also sets the
146 SVf_BREAK flag on the SV to indicate that the
147 refcnt has been artificially lowered, and thus
148 stopping sv_free() from giving spurious warnings
149 about SVs which unexpectedly have a refcnt
150 of zero. called repeatedly from perl_destruct()
151 until there are no SVs left.
153 =head2 Arena allocator API Summary
155 Private API to rest of sv.c
159 new_XPVNV(), del_XPVGV(),
164 sv_report_used(), sv_clean_objs(), sv_clean_all(), sv_free_arenas()
168 * ========================================================================= */
171 * "A time to plant, and a time to uproot what was planted..."
175 # define MEM_LOG_NEW_SV(sv, file, line, func) \
176 Perl_mem_log_new_sv(sv, file, line, func)
177 # define MEM_LOG_DEL_SV(sv, file, line, func) \
178 Perl_mem_log_del_sv(sv, file, line, func)
180 # define MEM_LOG_NEW_SV(sv, file, line, func) NOOP
181 # define MEM_LOG_DEL_SV(sv, file, line, func) NOOP
184 #ifdef DEBUG_LEAKING_SCALARS
185 # define FREE_SV_DEBUG_FILE(sv) Safefree((sv)->sv_debug_file)
186 # define DEBUG_SV_SERIAL(sv) \
187 DEBUG_m(PerlIO_printf(Perl_debug_log, "0x%"UVxf": (%05ld) del_SV\n", \
188 PTR2UV(sv), (long)(sv)->sv_debug_serial))
190 # define FREE_SV_DEBUG_FILE(sv)
191 # define DEBUG_SV_SERIAL(sv) NOOP
195 # define SvARENA_CHAIN(sv) ((sv)->sv_u.svu_rv)
196 # define SvARENA_CHAIN_SET(sv,val) (sv)->sv_u.svu_rv = MUTABLE_SV((val))
197 /* Whilst I'd love to do this, it seems that things like to check on
199 # define POSION_SV_HEAD(sv) PoisonNew(sv, 1, struct STRUCT_SV)
201 # define POSION_SV_HEAD(sv) PoisonNew(&SvANY(sv), 1, void *), \
202 PoisonNew(&SvREFCNT(sv), 1, U32)
204 # define SvARENA_CHAIN(sv) SvANY(sv)
205 # define SvARENA_CHAIN_SET(sv,val) SvANY(sv) = (void *)(val)
206 # define POSION_SV_HEAD(sv)
209 /* Mark an SV head as unused, and add to free list.
211 * If SVf_BREAK is set, skip adding it to the free list, as this SV had
212 * its refcount artificially decremented during global destruction, so
213 * there may be dangling pointers to it. The last thing we want in that
214 * case is for it to be reused. */
216 #define plant_SV(p) \
218 const U32 old_flags = SvFLAGS(p); \
219 MEM_LOG_DEL_SV(p, __FILE__, __LINE__, FUNCTION__); \
220 DEBUG_SV_SERIAL(p); \
221 FREE_SV_DEBUG_FILE(p); \
223 SvFLAGS(p) = SVTYPEMASK; \
224 if (!(old_flags & SVf_BREAK)) { \
225 SvARENA_CHAIN_SET(p, PL_sv_root); \
231 #define uproot_SV(p) \
234 PL_sv_root = MUTABLE_SV(SvARENA_CHAIN(p)); \
239 /* make some more SVs by adding another arena */
246 char *chunk; /* must use New here to match call to */
247 Newx(chunk,PERL_ARENA_SIZE,char); /* Safefree() in sv_free_arenas() */
248 sv_add_arena(chunk, PERL_ARENA_SIZE, 0);
253 /* new_SV(): return a new, empty SV head */
255 #ifdef DEBUG_LEAKING_SCALARS
256 /* provide a real function for a debugger to play with */
258 S_new_SV(pTHX_ const char *file, int line, const char *func)
265 sv = S_more_sv(aTHX);
269 sv->sv_debug_optype = PL_op ? PL_op->op_type : 0;
270 sv->sv_debug_line = (U16) (PL_parser && PL_parser->copline != NOLINE
276 sv->sv_debug_inpad = 0;
277 sv->sv_debug_parent = NULL;
278 sv->sv_debug_file = PL_curcop ? savepv(CopFILE(PL_curcop)): NULL;
280 sv->sv_debug_serial = PL_sv_serial++;
282 MEM_LOG_NEW_SV(sv, file, line, func);
283 DEBUG_m(PerlIO_printf(Perl_debug_log, "0x%"UVxf": (%05ld) new_SV (from %s:%d [%s])\n",
284 PTR2UV(sv), (long)sv->sv_debug_serial, file, line, func));
288 # define new_SV(p) (p)=S_new_SV(aTHX_ __FILE__, __LINE__, FUNCTION__)
296 (p) = S_more_sv(aTHX); \
300 MEM_LOG_NEW_SV(p, __FILE__, __LINE__, FUNCTION__); \
305 /* del_SV(): return an empty SV head to the free list */
318 S_del_sv(pTHX_ SV *p)
322 PERL_ARGS_ASSERT_DEL_SV;
327 for (sva = PL_sv_arenaroot; sva; sva = MUTABLE_SV(SvANY(sva))) {
328 const SV * const sv = sva + 1;
329 const SV * const svend = &sva[SvREFCNT(sva)];
330 if (p >= sv && p < svend) {
336 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL),
337 "Attempt to free non-arena SV: 0x%"UVxf
338 pTHX__FORMAT, PTR2UV(p) pTHX__VALUE);
345 #else /* ! DEBUGGING */
347 #define del_SV(p) plant_SV(p)
349 #endif /* DEBUGGING */
353 =head1 SV Manipulation Functions
355 =for apidoc sv_add_arena
357 Given a chunk of memory, link it to the head of the list of arenas,
358 and split it into a list of free SVs.
364 S_sv_add_arena(pTHX_ char *const ptr, const U32 size, const U32 flags)
367 SV *const sva = MUTABLE_SV(ptr);
371 PERL_ARGS_ASSERT_SV_ADD_ARENA;
373 /* The first SV in an arena isn't an SV. */
374 SvANY(sva) = (void *) PL_sv_arenaroot; /* ptr to next arena */
375 SvREFCNT(sva) = size / sizeof(SV); /* number of SV slots */
376 SvFLAGS(sva) = flags; /* FAKE if not to be freed */
378 PL_sv_arenaroot = sva;
379 PL_sv_root = sva + 1;
381 svend = &sva[SvREFCNT(sva) - 1];
384 SvARENA_CHAIN_SET(sv, (sv + 1));
388 /* Must always set typemask because it's always checked in on cleanup
389 when the arenas are walked looking for objects. */
390 SvFLAGS(sv) = SVTYPEMASK;
393 SvARENA_CHAIN_SET(sv, 0);
397 SvFLAGS(sv) = SVTYPEMASK;
400 /* visit(): call the named function for each non-free SV in the arenas
401 * whose flags field matches the flags/mask args. */
404 S_visit(pTHX_ SVFUNC_t f, const U32 flags, const U32 mask)
410 PERL_ARGS_ASSERT_VISIT;
412 for (sva = PL_sv_arenaroot; sva; sva = MUTABLE_SV(SvANY(sva))) {
413 register const SV * const svend = &sva[SvREFCNT(sva)];
415 for (sv = sva + 1; sv < svend; ++sv) {
416 if (SvTYPE(sv) != (svtype)SVTYPEMASK
417 && (sv->sv_flags & mask) == flags
430 /* called by sv_report_used() for each live SV */
433 do_report_used(pTHX_ SV *const sv)
435 if (SvTYPE(sv) != (svtype)SVTYPEMASK) {
436 PerlIO_printf(Perl_debug_log, "****\n");
443 =for apidoc sv_report_used
445 Dump the contents of all SVs not yet freed. (Debugging aid).
451 Perl_sv_report_used(pTHX)
454 visit(do_report_used, 0, 0);
460 /* called by sv_clean_objs() for each live SV */
463 do_clean_objs(pTHX_ SV *const ref)
468 SV * const target = SvRV(ref);
469 if (SvOBJECT(target)) {
470 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning object ref:\n "), sv_dump(ref)));
471 if (SvWEAKREF(ref)) {
472 sv_del_backref(target, ref);
478 SvREFCNT_dec(target);
483 /* XXX Might want to check arrays, etc. */
487 /* clear any slots in a GV which hold objects - except IO;
488 * called by sv_clean_objs() for each live GV */
491 do_clean_named_objs(pTHX_ SV *const sv)
495 assert(SvTYPE(sv) == SVt_PVGV);
496 assert(isGV_with_GP(sv));
500 /* freeing GP entries may indirectly free the current GV;
501 * hold onto it while we mess with the GP slots */
504 if ( ((obj = GvSV(sv) )) && SvOBJECT(obj)) {
505 DEBUG_D((PerlIO_printf(Perl_debug_log,
506 "Cleaning named glob SV object:\n "), sv_dump(obj)));
510 if ( ((obj = MUTABLE_SV(GvAV(sv)) )) && SvOBJECT(obj)) {
511 DEBUG_D((PerlIO_printf(Perl_debug_log,
512 "Cleaning named glob AV object:\n "), sv_dump(obj)));
516 if ( ((obj = MUTABLE_SV(GvHV(sv)) )) && SvOBJECT(obj)) {
517 DEBUG_D((PerlIO_printf(Perl_debug_log,
518 "Cleaning named glob HV object:\n "), sv_dump(obj)));
522 if ( ((obj = MUTABLE_SV(GvCV(sv)) )) && SvOBJECT(obj)) {
523 DEBUG_D((PerlIO_printf(Perl_debug_log,
524 "Cleaning named glob CV object:\n "), sv_dump(obj)));
528 SvREFCNT_dec(sv); /* undo the inc above */
531 /* clear any IO slots in a GV which hold objects (except stderr, defout);
532 * called by sv_clean_objs() for each live GV */
535 do_clean_named_io_objs(pTHX_ SV *const sv)
539 assert(SvTYPE(sv) == SVt_PVGV);
540 assert(isGV_with_GP(sv));
541 if (!GvGP(sv) || sv == (SV*)PL_stderrgv || sv == (SV*)PL_defoutgv)
545 if ( ((obj = MUTABLE_SV(GvIO(sv)) )) && SvOBJECT(obj)) {
546 DEBUG_D((PerlIO_printf(Perl_debug_log,
547 "Cleaning named glob IO object:\n "), sv_dump(obj)));
551 SvREFCNT_dec(sv); /* undo the inc above */
554 /* Void wrapper to pass to visit() */
556 do_curse(pTHX_ SV * const sv) {
557 if ((PL_stderrgv && GvGP(PL_stderrgv) && (SV*)GvIO(PL_stderrgv) == sv)
558 || (PL_defoutgv && GvGP(PL_defoutgv) && (SV*)GvIO(PL_defoutgv) == sv))
564 =for apidoc sv_clean_objs
566 Attempt to destroy all objects not yet freed
572 Perl_sv_clean_objs(pTHX)
576 PL_in_clean_objs = TRUE;
577 visit(do_clean_objs, SVf_ROK, SVf_ROK);
578 /* Some barnacles may yet remain, clinging to typeglobs.
579 * Run the non-IO destructors first: they may want to output
580 * error messages, close files etc */
581 visit(do_clean_named_objs, SVt_PVGV|SVpgv_GP, SVTYPEMASK|SVp_POK|SVpgv_GP);
582 visit(do_clean_named_io_objs, SVt_PVGV|SVpgv_GP, SVTYPEMASK|SVp_POK|SVpgv_GP);
583 /* And if there are some very tenacious barnacles clinging to arrays,
584 closures, or what have you.... */
585 visit(do_curse, SVs_OBJECT, SVs_OBJECT);
586 olddef = PL_defoutgv;
587 PL_defoutgv = NULL; /* disable skip of PL_defoutgv */
588 if (olddef && isGV_with_GP(olddef))
589 do_clean_named_io_objs(aTHX_ MUTABLE_SV(olddef));
590 olderr = PL_stderrgv;
591 PL_stderrgv = NULL; /* disable skip of PL_stderrgv */
592 if (olderr && isGV_with_GP(olderr))
593 do_clean_named_io_objs(aTHX_ MUTABLE_SV(olderr));
594 SvREFCNT_dec(olddef);
595 PL_in_clean_objs = FALSE;
598 /* called by sv_clean_all() for each live SV */
601 do_clean_all(pTHX_ SV *const sv)
604 if (sv == (const SV *) PL_fdpid || sv == (const SV *)PL_strtab) {
605 /* don't clean pid table and strtab */
608 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning loops: SV at 0x%"UVxf"\n", PTR2UV(sv)) ));
609 SvFLAGS(sv) |= SVf_BREAK;
614 =for apidoc sv_clean_all
616 Decrement the refcnt of each remaining SV, possibly triggering a
617 cleanup. This function may have to be called multiple times to free
618 SVs which are in complex self-referential hierarchies.
624 Perl_sv_clean_all(pTHX)
628 PL_in_clean_all = TRUE;
629 cleaned = visit(do_clean_all, 0,0);
634 ARENASETS: a meta-arena implementation which separates arena-info
635 into struct arena_set, which contains an array of struct
636 arena_descs, each holding info for a single arena. By separating
637 the meta-info from the arena, we recover the 1st slot, formerly
638 borrowed for list management. The arena_set is about the size of an
639 arena, avoiding the needless malloc overhead of a naive linked-list.
641 The cost is 1 arena-set malloc per ~320 arena-mallocs, + the unused
642 memory in the last arena-set (1/2 on average). In trade, we get
643 back the 1st slot in each arena (ie 1.7% of a CV-arena, less for
644 smaller types). The recovery of the wasted space allows use of
645 small arenas for large, rare body types, by changing array* fields
646 in body_details_by_type[] below.
649 char *arena; /* the raw storage, allocated aligned */
650 size_t size; /* its size ~4k typ */
651 svtype utype; /* bodytype stored in arena */
656 /* Get the maximum number of elements in set[] such that struct arena_set
657 will fit within PERL_ARENA_SIZE, which is probably just under 4K, and
658 therefore likely to be 1 aligned memory page. */
660 #define ARENAS_PER_SET ((PERL_ARENA_SIZE - sizeof(struct arena_set*) \
661 - 2 * sizeof(int)) / sizeof (struct arena_desc))
664 struct arena_set* next;
665 unsigned int set_size; /* ie ARENAS_PER_SET */
666 unsigned int curr; /* index of next available arena-desc */
667 struct arena_desc set[ARENAS_PER_SET];
671 =for apidoc sv_free_arenas
673 Deallocate the memory used by all arenas. Note that all the individual SV
674 heads and bodies within the arenas must already have been freed.
679 Perl_sv_free_arenas(pTHX)
686 /* Free arenas here, but be careful about fake ones. (We assume
687 contiguity of the fake ones with the corresponding real ones.) */
689 for (sva = PL_sv_arenaroot; sva; sva = svanext) {
690 svanext = MUTABLE_SV(SvANY(sva));
691 while (svanext && SvFAKE(svanext))
692 svanext = MUTABLE_SV(SvANY(svanext));
699 struct arena_set *aroot = (struct arena_set*) PL_body_arenas;
702 struct arena_set *current = aroot;
705 assert(aroot->set[i].arena);
706 Safefree(aroot->set[i].arena);
714 i = PERL_ARENA_ROOTS_SIZE;
716 PL_body_roots[i] = 0;
723 Here are mid-level routines that manage the allocation of bodies out
724 of the various arenas. There are 5 kinds of arenas:
726 1. SV-head arenas, which are discussed and handled above
727 2. regular body arenas
728 3. arenas for reduced-size bodies
731 Arena types 2 & 3 are chained by body-type off an array of
732 arena-root pointers, which is indexed by svtype. Some of the
733 larger/less used body types are malloced singly, since a large
734 unused block of them is wasteful. Also, several svtypes dont have
735 bodies; the data fits into the sv-head itself. The arena-root
736 pointer thus has a few unused root-pointers (which may be hijacked
737 later for arena types 4,5)
739 3 differs from 2 as an optimization; some body types have several
740 unused fields in the front of the structure (which are kept in-place
741 for consistency). These bodies can be allocated in smaller chunks,
742 because the leading fields arent accessed. Pointers to such bodies
743 are decremented to point at the unused 'ghost' memory, knowing that
744 the pointers are used with offsets to the real memory.
747 =head1 SV-Body Allocation
749 Allocation of SV-bodies is similar to SV-heads, differing as follows;
750 the allocation mechanism is used for many body types, so is somewhat
751 more complicated, it uses arena-sets, and has no need for still-live
754 At the outermost level, (new|del)_X*V macros return bodies of the
755 appropriate type. These macros call either (new|del)_body_type or
756 (new|del)_body_allocated macro pairs, depending on specifics of the
757 type. Most body types use the former pair, the latter pair is used to
758 allocate body types with "ghost fields".
760 "ghost fields" are fields that are unused in certain types, and
761 consequently don't need to actually exist. They are declared because
762 they're part of a "base type", which allows use of functions as
763 methods. The simplest examples are AVs and HVs, 2 aggregate types
764 which don't use the fields which support SCALAR semantics.
766 For these types, the arenas are carved up into appropriately sized
767 chunks, we thus avoid wasted memory for those unaccessed members.
768 When bodies are allocated, we adjust the pointer back in memory by the
769 size of the part not allocated, so it's as if we allocated the full
770 structure. (But things will all go boom if you write to the part that
771 is "not there", because you'll be overwriting the last members of the
772 preceding structure in memory.)
774 We calculate the correction using the STRUCT_OFFSET macro on the first
775 member present. If the allocated structure is smaller (no initial NV
776 actually allocated) then the net effect is to subtract the size of the NV
777 from the pointer, to return a new pointer as if an initial NV were actually
778 allocated. (We were using structures named *_allocated for this, but
779 this turned out to be a subtle bug, because a structure without an NV
780 could have a lower alignment constraint, but the compiler is allowed to
781 optimised accesses based on the alignment constraint of the actual pointer
782 to the full structure, for example, using a single 64 bit load instruction
783 because it "knows" that two adjacent 32 bit members will be 8-byte aligned.)
785 This is the same trick as was used for NV and IV bodies. Ironically it
786 doesn't need to be used for NV bodies any more, because NV is now at
787 the start of the structure. IV bodies don't need it either, because
788 they are no longer allocated.
790 In turn, the new_body_* allocators call S_new_body(), which invokes
791 new_body_inline macro, which takes a lock, and takes a body off the
792 linked list at PL_body_roots[sv_type], calling Perl_more_bodies() if
793 necessary to refresh an empty list. Then the lock is released, and
794 the body is returned.
796 Perl_more_bodies allocates a new arena, and carves it up into an array of N
797 bodies, which it strings into a linked list. It looks up arena-size
798 and body-size from the body_details table described below, thus
799 supporting the multiple body-types.
801 If PURIFY is defined, or PERL_ARENA_SIZE=0, arenas are not used, and
802 the (new|del)_X*V macros are mapped directly to malloc/free.
804 For each sv-type, struct body_details bodies_by_type[] carries
805 parameters which control these aspects of SV handling:
807 Arena_size determines whether arenas are used for this body type, and if
808 so, how big they are. PURIFY or PERL_ARENA_SIZE=0 set this field to
809 zero, forcing individual mallocs and frees.
811 Body_size determines how big a body is, and therefore how many fit into
812 each arena. Offset carries the body-pointer adjustment needed for
813 "ghost fields", and is used in *_allocated macros.
815 But its main purpose is to parameterize info needed in
816 Perl_sv_upgrade(). The info here dramatically simplifies the function
817 vs the implementation in 5.8.8, making it table-driven. All fields
818 are used for this, except for arena_size.
820 For the sv-types that have no bodies, arenas are not used, so those
821 PL_body_roots[sv_type] are unused, and can be overloaded. In
822 something of a special case, SVt_NULL is borrowed for HE arenas;
823 PL_body_roots[HE_SVSLOT=SVt_NULL] is filled by S_more_he, but the
824 bodies_by_type[SVt_NULL] slot is not used, as the table is not
829 struct body_details {
830 U8 body_size; /* Size to allocate */
831 U8 copy; /* Size of structure to copy (may be shorter) */
833 unsigned int type : 4; /* We have space for a sanity check. */
834 unsigned int cant_upgrade : 1; /* Cannot upgrade this type */
835 unsigned int zero_nv : 1; /* zero the NV when upgrading from this */
836 unsigned int arena : 1; /* Allocated from an arena */
837 size_t arena_size; /* Size of arena to allocate */
845 /* With -DPURFIY we allocate everything directly, and don't use arenas.
846 This seems a rather elegant way to simplify some of the code below. */
847 #define HASARENA FALSE
849 #define HASARENA TRUE
851 #define NOARENA FALSE
853 /* Size the arenas to exactly fit a given number of bodies. A count
854 of 0 fits the max number bodies into a PERL_ARENA_SIZE.block,
855 simplifying the default. If count > 0, the arena is sized to fit
856 only that many bodies, allowing arenas to be used for large, rare
857 bodies (XPVFM, XPVIO) without undue waste. The arena size is
858 limited by PERL_ARENA_SIZE, so we can safely oversize the
861 #define FIT_ARENA0(body_size) \
862 ((size_t)(PERL_ARENA_SIZE / body_size) * body_size)
863 #define FIT_ARENAn(count,body_size) \
864 ( count * body_size <= PERL_ARENA_SIZE) \
865 ? count * body_size \
866 : FIT_ARENA0 (body_size)
867 #define FIT_ARENA(count,body_size) \
869 ? FIT_ARENAn (count, body_size) \
870 : FIT_ARENA0 (body_size)
872 /* Calculate the length to copy. Specifically work out the length less any
873 final padding the compiler needed to add. See the comment in sv_upgrade
874 for why copying the padding proved to be a bug. */
876 #define copy_length(type, last_member) \
877 STRUCT_OFFSET(type, last_member) \
878 + sizeof (((type*)SvANY((const SV *)0))->last_member)
880 static const struct body_details bodies_by_type[] = {
881 /* HEs use this offset for their arena. */
882 { 0, 0, 0, SVt_NULL, FALSE, NONV, NOARENA, 0 },
884 /* The bind placeholder pretends to be an RV for now.
885 Also it's marked as "can't upgrade" to stop anyone using it before it's
887 { 0, 0, 0, SVt_BIND, TRUE, NONV, NOARENA, 0 },
889 /* IVs are in the head, so the allocation size is 0. */
891 sizeof(IV), /* This is used to copy out the IV body. */
892 STRUCT_OFFSET(XPVIV, xiv_iv), SVt_IV, FALSE, NONV,
893 NOARENA /* IVS don't need an arena */, 0
896 { sizeof(NV), sizeof(NV),
897 STRUCT_OFFSET(XPVNV, xnv_u),
898 SVt_NV, FALSE, HADNV, HASARENA, FIT_ARENA(0, sizeof(NV)) },
900 { sizeof(XPV) - STRUCT_OFFSET(XPV, xpv_cur),
901 copy_length(XPV, xpv_len) - STRUCT_OFFSET(XPV, xpv_cur),
902 + STRUCT_OFFSET(XPV, xpv_cur),
903 SVt_PV, FALSE, NONV, HASARENA,
904 FIT_ARENA(0, sizeof(XPV) - STRUCT_OFFSET(XPV, xpv_cur)) },
906 { sizeof(XPVIV) - STRUCT_OFFSET(XPV, xpv_cur),
907 copy_length(XPVIV, xiv_u) - STRUCT_OFFSET(XPV, xpv_cur),
908 + STRUCT_OFFSET(XPV, xpv_cur),
909 SVt_PVIV, FALSE, NONV, HASARENA,
910 FIT_ARENA(0, sizeof(XPVIV) - STRUCT_OFFSET(XPV, xpv_cur)) },
912 { sizeof(XPVNV) - STRUCT_OFFSET(XPV, xpv_cur),
913 copy_length(XPVNV, xnv_u) - STRUCT_OFFSET(XPV, xpv_cur),
914 + STRUCT_OFFSET(XPV, xpv_cur),
915 SVt_PVNV, FALSE, HADNV, HASARENA,
916 FIT_ARENA(0, sizeof(XPVNV) - STRUCT_OFFSET(XPV, xpv_cur)) },
918 { sizeof(XPVMG), copy_length(XPVMG, xnv_u), 0, SVt_PVMG, FALSE, HADNV,
919 HASARENA, FIT_ARENA(0, sizeof(XPVMG)) },
924 SVt_REGEXP, FALSE, NONV, HASARENA,
925 FIT_ARENA(0, sizeof(regexp))
928 { sizeof(XPVGV), sizeof(XPVGV), 0, SVt_PVGV, TRUE, HADNV,
929 HASARENA, FIT_ARENA(0, sizeof(XPVGV)) },
931 { sizeof(XPVLV), sizeof(XPVLV), 0, SVt_PVLV, TRUE, HADNV,
932 HASARENA, FIT_ARENA(0, sizeof(XPVLV)) },
935 copy_length(XPVAV, xav_alloc),
937 SVt_PVAV, TRUE, NONV, HASARENA,
938 FIT_ARENA(0, sizeof(XPVAV)) },
941 copy_length(XPVHV, xhv_max),
943 SVt_PVHV, TRUE, NONV, HASARENA,
944 FIT_ARENA(0, sizeof(XPVHV)) },
949 SVt_PVCV, TRUE, NONV, HASARENA,
950 FIT_ARENA(0, sizeof(XPVCV)) },
955 SVt_PVFM, TRUE, NONV, NOARENA,
956 FIT_ARENA(20, sizeof(XPVFM)) },
961 SVt_PVIO, TRUE, NONV, HASARENA,
962 FIT_ARENA(24, sizeof(XPVIO)) },
965 #define new_body_allocated(sv_type) \
966 (void *)((char *)S_new_body(aTHX_ sv_type) \
967 - bodies_by_type[sv_type].offset)
969 /* return a thing to the free list */
971 #define del_body(thing, root) \
973 void ** const thing_copy = (void **)thing; \
974 *thing_copy = *root; \
975 *root = (void*)thing_copy; \
980 #define new_XNV() safemalloc(sizeof(XPVNV))
981 #define new_XPVNV() safemalloc(sizeof(XPVNV))
982 #define new_XPVMG() safemalloc(sizeof(XPVMG))
984 #define del_XPVGV(p) safefree(p)
988 #define new_XNV() new_body_allocated(SVt_NV)
989 #define new_XPVNV() new_body_allocated(SVt_PVNV)
990 #define new_XPVMG() new_body_allocated(SVt_PVMG)
992 #define del_XPVGV(p) del_body(p + bodies_by_type[SVt_PVGV].offset, \
993 &PL_body_roots[SVt_PVGV])
997 /* no arena for you! */
999 #define new_NOARENA(details) \
1000 safemalloc((details)->body_size + (details)->offset)
1001 #define new_NOARENAZ(details) \
1002 safecalloc((details)->body_size + (details)->offset, 1)
1005 Perl_more_bodies (pTHX_ const svtype sv_type, const size_t body_size,
1006 const size_t arena_size)
1009 void ** const root = &PL_body_roots[sv_type];
1010 struct arena_desc *adesc;
1011 struct arena_set *aroot = (struct arena_set *) PL_body_arenas;
1015 const size_t good_arena_size = Perl_malloc_good_size(arena_size);
1016 #if defined(DEBUGGING) && !defined(PERL_GLOBAL_STRUCT_PRIVATE)
1017 static bool done_sanity_check;
1019 /* PERL_GLOBAL_STRUCT_PRIVATE cannot coexist with global
1020 * variables like done_sanity_check. */
1021 if (!done_sanity_check) {
1022 unsigned int i = SVt_LAST;
1024 done_sanity_check = TRUE;
1027 assert (bodies_by_type[i].type == i);
1033 /* may need new arena-set to hold new arena */
1034 if (!aroot || aroot->curr >= aroot->set_size) {
1035 struct arena_set *newroot;
1036 Newxz(newroot, 1, struct arena_set);
1037 newroot->set_size = ARENAS_PER_SET;
1038 newroot->next = aroot;
1040 PL_body_arenas = (void *) newroot;
1041 DEBUG_m(PerlIO_printf(Perl_debug_log, "new arenaset %p\n", (void*)aroot));
1044 /* ok, now have arena-set with at least 1 empty/available arena-desc */
1045 curr = aroot->curr++;
1046 adesc = &(aroot->set[curr]);
1047 assert(!adesc->arena);
1049 Newx(adesc->arena, good_arena_size, char);
1050 adesc->size = good_arena_size;
1051 adesc->utype = sv_type;
1052 DEBUG_m(PerlIO_printf(Perl_debug_log, "arena %d added: %p size %"UVuf"\n",
1053 curr, (void*)adesc->arena, (UV)good_arena_size));
1055 start = (char *) adesc->arena;
1057 /* Get the address of the byte after the end of the last body we can fit.
1058 Remember, this is integer division: */
1059 end = start + good_arena_size / body_size * body_size;
1061 /* computed count doesn't reflect the 1st slot reservation */
1062 #if defined(MYMALLOC) || defined(HAS_MALLOC_GOOD_SIZE)
1063 DEBUG_m(PerlIO_printf(Perl_debug_log,
1064 "arena %p end %p arena-size %d (from %d) type %d "
1066 (void*)start, (void*)end, (int)good_arena_size,
1067 (int)arena_size, sv_type, (int)body_size,
1068 (int)good_arena_size / (int)body_size));
1070 DEBUG_m(PerlIO_printf(Perl_debug_log,
1071 "arena %p end %p arena-size %d type %d size %d ct %d\n",
1072 (void*)start, (void*)end,
1073 (int)arena_size, sv_type, (int)body_size,
1074 (int)good_arena_size / (int)body_size));
1076 *root = (void *)start;
1079 /* Where the next body would start: */
1080 char * const next = start + body_size;
1083 /* This is the last body: */
1084 assert(next == end);
1086 *(void **)start = 0;
1090 *(void**) start = (void *)next;
1095 /* grab a new thing from the free list, allocating more if necessary.
1096 The inline version is used for speed in hot routines, and the
1097 function using it serves the rest (unless PURIFY).
1099 #define new_body_inline(xpv, sv_type) \
1101 void ** const r3wt = &PL_body_roots[sv_type]; \
1102 xpv = (PTR_TBL_ENT_t*) (*((void **)(r3wt)) \
1103 ? *((void **)(r3wt)) : Perl_more_bodies(aTHX_ sv_type, \
1104 bodies_by_type[sv_type].body_size,\
1105 bodies_by_type[sv_type].arena_size)); \
1106 *(r3wt) = *(void**)(xpv); \
1112 S_new_body(pTHX_ const svtype sv_type)
1116 new_body_inline(xpv, sv_type);
1122 static const struct body_details fake_rv =
1123 { 0, 0, 0, SVt_IV, FALSE, NONV, NOARENA, 0 };
1126 =for apidoc sv_upgrade
1128 Upgrade an SV to a more complex form. Generally adds a new body type to the
1129 SV, then copies across as much information as possible from the old body.
1130 You generally want to use the C<SvUPGRADE> macro wrapper. See also C<svtype>.
1136 Perl_sv_upgrade(pTHX_ register SV *const sv, svtype new_type)
1141 const svtype old_type = SvTYPE(sv);
1142 const struct body_details *new_type_details;
1143 const struct body_details *old_type_details
1144 = bodies_by_type + old_type;
1145 SV *referant = NULL;
1147 PERL_ARGS_ASSERT_SV_UPGRADE;
1149 if (old_type == new_type)
1152 /* This clause was purposefully added ahead of the early return above to
1153 the shared string hackery for (sort {$a <=> $b} keys %hash), with the
1154 inference by Nick I-S that it would fix other troublesome cases. See
1155 changes 7162, 7163 (f130fd4589cf5fbb24149cd4db4137c8326f49c1 and parent)
1157 Given that shared hash key scalars are no longer PVIV, but PV, there is
1158 no longer need to unshare so as to free up the IVX slot for its proper
1159 purpose. So it's safe to move the early return earlier. */
1161 if (new_type != SVt_PV && SvIsCOW(sv)) {
1162 sv_force_normal_flags(sv, 0);
1165 old_body = SvANY(sv);
1167 /* Copying structures onto other structures that have been neatly zeroed
1168 has a subtle gotcha. Consider XPVMG
1170 +------+------+------+------+------+-------+-------+
1171 | NV | CUR | LEN | IV | MAGIC | STASH |
1172 +------+------+------+------+------+-------+-------+
1173 0 4 8 12 16 20 24 28
1175 where NVs are aligned to 8 bytes, so that sizeof that structure is
1176 actually 32 bytes long, with 4 bytes of padding at the end:
1178 +------+------+------+------+------+-------+-------+------+
1179 | NV | CUR | LEN | IV | MAGIC | STASH | ??? |
1180 +------+------+------+------+------+-------+-------+------+
1181 0 4 8 12 16 20 24 28 32
1183 so what happens if you allocate memory for this structure:
1185 +------+------+------+------+------+-------+-------+------+------+...
1186 | NV | CUR | LEN | IV | MAGIC | STASH | GP | NAME |
1187 +------+------+------+------+------+-------+-------+------+------+...
1188 0 4 8 12 16 20 24 28 32 36
1190 zero it, then copy sizeof(XPVMG) bytes on top of it? Not quite what you
1191 expect, because you copy the area marked ??? onto GP. Now, ??? may have
1192 started out as zero once, but it's quite possible that it isn't. So now,
1193 rather than a nicely zeroed GP, you have it pointing somewhere random.
1196 (In fact, GP ends up pointing at a previous GP structure, because the
1197 principle cause of the padding in XPVMG getting garbage is a copy of
1198 sizeof(XPVMG) bytes from a XPVGV structure in sv_unglob. Right now
1199 this happens to be moot because XPVGV has been re-ordered, with GP
1200 no longer after STASH)
1202 So we are careful and work out the size of used parts of all the
1210 referant = SvRV(sv);
1211 old_type_details = &fake_rv;
1212 if (new_type == SVt_NV)
1213 new_type = SVt_PVNV;
1215 if (new_type < SVt_PVIV) {
1216 new_type = (new_type == SVt_NV)
1217 ? SVt_PVNV : SVt_PVIV;
1222 if (new_type < SVt_PVNV) {
1223 new_type = SVt_PVNV;
1227 assert(new_type > SVt_PV);
1228 assert(SVt_IV < SVt_PV);
1229 assert(SVt_NV < SVt_PV);
1236 /* Because the XPVMG of PL_mess_sv isn't allocated from the arena,
1237 there's no way that it can be safely upgraded, because perl.c
1238 expects to Safefree(SvANY(PL_mess_sv)) */
1239 assert(sv != PL_mess_sv);
1240 /* This flag bit is used to mean other things in other scalar types.
1241 Given that it only has meaning inside the pad, it shouldn't be set
1242 on anything that can get upgraded. */
1243 assert(!SvPAD_TYPED(sv));
1246 if (old_type_details->cant_upgrade)
1247 Perl_croak(aTHX_ "Can't upgrade %s (%" UVuf ") to %" UVuf,
1248 sv_reftype(sv, 0), (UV) old_type, (UV) new_type);
1251 if (old_type > new_type)
1252 Perl_croak(aTHX_ "sv_upgrade from type %d down to type %d",
1253 (int)old_type, (int)new_type);
1255 new_type_details = bodies_by_type + new_type;
1257 SvFLAGS(sv) &= ~SVTYPEMASK;
1258 SvFLAGS(sv) |= new_type;
1260 /* This can't happen, as SVt_NULL is <= all values of new_type, so one of
1261 the return statements above will have triggered. */
1262 assert (new_type != SVt_NULL);
1265 assert(old_type == SVt_NULL);
1266 SvANY(sv) = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
1270 assert(old_type == SVt_NULL);
1271 SvANY(sv) = new_XNV();
1276 assert(new_type_details->body_size);
1279 assert(new_type_details->arena);
1280 assert(new_type_details->arena_size);
1281 /* This points to the start of the allocated area. */
1282 new_body_inline(new_body, new_type);
1283 Zero(new_body, new_type_details->body_size, char);
1284 new_body = ((char *)new_body) - new_type_details->offset;
1286 /* We always allocated the full length item with PURIFY. To do this
1287 we fake things so that arena is false for all 16 types.. */
1288 new_body = new_NOARENAZ(new_type_details);
1290 SvANY(sv) = new_body;
1291 if (new_type == SVt_PVAV) {
1295 if (old_type_details->body_size) {
1298 /* It will have been zeroed when the new body was allocated.
1299 Lets not write to it, in case it confuses a write-back
1305 #ifndef NODEFAULT_SHAREKEYS
1306 HvSHAREKEYS_on(sv); /* key-sharing on by default */
1308 HvMAX(sv) = 7; /* (start with 8 buckets) */
1311 /* SVt_NULL isn't the only thing upgraded to AV or HV.
1312 The target created by newSVrv also is, and it can have magic.
1313 However, it never has SvPVX set.
1315 if (old_type == SVt_IV) {
1317 } else if (old_type >= SVt_PV) {
1318 assert(SvPVX_const(sv) == 0);
1321 if (old_type >= SVt_PVMG) {
1322 SvMAGIC_set(sv, ((XPVMG*)old_body)->xmg_u.xmg_magic);
1323 SvSTASH_set(sv, ((XPVMG*)old_body)->xmg_stash);
1325 sv->sv_u.svu_array = NULL; /* or svu_hash */
1331 /* This ensures that SvTHINKFIRST(sv) is true, and hence that
1332 sv_force_normal_flags(sv) is called. */
1335 /* XXX Is this still needed? Was it ever needed? Surely as there is
1336 no route from NV to PVIV, NOK can never be true */
1337 assert(!SvNOKp(sv));
1348 assert(new_type_details->body_size);
1349 /* We always allocated the full length item with PURIFY. To do this
1350 we fake things so that arena is false for all 16 types.. */
1351 if(new_type_details->arena) {
1352 /* This points to the start of the allocated area. */
1353 new_body_inline(new_body, new_type);
1354 Zero(new_body, new_type_details->body_size, char);
1355 new_body = ((char *)new_body) - new_type_details->offset;
1357 new_body = new_NOARENAZ(new_type_details);
1359 SvANY(sv) = new_body;
1361 if (old_type_details->copy) {
1362 /* There is now the potential for an upgrade from something without
1363 an offset (PVNV or PVMG) to something with one (PVCV, PVFM) */
1364 int offset = old_type_details->offset;
1365 int length = old_type_details->copy;
1367 if (new_type_details->offset > old_type_details->offset) {
1368 const int difference
1369 = new_type_details->offset - old_type_details->offset;
1370 offset += difference;
1371 length -= difference;
1373 assert (length >= 0);
1375 Copy((char *)old_body + offset, (char *)new_body + offset, length,
1379 #ifndef NV_ZERO_IS_ALLBITS_ZERO
1380 /* If NV 0.0 is stores as all bits 0 then Zero() already creates a
1381 * correct 0.0 for us. Otherwise, if the old body didn't have an
1382 * NV slot, but the new one does, then we need to initialise the
1383 * freshly created NV slot with whatever the correct bit pattern is
1385 if (old_type_details->zero_nv && !new_type_details->zero_nv
1386 && !isGV_with_GP(sv))
1390 if (new_type == SVt_PVIO) {
1391 IO * const io = MUTABLE_IO(sv);
1392 GV *iogv = gv_fetchpvs("IO::File::", GV_ADD, SVt_PVHV);
1395 /* Clear the stashcache because a new IO could overrule a package
1397 hv_clear(PL_stashcache);
1399 SvSTASH_set(io, MUTABLE_HV(SvREFCNT_inc(GvHV(iogv))));
1400 IoPAGE_LEN(sv) = 60;
1402 if (old_type < SVt_PV) {
1403 /* referant will be NULL unless the old type was SVt_IV emulating
1405 sv->sv_u.svu_rv = referant;
1409 Perl_croak(aTHX_ "panic: sv_upgrade to unknown type %lu",
1410 (unsigned long)new_type);
1413 if (old_type > SVt_IV) {
1417 /* Note that there is an assumption that all bodies of types that
1418 can be upgraded came from arenas. Only the more complex non-
1419 upgradable types are allowed to be directly malloc()ed. */
1420 assert(old_type_details->arena);
1421 del_body((void*)((char*)old_body + old_type_details->offset),
1422 &PL_body_roots[old_type]);
1428 =for apidoc sv_backoff
1430 Remove any string offset. You should normally use the C<SvOOK_off> macro
1437 Perl_sv_backoff(pTHX_ register SV *const sv)
1440 const char * const s = SvPVX_const(sv);
1442 PERL_ARGS_ASSERT_SV_BACKOFF;
1443 PERL_UNUSED_CONTEXT;
1446 assert(SvTYPE(sv) != SVt_PVHV);
1447 assert(SvTYPE(sv) != SVt_PVAV);
1449 SvOOK_offset(sv, delta);
1451 SvLEN_set(sv, SvLEN(sv) + delta);
1452 SvPV_set(sv, SvPVX(sv) - delta);
1453 Move(s, SvPVX(sv), SvCUR(sv)+1, char);
1454 SvFLAGS(sv) &= ~SVf_OOK;
1461 Expands the character buffer in the SV. If necessary, uses C<sv_unref> and
1462 upgrades the SV to C<SVt_PV>. Returns a pointer to the character buffer.
1463 Use the C<SvGROW> wrapper instead.
1469 Perl_sv_grow(pTHX_ register SV *const sv, register STRLEN newlen)
1473 PERL_ARGS_ASSERT_SV_GROW;
1475 if (PL_madskills && newlen >= 0x100000) {
1476 PerlIO_printf(Perl_debug_log,
1477 "Allocation too large: %"UVxf"\n", (UV)newlen);
1479 #ifdef HAS_64K_LIMIT
1480 if (newlen >= 0x10000) {
1481 PerlIO_printf(Perl_debug_log,
1482 "Allocation too large: %"UVxf"\n", (UV)newlen);
1485 #endif /* HAS_64K_LIMIT */
1488 if (SvTYPE(sv) < SVt_PV) {
1489 sv_upgrade(sv, SVt_PV);
1490 s = SvPVX_mutable(sv);
1492 else if (SvOOK(sv)) { /* pv is offset? */
1494 s = SvPVX_mutable(sv);
1495 if (newlen > SvLEN(sv))
1496 newlen += 10 * (newlen - SvCUR(sv)); /* avoid copy each time */
1497 #ifdef HAS_64K_LIMIT
1498 if (newlen >= 0x10000)
1503 s = SvPVX_mutable(sv);
1505 if (newlen > SvLEN(sv)) { /* need more room? */
1506 STRLEN minlen = SvCUR(sv);
1507 minlen += (minlen >> PERL_STRLEN_EXPAND_SHIFT) + 10;
1508 if (newlen < minlen)
1510 #ifndef Perl_safesysmalloc_size
1511 newlen = PERL_STRLEN_ROUNDUP(newlen);
1513 if (SvLEN(sv) && s) {
1514 s = (char*)saferealloc(s, newlen);
1517 s = (char*)safemalloc(newlen);
1518 if (SvPVX_const(sv) && SvCUR(sv)) {
1519 Move(SvPVX_const(sv), s, (newlen < SvCUR(sv)) ? newlen : SvCUR(sv), char);
1523 #ifdef Perl_safesysmalloc_size
1524 /* Do this here, do it once, do it right, and then we will never get
1525 called back into sv_grow() unless there really is some growing
1527 SvLEN_set(sv, Perl_safesysmalloc_size(s));
1529 SvLEN_set(sv, newlen);
1536 =for apidoc sv_setiv
1538 Copies an integer into the given SV, upgrading first if necessary.
1539 Does not handle 'set' magic. See also C<sv_setiv_mg>.
1545 Perl_sv_setiv(pTHX_ register SV *const sv, const IV i)
1549 PERL_ARGS_ASSERT_SV_SETIV;
1551 SV_CHECK_THINKFIRST_COW_DROP(sv);
1552 switch (SvTYPE(sv)) {
1555 sv_upgrade(sv, SVt_IV);
1558 sv_upgrade(sv, SVt_PVIV);
1562 if (!isGV_with_GP(sv))
1569 /* diag_listed_as: Can't coerce %s to %s in %s */
1570 Perl_croak(aTHX_ "Can't coerce %s to integer in %s", sv_reftype(sv,0),
1574 (void)SvIOK_only(sv); /* validate number */
1580 =for apidoc sv_setiv_mg
1582 Like C<sv_setiv>, but also handles 'set' magic.
1588 Perl_sv_setiv_mg(pTHX_ register SV *const sv, const IV i)
1590 PERL_ARGS_ASSERT_SV_SETIV_MG;
1597 =for apidoc sv_setuv
1599 Copies an unsigned integer into the given SV, upgrading first if necessary.
1600 Does not handle 'set' magic. See also C<sv_setuv_mg>.
1606 Perl_sv_setuv(pTHX_ register SV *const sv, const UV u)
1608 PERL_ARGS_ASSERT_SV_SETUV;
1610 /* With these two if statements:
1611 u=1.49 s=0.52 cu=72.49 cs=10.64 scripts=270 tests=20865
1614 u=1.35 s=0.47 cu=73.45 cs=11.43 scripts=270 tests=20865
1616 If you wish to remove them, please benchmark to see what the effect is
1618 if (u <= (UV)IV_MAX) {
1619 sv_setiv(sv, (IV)u);
1628 =for apidoc sv_setuv_mg
1630 Like C<sv_setuv>, but also handles 'set' magic.
1636 Perl_sv_setuv_mg(pTHX_ register SV *const sv, const UV u)
1638 PERL_ARGS_ASSERT_SV_SETUV_MG;
1645 =for apidoc sv_setnv
1647 Copies a double into the given SV, upgrading first if necessary.
1648 Does not handle 'set' magic. See also C<sv_setnv_mg>.
1654 Perl_sv_setnv(pTHX_ register SV *const sv, const NV num)
1658 PERL_ARGS_ASSERT_SV_SETNV;
1660 SV_CHECK_THINKFIRST_COW_DROP(sv);
1661 switch (SvTYPE(sv)) {
1664 sv_upgrade(sv, SVt_NV);
1668 sv_upgrade(sv, SVt_PVNV);
1672 if (!isGV_with_GP(sv))
1679 /* diag_listed_as: Can't coerce %s to %s in %s */
1680 Perl_croak(aTHX_ "Can't coerce %s to number in %s", sv_reftype(sv,0),
1685 (void)SvNOK_only(sv); /* validate number */
1690 =for apidoc sv_setnv_mg
1692 Like C<sv_setnv>, but also handles 'set' magic.
1698 Perl_sv_setnv_mg(pTHX_ register SV *const sv, const NV num)
1700 PERL_ARGS_ASSERT_SV_SETNV_MG;
1706 /* Print an "isn't numeric" warning, using a cleaned-up,
1707 * printable version of the offending string
1711 S_not_a_number(pTHX_ SV *const sv)
1718 PERL_ARGS_ASSERT_NOT_A_NUMBER;
1721 dsv = newSVpvs_flags("", SVs_TEMP);
1722 pv = sv_uni_display(dsv, sv, 10, 0);
1725 const char * const limit = tmpbuf + sizeof(tmpbuf) - 8;
1726 /* each *s can expand to 4 chars + "...\0",
1727 i.e. need room for 8 chars */
1729 const char *s = SvPVX_const(sv);
1730 const char * const end = s + SvCUR(sv);
1731 for ( ; s < end && d < limit; s++ ) {
1733 if (ch & 128 && !isPRINT_LC(ch)) {
1742 else if (ch == '\r') {
1746 else if (ch == '\f') {
1750 else if (ch == '\\') {
1754 else if (ch == '\0') {
1758 else if (isPRINT_LC(ch))
1775 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
1776 "Argument \"%s\" isn't numeric in %s", pv,
1779 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
1780 "Argument \"%s\" isn't numeric", pv);
1784 =for apidoc looks_like_number
1786 Test if the content of an SV looks like a number (or is a number).
1787 C<Inf> and C<Infinity> are treated as numbers (so will not issue a
1788 non-numeric warning), even if your atof() doesn't grok them.
1794 Perl_looks_like_number(pTHX_ SV *const sv)
1796 register const char *sbegin;
1799 PERL_ARGS_ASSERT_LOOKS_LIKE_NUMBER;
1802 sbegin = SvPVX_const(sv);
1805 else if (SvPOKp(sv))
1806 sbegin = SvPV_const(sv, len);
1808 return SvFLAGS(sv) & (SVf_NOK|SVp_NOK|SVf_IOK|SVp_IOK);
1809 return grok_number(sbegin, len, NULL);
1813 S_glob_2number(pTHX_ GV * const gv)
1815 const U32 wasfake = SvFLAGS(gv) & SVf_FAKE;
1816 SV *const buffer = sv_newmortal();
1818 PERL_ARGS_ASSERT_GLOB_2NUMBER;
1820 /* FAKE globs can get coerced, so need to turn this off temporarily if it
1823 gv_efullname3(buffer, gv, "*");
1824 SvFLAGS(gv) |= wasfake;
1826 /* We know that all GVs stringify to something that is not-a-number,
1827 so no need to test that. */
1828 if (ckWARN(WARN_NUMERIC))
1829 not_a_number(buffer);
1830 /* We just want something true to return, so that S_sv_2iuv_common
1831 can tail call us and return true. */
1835 /* Actually, ISO C leaves conversion of UV to IV undefined, but
1836 until proven guilty, assume that things are not that bad... */
1841 As 64 bit platforms often have an NV that doesn't preserve all bits of
1842 an IV (an assumption perl has been based on to date) it becomes necessary
1843 to remove the assumption that the NV always carries enough precision to
1844 recreate the IV whenever needed, and that the NV is the canonical form.
1845 Instead, IV/UV and NV need to be given equal rights. So as to not lose
1846 precision as a side effect of conversion (which would lead to insanity
1847 and the dragon(s) in t/op/numconvert.t getting very angry) the intent is
1848 1) to distinguish between IV/UV/NV slots that have cached a valid
1849 conversion where precision was lost and IV/UV/NV slots that have a
1850 valid conversion which has lost no precision
1851 2) to ensure that if a numeric conversion to one form is requested that
1852 would lose precision, the precise conversion (or differently
1853 imprecise conversion) is also performed and cached, to prevent
1854 requests for different numeric formats on the same SV causing
1855 lossy conversion chains. (lossless conversion chains are perfectly
1860 SvIOKp is true if the IV slot contains a valid value
1861 SvIOK is true only if the IV value is accurate (UV if SvIOK_UV true)
1862 SvNOKp is true if the NV slot contains a valid value
1863 SvNOK is true only if the NV value is accurate
1866 while converting from PV to NV, check to see if converting that NV to an
1867 IV(or UV) would lose accuracy over a direct conversion from PV to
1868 IV(or UV). If it would, cache both conversions, return NV, but mark
1869 SV as IOK NOKp (ie not NOK).
1871 While converting from PV to IV, check to see if converting that IV to an
1872 NV would lose accuracy over a direct conversion from PV to NV. If it
1873 would, cache both conversions, flag similarly.
1875 Before, the SV value "3.2" could become NV=3.2 IV=3 NOK, IOK quite
1876 correctly because if IV & NV were set NV *always* overruled.
1877 Now, "3.2" will become NV=3.2 IV=3 NOK, IOKp, because the flag's meaning
1878 changes - now IV and NV together means that the two are interchangeable:
1879 SvIVX == (IV) SvNVX && SvNVX == (NV) SvIVX;
1881 The benefit of this is that operations such as pp_add know that if
1882 SvIOK is true for both left and right operands, then integer addition
1883 can be used instead of floating point (for cases where the result won't
1884 overflow). Before, floating point was always used, which could lead to
1885 loss of precision compared with integer addition.
1887 * making IV and NV equal status should make maths accurate on 64 bit
1889 * may speed up maths somewhat if pp_add and friends start to use
1890 integers when possible instead of fp. (Hopefully the overhead in
1891 looking for SvIOK and checking for overflow will not outweigh the
1892 fp to integer speedup)
1893 * will slow down integer operations (callers of SvIV) on "inaccurate"
1894 values, as the change from SvIOK to SvIOKp will cause a call into
1895 sv_2iv each time rather than a macro access direct to the IV slot
1896 * should speed up number->string conversion on integers as IV is
1897 favoured when IV and NV are equally accurate
1899 ####################################################################
1900 You had better be using SvIOK_notUV if you want an IV for arithmetic:
1901 SvIOK is true if (IV or UV), so you might be getting (IV)SvUV.
1902 On the other hand, SvUOK is true iff UV.
1903 ####################################################################
1905 Your mileage will vary depending your CPU's relative fp to integer
1909 #ifndef NV_PRESERVES_UV
1910 # define IS_NUMBER_UNDERFLOW_IV 1
1911 # define IS_NUMBER_UNDERFLOW_UV 2
1912 # define IS_NUMBER_IV_AND_UV 2
1913 # define IS_NUMBER_OVERFLOW_IV 4
1914 # define IS_NUMBER_OVERFLOW_UV 5
1916 /* sv_2iuv_non_preserve(): private routine for use by sv_2iv() and sv_2uv() */
1918 /* For sv_2nv these three cases are "SvNOK and don't bother casting" */
1920 S_sv_2iuv_non_preserve(pTHX_ register SV *const sv
1928 PERL_ARGS_ASSERT_SV_2IUV_NON_PRESERVE;
1930 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_2iuv_non '%s', IV=0x%"UVxf" NV=%"NVgf" inttype=%"UVXf"\n", SvPVX_const(sv), SvIVX(sv), SvNVX(sv), (UV)numtype));
1931 if (SvNVX(sv) < (NV)IV_MIN) {
1932 (void)SvIOKp_on(sv);
1934 SvIV_set(sv, IV_MIN);
1935 return IS_NUMBER_UNDERFLOW_IV;
1937 if (SvNVX(sv) > (NV)UV_MAX) {
1938 (void)SvIOKp_on(sv);
1941 SvUV_set(sv, UV_MAX);
1942 return IS_NUMBER_OVERFLOW_UV;
1944 (void)SvIOKp_on(sv);
1946 /* Can't use strtol etc to convert this string. (See truth table in
1948 if (SvNVX(sv) <= (UV)IV_MAX) {
1949 SvIV_set(sv, I_V(SvNVX(sv)));
1950 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
1951 SvIOK_on(sv); /* Integer is precise. NOK, IOK */
1953 /* Integer is imprecise. NOK, IOKp */
1955 return SvNVX(sv) < 0 ? IS_NUMBER_UNDERFLOW_UV : IS_NUMBER_IV_AND_UV;
1958 SvUV_set(sv, U_V(SvNVX(sv)));
1959 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
1960 if (SvUVX(sv) == UV_MAX) {
1961 /* As we know that NVs don't preserve UVs, UV_MAX cannot
1962 possibly be preserved by NV. Hence, it must be overflow.
1964 return IS_NUMBER_OVERFLOW_UV;
1966 SvIOK_on(sv); /* Integer is precise. NOK, UOK */
1968 /* Integer is imprecise. NOK, IOKp */
1970 return IS_NUMBER_OVERFLOW_IV;
1972 #endif /* !NV_PRESERVES_UV*/
1975 S_sv_2iuv_common(pTHX_ SV *const sv)
1979 PERL_ARGS_ASSERT_SV_2IUV_COMMON;
1982 /* erm. not sure. *should* never get NOKp (without NOK) from sv_2nv
1983 * without also getting a cached IV/UV from it at the same time
1984 * (ie PV->NV conversion should detect loss of accuracy and cache
1985 * IV or UV at same time to avoid this. */
1986 /* IV-over-UV optimisation - choose to cache IV if possible */
1988 if (SvTYPE(sv) == SVt_NV)
1989 sv_upgrade(sv, SVt_PVNV);
1991 (void)SvIOKp_on(sv); /* Must do this first, to clear any SvOOK */
1992 /* < not <= as for NV doesn't preserve UV, ((NV)IV_MAX+1) will almost
1993 certainly cast into the IV range at IV_MAX, whereas the correct
1994 answer is the UV IV_MAX +1. Hence < ensures that dodgy boundary
1996 #if defined(NAN_COMPARE_BROKEN) && defined(Perl_isnan)
1997 if (Perl_isnan(SvNVX(sv))) {
2003 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2004 SvIV_set(sv, I_V(SvNVX(sv)));
2005 if (SvNVX(sv) == (NV) SvIVX(sv)
2006 #ifndef NV_PRESERVES_UV
2007 && (((UV)1 << NV_PRESERVES_UV_BITS) >
2008 (UV)(SvIVX(sv) > 0 ? SvIVX(sv) : -SvIVX(sv)))
2009 /* Don't flag it as "accurately an integer" if the number
2010 came from a (by definition imprecise) NV operation, and
2011 we're outside the range of NV integer precision */
2015 SvIOK_on(sv); /* Can this go wrong with rounding? NWC */
2017 /* scalar has trailing garbage, eg "42a" */
2019 DEBUG_c(PerlIO_printf(Perl_debug_log,
2020 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (precise)\n",
2026 /* IV not precise. No need to convert from PV, as NV
2027 conversion would already have cached IV if it detected
2028 that PV->IV would be better than PV->NV->IV
2029 flags already correct - don't set public IOK. */
2030 DEBUG_c(PerlIO_printf(Perl_debug_log,
2031 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (imprecise)\n",
2036 /* Can the above go wrong if SvIVX == IV_MIN and SvNVX < IV_MIN,
2037 but the cast (NV)IV_MIN rounds to a the value less (more
2038 negative) than IV_MIN which happens to be equal to SvNVX ??
2039 Analogous to 0xFFFFFFFFFFFFFFFF rounding up to NV (2**64) and
2040 NV rounding back to 0xFFFFFFFFFFFFFFFF, so UVX == UV(NVX) and
2041 (NV)UVX == NVX are both true, but the values differ. :-(
2042 Hopefully for 2s complement IV_MIN is something like
2043 0x8000000000000000 which will be exact. NWC */
2046 SvUV_set(sv, U_V(SvNVX(sv)));
2048 (SvNVX(sv) == (NV) SvUVX(sv))
2049 #ifndef NV_PRESERVES_UV
2050 /* Make sure it's not 0xFFFFFFFFFFFFFFFF */
2051 /*&& (SvUVX(sv) != UV_MAX) irrelevant with code below */
2052 && (((UV)1 << NV_PRESERVES_UV_BITS) > SvUVX(sv))
2053 /* Don't flag it as "accurately an integer" if the number
2054 came from a (by definition imprecise) NV operation, and
2055 we're outside the range of NV integer precision */
2061 DEBUG_c(PerlIO_printf(Perl_debug_log,
2062 "0x%"UVxf" 2iv(%"UVuf" => %"IVdf") (as unsigned)\n",
2068 else if (SvPOKp(sv) && SvLEN(sv)) {
2070 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2071 /* We want to avoid a possible problem when we cache an IV/ a UV which
2072 may be later translated to an NV, and the resulting NV is not
2073 the same as the direct translation of the initial string
2074 (eg 123.456 can shortcut to the IV 123 with atol(), but we must
2075 be careful to ensure that the value with the .456 is around if the
2076 NV value is requested in the future).
2078 This means that if we cache such an IV/a UV, we need to cache the
2079 NV as well. Moreover, we trade speed for space, and do not
2080 cache the NV if we are sure it's not needed.
2083 /* SVt_PVNV is one higher than SVt_PVIV, hence this order */
2084 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2085 == IS_NUMBER_IN_UV) {
2086 /* It's definitely an integer, only upgrade to PVIV */
2087 if (SvTYPE(sv) < SVt_PVIV)
2088 sv_upgrade(sv, SVt_PVIV);
2090 } else if (SvTYPE(sv) < SVt_PVNV)
2091 sv_upgrade(sv, SVt_PVNV);
2093 /* If NVs preserve UVs then we only use the UV value if we know that
2094 we aren't going to call atof() below. If NVs don't preserve UVs
2095 then the value returned may have more precision than atof() will
2096 return, even though value isn't perfectly accurate. */
2097 if ((numtype & (IS_NUMBER_IN_UV
2098 #ifdef NV_PRESERVES_UV
2101 )) == IS_NUMBER_IN_UV) {
2102 /* This won't turn off the public IOK flag if it was set above */
2103 (void)SvIOKp_on(sv);
2105 if (!(numtype & IS_NUMBER_NEG)) {
2107 if (value <= (UV)IV_MAX) {
2108 SvIV_set(sv, (IV)value);
2110 /* it didn't overflow, and it was positive. */
2111 SvUV_set(sv, value);
2115 /* 2s complement assumption */
2116 if (value <= (UV)IV_MIN) {
2117 SvIV_set(sv, -(IV)value);
2119 /* Too negative for an IV. This is a double upgrade, but
2120 I'm assuming it will be rare. */
2121 if (SvTYPE(sv) < SVt_PVNV)
2122 sv_upgrade(sv, SVt_PVNV);
2126 SvNV_set(sv, -(NV)value);
2127 SvIV_set(sv, IV_MIN);
2131 /* For !NV_PRESERVES_UV and IS_NUMBER_IN_UV and IS_NUMBER_NOT_INT we
2132 will be in the previous block to set the IV slot, and the next
2133 block to set the NV slot. So no else here. */
2135 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2136 != IS_NUMBER_IN_UV) {
2137 /* It wasn't an (integer that doesn't overflow the UV). */
2138 SvNV_set(sv, Atof(SvPVX_const(sv)));
2140 if (! numtype && ckWARN(WARN_NUMERIC))
2143 #if defined(USE_LONG_DOUBLE)
2144 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%" PERL_PRIgldbl ")\n",
2145 PTR2UV(sv), SvNVX(sv)));
2147 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"NVgf")\n",
2148 PTR2UV(sv), SvNVX(sv)));
2151 #ifdef NV_PRESERVES_UV
2152 (void)SvIOKp_on(sv);
2154 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2155 SvIV_set(sv, I_V(SvNVX(sv)));
2156 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
2159 NOOP; /* Integer is imprecise. NOK, IOKp */
2161 /* UV will not work better than IV */
2163 if (SvNVX(sv) > (NV)UV_MAX) {
2165 /* Integer is inaccurate. NOK, IOKp, is UV */
2166 SvUV_set(sv, UV_MAX);
2168 SvUV_set(sv, U_V(SvNVX(sv)));
2169 /* 0xFFFFFFFFFFFFFFFF not an issue in here, NVs
2170 NV preservse UV so can do correct comparison. */
2171 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
2174 NOOP; /* Integer is imprecise. NOK, IOKp, is UV */
2179 #else /* NV_PRESERVES_UV */
2180 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2181 == (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT)) {
2182 /* The IV/UV slot will have been set from value returned by
2183 grok_number above. The NV slot has just been set using
2186 assert (SvIOKp(sv));
2188 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2189 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2190 /* Small enough to preserve all bits. */
2191 (void)SvIOKp_on(sv);
2193 SvIV_set(sv, I_V(SvNVX(sv)));
2194 if ((NV)(SvIVX(sv)) == SvNVX(sv))
2196 /* Assumption: first non-preserved integer is < IV_MAX,
2197 this NV is in the preserved range, therefore: */
2198 if (!(U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))
2200 Perl_croak(aTHX_ "sv_2iv assumed (U_V(fabs((double)SvNVX(sv))) < (UV)IV_MAX) but SvNVX(sv)=%"NVgf" U_V is 0x%"UVxf", IV_MAX is 0x%"UVxf"\n", SvNVX(sv), U_V(SvNVX(sv)), (UV)IV_MAX);
2204 0 0 already failed to read UV.
2205 0 1 already failed to read UV.
2206 1 0 you won't get here in this case. IV/UV
2207 slot set, public IOK, Atof() unneeded.
2208 1 1 already read UV.
2209 so there's no point in sv_2iuv_non_preserve() attempting
2210 to use atol, strtol, strtoul etc. */
2212 sv_2iuv_non_preserve (sv, numtype);
2214 sv_2iuv_non_preserve (sv);
2218 #endif /* NV_PRESERVES_UV */
2219 /* It might be more code efficient to go through the entire logic above
2220 and conditionally set with SvIOKp_on() rather than SvIOK(), but it
2221 gets complex and potentially buggy, so more programmer efficient
2222 to do it this way, by turning off the public flags: */
2224 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK);
2228 if (isGV_with_GP(sv))
2229 return glob_2number(MUTABLE_GV(sv));
2231 if (!(SvFLAGS(sv) & SVs_PADTMP)) {
2232 if (!PL_localizing && ckWARN(WARN_UNINITIALIZED))
2235 if (SvTYPE(sv) < SVt_IV)
2236 /* Typically the caller expects that sv_any is not NULL now. */
2237 sv_upgrade(sv, SVt_IV);
2238 /* Return 0 from the caller. */
2245 =for apidoc sv_2iv_flags
2247 Return the integer value of an SV, doing any necessary string
2248 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2249 Normally used via the C<SvIV(sv)> and C<SvIVx(sv)> macros.
2255 Perl_sv_2iv_flags(pTHX_ register SV *const sv, const I32 flags)
2260 if (SvGMAGICAL(sv) || SvVALID(sv)) {
2261 /* FBMs use the space for SvIVX and SvNVX for other purposes, and use
2262 the same flag bit as SVf_IVisUV, so must not let them cache IVs.
2263 In practice they are extremely unlikely to actually get anywhere
2264 accessible by user Perl code - the only way that I'm aware of is when
2265 a constant subroutine which is used as the second argument to index.
2267 if (flags & SV_GMAGIC)
2272 return I_V(SvNVX(sv));
2274 if (SvPOKp(sv) && SvLEN(sv)) {
2277 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2279 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2280 == IS_NUMBER_IN_UV) {
2281 /* It's definitely an integer */
2282 if (numtype & IS_NUMBER_NEG) {
2283 if (value < (UV)IV_MIN)
2286 if (value < (UV)IV_MAX)
2291 if (ckWARN(WARN_NUMERIC))
2294 return I_V(Atof(SvPVX_const(sv)));
2299 assert(SvTYPE(sv) >= SVt_PVMG);
2300 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
2301 } else if (SvTHINKFIRST(sv)) {
2306 if (flags & SV_SKIP_OVERLOAD)
2308 tmpstr = AMG_CALLunary(sv, numer_amg);
2309 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2310 return SvIV(tmpstr);
2313 return PTR2IV(SvRV(sv));
2316 sv_force_normal_flags(sv, 0);
2318 if (SvREADONLY(sv) && !SvOK(sv)) {
2319 if (ckWARN(WARN_UNINITIALIZED))
2325 if (S_sv_2iuv_common(aTHX_ sv))
2328 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"IVdf")\n",
2329 PTR2UV(sv),SvIVX(sv)));
2330 return SvIsUV(sv) ? (IV)SvUVX(sv) : SvIVX(sv);
2334 =for apidoc sv_2uv_flags
2336 Return the unsigned integer value of an SV, doing any necessary string
2337 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2338 Normally used via the C<SvUV(sv)> and C<SvUVx(sv)> macros.
2344 Perl_sv_2uv_flags(pTHX_ register SV *const sv, const I32 flags)
2349 if (SvGMAGICAL(sv) || SvVALID(sv)) {
2350 /* FBMs use the space for SvIVX and SvNVX for other purposes, and use
2351 the same flag bit as SVf_IVisUV, so must not let them cache IVs. */
2352 if (flags & SV_GMAGIC)
2357 return U_V(SvNVX(sv));
2358 if (SvPOKp(sv) && SvLEN(sv)) {
2361 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2363 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2364 == IS_NUMBER_IN_UV) {
2365 /* It's definitely an integer */
2366 if (!(numtype & IS_NUMBER_NEG))
2370 if (ckWARN(WARN_NUMERIC))
2373 return U_V(Atof(SvPVX_const(sv)));
2378 assert(SvTYPE(sv) >= SVt_PVMG);
2379 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
2380 } else if (SvTHINKFIRST(sv)) {
2385 if (flags & SV_SKIP_OVERLOAD)
2387 tmpstr = AMG_CALLunary(sv, numer_amg);
2388 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2389 return SvUV(tmpstr);
2392 return PTR2UV(SvRV(sv));
2395 sv_force_normal_flags(sv, 0);
2397 if (SvREADONLY(sv) && !SvOK(sv)) {
2398 if (ckWARN(WARN_UNINITIALIZED))
2404 if (S_sv_2iuv_common(aTHX_ sv))
2408 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2uv(%"UVuf")\n",
2409 PTR2UV(sv),SvUVX(sv)));
2410 return SvIsUV(sv) ? SvUVX(sv) : (UV)SvIVX(sv);
2414 =for apidoc sv_2nv_flags
2416 Return the num value of an SV, doing any necessary string or integer
2417 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2418 Normally used via the C<SvNV(sv)> and C<SvNVx(sv)> macros.
2424 Perl_sv_2nv_flags(pTHX_ register SV *const sv, const I32 flags)
2429 if (SvGMAGICAL(sv) || SvVALID(sv)) {
2430 /* FBMs use the space for SvIVX and SvNVX for other purposes, and use
2431 the same flag bit as SVf_IVisUV, so must not let them cache NVs. */
2432 if (flags & SV_GMAGIC)
2436 if ((SvPOKp(sv) && SvLEN(sv)) && !SvIOKp(sv)) {
2437 if (!SvIOKp(sv) && ckWARN(WARN_NUMERIC) &&
2438 !grok_number(SvPVX_const(sv), SvCUR(sv), NULL))
2440 return Atof(SvPVX_const(sv));
2444 return (NV)SvUVX(sv);
2446 return (NV)SvIVX(sv);
2451 assert(SvTYPE(sv) >= SVt_PVMG);
2452 /* This falls through to the report_uninit near the end of the
2454 } else if (SvTHINKFIRST(sv)) {
2459 if (flags & SV_SKIP_OVERLOAD)
2461 tmpstr = AMG_CALLunary(sv, numer_amg);
2462 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2463 return SvNV(tmpstr);
2466 return PTR2NV(SvRV(sv));
2469 sv_force_normal_flags(sv, 0);
2471 if (SvREADONLY(sv) && !SvOK(sv)) {
2472 if (ckWARN(WARN_UNINITIALIZED))
2477 if (SvTYPE(sv) < SVt_NV) {
2478 /* The logic to use SVt_PVNV if necessary is in sv_upgrade. */
2479 sv_upgrade(sv, SVt_NV);
2480 #ifdef USE_LONG_DOUBLE
2482 STORE_NUMERIC_LOCAL_SET_STANDARD();
2483 PerlIO_printf(Perl_debug_log,
2484 "0x%"UVxf" num(%" PERL_PRIgldbl ")\n",
2485 PTR2UV(sv), SvNVX(sv));
2486 RESTORE_NUMERIC_LOCAL();
2490 STORE_NUMERIC_LOCAL_SET_STANDARD();
2491 PerlIO_printf(Perl_debug_log, "0x%"UVxf" num(%"NVgf")\n",
2492 PTR2UV(sv), SvNVX(sv));
2493 RESTORE_NUMERIC_LOCAL();
2497 else if (SvTYPE(sv) < SVt_PVNV)
2498 sv_upgrade(sv, SVt_PVNV);
2503 SvNV_set(sv, SvIsUV(sv) ? (NV)SvUVX(sv) : (NV)SvIVX(sv));
2504 #ifdef NV_PRESERVES_UV
2510 /* Only set the public NV OK flag if this NV preserves the IV */
2511 /* Check it's not 0xFFFFFFFFFFFFFFFF */
2513 SvIsUV(sv) ? ((SvUVX(sv) != UV_MAX)&&(SvUVX(sv) == U_V(SvNVX(sv))))
2514 : (SvIVX(sv) == I_V(SvNVX(sv))))
2520 else if (SvPOKp(sv) && SvLEN(sv)) {
2522 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2523 if (!SvIOKp(sv) && !numtype && ckWARN(WARN_NUMERIC))
2525 #ifdef NV_PRESERVES_UV
2526 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2527 == IS_NUMBER_IN_UV) {
2528 /* It's definitely an integer */
2529 SvNV_set(sv, (numtype & IS_NUMBER_NEG) ? -(NV)value : (NV)value);
2531 SvNV_set(sv, Atof(SvPVX_const(sv)));
2537 SvNV_set(sv, Atof(SvPVX_const(sv)));
2538 /* Only set the public NV OK flag if this NV preserves the value in
2539 the PV at least as well as an IV/UV would.
2540 Not sure how to do this 100% reliably. */
2541 /* if that shift count is out of range then Configure's test is
2542 wonky. We shouldn't be in here with NV_PRESERVES_UV_BITS ==
2544 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2545 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2546 SvNOK_on(sv); /* Definitely small enough to preserve all bits */
2547 } else if (!(numtype & IS_NUMBER_IN_UV)) {
2548 /* Can't use strtol etc to convert this string, so don't try.
2549 sv_2iv and sv_2uv will use the NV to convert, not the PV. */
2552 /* value has been set. It may not be precise. */
2553 if ((numtype & IS_NUMBER_NEG) && (value > (UV)IV_MIN)) {
2554 /* 2s complement assumption for (UV)IV_MIN */
2555 SvNOK_on(sv); /* Integer is too negative. */
2560 if (numtype & IS_NUMBER_NEG) {
2561 SvIV_set(sv, -(IV)value);
2562 } else if (value <= (UV)IV_MAX) {
2563 SvIV_set(sv, (IV)value);
2565 SvUV_set(sv, value);
2569 if (numtype & IS_NUMBER_NOT_INT) {
2570 /* I believe that even if the original PV had decimals,
2571 they are lost beyond the limit of the FP precision.
2572 However, neither is canonical, so both only get p
2573 flags. NWC, 2000/11/25 */
2574 /* Both already have p flags, so do nothing */
2576 const NV nv = SvNVX(sv);
2577 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2578 if (SvIVX(sv) == I_V(nv)) {
2581 /* It had no "." so it must be integer. */
2585 /* between IV_MAX and NV(UV_MAX).
2586 Could be slightly > UV_MAX */
2588 if (numtype & IS_NUMBER_NOT_INT) {
2589 /* UV and NV both imprecise. */
2591 const UV nv_as_uv = U_V(nv);
2593 if (value == nv_as_uv && SvUVX(sv) != UV_MAX) {
2602 /* It might be more code efficient to go through the entire logic above
2603 and conditionally set with SvNOKp_on() rather than SvNOK(), but it
2604 gets complex and potentially buggy, so more programmer efficient
2605 to do it this way, by turning off the public flags: */
2607 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK);
2608 #endif /* NV_PRESERVES_UV */
2611 if (isGV_with_GP(sv)) {
2612 glob_2number(MUTABLE_GV(sv));
2616 if (!PL_localizing && !(SvFLAGS(sv) & SVs_PADTMP) && ckWARN(WARN_UNINITIALIZED))
2618 assert (SvTYPE(sv) >= SVt_NV);
2619 /* Typically the caller expects that sv_any is not NULL now. */
2620 /* XXX Ilya implies that this is a bug in callers that assume this
2621 and ideally should be fixed. */
2624 #if defined(USE_LONG_DOUBLE)
2626 STORE_NUMERIC_LOCAL_SET_STANDARD();
2627 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2nv(%" PERL_PRIgldbl ")\n",
2628 PTR2UV(sv), SvNVX(sv));
2629 RESTORE_NUMERIC_LOCAL();
2633 STORE_NUMERIC_LOCAL_SET_STANDARD();
2634 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 1nv(%"NVgf")\n",
2635 PTR2UV(sv), SvNVX(sv));
2636 RESTORE_NUMERIC_LOCAL();
2645 Return an SV with the numeric value of the source SV, doing any necessary
2646 reference or overload conversion. You must use the C<SvNUM(sv)> macro to
2647 access this function.
2653 Perl_sv_2num(pTHX_ register SV *const sv)
2655 PERL_ARGS_ASSERT_SV_2NUM;
2660 SV * const tmpsv = AMG_CALLunary(sv, numer_amg);
2661 TAINT_IF(tmpsv && SvTAINTED(tmpsv));
2662 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
2663 return sv_2num(tmpsv);
2665 return sv_2mortal(newSVuv(PTR2UV(SvRV(sv))));
2668 /* uiv_2buf(): private routine for use by sv_2pv_flags(): print an IV or
2669 * UV as a string towards the end of buf, and return pointers to start and
2672 * We assume that buf is at least TYPE_CHARS(UV) long.
2676 S_uiv_2buf(char *const buf, const IV iv, UV uv, const int is_uv, char **const peob)
2678 char *ptr = buf + TYPE_CHARS(UV);
2679 char * const ebuf = ptr;
2682 PERL_ARGS_ASSERT_UIV_2BUF;
2694 *--ptr = '0' + (char)(uv % 10);
2703 =for apidoc sv_2pv_flags
2705 Returns a pointer to the string value of an SV, and sets *lp to its length.
2706 If flags includes SV_GMAGIC, does an mg_get() first. Coerces sv to a string
2708 Normally invoked via the C<SvPV_flags> macro. C<sv_2pv()> and C<sv_2pv_nomg>
2709 usually end up here too.
2715 Perl_sv_2pv_flags(pTHX_ register SV *const sv, STRLEN *const lp, const I32 flags)
2725 if (SvGMAGICAL(sv)) {
2726 if (flags & SV_GMAGIC)
2731 if (flags & SV_MUTABLE_RETURN)
2732 return SvPVX_mutable(sv);
2733 if (flags & SV_CONST_RETURN)
2734 return (char *)SvPVX_const(sv);
2737 if (SvIOKp(sv) || SvNOKp(sv)) {
2738 char tbuf[64]; /* Must fit sprintf/Gconvert of longest IV/NV */
2743 ? my_snprintf(tbuf, sizeof(tbuf), "%"UVuf, (UV)SvUVX(sv))
2744 : my_snprintf(tbuf, sizeof(tbuf), "%"IVdf, (IV)SvIVX(sv));
2745 } else if(SvNVX(sv) == 0.0) {
2750 Gconvert(SvNVX(sv), NV_DIG, 0, tbuf);
2757 SvUPGRADE(sv, SVt_PV);
2760 s = SvGROW_mutable(sv, len + 1);
2763 return (char*)memcpy(s, tbuf, len + 1);
2769 assert(SvTYPE(sv) >= SVt_PVMG);
2770 /* This falls through to the report_uninit near the end of the
2772 } else if (SvTHINKFIRST(sv)) {
2777 if (flags & SV_SKIP_OVERLOAD)
2779 tmpstr = AMG_CALLunary(sv, string_amg);
2780 TAINT_IF(tmpstr && SvTAINTED(tmpstr));
2781 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2783 /* char *pv = lp ? SvPV(tmpstr, *lp) : SvPV_nolen(tmpstr);
2787 if ((SvFLAGS(tmpstr) & (SVf_POK)) == SVf_POK) {
2788 if (flags & SV_CONST_RETURN) {
2789 pv = (char *) SvPVX_const(tmpstr);
2791 pv = (flags & SV_MUTABLE_RETURN)
2792 ? SvPVX_mutable(tmpstr) : SvPVX(tmpstr);
2795 *lp = SvCUR(tmpstr);
2797 pv = sv_2pv_flags(tmpstr, lp, flags);
2810 SV *const referent = SvRV(sv);
2814 retval = buffer = savepvn("NULLREF", len);
2815 } else if (SvTYPE(referent) == SVt_REGEXP) {
2816 REGEXP * const re = (REGEXP *)MUTABLE_PTR(referent);
2821 /* If the regex is UTF-8 we want the containing scalar to
2822 have an UTF-8 flag too */
2828 if ((seen_evals = RX_SEEN_EVALS(re)))
2829 PL_reginterp_cnt += seen_evals;
2832 *lp = RX_WRAPLEN(re);
2834 return RX_WRAPPED(re);
2836 const char *const typestr = sv_reftype(referent, 0);
2837 const STRLEN typelen = strlen(typestr);
2838 UV addr = PTR2UV(referent);
2839 const char *stashname = NULL;
2840 STRLEN stashnamelen = 0; /* hush, gcc */
2841 const char *buffer_end;
2843 if (SvOBJECT(referent)) {
2844 const HEK *const name = HvNAME_HEK(SvSTASH(referent));
2847 stashname = HEK_KEY(name);
2848 stashnamelen = HEK_LEN(name);
2850 if (HEK_UTF8(name)) {
2856 stashname = "__ANON__";
2859 len = stashnamelen + 1 /* = */ + typelen + 3 /* (0x */
2860 + 2 * sizeof(UV) + 2 /* )\0 */;
2862 len = typelen + 3 /* (0x */
2863 + 2 * sizeof(UV) + 2 /* )\0 */;
2866 Newx(buffer, len, char);
2867 buffer_end = retval = buffer + len;
2869 /* Working backwards */
2873 *--retval = PL_hexdigit[addr & 15];
2874 } while (addr >>= 4);
2880 memcpy(retval, typestr, typelen);
2884 retval -= stashnamelen;
2885 memcpy(retval, stashname, stashnamelen);
2887 /* retval may not necessarily have reached the start of the
2889 assert (retval >= buffer);
2891 len = buffer_end - retval - 1; /* -1 for that \0 */
2899 if (SvREADONLY(sv) && !SvOK(sv)) {
2902 if (flags & SV_UNDEF_RETURNS_NULL)
2904 if (ckWARN(WARN_UNINITIALIZED))
2909 if (SvIOK(sv) || ((SvIOKp(sv) && !SvNOKp(sv)))) {
2910 /* I'm assuming that if both IV and NV are equally valid then
2911 converting the IV is going to be more efficient */
2912 const U32 isUIOK = SvIsUV(sv);
2913 char buf[TYPE_CHARS(UV)];
2917 if (SvTYPE(sv) < SVt_PVIV)
2918 sv_upgrade(sv, SVt_PVIV);
2919 ptr = uiv_2buf(buf, SvIVX(sv), SvUVX(sv), isUIOK, &ebuf);
2921 /* inlined from sv_setpvn */
2922 s = SvGROW_mutable(sv, len + 1);
2923 Move(ptr, s, len, char);
2927 else if (SvNOKp(sv)) {
2928 if (SvTYPE(sv) < SVt_PVNV)
2929 sv_upgrade(sv, SVt_PVNV);
2930 if (SvNVX(sv) == 0.0) {
2931 s = SvGROW_mutable(sv, 2);
2936 /* The +20 is pure guesswork. Configure test needed. --jhi */
2937 s = SvGROW_mutable(sv, NV_DIG + 20);
2938 /* some Xenix systems wipe out errno here */
2939 Gconvert(SvNVX(sv), NV_DIG, 0, s);
2949 if (isGV_with_GP(sv)) {
2950 GV *const gv = MUTABLE_GV(sv);
2951 const U32 wasfake = SvFLAGS(gv) & SVf_FAKE;
2952 SV *const buffer = sv_newmortal();
2954 /* FAKE globs can get coerced, so need to turn this off temporarily
2957 gv_efullname3(buffer, gv, "*");
2958 SvFLAGS(gv) |= wasfake;
2960 if (SvPOK(buffer)) {
2962 *lp = SvCUR(buffer);
2964 return SvPVX(buffer);
2975 if (flags & SV_UNDEF_RETURNS_NULL)
2977 if (!PL_localizing && !(SvFLAGS(sv) & SVs_PADTMP) && ckWARN(WARN_UNINITIALIZED))
2979 if (SvTYPE(sv) < SVt_PV)
2980 /* Typically the caller expects that sv_any is not NULL now. */
2981 sv_upgrade(sv, SVt_PV);
2985 const STRLEN len = s - SvPVX_const(sv);
2991 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
2992 PTR2UV(sv),SvPVX_const(sv)));
2993 if (flags & SV_CONST_RETURN)
2994 return (char *)SvPVX_const(sv);
2995 if (flags & SV_MUTABLE_RETURN)
2996 return SvPVX_mutable(sv);
3001 =for apidoc sv_copypv
3003 Copies a stringified representation of the source SV into the
3004 destination SV. Automatically performs any necessary mg_get and
3005 coercion of numeric values into strings. Guaranteed to preserve
3006 UTF8 flag even from overloaded objects. Similar in nature to
3007 sv_2pv[_flags] but operates directly on an SV instead of just the
3008 string. Mostly uses sv_2pv_flags to do its work, except when that
3009 would lose the UTF-8'ness of the PV.
3015 Perl_sv_copypv(pTHX_ SV *const dsv, register SV *const ssv)
3018 const char * const s = SvPV_const(ssv,len);
3020 PERL_ARGS_ASSERT_SV_COPYPV;
3022 sv_setpvn(dsv,s,len);
3030 =for apidoc sv_2pvbyte
3032 Return a pointer to the byte-encoded representation of the SV, and set *lp
3033 to its length. May cause the SV to be downgraded from UTF-8 as a
3036 Usually accessed via the C<SvPVbyte> macro.
3042 Perl_sv_2pvbyte(pTHX_ register SV *const sv, STRLEN *const lp)
3044 PERL_ARGS_ASSERT_SV_2PVBYTE;
3047 sv_utf8_downgrade(sv,0);
3048 return lp ? SvPV_nomg(sv,*lp) : SvPV_nomg_nolen(sv);
3052 =for apidoc sv_2pvutf8
3054 Return a pointer to the UTF-8-encoded representation of the SV, and set *lp
3055 to its length. May cause the SV to be upgraded to UTF-8 as a side-effect.
3057 Usually accessed via the C<SvPVutf8> macro.
3063 Perl_sv_2pvutf8(pTHX_ register SV *const sv, STRLEN *const lp)
3065 PERL_ARGS_ASSERT_SV_2PVUTF8;
3067 sv_utf8_upgrade(sv);
3068 return lp ? SvPV(sv,*lp) : SvPV_nolen(sv);
3073 =for apidoc sv_2bool
3075 This macro is only used by sv_true() or its macro equivalent, and only if
3076 the latter's argument is neither SvPOK, SvIOK nor SvNOK.
3077 It calls sv_2bool_flags with the SV_GMAGIC flag.
3079 =for apidoc sv_2bool_flags
3081 This function is only used by sv_true() and friends, and only if
3082 the latter's argument is neither SvPOK, SvIOK nor SvNOK. If the flags
3083 contain SV_GMAGIC, then it does an mg_get() first.
3090 Perl_sv_2bool_flags(pTHX_ register SV *const sv, const I32 flags)
3094 PERL_ARGS_ASSERT_SV_2BOOL_FLAGS;
3096 if(flags & SV_GMAGIC) SvGETMAGIC(sv);
3102 SV * const tmpsv = AMG_CALLunary(sv, bool__amg);
3103 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
3104 return cBOOL(SvTRUE(tmpsv));
3106 return SvRV(sv) != 0;
3109 register XPV* const Xpvtmp = (XPV*)SvANY(sv);
3111 (*sv->sv_u.svu_pv > '0' ||
3112 Xpvtmp->xpv_cur > 1 ||
3113 (Xpvtmp->xpv_cur && *sv->sv_u.svu_pv != '0')))
3120 return SvIVX(sv) != 0;
3123 return SvNVX(sv) != 0.0;
3125 if (isGV_with_GP(sv))
3135 =for apidoc sv_utf8_upgrade
3137 Converts the PV of an SV to its UTF-8-encoded form.
3138 Forces the SV to string form if it is not already.
3139 Will C<mg_get> on C<sv> if appropriate.
3140 Always sets the SvUTF8 flag to avoid future validity checks even
3141 if the whole string is the same in UTF-8 as not.
3142 Returns the number of bytes in the converted string
3144 This is not as a general purpose byte encoding to Unicode interface:
3145 use the Encode extension for that.
3147 =for apidoc sv_utf8_upgrade_nomg
3149 Like sv_utf8_upgrade, but doesn't do magic on C<sv>
3151 =for apidoc sv_utf8_upgrade_flags
3153 Converts the PV of an SV to its UTF-8-encoded form.
3154 Forces the SV to string form if it is not already.
3155 Always sets the SvUTF8 flag to avoid future validity checks even
3156 if all the bytes are invariant in UTF-8. If C<flags> has C<SV_GMAGIC> bit set,
3157 will C<mg_get> on C<sv> if appropriate, else not.
3158 Returns the number of bytes in the converted string
3159 C<sv_utf8_upgrade> and
3160 C<sv_utf8_upgrade_nomg> are implemented in terms of this function.
3162 This is not as a general purpose byte encoding to Unicode interface:
3163 use the Encode extension for that.
3167 The grow version is currently not externally documented. It adds a parameter,
3168 extra, which is the number of unused bytes the string of 'sv' is guaranteed to
3169 have free after it upon return. This allows the caller to reserve extra space
3170 that it intends to fill, to avoid extra grows.
3172 Also externally undocumented for the moment is the flag SV_FORCE_UTF8_UPGRADE,
3173 which can be used to tell this function to not first check to see if there are
3174 any characters that are different in UTF-8 (variant characters) which would
3175 force it to allocate a new string to sv, but to assume there are. Typically
3176 this flag is used by a routine that has already parsed the string to find that
3177 there are such characters, and passes this information on so that the work
3178 doesn't have to be repeated.
3180 (One might think that the calling routine could pass in the position of the
3181 first such variant, so it wouldn't have to be found again. But that is not the
3182 case, because typically when the caller is likely to use this flag, it won't be
3183 calling this routine unless it finds something that won't fit into a byte.
3184 Otherwise it tries to not upgrade and just use bytes. But some things that
3185 do fit into a byte are variants in utf8, and the caller may not have been
3186 keeping track of these.)
3188 If the routine itself changes the string, it adds a trailing NUL. Such a NUL
3189 isn't guaranteed due to having other routines do the work in some input cases,
3190 or if the input is already flagged as being in utf8.
3192 The speed of this could perhaps be improved for many cases if someone wanted to
3193 write a fast function that counts the number of variant characters in a string,
3194 especially if it could return the position of the first one.
3199 Perl_sv_utf8_upgrade_flags_grow(pTHX_ register SV *const sv, const I32 flags, STRLEN extra)
3203 PERL_ARGS_ASSERT_SV_UTF8_UPGRADE_FLAGS_GROW;
3205 if (sv == &PL_sv_undef)
3209 if (SvREADONLY(sv) && (SvPOKp(sv) || SvIOKp(sv) || SvNOKp(sv))) {
3210 (void) sv_2pv_flags(sv,&len, flags);
3212 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3216 (void) SvPV_force_flags(sv,len,flags & SV_GMAGIC);
3221 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3226 sv_force_normal_flags(sv, 0);
3229 if (PL_encoding && !(flags & SV_UTF8_NO_ENCODING)) {
3230 sv_recode_to_utf8(sv, PL_encoding);
3231 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3235 if (SvCUR(sv) == 0) {
3236 if (extra) SvGROW(sv, extra);
3237 } else { /* Assume Latin-1/EBCDIC */
3238 /* This function could be much more efficient if we
3239 * had a FLAG in SVs to signal if there are any variant
3240 * chars in the PV. Given that there isn't such a flag
3241 * make the loop as fast as possible (although there are certainly ways
3242 * to speed this up, eg. through vectorization) */
3243 U8 * s = (U8 *) SvPVX_const(sv);
3244 U8 * e = (U8 *) SvEND(sv);
3246 STRLEN two_byte_count = 0;
3248 if (flags & SV_FORCE_UTF8_UPGRADE) goto must_be_utf8;
3250 /* See if really will need to convert to utf8. We mustn't rely on our
3251 * incoming SV being well formed and having a trailing '\0', as certain
3252 * code in pp_formline can send us partially built SVs. */
3256 if (NATIVE_IS_INVARIANT(ch)) continue;
3258 t--; /* t already incremented; re-point to first variant */
3263 /* utf8 conversion not needed because all are invariants. Mark as
3264 * UTF-8 even if no variant - saves scanning loop */
3270 /* Here, the string should be converted to utf8, either because of an
3271 * input flag (two_byte_count = 0), or because a character that
3272 * requires 2 bytes was found (two_byte_count = 1). t points either to
3273 * the beginning of the string (if we didn't examine anything), or to
3274 * the first variant. In either case, everything from s to t - 1 will
3275 * occupy only 1 byte each on output.
3277 * There are two main ways to convert. One is to create a new string
3278 * and go through the input starting from the beginning, appending each
3279 * converted value onto the new string as we go along. It's probably
3280 * best to allocate enough space in the string for the worst possible
3281 * case rather than possibly running out of space and having to
3282 * reallocate and then copy what we've done so far. Since everything
3283 * from s to t - 1 is invariant, the destination can be initialized
3284 * with these using a fast memory copy
3286 * The other way is to figure out exactly how big the string should be
3287 * by parsing the entire input. Then you don't have to make it big
3288 * enough to handle the worst possible case, and more importantly, if
3289 * the string you already have is large enough, you don't have to
3290 * allocate a new string, you can copy the last character in the input
3291 * string to the final position(s) that will be occupied by the
3292 * converted string and go backwards, stopping at t, since everything
3293 * before that is invariant.
3295 * There are advantages and disadvantages to each method.
3297 * In the first method, we can allocate a new string, do the memory
3298 * copy from the s to t - 1, and then proceed through the rest of the
3299 * string byte-by-byte.
3301 * In the second method, we proceed through the rest of the input
3302 * string just calculating how big the converted string will be. Then
3303 * there are two cases:
3304 * 1) if the string has enough extra space to handle the converted
3305 * value. We go backwards through the string, converting until we
3306 * get to the position we are at now, and then stop. If this
3307 * position is far enough along in the string, this method is
3308 * faster than the other method. If the memory copy were the same
3309 * speed as the byte-by-byte loop, that position would be about
3310 * half-way, as at the half-way mark, parsing to the end and back
3311 * is one complete string's parse, the same amount as starting
3312 * over and going all the way through. Actually, it would be
3313 * somewhat less than half-way, as it's faster to just count bytes
3314 * than to also copy, and we don't have the overhead of allocating
3315 * a new string, changing the scalar to use it, and freeing the
3316 * existing one. But if the memory copy is fast, the break-even
3317 * point is somewhere after half way. The counting loop could be
3318 * sped up by vectorization, etc, to move the break-even point
3319 * further towards the beginning.
3320 * 2) if the string doesn't have enough space to handle the converted
3321 * value. A new string will have to be allocated, and one might
3322 * as well, given that, start from the beginning doing the first
3323 * method. We've spent extra time parsing the string and in
3324 * exchange all we've gotten is that we know precisely how big to
3325 * make the new one. Perl is more optimized for time than space,
3326 * so this case is a loser.
3327 * So what I've decided to do is not use the 2nd method unless it is
3328 * guaranteed that a new string won't have to be allocated, assuming
3329 * the worst case. I also decided not to put any more conditions on it
3330 * than this, for now. It seems likely that, since the worst case is
3331 * twice as big as the unknown portion of the string (plus 1), we won't
3332 * be guaranteed enough space, causing us to go to the first method,
3333 * unless the string is short, or the first variant character is near
3334 * the end of it. In either of these cases, it seems best to use the
3335 * 2nd method. The only circumstance I can think of where this would
3336 * be really slower is if the string had once had much more data in it
3337 * than it does now, but there is still a substantial amount in it */
3340 STRLEN invariant_head = t - s;
3341 STRLEN size = invariant_head + (e - t) * 2 + 1 + extra;
3342 if (SvLEN(sv) < size) {
3344 /* Here, have decided to allocate a new string */
3349 Newx(dst, size, U8);
3351 /* If no known invariants at the beginning of the input string,
3352 * set so starts from there. Otherwise, can use memory copy to
3353 * get up to where we are now, and then start from here */
3355 if (invariant_head <= 0) {
3358 Copy(s, dst, invariant_head, char);
3359 d = dst + invariant_head;
3363 const UV uv = NATIVE8_TO_UNI(*t++);
3364 if (UNI_IS_INVARIANT(uv))
3365 *d++ = (U8)UNI_TO_NATIVE(uv);
3367 *d++ = (U8)UTF8_EIGHT_BIT_HI(uv);
3368 *d++ = (U8)UTF8_EIGHT_BIT_LO(uv);
3372 SvPV_free(sv); /* No longer using pre-existing string */
3373 SvPV_set(sv, (char*)dst);
3374 SvCUR_set(sv, d - dst);
3375 SvLEN_set(sv, size);
3378 /* Here, have decided to get the exact size of the string.
3379 * Currently this happens only when we know that there is
3380 * guaranteed enough space to fit the converted string, so
3381 * don't have to worry about growing. If two_byte_count is 0,
3382 * then t points to the first byte of the string which hasn't
3383 * been examined yet. Otherwise two_byte_count is 1, and t
3384 * points to the first byte in the string that will expand to
3385 * two. Depending on this, start examining at t or 1 after t.
3388 U8 *d = t + two_byte_count;
3391 /* Count up the remaining bytes that expand to two */
3394 const U8 chr = *d++;
3395 if (! NATIVE_IS_INVARIANT(chr)) two_byte_count++;
3398 /* The string will expand by just the number of bytes that
3399 * occupy two positions. But we are one afterwards because of
3400 * the increment just above. This is the place to put the
3401 * trailing NUL, and to set the length before we decrement */
3403 d += two_byte_count;
3404 SvCUR_set(sv, d - s);
3408 /* Having decremented d, it points to the position to put the
3409 * very last byte of the expanded string. Go backwards through
3410 * the string, copying and expanding as we go, stopping when we
3411 * get to the part that is invariant the rest of the way down */
3415 const U8 ch = NATIVE8_TO_UNI(*e--);
3416 if (UNI_IS_INVARIANT(ch)) {
3417 *d-- = UNI_TO_NATIVE(ch);
3419 *d-- = (U8)UTF8_EIGHT_BIT_LO(ch);
3420 *d-- = (U8)UTF8_EIGHT_BIT_HI(ch);
3425 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
3426 /* Update pos. We do it at the end rather than during
3427 * the upgrade, to avoid slowing down the common case
3428 * (upgrade without pos) */
3429 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
3431 I32 pos = mg->mg_len;
3432 if (pos > 0 && (U32)pos > invariant_head) {
3433 U8 *d = (U8*) SvPVX(sv) + invariant_head;
3434 STRLEN n = (U32)pos - invariant_head;
3436 if (UTF8_IS_START(*d))
3441 mg->mg_len = d - (U8*)SvPVX(sv);
3444 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
3445 magic_setutf8(sv,mg); /* clear UTF8 cache */
3450 /* Mark as UTF-8 even if no variant - saves scanning loop */
3456 =for apidoc sv_utf8_downgrade
3458 Attempts to convert the PV of an SV from characters to bytes.
3459 If the PV contains a character that cannot fit
3460 in a byte, this conversion will fail;
3461 in this case, either returns false or, if C<fail_ok> is not
3464 This is not as a general purpose Unicode to byte encoding interface:
3465 use the Encode extension for that.
3471 Perl_sv_utf8_downgrade(pTHX_ register SV *const sv, const bool fail_ok)
3475 PERL_ARGS_ASSERT_SV_UTF8_DOWNGRADE;
3477 if (SvPOKp(sv) && SvUTF8(sv)) {
3481 int mg_flags = SV_GMAGIC;
3484 sv_force_normal_flags(sv, 0);
3486 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
3488 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
3490 I32 pos = mg->mg_len;
3492 sv_pos_b2u(sv, &pos);
3493 mg_flags = 0; /* sv_pos_b2u does get magic */
3497 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
3498 magic_setutf8(sv,mg); /* clear UTF8 cache */
3501 s = (U8 *) SvPV_flags(sv, len, mg_flags);
3503 if (!utf8_to_bytes(s, &len)) {
3508 Perl_croak(aTHX_ "Wide character in %s",
3511 Perl_croak(aTHX_ "Wide character");
3522 =for apidoc sv_utf8_encode
3524 Converts the PV of an SV to UTF-8, but then turns the C<SvUTF8>
3525 flag off so that it looks like octets again.
3531 Perl_sv_utf8_encode(pTHX_ register SV *const sv)
3533 PERL_ARGS_ASSERT_SV_UTF8_ENCODE;
3536 sv_force_normal_flags(sv, 0);
3538 if (SvREADONLY(sv)) {
3539 Perl_croak_no_modify(aTHX);
3541 (void) sv_utf8_upgrade(sv);
3546 =for apidoc sv_utf8_decode
3548 If the PV of the SV is an octet sequence in UTF-8
3549 and contains a multiple-byte character, the C<SvUTF8> flag is turned on
3550 so that it looks like a character. If the PV contains only single-byte
3551 characters, the C<SvUTF8> flag stays off.
3552 Scans PV for validity and returns false if the PV is invalid UTF-8.
3558 Perl_sv_utf8_decode(pTHX_ register SV *const sv)
3560 PERL_ARGS_ASSERT_SV_UTF8_DECODE;
3563 const U8 *start, *c;
3566 /* The octets may have got themselves encoded - get them back as
3569 if (!sv_utf8_downgrade(sv, TRUE))
3572 /* it is actually just a matter of turning the utf8 flag on, but
3573 * we want to make sure everything inside is valid utf8 first.
3575 c = start = (const U8 *) SvPVX_const(sv);
3576 if (!is_utf8_string(c, SvCUR(sv)+1))
3578 e = (const U8 *) SvEND(sv);
3581 if (!UTF8_IS_INVARIANT(ch)) {
3586 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
3587 /* adjust pos to the start of a UTF8 char sequence */
3588 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
3590 I32 pos = mg->mg_len;
3592 for (c = start + pos; c > start; c--) {
3593 if (UTF8_IS_START(*c))
3596 mg->mg_len = c - start;
3599 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
3600 magic_setutf8(sv,mg); /* clear UTF8 cache */
3607 =for apidoc sv_setsv
3609 Copies the contents of the source SV C<ssv> into the destination SV
3610 C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3611 function if the source SV needs to be reused. Does not handle 'set' magic.
3612 Loosely speaking, it performs a copy-by-value, obliterating any previous
3613 content of the destination.
3615 You probably want to use one of the assortment of wrappers, such as
3616 C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3617 C<SvSetMagicSV_nosteal>.
3619 =for apidoc sv_setsv_flags
3621 Copies the contents of the source SV C<ssv> into the destination SV
3622 C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3623 function if the source SV needs to be reused. Does not handle 'set' magic.
3624 Loosely speaking, it performs a copy-by-value, obliterating any previous
3625 content of the destination.
3626 If the C<flags> parameter has the C<SV_GMAGIC> bit set, will C<mg_get> on
3627 C<ssv> if appropriate, else not. If the C<flags> parameter has the
3628 C<NOSTEAL> bit set then the buffers of temps will not be stolen. <sv_setsv>
3629 and C<sv_setsv_nomg> are implemented in terms of this function.
3631 You probably want to use one of the assortment of wrappers, such as
3632 C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3633 C<SvSetMagicSV_nosteal>.
3635 This is the primary function for copying scalars, and most other
3636 copy-ish functions and macros use this underneath.
3642 S_glob_assign_glob(pTHX_ SV *const dstr, SV *const sstr, const int dtype)
3644 I32 mro_changes = 0; /* 1 = method, 2 = isa, 3 = recursive isa */
3645 HV *old_stash = NULL;
3647 PERL_ARGS_ASSERT_GLOB_ASSIGN_GLOB;
3649 if (dtype != SVt_PVGV && !isGV_with_GP(dstr)) {
3650 const char * const name = GvNAME(sstr);
3651 const STRLEN len = GvNAMELEN(sstr);
3653 if (dtype >= SVt_PV) {
3659 SvUPGRADE(dstr, SVt_PVGV);
3660 (void)SvOK_off(dstr);
3661 /* FIXME - why are we doing this, then turning it off and on again
3663 isGV_with_GP_on(dstr);
3665 GvSTASH(dstr) = GvSTASH(sstr);
3667 Perl_sv_add_backref(aTHX_ MUTABLE_SV(GvSTASH(dstr)), dstr);
3668 gv_name_set(MUTABLE_GV(dstr), name, len, GV_ADD);
3669 SvFAKE_on(dstr); /* can coerce to non-glob */
3672 if(GvGP(MUTABLE_GV(sstr))) {
3673 /* If source has method cache entry, clear it */
3675 SvREFCNT_dec(GvCV(sstr));
3676 GvCV_set(sstr, NULL);
3679 /* If source has a real method, then a method is
3682 GvCV((const GV *)sstr) && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3688 /* If dest already had a real method, that's a change as well */
3690 !mro_changes && GvGP(MUTABLE_GV(dstr)) && GvCVu((const GV *)dstr)
3691 && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3696 /* We don’t need to check the name of the destination if it was not a
3697 glob to begin with. */
3698 if(dtype == SVt_PVGV) {
3699 const char * const name = GvNAME((const GV *)dstr);
3702 /* The stash may have been detached from the symbol table, so
3704 && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3705 && GvAV((const GV *)sstr)
3709 const STRLEN len = GvNAMELEN(dstr);
3710 if ((len > 1 && name[len-2] == ':' && name[len-1] == ':')
3711 || (len == 1 && name[0] == ':')) {
3714 /* Set aside the old stash, so we can reset isa caches on
3716 if((old_stash = GvHV(dstr)))
3717 /* Make sure we do not lose it early. */
3718 SvREFCNT_inc_simple_void_NN(
3719 sv_2mortal((SV *)old_stash)
3725 gp_free(MUTABLE_GV(dstr));
3726 isGV_with_GP_off(dstr);
3727 (void)SvOK_off(dstr);
3728 isGV_with_GP_on(dstr);
3729 GvINTRO_off(dstr); /* one-shot flag */
3730 GvGP_set(dstr, gp_ref(GvGP(sstr)));
3731 if (SvTAINTED(sstr))
3733 if (GvIMPORTED(dstr) != GVf_IMPORTED
3734 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3736 GvIMPORTED_on(dstr);
3739 if(mro_changes == 2) {
3741 SV * const sref = (SV *)GvAV((const GV *)dstr);
3742 if (SvSMAGICAL(sref) && (mg = mg_find(sref, PERL_MAGIC_isa))) {
3743 if (SvTYPE(mg->mg_obj) != SVt_PVAV) {
3744 AV * const ary = newAV();
3745 av_push(ary, mg->mg_obj); /* takes the refcount */
3746 mg->mg_obj = (SV *)ary;
3748 av_push((AV *)mg->mg_obj, SvREFCNT_inc_simple_NN(dstr));
3750 else sv_magic(sref, dstr, PERL_MAGIC_isa, NULL, 0);
3751 mro_isa_changed_in(GvSTASH(dstr));
3753 else if(mro_changes == 3) {
3754 HV * const stash = GvHV(dstr);
3755 if(old_stash ? (HV *)HvENAME_get(old_stash) : stash)
3761 else if(mro_changes) mro_method_changed_in(GvSTASH(dstr));
3766 S_glob_assign_ref(pTHX_ SV *const dstr, SV *const sstr)
3768 SV * const sref = SvREFCNT_inc(SvRV(sstr));
3770 const int intro = GvINTRO(dstr);
3773 const U32 stype = SvTYPE(sref);
3775 PERL_ARGS_ASSERT_GLOB_ASSIGN_REF;
3778 GvINTRO_off(dstr); /* one-shot flag */
3779 GvLINE(dstr) = CopLINE(PL_curcop);
3780 GvEGV(dstr) = MUTABLE_GV(dstr);
3785 location = (SV **) &(GvGP(dstr)->gp_cv); /* XXX bypassing GvCV_set */
3786 import_flag = GVf_IMPORTED_CV;
3789 location = (SV **) &GvHV(dstr);
3790 import_flag = GVf_IMPORTED_HV;
3793 location = (SV **) &GvAV(dstr);
3794 import_flag = GVf_IMPORTED_AV;
3797 location = (SV **) &GvIOp(dstr);
3800 location = (SV **) &GvFORM(dstr);
3803 location = &GvSV(dstr);
3804 import_flag = GVf_IMPORTED_SV;
3807 if (stype == SVt_PVCV) {
3808 /*if (GvCVGEN(dstr) && (GvCV(dstr) != (const CV *)sref || GvCVGEN(dstr))) {*/
3809 if (GvCVGEN(dstr)) {
3810 SvREFCNT_dec(GvCV(dstr));
3811 GvCV_set(dstr, NULL);
3812 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3815 SAVEGENERICSV(*location);
3819 if (stype == SVt_PVCV && (*location != sref || GvCVGEN(dstr))) {
3820 CV* const cv = MUTABLE_CV(*location);
3822 if (!GvCVGEN((const GV *)dstr) &&
3823 (CvROOT(cv) || CvXSUB(cv)))
3825 /* Redefining a sub - warning is mandatory if
3826 it was a const and its value changed. */
3827 if (CvCONST(cv) && CvCONST((const CV *)sref)
3829 == cv_const_sv((const CV *)sref)) {
3831 /* They are 2 constant subroutines generated from
3832 the same constant. This probably means that
3833 they are really the "same" proxy subroutine
3834 instantiated in 2 places. Most likely this is
3835 when a constant is exported twice. Don't warn.
3838 else if (ckWARN(WARN_REDEFINE)
3840 && (!CvCONST((const CV *)sref)
3841 || sv_cmp(cv_const_sv(cv),
3842 cv_const_sv((const CV *)
3844 Perl_warner(aTHX_ packWARN(WARN_REDEFINE),
3847 ? "Constant subroutine %s::%s redefined"
3848 : "Subroutine %s::%s redefined"),
3849 HvNAME_get(GvSTASH((const GV *)dstr)),
3850 GvENAME(MUTABLE_GV(dstr)));
3854 cv_ckproto_len(cv, (const GV *)dstr,
3855 SvPOK(sref) ? SvPVX_const(sref) : NULL,
3856 SvPOK(sref) ? SvCUR(sref) : 0);
3858 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3859 GvASSUMECV_on(dstr);
3860 if(GvSTASH(dstr)) mro_method_changed_in(GvSTASH(dstr)); /* sub foo { 1 } sub bar { 2 } *bar = \&foo */
3863 if (import_flag && !(GvFLAGS(dstr) & import_flag)
3864 && CopSTASH_ne(PL_curcop, GvSTASH(dstr))) {
3865 GvFLAGS(dstr) |= import_flag;
3867 if (stype == SVt_PVHV) {
3868 const char * const name = GvNAME((GV*)dstr);
3869 const STRLEN len = GvNAMELEN(dstr);
3872 (len > 1 && name[len-2] == ':' && name[len-1] == ':')
3873 || (len == 1 && name[0] == ':')
3875 && (!dref || HvENAME_get(dref))
3878 (HV *)sref, (HV *)dref,
3884 stype == SVt_PVAV && sref != dref
3885 && strEQ(GvNAME((GV*)dstr), "ISA")
3886 /* The stash may have been detached from the symbol table, so
3887 check its name before doing anything. */
3888 && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3891 MAGIC * const omg = dref && SvSMAGICAL(dref)
3892 ? mg_find(dref, PERL_MAGIC_isa)
3894 if (SvSMAGICAL(sref) && (mg = mg_find(sref, PERL_MAGIC_isa))) {
3895 if (SvTYPE(mg->mg_obj) != SVt_PVAV) {
3896 AV * const ary = newAV();
3897 av_push(ary, mg->mg_obj); /* takes the refcount */
3898 mg->mg_obj = (SV *)ary;
3901 if (SvTYPE(omg->mg_obj) == SVt_PVAV) {
3902 SV **svp = AvARRAY((AV *)omg->mg_obj);
3903 I32 items = AvFILLp((AV *)omg->mg_obj) + 1;
3907 SvREFCNT_inc_simple_NN(*svp++)
3913 SvREFCNT_inc_simple_NN(omg->mg_obj)
3917 av_push((AV *)mg->mg_obj,SvREFCNT_inc_simple_NN(dstr));
3922 sref, omg ? omg->mg_obj : dstr, PERL_MAGIC_isa, NULL, 0
3924 mg = mg_find(sref, PERL_MAGIC_isa);
3926 /* Since the *ISA assignment could have affected more than
3927 one stash, don’t call mro_isa_changed_in directly, but let
3928 magic_clearisa do it for us, as it already has the logic for
3929 dealing with globs vs arrays of globs. */
3931 Perl_magic_clearisa(aTHX_ NULL, mg);
3936 if (SvTAINTED(sstr))
3942 Perl_sv_setsv_flags(pTHX_ SV *dstr, register SV* sstr, const I32 flags)
3945 register U32 sflags;
3947 register svtype stype;
3949 PERL_ARGS_ASSERT_SV_SETSV_FLAGS;
3954 if (SvIS_FREED(dstr)) {
3955 Perl_croak(aTHX_ "panic: attempt to copy value %" SVf
3956 " to a freed scalar %p", SVfARG(sstr), (void *)dstr);
3958 SV_CHECK_THINKFIRST_COW_DROP(dstr);
3960 sstr = &PL_sv_undef;
3961 if (SvIS_FREED(sstr)) {
3962 Perl_croak(aTHX_ "panic: attempt to copy freed scalar %p to %p",
3963 (void*)sstr, (void*)dstr);
3965 stype = SvTYPE(sstr);
3966 dtype = SvTYPE(dstr);
3968 (void)SvAMAGIC_off(dstr);
3971 /* need to nuke the magic */
3975 /* There's a lot of redundancy below but we're going for speed here */
3980 if (dtype != SVt_PVGV && dtype != SVt_PVLV) {
3981 (void)SvOK_off(dstr);
3989 sv_upgrade(dstr, SVt_IV);
3993 sv_upgrade(dstr, SVt_PVIV);
3997 goto end_of_first_switch;
3999 (void)SvIOK_only(dstr);
4000 SvIV_set(dstr, SvIVX(sstr));
4003 /* SvTAINTED can only be true if the SV has taint magic, which in
4004 turn means that the SV type is PVMG (or greater). This is the
4005 case statement for SVt_IV, so this cannot be true (whatever gcov
4007 assert(!SvTAINTED(sstr));
4012 if (dtype < SVt_PV && dtype != SVt_IV)
4013 sv_upgrade(dstr, SVt_IV);
4021 sv_upgrade(dstr, SVt_NV);
4025 sv_upgrade(dstr, SVt_PVNV);
4029 goto end_of_first_switch;
4031 SvNV_set(dstr, SvNVX(sstr));
4032 (void)SvNOK_only(dstr);
4033 /* SvTAINTED can only be true if the SV has taint magic, which in
4034 turn means that the SV type is PVMG (or greater). This is the
4035 case statement for SVt_NV, so this cannot be true (whatever gcov
4037 assert(!SvTAINTED(sstr));
4043 #ifdef PERL_OLD_COPY_ON_WRITE
4044 if ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS) {
4045 if (dtype < SVt_PVIV)
4046 sv_upgrade(dstr, SVt_PVIV);
4053 sv_upgrade(dstr, SVt_PV);
4056 if (dtype < SVt_PVIV)
4057 sv_upgrade(dstr, SVt_PVIV);
4060 if (dtype < SVt_PVNV)
4061 sv_upgrade(dstr, SVt_PVNV);
4065 const char * const type = sv_reftype(sstr,0);
4067 Perl_croak(aTHX_ "Bizarre copy of %s in %s", type, OP_DESC(PL_op));
4069 Perl_croak(aTHX_ "Bizarre copy of %s", type);
4074 if (dtype < SVt_REGEXP)
4075 sv_upgrade(dstr, SVt_REGEXP);
4078 /* case SVt_BIND: */
4082 if (SvGMAGICAL(sstr) && (flags & SV_GMAGIC)) {
4084 if (SvTYPE(sstr) != stype)
4085 stype = SvTYPE(sstr);
4087 if (isGV_with_GP(sstr) && dtype <= SVt_PVLV) {
4088 glob_assign_glob(dstr, sstr, dtype);
4091 if (stype == SVt_PVLV)
4092 SvUPGRADE(dstr, SVt_PVNV);
4094 SvUPGRADE(dstr, (svtype)stype);
4096 end_of_first_switch:
4098 /* dstr may have been upgraded. */
4099 dtype = SvTYPE(dstr);
4100 sflags = SvFLAGS(sstr);
4102 if (dtype == SVt_PVCV || dtype == SVt_PVFM) {
4103 /* Assigning to a subroutine sets the prototype. */
4106 const char *const ptr = SvPV_const(sstr, len);
4108 SvGROW(dstr, len + 1);
4109 Copy(ptr, SvPVX(dstr), len + 1, char);
4110 SvCUR_set(dstr, len);
4112 SvFLAGS(dstr) |= sflags & SVf_UTF8;
4116 } else if (dtype == SVt_PVAV || dtype == SVt_PVHV) {
4117 const char * const type = sv_reftype(dstr,0);
4119 Perl_croak(aTHX_ "Cannot copy to %s in %s", type, OP_DESC(PL_op));
4121 Perl_croak(aTHX_ "Cannot copy to %s", type);
4122 } else if (sflags & SVf_ROK) {
4123 if (isGV_with_GP(dstr)
4124 && SvTYPE(SvRV(sstr)) == SVt_PVGV && isGV_with_GP(SvRV(sstr))) {
4127 if (GvIMPORTED(dstr) != GVf_IMPORTED
4128 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
4130 GvIMPORTED_on(dstr);
4135 glob_assign_glob(dstr, sstr, dtype);
4139 if (dtype >= SVt_PV) {
4140 if (isGV_with_GP(dstr)) {
4141 glob_assign_ref(dstr, sstr);
4144 if (SvPVX_const(dstr)) {
4150 (void)SvOK_off(dstr);
4151 SvRV_set(dstr, SvREFCNT_inc(SvRV(sstr)));
4152 SvFLAGS(dstr) |= sflags & SVf_ROK;
4153 assert(!(sflags & SVp_NOK));
4154 assert(!(sflags & SVp_IOK));
4155 assert(!(sflags & SVf_NOK));
4156 assert(!(sflags & SVf_IOK));
4158 else if (isGV_with_GP(dstr)) {
4159 if (!(sflags & SVf_OK)) {
4160 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
4161 "Undefined value assigned to typeglob");
4165 const char *nambeg = SvPV_nomg_const(sstr, len);
4166 GV *gv = gv_fetchpvn_flags(
4167 nambeg, len, SvUTF8(sstr)|GV_ADD, SVt_PVGV
4169 if (dstr != (const SV *)gv) {
4170 const char * const name = GvNAME((const GV *)dstr);
4171 const STRLEN len = GvNAMELEN(dstr);
4172 HV *old_stash = NULL;
4173 bool reset_isa = FALSE;
4174 if ((len > 1 && name[len-2] == ':' && name[len-1] == ':')
4175 || (len == 1 && name[0] == ':')) {
4176 /* Set aside the old stash, so we can reset isa caches
4177 on its subclasses. */
4178 if((old_stash = GvHV(dstr))) {
4179 /* Make sure we do not lose it early. */
4180 SvREFCNT_inc_simple_void_NN(
4181 sv_2mortal((SV *)old_stash)
4188 gp_free(MUTABLE_GV(dstr));
4189 GvGP_set(dstr, gp_ref(GvGP(gv)));
4192 HV * const stash = GvHV(dstr);
4194 old_stash ? (HV *)HvENAME_get(old_stash) : stash
4204 else if (dtype == SVt_REGEXP && stype == SVt_REGEXP) {
4205 reg_temp_copy((REGEXP*)dstr, (REGEXP*)sstr);
4207 else if (sflags & SVp_POK) {
4211 * Check to see if we can just swipe the string. If so, it's a
4212 * possible small lose on short strings, but a big win on long ones.
4213 * It might even be a win on short strings if SvPVX_const(dstr)
4214 * has to be allocated and SvPVX_const(sstr) has to be freed.
4215 * Likewise if we can set up COW rather than doing an actual copy, we
4216 * drop to the else clause, as the swipe code and the COW setup code
4217 * have much in common.
4220 /* Whichever path we take through the next code, we want this true,
4221 and doing it now facilitates the COW check. */
4222 (void)SvPOK_only(dstr);
4225 /* If we're already COW then this clause is not true, and if COW
4226 is allowed then we drop down to the else and make dest COW
4227 with us. If caller hasn't said that we're allowed to COW
4228 shared hash keys then we don't do the COW setup, even if the
4229 source scalar is a shared hash key scalar. */
4230 (((flags & SV_COW_SHARED_HASH_KEYS)
4231 ? (sflags & (SVf_FAKE|SVf_READONLY)) != (SVf_FAKE|SVf_READONLY)
4232 : 1 /* If making a COW copy is forbidden then the behaviour we
4233 desire is as if the source SV isn't actually already
4234 COW, even if it is. So we act as if the source flags
4235 are not COW, rather than actually testing them. */
4237 #ifndef PERL_OLD_COPY_ON_WRITE
4238 /* The change that added SV_COW_SHARED_HASH_KEYS makes the logic
4239 when PERL_OLD_COPY_ON_WRITE is defined a little wrong.
4240 Conceptually PERL_OLD_COPY_ON_WRITE being defined should
4241 override SV_COW_SHARED_HASH_KEYS, because it means "always COW"
4242 but in turn, it's somewhat dead code, never expected to go
4243 live, but more kept as a placeholder on how to do it better
4244 in a newer implementation. */
4245 /* If we are COW and dstr is a suitable target then we drop down
4246 into the else and make dest a COW of us. */
4247 || (SvFLAGS(dstr) & CAN_COW_MASK) != CAN_COW_FLAGS
4252 (sflags & SVs_TEMP) && /* slated for free anyway? */
4253 !(sflags & SVf_OOK) && /* and not involved in OOK hack? */
4254 (!(flags & SV_NOSTEAL)) &&
4255 /* and we're allowed to steal temps */
4256 SvREFCNT(sstr) == 1 && /* and no other references to it? */
4257 SvLEN(sstr)) /* and really is a string */
4258 #ifdef PERL_OLD_COPY_ON_WRITE
4259 && ((flags & SV_COW_SHARED_HASH_KEYS)
4260 ? (!((sflags & CAN_COW_MASK) == CAN_COW_FLAGS
4261 && (SvFLAGS(dstr) & CAN_COW_MASK) == CAN_COW_FLAGS
4262 && SvTYPE(sstr) >= SVt_PVIV && SvTYPE(sstr) != SVt_PVFM))
4266 /* Failed the swipe test, and it's not a shared hash key either.
4267 Have to copy the string. */
4268 STRLEN len = SvCUR(sstr);
4269 SvGROW(dstr, len + 1); /* inlined from sv_setpvn */
4270 Move(SvPVX_const(sstr),SvPVX(dstr),len,char);
4271 SvCUR_set(dstr, len);
4272 *SvEND(dstr) = '\0';
4274 /* If PERL_OLD_COPY_ON_WRITE is not defined, then isSwipe will always
4276 /* Either it's a shared hash key, or it's suitable for
4277 copy-on-write or we can swipe the string. */
4279 PerlIO_printf(Perl_debug_log, "Copy on write: sstr --> dstr\n");
4283 #ifdef PERL_OLD_COPY_ON_WRITE
4285 if ((sflags & (SVf_FAKE | SVf_READONLY))
4286 != (SVf_FAKE | SVf_READONLY)) {
4287 SvREADONLY_on(sstr);
4289 /* Make the source SV into a loop of 1.
4290 (about to become 2) */
4291 SV_COW_NEXT_SV_SET(sstr, sstr);
4295 /* Initial code is common. */
4296 if (SvPVX_const(dstr)) { /* we know that dtype >= SVt_PV */
4301 /* making another shared SV. */
4302 STRLEN cur = SvCUR(sstr);
4303 STRLEN len = SvLEN(sstr);
4304 #ifdef PERL_OLD_COPY_ON_WRITE
4306 assert (SvTYPE(dstr) >= SVt_PVIV);
4307 /* SvIsCOW_normal */
4308 /* splice us in between source and next-after-source. */
4309 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4310 SV_COW_NEXT_SV_SET(sstr, dstr);
4311 SvPV_set(dstr, SvPVX_mutable(sstr));
4315 /* SvIsCOW_shared_hash */
4316 DEBUG_C(PerlIO_printf(Perl_debug_log,
4317 "Copy on write: Sharing hash\n"));
4319 assert (SvTYPE(dstr) >= SVt_PV);
4321 HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)))));
4323 SvLEN_set(dstr, len);
4324 SvCUR_set(dstr, cur);
4325 SvREADONLY_on(dstr);
4329 { /* Passes the swipe test. */
4330 SvPV_set(dstr, SvPVX_mutable(sstr));
4331 SvLEN_set(dstr, SvLEN(sstr));
4332 SvCUR_set(dstr, SvCUR(sstr));
4335 (void)SvOK_off(sstr); /* NOTE: nukes most SvFLAGS on sstr */
4336 SvPV_set(sstr, NULL);
4342 if (sflags & SVp_NOK) {
4343 SvNV_set(dstr, SvNVX(sstr));
4345 if (sflags & SVp_IOK) {
4346 SvIV_set(dstr, SvIVX(sstr));
4347 /* Must do this otherwise some other overloaded use of 0x80000000
4348 gets confused. I guess SVpbm_VALID */
4349 if (sflags & SVf_IVisUV)
4352 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_NOK|SVp_NOK|SVf_UTF8);
4354 const MAGIC * const smg = SvVSTRING_mg(sstr);
4356 sv_magic(dstr, NULL, PERL_MAGIC_vstring,
4357 smg->mg_ptr, smg->mg_len);
4358 SvRMAGICAL_on(dstr);
4362 else if (sflags & (SVp_IOK|SVp_NOK)) {
4363 (void)SvOK_off(dstr);
4364 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_IVisUV|SVf_NOK|SVp_NOK);
4365 if (sflags & SVp_IOK) {
4366 /* XXXX Do we want to set IsUV for IV(ROK)? Be extra safe... */
4367 SvIV_set(dstr, SvIVX(sstr));
4369 if (sflags & SVp_NOK) {
4370 SvNV_set(dstr, SvNVX(sstr));
4374 if (isGV_with_GP(sstr)) {
4375 /* This stringification rule for globs is spread in 3 places.
4376 This feels bad. FIXME. */
4377 const U32 wasfake = sflags & SVf_FAKE;
4379 /* FAKE globs can get coerced, so need to turn this off
4380 temporarily if it is on. */
4382 gv_efullname3(dstr, MUTABLE_GV(sstr), "*");
4383 SvFLAGS(sstr) |= wasfake;
4386 (void)SvOK_off(dstr);
4388 if (SvTAINTED(sstr))
4393 =for apidoc sv_setsv_mg
4395 Like C<sv_setsv>, but also handles 'set' magic.
4401 Perl_sv_setsv_mg(pTHX_ SV *const dstr, register SV *const sstr)
4403 PERL_ARGS_ASSERT_SV_SETSV_MG;
4405 sv_setsv(dstr,sstr);
4409 #ifdef PERL_OLD_COPY_ON_WRITE
4411 Perl_sv_setsv_cow(pTHX_ SV *dstr, SV *sstr)
4413 STRLEN cur = SvCUR(sstr);
4414 STRLEN len = SvLEN(sstr);
4415 register char *new_pv;
4417 PERL_ARGS_ASSERT_SV_SETSV_COW;
4420 PerlIO_printf(Perl_debug_log, "Fast copy on write: %p -> %p\n",
4421 (void*)sstr, (void*)dstr);
4428 if (SvTHINKFIRST(dstr))
4429 sv_force_normal_flags(dstr, SV_COW_DROP_PV);
4430 else if (SvPVX_const(dstr))
4431 Safefree(SvPVX_const(dstr));
4435 SvUPGRADE(dstr, SVt_PVIV);
4437 assert (SvPOK(sstr));
4438 assert (SvPOKp(sstr));
4439 assert (!SvIOK(sstr));
4440 assert (!SvIOKp(sstr));
4441 assert (!SvNOK(sstr));
4442 assert (!SvNOKp(sstr));
4444 if (SvIsCOW(sstr)) {
4446 if (SvLEN(sstr) == 0) {
4447 /* source is a COW shared hash key. */
4448 DEBUG_C(PerlIO_printf(Perl_debug_log,
4449 "Fast copy on write: Sharing hash\n"));
4450 new_pv = HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr))));
4453 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4455 assert ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS);
4456 SvUPGRADE(sstr, SVt_PVIV);
4457 SvREADONLY_on(sstr);
4459 DEBUG_C(PerlIO_printf(Perl_debug_log,
4460 "Fast copy on write: Converting sstr to COW\n"));
4461 SV_COW_NEXT_SV_SET(dstr, sstr);
4463 SV_COW_NEXT_SV_SET(sstr, dstr);
4464 new_pv = SvPVX_mutable(sstr);
4467 SvPV_set(dstr, new_pv);
4468 SvFLAGS(dstr) = (SVt_PVIV|SVf_POK|SVp_POK|SVf_FAKE|SVf_READONLY);
4471 SvLEN_set(dstr, len);
4472 SvCUR_set(dstr, cur);
4481 =for apidoc sv_setpvn
4483 Copies a string into an SV. The C<len> parameter indicates the number of
4484 bytes to be copied. If the C<ptr> argument is NULL the SV will become
4485 undefined. Does not handle 'set' magic. See C<sv_setpvn_mg>.
4491 Perl_sv_setpvn(pTHX_ register SV *const sv, register const char *const ptr, register const STRLEN len)
4494 register char *dptr;
4496 PERL_ARGS_ASSERT_SV_SETPVN;
4498 SV_CHECK_THINKFIRST_COW_DROP(sv);
4504 /* len is STRLEN which is unsigned, need to copy to signed */
4507 Perl_croak(aTHX_ "panic: sv_setpvn called with negative strlen");
4509 SvUPGRADE(sv, SVt_PV);
4511 dptr = SvGROW(sv, len + 1);
4512 Move(ptr,dptr,len,char);
4515 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4520 =for apidoc sv_setpvn_mg
4522 Like C<sv_setpvn>, but also handles 'set' magic.
4528 Perl_sv_setpvn_mg(pTHX_ register SV *const sv, register const char *const ptr, register const STRLEN len)
4530 PERL_ARGS_ASSERT_SV_SETPVN_MG;
4532 sv_setpvn(sv,ptr,len);
4537 =for apidoc sv_setpv
4539 Copies a string into an SV. The string must be null-terminated. Does not
4540 handle 'set' magic. See C<sv_setpv_mg>.
4546 Perl_sv_setpv(pTHX_ register SV *const sv, register const char *const ptr)
4549 register STRLEN len;
4551 PERL_ARGS_ASSERT_SV_SETPV;
4553 SV_CHECK_THINKFIRST_COW_DROP(sv);
4559 SvUPGRADE(sv, SVt_PV);
4561 SvGROW(sv, len + 1);
4562 Move(ptr,SvPVX(sv),len+1,char);
4564 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4569 =for apidoc sv_setpv_mg
4571 Like C<sv_setpv>, but also handles 'set' magic.
4577 Perl_sv_setpv_mg(pTHX_ register SV *const sv, register const char *const ptr)
4579 PERL_ARGS_ASSERT_SV_SETPV_MG;
4586 =for apidoc sv_usepvn_flags
4588 Tells an SV to use C<ptr> to find its string value. Normally the
4589 string is stored inside the SV but sv_usepvn allows the SV to use an
4590 outside string. The C<ptr> should point to memory that was allocated
4591 by C<malloc>. The string length, C<len>, must be supplied. By default
4592 this function will realloc (i.e. move) the memory pointed to by C<ptr>,
4593 so that pointer should not be freed or used by the programmer after
4594 giving it to sv_usepvn, and neither should any pointers from "behind"
4595 that pointer (e.g. ptr + 1) be used.
4597 If C<flags> & SV_SMAGIC is true, will call SvSETMAGIC. If C<flags> &
4598 SV_HAS_TRAILING_NUL is true, then C<ptr[len]> must be NUL, and the realloc
4599 will be skipped. (i.e. the buffer is actually at least 1 byte longer than
4600 C<len>, and already meets the requirements for storing in C<SvPVX>)
4606 Perl_sv_usepvn_flags(pTHX_ SV *const sv, char *ptr, const STRLEN len, const U32 flags)
4611 PERL_ARGS_ASSERT_SV_USEPVN_FLAGS;
4613 SV_CHECK_THINKFIRST_COW_DROP(sv);
4614 SvUPGRADE(sv, SVt_PV);
4617 if (flags & SV_SMAGIC)
4621 if (SvPVX_const(sv))
4625 if (flags & SV_HAS_TRAILING_NUL)
4626 assert(ptr[len] == '\0');
4629 allocate = (flags & SV_HAS_TRAILING_NUL)
4631 #ifdef Perl_safesysmalloc_size
4634 PERL_STRLEN_ROUNDUP(len + 1);
4636 if (flags & SV_HAS_TRAILING_NUL) {
4637 /* It's long enough - do nothing.
4638 Specifically Perl_newCONSTSUB is relying on this. */
4641 /* Force a move to shake out bugs in callers. */
4642 char *new_ptr = (char*)safemalloc(allocate);
4643 Copy(ptr, new_ptr, len, char);
4644 PoisonFree(ptr,len,char);
4648 ptr = (char*) saferealloc (ptr, allocate);
4651 #ifdef Perl_safesysmalloc_size
4652 SvLEN_set(sv, Perl_safesysmalloc_size(ptr));
4654 SvLEN_set(sv, allocate);
4658 if (!(flags & SV_HAS_TRAILING_NUL)) {
4661 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4663 if (flags & SV_SMAGIC)
4667 #ifdef PERL_OLD_COPY_ON_WRITE
4668 /* Need to do this *after* making the SV normal, as we need the buffer
4669 pointer to remain valid until after we've copied it. If we let go too early,
4670 another thread could invalidate it by unsharing last of the same hash key
4671 (which it can do by means other than releasing copy-on-write Svs)
4672 or by changing the other copy-on-write SVs in the loop. */
4674 S_sv_release_COW(pTHX_ register SV *sv, const char *pvx, SV *after)
4676 PERL_ARGS_ASSERT_SV_RELEASE_COW;
4678 { /* this SV was SvIsCOW_normal(sv) */
4679 /* we need to find the SV pointing to us. */
4680 SV *current = SV_COW_NEXT_SV(after);
4682 if (current == sv) {
4683 /* The SV we point to points back to us (there were only two of us
4685 Hence other SV is no longer copy on write either. */
4687 SvREADONLY_off(after);
4689 /* We need to follow the pointers around the loop. */
4691 while ((next = SV_COW_NEXT_SV(current)) != sv) {
4694 /* don't loop forever if the structure is bust, and we have
4695 a pointer into a closed loop. */
4696 assert (current != after);
4697 assert (SvPVX_const(current) == pvx);
4699 /* Make the SV before us point to the SV after us. */
4700 SV_COW_NEXT_SV_SET(current, after);
4706 =for apidoc sv_force_normal_flags
4708 Undo various types of fakery on an SV: if the PV is a shared string, make
4709 a private copy; if we're a ref, stop refing; if we're a glob, downgrade to
4710 an xpvmg; if we're a copy-on-write scalar, this is the on-write time when
4711 we do the copy, and is also used locally. If C<SV_COW_DROP_PV> is set
4712 then a copy-on-write scalar drops its PV buffer (if any) and becomes
4713 SvPOK_off rather than making a copy. (Used where this scalar is about to be
4714 set to some other value.) In addition, the C<flags> parameter gets passed to
4715 C<sv_unref_flags()> when unreffing. C<sv_force_normal> calls this function
4716 with flags set to 0.
4722 Perl_sv_force_normal_flags(pTHX_ register SV *const sv, const U32 flags)
4726 PERL_ARGS_ASSERT_SV_FORCE_NORMAL_FLAGS;
4728 #ifdef PERL_OLD_COPY_ON_WRITE
4729 if (SvREADONLY(sv)) {
4731 const char * const pvx = SvPVX_const(sv);
4732 const STRLEN len = SvLEN(sv);
4733 const STRLEN cur = SvCUR(sv);
4734 /* next COW sv in the loop. If len is 0 then this is a shared-hash
4735 key scalar, so we mustn't attempt to call SV_COW_NEXT_SV(), as
4736 we'll fail an assertion. */
4737 SV * const next = len ? SV_COW_NEXT_SV(sv) : 0;
4740 PerlIO_printf(Perl_debug_log,
4741 "Copy on write: Force normal %ld\n",
4747 /* This SV doesn't own the buffer, so need to Newx() a new one: */
4750 if (flags & SV_COW_DROP_PV) {
4751 /* OK, so we don't need to copy our buffer. */
4754 SvGROW(sv, cur + 1);
4755 Move(pvx,SvPVX(sv),cur,char);
4760 sv_release_COW(sv, pvx, next);
4762 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4768 else if (IN_PERL_RUNTIME)
4769 Perl_croak_no_modify(aTHX);
4772 if (SvREADONLY(sv)) {
4773 if (SvFAKE(sv) && !isGV_with_GP(sv)) {
4774 const char * const pvx = SvPVX_const(sv);
4775 const STRLEN len = SvCUR(sv);
4780 SvGROW(sv, len + 1);
4781 Move(pvx,SvPVX(sv),len,char);
4783 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4785 else if (IN_PERL_RUNTIME)
4786 Perl_croak_no_modify(aTHX);
4790 sv_unref_flags(sv, flags);
4791 else if (SvFAKE(sv) && isGV_with_GP(sv))
4793 else if (SvFAKE(sv) && SvTYPE(sv) == SVt_REGEXP) {
4794 /* Need to downgrade the REGEXP to a simple(r) scalar. This is analogous
4795 to sv_unglob. We only need it here, so inline it. */
4796 const svtype new_type = SvMAGIC(sv) || SvSTASH(sv) ? SVt_PVMG : SVt_PV;
4797 SV *const temp = newSV_type(new_type);
4798 void *const temp_p = SvANY(sv);
4800 if (new_type == SVt_PVMG) {
4801 SvMAGIC_set(temp, SvMAGIC(sv));
4802 SvMAGIC_set(sv, NULL);
4803 SvSTASH_set(temp, SvSTASH(sv));
4804 SvSTASH_set(sv, NULL);
4806 SvCUR_set(temp, SvCUR(sv));
4807 /* Remember that SvPVX is in the head, not the body. */
4809 SvLEN_set(temp, SvLEN(sv));
4810 /* This signals "buffer is owned by someone else" in sv_clear,
4811 which is the least effort way to stop it freeing the buffer.
4813 SvLEN_set(sv, SvLEN(sv)+1);
4815 /* Their buffer is already owned by someone else. */
4816 SvPVX(sv) = savepvn(SvPVX(sv), SvCUR(sv));
4817 SvLEN_set(temp, SvCUR(sv)+1);
4820 /* Now swap the rest of the bodies. */
4822 SvFLAGS(sv) &= ~(SVf_FAKE|SVTYPEMASK);
4823 SvFLAGS(sv) |= new_type;
4824 SvANY(sv) = SvANY(temp);
4826 SvFLAGS(temp) &= ~(SVTYPEMASK);
4827 SvFLAGS(temp) |= SVt_REGEXP|SVf_FAKE;
4828 SvANY(temp) = temp_p;
4837 Efficient removal of characters from the beginning of the string buffer.
4838 SvPOK(sv) must be true and the C<ptr> must be a pointer to somewhere inside
4839 the string buffer. The C<ptr> becomes the first character of the adjusted
4840 string. Uses the "OOK hack".
4841 Beware: after this function returns, C<ptr> and SvPVX_const(sv) may no longer
4842 refer to the same chunk of data.
4848 Perl_sv_chop(pTHX_ register SV *const sv, register const char *const ptr)
4854 const U8 *real_start;
4858 PERL_ARGS_ASSERT_SV_CHOP;
4860 if (!ptr || !SvPOKp(sv))
4862 delta = ptr - SvPVX_const(sv);
4864 /* Nothing to do. */
4867 /* SvPVX(sv) may move in SV_CHECK_THINKFIRST(sv), but after this line,
4868 nothing uses the value of ptr any more. */
4869 max_delta = SvLEN(sv) ? SvLEN(sv) : SvCUR(sv);
4870 if (ptr <= SvPVX_const(sv))
4871 Perl_croak(aTHX_ "panic: sv_chop ptr=%p, start=%p, end=%p",
4872 ptr, SvPVX_const(sv), SvPVX_const(sv) + max_delta);
4873 SV_CHECK_THINKFIRST(sv);
4874 if (delta > max_delta)
4875 Perl_croak(aTHX_ "panic: sv_chop ptr=%p (was %p), start=%p, end=%p",
4876 SvPVX_const(sv) + delta, ptr, SvPVX_const(sv),
4877 SvPVX_const(sv) + max_delta);
4880 if (!SvLEN(sv)) { /* make copy of shared string */
4881 const char *pvx = SvPVX_const(sv);
4882 const STRLEN len = SvCUR(sv);
4883 SvGROW(sv, len + 1);
4884 Move(pvx,SvPVX(sv),len,char);
4887 SvFLAGS(sv) |= SVf_OOK;
4890 SvOOK_offset(sv, old_delta);
4892 SvLEN_set(sv, SvLEN(sv) - delta);
4893 SvCUR_set(sv, SvCUR(sv) - delta);
4894 SvPV_set(sv, SvPVX(sv) + delta);
4896 p = (U8 *)SvPVX_const(sv);
4901 real_start = p - delta;
4905 if (delta < 0x100) {
4909 p -= sizeof(STRLEN);
4910 Copy((U8*)&delta, p, sizeof(STRLEN), U8);
4914 /* Fill the preceding buffer with sentinals to verify that no-one is
4916 while (p > real_start) {
4924 =for apidoc sv_catpvn
4926 Concatenates the string onto the end of the string which is in the SV. The
4927 C<len> indicates number of bytes to copy. If the SV has the UTF-8
4928 status set, then the bytes appended should be valid UTF-8.
4929 Handles 'get' magic, but not 'set' magic. See C<sv_catpvn_mg>.
4931 =for apidoc sv_catpvn_flags
4933 Concatenates the string onto the end of the string which is in the SV. The
4934 C<len> indicates number of bytes to copy. If the SV has the UTF-8
4935 status set, then the bytes appended should be valid UTF-8.
4936 If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<dsv> if
4937 appropriate, else not. C<sv_catpvn> and C<sv_catpvn_nomg> are implemented
4938 in terms of this function.
4944 Perl_sv_catpvn_flags(pTHX_ register SV *const dsv, register const char *sstr, register const STRLEN slen, const I32 flags)
4948 const char * const dstr = SvPV_force_flags(dsv, dlen, flags);
4950 PERL_ARGS_ASSERT_SV_CATPVN_FLAGS;
4952 SvGROW(dsv, dlen + slen + 1);
4954 sstr = SvPVX_const(dsv);
4955 Move(sstr, SvPVX(dsv) + dlen, slen, char);
4956 SvCUR_set(dsv, SvCUR(dsv) + slen);
4958 (void)SvPOK_only_UTF8(dsv); /* validate pointer */
4960 if (flags & SV_SMAGIC)
4965 =for apidoc sv_catsv
4967 Concatenates the string from SV C<ssv> onto the end of the string in
4968 SV C<dsv>. Modifies C<dsv> but not C<ssv>. Handles 'get' magic, but
4969 not 'set' magic. See C<sv_catsv_mg>.
4971 =for apidoc sv_catsv_flags
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>. If C<flags> has C<SV_GMAGIC>
4975 bit set, will C<mg_get> on the SVs if appropriate, else not. C<sv_catsv>
4976 and C<sv_catsv_nomg> are implemented in terms of this function.
4981 Perl_sv_catsv_flags(pTHX_ SV *const dsv, register SV *const ssv, const I32 flags)
4985 PERL_ARGS_ASSERT_SV_CATSV_FLAGS;
4989 const char *spv = SvPV_flags_const(ssv, slen, flags);
4991 /* sutf8 and dutf8 were type bool, but under USE_ITHREADS,
4992 gcc version 2.95.2 20000220 (Debian GNU/Linux) for
4993 Linux xxx 2.2.17 on sparc64 with gcc -O2, we erroneously
4994 get dutf8 = 0x20000000, (i.e. SVf_UTF8) even though
4995 dsv->sv_flags doesn't have that bit set.
4996 Andy Dougherty 12 Oct 2001
4998 const I32 sutf8 = DO_UTF8(ssv);
5001 if (SvGMAGICAL(dsv) && (flags & SV_GMAGIC))
5003 dutf8 = DO_UTF8(dsv);
5005 if (dutf8 != sutf8) {
5007 /* Not modifying source SV, so taking a temporary copy. */
5008 SV* const csv = newSVpvn_flags(spv, slen, SVs_TEMP);
5010 sv_utf8_upgrade(csv);
5011 spv = SvPV_const(csv, slen);
5014 /* Leave enough space for the cat that's about to happen */
5015 sv_utf8_upgrade_flags_grow(dsv, 0, slen);
5017 sv_catpvn_nomg(dsv, spv, slen);
5020 if (flags & SV_SMAGIC)
5025 =for apidoc sv_catpv
5027 Concatenates the string onto the end of the string which is in the SV.
5028 If the SV has the UTF-8 status set, then the bytes appended should be
5029 valid UTF-8. Handles 'get' magic, but not 'set' magic. See C<sv_catpv_mg>.
5034 Perl_sv_catpv(pTHX_ register SV *const sv, register const char *ptr)
5037 register STRLEN len;
5041 PERL_ARGS_ASSERT_SV_CATPV;
5045 junk = SvPV_force(sv, tlen);
5047 SvGROW(sv, tlen + len + 1);
5049 ptr = SvPVX_const(sv);
5050 Move(ptr,SvPVX(sv)+tlen,len+1,char);
5051 SvCUR_set(sv, SvCUR(sv) + len);
5052 (void)SvPOK_only_UTF8(sv); /* validate pointer */
5057 =for apidoc sv_catpv_flags
5059 Concatenates the string onto the end of the string which is in the SV.
5060 If the SV has the UTF-8 status set, then the bytes appended should
5061 be valid UTF-8. If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get>
5062 on the SVs if appropriate, else not.
5068 Perl_sv_catpv_flags(pTHX_ SV *dstr, const char *sstr, const I32 flags)
5070 PERL_ARGS_ASSERT_SV_CATPV_FLAGS;
5071 sv_catpvn_flags(dstr, sstr, strlen(sstr), flags);
5075 =for apidoc sv_catpv_mg
5077 Like C<sv_catpv>, but also handles 'set' magic.
5083 Perl_sv_catpv_mg(pTHX_ register SV *const sv, register const char *const ptr)
5085 PERL_ARGS_ASSERT_SV_CATPV_MG;
5094 Creates a new SV. A non-zero C<len> parameter indicates the number of
5095 bytes of preallocated string space the SV should have. An extra byte for a
5096 trailing NUL is also reserved. (SvPOK is not set for the SV even if string
5097 space is allocated.) The reference count for the new SV is set to 1.
5099 In 5.9.3, newSV() replaces the older NEWSV() API, and drops the first
5100 parameter, I<x>, a debug aid which allowed callers to identify themselves.
5101 This aid has been superseded by a new build option, PERL_MEM_LOG (see
5102 L<perlhacktips/PERL_MEM_LOG>). The older API is still there for use in XS
5103 modules supporting older perls.
5109 Perl_newSV(pTHX_ const STRLEN len)
5116 sv_upgrade(sv, SVt_PV);
5117 SvGROW(sv, len + 1);
5122 =for apidoc sv_magicext
5124 Adds magic to an SV, upgrading it if necessary. Applies the
5125 supplied vtable and returns a pointer to the magic added.
5127 Note that C<sv_magicext> will allow things that C<sv_magic> will not.
5128 In particular, you can add magic to SvREADONLY SVs, and add more than
5129 one instance of the same 'how'.
5131 If C<namlen> is greater than zero then a C<savepvn> I<copy> of C<name> is
5132 stored, if C<namlen> is zero then C<name> is stored as-is and - as another
5133 special case - if C<(name && namlen == HEf_SVKEY)> then C<name> is assumed
5134 to contain an C<SV*> and is stored as-is with its REFCNT incremented.
5136 (This is now used as a subroutine by C<sv_magic>.)
5141 Perl_sv_magicext(pTHX_ SV *const sv, SV *const obj, const int how,
5142 const MGVTBL *const vtable, const char *const name, const I32 namlen)
5147 PERL_ARGS_ASSERT_SV_MAGICEXT;
5149 SvUPGRADE(sv, SVt_PVMG);
5150 Newxz(mg, 1, MAGIC);
5151 mg->mg_moremagic = SvMAGIC(sv);
5152 SvMAGIC_set(sv, mg);
5154 /* Sometimes a magic contains a reference loop, where the sv and
5155 object refer to each other. To prevent a reference loop that
5156 would prevent such objects being freed, we look for such loops
5157 and if we find one we avoid incrementing the object refcount.
5159 Note we cannot do this to avoid self-tie loops as intervening RV must
5160 have its REFCNT incremented to keep it in existence.
5163 if (!obj || obj == sv ||
5164 how == PERL_MAGIC_arylen ||
5165 how == PERL_MAGIC_symtab ||
5166 (SvTYPE(obj) == SVt_PVGV &&
5167 (GvSV(obj) == sv || GvHV(obj) == (const HV *)sv
5168 || GvAV(obj) == (const AV *)sv || GvCV(obj) == (const CV *)sv
5169 || GvIOp(obj) == (const IO *)sv || GvFORM(obj) == (const CV *)sv)))
5174 mg->mg_obj = SvREFCNT_inc_simple(obj);
5175 mg->mg_flags |= MGf_REFCOUNTED;
5178 /* Normal self-ties simply pass a null object, and instead of
5179 using mg_obj directly, use the SvTIED_obj macro to produce a
5180 new RV as needed. For glob "self-ties", we are tieing the PVIO
5181 with an RV obj pointing to the glob containing the PVIO. In
5182 this case, to avoid a reference loop, we need to weaken the
5186 if (how == PERL_MAGIC_tiedscalar && SvTYPE(sv) == SVt_PVIO &&
5187 obj && SvROK(obj) && GvIO(SvRV(obj)) == (const IO *)sv)
5193 mg->mg_len = namlen;
5196 mg->mg_ptr = savepvn(name, namlen);
5197 else if (namlen == HEf_SVKEY) {
5198 /* Yes, this is casting away const. This is only for the case of
5199 HEf_SVKEY. I think we need to document this aberation of the
5200 constness of the API, rather than making name non-const, as
5201 that change propagating outwards a long way. */
5202 mg->mg_ptr = (char*)SvREFCNT_inc_simple_NN((SV *)name);
5204 mg->mg_ptr = (char *) name;
5206 mg->mg_virtual = (MGVTBL *) vtable;
5210 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
5215 =for apidoc sv_magic
5217 Adds magic to an SV. First upgrades C<sv> to type C<SVt_PVMG> if necessary,
5218 then adds a new magic item of type C<how> to the head of the magic list.
5220 See C<sv_magicext> (which C<sv_magic> now calls) for a description of the
5221 handling of the C<name> and C<namlen> arguments.
5223 You need to use C<sv_magicext> to add magic to SvREADONLY SVs and also
5224 to add more than one instance of the same 'how'.
5230 Perl_sv_magic(pTHX_ register SV *const sv, SV *const obj, const int how,
5231 const char *const name, const I32 namlen)
5234 const MGVTBL *vtable;
5237 unsigned int vtable_index;
5239 PERL_ARGS_ASSERT_SV_MAGIC;
5241 if (how < 0 || (unsigned)how > C_ARRAY_LENGTH(PL_magic_data)
5242 || ((flags = PL_magic_data[how]),
5243 (vtable_index = flags & PERL_MAGIC_VTABLE_MASK)
5244 > magic_vtable_max))
5245 Perl_croak(aTHX_ "Don't know how to handle magic of type \\%o", how);
5247 /* PERL_MAGIC_ext is reserved for use by extensions not perl internals.
5248 Useful for attaching extension internal data to perl vars.
5249 Note that multiple extensions may clash if magical scalars
5250 etc holding private data from one are passed to another. */
5252 vtable = (vtable_index == magic_vtable_max)
5253 ? NULL : PL_magic_vtables + vtable_index;
5255 #ifdef PERL_OLD_COPY_ON_WRITE
5257 sv_force_normal_flags(sv, 0);
5259 if (SvREADONLY(sv)) {
5261 /* its okay to attach magic to shared strings; the subsequent
5262 * upgrade to PVMG will unshare the string */
5263 !(SvFAKE(sv) && SvTYPE(sv) < SVt_PVMG)
5266 && !PERL_MAGIC_TYPE_READONLY_ACCEPTABLE(how)
5269 Perl_croak_no_modify(aTHX);
5272 if (SvMAGICAL(sv) || (how == PERL_MAGIC_taint && SvTYPE(sv) >= SVt_PVMG)) {
5273 if (SvMAGIC(sv) && (mg = mg_find(sv, how))) {
5274 /* sv_magic() refuses to add a magic of the same 'how' as an
5277 if (how == PERL_MAGIC_taint) {
5279 /* Any scalar which already had taint magic on which someone
5280 (erroneously?) did SvIOK_on() or similar will now be
5281 incorrectly sporting public "OK" flags. */
5282 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
5288 /* Rest of work is done else where */
5289 mg = sv_magicext(sv,obj,how,vtable,name,namlen);
5292 case PERL_MAGIC_taint:
5295 case PERL_MAGIC_ext:
5296 case PERL_MAGIC_dbfile:
5303 S_sv_unmagicext_flags(pTHX_ SV *const sv, const int type, MGVTBL *vtbl, const U32 flags)
5310 if (SvTYPE(sv) < SVt_PVMG || !SvMAGIC(sv))
5312 mgp = &(((XPVMG*) SvANY(sv))->xmg_u.xmg_magic);
5313 for (mg = *mgp; mg; mg = *mgp) {
5314 const MGVTBL* const virt = mg->mg_virtual;
5315 if (mg->mg_type == type && (!flags || virt == vtbl)) {
5316 *mgp = mg->mg_moremagic;
5317 if (virt && virt->svt_free)
5318 virt->svt_free(aTHX_ sv, mg);
5319 if (mg->mg_ptr && mg->mg_type != PERL_MAGIC_regex_global) {
5321 Safefree(mg->mg_ptr);
5322 else if (mg->mg_len == HEf_SVKEY)
5323 SvREFCNT_dec(MUTABLE_SV(mg->mg_ptr));
5324 else if (mg->mg_type == PERL_MAGIC_utf8)
5325 Safefree(mg->mg_ptr);
5327 if (mg->mg_flags & MGf_REFCOUNTED)
5328 SvREFCNT_dec(mg->mg_obj);
5332 mgp = &mg->mg_moremagic;
5335 if (SvMAGICAL(sv)) /* if we're under save_magic, wait for restore_magic; */
5336 mg_magical(sv); /* else fix the flags now */
5340 SvFLAGS(sv) |= (SvFLAGS(sv) & (SVp_IOK|SVp_NOK|SVp_POK)) >> PRIVSHIFT;
5346 =for apidoc sv_unmagic
5348 Removes all magic of type C<type> from an SV.
5354 Perl_sv_unmagic(pTHX_ SV *const sv, const int type)
5356 PERL_ARGS_ASSERT_SV_UNMAGIC;
5357 return S_sv_unmagicext_flags(aTHX_ sv, type, NULL, 0);
5361 =for apidoc sv_unmagicext
5363 Removes all magic of type C<type> with the specified C<vtbl> from an SV.
5369 Perl_sv_unmagicext(pTHX_ SV *const sv, const int type, MGVTBL *vtbl)
5371 PERL_ARGS_ASSERT_SV_UNMAGICEXT;
5372 return S_sv_unmagicext_flags(aTHX_ sv, type, vtbl, 1);
5376 =for apidoc sv_rvweaken
5378 Weaken a reference: set the C<SvWEAKREF> flag on this RV; give the
5379 referred-to SV C<PERL_MAGIC_backref> magic if it hasn't already; and
5380 push a back-reference to this RV onto the array of backreferences
5381 associated with that magic. If the RV is magical, set magic will be
5382 called after the RV is cleared.
5388 Perl_sv_rvweaken(pTHX_ SV *const sv)
5392 PERL_ARGS_ASSERT_SV_RVWEAKEN;
5394 if (!SvOK(sv)) /* let undefs pass */
5397 Perl_croak(aTHX_ "Can't weaken a nonreference");
5398 else if (SvWEAKREF(sv)) {
5399 Perl_ck_warner(aTHX_ packWARN(WARN_MISC), "Reference is already weak");
5403 Perl_sv_add_backref(aTHX_ tsv, sv);
5409 /* Give tsv backref magic if it hasn't already got it, then push a
5410 * back-reference to sv onto the array associated with the backref magic.
5412 * As an optimisation, if there's only one backref and it's not an AV,
5413 * store it directly in the HvAUX or mg_obj slot, avoiding the need to
5414 * allocate an AV. (Whether the slot holds an AV tells us whether this is
5418 /* A discussion about the backreferences array and its refcount:
5420 * The AV holding the backreferences is pointed to either as the mg_obj of
5421 * PERL_MAGIC_backref, or in the specific case of a HV, from the
5422 * xhv_backreferences field. The array is created with a refcount
5423 * of 2. This means that if during global destruction the array gets
5424 * picked on before its parent to have its refcount decremented by the
5425 * random zapper, it won't actually be freed, meaning it's still there for
5426 * when its parent gets freed.
5428 * When the parent SV is freed, the extra ref is killed by
5429 * Perl_sv_kill_backrefs. The other ref is killed, in the case of magic,
5430 * by mg_free() / MGf_REFCOUNTED, or for a hash, by Perl_hv_kill_backrefs.
5432 * When a single backref SV is stored directly, it is not reference
5437 Perl_sv_add_backref(pTHX_ SV *const tsv, SV *const sv)
5444 PERL_ARGS_ASSERT_SV_ADD_BACKREF;
5446 /* find slot to store array or singleton backref */
5448 if (SvTYPE(tsv) == SVt_PVHV) {
5449 svp = (SV**)Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5452 (SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL))))
5454 sv_magic(tsv, NULL, PERL_MAGIC_backref, NULL, 0);
5455 mg = mg_find(tsv, PERL_MAGIC_backref);
5457 svp = &(mg->mg_obj);
5460 /* create or retrieve the array */
5462 if ( (!*svp && SvTYPE(sv) == SVt_PVAV)
5463 || (*svp && SvTYPE(*svp) != SVt_PVAV)
5468 SvREFCNT_inc_simple_void(av);
5469 /* av now has a refcnt of 2; see discussion above */
5471 /* move single existing backref to the array */
5473 AvARRAY(av)[++AvFILLp(av)] = *svp; /* av_push() */
5477 mg->mg_flags |= MGf_REFCOUNTED;
5480 av = MUTABLE_AV(*svp);
5483 /* optimisation: store single backref directly in HvAUX or mg_obj */
5487 /* push new backref */
5488 assert(SvTYPE(av) == SVt_PVAV);
5489 if (AvFILLp(av) >= AvMAX(av)) {
5490 av_extend(av, AvFILLp(av)+1);
5492 AvARRAY(av)[++AvFILLp(av)] = sv; /* av_push() */
5495 /* delete a back-reference to ourselves from the backref magic associated
5496 * with the SV we point to.
5500 Perl_sv_del_backref(pTHX_ SV *const tsv, SV *const sv)
5505 PERL_ARGS_ASSERT_SV_DEL_BACKREF;
5507 if (SvTYPE(tsv) == SVt_PVHV) {
5509 svp = (SV**)Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5513 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
5514 svp = mg ? &(mg->mg_obj) : NULL;
5518 Perl_croak(aTHX_ "panic: del_backref");
5520 if (SvTYPE(*svp) == SVt_PVAV) {
5524 AV * const av = (AV*)*svp;
5526 assert(!SvIS_FREED(av));
5530 /* for an SV with N weak references to it, if all those
5531 * weak refs are deleted, then sv_del_backref will be called
5532 * N times and O(N^2) compares will be done within the backref
5533 * array. To ameliorate this potential slowness, we:
5534 * 1) make sure this code is as tight as possible;
5535 * 2) when looking for SV, look for it at both the head and tail of the
5536 * array first before searching the rest, since some create/destroy
5537 * patterns will cause the backrefs to be freed in order.
5544 SV **p = &svp[fill];
5545 SV *const topsv = *p;
5552 /* We weren't the last entry.
5553 An unordered list has this property that you
5554 can take the last element off the end to fill
5555 the hole, and it's still an unordered list :-)
5561 break; /* should only be one */
5568 AvFILLp(av) = fill-1;
5571 /* optimisation: only a single backref, stored directly */
5573 Perl_croak(aTHX_ "panic: del_backref");
5580 Perl_sv_kill_backrefs(pTHX_ SV *const sv, AV *const av)
5586 PERL_ARGS_ASSERT_SV_KILL_BACKREFS;
5591 /* after multiple passes through Perl_sv_clean_all() for a thinngy
5592 * that has badly leaked, the backref array may have gotten freed,
5593 * since we only protect it against 1 round of cleanup */
5594 if (SvIS_FREED(av)) {
5595 if (PL_in_clean_all) /* All is fair */
5598 "panic: magic_killbackrefs (freed backref AV/SV)");
5602 is_array = (SvTYPE(av) == SVt_PVAV);
5604 assert(!SvIS_FREED(av));
5607 last = svp + AvFILLp(av);
5610 /* optimisation: only a single backref, stored directly */
5616 while (svp <= last) {
5618 SV *const referrer = *svp;
5619 if (SvWEAKREF(referrer)) {
5620 /* XXX Should we check that it hasn't changed? */
5621 assert(SvROK(referrer));
5622 SvRV_set(referrer, 0);
5624 SvWEAKREF_off(referrer);
5625 SvSETMAGIC(referrer);
5626 } else if (SvTYPE(referrer) == SVt_PVGV ||
5627 SvTYPE(referrer) == SVt_PVLV) {
5628 assert(SvTYPE(sv) == SVt_PVHV); /* stash backref */
5629 /* You lookin' at me? */
5630 assert(GvSTASH(referrer));
5631 assert(GvSTASH(referrer) == (const HV *)sv);
5632 GvSTASH(referrer) = 0;
5633 } else if (SvTYPE(referrer) == SVt_PVCV ||
5634 SvTYPE(referrer) == SVt_PVFM) {
5635 if (SvTYPE(sv) == SVt_PVHV) { /* stash backref */
5636 /* You lookin' at me? */
5637 assert(CvSTASH(referrer));
5638 assert(CvSTASH(referrer) == (const HV *)sv);
5639 SvANY(MUTABLE_CV(referrer))->xcv_stash = 0;
5642 assert(SvTYPE(sv) == SVt_PVGV);
5643 /* You lookin' at me? */
5644 assert(CvGV(referrer));
5645 assert(CvGV(referrer) == (const GV *)sv);
5646 anonymise_cv_maybe(MUTABLE_GV(sv),
5647 MUTABLE_CV(referrer));
5652 "panic: magic_killbackrefs (flags=%"UVxf")",
5653 (UV)SvFLAGS(referrer));
5664 SvREFCNT_dec(av); /* remove extra count added by sv_add_backref() */
5670 =for apidoc sv_insert
5672 Inserts a string at the specified offset/length within the SV. Similar to
5673 the Perl substr() function. Handles get magic.
5675 =for apidoc sv_insert_flags
5677 Same as C<sv_insert>, but the extra C<flags> are passed the C<SvPV_force_flags> that applies to C<bigstr>.
5683 Perl_sv_insert_flags(pTHX_ SV *const bigstr, const STRLEN offset, const STRLEN len, const char *const little, const STRLEN littlelen, const U32 flags)
5688 register char *midend;
5689 register char *bigend;
5693 PERL_ARGS_ASSERT_SV_INSERT_FLAGS;
5696 Perl_croak(aTHX_ "Can't modify non-existent substring");
5697 SvPV_force_flags(bigstr, curlen, flags);
5698 (void)SvPOK_only_UTF8(bigstr);
5699 if (offset + len > curlen) {
5700 SvGROW(bigstr, offset+len+1);
5701 Zero(SvPVX(bigstr)+curlen, offset+len-curlen, char);
5702 SvCUR_set(bigstr, offset+len);
5706 i = littlelen - len;
5707 if (i > 0) { /* string might grow */
5708 big = SvGROW(bigstr, SvCUR(bigstr) + i + 1);
5709 mid = big + offset + len;
5710 midend = bigend = big + SvCUR(bigstr);
5713 while (midend > mid) /* shove everything down */
5714 *--bigend = *--midend;
5715 Move(little,big+offset,littlelen,char);
5716 SvCUR_set(bigstr, SvCUR(bigstr) + i);
5721 Move(little,SvPVX(bigstr)+offset,len,char);
5726 big = SvPVX(bigstr);
5729 bigend = big + SvCUR(bigstr);
5731 if (midend > bigend)
5732 Perl_croak(aTHX_ "panic: sv_insert");
5734 if (mid - big > bigend - midend) { /* faster to shorten from end */
5736 Move(little, mid, littlelen,char);
5739 i = bigend - midend;
5741 Move(midend, mid, i,char);
5745 SvCUR_set(bigstr, mid - big);
5747 else if ((i = mid - big)) { /* faster from front */
5748 midend -= littlelen;
5750 Move(big, midend - i, i, char);
5751 sv_chop(bigstr,midend-i);
5753 Move(little, mid, littlelen,char);
5755 else if (littlelen) {
5756 midend -= littlelen;
5757 sv_chop(bigstr,midend);
5758 Move(little,midend,littlelen,char);
5761 sv_chop(bigstr,midend);
5767 =for apidoc sv_replace
5769 Make the first argument a copy of the second, then delete the original.
5770 The target SV physically takes over ownership of the body of the source SV
5771 and inherits its flags; however, the target keeps any magic it owns,
5772 and any magic in the source is discarded.
5773 Note that this is a rather specialist SV copying operation; most of the
5774 time you'll want to use C<sv_setsv> or one of its many macro front-ends.
5780 Perl_sv_replace(pTHX_ register SV *const sv, register SV *const nsv)
5783 const U32 refcnt = SvREFCNT(sv);
5785 PERL_ARGS_ASSERT_SV_REPLACE;
5787 SV_CHECK_THINKFIRST_COW_DROP(sv);
5788 if (SvREFCNT(nsv) != 1) {
5789 Perl_croak(aTHX_ "panic: reference miscount on nsv in sv_replace()"
5790 " (%" UVuf " != 1)", (UV) SvREFCNT(nsv));
5792 if (SvMAGICAL(sv)) {
5796 sv_upgrade(nsv, SVt_PVMG);
5797 SvMAGIC_set(nsv, SvMAGIC(sv));
5798 SvFLAGS(nsv) |= SvMAGICAL(sv);
5800 SvMAGIC_set(sv, NULL);
5804 assert(!SvREFCNT(sv));
5805 #ifdef DEBUG_LEAKING_SCALARS
5806 sv->sv_flags = nsv->sv_flags;
5807 sv->sv_any = nsv->sv_any;
5808 sv->sv_refcnt = nsv->sv_refcnt;
5809 sv->sv_u = nsv->sv_u;
5811 StructCopy(nsv,sv,SV);
5813 if(SvTYPE(sv) == SVt_IV) {
5815 = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
5819 #ifdef PERL_OLD_COPY_ON_WRITE
5820 if (SvIsCOW_normal(nsv)) {
5821 /* We need to follow the pointers around the loop to make the
5822 previous SV point to sv, rather than nsv. */
5825 while ((next = SV_COW_NEXT_SV(current)) != nsv) {
5828 assert(SvPVX_const(current) == SvPVX_const(nsv));
5830 /* Make the SV before us point to the SV after us. */
5832 PerlIO_printf(Perl_debug_log, "previous is\n");
5834 PerlIO_printf(Perl_debug_log,
5835 "move it from 0x%"UVxf" to 0x%"UVxf"\n",
5836 (UV) SV_COW_NEXT_SV(current), (UV) sv);
5838 SV_COW_NEXT_SV_SET(current, sv);
5841 SvREFCNT(sv) = refcnt;
5842 SvFLAGS(nsv) |= SVTYPEMASK; /* Mark as freed */
5847 /* We're about to free a GV which has a CV that refers back to us.
5848 * If that CV will outlive us, make it anonymous (i.e. fix up its CvGV
5852 S_anonymise_cv_maybe(pTHX_ GV *gv, CV* cv)
5858 PERL_ARGS_ASSERT_ANONYMISE_CV_MAYBE;
5861 assert(SvREFCNT(gv) == 0);
5862 assert(isGV(gv) && isGV_with_GP(gv));
5864 assert(!CvANON(cv));
5865 assert(CvGV(cv) == gv);
5867 /* will the CV shortly be freed by gp_free() ? */
5868 if (GvCV(gv) == cv && GvGP(gv)->gp_refcnt < 2 && SvREFCNT(cv) < 2) {
5869 SvANY(cv)->xcv_gv = NULL;
5873 /* if not, anonymise: */
5874 stash = GvSTASH(gv) && HvNAME(GvSTASH(gv))
5875 ? HvENAME(GvSTASH(gv)) : NULL;
5876 gvname = Perl_newSVpvf(aTHX_ "%s::__ANON__",
5877 stash ? stash : "__ANON__");
5878 anongv = gv_fetchsv(gvname, GV_ADDMULTI, SVt_PVCV);
5879 SvREFCNT_dec(gvname);
5883 SvANY(cv)->xcv_gv = MUTABLE_GV(SvREFCNT_inc(anongv));
5888 =for apidoc sv_clear
5890 Clear an SV: call any destructors, free up any memory used by the body,
5891 and free the body itself. The SV's head is I<not> freed, although
5892 its type is set to all 1's so that it won't inadvertently be assumed
5893 to be live during global destruction etc.
5894 This function should only be called when REFCNT is zero. Most of the time
5895 you'll want to call C<sv_free()> (or its macro wrapper C<SvREFCNT_dec>)
5902 Perl_sv_clear(pTHX_ SV *const orig_sv)
5907 const struct body_details *sv_type_details;
5910 register SV *sv = orig_sv;
5913 PERL_ARGS_ASSERT_SV_CLEAR;
5915 /* within this loop, sv is the SV currently being freed, and
5916 * iter_sv is the most recent AV or whatever that's being iterated
5917 * over to provide more SVs */
5923 assert(SvREFCNT(sv) == 0);
5924 assert(SvTYPE(sv) != (svtype)SVTYPEMASK);
5926 if (type <= SVt_IV) {
5927 /* See the comment in sv.h about the collusion between this
5928 * early return and the overloading of the NULL slots in the
5932 SvFLAGS(sv) &= SVf_BREAK;
5933 SvFLAGS(sv) |= SVTYPEMASK;
5937 assert(!SvOBJECT(sv) || type >= SVt_PVMG); /* objs are always >= MG */
5939 if (type >= SVt_PVMG) {
5941 if (!curse(sv, 1)) goto get_next_sv;
5942 type = SvTYPE(sv); /* destructor may have changed it */
5944 /* Free back-references before magic, in case the magic calls
5945 * Perl code that has weak references to sv. */
5946 if (type == SVt_PVHV) {
5947 Perl_hv_kill_backrefs(aTHX_ MUTABLE_HV(sv));
5951 else if (type == SVt_PVMG && SvPAD_OUR(sv)) {
5952 SvREFCNT_dec(SvOURSTASH(sv));
5953 } else if (SvMAGIC(sv)) {
5954 /* Free back-references before other types of magic. */
5955 sv_unmagic(sv, PERL_MAGIC_backref);
5958 if (type == SVt_PVMG && SvPAD_TYPED(sv))
5959 SvREFCNT_dec(SvSTASH(sv));
5962 /* case SVt_BIND: */
5965 IoIFP(sv) != PerlIO_stdin() &&
5966 IoIFP(sv) != PerlIO_stdout() &&
5967 IoIFP(sv) != PerlIO_stderr() &&
5968 !(IoFLAGS(sv) & IOf_FAKE_DIRP))
5970 io_close(MUTABLE_IO(sv), FALSE);
5972 if (IoDIRP(sv) && !(IoFLAGS(sv) & IOf_FAKE_DIRP))
5973 PerlDir_close(IoDIRP(sv));
5974 IoDIRP(sv) = (DIR*)NULL;
5975 Safefree(IoTOP_NAME(sv));
5976 Safefree(IoFMT_NAME(sv));
5977 Safefree(IoBOTTOM_NAME(sv));
5980 /* FIXME for plugins */
5981 pregfree2((REGEXP*) sv);
5985 cv_undef(MUTABLE_CV(sv));
5986 /* If we're in a stash, we don't own a reference to it.
5987 * However it does have a back reference to us, which needs to
5989 if ((stash = CvSTASH(sv)))
5990 sv_del_backref(MUTABLE_SV(stash), sv);
5993 if (PL_last_swash_hv == (const HV *)sv) {
5994 PL_last_swash_hv = NULL;
5996 if (HvTOTALKEYS((HV*)sv) > 0) {
5998 /* this statement should match the one at the beginning of
5999 * hv_undef_flags() */
6000 if ( PL_phase != PERL_PHASE_DESTRUCT
6001 && (name = HvNAME((HV*)sv)))
6004 (void)hv_delete(PL_stashcache, name,
6005 HvNAMELEN_get((HV*)sv), G_DISCARD);
6006 hv_name_set((HV*)sv, NULL, 0, 0);
6009 /* save old iter_sv in unused SvSTASH field */
6010 assert(!SvOBJECT(sv));
6011 SvSTASH(sv) = (HV*)iter_sv;
6014 /* XXX ideally we should save the old value of hash_index
6015 * too, but I can't think of any place to hide it. The
6016 * effect of not saving it is that for freeing hashes of
6017 * hashes, we become quadratic in scanning the HvARRAY of
6018 * the top hash looking for new entries to free; but
6019 * hopefully this will be dwarfed by the freeing of all
6020 * the nested hashes. */
6022 next_sv = Perl_hfree_next_entry(aTHX_ (HV*)sv, &hash_index);
6023 goto get_next_sv; /* process this new sv */
6025 /* free empty hash */
6026 Perl_hv_undef_flags(aTHX_ MUTABLE_HV(sv), HV_NAME_SETALL);
6027 assert(!HvARRAY((HV*)sv));
6031 AV* av = MUTABLE_AV(sv);
6032 if (PL_comppad == av) {
6036 if (AvREAL(av) && AvFILLp(av) > -1) {
6037 next_sv = AvARRAY(av)[AvFILLp(av)--];
6038 /* save old iter_sv in top-most slot of AV,
6039 * and pray that it doesn't get wiped in the meantime */
6040 AvARRAY(av)[AvMAX(av)] = iter_sv;
6042 goto get_next_sv; /* process this new sv */
6044 Safefree(AvALLOC(av));
6049 if (LvTYPE(sv) == 'T') { /* for tie: return HE to pool */
6050 SvREFCNT_dec(HeKEY_sv((HE*)LvTARG(sv)));
6051 HeNEXT((HE*)LvTARG(sv)) = PL_hv_fetch_ent_mh;
6052 PL_hv_fetch_ent_mh = (HE*)LvTARG(sv);
6054 else if (LvTYPE(sv) != 't') /* unless tie: unrefcnted fake SV** */
6055 SvREFCNT_dec(LvTARG(sv));
6057 if (isGV_with_GP(sv)) {
6058 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
6059 && HvENAME_get(stash))
6060 mro_method_changed_in(stash);
6061 gp_free(MUTABLE_GV(sv));
6063 unshare_hek(GvNAME_HEK(sv));
6064 /* If we're in a stash, we don't own a reference to it.
6065 * However it does have a back reference to us, which
6066 * needs to be cleared. */
6067 if (!SvVALID(sv) && (stash = GvSTASH(sv)))
6068 sv_del_backref(MUTABLE_SV(stash), sv);
6070 /* FIXME. There are probably more unreferenced pointers to SVs
6071 * in the interpreter struct that we should check and tidy in
6072 * a similar fashion to this: */
6073 if ((const GV *)sv == PL_last_in_gv)
6074 PL_last_in_gv = NULL;
6080 /* Don't bother with SvOOK_off(sv); as we're only going to
6084 SvOOK_offset(sv, offset);
6085 SvPV_set(sv, SvPVX_mutable(sv) - offset);
6086 /* Don't even bother with turning off the OOK flag. */
6091 SV * const target = SvRV(sv);
6093 sv_del_backref(target, sv);
6098 #ifdef PERL_OLD_COPY_ON_WRITE
6099 else if (SvPVX_const(sv)
6100 && !(SvTYPE(sv) == SVt_PVIO
6101 && !(IoFLAGS(sv) & IOf_FAKE_DIRP)))
6105 PerlIO_printf(Perl_debug_log, "Copy on write: clear\n");
6109 sv_release_COW(sv, SvPVX_const(sv), SV_COW_NEXT_SV(sv));
6111 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
6115 } else if (SvLEN(sv)) {
6116 Safefree(SvPVX_const(sv));
6120 else if (SvPVX_const(sv) && SvLEN(sv)
6121 && !(SvTYPE(sv) == SVt_PVIO
6122 && !(IoFLAGS(sv) & IOf_FAKE_DIRP)))
6123 Safefree(SvPVX_mutable(sv));
6124 else if (SvPVX_const(sv) && SvREADONLY(sv) && SvFAKE(sv)) {
6125 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
6136 SvFLAGS(sv) &= SVf_BREAK;
6137 SvFLAGS(sv) |= SVTYPEMASK;
6139 sv_type_details = bodies_by_type + type;
6140 if (sv_type_details->arena) {
6141 del_body(((char *)SvANY(sv) + sv_type_details->offset),
6142 &PL_body_roots[type]);
6144 else if (sv_type_details->body_size) {
6145 safefree(SvANY(sv));
6149 /* caller is responsible for freeing the head of the original sv */
6150 if (sv != orig_sv && !SvREFCNT(sv))
6153 /* grab and free next sv, if any */
6161 else if (!iter_sv) {
6163 } else if (SvTYPE(iter_sv) == SVt_PVAV) {
6164 AV *const av = (AV*)iter_sv;
6165 if (AvFILLp(av) > -1) {
6166 sv = AvARRAY(av)[AvFILLp(av)--];
6168 else { /* no more elements of current AV to free */
6171 /* restore previous value, squirrelled away */
6172 iter_sv = AvARRAY(av)[AvMAX(av)];
6173 Safefree(AvALLOC(av));
6176 } else if (SvTYPE(iter_sv) == SVt_PVHV) {
6177 sv = Perl_hfree_next_entry(aTHX_ (HV*)iter_sv, &hash_index);
6178 if (!sv && !HvTOTALKEYS((HV *)iter_sv)) {
6179 /* no more elements of current HV to free */
6182 /* Restore previous value of iter_sv, squirrelled away */
6183 assert(!SvOBJECT(sv));
6184 iter_sv = (SV*)SvSTASH(sv);
6186 /* ideally we should restore the old hash_index here,
6187 * but we don't currently save the old value */
6190 /* free any remaining detritus from the hash struct */
6191 Perl_hv_undef_flags(aTHX_ MUTABLE_HV(sv), HV_NAME_SETALL);
6192 assert(!HvARRAY((HV*)sv));
6197 /* unrolled SvREFCNT_dec and sv_free2 follows: */
6201 if (!SvREFCNT(sv)) {
6205 if (--(SvREFCNT(sv)))
6209 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEBUGGING),
6210 "Attempt to free temp prematurely: SV 0x%"UVxf
6211 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
6215 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
6216 /* make sure SvREFCNT(sv)==0 happens very seldom */
6217 SvREFCNT(sv) = (~(U32)0)/2;
6226 /* This routine curses the sv itself, not the object referenced by sv. So
6227 sv does not have to be ROK. */
6230 S_curse(pTHX_ SV * const sv, const bool check_refcnt) {
6233 PERL_ARGS_ASSERT_CURSE;
6234 assert(SvOBJECT(sv));
6236 if (PL_defstash && /* Still have a symbol table? */
6243 stash = SvSTASH(sv);
6244 destructor = StashHANDLER(stash,DESTROY);
6246 /* A constant subroutine can have no side effects, so
6247 don't bother calling it. */
6248 && !CvCONST(destructor)
6249 /* Don't bother calling an empty destructor */
6250 && (CvISXSUB(destructor)
6251 || (CvSTART(destructor)
6252 && (CvSTART(destructor)->op_next->op_type
6255 SV* const tmpref = newRV(sv);
6256 SvREADONLY_on(tmpref); /* DESTROY() could be naughty */
6258 PUSHSTACKi(PERLSI_DESTROY);
6263 call_sv(MUTABLE_SV(destructor),
6264 G_DISCARD|G_EVAL|G_KEEPERR|G_VOID);
6268 if(SvREFCNT(tmpref) < 2) {
6269 /* tmpref is not kept alive! */
6271 SvRV_set(tmpref, NULL);
6274 SvREFCNT_dec(tmpref);
6276 } while (SvOBJECT(sv) && SvSTASH(sv) != stash);
6279 if (check_refcnt && SvREFCNT(sv)) {
6280 if (PL_in_clean_objs)
6282 "DESTROY created new reference to dead object '%s'",
6284 /* DESTROY gave object new lease on life */
6290 SvREFCNT_dec(SvSTASH(sv)); /* possibly of changed persuasion */
6291 SvOBJECT_off(sv); /* Curse the object. */
6292 if (SvTYPE(sv) != SVt_PVIO)
6293 --PL_sv_objcount;/* XXX Might want something more general */
6299 =for apidoc sv_newref
6301 Increment an SV's reference count. Use the C<SvREFCNT_inc()> wrapper
6308 Perl_sv_newref(pTHX_ SV *const sv)
6310 PERL_UNUSED_CONTEXT;
6319 Decrement an SV's reference count, and if it drops to zero, call
6320 C<sv_clear> to invoke destructors and free up any memory used by
6321 the body; finally, deallocate the SV's head itself.
6322 Normally called via a wrapper macro C<SvREFCNT_dec>.
6328 Perl_sv_free(pTHX_ SV *const sv)
6333 if (SvREFCNT(sv) == 0) {
6334 if (SvFLAGS(sv) & SVf_BREAK)
6335 /* this SV's refcnt has been artificially decremented to
6336 * trigger cleanup */
6338 if (PL_in_clean_all) /* All is fair */
6340 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
6341 /* make sure SvREFCNT(sv)==0 happens very seldom */
6342 SvREFCNT(sv) = (~(U32)0)/2;
6345 if (ckWARN_d(WARN_INTERNAL)) {
6346 #ifdef DEBUG_LEAKING_SCALARS_FORK_DUMP
6347 Perl_dump_sv_child(aTHX_ sv);
6349 #ifdef DEBUG_LEAKING_SCALARS
6352 #ifdef DEBUG_LEAKING_SCALARS_ABORT
6353 if (PL_warnhook == PERL_WARNHOOK_FATAL
6354 || ckDEAD(packWARN(WARN_INTERNAL))) {
6355 /* Don't let Perl_warner cause us to escape our fate: */
6359 /* This may not return: */
6360 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
6361 "Attempt to free unreferenced scalar: SV 0x%"UVxf
6362 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
6365 #ifdef DEBUG_LEAKING_SCALARS_ABORT
6370 if (--(SvREFCNT(sv)) > 0)
6372 Perl_sv_free2(aTHX_ sv);
6376 Perl_sv_free2(pTHX_ SV *const sv)
6380 PERL_ARGS_ASSERT_SV_FREE2;
6384 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEBUGGING),
6385 "Attempt to free temp prematurely: SV 0x%"UVxf
6386 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
6390 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
6391 /* make sure SvREFCNT(sv)==0 happens very seldom */
6392 SvREFCNT(sv) = (~(U32)0)/2;
6403 Returns the length of the string in the SV. Handles magic and type
6404 coercion. See also C<SvCUR>, which gives raw access to the xpv_cur slot.
6410 Perl_sv_len(pTHX_ register SV *const sv)
6418 len = mg_length(sv);
6420 (void)SvPV_const(sv, len);
6425 =for apidoc sv_len_utf8
6427 Returns the number of characters in the string in an SV, counting wide
6428 UTF-8 bytes as a single character. Handles magic and type coercion.
6434 * The length is cached in PERL_MAGIC_utf8, in the mg_len field. Also the
6435 * mg_ptr is used, by sv_pos_u2b() and sv_pos_b2u() - see the comments below.
6436 * (Note that the mg_len is not the length of the mg_ptr field.
6437 * This allows the cache to store the character length of the string without
6438 * needing to malloc() extra storage to attach to the mg_ptr.)
6443 Perl_sv_len_utf8(pTHX_ register SV *const sv)
6449 return mg_length(sv);
6453 const U8 *s = (U8*)SvPV_const(sv, len);
6457 MAGIC *mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_utf8) : NULL;
6459 if (mg && (mg->mg_len != -1 || mg->mg_ptr)) {
6460 if (mg->mg_len != -1)
6463 /* We can use the offset cache for a headstart.
6464 The longer value is stored in the first pair. */
6465 STRLEN *cache = (STRLEN *) mg->mg_ptr;
6467 ulen = cache[0] + Perl_utf8_length(aTHX_ s + cache[1],
6471 if (PL_utf8cache < 0) {
6472 const STRLEN real = Perl_utf8_length(aTHX_ s, s + len);
6473 assert_uft8_cache_coherent("sv_len_utf8", ulen, real, sv);
6477 ulen = Perl_utf8_length(aTHX_ s, s + len);
6478 utf8_mg_len_cache_update(sv, &mg, ulen);
6482 return Perl_utf8_length(aTHX_ s, s + len);
6486 /* Walk forwards to find the byte corresponding to the passed in UTF-8
6489 S_sv_pos_u2b_forwards(const U8 *const start, const U8 *const send,
6490 STRLEN *const uoffset_p, bool *const at_end)
6492 const U8 *s = start;
6493 STRLEN uoffset = *uoffset_p;
6495 PERL_ARGS_ASSERT_SV_POS_U2B_FORWARDS;
6497 while (s < send && uoffset) {
6504 else if (s > send) {
6506 /* This is the existing behaviour. Possibly it should be a croak, as
6507 it's actually a bounds error */
6510 *uoffset_p -= uoffset;
6514 /* Given the length of the string in both bytes and UTF-8 characters, decide
6515 whether to walk forwards or backwards to find the byte corresponding to
6516 the passed in UTF-8 offset. */
6518 S_sv_pos_u2b_midway(const U8 *const start, const U8 *send,
6519 STRLEN uoffset, const STRLEN uend)
6521 STRLEN backw = uend - uoffset;
6523 PERL_ARGS_ASSERT_SV_POS_U2B_MIDWAY;
6525 if (uoffset < 2 * backw) {
6526 /* The assumption is that going forwards is twice the speed of going
6527 forward (that's where the 2 * backw comes from).
6528 (The real figure of course depends on the UTF-8 data.) */
6529 const U8 *s = start;
6531 while (s < send && uoffset--)
6541 while (UTF8_IS_CONTINUATION(*send))
6544 return send - start;
6547 /* For the string representation of the given scalar, find the byte
6548 corresponding to the passed in UTF-8 offset. uoffset0 and boffset0
6549 give another position in the string, *before* the sought offset, which
6550 (which is always true, as 0, 0 is a valid pair of positions), which should
6551 help reduce the amount of linear searching.
6552 If *mgp is non-NULL, it should point to the UTF-8 cache magic, which
6553 will be used to reduce the amount of linear searching. The cache will be
6554 created if necessary, and the found value offered to it for update. */
6556 S_sv_pos_u2b_cached(pTHX_ SV *const sv, MAGIC **const mgp, const U8 *const start,
6557 const U8 *const send, STRLEN uoffset,
6558 STRLEN uoffset0, STRLEN boffset0)
6560 STRLEN boffset = 0; /* Actually always set, but let's keep gcc happy. */
6562 bool at_end = FALSE;
6564 PERL_ARGS_ASSERT_SV_POS_U2B_CACHED;
6566 assert (uoffset >= uoffset0);
6573 && (*mgp || (SvTYPE(sv) >= SVt_PVMG &&
6574 (*mgp = mg_find(sv, PERL_MAGIC_utf8))))) {
6575 if ((*mgp)->mg_ptr) {
6576 STRLEN *cache = (STRLEN *) (*mgp)->mg_ptr;
6577 if (cache[0] == uoffset) {
6578 /* An exact match. */
6581 if (cache[2] == uoffset) {
6582 /* An exact match. */
6586 if (cache[0] < uoffset) {
6587 /* The cache already knows part of the way. */
6588 if (cache[0] > uoffset0) {
6589 /* The cache knows more than the passed in pair */
6590 uoffset0 = cache[0];
6591 boffset0 = cache[1];
6593 if ((*mgp)->mg_len != -1) {
6594 /* And we know the end too. */
6596 + sv_pos_u2b_midway(start + boffset0, send,
6598 (*mgp)->mg_len - uoffset0);
6600 uoffset -= uoffset0;
6602 + sv_pos_u2b_forwards(start + boffset0,
6603 send, &uoffset, &at_end);
6604 uoffset += uoffset0;
6607 else if (cache[2] < uoffset) {
6608 /* We're between the two cache entries. */
6609 if (cache[2] > uoffset0) {
6610 /* and the cache knows more than the passed in pair */
6611 uoffset0 = cache[2];
6612 boffset0 = cache[3];
6616 + sv_pos_u2b_midway(start + boffset0,
6619 cache[0] - uoffset0);
6622 + sv_pos_u2b_midway(start + boffset0,
6625 cache[2] - uoffset0);
6629 else if ((*mgp)->mg_len != -1) {
6630 /* If we can take advantage of a passed in offset, do so. */
6631 /* In fact, offset0 is either 0, or less than offset, so don't
6632 need to worry about the other possibility. */
6634 + sv_pos_u2b_midway(start + boffset0, send,
6636 (*mgp)->mg_len - uoffset0);
6641 if (!found || PL_utf8cache < 0) {
6642 STRLEN real_boffset;
6643 uoffset -= uoffset0;
6644 real_boffset = boffset0 + sv_pos_u2b_forwards(start + boffset0,
6645 send, &uoffset, &at_end);
6646 uoffset += uoffset0;
6648 if (found && PL_utf8cache < 0)
6649 assert_uft8_cache_coherent("sv_pos_u2b_cache", boffset,
6651 boffset = real_boffset;
6656 utf8_mg_len_cache_update(sv, mgp, uoffset);
6658 utf8_mg_pos_cache_update(sv, mgp, boffset, uoffset, send - start);
6665 =for apidoc sv_pos_u2b_flags
6667 Converts the value pointed to by offsetp from a count of UTF-8 chars from
6668 the start of the string, to a count of the equivalent number of bytes; if
6669 lenp is non-zero, it does the same to lenp, but this time starting from
6670 the offset, rather than from the start of the string. Handles type coercion.
6671 I<flags> is passed to C<SvPV_flags>, and usually should be
6672 C<SV_GMAGIC|SV_CONST_RETURN> to handle magic.
6678 * sv_pos_u2b_flags() uses, like sv_pos_b2u(), the mg_ptr of the potential
6679 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6680 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
6685 Perl_sv_pos_u2b_flags(pTHX_ SV *const sv, STRLEN uoffset, STRLEN *const lenp,
6692 PERL_ARGS_ASSERT_SV_POS_U2B_FLAGS;
6694 start = (U8*)SvPV_flags(sv, len, flags);
6696 const U8 * const send = start + len;
6698 boffset = sv_pos_u2b_cached(sv, &mg, start, send, uoffset, 0, 0);
6701 && *lenp /* don't bother doing work for 0, as its bytes equivalent
6702 is 0, and *lenp is already set to that. */) {
6703 /* Convert the relative offset to absolute. */
6704 const STRLEN uoffset2 = uoffset + *lenp;
6705 const STRLEN boffset2
6706 = sv_pos_u2b_cached(sv, &mg, start, send, uoffset2,
6707 uoffset, boffset) - boffset;
6721 =for apidoc sv_pos_u2b
6723 Converts the value pointed to by offsetp from a count of UTF-8 chars from
6724 the start of the string, to a count of the equivalent number of bytes; if
6725 lenp is non-zero, it does the same to lenp, but this time starting from
6726 the offset, rather than from the start of the string. Handles magic and
6729 Use C<sv_pos_u2b_flags> in preference, which correctly handles strings longer
6736 * sv_pos_u2b() uses, like sv_pos_b2u(), the mg_ptr of the potential
6737 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6738 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
6742 /* This function is subject to size and sign problems */
6745 Perl_sv_pos_u2b(pTHX_ register SV *const sv, I32 *const offsetp, I32 *const lenp)
6747 PERL_ARGS_ASSERT_SV_POS_U2B;
6750 STRLEN ulen = (STRLEN)*lenp;
6751 *offsetp = (I32)sv_pos_u2b_flags(sv, (STRLEN)*offsetp, &ulen,
6752 SV_GMAGIC|SV_CONST_RETURN);
6755 *offsetp = (I32)sv_pos_u2b_flags(sv, (STRLEN)*offsetp, NULL,
6756 SV_GMAGIC|SV_CONST_RETURN);
6761 S_utf8_mg_len_cache_update(pTHX_ SV *const sv, MAGIC **const mgp,
6764 PERL_ARGS_ASSERT_UTF8_MG_LEN_CACHE_UPDATE;
6768 if (!*mgp && (SvTYPE(sv) < SVt_PVMG ||
6769 !(*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
6770 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, &PL_vtbl_utf8, 0, 0);
6774 (*mgp)->mg_len = ulen;
6775 /* For now, treat "overflowed" as "still unknown". See RT #72924. */
6776 if (ulen != (STRLEN) (*mgp)->mg_len)
6777 (*mgp)->mg_len = -1;
6780 /* Create and update the UTF8 magic offset cache, with the proffered utf8/
6781 byte length pairing. The (byte) length of the total SV is passed in too,
6782 as blen, because for some (more esoteric) SVs, the call to SvPV_const()
6783 may not have updated SvCUR, so we can't rely on reading it directly.
6785 The proffered utf8/byte length pairing isn't used if the cache already has
6786 two pairs, and swapping either for the proffered pair would increase the
6787 RMS of the intervals between known byte offsets.
6789 The cache itself consists of 4 STRLEN values
6790 0: larger UTF-8 offset
6791 1: corresponding byte offset
6792 2: smaller UTF-8 offset
6793 3: corresponding byte offset
6795 Unused cache pairs have the value 0, 0.
6796 Keeping the cache "backwards" means that the invariant of
6797 cache[0] >= cache[2] is maintained even with empty slots, which means that
6798 the code that uses it doesn't need to worry if only 1 entry has actually
6799 been set to non-zero. It also makes the "position beyond the end of the
6800 cache" logic much simpler, as the first slot is always the one to start
6804 S_utf8_mg_pos_cache_update(pTHX_ SV *const sv, MAGIC **const mgp, const STRLEN byte,
6805 const STRLEN utf8, const STRLEN blen)
6809 PERL_ARGS_ASSERT_UTF8_MG_POS_CACHE_UPDATE;
6814 if (!*mgp && (SvTYPE(sv) < SVt_PVMG ||
6815 !(*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
6816 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, (MGVTBL*)&PL_vtbl_utf8, 0,
6818 (*mgp)->mg_len = -1;
6822 if (!(cache = (STRLEN *)(*mgp)->mg_ptr)) {
6823 Newxz(cache, PERL_MAGIC_UTF8_CACHESIZE * 2, STRLEN);
6824 (*mgp)->mg_ptr = (char *) cache;
6828 if (PL_utf8cache < 0 && SvPOKp(sv)) {
6829 /* SvPOKp() because it's possible that sv has string overloading, and
6830 therefore is a reference, hence SvPVX() is actually a pointer.
6831 This cures the (very real) symptoms of RT 69422, but I'm not actually
6832 sure whether we should even be caching the results of UTF-8
6833 operations on overloading, given that nothing stops overloading
6834 returning a different value every time it's called. */
6835 const U8 *start = (const U8 *) SvPVX_const(sv);
6836 const STRLEN realutf8 = utf8_length(start, start + byte);
6838 assert_uft8_cache_coherent("utf8_mg_pos_cache_update", utf8, realutf8,
6842 /* Cache is held with the later position first, to simplify the code
6843 that deals with unbounded ends. */
6845 ASSERT_UTF8_CACHE(cache);
6846 if (cache[1] == 0) {
6847 /* Cache is totally empty */
6850 } else if (cache[3] == 0) {
6851 if (byte > cache[1]) {
6852 /* New one is larger, so goes first. */
6853 cache[2] = cache[0];
6854 cache[3] = cache[1];
6862 #define THREEWAY_SQUARE(a,b,c,d) \
6863 ((float)((d) - (c))) * ((float)((d) - (c))) \
6864 + ((float)((c) - (b))) * ((float)((c) - (b))) \
6865 + ((float)((b) - (a))) * ((float)((b) - (a)))
6867 /* Cache has 2 slots in use, and we know three potential pairs.
6868 Keep the two that give the lowest RMS distance. Do the
6869 calculation in bytes simply because we always know the byte
6870 length. squareroot has the same ordering as the positive value,
6871 so don't bother with the actual square root. */
6872 const float existing = THREEWAY_SQUARE(0, cache[3], cache[1], blen);
6873 if (byte > cache[1]) {
6874 /* New position is after the existing pair of pairs. */
6875 const float keep_earlier
6876 = THREEWAY_SQUARE(0, cache[3], byte, blen);
6877 const float keep_later
6878 = THREEWAY_SQUARE(0, cache[1], byte, blen);
6880 if (keep_later < keep_earlier) {
6881 if (keep_later < existing) {
6882 cache[2] = cache[0];
6883 cache[3] = cache[1];
6889 if (keep_earlier < existing) {
6895 else if (byte > cache[3]) {
6896 /* New position is between the existing pair of pairs. */
6897 const float keep_earlier
6898 = THREEWAY_SQUARE(0, cache[3], byte, blen);
6899 const float keep_later
6900 = THREEWAY_SQUARE(0, byte, cache[1], blen);
6902 if (keep_later < keep_earlier) {
6903 if (keep_later < existing) {
6909 if (keep_earlier < existing) {
6916 /* New position is before the existing pair of pairs. */
6917 const float keep_earlier
6918 = THREEWAY_SQUARE(0, byte, cache[3], blen);
6919 const float keep_later
6920 = THREEWAY_SQUARE(0, byte, cache[1], blen);
6922 if (keep_later < keep_earlier) {
6923 if (keep_later < existing) {
6929 if (keep_earlier < existing) {
6930 cache[0] = cache[2];
6931 cache[1] = cache[3];
6938 ASSERT_UTF8_CACHE(cache);
6941 /* We already know all of the way, now we may be able to walk back. The same
6942 assumption is made as in S_sv_pos_u2b_midway(), namely that walking
6943 backward is half the speed of walking forward. */
6945 S_sv_pos_b2u_midway(pTHX_ const U8 *const s, const U8 *const target,
6946 const U8 *end, STRLEN endu)
6948 const STRLEN forw = target - s;
6949 STRLEN backw = end - target;
6951 PERL_ARGS_ASSERT_SV_POS_B2U_MIDWAY;
6953 if (forw < 2 * backw) {
6954 return utf8_length(s, target);
6957 while (end > target) {
6959 while (UTF8_IS_CONTINUATION(*end)) {
6968 =for apidoc sv_pos_b2u
6970 Converts the value pointed to by offsetp from a count of bytes from the
6971 start of the string, to a count of the equivalent number of UTF-8 chars.
6972 Handles magic and type coercion.
6978 * sv_pos_b2u() uses, like sv_pos_u2b(), the mg_ptr of the potential
6979 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6984 Perl_sv_pos_b2u(pTHX_ register SV *const sv, I32 *const offsetp)
6987 const STRLEN byte = *offsetp;
6988 STRLEN len = 0; /* Actually always set, but let's keep gcc happy. */
6994 PERL_ARGS_ASSERT_SV_POS_B2U;
6999 s = (const U8*)SvPV_const(sv, blen);
7002 Perl_croak(aTHX_ "panic: sv_pos_b2u: bad byte offset");
7008 && SvTYPE(sv) >= SVt_PVMG
7009 && (mg = mg_find(sv, PERL_MAGIC_utf8)))
7012 STRLEN * const cache = (STRLEN *) mg->mg_ptr;
7013 if (cache[1] == byte) {
7014 /* An exact match. */
7015 *offsetp = cache[0];
7018 if (cache[3] == byte) {
7019 /* An exact match. */
7020 *offsetp = cache[2];
7024 if (cache[1] < byte) {
7025 /* We already know part of the way. */
7026 if (mg->mg_len != -1) {
7027 /* Actually, we know the end too. */
7029 + S_sv_pos_b2u_midway(aTHX_ s + cache[1], send,
7030 s + blen, mg->mg_len - cache[0]);
7032 len = cache[0] + utf8_length(s + cache[1], send);
7035 else if (cache[3] < byte) {
7036 /* We're between the two cached pairs, so we do the calculation
7037 offset by the byte/utf-8 positions for the earlier pair,
7038 then add the utf-8 characters from the string start to
7040 len = S_sv_pos_b2u_midway(aTHX_ s + cache[3], send,
7041 s + cache[1], cache[0] - cache[2])
7045 else { /* cache[3] > byte */
7046 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + cache[3],
7050 ASSERT_UTF8_CACHE(cache);
7052 } else if (mg->mg_len != -1) {
7053 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + blen, mg->mg_len);
7057 if (!found || PL_utf8cache < 0) {
7058 const STRLEN real_len = utf8_length(s, send);
7060 if (found && PL_utf8cache < 0)
7061 assert_uft8_cache_coherent("sv_pos_b2u", len, real_len, sv);
7068 utf8_mg_len_cache_update(sv, &mg, len);
7070 utf8_mg_pos_cache_update(sv, &mg, byte, len, blen);
7075 S_assert_uft8_cache_coherent(pTHX_ const char *const func, STRLEN from_cache,
7076 STRLEN real, SV *const sv)
7078 PERL_ARGS_ASSERT_ASSERT_UFT8_CACHE_COHERENT;
7080 /* As this is debugging only code, save space by keeping this test here,
7081 rather than inlining it in all the callers. */
7082 if (from_cache == real)
7085 /* Need to turn the assertions off otherwise we may recurse infinitely
7086 while printing error messages. */
7087 SAVEI8(PL_utf8cache);
7089 Perl_croak(aTHX_ "panic: %s cache %"UVuf" real %"UVuf" for %"SVf,
7090 func, (UV) from_cache, (UV) real, SVfARG(sv));
7096 Returns a boolean indicating whether the strings in the two SVs are
7097 identical. Is UTF-8 and 'use bytes' aware, handles get magic, and will
7098 coerce its args to strings if necessary.
7100 =for apidoc sv_eq_flags
7102 Returns a boolean indicating whether the strings in the two SVs are
7103 identical. Is UTF-8 and 'use bytes' aware and coerces its args to strings
7104 if necessary. If the flags include SV_GMAGIC, it handles get-magic, too.
7110 Perl_sv_eq_flags(pTHX_ register SV *sv1, register SV *sv2, const U32 flags)
7119 SV* svrecode = NULL;
7126 /* if pv1 and pv2 are the same, second SvPV_const call may
7127 * invalidate pv1 (if we are handling magic), so we may need to
7129 if (sv1 == sv2 && flags & SV_GMAGIC
7130 && (SvTHINKFIRST(sv1) || SvGMAGICAL(sv1))) {
7131 pv1 = SvPV_const(sv1, cur1);
7132 sv1 = newSVpvn_flags(pv1, cur1, SVs_TEMP | SvUTF8(sv2));
7134 pv1 = SvPV_flags_const(sv1, cur1, flags);
7142 pv2 = SvPV_flags_const(sv2, cur2, flags);
7144 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
7145 /* Differing utf8ness.
7146 * Do not UTF8size the comparands as a side-effect. */
7149 svrecode = newSVpvn(pv2, cur2);
7150 sv_recode_to_utf8(svrecode, PL_encoding);
7151 pv2 = SvPV_const(svrecode, cur2);
7154 svrecode = newSVpvn(pv1, cur1);
7155 sv_recode_to_utf8(svrecode, PL_encoding);
7156 pv1 = SvPV_const(svrecode, cur1);
7158 /* Now both are in UTF-8. */
7160 SvREFCNT_dec(svrecode);
7166 /* sv1 is the UTF-8 one */
7167 return bytes_cmp_utf8((const U8*)pv2, cur2,
7168 (const U8*)pv1, cur1) == 0;
7171 /* sv2 is the UTF-8 one */
7172 return bytes_cmp_utf8((const U8*)pv1, cur1,
7173 (const U8*)pv2, cur2) == 0;
7179 eq = (pv1 == pv2) || memEQ(pv1, pv2, cur1);
7181 SvREFCNT_dec(svrecode);
7191 Compares the strings in two SVs. Returns -1, 0, or 1 indicating whether the
7192 string in C<sv1> is less than, equal to, or greater than the string in
7193 C<sv2>. Is UTF-8 and 'use bytes' aware, handles get magic, and will
7194 coerce its args to strings if necessary. See also C<sv_cmp_locale>.
7196 =for apidoc sv_cmp_flags
7198 Compares the strings in two SVs. Returns -1, 0, or 1 indicating whether the
7199 string in C<sv1> is less than, equal to, or greater than the string in
7200 C<sv2>. Is UTF-8 and 'use bytes' aware and will coerce its args to strings
7201 if necessary. If the flags include SV_GMAGIC, it handles get magic. See
7202 also C<sv_cmp_locale_flags>.
7208 Perl_sv_cmp(pTHX_ register SV *const sv1, register SV *const sv2)
7210 return sv_cmp_flags(sv1, sv2, SV_GMAGIC);
7214 Perl_sv_cmp_flags(pTHX_ register SV *const sv1, register SV *const sv2,
7219 const char *pv1, *pv2;
7222 SV *svrecode = NULL;
7229 pv1 = SvPV_flags_const(sv1, cur1, flags);
7236 pv2 = SvPV_flags_const(sv2, cur2, flags);
7238 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
7239 /* Differing utf8ness.
7240 * Do not UTF8size the comparands as a side-effect. */
7243 svrecode = newSVpvn(pv2, cur2);
7244 sv_recode_to_utf8(svrecode, PL_encoding);
7245 pv2 = SvPV_const(svrecode, cur2);
7248 const int retval = -bytes_cmp_utf8((const U8*)pv2, cur2,
7249 (const U8*)pv1, cur1);
7250 return retval ? retval < 0 ? -1 : +1 : 0;
7255 svrecode = newSVpvn(pv1, cur1);
7256 sv_recode_to_utf8(svrecode, PL_encoding);
7257 pv1 = SvPV_const(svrecode, cur1);
7260 const int retval = bytes_cmp_utf8((const U8*)pv1, cur1,
7261 (const U8*)pv2, cur2);
7262 return retval ? retval < 0 ? -1 : +1 : 0;
7268 cmp = cur2 ? -1 : 0;
7272 const I32 retval = memcmp((const void*)pv1, (const void*)pv2, cur1 < cur2 ? cur1 : cur2);
7275 cmp = retval < 0 ? -1 : 1;
7276 } else if (cur1 == cur2) {
7279 cmp = cur1 < cur2 ? -1 : 1;
7283 SvREFCNT_dec(svrecode);
7291 =for apidoc sv_cmp_locale
7293 Compares the strings in two SVs in a locale-aware manner. Is UTF-8 and
7294 'use bytes' aware, handles get magic, and will coerce its args to strings
7295 if necessary. See also C<sv_cmp>.
7297 =for apidoc sv_cmp_locale_flags
7299 Compares the strings in two SVs in a locale-aware manner. Is UTF-8 and
7300 'use bytes' aware and will coerce its args to strings if necessary. If the
7301 flags contain SV_GMAGIC, it handles get magic. See also C<sv_cmp_flags>.
7307 Perl_sv_cmp_locale(pTHX_ register SV *const sv1, register SV *const sv2)
7309 return sv_cmp_locale_flags(sv1, sv2, SV_GMAGIC);
7313 Perl_sv_cmp_locale_flags(pTHX_ register SV *const sv1, register SV *const sv2,
7317 #ifdef USE_LOCALE_COLLATE
7323 if (PL_collation_standard)
7327 pv1 = sv1 ? sv_collxfrm_flags(sv1, &len1, flags) : (char *) NULL;
7329 pv2 = sv2 ? sv_collxfrm_flags(sv2, &len2, flags) : (char *) NULL;
7331 if (!pv1 || !len1) {
7342 retval = memcmp((void*)pv1, (void*)pv2, len1 < len2 ? len1 : len2);
7345 return retval < 0 ? -1 : 1;
7348 * When the result of collation is equality, that doesn't mean
7349 * that there are no differences -- some locales exclude some
7350 * characters from consideration. So to avoid false equalities,
7351 * we use the raw string as a tiebreaker.
7357 #endif /* USE_LOCALE_COLLATE */
7359 return sv_cmp(sv1, sv2);
7363 #ifdef USE_LOCALE_COLLATE
7366 =for apidoc sv_collxfrm
7368 This calls C<sv_collxfrm_flags> with the SV_GMAGIC flag. See
7369 C<sv_collxfrm_flags>.
7371 =for apidoc sv_collxfrm_flags
7373 Add Collate Transform magic to an SV if it doesn't already have it. If the
7374 flags contain SV_GMAGIC, it handles get-magic.
7376 Any scalar variable may carry PERL_MAGIC_collxfrm magic that contains the
7377 scalar data of the variable, but transformed to such a format that a normal
7378 memory comparison can be used to compare the data according to the locale
7385 Perl_sv_collxfrm_flags(pTHX_ SV *const sv, STRLEN *const nxp, const I32 flags)
7390 PERL_ARGS_ASSERT_SV_COLLXFRM_FLAGS;
7392 mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_collxfrm) : (MAGIC *) NULL;
7393 if (!mg || !mg->mg_ptr || *(U32*)mg->mg_ptr != PL_collation_ix) {
7399 Safefree(mg->mg_ptr);
7400 s = SvPV_flags_const(sv, len, flags);
7401 if ((xf = mem_collxfrm(s, len, &xlen))) {
7403 #ifdef PERL_OLD_COPY_ON_WRITE
7405 sv_force_normal_flags(sv, 0);
7407 mg = sv_magicext(sv, 0, PERL_MAGIC_collxfrm, &PL_vtbl_collxfrm,
7421 if (mg && mg->mg_ptr) {
7423 return mg->mg_ptr + sizeof(PL_collation_ix);
7431 #endif /* USE_LOCALE_COLLATE */
7434 S_sv_gets_append_to_utf8(pTHX_ SV *const sv, PerlIO *const fp, I32 append)
7436 SV * const tsv = newSV(0);
7439 sv_gets(tsv, fp, 0);
7440 sv_utf8_upgrade_nomg(tsv);
7441 SvCUR_set(sv,append);
7444 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7448 S_sv_gets_read_record(pTHX_ SV *const sv, PerlIO *const fp, I32 append)
7451 const U32 recsize = SvUV(SvRV(PL_rs)); /* RsRECORD() guarantees > 0. */
7452 /* Grab the size of the record we're getting */
7453 char *const buffer = SvGROW(sv, (STRLEN)(recsize + append + 1)) + append;
7460 /* VMS wants read instead of fread, because fread doesn't respect */
7461 /* RMS record boundaries. This is not necessarily a good thing to be */
7462 /* doing, but we've got no other real choice - except avoid stdio
7463 as implementation - perhaps write a :vms layer ?
7465 fd = PerlIO_fileno(fp);
7467 bytesread = PerlLIO_read(fd, buffer, recsize);
7469 else /* in-memory file from PerlIO::Scalar */
7472 bytesread = PerlIO_read(fp, buffer, recsize);
7477 SvCUR_set(sv, bytesread + append);
7478 buffer[bytesread] = '\0';
7479 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7485 Get a line from the filehandle and store it into the SV, optionally
7486 appending to the currently-stored string.
7492 Perl_sv_gets(pTHX_ register SV *const sv, register PerlIO *const fp, I32 append)
7497 register STDCHAR rslast;
7498 register STDCHAR *bp;
7503 PERL_ARGS_ASSERT_SV_GETS;
7505 if (SvTHINKFIRST(sv))
7506 sv_force_normal_flags(sv, append ? 0 : SV_COW_DROP_PV);
7507 /* XXX. If you make this PVIV, then copy on write can copy scalars read
7509 However, perlbench says it's slower, because the existing swipe code
7510 is faster than copy on write.
7511 Swings and roundabouts. */
7512 SvUPGRADE(sv, SVt_PV);
7517 if (PerlIO_isutf8(fp)) {
7519 sv_utf8_upgrade_nomg(sv);
7520 sv_pos_u2b(sv,&append,0);
7522 } else if (SvUTF8(sv)) {
7523 return S_sv_gets_append_to_utf8(aTHX_ sv, fp, append);
7531 if (PerlIO_isutf8(fp))
7534 if (IN_PERL_COMPILETIME) {
7535 /* we always read code in line mode */
7539 else if (RsSNARF(PL_rs)) {
7540 /* If it is a regular disk file use size from stat() as estimate
7541 of amount we are going to read -- may result in mallocing
7542 more memory than we really need if the layers below reduce
7543 the size we read (e.g. CRLF or a gzip layer).
7546 if (!PerlLIO_fstat(PerlIO_fileno(fp), &st) && S_ISREG(st.st_mode)) {
7547 const Off_t offset = PerlIO_tell(fp);
7548 if (offset != (Off_t) -1 && st.st_size + append > offset) {
7549 (void) SvGROW(sv, (STRLEN)((st.st_size - offset) + append + 1));
7555 else if (RsRECORD(PL_rs)) {
7556 return S_sv_gets_read_record(aTHX_ sv, fp, append);
7558 else if (RsPARA(PL_rs)) {
7564 /* Get $/ i.e. PL_rs into same encoding as stream wants */
7565 if (PerlIO_isutf8(fp)) {
7566 rsptr = SvPVutf8(PL_rs, rslen);
7569 if (SvUTF8(PL_rs)) {
7570 if (!sv_utf8_downgrade(PL_rs, TRUE)) {
7571 Perl_croak(aTHX_ "Wide character in $/");
7574 rsptr = SvPV_const(PL_rs, rslen);
7578 rslast = rslen ? rsptr[rslen - 1] : '\0';
7580 if (rspara) { /* have to do this both before and after */
7581 do { /* to make sure file boundaries work right */
7584 i = PerlIO_getc(fp);
7588 PerlIO_ungetc(fp,i);
7594 /* See if we know enough about I/O mechanism to cheat it ! */
7596 /* This used to be #ifdef test - it is made run-time test for ease
7597 of abstracting out stdio interface. One call should be cheap
7598 enough here - and may even be a macro allowing compile
7602 if (PerlIO_fast_gets(fp)) {
7605 * We're going to steal some values from the stdio struct
7606 * and put EVERYTHING in the innermost loop into registers.
7608 register STDCHAR *ptr;
7612 #if defined(VMS) && defined(PERLIO_IS_STDIO)
7613 /* An ungetc()d char is handled separately from the regular
7614 * buffer, so we getc() it back out and stuff it in the buffer.
7616 i = PerlIO_getc(fp);
7617 if (i == EOF) return 0;
7618 *(--((*fp)->_ptr)) = (unsigned char) i;
7622 /* Here is some breathtakingly efficient cheating */
7624 cnt = PerlIO_get_cnt(fp); /* get count into register */
7625 /* make sure we have the room */
7626 if ((I32)(SvLEN(sv) - append) <= cnt + 1) {
7627 /* Not room for all of it
7628 if we are looking for a separator and room for some
7630 if (rslen && cnt > 80 && (I32)SvLEN(sv) > append) {
7631 /* just process what we have room for */
7632 shortbuffered = cnt - SvLEN(sv) + append + 1;
7633 cnt -= shortbuffered;
7637 /* remember that cnt can be negative */
7638 SvGROW(sv, (STRLEN)(append + (cnt <= 0 ? 2 : (cnt + 1))));
7643 bp = (STDCHAR*)SvPVX_const(sv) + append; /* move these two too to registers */
7644 ptr = (STDCHAR*)PerlIO_get_ptr(fp);
7645 DEBUG_P(PerlIO_printf(Perl_debug_log,
7646 "Screamer: entering, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7647 DEBUG_P(PerlIO_printf(Perl_debug_log,
7648 "Screamer: entering: PerlIO * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7649 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7650 PTR2UV(PerlIO_has_base(fp) ? PerlIO_get_base(fp) : 0)));
7655 while (cnt > 0) { /* this | eat */
7657 if ((*bp++ = *ptr++) == rslast) /* really | dust */
7658 goto thats_all_folks; /* screams | sed :-) */
7662 Copy(ptr, bp, cnt, char); /* this | eat */
7663 bp += cnt; /* screams | dust */
7664 ptr += cnt; /* louder | sed :-) */
7666 assert (!shortbuffered);
7667 goto cannot_be_shortbuffered;
7671 if (shortbuffered) { /* oh well, must extend */
7672 cnt = shortbuffered;
7674 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
7676 SvGROW(sv, SvLEN(sv) + append + cnt + 2);
7677 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
7681 cannot_be_shortbuffered:
7682 DEBUG_P(PerlIO_printf(Perl_debug_log,
7683 "Screamer: going to getc, ptr=%"UVuf", cnt=%ld\n",
7684 PTR2UV(ptr),(long)cnt));
7685 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* deregisterize cnt and ptr */
7687 DEBUG_Pv(PerlIO_printf(Perl_debug_log,
7688 "Screamer: pre: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7689 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7690 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7692 /* This used to call 'filbuf' in stdio form, but as that behaves like
7693 getc when cnt <= 0 we use PerlIO_getc here to avoid introducing
7694 another abstraction. */
7695 i = PerlIO_getc(fp); /* get more characters */
7697 DEBUG_Pv(PerlIO_printf(Perl_debug_log,
7698 "Screamer: post: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7699 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7700 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7702 cnt = PerlIO_get_cnt(fp);
7703 ptr = (STDCHAR*)PerlIO_get_ptr(fp); /* reregisterize cnt and ptr */
7704 DEBUG_P(PerlIO_printf(Perl_debug_log,
7705 "Screamer: after getc, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7707 if (i == EOF) /* all done for ever? */
7708 goto thats_really_all_folks;
7710 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
7712 SvGROW(sv, bpx + cnt + 2);
7713 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
7715 *bp++ = (STDCHAR)i; /* store character from PerlIO_getc */
7717 if (rslen && (STDCHAR)i == rslast) /* all done for now? */
7718 goto thats_all_folks;
7722 if ((rslen > 1 && (STRLEN)(bp - (STDCHAR*)SvPVX_const(sv)) < rslen) ||
7723 memNE((char*)bp - rslen, rsptr, rslen))
7724 goto screamer; /* go back to the fray */
7725 thats_really_all_folks:
7727 cnt += shortbuffered;
7728 DEBUG_P(PerlIO_printf(Perl_debug_log,
7729 "Screamer: quitting, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7730 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* put these back or we're in trouble */
7731 DEBUG_P(PerlIO_printf(Perl_debug_log,
7732 "Screamer: end: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7733 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7734 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7736 SvCUR_set(sv, bp - (STDCHAR*)SvPVX_const(sv)); /* set length */
7737 DEBUG_P(PerlIO_printf(Perl_debug_log,
7738 "Screamer: done, len=%ld, string=|%.*s|\n",
7739 (long)SvCUR(sv),(int)SvCUR(sv),SvPVX_const(sv)));
7743 /*The big, slow, and stupid way. */
7744 #ifdef USE_HEAP_INSTEAD_OF_STACK /* Even slower way. */
7745 STDCHAR *buf = NULL;
7746 Newx(buf, 8192, STDCHAR);
7754 register const STDCHAR * const bpe = buf + sizeof(buf);
7756 while ((i = PerlIO_getc(fp)) != EOF && (*bp++ = (STDCHAR)i) != rslast && bp < bpe)
7757 ; /* keep reading */
7761 cnt = PerlIO_read(fp,(char*)buf, sizeof(buf));
7762 /* Accommodate broken VAXC compiler, which applies U8 cast to
7763 * both args of ?: operator, causing EOF to change into 255
7766 i = (U8)buf[cnt - 1];
7772 cnt = 0; /* we do need to re-set the sv even when cnt <= 0 */
7774 sv_catpvn(sv, (char *) buf, cnt);
7776 sv_setpvn(sv, (char *) buf, cnt);
7778 if (i != EOF && /* joy */
7780 SvCUR(sv) < rslen ||
7781 memNE(SvPVX_const(sv) + SvCUR(sv) - rslen, rsptr, rslen)))
7785 * If we're reading from a TTY and we get a short read,
7786 * indicating that the user hit his EOF character, we need
7787 * to notice it now, because if we try to read from the TTY
7788 * again, the EOF condition will disappear.
7790 * The comparison of cnt to sizeof(buf) is an optimization
7791 * that prevents unnecessary calls to feof().
7795 if (!(cnt < (I32)sizeof(buf) && PerlIO_eof(fp)))
7799 #ifdef USE_HEAP_INSTEAD_OF_STACK
7804 if (rspara) { /* have to do this both before and after */
7805 while (i != EOF) { /* to make sure file boundaries work right */
7806 i = PerlIO_getc(fp);
7808 PerlIO_ungetc(fp,i);
7814 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7820 Auto-increment of the value in the SV, doing string to numeric conversion
7821 if necessary. Handles 'get' magic and operator overloading.
7827 Perl_sv_inc(pTHX_ register SV *const sv)
7836 =for apidoc sv_inc_nomg
7838 Auto-increment of the value in the SV, doing string to numeric conversion
7839 if necessary. Handles operator overloading. Skips handling 'get' magic.
7845 Perl_sv_inc_nomg(pTHX_ register SV *const sv)
7853 if (SvTHINKFIRST(sv)) {
7855 sv_force_normal_flags(sv, 0);
7856 if (SvREADONLY(sv)) {
7857 if (IN_PERL_RUNTIME)
7858 Perl_croak_no_modify(aTHX);
7862 if (SvAMAGIC(sv) && AMG_CALLunary(sv, inc_amg))
7864 i = PTR2IV(SvRV(sv));
7869 flags = SvFLAGS(sv);
7870 if ((flags & (SVp_NOK|SVp_IOK)) == SVp_NOK) {
7871 /* It's (privately or publicly) a float, but not tested as an
7872 integer, so test it to see. */
7874 flags = SvFLAGS(sv);
7876 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
7877 /* It's publicly an integer, or privately an integer-not-float */
7878 #ifdef PERL_PRESERVE_IVUV
7882 if (SvUVX(sv) == UV_MAX)
7883 sv_setnv(sv, UV_MAX_P1);
7885 (void)SvIOK_only_UV(sv);
7886 SvUV_set(sv, SvUVX(sv) + 1);
7888 if (SvIVX(sv) == IV_MAX)
7889 sv_setuv(sv, (UV)IV_MAX + 1);
7891 (void)SvIOK_only(sv);
7892 SvIV_set(sv, SvIVX(sv) + 1);
7897 if (flags & SVp_NOK) {
7898 const NV was = SvNVX(sv);
7899 if (NV_OVERFLOWS_INTEGERS_AT &&
7900 was >= NV_OVERFLOWS_INTEGERS_AT) {
7901 Perl_ck_warner(aTHX_ packWARN(WARN_IMPRECISION),
7902 "Lost precision when incrementing %" NVff " by 1",
7905 (void)SvNOK_only(sv);
7906 SvNV_set(sv, was + 1.0);
7910 if (!(flags & SVp_POK) || !*SvPVX_const(sv)) {
7911 if ((flags & SVTYPEMASK) < SVt_PVIV)
7912 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV ? SVt_PVIV : SVt_IV));
7913 (void)SvIOK_only(sv);
7918 while (isALPHA(*d)) d++;
7919 while (isDIGIT(*d)) d++;
7920 if (d < SvEND(sv)) {
7921 #ifdef PERL_PRESERVE_IVUV
7922 /* Got to punt this as an integer if needs be, but we don't issue
7923 warnings. Probably ought to make the sv_iv_please() that does
7924 the conversion if possible, and silently. */
7925 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
7926 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
7927 /* Need to try really hard to see if it's an integer.
7928 9.22337203685478e+18 is an integer.
7929 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
7930 so $a="9.22337203685478e+18"; $a+0; $a++
7931 needs to be the same as $a="9.22337203685478e+18"; $a++
7938 /* sv_2iv *should* have made this an NV */
7939 if (flags & SVp_NOK) {
7940 (void)SvNOK_only(sv);
7941 SvNV_set(sv, SvNVX(sv) + 1.0);
7944 /* I don't think we can get here. Maybe I should assert this
7945 And if we do get here I suspect that sv_setnv will croak. NWC
7947 #if defined(USE_LONG_DOUBLE)
7948 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",
7949 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7951 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_inc punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
7952 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7955 #endif /* PERL_PRESERVE_IVUV */
7956 sv_setnv(sv,Atof(SvPVX_const(sv)) + 1.0);
7960 while (d >= SvPVX_const(sv)) {
7968 /* MKS: The original code here died if letters weren't consecutive.
7969 * at least it didn't have to worry about non-C locales. The
7970 * new code assumes that ('z'-'a')==('Z'-'A'), letters are
7971 * arranged in order (although not consecutively) and that only
7972 * [A-Za-z] are accepted by isALPHA in the C locale.
7974 if (*d != 'z' && *d != 'Z') {
7975 do { ++*d; } while (!isALPHA(*d));
7978 *(d--) -= 'z' - 'a';
7983 *(d--) -= 'z' - 'a' + 1;
7987 /* oh,oh, the number grew */
7988 SvGROW(sv, SvCUR(sv) + 2);
7989 SvCUR_set(sv, SvCUR(sv) + 1);
7990 for (d = SvPVX(sv) + SvCUR(sv); d > SvPVX_const(sv); d--)
8001 Auto-decrement of the value in the SV, doing string to numeric conversion
8002 if necessary. Handles 'get' magic and operator overloading.
8008 Perl_sv_dec(pTHX_ register SV *const sv)
8018 =for apidoc sv_dec_nomg
8020 Auto-decrement of the value in the SV, doing string to numeric conversion
8021 if necessary. Handles operator overloading. Skips handling 'get' magic.
8027 Perl_sv_dec_nomg(pTHX_ register SV *const sv)
8034 if (SvTHINKFIRST(sv)) {
8036 sv_force_normal_flags(sv, 0);
8037 if (SvREADONLY(sv)) {
8038 if (IN_PERL_RUNTIME)
8039 Perl_croak_no_modify(aTHX);
8043 if (SvAMAGIC(sv) && AMG_CALLunary(sv, dec_amg))
8045 i = PTR2IV(SvRV(sv));
8050 /* Unlike sv_inc we don't have to worry about string-never-numbers
8051 and keeping them magic. But we mustn't warn on punting */
8052 flags = SvFLAGS(sv);
8053 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
8054 /* It's publicly an integer, or privately an integer-not-float */
8055 #ifdef PERL_PRESERVE_IVUV
8059 if (SvUVX(sv) == 0) {
8060 (void)SvIOK_only(sv);
8064 (void)SvIOK_only_UV(sv);
8065 SvUV_set(sv, SvUVX(sv) - 1);
8068 if (SvIVX(sv) == IV_MIN) {
8069 sv_setnv(sv, (NV)IV_MIN);
8073 (void)SvIOK_only(sv);
8074 SvIV_set(sv, SvIVX(sv) - 1);
8079 if (flags & SVp_NOK) {
8082 const NV was = SvNVX(sv);
8083 if (NV_OVERFLOWS_INTEGERS_AT &&
8084 was <= -NV_OVERFLOWS_INTEGERS_AT) {
8085 Perl_ck_warner(aTHX_ packWARN(WARN_IMPRECISION),
8086 "Lost precision when decrementing %" NVff " by 1",
8089 (void)SvNOK_only(sv);
8090 SvNV_set(sv, was - 1.0);
8094 if (!(flags & SVp_POK)) {
8095 if ((flags & SVTYPEMASK) < SVt_PVIV)
8096 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV) ? SVt_PVIV : SVt_IV);
8098 (void)SvIOK_only(sv);
8101 #ifdef PERL_PRESERVE_IVUV
8103 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
8104 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
8105 /* Need to try really hard to see if it's an integer.
8106 9.22337203685478e+18 is an integer.
8107 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
8108 so $a="9.22337203685478e+18"; $a+0; $a--
8109 needs to be the same as $a="9.22337203685478e+18"; $a--
8116 /* sv_2iv *should* have made this an NV */
8117 if (flags & SVp_NOK) {
8118 (void)SvNOK_only(sv);
8119 SvNV_set(sv, SvNVX(sv) - 1.0);
8122 /* I don't think we can get here. Maybe I should assert this
8123 And if we do get here I suspect that sv_setnv will croak. NWC
8125 #if defined(USE_LONG_DOUBLE)
8126 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",
8127 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
8129 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_dec punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
8130 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
8134 #endif /* PERL_PRESERVE_IVUV */
8135 sv_setnv(sv,Atof(SvPVX_const(sv)) - 1.0); /* punt */
8138 /* this define is used to eliminate a chunk of duplicated but shared logic
8139 * it has the suffix __SV_C to signal that it isnt API, and isnt meant to be
8140 * used anywhere but here - yves
8142 #define PUSH_EXTEND_MORTAL__SV_C(AnSv) \
8145 PL_tmps_stack[++PL_tmps_ix] = (AnSv); \
8149 =for apidoc sv_mortalcopy
8151 Creates a new SV which is a copy of the original SV (using C<sv_setsv>).
8152 The new SV is marked as mortal. It will be destroyed "soon", either by an
8153 explicit call to FREETMPS, or by an implicit call at places such as
8154 statement boundaries. See also C<sv_newmortal> and C<sv_2mortal>.
8159 /* Make a string that will exist for the duration of the expression
8160 * evaluation. Actually, it may have to last longer than that, but
8161 * hopefully we won't free it until it has been assigned to a
8162 * permanent location. */
8165 Perl_sv_mortalcopy(pTHX_ SV *const oldstr)
8171 sv_setsv(sv,oldstr);
8172 PUSH_EXTEND_MORTAL__SV_C(sv);
8178 =for apidoc sv_newmortal
8180 Creates a new null SV which is mortal. The reference count of the SV is
8181 set to 1. It will be destroyed "soon", either by an explicit call to
8182 FREETMPS, or by an implicit call at places such as statement boundaries.
8183 See also C<sv_mortalcopy> and C<sv_2mortal>.
8189 Perl_sv_newmortal(pTHX)
8195 SvFLAGS(sv) = SVs_TEMP;
8196 PUSH_EXTEND_MORTAL__SV_C(sv);
8202 =for apidoc newSVpvn_flags
8204 Creates a new SV and copies a string into it. The reference count for the
8205 SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
8206 string. You are responsible for ensuring that the source string is at least
8207 C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
8208 Currently the only flag bits accepted are C<SVf_UTF8> and C<SVs_TEMP>.
8209 If C<SVs_TEMP> is set, then C<sv_2mortal()> is called on the result before
8210 returning. If C<SVf_UTF8> is set, C<s> is considered to be in UTF-8 and the
8211 C<SVf_UTF8> flag will be set on the new SV.
8212 C<newSVpvn_utf8()> is a convenience wrapper for this function, defined as
8214 #define newSVpvn_utf8(s, len, u) \
8215 newSVpvn_flags((s), (len), (u) ? SVf_UTF8 : 0)
8221 Perl_newSVpvn_flags(pTHX_ const char *const s, const STRLEN len, const U32 flags)
8226 /* All the flags we don't support must be zero.
8227 And we're new code so I'm going to assert this from the start. */
8228 assert(!(flags & ~(SVf_UTF8|SVs_TEMP)));
8230 sv_setpvn(sv,s,len);
8232 /* This code used to a sv_2mortal(), however we now unroll the call to sv_2mortal()
8233 * and do what it does ourselves here.
8234 * Since we have asserted that flags can only have the SVf_UTF8 and/or SVs_TEMP flags
8235 * set above we can use it to enable the sv flags directly (bypassing SvTEMP_on), which
8236 * in turn means we dont need to mask out the SVf_UTF8 flag below, which means that we
8237 * eliminate quite a few steps than it looks - Yves (explaining patch by gfx)
8240 SvFLAGS(sv) |= flags;
8242 if(flags & SVs_TEMP){
8243 PUSH_EXTEND_MORTAL__SV_C(sv);
8250 =for apidoc sv_2mortal
8252 Marks an existing SV as mortal. The SV will be destroyed "soon", either
8253 by an explicit call to FREETMPS, or by an implicit call at places such as
8254 statement boundaries. SvTEMP() is turned on which means that the SV's
8255 string buffer can be "stolen" if this SV is copied. See also C<sv_newmortal>
8256 and C<sv_mortalcopy>.
8262 Perl_sv_2mortal(pTHX_ register SV *const sv)
8267 if (SvREADONLY(sv) && SvIMMORTAL(sv))
8269 PUSH_EXTEND_MORTAL__SV_C(sv);
8277 Creates a new SV and copies a string into it. The reference count for the
8278 SV is set to 1. If C<len> is zero, Perl will compute the length using
8279 strlen(). For efficiency, consider using C<newSVpvn> instead.
8285 Perl_newSVpv(pTHX_ const char *const s, const STRLEN len)
8291 sv_setpvn(sv, s, len || s == NULL ? len : strlen(s));
8296 =for apidoc newSVpvn
8298 Creates a new SV and copies a string into it. The reference count for the
8299 SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
8300 string. You are responsible for ensuring that the source string is at least
8301 C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
8307 Perl_newSVpvn(pTHX_ const char *const s, const STRLEN len)
8313 sv_setpvn(sv,s,len);
8318 =for apidoc newSVhek
8320 Creates a new SV from the hash key structure. It will generate scalars that
8321 point to the shared string table where possible. Returns a new (undefined)
8322 SV if the hek is NULL.
8328 Perl_newSVhek(pTHX_ const HEK *const hek)
8338 if (HEK_LEN(hek) == HEf_SVKEY) {
8339 return newSVsv(*(SV**)HEK_KEY(hek));
8341 const int flags = HEK_FLAGS(hek);
8342 if (flags & HVhek_WASUTF8) {
8344 Andreas would like keys he put in as utf8 to come back as utf8
8346 STRLEN utf8_len = HEK_LEN(hek);
8347 SV * const sv = newSV_type(SVt_PV);
8348 char *as_utf8 = (char *)bytes_to_utf8 ((U8*)HEK_KEY(hek), &utf8_len);
8349 /* bytes_to_utf8() allocates a new string, which we can repurpose: */
8350 sv_usepvn_flags(sv, as_utf8, utf8_len, SV_HAS_TRAILING_NUL);
8353 } else if (flags & (HVhek_REHASH|HVhek_UNSHARED)) {
8354 /* We don't have a pointer to the hv, so we have to replicate the
8355 flag into every HEK. This hv is using custom a hasing
8356 algorithm. Hence we can't return a shared string scalar, as
8357 that would contain the (wrong) hash value, and might get passed
8358 into an hv routine with a regular hash.
8359 Similarly, a hash that isn't using shared hash keys has to have
8360 the flag in every key so that we know not to try to call
8361 share_hek_kek on it. */
8363 SV * const sv = newSVpvn (HEK_KEY(hek), HEK_LEN(hek));
8368 /* This will be overwhelminly the most common case. */
8370 /* Inline most of newSVpvn_share(), because share_hek_hek() is far
8371 more efficient than sharepvn(). */
8375 sv_upgrade(sv, SVt_PV);
8376 SvPV_set(sv, (char *)HEK_KEY(share_hek_hek(hek)));
8377 SvCUR_set(sv, HEK_LEN(hek));
8390 =for apidoc newSVpvn_share
8392 Creates a new SV with its SvPVX_const pointing to a shared string in the string
8393 table. If the string does not already exist in the table, it is created
8394 first. Turns on READONLY and FAKE. If the C<hash> parameter is non-zero, that
8395 value is used; otherwise the hash is computed. The string's hash can be later
8396 be retrieved from the SV with the C<SvSHARED_HASH()> macro. The idea here is
8397 that as the string table is used for shared hash keys these strings will have
8398 SvPVX_const == HeKEY and hash lookup will avoid string compare.
8404 Perl_newSVpvn_share(pTHX_ const char *src, I32 len, U32 hash)
8408 bool is_utf8 = FALSE;
8409 const char *const orig_src = src;
8412 STRLEN tmplen = -len;
8414 /* See the note in hv.c:hv_fetch() --jhi */
8415 src = (char*)bytes_from_utf8((const U8*)src, &tmplen, &is_utf8);
8419 PERL_HASH(hash, src, len);
8421 /* The logic for this is inlined in S_mro_get_linear_isa_dfs(), so if it
8422 changes here, update it there too. */
8423 sv_upgrade(sv, SVt_PV);
8424 SvPV_set(sv, sharepvn(src, is_utf8?-len:len, hash));
8432 if (src != orig_src)
8438 =for apidoc newSVpv_share
8440 Like C<newSVpvn_share>, but takes a nul-terminated string instead of a
8447 Perl_newSVpv_share(pTHX_ const char *src, U32 hash)
8449 return newSVpvn_share(src, strlen(src), hash);
8452 #if defined(PERL_IMPLICIT_CONTEXT)
8454 /* pTHX_ magic can't cope with varargs, so this is a no-context
8455 * version of the main function, (which may itself be aliased to us).
8456 * Don't access this version directly.
8460 Perl_newSVpvf_nocontext(const char *const pat, ...)
8466 PERL_ARGS_ASSERT_NEWSVPVF_NOCONTEXT;
8468 va_start(args, pat);
8469 sv = vnewSVpvf(pat, &args);
8476 =for apidoc newSVpvf
8478 Creates a new SV and initializes it with the string formatted like
8485 Perl_newSVpvf(pTHX_ const char *const pat, ...)
8490 PERL_ARGS_ASSERT_NEWSVPVF;
8492 va_start(args, pat);
8493 sv = vnewSVpvf(pat, &args);
8498 /* backend for newSVpvf() and newSVpvf_nocontext() */
8501 Perl_vnewSVpvf(pTHX_ const char *const pat, va_list *const args)
8506 PERL_ARGS_ASSERT_VNEWSVPVF;
8509 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
8516 Creates a new SV and copies a floating point value into it.
8517 The reference count for the SV is set to 1.
8523 Perl_newSVnv(pTHX_ const NV n)
8536 Creates a new SV and copies an integer into it. The reference count for the
8543 Perl_newSViv(pTHX_ const IV i)
8556 Creates a new SV and copies an unsigned integer into it.
8557 The reference count for the SV is set to 1.
8563 Perl_newSVuv(pTHX_ const UV u)
8574 =for apidoc newSV_type
8576 Creates a new SV, of the type specified. The reference count for the new SV
8583 Perl_newSV_type(pTHX_ const svtype type)
8588 sv_upgrade(sv, type);
8593 =for apidoc newRV_noinc
8595 Creates an RV wrapper for an SV. The reference count for the original
8596 SV is B<not> incremented.
8602 Perl_newRV_noinc(pTHX_ SV *const tmpRef)
8605 register SV *sv = newSV_type(SVt_IV);
8607 PERL_ARGS_ASSERT_NEWRV_NOINC;
8610 SvRV_set(sv, tmpRef);
8615 /* newRV_inc is the official function name to use now.
8616 * newRV_inc is in fact #defined to newRV in sv.h
8620 Perl_newRV(pTHX_ SV *const sv)
8624 PERL_ARGS_ASSERT_NEWRV;
8626 return newRV_noinc(SvREFCNT_inc_simple_NN(sv));
8632 Creates a new SV which is an exact duplicate of the original SV.
8639 Perl_newSVsv(pTHX_ register SV *const old)
8646 if (SvTYPE(old) == (svtype)SVTYPEMASK) {
8647 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL), "semi-panic: attempt to dup freed string");
8651 /* SV_GMAGIC is the default for sv_setv()
8652 SV_NOSTEAL prevents TEMP buffers being, well, stolen, and saves games
8653 with SvTEMP_off and SvTEMP_on round a call to sv_setsv. */
8654 sv_setsv_flags(sv, old, SV_GMAGIC | SV_NOSTEAL);
8659 =for apidoc sv_reset
8661 Underlying implementation for the C<reset> Perl function.
8662 Note that the perl-level function is vaguely deprecated.
8668 Perl_sv_reset(pTHX_ register const char *s, HV *const stash)
8671 char todo[PERL_UCHAR_MAX+1];
8673 PERL_ARGS_ASSERT_SV_RESET;
8678 if (!*s) { /* reset ?? searches */
8679 MAGIC * const mg = mg_find((const SV *)stash, PERL_MAGIC_symtab);
8681 const U32 count = mg->mg_len / sizeof(PMOP**);
8682 PMOP **pmp = (PMOP**) mg->mg_ptr;
8683 PMOP *const *const end = pmp + count;
8687 SvREADONLY_off(PL_regex_pad[(*pmp)->op_pmoffset]);
8689 (*pmp)->op_pmflags &= ~PMf_USED;
8697 /* reset variables */
8699 if (!HvARRAY(stash))
8702 Zero(todo, 256, char);
8705 I32 i = (unsigned char)*s;
8709 max = (unsigned char)*s++;
8710 for ( ; i <= max; i++) {
8713 for (i = 0; i <= (I32) HvMAX(stash); i++) {
8715 for (entry = HvARRAY(stash)[i];
8717 entry = HeNEXT(entry))
8722 if (!todo[(U8)*HeKEY(entry)])
8724 gv = MUTABLE_GV(HeVAL(entry));
8727 if (SvTHINKFIRST(sv)) {
8728 if (!SvREADONLY(sv) && SvROK(sv))
8730 /* XXX Is this continue a bug? Why should THINKFIRST
8731 exempt us from resetting arrays and hashes? */
8735 if (SvTYPE(sv) >= SVt_PV) {
8737 if (SvPVX_const(sv) != NULL)
8745 if (GvHV(gv) && !HvNAME_get(GvHV(gv))) {
8747 Perl_die(aTHX_ "Can't reset %%ENV on this system");
8750 # if defined(USE_ENVIRON_ARRAY)
8753 # endif /* USE_ENVIRON_ARRAY */
8764 Using various gambits, try to get an IO from an SV: the IO slot if its a
8765 GV; or the recursive result if we're an RV; or the IO slot of the symbol
8766 named after the PV if we're a string.
8772 Perl_sv_2io(pTHX_ SV *const sv)
8777 PERL_ARGS_ASSERT_SV_2IO;
8779 switch (SvTYPE(sv)) {
8781 io = MUTABLE_IO(sv);
8785 if (isGV_with_GP(sv)) {
8786 gv = MUTABLE_GV(sv);
8789 Perl_croak(aTHX_ "Bad filehandle: %s", GvNAME(gv));
8795 Perl_croak(aTHX_ PL_no_usym, "filehandle");
8797 return sv_2io(SvRV(sv));
8798 gv = gv_fetchsv(sv, 0, SVt_PVIO);
8804 Perl_croak(aTHX_ "Bad filehandle: %"SVf, SVfARG(sv));
8813 Using various gambits, try to get a CV from an SV; in addition, try if
8814 possible to set C<*st> and C<*gvp> to the stash and GV associated with it.
8815 The flags in C<lref> are passed to gv_fetchsv.
8821 Perl_sv_2cv(pTHX_ SV *sv, HV **const st, GV **const gvp, const I32 lref)
8827 PERL_ARGS_ASSERT_SV_2CV;
8834 switch (SvTYPE(sv)) {
8838 return MUTABLE_CV(sv);
8845 if (isGV_with_GP(sv)) {
8846 gv = MUTABLE_GV(sv);
8857 sv = amagic_deref_call(sv, to_cv_amg);
8858 /* At this point I'd like to do SPAGAIN, but really I need to
8859 force it upon my callers. Hmmm. This is a mess... */
8862 if (SvTYPE(sv) == SVt_PVCV) {
8863 cv = MUTABLE_CV(sv);
8868 else if(isGV_with_GP(sv))
8869 gv = MUTABLE_GV(sv);
8871 Perl_croak(aTHX_ "Not a subroutine reference");
8873 else if (isGV_with_GP(sv)) {
8874 gv = MUTABLE_GV(sv);
8878 const char * const nambeg = SvPV_nomg_const(sv, len);
8879 gv = gv_fetchpvn_flags(
8880 nambeg, len, lref | SvUTF8(sv), SVt_PVCV
8888 /* Some flags to gv_fetchsv mean don't really create the GV */
8889 if (!isGV_with_GP(gv)) {
8895 if (lref & ~GV_ADDMG && !GvCVu(gv)) {
8899 gv_efullname3(tmpsv, gv, NULL);
8900 /* XXX this is probably not what they think they're getting.
8901 * It has the same effect as "sub name;", i.e. just a forward
8903 newSUB(start_subparse(FALSE, 0),
8904 newSVOP(OP_CONST, 0, tmpsv),
8908 Perl_croak(aTHX_ "Unable to create sub named \"%"SVf"\"",
8909 SVfARG(SvOK(sv) ? sv : &PL_sv_no));
8918 Returns true if the SV has a true value by Perl's rules.
8919 Use the C<SvTRUE> macro instead, which may call C<sv_true()> or may
8920 instead use an in-line version.
8926 Perl_sv_true(pTHX_ register SV *const sv)
8931 register const XPV* const tXpv = (XPV*)SvANY(sv);
8933 (tXpv->xpv_cur > 1 ||
8934 (tXpv->xpv_cur && *sv->sv_u.svu_pv != '0')))
8941 return SvIVX(sv) != 0;
8944 return SvNVX(sv) != 0.0;
8946 return sv_2bool(sv);
8952 =for apidoc sv_pvn_force
8954 Get a sensible string out of the SV somehow.
8955 A private implementation of the C<SvPV_force> macro for compilers which
8956 can't cope with complex macro expressions. Always use the macro instead.
8958 =for apidoc sv_pvn_force_flags
8960 Get a sensible string out of the SV somehow.
8961 If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<sv> if
8962 appropriate, else not. C<sv_pvn_force> and C<sv_pvn_force_nomg> are
8963 implemented in terms of this function.
8964 You normally want to use the various wrapper macros instead: see
8965 C<SvPV_force> and C<SvPV_force_nomg>
8971 Perl_sv_pvn_force_flags(pTHX_ SV *const sv, STRLEN *const lp, const I32 flags)
8975 PERL_ARGS_ASSERT_SV_PVN_FORCE_FLAGS;
8977 if (SvTHINKFIRST(sv) && !SvROK(sv))
8978 sv_force_normal_flags(sv, 0);
8988 if (SvREADONLY(sv) && !(flags & SV_MUTABLE_RETURN)) {
8989 const char * const ref = sv_reftype(sv,0);
8991 Perl_croak(aTHX_ "Can't coerce readonly %s to string in %s",
8992 ref, OP_DESC(PL_op));
8994 Perl_croak(aTHX_ "Can't coerce readonly %s to string", ref);
8996 if ((SvTYPE(sv) > SVt_PVLV && SvTYPE(sv) != SVt_PVFM)
8997 || isGV_with_GP(sv))
8998 /* diag_listed_as: Can't coerce %s to %s in %s */
8999 Perl_croak(aTHX_ "Can't coerce %s to string in %s", sv_reftype(sv,0),
9001 s = sv_2pv_flags(sv, &len, flags);
9005 if (s != SvPVX_const(sv)) { /* Almost, but not quite, sv_setpvn() */
9008 SvUPGRADE(sv, SVt_PV); /* Never FALSE */
9009 SvGROW(sv, len + 1);
9010 Move(s,SvPVX(sv),len,char);
9012 SvPVX(sv)[len] = '\0';
9015 SvPOK_on(sv); /* validate pointer */
9017 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
9018 PTR2UV(sv),SvPVX_const(sv)));
9021 return SvPVX_mutable(sv);
9025 =for apidoc sv_pvbyten_force
9027 The backend for the C<SvPVbytex_force> macro. Always use the macro instead.
9033 Perl_sv_pvbyten_force(pTHX_ SV *const sv, STRLEN *const lp)
9035 PERL_ARGS_ASSERT_SV_PVBYTEN_FORCE;
9037 sv_pvn_force(sv,lp);
9038 sv_utf8_downgrade(sv,0);
9044 =for apidoc sv_pvutf8n_force
9046 The backend for the C<SvPVutf8x_force> macro. Always use the macro instead.
9052 Perl_sv_pvutf8n_force(pTHX_ SV *const sv, STRLEN *const lp)
9054 PERL_ARGS_ASSERT_SV_PVUTF8N_FORCE;
9056 sv_pvn_force(sv,lp);
9057 sv_utf8_upgrade(sv);
9063 =for apidoc sv_reftype
9065 Returns a string describing what the SV is a reference to.
9071 Perl_sv_reftype(pTHX_ const SV *const sv, const int ob)
9073 PERL_ARGS_ASSERT_SV_REFTYPE;
9075 /* The fact that I don't need to downcast to char * everywhere, only in ?:
9076 inside return suggests a const propagation bug in g++. */
9077 if (ob && SvOBJECT(sv)) {
9078 char * const name = HvNAME_get(SvSTASH(sv));
9079 return name ? name : (char *) "__ANON__";
9082 switch (SvTYPE(sv)) {
9097 case SVt_PVLV: return (char *) (SvROK(sv) ? "REF"
9098 /* tied lvalues should appear to be
9099 * scalars for backwards compatibility */
9100 : (LvTYPE(sv) == 't' || LvTYPE(sv) == 'T')
9101 ? "SCALAR" : "LVALUE");
9102 case SVt_PVAV: return "ARRAY";
9103 case SVt_PVHV: return "HASH";
9104 case SVt_PVCV: return "CODE";
9105 case SVt_PVGV: return (char *) (isGV_with_GP(sv)
9106 ? "GLOB" : "SCALAR");
9107 case SVt_PVFM: return "FORMAT";
9108 case SVt_PVIO: return "IO";
9109 case SVt_BIND: return "BIND";
9110 case SVt_REGEXP: return "REGEXP";
9111 default: return "UNKNOWN";
9117 =for apidoc sv_isobject
9119 Returns a boolean indicating whether the SV is an RV pointing to a blessed
9120 object. If the SV is not an RV, or if the object is not blessed, then this
9127 Perl_sv_isobject(pTHX_ SV *sv)
9143 Returns a boolean indicating whether the SV is blessed into the specified
9144 class. This does not check for subtypes; use C<sv_derived_from> to verify
9145 an inheritance relationship.
9151 Perl_sv_isa(pTHX_ SV *sv, const char *const name)
9155 PERL_ARGS_ASSERT_SV_ISA;
9165 hvname = HvNAME_get(SvSTASH(sv));
9169 return strEQ(hvname, name);
9175 Creates a new SV for the RV, C<rv>, to point to. If C<rv> is not an RV then
9176 it will be upgraded to one. If C<classname> is non-null then the new SV will
9177 be blessed in the specified package. The new SV is returned and its
9178 reference count is 1.
9184 Perl_newSVrv(pTHX_ SV *const rv, const char *const classname)
9189 PERL_ARGS_ASSERT_NEWSVRV;
9193 SV_CHECK_THINKFIRST_COW_DROP(rv);
9194 (void)SvAMAGIC_off(rv);
9196 if (SvTYPE(rv) >= SVt_PVMG) {
9197 const U32 refcnt = SvREFCNT(rv);
9201 SvREFCNT(rv) = refcnt;
9203 sv_upgrade(rv, SVt_IV);
9204 } else if (SvROK(rv)) {
9205 SvREFCNT_dec(SvRV(rv));
9207 prepare_SV_for_RV(rv);
9215 HV* const stash = gv_stashpv(classname, GV_ADD);
9216 (void)sv_bless(rv, stash);
9222 =for apidoc sv_setref_pv
9224 Copies a pointer into a new SV, optionally blessing the SV. The C<rv>
9225 argument will be upgraded to an RV. That RV will be modified to point to
9226 the new SV. If the C<pv> argument is NULL then C<PL_sv_undef> will be placed
9227 into the SV. The C<classname> argument indicates the package for the
9228 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9229 will have a reference count of 1, and the RV will be returned.
9231 Do not use with other Perl types such as HV, AV, SV, CV, because those
9232 objects will become corrupted by the pointer copy process.
9234 Note that C<sv_setref_pvn> copies the string while this copies the pointer.
9240 Perl_sv_setref_pv(pTHX_ SV *const rv, const char *const classname, void *const pv)
9244 PERL_ARGS_ASSERT_SV_SETREF_PV;
9247 sv_setsv(rv, &PL_sv_undef);
9251 sv_setiv(newSVrv(rv,classname), PTR2IV(pv));
9256 =for apidoc sv_setref_iv
9258 Copies an integer into a new SV, optionally blessing the SV. The C<rv>
9259 argument will be upgraded to an RV. That RV will be modified to point to
9260 the new SV. The C<classname> argument indicates the package for the
9261 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9262 will have a reference count of 1, and the RV will be returned.
9268 Perl_sv_setref_iv(pTHX_ SV *const rv, const char *const classname, const IV iv)
9270 PERL_ARGS_ASSERT_SV_SETREF_IV;
9272 sv_setiv(newSVrv(rv,classname), iv);
9277 =for apidoc sv_setref_uv
9279 Copies an unsigned integer into a new SV, optionally blessing the SV. The C<rv>
9280 argument will be upgraded to an RV. That RV will be modified to point to
9281 the new SV. The C<classname> argument indicates the package for the
9282 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9283 will have a reference count of 1, and the RV will be returned.
9289 Perl_sv_setref_uv(pTHX_ SV *const rv, const char *const classname, const UV uv)
9291 PERL_ARGS_ASSERT_SV_SETREF_UV;
9293 sv_setuv(newSVrv(rv,classname), uv);
9298 =for apidoc sv_setref_nv
9300 Copies a double into a new SV, optionally blessing the SV. The C<rv>
9301 argument will be upgraded to an RV. That RV will be modified to point to
9302 the new SV. The C<classname> argument indicates the package for the
9303 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9304 will have a reference count of 1, and the RV will be returned.
9310 Perl_sv_setref_nv(pTHX_ SV *const rv, const char *const classname, const NV nv)
9312 PERL_ARGS_ASSERT_SV_SETREF_NV;
9314 sv_setnv(newSVrv(rv,classname), nv);
9319 =for apidoc sv_setref_pvn
9321 Copies a string into a new SV, optionally blessing the SV. The length of the
9322 string must be specified with C<n>. The C<rv> argument will be upgraded to
9323 an RV. That RV will be modified to point to the new SV. The C<classname>
9324 argument indicates the package for the blessing. Set C<classname> to
9325 C<NULL> to avoid the blessing. The new SV will have a reference count
9326 of 1, and the RV will be returned.
9328 Note that C<sv_setref_pv> copies the pointer while this copies the string.
9334 Perl_sv_setref_pvn(pTHX_ SV *const rv, const char *const classname,
9335 const char *const pv, const STRLEN n)
9337 PERL_ARGS_ASSERT_SV_SETREF_PVN;
9339 sv_setpvn(newSVrv(rv,classname), pv, n);
9344 =for apidoc sv_bless
9346 Blesses an SV into a specified package. The SV must be an RV. The package
9347 must be designated by its stash (see C<gv_stashpv()>). The reference count
9348 of the SV is unaffected.
9354 Perl_sv_bless(pTHX_ SV *const sv, HV *const stash)
9359 PERL_ARGS_ASSERT_SV_BLESS;
9362 Perl_croak(aTHX_ "Can't bless non-reference value");
9364 if (SvFLAGS(tmpRef) & (SVs_OBJECT|SVf_READONLY)) {
9365 if (SvIsCOW(tmpRef))
9366 sv_force_normal_flags(tmpRef, 0);
9367 if (SvREADONLY(tmpRef))
9368 Perl_croak_no_modify(aTHX);
9369 if (SvOBJECT(tmpRef)) {
9370 if (SvTYPE(tmpRef) != SVt_PVIO)
9372 SvREFCNT_dec(SvSTASH(tmpRef));
9375 SvOBJECT_on(tmpRef);
9376 if (SvTYPE(tmpRef) != SVt_PVIO)
9378 SvUPGRADE(tmpRef, SVt_PVMG);
9379 SvSTASH_set(tmpRef, MUTABLE_HV(SvREFCNT_inc_simple(stash)));
9384 (void)SvAMAGIC_off(sv);
9386 if(SvSMAGICAL(tmpRef))
9387 if(mg_find(tmpRef, PERL_MAGIC_ext) || mg_find(tmpRef, PERL_MAGIC_uvar))
9395 /* Downgrades a PVGV to a PVMG. If it’s actually a PVLV, we leave the type
9396 * as it is after unglobbing it.
9400 S_sv_unglob(pTHX_ SV *const sv)
9405 SV * const temp = sv_newmortal();
9407 PERL_ARGS_ASSERT_SV_UNGLOB;
9409 assert(SvTYPE(sv) == SVt_PVGV || SvTYPE(sv) == SVt_PVLV);
9411 gv_efullname3(temp, MUTABLE_GV(sv), "*");
9414 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
9415 && HvNAME_get(stash))
9416 mro_method_changed_in(stash);
9417 gp_free(MUTABLE_GV(sv));
9420 sv_del_backref(MUTABLE_SV(GvSTASH(sv)), sv);
9424 if (GvNAME_HEK(sv)) {
9425 unshare_hek(GvNAME_HEK(sv));
9427 isGV_with_GP_off(sv);
9429 if(SvTYPE(sv) == SVt_PVGV) {
9430 /* need to keep SvANY(sv) in the right arena */
9431 xpvmg = new_XPVMG();
9432 StructCopy(SvANY(sv), xpvmg, XPVMG);
9433 del_XPVGV(SvANY(sv));
9436 SvFLAGS(sv) &= ~SVTYPEMASK;
9437 SvFLAGS(sv) |= SVt_PVMG;
9440 /* Intentionally not calling any local SET magic, as this isn't so much a
9441 set operation as merely an internal storage change. */
9442 sv_setsv_flags(sv, temp, 0);
9446 =for apidoc sv_unref_flags
9448 Unsets the RV status of the SV, and decrements the reference count of
9449 whatever was being referenced by the RV. This can almost be thought of
9450 as a reversal of C<newSVrv>. The C<cflags> argument can contain
9451 C<SV_IMMEDIATE_UNREF> to force the reference count to be decremented
9452 (otherwise the decrementing is conditional on the reference count being
9453 different from one or the reference being a readonly SV).
9460 Perl_sv_unref_flags(pTHX_ SV *const ref, const U32 flags)
9462 SV* const target = SvRV(ref);
9464 PERL_ARGS_ASSERT_SV_UNREF_FLAGS;
9466 if (SvWEAKREF(ref)) {
9467 sv_del_backref(target, ref);
9469 SvRV_set(ref, NULL);
9472 SvRV_set(ref, NULL);
9474 /* You can't have a || SvREADONLY(target) here, as $a = $$a, where $a was
9475 assigned to as BEGIN {$a = \"Foo"} will fail. */
9476 if (SvREFCNT(target) != 1 || (flags & SV_IMMEDIATE_UNREF))
9477 SvREFCNT_dec(target);
9478 else /* XXX Hack, but hard to make $a=$a->[1] work otherwise */
9479 sv_2mortal(target); /* Schedule for freeing later */
9483 =for apidoc sv_untaint
9485 Untaint an SV. Use C<SvTAINTED_off> instead.
9491 Perl_sv_untaint(pTHX_ SV *const sv)
9493 PERL_ARGS_ASSERT_SV_UNTAINT;
9495 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
9496 MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
9503 =for apidoc sv_tainted
9505 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) == (svtype)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 } else if (CvCONST(dstr)) {
12028 CvXSUBANY(dstr).any_ptr =
12029 sv_dup_inc((const SV *)CvXSUBANY(dstr).any_ptr, param);
12031 if (CvDYNFILE(dstr)) CvFILE(dstr) = SAVEPV(CvFILE(dstr));
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_regdummy = proto_perl->Iregdummy;
12998 PL_colorset = 0; /* reinits PL_colors[] */
12999 /*PL_colors[6] = {0,0,0,0,0,0};*/
13001 /* Pluggable optimizer */
13002 PL_peepp = proto_perl->Ipeepp;
13003 PL_rpeepp = proto_perl->Irpeepp;
13004 /* op_free() hook */
13005 PL_opfreehook = proto_perl->Iopfreehook;
13007 #ifdef USE_REENTRANT_API
13008 /* XXX: things like -Dm will segfault here in perlio, but doing
13009 * PERL_SET_CONTEXT(proto_perl);
13010 * breaks too many other things
13012 Perl_reentrant_init(aTHX);
13015 /* create SV map for pointer relocation */
13016 PL_ptr_table = ptr_table_new();
13018 /* initialize these special pointers as early as possible */
13019 ptr_table_store(PL_ptr_table, &proto_perl->Isv_undef, &PL_sv_undef);
13021 SvANY(&PL_sv_no) = new_XPVNV();
13022 SvPV_set(&PL_sv_no, savepvn(PL_No, 0));
13023 SvCUR_set(&PL_sv_no, 0);
13024 SvLEN_set(&PL_sv_no, 1);
13025 SvIV_set(&PL_sv_no, 0);
13026 SvNV_set(&PL_sv_no, 0);
13027 ptr_table_store(PL_ptr_table, &proto_perl->Isv_no, &PL_sv_no);
13029 SvPV_set(&PL_sv_yes, savepvn(PL_Yes, 1));
13030 SvCUR_set(&PL_sv_yes, 1);
13031 SvLEN_set(&PL_sv_yes, 2);
13032 SvIV_set(&PL_sv_yes, 1);
13033 SvNV_set(&PL_sv_yes, 1);
13034 ptr_table_store(PL_ptr_table, &proto_perl->Isv_yes, &PL_sv_yes);
13036 /* create (a non-shared!) shared string table */
13037 PL_strtab = newHV();
13038 HvSHAREKEYS_off(PL_strtab);
13039 hv_ksplit(PL_strtab, HvTOTALKEYS(proto_perl->Istrtab));
13040 ptr_table_store(PL_ptr_table, proto_perl->Istrtab, PL_strtab);
13042 /* These two PVs will be free'd special way so must set them same way op.c does */
13043 PL_compiling.cop_stashpv = savesharedpv(PL_compiling.cop_stashpv);
13044 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_stashpv, PL_compiling.cop_stashpv);
13046 PL_compiling.cop_file = savesharedpv(PL_compiling.cop_file);
13047 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_file, PL_compiling.cop_file);
13049 ptr_table_store(PL_ptr_table, &proto_perl->Icompiling, &PL_compiling);
13050 PL_compiling.cop_warnings = DUP_WARNINGS(PL_compiling.cop_warnings);
13051 CopHINTHASH_set(&PL_compiling, cophh_copy(CopHINTHASH_get(&PL_compiling)));
13052 PL_curcop = (COP*)any_dup(proto_perl->Icurcop, proto_perl);
13054 param->stashes = newAV(); /* Setup array of objects to call clone on */
13055 /* This makes no difference to the implementation, as it always pushes
13056 and shifts pointers to other SVs without changing their reference
13057 count, with the array becoming empty before it is freed. However, it
13058 makes it conceptually clear what is going on, and will avoid some
13059 work inside av.c, filling slots between AvFILL() and AvMAX() with
13060 &PL_sv_undef, and SvREFCNT_dec()ing those. */
13061 AvREAL_off(param->stashes);
13063 if (!(flags & CLONEf_COPY_STACKS)) {
13064 param->unreferenced = newAV();
13067 #ifdef PERLIO_LAYERS
13068 /* Clone PerlIO tables as soon as we can handle general xx_dup() */
13069 PerlIO_clone(aTHX_ proto_perl, param);
13072 PL_envgv = gv_dup(proto_perl->Ienvgv, param);
13073 PL_incgv = gv_dup(proto_perl->Iincgv, param);
13074 PL_hintgv = gv_dup(proto_perl->Ihintgv, param);
13075 PL_origfilename = SAVEPV(proto_perl->Iorigfilename);
13076 PL_diehook = sv_dup_inc(proto_perl->Idiehook, param);
13077 PL_warnhook = sv_dup_inc(proto_perl->Iwarnhook, param);
13080 PL_patchlevel = sv_dup_inc(proto_perl->Ipatchlevel, param);
13081 PL_apiversion = sv_dup_inc(proto_perl->Iapiversion, param);
13082 PL_inplace = SAVEPV(proto_perl->Iinplace);
13083 PL_e_script = sv_dup_inc(proto_perl->Ie_script, param);
13085 /* magical thingies */
13086 PL_formfeed = sv_dup(proto_perl->Iformfeed, param);
13088 PL_encoding = sv_dup(proto_perl->Iencoding, param);
13090 sv_setpvs(PERL_DEBUG_PAD(0), ""); /* For regex debugging. */
13091 sv_setpvs(PERL_DEBUG_PAD(1), ""); /* ext/re needs these */
13092 sv_setpvs(PERL_DEBUG_PAD(2), ""); /* even without DEBUGGING. */
13095 /* Clone the regex array */
13096 /* ORANGE FIXME for plugins, probably in the SV dup code.
13097 newSViv(PTR2IV(CALLREGDUPE(
13098 INT2PTR(REGEXP *, SvIVX(regex)), param))))
13100 PL_regex_padav = av_dup_inc(proto_perl->Iregex_padav, param);
13101 PL_regex_pad = AvARRAY(PL_regex_padav);
13103 /* shortcuts to various I/O objects */
13104 PL_ofsgv = gv_dup_inc(proto_perl->Iofsgv, param);
13105 PL_stdingv = gv_dup(proto_perl->Istdingv, param);
13106 PL_stderrgv = gv_dup(proto_perl->Istderrgv, param);
13107 PL_defgv = gv_dup(proto_perl->Idefgv, param);
13108 PL_argvgv = gv_dup(proto_perl->Iargvgv, param);
13109 PL_argvoutgv = gv_dup(proto_perl->Iargvoutgv, param);
13110 PL_argvout_stack = av_dup_inc(proto_perl->Iargvout_stack, param);
13112 /* shortcuts to regexp stuff */
13113 PL_replgv = gv_dup(proto_perl->Ireplgv, param);
13115 /* shortcuts to misc objects */
13116 PL_errgv = gv_dup(proto_perl->Ierrgv, param);
13118 /* shortcuts to debugging objects */
13119 PL_DBgv = gv_dup(proto_perl->IDBgv, param);
13120 PL_DBline = gv_dup(proto_perl->IDBline, param);
13121 PL_DBsub = gv_dup(proto_perl->IDBsub, param);
13122 PL_DBsingle = sv_dup(proto_perl->IDBsingle, param);
13123 PL_DBtrace = sv_dup(proto_perl->IDBtrace, param);
13124 PL_DBsignal = sv_dup(proto_perl->IDBsignal, param);
13126 /* symbol tables */
13127 PL_defstash = hv_dup_inc(proto_perl->Idefstash, param);
13128 PL_curstash = hv_dup(proto_perl->Icurstash, param);
13129 PL_debstash = hv_dup(proto_perl->Idebstash, param);
13130 PL_globalstash = hv_dup(proto_perl->Iglobalstash, param);
13131 PL_curstname = sv_dup_inc(proto_perl->Icurstname, param);
13133 PL_beginav = av_dup_inc(proto_perl->Ibeginav, param);
13134 PL_beginav_save = av_dup_inc(proto_perl->Ibeginav_save, param);
13135 PL_checkav_save = av_dup_inc(proto_perl->Icheckav_save, param);
13136 PL_unitcheckav = av_dup_inc(proto_perl->Iunitcheckav, param);
13137 PL_unitcheckav_save = av_dup_inc(proto_perl->Iunitcheckav_save, param);
13138 PL_endav = av_dup_inc(proto_perl->Iendav, param);
13139 PL_checkav = av_dup_inc(proto_perl->Icheckav, param);
13140 PL_initav = av_dup_inc(proto_perl->Iinitav, param);
13142 PL_isarev = hv_dup_inc(proto_perl->Iisarev, param);
13144 /* subprocess state */
13145 PL_fdpid = av_dup_inc(proto_perl->Ifdpid, param);
13147 if (proto_perl->Iop_mask)
13148 PL_op_mask = SAVEPVN(proto_perl->Iop_mask, PL_maxo);
13151 /* PL_asserting = proto_perl->Iasserting; */
13153 /* current interpreter roots */
13154 PL_main_cv = cv_dup_inc(proto_perl->Imain_cv, param);
13156 PL_main_root = OpREFCNT_inc(proto_perl->Imain_root);
13159 /* runtime control stuff */
13160 PL_curcopdb = (COP*)any_dup(proto_perl->Icurcopdb, proto_perl);
13162 PL_preambleav = av_dup_inc(proto_perl->Ipreambleav, param);
13164 PL_ors_sv = sv_dup_inc(proto_perl->Iors_sv, param);
13166 /* interpreter atexit processing */
13167 PL_exitlistlen = proto_perl->Iexitlistlen;
13168 if (PL_exitlistlen) {
13169 Newx(PL_exitlist, PL_exitlistlen, PerlExitListEntry);
13170 Copy(proto_perl->Iexitlist, PL_exitlist, PL_exitlistlen, PerlExitListEntry);
13173 PL_exitlist = (PerlExitListEntry*)NULL;
13175 PL_my_cxt_size = proto_perl->Imy_cxt_size;
13176 if (PL_my_cxt_size) {
13177 Newx(PL_my_cxt_list, PL_my_cxt_size, void *);
13178 Copy(proto_perl->Imy_cxt_list, PL_my_cxt_list, PL_my_cxt_size, void *);
13179 #ifdef PERL_GLOBAL_STRUCT_PRIVATE
13180 Newx(PL_my_cxt_keys, PL_my_cxt_size, const char *);
13181 Copy(proto_perl->Imy_cxt_keys, PL_my_cxt_keys, PL_my_cxt_size, char *);
13185 PL_my_cxt_list = (void**)NULL;
13186 #ifdef PERL_GLOBAL_STRUCT_PRIVATE
13187 PL_my_cxt_keys = (const char**)NULL;
13190 PL_modglobal = hv_dup_inc(proto_perl->Imodglobal, param);
13191 PL_custom_op_names = hv_dup_inc(proto_perl->Icustom_op_names,param);
13192 PL_custom_op_descs = hv_dup_inc(proto_perl->Icustom_op_descs,param);
13193 PL_custom_ops = hv_dup_inc(proto_perl->Icustom_ops, param);
13195 PL_compcv = cv_dup(proto_perl->Icompcv, param);
13197 PAD_CLONE_VARS(proto_perl, param);
13199 #ifdef HAVE_INTERP_INTERN
13200 sys_intern_dup(&proto_perl->Isys_intern, &PL_sys_intern);
13203 PL_DBcv = cv_dup(proto_perl->IDBcv, param);
13205 #ifdef PERL_USES_PL_PIDSTATUS
13206 PL_pidstatus = newHV(); /* XXX flag for cloning? */
13208 PL_osname = SAVEPV(proto_perl->Iosname);
13209 PL_parser = parser_dup(proto_perl->Iparser, param);
13211 /* XXX this only works if the saved cop has already been cloned */
13212 if (proto_perl->Iparser) {
13213 PL_parser->saved_curcop = (COP*)any_dup(
13214 proto_perl->Iparser->saved_curcop,
13218 PL_subname = sv_dup_inc(proto_perl->Isubname, param);
13220 #ifdef USE_LOCALE_COLLATE
13221 PL_collation_name = SAVEPV(proto_perl->Icollation_name);
13222 #endif /* USE_LOCALE_COLLATE */
13224 #ifdef USE_LOCALE_NUMERIC
13225 PL_numeric_name = SAVEPV(proto_perl->Inumeric_name);
13226 PL_numeric_radix_sv = sv_dup_inc(proto_perl->Inumeric_radix_sv, param);
13227 #endif /* !USE_LOCALE_NUMERIC */
13229 /* utf8 character classes */
13230 PL_utf8_alnum = sv_dup_inc(proto_perl->Iutf8_alnum, param);
13231 PL_utf8_ascii = sv_dup_inc(proto_perl->Iutf8_ascii, param);
13232 PL_utf8_alpha = sv_dup_inc(proto_perl->Iutf8_alpha, param);
13233 PL_utf8_space = sv_dup_inc(proto_perl->Iutf8_space, param);
13234 PL_utf8_cntrl = sv_dup_inc(proto_perl->Iutf8_cntrl, param);
13235 PL_utf8_graph = sv_dup_inc(proto_perl->Iutf8_graph, param);
13236 PL_utf8_digit = sv_dup_inc(proto_perl->Iutf8_digit, param);
13237 PL_utf8_upper = sv_dup_inc(proto_perl->Iutf8_upper, param);
13238 PL_utf8_lower = sv_dup_inc(proto_perl->Iutf8_lower, param);
13239 PL_utf8_print = sv_dup_inc(proto_perl->Iutf8_print, param);
13240 PL_utf8_punct = sv_dup_inc(proto_perl->Iutf8_punct, param);
13241 PL_utf8_xdigit = sv_dup_inc(proto_perl->Iutf8_xdigit, param);
13242 PL_utf8_mark = sv_dup_inc(proto_perl->Iutf8_mark, param);
13243 PL_utf8_X_begin = sv_dup_inc(proto_perl->Iutf8_X_begin, param);
13244 PL_utf8_X_extend = sv_dup_inc(proto_perl->Iutf8_X_extend, param);
13245 PL_utf8_X_prepend = sv_dup_inc(proto_perl->Iutf8_X_prepend, param);
13246 PL_utf8_X_non_hangul = sv_dup_inc(proto_perl->Iutf8_X_non_hangul, param);
13247 PL_utf8_X_L = sv_dup_inc(proto_perl->Iutf8_X_L, param);
13248 PL_utf8_X_LV = sv_dup_inc(proto_perl->Iutf8_X_LV, param);
13249 PL_utf8_X_LVT = sv_dup_inc(proto_perl->Iutf8_X_LVT, param);
13250 PL_utf8_X_T = sv_dup_inc(proto_perl->Iutf8_X_T, param);
13251 PL_utf8_X_V = sv_dup_inc(proto_perl->Iutf8_X_V, param);
13252 PL_utf8_X_LV_LVT_V = sv_dup_inc(proto_perl->Iutf8_X_LV_LVT_V, param);
13253 PL_utf8_toupper = sv_dup_inc(proto_perl->Iutf8_toupper, param);
13254 PL_utf8_totitle = sv_dup_inc(proto_perl->Iutf8_totitle, param);
13255 PL_utf8_tolower = sv_dup_inc(proto_perl->Iutf8_tolower, param);
13256 PL_utf8_tofold = sv_dup_inc(proto_perl->Iutf8_tofold, param);
13257 PL_utf8_idstart = sv_dup_inc(proto_perl->Iutf8_idstart, param);
13258 PL_utf8_xidstart = sv_dup_inc(proto_perl->Iutf8_xidstart, param);
13259 PL_utf8_idcont = sv_dup_inc(proto_perl->Iutf8_idcont, param);
13260 PL_utf8_xidcont = sv_dup_inc(proto_perl->Iutf8_xidcont, param);
13261 PL_utf8_foldable = sv_dup_inc(proto_perl->Iutf8_foldable, param);
13264 if (proto_perl->Ipsig_pend) {
13265 Newxz(PL_psig_pend, SIG_SIZE, int);
13268 PL_psig_pend = (int*)NULL;
13271 if (proto_perl->Ipsig_name) {
13272 Newx(PL_psig_name, 2 * SIG_SIZE, SV*);
13273 sv_dup_inc_multiple(proto_perl->Ipsig_name, PL_psig_name, 2 * SIG_SIZE,
13275 PL_psig_ptr = PL_psig_name + SIG_SIZE;
13278 PL_psig_ptr = (SV**)NULL;
13279 PL_psig_name = (SV**)NULL;
13282 if (flags & CLONEf_COPY_STACKS) {
13283 Newx(PL_tmps_stack, PL_tmps_max, SV*);
13284 sv_dup_inc_multiple(proto_perl->Itmps_stack, PL_tmps_stack,
13285 PL_tmps_ix+1, param);
13287 /* next PUSHMARK() sets *(PL_markstack_ptr+1) */
13288 i = proto_perl->Imarkstack_max - proto_perl->Imarkstack;
13289 Newxz(PL_markstack, i, I32);
13290 PL_markstack_max = PL_markstack + (proto_perl->Imarkstack_max
13291 - proto_perl->Imarkstack);
13292 PL_markstack_ptr = PL_markstack + (proto_perl->Imarkstack_ptr
13293 - proto_perl->Imarkstack);
13294 Copy(proto_perl->Imarkstack, PL_markstack,
13295 PL_markstack_ptr - PL_markstack + 1, I32);
13297 /* next push_scope()/ENTER sets PL_scopestack[PL_scopestack_ix]
13298 * NOTE: unlike the others! */
13299 Newxz(PL_scopestack, PL_scopestack_max, I32);
13300 Copy(proto_perl->Iscopestack, PL_scopestack, PL_scopestack_ix, I32);
13303 Newxz(PL_scopestack_name, PL_scopestack_max, const char *);
13304 Copy(proto_perl->Iscopestack_name, PL_scopestack_name, PL_scopestack_ix, const char *);
13306 /* NOTE: si_dup() looks at PL_markstack */
13307 PL_curstackinfo = si_dup(proto_perl->Icurstackinfo, param);
13309 /* PL_curstack = PL_curstackinfo->si_stack; */
13310 PL_curstack = av_dup(proto_perl->Icurstack, param);
13311 PL_mainstack = av_dup(proto_perl->Imainstack, param);
13313 /* next PUSHs() etc. set *(PL_stack_sp+1) */
13314 PL_stack_base = AvARRAY(PL_curstack);
13315 PL_stack_sp = PL_stack_base + (proto_perl->Istack_sp
13316 - proto_perl->Istack_base);
13317 PL_stack_max = PL_stack_base + AvMAX(PL_curstack);
13319 /*Newxz(PL_savestack, PL_savestack_max, ANY);*/
13320 PL_savestack = ss_dup(proto_perl, param);
13324 ENTER; /* perl_destruct() wants to LEAVE; */
13327 PL_statgv = gv_dup(proto_perl->Istatgv, param);
13328 PL_statname = sv_dup_inc(proto_perl->Istatname, param);
13330 PL_rs = sv_dup_inc(proto_perl->Irs, param);
13331 PL_last_in_gv = gv_dup(proto_perl->Ilast_in_gv, param);
13332 PL_defoutgv = gv_dup_inc(proto_perl->Idefoutgv, param);
13333 PL_toptarget = sv_dup_inc(proto_perl->Itoptarget, param);
13334 PL_bodytarget = sv_dup_inc(proto_perl->Ibodytarget, param);
13335 PL_formtarget = sv_dup(proto_perl->Iformtarget, param);
13337 PL_errors = sv_dup_inc(proto_perl->Ierrors, param);
13339 PL_sortcop = (OP*)any_dup(proto_perl->Isortcop, proto_perl);
13340 PL_sortstash = hv_dup(proto_perl->Isortstash, param);
13341 PL_firstgv = gv_dup(proto_perl->Ifirstgv, param);
13342 PL_secondgv = gv_dup(proto_perl->Isecondgv, param);
13344 PL_stashcache = newHV();
13346 PL_watchaddr = (char **) ptr_table_fetch(PL_ptr_table,
13347 proto_perl->Iwatchaddr);
13348 PL_watchok = PL_watchaddr ? * PL_watchaddr : NULL;
13349 if (PL_debug && PL_watchaddr) {
13350 PerlIO_printf(Perl_debug_log,
13351 "WATCHING: %"UVxf" cloned as %"UVxf" with value %"UVxf"\n",
13352 PTR2UV(proto_perl->Iwatchaddr), PTR2UV(PL_watchaddr),
13353 PTR2UV(PL_watchok));
13356 PL_registered_mros = hv_dup_inc(proto_perl->Iregistered_mros, param);
13357 PL_blockhooks = av_dup_inc(proto_perl->Iblockhooks, param);
13358 PL_utf8_foldclosures = hv_dup_inc(proto_perl->Iutf8_foldclosures, param);
13360 /* Call the ->CLONE method, if it exists, for each of the stashes
13361 identified by sv_dup() above.
13363 while(av_len(param->stashes) != -1) {
13364 HV* const stash = MUTABLE_HV(av_shift(param->stashes));
13365 GV* const cloner = gv_fetchmethod_autoload(stash, "CLONE", 0);
13366 if (cloner && GvCV(cloner)) {
13371 mXPUSHs(newSVhek(HvNAME_HEK(stash)));
13373 call_sv(MUTABLE_SV(GvCV(cloner)), G_DISCARD);
13379 if (!(flags & CLONEf_KEEP_PTR_TABLE)) {
13380 ptr_table_free(PL_ptr_table);
13381 PL_ptr_table = NULL;
13384 if (!(flags & CLONEf_COPY_STACKS)) {
13385 unreferenced_to_tmp_stack(param->unreferenced);
13388 SvREFCNT_dec(param->stashes);
13390 /* orphaned? eg threads->new inside BEGIN or use */
13391 if (PL_compcv && ! SvREFCNT(PL_compcv)) {
13392 SvREFCNT_inc_simple_void(PL_compcv);
13393 SAVEFREESV(PL_compcv);
13400 S_unreferenced_to_tmp_stack(pTHX_ AV *const unreferenced)
13402 PERL_ARGS_ASSERT_UNREFERENCED_TO_TMP_STACK;
13404 if (AvFILLp(unreferenced) > -1) {
13405 SV **svp = AvARRAY(unreferenced);
13406 SV **const last = svp + AvFILLp(unreferenced);
13410 if (SvREFCNT(*svp) == 1)
13412 } while (++svp <= last);
13414 EXTEND_MORTAL(count);
13415 svp = AvARRAY(unreferenced);
13418 if (SvREFCNT(*svp) == 1) {
13419 /* Our reference is the only one to this SV. This means that
13420 in this thread, the scalar effectively has a 0 reference.
13421 That doesn't work (cleanup never happens), so donate our
13422 reference to it onto the save stack. */
13423 PL_tmps_stack[++PL_tmps_ix] = *svp;
13425 /* As an optimisation, because we are already walking the
13426 entire array, instead of above doing either
13427 SvREFCNT_inc(*svp) or *svp = &PL_sv_undef, we can instead
13428 release our reference to the scalar, so that at the end of
13429 the array owns zero references to the scalars it happens to
13430 point to. We are effectively converting the array from
13431 AvREAL() on to AvREAL() off. This saves the av_clear()
13432 (triggered by the SvREFCNT_dec(unreferenced) below) from
13433 walking the array a second time. */
13434 SvREFCNT_dec(*svp);
13437 } while (++svp <= last);
13438 AvREAL_off(unreferenced);
13440 SvREFCNT_dec(unreferenced);
13444 Perl_clone_params_del(CLONE_PARAMS *param)
13446 /* This seemingly funky ordering keeps the build with PERL_GLOBAL_STRUCT
13448 PerlInterpreter *const to = param->new_perl;
13450 PerlInterpreter *const was = PERL_GET_THX;
13452 PERL_ARGS_ASSERT_CLONE_PARAMS_DEL;
13458 SvREFCNT_dec(param->stashes);
13459 if (param->unreferenced)
13460 unreferenced_to_tmp_stack(param->unreferenced);
13470 Perl_clone_params_new(PerlInterpreter *const from, PerlInterpreter *const to)
13473 /* Need to play this game, as newAV() can call safesysmalloc(), and that
13474 does a dTHX; to get the context from thread local storage.
13475 FIXME - under PERL_CORE Newx(), Safefree() and friends should expand to
13476 a version that passes in my_perl. */
13477 PerlInterpreter *const was = PERL_GET_THX;
13478 CLONE_PARAMS *param;
13480 PERL_ARGS_ASSERT_CLONE_PARAMS_NEW;
13486 /* Given that we've set the context, we can do this unshared. */
13487 Newx(param, 1, CLONE_PARAMS);
13490 param->proto_perl = from;
13491 param->new_perl = to;
13492 param->stashes = (AV *)Perl_newSV_type(to, SVt_PVAV);
13493 AvREAL_off(param->stashes);
13494 param->unreferenced = (AV *)Perl_newSV_type(to, SVt_PVAV);
13502 #endif /* USE_ITHREADS */
13505 =head1 Unicode Support
13507 =for apidoc sv_recode_to_utf8
13509 The encoding is assumed to be an Encode object, on entry the PV
13510 of the sv is assumed to be octets in that encoding, and the sv
13511 will be converted into Unicode (and UTF-8).
13513 If the sv already is UTF-8 (or if it is not POK), or if the encoding
13514 is not a reference, nothing is done to the sv. If the encoding is not
13515 an C<Encode::XS> Encoding object, bad things will happen.
13516 (See F<lib/encoding.pm> and L<Encode>).
13518 The PV of the sv is returned.
13523 Perl_sv_recode_to_utf8(pTHX_ SV *sv, SV *encoding)
13527 PERL_ARGS_ASSERT_SV_RECODE_TO_UTF8;
13529 if (SvPOK(sv) && !SvUTF8(sv) && !IN_BYTES && SvROK(encoding)) {
13543 Passing sv_yes is wrong - it needs to be or'ed set of constants
13544 for Encode::XS, while UTf-8 decode (currently) assumes a true value means
13545 remove converted chars from source.
13547 Both will default the value - let them.
13549 XPUSHs(&PL_sv_yes);
13552 call_method("decode", G_SCALAR);
13556 s = SvPV_const(uni, len);
13557 if (s != SvPVX_const(sv)) {
13558 SvGROW(sv, len + 1);
13559 Move(s, SvPVX(sv), len + 1, char);
13560 SvCUR_set(sv, len);
13564 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
13565 /* clear pos and any utf8 cache */
13566 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
13569 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
13570 magic_setutf8(sv,mg); /* clear UTF8 cache */
13575 return SvPOKp(sv) ? SvPVX(sv) : NULL;
13579 =for apidoc sv_cat_decode
13581 The encoding is assumed to be an Encode object, the PV of the ssv is
13582 assumed to be octets in that encoding and decoding the input starts
13583 from the position which (PV + *offset) pointed to. The dsv will be
13584 concatenated the decoded UTF-8 string from ssv. Decoding will terminate
13585 when the string tstr appears in decoding output or the input ends on
13586 the PV of the ssv. The value which the offset points will be modified
13587 to the last input position on the ssv.
13589 Returns TRUE if the terminator was found, else returns FALSE.
13594 Perl_sv_cat_decode(pTHX_ SV *dsv, SV *encoding,
13595 SV *ssv, int *offset, char *tstr, int tlen)
13600 PERL_ARGS_ASSERT_SV_CAT_DECODE;
13602 if (SvPOK(ssv) && SvPOK(dsv) && SvROK(encoding) && offset) {
13613 offsv = newSViv(*offset);
13615 mXPUSHp(tstr, tlen);
13617 call_method("cat_decode", G_SCALAR);
13619 ret = SvTRUE(TOPs);
13620 *offset = SvIV(offsv);
13626 Perl_croak(aTHX_ "Invalid argument to sv_cat_decode");
13631 /* ---------------------------------------------------------------------
13633 * support functions for report_uninit()
13636 /* the maxiumum size of array or hash where we will scan looking
13637 * for the undefined element that triggered the warning */
13639 #define FUV_MAX_SEARCH_SIZE 1000
13641 /* Look for an entry in the hash whose value has the same SV as val;
13642 * If so, return a mortal copy of the key. */
13645 S_find_hash_subscript(pTHX_ const HV *const hv, const SV *const val)
13648 register HE **array;
13651 PERL_ARGS_ASSERT_FIND_HASH_SUBSCRIPT;
13653 if (!hv || SvMAGICAL(hv) || !HvARRAY(hv) ||
13654 (HvTOTALKEYS(hv) > FUV_MAX_SEARCH_SIZE))
13657 array = HvARRAY(hv);
13659 for (i=HvMAX(hv); i>0; i--) {
13660 register HE *entry;
13661 for (entry = array[i]; entry; entry = HeNEXT(entry)) {
13662 if (HeVAL(entry) != val)
13664 if ( HeVAL(entry) == &PL_sv_undef ||
13665 HeVAL(entry) == &PL_sv_placeholder)
13669 if (HeKLEN(entry) == HEf_SVKEY)
13670 return sv_mortalcopy(HeKEY_sv(entry));
13671 return sv_2mortal(newSVhek(HeKEY_hek(entry)));
13677 /* Look for an entry in the array whose value has the same SV as val;
13678 * If so, return the index, otherwise return -1. */
13681 S_find_array_subscript(pTHX_ const AV *const av, const SV *const val)
13685 PERL_ARGS_ASSERT_FIND_ARRAY_SUBSCRIPT;
13687 if (!av || SvMAGICAL(av) || !AvARRAY(av) ||
13688 (AvFILLp(av) > FUV_MAX_SEARCH_SIZE))
13691 if (val != &PL_sv_undef) {
13692 SV ** const svp = AvARRAY(av);
13695 for (i=AvFILLp(av); i>=0; i--)
13702 /* S_varname(): return the name of a variable, optionally with a subscript.
13703 * If gv is non-zero, use the name of that global, along with gvtype (one
13704 * of "$", "@", "%"); otherwise use the name of the lexical at pad offset
13705 * targ. Depending on the value of the subscript_type flag, return:
13708 #define FUV_SUBSCRIPT_NONE 1 /* "@foo" */
13709 #define FUV_SUBSCRIPT_ARRAY 2 /* "$foo[aindex]" */
13710 #define FUV_SUBSCRIPT_HASH 3 /* "$foo{keyname}" */
13711 #define FUV_SUBSCRIPT_WITHIN 4 /* "within @foo" */
13714 S_varname(pTHX_ const GV *const gv, const char gvtype, PADOFFSET targ,
13715 const SV *const keyname, I32 aindex, int subscript_type)
13718 SV * const name = sv_newmortal();
13721 buffer[0] = gvtype;
13724 /* as gv_fullname4(), but add literal '^' for $^FOO names */
13726 gv_fullname4(name, gv, buffer, 0);
13728 if ((unsigned int)SvPVX(name)[1] <= 26) {
13730 buffer[1] = SvPVX(name)[1] + 'A' - 1;
13732 /* Swap the 1 unprintable control character for the 2 byte pretty
13733 version - ie substr($name, 1, 1) = $buffer; */
13734 sv_insert(name, 1, 1, buffer, 2);
13738 CV * const cv = find_runcv(NULL);
13742 if (!cv || !CvPADLIST(cv))
13744 av = MUTABLE_AV((*av_fetch(CvPADLIST(cv), 0, FALSE)));
13745 sv = *av_fetch(av, targ, FALSE);
13746 sv_setpvn(name, SvPV_nolen_const(sv), SvCUR(sv));
13749 if (subscript_type == FUV_SUBSCRIPT_HASH) {
13750 SV * const sv = newSV(0);
13751 *SvPVX(name) = '$';
13752 Perl_sv_catpvf(aTHX_ name, "{%s}",
13753 pv_display(sv,SvPVX_const(keyname), SvCUR(keyname), 0, 32));
13756 else if (subscript_type == FUV_SUBSCRIPT_ARRAY) {
13757 *SvPVX(name) = '$';
13758 Perl_sv_catpvf(aTHX_ name, "[%"IVdf"]", (IV)aindex);
13760 else if (subscript_type == FUV_SUBSCRIPT_WITHIN) {
13761 /* We know that name has no magic, so can use 0 instead of SV_GMAGIC */
13762 Perl_sv_insert_flags(aTHX_ name, 0, 0, STR_WITH_LEN("within "), 0);
13770 =for apidoc find_uninit_var
13772 Find the name of the undefined variable (if any) that caused the operator o
13773 to issue a "Use of uninitialized value" warning.
13774 If match is true, only return a name if it's value matches uninit_sv.
13775 So roughly speaking, if a unary operator (such as OP_COS) generates a
13776 warning, then following the direct child of the op may yield an
13777 OP_PADSV or OP_GV that gives the name of the undefined variable. On the
13778 other hand, with OP_ADD there are two branches to follow, so we only print
13779 the variable name if we get an exact match.
13781 The name is returned as a mortal SV.
13783 Assumes that PL_op is the op that originally triggered the error, and that
13784 PL_comppad/PL_curpad points to the currently executing pad.
13790 S_find_uninit_var(pTHX_ const OP *const obase, const SV *const uninit_sv,
13796 const OP *o, *o2, *kid;
13798 if (!obase || (match && (!uninit_sv || uninit_sv == &PL_sv_undef ||
13799 uninit_sv == &PL_sv_placeholder)))
13802 switch (obase->op_type) {
13809 const bool pad = (obase->op_type == OP_PADAV || obase->op_type == OP_PADHV);
13810 const bool hash = (obase->op_type == OP_PADHV || obase->op_type == OP_RV2HV);
13813 int subscript_type = FUV_SUBSCRIPT_WITHIN;
13815 if (pad) { /* @lex, %lex */
13816 sv = PAD_SVl(obase->op_targ);
13820 if (cUNOPx(obase)->op_first->op_type == OP_GV) {
13821 /* @global, %global */
13822 gv = cGVOPx_gv(cUNOPx(obase)->op_first);
13825 sv = hash ? MUTABLE_SV(GvHV(gv)): MUTABLE_SV(GvAV(gv));
13827 else /* @{expr}, %{expr} */
13828 return find_uninit_var(cUNOPx(obase)->op_first,
13832 /* attempt to find a match within the aggregate */
13834 keysv = find_hash_subscript((const HV*)sv, uninit_sv);
13836 subscript_type = FUV_SUBSCRIPT_HASH;
13839 index = find_array_subscript((const AV *)sv, uninit_sv);
13841 subscript_type = FUV_SUBSCRIPT_ARRAY;
13844 if (match && subscript_type == FUV_SUBSCRIPT_WITHIN)
13847 return varname(gv, hash ? '%' : '@', obase->op_targ,
13848 keysv, index, subscript_type);
13852 if (cUNOPx(obase)->op_first->op_type == OP_GV) {
13854 gv = cGVOPx_gv(cUNOPx(obase)->op_first);
13855 if (!gv || !GvSTASH(gv))
13857 if (match && (GvSV(gv) != uninit_sv))
13859 return varname(gv, '$', 0, NULL, 0, FUV_SUBSCRIPT_NONE);
13862 return find_uninit_var(cUNOPx(obase)->op_first, uninit_sv, 1);
13865 if (match && PAD_SVl(obase->op_targ) != uninit_sv)
13867 return varname(NULL, '$', obase->op_targ,
13868 NULL, 0, FUV_SUBSCRIPT_NONE);
13871 gv = cGVOPx_gv(obase);
13872 if (!gv || (match && GvSV(gv) != uninit_sv) || !GvSTASH(gv))
13874 return varname(gv, '$', 0, NULL, 0, FUV_SUBSCRIPT_NONE);
13876 case OP_AELEMFAST_LEX:
13879 AV *av = MUTABLE_AV(PAD_SV(obase->op_targ));
13880 if (!av || SvRMAGICAL(av))
13882 svp = av_fetch(av, (I32)obase->op_private, FALSE);
13883 if (!svp || *svp != uninit_sv)
13886 return varname(NULL, '$', obase->op_targ,
13887 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
13890 gv = cGVOPx_gv(obase);
13895 AV *const av = GvAV(gv);
13896 if (!av || SvRMAGICAL(av))
13898 svp = av_fetch(av, (I32)obase->op_private, FALSE);
13899 if (!svp || *svp != uninit_sv)
13902 return varname(gv, '$', 0,
13903 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
13908 o = cUNOPx(obase)->op_first;
13909 if (!o || o->op_type != OP_NULL ||
13910 ! (o->op_targ == OP_AELEM || o->op_targ == OP_HELEM))
13912 return find_uninit_var(cBINOPo->op_last, uninit_sv, match);
13917 bool negate = FALSE;
13919 if (PL_op == obase)
13920 /* $a[uninit_expr] or $h{uninit_expr} */
13921 return find_uninit_var(cBINOPx(obase)->op_last, uninit_sv, match);
13924 o = cBINOPx(obase)->op_first;
13925 kid = cBINOPx(obase)->op_last;
13927 /* get the av or hv, and optionally the gv */
13929 if (o->op_type == OP_PADAV || o->op_type == OP_PADHV) {
13930 sv = PAD_SV(o->op_targ);
13932 else if ((o->op_type == OP_RV2AV || o->op_type == OP_RV2HV)
13933 && cUNOPo->op_first->op_type == OP_GV)
13935 gv = cGVOPx_gv(cUNOPo->op_first);
13939 == OP_RV2HV ? MUTABLE_SV(GvHV(gv)) : MUTABLE_SV(GvAV(gv));
13944 if (kid && kid->op_type == OP_NEGATE) {
13946 kid = cUNOPx(kid)->op_first;
13949 if (kid && kid->op_type == OP_CONST && SvOK(cSVOPx_sv(kid))) {
13950 /* index is constant */
13953 kidsv = sv_2mortal(newSVpvs("-"));
13954 sv_catsv(kidsv, cSVOPx_sv(kid));
13957 kidsv = cSVOPx_sv(kid);
13961 if (obase->op_type == OP_HELEM) {
13962 HE* he = hv_fetch_ent(MUTABLE_HV(sv), kidsv, 0, 0);
13963 if (!he || HeVAL(he) != uninit_sv)
13967 SV * const * const svp = av_fetch(MUTABLE_AV(sv),
13968 negate ? - SvIV(cSVOPx_sv(kid)) : SvIV(cSVOPx_sv(kid)),
13970 if (!svp || *svp != uninit_sv)
13974 if (obase->op_type == OP_HELEM)
13975 return varname(gv, '%', o->op_targ,
13976 kidsv, 0, FUV_SUBSCRIPT_HASH);
13978 return varname(gv, '@', o->op_targ, NULL,
13979 negate ? - SvIV(cSVOPx_sv(kid)) : SvIV(cSVOPx_sv(kid)),
13980 FUV_SUBSCRIPT_ARRAY);
13983 /* index is an expression;
13984 * attempt to find a match within the aggregate */
13985 if (obase->op_type == OP_HELEM) {
13986 SV * const keysv = find_hash_subscript((const HV*)sv, uninit_sv);
13988 return varname(gv, '%', o->op_targ,
13989 keysv, 0, FUV_SUBSCRIPT_HASH);
13993 = find_array_subscript((const AV *)sv, uninit_sv);
13995 return varname(gv, '@', o->op_targ,
13996 NULL, index, FUV_SUBSCRIPT_ARRAY);
14001 (o->op_type == OP_PADAV || o->op_type == OP_RV2AV)
14003 o->op_targ, NULL, 0, FUV_SUBSCRIPT_WITHIN);
14009 /* only examine RHS */
14010 return find_uninit_var(cBINOPx(obase)->op_first, uninit_sv, match);
14013 o = cUNOPx(obase)->op_first;
14014 if (o->op_type == OP_PUSHMARK)
14017 if (!o->op_sibling) {
14018 /* one-arg version of open is highly magical */
14020 if (o->op_type == OP_GV) { /* open FOO; */
14022 if (match && GvSV(gv) != uninit_sv)
14024 return varname(gv, '$', 0,
14025 NULL, 0, FUV_SUBSCRIPT_NONE);
14027 /* other possibilities not handled are:
14028 * open $x; or open my $x; should return '${*$x}'
14029 * open expr; should return '$'.expr ideally
14035 /* ops where $_ may be an implicit arg */
14039 if ( !(obase->op_flags & OPf_STACKED)) {
14040 if (uninit_sv == ((obase->op_private & OPpTARGET_MY)
14041 ? PAD_SVl(obase->op_targ)
14044 sv = sv_newmortal();
14045 sv_setpvs(sv, "$_");
14054 match = 1; /* print etc can return undef on defined args */
14055 /* skip filehandle as it can't produce 'undef' warning */
14056 o = cUNOPx(obase)->op_first;
14057 if ((obase->op_flags & OPf_STACKED) && o->op_type == OP_PUSHMARK)
14058 o = o->op_sibling->op_sibling;
14062 case OP_ENTEREVAL: /* could be eval $undef or $x='$undef'; eval $x */
14063 case OP_CUSTOM: /* XS or custom code could trigger random warnings */
14065 /* the following ops are capable of returning PL_sv_undef even for
14066 * defined arg(s) */
14085 case OP_GETPEERNAME:
14133 case OP_SMARTMATCH:
14142 /* XXX tmp hack: these two may call an XS sub, and currently
14143 XS subs don't have a SUB entry on the context stack, so CV and
14144 pad determination goes wrong, and BAD things happen. So, just
14145 don't try to determine the value under those circumstances.
14146 Need a better fix at dome point. DAPM 11/2007 */
14152 GV * const gv = gv_fetchpvs(".", GV_NOTQUAL, SVt_PV);
14153 if (gv && GvSV(gv) == uninit_sv)
14154 return newSVpvs_flags("$.", SVs_TEMP);
14159 /* def-ness of rval pos() is independent of the def-ness of its arg */
14160 if ( !(obase->op_flags & OPf_MOD))
14165 if (SvROK(PL_rs) && uninit_sv == SvRV(PL_rs))
14166 return newSVpvs_flags("${$/}", SVs_TEMP);
14171 if (!(obase->op_flags & OPf_KIDS))
14173 o = cUNOPx(obase)->op_first;
14179 /* if all except one arg are constant, or have no side-effects,
14180 * or are optimized away, then it's unambiguous */
14182 for (kid=o; kid; kid = kid->op_sibling) {
14184 const OPCODE type = kid->op_type;
14185 if ( (type == OP_CONST && SvOK(cSVOPx_sv(kid)))
14186 || (type == OP_NULL && ! (kid->op_flags & OPf_KIDS))
14187 || (type == OP_PUSHMARK)
14189 /* @$a and %$a, but not @a or %a */
14190 (type == OP_RV2AV || type == OP_RV2HV)
14191 && cUNOPx(kid)->op_first
14192 && cUNOPx(kid)->op_first->op_type != OP_GV
14197 if (o2) { /* more than one found */
14204 return find_uninit_var(o2, uninit_sv, match);
14206 /* scan all args */
14208 sv = find_uninit_var(o, uninit_sv, 1);
14220 =for apidoc report_uninit
14222 Print appropriate "Use of uninitialized variable" warning
14228 Perl_report_uninit(pTHX_ const SV *uninit_sv)
14232 SV* varname = NULL;
14234 varname = find_uninit_var(PL_op, uninit_sv,0);
14236 sv_insert(varname, 0, 0, " ", 1);
14238 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
14239 varname ? SvPV_nolen_const(varname) : "",
14240 " in ", OP_DESC(PL_op));
14243 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
14249 * c-indentation-style: bsd
14250 * c-basic-offset: 4
14251 * indent-tabs-mode: t
14254 * ex: set ts=8 sts=4 sw=4 noet: