3 * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, 2000,
4 * 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009 by Larry Wall
7 * You may distribute under the terms of either the GNU General Public
8 * License or the Artistic License, as specified in the README file.
13 * 'I wonder what the Entish is for "yes" and "no",' he thought.
16 * [p.480 of _The Lord of the Rings_, III/iv: "Treebeard"]
22 * This file contains the code that creates, manipulates and destroys
23 * scalar values (SVs). The other types (AV, HV, GV, etc.) reuse the
24 * structure of an SV, so their creation and destruction is handled
25 * here; higher-level functions are in av.c, hv.c, and so on. Opcode
26 * level functions (eg. substr, split, join) for each of the types are
36 # if __STDC_VERSION__ >= 199901L && !defined(VMS)
47 /* Missing proto on LynxOS */
48 char *gconvert(double, int, int, char *);
51 #ifdef PERL_UTF8_CACHE_ASSERT
52 /* if adding more checks watch out for the following tests:
53 * t/op/index.t t/op/length.t t/op/pat.t t/op/substr.t
54 * lib/utf8.t lib/Unicode/Collate/t/index.t
57 # define ASSERT_UTF8_CACHE(cache) \
58 STMT_START { if (cache) { assert((cache)[0] <= (cache)[1]); \
59 assert((cache)[2] <= (cache)[3]); \
60 assert((cache)[3] <= (cache)[1]);} \
63 # define ASSERT_UTF8_CACHE(cache) NOOP
66 #ifdef PERL_OLD_COPY_ON_WRITE
67 #define SV_COW_NEXT_SV(sv) INT2PTR(SV *,SvUVX(sv))
68 #define SV_COW_NEXT_SV_SET(current,next) SvUV_set(current, PTR2UV(next))
69 /* This is a pessimistic view. Scalar must be purely a read-write PV to copy-
73 /* ============================================================================
75 =head1 Allocation and deallocation of SVs.
77 An SV (or AV, HV, etc.) is allocated in two parts: the head (struct
78 sv, av, hv...) contains type and reference count information, and for
79 many types, a pointer to the body (struct xrv, xpv, xpviv...), which
80 contains fields specific to each type. Some types store all they need
81 in the head, so don't have a body.
83 In all but the most memory-paranoid configurations (ex: PURIFY), heads
84 and bodies are allocated out of arenas, which by default are
85 approximately 4K chunks of memory parcelled up into N heads or bodies.
86 Sv-bodies are allocated by their sv-type, guaranteeing size
87 consistency needed to allocate safely from arrays.
89 For SV-heads, the first slot in each arena is reserved, and holds a
90 link to the next arena, some flags, and a note of the number of slots.
91 Snaked through each arena chain is a linked list of free items; when
92 this becomes empty, an extra arena is allocated and divided up into N
93 items which are threaded into the free list.
95 SV-bodies are similar, but they use arena-sets by default, which
96 separate the link and info from the arena itself, and reclaim the 1st
97 slot in the arena. SV-bodies are further described later.
99 The following global variables are associated with arenas:
101 PL_sv_arenaroot pointer to list of SV arenas
102 PL_sv_root pointer to list of free SV structures
104 PL_body_arenas head of linked-list of body arenas
105 PL_body_roots[] array of pointers to list of free bodies of svtype
106 arrays are indexed by the svtype needed
108 A few special SV heads are not allocated from an arena, but are
109 instead directly created in the interpreter structure, eg PL_sv_undef.
110 The size of arenas can be changed from the default by setting
111 PERL_ARENA_SIZE appropriately at compile time.
113 The SV arena serves the secondary purpose of allowing still-live SVs
114 to be located and destroyed during final cleanup.
116 At the lowest level, the macros new_SV() and del_SV() grab and free
117 an SV head. (If debugging with -DD, del_SV() calls the function S_del_sv()
118 to return the SV to the free list with error checking.) new_SV() calls
119 more_sv() / sv_add_arena() to add an extra arena if the free list is empty.
120 SVs in the free list have their SvTYPE field set to all ones.
122 At the time of very final cleanup, sv_free_arenas() is called from
123 perl_destruct() to physically free all the arenas allocated since the
124 start of the interpreter.
126 The function visit() scans the SV arenas list, and calls a specified
127 function for each SV it finds which is still live - ie which has an SvTYPE
128 other than all 1's, and a non-zero SvREFCNT. visit() is used by the
129 following functions (specified as [function that calls visit()] / [function
130 called by visit() for each SV]):
132 sv_report_used() / do_report_used()
133 dump all remaining SVs (debugging aid)
135 sv_clean_objs() / do_clean_objs(),do_clean_named_objs(),
136 do_clean_named_io_objs()
137 Attempt to free all objects pointed to by RVs,
138 and try to do the same for all objects indirectly
139 referenced by typeglobs too. Called once from
140 perl_destruct(), prior to calling sv_clean_all()
143 sv_clean_all() / do_clean_all()
144 SvREFCNT_dec(sv) each remaining SV, possibly
145 triggering an sv_free(). It also sets the
146 SVf_BREAK flag on the SV to indicate that the
147 refcnt has been artificially lowered, and thus
148 stopping sv_free() from giving spurious warnings
149 about SVs which unexpectedly have a refcnt
150 of zero. called repeatedly from perl_destruct()
151 until there are no SVs left.
153 =head2 Arena allocator API Summary
155 Private API to rest of sv.c
159 new_XPVNV(), del_XPVGV(),
164 sv_report_used(), sv_clean_objs(), sv_clean_all(), sv_free_arenas()
168 * ========================================================================= */
171 * "A time to plant, and a time to uproot what was planted..."
175 # define MEM_LOG_NEW_SV(sv, file, line, func) \
176 Perl_mem_log_new_sv(sv, file, line, func)
177 # define MEM_LOG_DEL_SV(sv, file, line, func) \
178 Perl_mem_log_del_sv(sv, file, line, func)
180 # define MEM_LOG_NEW_SV(sv, file, line, func) NOOP
181 # define MEM_LOG_DEL_SV(sv, file, line, func) NOOP
184 #ifdef DEBUG_LEAKING_SCALARS
185 # define FREE_SV_DEBUG_FILE(sv) Safefree((sv)->sv_debug_file)
186 # define DEBUG_SV_SERIAL(sv) \
187 DEBUG_m(PerlIO_printf(Perl_debug_log, "0x%"UVxf": (%05ld) del_SV\n", \
188 PTR2UV(sv), (long)(sv)->sv_debug_serial))
190 # define FREE_SV_DEBUG_FILE(sv)
191 # define DEBUG_SV_SERIAL(sv) NOOP
195 # define SvARENA_CHAIN(sv) ((sv)->sv_u.svu_rv)
196 # define SvARENA_CHAIN_SET(sv,val) (sv)->sv_u.svu_rv = MUTABLE_SV((val))
197 /* Whilst I'd love to do this, it seems that things like to check on
199 # define POSION_SV_HEAD(sv) PoisonNew(sv, 1, struct STRUCT_SV)
201 # define POSION_SV_HEAD(sv) PoisonNew(&SvANY(sv), 1, void *), \
202 PoisonNew(&SvREFCNT(sv), 1, U32)
204 # define SvARENA_CHAIN(sv) SvANY(sv)
205 # define SvARENA_CHAIN_SET(sv,val) SvANY(sv) = (void *)(val)
206 # define POSION_SV_HEAD(sv)
209 /* Mark an SV head as unused, and add to free list.
211 * If SVf_BREAK is set, skip adding it to the free list, as this SV had
212 * its refcount artificially decremented during global destruction, so
213 * there may be dangling pointers to it. The last thing we want in that
214 * case is for it to be reused. */
216 #define plant_SV(p) \
218 const U32 old_flags = SvFLAGS(p); \
219 MEM_LOG_DEL_SV(p, __FILE__, __LINE__, FUNCTION__); \
220 DEBUG_SV_SERIAL(p); \
221 FREE_SV_DEBUG_FILE(p); \
223 SvFLAGS(p) = SVTYPEMASK; \
224 if (!(old_flags & SVf_BREAK)) { \
225 SvARENA_CHAIN_SET(p, PL_sv_root); \
231 #define uproot_SV(p) \
234 PL_sv_root = MUTABLE_SV(SvARENA_CHAIN(p)); \
239 /* make some more SVs by adding another arena */
246 char *chunk; /* must use New here to match call to */
247 Newx(chunk,PERL_ARENA_SIZE,char); /* Safefree() in sv_free_arenas() */
248 sv_add_arena(chunk, PERL_ARENA_SIZE, 0);
253 /* new_SV(): return a new, empty SV head */
255 #ifdef DEBUG_LEAKING_SCALARS
256 /* provide a real function for a debugger to play with */
258 S_new_SV(pTHX_ const char *file, int line, const char *func)
265 sv = S_more_sv(aTHX);
269 sv->sv_debug_optype = PL_op ? PL_op->op_type : 0;
270 sv->sv_debug_line = (U16) (PL_parser && PL_parser->copline != NOLINE
276 sv->sv_debug_inpad = 0;
277 sv->sv_debug_parent = NULL;
278 sv->sv_debug_file = PL_curcop ? savepv(CopFILE(PL_curcop)): NULL;
280 sv->sv_debug_serial = PL_sv_serial++;
282 MEM_LOG_NEW_SV(sv, file, line, func);
283 DEBUG_m(PerlIO_printf(Perl_debug_log, "0x%"UVxf": (%05ld) new_SV (from %s:%d [%s])\n",
284 PTR2UV(sv), (long)sv->sv_debug_serial, file, line, func));
288 # define new_SV(p) (p)=S_new_SV(aTHX_ __FILE__, __LINE__, FUNCTION__)
296 (p) = S_more_sv(aTHX); \
300 MEM_LOG_NEW_SV(p, __FILE__, __LINE__, FUNCTION__); \
305 /* del_SV(): return an empty SV head to the free list */
318 S_del_sv(pTHX_ SV *p)
322 PERL_ARGS_ASSERT_DEL_SV;
327 for (sva = PL_sv_arenaroot; sva; sva = MUTABLE_SV(SvANY(sva))) {
328 const SV * const sv = sva + 1;
329 const SV * const svend = &sva[SvREFCNT(sva)];
330 if (p >= sv && p < svend) {
336 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL),
337 "Attempt to free non-arena SV: 0x%"UVxf
338 pTHX__FORMAT, PTR2UV(p) pTHX__VALUE);
345 #else /* ! DEBUGGING */
347 #define del_SV(p) plant_SV(p)
349 #endif /* DEBUGGING */
353 =head1 SV Manipulation Functions
355 =for apidoc sv_add_arena
357 Given a chunk of memory, link it to the head of the list of arenas,
358 and split it into a list of free SVs.
364 S_sv_add_arena(pTHX_ char *const ptr, const U32 size, const U32 flags)
367 SV *const sva = MUTABLE_SV(ptr);
371 PERL_ARGS_ASSERT_SV_ADD_ARENA;
373 /* The first SV in an arena isn't an SV. */
374 SvANY(sva) = (void *) PL_sv_arenaroot; /* ptr to next arena */
375 SvREFCNT(sva) = size / sizeof(SV); /* number of SV slots */
376 SvFLAGS(sva) = flags; /* FAKE if not to be freed */
378 PL_sv_arenaroot = sva;
379 PL_sv_root = sva + 1;
381 svend = &sva[SvREFCNT(sva) - 1];
384 SvARENA_CHAIN_SET(sv, (sv + 1));
388 /* Must always set typemask because it's always checked in on cleanup
389 when the arenas are walked looking for objects. */
390 SvFLAGS(sv) = SVTYPEMASK;
393 SvARENA_CHAIN_SET(sv, 0);
397 SvFLAGS(sv) = SVTYPEMASK;
400 /* visit(): call the named function for each non-free SV in the arenas
401 * whose flags field matches the flags/mask args. */
404 S_visit(pTHX_ SVFUNC_t f, const U32 flags, const U32 mask)
410 PERL_ARGS_ASSERT_VISIT;
412 for (sva = PL_sv_arenaroot; sva; sva = MUTABLE_SV(SvANY(sva))) {
413 register const SV * const svend = &sva[SvREFCNT(sva)];
415 for (sv = sva + 1; sv < svend; ++sv) {
416 if (SvTYPE(sv) != (svtype)SVTYPEMASK
417 && (sv->sv_flags & mask) == flags
430 /* called by sv_report_used() for each live SV */
433 do_report_used(pTHX_ SV *const sv)
435 if (SvTYPE(sv) != (svtype)SVTYPEMASK) {
436 PerlIO_printf(Perl_debug_log, "****\n");
443 =for apidoc sv_report_used
445 Dump the contents of all SVs not yet freed. (Debugging aid).
451 Perl_sv_report_used(pTHX)
454 visit(do_report_used, 0, 0);
460 /* called by sv_clean_objs() for each live SV */
463 do_clean_objs(pTHX_ SV *const ref)
468 SV * const target = SvRV(ref);
469 if (SvOBJECT(target)) {
470 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning object ref:\n "), sv_dump(ref)));
471 if (SvWEAKREF(ref)) {
472 sv_del_backref(target, ref);
478 SvREFCNT_dec(target);
483 /* XXX Might want to check arrays, etc. */
487 /* clear any slots in a GV which hold objects - except IO;
488 * called by sv_clean_objs() for each live GV */
491 do_clean_named_objs(pTHX_ SV *const sv)
495 assert(SvTYPE(sv) == SVt_PVGV);
496 assert(isGV_with_GP(sv));
500 /* freeing GP entries may indirectly free the current GV;
501 * hold onto it while we mess with the GP slots */
504 if ( ((obj = GvSV(sv) )) && SvOBJECT(obj)) {
505 DEBUG_D((PerlIO_printf(Perl_debug_log,
506 "Cleaning named glob SV object:\n "), sv_dump(obj)));
510 if ( ((obj = MUTABLE_SV(GvAV(sv)) )) && SvOBJECT(obj)) {
511 DEBUG_D((PerlIO_printf(Perl_debug_log,
512 "Cleaning named glob AV object:\n "), sv_dump(obj)));
516 if ( ((obj = MUTABLE_SV(GvHV(sv)) )) && SvOBJECT(obj)) {
517 DEBUG_D((PerlIO_printf(Perl_debug_log,
518 "Cleaning named glob HV object:\n "), sv_dump(obj)));
522 if ( ((obj = MUTABLE_SV(GvCV(sv)) )) && SvOBJECT(obj)) {
523 DEBUG_D((PerlIO_printf(Perl_debug_log,
524 "Cleaning named glob CV object:\n "), sv_dump(obj)));
528 SvREFCNT_dec(sv); /* undo the inc above */
531 /* clear any IO slots in a GV which hold objects (except stderr, defout);
532 * called by sv_clean_objs() for each live GV */
535 do_clean_named_io_objs(pTHX_ SV *const sv)
539 assert(SvTYPE(sv) == SVt_PVGV);
540 assert(isGV_with_GP(sv));
541 if (!GvGP(sv) || sv == (SV*)PL_stderrgv || sv == (SV*)PL_defoutgv)
545 if ( ((obj = MUTABLE_SV(GvIO(sv)) )) && SvOBJECT(obj)) {
546 DEBUG_D((PerlIO_printf(Perl_debug_log,
547 "Cleaning named glob IO object:\n "), sv_dump(obj)));
551 SvREFCNT_dec(sv); /* undo the inc above */
554 /* Void wrapper to pass to visit() */
556 do_curse(pTHX_ SV * const sv) {
557 if ((PL_stderrgv && GvGP(PL_stderrgv) && (SV*)GvIO(PL_stderrgv) == sv)
558 || (PL_defoutgv && GvGP(PL_defoutgv) && (SV*)GvIO(PL_defoutgv) == sv))
564 =for apidoc sv_clean_objs
566 Attempt to destroy all objects not yet freed
572 Perl_sv_clean_objs(pTHX)
576 PL_in_clean_objs = TRUE;
577 visit(do_clean_objs, SVf_ROK, SVf_ROK);
578 /* Some barnacles may yet remain, clinging to typeglobs.
579 * Run the non-IO destructors first: they may want to output
580 * error messages, close files etc */
581 visit(do_clean_named_objs, SVt_PVGV|SVpgv_GP, SVTYPEMASK|SVp_POK|SVpgv_GP);
582 visit(do_clean_named_io_objs, SVt_PVGV|SVpgv_GP, SVTYPEMASK|SVp_POK|SVpgv_GP);
583 /* And if there are some very tenacious barnacles clinging to arrays,
584 closures, or what have you.... */
585 visit(do_curse, SVs_OBJECT, SVs_OBJECT);
586 olddef = PL_defoutgv;
587 PL_defoutgv = NULL; /* disable skip of PL_defoutgv */
588 if (olddef && isGV_with_GP(olddef))
589 do_clean_named_io_objs(aTHX_ MUTABLE_SV(olddef));
590 olderr = PL_stderrgv;
591 PL_stderrgv = NULL; /* disable skip of PL_stderrgv */
592 if (olderr && isGV_with_GP(olderr))
593 do_clean_named_io_objs(aTHX_ MUTABLE_SV(olderr));
594 SvREFCNT_dec(olddef);
595 PL_in_clean_objs = FALSE;
598 /* called by sv_clean_all() for each live SV */
601 do_clean_all(pTHX_ SV *const sv)
604 if (sv == (const SV *) PL_fdpid || sv == (const SV *)PL_strtab) {
605 /* don't clean pid table and strtab */
608 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning loops: SV at 0x%"UVxf"\n", PTR2UV(sv)) ));
609 SvFLAGS(sv) |= SVf_BREAK;
614 =for apidoc sv_clean_all
616 Decrement the refcnt of each remaining SV, possibly triggering a
617 cleanup. This function may have to be called multiple times to free
618 SVs which are in complex self-referential hierarchies.
624 Perl_sv_clean_all(pTHX)
628 PL_in_clean_all = TRUE;
629 cleaned = visit(do_clean_all, 0,0);
634 ARENASETS: a meta-arena implementation which separates arena-info
635 into struct arena_set, which contains an array of struct
636 arena_descs, each holding info for a single arena. By separating
637 the meta-info from the arena, we recover the 1st slot, formerly
638 borrowed for list management. The arena_set is about the size of an
639 arena, avoiding the needless malloc overhead of a naive linked-list.
641 The cost is 1 arena-set malloc per ~320 arena-mallocs, + the unused
642 memory in the last arena-set (1/2 on average). In trade, we get
643 back the 1st slot in each arena (ie 1.7% of a CV-arena, less for
644 smaller types). The recovery of the wasted space allows use of
645 small arenas for large, rare body types, by changing array* fields
646 in body_details_by_type[] below.
649 char *arena; /* the raw storage, allocated aligned */
650 size_t size; /* its size ~4k typ */
651 svtype utype; /* bodytype stored in arena */
656 /* Get the maximum number of elements in set[] such that struct arena_set
657 will fit within PERL_ARENA_SIZE, which is probably just under 4K, and
658 therefore likely to be 1 aligned memory page. */
660 #define ARENAS_PER_SET ((PERL_ARENA_SIZE - sizeof(struct arena_set*) \
661 - 2 * sizeof(int)) / sizeof (struct arena_desc))
664 struct arena_set* next;
665 unsigned int set_size; /* ie ARENAS_PER_SET */
666 unsigned int curr; /* index of next available arena-desc */
667 struct arena_desc set[ARENAS_PER_SET];
671 =for apidoc sv_free_arenas
673 Deallocate the memory used by all arenas. Note that all the individual SV
674 heads and bodies within the arenas must already have been freed.
679 Perl_sv_free_arenas(pTHX)
686 /* Free arenas here, but be careful about fake ones. (We assume
687 contiguity of the fake ones with the corresponding real ones.) */
689 for (sva = PL_sv_arenaroot; sva; sva = svanext) {
690 svanext = MUTABLE_SV(SvANY(sva));
691 while (svanext && SvFAKE(svanext))
692 svanext = MUTABLE_SV(SvANY(svanext));
699 struct arena_set *aroot = (struct arena_set*) PL_body_arenas;
702 struct arena_set *current = aroot;
705 assert(aroot->set[i].arena);
706 Safefree(aroot->set[i].arena);
714 i = PERL_ARENA_ROOTS_SIZE;
716 PL_body_roots[i] = 0;
723 Here are mid-level routines that manage the allocation of bodies out
724 of the various arenas. There are 5 kinds of arenas:
726 1. SV-head arenas, which are discussed and handled above
727 2. regular body arenas
728 3. arenas for reduced-size bodies
731 Arena types 2 & 3 are chained by body-type off an array of
732 arena-root pointers, which is indexed by svtype. Some of the
733 larger/less used body types are malloced singly, since a large
734 unused block of them is wasteful. Also, several svtypes dont have
735 bodies; the data fits into the sv-head itself. The arena-root
736 pointer thus has a few unused root-pointers (which may be hijacked
737 later for arena types 4,5)
739 3 differs from 2 as an optimization; some body types have several
740 unused fields in the front of the structure (which are kept in-place
741 for consistency). These bodies can be allocated in smaller chunks,
742 because the leading fields arent accessed. Pointers to such bodies
743 are decremented to point at the unused 'ghost' memory, knowing that
744 the pointers are used with offsets to the real memory.
747 =head1 SV-Body Allocation
749 Allocation of SV-bodies is similar to SV-heads, differing as follows;
750 the allocation mechanism is used for many body types, so is somewhat
751 more complicated, it uses arena-sets, and has no need for still-live
754 At the outermost level, (new|del)_X*V macros return bodies of the
755 appropriate type. These macros call either (new|del)_body_type or
756 (new|del)_body_allocated macro pairs, depending on specifics of the
757 type. Most body types use the former pair, the latter pair is used to
758 allocate body types with "ghost fields".
760 "ghost fields" are fields that are unused in certain types, and
761 consequently don't need to actually exist. They are declared because
762 they're part of a "base type", which allows use of functions as
763 methods. The simplest examples are AVs and HVs, 2 aggregate types
764 which don't use the fields which support SCALAR semantics.
766 For these types, the arenas are carved up into appropriately sized
767 chunks, we thus avoid wasted memory for those unaccessed members.
768 When bodies are allocated, we adjust the pointer back in memory by the
769 size of the part not allocated, so it's as if we allocated the full
770 structure. (But things will all go boom if you write to the part that
771 is "not there", because you'll be overwriting the last members of the
772 preceding structure in memory.)
774 We calculate the correction using the STRUCT_OFFSET macro on the first
775 member present. If the allocated structure is smaller (no initial NV
776 actually allocated) then the net effect is to subtract the size of the NV
777 from the pointer, to return a new pointer as if an initial NV were actually
778 allocated. (We were using structures named *_allocated for this, but
779 this turned out to be a subtle bug, because a structure without an NV
780 could have a lower alignment constraint, but the compiler is allowed to
781 optimised accesses based on the alignment constraint of the actual pointer
782 to the full structure, for example, using a single 64 bit load instruction
783 because it "knows" that two adjacent 32 bit members will be 8-byte aligned.)
785 This is the same trick as was used for NV and IV bodies. Ironically it
786 doesn't need to be used for NV bodies any more, because NV is now at
787 the start of the structure. IV bodies don't need it either, because
788 they are no longer allocated.
790 In turn, the new_body_* allocators call S_new_body(), which invokes
791 new_body_inline macro, which takes a lock, and takes a body off the
792 linked list at PL_body_roots[sv_type], calling Perl_more_bodies() if
793 necessary to refresh an empty list. Then the lock is released, and
794 the body is returned.
796 Perl_more_bodies allocates a new arena, and carves it up into an array of N
797 bodies, which it strings into a linked list. It looks up arena-size
798 and body-size from the body_details table described below, thus
799 supporting the multiple body-types.
801 If PURIFY is defined, or PERL_ARENA_SIZE=0, arenas are not used, and
802 the (new|del)_X*V macros are mapped directly to malloc/free.
804 For each sv-type, struct body_details bodies_by_type[] carries
805 parameters which control these aspects of SV handling:
807 Arena_size determines whether arenas are used for this body type, and if
808 so, how big they are. PURIFY or PERL_ARENA_SIZE=0 set this field to
809 zero, forcing individual mallocs and frees.
811 Body_size determines how big a body is, and therefore how many fit into
812 each arena. Offset carries the body-pointer adjustment needed for
813 "ghost fields", and is used in *_allocated macros.
815 But its main purpose is to parameterize info needed in
816 Perl_sv_upgrade(). The info here dramatically simplifies the function
817 vs the implementation in 5.8.8, making it table-driven. All fields
818 are used for this, except for arena_size.
820 For the sv-types that have no bodies, arenas are not used, so those
821 PL_body_roots[sv_type] are unused, and can be overloaded. In
822 something of a special case, SVt_NULL is borrowed for HE arenas;
823 PL_body_roots[HE_SVSLOT=SVt_NULL] is filled by S_more_he, but the
824 bodies_by_type[SVt_NULL] slot is not used, as the table is not
829 struct body_details {
830 U8 body_size; /* Size to allocate */
831 U8 copy; /* Size of structure to copy (may be shorter) */
833 unsigned int type : 4; /* We have space for a sanity check. */
834 unsigned int cant_upgrade : 1; /* Cannot upgrade this type */
835 unsigned int zero_nv : 1; /* zero the NV when upgrading from this */
836 unsigned int arena : 1; /* Allocated from an arena */
837 size_t arena_size; /* Size of arena to allocate */
845 /* With -DPURFIY we allocate everything directly, and don't use arenas.
846 This seems a rather elegant way to simplify some of the code below. */
847 #define HASARENA FALSE
849 #define HASARENA TRUE
851 #define NOARENA FALSE
853 /* Size the arenas to exactly fit a given number of bodies. A count
854 of 0 fits the max number bodies into a PERL_ARENA_SIZE.block,
855 simplifying the default. If count > 0, the arena is sized to fit
856 only that many bodies, allowing arenas to be used for large, rare
857 bodies (XPVFM, XPVIO) without undue waste. The arena size is
858 limited by PERL_ARENA_SIZE, so we can safely oversize the
861 #define FIT_ARENA0(body_size) \
862 ((size_t)(PERL_ARENA_SIZE / body_size) * body_size)
863 #define FIT_ARENAn(count,body_size) \
864 ( count * body_size <= PERL_ARENA_SIZE) \
865 ? count * body_size \
866 : FIT_ARENA0 (body_size)
867 #define FIT_ARENA(count,body_size) \
869 ? FIT_ARENAn (count, body_size) \
870 : FIT_ARENA0 (body_size)
872 /* Calculate the length to copy. Specifically work out the length less any
873 final padding the compiler needed to add. See the comment in sv_upgrade
874 for why copying the padding proved to be a bug. */
876 #define copy_length(type, last_member) \
877 STRUCT_OFFSET(type, last_member) \
878 + sizeof (((type*)SvANY((const SV *)0))->last_member)
880 static const struct body_details bodies_by_type[] = {
881 /* HEs use this offset for their arena. */
882 { 0, 0, 0, SVt_NULL, FALSE, NONV, NOARENA, 0 },
884 /* The bind placeholder pretends to be an RV for now.
885 Also it's marked as "can't upgrade" to stop anyone using it before it's
887 { 0, 0, 0, SVt_BIND, TRUE, NONV, NOARENA, 0 },
889 /* IVs are in the head, so the allocation size is 0. */
891 sizeof(IV), /* This is used to copy out the IV body. */
892 STRUCT_OFFSET(XPVIV, xiv_iv), SVt_IV, FALSE, NONV,
893 NOARENA /* IVS don't need an arena */, 0
896 { sizeof(NV), sizeof(NV),
897 STRUCT_OFFSET(XPVNV, xnv_u),
898 SVt_NV, FALSE, HADNV, HASARENA, FIT_ARENA(0, sizeof(NV)) },
900 { sizeof(XPV) - STRUCT_OFFSET(XPV, xpv_cur),
901 copy_length(XPV, xpv_len) - STRUCT_OFFSET(XPV, xpv_cur),
902 + STRUCT_OFFSET(XPV, xpv_cur),
903 SVt_PV, FALSE, NONV, HASARENA,
904 FIT_ARENA(0, sizeof(XPV) - STRUCT_OFFSET(XPV, xpv_cur)) },
906 { sizeof(XPVIV) - STRUCT_OFFSET(XPV, xpv_cur),
907 copy_length(XPVIV, xiv_u) - STRUCT_OFFSET(XPV, xpv_cur),
908 + STRUCT_OFFSET(XPV, xpv_cur),
909 SVt_PVIV, FALSE, NONV, HASARENA,
910 FIT_ARENA(0, sizeof(XPVIV) - STRUCT_OFFSET(XPV, xpv_cur)) },
912 { sizeof(XPVNV) - STRUCT_OFFSET(XPV, xpv_cur),
913 copy_length(XPVNV, xnv_u) - STRUCT_OFFSET(XPV, xpv_cur),
914 + STRUCT_OFFSET(XPV, xpv_cur),
915 SVt_PVNV, FALSE, HADNV, HASARENA,
916 FIT_ARENA(0, sizeof(XPVNV) - STRUCT_OFFSET(XPV, xpv_cur)) },
918 { sizeof(XPVMG), copy_length(XPVMG, xnv_u), 0, SVt_PVMG, FALSE, HADNV,
919 HASARENA, FIT_ARENA(0, sizeof(XPVMG)) },
924 SVt_REGEXP, FALSE, NONV, HASARENA,
925 FIT_ARENA(0, sizeof(regexp))
928 { sizeof(XPVGV), sizeof(XPVGV), 0, SVt_PVGV, TRUE, HADNV,
929 HASARENA, FIT_ARENA(0, sizeof(XPVGV)) },
931 { sizeof(XPVLV), sizeof(XPVLV), 0, SVt_PVLV, TRUE, HADNV,
932 HASARENA, FIT_ARENA(0, sizeof(XPVLV)) },
935 copy_length(XPVAV, xav_alloc),
937 SVt_PVAV, TRUE, NONV, HASARENA,
938 FIT_ARENA(0, sizeof(XPVAV)) },
941 copy_length(XPVHV, xhv_max),
943 SVt_PVHV, TRUE, NONV, HASARENA,
944 FIT_ARENA(0, sizeof(XPVHV)) },
949 SVt_PVCV, TRUE, NONV, HASARENA,
950 FIT_ARENA(0, sizeof(XPVCV)) },
955 SVt_PVFM, TRUE, NONV, NOARENA,
956 FIT_ARENA(20, sizeof(XPVFM)) },
961 SVt_PVIO, TRUE, NONV, HASARENA,
962 FIT_ARENA(24, sizeof(XPVIO)) },
965 #define new_body_allocated(sv_type) \
966 (void *)((char *)S_new_body(aTHX_ sv_type) \
967 - bodies_by_type[sv_type].offset)
969 /* return a thing to the free list */
971 #define del_body(thing, root) \
973 void ** const thing_copy = (void **)thing; \
974 *thing_copy = *root; \
975 *root = (void*)thing_copy; \
980 #define new_XNV() safemalloc(sizeof(XPVNV))
981 #define new_XPVNV() safemalloc(sizeof(XPVNV))
982 #define new_XPVMG() safemalloc(sizeof(XPVMG))
984 #define del_XPVGV(p) safefree(p)
988 #define new_XNV() new_body_allocated(SVt_NV)
989 #define new_XPVNV() new_body_allocated(SVt_PVNV)
990 #define new_XPVMG() new_body_allocated(SVt_PVMG)
992 #define del_XPVGV(p) del_body(p + bodies_by_type[SVt_PVGV].offset, \
993 &PL_body_roots[SVt_PVGV])
997 /* no arena for you! */
999 #define new_NOARENA(details) \
1000 safemalloc((details)->body_size + (details)->offset)
1001 #define new_NOARENAZ(details) \
1002 safecalloc((details)->body_size + (details)->offset, 1)
1005 Perl_more_bodies (pTHX_ const svtype sv_type, const size_t body_size,
1006 const size_t arena_size)
1009 void ** const root = &PL_body_roots[sv_type];
1010 struct arena_desc *adesc;
1011 struct arena_set *aroot = (struct arena_set *) PL_body_arenas;
1015 const size_t good_arena_size = Perl_malloc_good_size(arena_size);
1016 #if defined(DEBUGGING) && !defined(PERL_GLOBAL_STRUCT_PRIVATE)
1017 static bool done_sanity_check;
1019 /* PERL_GLOBAL_STRUCT_PRIVATE cannot coexist with global
1020 * variables like done_sanity_check. */
1021 if (!done_sanity_check) {
1022 unsigned int i = SVt_LAST;
1024 done_sanity_check = TRUE;
1027 assert (bodies_by_type[i].type == i);
1033 /* may need new arena-set to hold new arena */
1034 if (!aroot || aroot->curr >= aroot->set_size) {
1035 struct arena_set *newroot;
1036 Newxz(newroot, 1, struct arena_set);
1037 newroot->set_size = ARENAS_PER_SET;
1038 newroot->next = aroot;
1040 PL_body_arenas = (void *) newroot;
1041 DEBUG_m(PerlIO_printf(Perl_debug_log, "new arenaset %p\n", (void*)aroot));
1044 /* ok, now have arena-set with at least 1 empty/available arena-desc */
1045 curr = aroot->curr++;
1046 adesc = &(aroot->set[curr]);
1047 assert(!adesc->arena);
1049 Newx(adesc->arena, good_arena_size, char);
1050 adesc->size = good_arena_size;
1051 adesc->utype = sv_type;
1052 DEBUG_m(PerlIO_printf(Perl_debug_log, "arena %d added: %p size %"UVuf"\n",
1053 curr, (void*)adesc->arena, (UV)good_arena_size));
1055 start = (char *) adesc->arena;
1057 /* Get the address of the byte after the end of the last body we can fit.
1058 Remember, this is integer division: */
1059 end = start + good_arena_size / body_size * body_size;
1061 /* computed count doesn't reflect the 1st slot reservation */
1062 #if defined(MYMALLOC) || defined(HAS_MALLOC_GOOD_SIZE)
1063 DEBUG_m(PerlIO_printf(Perl_debug_log,
1064 "arena %p end %p arena-size %d (from %d) type %d "
1066 (void*)start, (void*)end, (int)good_arena_size,
1067 (int)arena_size, sv_type, (int)body_size,
1068 (int)good_arena_size / (int)body_size));
1070 DEBUG_m(PerlIO_printf(Perl_debug_log,
1071 "arena %p end %p arena-size %d type %d size %d ct %d\n",
1072 (void*)start, (void*)end,
1073 (int)arena_size, sv_type, (int)body_size,
1074 (int)good_arena_size / (int)body_size));
1076 *root = (void *)start;
1079 /* Where the next body would start: */
1080 char * const next = start + body_size;
1083 /* This is the last body: */
1084 assert(next == end);
1086 *(void **)start = 0;
1090 *(void**) start = (void *)next;
1095 /* grab a new thing from the free list, allocating more if necessary.
1096 The inline version is used for speed in hot routines, and the
1097 function using it serves the rest (unless PURIFY).
1099 #define new_body_inline(xpv, sv_type) \
1101 void ** const r3wt = &PL_body_roots[sv_type]; \
1102 xpv = (PTR_TBL_ENT_t*) (*((void **)(r3wt)) \
1103 ? *((void **)(r3wt)) : Perl_more_bodies(aTHX_ sv_type, \
1104 bodies_by_type[sv_type].body_size,\
1105 bodies_by_type[sv_type].arena_size)); \
1106 *(r3wt) = *(void**)(xpv); \
1112 S_new_body(pTHX_ const svtype sv_type)
1116 new_body_inline(xpv, sv_type);
1122 static const struct body_details fake_rv =
1123 { 0, 0, 0, SVt_IV, FALSE, NONV, NOARENA, 0 };
1126 =for apidoc sv_upgrade
1128 Upgrade an SV to a more complex form. Generally adds a new body type to the
1129 SV, then copies across as much information as possible from the old body.
1130 You generally want to use the C<SvUPGRADE> macro wrapper. See also C<svtype>.
1136 Perl_sv_upgrade(pTHX_ register SV *const sv, svtype new_type)
1141 const svtype old_type = SvTYPE(sv);
1142 const struct body_details *new_type_details;
1143 const struct body_details *old_type_details
1144 = bodies_by_type + old_type;
1145 SV *referant = NULL;
1147 PERL_ARGS_ASSERT_SV_UPGRADE;
1149 if (old_type == new_type)
1152 /* This clause was purposefully added ahead of the early return above to
1153 the shared string hackery for (sort {$a <=> $b} keys %hash), with the
1154 inference by Nick I-S that it would fix other troublesome cases. See
1155 changes 7162, 7163 (f130fd4589cf5fbb24149cd4db4137c8326f49c1 and parent)
1157 Given that shared hash key scalars are no longer PVIV, but PV, there is
1158 no longer need to unshare so as to free up the IVX slot for its proper
1159 purpose. So it's safe to move the early return earlier. */
1161 if (new_type != SVt_PV && SvIsCOW(sv)) {
1162 sv_force_normal_flags(sv, 0);
1165 old_body = SvANY(sv);
1167 /* Copying structures onto other structures that have been neatly zeroed
1168 has a subtle gotcha. Consider XPVMG
1170 +------+------+------+------+------+-------+-------+
1171 | NV | CUR | LEN | IV | MAGIC | STASH |
1172 +------+------+------+------+------+-------+-------+
1173 0 4 8 12 16 20 24 28
1175 where NVs are aligned to 8 bytes, so that sizeof that structure is
1176 actually 32 bytes long, with 4 bytes of padding at the end:
1178 +------+------+------+------+------+-------+-------+------+
1179 | NV | CUR | LEN | IV | MAGIC | STASH | ??? |
1180 +------+------+------+------+------+-------+-------+------+
1181 0 4 8 12 16 20 24 28 32
1183 so what happens if you allocate memory for this structure:
1185 +------+------+------+------+------+-------+-------+------+------+...
1186 | NV | CUR | LEN | IV | MAGIC | STASH | GP | NAME |
1187 +------+------+------+------+------+-------+-------+------+------+...
1188 0 4 8 12 16 20 24 28 32 36
1190 zero it, then copy sizeof(XPVMG) bytes on top of it? Not quite what you
1191 expect, because you copy the area marked ??? onto GP. Now, ??? may have
1192 started out as zero once, but it's quite possible that it isn't. So now,
1193 rather than a nicely zeroed GP, you have it pointing somewhere random.
1196 (In fact, GP ends up pointing at a previous GP structure, because the
1197 principle cause of the padding in XPVMG getting garbage is a copy of
1198 sizeof(XPVMG) bytes from a XPVGV structure in sv_unglob. Right now
1199 this happens to be moot because XPVGV has been re-ordered, with GP
1200 no longer after STASH)
1202 So we are careful and work out the size of used parts of all the
1210 referant = SvRV(sv);
1211 old_type_details = &fake_rv;
1212 if (new_type == SVt_NV)
1213 new_type = SVt_PVNV;
1215 if (new_type < SVt_PVIV) {
1216 new_type = (new_type == SVt_NV)
1217 ? SVt_PVNV : SVt_PVIV;
1222 if (new_type < SVt_PVNV) {
1223 new_type = SVt_PVNV;
1227 assert(new_type > SVt_PV);
1228 assert(SVt_IV < SVt_PV);
1229 assert(SVt_NV < SVt_PV);
1236 /* Because the XPVMG of PL_mess_sv isn't allocated from the arena,
1237 there's no way that it can be safely upgraded, because perl.c
1238 expects to Safefree(SvANY(PL_mess_sv)) */
1239 assert(sv != PL_mess_sv);
1240 /* This flag bit is used to mean other things in other scalar types.
1241 Given that it only has meaning inside the pad, it shouldn't be set
1242 on anything that can get upgraded. */
1243 assert(!SvPAD_TYPED(sv));
1246 if (old_type_details->cant_upgrade)
1247 Perl_croak(aTHX_ "Can't upgrade %s (%" UVuf ") to %" UVuf,
1248 sv_reftype(sv, 0), (UV) old_type, (UV) new_type);
1251 if (old_type > new_type)
1252 Perl_croak(aTHX_ "sv_upgrade from type %d down to type %d",
1253 (int)old_type, (int)new_type);
1255 new_type_details = bodies_by_type + new_type;
1257 SvFLAGS(sv) &= ~SVTYPEMASK;
1258 SvFLAGS(sv) |= new_type;
1260 /* This can't happen, as SVt_NULL is <= all values of new_type, so one of
1261 the return statements above will have triggered. */
1262 assert (new_type != SVt_NULL);
1265 assert(old_type == SVt_NULL);
1266 SvANY(sv) = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
1270 assert(old_type == SVt_NULL);
1271 SvANY(sv) = new_XNV();
1276 assert(new_type_details->body_size);
1279 assert(new_type_details->arena);
1280 assert(new_type_details->arena_size);
1281 /* This points to the start of the allocated area. */
1282 new_body_inline(new_body, new_type);
1283 Zero(new_body, new_type_details->body_size, char);
1284 new_body = ((char *)new_body) - new_type_details->offset;
1286 /* We always allocated the full length item with PURIFY. To do this
1287 we fake things so that arena is false for all 16 types.. */
1288 new_body = new_NOARENAZ(new_type_details);
1290 SvANY(sv) = new_body;
1291 if (new_type == SVt_PVAV) {
1295 if (old_type_details->body_size) {
1298 /* It will have been zeroed when the new body was allocated.
1299 Lets not write to it, in case it confuses a write-back
1305 #ifndef NODEFAULT_SHAREKEYS
1306 HvSHAREKEYS_on(sv); /* key-sharing on by default */
1308 HvMAX(sv) = 7; /* (start with 8 buckets) */
1311 /* SVt_NULL isn't the only thing upgraded to AV or HV.
1312 The target created by newSVrv also is, and it can have magic.
1313 However, it never has SvPVX set.
1315 if (old_type == SVt_IV) {
1317 } else if (old_type >= SVt_PV) {
1318 assert(SvPVX_const(sv) == 0);
1321 if (old_type >= SVt_PVMG) {
1322 SvMAGIC_set(sv, ((XPVMG*)old_body)->xmg_u.xmg_magic);
1323 SvSTASH_set(sv, ((XPVMG*)old_body)->xmg_stash);
1325 sv->sv_u.svu_array = NULL; /* or svu_hash */
1331 /* This ensures that SvTHINKFIRST(sv) is true, and hence that
1332 sv_force_normal_flags(sv) is called. */
1335 /* XXX Is this still needed? Was it ever needed? Surely as there is
1336 no route from NV to PVIV, NOK can never be true */
1337 assert(!SvNOKp(sv));
1348 assert(new_type_details->body_size);
1349 /* We always allocated the full length item with PURIFY. To do this
1350 we fake things so that arena is false for all 16 types.. */
1351 if(new_type_details->arena) {
1352 /* This points to the start of the allocated area. */
1353 new_body_inline(new_body, new_type);
1354 Zero(new_body, new_type_details->body_size, char);
1355 new_body = ((char *)new_body) - new_type_details->offset;
1357 new_body = new_NOARENAZ(new_type_details);
1359 SvANY(sv) = new_body;
1361 if (old_type_details->copy) {
1362 /* There is now the potential for an upgrade from something without
1363 an offset (PVNV or PVMG) to something with one (PVCV, PVFM) */
1364 int offset = old_type_details->offset;
1365 int length = old_type_details->copy;
1367 if (new_type_details->offset > old_type_details->offset) {
1368 const int difference
1369 = new_type_details->offset - old_type_details->offset;
1370 offset += difference;
1371 length -= difference;
1373 assert (length >= 0);
1375 Copy((char *)old_body + offset, (char *)new_body + offset, length,
1379 #ifndef NV_ZERO_IS_ALLBITS_ZERO
1380 /* If NV 0.0 is stores as all bits 0 then Zero() already creates a
1381 * correct 0.0 for us. Otherwise, if the old body didn't have an
1382 * NV slot, but the new one does, then we need to initialise the
1383 * freshly created NV slot with whatever the correct bit pattern is
1385 if (old_type_details->zero_nv && !new_type_details->zero_nv
1386 && !isGV_with_GP(sv))
1390 if (new_type == SVt_PVIO) {
1391 IO * const io = MUTABLE_IO(sv);
1392 GV *iogv = gv_fetchpvs("IO::File::", GV_ADD, SVt_PVHV);
1395 /* Clear the stashcache because a new IO could overrule a package
1397 hv_clear(PL_stashcache);
1399 SvSTASH_set(io, MUTABLE_HV(SvREFCNT_inc(GvHV(iogv))));
1400 IoPAGE_LEN(sv) = 60;
1402 if (old_type < SVt_PV) {
1403 /* referant will be NULL unless the old type was SVt_IV emulating
1405 sv->sv_u.svu_rv = referant;
1409 Perl_croak(aTHX_ "panic: sv_upgrade to unknown type %lu",
1410 (unsigned long)new_type);
1413 if (old_type > SVt_IV) {
1417 /* Note that there is an assumption that all bodies of types that
1418 can be upgraded came from arenas. Only the more complex non-
1419 upgradable types are allowed to be directly malloc()ed. */
1420 assert(old_type_details->arena);
1421 del_body((void*)((char*)old_body + old_type_details->offset),
1422 &PL_body_roots[old_type]);
1428 =for apidoc sv_backoff
1430 Remove any string offset. You should normally use the C<SvOOK_off> macro
1437 Perl_sv_backoff(pTHX_ register SV *const sv)
1440 const char * const s = SvPVX_const(sv);
1442 PERL_ARGS_ASSERT_SV_BACKOFF;
1443 PERL_UNUSED_CONTEXT;
1446 assert(SvTYPE(sv) != SVt_PVHV);
1447 assert(SvTYPE(sv) != SVt_PVAV);
1449 SvOOK_offset(sv, delta);
1451 SvLEN_set(sv, SvLEN(sv) + delta);
1452 SvPV_set(sv, SvPVX(sv) - delta);
1453 Move(s, SvPVX(sv), SvCUR(sv)+1, char);
1454 SvFLAGS(sv) &= ~SVf_OOK;
1461 Expands the character buffer in the SV. If necessary, uses C<sv_unref> and
1462 upgrades the SV to C<SVt_PV>. Returns a pointer to the character buffer.
1463 Use the C<SvGROW> wrapper instead.
1469 Perl_sv_grow(pTHX_ register SV *const sv, register STRLEN newlen)
1473 PERL_ARGS_ASSERT_SV_GROW;
1475 if (PL_madskills && newlen >= 0x100000) {
1476 PerlIO_printf(Perl_debug_log,
1477 "Allocation too large: %"UVxf"\n", (UV)newlen);
1479 #ifdef HAS_64K_LIMIT
1480 if (newlen >= 0x10000) {
1481 PerlIO_printf(Perl_debug_log,
1482 "Allocation too large: %"UVxf"\n", (UV)newlen);
1485 #endif /* HAS_64K_LIMIT */
1488 if (SvTYPE(sv) < SVt_PV) {
1489 sv_upgrade(sv, SVt_PV);
1490 s = SvPVX_mutable(sv);
1492 else if (SvOOK(sv)) { /* pv is offset? */
1494 s = SvPVX_mutable(sv);
1495 if (newlen > SvLEN(sv))
1496 newlen += 10 * (newlen - SvCUR(sv)); /* avoid copy each time */
1497 #ifdef HAS_64K_LIMIT
1498 if (newlen >= 0x10000)
1503 s = SvPVX_mutable(sv);
1505 if (newlen > SvLEN(sv)) { /* need more room? */
1506 STRLEN minlen = SvCUR(sv);
1507 minlen += (minlen >> PERL_STRLEN_EXPAND_SHIFT) + 10;
1508 if (newlen < minlen)
1510 #ifndef Perl_safesysmalloc_size
1511 newlen = PERL_STRLEN_ROUNDUP(newlen);
1513 if (SvLEN(sv) && s) {
1514 s = (char*)saferealloc(s, newlen);
1517 s = (char*)safemalloc(newlen);
1518 if (SvPVX_const(sv) && SvCUR(sv)) {
1519 Move(SvPVX_const(sv), s, (newlen < SvCUR(sv)) ? newlen : SvCUR(sv), char);
1523 #ifdef Perl_safesysmalloc_size
1524 /* Do this here, do it once, do it right, and then we will never get
1525 called back into sv_grow() unless there really is some growing
1527 SvLEN_set(sv, Perl_safesysmalloc_size(s));
1529 SvLEN_set(sv, newlen);
1536 =for apidoc sv_setiv
1538 Copies an integer into the given SV, upgrading first if necessary.
1539 Does not handle 'set' magic. See also C<sv_setiv_mg>.
1545 Perl_sv_setiv(pTHX_ register SV *const sv, const IV i)
1549 PERL_ARGS_ASSERT_SV_SETIV;
1551 SV_CHECK_THINKFIRST_COW_DROP(sv);
1552 switch (SvTYPE(sv)) {
1555 sv_upgrade(sv, SVt_IV);
1558 sv_upgrade(sv, SVt_PVIV);
1562 if (!isGV_with_GP(sv))
1569 /* diag_listed_as: Can't coerce %s to %s in %s */
1570 Perl_croak(aTHX_ "Can't coerce %s to integer in %s", sv_reftype(sv,0),
1574 (void)SvIOK_only(sv); /* validate number */
1580 =for apidoc sv_setiv_mg
1582 Like C<sv_setiv>, but also handles 'set' magic.
1588 Perl_sv_setiv_mg(pTHX_ register SV *const sv, const IV i)
1590 PERL_ARGS_ASSERT_SV_SETIV_MG;
1597 =for apidoc sv_setuv
1599 Copies an unsigned integer into the given SV, upgrading first if necessary.
1600 Does not handle 'set' magic. See also C<sv_setuv_mg>.
1606 Perl_sv_setuv(pTHX_ register SV *const sv, const UV u)
1608 PERL_ARGS_ASSERT_SV_SETUV;
1610 /* With these two if statements:
1611 u=1.49 s=0.52 cu=72.49 cs=10.64 scripts=270 tests=20865
1614 u=1.35 s=0.47 cu=73.45 cs=11.43 scripts=270 tests=20865
1616 If you wish to remove them, please benchmark to see what the effect is
1618 if (u <= (UV)IV_MAX) {
1619 sv_setiv(sv, (IV)u);
1628 =for apidoc sv_setuv_mg
1630 Like C<sv_setuv>, but also handles 'set' magic.
1636 Perl_sv_setuv_mg(pTHX_ register SV *const sv, const UV u)
1638 PERL_ARGS_ASSERT_SV_SETUV_MG;
1645 =for apidoc sv_setnv
1647 Copies a double into the given SV, upgrading first if necessary.
1648 Does not handle 'set' magic. See also C<sv_setnv_mg>.
1654 Perl_sv_setnv(pTHX_ register SV *const sv, const NV num)
1658 PERL_ARGS_ASSERT_SV_SETNV;
1660 SV_CHECK_THINKFIRST_COW_DROP(sv);
1661 switch (SvTYPE(sv)) {
1664 sv_upgrade(sv, SVt_NV);
1668 sv_upgrade(sv, SVt_PVNV);
1672 if (!isGV_with_GP(sv))
1679 /* diag_listed_as: Can't coerce %s to %s in %s */
1680 Perl_croak(aTHX_ "Can't coerce %s to number in %s", sv_reftype(sv,0),
1685 (void)SvNOK_only(sv); /* validate number */
1690 =for apidoc sv_setnv_mg
1692 Like C<sv_setnv>, but also handles 'set' magic.
1698 Perl_sv_setnv_mg(pTHX_ register SV *const sv, const NV num)
1700 PERL_ARGS_ASSERT_SV_SETNV_MG;
1706 /* Print an "isn't numeric" warning, using a cleaned-up,
1707 * printable version of the offending string
1711 S_not_a_number(pTHX_ SV *const sv)
1718 PERL_ARGS_ASSERT_NOT_A_NUMBER;
1721 dsv = newSVpvs_flags("", SVs_TEMP);
1722 pv = sv_uni_display(dsv, sv, 10, 0);
1725 const char * const limit = tmpbuf + sizeof(tmpbuf) - 8;
1726 /* each *s can expand to 4 chars + "...\0",
1727 i.e. need room for 8 chars */
1729 const char *s = SvPVX_const(sv);
1730 const char * const end = s + SvCUR(sv);
1731 for ( ; s < end && d < limit; s++ ) {
1733 if (ch & 128 && !isPRINT_LC(ch)) {
1742 else if (ch == '\r') {
1746 else if (ch == '\f') {
1750 else if (ch == '\\') {
1754 else if (ch == '\0') {
1758 else if (isPRINT_LC(ch))
1775 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
1776 "Argument \"%s\" isn't numeric in %s", pv,
1779 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
1780 "Argument \"%s\" isn't numeric", pv);
1784 =for apidoc looks_like_number
1786 Test if the content of an SV looks like a number (or is a number).
1787 C<Inf> and C<Infinity> are treated as numbers (so will not issue a
1788 non-numeric warning), even if your atof() doesn't grok them.
1794 Perl_looks_like_number(pTHX_ SV *const sv)
1796 register const char *sbegin;
1799 PERL_ARGS_ASSERT_LOOKS_LIKE_NUMBER;
1802 sbegin = SvPVX_const(sv);
1805 else if (SvPOKp(sv))
1806 sbegin = SvPV_const(sv, len);
1808 return SvFLAGS(sv) & (SVf_NOK|SVp_NOK|SVf_IOK|SVp_IOK);
1809 return grok_number(sbegin, len, NULL);
1813 S_glob_2number(pTHX_ GV * const gv)
1815 const U32 wasfake = SvFLAGS(gv) & SVf_FAKE;
1816 SV *const buffer = sv_newmortal();
1818 PERL_ARGS_ASSERT_GLOB_2NUMBER;
1820 /* FAKE globs can get coerced, so need to turn this off temporarily if it
1823 gv_efullname3(buffer, gv, "*");
1824 SvFLAGS(gv) |= wasfake;
1826 /* We know that all GVs stringify to something that is not-a-number,
1827 so no need to test that. */
1828 if (ckWARN(WARN_NUMERIC))
1829 not_a_number(buffer);
1830 /* We just want something true to return, so that S_sv_2iuv_common
1831 can tail call us and return true. */
1835 /* Actually, ISO C leaves conversion of UV to IV undefined, but
1836 until proven guilty, assume that things are not that bad... */
1841 As 64 bit platforms often have an NV that doesn't preserve all bits of
1842 an IV (an assumption perl has been based on to date) it becomes necessary
1843 to remove the assumption that the NV always carries enough precision to
1844 recreate the IV whenever needed, and that the NV is the canonical form.
1845 Instead, IV/UV and NV need to be given equal rights. So as to not lose
1846 precision as a side effect of conversion (which would lead to insanity
1847 and the dragon(s) in t/op/numconvert.t getting very angry) the intent is
1848 1) to distinguish between IV/UV/NV slots that have cached a valid
1849 conversion where precision was lost and IV/UV/NV slots that have a
1850 valid conversion which has lost no precision
1851 2) to ensure that if a numeric conversion to one form is requested that
1852 would lose precision, the precise conversion (or differently
1853 imprecise conversion) is also performed and cached, to prevent
1854 requests for different numeric formats on the same SV causing
1855 lossy conversion chains. (lossless conversion chains are perfectly
1860 SvIOKp is true if the IV slot contains a valid value
1861 SvIOK is true only if the IV value is accurate (UV if SvIOK_UV true)
1862 SvNOKp is true if the NV slot contains a valid value
1863 SvNOK is true only if the NV value is accurate
1866 while converting from PV to NV, check to see if converting that NV to an
1867 IV(or UV) would lose accuracy over a direct conversion from PV to
1868 IV(or UV). If it would, cache both conversions, return NV, but mark
1869 SV as IOK NOKp (ie not NOK).
1871 While converting from PV to IV, check to see if converting that IV to an
1872 NV would lose accuracy over a direct conversion from PV to NV. If it
1873 would, cache both conversions, flag similarly.
1875 Before, the SV value "3.2" could become NV=3.2 IV=3 NOK, IOK quite
1876 correctly because if IV & NV were set NV *always* overruled.
1877 Now, "3.2" will become NV=3.2 IV=3 NOK, IOKp, because the flag's meaning
1878 changes - now IV and NV together means that the two are interchangeable:
1879 SvIVX == (IV) SvNVX && SvNVX == (NV) SvIVX;
1881 The benefit of this is that operations such as pp_add know that if
1882 SvIOK is true for both left and right operands, then integer addition
1883 can be used instead of floating point (for cases where the result won't
1884 overflow). Before, floating point was always used, which could lead to
1885 loss of precision compared with integer addition.
1887 * making IV and NV equal status should make maths accurate on 64 bit
1889 * may speed up maths somewhat if pp_add and friends start to use
1890 integers when possible instead of fp. (Hopefully the overhead in
1891 looking for SvIOK and checking for overflow will not outweigh the
1892 fp to integer speedup)
1893 * will slow down integer operations (callers of SvIV) on "inaccurate"
1894 values, as the change from SvIOK to SvIOKp will cause a call into
1895 sv_2iv each time rather than a macro access direct to the IV slot
1896 * should speed up number->string conversion on integers as IV is
1897 favoured when IV and NV are equally accurate
1899 ####################################################################
1900 You had better be using SvIOK_notUV if you want an IV for arithmetic:
1901 SvIOK is true if (IV or UV), so you might be getting (IV)SvUV.
1902 On the other hand, SvUOK is true iff UV.
1903 ####################################################################
1905 Your mileage will vary depending your CPU's relative fp to integer
1909 #ifndef NV_PRESERVES_UV
1910 # define IS_NUMBER_UNDERFLOW_IV 1
1911 # define IS_NUMBER_UNDERFLOW_UV 2
1912 # define IS_NUMBER_IV_AND_UV 2
1913 # define IS_NUMBER_OVERFLOW_IV 4
1914 # define IS_NUMBER_OVERFLOW_UV 5
1916 /* sv_2iuv_non_preserve(): private routine for use by sv_2iv() and sv_2uv() */
1918 /* For sv_2nv these three cases are "SvNOK and don't bother casting" */
1920 S_sv_2iuv_non_preserve(pTHX_ register SV *const sv
1928 PERL_ARGS_ASSERT_SV_2IUV_NON_PRESERVE;
1930 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_2iuv_non '%s', IV=0x%"UVxf" NV=%"NVgf" inttype=%"UVXf"\n", SvPVX_const(sv), SvIVX(sv), SvNVX(sv), (UV)numtype));
1931 if (SvNVX(sv) < (NV)IV_MIN) {
1932 (void)SvIOKp_on(sv);
1934 SvIV_set(sv, IV_MIN);
1935 return IS_NUMBER_UNDERFLOW_IV;
1937 if (SvNVX(sv) > (NV)UV_MAX) {
1938 (void)SvIOKp_on(sv);
1941 SvUV_set(sv, UV_MAX);
1942 return IS_NUMBER_OVERFLOW_UV;
1944 (void)SvIOKp_on(sv);
1946 /* Can't use strtol etc to convert this string. (See truth table in
1948 if (SvNVX(sv) <= (UV)IV_MAX) {
1949 SvIV_set(sv, I_V(SvNVX(sv)));
1950 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
1951 SvIOK_on(sv); /* Integer is precise. NOK, IOK */
1953 /* Integer is imprecise. NOK, IOKp */
1955 return SvNVX(sv) < 0 ? IS_NUMBER_UNDERFLOW_UV : IS_NUMBER_IV_AND_UV;
1958 SvUV_set(sv, U_V(SvNVX(sv)));
1959 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
1960 if (SvUVX(sv) == UV_MAX) {
1961 /* As we know that NVs don't preserve UVs, UV_MAX cannot
1962 possibly be preserved by NV. Hence, it must be overflow.
1964 return IS_NUMBER_OVERFLOW_UV;
1966 SvIOK_on(sv); /* Integer is precise. NOK, UOK */
1968 /* Integer is imprecise. NOK, IOKp */
1970 return IS_NUMBER_OVERFLOW_IV;
1972 #endif /* !NV_PRESERVES_UV*/
1975 S_sv_2iuv_common(pTHX_ SV *const sv)
1979 PERL_ARGS_ASSERT_SV_2IUV_COMMON;
1982 /* erm. not sure. *should* never get NOKp (without NOK) from sv_2nv
1983 * without also getting a cached IV/UV from it at the same time
1984 * (ie PV->NV conversion should detect loss of accuracy and cache
1985 * IV or UV at same time to avoid this. */
1986 /* IV-over-UV optimisation - choose to cache IV if possible */
1988 if (SvTYPE(sv) == SVt_NV)
1989 sv_upgrade(sv, SVt_PVNV);
1991 (void)SvIOKp_on(sv); /* Must do this first, to clear any SvOOK */
1992 /* < not <= as for NV doesn't preserve UV, ((NV)IV_MAX+1) will almost
1993 certainly cast into the IV range at IV_MAX, whereas the correct
1994 answer is the UV IV_MAX +1. Hence < ensures that dodgy boundary
1996 #if defined(NAN_COMPARE_BROKEN) && defined(Perl_isnan)
1997 if (Perl_isnan(SvNVX(sv))) {
2003 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2004 SvIV_set(sv, I_V(SvNVX(sv)));
2005 if (SvNVX(sv) == (NV) SvIVX(sv)
2006 #ifndef NV_PRESERVES_UV
2007 && (((UV)1 << NV_PRESERVES_UV_BITS) >
2008 (UV)(SvIVX(sv) > 0 ? SvIVX(sv) : -SvIVX(sv)))
2009 /* Don't flag it as "accurately an integer" if the number
2010 came from a (by definition imprecise) NV operation, and
2011 we're outside the range of NV integer precision */
2015 SvIOK_on(sv); /* Can this go wrong with rounding? NWC */
2017 /* scalar has trailing garbage, eg "42a" */
2019 DEBUG_c(PerlIO_printf(Perl_debug_log,
2020 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (precise)\n",
2026 /* IV not precise. No need to convert from PV, as NV
2027 conversion would already have cached IV if it detected
2028 that PV->IV would be better than PV->NV->IV
2029 flags already correct - don't set public IOK. */
2030 DEBUG_c(PerlIO_printf(Perl_debug_log,
2031 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (imprecise)\n",
2036 /* Can the above go wrong if SvIVX == IV_MIN and SvNVX < IV_MIN,
2037 but the cast (NV)IV_MIN rounds to a the value less (more
2038 negative) than IV_MIN which happens to be equal to SvNVX ??
2039 Analogous to 0xFFFFFFFFFFFFFFFF rounding up to NV (2**64) and
2040 NV rounding back to 0xFFFFFFFFFFFFFFFF, so UVX == UV(NVX) and
2041 (NV)UVX == NVX are both true, but the values differ. :-(
2042 Hopefully for 2s complement IV_MIN is something like
2043 0x8000000000000000 which will be exact. NWC */
2046 SvUV_set(sv, U_V(SvNVX(sv)));
2048 (SvNVX(sv) == (NV) SvUVX(sv))
2049 #ifndef NV_PRESERVES_UV
2050 /* Make sure it's not 0xFFFFFFFFFFFFFFFF */
2051 /*&& (SvUVX(sv) != UV_MAX) irrelevant with code below */
2052 && (((UV)1 << NV_PRESERVES_UV_BITS) > SvUVX(sv))
2053 /* Don't flag it as "accurately an integer" if the number
2054 came from a (by definition imprecise) NV operation, and
2055 we're outside the range of NV integer precision */
2061 DEBUG_c(PerlIO_printf(Perl_debug_log,
2062 "0x%"UVxf" 2iv(%"UVuf" => %"IVdf") (as unsigned)\n",
2068 else if (SvPOKp(sv) && SvLEN(sv)) {
2070 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2071 /* We want to avoid a possible problem when we cache an IV/ a UV which
2072 may be later translated to an NV, and the resulting NV is not
2073 the same as the direct translation of the initial string
2074 (eg 123.456 can shortcut to the IV 123 with atol(), but we must
2075 be careful to ensure that the value with the .456 is around if the
2076 NV value is requested in the future).
2078 This means that if we cache such an IV/a UV, we need to cache the
2079 NV as well. Moreover, we trade speed for space, and do not
2080 cache the NV if we are sure it's not needed.
2083 /* SVt_PVNV is one higher than SVt_PVIV, hence this order */
2084 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2085 == IS_NUMBER_IN_UV) {
2086 /* It's definitely an integer, only upgrade to PVIV */
2087 if (SvTYPE(sv) < SVt_PVIV)
2088 sv_upgrade(sv, SVt_PVIV);
2090 } else if (SvTYPE(sv) < SVt_PVNV)
2091 sv_upgrade(sv, SVt_PVNV);
2093 /* If NVs preserve UVs then we only use the UV value if we know that
2094 we aren't going to call atof() below. If NVs don't preserve UVs
2095 then the value returned may have more precision than atof() will
2096 return, even though value isn't perfectly accurate. */
2097 if ((numtype & (IS_NUMBER_IN_UV
2098 #ifdef NV_PRESERVES_UV
2101 )) == IS_NUMBER_IN_UV) {
2102 /* This won't turn off the public IOK flag if it was set above */
2103 (void)SvIOKp_on(sv);
2105 if (!(numtype & IS_NUMBER_NEG)) {
2107 if (value <= (UV)IV_MAX) {
2108 SvIV_set(sv, (IV)value);
2110 /* it didn't overflow, and it was positive. */
2111 SvUV_set(sv, value);
2115 /* 2s complement assumption */
2116 if (value <= (UV)IV_MIN) {
2117 SvIV_set(sv, -(IV)value);
2119 /* Too negative for an IV. This is a double upgrade, but
2120 I'm assuming it will be rare. */
2121 if (SvTYPE(sv) < SVt_PVNV)
2122 sv_upgrade(sv, SVt_PVNV);
2126 SvNV_set(sv, -(NV)value);
2127 SvIV_set(sv, IV_MIN);
2131 /* For !NV_PRESERVES_UV and IS_NUMBER_IN_UV and IS_NUMBER_NOT_INT we
2132 will be in the previous block to set the IV slot, and the next
2133 block to set the NV slot. So no else here. */
2135 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2136 != IS_NUMBER_IN_UV) {
2137 /* It wasn't an (integer that doesn't overflow the UV). */
2138 SvNV_set(sv, Atof(SvPVX_const(sv)));
2140 if (! numtype && ckWARN(WARN_NUMERIC))
2143 #if defined(USE_LONG_DOUBLE)
2144 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%" PERL_PRIgldbl ")\n",
2145 PTR2UV(sv), SvNVX(sv)));
2147 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"NVgf")\n",
2148 PTR2UV(sv), SvNVX(sv)));
2151 #ifdef NV_PRESERVES_UV
2152 (void)SvIOKp_on(sv);
2154 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2155 SvIV_set(sv, I_V(SvNVX(sv)));
2156 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
2159 NOOP; /* Integer is imprecise. NOK, IOKp */
2161 /* UV will not work better than IV */
2163 if (SvNVX(sv) > (NV)UV_MAX) {
2165 /* Integer is inaccurate. NOK, IOKp, is UV */
2166 SvUV_set(sv, UV_MAX);
2168 SvUV_set(sv, U_V(SvNVX(sv)));
2169 /* 0xFFFFFFFFFFFFFFFF not an issue in here, NVs
2170 NV preservse UV so can do correct comparison. */
2171 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
2174 NOOP; /* Integer is imprecise. NOK, IOKp, is UV */
2179 #else /* NV_PRESERVES_UV */
2180 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2181 == (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT)) {
2182 /* The IV/UV slot will have been set from value returned by
2183 grok_number above. The NV slot has just been set using
2186 assert (SvIOKp(sv));
2188 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2189 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2190 /* Small enough to preserve all bits. */
2191 (void)SvIOKp_on(sv);
2193 SvIV_set(sv, I_V(SvNVX(sv)));
2194 if ((NV)(SvIVX(sv)) == SvNVX(sv))
2196 /* Assumption: first non-preserved integer is < IV_MAX,
2197 this NV is in the preserved range, therefore: */
2198 if (!(U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))
2200 Perl_croak(aTHX_ "sv_2iv assumed (U_V(fabs((double)SvNVX(sv))) < (UV)IV_MAX) but SvNVX(sv)=%"NVgf" U_V is 0x%"UVxf", IV_MAX is 0x%"UVxf"\n", SvNVX(sv), U_V(SvNVX(sv)), (UV)IV_MAX);
2204 0 0 already failed to read UV.
2205 0 1 already failed to read UV.
2206 1 0 you won't get here in this case. IV/UV
2207 slot set, public IOK, Atof() unneeded.
2208 1 1 already read UV.
2209 so there's no point in sv_2iuv_non_preserve() attempting
2210 to use atol, strtol, strtoul etc. */
2212 sv_2iuv_non_preserve (sv, numtype);
2214 sv_2iuv_non_preserve (sv);
2218 #endif /* NV_PRESERVES_UV */
2219 /* It might be more code efficient to go through the entire logic above
2220 and conditionally set with SvIOKp_on() rather than SvIOK(), but it
2221 gets complex and potentially buggy, so more programmer efficient
2222 to do it this way, by turning off the public flags: */
2224 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK);
2228 if (isGV_with_GP(sv))
2229 return glob_2number(MUTABLE_GV(sv));
2231 if (!(SvFLAGS(sv) & SVs_PADTMP)) {
2232 if (!PL_localizing && ckWARN(WARN_UNINITIALIZED))
2235 if (SvTYPE(sv) < SVt_IV)
2236 /* Typically the caller expects that sv_any is not NULL now. */
2237 sv_upgrade(sv, SVt_IV);
2238 /* Return 0 from the caller. */
2245 =for apidoc sv_2iv_flags
2247 Return the integer value of an SV, doing any necessary string
2248 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2249 Normally used via the C<SvIV(sv)> and C<SvIVx(sv)> macros.
2255 Perl_sv_2iv_flags(pTHX_ register SV *const sv, const I32 flags)
2260 if (SvGMAGICAL(sv) || SvVALID(sv)) {
2261 /* FBMs use the space for SvIVX and SvNVX for other purposes, and use
2262 the same flag bit as SVf_IVisUV, so must not let them cache IVs.
2263 In practice they are extremely unlikely to actually get anywhere
2264 accessible by user Perl code - the only way that I'm aware of is when
2265 a constant subroutine which is used as the second argument to index.
2267 if (flags & SV_GMAGIC)
2272 return I_V(SvNVX(sv));
2274 if (SvPOKp(sv) && SvLEN(sv)) {
2277 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2279 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2280 == IS_NUMBER_IN_UV) {
2281 /* It's definitely an integer */
2282 if (numtype & IS_NUMBER_NEG) {
2283 if (value < (UV)IV_MIN)
2286 if (value < (UV)IV_MAX)
2291 if (ckWARN(WARN_NUMERIC))
2294 return I_V(Atof(SvPVX_const(sv)));
2299 assert(SvTYPE(sv) >= SVt_PVMG);
2300 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
2301 } else if (SvTHINKFIRST(sv)) {
2306 if (flags & SV_SKIP_OVERLOAD)
2308 tmpstr = AMG_CALLunary(sv, numer_amg);
2309 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2310 return SvIV(tmpstr);
2313 return PTR2IV(SvRV(sv));
2316 sv_force_normal_flags(sv, 0);
2318 if (SvREADONLY(sv) && !SvOK(sv)) {
2319 if (ckWARN(WARN_UNINITIALIZED))
2325 if (S_sv_2iuv_common(aTHX_ sv))
2328 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"IVdf")\n",
2329 PTR2UV(sv),SvIVX(sv)));
2330 return SvIsUV(sv) ? (IV)SvUVX(sv) : SvIVX(sv);
2334 =for apidoc sv_2uv_flags
2336 Return the unsigned integer value of an SV, doing any necessary string
2337 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2338 Normally used via the C<SvUV(sv)> and C<SvUVx(sv)> macros.
2344 Perl_sv_2uv_flags(pTHX_ register SV *const sv, const I32 flags)
2349 if (SvGMAGICAL(sv) || SvVALID(sv)) {
2350 /* FBMs use the space for SvIVX and SvNVX for other purposes, and use
2351 the same flag bit as SVf_IVisUV, so must not let them cache IVs. */
2352 if (flags & SV_GMAGIC)
2357 return U_V(SvNVX(sv));
2358 if (SvPOKp(sv) && SvLEN(sv)) {
2361 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2363 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2364 == IS_NUMBER_IN_UV) {
2365 /* It's definitely an integer */
2366 if (!(numtype & IS_NUMBER_NEG))
2370 if (ckWARN(WARN_NUMERIC))
2373 return U_V(Atof(SvPVX_const(sv)));
2378 assert(SvTYPE(sv) >= SVt_PVMG);
2379 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
2380 } else if (SvTHINKFIRST(sv)) {
2385 if (flags & SV_SKIP_OVERLOAD)
2387 tmpstr = AMG_CALLunary(sv, numer_amg);
2388 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2389 return SvUV(tmpstr);
2392 return PTR2UV(SvRV(sv));
2395 sv_force_normal_flags(sv, 0);
2397 if (SvREADONLY(sv) && !SvOK(sv)) {
2398 if (ckWARN(WARN_UNINITIALIZED))
2404 if (S_sv_2iuv_common(aTHX_ sv))
2408 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2uv(%"UVuf")\n",
2409 PTR2UV(sv),SvUVX(sv)));
2410 return SvIsUV(sv) ? SvUVX(sv) : (UV)SvIVX(sv);
2414 =for apidoc sv_2nv_flags
2416 Return the num value of an SV, doing any necessary string or integer
2417 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2418 Normally used via the C<SvNV(sv)> and C<SvNVx(sv)> macros.
2424 Perl_sv_2nv_flags(pTHX_ register SV *const sv, const I32 flags)
2429 if (SvGMAGICAL(sv) || SvVALID(sv)) {
2430 /* FBMs use the space for SvIVX and SvNVX for other purposes, and use
2431 the same flag bit as SVf_IVisUV, so must not let them cache NVs. */
2432 if (flags & SV_GMAGIC)
2436 if ((SvPOKp(sv) && SvLEN(sv)) && !SvIOKp(sv)) {
2437 if (!SvIOKp(sv) && ckWARN(WARN_NUMERIC) &&
2438 !grok_number(SvPVX_const(sv), SvCUR(sv), NULL))
2440 return Atof(SvPVX_const(sv));
2444 return (NV)SvUVX(sv);
2446 return (NV)SvIVX(sv);
2451 assert(SvTYPE(sv) >= SVt_PVMG);
2452 /* This falls through to the report_uninit near the end of the
2454 } else if (SvTHINKFIRST(sv)) {
2459 if (flags & SV_SKIP_OVERLOAD)
2461 tmpstr = AMG_CALLunary(sv, numer_amg);
2462 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2463 return SvNV(tmpstr);
2466 return PTR2NV(SvRV(sv));
2469 sv_force_normal_flags(sv, 0);
2471 if (SvREADONLY(sv) && !SvOK(sv)) {
2472 if (ckWARN(WARN_UNINITIALIZED))
2477 if (SvTYPE(sv) < SVt_NV) {
2478 /* The logic to use SVt_PVNV if necessary is in sv_upgrade. */
2479 sv_upgrade(sv, SVt_NV);
2480 #ifdef USE_LONG_DOUBLE
2482 STORE_NUMERIC_LOCAL_SET_STANDARD();
2483 PerlIO_printf(Perl_debug_log,
2484 "0x%"UVxf" num(%" PERL_PRIgldbl ")\n",
2485 PTR2UV(sv), SvNVX(sv));
2486 RESTORE_NUMERIC_LOCAL();
2490 STORE_NUMERIC_LOCAL_SET_STANDARD();
2491 PerlIO_printf(Perl_debug_log, "0x%"UVxf" num(%"NVgf")\n",
2492 PTR2UV(sv), SvNVX(sv));
2493 RESTORE_NUMERIC_LOCAL();
2497 else if (SvTYPE(sv) < SVt_PVNV)
2498 sv_upgrade(sv, SVt_PVNV);
2503 SvNV_set(sv, SvIsUV(sv) ? (NV)SvUVX(sv) : (NV)SvIVX(sv));
2504 #ifdef NV_PRESERVES_UV
2510 /* Only set the public NV OK flag if this NV preserves the IV */
2511 /* Check it's not 0xFFFFFFFFFFFFFFFF */
2513 SvIsUV(sv) ? ((SvUVX(sv) != UV_MAX)&&(SvUVX(sv) == U_V(SvNVX(sv))))
2514 : (SvIVX(sv) == I_V(SvNVX(sv))))
2520 else if (SvPOKp(sv) && SvLEN(sv)) {
2522 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2523 if (!SvIOKp(sv) && !numtype && ckWARN(WARN_NUMERIC))
2525 #ifdef NV_PRESERVES_UV
2526 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2527 == IS_NUMBER_IN_UV) {
2528 /* It's definitely an integer */
2529 SvNV_set(sv, (numtype & IS_NUMBER_NEG) ? -(NV)value : (NV)value);
2531 SvNV_set(sv, Atof(SvPVX_const(sv)));
2537 SvNV_set(sv, Atof(SvPVX_const(sv)));
2538 /* Only set the public NV OK flag if this NV preserves the value in
2539 the PV at least as well as an IV/UV would.
2540 Not sure how to do this 100% reliably. */
2541 /* if that shift count is out of range then Configure's test is
2542 wonky. We shouldn't be in here with NV_PRESERVES_UV_BITS ==
2544 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2545 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2546 SvNOK_on(sv); /* Definitely small enough to preserve all bits */
2547 } else if (!(numtype & IS_NUMBER_IN_UV)) {
2548 /* Can't use strtol etc to convert this string, so don't try.
2549 sv_2iv and sv_2uv will use the NV to convert, not the PV. */
2552 /* value has been set. It may not be precise. */
2553 if ((numtype & IS_NUMBER_NEG) && (value > (UV)IV_MIN)) {
2554 /* 2s complement assumption for (UV)IV_MIN */
2555 SvNOK_on(sv); /* Integer is too negative. */
2560 if (numtype & IS_NUMBER_NEG) {
2561 SvIV_set(sv, -(IV)value);
2562 } else if (value <= (UV)IV_MAX) {
2563 SvIV_set(sv, (IV)value);
2565 SvUV_set(sv, value);
2569 if (numtype & IS_NUMBER_NOT_INT) {
2570 /* I believe that even if the original PV had decimals,
2571 they are lost beyond the limit of the FP precision.
2572 However, neither is canonical, so both only get p
2573 flags. NWC, 2000/11/25 */
2574 /* Both already have p flags, so do nothing */
2576 const NV nv = SvNVX(sv);
2577 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2578 if (SvIVX(sv) == I_V(nv)) {
2581 /* It had no "." so it must be integer. */
2585 /* between IV_MAX and NV(UV_MAX).
2586 Could be slightly > UV_MAX */
2588 if (numtype & IS_NUMBER_NOT_INT) {
2589 /* UV and NV both imprecise. */
2591 const UV nv_as_uv = U_V(nv);
2593 if (value == nv_as_uv && SvUVX(sv) != UV_MAX) {
2602 /* It might be more code efficient to go through the entire logic above
2603 and conditionally set with SvNOKp_on() rather than SvNOK(), but it
2604 gets complex and potentially buggy, so more programmer efficient
2605 to do it this way, by turning off the public flags: */
2607 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK);
2608 #endif /* NV_PRESERVES_UV */
2611 if (isGV_with_GP(sv)) {
2612 glob_2number(MUTABLE_GV(sv));
2616 if (!PL_localizing && !(SvFLAGS(sv) & SVs_PADTMP) && ckWARN(WARN_UNINITIALIZED))
2618 assert (SvTYPE(sv) >= SVt_NV);
2619 /* Typically the caller expects that sv_any is not NULL now. */
2620 /* XXX Ilya implies that this is a bug in callers that assume this
2621 and ideally should be fixed. */
2624 #if defined(USE_LONG_DOUBLE)
2626 STORE_NUMERIC_LOCAL_SET_STANDARD();
2627 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2nv(%" PERL_PRIgldbl ")\n",
2628 PTR2UV(sv), SvNVX(sv));
2629 RESTORE_NUMERIC_LOCAL();
2633 STORE_NUMERIC_LOCAL_SET_STANDARD();
2634 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 1nv(%"NVgf")\n",
2635 PTR2UV(sv), SvNVX(sv));
2636 RESTORE_NUMERIC_LOCAL();
2645 Return an SV with the numeric value of the source SV, doing any necessary
2646 reference or overload conversion. You must use the C<SvNUM(sv)> macro to
2647 access this function.
2653 Perl_sv_2num(pTHX_ register SV *const sv)
2655 PERL_ARGS_ASSERT_SV_2NUM;
2660 SV * const tmpsv = AMG_CALLunary(sv, numer_amg);
2661 TAINT_IF(tmpsv && SvTAINTED(tmpsv));
2662 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
2663 return sv_2num(tmpsv);
2665 return sv_2mortal(newSVuv(PTR2UV(SvRV(sv))));
2668 /* uiv_2buf(): private routine for use by sv_2pv_flags(): print an IV or
2669 * UV as a string towards the end of buf, and return pointers to start and
2672 * We assume that buf is at least TYPE_CHARS(UV) long.
2676 S_uiv_2buf(char *const buf, const IV iv, UV uv, const int is_uv, char **const peob)
2678 char *ptr = buf + TYPE_CHARS(UV);
2679 char * const ebuf = ptr;
2682 PERL_ARGS_ASSERT_UIV_2BUF;
2694 *--ptr = '0' + (char)(uv % 10);
2703 =for apidoc sv_2pv_flags
2705 Returns a pointer to the string value of an SV, and sets *lp to its length.
2706 If flags includes SV_GMAGIC, does an mg_get() first. Coerces sv to a string
2708 Normally invoked via the C<SvPV_flags> macro. C<sv_2pv()> and C<sv_2pv_nomg>
2709 usually end up here too.
2715 Perl_sv_2pv_flags(pTHX_ register SV *const sv, STRLEN *const lp, const I32 flags)
2725 if (SvGMAGICAL(sv)) {
2726 if (flags & SV_GMAGIC)
2731 if (flags & SV_MUTABLE_RETURN)
2732 return SvPVX_mutable(sv);
2733 if (flags & SV_CONST_RETURN)
2734 return (char *)SvPVX_const(sv);
2737 if (SvIOKp(sv) || SvNOKp(sv)) {
2738 char tbuf[64]; /* Must fit sprintf/Gconvert of longest IV/NV */
2743 ? my_snprintf(tbuf, sizeof(tbuf), "%"UVuf, (UV)SvUVX(sv))
2744 : my_snprintf(tbuf, sizeof(tbuf), "%"IVdf, (IV)SvIVX(sv));
2745 } else if(SvNVX(sv) == 0.0) {
2750 Gconvert(SvNVX(sv), NV_DIG, 0, tbuf);
2757 SvUPGRADE(sv, SVt_PV);
2760 s = SvGROW_mutable(sv, len + 1);
2763 return (char*)memcpy(s, tbuf, len + 1);
2769 assert(SvTYPE(sv) >= SVt_PVMG);
2770 /* This falls through to the report_uninit near the end of the
2772 } else if (SvTHINKFIRST(sv)) {
2777 if (flags & SV_SKIP_OVERLOAD)
2779 tmpstr = AMG_CALLunary(sv, string_amg);
2780 TAINT_IF(tmpstr && SvTAINTED(tmpstr));
2781 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2783 /* char *pv = lp ? SvPV(tmpstr, *lp) : SvPV_nolen(tmpstr);
2787 if ((SvFLAGS(tmpstr) & (SVf_POK)) == SVf_POK) {
2788 if (flags & SV_CONST_RETURN) {
2789 pv = (char *) SvPVX_const(tmpstr);
2791 pv = (flags & SV_MUTABLE_RETURN)
2792 ? SvPVX_mutable(tmpstr) : SvPVX(tmpstr);
2795 *lp = SvCUR(tmpstr);
2797 pv = sv_2pv_flags(tmpstr, lp, flags);
2810 SV *const referent = SvRV(sv);
2814 retval = buffer = savepvn("NULLREF", len);
2815 } else if (SvTYPE(referent) == SVt_REGEXP) {
2816 REGEXP * const re = (REGEXP *)MUTABLE_PTR(referent);
2821 /* If the regex is UTF-8 we want the containing scalar to
2822 have an UTF-8 flag too */
2828 if ((seen_evals = RX_SEEN_EVALS(re)))
2829 PL_reginterp_cnt += seen_evals;
2832 *lp = RX_WRAPLEN(re);
2834 return RX_WRAPPED(re);
2836 const char *const typestr = sv_reftype(referent, 0);
2837 const STRLEN typelen = strlen(typestr);
2838 UV addr = PTR2UV(referent);
2839 const char *stashname = NULL;
2840 STRLEN stashnamelen = 0; /* hush, gcc */
2841 const char *buffer_end;
2843 if (SvOBJECT(referent)) {
2844 const HEK *const name = HvNAME_HEK(SvSTASH(referent));
2847 stashname = HEK_KEY(name);
2848 stashnamelen = HEK_LEN(name);
2850 if (HEK_UTF8(name)) {
2856 stashname = "__ANON__";
2859 len = stashnamelen + 1 /* = */ + typelen + 3 /* (0x */
2860 + 2 * sizeof(UV) + 2 /* )\0 */;
2862 len = typelen + 3 /* (0x */
2863 + 2 * sizeof(UV) + 2 /* )\0 */;
2866 Newx(buffer, len, char);
2867 buffer_end = retval = buffer + len;
2869 /* Working backwards */
2873 *--retval = PL_hexdigit[addr & 15];
2874 } while (addr >>= 4);
2880 memcpy(retval, typestr, typelen);
2884 retval -= stashnamelen;
2885 memcpy(retval, stashname, stashnamelen);
2887 /* retval may not necessarily have reached the start of the
2889 assert (retval >= buffer);
2891 len = buffer_end - retval - 1; /* -1 for that \0 */
2899 if (SvREADONLY(sv) && !SvOK(sv)) {
2902 if (flags & SV_UNDEF_RETURNS_NULL)
2904 if (ckWARN(WARN_UNINITIALIZED))
2909 if (SvIOK(sv) || ((SvIOKp(sv) && !SvNOKp(sv)))) {
2910 /* I'm assuming that if both IV and NV are equally valid then
2911 converting the IV is going to be more efficient */
2912 const U32 isUIOK = SvIsUV(sv);
2913 char buf[TYPE_CHARS(UV)];
2917 if (SvTYPE(sv) < SVt_PVIV)
2918 sv_upgrade(sv, SVt_PVIV);
2919 ptr = uiv_2buf(buf, SvIVX(sv), SvUVX(sv), isUIOK, &ebuf);
2921 /* inlined from sv_setpvn */
2922 s = SvGROW_mutable(sv, len + 1);
2923 Move(ptr, s, len, char);
2927 else if (SvNOKp(sv)) {
2928 if (SvTYPE(sv) < SVt_PVNV)
2929 sv_upgrade(sv, SVt_PVNV);
2930 if (SvNVX(sv) == 0.0) {
2931 s = SvGROW_mutable(sv, 2);
2936 /* The +20 is pure guesswork. Configure test needed. --jhi */
2937 s = SvGROW_mutable(sv, NV_DIG + 20);
2938 /* some Xenix systems wipe out errno here */
2939 Gconvert(SvNVX(sv), NV_DIG, 0, s);
2949 if (isGV_with_GP(sv)) {
2950 GV *const gv = MUTABLE_GV(sv);
2951 const U32 wasfake = SvFLAGS(gv) & SVf_FAKE;
2952 SV *const buffer = sv_newmortal();
2954 /* FAKE globs can get coerced, so need to turn this off temporarily
2957 gv_efullname3(buffer, gv, "*");
2958 SvFLAGS(gv) |= wasfake;
2960 if (SvPOK(buffer)) {
2962 *lp = SvCUR(buffer);
2964 if ( SvUTF8(buffer) ) SvUTF8_on(sv);
2965 return SvPVX(buffer);
2976 if (flags & SV_UNDEF_RETURNS_NULL)
2978 if (!PL_localizing && !(SvFLAGS(sv) & SVs_PADTMP) && ckWARN(WARN_UNINITIALIZED))
2980 if (SvTYPE(sv) < SVt_PV)
2981 /* Typically the caller expects that sv_any is not NULL now. */
2982 sv_upgrade(sv, SVt_PV);
2986 const STRLEN len = s - SvPVX_const(sv);
2992 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
2993 PTR2UV(sv),SvPVX_const(sv)));
2994 if (flags & SV_CONST_RETURN)
2995 return (char *)SvPVX_const(sv);
2996 if (flags & SV_MUTABLE_RETURN)
2997 return SvPVX_mutable(sv);
3002 =for apidoc sv_copypv
3004 Copies a stringified representation of the source SV into the
3005 destination SV. Automatically performs any necessary mg_get and
3006 coercion of numeric values into strings. Guaranteed to preserve
3007 UTF8 flag even from overloaded objects. Similar in nature to
3008 sv_2pv[_flags] but operates directly on an SV instead of just the
3009 string. Mostly uses sv_2pv_flags to do its work, except when that
3010 would lose the UTF-8'ness of the PV.
3016 Perl_sv_copypv(pTHX_ SV *const dsv, register SV *const ssv)
3019 const char * const s = SvPV_const(ssv,len);
3021 PERL_ARGS_ASSERT_SV_COPYPV;
3023 sv_setpvn(dsv,s,len);
3031 =for apidoc sv_2pvbyte
3033 Return a pointer to the byte-encoded representation of the SV, and set *lp
3034 to its length. May cause the SV to be downgraded from UTF-8 as a
3037 Usually accessed via the C<SvPVbyte> macro.
3043 Perl_sv_2pvbyte(pTHX_ register SV *const sv, STRLEN *const lp)
3045 PERL_ARGS_ASSERT_SV_2PVBYTE;
3048 sv_utf8_downgrade(sv,0);
3049 return lp ? SvPV_nomg(sv,*lp) : SvPV_nomg_nolen(sv);
3053 =for apidoc sv_2pvutf8
3055 Return a pointer to the UTF-8-encoded representation of the SV, and set *lp
3056 to its length. May cause the SV to be upgraded to UTF-8 as a side-effect.
3058 Usually accessed via the C<SvPVutf8> macro.
3064 Perl_sv_2pvutf8(pTHX_ register SV *const sv, STRLEN *const lp)
3066 PERL_ARGS_ASSERT_SV_2PVUTF8;
3068 sv_utf8_upgrade(sv);
3069 return lp ? SvPV(sv,*lp) : SvPV_nolen(sv);
3074 =for apidoc sv_2bool
3076 This macro is only used by sv_true() or its macro equivalent, and only if
3077 the latter's argument is neither SvPOK, SvIOK nor SvNOK.
3078 It calls sv_2bool_flags with the SV_GMAGIC flag.
3080 =for apidoc sv_2bool_flags
3082 This function is only used by sv_true() and friends, and only if
3083 the latter's argument is neither SvPOK, SvIOK nor SvNOK. If the flags
3084 contain SV_GMAGIC, then it does an mg_get() first.
3091 Perl_sv_2bool_flags(pTHX_ register SV *const sv, const I32 flags)
3095 PERL_ARGS_ASSERT_SV_2BOOL_FLAGS;
3097 if(flags & SV_GMAGIC) SvGETMAGIC(sv);
3103 SV * const tmpsv = AMG_CALLunary(sv, bool__amg);
3104 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
3105 return cBOOL(SvTRUE(tmpsv));
3107 return SvRV(sv) != 0;
3110 register XPV* const Xpvtmp = (XPV*)SvANY(sv);
3112 (*sv->sv_u.svu_pv > '0' ||
3113 Xpvtmp->xpv_cur > 1 ||
3114 (Xpvtmp->xpv_cur && *sv->sv_u.svu_pv != '0')))
3121 return SvIVX(sv) != 0;
3124 return SvNVX(sv) != 0.0;
3126 if (isGV_with_GP(sv))
3136 =for apidoc sv_utf8_upgrade
3138 Converts the PV of an SV to its UTF-8-encoded form.
3139 Forces the SV to string form if it is not already.
3140 Will C<mg_get> on C<sv> if appropriate.
3141 Always sets the SvUTF8 flag to avoid future validity checks even
3142 if the whole string is the same in UTF-8 as not.
3143 Returns the number of bytes in the converted string
3145 This is not as a general purpose byte encoding to Unicode interface:
3146 use the Encode extension for that.
3148 =for apidoc sv_utf8_upgrade_nomg
3150 Like sv_utf8_upgrade, but doesn't do magic on C<sv>
3152 =for apidoc sv_utf8_upgrade_flags
3154 Converts the PV of an SV to its UTF-8-encoded form.
3155 Forces the SV to string form if it is not already.
3156 Always sets the SvUTF8 flag to avoid future validity checks even
3157 if all the bytes are invariant in UTF-8. If C<flags> has C<SV_GMAGIC> bit set,
3158 will C<mg_get> on C<sv> if appropriate, else not.
3159 Returns the number of bytes in the converted string
3160 C<sv_utf8_upgrade> and
3161 C<sv_utf8_upgrade_nomg> are implemented in terms of this function.
3163 This is not as a general purpose byte encoding to Unicode interface:
3164 use the Encode extension for that.
3168 The grow version is currently not externally documented. It adds a parameter,
3169 extra, which is the number of unused bytes the string of 'sv' is guaranteed to
3170 have free after it upon return. This allows the caller to reserve extra space
3171 that it intends to fill, to avoid extra grows.
3173 Also externally undocumented for the moment is the flag SV_FORCE_UTF8_UPGRADE,
3174 which can be used to tell this function to not first check to see if there are
3175 any characters that are different in UTF-8 (variant characters) which would
3176 force it to allocate a new string to sv, but to assume there are. Typically
3177 this flag is used by a routine that has already parsed the string to find that
3178 there are such characters, and passes this information on so that the work
3179 doesn't have to be repeated.
3181 (One might think that the calling routine could pass in the position of the
3182 first such variant, so it wouldn't have to be found again. But that is not the
3183 case, because typically when the caller is likely to use this flag, it won't be
3184 calling this routine unless it finds something that won't fit into a byte.
3185 Otherwise it tries to not upgrade and just use bytes. But some things that
3186 do fit into a byte are variants in utf8, and the caller may not have been
3187 keeping track of these.)
3189 If the routine itself changes the string, it adds a trailing NUL. Such a NUL
3190 isn't guaranteed due to having other routines do the work in some input cases,
3191 or if the input is already flagged as being in utf8.
3193 The speed of this could perhaps be improved for many cases if someone wanted to
3194 write a fast function that counts the number of variant characters in a string,
3195 especially if it could return the position of the first one.
3200 Perl_sv_utf8_upgrade_flags_grow(pTHX_ register SV *const sv, const I32 flags, STRLEN extra)
3204 PERL_ARGS_ASSERT_SV_UTF8_UPGRADE_FLAGS_GROW;
3206 if (sv == &PL_sv_undef)
3210 if (SvREADONLY(sv) && (SvPOKp(sv) || SvIOKp(sv) || SvNOKp(sv))) {
3211 (void) sv_2pv_flags(sv,&len, flags);
3213 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3217 (void) SvPV_force_flags(sv,len,flags & SV_GMAGIC);
3222 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3227 sv_force_normal_flags(sv, 0);
3230 if (PL_encoding && !(flags & SV_UTF8_NO_ENCODING)) {
3231 sv_recode_to_utf8(sv, PL_encoding);
3232 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3236 if (SvCUR(sv) == 0) {
3237 if (extra) SvGROW(sv, extra);
3238 } else { /* Assume Latin-1/EBCDIC */
3239 /* This function could be much more efficient if we
3240 * had a FLAG in SVs to signal if there are any variant
3241 * chars in the PV. Given that there isn't such a flag
3242 * make the loop as fast as possible (although there are certainly ways
3243 * to speed this up, eg. through vectorization) */
3244 U8 * s = (U8 *) SvPVX_const(sv);
3245 U8 * e = (U8 *) SvEND(sv);
3247 STRLEN two_byte_count = 0;
3249 if (flags & SV_FORCE_UTF8_UPGRADE) goto must_be_utf8;
3251 /* See if really will need to convert to utf8. We mustn't rely on our
3252 * incoming SV being well formed and having a trailing '\0', as certain
3253 * code in pp_formline can send us partially built SVs. */
3257 if (NATIVE_IS_INVARIANT(ch)) continue;
3259 t--; /* t already incremented; re-point to first variant */
3264 /* utf8 conversion not needed because all are invariants. Mark as
3265 * UTF-8 even if no variant - saves scanning loop */
3271 /* Here, the string should be converted to utf8, either because of an
3272 * input flag (two_byte_count = 0), or because a character that
3273 * requires 2 bytes was found (two_byte_count = 1). t points either to
3274 * the beginning of the string (if we didn't examine anything), or to
3275 * the first variant. In either case, everything from s to t - 1 will
3276 * occupy only 1 byte each on output.
3278 * There are two main ways to convert. One is to create a new string
3279 * and go through the input starting from the beginning, appending each
3280 * converted value onto the new string as we go along. It's probably
3281 * best to allocate enough space in the string for the worst possible
3282 * case rather than possibly running out of space and having to
3283 * reallocate and then copy what we've done so far. Since everything
3284 * from s to t - 1 is invariant, the destination can be initialized
3285 * with these using a fast memory copy
3287 * The other way is to figure out exactly how big the string should be
3288 * by parsing the entire input. Then you don't have to make it big
3289 * enough to handle the worst possible case, and more importantly, if
3290 * the string you already have is large enough, you don't have to
3291 * allocate a new string, you can copy the last character in the input
3292 * string to the final position(s) that will be occupied by the
3293 * converted string and go backwards, stopping at t, since everything
3294 * before that is invariant.
3296 * There are advantages and disadvantages to each method.
3298 * In the first method, we can allocate a new string, do the memory
3299 * copy from the s to t - 1, and then proceed through the rest of the
3300 * string byte-by-byte.
3302 * In the second method, we proceed through the rest of the input
3303 * string just calculating how big the converted string will be. Then
3304 * there are two cases:
3305 * 1) if the string has enough extra space to handle the converted
3306 * value. We go backwards through the string, converting until we
3307 * get to the position we are at now, and then stop. If this
3308 * position is far enough along in the string, this method is
3309 * faster than the other method. If the memory copy were the same
3310 * speed as the byte-by-byte loop, that position would be about
3311 * half-way, as at the half-way mark, parsing to the end and back
3312 * is one complete string's parse, the same amount as starting
3313 * over and going all the way through. Actually, it would be
3314 * somewhat less than half-way, as it's faster to just count bytes
3315 * than to also copy, and we don't have the overhead of allocating
3316 * a new string, changing the scalar to use it, and freeing the
3317 * existing one. But if the memory copy is fast, the break-even
3318 * point is somewhere after half way. The counting loop could be
3319 * sped up by vectorization, etc, to move the break-even point
3320 * further towards the beginning.
3321 * 2) if the string doesn't have enough space to handle the converted
3322 * value. A new string will have to be allocated, and one might
3323 * as well, given that, start from the beginning doing the first
3324 * method. We've spent extra time parsing the string and in
3325 * exchange all we've gotten is that we know precisely how big to
3326 * make the new one. Perl is more optimized for time than space,
3327 * so this case is a loser.
3328 * So what I've decided to do is not use the 2nd method unless it is
3329 * guaranteed that a new string won't have to be allocated, assuming
3330 * the worst case. I also decided not to put any more conditions on it
3331 * than this, for now. It seems likely that, since the worst case is
3332 * twice as big as the unknown portion of the string (plus 1), we won't
3333 * be guaranteed enough space, causing us to go to the first method,
3334 * unless the string is short, or the first variant character is near
3335 * the end of it. In either of these cases, it seems best to use the
3336 * 2nd method. The only circumstance I can think of where this would
3337 * be really slower is if the string had once had much more data in it
3338 * than it does now, but there is still a substantial amount in it */
3341 STRLEN invariant_head = t - s;
3342 STRLEN size = invariant_head + (e - t) * 2 + 1 + extra;
3343 if (SvLEN(sv) < size) {
3345 /* Here, have decided to allocate a new string */
3350 Newx(dst, size, U8);
3352 /* If no known invariants at the beginning of the input string,
3353 * set so starts from there. Otherwise, can use memory copy to
3354 * get up to where we are now, and then start from here */
3356 if (invariant_head <= 0) {
3359 Copy(s, dst, invariant_head, char);
3360 d = dst + invariant_head;
3364 const UV uv = NATIVE8_TO_UNI(*t++);
3365 if (UNI_IS_INVARIANT(uv))
3366 *d++ = (U8)UNI_TO_NATIVE(uv);
3368 *d++ = (U8)UTF8_EIGHT_BIT_HI(uv);
3369 *d++ = (U8)UTF8_EIGHT_BIT_LO(uv);
3373 SvPV_free(sv); /* No longer using pre-existing string */
3374 SvPV_set(sv, (char*)dst);
3375 SvCUR_set(sv, d - dst);
3376 SvLEN_set(sv, size);
3379 /* Here, have decided to get the exact size of the string.
3380 * Currently this happens only when we know that there is
3381 * guaranteed enough space to fit the converted string, so
3382 * don't have to worry about growing. If two_byte_count is 0,
3383 * then t points to the first byte of the string which hasn't
3384 * been examined yet. Otherwise two_byte_count is 1, and t
3385 * points to the first byte in the string that will expand to
3386 * two. Depending on this, start examining at t or 1 after t.
3389 U8 *d = t + two_byte_count;
3392 /* Count up the remaining bytes that expand to two */
3395 const U8 chr = *d++;
3396 if (! NATIVE_IS_INVARIANT(chr)) two_byte_count++;
3399 /* The string will expand by just the number of bytes that
3400 * occupy two positions. But we are one afterwards because of
3401 * the increment just above. This is the place to put the
3402 * trailing NUL, and to set the length before we decrement */
3404 d += two_byte_count;
3405 SvCUR_set(sv, d - s);
3409 /* Having decremented d, it points to the position to put the
3410 * very last byte of the expanded string. Go backwards through
3411 * the string, copying and expanding as we go, stopping when we
3412 * get to the part that is invariant the rest of the way down */
3416 const U8 ch = NATIVE8_TO_UNI(*e--);
3417 if (UNI_IS_INVARIANT(ch)) {
3418 *d-- = UNI_TO_NATIVE(ch);
3420 *d-- = (U8)UTF8_EIGHT_BIT_LO(ch);
3421 *d-- = (U8)UTF8_EIGHT_BIT_HI(ch);
3426 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
3427 /* Update pos. We do it at the end rather than during
3428 * the upgrade, to avoid slowing down the common case
3429 * (upgrade without pos) */
3430 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
3432 I32 pos = mg->mg_len;
3433 if (pos > 0 && (U32)pos > invariant_head) {
3434 U8 *d = (U8*) SvPVX(sv) + invariant_head;
3435 STRLEN n = (U32)pos - invariant_head;
3437 if (UTF8_IS_START(*d))
3442 mg->mg_len = d - (U8*)SvPVX(sv);
3445 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
3446 magic_setutf8(sv,mg); /* clear UTF8 cache */
3451 /* Mark as UTF-8 even if no variant - saves scanning loop */
3457 =for apidoc sv_utf8_downgrade
3459 Attempts to convert the PV of an SV from characters to bytes.
3460 If the PV contains a character that cannot fit
3461 in a byte, this conversion will fail;
3462 in this case, either returns false or, if C<fail_ok> is not
3465 This is not as a general purpose Unicode to byte encoding interface:
3466 use the Encode extension for that.
3472 Perl_sv_utf8_downgrade(pTHX_ register SV *const sv, const bool fail_ok)
3476 PERL_ARGS_ASSERT_SV_UTF8_DOWNGRADE;
3478 if (SvPOKp(sv) && SvUTF8(sv)) {
3482 int mg_flags = SV_GMAGIC;
3485 sv_force_normal_flags(sv, 0);
3487 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
3489 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
3491 I32 pos = mg->mg_len;
3493 sv_pos_b2u(sv, &pos);
3494 mg_flags = 0; /* sv_pos_b2u does get magic */
3498 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
3499 magic_setutf8(sv,mg); /* clear UTF8 cache */
3502 s = (U8 *) SvPV_flags(sv, len, mg_flags);
3504 if (!utf8_to_bytes(s, &len)) {
3509 Perl_croak(aTHX_ "Wide character in %s",
3512 Perl_croak(aTHX_ "Wide character");
3523 =for apidoc sv_utf8_encode
3525 Converts the PV of an SV to UTF-8, but then turns the C<SvUTF8>
3526 flag off so that it looks like octets again.
3532 Perl_sv_utf8_encode(pTHX_ register SV *const sv)
3534 PERL_ARGS_ASSERT_SV_UTF8_ENCODE;
3537 sv_force_normal_flags(sv, 0);
3539 if (SvREADONLY(sv)) {
3540 Perl_croak_no_modify(aTHX);
3542 (void) sv_utf8_upgrade(sv);
3547 =for apidoc sv_utf8_decode
3549 If the PV of the SV is an octet sequence in UTF-8
3550 and contains a multiple-byte character, the C<SvUTF8> flag is turned on
3551 so that it looks like a character. If the PV contains only single-byte
3552 characters, the C<SvUTF8> flag stays off.
3553 Scans PV for validity and returns false if the PV is invalid UTF-8.
3559 Perl_sv_utf8_decode(pTHX_ register SV *const sv)
3561 PERL_ARGS_ASSERT_SV_UTF8_DECODE;
3564 const U8 *start, *c;
3567 /* The octets may have got themselves encoded - get them back as
3570 if (!sv_utf8_downgrade(sv, TRUE))
3573 /* it is actually just a matter of turning the utf8 flag on, but
3574 * we want to make sure everything inside is valid utf8 first.
3576 c = start = (const U8 *) SvPVX_const(sv);
3577 if (!is_utf8_string(c, SvCUR(sv)+1))
3579 e = (const U8 *) SvEND(sv);
3582 if (!UTF8_IS_INVARIANT(ch)) {
3587 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
3588 /* adjust pos to the start of a UTF8 char sequence */
3589 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
3591 I32 pos = mg->mg_len;
3593 for (c = start + pos; c > start; c--) {
3594 if (UTF8_IS_START(*c))
3597 mg->mg_len = c - start;
3600 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
3601 magic_setutf8(sv,mg); /* clear UTF8 cache */
3608 =for apidoc sv_setsv
3610 Copies the contents of the source SV C<ssv> into the destination SV
3611 C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3612 function if the source SV needs to be reused. Does not handle 'set' magic.
3613 Loosely speaking, it performs a copy-by-value, obliterating any previous
3614 content of the destination.
3616 You probably want to use one of the assortment of wrappers, such as
3617 C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3618 C<SvSetMagicSV_nosteal>.
3620 =for apidoc sv_setsv_flags
3622 Copies the contents of the source SV C<ssv> into the destination SV
3623 C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3624 function if the source SV needs to be reused. Does not handle 'set' magic.
3625 Loosely speaking, it performs a copy-by-value, obliterating any previous
3626 content of the destination.
3627 If the C<flags> parameter has the C<SV_GMAGIC> bit set, will C<mg_get> on
3628 C<ssv> if appropriate, else not. If the C<flags> parameter has the
3629 C<NOSTEAL> bit set then the buffers of temps will not be stolen. <sv_setsv>
3630 and C<sv_setsv_nomg> are implemented in terms of this function.
3632 You probably want to use one of the assortment of wrappers, such as
3633 C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3634 C<SvSetMagicSV_nosteal>.
3636 This is the primary function for copying scalars, and most other
3637 copy-ish functions and macros use this underneath.
3643 S_glob_assign_glob(pTHX_ SV *const dstr, SV *const sstr, const int dtype)
3645 I32 mro_changes = 0; /* 1 = method, 2 = isa, 3 = recursive isa */
3646 HV *old_stash = NULL;
3648 PERL_ARGS_ASSERT_GLOB_ASSIGN_GLOB;
3650 if (dtype != SVt_PVGV && !isGV_with_GP(dstr)) {
3651 const char * const name = GvNAME(sstr);
3652 const STRLEN len = GvNAMELEN(sstr);
3654 if (dtype >= SVt_PV) {
3660 SvUPGRADE(dstr, SVt_PVGV);
3661 (void)SvOK_off(dstr);
3662 /* FIXME - why are we doing this, then turning it off and on again
3664 isGV_with_GP_on(dstr);
3666 GvSTASH(dstr) = GvSTASH(sstr);
3668 Perl_sv_add_backref(aTHX_ MUTABLE_SV(GvSTASH(dstr)), dstr);
3669 gv_name_set(MUTABLE_GV(dstr), name, len, GV_ADD);
3670 SvFAKE_on(dstr); /* can coerce to non-glob */
3673 if(GvGP(MUTABLE_GV(sstr))) {
3674 /* If source has method cache entry, clear it */
3676 SvREFCNT_dec(GvCV(sstr));
3677 GvCV_set(sstr, NULL);
3680 /* If source has a real method, then a method is
3683 GvCV((const GV *)sstr) && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3689 /* If dest already had a real method, that's a change as well */
3691 !mro_changes && GvGP(MUTABLE_GV(dstr)) && GvCVu((const GV *)dstr)
3692 && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3697 /* We don’t need to check the name of the destination if it was not a
3698 glob to begin with. */
3699 if(dtype == SVt_PVGV) {
3700 const char * const name = GvNAME((const GV *)dstr);
3703 /* The stash may have been detached from the symbol table, so
3705 && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3706 && GvAV((const GV *)sstr)
3710 const STRLEN len = GvNAMELEN(dstr);
3711 if ((len > 1 && name[len-2] == ':' && name[len-1] == ':')
3712 || (len == 1 && name[0] == ':')) {
3715 /* Set aside the old stash, so we can reset isa caches on
3717 if((old_stash = GvHV(dstr)))
3718 /* Make sure we do not lose it early. */
3719 SvREFCNT_inc_simple_void_NN(
3720 sv_2mortal((SV *)old_stash)
3726 gp_free(MUTABLE_GV(dstr));
3727 isGV_with_GP_off(dstr);
3728 (void)SvOK_off(dstr);
3729 isGV_with_GP_on(dstr);
3730 GvINTRO_off(dstr); /* one-shot flag */
3731 GvGP_set(dstr, gp_ref(GvGP(sstr)));
3732 if (SvTAINTED(sstr))
3734 if (GvIMPORTED(dstr) != GVf_IMPORTED
3735 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3737 GvIMPORTED_on(dstr);
3740 if(mro_changes == 2) {
3742 SV * const sref = (SV *)GvAV((const GV *)dstr);
3743 if (SvSMAGICAL(sref) && (mg = mg_find(sref, PERL_MAGIC_isa))) {
3744 if (SvTYPE(mg->mg_obj) != SVt_PVAV) {
3745 AV * const ary = newAV();
3746 av_push(ary, mg->mg_obj); /* takes the refcount */
3747 mg->mg_obj = (SV *)ary;
3749 av_push((AV *)mg->mg_obj, SvREFCNT_inc_simple_NN(dstr));
3751 else sv_magic(sref, dstr, PERL_MAGIC_isa, NULL, 0);
3752 mro_isa_changed_in(GvSTASH(dstr));
3754 else if(mro_changes == 3) {
3755 HV * const stash = GvHV(dstr);
3756 if(old_stash ? (HV *)HvENAME_get(old_stash) : stash)
3762 else if(mro_changes) mro_method_changed_in(GvSTASH(dstr));
3767 S_glob_assign_ref(pTHX_ SV *const dstr, SV *const sstr)
3769 SV * const sref = SvREFCNT_inc(SvRV(sstr));
3771 const int intro = GvINTRO(dstr);
3774 const U32 stype = SvTYPE(sref);
3776 PERL_ARGS_ASSERT_GLOB_ASSIGN_REF;
3779 GvINTRO_off(dstr); /* one-shot flag */
3780 GvLINE(dstr) = CopLINE(PL_curcop);
3781 GvEGV(dstr) = MUTABLE_GV(dstr);
3786 location = (SV **) &(GvGP(dstr)->gp_cv); /* XXX bypassing GvCV_set */
3787 import_flag = GVf_IMPORTED_CV;
3790 location = (SV **) &GvHV(dstr);
3791 import_flag = GVf_IMPORTED_HV;
3794 location = (SV **) &GvAV(dstr);
3795 import_flag = GVf_IMPORTED_AV;
3798 location = (SV **) &GvIOp(dstr);
3801 location = (SV **) &GvFORM(dstr);
3804 location = &GvSV(dstr);
3805 import_flag = GVf_IMPORTED_SV;
3808 if (stype == SVt_PVCV) {
3809 /*if (GvCVGEN(dstr) && (GvCV(dstr) != (const CV *)sref || GvCVGEN(dstr))) {*/
3810 if (GvCVGEN(dstr)) {
3811 SvREFCNT_dec(GvCV(dstr));
3812 GvCV_set(dstr, NULL);
3813 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3816 SAVEGENERICSV(*location);
3820 if (stype == SVt_PVCV && (*location != sref || GvCVGEN(dstr))) {
3821 CV* const cv = MUTABLE_CV(*location);
3823 if (!GvCVGEN((const GV *)dstr) &&
3824 (CvROOT(cv) || CvXSUB(cv)))
3826 /* Redefining a sub - warning is mandatory if
3827 it was a const and its value changed. */
3828 if (CvCONST(cv) && CvCONST((const CV *)sref)
3830 == cv_const_sv((const CV *)sref)) {
3832 /* They are 2 constant subroutines generated from
3833 the same constant. This probably means that
3834 they are really the "same" proxy subroutine
3835 instantiated in 2 places. Most likely this is
3836 when a constant is exported twice. Don't warn.
3839 else if (ckWARN(WARN_REDEFINE)
3841 && (!CvCONST((const CV *)sref)
3842 || sv_cmp(cv_const_sv(cv),
3843 cv_const_sv((const CV *)
3845 Perl_warner(aTHX_ packWARN(WARN_REDEFINE),
3848 ? "Constant subroutine %s::%s redefined"
3849 : "Subroutine %s::%s redefined"),
3850 HvNAME_get(GvSTASH((const GV *)dstr)),
3851 GvENAME(MUTABLE_GV(dstr)));
3855 cv_ckproto_len(cv, (const GV *)dstr,
3856 SvPOK(sref) ? SvPVX_const(sref) : NULL,
3857 SvPOK(sref) ? SvCUR(sref) : 0);
3859 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3860 GvASSUMECV_on(dstr);
3861 if(GvSTASH(dstr)) mro_method_changed_in(GvSTASH(dstr)); /* sub foo { 1 } sub bar { 2 } *bar = \&foo */
3864 if (import_flag && !(GvFLAGS(dstr) & import_flag)
3865 && CopSTASH_ne(PL_curcop, GvSTASH(dstr))) {
3866 GvFLAGS(dstr) |= import_flag;
3868 if (stype == SVt_PVHV) {
3869 const char * const name = GvNAME((GV*)dstr);
3870 const STRLEN len = GvNAMELEN(dstr);
3873 (len > 1 && name[len-2] == ':' && name[len-1] == ':')
3874 || (len == 1 && name[0] == ':')
3876 && (!dref || HvENAME_get(dref))
3879 (HV *)sref, (HV *)dref,
3885 stype == SVt_PVAV && sref != dref
3886 && strEQ(GvNAME((GV*)dstr), "ISA")
3887 /* The stash may have been detached from the symbol table, so
3888 check its name before doing anything. */
3889 && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3892 MAGIC * const omg = dref && SvSMAGICAL(dref)
3893 ? mg_find(dref, PERL_MAGIC_isa)
3895 if (SvSMAGICAL(sref) && (mg = mg_find(sref, PERL_MAGIC_isa))) {
3896 if (SvTYPE(mg->mg_obj) != SVt_PVAV) {
3897 AV * const ary = newAV();
3898 av_push(ary, mg->mg_obj); /* takes the refcount */
3899 mg->mg_obj = (SV *)ary;
3902 if (SvTYPE(omg->mg_obj) == SVt_PVAV) {
3903 SV **svp = AvARRAY((AV *)omg->mg_obj);
3904 I32 items = AvFILLp((AV *)omg->mg_obj) + 1;
3908 SvREFCNT_inc_simple_NN(*svp++)
3914 SvREFCNT_inc_simple_NN(omg->mg_obj)
3918 av_push((AV *)mg->mg_obj,SvREFCNT_inc_simple_NN(dstr));
3923 sref, omg ? omg->mg_obj : dstr, PERL_MAGIC_isa, NULL, 0
3925 mg = mg_find(sref, PERL_MAGIC_isa);
3927 /* Since the *ISA assignment could have affected more than
3928 one stash, don’t call mro_isa_changed_in directly, but let
3929 magic_clearisa do it for us, as it already has the logic for
3930 dealing with globs vs arrays of globs. */
3932 Perl_magic_clearisa(aTHX_ NULL, mg);
3937 if (SvTAINTED(sstr))
3943 Perl_sv_setsv_flags(pTHX_ SV *dstr, register SV* sstr, const I32 flags)
3946 register U32 sflags;
3948 register svtype stype;
3950 PERL_ARGS_ASSERT_SV_SETSV_FLAGS;
3955 if (SvIS_FREED(dstr)) {
3956 Perl_croak(aTHX_ "panic: attempt to copy value %" SVf
3957 " to a freed scalar %p", SVfARG(sstr), (void *)dstr);
3959 SV_CHECK_THINKFIRST_COW_DROP(dstr);
3961 sstr = &PL_sv_undef;
3962 if (SvIS_FREED(sstr)) {
3963 Perl_croak(aTHX_ "panic: attempt to copy freed scalar %p to %p",
3964 (void*)sstr, (void*)dstr);
3966 stype = SvTYPE(sstr);
3967 dtype = SvTYPE(dstr);
3969 (void)SvAMAGIC_off(dstr);
3972 /* need to nuke the magic */
3976 /* There's a lot of redundancy below but we're going for speed here */
3981 if (dtype != SVt_PVGV && dtype != SVt_PVLV) {
3982 (void)SvOK_off(dstr);
3990 sv_upgrade(dstr, SVt_IV);
3994 sv_upgrade(dstr, SVt_PVIV);
3998 goto end_of_first_switch;
4000 (void)SvIOK_only(dstr);
4001 SvIV_set(dstr, SvIVX(sstr));
4004 /* SvTAINTED can only be true if the SV has taint magic, which in
4005 turn means that the SV type is PVMG (or greater). This is the
4006 case statement for SVt_IV, so this cannot be true (whatever gcov
4008 assert(!SvTAINTED(sstr));
4013 if (dtype < SVt_PV && dtype != SVt_IV)
4014 sv_upgrade(dstr, SVt_IV);
4022 sv_upgrade(dstr, SVt_NV);
4026 sv_upgrade(dstr, SVt_PVNV);
4030 goto end_of_first_switch;
4032 SvNV_set(dstr, SvNVX(sstr));
4033 (void)SvNOK_only(dstr);
4034 /* SvTAINTED can only be true if the SV has taint magic, which in
4035 turn means that the SV type is PVMG (or greater). This is the
4036 case statement for SVt_NV, so this cannot be true (whatever gcov
4038 assert(!SvTAINTED(sstr));
4044 #ifdef PERL_OLD_COPY_ON_WRITE
4045 if ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS) {
4046 if (dtype < SVt_PVIV)
4047 sv_upgrade(dstr, SVt_PVIV);
4054 sv_upgrade(dstr, SVt_PV);
4057 if (dtype < SVt_PVIV)
4058 sv_upgrade(dstr, SVt_PVIV);
4061 if (dtype < SVt_PVNV)
4062 sv_upgrade(dstr, SVt_PVNV);
4066 const char * const type = sv_reftype(sstr,0);
4068 Perl_croak(aTHX_ "Bizarre copy of %s in %s", type, OP_DESC(PL_op));
4070 Perl_croak(aTHX_ "Bizarre copy of %s", type);
4075 if (dtype < SVt_REGEXP)
4076 sv_upgrade(dstr, SVt_REGEXP);
4079 /* case SVt_BIND: */
4083 if (SvGMAGICAL(sstr) && (flags & SV_GMAGIC)) {
4085 if (SvTYPE(sstr) != stype)
4086 stype = SvTYPE(sstr);
4088 if (isGV_with_GP(sstr) && dtype <= SVt_PVLV) {
4089 glob_assign_glob(dstr, sstr, dtype);
4092 if (stype == SVt_PVLV)
4093 SvUPGRADE(dstr, SVt_PVNV);
4095 SvUPGRADE(dstr, (svtype)stype);
4097 end_of_first_switch:
4099 /* dstr may have been upgraded. */
4100 dtype = SvTYPE(dstr);
4101 sflags = SvFLAGS(sstr);
4103 if (dtype == SVt_PVCV || dtype == SVt_PVFM) {
4104 /* Assigning to a subroutine sets the prototype. */
4107 const char *const ptr = SvPV_const(sstr, len);
4109 SvGROW(dstr, len + 1);
4110 Copy(ptr, SvPVX(dstr), len + 1, char);
4111 SvCUR_set(dstr, len);
4113 SvFLAGS(dstr) |= sflags & SVf_UTF8;
4117 } else if (dtype == SVt_PVAV || dtype == SVt_PVHV) {
4118 const char * const type = sv_reftype(dstr,0);
4120 Perl_croak(aTHX_ "Cannot copy to %s in %s", type, OP_DESC(PL_op));
4122 Perl_croak(aTHX_ "Cannot copy to %s", type);
4123 } else if (sflags & SVf_ROK) {
4124 if (isGV_with_GP(dstr)
4125 && SvTYPE(SvRV(sstr)) == SVt_PVGV && isGV_with_GP(SvRV(sstr))) {
4128 if (GvIMPORTED(dstr) != GVf_IMPORTED
4129 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
4131 GvIMPORTED_on(dstr);
4136 glob_assign_glob(dstr, sstr, dtype);
4140 if (dtype >= SVt_PV) {
4141 if (isGV_with_GP(dstr)) {
4142 glob_assign_ref(dstr, sstr);
4145 if (SvPVX_const(dstr)) {
4151 (void)SvOK_off(dstr);
4152 SvRV_set(dstr, SvREFCNT_inc(SvRV(sstr)));
4153 SvFLAGS(dstr) |= sflags & SVf_ROK;
4154 assert(!(sflags & SVp_NOK));
4155 assert(!(sflags & SVp_IOK));
4156 assert(!(sflags & SVf_NOK));
4157 assert(!(sflags & SVf_IOK));
4159 else if (isGV_with_GP(dstr)) {
4160 if (!(sflags & SVf_OK)) {
4161 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
4162 "Undefined value assigned to typeglob");
4165 GV *gv = gv_fetchsv_nomg(sstr, GV_ADD, SVt_PVGV);
4166 if (dstr != (const SV *)gv) {
4167 const char * const name = GvNAME((const GV *)dstr);
4168 const STRLEN len = GvNAMELEN(dstr);
4169 HV *old_stash = NULL;
4170 bool reset_isa = FALSE;
4171 if ((len > 1 && name[len-2] == ':' && name[len-1] == ':')
4172 || (len == 1 && name[0] == ':')) {
4173 /* Set aside the old stash, so we can reset isa caches
4174 on its subclasses. */
4175 if((old_stash = GvHV(dstr))) {
4176 /* Make sure we do not lose it early. */
4177 SvREFCNT_inc_simple_void_NN(
4178 sv_2mortal((SV *)old_stash)
4185 gp_free(MUTABLE_GV(dstr));
4186 GvGP_set(dstr, gp_ref(GvGP(gv)));
4189 HV * const stash = GvHV(dstr);
4191 old_stash ? (HV *)HvENAME_get(old_stash) : stash
4201 else if (dtype == SVt_REGEXP && stype == SVt_REGEXP) {
4202 reg_temp_copy((REGEXP*)dstr, (REGEXP*)sstr);
4204 else if (sflags & SVp_POK) {
4208 * Check to see if we can just swipe the string. If so, it's a
4209 * possible small lose on short strings, but a big win on long ones.
4210 * It might even be a win on short strings if SvPVX_const(dstr)
4211 * has to be allocated and SvPVX_const(sstr) has to be freed.
4212 * Likewise if we can set up COW rather than doing an actual copy, we
4213 * drop to the else clause, as the swipe code and the COW setup code
4214 * have much in common.
4217 /* Whichever path we take through the next code, we want this true,
4218 and doing it now facilitates the COW check. */
4219 (void)SvPOK_only(dstr);
4222 /* If we're already COW then this clause is not true, and if COW
4223 is allowed then we drop down to the else and make dest COW
4224 with us. If caller hasn't said that we're allowed to COW
4225 shared hash keys then we don't do the COW setup, even if the
4226 source scalar is a shared hash key scalar. */
4227 (((flags & SV_COW_SHARED_HASH_KEYS)
4228 ? (sflags & (SVf_FAKE|SVf_READONLY)) != (SVf_FAKE|SVf_READONLY)
4229 : 1 /* If making a COW copy is forbidden then the behaviour we
4230 desire is as if the source SV isn't actually already
4231 COW, even if it is. So we act as if the source flags
4232 are not COW, rather than actually testing them. */
4234 #ifndef PERL_OLD_COPY_ON_WRITE
4235 /* The change that added SV_COW_SHARED_HASH_KEYS makes the logic
4236 when PERL_OLD_COPY_ON_WRITE is defined a little wrong.
4237 Conceptually PERL_OLD_COPY_ON_WRITE being defined should
4238 override SV_COW_SHARED_HASH_KEYS, because it means "always COW"
4239 but in turn, it's somewhat dead code, never expected to go
4240 live, but more kept as a placeholder on how to do it better
4241 in a newer implementation. */
4242 /* If we are COW and dstr is a suitable target then we drop down
4243 into the else and make dest a COW of us. */
4244 || (SvFLAGS(dstr) & CAN_COW_MASK) != CAN_COW_FLAGS
4249 (sflags & SVs_TEMP) && /* slated for free anyway? */
4250 !(sflags & SVf_OOK) && /* and not involved in OOK hack? */
4251 (!(flags & SV_NOSTEAL)) &&
4252 /* and we're allowed to steal temps */
4253 SvREFCNT(sstr) == 1 && /* and no other references to it? */
4254 SvLEN(sstr)) /* and really is a string */
4255 #ifdef PERL_OLD_COPY_ON_WRITE
4256 && ((flags & SV_COW_SHARED_HASH_KEYS)
4257 ? (!((sflags & CAN_COW_MASK) == CAN_COW_FLAGS
4258 && (SvFLAGS(dstr) & CAN_COW_MASK) == CAN_COW_FLAGS
4259 && SvTYPE(sstr) >= SVt_PVIV && SvTYPE(sstr) != SVt_PVFM))
4263 /* Failed the swipe test, and it's not a shared hash key either.
4264 Have to copy the string. */
4265 STRLEN len = SvCUR(sstr);
4266 SvGROW(dstr, len + 1); /* inlined from sv_setpvn */
4267 Move(SvPVX_const(sstr),SvPVX(dstr),len,char);
4268 SvCUR_set(dstr, len);
4269 *SvEND(dstr) = '\0';
4271 /* If PERL_OLD_COPY_ON_WRITE is not defined, then isSwipe will always
4273 /* Either it's a shared hash key, or it's suitable for
4274 copy-on-write or we can swipe the string. */
4276 PerlIO_printf(Perl_debug_log, "Copy on write: sstr --> dstr\n");
4280 #ifdef PERL_OLD_COPY_ON_WRITE
4282 if ((sflags & (SVf_FAKE | SVf_READONLY))
4283 != (SVf_FAKE | SVf_READONLY)) {
4284 SvREADONLY_on(sstr);
4286 /* Make the source SV into a loop of 1.
4287 (about to become 2) */
4288 SV_COW_NEXT_SV_SET(sstr, sstr);
4292 /* Initial code is common. */
4293 if (SvPVX_const(dstr)) { /* we know that dtype >= SVt_PV */
4298 /* making another shared SV. */
4299 STRLEN cur = SvCUR(sstr);
4300 STRLEN len = SvLEN(sstr);
4301 #ifdef PERL_OLD_COPY_ON_WRITE
4303 assert (SvTYPE(dstr) >= SVt_PVIV);
4304 /* SvIsCOW_normal */
4305 /* splice us in between source and next-after-source. */
4306 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4307 SV_COW_NEXT_SV_SET(sstr, dstr);
4308 SvPV_set(dstr, SvPVX_mutable(sstr));
4312 /* SvIsCOW_shared_hash */
4313 DEBUG_C(PerlIO_printf(Perl_debug_log,
4314 "Copy on write: Sharing hash\n"));
4316 assert (SvTYPE(dstr) >= SVt_PV);
4318 HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)))));
4320 SvLEN_set(dstr, len);
4321 SvCUR_set(dstr, cur);
4322 SvREADONLY_on(dstr);
4326 { /* Passes the swipe test. */
4327 SvPV_set(dstr, SvPVX_mutable(sstr));
4328 SvLEN_set(dstr, SvLEN(sstr));
4329 SvCUR_set(dstr, SvCUR(sstr));
4332 (void)SvOK_off(sstr); /* NOTE: nukes most SvFLAGS on sstr */
4333 SvPV_set(sstr, NULL);
4339 if (sflags & SVp_NOK) {
4340 SvNV_set(dstr, SvNVX(sstr));
4342 if (sflags & SVp_IOK) {
4343 SvIV_set(dstr, SvIVX(sstr));
4344 /* Must do this otherwise some other overloaded use of 0x80000000
4345 gets confused. I guess SVpbm_VALID */
4346 if (sflags & SVf_IVisUV)
4349 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_NOK|SVp_NOK|SVf_UTF8);
4351 const MAGIC * const smg = SvVSTRING_mg(sstr);
4353 sv_magic(dstr, NULL, PERL_MAGIC_vstring,
4354 smg->mg_ptr, smg->mg_len);
4355 SvRMAGICAL_on(dstr);
4359 else if (sflags & (SVp_IOK|SVp_NOK)) {
4360 (void)SvOK_off(dstr);
4361 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_IVisUV|SVf_NOK|SVp_NOK);
4362 if (sflags & SVp_IOK) {
4363 /* XXXX Do we want to set IsUV for IV(ROK)? Be extra safe... */
4364 SvIV_set(dstr, SvIVX(sstr));
4366 if (sflags & SVp_NOK) {
4367 SvNV_set(dstr, SvNVX(sstr));
4371 if (isGV_with_GP(sstr)) {
4372 /* This stringification rule for globs is spread in 3 places.
4373 This feels bad. FIXME. */
4374 const U32 wasfake = sflags & SVf_FAKE;
4376 /* FAKE globs can get coerced, so need to turn this off
4377 temporarily if it is on. */
4379 gv_efullname3(dstr, MUTABLE_GV(sstr), "*");
4380 SvFLAGS(sstr) |= wasfake;
4383 (void)SvOK_off(dstr);
4385 if (SvTAINTED(sstr))
4390 =for apidoc sv_setsv_mg
4392 Like C<sv_setsv>, but also handles 'set' magic.
4398 Perl_sv_setsv_mg(pTHX_ SV *const dstr, register SV *const sstr)
4400 PERL_ARGS_ASSERT_SV_SETSV_MG;
4402 sv_setsv(dstr,sstr);
4406 #ifdef PERL_OLD_COPY_ON_WRITE
4408 Perl_sv_setsv_cow(pTHX_ SV *dstr, SV *sstr)
4410 STRLEN cur = SvCUR(sstr);
4411 STRLEN len = SvLEN(sstr);
4412 register char *new_pv;
4414 PERL_ARGS_ASSERT_SV_SETSV_COW;
4417 PerlIO_printf(Perl_debug_log, "Fast copy on write: %p -> %p\n",
4418 (void*)sstr, (void*)dstr);
4425 if (SvTHINKFIRST(dstr))
4426 sv_force_normal_flags(dstr, SV_COW_DROP_PV);
4427 else if (SvPVX_const(dstr))
4428 Safefree(SvPVX_const(dstr));
4432 SvUPGRADE(dstr, SVt_PVIV);
4434 assert (SvPOK(sstr));
4435 assert (SvPOKp(sstr));
4436 assert (!SvIOK(sstr));
4437 assert (!SvIOKp(sstr));
4438 assert (!SvNOK(sstr));
4439 assert (!SvNOKp(sstr));
4441 if (SvIsCOW(sstr)) {
4443 if (SvLEN(sstr) == 0) {
4444 /* source is a COW shared hash key. */
4445 DEBUG_C(PerlIO_printf(Perl_debug_log,
4446 "Fast copy on write: Sharing hash\n"));
4447 new_pv = HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr))));
4450 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4452 assert ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS);
4453 SvUPGRADE(sstr, SVt_PVIV);
4454 SvREADONLY_on(sstr);
4456 DEBUG_C(PerlIO_printf(Perl_debug_log,
4457 "Fast copy on write: Converting sstr to COW\n"));
4458 SV_COW_NEXT_SV_SET(dstr, sstr);
4460 SV_COW_NEXT_SV_SET(sstr, dstr);
4461 new_pv = SvPVX_mutable(sstr);
4464 SvPV_set(dstr, new_pv);
4465 SvFLAGS(dstr) = (SVt_PVIV|SVf_POK|SVp_POK|SVf_FAKE|SVf_READONLY);
4468 SvLEN_set(dstr, len);
4469 SvCUR_set(dstr, cur);
4478 =for apidoc sv_setpvn
4480 Copies a string into an SV. The C<len> parameter indicates the number of
4481 bytes to be copied. If the C<ptr> argument is NULL the SV will become
4482 undefined. Does not handle 'set' magic. See C<sv_setpvn_mg>.
4488 Perl_sv_setpvn(pTHX_ register SV *const sv, register const char *const ptr, register const STRLEN len)
4491 register char *dptr;
4493 PERL_ARGS_ASSERT_SV_SETPVN;
4495 SV_CHECK_THINKFIRST_COW_DROP(sv);
4501 /* len is STRLEN which is unsigned, need to copy to signed */
4504 Perl_croak(aTHX_ "panic: sv_setpvn called with negative strlen");
4506 SvUPGRADE(sv, SVt_PV);
4508 dptr = SvGROW(sv, len + 1);
4509 Move(ptr,dptr,len,char);
4512 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4517 =for apidoc sv_setpvn_mg
4519 Like C<sv_setpvn>, but also handles 'set' magic.
4525 Perl_sv_setpvn_mg(pTHX_ register SV *const sv, register const char *const ptr, register const STRLEN len)
4527 PERL_ARGS_ASSERT_SV_SETPVN_MG;
4529 sv_setpvn(sv,ptr,len);
4534 =for apidoc sv_setpv
4536 Copies a string into an SV. The string must be null-terminated. Does not
4537 handle 'set' magic. See C<sv_setpv_mg>.
4543 Perl_sv_setpv(pTHX_ register SV *const sv, register const char *const ptr)
4546 register STRLEN len;
4548 PERL_ARGS_ASSERT_SV_SETPV;
4550 SV_CHECK_THINKFIRST_COW_DROP(sv);
4556 SvUPGRADE(sv, SVt_PV);
4558 SvGROW(sv, len + 1);
4559 Move(ptr,SvPVX(sv),len+1,char);
4561 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4566 =for apidoc sv_setpv_mg
4568 Like C<sv_setpv>, but also handles 'set' magic.
4574 Perl_sv_setpv_mg(pTHX_ register SV *const sv, register const char *const ptr)
4576 PERL_ARGS_ASSERT_SV_SETPV_MG;
4583 =for apidoc sv_usepvn_flags
4585 Tells an SV to use C<ptr> to find its string value. Normally the
4586 string is stored inside the SV but sv_usepvn allows the SV to use an
4587 outside string. The C<ptr> should point to memory that was allocated
4588 by C<malloc>. The string length, C<len>, must be supplied. By default
4589 this function will realloc (i.e. move) the memory pointed to by C<ptr>,
4590 so that pointer should not be freed or used by the programmer after
4591 giving it to sv_usepvn, and neither should any pointers from "behind"
4592 that pointer (e.g. ptr + 1) be used.
4594 If C<flags> & SV_SMAGIC is true, will call SvSETMAGIC. If C<flags> &
4595 SV_HAS_TRAILING_NUL is true, then C<ptr[len]> must be NUL, and the realloc
4596 will be skipped. (i.e. the buffer is actually at least 1 byte longer than
4597 C<len>, and already meets the requirements for storing in C<SvPVX>)
4603 Perl_sv_usepvn_flags(pTHX_ SV *const sv, char *ptr, const STRLEN len, const U32 flags)
4608 PERL_ARGS_ASSERT_SV_USEPVN_FLAGS;
4610 SV_CHECK_THINKFIRST_COW_DROP(sv);
4611 SvUPGRADE(sv, SVt_PV);
4614 if (flags & SV_SMAGIC)
4618 if (SvPVX_const(sv))
4622 if (flags & SV_HAS_TRAILING_NUL)
4623 assert(ptr[len] == '\0');
4626 allocate = (flags & SV_HAS_TRAILING_NUL)
4628 #ifdef Perl_safesysmalloc_size
4631 PERL_STRLEN_ROUNDUP(len + 1);
4633 if (flags & SV_HAS_TRAILING_NUL) {
4634 /* It's long enough - do nothing.
4635 Specifically Perl_newCONSTSUB is relying on this. */
4638 /* Force a move to shake out bugs in callers. */
4639 char *new_ptr = (char*)safemalloc(allocate);
4640 Copy(ptr, new_ptr, len, char);
4641 PoisonFree(ptr,len,char);
4645 ptr = (char*) saferealloc (ptr, allocate);
4648 #ifdef Perl_safesysmalloc_size
4649 SvLEN_set(sv, Perl_safesysmalloc_size(ptr));
4651 SvLEN_set(sv, allocate);
4655 if (!(flags & SV_HAS_TRAILING_NUL)) {
4658 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4660 if (flags & SV_SMAGIC)
4664 #ifdef PERL_OLD_COPY_ON_WRITE
4665 /* Need to do this *after* making the SV normal, as we need the buffer
4666 pointer to remain valid until after we've copied it. If we let go too early,
4667 another thread could invalidate it by unsharing last of the same hash key
4668 (which it can do by means other than releasing copy-on-write Svs)
4669 or by changing the other copy-on-write SVs in the loop. */
4671 S_sv_release_COW(pTHX_ register SV *sv, const char *pvx, SV *after)
4673 PERL_ARGS_ASSERT_SV_RELEASE_COW;
4675 { /* this SV was SvIsCOW_normal(sv) */
4676 /* we need to find the SV pointing to us. */
4677 SV *current = SV_COW_NEXT_SV(after);
4679 if (current == sv) {
4680 /* The SV we point to points back to us (there were only two of us
4682 Hence other SV is no longer copy on write either. */
4684 SvREADONLY_off(after);
4686 /* We need to follow the pointers around the loop. */
4688 while ((next = SV_COW_NEXT_SV(current)) != sv) {
4691 /* don't loop forever if the structure is bust, and we have
4692 a pointer into a closed loop. */
4693 assert (current != after);
4694 assert (SvPVX_const(current) == pvx);
4696 /* Make the SV before us point to the SV after us. */
4697 SV_COW_NEXT_SV_SET(current, after);
4703 =for apidoc sv_force_normal_flags
4705 Undo various types of fakery on an SV: if the PV is a shared string, make
4706 a private copy; if we're a ref, stop refing; if we're a glob, downgrade to
4707 an xpvmg; if we're a copy-on-write scalar, this is the on-write time when
4708 we do the copy, and is also used locally. If C<SV_COW_DROP_PV> is set
4709 then a copy-on-write scalar drops its PV buffer (if any) and becomes
4710 SvPOK_off rather than making a copy. (Used where this scalar is about to be
4711 set to some other value.) In addition, the C<flags> parameter gets passed to
4712 C<sv_unref_flags()> when unreffing. C<sv_force_normal> calls this function
4713 with flags set to 0.
4719 Perl_sv_force_normal_flags(pTHX_ register SV *const sv, const U32 flags)
4723 PERL_ARGS_ASSERT_SV_FORCE_NORMAL_FLAGS;
4725 #ifdef PERL_OLD_COPY_ON_WRITE
4726 if (SvREADONLY(sv)) {
4728 const char * const pvx = SvPVX_const(sv);
4729 const STRLEN len = SvLEN(sv);
4730 const STRLEN cur = SvCUR(sv);
4731 /* next COW sv in the loop. If len is 0 then this is a shared-hash
4732 key scalar, so we mustn't attempt to call SV_COW_NEXT_SV(), as
4733 we'll fail an assertion. */
4734 SV * const next = len ? SV_COW_NEXT_SV(sv) : 0;
4737 PerlIO_printf(Perl_debug_log,
4738 "Copy on write: Force normal %ld\n",
4744 /* This SV doesn't own the buffer, so need to Newx() a new one: */
4747 if (flags & SV_COW_DROP_PV) {
4748 /* OK, so we don't need to copy our buffer. */
4751 SvGROW(sv, cur + 1);
4752 Move(pvx,SvPVX(sv),cur,char);
4757 sv_release_COW(sv, pvx, next);
4759 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4765 else if (IN_PERL_RUNTIME)
4766 Perl_croak_no_modify(aTHX);
4769 if (SvREADONLY(sv)) {
4770 if (SvFAKE(sv) && !isGV_with_GP(sv)) {
4771 const char * const pvx = SvPVX_const(sv);
4772 const STRLEN len = SvCUR(sv);
4777 SvGROW(sv, len + 1);
4778 Move(pvx,SvPVX(sv),len,char);
4780 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4782 else if (IN_PERL_RUNTIME)
4783 Perl_croak_no_modify(aTHX);
4787 sv_unref_flags(sv, flags);
4788 else if (SvFAKE(sv) && isGV_with_GP(sv))
4790 else if (SvFAKE(sv) && SvTYPE(sv) == SVt_REGEXP) {
4791 /* Need to downgrade the REGEXP to a simple(r) scalar. This is analogous
4792 to sv_unglob. We only need it here, so inline it. */
4793 const svtype new_type = SvMAGIC(sv) || SvSTASH(sv) ? SVt_PVMG : SVt_PV;
4794 SV *const temp = newSV_type(new_type);
4795 void *const temp_p = SvANY(sv);
4797 if (new_type == SVt_PVMG) {
4798 SvMAGIC_set(temp, SvMAGIC(sv));
4799 SvMAGIC_set(sv, NULL);
4800 SvSTASH_set(temp, SvSTASH(sv));
4801 SvSTASH_set(sv, NULL);
4803 SvCUR_set(temp, SvCUR(sv));
4804 /* Remember that SvPVX is in the head, not the body. */
4806 SvLEN_set(temp, SvLEN(sv));
4807 /* This signals "buffer is owned by someone else" in sv_clear,
4808 which is the least effort way to stop it freeing the buffer.
4810 SvLEN_set(sv, SvLEN(sv)+1);
4812 /* Their buffer is already owned by someone else. */
4813 SvPVX(sv) = savepvn(SvPVX(sv), SvCUR(sv));
4814 SvLEN_set(temp, SvCUR(sv)+1);
4817 /* Now swap the rest of the bodies. */
4819 SvFLAGS(sv) &= ~(SVf_FAKE|SVTYPEMASK);
4820 SvFLAGS(sv) |= new_type;
4821 SvANY(sv) = SvANY(temp);
4823 SvFLAGS(temp) &= ~(SVTYPEMASK);
4824 SvFLAGS(temp) |= SVt_REGEXP|SVf_FAKE;
4825 SvANY(temp) = temp_p;
4834 Efficient removal of characters from the beginning of the string buffer.
4835 SvPOK(sv) must be true and the C<ptr> must be a pointer to somewhere inside
4836 the string buffer. The C<ptr> becomes the first character of the adjusted
4837 string. Uses the "OOK hack".
4838 Beware: after this function returns, C<ptr> and SvPVX_const(sv) may no longer
4839 refer to the same chunk of data.
4845 Perl_sv_chop(pTHX_ register SV *const sv, register const char *const ptr)
4851 const U8 *real_start;
4855 PERL_ARGS_ASSERT_SV_CHOP;
4857 if (!ptr || !SvPOKp(sv))
4859 delta = ptr - SvPVX_const(sv);
4861 /* Nothing to do. */
4864 /* SvPVX(sv) may move in SV_CHECK_THINKFIRST(sv), but after this line,
4865 nothing uses the value of ptr any more. */
4866 max_delta = SvLEN(sv) ? SvLEN(sv) : SvCUR(sv);
4867 if (ptr <= SvPVX_const(sv))
4868 Perl_croak(aTHX_ "panic: sv_chop ptr=%p, start=%p, end=%p",
4869 ptr, SvPVX_const(sv), SvPVX_const(sv) + max_delta);
4870 SV_CHECK_THINKFIRST(sv);
4871 if (delta > max_delta)
4872 Perl_croak(aTHX_ "panic: sv_chop ptr=%p (was %p), start=%p, end=%p",
4873 SvPVX_const(sv) + delta, ptr, SvPVX_const(sv),
4874 SvPVX_const(sv) + max_delta);
4877 if (!SvLEN(sv)) { /* make copy of shared string */
4878 const char *pvx = SvPVX_const(sv);
4879 const STRLEN len = SvCUR(sv);
4880 SvGROW(sv, len + 1);
4881 Move(pvx,SvPVX(sv),len,char);
4884 SvFLAGS(sv) |= SVf_OOK;
4887 SvOOK_offset(sv, old_delta);
4889 SvLEN_set(sv, SvLEN(sv) - delta);
4890 SvCUR_set(sv, SvCUR(sv) - delta);
4891 SvPV_set(sv, SvPVX(sv) + delta);
4893 p = (U8 *)SvPVX_const(sv);
4898 real_start = p - delta;
4902 if (delta < 0x100) {
4906 p -= sizeof(STRLEN);
4907 Copy((U8*)&delta, p, sizeof(STRLEN), U8);
4911 /* Fill the preceding buffer with sentinals to verify that no-one is
4913 while (p > real_start) {
4921 =for apidoc sv_catpvn
4923 Concatenates the string onto the end of the string which is in the SV. The
4924 C<len> indicates number of bytes to copy. If the SV has the UTF-8
4925 status set, then the bytes appended should be valid UTF-8.
4926 Handles 'get' magic, but not 'set' magic. See C<sv_catpvn_mg>.
4928 =for apidoc sv_catpvn_flags
4930 Concatenates the string onto the end of the string which is in the SV. The
4931 C<len> indicates number of bytes to copy. If the SV has the UTF-8
4932 status set, then the bytes appended should be valid UTF-8.
4933 If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<dsv> if
4934 appropriate, else not. C<sv_catpvn> and C<sv_catpvn_nomg> are implemented
4935 in terms of this function.
4941 Perl_sv_catpvn_flags(pTHX_ register SV *const dsv, register const char *sstr, register const STRLEN slen, const I32 flags)
4945 const char * const dstr = SvPV_force_flags(dsv, dlen, flags);
4947 PERL_ARGS_ASSERT_SV_CATPVN_FLAGS;
4949 SvGROW(dsv, dlen + slen + 1);
4951 sstr = SvPVX_const(dsv);
4952 Move(sstr, SvPVX(dsv) + dlen, slen, char);
4953 SvCUR_set(dsv, SvCUR(dsv) + slen);
4955 (void)SvPOK_only_UTF8(dsv); /* validate pointer */
4957 if (flags & SV_SMAGIC)
4962 =for apidoc sv_catsv
4964 Concatenates the string from SV C<ssv> onto the end of the string in
4965 SV C<dsv>. Modifies C<dsv> but not C<ssv>. Handles 'get' magic, but
4966 not 'set' magic. See C<sv_catsv_mg>.
4968 =for apidoc sv_catsv_flags
4970 Concatenates the string from SV C<ssv> onto the end of the string in
4971 SV C<dsv>. Modifies C<dsv> but not C<ssv>. If C<flags> has C<SV_GMAGIC>
4972 bit set, will C<mg_get> on the SVs if appropriate, else not. C<sv_catsv>
4973 and C<sv_catsv_nomg> are implemented in terms of this function.
4978 Perl_sv_catsv_flags(pTHX_ SV *const dsv, register SV *const ssv, const I32 flags)
4982 PERL_ARGS_ASSERT_SV_CATSV_FLAGS;
4986 const char *spv = SvPV_flags_const(ssv, slen, flags);
4988 /* sutf8 and dutf8 were type bool, but under USE_ITHREADS,
4989 gcc version 2.95.2 20000220 (Debian GNU/Linux) for
4990 Linux xxx 2.2.17 on sparc64 with gcc -O2, we erroneously
4991 get dutf8 = 0x20000000, (i.e. SVf_UTF8) even though
4992 dsv->sv_flags doesn't have that bit set.
4993 Andy Dougherty 12 Oct 2001
4995 const I32 sutf8 = DO_UTF8(ssv);
4998 if (SvGMAGICAL(dsv) && (flags & SV_GMAGIC))
5000 dutf8 = DO_UTF8(dsv);
5002 if (dutf8 != sutf8) {
5004 /* Not modifying source SV, so taking a temporary copy. */
5005 SV* const csv = newSVpvn_flags(spv, slen, SVs_TEMP);
5007 sv_utf8_upgrade(csv);
5008 spv = SvPV_const(csv, slen);
5011 /* Leave enough space for the cat that's about to happen */
5012 sv_utf8_upgrade_flags_grow(dsv, 0, slen);
5014 sv_catpvn_nomg(dsv, spv, slen);
5017 if (flags & SV_SMAGIC)
5022 =for apidoc sv_catpv
5024 Concatenates the string onto the end of the string which is in the SV.
5025 If the SV has the UTF-8 status set, then the bytes appended should be
5026 valid UTF-8. Handles 'get' magic, but not 'set' magic. See C<sv_catpv_mg>.
5031 Perl_sv_catpv(pTHX_ register SV *const sv, register const char *ptr)
5034 register STRLEN len;
5038 PERL_ARGS_ASSERT_SV_CATPV;
5042 junk = SvPV_force(sv, tlen);
5044 SvGROW(sv, tlen + len + 1);
5046 ptr = SvPVX_const(sv);
5047 Move(ptr,SvPVX(sv)+tlen,len+1,char);
5048 SvCUR_set(sv, SvCUR(sv) + len);
5049 (void)SvPOK_only_UTF8(sv); /* validate pointer */
5054 =for apidoc sv_catpv_flags
5056 Concatenates the string onto the end of the string which is in the SV.
5057 If the SV has the UTF-8 status set, then the bytes appended should
5058 be valid UTF-8. If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get>
5059 on the SVs if appropriate, else not.
5065 Perl_sv_catpv_flags(pTHX_ SV *dstr, const char *sstr, const I32 flags)
5067 PERL_ARGS_ASSERT_SV_CATPV_FLAGS;
5068 sv_catpvn_flags(dstr, sstr, strlen(sstr), flags);
5072 =for apidoc sv_catpv_mg
5074 Like C<sv_catpv>, but also handles 'set' magic.
5080 Perl_sv_catpv_mg(pTHX_ register SV *const sv, register const char *const ptr)
5082 PERL_ARGS_ASSERT_SV_CATPV_MG;
5091 Creates a new SV. A non-zero C<len> parameter indicates the number of
5092 bytes of preallocated string space the SV should have. An extra byte for a
5093 trailing NUL is also reserved. (SvPOK is not set for the SV even if string
5094 space is allocated.) The reference count for the new SV is set to 1.
5096 In 5.9.3, newSV() replaces the older NEWSV() API, and drops the first
5097 parameter, I<x>, a debug aid which allowed callers to identify themselves.
5098 This aid has been superseded by a new build option, PERL_MEM_LOG (see
5099 L<perlhacktips/PERL_MEM_LOG>). The older API is still there for use in XS
5100 modules supporting older perls.
5106 Perl_newSV(pTHX_ const STRLEN len)
5113 sv_upgrade(sv, SVt_PV);
5114 SvGROW(sv, len + 1);
5119 =for apidoc sv_magicext
5121 Adds magic to an SV, upgrading it if necessary. Applies the
5122 supplied vtable and returns a pointer to the magic added.
5124 Note that C<sv_magicext> will allow things that C<sv_magic> will not.
5125 In particular, you can add magic to SvREADONLY SVs, and add more than
5126 one instance of the same 'how'.
5128 If C<namlen> is greater than zero then a C<savepvn> I<copy> of C<name> is
5129 stored, if C<namlen> is zero then C<name> is stored as-is and - as another
5130 special case - if C<(name && namlen == HEf_SVKEY)> then C<name> is assumed
5131 to contain an C<SV*> and is stored as-is with its REFCNT incremented.
5133 (This is now used as a subroutine by C<sv_magic>.)
5138 Perl_sv_magicext(pTHX_ SV *const sv, SV *const obj, const int how,
5139 const MGVTBL *const vtable, const char *const name, const I32 namlen)
5144 PERL_ARGS_ASSERT_SV_MAGICEXT;
5146 SvUPGRADE(sv, SVt_PVMG);
5147 Newxz(mg, 1, MAGIC);
5148 mg->mg_moremagic = SvMAGIC(sv);
5149 SvMAGIC_set(sv, mg);
5151 /* Sometimes a magic contains a reference loop, where the sv and
5152 object refer to each other. To prevent a reference loop that
5153 would prevent such objects being freed, we look for such loops
5154 and if we find one we avoid incrementing the object refcount.
5156 Note we cannot do this to avoid self-tie loops as intervening RV must
5157 have its REFCNT incremented to keep it in existence.
5160 if (!obj || obj == sv ||
5161 how == PERL_MAGIC_arylen ||
5162 how == PERL_MAGIC_symtab ||
5163 (SvTYPE(obj) == SVt_PVGV &&
5164 (GvSV(obj) == sv || GvHV(obj) == (const HV *)sv
5165 || GvAV(obj) == (const AV *)sv || GvCV(obj) == (const CV *)sv
5166 || GvIOp(obj) == (const IO *)sv || GvFORM(obj) == (const CV *)sv)))
5171 mg->mg_obj = SvREFCNT_inc_simple(obj);
5172 mg->mg_flags |= MGf_REFCOUNTED;
5175 /* Normal self-ties simply pass a null object, and instead of
5176 using mg_obj directly, use the SvTIED_obj macro to produce a
5177 new RV as needed. For glob "self-ties", we are tieing the PVIO
5178 with an RV obj pointing to the glob containing the PVIO. In
5179 this case, to avoid a reference loop, we need to weaken the
5183 if (how == PERL_MAGIC_tiedscalar && SvTYPE(sv) == SVt_PVIO &&
5184 obj && SvROK(obj) && GvIO(SvRV(obj)) == (const IO *)sv)
5190 mg->mg_len = namlen;
5193 mg->mg_ptr = savepvn(name, namlen);
5194 else if (namlen == HEf_SVKEY) {
5195 /* Yes, this is casting away const. This is only for the case of
5196 HEf_SVKEY. I think we need to document this aberation of the
5197 constness of the API, rather than making name non-const, as
5198 that change propagating outwards a long way. */
5199 mg->mg_ptr = (char*)SvREFCNT_inc_simple_NN((SV *)name);
5201 mg->mg_ptr = (char *) name;
5203 mg->mg_virtual = (MGVTBL *) vtable;
5207 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
5212 =for apidoc sv_magic
5214 Adds magic to an SV. First upgrades C<sv> to type C<SVt_PVMG> if necessary,
5215 then adds a new magic item of type C<how> to the head of the magic list.
5217 See C<sv_magicext> (which C<sv_magic> now calls) for a description of the
5218 handling of the C<name> and C<namlen> arguments.
5220 You need to use C<sv_magicext> to add magic to SvREADONLY SVs and also
5221 to add more than one instance of the same 'how'.
5227 Perl_sv_magic(pTHX_ register SV *const sv, SV *const obj, const int how,
5228 const char *const name, const I32 namlen)
5231 const MGVTBL *vtable;
5234 unsigned int vtable_index;
5236 PERL_ARGS_ASSERT_SV_MAGIC;
5238 if (how < 0 || (unsigned)how > C_ARRAY_LENGTH(PL_magic_data)
5239 || ((flags = PL_magic_data[how]),
5240 (vtable_index = flags & PERL_MAGIC_VTABLE_MASK)
5241 > magic_vtable_max))
5242 Perl_croak(aTHX_ "Don't know how to handle magic of type \\%o", how);
5244 /* PERL_MAGIC_ext is reserved for use by extensions not perl internals.
5245 Useful for attaching extension internal data to perl vars.
5246 Note that multiple extensions may clash if magical scalars
5247 etc holding private data from one are passed to another. */
5249 vtable = (vtable_index == magic_vtable_max)
5250 ? NULL : PL_magic_vtables + vtable_index;
5252 #ifdef PERL_OLD_COPY_ON_WRITE
5254 sv_force_normal_flags(sv, 0);
5256 if (SvREADONLY(sv)) {
5258 /* its okay to attach magic to shared strings; the subsequent
5259 * upgrade to PVMG will unshare the string */
5260 !(SvFAKE(sv) && SvTYPE(sv) < SVt_PVMG)
5263 && !PERL_MAGIC_TYPE_READONLY_ACCEPTABLE(how)
5266 Perl_croak_no_modify(aTHX);
5269 if (SvMAGICAL(sv) || (how == PERL_MAGIC_taint && SvTYPE(sv) >= SVt_PVMG)) {
5270 if (SvMAGIC(sv) && (mg = mg_find(sv, how))) {
5271 /* sv_magic() refuses to add a magic of the same 'how' as an
5274 if (how == PERL_MAGIC_taint) {
5276 /* Any scalar which already had taint magic on which someone
5277 (erroneously?) did SvIOK_on() or similar will now be
5278 incorrectly sporting public "OK" flags. */
5279 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
5285 /* Rest of work is done else where */
5286 mg = sv_magicext(sv,obj,how,vtable,name,namlen);
5289 case PERL_MAGIC_taint:
5292 case PERL_MAGIC_ext:
5293 case PERL_MAGIC_dbfile:
5300 S_sv_unmagicext_flags(pTHX_ SV *const sv, const int type, MGVTBL *vtbl, const U32 flags)
5307 if (SvTYPE(sv) < SVt_PVMG || !SvMAGIC(sv))
5309 mgp = &(((XPVMG*) SvANY(sv))->xmg_u.xmg_magic);
5310 for (mg = *mgp; mg; mg = *mgp) {
5311 const MGVTBL* const virt = mg->mg_virtual;
5312 if (mg->mg_type == type && (!flags || virt == vtbl)) {
5313 *mgp = mg->mg_moremagic;
5314 if (virt && virt->svt_free)
5315 virt->svt_free(aTHX_ sv, mg);
5316 if (mg->mg_ptr && mg->mg_type != PERL_MAGIC_regex_global) {
5318 Safefree(mg->mg_ptr);
5319 else if (mg->mg_len == HEf_SVKEY)
5320 SvREFCNT_dec(MUTABLE_SV(mg->mg_ptr));
5321 else if (mg->mg_type == PERL_MAGIC_utf8)
5322 Safefree(mg->mg_ptr);
5324 if (mg->mg_flags & MGf_REFCOUNTED)
5325 SvREFCNT_dec(mg->mg_obj);
5329 mgp = &mg->mg_moremagic;
5332 if (SvMAGICAL(sv)) /* if we're under save_magic, wait for restore_magic; */
5333 mg_magical(sv); /* else fix the flags now */
5337 SvFLAGS(sv) |= (SvFLAGS(sv) & (SVp_IOK|SVp_NOK|SVp_POK)) >> PRIVSHIFT;
5343 =for apidoc sv_unmagic
5345 Removes all magic of type C<type> from an SV.
5351 Perl_sv_unmagic(pTHX_ SV *const sv, const int type)
5353 PERL_ARGS_ASSERT_SV_UNMAGIC;
5354 return S_sv_unmagicext_flags(aTHX_ sv, type, NULL, 0);
5358 =for apidoc sv_unmagicext
5360 Removes all magic of type C<type> with the specified C<vtbl> from an SV.
5366 Perl_sv_unmagicext(pTHX_ SV *const sv, const int type, MGVTBL *vtbl)
5368 PERL_ARGS_ASSERT_SV_UNMAGICEXT;
5369 return S_sv_unmagicext_flags(aTHX_ sv, type, vtbl, 1);
5373 =for apidoc sv_rvweaken
5375 Weaken a reference: set the C<SvWEAKREF> flag on this RV; give the
5376 referred-to SV C<PERL_MAGIC_backref> magic if it hasn't already; and
5377 push a back-reference to this RV onto the array of backreferences
5378 associated with that magic. If the RV is magical, set magic will be
5379 called after the RV is cleared.
5385 Perl_sv_rvweaken(pTHX_ SV *const sv)
5389 PERL_ARGS_ASSERT_SV_RVWEAKEN;
5391 if (!SvOK(sv)) /* let undefs pass */
5394 Perl_croak(aTHX_ "Can't weaken a nonreference");
5395 else if (SvWEAKREF(sv)) {
5396 Perl_ck_warner(aTHX_ packWARN(WARN_MISC), "Reference is already weak");
5399 else if (SvREADONLY(sv)) croak_no_modify();
5401 Perl_sv_add_backref(aTHX_ tsv, sv);
5407 /* Give tsv backref magic if it hasn't already got it, then push a
5408 * back-reference to sv onto the array associated with the backref magic.
5410 * As an optimisation, if there's only one backref and it's not an AV,
5411 * store it directly in the HvAUX or mg_obj slot, avoiding the need to
5412 * allocate an AV. (Whether the slot holds an AV tells us whether this is
5416 /* A discussion about the backreferences array and its refcount:
5418 * The AV holding the backreferences is pointed to either as the mg_obj of
5419 * PERL_MAGIC_backref, or in the specific case of a HV, from the
5420 * xhv_backreferences field. The array is created with a refcount
5421 * of 2. This means that if during global destruction the array gets
5422 * picked on before its parent to have its refcount decremented by the
5423 * random zapper, it won't actually be freed, meaning it's still there for
5424 * when its parent gets freed.
5426 * When the parent SV is freed, the extra ref is killed by
5427 * Perl_sv_kill_backrefs. The other ref is killed, in the case of magic,
5428 * by mg_free() / MGf_REFCOUNTED, or for a hash, by Perl_hv_kill_backrefs.
5430 * When a single backref SV is stored directly, it is not reference
5435 Perl_sv_add_backref(pTHX_ SV *const tsv, SV *const sv)
5442 PERL_ARGS_ASSERT_SV_ADD_BACKREF;
5444 /* find slot to store array or singleton backref */
5446 if (SvTYPE(tsv) == SVt_PVHV) {
5447 svp = (SV**)Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5450 (SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL))))
5452 sv_magic(tsv, NULL, PERL_MAGIC_backref, NULL, 0);
5453 mg = mg_find(tsv, PERL_MAGIC_backref);
5455 svp = &(mg->mg_obj);
5458 /* create or retrieve the array */
5460 if ( (!*svp && SvTYPE(sv) == SVt_PVAV)
5461 || (*svp && SvTYPE(*svp) != SVt_PVAV)
5466 SvREFCNT_inc_simple_void(av);
5467 /* av now has a refcnt of 2; see discussion above */
5469 /* move single existing backref to the array */
5471 AvARRAY(av)[++AvFILLp(av)] = *svp; /* av_push() */
5475 mg->mg_flags |= MGf_REFCOUNTED;
5478 av = MUTABLE_AV(*svp);
5481 /* optimisation: store single backref directly in HvAUX or mg_obj */
5485 /* push new backref */
5486 assert(SvTYPE(av) == SVt_PVAV);
5487 if (AvFILLp(av) >= AvMAX(av)) {
5488 av_extend(av, AvFILLp(av)+1);
5490 AvARRAY(av)[++AvFILLp(av)] = sv; /* av_push() */
5493 /* delete a back-reference to ourselves from the backref magic associated
5494 * with the SV we point to.
5498 Perl_sv_del_backref(pTHX_ SV *const tsv, SV *const sv)
5503 PERL_ARGS_ASSERT_SV_DEL_BACKREF;
5505 if (SvTYPE(tsv) == SVt_PVHV) {
5507 svp = (SV**)Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5511 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
5512 svp = mg ? &(mg->mg_obj) : NULL;
5516 Perl_croak(aTHX_ "panic: del_backref");
5518 if (SvTYPE(*svp) == SVt_PVAV) {
5522 AV * const av = (AV*)*svp;
5524 assert(!SvIS_FREED(av));
5528 /* for an SV with N weak references to it, if all those
5529 * weak refs are deleted, then sv_del_backref will be called
5530 * N times and O(N^2) compares will be done within the backref
5531 * array. To ameliorate this potential slowness, we:
5532 * 1) make sure this code is as tight as possible;
5533 * 2) when looking for SV, look for it at both the head and tail of the
5534 * array first before searching the rest, since some create/destroy
5535 * patterns will cause the backrefs to be freed in order.
5542 SV **p = &svp[fill];
5543 SV *const topsv = *p;
5550 /* We weren't the last entry.
5551 An unordered list has this property that you
5552 can take the last element off the end to fill
5553 the hole, and it's still an unordered list :-)
5559 break; /* should only be one */
5566 AvFILLp(av) = fill-1;
5569 /* optimisation: only a single backref, stored directly */
5571 Perl_croak(aTHX_ "panic: del_backref");
5578 Perl_sv_kill_backrefs(pTHX_ SV *const sv, AV *const av)
5584 PERL_ARGS_ASSERT_SV_KILL_BACKREFS;
5589 /* after multiple passes through Perl_sv_clean_all() for a thinngy
5590 * that has badly leaked, the backref array may have gotten freed,
5591 * since we only protect it against 1 round of cleanup */
5592 if (SvIS_FREED(av)) {
5593 if (PL_in_clean_all) /* All is fair */
5596 "panic: magic_killbackrefs (freed backref AV/SV)");
5600 is_array = (SvTYPE(av) == SVt_PVAV);
5602 assert(!SvIS_FREED(av));
5605 last = svp + AvFILLp(av);
5608 /* optimisation: only a single backref, stored directly */
5614 while (svp <= last) {
5616 SV *const referrer = *svp;
5617 if (SvWEAKREF(referrer)) {
5618 /* XXX Should we check that it hasn't changed? */
5619 assert(SvROK(referrer));
5620 SvRV_set(referrer, 0);
5622 SvWEAKREF_off(referrer);
5623 SvSETMAGIC(referrer);
5624 } else if (SvTYPE(referrer) == SVt_PVGV ||
5625 SvTYPE(referrer) == SVt_PVLV) {
5626 assert(SvTYPE(sv) == SVt_PVHV); /* stash backref */
5627 /* You lookin' at me? */
5628 assert(GvSTASH(referrer));
5629 assert(GvSTASH(referrer) == (const HV *)sv);
5630 GvSTASH(referrer) = 0;
5631 } else if (SvTYPE(referrer) == SVt_PVCV ||
5632 SvTYPE(referrer) == SVt_PVFM) {
5633 if (SvTYPE(sv) == SVt_PVHV) { /* stash backref */
5634 /* You lookin' at me? */
5635 assert(CvSTASH(referrer));
5636 assert(CvSTASH(referrer) == (const HV *)sv);
5637 SvANY(MUTABLE_CV(referrer))->xcv_stash = 0;
5640 assert(SvTYPE(sv) == SVt_PVGV);
5641 /* You lookin' at me? */
5642 assert(CvGV(referrer));
5643 assert(CvGV(referrer) == (const GV *)sv);
5644 anonymise_cv_maybe(MUTABLE_GV(sv),
5645 MUTABLE_CV(referrer));
5650 "panic: magic_killbackrefs (flags=%"UVxf")",
5651 (UV)SvFLAGS(referrer));
5662 SvREFCNT_dec(av); /* remove extra count added by sv_add_backref() */
5668 =for apidoc sv_insert
5670 Inserts a string at the specified offset/length within the SV. Similar to
5671 the Perl substr() function. Handles get magic.
5673 =for apidoc sv_insert_flags
5675 Same as C<sv_insert>, but the extra C<flags> are passed the C<SvPV_force_flags> that applies to C<bigstr>.
5681 Perl_sv_insert_flags(pTHX_ SV *const bigstr, const STRLEN offset, const STRLEN len, const char *const little, const STRLEN littlelen, const U32 flags)
5686 register char *midend;
5687 register char *bigend;
5691 PERL_ARGS_ASSERT_SV_INSERT_FLAGS;
5694 Perl_croak(aTHX_ "Can't modify non-existent substring");
5695 SvPV_force_flags(bigstr, curlen, flags);
5696 (void)SvPOK_only_UTF8(bigstr);
5697 if (offset + len > curlen) {
5698 SvGROW(bigstr, offset+len+1);
5699 Zero(SvPVX(bigstr)+curlen, offset+len-curlen, char);
5700 SvCUR_set(bigstr, offset+len);
5704 i = littlelen - len;
5705 if (i > 0) { /* string might grow */
5706 big = SvGROW(bigstr, SvCUR(bigstr) + i + 1);
5707 mid = big + offset + len;
5708 midend = bigend = big + SvCUR(bigstr);
5711 while (midend > mid) /* shove everything down */
5712 *--bigend = *--midend;
5713 Move(little,big+offset,littlelen,char);
5714 SvCUR_set(bigstr, SvCUR(bigstr) + i);
5719 Move(little,SvPVX(bigstr)+offset,len,char);
5724 big = SvPVX(bigstr);
5727 bigend = big + SvCUR(bigstr);
5729 if (midend > bigend)
5730 Perl_croak(aTHX_ "panic: sv_insert");
5732 if (mid - big > bigend - midend) { /* faster to shorten from end */
5734 Move(little, mid, littlelen,char);
5737 i = bigend - midend;
5739 Move(midend, mid, i,char);
5743 SvCUR_set(bigstr, mid - big);
5745 else if ((i = mid - big)) { /* faster from front */
5746 midend -= littlelen;
5748 Move(big, midend - i, i, char);
5749 sv_chop(bigstr,midend-i);
5751 Move(little, mid, littlelen,char);
5753 else if (littlelen) {
5754 midend -= littlelen;
5755 sv_chop(bigstr,midend);
5756 Move(little,midend,littlelen,char);
5759 sv_chop(bigstr,midend);
5765 =for apidoc sv_replace
5767 Make the first argument a copy of the second, then delete the original.
5768 The target SV physically takes over ownership of the body of the source SV
5769 and inherits its flags; however, the target keeps any magic it owns,
5770 and any magic in the source is discarded.
5771 Note that this is a rather specialist SV copying operation; most of the
5772 time you'll want to use C<sv_setsv> or one of its many macro front-ends.
5778 Perl_sv_replace(pTHX_ register SV *const sv, register SV *const nsv)
5781 const U32 refcnt = SvREFCNT(sv);
5783 PERL_ARGS_ASSERT_SV_REPLACE;
5785 SV_CHECK_THINKFIRST_COW_DROP(sv);
5786 if (SvREFCNT(nsv) != 1) {
5787 Perl_croak(aTHX_ "panic: reference miscount on nsv in sv_replace()"
5788 " (%" UVuf " != 1)", (UV) SvREFCNT(nsv));
5790 if (SvMAGICAL(sv)) {
5794 sv_upgrade(nsv, SVt_PVMG);
5795 SvMAGIC_set(nsv, SvMAGIC(sv));
5796 SvFLAGS(nsv) |= SvMAGICAL(sv);
5798 SvMAGIC_set(sv, NULL);
5802 assert(!SvREFCNT(sv));
5803 #ifdef DEBUG_LEAKING_SCALARS
5804 sv->sv_flags = nsv->sv_flags;
5805 sv->sv_any = nsv->sv_any;
5806 sv->sv_refcnt = nsv->sv_refcnt;
5807 sv->sv_u = nsv->sv_u;
5809 StructCopy(nsv,sv,SV);
5811 if(SvTYPE(sv) == SVt_IV) {
5813 = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
5817 #ifdef PERL_OLD_COPY_ON_WRITE
5818 if (SvIsCOW_normal(nsv)) {
5819 /* We need to follow the pointers around the loop to make the
5820 previous SV point to sv, rather than nsv. */
5823 while ((next = SV_COW_NEXT_SV(current)) != nsv) {
5826 assert(SvPVX_const(current) == SvPVX_const(nsv));
5828 /* Make the SV before us point to the SV after us. */
5830 PerlIO_printf(Perl_debug_log, "previous is\n");
5832 PerlIO_printf(Perl_debug_log,
5833 "move it from 0x%"UVxf" to 0x%"UVxf"\n",
5834 (UV) SV_COW_NEXT_SV(current), (UV) sv);
5836 SV_COW_NEXT_SV_SET(current, sv);
5839 SvREFCNT(sv) = refcnt;
5840 SvFLAGS(nsv) |= SVTYPEMASK; /* Mark as freed */
5845 /* We're about to free a GV which has a CV that refers back to us.
5846 * If that CV will outlive us, make it anonymous (i.e. fix up its CvGV
5850 S_anonymise_cv_maybe(pTHX_ GV *gv, CV* cv)
5856 PERL_ARGS_ASSERT_ANONYMISE_CV_MAYBE;
5859 assert(SvREFCNT(gv) == 0);
5860 assert(isGV(gv) && isGV_with_GP(gv));
5862 assert(!CvANON(cv));
5863 assert(CvGV(cv) == gv);
5865 /* will the CV shortly be freed by gp_free() ? */
5866 if (GvCV(gv) == cv && GvGP(gv)->gp_refcnt < 2 && SvREFCNT(cv) < 2) {
5867 SvANY(cv)->xcv_gv = NULL;
5871 /* if not, anonymise: */
5872 stash = GvSTASH(gv) && HvNAME(GvSTASH(gv))
5873 ? HvENAME(GvSTASH(gv)) : NULL;
5874 gvname = Perl_newSVpvf(aTHX_ "%s::__ANON__",
5875 stash ? stash : "__ANON__");
5876 anongv = gv_fetchsv(gvname, GV_ADDMULTI, SVt_PVCV);
5877 SvREFCNT_dec(gvname);
5881 SvANY(cv)->xcv_gv = MUTABLE_GV(SvREFCNT_inc(anongv));
5886 =for apidoc sv_clear
5888 Clear an SV: call any destructors, free up any memory used by the body,
5889 and free the body itself. The SV's head is I<not> freed, although
5890 its type is set to all 1's so that it won't inadvertently be assumed
5891 to be live during global destruction etc.
5892 This function should only be called when REFCNT is zero. Most of the time
5893 you'll want to call C<sv_free()> (or its macro wrapper C<SvREFCNT_dec>)
5900 Perl_sv_clear(pTHX_ SV *const orig_sv)
5905 const struct body_details *sv_type_details;
5908 register SV *sv = orig_sv;
5911 PERL_ARGS_ASSERT_SV_CLEAR;
5913 /* within this loop, sv is the SV currently being freed, and
5914 * iter_sv is the most recent AV or whatever that's being iterated
5915 * over to provide more SVs */
5921 assert(SvREFCNT(sv) == 0);
5922 assert(SvTYPE(sv) != (svtype)SVTYPEMASK);
5924 if (type <= SVt_IV) {
5925 /* See the comment in sv.h about the collusion between this
5926 * early return and the overloading of the NULL slots in the
5930 SvFLAGS(sv) &= SVf_BREAK;
5931 SvFLAGS(sv) |= SVTYPEMASK;
5935 assert(!SvOBJECT(sv) || type >= SVt_PVMG); /* objs are always >= MG */
5937 if (type >= SVt_PVMG) {
5939 if (!curse(sv, 1)) goto get_next_sv;
5940 type = SvTYPE(sv); /* destructor may have changed it */
5942 /* Free back-references before magic, in case the magic calls
5943 * Perl code that has weak references to sv. */
5944 if (type == SVt_PVHV) {
5945 Perl_hv_kill_backrefs(aTHX_ MUTABLE_HV(sv));
5949 else if (type == SVt_PVMG && SvPAD_OUR(sv)) {
5950 SvREFCNT_dec(SvOURSTASH(sv));
5951 } else if (SvMAGIC(sv)) {
5952 /* Free back-references before other types of magic. */
5953 sv_unmagic(sv, PERL_MAGIC_backref);
5956 if (type == SVt_PVMG && SvPAD_TYPED(sv))
5957 SvREFCNT_dec(SvSTASH(sv));
5960 /* case SVt_BIND: */
5963 IoIFP(sv) != PerlIO_stdin() &&
5964 IoIFP(sv) != PerlIO_stdout() &&
5965 IoIFP(sv) != PerlIO_stderr() &&
5966 !(IoFLAGS(sv) & IOf_FAKE_DIRP))
5968 io_close(MUTABLE_IO(sv), FALSE);
5970 if (IoDIRP(sv) && !(IoFLAGS(sv) & IOf_FAKE_DIRP))
5971 PerlDir_close(IoDIRP(sv));
5972 IoDIRP(sv) = (DIR*)NULL;
5973 Safefree(IoTOP_NAME(sv));
5974 Safefree(IoFMT_NAME(sv));
5975 Safefree(IoBOTTOM_NAME(sv));
5978 /* FIXME for plugins */
5979 pregfree2((REGEXP*) sv);
5983 cv_undef(MUTABLE_CV(sv));
5984 /* If we're in a stash, we don't own a reference to it.
5985 * However it does have a back reference to us, which needs to
5987 if ((stash = CvSTASH(sv)))
5988 sv_del_backref(MUTABLE_SV(stash), sv);
5991 if (PL_last_swash_hv == (const HV *)sv) {
5992 PL_last_swash_hv = NULL;
5994 if (HvTOTALKEYS((HV*)sv) > 0) {
5996 /* this statement should match the one at the beginning of
5997 * hv_undef_flags() */
5998 if ( PL_phase != PERL_PHASE_DESTRUCT
5999 && (name = HvNAME((HV*)sv)))
6002 (void)hv_delete(PL_stashcache, name,
6003 HvNAMELEN_get((HV*)sv), G_DISCARD);
6004 hv_name_set((HV*)sv, NULL, 0, 0);
6007 /* save old iter_sv in unused SvSTASH field */
6008 assert(!SvOBJECT(sv));
6009 SvSTASH(sv) = (HV*)iter_sv;
6012 /* XXX ideally we should save the old value of hash_index
6013 * too, but I can't think of any place to hide it. The
6014 * effect of not saving it is that for freeing hashes of
6015 * hashes, we become quadratic in scanning the HvARRAY of
6016 * the top hash looking for new entries to free; but
6017 * hopefully this will be dwarfed by the freeing of all
6018 * the nested hashes. */
6020 next_sv = Perl_hfree_next_entry(aTHX_ (HV*)sv, &hash_index);
6021 goto get_next_sv; /* process this new sv */
6023 /* free empty hash */
6024 Perl_hv_undef_flags(aTHX_ MUTABLE_HV(sv), HV_NAME_SETALL);
6025 assert(!HvARRAY((HV*)sv));
6029 AV* av = MUTABLE_AV(sv);
6030 if (PL_comppad == av) {
6034 if (AvREAL(av) && AvFILLp(av) > -1) {
6035 next_sv = AvARRAY(av)[AvFILLp(av)--];
6036 /* save old iter_sv in top-most slot of AV,
6037 * and pray that it doesn't get wiped in the meantime */
6038 AvARRAY(av)[AvMAX(av)] = iter_sv;
6040 goto get_next_sv; /* process this new sv */
6042 Safefree(AvALLOC(av));
6047 if (LvTYPE(sv) == 'T') { /* for tie: return HE to pool */
6048 SvREFCNT_dec(HeKEY_sv((HE*)LvTARG(sv)));
6049 HeNEXT((HE*)LvTARG(sv)) = PL_hv_fetch_ent_mh;
6050 PL_hv_fetch_ent_mh = (HE*)LvTARG(sv);
6052 else if (LvTYPE(sv) != 't') /* unless tie: unrefcnted fake SV** */
6053 SvREFCNT_dec(LvTARG(sv));
6055 if (isGV_with_GP(sv)) {
6056 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
6057 && HvENAME_get(stash))
6058 mro_method_changed_in(stash);
6059 gp_free(MUTABLE_GV(sv));
6061 unshare_hek(GvNAME_HEK(sv));
6062 /* If we're in a stash, we don't own a reference to it.
6063 * However it does have a back reference to us, which
6064 * needs to be cleared. */
6065 if (!SvVALID(sv) && (stash = GvSTASH(sv)))
6066 sv_del_backref(MUTABLE_SV(stash), sv);
6068 /* FIXME. There are probably more unreferenced pointers to SVs
6069 * in the interpreter struct that we should check and tidy in
6070 * a similar fashion to this: */
6071 if ((const GV *)sv == PL_last_in_gv)
6072 PL_last_in_gv = NULL;
6078 /* Don't bother with SvOOK_off(sv); as we're only going to
6082 SvOOK_offset(sv, offset);
6083 SvPV_set(sv, SvPVX_mutable(sv) - offset);
6084 /* Don't even bother with turning off the OOK flag. */
6089 SV * const target = SvRV(sv);
6091 sv_del_backref(target, sv);
6096 #ifdef PERL_OLD_COPY_ON_WRITE
6097 else if (SvPVX_const(sv)
6098 && !(SvTYPE(sv) == SVt_PVIO
6099 && !(IoFLAGS(sv) & IOf_FAKE_DIRP)))
6103 PerlIO_printf(Perl_debug_log, "Copy on write: clear\n");
6107 sv_release_COW(sv, SvPVX_const(sv), SV_COW_NEXT_SV(sv));
6109 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
6113 } else if (SvLEN(sv)) {
6114 Safefree(SvPVX_const(sv));
6118 else if (SvPVX_const(sv) && SvLEN(sv)
6119 && !(SvTYPE(sv) == SVt_PVIO
6120 && !(IoFLAGS(sv) & IOf_FAKE_DIRP)))
6121 Safefree(SvPVX_mutable(sv));
6122 else if (SvPVX_const(sv) && SvREADONLY(sv) && SvFAKE(sv)) {
6123 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
6134 SvFLAGS(sv) &= SVf_BREAK;
6135 SvFLAGS(sv) |= SVTYPEMASK;
6137 sv_type_details = bodies_by_type + type;
6138 if (sv_type_details->arena) {
6139 del_body(((char *)SvANY(sv) + sv_type_details->offset),
6140 &PL_body_roots[type]);
6142 else if (sv_type_details->body_size) {
6143 safefree(SvANY(sv));
6147 /* caller is responsible for freeing the head of the original sv */
6148 if (sv != orig_sv && !SvREFCNT(sv))
6151 /* grab and free next sv, if any */
6159 else if (!iter_sv) {
6161 } else if (SvTYPE(iter_sv) == SVt_PVAV) {
6162 AV *const av = (AV*)iter_sv;
6163 if (AvFILLp(av) > -1) {
6164 sv = AvARRAY(av)[AvFILLp(av)--];
6166 else { /* no more elements of current AV to free */
6169 /* restore previous value, squirrelled away */
6170 iter_sv = AvARRAY(av)[AvMAX(av)];
6171 Safefree(AvALLOC(av));
6174 } else if (SvTYPE(iter_sv) == SVt_PVHV) {
6175 sv = Perl_hfree_next_entry(aTHX_ (HV*)iter_sv, &hash_index);
6176 if (!sv && !HvTOTALKEYS((HV *)iter_sv)) {
6177 /* no more elements of current HV to free */
6180 /* Restore previous value of iter_sv, squirrelled away */
6181 assert(!SvOBJECT(sv));
6182 iter_sv = (SV*)SvSTASH(sv);
6184 /* ideally we should restore the old hash_index here,
6185 * but we don't currently save the old value */
6188 /* free any remaining detritus from the hash struct */
6189 Perl_hv_undef_flags(aTHX_ MUTABLE_HV(sv), HV_NAME_SETALL);
6190 assert(!HvARRAY((HV*)sv));
6195 /* unrolled SvREFCNT_dec and sv_free2 follows: */
6199 if (!SvREFCNT(sv)) {
6203 if (--(SvREFCNT(sv)))
6207 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEBUGGING),
6208 "Attempt to free temp prematurely: SV 0x%"UVxf
6209 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
6213 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
6214 /* make sure SvREFCNT(sv)==0 happens very seldom */
6215 SvREFCNT(sv) = (~(U32)0)/2;
6224 /* This routine curses the sv itself, not the object referenced by sv. So
6225 sv does not have to be ROK. */
6228 S_curse(pTHX_ SV * const sv, const bool check_refcnt) {
6231 PERL_ARGS_ASSERT_CURSE;
6232 assert(SvOBJECT(sv));
6234 if (PL_defstash && /* Still have a symbol table? */
6241 stash = SvSTASH(sv);
6242 destructor = StashHANDLER(stash,DESTROY);
6244 /* A constant subroutine can have no side effects, so
6245 don't bother calling it. */
6246 && !CvCONST(destructor)
6247 /* Don't bother calling an empty destructor */
6248 && (CvISXSUB(destructor)
6249 || (CvSTART(destructor)
6250 && (CvSTART(destructor)->op_next->op_type
6253 SV* const tmpref = newRV(sv);
6254 SvREADONLY_on(tmpref); /* DESTROY() could be naughty */
6256 PUSHSTACKi(PERLSI_DESTROY);
6261 call_sv(MUTABLE_SV(destructor),
6262 G_DISCARD|G_EVAL|G_KEEPERR|G_VOID);
6266 if(SvREFCNT(tmpref) < 2) {
6267 /* tmpref is not kept alive! */
6269 SvRV_set(tmpref, NULL);
6272 SvREFCNT_dec(tmpref);
6274 } while (SvOBJECT(sv) && SvSTASH(sv) != stash);
6277 if (check_refcnt && SvREFCNT(sv)) {
6278 if (PL_in_clean_objs)
6280 "DESTROY created new reference to dead object '%s'",
6282 /* DESTROY gave object new lease on life */
6288 SvREFCNT_dec(SvSTASH(sv)); /* possibly of changed persuasion */
6289 SvOBJECT_off(sv); /* Curse the object. */
6290 if (SvTYPE(sv) != SVt_PVIO)
6291 --PL_sv_objcount;/* XXX Might want something more general */
6297 =for apidoc sv_newref
6299 Increment an SV's reference count. Use the C<SvREFCNT_inc()> wrapper
6306 Perl_sv_newref(pTHX_ SV *const sv)
6308 PERL_UNUSED_CONTEXT;
6317 Decrement an SV's reference count, and if it drops to zero, call
6318 C<sv_clear> to invoke destructors and free up any memory used by
6319 the body; finally, deallocate the SV's head itself.
6320 Normally called via a wrapper macro C<SvREFCNT_dec>.
6326 Perl_sv_free(pTHX_ SV *const sv)
6331 if (SvREFCNT(sv) == 0) {
6332 if (SvFLAGS(sv) & SVf_BREAK)
6333 /* this SV's refcnt has been artificially decremented to
6334 * trigger cleanup */
6336 if (PL_in_clean_all) /* All is fair */
6338 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
6339 /* make sure SvREFCNT(sv)==0 happens very seldom */
6340 SvREFCNT(sv) = (~(U32)0)/2;
6343 if (ckWARN_d(WARN_INTERNAL)) {
6344 #ifdef DEBUG_LEAKING_SCALARS_FORK_DUMP
6345 Perl_dump_sv_child(aTHX_ sv);
6347 #ifdef DEBUG_LEAKING_SCALARS
6350 #ifdef DEBUG_LEAKING_SCALARS_ABORT
6351 if (PL_warnhook == PERL_WARNHOOK_FATAL
6352 || ckDEAD(packWARN(WARN_INTERNAL))) {
6353 /* Don't let Perl_warner cause us to escape our fate: */
6357 /* This may not return: */
6358 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
6359 "Attempt to free unreferenced scalar: SV 0x%"UVxf
6360 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
6363 #ifdef DEBUG_LEAKING_SCALARS_ABORT
6368 if (--(SvREFCNT(sv)) > 0)
6370 Perl_sv_free2(aTHX_ sv);
6374 Perl_sv_free2(pTHX_ SV *const sv)
6378 PERL_ARGS_ASSERT_SV_FREE2;
6382 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEBUGGING),
6383 "Attempt to free temp prematurely: SV 0x%"UVxf
6384 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
6388 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
6389 /* make sure SvREFCNT(sv)==0 happens very seldom */
6390 SvREFCNT(sv) = (~(U32)0)/2;
6401 Returns the length of the string in the SV. Handles magic and type
6402 coercion. See also C<SvCUR>, which gives raw access to the xpv_cur slot.
6408 Perl_sv_len(pTHX_ register SV *const sv)
6416 len = mg_length(sv);
6418 (void)SvPV_const(sv, len);
6423 =for apidoc sv_len_utf8
6425 Returns the number of characters in the string in an SV, counting wide
6426 UTF-8 bytes as a single character. Handles magic and type coercion.
6432 * The length is cached in PERL_MAGIC_utf8, in the mg_len field. Also the
6433 * mg_ptr is used, by sv_pos_u2b() and sv_pos_b2u() - see the comments below.
6434 * (Note that the mg_len is not the length of the mg_ptr field.
6435 * This allows the cache to store the character length of the string without
6436 * needing to malloc() extra storage to attach to the mg_ptr.)
6441 Perl_sv_len_utf8(pTHX_ register SV *const sv)
6447 return mg_length(sv);
6451 const U8 *s = (U8*)SvPV_const(sv, len);
6455 MAGIC *mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_utf8) : NULL;
6457 if (mg && (mg->mg_len != -1 || mg->mg_ptr)) {
6458 if (mg->mg_len != -1)
6461 /* We can use the offset cache for a headstart.
6462 The longer value is stored in the first pair. */
6463 STRLEN *cache = (STRLEN *) mg->mg_ptr;
6465 ulen = cache[0] + Perl_utf8_length(aTHX_ s + cache[1],
6469 if (PL_utf8cache < 0) {
6470 const STRLEN real = Perl_utf8_length(aTHX_ s, s + len);
6471 assert_uft8_cache_coherent("sv_len_utf8", ulen, real, sv);
6475 ulen = Perl_utf8_length(aTHX_ s, s + len);
6476 utf8_mg_len_cache_update(sv, &mg, ulen);
6480 return Perl_utf8_length(aTHX_ s, s + len);
6484 /* Walk forwards to find the byte corresponding to the passed in UTF-8
6487 S_sv_pos_u2b_forwards(const U8 *const start, const U8 *const send,
6488 STRLEN *const uoffset_p, bool *const at_end)
6490 const U8 *s = start;
6491 STRLEN uoffset = *uoffset_p;
6493 PERL_ARGS_ASSERT_SV_POS_U2B_FORWARDS;
6495 while (s < send && uoffset) {
6502 else if (s > send) {
6504 /* This is the existing behaviour. Possibly it should be a croak, as
6505 it's actually a bounds error */
6508 *uoffset_p -= uoffset;
6512 /* Given the length of the string in both bytes and UTF-8 characters, decide
6513 whether to walk forwards or backwards to find the byte corresponding to
6514 the passed in UTF-8 offset. */
6516 S_sv_pos_u2b_midway(const U8 *const start, const U8 *send,
6517 STRLEN uoffset, const STRLEN uend)
6519 STRLEN backw = uend - uoffset;
6521 PERL_ARGS_ASSERT_SV_POS_U2B_MIDWAY;
6523 if (uoffset < 2 * backw) {
6524 /* The assumption is that going forwards is twice the speed of going
6525 forward (that's where the 2 * backw comes from).
6526 (The real figure of course depends on the UTF-8 data.) */
6527 const U8 *s = start;
6529 while (s < send && uoffset--)
6539 while (UTF8_IS_CONTINUATION(*send))
6542 return send - start;
6545 /* For the string representation of the given scalar, find the byte
6546 corresponding to the passed in UTF-8 offset. uoffset0 and boffset0
6547 give another position in the string, *before* the sought offset, which
6548 (which is always true, as 0, 0 is a valid pair of positions), which should
6549 help reduce the amount of linear searching.
6550 If *mgp is non-NULL, it should point to the UTF-8 cache magic, which
6551 will be used to reduce the amount of linear searching. The cache will be
6552 created if necessary, and the found value offered to it for update. */
6554 S_sv_pos_u2b_cached(pTHX_ SV *const sv, MAGIC **const mgp, const U8 *const start,
6555 const U8 *const send, STRLEN uoffset,
6556 STRLEN uoffset0, STRLEN boffset0)
6558 STRLEN boffset = 0; /* Actually always set, but let's keep gcc happy. */
6560 bool at_end = FALSE;
6562 PERL_ARGS_ASSERT_SV_POS_U2B_CACHED;
6564 assert (uoffset >= uoffset0);
6571 && (*mgp || (SvTYPE(sv) >= SVt_PVMG &&
6572 (*mgp = mg_find(sv, PERL_MAGIC_utf8))))) {
6573 if ((*mgp)->mg_ptr) {
6574 STRLEN *cache = (STRLEN *) (*mgp)->mg_ptr;
6575 if (cache[0] == uoffset) {
6576 /* An exact match. */
6579 if (cache[2] == uoffset) {
6580 /* An exact match. */
6584 if (cache[0] < uoffset) {
6585 /* The cache already knows part of the way. */
6586 if (cache[0] > uoffset0) {
6587 /* The cache knows more than the passed in pair */
6588 uoffset0 = cache[0];
6589 boffset0 = cache[1];
6591 if ((*mgp)->mg_len != -1) {
6592 /* And we know the end too. */
6594 + sv_pos_u2b_midway(start + boffset0, send,
6596 (*mgp)->mg_len - uoffset0);
6598 uoffset -= uoffset0;
6600 + sv_pos_u2b_forwards(start + boffset0,
6601 send, &uoffset, &at_end);
6602 uoffset += uoffset0;
6605 else if (cache[2] < uoffset) {
6606 /* We're between the two cache entries. */
6607 if (cache[2] > uoffset0) {
6608 /* and the cache knows more than the passed in pair */
6609 uoffset0 = cache[2];
6610 boffset0 = cache[3];
6614 + sv_pos_u2b_midway(start + boffset0,
6617 cache[0] - uoffset0);
6620 + sv_pos_u2b_midway(start + boffset0,
6623 cache[2] - uoffset0);
6627 else if ((*mgp)->mg_len != -1) {
6628 /* If we can take advantage of a passed in offset, do so. */
6629 /* In fact, offset0 is either 0, or less than offset, so don't
6630 need to worry about the other possibility. */
6632 + sv_pos_u2b_midway(start + boffset0, send,
6634 (*mgp)->mg_len - uoffset0);
6639 if (!found || PL_utf8cache < 0) {
6640 STRLEN real_boffset;
6641 uoffset -= uoffset0;
6642 real_boffset = boffset0 + sv_pos_u2b_forwards(start + boffset0,
6643 send, &uoffset, &at_end);
6644 uoffset += uoffset0;
6646 if (found && PL_utf8cache < 0)
6647 assert_uft8_cache_coherent("sv_pos_u2b_cache", boffset,
6649 boffset = real_boffset;
6654 utf8_mg_len_cache_update(sv, mgp, uoffset);
6656 utf8_mg_pos_cache_update(sv, mgp, boffset, uoffset, send - start);
6663 =for apidoc sv_pos_u2b_flags
6665 Converts the value pointed to by offsetp from a count of UTF-8 chars from
6666 the start of the string, to a count of the equivalent number of bytes; if
6667 lenp is non-zero, it does the same to lenp, but this time starting from
6668 the offset, rather than from the start of the string. Handles type coercion.
6669 I<flags> is passed to C<SvPV_flags>, and usually should be
6670 C<SV_GMAGIC|SV_CONST_RETURN> to handle magic.
6676 * sv_pos_u2b_flags() uses, like sv_pos_b2u(), the mg_ptr of the potential
6677 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6678 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
6683 Perl_sv_pos_u2b_flags(pTHX_ SV *const sv, STRLEN uoffset, STRLEN *const lenp,
6690 PERL_ARGS_ASSERT_SV_POS_U2B_FLAGS;
6692 start = (U8*)SvPV_flags(sv, len, flags);
6694 const U8 * const send = start + len;
6696 boffset = sv_pos_u2b_cached(sv, &mg, start, send, uoffset, 0, 0);
6699 && *lenp /* don't bother doing work for 0, as its bytes equivalent
6700 is 0, and *lenp is already set to that. */) {
6701 /* Convert the relative offset to absolute. */
6702 const STRLEN uoffset2 = uoffset + *lenp;
6703 const STRLEN boffset2
6704 = sv_pos_u2b_cached(sv, &mg, start, send, uoffset2,
6705 uoffset, boffset) - boffset;
6719 =for apidoc sv_pos_u2b
6721 Converts the value pointed to by offsetp from a count of UTF-8 chars from
6722 the start of the string, to a count of the equivalent number of bytes; if
6723 lenp is non-zero, it does the same to lenp, but this time starting from
6724 the offset, rather than from the start of the string. Handles magic and
6727 Use C<sv_pos_u2b_flags> in preference, which correctly handles strings longer
6734 * sv_pos_u2b() uses, like sv_pos_b2u(), the mg_ptr of the potential
6735 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6736 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
6740 /* This function is subject to size and sign problems */
6743 Perl_sv_pos_u2b(pTHX_ register SV *const sv, I32 *const offsetp, I32 *const lenp)
6745 PERL_ARGS_ASSERT_SV_POS_U2B;
6748 STRLEN ulen = (STRLEN)*lenp;
6749 *offsetp = (I32)sv_pos_u2b_flags(sv, (STRLEN)*offsetp, &ulen,
6750 SV_GMAGIC|SV_CONST_RETURN);
6753 *offsetp = (I32)sv_pos_u2b_flags(sv, (STRLEN)*offsetp, NULL,
6754 SV_GMAGIC|SV_CONST_RETURN);
6759 S_utf8_mg_len_cache_update(pTHX_ SV *const sv, MAGIC **const mgp,
6762 PERL_ARGS_ASSERT_UTF8_MG_LEN_CACHE_UPDATE;
6766 if (!*mgp && (SvTYPE(sv) < SVt_PVMG ||
6767 !(*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
6768 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, &PL_vtbl_utf8, 0, 0);
6772 (*mgp)->mg_len = ulen;
6773 /* For now, treat "overflowed" as "still unknown". See RT #72924. */
6774 if (ulen != (STRLEN) (*mgp)->mg_len)
6775 (*mgp)->mg_len = -1;
6778 /* Create and update the UTF8 magic offset cache, with the proffered utf8/
6779 byte length pairing. The (byte) length of the total SV is passed in too,
6780 as blen, because for some (more esoteric) SVs, the call to SvPV_const()
6781 may not have updated SvCUR, so we can't rely on reading it directly.
6783 The proffered utf8/byte length pairing isn't used if the cache already has
6784 two pairs, and swapping either for the proffered pair would increase the
6785 RMS of the intervals between known byte offsets.
6787 The cache itself consists of 4 STRLEN values
6788 0: larger UTF-8 offset
6789 1: corresponding byte offset
6790 2: smaller UTF-8 offset
6791 3: corresponding byte offset
6793 Unused cache pairs have the value 0, 0.
6794 Keeping the cache "backwards" means that the invariant of
6795 cache[0] >= cache[2] is maintained even with empty slots, which means that
6796 the code that uses it doesn't need to worry if only 1 entry has actually
6797 been set to non-zero. It also makes the "position beyond the end of the
6798 cache" logic much simpler, as the first slot is always the one to start
6802 S_utf8_mg_pos_cache_update(pTHX_ SV *const sv, MAGIC **const mgp, const STRLEN byte,
6803 const STRLEN utf8, const STRLEN blen)
6807 PERL_ARGS_ASSERT_UTF8_MG_POS_CACHE_UPDATE;
6812 if (!*mgp && (SvTYPE(sv) < SVt_PVMG ||
6813 !(*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
6814 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, (MGVTBL*)&PL_vtbl_utf8, 0,
6816 (*mgp)->mg_len = -1;
6820 if (!(cache = (STRLEN *)(*mgp)->mg_ptr)) {
6821 Newxz(cache, PERL_MAGIC_UTF8_CACHESIZE * 2, STRLEN);
6822 (*mgp)->mg_ptr = (char *) cache;
6826 if (PL_utf8cache < 0 && SvPOKp(sv)) {
6827 /* SvPOKp() because it's possible that sv has string overloading, and
6828 therefore is a reference, hence SvPVX() is actually a pointer.
6829 This cures the (very real) symptoms of RT 69422, but I'm not actually
6830 sure whether we should even be caching the results of UTF-8
6831 operations on overloading, given that nothing stops overloading
6832 returning a different value every time it's called. */
6833 const U8 *start = (const U8 *) SvPVX_const(sv);
6834 const STRLEN realutf8 = utf8_length(start, start + byte);
6836 assert_uft8_cache_coherent("utf8_mg_pos_cache_update", utf8, realutf8,
6840 /* Cache is held with the later position first, to simplify the code
6841 that deals with unbounded ends. */
6843 ASSERT_UTF8_CACHE(cache);
6844 if (cache[1] == 0) {
6845 /* Cache is totally empty */
6848 } else if (cache[3] == 0) {
6849 if (byte > cache[1]) {
6850 /* New one is larger, so goes first. */
6851 cache[2] = cache[0];
6852 cache[3] = cache[1];
6860 #define THREEWAY_SQUARE(a,b,c,d) \
6861 ((float)((d) - (c))) * ((float)((d) - (c))) \
6862 + ((float)((c) - (b))) * ((float)((c) - (b))) \
6863 + ((float)((b) - (a))) * ((float)((b) - (a)))
6865 /* Cache has 2 slots in use, and we know three potential pairs.
6866 Keep the two that give the lowest RMS distance. Do the
6867 calculation in bytes simply because we always know the byte
6868 length. squareroot has the same ordering as the positive value,
6869 so don't bother with the actual square root. */
6870 const float existing = THREEWAY_SQUARE(0, cache[3], cache[1], blen);
6871 if (byte > cache[1]) {
6872 /* New position is after the existing pair of pairs. */
6873 const float keep_earlier
6874 = THREEWAY_SQUARE(0, cache[3], byte, blen);
6875 const float keep_later
6876 = THREEWAY_SQUARE(0, cache[1], byte, blen);
6878 if (keep_later < keep_earlier) {
6879 if (keep_later < existing) {
6880 cache[2] = cache[0];
6881 cache[3] = cache[1];
6887 if (keep_earlier < existing) {
6893 else if (byte > cache[3]) {
6894 /* New position is between the existing pair of pairs. */
6895 const float keep_earlier
6896 = THREEWAY_SQUARE(0, cache[3], byte, blen);
6897 const float keep_later
6898 = THREEWAY_SQUARE(0, byte, cache[1], blen);
6900 if (keep_later < keep_earlier) {
6901 if (keep_later < existing) {
6907 if (keep_earlier < existing) {
6914 /* New position is before the existing pair of pairs. */
6915 const float keep_earlier
6916 = THREEWAY_SQUARE(0, byte, cache[3], blen);
6917 const float keep_later
6918 = THREEWAY_SQUARE(0, byte, cache[1], blen);
6920 if (keep_later < keep_earlier) {
6921 if (keep_later < existing) {
6927 if (keep_earlier < existing) {
6928 cache[0] = cache[2];
6929 cache[1] = cache[3];
6936 ASSERT_UTF8_CACHE(cache);
6939 /* We already know all of the way, now we may be able to walk back. The same
6940 assumption is made as in S_sv_pos_u2b_midway(), namely that walking
6941 backward is half the speed of walking forward. */
6943 S_sv_pos_b2u_midway(pTHX_ const U8 *const s, const U8 *const target,
6944 const U8 *end, STRLEN endu)
6946 const STRLEN forw = target - s;
6947 STRLEN backw = end - target;
6949 PERL_ARGS_ASSERT_SV_POS_B2U_MIDWAY;
6951 if (forw < 2 * backw) {
6952 return utf8_length(s, target);
6955 while (end > target) {
6957 while (UTF8_IS_CONTINUATION(*end)) {
6966 =for apidoc sv_pos_b2u
6968 Converts the value pointed to by offsetp from a count of bytes from the
6969 start of the string, to a count of the equivalent number of UTF-8 chars.
6970 Handles magic and type coercion.
6976 * sv_pos_b2u() uses, like sv_pos_u2b(), the mg_ptr of the potential
6977 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6982 Perl_sv_pos_b2u(pTHX_ register SV *const sv, I32 *const offsetp)
6985 const STRLEN byte = *offsetp;
6986 STRLEN len = 0; /* Actually always set, but let's keep gcc happy. */
6992 PERL_ARGS_ASSERT_SV_POS_B2U;
6997 s = (const U8*)SvPV_const(sv, blen);
7000 Perl_croak(aTHX_ "panic: sv_pos_b2u: bad byte offset");
7006 && SvTYPE(sv) >= SVt_PVMG
7007 && (mg = mg_find(sv, PERL_MAGIC_utf8)))
7010 STRLEN * const cache = (STRLEN *) mg->mg_ptr;
7011 if (cache[1] == byte) {
7012 /* An exact match. */
7013 *offsetp = cache[0];
7016 if (cache[3] == byte) {
7017 /* An exact match. */
7018 *offsetp = cache[2];
7022 if (cache[1] < byte) {
7023 /* We already know part of the way. */
7024 if (mg->mg_len != -1) {
7025 /* Actually, we know the end too. */
7027 + S_sv_pos_b2u_midway(aTHX_ s + cache[1], send,
7028 s + blen, mg->mg_len - cache[0]);
7030 len = cache[0] + utf8_length(s + cache[1], send);
7033 else if (cache[3] < byte) {
7034 /* We're between the two cached pairs, so we do the calculation
7035 offset by the byte/utf-8 positions for the earlier pair,
7036 then add the utf-8 characters from the string start to
7038 len = S_sv_pos_b2u_midway(aTHX_ s + cache[3], send,
7039 s + cache[1], cache[0] - cache[2])
7043 else { /* cache[3] > byte */
7044 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + cache[3],
7048 ASSERT_UTF8_CACHE(cache);
7050 } else if (mg->mg_len != -1) {
7051 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + blen, mg->mg_len);
7055 if (!found || PL_utf8cache < 0) {
7056 const STRLEN real_len = utf8_length(s, send);
7058 if (found && PL_utf8cache < 0)
7059 assert_uft8_cache_coherent("sv_pos_b2u", len, real_len, sv);
7066 utf8_mg_len_cache_update(sv, &mg, len);
7068 utf8_mg_pos_cache_update(sv, &mg, byte, len, blen);
7073 S_assert_uft8_cache_coherent(pTHX_ const char *const func, STRLEN from_cache,
7074 STRLEN real, SV *const sv)
7076 PERL_ARGS_ASSERT_ASSERT_UFT8_CACHE_COHERENT;
7078 /* As this is debugging only code, save space by keeping this test here,
7079 rather than inlining it in all the callers. */
7080 if (from_cache == real)
7083 /* Need to turn the assertions off otherwise we may recurse infinitely
7084 while printing error messages. */
7085 SAVEI8(PL_utf8cache);
7087 Perl_croak(aTHX_ "panic: %s cache %"UVuf" real %"UVuf" for %"SVf,
7088 func, (UV) from_cache, (UV) real, SVfARG(sv));
7094 Returns a boolean indicating whether the strings in the two SVs are
7095 identical. Is UTF-8 and 'use bytes' aware, handles get magic, and will
7096 coerce its args to strings if necessary.
7098 =for apidoc sv_eq_flags
7100 Returns a boolean indicating whether the strings in the two SVs are
7101 identical. Is UTF-8 and 'use bytes' aware and coerces its args to strings
7102 if necessary. If the flags include SV_GMAGIC, it handles get-magic, too.
7108 Perl_sv_eq_flags(pTHX_ register SV *sv1, register SV *sv2, const U32 flags)
7117 SV* svrecode = NULL;
7124 /* if pv1 and pv2 are the same, second SvPV_const call may
7125 * invalidate pv1 (if we are handling magic), so we may need to
7127 if (sv1 == sv2 && flags & SV_GMAGIC
7128 && (SvTHINKFIRST(sv1) || SvGMAGICAL(sv1))) {
7129 pv1 = SvPV_const(sv1, cur1);
7130 sv1 = newSVpvn_flags(pv1, cur1, SVs_TEMP | SvUTF8(sv2));
7132 pv1 = SvPV_flags_const(sv1, cur1, flags);
7140 pv2 = SvPV_flags_const(sv2, cur2, flags);
7142 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
7143 /* Differing utf8ness.
7144 * Do not UTF8size the comparands as a side-effect. */
7147 svrecode = newSVpvn(pv2, cur2);
7148 sv_recode_to_utf8(svrecode, PL_encoding);
7149 pv2 = SvPV_const(svrecode, cur2);
7152 svrecode = newSVpvn(pv1, cur1);
7153 sv_recode_to_utf8(svrecode, PL_encoding);
7154 pv1 = SvPV_const(svrecode, cur1);
7156 /* Now both are in UTF-8. */
7158 SvREFCNT_dec(svrecode);
7164 /* sv1 is the UTF-8 one */
7165 return bytes_cmp_utf8((const U8*)pv2, cur2,
7166 (const U8*)pv1, cur1) == 0;
7169 /* sv2 is the UTF-8 one */
7170 return bytes_cmp_utf8((const U8*)pv1, cur1,
7171 (const U8*)pv2, cur2) == 0;
7177 eq = (pv1 == pv2) || memEQ(pv1, pv2, cur1);
7179 SvREFCNT_dec(svrecode);
7189 Compares the strings in two SVs. Returns -1, 0, or 1 indicating whether the
7190 string in C<sv1> is less than, equal to, or greater than the string in
7191 C<sv2>. Is UTF-8 and 'use bytes' aware, handles get magic, and will
7192 coerce its args to strings if necessary. See also C<sv_cmp_locale>.
7194 =for apidoc sv_cmp_flags
7196 Compares the strings in two SVs. Returns -1, 0, or 1 indicating whether the
7197 string in C<sv1> is less than, equal to, or greater than the string in
7198 C<sv2>. Is UTF-8 and 'use bytes' aware and will coerce its args to strings
7199 if necessary. If the flags include SV_GMAGIC, it handles get magic. See
7200 also C<sv_cmp_locale_flags>.
7206 Perl_sv_cmp(pTHX_ register SV *const sv1, register SV *const sv2)
7208 return sv_cmp_flags(sv1, sv2, SV_GMAGIC);
7212 Perl_sv_cmp_flags(pTHX_ register SV *const sv1, register SV *const sv2,
7217 const char *pv1, *pv2;
7220 SV *svrecode = NULL;
7227 pv1 = SvPV_flags_const(sv1, cur1, flags);
7234 pv2 = SvPV_flags_const(sv2, cur2, flags);
7236 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
7237 /* Differing utf8ness.
7238 * Do not UTF8size the comparands as a side-effect. */
7241 svrecode = newSVpvn(pv2, cur2);
7242 sv_recode_to_utf8(svrecode, PL_encoding);
7243 pv2 = SvPV_const(svrecode, cur2);
7246 const int retval = -bytes_cmp_utf8((const U8*)pv2, cur2,
7247 (const U8*)pv1, cur1);
7248 return retval ? retval < 0 ? -1 : +1 : 0;
7253 svrecode = newSVpvn(pv1, cur1);
7254 sv_recode_to_utf8(svrecode, PL_encoding);
7255 pv1 = SvPV_const(svrecode, cur1);
7258 const int retval = bytes_cmp_utf8((const U8*)pv1, cur1,
7259 (const U8*)pv2, cur2);
7260 return retval ? retval < 0 ? -1 : +1 : 0;
7266 cmp = cur2 ? -1 : 0;
7270 const I32 retval = memcmp((const void*)pv1, (const void*)pv2, cur1 < cur2 ? cur1 : cur2);
7273 cmp = retval < 0 ? -1 : 1;
7274 } else if (cur1 == cur2) {
7277 cmp = cur1 < cur2 ? -1 : 1;
7281 SvREFCNT_dec(svrecode);
7289 =for apidoc sv_cmp_locale
7291 Compares the strings in two SVs in a locale-aware manner. Is UTF-8 and
7292 'use bytes' aware, handles get magic, and will coerce its args to strings
7293 if necessary. See also C<sv_cmp>.
7295 =for apidoc sv_cmp_locale_flags
7297 Compares the strings in two SVs in a locale-aware manner. Is UTF-8 and
7298 'use bytes' aware and will coerce its args to strings if necessary. If the
7299 flags contain SV_GMAGIC, it handles get magic. See also C<sv_cmp_flags>.
7305 Perl_sv_cmp_locale(pTHX_ register SV *const sv1, register SV *const sv2)
7307 return sv_cmp_locale_flags(sv1, sv2, SV_GMAGIC);
7311 Perl_sv_cmp_locale_flags(pTHX_ register SV *const sv1, register SV *const sv2,
7315 #ifdef USE_LOCALE_COLLATE
7321 if (PL_collation_standard)
7325 pv1 = sv1 ? sv_collxfrm_flags(sv1, &len1, flags) : (char *) NULL;
7327 pv2 = sv2 ? sv_collxfrm_flags(sv2, &len2, flags) : (char *) NULL;
7329 if (!pv1 || !len1) {
7340 retval = memcmp((void*)pv1, (void*)pv2, len1 < len2 ? len1 : len2);
7343 return retval < 0 ? -1 : 1;
7346 * When the result of collation is equality, that doesn't mean
7347 * that there are no differences -- some locales exclude some
7348 * characters from consideration. So to avoid false equalities,
7349 * we use the raw string as a tiebreaker.
7355 #endif /* USE_LOCALE_COLLATE */
7357 return sv_cmp(sv1, sv2);
7361 #ifdef USE_LOCALE_COLLATE
7364 =for apidoc sv_collxfrm
7366 This calls C<sv_collxfrm_flags> with the SV_GMAGIC flag. See
7367 C<sv_collxfrm_flags>.
7369 =for apidoc sv_collxfrm_flags
7371 Add Collate Transform magic to an SV if it doesn't already have it. If the
7372 flags contain SV_GMAGIC, it handles get-magic.
7374 Any scalar variable may carry PERL_MAGIC_collxfrm magic that contains the
7375 scalar data of the variable, but transformed to such a format that a normal
7376 memory comparison can be used to compare the data according to the locale
7383 Perl_sv_collxfrm_flags(pTHX_ SV *const sv, STRLEN *const nxp, const I32 flags)
7388 PERL_ARGS_ASSERT_SV_COLLXFRM_FLAGS;
7390 mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_collxfrm) : (MAGIC *) NULL;
7391 if (!mg || !mg->mg_ptr || *(U32*)mg->mg_ptr != PL_collation_ix) {
7397 Safefree(mg->mg_ptr);
7398 s = SvPV_flags_const(sv, len, flags);
7399 if ((xf = mem_collxfrm(s, len, &xlen))) {
7401 #ifdef PERL_OLD_COPY_ON_WRITE
7403 sv_force_normal_flags(sv, 0);
7405 mg = sv_magicext(sv, 0, PERL_MAGIC_collxfrm, &PL_vtbl_collxfrm,
7419 if (mg && mg->mg_ptr) {
7421 return mg->mg_ptr + sizeof(PL_collation_ix);
7429 #endif /* USE_LOCALE_COLLATE */
7432 S_sv_gets_append_to_utf8(pTHX_ SV *const sv, PerlIO *const fp, I32 append)
7434 SV * const tsv = newSV(0);
7437 sv_gets(tsv, fp, 0);
7438 sv_utf8_upgrade_nomg(tsv);
7439 SvCUR_set(sv,append);
7442 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7446 S_sv_gets_read_record(pTHX_ SV *const sv, PerlIO *const fp, I32 append)
7449 const U32 recsize = SvUV(SvRV(PL_rs)); /* RsRECORD() guarantees > 0. */
7450 /* Grab the size of the record we're getting */
7451 char *const buffer = SvGROW(sv, (STRLEN)(recsize + append + 1)) + append;
7458 /* VMS wants read instead of fread, because fread doesn't respect */
7459 /* RMS record boundaries. This is not necessarily a good thing to be */
7460 /* doing, but we've got no other real choice - except avoid stdio
7461 as implementation - perhaps write a :vms layer ?
7463 fd = PerlIO_fileno(fp);
7465 bytesread = PerlLIO_read(fd, buffer, recsize);
7467 else /* in-memory file from PerlIO::Scalar */
7470 bytesread = PerlIO_read(fp, buffer, recsize);
7475 SvCUR_set(sv, bytesread + append);
7476 buffer[bytesread] = '\0';
7477 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7483 Get a line from the filehandle and store it into the SV, optionally
7484 appending to the currently-stored string.
7490 Perl_sv_gets(pTHX_ register SV *const sv, register PerlIO *const fp, I32 append)
7495 register STDCHAR rslast;
7496 register STDCHAR *bp;
7501 PERL_ARGS_ASSERT_SV_GETS;
7503 if (SvTHINKFIRST(sv))
7504 sv_force_normal_flags(sv, append ? 0 : SV_COW_DROP_PV);
7505 /* XXX. If you make this PVIV, then copy on write can copy scalars read
7507 However, perlbench says it's slower, because the existing swipe code
7508 is faster than copy on write.
7509 Swings and roundabouts. */
7510 SvUPGRADE(sv, SVt_PV);
7515 if (PerlIO_isutf8(fp)) {
7517 sv_utf8_upgrade_nomg(sv);
7518 sv_pos_u2b(sv,&append,0);
7520 } else if (SvUTF8(sv)) {
7521 return S_sv_gets_append_to_utf8(aTHX_ sv, fp, append);
7529 if (PerlIO_isutf8(fp))
7532 if (IN_PERL_COMPILETIME) {
7533 /* we always read code in line mode */
7537 else if (RsSNARF(PL_rs)) {
7538 /* If it is a regular disk file use size from stat() as estimate
7539 of amount we are going to read -- may result in mallocing
7540 more memory than we really need if the layers below reduce
7541 the size we read (e.g. CRLF or a gzip layer).
7544 if (!PerlLIO_fstat(PerlIO_fileno(fp), &st) && S_ISREG(st.st_mode)) {
7545 const Off_t offset = PerlIO_tell(fp);
7546 if (offset != (Off_t) -1 && st.st_size + append > offset) {
7547 (void) SvGROW(sv, (STRLEN)((st.st_size - offset) + append + 1));
7553 else if (RsRECORD(PL_rs)) {
7554 return S_sv_gets_read_record(aTHX_ sv, fp, append);
7556 else if (RsPARA(PL_rs)) {
7562 /* Get $/ i.e. PL_rs into same encoding as stream wants */
7563 if (PerlIO_isutf8(fp)) {
7564 rsptr = SvPVutf8(PL_rs, rslen);
7567 if (SvUTF8(PL_rs)) {
7568 if (!sv_utf8_downgrade(PL_rs, TRUE)) {
7569 Perl_croak(aTHX_ "Wide character in $/");
7572 rsptr = SvPV_const(PL_rs, rslen);
7576 rslast = rslen ? rsptr[rslen - 1] : '\0';
7578 if (rspara) { /* have to do this both before and after */
7579 do { /* to make sure file boundaries work right */
7582 i = PerlIO_getc(fp);
7586 PerlIO_ungetc(fp,i);
7592 /* See if we know enough about I/O mechanism to cheat it ! */
7594 /* This used to be #ifdef test - it is made run-time test for ease
7595 of abstracting out stdio interface. One call should be cheap
7596 enough here - and may even be a macro allowing compile
7600 if (PerlIO_fast_gets(fp)) {
7603 * We're going to steal some values from the stdio struct
7604 * and put EVERYTHING in the innermost loop into registers.
7606 register STDCHAR *ptr;
7610 #if defined(VMS) && defined(PERLIO_IS_STDIO)
7611 /* An ungetc()d char is handled separately from the regular
7612 * buffer, so we getc() it back out and stuff it in the buffer.
7614 i = PerlIO_getc(fp);
7615 if (i == EOF) return 0;
7616 *(--((*fp)->_ptr)) = (unsigned char) i;
7620 /* Here is some breathtakingly efficient cheating */
7622 cnt = PerlIO_get_cnt(fp); /* get count into register */
7623 /* make sure we have the room */
7624 if ((I32)(SvLEN(sv) - append) <= cnt + 1) {
7625 /* Not room for all of it
7626 if we are looking for a separator and room for some
7628 if (rslen && cnt > 80 && (I32)SvLEN(sv) > append) {
7629 /* just process what we have room for */
7630 shortbuffered = cnt - SvLEN(sv) + append + 1;
7631 cnt -= shortbuffered;
7635 /* remember that cnt can be negative */
7636 SvGROW(sv, (STRLEN)(append + (cnt <= 0 ? 2 : (cnt + 1))));
7641 bp = (STDCHAR*)SvPVX_const(sv) + append; /* move these two too to registers */
7642 ptr = (STDCHAR*)PerlIO_get_ptr(fp);
7643 DEBUG_P(PerlIO_printf(Perl_debug_log,
7644 "Screamer: entering, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7645 DEBUG_P(PerlIO_printf(Perl_debug_log,
7646 "Screamer: entering: PerlIO * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7647 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7648 PTR2UV(PerlIO_has_base(fp) ? PerlIO_get_base(fp) : 0)));
7653 while (cnt > 0) { /* this | eat */
7655 if ((*bp++ = *ptr++) == rslast) /* really | dust */
7656 goto thats_all_folks; /* screams | sed :-) */
7660 Copy(ptr, bp, cnt, char); /* this | eat */
7661 bp += cnt; /* screams | dust */
7662 ptr += cnt; /* louder | sed :-) */
7664 assert (!shortbuffered);
7665 goto cannot_be_shortbuffered;
7669 if (shortbuffered) { /* oh well, must extend */
7670 cnt = shortbuffered;
7672 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
7674 SvGROW(sv, SvLEN(sv) + append + cnt + 2);
7675 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
7679 cannot_be_shortbuffered:
7680 DEBUG_P(PerlIO_printf(Perl_debug_log,
7681 "Screamer: going to getc, ptr=%"UVuf", cnt=%ld\n",
7682 PTR2UV(ptr),(long)cnt));
7683 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* deregisterize cnt and ptr */
7685 DEBUG_Pv(PerlIO_printf(Perl_debug_log,
7686 "Screamer: pre: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7687 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7688 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7690 /* This used to call 'filbuf' in stdio form, but as that behaves like
7691 getc when cnt <= 0 we use PerlIO_getc here to avoid introducing
7692 another abstraction. */
7693 i = PerlIO_getc(fp); /* get more characters */
7695 DEBUG_Pv(PerlIO_printf(Perl_debug_log,
7696 "Screamer: post: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7697 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7698 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7700 cnt = PerlIO_get_cnt(fp);
7701 ptr = (STDCHAR*)PerlIO_get_ptr(fp); /* reregisterize cnt and ptr */
7702 DEBUG_P(PerlIO_printf(Perl_debug_log,
7703 "Screamer: after getc, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7705 if (i == EOF) /* all done for ever? */
7706 goto thats_really_all_folks;
7708 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
7710 SvGROW(sv, bpx + cnt + 2);
7711 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
7713 *bp++ = (STDCHAR)i; /* store character from PerlIO_getc */
7715 if (rslen && (STDCHAR)i == rslast) /* all done for now? */
7716 goto thats_all_folks;
7720 if ((rslen > 1 && (STRLEN)(bp - (STDCHAR*)SvPVX_const(sv)) < rslen) ||
7721 memNE((char*)bp - rslen, rsptr, rslen))
7722 goto screamer; /* go back to the fray */
7723 thats_really_all_folks:
7725 cnt += shortbuffered;
7726 DEBUG_P(PerlIO_printf(Perl_debug_log,
7727 "Screamer: quitting, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7728 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* put these back or we're in trouble */
7729 DEBUG_P(PerlIO_printf(Perl_debug_log,
7730 "Screamer: end: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7731 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7732 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7734 SvCUR_set(sv, bp - (STDCHAR*)SvPVX_const(sv)); /* set length */
7735 DEBUG_P(PerlIO_printf(Perl_debug_log,
7736 "Screamer: done, len=%ld, string=|%.*s|\n",
7737 (long)SvCUR(sv),(int)SvCUR(sv),SvPVX_const(sv)));
7741 /*The big, slow, and stupid way. */
7742 #ifdef USE_HEAP_INSTEAD_OF_STACK /* Even slower way. */
7743 STDCHAR *buf = NULL;
7744 Newx(buf, 8192, STDCHAR);
7752 register const STDCHAR * const bpe = buf + sizeof(buf);
7754 while ((i = PerlIO_getc(fp)) != EOF && (*bp++ = (STDCHAR)i) != rslast && bp < bpe)
7755 ; /* keep reading */
7759 cnt = PerlIO_read(fp,(char*)buf, sizeof(buf));
7760 /* Accommodate broken VAXC compiler, which applies U8 cast to
7761 * both args of ?: operator, causing EOF to change into 255
7764 i = (U8)buf[cnt - 1];
7770 cnt = 0; /* we do need to re-set the sv even when cnt <= 0 */
7772 sv_catpvn(sv, (char *) buf, cnt);
7774 sv_setpvn(sv, (char *) buf, cnt);
7776 if (i != EOF && /* joy */
7778 SvCUR(sv) < rslen ||
7779 memNE(SvPVX_const(sv) + SvCUR(sv) - rslen, rsptr, rslen)))
7783 * If we're reading from a TTY and we get a short read,
7784 * indicating that the user hit his EOF character, we need
7785 * to notice it now, because if we try to read from the TTY
7786 * again, the EOF condition will disappear.
7788 * The comparison of cnt to sizeof(buf) is an optimization
7789 * that prevents unnecessary calls to feof().
7793 if (!(cnt < (I32)sizeof(buf) && PerlIO_eof(fp)))
7797 #ifdef USE_HEAP_INSTEAD_OF_STACK
7802 if (rspara) { /* have to do this both before and after */
7803 while (i != EOF) { /* to make sure file boundaries work right */
7804 i = PerlIO_getc(fp);
7806 PerlIO_ungetc(fp,i);
7812 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7818 Auto-increment of the value in the SV, doing string to numeric conversion
7819 if necessary. Handles 'get' magic and operator overloading.
7825 Perl_sv_inc(pTHX_ register SV *const sv)
7834 =for apidoc sv_inc_nomg
7836 Auto-increment of the value in the SV, doing string to numeric conversion
7837 if necessary. Handles operator overloading. Skips handling 'get' magic.
7843 Perl_sv_inc_nomg(pTHX_ register SV *const sv)
7851 if (SvTHINKFIRST(sv)) {
7852 if (SvIsCOW(sv) || isGV_with_GP(sv))
7853 sv_force_normal_flags(sv, 0);
7854 if (SvREADONLY(sv)) {
7855 if (IN_PERL_RUNTIME)
7856 Perl_croak_no_modify(aTHX);
7860 if (SvAMAGIC(sv) && AMG_CALLunary(sv, inc_amg))
7862 i = PTR2IV(SvRV(sv));
7867 flags = SvFLAGS(sv);
7868 if ((flags & (SVp_NOK|SVp_IOK)) == SVp_NOK) {
7869 /* It's (privately or publicly) a float, but not tested as an
7870 integer, so test it to see. */
7872 flags = SvFLAGS(sv);
7874 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
7875 /* It's publicly an integer, or privately an integer-not-float */
7876 #ifdef PERL_PRESERVE_IVUV
7880 if (SvUVX(sv) == UV_MAX)
7881 sv_setnv(sv, UV_MAX_P1);
7883 (void)SvIOK_only_UV(sv);
7884 SvUV_set(sv, SvUVX(sv) + 1);
7886 if (SvIVX(sv) == IV_MAX)
7887 sv_setuv(sv, (UV)IV_MAX + 1);
7889 (void)SvIOK_only(sv);
7890 SvIV_set(sv, SvIVX(sv) + 1);
7895 if (flags & SVp_NOK) {
7896 const NV was = SvNVX(sv);
7897 if (NV_OVERFLOWS_INTEGERS_AT &&
7898 was >= NV_OVERFLOWS_INTEGERS_AT) {
7899 Perl_ck_warner(aTHX_ packWARN(WARN_IMPRECISION),
7900 "Lost precision when incrementing %" NVff " by 1",
7903 (void)SvNOK_only(sv);
7904 SvNV_set(sv, was + 1.0);
7908 if (!(flags & SVp_POK) || !*SvPVX_const(sv)) {
7909 if ((flags & SVTYPEMASK) < SVt_PVIV)
7910 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV ? SVt_PVIV : SVt_IV));
7911 (void)SvIOK_only(sv);
7916 while (isALPHA(*d)) d++;
7917 while (isDIGIT(*d)) d++;
7918 if (d < SvEND(sv)) {
7919 #ifdef PERL_PRESERVE_IVUV
7920 /* Got to punt this as an integer if needs be, but we don't issue
7921 warnings. Probably ought to make the sv_iv_please() that does
7922 the conversion if possible, and silently. */
7923 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
7924 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
7925 /* Need to try really hard to see if it's an integer.
7926 9.22337203685478e+18 is an integer.
7927 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
7928 so $a="9.22337203685478e+18"; $a+0; $a++
7929 needs to be the same as $a="9.22337203685478e+18"; $a++
7936 /* sv_2iv *should* have made this an NV */
7937 if (flags & SVp_NOK) {
7938 (void)SvNOK_only(sv);
7939 SvNV_set(sv, SvNVX(sv) + 1.0);
7942 /* I don't think we can get here. Maybe I should assert this
7943 And if we do get here I suspect that sv_setnv will croak. NWC
7945 #if defined(USE_LONG_DOUBLE)
7946 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",
7947 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7949 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_inc punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
7950 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7953 #endif /* PERL_PRESERVE_IVUV */
7954 sv_setnv(sv,Atof(SvPVX_const(sv)) + 1.0);
7958 while (d >= SvPVX_const(sv)) {
7966 /* MKS: The original code here died if letters weren't consecutive.
7967 * at least it didn't have to worry about non-C locales. The
7968 * new code assumes that ('z'-'a')==('Z'-'A'), letters are
7969 * arranged in order (although not consecutively) and that only
7970 * [A-Za-z] are accepted by isALPHA in the C locale.
7972 if (*d != 'z' && *d != 'Z') {
7973 do { ++*d; } while (!isALPHA(*d));
7976 *(d--) -= 'z' - 'a';
7981 *(d--) -= 'z' - 'a' + 1;
7985 /* oh,oh, the number grew */
7986 SvGROW(sv, SvCUR(sv) + 2);
7987 SvCUR_set(sv, SvCUR(sv) + 1);
7988 for (d = SvPVX(sv) + SvCUR(sv); d > SvPVX_const(sv); d--)
7999 Auto-decrement of the value in the SV, doing string to numeric conversion
8000 if necessary. Handles 'get' magic and operator overloading.
8006 Perl_sv_dec(pTHX_ register SV *const sv)
8016 =for apidoc sv_dec_nomg
8018 Auto-decrement of the value in the SV, doing string to numeric conversion
8019 if necessary. Handles operator overloading. Skips handling 'get' magic.
8025 Perl_sv_dec_nomg(pTHX_ register SV *const sv)
8032 if (SvTHINKFIRST(sv)) {
8033 if (SvIsCOW(sv) || isGV_with_GP(sv))
8034 sv_force_normal_flags(sv, 0);
8035 if (SvREADONLY(sv)) {
8036 if (IN_PERL_RUNTIME)
8037 Perl_croak_no_modify(aTHX);
8041 if (SvAMAGIC(sv) && AMG_CALLunary(sv, dec_amg))
8043 i = PTR2IV(SvRV(sv));
8048 /* Unlike sv_inc we don't have to worry about string-never-numbers
8049 and keeping them magic. But we mustn't warn on punting */
8050 flags = SvFLAGS(sv);
8051 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
8052 /* It's publicly an integer, or privately an integer-not-float */
8053 #ifdef PERL_PRESERVE_IVUV
8057 if (SvUVX(sv) == 0) {
8058 (void)SvIOK_only(sv);
8062 (void)SvIOK_only_UV(sv);
8063 SvUV_set(sv, SvUVX(sv) - 1);
8066 if (SvIVX(sv) == IV_MIN) {
8067 sv_setnv(sv, (NV)IV_MIN);
8071 (void)SvIOK_only(sv);
8072 SvIV_set(sv, SvIVX(sv) - 1);
8077 if (flags & SVp_NOK) {
8080 const NV was = SvNVX(sv);
8081 if (NV_OVERFLOWS_INTEGERS_AT &&
8082 was <= -NV_OVERFLOWS_INTEGERS_AT) {
8083 Perl_ck_warner(aTHX_ packWARN(WARN_IMPRECISION),
8084 "Lost precision when decrementing %" NVff " by 1",
8087 (void)SvNOK_only(sv);
8088 SvNV_set(sv, was - 1.0);
8092 if (!(flags & SVp_POK)) {
8093 if ((flags & SVTYPEMASK) < SVt_PVIV)
8094 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV) ? SVt_PVIV : SVt_IV);
8096 (void)SvIOK_only(sv);
8099 #ifdef PERL_PRESERVE_IVUV
8101 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
8102 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
8103 /* Need to try really hard to see if it's an integer.
8104 9.22337203685478e+18 is an integer.
8105 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
8106 so $a="9.22337203685478e+18"; $a+0; $a--
8107 needs to be the same as $a="9.22337203685478e+18"; $a--
8114 /* sv_2iv *should* have made this an NV */
8115 if (flags & SVp_NOK) {
8116 (void)SvNOK_only(sv);
8117 SvNV_set(sv, SvNVX(sv) - 1.0);
8120 /* I don't think we can get here. Maybe I should assert this
8121 And if we do get here I suspect that sv_setnv will croak. NWC
8123 #if defined(USE_LONG_DOUBLE)
8124 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",
8125 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
8127 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_dec punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
8128 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
8132 #endif /* PERL_PRESERVE_IVUV */
8133 sv_setnv(sv,Atof(SvPVX_const(sv)) - 1.0); /* punt */
8136 /* this define is used to eliminate a chunk of duplicated but shared logic
8137 * it has the suffix __SV_C to signal that it isnt API, and isnt meant to be
8138 * used anywhere but here - yves
8140 #define PUSH_EXTEND_MORTAL__SV_C(AnSv) \
8143 PL_tmps_stack[++PL_tmps_ix] = (AnSv); \
8147 =for apidoc sv_mortalcopy
8149 Creates a new SV which is a copy of the original SV (using C<sv_setsv>).
8150 The new SV is marked as mortal. It will be destroyed "soon", either by an
8151 explicit call to FREETMPS, or by an implicit call at places such as
8152 statement boundaries. See also C<sv_newmortal> and C<sv_2mortal>.
8157 /* Make a string that will exist for the duration of the expression
8158 * evaluation. Actually, it may have to last longer than that, but
8159 * hopefully we won't free it until it has been assigned to a
8160 * permanent location. */
8163 Perl_sv_mortalcopy(pTHX_ SV *const oldstr)
8169 sv_setsv(sv,oldstr);
8170 PUSH_EXTEND_MORTAL__SV_C(sv);
8176 =for apidoc sv_newmortal
8178 Creates a new null SV which is mortal. The reference count of the SV is
8179 set to 1. It will be destroyed "soon", either by an explicit call to
8180 FREETMPS, or by an implicit call at places such as statement boundaries.
8181 See also C<sv_mortalcopy> and C<sv_2mortal>.
8187 Perl_sv_newmortal(pTHX)
8193 SvFLAGS(sv) = SVs_TEMP;
8194 PUSH_EXTEND_MORTAL__SV_C(sv);
8200 =for apidoc newSVpvn_flags
8202 Creates a new SV and copies a string into it. The reference count for the
8203 SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
8204 string. You are responsible for ensuring that the source string is at least
8205 C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
8206 Currently the only flag bits accepted are C<SVf_UTF8> and C<SVs_TEMP>.
8207 If C<SVs_TEMP> is set, then C<sv_2mortal()> is called on the result before
8208 returning. If C<SVf_UTF8> is set, C<s> is considered to be in UTF-8 and the
8209 C<SVf_UTF8> flag will be set on the new SV.
8210 C<newSVpvn_utf8()> is a convenience wrapper for this function, defined as
8212 #define newSVpvn_utf8(s, len, u) \
8213 newSVpvn_flags((s), (len), (u) ? SVf_UTF8 : 0)
8219 Perl_newSVpvn_flags(pTHX_ const char *const s, const STRLEN len, const U32 flags)
8224 /* All the flags we don't support must be zero.
8225 And we're new code so I'm going to assert this from the start. */
8226 assert(!(flags & ~(SVf_UTF8|SVs_TEMP)));
8228 sv_setpvn(sv,s,len);
8230 /* This code used to a sv_2mortal(), however we now unroll the call to sv_2mortal()
8231 * and do what it does ourselves here.
8232 * Since we have asserted that flags can only have the SVf_UTF8 and/or SVs_TEMP flags
8233 * set above we can use it to enable the sv flags directly (bypassing SvTEMP_on), which
8234 * in turn means we dont need to mask out the SVf_UTF8 flag below, which means that we
8235 * eliminate quite a few steps than it looks - Yves (explaining patch by gfx)
8238 SvFLAGS(sv) |= flags;
8240 if(flags & SVs_TEMP){
8241 PUSH_EXTEND_MORTAL__SV_C(sv);
8248 =for apidoc sv_2mortal
8250 Marks an existing SV as mortal. The SV will be destroyed "soon", either
8251 by an explicit call to FREETMPS, or by an implicit call at places such as
8252 statement boundaries. SvTEMP() is turned on which means that the SV's
8253 string buffer can be "stolen" if this SV is copied. See also C<sv_newmortal>
8254 and C<sv_mortalcopy>.
8260 Perl_sv_2mortal(pTHX_ register SV *const sv)
8265 if (SvREADONLY(sv) && SvIMMORTAL(sv))
8267 PUSH_EXTEND_MORTAL__SV_C(sv);
8275 Creates a new SV and copies a string into it. The reference count for the
8276 SV is set to 1. If C<len> is zero, Perl will compute the length using
8277 strlen(). For efficiency, consider using C<newSVpvn> instead.
8283 Perl_newSVpv(pTHX_ const char *const s, const STRLEN len)
8289 sv_setpvn(sv, s, len || s == NULL ? len : strlen(s));
8294 =for apidoc newSVpvn
8296 Creates a new SV and copies a string into it. The reference count for the
8297 SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
8298 string. You are responsible for ensuring that the source string is at least
8299 C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
8305 Perl_newSVpvn(pTHX_ const char *const s, const STRLEN len)
8311 sv_setpvn(sv,s,len);
8316 =for apidoc newSVhek
8318 Creates a new SV from the hash key structure. It will generate scalars that
8319 point to the shared string table where possible. Returns a new (undefined)
8320 SV if the hek is NULL.
8326 Perl_newSVhek(pTHX_ const HEK *const hek)
8336 if (HEK_LEN(hek) == HEf_SVKEY) {
8337 return newSVsv(*(SV**)HEK_KEY(hek));
8339 const int flags = HEK_FLAGS(hek);
8340 if (flags & HVhek_WASUTF8) {
8342 Andreas would like keys he put in as utf8 to come back as utf8
8344 STRLEN utf8_len = HEK_LEN(hek);
8345 SV * const sv = newSV_type(SVt_PV);
8346 char *as_utf8 = (char *)bytes_to_utf8 ((U8*)HEK_KEY(hek), &utf8_len);
8347 /* bytes_to_utf8() allocates a new string, which we can repurpose: */
8348 sv_usepvn_flags(sv, as_utf8, utf8_len, SV_HAS_TRAILING_NUL);
8351 } else if (flags & (HVhek_REHASH|HVhek_UNSHARED)) {
8352 /* We don't have a pointer to the hv, so we have to replicate the
8353 flag into every HEK. This hv is using custom a hasing
8354 algorithm. Hence we can't return a shared string scalar, as
8355 that would contain the (wrong) hash value, and might get passed
8356 into an hv routine with a regular hash.
8357 Similarly, a hash that isn't using shared hash keys has to have
8358 the flag in every key so that we know not to try to call
8359 share_hek_hek on it. */
8361 SV * const sv = newSVpvn (HEK_KEY(hek), HEK_LEN(hek));
8366 /* This will be overwhelminly the most common case. */
8368 /* Inline most of newSVpvn_share(), because share_hek_hek() is far
8369 more efficient than sharepvn(). */
8373 sv_upgrade(sv, SVt_PV);
8374 SvPV_set(sv, (char *)HEK_KEY(share_hek_hek(hek)));
8375 SvCUR_set(sv, HEK_LEN(hek));
8388 =for apidoc newSVpvn_share
8390 Creates a new SV with its SvPVX_const pointing to a shared string in the string
8391 table. If the string does not already exist in the table, it is created
8392 first. Turns on READONLY and FAKE. If the C<hash> parameter is non-zero, that
8393 value is used; otherwise the hash is computed. The string's hash can be later
8394 be retrieved from the SV with the C<SvSHARED_HASH()> macro. The idea here is
8395 that as the string table is used for shared hash keys these strings will have
8396 SvPVX_const == HeKEY and hash lookup will avoid string compare.
8402 Perl_newSVpvn_share(pTHX_ const char *src, I32 len, U32 hash)
8406 bool is_utf8 = FALSE;
8407 const char *const orig_src = src;
8410 STRLEN tmplen = -len;
8412 /* See the note in hv.c:hv_fetch() --jhi */
8413 src = (char*)bytes_from_utf8((const U8*)src, &tmplen, &is_utf8);
8417 PERL_HASH(hash, src, len);
8419 /* The logic for this is inlined in S_mro_get_linear_isa_dfs(), so if it
8420 changes here, update it there too. */
8421 sv_upgrade(sv, SVt_PV);
8422 SvPV_set(sv, sharepvn(src, is_utf8?-len:len, hash));
8430 if (src != orig_src)
8436 =for apidoc newSVpv_share
8438 Like C<newSVpvn_share>, but takes a nul-terminated string instead of a
8445 Perl_newSVpv_share(pTHX_ const char *src, U32 hash)
8447 return newSVpvn_share(src, strlen(src), hash);
8450 #if defined(PERL_IMPLICIT_CONTEXT)
8452 /* pTHX_ magic can't cope with varargs, so this is a no-context
8453 * version of the main function, (which may itself be aliased to us).
8454 * Don't access this version directly.
8458 Perl_newSVpvf_nocontext(const char *const pat, ...)
8464 PERL_ARGS_ASSERT_NEWSVPVF_NOCONTEXT;
8466 va_start(args, pat);
8467 sv = vnewSVpvf(pat, &args);
8474 =for apidoc newSVpvf
8476 Creates a new SV and initializes it with the string formatted like
8483 Perl_newSVpvf(pTHX_ const char *const pat, ...)
8488 PERL_ARGS_ASSERT_NEWSVPVF;
8490 va_start(args, pat);
8491 sv = vnewSVpvf(pat, &args);
8496 /* backend for newSVpvf() and newSVpvf_nocontext() */
8499 Perl_vnewSVpvf(pTHX_ const char *const pat, va_list *const args)
8504 PERL_ARGS_ASSERT_VNEWSVPVF;
8507 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
8514 Creates a new SV and copies a floating point value into it.
8515 The reference count for the SV is set to 1.
8521 Perl_newSVnv(pTHX_ const NV n)
8534 Creates a new SV and copies an integer into it. The reference count for the
8541 Perl_newSViv(pTHX_ const IV i)
8554 Creates a new SV and copies an unsigned integer into it.
8555 The reference count for the SV is set to 1.
8561 Perl_newSVuv(pTHX_ const UV u)
8572 =for apidoc newSV_type
8574 Creates a new SV, of the type specified. The reference count for the new SV
8581 Perl_newSV_type(pTHX_ const svtype type)
8586 sv_upgrade(sv, type);
8591 =for apidoc newRV_noinc
8593 Creates an RV wrapper for an SV. The reference count for the original
8594 SV is B<not> incremented.
8600 Perl_newRV_noinc(pTHX_ SV *const tmpRef)
8603 register SV *sv = newSV_type(SVt_IV);
8605 PERL_ARGS_ASSERT_NEWRV_NOINC;
8608 SvRV_set(sv, tmpRef);
8613 /* newRV_inc is the official function name to use now.
8614 * newRV_inc is in fact #defined to newRV in sv.h
8618 Perl_newRV(pTHX_ SV *const sv)
8622 PERL_ARGS_ASSERT_NEWRV;
8624 return newRV_noinc(SvREFCNT_inc_simple_NN(sv));
8630 Creates a new SV which is an exact duplicate of the original SV.
8637 Perl_newSVsv(pTHX_ register SV *const old)
8644 if (SvTYPE(old) == (svtype)SVTYPEMASK) {
8645 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL), "semi-panic: attempt to dup freed string");
8649 /* SV_GMAGIC is the default for sv_setv()
8650 SV_NOSTEAL prevents TEMP buffers being, well, stolen, and saves games
8651 with SvTEMP_off and SvTEMP_on round a call to sv_setsv. */
8652 sv_setsv_flags(sv, old, SV_GMAGIC | SV_NOSTEAL);
8657 =for apidoc sv_reset
8659 Underlying implementation for the C<reset> Perl function.
8660 Note that the perl-level function is vaguely deprecated.
8666 Perl_sv_reset(pTHX_ register const char *s, HV *const stash)
8669 char todo[PERL_UCHAR_MAX+1];
8671 PERL_ARGS_ASSERT_SV_RESET;
8676 if (!*s) { /* reset ?? searches */
8677 MAGIC * const mg = mg_find((const SV *)stash, PERL_MAGIC_symtab);
8679 const U32 count = mg->mg_len / sizeof(PMOP**);
8680 PMOP **pmp = (PMOP**) mg->mg_ptr;
8681 PMOP *const *const end = pmp + count;
8685 SvREADONLY_off(PL_regex_pad[(*pmp)->op_pmoffset]);
8687 (*pmp)->op_pmflags &= ~PMf_USED;
8695 /* reset variables */
8697 if (!HvARRAY(stash))
8700 Zero(todo, 256, char);
8703 I32 i = (unsigned char)*s;
8707 max = (unsigned char)*s++;
8708 for ( ; i <= max; i++) {
8711 for (i = 0; i <= (I32) HvMAX(stash); i++) {
8713 for (entry = HvARRAY(stash)[i];
8715 entry = HeNEXT(entry))
8720 if (!todo[(U8)*HeKEY(entry)])
8722 gv = MUTABLE_GV(HeVAL(entry));
8725 if (SvTHINKFIRST(sv)) {
8726 if (!SvREADONLY(sv) && SvROK(sv))
8728 /* XXX Is this continue a bug? Why should THINKFIRST
8729 exempt us from resetting arrays and hashes? */
8733 if (SvTYPE(sv) >= SVt_PV) {
8735 if (SvPVX_const(sv) != NULL)
8743 if (GvHV(gv) && !HvNAME_get(GvHV(gv))) {
8745 Perl_die(aTHX_ "Can't reset %%ENV on this system");
8748 # if defined(USE_ENVIRON_ARRAY)
8751 # endif /* USE_ENVIRON_ARRAY */
8762 Using various gambits, try to get an IO from an SV: the IO slot if its a
8763 GV; or the recursive result if we're an RV; or the IO slot of the symbol
8764 named after the PV if we're a string.
8770 Perl_sv_2io(pTHX_ SV *const sv)
8775 PERL_ARGS_ASSERT_SV_2IO;
8777 switch (SvTYPE(sv)) {
8779 io = MUTABLE_IO(sv);
8783 if (isGV_with_GP(sv)) {
8784 gv = MUTABLE_GV(sv);
8787 Perl_croak(aTHX_ "Bad filehandle: %s", GvNAME(gv));
8793 Perl_croak(aTHX_ PL_no_usym, "filehandle");
8795 return sv_2io(SvRV(sv));
8796 gv = gv_fetchsv(sv, 0, SVt_PVIO);
8802 Perl_croak(aTHX_ "Bad filehandle: %"SVf, SVfARG(sv));
8811 Using various gambits, try to get a CV from an SV; in addition, try if
8812 possible to set C<*st> and C<*gvp> to the stash and GV associated with it.
8813 The flags in C<lref> are passed to gv_fetchsv.
8819 Perl_sv_2cv(pTHX_ SV *sv, HV **const st, GV **const gvp, const I32 lref)
8825 PERL_ARGS_ASSERT_SV_2CV;
8832 switch (SvTYPE(sv)) {
8836 return MUTABLE_CV(sv);
8846 sv = amagic_deref_call(sv, to_cv_amg);
8847 /* At this point I'd like to do SPAGAIN, but really I need to
8848 force it upon my callers. Hmmm. This is a mess... */
8851 if (SvTYPE(sv) == SVt_PVCV) {
8852 cv = MUTABLE_CV(sv);
8857 else if(isGV_with_GP(sv))
8858 gv = MUTABLE_GV(sv);
8860 Perl_croak(aTHX_ "Not a subroutine reference");
8862 else if (isGV_with_GP(sv)) {
8863 gv = MUTABLE_GV(sv);
8866 gv = gv_fetchsv_nomg(sv, lref, SVt_PVCV);
8873 /* Some flags to gv_fetchsv mean don't really create the GV */
8874 if (!isGV_with_GP(gv)) {
8879 if (lref & ~GV_ADDMG && !GvCVu(gv)) {
8883 gv_efullname3(tmpsv, gv, NULL);
8884 /* XXX this is probably not what they think they're getting.
8885 * It has the same effect as "sub name;", i.e. just a forward
8887 newSUB(start_subparse(FALSE, 0),
8888 newSVOP(OP_CONST, 0, tmpsv),
8892 Perl_croak(aTHX_ "Unable to create sub named \"%"SVf"\"",
8893 SVfARG(SvOK(sv) ? sv : &PL_sv_no));
8902 Returns true if the SV has a true value by Perl's rules.
8903 Use the C<SvTRUE> macro instead, which may call C<sv_true()> or may
8904 instead use an in-line version.
8910 Perl_sv_true(pTHX_ register SV *const sv)
8915 register const XPV* const tXpv = (XPV*)SvANY(sv);
8917 (tXpv->xpv_cur > 1 ||
8918 (tXpv->xpv_cur && *sv->sv_u.svu_pv != '0')))
8925 return SvIVX(sv) != 0;
8928 return SvNVX(sv) != 0.0;
8930 return sv_2bool(sv);
8936 =for apidoc sv_pvn_force
8938 Get a sensible string out of the SV somehow.
8939 A private implementation of the C<SvPV_force> macro for compilers which
8940 can't cope with complex macro expressions. Always use the macro instead.
8942 =for apidoc sv_pvn_force_flags
8944 Get a sensible string out of the SV somehow.
8945 If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<sv> if
8946 appropriate, else not. C<sv_pvn_force> and C<sv_pvn_force_nomg> are
8947 implemented in terms of this function.
8948 You normally want to use the various wrapper macros instead: see
8949 C<SvPV_force> and C<SvPV_force_nomg>
8955 Perl_sv_pvn_force_flags(pTHX_ SV *const sv, STRLEN *const lp, const I32 flags)
8959 PERL_ARGS_ASSERT_SV_PVN_FORCE_FLAGS;
8961 if (SvTHINKFIRST(sv) && !SvROK(sv))
8962 sv_force_normal_flags(sv, 0);
8972 if (SvREADONLY(sv) && !(flags & SV_MUTABLE_RETURN)) {
8973 const char * const ref = sv_reftype(sv,0);
8975 Perl_croak(aTHX_ "Can't coerce readonly %s to string in %s",
8976 ref, OP_DESC(PL_op));
8978 Perl_croak(aTHX_ "Can't coerce readonly %s to string", ref);
8980 if ((SvTYPE(sv) > SVt_PVLV && SvTYPE(sv) != SVt_PVFM)
8981 || isGV_with_GP(sv))
8982 /* diag_listed_as: Can't coerce %s to %s in %s */
8983 Perl_croak(aTHX_ "Can't coerce %s to string in %s", sv_reftype(sv,0),
8985 s = sv_2pv_flags(sv, &len, flags);
8989 if (s != SvPVX_const(sv)) { /* Almost, but not quite, sv_setpvn() */
8992 SvUPGRADE(sv, SVt_PV); /* Never FALSE */
8993 SvGROW(sv, len + 1);
8994 Move(s,SvPVX(sv),len,char);
8996 SvPVX(sv)[len] = '\0';
8999 SvPOK_on(sv); /* validate pointer */
9001 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
9002 PTR2UV(sv),SvPVX_const(sv)));
9005 return SvPVX_mutable(sv);
9009 =for apidoc sv_pvbyten_force
9011 The backend for the C<SvPVbytex_force> macro. Always use the macro instead.
9017 Perl_sv_pvbyten_force(pTHX_ SV *const sv, STRLEN *const lp)
9019 PERL_ARGS_ASSERT_SV_PVBYTEN_FORCE;
9021 sv_pvn_force(sv,lp);
9022 sv_utf8_downgrade(sv,0);
9028 =for apidoc sv_pvutf8n_force
9030 The backend for the C<SvPVutf8x_force> macro. Always use the macro instead.
9036 Perl_sv_pvutf8n_force(pTHX_ SV *const sv, STRLEN *const lp)
9038 PERL_ARGS_ASSERT_SV_PVUTF8N_FORCE;
9040 sv_pvn_force(sv,lp);
9041 sv_utf8_upgrade(sv);
9047 =for apidoc sv_reftype
9049 Returns a string describing what the SV is a reference to.
9055 Perl_sv_reftype(pTHX_ const SV *const sv, const int ob)
9057 PERL_ARGS_ASSERT_SV_REFTYPE;
9059 /* The fact that I don't need to downcast to char * everywhere, only in ?:
9060 inside return suggests a const propagation bug in g++. */
9061 if (ob && SvOBJECT(sv)) {
9062 char * const name = HvNAME_get(SvSTASH(sv));
9063 return name ? name : (char *) "__ANON__";
9066 switch (SvTYPE(sv)) {
9081 case SVt_PVLV: return (char *) (SvROK(sv) ? "REF"
9082 /* tied lvalues should appear to be
9083 * scalars for backwards compatibility */
9084 : (LvTYPE(sv) == 't' || LvTYPE(sv) == 'T')
9085 ? "SCALAR" : "LVALUE");
9086 case SVt_PVAV: return "ARRAY";
9087 case SVt_PVHV: return "HASH";
9088 case SVt_PVCV: return "CODE";
9089 case SVt_PVGV: return (char *) (isGV_with_GP(sv)
9090 ? "GLOB" : "SCALAR");
9091 case SVt_PVFM: return "FORMAT";
9092 case SVt_PVIO: return "IO";
9093 case SVt_BIND: return "BIND";
9094 case SVt_REGEXP: return "REGEXP";
9095 default: return "UNKNOWN";
9101 =for apidoc sv_isobject
9103 Returns a boolean indicating whether the SV is an RV pointing to a blessed
9104 object. If the SV is not an RV, or if the object is not blessed, then this
9111 Perl_sv_isobject(pTHX_ SV *sv)
9127 Returns a boolean indicating whether the SV is blessed into the specified
9128 class. This does not check for subtypes; use C<sv_derived_from> to verify
9129 an inheritance relationship.
9135 Perl_sv_isa(pTHX_ SV *sv, const char *const name)
9139 PERL_ARGS_ASSERT_SV_ISA;
9149 hvname = HvNAME_get(SvSTASH(sv));
9153 return strEQ(hvname, name);
9159 Creates a new SV for the RV, C<rv>, to point to. If C<rv> is not an RV then
9160 it will be upgraded to one. If C<classname> is non-null then the new SV will
9161 be blessed in the specified package. The new SV is returned and its
9162 reference count is 1.
9168 Perl_newSVrv(pTHX_ SV *const rv, const char *const classname)
9173 PERL_ARGS_ASSERT_NEWSVRV;
9177 SV_CHECK_THINKFIRST_COW_DROP(rv);
9178 (void)SvAMAGIC_off(rv);
9180 if (SvTYPE(rv) >= SVt_PVMG) {
9181 const U32 refcnt = SvREFCNT(rv);
9185 SvREFCNT(rv) = refcnt;
9187 sv_upgrade(rv, SVt_IV);
9188 } else if (SvROK(rv)) {
9189 SvREFCNT_dec(SvRV(rv));
9191 prepare_SV_for_RV(rv);
9199 HV* const stash = gv_stashpv(classname, GV_ADD);
9200 (void)sv_bless(rv, stash);
9206 =for apidoc sv_setref_pv
9208 Copies a pointer into a new SV, optionally blessing the SV. The C<rv>
9209 argument will be upgraded to an RV. That RV will be modified to point to
9210 the new SV. If the C<pv> argument is NULL then C<PL_sv_undef> will be placed
9211 into the SV. The C<classname> argument indicates the package for the
9212 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9213 will have a reference count of 1, and the RV will be returned.
9215 Do not use with other Perl types such as HV, AV, SV, CV, because those
9216 objects will become corrupted by the pointer copy process.
9218 Note that C<sv_setref_pvn> copies the string while this copies the pointer.
9224 Perl_sv_setref_pv(pTHX_ SV *const rv, const char *const classname, void *const pv)
9228 PERL_ARGS_ASSERT_SV_SETREF_PV;
9231 sv_setsv(rv, &PL_sv_undef);
9235 sv_setiv(newSVrv(rv,classname), PTR2IV(pv));
9240 =for apidoc sv_setref_iv
9242 Copies an integer into a new SV, optionally blessing the SV. The C<rv>
9243 argument will be upgraded to an RV. That RV will be modified to point to
9244 the new SV. The C<classname> argument indicates the package for the
9245 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9246 will have a reference count of 1, and the RV will be returned.
9252 Perl_sv_setref_iv(pTHX_ SV *const rv, const char *const classname, const IV iv)
9254 PERL_ARGS_ASSERT_SV_SETREF_IV;
9256 sv_setiv(newSVrv(rv,classname), iv);
9261 =for apidoc sv_setref_uv
9263 Copies an unsigned integer into a new SV, optionally blessing the SV. The C<rv>
9264 argument will be upgraded to an RV. That RV will be modified to point to
9265 the new SV. The C<classname> argument indicates the package for the
9266 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9267 will have a reference count of 1, and the RV will be returned.
9273 Perl_sv_setref_uv(pTHX_ SV *const rv, const char *const classname, const UV uv)
9275 PERL_ARGS_ASSERT_SV_SETREF_UV;
9277 sv_setuv(newSVrv(rv,classname), uv);
9282 =for apidoc sv_setref_nv
9284 Copies a double into a new SV, optionally blessing the SV. The C<rv>
9285 argument will be upgraded to an RV. That RV will be modified to point to
9286 the new SV. The C<classname> argument indicates the package for the
9287 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9288 will have a reference count of 1, and the RV will be returned.
9294 Perl_sv_setref_nv(pTHX_ SV *const rv, const char *const classname, const NV nv)
9296 PERL_ARGS_ASSERT_SV_SETREF_NV;
9298 sv_setnv(newSVrv(rv,classname), nv);
9303 =for apidoc sv_setref_pvn
9305 Copies a string into a new SV, optionally blessing the SV. The length of the
9306 string must be specified with C<n>. The C<rv> argument will be upgraded to
9307 an RV. That RV will be modified to point to the new SV. The C<classname>
9308 argument indicates the package for the blessing. Set C<classname> to
9309 C<NULL> to avoid the blessing. The new SV will have a reference count
9310 of 1, and the RV will be returned.
9312 Note that C<sv_setref_pv> copies the pointer while this copies the string.
9318 Perl_sv_setref_pvn(pTHX_ SV *const rv, const char *const classname,
9319 const char *const pv, const STRLEN n)
9321 PERL_ARGS_ASSERT_SV_SETREF_PVN;
9323 sv_setpvn(newSVrv(rv,classname), pv, n);
9328 =for apidoc sv_bless
9330 Blesses an SV into a specified package. The SV must be an RV. The package
9331 must be designated by its stash (see C<gv_stashpv()>). The reference count
9332 of the SV is unaffected.
9338 Perl_sv_bless(pTHX_ SV *const sv, HV *const stash)
9343 PERL_ARGS_ASSERT_SV_BLESS;
9346 Perl_croak(aTHX_ "Can't bless non-reference value");
9348 if (SvFLAGS(tmpRef) & (SVs_OBJECT|SVf_READONLY)) {
9349 if (SvIsCOW(tmpRef))
9350 sv_force_normal_flags(tmpRef, 0);
9351 if (SvREADONLY(tmpRef))
9352 Perl_croak_no_modify(aTHX);
9353 if (SvOBJECT(tmpRef)) {
9354 if (SvTYPE(tmpRef) != SVt_PVIO)
9356 SvREFCNT_dec(SvSTASH(tmpRef));
9359 SvOBJECT_on(tmpRef);
9360 if (SvTYPE(tmpRef) != SVt_PVIO)
9362 SvUPGRADE(tmpRef, SVt_PVMG);
9363 SvSTASH_set(tmpRef, MUTABLE_HV(SvREFCNT_inc_simple(stash)));
9368 (void)SvAMAGIC_off(sv);
9370 if(SvSMAGICAL(tmpRef))
9371 if(mg_find(tmpRef, PERL_MAGIC_ext) || mg_find(tmpRef, PERL_MAGIC_uvar))
9379 /* Downgrades a PVGV to a PVMG. If it’s actually a PVLV, we leave the type
9380 * as it is after unglobbing it.
9384 S_sv_unglob(pTHX_ SV *const sv)
9389 SV * const temp = sv_newmortal();
9391 PERL_ARGS_ASSERT_SV_UNGLOB;
9393 assert(SvTYPE(sv) == SVt_PVGV || SvTYPE(sv) == SVt_PVLV);
9395 gv_efullname3(temp, MUTABLE_GV(sv), "*");
9398 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
9399 && HvNAME_get(stash))
9400 mro_method_changed_in(stash);
9401 gp_free(MUTABLE_GV(sv));
9404 sv_del_backref(MUTABLE_SV(GvSTASH(sv)), sv);
9408 if (GvNAME_HEK(sv)) {
9409 unshare_hek(GvNAME_HEK(sv));
9411 isGV_with_GP_off(sv);
9413 if(SvTYPE(sv) == SVt_PVGV) {
9414 /* need to keep SvANY(sv) in the right arena */
9415 xpvmg = new_XPVMG();
9416 StructCopy(SvANY(sv), xpvmg, XPVMG);
9417 del_XPVGV(SvANY(sv));
9420 SvFLAGS(sv) &= ~SVTYPEMASK;
9421 SvFLAGS(sv) |= SVt_PVMG;
9424 /* Intentionally not calling any local SET magic, as this isn't so much a
9425 set operation as merely an internal storage change. */
9426 sv_setsv_flags(sv, temp, 0);
9430 =for apidoc sv_unref_flags
9432 Unsets the RV status of the SV, and decrements the reference count of
9433 whatever was being referenced by the RV. This can almost be thought of
9434 as a reversal of C<newSVrv>. The C<cflags> argument can contain
9435 C<SV_IMMEDIATE_UNREF> to force the reference count to be decremented
9436 (otherwise the decrementing is conditional on the reference count being
9437 different from one or the reference being a readonly SV).
9444 Perl_sv_unref_flags(pTHX_ SV *const ref, const U32 flags)
9446 SV* const target = SvRV(ref);
9448 PERL_ARGS_ASSERT_SV_UNREF_FLAGS;
9450 if (SvWEAKREF(ref)) {
9451 sv_del_backref(target, ref);
9453 SvRV_set(ref, NULL);
9456 SvRV_set(ref, NULL);
9458 /* You can't have a || SvREADONLY(target) here, as $a = $$a, where $a was
9459 assigned to as BEGIN {$a = \"Foo"} will fail. */
9460 if (SvREFCNT(target) != 1 || (flags & SV_IMMEDIATE_UNREF))
9461 SvREFCNT_dec(target);
9462 else /* XXX Hack, but hard to make $a=$a->[1] work otherwise */
9463 sv_2mortal(target); /* Schedule for freeing later */
9467 =for apidoc sv_untaint
9469 Untaint an SV. Use C<SvTAINTED_off> instead.
9475 Perl_sv_untaint(pTHX_ SV *const sv)
9477 PERL_ARGS_ASSERT_SV_UNTAINT;
9479 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
9480 MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
9487 =for apidoc sv_tainted
9489 Test an SV for taintedness. Use C<SvTAINTED> instead.
9495 Perl_sv_tainted(pTHX_ SV *const sv)
9497 PERL_ARGS_ASSERT_SV_TAINTED;
9499 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
9500 const MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
9501 if (mg && (mg->mg_len & 1) )
9508 =for apidoc sv_setpviv
9510 Copies an integer into the given SV, also updating its string value.
9511 Does not handle 'set' magic. See C<sv_setpviv_mg>.
9517 Perl_sv_setpviv(pTHX_ SV *const sv, const IV iv)
9519 char buf[TYPE_CHARS(UV)];
9521 char * const ptr = uiv_2buf(buf, iv, 0, 0, &ebuf);
9523 PERL_ARGS_ASSERT_SV_SETPVIV;
9525 sv_setpvn(sv, ptr, ebuf - ptr);
9529 =for apidoc sv_setpviv_mg
9531 Like C<sv_setpviv>, but also handles 'set' magic.
9537 Perl_sv_setpviv_mg(pTHX_ SV *const sv, const IV iv)
9539 PERL_ARGS_ASSERT_SV_SETPVIV_MG;
9545 #if defined(PERL_IMPLICIT_CONTEXT)
9547 /* pTHX_ magic can't cope with varargs, so this is a no-context
9548 * version of the main function, (which may itself be aliased to us).
9549 * Don't access this version directly.
9553 Perl_sv_setpvf_nocontext(SV *const sv, const char *const pat, ...)
9558 PERL_ARGS_ASSERT_SV_SETPVF_NOCONTEXT;
9560 va_start(args, pat);
9561 sv_vsetpvf(sv, pat, &args);
9565 /* pTHX_ magic can't cope with varargs, so this is a no-context
9566 * version of the main function, (which may itself be aliased to us).
9567 * Don't access this version directly.
9571 Perl_sv_setpvf_mg_nocontext(SV *const sv, const char *const pat, ...)
9576 PERL_ARGS_ASSERT_SV_SETPVF_MG_NOCONTEXT;
9578 va_start(args, pat);
9579 sv_vsetpvf_mg(sv, pat, &args);
9585 =for apidoc sv_setpvf
9587 Works like C<sv_catpvf> but copies the text into the SV instead of
9588 appending it. Does not handle 'set' magic. See C<sv_setpvf_mg>.
9594 Perl_sv_setpvf(pTHX_ SV *const sv, const char *const pat, ...)
9598 PERL_ARGS_ASSERT_SV_SETPVF;
9600 va_start(args, pat);
9601 sv_vsetpvf(sv, pat, &args);
9606 =for apidoc sv_vsetpvf
9608 Works like C<sv_vcatpvf> but copies the text into the SV instead of
9609 appending it. Does not handle 'set' magic. See C<sv_vsetpvf_mg>.
9611 Usually used via its frontend C<sv_setpvf>.
9617 Perl_sv_vsetpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9619 PERL_ARGS_ASSERT_SV_VSETPVF;
9621 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9625 =for apidoc sv_setpvf_mg
9627 Like C<sv_setpvf>, but also handles 'set' magic.
9633 Perl_sv_setpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
9637 PERL_ARGS_ASSERT_SV_SETPVF_MG;
9639 va_start(args, pat);
9640 sv_vsetpvf_mg(sv, pat, &args);
9645 =for apidoc sv_vsetpvf_mg
9647 Like C<sv_vsetpvf>, but also handles 'set' magic.
9649 Usually used via its frontend C<sv_setpvf_mg>.
9655 Perl_sv_vsetpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9657 PERL_ARGS_ASSERT_SV_VSETPVF_MG;
9659 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9663 #if defined(PERL_IMPLICIT_CONTEXT)
9665 /* pTHX_ magic can't cope with varargs, so this is a no-context
9666 * version of the main function, (which may itself be aliased to us).
9667 * Don't access this version directly.
9671 Perl_sv_catpvf_nocontext(SV *const sv, const char *const pat, ...)
9676 PERL_ARGS_ASSERT_SV_CATPVF_NOCONTEXT;
9678 va_start(args, pat);
9679 sv_vcatpvf(sv, pat, &args);
9683 /* pTHX_ magic can't cope with varargs, so this is a no-context
9684 * version of the main function, (which may itself be aliased to us).
9685 * Don't access this version directly.
9689 Perl_sv_catpvf_mg_nocontext(SV *const sv, const char *const pat, ...)
9694 PERL_ARGS_ASSERT_SV_CATPVF_MG_NOCONTEXT;
9696 va_start(args, pat);
9697 sv_vcatpvf_mg(sv, pat, &args);
9703 =for apidoc sv_catpvf
9705 Processes its arguments like C<sprintf> and appends the formatted
9706 output to an SV. If the appended data contains "wide" characters
9707 (including, but not limited to, SVs with a UTF-8 PV formatted with %s,
9708 and characters >255 formatted with %c), the original SV might get
9709 upgraded to UTF-8. Handles 'get' magic, but not 'set' magic. See
9710 C<sv_catpvf_mg>. If the original SV was UTF-8, the pattern should be
9711 valid UTF-8; if the original SV was bytes, the pattern should be too.
9716 Perl_sv_catpvf(pTHX_ SV *const sv, const char *const pat, ...)
9720 PERL_ARGS_ASSERT_SV_CATPVF;
9722 va_start(args, pat);
9723 sv_vcatpvf(sv, pat, &args);
9728 =for apidoc sv_vcatpvf
9730 Processes its arguments like C<vsprintf> and appends the formatted output
9731 to an SV. Does not handle 'set' magic. See C<sv_vcatpvf_mg>.
9733 Usually used via its frontend C<sv_catpvf>.
9739 Perl_sv_vcatpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9741 PERL_ARGS_ASSERT_SV_VCATPVF;
9743 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9747 =for apidoc sv_catpvf_mg
9749 Like C<sv_catpvf>, but also handles 'set' magic.
9755 Perl_sv_catpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
9759 PERL_ARGS_ASSERT_SV_CATPVF_MG;
9761 va_start(args, pat);
9762 sv_vcatpvf_mg(sv, pat, &args);
9767 =for apidoc sv_vcatpvf_mg
9769 Like C<sv_vcatpvf>, but also handles 'set' magic.
9771 Usually used via its frontend C<sv_catpvf_mg>.
9777 Perl_sv_vcatpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9779 PERL_ARGS_ASSERT_SV_VCATPVF_MG;
9781 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9786 =for apidoc sv_vsetpvfn
9788 Works like C<sv_vcatpvfn> but copies the text into the SV instead of
9791 Usually used via one of its frontends C<sv_vsetpvf> and C<sv_vsetpvf_mg>.
9797 Perl_sv_vsetpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
9798 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
9800 PERL_ARGS_ASSERT_SV_VSETPVFN;
9803 sv_vcatpvfn(sv, pat, patlen, args, svargs, svmax, maybe_tainted);
9808 * Warn of missing argument to sprintf, and then return a defined value
9809 * to avoid inappropriate "use of uninit" warnings [perl #71000].
9811 #define WARN_MISSING WARN_UNINITIALIZED /* Not sure we want a new category */
9813 S_vcatpvfn_missing_argument(pTHX) {
9814 if (ckWARN(WARN_MISSING)) {
9815 Perl_warner(aTHX_ packWARN(WARN_MISSING), "Missing argument in %s",
9816 PL_op ? OP_DESC(PL_op) : "sv_vcatpvfn()");
9823 S_expect_number(pTHX_ char **const pattern)
9828 PERL_ARGS_ASSERT_EXPECT_NUMBER;
9830 switch (**pattern) {
9831 case '1': case '2': case '3':
9832 case '4': case '5': case '6':
9833 case '7': case '8': case '9':
9834 var = *(*pattern)++ - '0';
9835 while (isDIGIT(**pattern)) {
9836 const I32 tmp = var * 10 + (*(*pattern)++ - '0');
9838 Perl_croak(aTHX_ "Integer overflow in format string for %s", (PL_op ? OP_DESC(PL_op) : "sv_vcatpvfn"));
9846 S_F0convert(NV nv, char *const endbuf, STRLEN *const len)
9848 const int neg = nv < 0;
9851 PERL_ARGS_ASSERT_F0CONVERT;
9859 if (uv & 1 && uv == nv)
9860 uv--; /* Round to even */
9862 const unsigned dig = uv % 10;
9875 =for apidoc sv_vcatpvfn
9877 Processes its arguments like C<vsprintf> and appends the formatted output
9878 to an SV. Uses an array of SVs if the C style variable argument list is
9879 missing (NULL). When running with taint checks enabled, indicates via
9880 C<maybe_tainted> if results are untrustworthy (often due to the use of
9883 Usually used via one of its frontends C<sv_vcatpvf> and C<sv_vcatpvf_mg>.
9889 #define VECTORIZE_ARGS vecsv = va_arg(*args, SV*);\
9890 vecstr = (U8*)SvPV_const(vecsv,veclen);\
9891 vec_utf8 = DO_UTF8(vecsv);
9893 /* XXX maybe_tainted is never assigned to, so the doc above is lying. */
9896 Perl_sv_vcatpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
9897 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
9905 static const char nullstr[] = "(null)";
9907 bool has_utf8 = DO_UTF8(sv); /* has the result utf8? */
9908 const bool pat_utf8 = has_utf8; /* the pattern is in utf8? */
9910 /* Times 4: a decimal digit takes more than 3 binary digits.
9911 * NV_DIG: mantissa takes than many decimal digits.
9912 * Plus 32: Playing safe. */
9913 char ebuf[IV_DIG * 4 + NV_DIG + 32];
9914 /* large enough for "%#.#f" --chip */
9915 /* what about long double NVs? --jhi */
9917 PERL_ARGS_ASSERT_SV_VCATPVFN;
9918 PERL_UNUSED_ARG(maybe_tainted);
9920 /* no matter what, this is a string now */
9921 (void)SvPV_force(sv, origlen);
9923 /* special-case "", "%s", and "%-p" (SVf - see below) */
9926 if (patlen == 2 && pat[0] == '%' && pat[1] == 's') {
9928 const char * const s = va_arg(*args, char*);
9929 sv_catpv(sv, s ? s : nullstr);
9931 else if (svix < svmax) {
9932 sv_catsv(sv, *svargs);
9935 S_vcatpvfn_missing_argument(aTHX);
9938 if (args && patlen == 3 && pat[0] == '%' &&
9939 pat[1] == '-' && pat[2] == 'p') {
9940 argsv = MUTABLE_SV(va_arg(*args, void*));
9941 sv_catsv(sv, argsv);
9945 #ifndef USE_LONG_DOUBLE
9946 /* special-case "%.<number>[gf]" */
9947 if ( !args && patlen <= 5 && pat[0] == '%' && pat[1] == '.'
9948 && (pat[patlen-1] == 'g' || pat[patlen-1] == 'f') ) {
9949 unsigned digits = 0;
9953 while (*pp >= '0' && *pp <= '9')
9954 digits = 10 * digits + (*pp++ - '0');
9955 if (pp - pat == (int)patlen - 1 && svix < svmax) {
9956 const NV nv = SvNV(*svargs);
9958 /* Add check for digits != 0 because it seems that some
9959 gconverts are buggy in this case, and we don't yet have
9960 a Configure test for this. */
9961 if (digits && digits < sizeof(ebuf) - NV_DIG - 10) {
9962 /* 0, point, slack */
9963 Gconvert(nv, (int)digits, 0, ebuf);
9965 if (*ebuf) /* May return an empty string for digits==0 */
9968 } else if (!digits) {
9971 if ((p = F0convert(nv, ebuf + sizeof ebuf, &l))) {
9972 sv_catpvn(sv, p, l);
9978 #endif /* !USE_LONG_DOUBLE */
9980 if (!args && svix < svmax && DO_UTF8(*svargs))
9983 patend = (char*)pat + patlen;
9984 for (p = (char*)pat; p < patend; p = q) {
9987 bool vectorize = FALSE;
9988 bool vectorarg = FALSE;
9989 bool vec_utf8 = FALSE;
9995 bool has_precis = FALSE;
9997 const I32 osvix = svix;
9998 bool is_utf8 = FALSE; /* is this item utf8? */
9999 #ifdef HAS_LDBL_SPRINTF_BUG
10000 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
10001 with sfio - Allen <allens@cpan.org> */
10002 bool fix_ldbl_sprintf_bug = FALSE;
10006 U8 utf8buf[UTF8_MAXBYTES+1];
10007 STRLEN esignlen = 0;
10009 const char *eptr = NULL;
10010 const char *fmtstart;
10013 const U8 *vecstr = NULL;
10020 /* we need a long double target in case HAS_LONG_DOUBLE but
10021 not USE_LONG_DOUBLE
10023 #if defined(HAS_LONG_DOUBLE) && LONG_DOUBLESIZE > DOUBLESIZE
10031 const char *dotstr = ".";
10032 STRLEN dotstrlen = 1;
10033 I32 efix = 0; /* explicit format parameter index */
10034 I32 ewix = 0; /* explicit width index */
10035 I32 epix = 0; /* explicit precision index */
10036 I32 evix = 0; /* explicit vector index */
10037 bool asterisk = FALSE;
10039 /* echo everything up to the next format specification */
10040 for (q = p; q < patend && *q != '%'; ++q) ;
10042 if (has_utf8 && !pat_utf8)
10043 sv_catpvn_utf8_upgrade(sv, p, q - p, nsv);
10045 sv_catpvn(sv, p, q - p);
10054 We allow format specification elements in this order:
10055 \d+\$ explicit format parameter index
10057 v|\*(\d+\$)?v vector with optional (optionally specified) arg
10058 0 flag (as above): repeated to allow "v02"
10059 \d+|\*(\d+\$)? width using optional (optionally specified) arg
10060 \.(\d*|\*(\d+\$)?) precision using optional (optionally specified) arg
10062 [%bcdefginopsuxDFOUX] format (mandatory)
10067 As of perl5.9.3, printf format checking is on by default.
10068 Internally, perl uses %p formats to provide an escape to
10069 some extended formatting. This block deals with those
10070 extensions: if it does not match, (char*)q is reset and
10071 the normal format processing code is used.
10073 Currently defined extensions are:
10074 %p include pointer address (standard)
10075 %-p (SVf) include an SV (previously %_)
10076 %-<num>p include an SV with precision <num>
10077 %<num>p reserved for future extensions
10079 Robin Barker 2005-07-14
10081 %1p (VDf) removed. RMB 2007-10-19
10088 n = expect_number(&q);
10090 if (sv) { /* SVf */
10095 argsv = MUTABLE_SV(va_arg(*args, void*));
10096 eptr = SvPV_const(argsv, elen);
10097 if (DO_UTF8(argsv))
10102 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL),
10103 "internal %%<num>p might conflict with future printf extensions");
10109 if ( (width = expect_number(&q)) ) {
10124 if (plus == '+' && *q == ' ') /* '+' over ' ' */
10153 if ( (ewix = expect_number(&q)) )
10162 if ((vectorarg = asterisk)) {
10175 width = expect_number(&q);
10178 if (vectorize && vectorarg) {
10179 /* vectorizing, but not with the default "." */
10181 vecsv = va_arg(*args, SV*);
10183 vecsv = (evix > 0 && evix <= svmax)
10184 ? svargs[evix-1] : S_vcatpvfn_missing_argument(aTHX);
10186 vecsv = svix < svmax
10187 ? svargs[svix++] : S_vcatpvfn_missing_argument(aTHX);
10189 dotstr = SvPV_const(vecsv, dotstrlen);
10190 /* Keep the DO_UTF8 test *after* the SvPV call, else things go
10191 bad with tied or overloaded values that return UTF8. */
10192 if (DO_UTF8(vecsv))
10194 else if (has_utf8) {
10195 vecsv = sv_mortalcopy(vecsv);
10196 sv_utf8_upgrade(vecsv);
10197 dotstr = SvPV_const(vecsv, dotstrlen);
10204 i = va_arg(*args, int);
10206 i = (ewix ? ewix <= svmax : svix < svmax) ?
10207 SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
10209 width = (i < 0) ? -i : i;
10219 if ( ((epix = expect_number(&q))) && (*q++ != '$') )
10221 /* XXX: todo, support specified precision parameter */
10225 i = va_arg(*args, int);
10227 i = (ewix ? ewix <= svmax : svix < svmax)
10228 ? SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
10230 has_precis = !(i < 0);
10234 while (isDIGIT(*q))
10235 precis = precis * 10 + (*q++ - '0');
10244 else if (efix ? (efix > 0 && efix <= svmax) : svix < svmax) {
10245 vecsv = svargs[efix ? efix-1 : svix++];
10246 vecstr = (U8*)SvPV_const(vecsv,veclen);
10247 vec_utf8 = DO_UTF8(vecsv);
10249 /* if this is a version object, we need to convert
10250 * back into v-string notation and then let the
10251 * vectorize happen normally
10253 if (sv_derived_from(vecsv, "version")) {
10254 char *version = savesvpv(vecsv);
10255 if ( hv_exists(MUTABLE_HV(SvRV(vecsv)), "alpha", 5 ) ) {
10256 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
10257 "vector argument not supported with alpha versions");
10260 vecsv = sv_newmortal();
10261 scan_vstring(version, version + veclen, vecsv);
10262 vecstr = (U8*)SvPV_const(vecsv, veclen);
10263 vec_utf8 = DO_UTF8(vecsv);
10277 case 'I': /* Ix, I32x, and I64x */
10279 if (q[1] == '6' && q[2] == '4') {
10285 if (q[1] == '3' && q[2] == '2') {
10295 #if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
10307 #if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
10308 if (*q == 'l') { /* lld, llf */
10317 if (*++q == 'h') { /* hhd, hhu */
10346 if (!vectorize && !args) {
10348 const I32 i = efix-1;
10349 argsv = (i >= 0 && i < svmax)
10350 ? svargs[i] : S_vcatpvfn_missing_argument(aTHX);
10352 argsv = (svix >= 0 && svix < svmax)
10353 ? svargs[svix++] : S_vcatpvfn_missing_argument(aTHX);
10357 switch (c = *q++) {
10364 uv = (args) ? va_arg(*args, int) : SvIV(argsv);
10366 (!UNI_IS_INVARIANT(uv) && SvUTF8(sv)))
10368 eptr = (char*)utf8buf;
10369 elen = uvchr_to_utf8((U8*)eptr, uv) - utf8buf;
10383 eptr = va_arg(*args, char*);
10385 elen = strlen(eptr);
10387 eptr = (char *)nullstr;
10388 elen = sizeof nullstr - 1;
10392 eptr = SvPV_const(argsv, elen);
10393 if (DO_UTF8(argsv)) {
10394 STRLEN old_precis = precis;
10395 if (has_precis && precis < elen) {
10396 STRLEN ulen = sv_len_utf8(argsv);
10397 I32 p = precis > ulen ? ulen : precis;
10398 sv_pos_u2b(argsv, &p, 0); /* sticks at end */
10401 if (width) { /* fudge width (can't fudge elen) */
10402 if (has_precis && precis < elen)
10403 width += precis - old_precis;
10405 width += elen - sv_len_utf8(argsv);
10412 if (has_precis && precis < elen)
10419 if (alt || vectorize)
10421 uv = PTR2UV(args ? va_arg(*args, void*) : argsv);
10442 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
10451 esignbuf[esignlen++] = plus;
10455 case 'c': iv = (char)va_arg(*args, int); break;
10456 case 'h': iv = (short)va_arg(*args, int); break;
10457 case 'l': iv = va_arg(*args, long); break;
10458 case 'V': iv = va_arg(*args, IV); break;
10459 case 'z': iv = va_arg(*args, SSize_t); break;
10460 case 't': iv = va_arg(*args, ptrdiff_t); break;
10461 default: iv = va_arg(*args, int); break;
10463 case 'j': iv = va_arg(*args, intmax_t); break;
10467 iv = va_arg(*args, Quad_t); break;
10474 IV tiv = SvIV(argsv); /* work around GCC bug #13488 */
10476 case 'c': iv = (char)tiv; break;
10477 case 'h': iv = (short)tiv; break;
10478 case 'l': iv = (long)tiv; break;
10480 default: iv = tiv; break;
10483 iv = (Quad_t)tiv; break;
10489 if ( !vectorize ) /* we already set uv above */
10494 esignbuf[esignlen++] = plus;
10498 esignbuf[esignlen++] = '-';
10542 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
10553 case 'c': uv = (unsigned char)va_arg(*args, unsigned); break;
10554 case 'h': uv = (unsigned short)va_arg(*args, unsigned); break;
10555 case 'l': uv = va_arg(*args, unsigned long); break;
10556 case 'V': uv = va_arg(*args, UV); break;
10557 case 'z': uv = va_arg(*args, Size_t); break;
10558 case 't': uv = va_arg(*args, ptrdiff_t); break; /* will sign extend, but there is no uptrdiff_t, so oh well */
10560 case 'j': uv = va_arg(*args, uintmax_t); break;
10562 default: uv = va_arg(*args, unsigned); break;
10565 uv = va_arg(*args, Uquad_t); break;
10572 UV tuv = SvUV(argsv); /* work around GCC bug #13488 */
10574 case 'c': uv = (unsigned char)tuv; break;
10575 case 'h': uv = (unsigned short)tuv; break;
10576 case 'l': uv = (unsigned long)tuv; break;
10578 default: uv = tuv; break;
10581 uv = (Uquad_t)tuv; break;
10590 char *ptr = ebuf + sizeof ebuf;
10591 bool tempalt = uv ? alt : FALSE; /* Vectors can't change alt */
10597 p = (char *)((c == 'X') ? PL_hexdigit + 16 : PL_hexdigit);
10601 } while (uv >>= 4);
10603 esignbuf[esignlen++] = '0';
10604 esignbuf[esignlen++] = c; /* 'x' or 'X' */
10610 *--ptr = '0' + dig;
10611 } while (uv >>= 3);
10612 if (alt && *ptr != '0')
10618 *--ptr = '0' + dig;
10619 } while (uv >>= 1);
10621 esignbuf[esignlen++] = '0';
10622 esignbuf[esignlen++] = c;
10625 default: /* it had better be ten or less */
10628 *--ptr = '0' + dig;
10629 } while (uv /= base);
10632 elen = (ebuf + sizeof ebuf) - ptr;
10636 zeros = precis - elen;
10637 else if (precis == 0 && elen == 1 && *eptr == '0'
10638 && !(base == 8 && alt)) /* "%#.0o" prints "0" */
10641 /* a precision nullifies the 0 flag. */
10648 /* FLOATING POINT */
10651 c = 'f'; /* maybe %F isn't supported here */
10653 case 'e': case 'E':
10655 case 'g': case 'G':
10659 /* This is evil, but floating point is even more evil */
10661 /* for SV-style calling, we can only get NV
10662 for C-style calling, we assume %f is double;
10663 for simplicity we allow any of %Lf, %llf, %qf for long double
10667 #if defined(USE_LONG_DOUBLE)
10671 /* [perl #20339] - we should accept and ignore %lf rather than die */
10675 #if defined(USE_LONG_DOUBLE)
10676 intsize = args ? 0 : 'q';
10680 #if defined(HAS_LONG_DOUBLE)
10693 /* now we need (long double) if intsize == 'q', else (double) */
10695 #if LONG_DOUBLESIZE > DOUBLESIZE
10697 va_arg(*args, long double) :
10698 va_arg(*args, double)
10700 va_arg(*args, double)
10705 /* nv * 0 will be NaN for NaN, +Inf and -Inf, and 0 for anything
10706 else. frexp() has some unspecified behaviour for those three */
10707 if (c != 'e' && c != 'E' && (nv * 0) == 0) {
10709 /* FIXME: if HAS_LONG_DOUBLE but not USE_LONG_DOUBLE this
10710 will cast our (long double) to (double) */
10711 (void)Perl_frexp(nv, &i);
10712 if (i == PERL_INT_MIN)
10713 Perl_die(aTHX_ "panic: frexp");
10715 need = BIT_DIGITS(i);
10717 need += has_precis ? precis : 6; /* known default */
10722 #ifdef HAS_LDBL_SPRINTF_BUG
10723 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
10724 with sfio - Allen <allens@cpan.org> */
10727 # define MY_DBL_MAX DBL_MAX
10728 # else /* XXX guessing! HUGE_VAL may be defined as infinity, so not using */
10729 # if DOUBLESIZE >= 8
10730 # define MY_DBL_MAX 1.7976931348623157E+308L
10732 # define MY_DBL_MAX 3.40282347E+38L
10736 # ifdef HAS_LDBL_SPRINTF_BUG_LESS1 /* only between -1L & 1L - Allen */
10737 # define MY_DBL_MAX_BUG 1L
10739 # define MY_DBL_MAX_BUG MY_DBL_MAX
10743 # define MY_DBL_MIN DBL_MIN
10744 # else /* XXX guessing! -Allen */
10745 # if DOUBLESIZE >= 8
10746 # define MY_DBL_MIN 2.2250738585072014E-308L
10748 # define MY_DBL_MIN 1.17549435E-38L
10752 if ((intsize == 'q') && (c == 'f') &&
10753 ((nv < MY_DBL_MAX_BUG) && (nv > -MY_DBL_MAX_BUG)) &&
10754 (need < DBL_DIG)) {
10755 /* it's going to be short enough that
10756 * long double precision is not needed */
10758 if ((nv <= 0L) && (nv >= -0L))
10759 fix_ldbl_sprintf_bug = TRUE; /* 0 is 0 - easiest */
10761 /* would use Perl_fp_class as a double-check but not
10762 * functional on IRIX - see perl.h comments */
10764 if ((nv >= MY_DBL_MIN) || (nv <= -MY_DBL_MIN)) {
10765 /* It's within the range that a double can represent */
10766 #if defined(DBL_MAX) && !defined(DBL_MIN)
10767 if ((nv >= ((long double)1/DBL_MAX)) ||
10768 (nv <= (-(long double)1/DBL_MAX)))
10770 fix_ldbl_sprintf_bug = TRUE;
10773 if (fix_ldbl_sprintf_bug == TRUE) {
10783 # undef MY_DBL_MAX_BUG
10786 #endif /* HAS_LDBL_SPRINTF_BUG */
10788 need += 20; /* fudge factor */
10789 if (PL_efloatsize < need) {
10790 Safefree(PL_efloatbuf);
10791 PL_efloatsize = need + 20; /* more fudge */
10792 Newx(PL_efloatbuf, PL_efloatsize, char);
10793 PL_efloatbuf[0] = '\0';
10796 if ( !(width || left || plus || alt) && fill != '0'
10797 && has_precis && intsize != 'q' ) { /* Shortcuts */
10798 /* See earlier comment about buggy Gconvert when digits,
10800 if ( c == 'g' && precis) {
10801 Gconvert((NV)nv, (int)precis, 0, PL_efloatbuf);
10802 /* May return an empty string for digits==0 */
10803 if (*PL_efloatbuf) {
10804 elen = strlen(PL_efloatbuf);
10805 goto float_converted;
10807 } else if ( c == 'f' && !precis) {
10808 if ((eptr = F0convert(nv, ebuf + sizeof ebuf, &elen)))
10813 char *ptr = ebuf + sizeof ebuf;
10816 /* FIXME: what to do if HAS_LONG_DOUBLE but not PERL_PRIfldbl? */
10817 #if defined(HAS_LONG_DOUBLE) && defined(PERL_PRIfldbl)
10818 if (intsize == 'q') {
10819 /* Copy the one or more characters in a long double
10820 * format before the 'base' ([efgEFG]) character to
10821 * the format string. */
10822 static char const prifldbl[] = PERL_PRIfldbl;
10823 char const *p = prifldbl + sizeof(prifldbl) - 3;
10824 while (p >= prifldbl) { *--ptr = *p--; }
10829 do { *--ptr = '0' + (base % 10); } while (base /= 10);
10834 do { *--ptr = '0' + (base % 10); } while (base /= 10);
10846 /* No taint. Otherwise we are in the strange situation
10847 * where printf() taints but print($float) doesn't.
10849 #if defined(HAS_LONG_DOUBLE)
10850 elen = ((intsize == 'q')
10851 ? my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, nv)
10852 : my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, (double)nv));
10854 elen = my_sprintf(PL_efloatbuf, ptr, nv);
10858 eptr = PL_efloatbuf;
10866 i = SvCUR(sv) - origlen;
10869 case 'c': *(va_arg(*args, char*)) = i; break;
10870 case 'h': *(va_arg(*args, short*)) = i; break;
10871 default: *(va_arg(*args, int*)) = i; break;
10872 case 'l': *(va_arg(*args, long*)) = i; break;
10873 case 'V': *(va_arg(*args, IV*)) = i; break;
10874 case 'z': *(va_arg(*args, SSize_t*)) = i; break;
10875 case 't': *(va_arg(*args, ptrdiff_t*)) = i; break;
10877 case 'j': *(va_arg(*args, intmax_t*)) = i; break;
10881 *(va_arg(*args, Quad_t*)) = i; break;
10888 sv_setuv_mg(argsv, (UV)i);
10889 continue; /* not "break" */
10896 && (PL_op->op_type == OP_PRTF || PL_op->op_type == OP_SPRINTF)
10897 && ckWARN(WARN_PRINTF))
10899 SV * const msg = sv_newmortal();
10900 Perl_sv_setpvf(aTHX_ msg, "Invalid conversion in %sprintf: ",
10901 (PL_op->op_type == OP_PRTF) ? "" : "s");
10902 if (fmtstart < patend) {
10903 const char * const fmtend = q < patend ? q : patend;
10905 sv_catpvs(msg, "\"%");
10906 for (f = fmtstart; f < fmtend; f++) {
10908 sv_catpvn(msg, f, 1);
10910 Perl_sv_catpvf(aTHX_ msg,
10911 "\\%03"UVof, (UV)*f & 0xFF);
10914 sv_catpvs(msg, "\"");
10916 sv_catpvs(msg, "end of string");
10918 Perl_warner(aTHX_ packWARN(WARN_PRINTF), "%"SVf, SVfARG(msg)); /* yes, this is reentrant */
10921 /* output mangled stuff ... */
10927 /* ... right here, because formatting flags should not apply */
10928 SvGROW(sv, SvCUR(sv) + elen + 1);
10930 Copy(eptr, p, elen, char);
10933 SvCUR_set(sv, p - SvPVX_const(sv));
10935 continue; /* not "break" */
10938 if (is_utf8 != has_utf8) {
10941 sv_utf8_upgrade(sv);
10944 const STRLEN old_elen = elen;
10945 SV * const nsv = newSVpvn_flags(eptr, elen, SVs_TEMP);
10946 sv_utf8_upgrade(nsv);
10947 eptr = SvPVX_const(nsv);
10950 if (width) { /* fudge width (can't fudge elen) */
10951 width += elen - old_elen;
10957 have = esignlen + zeros + elen;
10959 Perl_croak_nocontext("%s", PL_memory_wrap);
10961 need = (have > width ? have : width);
10964 if (need >= (((STRLEN)~0) - SvCUR(sv) - dotstrlen - 1))
10965 Perl_croak_nocontext("%s", PL_memory_wrap);
10966 SvGROW(sv, SvCUR(sv) + need + dotstrlen + 1);
10968 if (esignlen && fill == '0') {
10970 for (i = 0; i < (int)esignlen; i++)
10971 *p++ = esignbuf[i];
10973 if (gap && !left) {
10974 memset(p, fill, gap);
10977 if (esignlen && fill != '0') {
10979 for (i = 0; i < (int)esignlen; i++)
10980 *p++ = esignbuf[i];
10984 for (i = zeros; i; i--)
10988 Copy(eptr, p, elen, char);
10992 memset(p, ' ', gap);
10997 Copy(dotstr, p, dotstrlen, char);
11001 vectorize = FALSE; /* done iterating over vecstr */
11008 SvCUR_set(sv, p - SvPVX_const(sv));
11017 /* =========================================================================
11019 =head1 Cloning an interpreter
11021 All the macros and functions in this section are for the private use of
11022 the main function, perl_clone().
11024 The foo_dup() functions make an exact copy of an existing foo thingy.
11025 During the course of a cloning, a hash table is used to map old addresses
11026 to new addresses. The table is created and manipulated with the
11027 ptr_table_* functions.
11031 * =========================================================================*/
11034 #if defined(USE_ITHREADS)
11036 /* XXX Remove this so it doesn't have to go thru the macro and return for nothing */
11037 #ifndef GpREFCNT_inc
11038 # define GpREFCNT_inc(gp) ((gp) ? (++(gp)->gp_refcnt, (gp)) : (GP*)NULL)
11042 /* Certain cases in Perl_ss_dup have been merged, by relying on the fact
11043 that currently av_dup, gv_dup and hv_dup are the same as sv_dup.
11044 If this changes, please unmerge ss_dup.
11045 Likewise, sv_dup_inc_multiple() relies on this fact. */
11046 #define sv_dup_inc_NN(s,t) SvREFCNT_inc_NN(sv_dup_inc(s,t))
11047 #define av_dup(s,t) MUTABLE_AV(sv_dup((const SV *)s,t))
11048 #define av_dup_inc(s,t) MUTABLE_AV(sv_dup_inc((const SV *)s,t))
11049 #define hv_dup(s,t) MUTABLE_HV(sv_dup((const SV *)s,t))
11050 #define hv_dup_inc(s,t) MUTABLE_HV(sv_dup_inc((const SV *)s,t))
11051 #define cv_dup(s,t) MUTABLE_CV(sv_dup((const SV *)s,t))
11052 #define cv_dup_inc(s,t) MUTABLE_CV(sv_dup_inc((const SV *)s,t))
11053 #define io_dup(s,t) MUTABLE_IO(sv_dup((const SV *)s,t))
11054 #define io_dup_inc(s,t) MUTABLE_IO(sv_dup_inc((const SV *)s,t))
11055 #define gv_dup(s,t) MUTABLE_GV(sv_dup((const SV *)s,t))
11056 #define gv_dup_inc(s,t) MUTABLE_GV(sv_dup_inc((const SV *)s,t))
11057 #define SAVEPV(p) ((p) ? savepv(p) : NULL)
11058 #define SAVEPVN(p,n) ((p) ? savepvn(p,n) : NULL)
11060 /* clone a parser */
11063 Perl_parser_dup(pTHX_ const yy_parser *const proto, CLONE_PARAMS *const param)
11067 PERL_ARGS_ASSERT_PARSER_DUP;
11072 /* look for it in the table first */
11073 parser = (yy_parser *)ptr_table_fetch(PL_ptr_table, proto);
11077 /* create anew and remember what it is */
11078 Newxz(parser, 1, yy_parser);
11079 ptr_table_store(PL_ptr_table, proto, parser);
11081 /* XXX these not yet duped */
11082 parser->old_parser = NULL;
11083 parser->stack = NULL;
11085 parser->stack_size = 0;
11086 /* XXX parser->stack->state = 0; */
11088 /* XXX eventually, just Copy() most of the parser struct ? */
11090 parser->lex_brackets = proto->lex_brackets;
11091 parser->lex_casemods = proto->lex_casemods;
11092 parser->lex_brackstack = savepvn(proto->lex_brackstack,
11093 (proto->lex_brackets < 120 ? 120 : proto->lex_brackets));
11094 parser->lex_casestack = savepvn(proto->lex_casestack,
11095 (proto->lex_casemods < 12 ? 12 : proto->lex_casemods));
11096 parser->lex_defer = proto->lex_defer;
11097 parser->lex_dojoin = proto->lex_dojoin;
11098 parser->lex_expect = proto->lex_expect;
11099 parser->lex_formbrack = proto->lex_formbrack;
11100 parser->lex_inpat = proto->lex_inpat;
11101 parser->lex_inwhat = proto->lex_inwhat;
11102 parser->lex_op = proto->lex_op;
11103 parser->lex_repl = sv_dup_inc(proto->lex_repl, param);
11104 parser->lex_starts = proto->lex_starts;
11105 parser->lex_stuff = sv_dup_inc(proto->lex_stuff, param);
11106 parser->multi_close = proto->multi_close;
11107 parser->multi_open = proto->multi_open;
11108 parser->multi_start = proto->multi_start;
11109 parser->multi_end = proto->multi_end;
11110 parser->pending_ident = proto->pending_ident;
11111 parser->preambled = proto->preambled;
11112 parser->sublex_info = proto->sublex_info; /* XXX not quite right */
11113 parser->linestr = sv_dup_inc(proto->linestr, param);
11114 parser->expect = proto->expect;
11115 parser->copline = proto->copline;
11116 parser->last_lop_op = proto->last_lop_op;
11117 parser->lex_state = proto->lex_state;
11118 parser->rsfp = fp_dup(proto->rsfp, '<', param);
11119 /* rsfp_filters entries have fake IoDIRP() */
11120 parser->rsfp_filters= av_dup_inc(proto->rsfp_filters, param);
11121 parser->in_my = proto->in_my;
11122 parser->in_my_stash = hv_dup(proto->in_my_stash, param);
11123 parser->error_count = proto->error_count;
11126 parser->linestr = sv_dup_inc(proto->linestr, param);
11129 char * const ols = SvPVX(proto->linestr);
11130 char * const ls = SvPVX(parser->linestr);
11132 parser->bufptr = ls + (proto->bufptr >= ols ?
11133 proto->bufptr - ols : 0);
11134 parser->oldbufptr = ls + (proto->oldbufptr >= ols ?
11135 proto->oldbufptr - ols : 0);
11136 parser->oldoldbufptr= ls + (proto->oldoldbufptr >= ols ?
11137 proto->oldoldbufptr - ols : 0);
11138 parser->linestart = ls + (proto->linestart >= ols ?
11139 proto->linestart - ols : 0);
11140 parser->last_uni = ls + (proto->last_uni >= ols ?
11141 proto->last_uni - ols : 0);
11142 parser->last_lop = ls + (proto->last_lop >= ols ?
11143 proto->last_lop - ols : 0);
11145 parser->bufend = ls + SvCUR(parser->linestr);
11148 Copy(proto->tokenbuf, parser->tokenbuf, 256, char);
11152 parser->endwhite = proto->endwhite;
11153 parser->faketokens = proto->faketokens;
11154 parser->lasttoke = proto->lasttoke;
11155 parser->nextwhite = proto->nextwhite;
11156 parser->realtokenstart = proto->realtokenstart;
11157 parser->skipwhite = proto->skipwhite;
11158 parser->thisclose = proto->thisclose;
11159 parser->thismad = proto->thismad;
11160 parser->thisopen = proto->thisopen;
11161 parser->thisstuff = proto->thisstuff;
11162 parser->thistoken = proto->thistoken;
11163 parser->thiswhite = proto->thiswhite;
11165 Copy(proto->nexttoke, parser->nexttoke, 5, NEXTTOKE);
11166 parser->curforce = proto->curforce;
11168 Copy(proto->nextval, parser->nextval, 5, YYSTYPE);
11169 Copy(proto->nexttype, parser->nexttype, 5, I32);
11170 parser->nexttoke = proto->nexttoke;
11173 /* XXX should clone saved_curcop here, but we aren't passed
11174 * proto_perl; so do it in perl_clone_using instead */
11180 /* duplicate a file handle */
11183 Perl_fp_dup(pTHX_ PerlIO *const fp, const char type, CLONE_PARAMS *const param)
11187 PERL_ARGS_ASSERT_FP_DUP;
11188 PERL_UNUSED_ARG(type);
11191 return (PerlIO*)NULL;
11193 /* look for it in the table first */
11194 ret = (PerlIO*)ptr_table_fetch(PL_ptr_table, fp);
11198 /* create anew and remember what it is */
11199 ret = PerlIO_fdupopen(aTHX_ fp, param, PERLIO_DUP_CLONE);
11200 ptr_table_store(PL_ptr_table, fp, ret);
11204 /* duplicate a directory handle */
11207 Perl_dirp_dup(pTHX_ DIR *const dp, CLONE_PARAMS *const param)
11213 register const Direntry_t *dirent;
11214 char smallbuf[256];
11220 PERL_UNUSED_CONTEXT;
11221 PERL_ARGS_ASSERT_DIRP_DUP;
11226 /* look for it in the table first */
11227 ret = (DIR*)ptr_table_fetch(PL_ptr_table, dp);
11233 PERL_UNUSED_ARG(param);
11237 /* open the current directory (so we can switch back) */
11238 if (!(pwd = PerlDir_open("."))) return (DIR *)NULL;
11240 /* chdir to our dir handle and open the present working directory */
11241 if (fchdir(my_dirfd(dp)) < 0 || !(ret = PerlDir_open("."))) {
11242 PerlDir_close(pwd);
11243 return (DIR *)NULL;
11245 /* Now we should have two dir handles pointing to the same dir. */
11247 /* Be nice to the calling code and chdir back to where we were. */
11248 fchdir(my_dirfd(pwd)); /* If this fails, then what? */
11250 /* We have no need of the pwd handle any more. */
11251 PerlDir_close(pwd);
11254 # define d_namlen(d) (d)->d_namlen
11256 # define d_namlen(d) strlen((d)->d_name)
11258 /* Iterate once through dp, to get the file name at the current posi-
11259 tion. Then step back. */
11260 pos = PerlDir_tell(dp);
11261 if ((dirent = PerlDir_read(dp))) {
11262 len = d_namlen(dirent);
11263 if (len <= sizeof smallbuf) name = smallbuf;
11264 else Newx(name, len, char);
11265 Move(dirent->d_name, name, len, char);
11267 PerlDir_seek(dp, pos);
11269 /* Iterate through the new dir handle, till we find a file with the
11271 if (!dirent) /* just before the end */
11273 pos = PerlDir_tell(ret);
11274 if (PerlDir_read(ret)) continue; /* not there yet */
11275 PerlDir_seek(ret, pos); /* step back */
11279 const long pos0 = PerlDir_tell(ret);
11281 pos = PerlDir_tell(ret);
11282 if ((dirent = PerlDir_read(ret))) {
11283 if (len == d_namlen(dirent)
11284 && memEQ(name, dirent->d_name, len)) {
11286 PerlDir_seek(ret, pos); /* step back */
11289 /* else we are not there yet; keep iterating */
11291 else { /* This is not meant to happen. The best we can do is
11292 reset the iterator to the beginning. */
11293 PerlDir_seek(ret, pos0);
11300 if (name && name != smallbuf)
11305 ret = win32_dirp_dup(dp, param);
11308 /* pop it in the pointer table */
11310 ptr_table_store(PL_ptr_table, dp, ret);
11315 /* duplicate a typeglob */
11318 Perl_gp_dup(pTHX_ GP *const gp, CLONE_PARAMS *const param)
11322 PERL_ARGS_ASSERT_GP_DUP;
11326 /* look for it in the table first */
11327 ret = (GP*)ptr_table_fetch(PL_ptr_table, gp);
11331 /* create anew and remember what it is */
11333 ptr_table_store(PL_ptr_table, gp, ret);
11336 /* ret->gp_refcnt must be 0 before any other dups are called. We're relying
11337 on Newxz() to do this for us. */
11338 ret->gp_sv = sv_dup_inc(gp->gp_sv, param);
11339 ret->gp_io = io_dup_inc(gp->gp_io, param);
11340 ret->gp_form = cv_dup_inc(gp->gp_form, param);
11341 ret->gp_av = av_dup_inc(gp->gp_av, param);
11342 ret->gp_hv = hv_dup_inc(gp->gp_hv, param);
11343 ret->gp_egv = gv_dup(gp->gp_egv, param);/* GvEGV is not refcounted */
11344 ret->gp_cv = cv_dup_inc(gp->gp_cv, param);
11345 ret->gp_cvgen = gp->gp_cvgen;
11346 ret->gp_line = gp->gp_line;
11347 ret->gp_file_hek = hek_dup(gp->gp_file_hek, param);
11351 /* duplicate a chain of magic */
11354 Perl_mg_dup(pTHX_ MAGIC *mg, CLONE_PARAMS *const param)
11356 MAGIC *mgret = NULL;
11357 MAGIC **mgprev_p = &mgret;
11359 PERL_ARGS_ASSERT_MG_DUP;
11361 for (; mg; mg = mg->mg_moremagic) {
11364 if ((param->flags & CLONEf_JOIN_IN)
11365 && mg->mg_type == PERL_MAGIC_backref)
11366 /* when joining, we let the individual SVs add themselves to
11367 * backref as needed. */
11370 Newx(nmg, 1, MAGIC);
11372 mgprev_p = &(nmg->mg_moremagic);
11374 /* There was a comment "XXX copy dynamic vtable?" but as we don't have
11375 dynamic vtables, I'm not sure why Sarathy wrote it. The comment dates
11376 from the original commit adding Perl_mg_dup() - revision 4538.
11377 Similarly there is the annotation "XXX random ptr?" next to the
11378 assignment to nmg->mg_ptr. */
11381 /* FIXME for plugins
11382 if (nmg->mg_type == PERL_MAGIC_qr) {
11383 nmg->mg_obj = MUTABLE_SV(CALLREGDUPE((REGEXP*)nmg->mg_obj, param));
11387 nmg->mg_obj = (nmg->mg_flags & MGf_REFCOUNTED)
11388 ? nmg->mg_type == PERL_MAGIC_backref
11389 /* The backref AV has its reference
11390 * count deliberately bumped by 1 */
11391 ? SvREFCNT_inc(av_dup_inc((const AV *)
11392 nmg->mg_obj, param))
11393 : sv_dup_inc(nmg->mg_obj, param)
11394 : sv_dup(nmg->mg_obj, param);
11396 if (nmg->mg_ptr && nmg->mg_type != PERL_MAGIC_regex_global) {
11397 if (nmg->mg_len > 0) {
11398 nmg->mg_ptr = SAVEPVN(nmg->mg_ptr, nmg->mg_len);
11399 if (nmg->mg_type == PERL_MAGIC_overload_table &&
11400 AMT_AMAGIC((AMT*)nmg->mg_ptr))
11402 AMT * const namtp = (AMT*)nmg->mg_ptr;
11403 sv_dup_inc_multiple((SV**)(namtp->table),
11404 (SV**)(namtp->table), NofAMmeth, param);
11407 else if (nmg->mg_len == HEf_SVKEY)
11408 nmg->mg_ptr = (char*)sv_dup_inc((const SV *)nmg->mg_ptr, param);
11410 if ((nmg->mg_flags & MGf_DUP) && nmg->mg_virtual && nmg->mg_virtual->svt_dup) {
11411 nmg->mg_virtual->svt_dup(aTHX_ nmg, param);
11417 #endif /* USE_ITHREADS */
11419 struct ptr_tbl_arena {
11420 struct ptr_tbl_arena *next;
11421 struct ptr_tbl_ent array[1023/3]; /* as ptr_tbl_ent has 3 pointers. */
11424 /* create a new pointer-mapping table */
11427 Perl_ptr_table_new(pTHX)
11430 PERL_UNUSED_CONTEXT;
11432 Newx(tbl, 1, PTR_TBL_t);
11433 tbl->tbl_max = 511;
11434 tbl->tbl_items = 0;
11435 tbl->tbl_arena = NULL;
11436 tbl->tbl_arena_next = NULL;
11437 tbl->tbl_arena_end = NULL;
11438 Newxz(tbl->tbl_ary, tbl->tbl_max + 1, PTR_TBL_ENT_t*);
11442 #define PTR_TABLE_HASH(ptr) \
11443 ((PTR2UV(ptr) >> 3) ^ (PTR2UV(ptr) >> (3 + 7)) ^ (PTR2UV(ptr) >> (3 + 17)))
11445 /* map an existing pointer using a table */
11447 STATIC PTR_TBL_ENT_t *
11448 S_ptr_table_find(PTR_TBL_t *const tbl, const void *const sv)
11450 PTR_TBL_ENT_t *tblent;
11451 const UV hash = PTR_TABLE_HASH(sv);
11453 PERL_ARGS_ASSERT_PTR_TABLE_FIND;
11455 tblent = tbl->tbl_ary[hash & tbl->tbl_max];
11456 for (; tblent; tblent = tblent->next) {
11457 if (tblent->oldval == sv)
11464 Perl_ptr_table_fetch(pTHX_ PTR_TBL_t *const tbl, const void *const sv)
11466 PTR_TBL_ENT_t const *const tblent = ptr_table_find(tbl, sv);
11468 PERL_ARGS_ASSERT_PTR_TABLE_FETCH;
11469 PERL_UNUSED_CONTEXT;
11471 return tblent ? tblent->newval : NULL;
11474 /* add a new entry to a pointer-mapping table */
11477 Perl_ptr_table_store(pTHX_ PTR_TBL_t *const tbl, const void *const oldsv, void *const newsv)
11479 PTR_TBL_ENT_t *tblent = ptr_table_find(tbl, oldsv);
11481 PERL_ARGS_ASSERT_PTR_TABLE_STORE;
11482 PERL_UNUSED_CONTEXT;
11485 tblent->newval = newsv;
11487 const UV entry = PTR_TABLE_HASH(oldsv) & tbl->tbl_max;
11489 if (tbl->tbl_arena_next == tbl->tbl_arena_end) {
11490 struct ptr_tbl_arena *new_arena;
11492 Newx(new_arena, 1, struct ptr_tbl_arena);
11493 new_arena->next = tbl->tbl_arena;
11494 tbl->tbl_arena = new_arena;
11495 tbl->tbl_arena_next = new_arena->array;
11496 tbl->tbl_arena_end = new_arena->array
11497 + sizeof(new_arena->array) / sizeof(new_arena->array[0]);
11500 tblent = tbl->tbl_arena_next++;
11502 tblent->oldval = oldsv;
11503 tblent->newval = newsv;
11504 tblent->next = tbl->tbl_ary[entry];
11505 tbl->tbl_ary[entry] = tblent;
11507 if (tblent->next && tbl->tbl_items > tbl->tbl_max)
11508 ptr_table_split(tbl);
11512 /* double the hash bucket size of an existing ptr table */
11515 Perl_ptr_table_split(pTHX_ PTR_TBL_t *const tbl)
11517 PTR_TBL_ENT_t **ary = tbl->tbl_ary;
11518 const UV oldsize = tbl->tbl_max + 1;
11519 UV newsize = oldsize * 2;
11522 PERL_ARGS_ASSERT_PTR_TABLE_SPLIT;
11523 PERL_UNUSED_CONTEXT;
11525 Renew(ary, newsize, PTR_TBL_ENT_t*);
11526 Zero(&ary[oldsize], newsize-oldsize, PTR_TBL_ENT_t*);
11527 tbl->tbl_max = --newsize;
11528 tbl->tbl_ary = ary;
11529 for (i=0; i < oldsize; i++, ary++) {
11530 PTR_TBL_ENT_t **entp = ary;
11531 PTR_TBL_ENT_t *ent = *ary;
11532 PTR_TBL_ENT_t **curentp;
11535 curentp = ary + oldsize;
11537 if ((newsize & PTR_TABLE_HASH(ent->oldval)) != i) {
11539 ent->next = *curentp;
11549 /* remove all the entries from a ptr table */
11550 /* Deprecated - will be removed post 5.14 */
11553 Perl_ptr_table_clear(pTHX_ PTR_TBL_t *const tbl)
11555 if (tbl && tbl->tbl_items) {
11556 struct ptr_tbl_arena *arena = tbl->tbl_arena;
11558 Zero(tbl->tbl_ary, tbl->tbl_max + 1, struct ptr_tbl_ent **);
11561 struct ptr_tbl_arena *next = arena->next;
11567 tbl->tbl_items = 0;
11568 tbl->tbl_arena = NULL;
11569 tbl->tbl_arena_next = NULL;
11570 tbl->tbl_arena_end = NULL;
11574 /* clear and free a ptr table */
11577 Perl_ptr_table_free(pTHX_ PTR_TBL_t *const tbl)
11579 struct ptr_tbl_arena *arena;
11585 arena = tbl->tbl_arena;
11588 struct ptr_tbl_arena *next = arena->next;
11594 Safefree(tbl->tbl_ary);
11598 #if defined(USE_ITHREADS)
11601 Perl_rvpv_dup(pTHX_ SV *const dstr, const SV *const sstr, CLONE_PARAMS *const param)
11603 PERL_ARGS_ASSERT_RVPV_DUP;
11606 if (SvWEAKREF(sstr)) {
11607 SvRV_set(dstr, sv_dup(SvRV_const(sstr), param));
11608 if (param->flags & CLONEf_JOIN_IN) {
11609 /* if joining, we add any back references individually rather
11610 * than copying the whole backref array */
11611 Perl_sv_add_backref(aTHX_ SvRV(dstr), dstr);
11615 SvRV_set(dstr, sv_dup_inc(SvRV_const(sstr), param));
11617 else if (SvPVX_const(sstr)) {
11618 /* Has something there */
11620 /* Normal PV - clone whole allocated space */
11621 SvPV_set(dstr, SAVEPVN(SvPVX_const(sstr), SvLEN(sstr)-1));
11622 if (SvREADONLY(sstr) && SvFAKE(sstr)) {
11623 /* Not that normal - actually sstr is copy on write.
11624 But we are a true, independent SV, so: */
11625 SvREADONLY_off(dstr);
11630 /* Special case - not normally malloced for some reason */
11631 if (isGV_with_GP(sstr)) {
11632 /* Don't need to do anything here. */
11634 else if ((SvREADONLY(sstr) && SvFAKE(sstr))) {
11635 /* A "shared" PV - clone it as "shared" PV */
11637 HEK_KEY(hek_dup(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)),
11641 /* Some other special case - random pointer */
11642 SvPV_set(dstr, (char *) SvPVX_const(sstr));
11647 /* Copy the NULL */
11648 SvPV_set(dstr, NULL);
11652 /* duplicate a list of SVs. source and dest may point to the same memory. */
11654 S_sv_dup_inc_multiple(pTHX_ SV *const *source, SV **dest,
11655 SSize_t items, CLONE_PARAMS *const param)
11657 PERL_ARGS_ASSERT_SV_DUP_INC_MULTIPLE;
11659 while (items-- > 0) {
11660 *dest++ = sv_dup_inc(*source++, param);
11666 /* duplicate an SV of any type (including AV, HV etc) */
11669 S_sv_dup_common(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
11674 PERL_ARGS_ASSERT_SV_DUP_COMMON;
11676 if (SvTYPE(sstr) == (svtype)SVTYPEMASK) {
11677 #ifdef DEBUG_LEAKING_SCALARS_ABORT
11682 /* look for it in the table first */
11683 dstr = MUTABLE_SV(ptr_table_fetch(PL_ptr_table, sstr));
11687 if(param->flags & CLONEf_JOIN_IN) {
11688 /** We are joining here so we don't want do clone
11689 something that is bad **/
11690 if (SvTYPE(sstr) == SVt_PVHV) {
11691 const HEK * const hvname = HvNAME_HEK(sstr);
11693 /** don't clone stashes if they already exist **/
11694 dstr = MUTABLE_SV(gv_stashpvn(HEK_KEY(hvname), HEK_LEN(hvname), 0));
11695 ptr_table_store(PL_ptr_table, sstr, dstr);
11701 /* create anew and remember what it is */
11704 #ifdef DEBUG_LEAKING_SCALARS
11705 dstr->sv_debug_optype = sstr->sv_debug_optype;
11706 dstr->sv_debug_line = sstr->sv_debug_line;
11707 dstr->sv_debug_inpad = sstr->sv_debug_inpad;
11708 dstr->sv_debug_parent = (SV*)sstr;
11709 FREE_SV_DEBUG_FILE(dstr);
11710 dstr->sv_debug_file = savepv(sstr->sv_debug_file);
11713 ptr_table_store(PL_ptr_table, sstr, dstr);
11716 SvFLAGS(dstr) = SvFLAGS(sstr);
11717 SvFLAGS(dstr) &= ~SVf_OOK; /* don't propagate OOK hack */
11718 SvREFCNT(dstr) = 0; /* must be before any other dups! */
11721 if (SvANY(sstr) && PL_watch_pvx && SvPVX_const(sstr) == PL_watch_pvx)
11722 PerlIO_printf(Perl_debug_log, "watch at %p hit, found string \"%s\"\n",
11723 (void*)PL_watch_pvx, SvPVX_const(sstr));
11726 /* don't clone objects whose class has asked us not to */
11727 if (SvOBJECT(sstr) && ! (SvFLAGS(SvSTASH(sstr)) & SVphv_CLONEABLE)) {
11732 switch (SvTYPE(sstr)) {
11734 SvANY(dstr) = NULL;
11737 SvANY(dstr) = (XPVIV*)((char*)&(dstr->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
11739 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
11741 SvIV_set(dstr, SvIVX(sstr));
11745 SvANY(dstr) = new_XNV();
11746 SvNV_set(dstr, SvNVX(sstr));
11748 /* case SVt_BIND: */
11751 /* These are all the types that need complex bodies allocating. */
11753 const svtype sv_type = SvTYPE(sstr);
11754 const struct body_details *const sv_type_details
11755 = bodies_by_type + sv_type;
11759 Perl_croak(aTHX_ "Bizarre SvTYPE [%" IVdf "]", (IV)SvTYPE(sstr));
11774 assert(sv_type_details->body_size);
11775 if (sv_type_details->arena) {
11776 new_body_inline(new_body, sv_type);
11778 = (void*)((char*)new_body - sv_type_details->offset);
11780 new_body = new_NOARENA(sv_type_details);
11784 SvANY(dstr) = new_body;
11787 Copy(((char*)SvANY(sstr)) + sv_type_details->offset,
11788 ((char*)SvANY(dstr)) + sv_type_details->offset,
11789 sv_type_details->copy, char);
11791 Copy(((char*)SvANY(sstr)),
11792 ((char*)SvANY(dstr)),
11793 sv_type_details->body_size + sv_type_details->offset, char);
11796 if (sv_type != SVt_PVAV && sv_type != SVt_PVHV
11797 && !isGV_with_GP(dstr)
11798 && !(sv_type == SVt_PVIO && !(IoFLAGS(dstr) & IOf_FAKE_DIRP)))
11799 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
11801 /* The Copy above means that all the source (unduplicated) pointers
11802 are now in the destination. We can check the flags and the
11803 pointers in either, but it's possible that there's less cache
11804 missing by always going for the destination.
11805 FIXME - instrument and check that assumption */
11806 if (sv_type >= SVt_PVMG) {
11807 if ((sv_type == SVt_PVMG) && SvPAD_OUR(dstr)) {
11808 SvOURSTASH_set(dstr, hv_dup_inc(SvOURSTASH(dstr), param));
11809 } else if (SvMAGIC(dstr))
11810 SvMAGIC_set(dstr, mg_dup(SvMAGIC(dstr), param));
11812 SvSTASH_set(dstr, hv_dup_inc(SvSTASH(dstr), param));
11815 /* The cast silences a GCC warning about unhandled types. */
11816 switch ((int)sv_type) {
11826 /* FIXME for plugins */
11827 re_dup_guts((REGEXP*) sstr, (REGEXP*) dstr, param);
11830 /* XXX LvTARGOFF sometimes holds PMOP* when DEBUGGING */
11831 if (LvTYPE(dstr) == 't') /* for tie: unrefcnted fake (SV**) */
11832 LvTARG(dstr) = dstr;
11833 else if (LvTYPE(dstr) == 'T') /* for tie: fake HE */
11834 LvTARG(dstr) = MUTABLE_SV(he_dup((HE*)LvTARG(dstr), 0, param));
11836 LvTARG(dstr) = sv_dup_inc(LvTARG(dstr), param);
11838 /* non-GP case already handled above */
11839 if(isGV_with_GP(sstr)) {
11840 GvNAME_HEK(dstr) = hek_dup(GvNAME_HEK(dstr), param);
11841 /* Don't call sv_add_backref here as it's going to be
11842 created as part of the magic cloning of the symbol
11843 table--unless this is during a join and the stash
11844 is not actually being cloned. */
11845 /* Danger Will Robinson - GvGP(dstr) isn't initialised
11846 at the point of this comment. */
11847 GvSTASH(dstr) = hv_dup(GvSTASH(dstr), param);
11848 if (param->flags & CLONEf_JOIN_IN)
11849 Perl_sv_add_backref(aTHX_ MUTABLE_SV(GvSTASH(dstr)), dstr);
11850 GvGP_set(dstr, gp_dup(GvGP(sstr), param));
11851 (void)GpREFCNT_inc(GvGP(dstr));
11855 /* PL_parser->rsfp_filters entries have fake IoDIRP() */
11856 if(IoFLAGS(dstr) & IOf_FAKE_DIRP) {
11857 /* I have no idea why fake dirp (rsfps)
11858 should be treated differently but otherwise
11859 we end up with leaks -- sky*/
11860 IoTOP_GV(dstr) = gv_dup_inc(IoTOP_GV(dstr), param);
11861 IoFMT_GV(dstr) = gv_dup_inc(IoFMT_GV(dstr), param);
11862 IoBOTTOM_GV(dstr) = gv_dup_inc(IoBOTTOM_GV(dstr), param);
11864 IoTOP_GV(dstr) = gv_dup(IoTOP_GV(dstr), param);
11865 IoFMT_GV(dstr) = gv_dup(IoFMT_GV(dstr), param);
11866 IoBOTTOM_GV(dstr) = gv_dup(IoBOTTOM_GV(dstr), param);
11867 if (IoDIRP(dstr)) {
11868 IoDIRP(dstr) = dirp_dup(IoDIRP(dstr), param);
11871 /* IoDIRP(dstr) is already a copy of IoDIRP(sstr) */
11873 IoIFP(dstr) = fp_dup(IoIFP(sstr), IoTYPE(dstr), param);
11875 if (IoOFP(dstr) == IoIFP(sstr))
11876 IoOFP(dstr) = IoIFP(dstr);
11878 IoOFP(dstr) = fp_dup(IoOFP(dstr), IoTYPE(dstr), param);
11879 IoTOP_NAME(dstr) = SAVEPV(IoTOP_NAME(dstr));
11880 IoFMT_NAME(dstr) = SAVEPV(IoFMT_NAME(dstr));
11881 IoBOTTOM_NAME(dstr) = SAVEPV(IoBOTTOM_NAME(dstr));
11884 /* avoid cloning an empty array */
11885 if (AvARRAY((const AV *)sstr) && AvFILLp((const AV *)sstr) >= 0) {
11886 SV **dst_ary, **src_ary;
11887 SSize_t items = AvFILLp((const AV *)sstr) + 1;
11889 src_ary = AvARRAY((const AV *)sstr);
11890 Newxz(dst_ary, AvMAX((const AV *)sstr)+1, SV*);
11891 ptr_table_store(PL_ptr_table, src_ary, dst_ary);
11892 AvARRAY(MUTABLE_AV(dstr)) = dst_ary;
11893 AvALLOC((const AV *)dstr) = dst_ary;
11894 if (AvREAL((const AV *)sstr)) {
11895 dst_ary = sv_dup_inc_multiple(src_ary, dst_ary, items,
11899 while (items-- > 0)
11900 *dst_ary++ = sv_dup(*src_ary++, param);
11902 items = AvMAX((const AV *)sstr) - AvFILLp((const AV *)sstr);
11903 while (items-- > 0) {
11904 *dst_ary++ = &PL_sv_undef;
11908 AvARRAY(MUTABLE_AV(dstr)) = NULL;
11909 AvALLOC((const AV *)dstr) = (SV**)NULL;
11910 AvMAX( (const AV *)dstr) = -1;
11911 AvFILLp((const AV *)dstr) = -1;
11915 if (HvARRAY((const HV *)sstr)) {
11917 const bool sharekeys = !!HvSHAREKEYS(sstr);
11918 XPVHV * const dxhv = (XPVHV*)SvANY(dstr);
11919 XPVHV * const sxhv = (XPVHV*)SvANY(sstr);
11921 Newx(darray, PERL_HV_ARRAY_ALLOC_BYTES(dxhv->xhv_max+1)
11922 + (SvOOK(sstr) ? sizeof(struct xpvhv_aux) : 0),
11924 HvARRAY(dstr) = (HE**)darray;
11925 while (i <= sxhv->xhv_max) {
11926 const HE * const source = HvARRAY(sstr)[i];
11927 HvARRAY(dstr)[i] = source
11928 ? he_dup(source, sharekeys, param) : 0;
11932 const struct xpvhv_aux * const saux = HvAUX(sstr);
11933 struct xpvhv_aux * const daux = HvAUX(dstr);
11934 /* This flag isn't copied. */
11935 /* SvOOK_on(hv) attacks the IV flags. */
11936 SvFLAGS(dstr) |= SVf_OOK;
11938 if (saux->xhv_name_count) {
11939 HEK ** const sname = saux->xhv_name_u.xhvnameu_names;
11941 = saux->xhv_name_count < 0
11942 ? -saux->xhv_name_count
11943 : saux->xhv_name_count;
11944 HEK **shekp = sname + count;
11946 Newx(daux->xhv_name_u.xhvnameu_names, count, HEK *);
11947 dhekp = daux->xhv_name_u.xhvnameu_names + count;
11948 while (shekp-- > sname) {
11950 *dhekp = hek_dup(*shekp, param);
11954 daux->xhv_name_u.xhvnameu_name
11955 = hek_dup(saux->xhv_name_u.xhvnameu_name,
11958 daux->xhv_name_count = saux->xhv_name_count;
11960 daux->xhv_riter = saux->xhv_riter;
11961 daux->xhv_eiter = saux->xhv_eiter
11962 ? he_dup(saux->xhv_eiter,
11963 cBOOL(HvSHAREKEYS(sstr)), param) : 0;
11964 /* backref array needs refcnt=2; see sv_add_backref */
11965 daux->xhv_backreferences =
11966 (param->flags & CLONEf_JOIN_IN)
11967 /* when joining, we let the individual GVs and
11968 * CVs add themselves to backref as
11969 * needed. This avoids pulling in stuff
11970 * that isn't required, and simplifies the
11971 * case where stashes aren't cloned back
11972 * if they already exist in the parent
11975 : saux->xhv_backreferences
11976 ? (SvTYPE(saux->xhv_backreferences) == SVt_PVAV)
11977 ? MUTABLE_AV(SvREFCNT_inc(
11978 sv_dup_inc((const SV *)
11979 saux->xhv_backreferences, param)))
11980 : MUTABLE_AV(sv_dup((const SV *)
11981 saux->xhv_backreferences, param))
11984 daux->xhv_mro_meta = saux->xhv_mro_meta
11985 ? mro_meta_dup(saux->xhv_mro_meta, param)
11988 /* Record stashes for possible cloning in Perl_clone(). */
11990 av_push(param->stashes, dstr);
11994 HvARRAY(MUTABLE_HV(dstr)) = NULL;
11997 if (!(param->flags & CLONEf_COPY_STACKS)) {
12002 /* NOTE: not refcounted */
12003 SvANY(MUTABLE_CV(dstr))->xcv_stash =
12004 hv_dup(CvSTASH(dstr), param);
12005 if ((param->flags & CLONEf_JOIN_IN) && CvSTASH(dstr))
12006 Perl_sv_add_backref(aTHX_ MUTABLE_SV(CvSTASH(dstr)), dstr);
12007 if (!CvISXSUB(dstr)) {
12009 CvROOT(dstr) = OpREFCNT_inc(CvROOT(dstr));
12011 } else if (CvCONST(dstr)) {
12012 CvXSUBANY(dstr).any_ptr =
12013 sv_dup_inc((const SV *)CvXSUBANY(dstr).any_ptr, param);
12015 if (CvDYNFILE(dstr)) CvFILE(dstr) = SAVEPV(CvFILE(dstr));
12016 /* don't dup if copying back - CvGV isn't refcounted, so the
12017 * duped GV may never be freed. A bit of a hack! DAPM */
12018 SvANY(MUTABLE_CV(dstr))->xcv_gv =
12020 ? gv_dup_inc(CvGV(sstr), param)
12021 : (param->flags & CLONEf_JOIN_IN)
12023 : gv_dup(CvGV(sstr), param);
12025 CvPADLIST(dstr) = padlist_dup(CvPADLIST(sstr), param);
12027 CvWEAKOUTSIDE(sstr)
12028 ? cv_dup( CvOUTSIDE(dstr), param)
12029 : cv_dup_inc(CvOUTSIDE(dstr), param);
12035 if (SvOBJECT(dstr) && SvTYPE(dstr) != SVt_PVIO)
12042 Perl_sv_dup_inc(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
12044 PERL_ARGS_ASSERT_SV_DUP_INC;
12045 return sstr ? SvREFCNT_inc(sv_dup_common(sstr, param)) : NULL;
12049 Perl_sv_dup(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
12051 SV *dstr = sstr ? sv_dup_common(sstr, param) : NULL;
12052 PERL_ARGS_ASSERT_SV_DUP;
12054 /* Track every SV that (at least initially) had a reference count of 0.
12055 We need to do this by holding an actual reference to it in this array.
12056 If we attempt to cheat, turn AvREAL_off(), and store only pointers
12057 (akin to the stashes hash, and the perl stack), we come unstuck if
12058 a weak reference (or other SV legitimately SvREFCNT() == 0 for this
12059 thread) is manipulated in a CLONE method, because CLONE runs before the
12060 unreferenced array is walked to find SVs still with SvREFCNT() == 0
12061 (and fix things up by giving each a reference via the temps stack).
12062 Instead, during CLONE, if the 0-referenced SV has SvREFCNT_inc() and
12063 then SvREFCNT_dec(), it will be cleaned up (and added to the free list)
12064 before the walk of unreferenced happens and a reference to that is SV
12065 added to the temps stack. At which point we have the same SV considered
12066 to be in use, and free to be re-used. Not good.
12068 if (dstr && !(param->flags & CLONEf_COPY_STACKS) && !SvREFCNT(dstr)) {
12069 assert(param->unreferenced);
12070 av_push(param->unreferenced, SvREFCNT_inc(dstr));
12076 /* duplicate a context */
12079 Perl_cx_dup(pTHX_ PERL_CONTEXT *cxs, I32 ix, I32 max, CLONE_PARAMS* param)
12081 PERL_CONTEXT *ncxs;
12083 PERL_ARGS_ASSERT_CX_DUP;
12086 return (PERL_CONTEXT*)NULL;
12088 /* look for it in the table first */
12089 ncxs = (PERL_CONTEXT*)ptr_table_fetch(PL_ptr_table, cxs);
12093 /* create anew and remember what it is */
12094 Newx(ncxs, max + 1, PERL_CONTEXT);
12095 ptr_table_store(PL_ptr_table, cxs, ncxs);
12096 Copy(cxs, ncxs, max + 1, PERL_CONTEXT);
12099 PERL_CONTEXT * const ncx = &ncxs[ix];
12100 if (CxTYPE(ncx) == CXt_SUBST) {
12101 Perl_croak(aTHX_ "Cloning substitution context is unimplemented");
12104 switch (CxTYPE(ncx)) {
12106 ncx->blk_sub.cv = (ncx->blk_sub.olddepth == 0
12107 ? cv_dup_inc(ncx->blk_sub.cv, param)
12108 : cv_dup(ncx->blk_sub.cv,param));
12109 ncx->blk_sub.argarray = (CxHASARGS(ncx)
12110 ? av_dup_inc(ncx->blk_sub.argarray,
12113 ncx->blk_sub.savearray = av_dup_inc(ncx->blk_sub.savearray,
12115 ncx->blk_sub.oldcomppad = (PAD*)ptr_table_fetch(PL_ptr_table,
12116 ncx->blk_sub.oldcomppad);
12119 ncx->blk_eval.old_namesv = sv_dup_inc(ncx->blk_eval.old_namesv,
12121 ncx->blk_eval.cur_text = sv_dup(ncx->blk_eval.cur_text, param);
12123 case CXt_LOOP_LAZYSV:
12124 ncx->blk_loop.state_u.lazysv.end
12125 = sv_dup_inc(ncx->blk_loop.state_u.lazysv.end, param);
12126 /* We are taking advantage of av_dup_inc and sv_dup_inc
12127 actually being the same function, and order equivalence of
12129 We can assert the later [but only at run time :-(] */
12130 assert ((void *) &ncx->blk_loop.state_u.ary.ary ==
12131 (void *) &ncx->blk_loop.state_u.lazysv.cur);
12133 ncx->blk_loop.state_u.ary.ary
12134 = av_dup_inc(ncx->blk_loop.state_u.ary.ary, param);
12135 case CXt_LOOP_LAZYIV:
12136 case CXt_LOOP_PLAIN:
12137 if (CxPADLOOP(ncx)) {
12138 ncx->blk_loop.itervar_u.oldcomppad
12139 = (PAD*)ptr_table_fetch(PL_ptr_table,
12140 ncx->blk_loop.itervar_u.oldcomppad);
12142 ncx->blk_loop.itervar_u.gv
12143 = gv_dup((const GV *)ncx->blk_loop.itervar_u.gv,
12148 ncx->blk_format.cv = cv_dup(ncx->blk_format.cv, param);
12149 ncx->blk_format.gv = gv_dup(ncx->blk_format.gv, param);
12150 ncx->blk_format.dfoutgv = gv_dup_inc(ncx->blk_format.dfoutgv,
12163 /* duplicate a stack info structure */
12166 Perl_si_dup(pTHX_ PERL_SI *si, CLONE_PARAMS* param)
12170 PERL_ARGS_ASSERT_SI_DUP;
12173 return (PERL_SI*)NULL;
12175 /* look for it in the table first */
12176 nsi = (PERL_SI*)ptr_table_fetch(PL_ptr_table, si);
12180 /* create anew and remember what it is */
12181 Newxz(nsi, 1, PERL_SI);
12182 ptr_table_store(PL_ptr_table, si, nsi);
12184 nsi->si_stack = av_dup_inc(si->si_stack, param);
12185 nsi->si_cxix = si->si_cxix;
12186 nsi->si_cxmax = si->si_cxmax;
12187 nsi->si_cxstack = cx_dup(si->si_cxstack, si->si_cxix, si->si_cxmax, param);
12188 nsi->si_type = si->si_type;
12189 nsi->si_prev = si_dup(si->si_prev, param);
12190 nsi->si_next = si_dup(si->si_next, param);
12191 nsi->si_markoff = si->si_markoff;
12196 #define POPINT(ss,ix) ((ss)[--(ix)].any_i32)
12197 #define TOPINT(ss,ix) ((ss)[ix].any_i32)
12198 #define POPLONG(ss,ix) ((ss)[--(ix)].any_long)
12199 #define TOPLONG(ss,ix) ((ss)[ix].any_long)
12200 #define POPIV(ss,ix) ((ss)[--(ix)].any_iv)
12201 #define TOPIV(ss,ix) ((ss)[ix].any_iv)
12202 #define POPUV(ss,ix) ((ss)[--(ix)].any_uv)
12203 #define TOPUV(ss,ix) ((ss)[ix].any_uv)
12204 #define POPBOOL(ss,ix) ((ss)[--(ix)].any_bool)
12205 #define TOPBOOL(ss,ix) ((ss)[ix].any_bool)
12206 #define POPPTR(ss,ix) ((ss)[--(ix)].any_ptr)
12207 #define TOPPTR(ss,ix) ((ss)[ix].any_ptr)
12208 #define POPDPTR(ss,ix) ((ss)[--(ix)].any_dptr)
12209 #define TOPDPTR(ss,ix) ((ss)[ix].any_dptr)
12210 #define POPDXPTR(ss,ix) ((ss)[--(ix)].any_dxptr)
12211 #define TOPDXPTR(ss,ix) ((ss)[ix].any_dxptr)
12214 #define pv_dup_inc(p) SAVEPV(p)
12215 #define pv_dup(p) SAVEPV(p)
12216 #define svp_dup_inc(p,pp) any_dup(p,pp)
12218 /* map any object to the new equivent - either something in the
12219 * ptr table, or something in the interpreter structure
12223 Perl_any_dup(pTHX_ void *v, const PerlInterpreter *proto_perl)
12227 PERL_ARGS_ASSERT_ANY_DUP;
12230 return (void*)NULL;
12232 /* look for it in the table first */
12233 ret = ptr_table_fetch(PL_ptr_table, v);
12237 /* see if it is part of the interpreter structure */
12238 if (v >= (void*)proto_perl && v < (void*)(proto_perl+1))
12239 ret = (void*)(((char*)aTHX) + (((char*)v) - (char*)proto_perl));
12247 /* duplicate the save stack */
12250 Perl_ss_dup(pTHX_ PerlInterpreter *proto_perl, CLONE_PARAMS* param)
12253 ANY * const ss = proto_perl->Isavestack;
12254 const I32 max = proto_perl->Isavestack_max;
12255 I32 ix = proto_perl->Isavestack_ix;
12268 void (*dptr) (void*);
12269 void (*dxptr) (pTHX_ void*);
12271 PERL_ARGS_ASSERT_SS_DUP;
12273 Newxz(nss, max, ANY);
12276 const UV uv = POPUV(ss,ix);
12277 const U8 type = (U8)uv & SAVE_MASK;
12279 TOPUV(nss,ix) = uv;
12281 case SAVEt_CLEARSV:
12283 case SAVEt_HELEM: /* hash element */
12284 sv = (const SV *)POPPTR(ss,ix);
12285 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12287 case SAVEt_ITEM: /* normal string */
12288 case SAVEt_GVSV: /* scalar slot in GV */
12289 case SAVEt_SV: /* scalar reference */
12290 sv = (const SV *)POPPTR(ss,ix);
12291 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12294 case SAVEt_MORTALIZESV:
12295 sv = (const SV *)POPPTR(ss,ix);
12296 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12298 case SAVEt_SHARED_PVREF: /* char* in shared space */
12299 c = (char*)POPPTR(ss,ix);
12300 TOPPTR(nss,ix) = savesharedpv(c);
12301 ptr = POPPTR(ss,ix);
12302 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12304 case SAVEt_GENERIC_SVREF: /* generic sv */
12305 case SAVEt_SVREF: /* scalar reference */
12306 sv = (const SV *)POPPTR(ss,ix);
12307 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12308 ptr = POPPTR(ss,ix);
12309 TOPPTR(nss,ix) = svp_dup_inc((SV**)ptr, proto_perl);/* XXXXX */
12311 case SAVEt_HV: /* hash reference */
12312 case SAVEt_AV: /* array reference */
12313 sv = (const SV *) POPPTR(ss,ix);
12314 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12316 case SAVEt_COMPPAD:
12318 sv = (const SV *) POPPTR(ss,ix);
12319 TOPPTR(nss,ix) = sv_dup(sv, param);
12321 case SAVEt_INT: /* int reference */
12322 ptr = POPPTR(ss,ix);
12323 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12324 intval = (int)POPINT(ss,ix);
12325 TOPINT(nss,ix) = intval;
12327 case SAVEt_LONG: /* long reference */
12328 ptr = POPPTR(ss,ix);
12329 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12330 longval = (long)POPLONG(ss,ix);
12331 TOPLONG(nss,ix) = longval;
12333 case SAVEt_I32: /* I32 reference */
12334 ptr = POPPTR(ss,ix);
12335 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12337 TOPINT(nss,ix) = i;
12339 case SAVEt_IV: /* IV reference */
12340 ptr = POPPTR(ss,ix);
12341 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12343 TOPIV(nss,ix) = iv;
12345 case SAVEt_HPTR: /* HV* reference */
12346 case SAVEt_APTR: /* AV* reference */
12347 case SAVEt_SPTR: /* SV* reference */
12348 ptr = POPPTR(ss,ix);
12349 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12350 sv = (const SV *)POPPTR(ss,ix);
12351 TOPPTR(nss,ix) = sv_dup(sv, param);
12353 case SAVEt_VPTR: /* random* reference */
12354 ptr = POPPTR(ss,ix);
12355 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12357 case SAVEt_INT_SMALL:
12358 case SAVEt_I32_SMALL:
12359 case SAVEt_I16: /* I16 reference */
12360 case SAVEt_I8: /* I8 reference */
12362 ptr = POPPTR(ss,ix);
12363 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12365 case SAVEt_GENERIC_PVREF: /* generic char* */
12366 case SAVEt_PPTR: /* char* reference */
12367 ptr = POPPTR(ss,ix);
12368 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12369 c = (char*)POPPTR(ss,ix);
12370 TOPPTR(nss,ix) = pv_dup(c);
12372 case SAVEt_GP: /* scalar reference */
12373 gp = (GP*)POPPTR(ss,ix);
12374 TOPPTR(nss,ix) = gp = gp_dup(gp, param);
12375 (void)GpREFCNT_inc(gp);
12376 gv = (const GV *)POPPTR(ss,ix);
12377 TOPPTR(nss,ix) = gv_dup_inc(gv, param);
12380 ptr = POPPTR(ss,ix);
12381 if (ptr && (((OP*)ptr)->op_private & OPpREFCOUNTED)) {
12382 /* these are assumed to be refcounted properly */
12384 switch (((OP*)ptr)->op_type) {
12386 case OP_LEAVESUBLV:
12390 case OP_LEAVEWRITE:
12391 TOPPTR(nss,ix) = ptr;
12394 (void) OpREFCNT_inc(o);
12398 TOPPTR(nss,ix) = NULL;
12403 TOPPTR(nss,ix) = NULL;
12405 case SAVEt_FREECOPHH:
12406 ptr = POPPTR(ss,ix);
12407 TOPPTR(nss,ix) = cophh_copy((COPHH *)ptr);
12410 hv = (const HV *)POPPTR(ss,ix);
12411 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
12413 TOPINT(nss,ix) = i;
12416 c = (char*)POPPTR(ss,ix);
12417 TOPPTR(nss,ix) = pv_dup_inc(c);
12419 case SAVEt_STACK_POS: /* Position on Perl stack */
12421 TOPINT(nss,ix) = i;
12423 case SAVEt_DESTRUCTOR:
12424 ptr = POPPTR(ss,ix);
12425 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
12426 dptr = POPDPTR(ss,ix);
12427 TOPDPTR(nss,ix) = DPTR2FPTR(void (*)(void*),
12428 any_dup(FPTR2DPTR(void *, dptr),
12431 case SAVEt_DESTRUCTOR_X:
12432 ptr = POPPTR(ss,ix);
12433 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
12434 dxptr = POPDXPTR(ss,ix);
12435 TOPDXPTR(nss,ix) = DPTR2FPTR(void (*)(pTHX_ void*),
12436 any_dup(FPTR2DPTR(void *, dxptr),
12439 case SAVEt_REGCONTEXT:
12441 ix -= uv >> SAVE_TIGHT_SHIFT;
12443 case SAVEt_AELEM: /* array element */
12444 sv = (const SV *)POPPTR(ss,ix);
12445 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12447 TOPINT(nss,ix) = i;
12448 av = (const AV *)POPPTR(ss,ix);
12449 TOPPTR(nss,ix) = av_dup_inc(av, param);
12452 ptr = POPPTR(ss,ix);
12453 TOPPTR(nss,ix) = ptr;
12456 ptr = POPPTR(ss,ix);
12457 ptr = cophh_copy((COPHH*)ptr);
12458 TOPPTR(nss,ix) = ptr;
12460 TOPINT(nss,ix) = i;
12461 if (i & HINT_LOCALIZE_HH) {
12462 hv = (const HV *)POPPTR(ss,ix);
12463 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
12466 case SAVEt_PADSV_AND_MORTALIZE:
12467 longval = (long)POPLONG(ss,ix);
12468 TOPLONG(nss,ix) = longval;
12469 ptr = POPPTR(ss,ix);
12470 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12471 sv = (const SV *)POPPTR(ss,ix);
12472 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12474 case SAVEt_SET_SVFLAGS:
12476 TOPINT(nss,ix) = i;
12478 TOPINT(nss,ix) = i;
12479 sv = (const SV *)POPPTR(ss,ix);
12480 TOPPTR(nss,ix) = sv_dup(sv, param);
12482 case SAVEt_RE_STATE:
12484 const struct re_save_state *const old_state
12485 = (struct re_save_state *)
12486 (ss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
12487 struct re_save_state *const new_state
12488 = (struct re_save_state *)
12489 (nss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
12491 Copy(old_state, new_state, 1, struct re_save_state);
12492 ix -= SAVESTACK_ALLOC_FOR_RE_SAVE_STATE;
12494 new_state->re_state_bostr
12495 = pv_dup(old_state->re_state_bostr);
12496 new_state->re_state_reginput
12497 = pv_dup(old_state->re_state_reginput);
12498 new_state->re_state_regeol
12499 = pv_dup(old_state->re_state_regeol);
12500 new_state->re_state_regoffs
12501 = (regexp_paren_pair*)
12502 any_dup(old_state->re_state_regoffs, proto_perl);
12503 new_state->re_state_reglastparen
12504 = (U32*) any_dup(old_state->re_state_reglastparen,
12506 new_state->re_state_reglastcloseparen
12507 = (U32*)any_dup(old_state->re_state_reglastcloseparen,
12509 /* XXX This just has to be broken. The old save_re_context
12510 code did SAVEGENERICPV(PL_reg_start_tmp);
12511 PL_reg_start_tmp is char **.
12512 Look above to what the dup code does for
12513 SAVEt_GENERIC_PVREF
12514 It can never have worked.
12515 So this is merely a faithful copy of the exiting bug: */
12516 new_state->re_state_reg_start_tmp
12517 = (char **) pv_dup((char *)
12518 old_state->re_state_reg_start_tmp);
12519 /* I assume that it only ever "worked" because no-one called
12520 (pseudo)fork while the regexp engine had re-entered itself.
12522 #ifdef PERL_OLD_COPY_ON_WRITE
12523 new_state->re_state_nrs
12524 = sv_dup(old_state->re_state_nrs, param);
12526 new_state->re_state_reg_magic
12527 = (MAGIC*) any_dup(old_state->re_state_reg_magic,
12529 new_state->re_state_reg_oldcurpm
12530 = (PMOP*) any_dup(old_state->re_state_reg_oldcurpm,
12532 new_state->re_state_reg_curpm
12533 = (PMOP*) any_dup(old_state->re_state_reg_curpm,
12535 new_state->re_state_reg_oldsaved
12536 = pv_dup(old_state->re_state_reg_oldsaved);
12537 new_state->re_state_reg_poscache
12538 = pv_dup(old_state->re_state_reg_poscache);
12539 new_state->re_state_reg_starttry
12540 = pv_dup(old_state->re_state_reg_starttry);
12543 case SAVEt_COMPILE_WARNINGS:
12544 ptr = POPPTR(ss,ix);
12545 TOPPTR(nss,ix) = DUP_WARNINGS((STRLEN*)ptr);
12548 ptr = POPPTR(ss,ix);
12549 TOPPTR(nss,ix) = parser_dup((const yy_parser*)ptr, param);
12553 "panic: ss_dup inconsistency (%"IVdf")", (IV) type);
12561 /* if sv is a stash, call $class->CLONE_SKIP(), and set the SVphv_CLONEABLE
12562 * flag to the result. This is done for each stash before cloning starts,
12563 * so we know which stashes want their objects cloned */
12566 do_mark_cloneable_stash(pTHX_ SV *const sv)
12568 const HEK * const hvname = HvNAME_HEK((const HV *)sv);
12570 GV* const cloner = gv_fetchmethod_autoload(MUTABLE_HV(sv), "CLONE_SKIP", 0);
12571 SvFLAGS(sv) |= SVphv_CLONEABLE; /* clone objects by default */
12572 if (cloner && GvCV(cloner)) {
12579 mXPUSHs(newSVhek(hvname));
12581 call_sv(MUTABLE_SV(GvCV(cloner)), G_SCALAR);
12588 SvFLAGS(sv) &= ~SVphv_CLONEABLE;
12596 =for apidoc perl_clone
12598 Create and return a new interpreter by cloning the current one.
12600 perl_clone takes these flags as parameters:
12602 CLONEf_COPY_STACKS - is used to, well, copy the stacks also,
12603 without it we only clone the data and zero the stacks,
12604 with it we copy the stacks and the new perl interpreter is
12605 ready to run at the exact same point as the previous one.
12606 The pseudo-fork code uses COPY_STACKS while the
12607 threads->create doesn't.
12609 CLONEf_KEEP_PTR_TABLE
12610 perl_clone keeps a ptr_table with the pointer of the old
12611 variable as a key and the new variable as a value,
12612 this allows it to check if something has been cloned and not
12613 clone it again but rather just use the value and increase the
12614 refcount. If KEEP_PTR_TABLE is not set then perl_clone will kill
12615 the ptr_table using the function
12616 C<ptr_table_free(PL_ptr_table); PL_ptr_table = NULL;>,
12617 reason to keep it around is if you want to dup some of your own
12618 variable who are outside the graph perl scans, example of this
12619 code is in threads.xs create
12622 This is a win32 thing, it is ignored on unix, it tells perls
12623 win32host code (which is c++) to clone itself, this is needed on
12624 win32 if you want to run two threads at the same time,
12625 if you just want to do some stuff in a separate perl interpreter
12626 and then throw it away and return to the original one,
12627 you don't need to do anything.
12632 /* XXX the above needs expanding by someone who actually understands it ! */
12633 EXTERN_C PerlInterpreter *
12634 perl_clone_host(PerlInterpreter* proto_perl, UV flags);
12637 perl_clone(PerlInterpreter *proto_perl, UV flags)
12640 #ifdef PERL_IMPLICIT_SYS
12642 PERL_ARGS_ASSERT_PERL_CLONE;
12644 /* perlhost.h so we need to call into it
12645 to clone the host, CPerlHost should have a c interface, sky */
12647 if (flags & CLONEf_CLONE_HOST) {
12648 return perl_clone_host(proto_perl,flags);
12650 return perl_clone_using(proto_perl, flags,
12652 proto_perl->IMemShared,
12653 proto_perl->IMemParse,
12655 proto_perl->IStdIO,
12659 proto_perl->IProc);
12663 perl_clone_using(PerlInterpreter *proto_perl, UV flags,
12664 struct IPerlMem* ipM, struct IPerlMem* ipMS,
12665 struct IPerlMem* ipMP, struct IPerlEnv* ipE,
12666 struct IPerlStdIO* ipStd, struct IPerlLIO* ipLIO,
12667 struct IPerlDir* ipD, struct IPerlSock* ipS,
12668 struct IPerlProc* ipP)
12670 /* XXX many of the string copies here can be optimized if they're
12671 * constants; they need to be allocated as common memory and just
12672 * their pointers copied. */
12675 CLONE_PARAMS clone_params;
12676 CLONE_PARAMS* const param = &clone_params;
12678 PerlInterpreter * const my_perl = (PerlInterpreter*)(*ipM->pMalloc)(ipM, sizeof(PerlInterpreter));
12680 PERL_ARGS_ASSERT_PERL_CLONE_USING;
12681 #else /* !PERL_IMPLICIT_SYS */
12683 CLONE_PARAMS clone_params;
12684 CLONE_PARAMS* param = &clone_params;
12685 PerlInterpreter * const my_perl = (PerlInterpreter*)PerlMem_malloc(sizeof(PerlInterpreter));
12687 PERL_ARGS_ASSERT_PERL_CLONE;
12688 #endif /* PERL_IMPLICIT_SYS */
12690 /* for each stash, determine whether its objects should be cloned */
12691 S_visit(proto_perl, do_mark_cloneable_stash, SVt_PVHV, SVTYPEMASK);
12692 PERL_SET_THX(my_perl);
12695 PoisonNew(my_perl, 1, PerlInterpreter);
12698 PL_defstash = NULL; /* may be used by perl malloc() */
12701 PL_scopestack_name = 0;
12703 PL_savestack_ix = 0;
12704 PL_savestack_max = -1;
12705 PL_sig_pending = 0;
12707 Zero(&PL_debug_pad, 1, struct perl_debug_pad);
12708 # ifdef DEBUG_LEAKING_SCALARS
12709 PL_sv_serial = (((UV)my_perl >> 2) & 0xfff) * 1000000;
12711 #else /* !DEBUGGING */
12712 Zero(my_perl, 1, PerlInterpreter);
12713 #endif /* DEBUGGING */
12715 #ifdef PERL_IMPLICIT_SYS
12716 /* host pointers */
12718 PL_MemShared = ipMS;
12719 PL_MemParse = ipMP;
12726 #endif /* PERL_IMPLICIT_SYS */
12728 param->flags = flags;
12729 /* Nothing in the core code uses this, but we make it available to
12730 extensions (using mg_dup). */
12731 param->proto_perl = proto_perl;
12732 /* Likely nothing will use this, but it is initialised to be consistent
12733 with Perl_clone_params_new(). */
12734 param->new_perl = my_perl;
12735 param->unreferenced = NULL;
12737 INIT_TRACK_MEMPOOL(my_perl->Imemory_debug_header, my_perl);
12739 PL_body_arenas = NULL;
12740 Zero(&PL_body_roots, 1, PL_body_roots);
12743 PL_sv_objcount = 0;
12745 PL_sv_arenaroot = NULL;
12747 PL_debug = proto_perl->Idebug;
12749 PL_hash_seed = proto_perl->Ihash_seed;
12750 PL_rehash_seed = proto_perl->Irehash_seed;
12752 SvANY(&PL_sv_undef) = NULL;
12753 SvREFCNT(&PL_sv_undef) = (~(U32)0)/2;
12754 SvFLAGS(&PL_sv_undef) = SVf_READONLY|SVt_NULL;
12755 SvREFCNT(&PL_sv_no) = (~(U32)0)/2;
12756 SvFLAGS(&PL_sv_no) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
12757 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
12759 SvANY(&PL_sv_yes) = new_XPVNV();
12760 SvREFCNT(&PL_sv_yes) = (~(U32)0)/2;
12761 SvFLAGS(&PL_sv_yes) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
12762 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
12764 /* dbargs array probably holds garbage */
12767 PL_compiling = proto_perl->Icompiling;
12769 #ifdef PERL_DEBUG_READONLY_OPS
12774 /* pseudo environmental stuff */
12775 PL_origargc = proto_perl->Iorigargc;
12776 PL_origargv = proto_perl->Iorigargv;
12778 /* Set tainting stuff before PerlIO_debug can possibly get called */
12779 PL_tainting = proto_perl->Itainting;
12780 PL_taint_warn = proto_perl->Itaint_warn;
12782 PL_minus_c = proto_perl->Iminus_c;
12784 PL_localpatches = proto_perl->Ilocalpatches;
12785 PL_splitstr = proto_perl->Isplitstr;
12786 PL_minus_n = proto_perl->Iminus_n;
12787 PL_minus_p = proto_perl->Iminus_p;
12788 PL_minus_l = proto_perl->Iminus_l;
12789 PL_minus_a = proto_perl->Iminus_a;
12790 PL_minus_E = proto_perl->Iminus_E;
12791 PL_minus_F = proto_perl->Iminus_F;
12792 PL_doswitches = proto_perl->Idoswitches;
12793 PL_dowarn = proto_perl->Idowarn;
12794 PL_sawampersand = proto_perl->Isawampersand;
12795 PL_unsafe = proto_perl->Iunsafe;
12796 PL_perldb = proto_perl->Iperldb;
12797 PL_perl_destruct_level = proto_perl->Iperl_destruct_level;
12798 PL_exit_flags = proto_perl->Iexit_flags;
12800 /* XXX time(&PL_basetime) when asked for? */
12801 PL_basetime = proto_perl->Ibasetime;
12803 PL_maxsysfd = proto_perl->Imaxsysfd;
12804 PL_statusvalue = proto_perl->Istatusvalue;
12806 PL_statusvalue_vms = proto_perl->Istatusvalue_vms;
12808 PL_statusvalue_posix = proto_perl->Istatusvalue_posix;
12811 /* RE engine related */
12812 Zero(&PL_reg_state, 1, struct re_save_state);
12813 PL_reginterp_cnt = 0;
12814 PL_regmatch_slab = NULL;
12816 PL_sub_generation = proto_perl->Isub_generation;
12818 /* funky return mechanisms */
12819 PL_forkprocess = proto_perl->Iforkprocess;
12821 /* internal state */
12822 PL_maxo = proto_perl->Imaxo;
12824 PL_main_start = proto_perl->Imain_start;
12825 PL_eval_root = proto_perl->Ieval_root;
12826 PL_eval_start = proto_perl->Ieval_start;
12828 PL_filemode = proto_perl->Ifilemode;
12829 PL_lastfd = proto_perl->Ilastfd;
12830 PL_oldname = proto_perl->Ioldname; /* XXX not quite right */
12833 PL_gensym = proto_perl->Igensym;
12835 PL_laststatval = proto_perl->Ilaststatval;
12836 PL_laststype = proto_perl->Ilaststype;
12839 PL_profiledata = NULL;
12841 PL_generation = proto_perl->Igeneration;
12843 PL_in_clean_objs = proto_perl->Iin_clean_objs;
12844 PL_in_clean_all = proto_perl->Iin_clean_all;
12846 PL_uid = proto_perl->Iuid;
12847 PL_euid = proto_perl->Ieuid;
12848 PL_gid = proto_perl->Igid;
12849 PL_egid = proto_perl->Iegid;
12850 PL_nomemok = proto_perl->Inomemok;
12851 PL_an = proto_perl->Ian;
12852 PL_evalseq = proto_perl->Ievalseq;
12853 PL_origenviron = proto_perl->Iorigenviron; /* XXX not quite right */
12854 PL_origalen = proto_perl->Iorigalen;
12856 PL_sighandlerp = proto_perl->Isighandlerp;
12858 PL_runops = proto_perl->Irunops;
12860 PL_subline = proto_perl->Isubline;
12863 PL_cryptseen = proto_perl->Icryptseen;
12866 PL_hints = proto_perl->Ihints;
12868 PL_amagic_generation = proto_perl->Iamagic_generation;
12870 #ifdef USE_LOCALE_COLLATE
12871 PL_collation_ix = proto_perl->Icollation_ix;
12872 PL_collation_standard = proto_perl->Icollation_standard;
12873 PL_collxfrm_base = proto_perl->Icollxfrm_base;
12874 PL_collxfrm_mult = proto_perl->Icollxfrm_mult;
12875 #endif /* USE_LOCALE_COLLATE */
12877 #ifdef USE_LOCALE_NUMERIC
12878 PL_numeric_standard = proto_perl->Inumeric_standard;
12879 PL_numeric_local = proto_perl->Inumeric_local;
12880 #endif /* !USE_LOCALE_NUMERIC */
12882 /* Did the locale setup indicate UTF-8? */
12883 PL_utf8locale = proto_perl->Iutf8locale;
12884 /* Unicode features (see perlrun/-C) */
12885 PL_unicode = proto_perl->Iunicode;
12887 /* Pre-5.8 signals control */
12888 PL_signals = proto_perl->Isignals;
12890 /* times() ticks per second */
12891 PL_clocktick = proto_perl->Iclocktick;
12893 /* Recursion stopper for PerlIO_find_layer */
12894 PL_in_load_module = proto_perl->Iin_load_module;
12896 /* sort() routine */
12897 PL_sort_RealCmp = proto_perl->Isort_RealCmp;
12899 /* Not really needed/useful since the reenrant_retint is "volatile",
12900 * but do it for consistency's sake. */
12901 PL_reentrant_retint = proto_perl->Ireentrant_retint;
12903 /* Hooks to shared SVs and locks. */
12904 PL_sharehook = proto_perl->Isharehook;
12905 PL_lockhook = proto_perl->Ilockhook;
12906 PL_unlockhook = proto_perl->Iunlockhook;
12907 PL_threadhook = proto_perl->Ithreadhook;
12908 PL_destroyhook = proto_perl->Idestroyhook;
12909 PL_signalhook = proto_perl->Isignalhook;
12911 #ifdef THREADS_HAVE_PIDS
12912 PL_ppid = proto_perl->Ippid;
12916 PL_last_swash_hv = NULL; /* reinits on demand */
12917 PL_last_swash_klen = 0;
12918 PL_last_swash_key[0]= '\0';
12919 PL_last_swash_tmps = (U8*)NULL;
12920 PL_last_swash_slen = 0;
12922 PL_glob_index = proto_perl->Iglob_index;
12923 PL_srand_called = proto_perl->Isrand_called;
12925 if (flags & CLONEf_COPY_STACKS) {
12926 /* next allocation will be PL_tmps_stack[PL_tmps_ix+1] */
12927 PL_tmps_ix = proto_perl->Itmps_ix;
12928 PL_tmps_max = proto_perl->Itmps_max;
12929 PL_tmps_floor = proto_perl->Itmps_floor;
12931 /* next push_scope()/ENTER sets PL_scopestack[PL_scopestack_ix]
12932 * NOTE: unlike the others! */
12933 PL_scopestack_ix = proto_perl->Iscopestack_ix;
12934 PL_scopestack_max = proto_perl->Iscopestack_max;
12936 /* next SSPUSHFOO() sets PL_savestack[PL_savestack_ix]
12937 * NOTE: unlike the others! */
12938 PL_savestack_ix = proto_perl->Isavestack_ix;
12939 PL_savestack_max = proto_perl->Isavestack_max;
12942 PL_start_env = proto_perl->Istart_env; /* XXXXXX */
12943 PL_top_env = &PL_start_env;
12945 PL_op = proto_perl->Iop;
12948 PL_Xpv = (XPV*)NULL;
12949 my_perl->Ina = proto_perl->Ina;
12951 PL_statbuf = proto_perl->Istatbuf;
12952 PL_statcache = proto_perl->Istatcache;
12955 PL_timesbuf = proto_perl->Itimesbuf;
12958 PL_tainted = proto_perl->Itainted;
12959 PL_curpm = proto_perl->Icurpm; /* XXX No PMOP ref count */
12961 PL_chopset = proto_perl->Ichopset; /* XXX never deallocated */
12963 PL_restartjmpenv = proto_perl->Irestartjmpenv;
12964 PL_restartop = proto_perl->Irestartop;
12965 PL_in_eval = proto_perl->Iin_eval;
12966 PL_delaymagic = proto_perl->Idelaymagic;
12967 PL_phase = proto_perl->Iphase;
12968 PL_localizing = proto_perl->Ilocalizing;
12970 PL_hv_fetch_ent_mh = NULL;
12971 PL_modcount = proto_perl->Imodcount;
12972 PL_lastgotoprobe = NULL;
12973 PL_dumpindent = proto_perl->Idumpindent;
12975 PL_efloatbuf = NULL; /* reinits on demand */
12976 PL_efloatsize = 0; /* reinits on demand */
12980 PL_regdummy = proto_perl->Iregdummy;
12981 PL_colorset = 0; /* reinits PL_colors[] */
12982 /*PL_colors[6] = {0,0,0,0,0,0};*/
12984 /* Pluggable optimizer */
12985 PL_peepp = proto_perl->Ipeepp;
12986 PL_rpeepp = proto_perl->Irpeepp;
12987 /* op_free() hook */
12988 PL_opfreehook = proto_perl->Iopfreehook;
12990 #ifdef USE_REENTRANT_API
12991 /* XXX: things like -Dm will segfault here in perlio, but doing
12992 * PERL_SET_CONTEXT(proto_perl);
12993 * breaks too many other things
12995 Perl_reentrant_init(aTHX);
12998 /* create SV map for pointer relocation */
12999 PL_ptr_table = ptr_table_new();
13001 /* initialize these special pointers as early as possible */
13002 ptr_table_store(PL_ptr_table, &proto_perl->Isv_undef, &PL_sv_undef);
13004 SvANY(&PL_sv_no) = new_XPVNV();
13005 SvPV_set(&PL_sv_no, savepvn(PL_No, 0));
13006 SvCUR_set(&PL_sv_no, 0);
13007 SvLEN_set(&PL_sv_no, 1);
13008 SvIV_set(&PL_sv_no, 0);
13009 SvNV_set(&PL_sv_no, 0);
13010 ptr_table_store(PL_ptr_table, &proto_perl->Isv_no, &PL_sv_no);
13012 SvPV_set(&PL_sv_yes, savepvn(PL_Yes, 1));
13013 SvCUR_set(&PL_sv_yes, 1);
13014 SvLEN_set(&PL_sv_yes, 2);
13015 SvIV_set(&PL_sv_yes, 1);
13016 SvNV_set(&PL_sv_yes, 1);
13017 ptr_table_store(PL_ptr_table, &proto_perl->Isv_yes, &PL_sv_yes);
13019 /* create (a non-shared!) shared string table */
13020 PL_strtab = newHV();
13021 HvSHAREKEYS_off(PL_strtab);
13022 hv_ksplit(PL_strtab, HvTOTALKEYS(proto_perl->Istrtab));
13023 ptr_table_store(PL_ptr_table, proto_perl->Istrtab, PL_strtab);
13025 /* These two PVs will be free'd special way so must set them same way op.c does */
13026 PL_compiling.cop_stashpv = savesharedpv(PL_compiling.cop_stashpv);
13027 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_stashpv, PL_compiling.cop_stashpv);
13029 PL_compiling.cop_file = savesharedpv(PL_compiling.cop_file);
13030 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_file, PL_compiling.cop_file);
13032 ptr_table_store(PL_ptr_table, &proto_perl->Icompiling, &PL_compiling);
13033 PL_compiling.cop_warnings = DUP_WARNINGS(PL_compiling.cop_warnings);
13034 CopHINTHASH_set(&PL_compiling, cophh_copy(CopHINTHASH_get(&PL_compiling)));
13035 PL_curcop = (COP*)any_dup(proto_perl->Icurcop, proto_perl);
13037 param->stashes = newAV(); /* Setup array of objects to call clone on */
13038 /* This makes no difference to the implementation, as it always pushes
13039 and shifts pointers to other SVs without changing their reference
13040 count, with the array becoming empty before it is freed. However, it
13041 makes it conceptually clear what is going on, and will avoid some
13042 work inside av.c, filling slots between AvFILL() and AvMAX() with
13043 &PL_sv_undef, and SvREFCNT_dec()ing those. */
13044 AvREAL_off(param->stashes);
13046 if (!(flags & CLONEf_COPY_STACKS)) {
13047 param->unreferenced = newAV();
13050 #ifdef PERLIO_LAYERS
13051 /* Clone PerlIO tables as soon as we can handle general xx_dup() */
13052 PerlIO_clone(aTHX_ proto_perl, param);
13055 PL_envgv = gv_dup(proto_perl->Ienvgv, param);
13056 PL_incgv = gv_dup(proto_perl->Iincgv, param);
13057 PL_hintgv = gv_dup(proto_perl->Ihintgv, param);
13058 PL_origfilename = SAVEPV(proto_perl->Iorigfilename);
13059 PL_diehook = sv_dup_inc(proto_perl->Idiehook, param);
13060 PL_warnhook = sv_dup_inc(proto_perl->Iwarnhook, param);
13063 PL_patchlevel = sv_dup_inc(proto_perl->Ipatchlevel, param);
13064 PL_apiversion = sv_dup_inc(proto_perl->Iapiversion, param);
13065 PL_inplace = SAVEPV(proto_perl->Iinplace);
13066 PL_e_script = sv_dup_inc(proto_perl->Ie_script, param);
13068 /* magical thingies */
13069 PL_formfeed = sv_dup(proto_perl->Iformfeed, param);
13071 PL_encoding = sv_dup(proto_perl->Iencoding, param);
13073 sv_setpvs(PERL_DEBUG_PAD(0), ""); /* For regex debugging. */
13074 sv_setpvs(PERL_DEBUG_PAD(1), ""); /* ext/re needs these */
13075 sv_setpvs(PERL_DEBUG_PAD(2), ""); /* even without DEBUGGING. */
13078 /* Clone the regex array */
13079 /* ORANGE FIXME for plugins, probably in the SV dup code.
13080 newSViv(PTR2IV(CALLREGDUPE(
13081 INT2PTR(REGEXP *, SvIVX(regex)), param))))
13083 PL_regex_padav = av_dup_inc(proto_perl->Iregex_padav, param);
13084 PL_regex_pad = AvARRAY(PL_regex_padav);
13086 /* shortcuts to various I/O objects */
13087 PL_ofsgv = gv_dup_inc(proto_perl->Iofsgv, param);
13088 PL_stdingv = gv_dup(proto_perl->Istdingv, param);
13089 PL_stderrgv = gv_dup(proto_perl->Istderrgv, param);
13090 PL_defgv = gv_dup(proto_perl->Idefgv, param);
13091 PL_argvgv = gv_dup(proto_perl->Iargvgv, param);
13092 PL_argvoutgv = gv_dup(proto_perl->Iargvoutgv, param);
13093 PL_argvout_stack = av_dup_inc(proto_perl->Iargvout_stack, param);
13095 /* shortcuts to regexp stuff */
13096 PL_replgv = gv_dup(proto_perl->Ireplgv, param);
13098 /* shortcuts to misc objects */
13099 PL_errgv = gv_dup(proto_perl->Ierrgv, param);
13101 /* shortcuts to debugging objects */
13102 PL_DBgv = gv_dup(proto_perl->IDBgv, param);
13103 PL_DBline = gv_dup(proto_perl->IDBline, param);
13104 PL_DBsub = gv_dup(proto_perl->IDBsub, param);
13105 PL_DBsingle = sv_dup(proto_perl->IDBsingle, param);
13106 PL_DBtrace = sv_dup(proto_perl->IDBtrace, param);
13107 PL_DBsignal = sv_dup(proto_perl->IDBsignal, param);
13109 /* symbol tables */
13110 PL_defstash = hv_dup_inc(proto_perl->Idefstash, param);
13111 PL_curstash = hv_dup(proto_perl->Icurstash, param);
13112 PL_debstash = hv_dup(proto_perl->Idebstash, param);
13113 PL_globalstash = hv_dup(proto_perl->Iglobalstash, param);
13114 PL_curstname = sv_dup_inc(proto_perl->Icurstname, param);
13116 PL_beginav = av_dup_inc(proto_perl->Ibeginav, param);
13117 PL_beginav_save = av_dup_inc(proto_perl->Ibeginav_save, param);
13118 PL_checkav_save = av_dup_inc(proto_perl->Icheckav_save, param);
13119 PL_unitcheckav = av_dup_inc(proto_perl->Iunitcheckav, param);
13120 PL_unitcheckav_save = av_dup_inc(proto_perl->Iunitcheckav_save, param);
13121 PL_endav = av_dup_inc(proto_perl->Iendav, param);
13122 PL_checkav = av_dup_inc(proto_perl->Icheckav, param);
13123 PL_initav = av_dup_inc(proto_perl->Iinitav, param);
13125 PL_isarev = hv_dup_inc(proto_perl->Iisarev, param);
13127 /* subprocess state */
13128 PL_fdpid = av_dup_inc(proto_perl->Ifdpid, param);
13130 if (proto_perl->Iop_mask)
13131 PL_op_mask = SAVEPVN(proto_perl->Iop_mask, PL_maxo);
13134 /* PL_asserting = proto_perl->Iasserting; */
13136 /* current interpreter roots */
13137 PL_main_cv = cv_dup_inc(proto_perl->Imain_cv, param);
13139 PL_main_root = OpREFCNT_inc(proto_perl->Imain_root);
13142 /* runtime control stuff */
13143 PL_curcopdb = (COP*)any_dup(proto_perl->Icurcopdb, proto_perl);
13145 PL_preambleav = av_dup_inc(proto_perl->Ipreambleav, param);
13147 PL_ors_sv = sv_dup_inc(proto_perl->Iors_sv, param);
13149 /* interpreter atexit processing */
13150 PL_exitlistlen = proto_perl->Iexitlistlen;
13151 if (PL_exitlistlen) {
13152 Newx(PL_exitlist, PL_exitlistlen, PerlExitListEntry);
13153 Copy(proto_perl->Iexitlist, PL_exitlist, PL_exitlistlen, PerlExitListEntry);
13156 PL_exitlist = (PerlExitListEntry*)NULL;
13158 PL_my_cxt_size = proto_perl->Imy_cxt_size;
13159 if (PL_my_cxt_size) {
13160 Newx(PL_my_cxt_list, PL_my_cxt_size, void *);
13161 Copy(proto_perl->Imy_cxt_list, PL_my_cxt_list, PL_my_cxt_size, void *);
13162 #ifdef PERL_GLOBAL_STRUCT_PRIVATE
13163 Newx(PL_my_cxt_keys, PL_my_cxt_size, const char *);
13164 Copy(proto_perl->Imy_cxt_keys, PL_my_cxt_keys, PL_my_cxt_size, char *);
13168 PL_my_cxt_list = (void**)NULL;
13169 #ifdef PERL_GLOBAL_STRUCT_PRIVATE
13170 PL_my_cxt_keys = (const char**)NULL;
13173 PL_modglobal = hv_dup_inc(proto_perl->Imodglobal, param);
13174 PL_custom_op_names = hv_dup_inc(proto_perl->Icustom_op_names,param);
13175 PL_custom_op_descs = hv_dup_inc(proto_perl->Icustom_op_descs,param);
13176 PL_custom_ops = hv_dup_inc(proto_perl->Icustom_ops, param);
13178 PL_compcv = cv_dup(proto_perl->Icompcv, param);
13180 PAD_CLONE_VARS(proto_perl, param);
13182 #ifdef HAVE_INTERP_INTERN
13183 sys_intern_dup(&proto_perl->Isys_intern, &PL_sys_intern);
13186 PL_DBcv = cv_dup(proto_perl->IDBcv, param);
13188 #ifdef PERL_USES_PL_PIDSTATUS
13189 PL_pidstatus = newHV(); /* XXX flag for cloning? */
13191 PL_osname = SAVEPV(proto_perl->Iosname);
13192 PL_parser = parser_dup(proto_perl->Iparser, param);
13194 /* XXX this only works if the saved cop has already been cloned */
13195 if (proto_perl->Iparser) {
13196 PL_parser->saved_curcop = (COP*)any_dup(
13197 proto_perl->Iparser->saved_curcop,
13201 PL_subname = sv_dup_inc(proto_perl->Isubname, param);
13203 #ifdef USE_LOCALE_COLLATE
13204 PL_collation_name = SAVEPV(proto_perl->Icollation_name);
13205 #endif /* USE_LOCALE_COLLATE */
13207 #ifdef USE_LOCALE_NUMERIC
13208 PL_numeric_name = SAVEPV(proto_perl->Inumeric_name);
13209 PL_numeric_radix_sv = sv_dup_inc(proto_perl->Inumeric_radix_sv, param);
13210 #endif /* !USE_LOCALE_NUMERIC */
13212 /* utf8 character classes */
13213 PL_utf8_alnum = sv_dup_inc(proto_perl->Iutf8_alnum, param);
13214 PL_utf8_alpha = sv_dup_inc(proto_perl->Iutf8_alpha, param);
13215 PL_utf8_space = sv_dup_inc(proto_perl->Iutf8_space, param);
13216 PL_utf8_graph = sv_dup_inc(proto_perl->Iutf8_graph, param);
13217 PL_utf8_digit = sv_dup_inc(proto_perl->Iutf8_digit, param);
13218 PL_utf8_upper = sv_dup_inc(proto_perl->Iutf8_upper, param);
13219 PL_utf8_lower = sv_dup_inc(proto_perl->Iutf8_lower, param);
13220 PL_utf8_print = sv_dup_inc(proto_perl->Iutf8_print, param);
13221 PL_utf8_punct = sv_dup_inc(proto_perl->Iutf8_punct, param);
13222 PL_utf8_xdigit = sv_dup_inc(proto_perl->Iutf8_xdigit, param);
13223 PL_utf8_mark = sv_dup_inc(proto_perl->Iutf8_mark, param);
13224 PL_utf8_X_begin = sv_dup_inc(proto_perl->Iutf8_X_begin, param);
13225 PL_utf8_X_extend = sv_dup_inc(proto_perl->Iutf8_X_extend, param);
13226 PL_utf8_X_prepend = sv_dup_inc(proto_perl->Iutf8_X_prepend, param);
13227 PL_utf8_X_non_hangul = sv_dup_inc(proto_perl->Iutf8_X_non_hangul, param);
13228 PL_utf8_X_L = sv_dup_inc(proto_perl->Iutf8_X_L, param);
13229 PL_utf8_X_LV = sv_dup_inc(proto_perl->Iutf8_X_LV, param);
13230 PL_utf8_X_LVT = sv_dup_inc(proto_perl->Iutf8_X_LVT, param);
13231 PL_utf8_X_T = sv_dup_inc(proto_perl->Iutf8_X_T, param);
13232 PL_utf8_X_V = sv_dup_inc(proto_perl->Iutf8_X_V, param);
13233 PL_utf8_X_LV_LVT_V = sv_dup_inc(proto_perl->Iutf8_X_LV_LVT_V, param);
13234 PL_utf8_toupper = sv_dup_inc(proto_perl->Iutf8_toupper, param);
13235 PL_utf8_totitle = sv_dup_inc(proto_perl->Iutf8_totitle, param);
13236 PL_utf8_tolower = sv_dup_inc(proto_perl->Iutf8_tolower, param);
13237 PL_utf8_tofold = sv_dup_inc(proto_perl->Iutf8_tofold, param);
13238 PL_utf8_idstart = sv_dup_inc(proto_perl->Iutf8_idstart, param);
13239 PL_utf8_xidstart = sv_dup_inc(proto_perl->Iutf8_xidstart, param);
13240 PL_utf8_perl_idstart = sv_dup_inc(proto_perl->Iutf8_perl_idstart, param);
13241 PL_utf8_idcont = sv_dup_inc(proto_perl->Iutf8_idcont, param);
13242 PL_utf8_xidcont = sv_dup_inc(proto_perl->Iutf8_xidcont, param);
13243 PL_utf8_foldable = sv_dup_inc(proto_perl->Iutf8_foldable, param);
13246 if (proto_perl->Ipsig_pend) {
13247 Newxz(PL_psig_pend, SIG_SIZE, int);
13250 PL_psig_pend = (int*)NULL;
13253 if (proto_perl->Ipsig_name) {
13254 Newx(PL_psig_name, 2 * SIG_SIZE, SV*);
13255 sv_dup_inc_multiple(proto_perl->Ipsig_name, PL_psig_name, 2 * SIG_SIZE,
13257 PL_psig_ptr = PL_psig_name + SIG_SIZE;
13260 PL_psig_ptr = (SV**)NULL;
13261 PL_psig_name = (SV**)NULL;
13264 if (flags & CLONEf_COPY_STACKS) {
13265 Newx(PL_tmps_stack, PL_tmps_max, SV*);
13266 sv_dup_inc_multiple(proto_perl->Itmps_stack, PL_tmps_stack,
13267 PL_tmps_ix+1, param);
13269 /* next PUSHMARK() sets *(PL_markstack_ptr+1) */
13270 i = proto_perl->Imarkstack_max - proto_perl->Imarkstack;
13271 Newxz(PL_markstack, i, I32);
13272 PL_markstack_max = PL_markstack + (proto_perl->Imarkstack_max
13273 - proto_perl->Imarkstack);
13274 PL_markstack_ptr = PL_markstack + (proto_perl->Imarkstack_ptr
13275 - proto_perl->Imarkstack);
13276 Copy(proto_perl->Imarkstack, PL_markstack,
13277 PL_markstack_ptr - PL_markstack + 1, I32);
13279 /* next push_scope()/ENTER sets PL_scopestack[PL_scopestack_ix]
13280 * NOTE: unlike the others! */
13281 Newxz(PL_scopestack, PL_scopestack_max, I32);
13282 Copy(proto_perl->Iscopestack, PL_scopestack, PL_scopestack_ix, I32);
13285 Newxz(PL_scopestack_name, PL_scopestack_max, const char *);
13286 Copy(proto_perl->Iscopestack_name, PL_scopestack_name, PL_scopestack_ix, const char *);
13288 /* NOTE: si_dup() looks at PL_markstack */
13289 PL_curstackinfo = si_dup(proto_perl->Icurstackinfo, param);
13291 /* PL_curstack = PL_curstackinfo->si_stack; */
13292 PL_curstack = av_dup(proto_perl->Icurstack, param);
13293 PL_mainstack = av_dup(proto_perl->Imainstack, param);
13295 /* next PUSHs() etc. set *(PL_stack_sp+1) */
13296 PL_stack_base = AvARRAY(PL_curstack);
13297 PL_stack_sp = PL_stack_base + (proto_perl->Istack_sp
13298 - proto_perl->Istack_base);
13299 PL_stack_max = PL_stack_base + AvMAX(PL_curstack);
13301 /*Newxz(PL_savestack, PL_savestack_max, ANY);*/
13302 PL_savestack = ss_dup(proto_perl, param);
13306 ENTER; /* perl_destruct() wants to LEAVE; */
13309 PL_statgv = gv_dup(proto_perl->Istatgv, param);
13310 PL_statname = sv_dup_inc(proto_perl->Istatname, param);
13312 PL_rs = sv_dup_inc(proto_perl->Irs, param);
13313 PL_last_in_gv = gv_dup(proto_perl->Ilast_in_gv, param);
13314 PL_defoutgv = gv_dup_inc(proto_perl->Idefoutgv, param);
13315 PL_toptarget = sv_dup_inc(proto_perl->Itoptarget, param);
13316 PL_bodytarget = sv_dup_inc(proto_perl->Ibodytarget, param);
13317 PL_formtarget = sv_dup(proto_perl->Iformtarget, param);
13319 PL_errors = sv_dup_inc(proto_perl->Ierrors, param);
13321 PL_sortcop = (OP*)any_dup(proto_perl->Isortcop, proto_perl);
13322 PL_sortstash = hv_dup(proto_perl->Isortstash, param);
13323 PL_firstgv = gv_dup(proto_perl->Ifirstgv, param);
13324 PL_secondgv = gv_dup(proto_perl->Isecondgv, param);
13326 PL_stashcache = newHV();
13328 PL_watchaddr = (char **) ptr_table_fetch(PL_ptr_table,
13329 proto_perl->Iwatchaddr);
13330 PL_watchok = PL_watchaddr ? * PL_watchaddr : NULL;
13331 if (PL_debug && PL_watchaddr) {
13332 PerlIO_printf(Perl_debug_log,
13333 "WATCHING: %"UVxf" cloned as %"UVxf" with value %"UVxf"\n",
13334 PTR2UV(proto_perl->Iwatchaddr), PTR2UV(PL_watchaddr),
13335 PTR2UV(PL_watchok));
13338 PL_registered_mros = hv_dup_inc(proto_perl->Iregistered_mros, param);
13339 PL_blockhooks = av_dup_inc(proto_perl->Iblockhooks, param);
13340 PL_utf8_foldclosures = hv_dup_inc(proto_perl->Iutf8_foldclosures, param);
13342 /* Call the ->CLONE method, if it exists, for each of the stashes
13343 identified by sv_dup() above.
13345 while(av_len(param->stashes) != -1) {
13346 HV* const stash = MUTABLE_HV(av_shift(param->stashes));
13347 GV* const cloner = gv_fetchmethod_autoload(stash, "CLONE", 0);
13348 if (cloner && GvCV(cloner)) {
13353 mXPUSHs(newSVhek(HvNAME_HEK(stash)));
13355 call_sv(MUTABLE_SV(GvCV(cloner)), G_DISCARD);
13361 if (!(flags & CLONEf_KEEP_PTR_TABLE)) {
13362 ptr_table_free(PL_ptr_table);
13363 PL_ptr_table = NULL;
13366 if (!(flags & CLONEf_COPY_STACKS)) {
13367 unreferenced_to_tmp_stack(param->unreferenced);
13370 SvREFCNT_dec(param->stashes);
13372 /* orphaned? eg threads->new inside BEGIN or use */
13373 if (PL_compcv && ! SvREFCNT(PL_compcv)) {
13374 SvREFCNT_inc_simple_void(PL_compcv);
13375 SAVEFREESV(PL_compcv);
13382 S_unreferenced_to_tmp_stack(pTHX_ AV *const unreferenced)
13384 PERL_ARGS_ASSERT_UNREFERENCED_TO_TMP_STACK;
13386 if (AvFILLp(unreferenced) > -1) {
13387 SV **svp = AvARRAY(unreferenced);
13388 SV **const last = svp + AvFILLp(unreferenced);
13392 if (SvREFCNT(*svp) == 1)
13394 } while (++svp <= last);
13396 EXTEND_MORTAL(count);
13397 svp = AvARRAY(unreferenced);
13400 if (SvREFCNT(*svp) == 1) {
13401 /* Our reference is the only one to this SV. This means that
13402 in this thread, the scalar effectively has a 0 reference.
13403 That doesn't work (cleanup never happens), so donate our
13404 reference to it onto the save stack. */
13405 PL_tmps_stack[++PL_tmps_ix] = *svp;
13407 /* As an optimisation, because we are already walking the
13408 entire array, instead of above doing either
13409 SvREFCNT_inc(*svp) or *svp = &PL_sv_undef, we can instead
13410 release our reference to the scalar, so that at the end of
13411 the array owns zero references to the scalars it happens to
13412 point to. We are effectively converting the array from
13413 AvREAL() on to AvREAL() off. This saves the av_clear()
13414 (triggered by the SvREFCNT_dec(unreferenced) below) from
13415 walking the array a second time. */
13416 SvREFCNT_dec(*svp);
13419 } while (++svp <= last);
13420 AvREAL_off(unreferenced);
13422 SvREFCNT_dec(unreferenced);
13426 Perl_clone_params_del(CLONE_PARAMS *param)
13428 /* This seemingly funky ordering keeps the build with PERL_GLOBAL_STRUCT
13430 PerlInterpreter *const to = param->new_perl;
13432 PerlInterpreter *const was = PERL_GET_THX;
13434 PERL_ARGS_ASSERT_CLONE_PARAMS_DEL;
13440 SvREFCNT_dec(param->stashes);
13441 if (param->unreferenced)
13442 unreferenced_to_tmp_stack(param->unreferenced);
13452 Perl_clone_params_new(PerlInterpreter *const from, PerlInterpreter *const to)
13455 /* Need to play this game, as newAV() can call safesysmalloc(), and that
13456 does a dTHX; to get the context from thread local storage.
13457 FIXME - under PERL_CORE Newx(), Safefree() and friends should expand to
13458 a version that passes in my_perl. */
13459 PerlInterpreter *const was = PERL_GET_THX;
13460 CLONE_PARAMS *param;
13462 PERL_ARGS_ASSERT_CLONE_PARAMS_NEW;
13468 /* Given that we've set the context, we can do this unshared. */
13469 Newx(param, 1, CLONE_PARAMS);
13472 param->proto_perl = from;
13473 param->new_perl = to;
13474 param->stashes = (AV *)Perl_newSV_type(to, SVt_PVAV);
13475 AvREAL_off(param->stashes);
13476 param->unreferenced = (AV *)Perl_newSV_type(to, SVt_PVAV);
13484 #endif /* USE_ITHREADS */
13487 =head1 Unicode Support
13489 =for apidoc sv_recode_to_utf8
13491 The encoding is assumed to be an Encode object, on entry the PV
13492 of the sv is assumed to be octets in that encoding, and the sv
13493 will be converted into Unicode (and UTF-8).
13495 If the sv already is UTF-8 (or if it is not POK), or if the encoding
13496 is not a reference, nothing is done to the sv. If the encoding is not
13497 an C<Encode::XS> Encoding object, bad things will happen.
13498 (See F<lib/encoding.pm> and L<Encode>).
13500 The PV of the sv is returned.
13505 Perl_sv_recode_to_utf8(pTHX_ SV *sv, SV *encoding)
13509 PERL_ARGS_ASSERT_SV_RECODE_TO_UTF8;
13511 if (SvPOK(sv) && !SvUTF8(sv) && !IN_BYTES && SvROK(encoding)) {
13525 Passing sv_yes is wrong - it needs to be or'ed set of constants
13526 for Encode::XS, while UTf-8 decode (currently) assumes a true value means
13527 remove converted chars from source.
13529 Both will default the value - let them.
13531 XPUSHs(&PL_sv_yes);
13534 call_method("decode", G_SCALAR);
13538 s = SvPV_const(uni, len);
13539 if (s != SvPVX_const(sv)) {
13540 SvGROW(sv, len + 1);
13541 Move(s, SvPVX(sv), len + 1, char);
13542 SvCUR_set(sv, len);
13546 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
13547 /* clear pos and any utf8 cache */
13548 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
13551 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
13552 magic_setutf8(sv,mg); /* clear UTF8 cache */
13557 return SvPOKp(sv) ? SvPVX(sv) : NULL;
13561 =for apidoc sv_cat_decode
13563 The encoding is assumed to be an Encode object, the PV of the ssv is
13564 assumed to be octets in that encoding and decoding the input starts
13565 from the position which (PV + *offset) pointed to. The dsv will be
13566 concatenated the decoded UTF-8 string from ssv. Decoding will terminate
13567 when the string tstr appears in decoding output or the input ends on
13568 the PV of the ssv. The value which the offset points will be modified
13569 to the last input position on the ssv.
13571 Returns TRUE if the terminator was found, else returns FALSE.
13576 Perl_sv_cat_decode(pTHX_ SV *dsv, SV *encoding,
13577 SV *ssv, int *offset, char *tstr, int tlen)
13582 PERL_ARGS_ASSERT_SV_CAT_DECODE;
13584 if (SvPOK(ssv) && SvPOK(dsv) && SvROK(encoding) && offset) {
13595 offsv = newSViv(*offset);
13597 mXPUSHp(tstr, tlen);
13599 call_method("cat_decode", G_SCALAR);
13601 ret = SvTRUE(TOPs);
13602 *offset = SvIV(offsv);
13608 Perl_croak(aTHX_ "Invalid argument to sv_cat_decode");
13613 /* ---------------------------------------------------------------------
13615 * support functions for report_uninit()
13618 /* the maxiumum size of array or hash where we will scan looking
13619 * for the undefined element that triggered the warning */
13621 #define FUV_MAX_SEARCH_SIZE 1000
13623 /* Look for an entry in the hash whose value has the same SV as val;
13624 * If so, return a mortal copy of the key. */
13627 S_find_hash_subscript(pTHX_ const HV *const hv, const SV *const val)
13630 register HE **array;
13633 PERL_ARGS_ASSERT_FIND_HASH_SUBSCRIPT;
13635 if (!hv || SvMAGICAL(hv) || !HvARRAY(hv) ||
13636 (HvTOTALKEYS(hv) > FUV_MAX_SEARCH_SIZE))
13639 array = HvARRAY(hv);
13641 for (i=HvMAX(hv); i>0; i--) {
13642 register HE *entry;
13643 for (entry = array[i]; entry; entry = HeNEXT(entry)) {
13644 if (HeVAL(entry) != val)
13646 if ( HeVAL(entry) == &PL_sv_undef ||
13647 HeVAL(entry) == &PL_sv_placeholder)
13651 if (HeKLEN(entry) == HEf_SVKEY)
13652 return sv_mortalcopy(HeKEY_sv(entry));
13653 return sv_2mortal(newSVhek(HeKEY_hek(entry)));
13659 /* Look for an entry in the array whose value has the same SV as val;
13660 * If so, return the index, otherwise return -1. */
13663 S_find_array_subscript(pTHX_ const AV *const av, const SV *const val)
13667 PERL_ARGS_ASSERT_FIND_ARRAY_SUBSCRIPT;
13669 if (!av || SvMAGICAL(av) || !AvARRAY(av) ||
13670 (AvFILLp(av) > FUV_MAX_SEARCH_SIZE))
13673 if (val != &PL_sv_undef) {
13674 SV ** const svp = AvARRAY(av);
13677 for (i=AvFILLp(av); i>=0; i--)
13684 /* S_varname(): return the name of a variable, optionally with a subscript.
13685 * If gv is non-zero, use the name of that global, along with gvtype (one
13686 * of "$", "@", "%"); otherwise use the name of the lexical at pad offset
13687 * targ. Depending on the value of the subscript_type flag, return:
13690 #define FUV_SUBSCRIPT_NONE 1 /* "@foo" */
13691 #define FUV_SUBSCRIPT_ARRAY 2 /* "$foo[aindex]" */
13692 #define FUV_SUBSCRIPT_HASH 3 /* "$foo{keyname}" */
13693 #define FUV_SUBSCRIPT_WITHIN 4 /* "within @foo" */
13696 S_varname(pTHX_ const GV *const gv, const char gvtype, PADOFFSET targ,
13697 const SV *const keyname, I32 aindex, int subscript_type)
13700 SV * const name = sv_newmortal();
13703 buffer[0] = gvtype;
13706 /* as gv_fullname4(), but add literal '^' for $^FOO names */
13708 gv_fullname4(name, gv, buffer, 0);
13710 if ((unsigned int)SvPVX(name)[1] <= 26) {
13712 buffer[1] = SvPVX(name)[1] + 'A' - 1;
13714 /* Swap the 1 unprintable control character for the 2 byte pretty
13715 version - ie substr($name, 1, 1) = $buffer; */
13716 sv_insert(name, 1, 1, buffer, 2);
13720 CV * const cv = find_runcv(NULL);
13724 if (!cv || !CvPADLIST(cv))
13726 av = MUTABLE_AV((*av_fetch(CvPADLIST(cv), 0, FALSE)));
13727 sv = *av_fetch(av, targ, FALSE);
13728 sv_setpvn(name, SvPV_nolen_const(sv), SvCUR(sv));
13731 if (subscript_type == FUV_SUBSCRIPT_HASH) {
13732 SV * const sv = newSV(0);
13733 *SvPVX(name) = '$';
13734 Perl_sv_catpvf(aTHX_ name, "{%s}",
13735 pv_display(sv,SvPVX_const(keyname), SvCUR(keyname), 0, 32));
13738 else if (subscript_type == FUV_SUBSCRIPT_ARRAY) {
13739 *SvPVX(name) = '$';
13740 Perl_sv_catpvf(aTHX_ name, "[%"IVdf"]", (IV)aindex);
13742 else if (subscript_type == FUV_SUBSCRIPT_WITHIN) {
13743 /* We know that name has no magic, so can use 0 instead of SV_GMAGIC */
13744 Perl_sv_insert_flags(aTHX_ name, 0, 0, STR_WITH_LEN("within "), 0);
13752 =for apidoc find_uninit_var
13754 Find the name of the undefined variable (if any) that caused the operator o
13755 to issue a "Use of uninitialized value" warning.
13756 If match is true, only return a name if it's value matches uninit_sv.
13757 So roughly speaking, if a unary operator (such as OP_COS) generates a
13758 warning, then following the direct child of the op may yield an
13759 OP_PADSV or OP_GV that gives the name of the undefined variable. On the
13760 other hand, with OP_ADD there are two branches to follow, so we only print
13761 the variable name if we get an exact match.
13763 The name is returned as a mortal SV.
13765 Assumes that PL_op is the op that originally triggered the error, and that
13766 PL_comppad/PL_curpad points to the currently executing pad.
13772 S_find_uninit_var(pTHX_ const OP *const obase, const SV *const uninit_sv,
13778 const OP *o, *o2, *kid;
13780 if (!obase || (match && (!uninit_sv || uninit_sv == &PL_sv_undef ||
13781 uninit_sv == &PL_sv_placeholder)))
13784 switch (obase->op_type) {
13791 const bool pad = (obase->op_type == OP_PADAV || obase->op_type == OP_PADHV);
13792 const bool hash = (obase->op_type == OP_PADHV || obase->op_type == OP_RV2HV);
13795 int subscript_type = FUV_SUBSCRIPT_WITHIN;
13797 if (pad) { /* @lex, %lex */
13798 sv = PAD_SVl(obase->op_targ);
13802 if (cUNOPx(obase)->op_first->op_type == OP_GV) {
13803 /* @global, %global */
13804 gv = cGVOPx_gv(cUNOPx(obase)->op_first);
13807 sv = hash ? MUTABLE_SV(GvHV(gv)): MUTABLE_SV(GvAV(gv));
13809 else /* @{expr}, %{expr} */
13810 return find_uninit_var(cUNOPx(obase)->op_first,
13814 /* attempt to find a match within the aggregate */
13816 keysv = find_hash_subscript((const HV*)sv, uninit_sv);
13818 subscript_type = FUV_SUBSCRIPT_HASH;
13821 index = find_array_subscript((const AV *)sv, uninit_sv);
13823 subscript_type = FUV_SUBSCRIPT_ARRAY;
13826 if (match && subscript_type == FUV_SUBSCRIPT_WITHIN)
13829 return varname(gv, hash ? '%' : '@', obase->op_targ,
13830 keysv, index, subscript_type);
13834 if (cUNOPx(obase)->op_first->op_type == OP_GV) {
13836 gv = cGVOPx_gv(cUNOPx(obase)->op_first);
13837 if (!gv || !GvSTASH(gv))
13839 if (match && (GvSV(gv) != uninit_sv))
13841 return varname(gv, '$', 0, NULL, 0, FUV_SUBSCRIPT_NONE);
13844 return find_uninit_var(cUNOPx(obase)->op_first, uninit_sv, 1);
13847 if (match && PAD_SVl(obase->op_targ) != uninit_sv)
13849 return varname(NULL, '$', obase->op_targ,
13850 NULL, 0, FUV_SUBSCRIPT_NONE);
13853 gv = cGVOPx_gv(obase);
13854 if (!gv || (match && GvSV(gv) != uninit_sv) || !GvSTASH(gv))
13856 return varname(gv, '$', 0, NULL, 0, FUV_SUBSCRIPT_NONE);
13858 case OP_AELEMFAST_LEX:
13861 AV *av = MUTABLE_AV(PAD_SV(obase->op_targ));
13862 if (!av || SvRMAGICAL(av))
13864 svp = av_fetch(av, (I32)obase->op_private, FALSE);
13865 if (!svp || *svp != uninit_sv)
13868 return varname(NULL, '$', obase->op_targ,
13869 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
13872 gv = cGVOPx_gv(obase);
13877 AV *const av = GvAV(gv);
13878 if (!av || SvRMAGICAL(av))
13880 svp = av_fetch(av, (I32)obase->op_private, FALSE);
13881 if (!svp || *svp != uninit_sv)
13884 return varname(gv, '$', 0,
13885 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
13890 o = cUNOPx(obase)->op_first;
13891 if (!o || o->op_type != OP_NULL ||
13892 ! (o->op_targ == OP_AELEM || o->op_targ == OP_HELEM))
13894 return find_uninit_var(cBINOPo->op_last, uninit_sv, match);
13899 bool negate = FALSE;
13901 if (PL_op == obase)
13902 /* $a[uninit_expr] or $h{uninit_expr} */
13903 return find_uninit_var(cBINOPx(obase)->op_last, uninit_sv, match);
13906 o = cBINOPx(obase)->op_first;
13907 kid = cBINOPx(obase)->op_last;
13909 /* get the av or hv, and optionally the gv */
13911 if (o->op_type == OP_PADAV || o->op_type == OP_PADHV) {
13912 sv = PAD_SV(o->op_targ);
13914 else if ((o->op_type == OP_RV2AV || o->op_type == OP_RV2HV)
13915 && cUNOPo->op_first->op_type == OP_GV)
13917 gv = cGVOPx_gv(cUNOPo->op_first);
13921 == OP_RV2HV ? MUTABLE_SV(GvHV(gv)) : MUTABLE_SV(GvAV(gv));
13926 if (kid && kid->op_type == OP_NEGATE) {
13928 kid = cUNOPx(kid)->op_first;
13931 if (kid && kid->op_type == OP_CONST && SvOK(cSVOPx_sv(kid))) {
13932 /* index is constant */
13935 kidsv = sv_2mortal(newSVpvs("-"));
13936 sv_catsv(kidsv, cSVOPx_sv(kid));
13939 kidsv = cSVOPx_sv(kid);
13943 if (obase->op_type == OP_HELEM) {
13944 HE* he = hv_fetch_ent(MUTABLE_HV(sv), kidsv, 0, 0);
13945 if (!he || HeVAL(he) != uninit_sv)
13949 SV * const * const svp = av_fetch(MUTABLE_AV(sv),
13950 negate ? - SvIV(cSVOPx_sv(kid)) : SvIV(cSVOPx_sv(kid)),
13952 if (!svp || *svp != uninit_sv)
13956 if (obase->op_type == OP_HELEM)
13957 return varname(gv, '%', o->op_targ,
13958 kidsv, 0, FUV_SUBSCRIPT_HASH);
13960 return varname(gv, '@', o->op_targ, NULL,
13961 negate ? - SvIV(cSVOPx_sv(kid)) : SvIV(cSVOPx_sv(kid)),
13962 FUV_SUBSCRIPT_ARRAY);
13965 /* index is an expression;
13966 * attempt to find a match within the aggregate */
13967 if (obase->op_type == OP_HELEM) {
13968 SV * const keysv = find_hash_subscript((const HV*)sv, uninit_sv);
13970 return varname(gv, '%', o->op_targ,
13971 keysv, 0, FUV_SUBSCRIPT_HASH);
13975 = find_array_subscript((const AV *)sv, uninit_sv);
13977 return varname(gv, '@', o->op_targ,
13978 NULL, index, FUV_SUBSCRIPT_ARRAY);
13983 (o->op_type == OP_PADAV || o->op_type == OP_RV2AV)
13985 o->op_targ, NULL, 0, FUV_SUBSCRIPT_WITHIN);
13991 /* only examine RHS */
13992 return find_uninit_var(cBINOPx(obase)->op_first, uninit_sv, match);
13995 o = cUNOPx(obase)->op_first;
13996 if (o->op_type == OP_PUSHMARK)
13999 if (!o->op_sibling) {
14000 /* one-arg version of open is highly magical */
14002 if (o->op_type == OP_GV) { /* open FOO; */
14004 if (match && GvSV(gv) != uninit_sv)
14006 return varname(gv, '$', 0,
14007 NULL, 0, FUV_SUBSCRIPT_NONE);
14009 /* other possibilities not handled are:
14010 * open $x; or open my $x; should return '${*$x}'
14011 * open expr; should return '$'.expr ideally
14017 /* ops where $_ may be an implicit arg */
14021 if ( !(obase->op_flags & OPf_STACKED)) {
14022 if (uninit_sv == ((obase->op_private & OPpTARGET_MY)
14023 ? PAD_SVl(obase->op_targ)
14026 sv = sv_newmortal();
14027 sv_setpvs(sv, "$_");
14036 match = 1; /* print etc can return undef on defined args */
14037 /* skip filehandle as it can't produce 'undef' warning */
14038 o = cUNOPx(obase)->op_first;
14039 if ((obase->op_flags & OPf_STACKED) && o->op_type == OP_PUSHMARK)
14040 o = o->op_sibling->op_sibling;
14044 case OP_ENTEREVAL: /* could be eval $undef or $x='$undef'; eval $x */
14045 case OP_CUSTOM: /* XS or custom code could trigger random warnings */
14047 /* the following ops are capable of returning PL_sv_undef even for
14048 * defined arg(s) */
14067 case OP_GETPEERNAME:
14115 case OP_SMARTMATCH:
14124 /* XXX tmp hack: these two may call an XS sub, and currently
14125 XS subs don't have a SUB entry on the context stack, so CV and
14126 pad determination goes wrong, and BAD things happen. So, just
14127 don't try to determine the value under those circumstances.
14128 Need a better fix at dome point. DAPM 11/2007 */
14134 GV * const gv = gv_fetchpvs(".", GV_NOTQUAL, SVt_PV);
14135 if (gv && GvSV(gv) == uninit_sv)
14136 return newSVpvs_flags("$.", SVs_TEMP);
14141 /* def-ness of rval pos() is independent of the def-ness of its arg */
14142 if ( !(obase->op_flags & OPf_MOD))
14147 if (SvROK(PL_rs) && uninit_sv == SvRV(PL_rs))
14148 return newSVpvs_flags("${$/}", SVs_TEMP);
14153 if (!(obase->op_flags & OPf_KIDS))
14155 o = cUNOPx(obase)->op_first;
14161 /* if all except one arg are constant, or have no side-effects,
14162 * or are optimized away, then it's unambiguous */
14164 for (kid=o; kid; kid = kid->op_sibling) {
14166 const OPCODE type = kid->op_type;
14167 if ( (type == OP_CONST && SvOK(cSVOPx_sv(kid)))
14168 || (type == OP_NULL && ! (kid->op_flags & OPf_KIDS))
14169 || (type == OP_PUSHMARK)
14171 /* @$a and %$a, but not @a or %a */
14172 (type == OP_RV2AV || type == OP_RV2HV)
14173 && cUNOPx(kid)->op_first
14174 && cUNOPx(kid)->op_first->op_type != OP_GV
14179 if (o2) { /* more than one found */
14186 return find_uninit_var(o2, uninit_sv, match);
14188 /* scan all args */
14190 sv = find_uninit_var(o, uninit_sv, 1);
14202 =for apidoc report_uninit
14204 Print appropriate "Use of uninitialized variable" warning
14210 Perl_report_uninit(pTHX_ const SV *uninit_sv)
14214 SV* varname = NULL;
14216 varname = find_uninit_var(PL_op, uninit_sv,0);
14218 sv_insert(varname, 0, 0, " ", 1);
14220 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
14221 varname ? SvPV_nolen_const(varname) : "",
14222 " in ", OP_DESC(PL_op));
14225 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
14231 * c-indentation-style: bsd
14232 * c-basic-offset: 4
14233 * indent-tabs-mode: t
14236 * ex: set ts=8 sts=4 sw=4 noet: