3 * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, 2000,
4 * 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009 by Larry Wall
7 * You may distribute under the terms of either the GNU General Public
8 * License or the Artistic License, as specified in the README file.
13 * 'I wonder what the Entish is for "yes" and "no",' he thought.
16 * [p.480 of _The Lord of the Rings_, III/iv: "Treebeard"]
22 * This file contains the code that creates, manipulates and destroys
23 * scalar values (SVs). The other types (AV, HV, GV, etc.) reuse the
24 * structure of an SV, so their creation and destruction is handled
25 * here; higher-level functions are in av.c, hv.c, and so on. Opcode
26 * level functions (eg. substr, split, join) for each of the types are
36 # if __STDC_VERSION__ >= 199901L && !defined(VMS)
47 /* Missing proto on LynxOS */
48 char *gconvert(double, int, int, char *);
51 #ifdef PERL_UTF8_CACHE_ASSERT
52 /* if adding more checks watch out for the following tests:
53 * t/op/index.t t/op/length.t t/op/pat.t t/op/substr.t
54 * lib/utf8.t lib/Unicode/Collate/t/index.t
57 # define ASSERT_UTF8_CACHE(cache) \
58 STMT_START { if (cache) { assert((cache)[0] <= (cache)[1]); \
59 assert((cache)[2] <= (cache)[3]); \
60 assert((cache)[3] <= (cache)[1]);} \
63 # define ASSERT_UTF8_CACHE(cache) NOOP
66 #ifdef PERL_OLD_COPY_ON_WRITE
67 #define SV_COW_NEXT_SV(sv) INT2PTR(SV *,SvUVX(sv))
68 #define SV_COW_NEXT_SV_SET(current,next) SvUV_set(current, PTR2UV(next))
69 /* This is a pessimistic view. Scalar must be purely a read-write PV to copy-
73 /* ============================================================================
75 =head1 Allocation and deallocation of SVs.
77 An SV (or AV, HV, etc.) is allocated in two parts: the head (struct
78 sv, av, hv...) contains type and reference count information, and for
79 many types, a pointer to the body (struct xrv, xpv, xpviv...), which
80 contains fields specific to each type. Some types store all they need
81 in the head, so don't have a body.
83 In all but the most memory-paranoid configurations (ex: PURIFY), heads
84 and bodies are allocated out of arenas, which by default are
85 approximately 4K chunks of memory parcelled up into N heads or bodies.
86 Sv-bodies are allocated by their sv-type, guaranteeing size
87 consistency needed to allocate safely from arrays.
89 For SV-heads, the first slot in each arena is reserved, and holds a
90 link to the next arena, some flags, and a note of the number of slots.
91 Snaked through each arena chain is a linked list of free items; when
92 this becomes empty, an extra arena is allocated and divided up into N
93 items which are threaded into the free list.
95 SV-bodies are similar, but they use arena-sets by default, which
96 separate the link and info from the arena itself, and reclaim the 1st
97 slot in the arena. SV-bodies are further described later.
99 The following global variables are associated with arenas:
101 PL_sv_arenaroot pointer to list of SV arenas
102 PL_sv_root pointer to list of free SV structures
104 PL_body_arenas head of linked-list of body arenas
105 PL_body_roots[] array of pointers to list of free bodies of svtype
106 arrays are indexed by the svtype needed
108 A few special SV heads are not allocated from an arena, but are
109 instead directly created in the interpreter structure, eg PL_sv_undef.
110 The size of arenas can be changed from the default by setting
111 PERL_ARENA_SIZE appropriately at compile time.
113 The SV arena serves the secondary purpose of allowing still-live SVs
114 to be located and destroyed during final cleanup.
116 At the lowest level, the macros new_SV() and del_SV() grab and free
117 an SV head. (If debugging with -DD, del_SV() calls the function S_del_sv()
118 to return the SV to the free list with error checking.) new_SV() calls
119 more_sv() / sv_add_arena() to add an extra arena if the free list is empty.
120 SVs in the free list have their SvTYPE field set to all ones.
122 At the time of very final cleanup, sv_free_arenas() is called from
123 perl_destruct() to physically free all the arenas allocated since the
124 start of the interpreter.
126 The function visit() scans the SV arenas list, and calls a specified
127 function for each SV it finds which is still live - ie which has an SvTYPE
128 other than all 1's, and a non-zero SvREFCNT. visit() is used by the
129 following functions (specified as [function that calls visit()] / [function
130 called by visit() for each SV]):
132 sv_report_used() / do_report_used()
133 dump all remaining SVs (debugging aid)
135 sv_clean_objs() / do_clean_objs(),do_clean_named_objs(),
136 do_clean_named_io_objs()
137 Attempt to free all objects pointed to by RVs,
138 and try to do the same for all objects indirectly
139 referenced by typeglobs too. Called once from
140 perl_destruct(), prior to calling sv_clean_all()
143 sv_clean_all() / do_clean_all()
144 SvREFCNT_dec(sv) each remaining SV, possibly
145 triggering an sv_free(). It also sets the
146 SVf_BREAK flag on the SV to indicate that the
147 refcnt has been artificially lowered, and thus
148 stopping sv_free() from giving spurious warnings
149 about SVs which unexpectedly have a refcnt
150 of zero. called repeatedly from perl_destruct()
151 until there are no SVs left.
153 =head2 Arena allocator API Summary
155 Private API to rest of sv.c
159 new_XPVNV(), del_XPVGV(),
164 sv_report_used(), sv_clean_objs(), sv_clean_all(), sv_free_arenas()
168 * ========================================================================= */
171 * "A time to plant, and a time to uproot what was planted..."
175 # define MEM_LOG_NEW_SV(sv, file, line, func) \
176 Perl_mem_log_new_sv(sv, file, line, func)
177 # define MEM_LOG_DEL_SV(sv, file, line, func) \
178 Perl_mem_log_del_sv(sv, file, line, func)
180 # define MEM_LOG_NEW_SV(sv, file, line, func) NOOP
181 # define MEM_LOG_DEL_SV(sv, file, line, func) NOOP
184 #ifdef DEBUG_LEAKING_SCALARS
185 # define FREE_SV_DEBUG_FILE(sv) Safefree((sv)->sv_debug_file)
186 # define DEBUG_SV_SERIAL(sv) \
187 DEBUG_m(PerlIO_printf(Perl_debug_log, "0x%"UVxf": (%05ld) del_SV\n", \
188 PTR2UV(sv), (long)(sv)->sv_debug_serial))
190 # define FREE_SV_DEBUG_FILE(sv)
191 # define DEBUG_SV_SERIAL(sv) NOOP
195 # define SvARENA_CHAIN(sv) ((sv)->sv_u.svu_rv)
196 # define SvARENA_CHAIN_SET(sv,val) (sv)->sv_u.svu_rv = MUTABLE_SV((val))
197 /* Whilst I'd love to do this, it seems that things like to check on
199 # define POSION_SV_HEAD(sv) PoisonNew(sv, 1, struct STRUCT_SV)
201 # define POSION_SV_HEAD(sv) PoisonNew(&SvANY(sv), 1, void *), \
202 PoisonNew(&SvREFCNT(sv), 1, U32)
204 # define SvARENA_CHAIN(sv) SvANY(sv)
205 # define SvARENA_CHAIN_SET(sv,val) SvANY(sv) = (void *)(val)
206 # define POSION_SV_HEAD(sv)
209 /* Mark an SV head as unused, and add to free list.
211 * If SVf_BREAK is set, skip adding it to the free list, as this SV had
212 * its refcount artificially decremented during global destruction, so
213 * there may be dangling pointers to it. The last thing we want in that
214 * case is for it to be reused. */
216 #define plant_SV(p) \
218 const U32 old_flags = SvFLAGS(p); \
219 MEM_LOG_DEL_SV(p, __FILE__, __LINE__, FUNCTION__); \
220 DEBUG_SV_SERIAL(p); \
221 FREE_SV_DEBUG_FILE(p); \
223 SvFLAGS(p) = SVTYPEMASK; \
224 if (!(old_flags & SVf_BREAK)) { \
225 SvARENA_CHAIN_SET(p, PL_sv_root); \
231 #define uproot_SV(p) \
234 PL_sv_root = MUTABLE_SV(SvARENA_CHAIN(p)); \
239 /* make some more SVs by adding another arena */
246 char *chunk; /* must use New here to match call to */
247 Newx(chunk,PERL_ARENA_SIZE,char); /* Safefree() in sv_free_arenas() */
248 sv_add_arena(chunk, PERL_ARENA_SIZE, 0);
253 /* new_SV(): return a new, empty SV head */
255 #ifdef DEBUG_LEAKING_SCALARS
256 /* provide a real function for a debugger to play with */
258 S_new_SV(pTHX_ const char *file, int line, const char *func)
265 sv = S_more_sv(aTHX);
269 sv->sv_debug_optype = PL_op ? PL_op->op_type : 0;
270 sv->sv_debug_line = (U16) (PL_parser && PL_parser->copline != NOLINE
276 sv->sv_debug_inpad = 0;
277 sv->sv_debug_parent = NULL;
278 sv->sv_debug_file = PL_curcop ? savepv(CopFILE(PL_curcop)): NULL;
280 sv->sv_debug_serial = PL_sv_serial++;
282 MEM_LOG_NEW_SV(sv, file, line, func);
283 DEBUG_m(PerlIO_printf(Perl_debug_log, "0x%"UVxf": (%05ld) new_SV (from %s:%d [%s])\n",
284 PTR2UV(sv), (long)sv->sv_debug_serial, file, line, func));
288 # define new_SV(p) (p)=S_new_SV(aTHX_ __FILE__, __LINE__, FUNCTION__)
296 (p) = S_more_sv(aTHX); \
300 MEM_LOG_NEW_SV(p, __FILE__, __LINE__, FUNCTION__); \
305 /* del_SV(): return an empty SV head to the free list */
318 S_del_sv(pTHX_ SV *p)
322 PERL_ARGS_ASSERT_DEL_SV;
327 for (sva = PL_sv_arenaroot; sva; sva = MUTABLE_SV(SvANY(sva))) {
328 const SV * const sv = sva + 1;
329 const SV * const svend = &sva[SvREFCNT(sva)];
330 if (p >= sv && p < svend) {
336 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL),
337 "Attempt to free non-arena SV: 0x%"UVxf
338 pTHX__FORMAT, PTR2UV(p) pTHX__VALUE);
345 #else /* ! DEBUGGING */
347 #define del_SV(p) plant_SV(p)
349 #endif /* DEBUGGING */
353 =head1 SV Manipulation Functions
355 =for apidoc sv_add_arena
357 Given a chunk of memory, link it to the head of the list of arenas,
358 and split it into a list of free SVs.
364 S_sv_add_arena(pTHX_ char *const ptr, const U32 size, const U32 flags)
367 SV *const sva = MUTABLE_SV(ptr);
371 PERL_ARGS_ASSERT_SV_ADD_ARENA;
373 /* The first SV in an arena isn't an SV. */
374 SvANY(sva) = (void *) PL_sv_arenaroot; /* ptr to next arena */
375 SvREFCNT(sva) = size / sizeof(SV); /* number of SV slots */
376 SvFLAGS(sva) = flags; /* FAKE if not to be freed */
378 PL_sv_arenaroot = sva;
379 PL_sv_root = sva + 1;
381 svend = &sva[SvREFCNT(sva) - 1];
384 SvARENA_CHAIN_SET(sv, (sv + 1));
388 /* Must always set typemask because it's always checked in on cleanup
389 when the arenas are walked looking for objects. */
390 SvFLAGS(sv) = SVTYPEMASK;
393 SvARENA_CHAIN_SET(sv, 0);
397 SvFLAGS(sv) = SVTYPEMASK;
400 /* visit(): call the named function for each non-free SV in the arenas
401 * whose flags field matches the flags/mask args. */
404 S_visit(pTHX_ SVFUNC_t f, const U32 flags, const U32 mask)
410 PERL_ARGS_ASSERT_VISIT;
412 for (sva = PL_sv_arenaroot; sva; sva = MUTABLE_SV(SvANY(sva))) {
413 register const SV * const svend = &sva[SvREFCNT(sva)];
415 for (sv = sva + 1; sv < svend; ++sv) {
416 if (SvTYPE(sv) != SVTYPEMASK
417 && (sv->sv_flags & mask) == flags
430 /* called by sv_report_used() for each live SV */
433 do_report_used(pTHX_ SV *const sv)
435 if (SvTYPE(sv) != SVTYPEMASK) {
436 PerlIO_printf(Perl_debug_log, "****\n");
443 =for apidoc sv_report_used
445 Dump the contents of all SVs not yet freed. (Debugging aid).
451 Perl_sv_report_used(pTHX)
454 visit(do_report_used, 0, 0);
460 /* called by sv_clean_objs() for each live SV */
463 do_clean_objs(pTHX_ SV *const ref)
468 SV * const target = SvRV(ref);
469 if (SvOBJECT(target)) {
470 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning object ref:\n "), sv_dump(ref)));
471 if (SvWEAKREF(ref)) {
472 sv_del_backref(target, ref);
478 SvREFCNT_dec(target);
483 /* XXX Might want to check arrays, etc. */
487 /* clear any slots in a GV which hold objects - except IO;
488 * called by sv_clean_objs() for each live GV */
491 do_clean_named_objs(pTHX_ SV *const sv)
495 assert(SvTYPE(sv) == SVt_PVGV);
496 assert(isGV_with_GP(sv));
500 /* freeing GP entries may indirectly free the current GV;
501 * hold onto it while we mess with the GP slots */
504 if ( ((obj = GvSV(sv) )) && SvOBJECT(obj)) {
505 DEBUG_D((PerlIO_printf(Perl_debug_log,
506 "Cleaning named glob SV object:\n "), sv_dump(obj)));
510 if ( ((obj = MUTABLE_SV(GvAV(sv)) )) && SvOBJECT(obj)) {
511 DEBUG_D((PerlIO_printf(Perl_debug_log,
512 "Cleaning named glob AV object:\n "), sv_dump(obj)));
516 if ( ((obj = MUTABLE_SV(GvHV(sv)) )) && SvOBJECT(obj)) {
517 DEBUG_D((PerlIO_printf(Perl_debug_log,
518 "Cleaning named glob HV object:\n "), sv_dump(obj)));
522 if ( ((obj = MUTABLE_SV(GvCV(sv)) )) && SvOBJECT(obj)) {
523 DEBUG_D((PerlIO_printf(Perl_debug_log,
524 "Cleaning named glob CV object:\n "), sv_dump(obj)));
528 SvREFCNT_dec(sv); /* undo the inc above */
531 /* clear any IO slots in a GV which hold objects (except stderr, defout);
532 * called by sv_clean_objs() for each live GV */
535 do_clean_named_io_objs(pTHX_ SV *const sv)
539 assert(SvTYPE(sv) == SVt_PVGV);
540 assert(isGV_with_GP(sv));
541 if (!GvGP(sv) || sv == (SV*)PL_stderrgv || sv == (SV*)PL_defoutgv)
545 if ( ((obj = MUTABLE_SV(GvIO(sv)) )) && SvOBJECT(obj)) {
546 DEBUG_D((PerlIO_printf(Perl_debug_log,
547 "Cleaning named glob IO object:\n "), sv_dump(obj)));
551 SvREFCNT_dec(sv); /* undo the inc above */
554 /* Void wrapper to pass to visit() */
556 do_curse(pTHX_ SV * const sv) {
557 if ((PL_stderrgv && GvGP(PL_stderrgv) && (SV*)GvIO(PL_stderrgv) == sv)
558 || (PL_defoutgv && GvGP(PL_defoutgv) && (SV*)GvIO(PL_defoutgv) == sv))
564 =for apidoc sv_clean_objs
566 Attempt to destroy all objects not yet freed
572 Perl_sv_clean_objs(pTHX)
576 PL_in_clean_objs = TRUE;
577 visit(do_clean_objs, SVf_ROK, SVf_ROK);
578 /* Some barnacles may yet remain, clinging to typeglobs.
579 * Run the non-IO destructors first: they may want to output
580 * error messages, close files etc */
581 visit(do_clean_named_objs, SVt_PVGV|SVpgv_GP, SVTYPEMASK|SVp_POK|SVpgv_GP);
582 visit(do_clean_named_io_objs, SVt_PVGV|SVpgv_GP, SVTYPEMASK|SVp_POK|SVpgv_GP);
583 /* And if there are some very tenacious barnacles clinging to arrays,
584 closures, or what have you.... */
585 visit(do_curse, SVs_OBJECT, SVs_OBJECT);
586 olddef = PL_defoutgv;
587 PL_defoutgv = NULL; /* disable skip of PL_defoutgv */
588 if (olddef && isGV_with_GP(olddef))
589 do_clean_named_io_objs(aTHX_ MUTABLE_SV(olddef));
590 olderr = PL_stderrgv;
591 PL_stderrgv = NULL; /* disable skip of PL_stderrgv */
592 if (olderr && isGV_with_GP(olderr))
593 do_clean_named_io_objs(aTHX_ MUTABLE_SV(olderr));
594 SvREFCNT_dec(olddef);
595 PL_in_clean_objs = FALSE;
598 /* called by sv_clean_all() for each live SV */
601 do_clean_all(pTHX_ SV *const sv)
604 if (sv == (const SV *) PL_fdpid || sv == (const SV *)PL_strtab) {
605 /* don't clean pid table and strtab */
608 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning loops: SV at 0x%"UVxf"\n", PTR2UV(sv)) ));
609 SvFLAGS(sv) |= SVf_BREAK;
614 =for apidoc sv_clean_all
616 Decrement the refcnt of each remaining SV, possibly triggering a
617 cleanup. This function may have to be called multiple times to free
618 SVs which are in complex self-referential hierarchies.
624 Perl_sv_clean_all(pTHX)
628 PL_in_clean_all = TRUE;
629 cleaned = visit(do_clean_all, 0,0);
634 ARENASETS: a meta-arena implementation which separates arena-info
635 into struct arena_set, which contains an array of struct
636 arena_descs, each holding info for a single arena. By separating
637 the meta-info from the arena, we recover the 1st slot, formerly
638 borrowed for list management. The arena_set is about the size of an
639 arena, avoiding the needless malloc overhead of a naive linked-list.
641 The cost is 1 arena-set malloc per ~320 arena-mallocs, + the unused
642 memory in the last arena-set (1/2 on average). In trade, we get
643 back the 1st slot in each arena (ie 1.7% of a CV-arena, less for
644 smaller types). The recovery of the wasted space allows use of
645 small arenas for large, rare body types, by changing array* fields
646 in body_details_by_type[] below.
649 char *arena; /* the raw storage, allocated aligned */
650 size_t size; /* its size ~4k typ */
651 svtype utype; /* bodytype stored in arena */
656 /* Get the maximum number of elements in set[] such that struct arena_set
657 will fit within PERL_ARENA_SIZE, which is probably just under 4K, and
658 therefore likely to be 1 aligned memory page. */
660 #define ARENAS_PER_SET ((PERL_ARENA_SIZE - sizeof(struct arena_set*) \
661 - 2 * sizeof(int)) / sizeof (struct arena_desc))
664 struct arena_set* next;
665 unsigned int set_size; /* ie ARENAS_PER_SET */
666 unsigned int curr; /* index of next available arena-desc */
667 struct arena_desc set[ARENAS_PER_SET];
671 =for apidoc sv_free_arenas
673 Deallocate the memory used by all arenas. Note that all the individual SV
674 heads and bodies within the arenas must already have been freed.
679 Perl_sv_free_arenas(pTHX)
686 /* Free arenas here, but be careful about fake ones. (We assume
687 contiguity of the fake ones with the corresponding real ones.) */
689 for (sva = PL_sv_arenaroot; sva; sva = svanext) {
690 svanext = MUTABLE_SV(SvANY(sva));
691 while (svanext && SvFAKE(svanext))
692 svanext = MUTABLE_SV(SvANY(svanext));
699 struct arena_set *aroot = (struct arena_set*) PL_body_arenas;
702 struct arena_set *current = aroot;
705 assert(aroot->set[i].arena);
706 Safefree(aroot->set[i].arena);
714 i = PERL_ARENA_ROOTS_SIZE;
716 PL_body_roots[i] = 0;
723 Here are mid-level routines that manage the allocation of bodies out
724 of the various arenas. There are 5 kinds of arenas:
726 1. SV-head arenas, which are discussed and handled above
727 2. regular body arenas
728 3. arenas for reduced-size bodies
731 Arena types 2 & 3 are chained by body-type off an array of
732 arena-root pointers, which is indexed by svtype. Some of the
733 larger/less used body types are malloced singly, since a large
734 unused block of them is wasteful. Also, several svtypes dont have
735 bodies; the data fits into the sv-head itself. The arena-root
736 pointer thus has a few unused root-pointers (which may be hijacked
737 later for arena types 4,5)
739 3 differs from 2 as an optimization; some body types have several
740 unused fields in the front of the structure (which are kept in-place
741 for consistency). These bodies can be allocated in smaller chunks,
742 because the leading fields arent accessed. Pointers to such bodies
743 are decremented to point at the unused 'ghost' memory, knowing that
744 the pointers are used with offsets to the real memory.
747 =head1 SV-Body Allocation
749 Allocation of SV-bodies is similar to SV-heads, differing as follows;
750 the allocation mechanism is used for many body types, so is somewhat
751 more complicated, it uses arena-sets, and has no need for still-live
754 At the outermost level, (new|del)_X*V macros return bodies of the
755 appropriate type. These macros call either (new|del)_body_type or
756 (new|del)_body_allocated macro pairs, depending on specifics of the
757 type. Most body types use the former pair, the latter pair is used to
758 allocate body types with "ghost fields".
760 "ghost fields" are fields that are unused in certain types, and
761 consequently don't need to actually exist. They are declared because
762 they're part of a "base type", which allows use of functions as
763 methods. The simplest examples are AVs and HVs, 2 aggregate types
764 which don't use the fields which support SCALAR semantics.
766 For these types, the arenas are carved up into appropriately sized
767 chunks, we thus avoid wasted memory for those unaccessed members.
768 When bodies are allocated, we adjust the pointer back in memory by the
769 size of the part not allocated, so it's as if we allocated the full
770 structure. (But things will all go boom if you write to the part that
771 is "not there", because you'll be overwriting the last members of the
772 preceding structure in memory.)
774 We calculate the correction using the STRUCT_OFFSET macro on the first
775 member present. If the allocated structure is smaller (no initial NV
776 actually allocated) then the net effect is to subtract the size of the NV
777 from the pointer, to return a new pointer as if an initial NV were actually
778 allocated. (We were using structures named *_allocated for this, but
779 this turned out to be a subtle bug, because a structure without an NV
780 could have a lower alignment constraint, but the compiler is allowed to
781 optimised accesses based on the alignment constraint of the actual pointer
782 to the full structure, for example, using a single 64 bit load instruction
783 because it "knows" that two adjacent 32 bit members will be 8-byte aligned.)
785 This is the same trick as was used for NV and IV bodies. Ironically it
786 doesn't need to be used for NV bodies any more, because NV is now at
787 the start of the structure. IV bodies don't need it either, because
788 they are no longer allocated.
790 In turn, the new_body_* allocators call S_new_body(), which invokes
791 new_body_inline macro, which takes a lock, and takes a body off the
792 linked list at PL_body_roots[sv_type], calling Perl_more_bodies() if
793 necessary to refresh an empty list. Then the lock is released, and
794 the body is returned.
796 Perl_more_bodies allocates a new arena, and carves it up into an array of N
797 bodies, which it strings into a linked list. It looks up arena-size
798 and body-size from the body_details table described below, thus
799 supporting the multiple body-types.
801 If PURIFY is defined, or PERL_ARENA_SIZE=0, arenas are not used, and
802 the (new|del)_X*V macros are mapped directly to malloc/free.
804 For each sv-type, struct body_details bodies_by_type[] carries
805 parameters which control these aspects of SV handling:
807 Arena_size determines whether arenas are used for this body type, and if
808 so, how big they are. PURIFY or PERL_ARENA_SIZE=0 set this field to
809 zero, forcing individual mallocs and frees.
811 Body_size determines how big a body is, and therefore how many fit into
812 each arena. Offset carries the body-pointer adjustment needed for
813 "ghost fields", and is used in *_allocated macros.
815 But its main purpose is to parameterize info needed in
816 Perl_sv_upgrade(). The info here dramatically simplifies the function
817 vs the implementation in 5.8.8, making it table-driven. All fields
818 are used for this, except for arena_size.
820 For the sv-types that have no bodies, arenas are not used, so those
821 PL_body_roots[sv_type] are unused, and can be overloaded. In
822 something of a special case, SVt_NULL is borrowed for HE arenas;
823 PL_body_roots[HE_SVSLOT=SVt_NULL] is filled by S_more_he, but the
824 bodies_by_type[SVt_NULL] slot is not used, as the table is not
829 struct body_details {
830 U8 body_size; /* Size to allocate */
831 U8 copy; /* Size of structure to copy (may be shorter) */
833 unsigned int type : 4; /* We have space for a sanity check. */
834 unsigned int cant_upgrade : 1; /* Cannot upgrade this type */
835 unsigned int zero_nv : 1; /* zero the NV when upgrading from this */
836 unsigned int arena : 1; /* Allocated from an arena */
837 size_t arena_size; /* Size of arena to allocate */
845 /* With -DPURFIY we allocate everything directly, and don't use arenas.
846 This seems a rather elegant way to simplify some of the code below. */
847 #define HASARENA FALSE
849 #define HASARENA TRUE
851 #define NOARENA FALSE
853 /* Size the arenas to exactly fit a given number of bodies. A count
854 of 0 fits the max number bodies into a PERL_ARENA_SIZE.block,
855 simplifying the default. If count > 0, the arena is sized to fit
856 only that many bodies, allowing arenas to be used for large, rare
857 bodies (XPVFM, XPVIO) without undue waste. The arena size is
858 limited by PERL_ARENA_SIZE, so we can safely oversize the
861 #define FIT_ARENA0(body_size) \
862 ((size_t)(PERL_ARENA_SIZE / body_size) * body_size)
863 #define FIT_ARENAn(count,body_size) \
864 ( count * body_size <= PERL_ARENA_SIZE) \
865 ? count * body_size \
866 : FIT_ARENA0 (body_size)
867 #define FIT_ARENA(count,body_size) \
869 ? FIT_ARENAn (count, body_size) \
870 : FIT_ARENA0 (body_size)
872 /* Calculate the length to copy. Specifically work out the length less any
873 final padding the compiler needed to add. See the comment in sv_upgrade
874 for why copying the padding proved to be a bug. */
876 #define copy_length(type, last_member) \
877 STRUCT_OFFSET(type, last_member) \
878 + sizeof (((type*)SvANY((const SV *)0))->last_member)
880 static const struct body_details bodies_by_type[] = {
881 /* HEs use this offset for their arena. */
882 { 0, 0, 0, SVt_NULL, FALSE, NONV, NOARENA, 0 },
884 /* The bind placeholder pretends to be an RV for now.
885 Also it's marked as "can't upgrade" to stop anyone using it before it's
887 { 0, 0, 0, SVt_BIND, TRUE, NONV, NOARENA, 0 },
889 /* IVs are in the head, so the allocation size is 0. */
891 sizeof(IV), /* This is used to copy out the IV body. */
892 STRUCT_OFFSET(XPVIV, xiv_iv), SVt_IV, FALSE, NONV,
893 NOARENA /* IVS don't need an arena */, 0
896 { sizeof(NV), sizeof(NV),
897 STRUCT_OFFSET(XPVNV, xnv_u),
898 SVt_NV, FALSE, HADNV, HASARENA, FIT_ARENA(0, sizeof(NV)) },
900 { sizeof(XPV) - STRUCT_OFFSET(XPV, xpv_cur),
901 copy_length(XPV, xpv_len) - STRUCT_OFFSET(XPV, xpv_cur),
902 + STRUCT_OFFSET(XPV, xpv_cur),
903 SVt_PV, FALSE, NONV, HASARENA,
904 FIT_ARENA(0, sizeof(XPV) - STRUCT_OFFSET(XPV, xpv_cur)) },
906 { sizeof(XPVIV) - STRUCT_OFFSET(XPV, xpv_cur),
907 copy_length(XPVIV, xiv_u) - STRUCT_OFFSET(XPV, xpv_cur),
908 + STRUCT_OFFSET(XPV, xpv_cur),
909 SVt_PVIV, FALSE, NONV, HASARENA,
910 FIT_ARENA(0, sizeof(XPVIV) - STRUCT_OFFSET(XPV, xpv_cur)) },
912 { sizeof(XPVNV) - STRUCT_OFFSET(XPV, xpv_cur),
913 copy_length(XPVNV, xnv_u) - STRUCT_OFFSET(XPV, xpv_cur),
914 + STRUCT_OFFSET(XPV, xpv_cur),
915 SVt_PVNV, FALSE, HADNV, HASARENA,
916 FIT_ARENA(0, sizeof(XPVNV) - STRUCT_OFFSET(XPV, xpv_cur)) },
918 { sizeof(XPVMG), copy_length(XPVMG, xnv_u), 0, SVt_PVMG, FALSE, HADNV,
919 HASARENA, FIT_ARENA(0, sizeof(XPVMG)) },
924 SVt_REGEXP, FALSE, NONV, HASARENA,
925 FIT_ARENA(0, sizeof(regexp))
928 { sizeof(XPVGV), sizeof(XPVGV), 0, SVt_PVGV, TRUE, HADNV,
929 HASARENA, FIT_ARENA(0, sizeof(XPVGV)) },
931 { sizeof(XPVLV), sizeof(XPVLV), 0, SVt_PVLV, TRUE, HADNV,
932 HASARENA, FIT_ARENA(0, sizeof(XPVLV)) },
935 copy_length(XPVAV, xav_alloc),
937 SVt_PVAV, TRUE, NONV, HASARENA,
938 FIT_ARENA(0, sizeof(XPVAV)) },
941 copy_length(XPVHV, xhv_max),
943 SVt_PVHV, TRUE, NONV, HASARENA,
944 FIT_ARENA(0, sizeof(XPVHV)) },
949 SVt_PVCV, TRUE, NONV, HASARENA,
950 FIT_ARENA(0, sizeof(XPVCV)) },
955 SVt_PVFM, TRUE, NONV, NOARENA,
956 FIT_ARENA(20, sizeof(XPVFM)) },
961 SVt_PVIO, TRUE, NONV, HASARENA,
962 FIT_ARENA(24, sizeof(XPVIO)) },
965 #define new_body_allocated(sv_type) \
966 (void *)((char *)S_new_body(aTHX_ sv_type) \
967 - bodies_by_type[sv_type].offset)
969 /* return a thing to the free list */
971 #define del_body(thing, root) \
973 void ** const thing_copy = (void **)thing; \
974 *thing_copy = *root; \
975 *root = (void*)thing_copy; \
980 #define new_XNV() safemalloc(sizeof(XPVNV))
981 #define new_XPVNV() safemalloc(sizeof(XPVNV))
982 #define new_XPVMG() safemalloc(sizeof(XPVMG))
984 #define del_XPVGV(p) safefree(p)
988 #define new_XNV() new_body_allocated(SVt_NV)
989 #define new_XPVNV() new_body_allocated(SVt_PVNV)
990 #define new_XPVMG() new_body_allocated(SVt_PVMG)
992 #define del_XPVGV(p) del_body(p + bodies_by_type[SVt_PVGV].offset, \
993 &PL_body_roots[SVt_PVGV])
997 /* no arena for you! */
999 #define new_NOARENA(details) \
1000 safemalloc((details)->body_size + (details)->offset)
1001 #define new_NOARENAZ(details) \
1002 safecalloc((details)->body_size + (details)->offset, 1)
1005 Perl_more_bodies (pTHX_ const svtype sv_type, const size_t body_size,
1006 const size_t arena_size)
1009 void ** const root = &PL_body_roots[sv_type];
1010 struct arena_desc *adesc;
1011 struct arena_set *aroot = (struct arena_set *) PL_body_arenas;
1015 const size_t good_arena_size = Perl_malloc_good_size(arena_size);
1016 #if defined(DEBUGGING) && !defined(PERL_GLOBAL_STRUCT_PRIVATE)
1017 static bool done_sanity_check;
1019 /* PERL_GLOBAL_STRUCT_PRIVATE cannot coexist with global
1020 * variables like done_sanity_check. */
1021 if (!done_sanity_check) {
1022 unsigned int i = SVt_LAST;
1024 done_sanity_check = TRUE;
1027 assert (bodies_by_type[i].type == i);
1033 /* may need new arena-set to hold new arena */
1034 if (!aroot || aroot->curr >= aroot->set_size) {
1035 struct arena_set *newroot;
1036 Newxz(newroot, 1, struct arena_set);
1037 newroot->set_size = ARENAS_PER_SET;
1038 newroot->next = aroot;
1040 PL_body_arenas = (void *) newroot;
1041 DEBUG_m(PerlIO_printf(Perl_debug_log, "new arenaset %p\n", (void*)aroot));
1044 /* ok, now have arena-set with at least 1 empty/available arena-desc */
1045 curr = aroot->curr++;
1046 adesc = &(aroot->set[curr]);
1047 assert(!adesc->arena);
1049 Newx(adesc->arena, good_arena_size, char);
1050 adesc->size = good_arena_size;
1051 adesc->utype = sv_type;
1052 DEBUG_m(PerlIO_printf(Perl_debug_log, "arena %d added: %p size %"UVuf"\n",
1053 curr, (void*)adesc->arena, (UV)good_arena_size));
1055 start = (char *) adesc->arena;
1057 /* Get the address of the byte after the end of the last body we can fit.
1058 Remember, this is integer division: */
1059 end = start + good_arena_size / body_size * body_size;
1061 /* computed count doesn't reflect the 1st slot reservation */
1062 #if defined(MYMALLOC) || defined(HAS_MALLOC_GOOD_SIZE)
1063 DEBUG_m(PerlIO_printf(Perl_debug_log,
1064 "arena %p end %p arena-size %d (from %d) type %d "
1066 (void*)start, (void*)end, (int)good_arena_size,
1067 (int)arena_size, sv_type, (int)body_size,
1068 (int)good_arena_size / (int)body_size));
1070 DEBUG_m(PerlIO_printf(Perl_debug_log,
1071 "arena %p end %p arena-size %d type %d size %d ct %d\n",
1072 (void*)start, (void*)end,
1073 (int)arena_size, sv_type, (int)body_size,
1074 (int)good_arena_size / (int)body_size));
1076 *root = (void *)start;
1079 /* Where the next body would start: */
1080 char * const next = start + body_size;
1083 /* This is the last body: */
1084 assert(next == end);
1086 *(void **)start = 0;
1090 *(void**) start = (void *)next;
1095 /* grab a new thing from the free list, allocating more if necessary.
1096 The inline version is used for speed in hot routines, and the
1097 function using it serves the rest (unless PURIFY).
1099 #define new_body_inline(xpv, sv_type) \
1101 void ** const r3wt = &PL_body_roots[sv_type]; \
1102 xpv = (PTR_TBL_ENT_t*) (*((void **)(r3wt)) \
1103 ? *((void **)(r3wt)) : Perl_more_bodies(aTHX_ sv_type, \
1104 bodies_by_type[sv_type].body_size,\
1105 bodies_by_type[sv_type].arena_size)); \
1106 *(r3wt) = *(void**)(xpv); \
1112 S_new_body(pTHX_ const svtype sv_type)
1116 new_body_inline(xpv, sv_type);
1122 static const struct body_details fake_rv =
1123 { 0, 0, 0, SVt_IV, FALSE, NONV, NOARENA, 0 };
1126 =for apidoc sv_upgrade
1128 Upgrade an SV to a more complex form. Generally adds a new body type to the
1129 SV, then copies across as much information as possible from the old body.
1130 You generally want to use the C<SvUPGRADE> macro wrapper. See also C<svtype>.
1136 Perl_sv_upgrade(pTHX_ register SV *const sv, svtype new_type)
1141 const svtype old_type = SvTYPE(sv);
1142 const struct body_details *new_type_details;
1143 const struct body_details *old_type_details
1144 = bodies_by_type + old_type;
1145 SV *referant = NULL;
1147 PERL_ARGS_ASSERT_SV_UPGRADE;
1149 if (old_type == new_type)
1152 /* This clause was purposefully added ahead of the early return above to
1153 the shared string hackery for (sort {$a <=> $b} keys %hash), with the
1154 inference by Nick I-S that it would fix other troublesome cases. See
1155 changes 7162, 7163 (f130fd4589cf5fbb24149cd4db4137c8326f49c1 and parent)
1157 Given that shared hash key scalars are no longer PVIV, but PV, there is
1158 no longer need to unshare so as to free up the IVX slot for its proper
1159 purpose. So it's safe to move the early return earlier. */
1161 if (new_type != SVt_PV && SvIsCOW(sv)) {
1162 sv_force_normal_flags(sv, 0);
1165 old_body = SvANY(sv);
1167 /* Copying structures onto other structures that have been neatly zeroed
1168 has a subtle gotcha. Consider XPVMG
1170 +------+------+------+------+------+-------+-------+
1171 | NV | CUR | LEN | IV | MAGIC | STASH |
1172 +------+------+------+------+------+-------+-------+
1173 0 4 8 12 16 20 24 28
1175 where NVs are aligned to 8 bytes, so that sizeof that structure is
1176 actually 32 bytes long, with 4 bytes of padding at the end:
1178 +------+------+------+------+------+-------+-------+------+
1179 | NV | CUR | LEN | IV | MAGIC | STASH | ??? |
1180 +------+------+------+------+------+-------+-------+------+
1181 0 4 8 12 16 20 24 28 32
1183 so what happens if you allocate memory for this structure:
1185 +------+------+------+------+------+-------+-------+------+------+...
1186 | NV | CUR | LEN | IV | MAGIC | STASH | GP | NAME |
1187 +------+------+------+------+------+-------+-------+------+------+...
1188 0 4 8 12 16 20 24 28 32 36
1190 zero it, then copy sizeof(XPVMG) bytes on top of it? Not quite what you
1191 expect, because you copy the area marked ??? onto GP. Now, ??? may have
1192 started out as zero once, but it's quite possible that it isn't. So now,
1193 rather than a nicely zeroed GP, you have it pointing somewhere random.
1196 (In fact, GP ends up pointing at a previous GP structure, because the
1197 principle cause of the padding in XPVMG getting garbage is a copy of
1198 sizeof(XPVMG) bytes from a XPVGV structure in sv_unglob. Right now
1199 this happens to be moot because XPVGV has been re-ordered, with GP
1200 no longer after STASH)
1202 So we are careful and work out the size of used parts of all the
1210 referant = SvRV(sv);
1211 old_type_details = &fake_rv;
1212 if (new_type == SVt_NV)
1213 new_type = SVt_PVNV;
1215 if (new_type < SVt_PVIV) {
1216 new_type = (new_type == SVt_NV)
1217 ? SVt_PVNV : SVt_PVIV;
1222 if (new_type < SVt_PVNV) {
1223 new_type = SVt_PVNV;
1227 assert(new_type > SVt_PV);
1228 assert(SVt_IV < SVt_PV);
1229 assert(SVt_NV < SVt_PV);
1236 /* Because the XPVMG of PL_mess_sv isn't allocated from the arena,
1237 there's no way that it can be safely upgraded, because perl.c
1238 expects to Safefree(SvANY(PL_mess_sv)) */
1239 assert(sv != PL_mess_sv);
1240 /* This flag bit is used to mean other things in other scalar types.
1241 Given that it only has meaning inside the pad, it shouldn't be set
1242 on anything that can get upgraded. */
1243 assert(!SvPAD_TYPED(sv));
1246 if (old_type_details->cant_upgrade)
1247 Perl_croak(aTHX_ "Can't upgrade %s (%" UVuf ") to %" UVuf,
1248 sv_reftype(sv, 0), (UV) old_type, (UV) new_type);
1251 if (old_type > new_type)
1252 Perl_croak(aTHX_ "sv_upgrade from type %d down to type %d",
1253 (int)old_type, (int)new_type);
1255 new_type_details = bodies_by_type + new_type;
1257 SvFLAGS(sv) &= ~SVTYPEMASK;
1258 SvFLAGS(sv) |= new_type;
1260 /* This can't happen, as SVt_NULL is <= all values of new_type, so one of
1261 the return statements above will have triggered. */
1262 assert (new_type != SVt_NULL);
1265 assert(old_type == SVt_NULL);
1266 SvANY(sv) = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
1270 assert(old_type == SVt_NULL);
1271 SvANY(sv) = new_XNV();
1276 assert(new_type_details->body_size);
1279 assert(new_type_details->arena);
1280 assert(new_type_details->arena_size);
1281 /* This points to the start of the allocated area. */
1282 new_body_inline(new_body, new_type);
1283 Zero(new_body, new_type_details->body_size, char);
1284 new_body = ((char *)new_body) - new_type_details->offset;
1286 /* We always allocated the full length item with PURIFY. To do this
1287 we fake things so that arena is false for all 16 types.. */
1288 new_body = new_NOARENAZ(new_type_details);
1290 SvANY(sv) = new_body;
1291 if (new_type == SVt_PVAV) {
1295 if (old_type_details->body_size) {
1298 /* It will have been zeroed when the new body was allocated.
1299 Lets not write to it, in case it confuses a write-back
1305 #ifndef NODEFAULT_SHAREKEYS
1306 HvSHAREKEYS_on(sv); /* key-sharing on by default */
1308 HvMAX(sv) = 7; /* (start with 8 buckets) */
1311 /* SVt_NULL isn't the only thing upgraded to AV or HV.
1312 The target created by newSVrv also is, and it can have magic.
1313 However, it never has SvPVX set.
1315 if (old_type == SVt_IV) {
1317 } else if (old_type >= SVt_PV) {
1318 assert(SvPVX_const(sv) == 0);
1321 if (old_type >= SVt_PVMG) {
1322 SvMAGIC_set(sv, ((XPVMG*)old_body)->xmg_u.xmg_magic);
1323 SvSTASH_set(sv, ((XPVMG*)old_body)->xmg_stash);
1325 sv->sv_u.svu_array = NULL; /* or svu_hash */
1331 /* This ensures that SvTHINKFIRST(sv) is true, and hence that
1332 sv_force_normal_flags(sv) is called. */
1335 /* XXX Is this still needed? Was it ever needed? Surely as there is
1336 no route from NV to PVIV, NOK can never be true */
1337 assert(!SvNOKp(sv));
1348 assert(new_type_details->body_size);
1349 /* We always allocated the full length item with PURIFY. To do this
1350 we fake things so that arena is false for all 16 types.. */
1351 if(new_type_details->arena) {
1352 /* This points to the start of the allocated area. */
1353 new_body_inline(new_body, new_type);
1354 Zero(new_body, new_type_details->body_size, char);
1355 new_body = ((char *)new_body) - new_type_details->offset;
1357 new_body = new_NOARENAZ(new_type_details);
1359 SvANY(sv) = new_body;
1361 if (old_type_details->copy) {
1362 /* There is now the potential for an upgrade from something without
1363 an offset (PVNV or PVMG) to something with one (PVCV, PVFM) */
1364 int offset = old_type_details->offset;
1365 int length = old_type_details->copy;
1367 if (new_type_details->offset > old_type_details->offset) {
1368 const int difference
1369 = new_type_details->offset - old_type_details->offset;
1370 offset += difference;
1371 length -= difference;
1373 assert (length >= 0);
1375 Copy((char *)old_body + offset, (char *)new_body + offset, length,
1379 #ifndef NV_ZERO_IS_ALLBITS_ZERO
1380 /* If NV 0.0 is stores as all bits 0 then Zero() already creates a
1381 * correct 0.0 for us. Otherwise, if the old body didn't have an
1382 * NV slot, but the new one does, then we need to initialise the
1383 * freshly created NV slot with whatever the correct bit pattern is
1385 if (old_type_details->zero_nv && !new_type_details->zero_nv
1386 && !isGV_with_GP(sv))
1390 if (new_type == SVt_PVIO) {
1391 IO * const io = MUTABLE_IO(sv);
1392 GV *iogv = gv_fetchpvs("IO::File::", GV_ADD, SVt_PVHV);
1395 /* Clear the stashcache because a new IO could overrule a package
1397 hv_clear(PL_stashcache);
1399 SvSTASH_set(io, MUTABLE_HV(SvREFCNT_inc(GvHV(iogv))));
1400 IoPAGE_LEN(sv) = 60;
1402 if (old_type < SVt_PV) {
1403 /* referant will be NULL unless the old type was SVt_IV emulating
1405 sv->sv_u.svu_rv = referant;
1409 Perl_croak(aTHX_ "panic: sv_upgrade to unknown type %lu",
1410 (unsigned long)new_type);
1413 if (old_type > SVt_IV) {
1417 /* Note that there is an assumption that all bodies of types that
1418 can be upgraded came from arenas. Only the more complex non-
1419 upgradable types are allowed to be directly malloc()ed. */
1420 assert(old_type_details->arena);
1421 del_body((void*)((char*)old_body + old_type_details->offset),
1422 &PL_body_roots[old_type]);
1428 =for apidoc sv_backoff
1430 Remove any string offset. You should normally use the C<SvOOK_off> macro
1437 Perl_sv_backoff(pTHX_ register SV *const sv)
1440 const char * const s = SvPVX_const(sv);
1442 PERL_ARGS_ASSERT_SV_BACKOFF;
1443 PERL_UNUSED_CONTEXT;
1446 assert(SvTYPE(sv) != SVt_PVHV);
1447 assert(SvTYPE(sv) != SVt_PVAV);
1449 SvOOK_offset(sv, delta);
1451 SvLEN_set(sv, SvLEN(sv) + delta);
1452 SvPV_set(sv, SvPVX(sv) - delta);
1453 Move(s, SvPVX(sv), SvCUR(sv)+1, char);
1454 SvFLAGS(sv) &= ~SVf_OOK;
1461 Expands the character buffer in the SV. If necessary, uses C<sv_unref> and
1462 upgrades the SV to C<SVt_PV>. Returns a pointer to the character buffer.
1463 Use the C<SvGROW> wrapper instead.
1469 Perl_sv_grow(pTHX_ register SV *const sv, register STRLEN newlen)
1473 PERL_ARGS_ASSERT_SV_GROW;
1475 if (PL_madskills && newlen >= 0x100000) {
1476 PerlIO_printf(Perl_debug_log,
1477 "Allocation too large: %"UVxf"\n", (UV)newlen);
1479 #ifdef HAS_64K_LIMIT
1480 if (newlen >= 0x10000) {
1481 PerlIO_printf(Perl_debug_log,
1482 "Allocation too large: %"UVxf"\n", (UV)newlen);
1485 #endif /* HAS_64K_LIMIT */
1488 if (SvTYPE(sv) < SVt_PV) {
1489 sv_upgrade(sv, SVt_PV);
1490 s = SvPVX_mutable(sv);
1492 else if (SvOOK(sv)) { /* pv is offset? */
1494 s = SvPVX_mutable(sv);
1495 if (newlen > SvLEN(sv))
1496 newlen += 10 * (newlen - SvCUR(sv)); /* avoid copy each time */
1497 #ifdef HAS_64K_LIMIT
1498 if (newlen >= 0x10000)
1503 s = SvPVX_mutable(sv);
1505 if (newlen > SvLEN(sv)) { /* need more room? */
1506 STRLEN minlen = SvCUR(sv);
1507 minlen += (minlen >> PERL_STRLEN_EXPAND_SHIFT) + 10;
1508 if (newlen < minlen)
1510 #ifndef Perl_safesysmalloc_size
1511 newlen = PERL_STRLEN_ROUNDUP(newlen);
1513 if (SvLEN(sv) && s) {
1514 s = (char*)saferealloc(s, newlen);
1517 s = (char*)safemalloc(newlen);
1518 if (SvPVX_const(sv) && SvCUR(sv)) {
1519 Move(SvPVX_const(sv), s, (newlen < SvCUR(sv)) ? newlen : SvCUR(sv), char);
1523 #ifdef Perl_safesysmalloc_size
1524 /* Do this here, do it once, do it right, and then we will never get
1525 called back into sv_grow() unless there really is some growing
1527 SvLEN_set(sv, Perl_safesysmalloc_size(s));
1529 SvLEN_set(sv, newlen);
1536 =for apidoc sv_setiv
1538 Copies an integer into the given SV, upgrading first if necessary.
1539 Does not handle 'set' magic. See also C<sv_setiv_mg>.
1545 Perl_sv_setiv(pTHX_ register SV *const sv, const IV i)
1549 PERL_ARGS_ASSERT_SV_SETIV;
1551 SV_CHECK_THINKFIRST_COW_DROP(sv);
1552 switch (SvTYPE(sv)) {
1555 sv_upgrade(sv, SVt_IV);
1558 sv_upgrade(sv, SVt_PVIV);
1562 if (!isGV_with_GP(sv))
1569 /* diag_listed_as: Can't coerce %s to %s in %s */
1570 Perl_croak(aTHX_ "Can't coerce %s to integer in %s", sv_reftype(sv,0),
1574 (void)SvIOK_only(sv); /* validate number */
1580 =for apidoc sv_setiv_mg
1582 Like C<sv_setiv>, but also handles 'set' magic.
1588 Perl_sv_setiv_mg(pTHX_ register SV *const sv, const IV i)
1590 PERL_ARGS_ASSERT_SV_SETIV_MG;
1597 =for apidoc sv_setuv
1599 Copies an unsigned integer into the given SV, upgrading first if necessary.
1600 Does not handle 'set' magic. See also C<sv_setuv_mg>.
1606 Perl_sv_setuv(pTHX_ register SV *const sv, const UV u)
1608 PERL_ARGS_ASSERT_SV_SETUV;
1610 /* With these two if statements:
1611 u=1.49 s=0.52 cu=72.49 cs=10.64 scripts=270 tests=20865
1614 u=1.35 s=0.47 cu=73.45 cs=11.43 scripts=270 tests=20865
1616 If you wish to remove them, please benchmark to see what the effect is
1618 if (u <= (UV)IV_MAX) {
1619 sv_setiv(sv, (IV)u);
1628 =for apidoc sv_setuv_mg
1630 Like C<sv_setuv>, but also handles 'set' magic.
1636 Perl_sv_setuv_mg(pTHX_ register SV *const sv, const UV u)
1638 PERL_ARGS_ASSERT_SV_SETUV_MG;
1645 =for apidoc sv_setnv
1647 Copies a double into the given SV, upgrading first if necessary.
1648 Does not handle 'set' magic. See also C<sv_setnv_mg>.
1654 Perl_sv_setnv(pTHX_ register SV *const sv, const NV num)
1658 PERL_ARGS_ASSERT_SV_SETNV;
1660 SV_CHECK_THINKFIRST_COW_DROP(sv);
1661 switch (SvTYPE(sv)) {
1664 sv_upgrade(sv, SVt_NV);
1668 sv_upgrade(sv, SVt_PVNV);
1672 if (!isGV_with_GP(sv))
1679 /* diag_listed_as: Can't coerce %s to %s in %s */
1680 Perl_croak(aTHX_ "Can't coerce %s to number in %s", sv_reftype(sv,0),
1685 (void)SvNOK_only(sv); /* validate number */
1690 =for apidoc sv_setnv_mg
1692 Like C<sv_setnv>, but also handles 'set' magic.
1698 Perl_sv_setnv_mg(pTHX_ register SV *const sv, const NV num)
1700 PERL_ARGS_ASSERT_SV_SETNV_MG;
1706 /* Print an "isn't numeric" warning, using a cleaned-up,
1707 * printable version of the offending string
1711 S_not_a_number(pTHX_ SV *const sv)
1718 PERL_ARGS_ASSERT_NOT_A_NUMBER;
1721 dsv = newSVpvs_flags("", SVs_TEMP);
1722 pv = sv_uni_display(dsv, sv, 10, 0);
1725 const char * const limit = tmpbuf + sizeof(tmpbuf) - 8;
1726 /* each *s can expand to 4 chars + "...\0",
1727 i.e. need room for 8 chars */
1729 const char *s = SvPVX_const(sv);
1730 const char * const end = s + SvCUR(sv);
1731 for ( ; s < end && d < limit; s++ ) {
1733 if (ch & 128 && !isPRINT_LC(ch)) {
1742 else if (ch == '\r') {
1746 else if (ch == '\f') {
1750 else if (ch == '\\') {
1754 else if (ch == '\0') {
1758 else if (isPRINT_LC(ch))
1775 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
1776 "Argument \"%s\" isn't numeric in %s", pv,
1779 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
1780 "Argument \"%s\" isn't numeric", pv);
1784 =for apidoc looks_like_number
1786 Test if the content of an SV looks like a number (or is a number).
1787 C<Inf> and C<Infinity> are treated as numbers (so will not issue a
1788 non-numeric warning), even if your atof() doesn't grok them.
1794 Perl_looks_like_number(pTHX_ SV *const sv)
1796 register const char *sbegin;
1799 PERL_ARGS_ASSERT_LOOKS_LIKE_NUMBER;
1802 sbegin = SvPVX_const(sv);
1805 else if (SvPOKp(sv))
1806 sbegin = SvPV_const(sv, len);
1808 return SvFLAGS(sv) & (SVf_NOK|SVp_NOK|SVf_IOK|SVp_IOK);
1809 return grok_number(sbegin, len, NULL);
1813 S_glob_2number(pTHX_ GV * const gv)
1815 const U32 wasfake = SvFLAGS(gv) & SVf_FAKE;
1816 SV *const buffer = sv_newmortal();
1818 PERL_ARGS_ASSERT_GLOB_2NUMBER;
1820 /* FAKE globs can get coerced, so need to turn this off temporarily if it
1823 gv_efullname3(buffer, gv, "*");
1824 SvFLAGS(gv) |= wasfake;
1826 /* We know that all GVs stringify to something that is not-a-number,
1827 so no need to test that. */
1828 if (ckWARN(WARN_NUMERIC))
1829 not_a_number(buffer);
1830 /* We just want something true to return, so that S_sv_2iuv_common
1831 can tail call us and return true. */
1835 /* Actually, ISO C leaves conversion of UV to IV undefined, but
1836 until proven guilty, assume that things are not that bad... */
1841 As 64 bit platforms often have an NV that doesn't preserve all bits of
1842 an IV (an assumption perl has been based on to date) it becomes necessary
1843 to remove the assumption that the NV always carries enough precision to
1844 recreate the IV whenever needed, and that the NV is the canonical form.
1845 Instead, IV/UV and NV need to be given equal rights. So as to not lose
1846 precision as a side effect of conversion (which would lead to insanity
1847 and the dragon(s) in t/op/numconvert.t getting very angry) the intent is
1848 1) to distinguish between IV/UV/NV slots that have cached a valid
1849 conversion where precision was lost and IV/UV/NV slots that have a
1850 valid conversion which has lost no precision
1851 2) to ensure that if a numeric conversion to one form is requested that
1852 would lose precision, the precise conversion (or differently
1853 imprecise conversion) is also performed and cached, to prevent
1854 requests for different numeric formats on the same SV causing
1855 lossy conversion chains. (lossless conversion chains are perfectly
1860 SvIOKp is true if the IV slot contains a valid value
1861 SvIOK is true only if the IV value is accurate (UV if SvIOK_UV true)
1862 SvNOKp is true if the NV slot contains a valid value
1863 SvNOK is true only if the NV value is accurate
1866 while converting from PV to NV, check to see if converting that NV to an
1867 IV(or UV) would lose accuracy over a direct conversion from PV to
1868 IV(or UV). If it would, cache both conversions, return NV, but mark
1869 SV as IOK NOKp (ie not NOK).
1871 While converting from PV to IV, check to see if converting that IV to an
1872 NV would lose accuracy over a direct conversion from PV to NV. If it
1873 would, cache both conversions, flag similarly.
1875 Before, the SV value "3.2" could become NV=3.2 IV=3 NOK, IOK quite
1876 correctly because if IV & NV were set NV *always* overruled.
1877 Now, "3.2" will become NV=3.2 IV=3 NOK, IOKp, because the flag's meaning
1878 changes - now IV and NV together means that the two are interchangeable:
1879 SvIVX == (IV) SvNVX && SvNVX == (NV) SvIVX;
1881 The benefit of this is that operations such as pp_add know that if
1882 SvIOK is true for both left and right operands, then integer addition
1883 can be used instead of floating point (for cases where the result won't
1884 overflow). Before, floating point was always used, which could lead to
1885 loss of precision compared with integer addition.
1887 * making IV and NV equal status should make maths accurate on 64 bit
1889 * may speed up maths somewhat if pp_add and friends start to use
1890 integers when possible instead of fp. (Hopefully the overhead in
1891 looking for SvIOK and checking for overflow will not outweigh the
1892 fp to integer speedup)
1893 * will slow down integer operations (callers of SvIV) on "inaccurate"
1894 values, as the change from SvIOK to SvIOKp will cause a call into
1895 sv_2iv each time rather than a macro access direct to the IV slot
1896 * should speed up number->string conversion on integers as IV is
1897 favoured when IV and NV are equally accurate
1899 ####################################################################
1900 You had better be using SvIOK_notUV if you want an IV for arithmetic:
1901 SvIOK is true if (IV or UV), so you might be getting (IV)SvUV.
1902 On the other hand, SvUOK is true iff UV.
1903 ####################################################################
1905 Your mileage will vary depending your CPU's relative fp to integer
1909 #ifndef NV_PRESERVES_UV
1910 # define IS_NUMBER_UNDERFLOW_IV 1
1911 # define IS_NUMBER_UNDERFLOW_UV 2
1912 # define IS_NUMBER_IV_AND_UV 2
1913 # define IS_NUMBER_OVERFLOW_IV 4
1914 # define IS_NUMBER_OVERFLOW_UV 5
1916 /* sv_2iuv_non_preserve(): private routine for use by sv_2iv() and sv_2uv() */
1918 /* For sv_2nv these three cases are "SvNOK and don't bother casting" */
1920 S_sv_2iuv_non_preserve(pTHX_ register SV *const sv
1928 PERL_ARGS_ASSERT_SV_2IUV_NON_PRESERVE;
1930 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_2iuv_non '%s', IV=0x%"UVxf" NV=%"NVgf" inttype=%"UVXf"\n", SvPVX_const(sv), SvIVX(sv), SvNVX(sv), (UV)numtype));
1931 if (SvNVX(sv) < (NV)IV_MIN) {
1932 (void)SvIOKp_on(sv);
1934 SvIV_set(sv, IV_MIN);
1935 return IS_NUMBER_UNDERFLOW_IV;
1937 if (SvNVX(sv) > (NV)UV_MAX) {
1938 (void)SvIOKp_on(sv);
1941 SvUV_set(sv, UV_MAX);
1942 return IS_NUMBER_OVERFLOW_UV;
1944 (void)SvIOKp_on(sv);
1946 /* Can't use strtol etc to convert this string. (See truth table in
1948 if (SvNVX(sv) <= (UV)IV_MAX) {
1949 SvIV_set(sv, I_V(SvNVX(sv)));
1950 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
1951 SvIOK_on(sv); /* Integer is precise. NOK, IOK */
1953 /* Integer is imprecise. NOK, IOKp */
1955 return SvNVX(sv) < 0 ? IS_NUMBER_UNDERFLOW_UV : IS_NUMBER_IV_AND_UV;
1958 SvUV_set(sv, U_V(SvNVX(sv)));
1959 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
1960 if (SvUVX(sv) == UV_MAX) {
1961 /* As we know that NVs don't preserve UVs, UV_MAX cannot
1962 possibly be preserved by NV. Hence, it must be overflow.
1964 return IS_NUMBER_OVERFLOW_UV;
1966 SvIOK_on(sv); /* Integer is precise. NOK, UOK */
1968 /* Integer is imprecise. NOK, IOKp */
1970 return IS_NUMBER_OVERFLOW_IV;
1972 #endif /* !NV_PRESERVES_UV*/
1975 S_sv_2iuv_common(pTHX_ SV *const sv)
1979 PERL_ARGS_ASSERT_SV_2IUV_COMMON;
1982 /* erm. not sure. *should* never get NOKp (without NOK) from sv_2nv
1983 * without also getting a cached IV/UV from it at the same time
1984 * (ie PV->NV conversion should detect loss of accuracy and cache
1985 * IV or UV at same time to avoid this. */
1986 /* IV-over-UV optimisation - choose to cache IV if possible */
1988 if (SvTYPE(sv) == SVt_NV)
1989 sv_upgrade(sv, SVt_PVNV);
1991 (void)SvIOKp_on(sv); /* Must do this first, to clear any SvOOK */
1992 /* < not <= as for NV doesn't preserve UV, ((NV)IV_MAX+1) will almost
1993 certainly cast into the IV range at IV_MAX, whereas the correct
1994 answer is the UV IV_MAX +1. Hence < ensures that dodgy boundary
1996 #if defined(NAN_COMPARE_BROKEN) && defined(Perl_isnan)
1997 if (Perl_isnan(SvNVX(sv))) {
2003 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2004 SvIV_set(sv, I_V(SvNVX(sv)));
2005 if (SvNVX(sv) == (NV) SvIVX(sv)
2006 #ifndef NV_PRESERVES_UV
2007 && (((UV)1 << NV_PRESERVES_UV_BITS) >
2008 (UV)(SvIVX(sv) > 0 ? SvIVX(sv) : -SvIVX(sv)))
2009 /* Don't flag it as "accurately an integer" if the number
2010 came from a (by definition imprecise) NV operation, and
2011 we're outside the range of NV integer precision */
2015 SvIOK_on(sv); /* Can this go wrong with rounding? NWC */
2017 /* scalar has trailing garbage, eg "42a" */
2019 DEBUG_c(PerlIO_printf(Perl_debug_log,
2020 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (precise)\n",
2026 /* IV not precise. No need to convert from PV, as NV
2027 conversion would already have cached IV if it detected
2028 that PV->IV would be better than PV->NV->IV
2029 flags already correct - don't set public IOK. */
2030 DEBUG_c(PerlIO_printf(Perl_debug_log,
2031 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (imprecise)\n",
2036 /* Can the above go wrong if SvIVX == IV_MIN and SvNVX < IV_MIN,
2037 but the cast (NV)IV_MIN rounds to a the value less (more
2038 negative) than IV_MIN which happens to be equal to SvNVX ??
2039 Analogous to 0xFFFFFFFFFFFFFFFF rounding up to NV (2**64) and
2040 NV rounding back to 0xFFFFFFFFFFFFFFFF, so UVX == UV(NVX) and
2041 (NV)UVX == NVX are both true, but the values differ. :-(
2042 Hopefully for 2s complement IV_MIN is something like
2043 0x8000000000000000 which will be exact. NWC */
2046 SvUV_set(sv, U_V(SvNVX(sv)));
2048 (SvNVX(sv) == (NV) SvUVX(sv))
2049 #ifndef NV_PRESERVES_UV
2050 /* Make sure it's not 0xFFFFFFFFFFFFFFFF */
2051 /*&& (SvUVX(sv) != UV_MAX) irrelevant with code below */
2052 && (((UV)1 << NV_PRESERVES_UV_BITS) > SvUVX(sv))
2053 /* Don't flag it as "accurately an integer" if the number
2054 came from a (by definition imprecise) NV operation, and
2055 we're outside the range of NV integer precision */
2061 DEBUG_c(PerlIO_printf(Perl_debug_log,
2062 "0x%"UVxf" 2iv(%"UVuf" => %"IVdf") (as unsigned)\n",
2068 else if (SvPOKp(sv) && SvLEN(sv)) {
2070 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2071 /* We want to avoid a possible problem when we cache an IV/ a UV which
2072 may be later translated to an NV, and the resulting NV is not
2073 the same as the direct translation of the initial string
2074 (eg 123.456 can shortcut to the IV 123 with atol(), but we must
2075 be careful to ensure that the value with the .456 is around if the
2076 NV value is requested in the future).
2078 This means that if we cache such an IV/a UV, we need to cache the
2079 NV as well. Moreover, we trade speed for space, and do not
2080 cache the NV if we are sure it's not needed.
2083 /* SVt_PVNV is one higher than SVt_PVIV, hence this order */
2084 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2085 == IS_NUMBER_IN_UV) {
2086 /* It's definitely an integer, only upgrade to PVIV */
2087 if (SvTYPE(sv) < SVt_PVIV)
2088 sv_upgrade(sv, SVt_PVIV);
2090 } else if (SvTYPE(sv) < SVt_PVNV)
2091 sv_upgrade(sv, SVt_PVNV);
2093 /* If NVs preserve UVs then we only use the UV value if we know that
2094 we aren't going to call atof() below. If NVs don't preserve UVs
2095 then the value returned may have more precision than atof() will
2096 return, even though value isn't perfectly accurate. */
2097 if ((numtype & (IS_NUMBER_IN_UV
2098 #ifdef NV_PRESERVES_UV
2101 )) == IS_NUMBER_IN_UV) {
2102 /* This won't turn off the public IOK flag if it was set above */
2103 (void)SvIOKp_on(sv);
2105 if (!(numtype & IS_NUMBER_NEG)) {
2107 if (value <= (UV)IV_MAX) {
2108 SvIV_set(sv, (IV)value);
2110 /* it didn't overflow, and it was positive. */
2111 SvUV_set(sv, value);
2115 /* 2s complement assumption */
2116 if (value <= (UV)IV_MIN) {
2117 SvIV_set(sv, -(IV)value);
2119 /* Too negative for an IV. This is a double upgrade, but
2120 I'm assuming it will be rare. */
2121 if (SvTYPE(sv) < SVt_PVNV)
2122 sv_upgrade(sv, SVt_PVNV);
2126 SvNV_set(sv, -(NV)value);
2127 SvIV_set(sv, IV_MIN);
2131 /* For !NV_PRESERVES_UV and IS_NUMBER_IN_UV and IS_NUMBER_NOT_INT we
2132 will be in the previous block to set the IV slot, and the next
2133 block to set the NV slot. So no else here. */
2135 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2136 != IS_NUMBER_IN_UV) {
2137 /* It wasn't an (integer that doesn't overflow the UV). */
2138 SvNV_set(sv, Atof(SvPVX_const(sv)));
2140 if (! numtype && ckWARN(WARN_NUMERIC))
2143 #if defined(USE_LONG_DOUBLE)
2144 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%" PERL_PRIgldbl ")\n",
2145 PTR2UV(sv), SvNVX(sv)));
2147 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"NVgf")\n",
2148 PTR2UV(sv), SvNVX(sv)));
2151 #ifdef NV_PRESERVES_UV
2152 (void)SvIOKp_on(sv);
2154 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2155 SvIV_set(sv, I_V(SvNVX(sv)));
2156 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
2159 NOOP; /* Integer is imprecise. NOK, IOKp */
2161 /* UV will not work better than IV */
2163 if (SvNVX(sv) > (NV)UV_MAX) {
2165 /* Integer is inaccurate. NOK, IOKp, is UV */
2166 SvUV_set(sv, UV_MAX);
2168 SvUV_set(sv, U_V(SvNVX(sv)));
2169 /* 0xFFFFFFFFFFFFFFFF not an issue in here, NVs
2170 NV preservse UV so can do correct comparison. */
2171 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
2174 NOOP; /* Integer is imprecise. NOK, IOKp, is UV */
2179 #else /* NV_PRESERVES_UV */
2180 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2181 == (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT)) {
2182 /* The IV/UV slot will have been set from value returned by
2183 grok_number above. The NV slot has just been set using
2186 assert (SvIOKp(sv));
2188 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2189 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2190 /* Small enough to preserve all bits. */
2191 (void)SvIOKp_on(sv);
2193 SvIV_set(sv, I_V(SvNVX(sv)));
2194 if ((NV)(SvIVX(sv)) == SvNVX(sv))
2196 /* Assumption: first non-preserved integer is < IV_MAX,
2197 this NV is in the preserved range, therefore: */
2198 if (!(U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))
2200 Perl_croak(aTHX_ "sv_2iv assumed (U_V(fabs((double)SvNVX(sv))) < (UV)IV_MAX) but SvNVX(sv)=%"NVgf" U_V is 0x%"UVxf", IV_MAX is 0x%"UVxf"\n", SvNVX(sv), U_V(SvNVX(sv)), (UV)IV_MAX);
2204 0 0 already failed to read UV.
2205 0 1 already failed to read UV.
2206 1 0 you won't get here in this case. IV/UV
2207 slot set, public IOK, Atof() unneeded.
2208 1 1 already read UV.
2209 so there's no point in sv_2iuv_non_preserve() attempting
2210 to use atol, strtol, strtoul etc. */
2212 sv_2iuv_non_preserve (sv, numtype);
2214 sv_2iuv_non_preserve (sv);
2218 #endif /* NV_PRESERVES_UV */
2219 /* It might be more code efficient to go through the entire logic above
2220 and conditionally set with SvIOKp_on() rather than SvIOK(), but it
2221 gets complex and potentially buggy, so more programmer efficient
2222 to do it this way, by turning off the public flags: */
2224 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK);
2228 if (isGV_with_GP(sv))
2229 return glob_2number(MUTABLE_GV(sv));
2231 if (!(SvFLAGS(sv) & SVs_PADTMP)) {
2232 if (!PL_localizing && ckWARN(WARN_UNINITIALIZED))
2235 if (SvTYPE(sv) < SVt_IV)
2236 /* Typically the caller expects that sv_any is not NULL now. */
2237 sv_upgrade(sv, SVt_IV);
2238 /* Return 0 from the caller. */
2245 =for apidoc sv_2iv_flags
2247 Return the integer value of an SV, doing any necessary string
2248 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2249 Normally used via the C<SvIV(sv)> and C<SvIVx(sv)> macros.
2255 Perl_sv_2iv_flags(pTHX_ register SV *const sv, const I32 flags)
2260 if (SvGMAGICAL(sv) || SvVALID(sv)) {
2261 /* FBMs use the space for SvIVX and SvNVX for other purposes, and use
2262 the same flag bit as SVf_IVisUV, so must not let them cache IVs.
2263 In practice they are extremely unlikely to actually get anywhere
2264 accessible by user Perl code - the only way that I'm aware of is when
2265 a constant subroutine which is used as the second argument to index.
2267 if (flags & SV_GMAGIC)
2272 return I_V(SvNVX(sv));
2274 if (SvPOKp(sv) && SvLEN(sv)) {
2277 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2279 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2280 == IS_NUMBER_IN_UV) {
2281 /* It's definitely an integer */
2282 if (numtype & IS_NUMBER_NEG) {
2283 if (value < (UV)IV_MIN)
2286 if (value < (UV)IV_MAX)
2291 if (ckWARN(WARN_NUMERIC))
2294 return I_V(Atof(SvPVX_const(sv)));
2299 assert(SvTYPE(sv) >= SVt_PVMG);
2300 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
2301 } else if (SvTHINKFIRST(sv)) {
2306 if (flags & SV_SKIP_OVERLOAD)
2308 tmpstr = AMG_CALLunary(sv, numer_amg);
2309 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2310 return SvIV(tmpstr);
2313 return PTR2IV(SvRV(sv));
2316 sv_force_normal_flags(sv, 0);
2318 if (SvREADONLY(sv) && !SvOK(sv)) {
2319 if (ckWARN(WARN_UNINITIALIZED))
2325 if (S_sv_2iuv_common(aTHX_ sv))
2328 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"IVdf")\n",
2329 PTR2UV(sv),SvIVX(sv)));
2330 return SvIsUV(sv) ? (IV)SvUVX(sv) : SvIVX(sv);
2334 =for apidoc sv_2uv_flags
2336 Return the unsigned integer value of an SV, doing any necessary string
2337 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2338 Normally used via the C<SvUV(sv)> and C<SvUVx(sv)> macros.
2344 Perl_sv_2uv_flags(pTHX_ register SV *const sv, const I32 flags)
2349 if (SvGMAGICAL(sv) || SvVALID(sv)) {
2350 /* FBMs use the space for SvIVX and SvNVX for other purposes, and use
2351 the same flag bit as SVf_IVisUV, so must not let them cache IVs. */
2352 if (flags & SV_GMAGIC)
2357 return U_V(SvNVX(sv));
2358 if (SvPOKp(sv) && SvLEN(sv)) {
2361 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2363 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2364 == IS_NUMBER_IN_UV) {
2365 /* It's definitely an integer */
2366 if (!(numtype & IS_NUMBER_NEG))
2370 if (ckWARN(WARN_NUMERIC))
2373 return U_V(Atof(SvPVX_const(sv)));
2378 assert(SvTYPE(sv) >= SVt_PVMG);
2379 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
2380 } else if (SvTHINKFIRST(sv)) {
2385 if (flags & SV_SKIP_OVERLOAD)
2387 tmpstr = AMG_CALLunary(sv, numer_amg);
2388 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2389 return SvUV(tmpstr);
2392 return PTR2UV(SvRV(sv));
2395 sv_force_normal_flags(sv, 0);
2397 if (SvREADONLY(sv) && !SvOK(sv)) {
2398 if (ckWARN(WARN_UNINITIALIZED))
2404 if (S_sv_2iuv_common(aTHX_ sv))
2408 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2uv(%"UVuf")\n",
2409 PTR2UV(sv),SvUVX(sv)));
2410 return SvIsUV(sv) ? SvUVX(sv) : (UV)SvIVX(sv);
2414 =for apidoc sv_2nv_flags
2416 Return the num value of an SV, doing any necessary string or integer
2417 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2418 Normally used via the C<SvNV(sv)> and C<SvNVx(sv)> macros.
2424 Perl_sv_2nv_flags(pTHX_ register SV *const sv, const I32 flags)
2429 if (SvGMAGICAL(sv) || SvVALID(sv)) {
2430 /* FBMs use the space for SvIVX and SvNVX for other purposes, and use
2431 the same flag bit as SVf_IVisUV, so must not let them cache NVs. */
2432 if (flags & SV_GMAGIC)
2436 if ((SvPOKp(sv) && SvLEN(sv)) && !SvIOKp(sv)) {
2437 if (!SvIOKp(sv) && ckWARN(WARN_NUMERIC) &&
2438 !grok_number(SvPVX_const(sv), SvCUR(sv), NULL))
2440 return Atof(SvPVX_const(sv));
2444 return (NV)SvUVX(sv);
2446 return (NV)SvIVX(sv);
2451 assert(SvTYPE(sv) >= SVt_PVMG);
2452 /* This falls through to the report_uninit near the end of the
2454 } else if (SvTHINKFIRST(sv)) {
2459 if (flags & SV_SKIP_OVERLOAD)
2461 tmpstr = AMG_CALLunary(sv, numer_amg);
2462 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2463 return SvNV(tmpstr);
2466 return PTR2NV(SvRV(sv));
2469 sv_force_normal_flags(sv, 0);
2471 if (SvREADONLY(sv) && !SvOK(sv)) {
2472 if (ckWARN(WARN_UNINITIALIZED))
2477 if (SvTYPE(sv) < SVt_NV) {
2478 /* The logic to use SVt_PVNV if necessary is in sv_upgrade. */
2479 sv_upgrade(sv, SVt_NV);
2480 #ifdef USE_LONG_DOUBLE
2482 STORE_NUMERIC_LOCAL_SET_STANDARD();
2483 PerlIO_printf(Perl_debug_log,
2484 "0x%"UVxf" num(%" PERL_PRIgldbl ")\n",
2485 PTR2UV(sv), SvNVX(sv));
2486 RESTORE_NUMERIC_LOCAL();
2490 STORE_NUMERIC_LOCAL_SET_STANDARD();
2491 PerlIO_printf(Perl_debug_log, "0x%"UVxf" num(%"NVgf")\n",
2492 PTR2UV(sv), SvNVX(sv));
2493 RESTORE_NUMERIC_LOCAL();
2497 else if (SvTYPE(sv) < SVt_PVNV)
2498 sv_upgrade(sv, SVt_PVNV);
2503 SvNV_set(sv, SvIsUV(sv) ? (NV)SvUVX(sv) : (NV)SvIVX(sv));
2504 #ifdef NV_PRESERVES_UV
2510 /* Only set the public NV OK flag if this NV preserves the IV */
2511 /* Check it's not 0xFFFFFFFFFFFFFFFF */
2513 SvIsUV(sv) ? ((SvUVX(sv) != UV_MAX)&&(SvUVX(sv) == U_V(SvNVX(sv))))
2514 : (SvIVX(sv) == I_V(SvNVX(sv))))
2520 else if (SvPOKp(sv) && SvLEN(sv)) {
2522 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2523 if (!SvIOKp(sv) && !numtype && ckWARN(WARN_NUMERIC))
2525 #ifdef NV_PRESERVES_UV
2526 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2527 == IS_NUMBER_IN_UV) {
2528 /* It's definitely an integer */
2529 SvNV_set(sv, (numtype & IS_NUMBER_NEG) ? -(NV)value : (NV)value);
2531 SvNV_set(sv, Atof(SvPVX_const(sv)));
2537 SvNV_set(sv, Atof(SvPVX_const(sv)));
2538 /* Only set the public NV OK flag if this NV preserves the value in
2539 the PV at least as well as an IV/UV would.
2540 Not sure how to do this 100% reliably. */
2541 /* if that shift count is out of range then Configure's test is
2542 wonky. We shouldn't be in here with NV_PRESERVES_UV_BITS ==
2544 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2545 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2546 SvNOK_on(sv); /* Definitely small enough to preserve all bits */
2547 } else if (!(numtype & IS_NUMBER_IN_UV)) {
2548 /* Can't use strtol etc to convert this string, so don't try.
2549 sv_2iv and sv_2uv will use the NV to convert, not the PV. */
2552 /* value has been set. It may not be precise. */
2553 if ((numtype & IS_NUMBER_NEG) && (value > (UV)IV_MIN)) {
2554 /* 2s complement assumption for (UV)IV_MIN */
2555 SvNOK_on(sv); /* Integer is too negative. */
2560 if (numtype & IS_NUMBER_NEG) {
2561 SvIV_set(sv, -(IV)value);
2562 } else if (value <= (UV)IV_MAX) {
2563 SvIV_set(sv, (IV)value);
2565 SvUV_set(sv, value);
2569 if (numtype & IS_NUMBER_NOT_INT) {
2570 /* I believe that even if the original PV had decimals,
2571 they are lost beyond the limit of the FP precision.
2572 However, neither is canonical, so both only get p
2573 flags. NWC, 2000/11/25 */
2574 /* Both already have p flags, so do nothing */
2576 const NV nv = SvNVX(sv);
2577 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2578 if (SvIVX(sv) == I_V(nv)) {
2581 /* It had no "." so it must be integer. */
2585 /* between IV_MAX and NV(UV_MAX).
2586 Could be slightly > UV_MAX */
2588 if (numtype & IS_NUMBER_NOT_INT) {
2589 /* UV and NV both imprecise. */
2591 const UV nv_as_uv = U_V(nv);
2593 if (value == nv_as_uv && SvUVX(sv) != UV_MAX) {
2602 /* It might be more code efficient to go through the entire logic above
2603 and conditionally set with SvNOKp_on() rather than SvNOK(), but it
2604 gets complex and potentially buggy, so more programmer efficient
2605 to do it this way, by turning off the public flags: */
2607 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK);
2608 #endif /* NV_PRESERVES_UV */
2611 if (isGV_with_GP(sv)) {
2612 glob_2number(MUTABLE_GV(sv));
2616 if (!PL_localizing && !(SvFLAGS(sv) & SVs_PADTMP) && ckWARN(WARN_UNINITIALIZED))
2618 assert (SvTYPE(sv) >= SVt_NV);
2619 /* Typically the caller expects that sv_any is not NULL now. */
2620 /* XXX Ilya implies that this is a bug in callers that assume this
2621 and ideally should be fixed. */
2624 #if defined(USE_LONG_DOUBLE)
2626 STORE_NUMERIC_LOCAL_SET_STANDARD();
2627 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2nv(%" PERL_PRIgldbl ")\n",
2628 PTR2UV(sv), SvNVX(sv));
2629 RESTORE_NUMERIC_LOCAL();
2633 STORE_NUMERIC_LOCAL_SET_STANDARD();
2634 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 1nv(%"NVgf")\n",
2635 PTR2UV(sv), SvNVX(sv));
2636 RESTORE_NUMERIC_LOCAL();
2645 Return an SV with the numeric value of the source SV, doing any necessary
2646 reference or overload conversion. You must use the C<SvNUM(sv)> macro to
2647 access this function.
2653 Perl_sv_2num(pTHX_ register SV *const sv)
2655 PERL_ARGS_ASSERT_SV_2NUM;
2660 SV * const tmpsv = AMG_CALLunary(sv, numer_amg);
2661 TAINT_IF(tmpsv && SvTAINTED(tmpsv));
2662 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
2663 return sv_2num(tmpsv);
2665 return sv_2mortal(newSVuv(PTR2UV(SvRV(sv))));
2668 /* uiv_2buf(): private routine for use by sv_2pv_flags(): print an IV or
2669 * UV as a string towards the end of buf, and return pointers to start and
2672 * We assume that buf is at least TYPE_CHARS(UV) long.
2676 S_uiv_2buf(char *const buf, const IV iv, UV uv, const int is_uv, char **const peob)
2678 char *ptr = buf + TYPE_CHARS(UV);
2679 char * const ebuf = ptr;
2682 PERL_ARGS_ASSERT_UIV_2BUF;
2694 *--ptr = '0' + (char)(uv % 10);
2703 =for apidoc sv_2pv_flags
2705 Returns a pointer to the string value of an SV, and sets *lp to its length.
2706 If flags includes SV_GMAGIC, does an mg_get() first. Coerces sv to a string
2708 Normally invoked via the C<SvPV_flags> macro. C<sv_2pv()> and C<sv_2pv_nomg>
2709 usually end up here too.
2715 Perl_sv_2pv_flags(pTHX_ register SV *const sv, STRLEN *const lp, const I32 flags)
2725 if (SvGMAGICAL(sv)) {
2726 if (flags & SV_GMAGIC)
2731 if (flags & SV_MUTABLE_RETURN)
2732 return SvPVX_mutable(sv);
2733 if (flags & SV_CONST_RETURN)
2734 return (char *)SvPVX_const(sv);
2737 if (SvIOKp(sv) || SvNOKp(sv)) {
2738 char tbuf[64]; /* Must fit sprintf/Gconvert of longest IV/NV */
2743 ? my_snprintf(tbuf, sizeof(tbuf), "%"UVuf, (UV)SvUVX(sv))
2744 : my_snprintf(tbuf, sizeof(tbuf), "%"IVdf, (IV)SvIVX(sv));
2745 } else if(SvNVX(sv) == 0.0) {
2750 Gconvert(SvNVX(sv), NV_DIG, 0, tbuf);
2757 SvUPGRADE(sv, SVt_PV);
2760 s = SvGROW_mutable(sv, len + 1);
2763 return (char*)memcpy(s, tbuf, len + 1);
2769 assert(SvTYPE(sv) >= SVt_PVMG);
2770 /* This falls through to the report_uninit near the end of the
2772 } else if (SvTHINKFIRST(sv)) {
2777 if (flags & SV_SKIP_OVERLOAD)
2779 tmpstr = AMG_CALLunary(sv, string_amg);
2780 TAINT_IF(tmpstr && SvTAINTED(tmpstr));
2781 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2783 /* char *pv = lp ? SvPV(tmpstr, *lp) : SvPV_nolen(tmpstr);
2787 if ((SvFLAGS(tmpstr) & (SVf_POK)) == SVf_POK) {
2788 if (flags & SV_CONST_RETURN) {
2789 pv = (char *) SvPVX_const(tmpstr);
2791 pv = (flags & SV_MUTABLE_RETURN)
2792 ? SvPVX_mutable(tmpstr) : SvPVX(tmpstr);
2795 *lp = SvCUR(tmpstr);
2797 pv = sv_2pv_flags(tmpstr, lp, flags);
2810 SV *const referent = SvRV(sv);
2814 retval = buffer = savepvn("NULLREF", len);
2815 } else if (SvTYPE(referent) == SVt_REGEXP) {
2816 REGEXP * const re = (REGEXP *)MUTABLE_PTR(referent);
2821 /* If the regex is UTF-8 we want the containing scalar to
2822 have an UTF-8 flag too */
2828 if ((seen_evals = RX_SEEN_EVALS(re)))
2829 PL_reginterp_cnt += seen_evals;
2832 *lp = RX_WRAPLEN(re);
2834 return RX_WRAPPED(re);
2836 const char *const typestr = sv_reftype(referent, 0);
2837 const STRLEN typelen = strlen(typestr);
2838 UV addr = PTR2UV(referent);
2839 const char *stashname = NULL;
2840 STRLEN stashnamelen = 0; /* hush, gcc */
2841 const char *buffer_end;
2843 if (SvOBJECT(referent)) {
2844 const HEK *const name = HvNAME_HEK(SvSTASH(referent));
2847 stashname = HEK_KEY(name);
2848 stashnamelen = HEK_LEN(name);
2850 if (HEK_UTF8(name)) {
2856 stashname = "__ANON__";
2859 len = stashnamelen + 1 /* = */ + typelen + 3 /* (0x */
2860 + 2 * sizeof(UV) + 2 /* )\0 */;
2862 len = typelen + 3 /* (0x */
2863 + 2 * sizeof(UV) + 2 /* )\0 */;
2866 Newx(buffer, len, char);
2867 buffer_end = retval = buffer + len;
2869 /* Working backwards */
2873 *--retval = PL_hexdigit[addr & 15];
2874 } while (addr >>= 4);
2880 memcpy(retval, typestr, typelen);
2884 retval -= stashnamelen;
2885 memcpy(retval, stashname, stashnamelen);
2887 /* retval may not necessarily have reached the start of the
2889 assert (retval >= buffer);
2891 len = buffer_end - retval - 1; /* -1 for that \0 */
2899 if (SvREADONLY(sv) && !SvOK(sv)) {
2902 if (flags & SV_UNDEF_RETURNS_NULL)
2904 if (ckWARN(WARN_UNINITIALIZED))
2909 if (SvIOK(sv) || ((SvIOKp(sv) && !SvNOKp(sv)))) {
2910 /* I'm assuming that if both IV and NV are equally valid then
2911 converting the IV is going to be more efficient */
2912 const U32 isUIOK = SvIsUV(sv);
2913 char buf[TYPE_CHARS(UV)];
2917 if (SvTYPE(sv) < SVt_PVIV)
2918 sv_upgrade(sv, SVt_PVIV);
2919 ptr = uiv_2buf(buf, SvIVX(sv), SvUVX(sv), isUIOK, &ebuf);
2921 /* inlined from sv_setpvn */
2922 s = SvGROW_mutable(sv, len + 1);
2923 Move(ptr, s, len, char);
2927 else if (SvNOKp(sv)) {
2928 if (SvTYPE(sv) < SVt_PVNV)
2929 sv_upgrade(sv, SVt_PVNV);
2930 if (SvNVX(sv) == 0.0) {
2931 s = SvGROW_mutable(sv, 2);
2936 /* The +20 is pure guesswork. Configure test needed. --jhi */
2937 s = SvGROW_mutable(sv, NV_DIG + 20);
2938 /* some Xenix systems wipe out errno here */
2939 Gconvert(SvNVX(sv), NV_DIG, 0, s);
2949 if (isGV_with_GP(sv)) {
2950 GV *const gv = MUTABLE_GV(sv);
2951 const U32 wasfake = SvFLAGS(gv) & SVf_FAKE;
2952 SV *const buffer = sv_newmortal();
2954 /* FAKE globs can get coerced, so need to turn this off temporarily
2957 gv_efullname3(buffer, gv, "*");
2958 SvFLAGS(gv) |= wasfake;
2960 if (SvPOK(buffer)) {
2962 *lp = SvCUR(buffer);
2964 return SvPVX(buffer);
2975 if (flags & SV_UNDEF_RETURNS_NULL)
2977 if (!PL_localizing && !(SvFLAGS(sv) & SVs_PADTMP) && ckWARN(WARN_UNINITIALIZED))
2979 if (SvTYPE(sv) < SVt_PV)
2980 /* Typically the caller expects that sv_any is not NULL now. */
2981 sv_upgrade(sv, SVt_PV);
2985 const STRLEN len = s - SvPVX_const(sv);
2991 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
2992 PTR2UV(sv),SvPVX_const(sv)));
2993 if (flags & SV_CONST_RETURN)
2994 return (char *)SvPVX_const(sv);
2995 if (flags & SV_MUTABLE_RETURN)
2996 return SvPVX_mutable(sv);
3001 =for apidoc sv_copypv
3003 Copies a stringified representation of the source SV into the
3004 destination SV. Automatically performs any necessary mg_get and
3005 coercion of numeric values into strings. Guaranteed to preserve
3006 UTF8 flag even from overloaded objects. Similar in nature to
3007 sv_2pv[_flags] but operates directly on an SV instead of just the
3008 string. Mostly uses sv_2pv_flags to do its work, except when that
3009 would lose the UTF-8'ness of the PV.
3015 Perl_sv_copypv(pTHX_ SV *const dsv, register SV *const ssv)
3018 const char * const s = SvPV_const(ssv,len);
3020 PERL_ARGS_ASSERT_SV_COPYPV;
3022 sv_setpvn(dsv,s,len);
3030 =for apidoc sv_2pvbyte
3032 Return a pointer to the byte-encoded representation of the SV, and set *lp
3033 to its length. May cause the SV to be downgraded from UTF-8 as a
3036 Usually accessed via the C<SvPVbyte> macro.
3042 Perl_sv_2pvbyte(pTHX_ register SV *const sv, STRLEN *const lp)
3044 PERL_ARGS_ASSERT_SV_2PVBYTE;
3047 sv_utf8_downgrade(sv,0);
3048 return lp ? SvPV_nomg(sv,*lp) : SvPV_nomg_nolen(sv);
3052 =for apidoc sv_2pvutf8
3054 Return a pointer to the UTF-8-encoded representation of the SV, and set *lp
3055 to its length. May cause the SV to be upgraded to UTF-8 as a side-effect.
3057 Usually accessed via the C<SvPVutf8> macro.
3063 Perl_sv_2pvutf8(pTHX_ register SV *const sv, STRLEN *const lp)
3065 PERL_ARGS_ASSERT_SV_2PVUTF8;
3067 sv_utf8_upgrade(sv);
3068 return lp ? SvPV(sv,*lp) : SvPV_nolen(sv);
3073 =for apidoc sv_2bool
3075 This macro is only used by sv_true() or its macro equivalent, and only if
3076 the latter's argument is neither SvPOK, SvIOK nor SvNOK.
3077 It calls sv_2bool_flags with the SV_GMAGIC flag.
3079 =for apidoc sv_2bool_flags
3081 This function is only used by sv_true() and friends, and only if
3082 the latter's argument is neither SvPOK, SvIOK nor SvNOK. If the flags
3083 contain SV_GMAGIC, then it does an mg_get() first.
3090 Perl_sv_2bool_flags(pTHX_ register SV *const sv, const I32 flags)
3094 PERL_ARGS_ASSERT_SV_2BOOL_FLAGS;
3096 if(flags & SV_GMAGIC) SvGETMAGIC(sv);
3102 SV * const tmpsv = AMG_CALLunary(sv, bool__amg);
3103 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
3104 return cBOOL(SvTRUE(tmpsv));
3106 return SvRV(sv) != 0;
3109 register XPV* const Xpvtmp = (XPV*)SvANY(sv);
3111 (*sv->sv_u.svu_pv > '0' ||
3112 Xpvtmp->xpv_cur > 1 ||
3113 (Xpvtmp->xpv_cur && *sv->sv_u.svu_pv != '0')))
3120 return SvIVX(sv) != 0;
3123 return SvNVX(sv) != 0.0;
3125 if (isGV_with_GP(sv))
3135 =for apidoc sv_utf8_upgrade
3137 Converts the PV of an SV to its UTF-8-encoded form.
3138 Forces the SV to string form if it is not already.
3139 Will C<mg_get> on C<sv> if appropriate.
3140 Always sets the SvUTF8 flag to avoid future validity checks even
3141 if the whole string is the same in UTF-8 as not.
3142 Returns the number of bytes in the converted string
3144 This is not as a general purpose byte encoding to Unicode interface:
3145 use the Encode extension for that.
3147 =for apidoc sv_utf8_upgrade_nomg
3149 Like sv_utf8_upgrade, but doesn't do magic on C<sv>
3151 =for apidoc sv_utf8_upgrade_flags
3153 Converts the PV of an SV to its UTF-8-encoded form.
3154 Forces the SV to string form if it is not already.
3155 Always sets the SvUTF8 flag to avoid future validity checks even
3156 if all the bytes are invariant in UTF-8. If C<flags> has C<SV_GMAGIC> bit set,
3157 will C<mg_get> on C<sv> if appropriate, else not.
3158 Returns the number of bytes in the converted string
3159 C<sv_utf8_upgrade> and
3160 C<sv_utf8_upgrade_nomg> are implemented in terms of this function.
3162 This is not as a general purpose byte encoding to Unicode interface:
3163 use the Encode extension for that.
3167 The grow version is currently not externally documented. It adds a parameter,
3168 extra, which is the number of unused bytes the string of 'sv' is guaranteed to
3169 have free after it upon return. This allows the caller to reserve extra space
3170 that it intends to fill, to avoid extra grows.
3172 Also externally undocumented for the moment is the flag SV_FORCE_UTF8_UPGRADE,
3173 which can be used to tell this function to not first check to see if there are
3174 any characters that are different in UTF-8 (variant characters) which would
3175 force it to allocate a new string to sv, but to assume there are. Typically
3176 this flag is used by a routine that has already parsed the string to find that
3177 there are such characters, and passes this information on so that the work
3178 doesn't have to be repeated.
3180 (One might think that the calling routine could pass in the position of the
3181 first such variant, so it wouldn't have to be found again. But that is not the
3182 case, because typically when the caller is likely to use this flag, it won't be
3183 calling this routine unless it finds something that won't fit into a byte.
3184 Otherwise it tries to not upgrade and just use bytes. But some things that
3185 do fit into a byte are variants in utf8, and the caller may not have been
3186 keeping track of these.)
3188 If the routine itself changes the string, it adds a trailing NUL. Such a NUL
3189 isn't guaranteed due to having other routines do the work in some input cases,
3190 or if the input is already flagged as being in utf8.
3192 The speed of this could perhaps be improved for many cases if someone wanted to
3193 write a fast function that counts the number of variant characters in a string,
3194 especially if it could return the position of the first one.
3199 Perl_sv_utf8_upgrade_flags_grow(pTHX_ register SV *const sv, const I32 flags, STRLEN extra)
3203 PERL_ARGS_ASSERT_SV_UTF8_UPGRADE_FLAGS_GROW;
3205 if (sv == &PL_sv_undef)
3209 if (SvREADONLY(sv) && (SvPOKp(sv) || SvIOKp(sv) || SvNOKp(sv))) {
3210 (void) sv_2pv_flags(sv,&len, flags);
3212 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3216 (void) SvPV_force_flags(sv,len,flags & SV_GMAGIC);
3221 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3226 sv_force_normal_flags(sv, 0);
3229 if (PL_encoding && !(flags & SV_UTF8_NO_ENCODING)) {
3230 sv_recode_to_utf8(sv, PL_encoding);
3231 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3235 if (SvCUR(sv) == 0) {
3236 if (extra) SvGROW(sv, extra);
3237 } else { /* Assume Latin-1/EBCDIC */
3238 /* This function could be much more efficient if we
3239 * had a FLAG in SVs to signal if there are any variant
3240 * chars in the PV. Given that there isn't such a flag
3241 * make the loop as fast as possible (although there are certainly ways
3242 * to speed this up, eg. through vectorization) */
3243 U8 * s = (U8 *) SvPVX_const(sv);
3244 U8 * e = (U8 *) SvEND(sv);
3246 STRLEN two_byte_count = 0;
3248 if (flags & SV_FORCE_UTF8_UPGRADE) goto must_be_utf8;
3250 /* See if really will need to convert to utf8. We mustn't rely on our
3251 * incoming SV being well formed and having a trailing '\0', as certain
3252 * code in pp_formline can send us partially built SVs. */
3256 if (NATIVE_IS_INVARIANT(ch)) continue;
3258 t--; /* t already incremented; re-point to first variant */
3263 /* utf8 conversion not needed because all are invariants. Mark as
3264 * UTF-8 even if no variant - saves scanning loop */
3270 /* Here, the string should be converted to utf8, either because of an
3271 * input flag (two_byte_count = 0), or because a character that
3272 * requires 2 bytes was found (two_byte_count = 1). t points either to
3273 * the beginning of the string (if we didn't examine anything), or to
3274 * the first variant. In either case, everything from s to t - 1 will
3275 * occupy only 1 byte each on output.
3277 * There are two main ways to convert. One is to create a new string
3278 * and go through the input starting from the beginning, appending each
3279 * converted value onto the new string as we go along. It's probably
3280 * best to allocate enough space in the string for the worst possible
3281 * case rather than possibly running out of space and having to
3282 * reallocate and then copy what we've done so far. Since everything
3283 * from s to t - 1 is invariant, the destination can be initialized
3284 * with these using a fast memory copy
3286 * The other way is to figure out exactly how big the string should be
3287 * by parsing the entire input. Then you don't have to make it big
3288 * enough to handle the worst possible case, and more importantly, if
3289 * the string you already have is large enough, you don't have to
3290 * allocate a new string, you can copy the last character in the input
3291 * string to the final position(s) that will be occupied by the
3292 * converted string and go backwards, stopping at t, since everything
3293 * before that is invariant.
3295 * There are advantages and disadvantages to each method.
3297 * In the first method, we can allocate a new string, do the memory
3298 * copy from the s to t - 1, and then proceed through the rest of the
3299 * string byte-by-byte.
3301 * In the second method, we proceed through the rest of the input
3302 * string just calculating how big the converted string will be. Then
3303 * there are two cases:
3304 * 1) if the string has enough extra space to handle the converted
3305 * value. We go backwards through the string, converting until we
3306 * get to the position we are at now, and then stop. If this
3307 * position is far enough along in the string, this method is
3308 * faster than the other method. If the memory copy were the same
3309 * speed as the byte-by-byte loop, that position would be about
3310 * half-way, as at the half-way mark, parsing to the end and back
3311 * is one complete string's parse, the same amount as starting
3312 * over and going all the way through. Actually, it would be
3313 * somewhat less than half-way, as it's faster to just count bytes
3314 * than to also copy, and we don't have the overhead of allocating
3315 * a new string, changing the scalar to use it, and freeing the
3316 * existing one. But if the memory copy is fast, the break-even
3317 * point is somewhere after half way. The counting loop could be
3318 * sped up by vectorization, etc, to move the break-even point
3319 * further towards the beginning.
3320 * 2) if the string doesn't have enough space to handle the converted
3321 * value. A new string will have to be allocated, and one might
3322 * as well, given that, start from the beginning doing the first
3323 * method. We've spent extra time parsing the string and in
3324 * exchange all we've gotten is that we know precisely how big to
3325 * make the new one. Perl is more optimized for time than space,
3326 * so this case is a loser.
3327 * So what I've decided to do is not use the 2nd method unless it is
3328 * guaranteed that a new string won't have to be allocated, assuming
3329 * the worst case. I also decided not to put any more conditions on it
3330 * than this, for now. It seems likely that, since the worst case is
3331 * twice as big as the unknown portion of the string (plus 1), we won't
3332 * be guaranteed enough space, causing us to go to the first method,
3333 * unless the string is short, or the first variant character is near
3334 * the end of it. In either of these cases, it seems best to use the
3335 * 2nd method. The only circumstance I can think of where this would
3336 * be really slower is if the string had once had much more data in it
3337 * than it does now, but there is still a substantial amount in it */
3340 STRLEN invariant_head = t - s;
3341 STRLEN size = invariant_head + (e - t) * 2 + 1 + extra;
3342 if (SvLEN(sv) < size) {
3344 /* Here, have decided to allocate a new string */
3349 Newx(dst, size, U8);
3351 /* If no known invariants at the beginning of the input string,
3352 * set so starts from there. Otherwise, can use memory copy to
3353 * get up to where we are now, and then start from here */
3355 if (invariant_head <= 0) {
3358 Copy(s, dst, invariant_head, char);
3359 d = dst + invariant_head;
3363 const UV uv = NATIVE8_TO_UNI(*t++);
3364 if (UNI_IS_INVARIANT(uv))
3365 *d++ = (U8)UNI_TO_NATIVE(uv);
3367 *d++ = (U8)UTF8_EIGHT_BIT_HI(uv);
3368 *d++ = (U8)UTF8_EIGHT_BIT_LO(uv);
3372 SvPV_free(sv); /* No longer using pre-existing string */
3373 SvPV_set(sv, (char*)dst);
3374 SvCUR_set(sv, d - dst);
3375 SvLEN_set(sv, size);
3378 /* Here, have decided to get the exact size of the string.
3379 * Currently this happens only when we know that there is
3380 * guaranteed enough space to fit the converted string, so
3381 * don't have to worry about growing. If two_byte_count is 0,
3382 * then t points to the first byte of the string which hasn't
3383 * been examined yet. Otherwise two_byte_count is 1, and t
3384 * points to the first byte in the string that will expand to
3385 * two. Depending on this, start examining at t or 1 after t.
3388 U8 *d = t + two_byte_count;
3391 /* Count up the remaining bytes that expand to two */
3394 const U8 chr = *d++;
3395 if (! NATIVE_IS_INVARIANT(chr)) two_byte_count++;
3398 /* The string will expand by just the number of bytes that
3399 * occupy two positions. But we are one afterwards because of
3400 * the increment just above. This is the place to put the
3401 * trailing NUL, and to set the length before we decrement */
3403 d += two_byte_count;
3404 SvCUR_set(sv, d - s);
3408 /* Having decremented d, it points to the position to put the
3409 * very last byte of the expanded string. Go backwards through
3410 * the string, copying and expanding as we go, stopping when we
3411 * get to the part that is invariant the rest of the way down */
3415 const U8 ch = NATIVE8_TO_UNI(*e--);
3416 if (UNI_IS_INVARIANT(ch)) {
3417 *d-- = UNI_TO_NATIVE(ch);
3419 *d-- = (U8)UTF8_EIGHT_BIT_LO(ch);
3420 *d-- = (U8)UTF8_EIGHT_BIT_HI(ch);
3425 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
3426 /* Update pos. We do it at the end rather than during
3427 * the upgrade, to avoid slowing down the common case
3428 * (upgrade without pos) */
3429 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
3431 I32 pos = mg->mg_len;
3432 if (pos > 0 && (U32)pos > invariant_head) {
3433 U8 *d = (U8*) SvPVX(sv) + invariant_head;
3434 STRLEN n = (U32)pos - invariant_head;
3436 if (UTF8_IS_START(*d))
3441 mg->mg_len = d - (U8*)SvPVX(sv);
3444 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
3445 magic_setutf8(sv,mg); /* clear UTF8 cache */
3450 /* Mark as UTF-8 even if no variant - saves scanning loop */
3456 =for apidoc sv_utf8_downgrade
3458 Attempts to convert the PV of an SV from characters to bytes.
3459 If the PV contains a character that cannot fit
3460 in a byte, this conversion will fail;
3461 in this case, either returns false or, if C<fail_ok> is not
3464 This is not as a general purpose Unicode to byte encoding interface:
3465 use the Encode extension for that.
3471 Perl_sv_utf8_downgrade(pTHX_ register SV *const sv, const bool fail_ok)
3475 PERL_ARGS_ASSERT_SV_UTF8_DOWNGRADE;
3477 if (SvPOKp(sv) && SvUTF8(sv)) {
3481 int mg_flags = SV_GMAGIC;
3484 sv_force_normal_flags(sv, 0);
3486 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
3488 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
3490 I32 pos = mg->mg_len;
3492 sv_pos_b2u(sv, &pos);
3493 mg_flags = 0; /* sv_pos_b2u does get magic */
3497 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
3498 magic_setutf8(sv,mg); /* clear UTF8 cache */
3501 s = (U8 *) SvPV_flags(sv, len, mg_flags);
3503 if (!utf8_to_bytes(s, &len)) {
3508 Perl_croak(aTHX_ "Wide character in %s",
3511 Perl_croak(aTHX_ "Wide character");
3522 =for apidoc sv_utf8_encode
3524 Converts the PV of an SV to UTF-8, but then turns the C<SvUTF8>
3525 flag off so that it looks like octets again.
3531 Perl_sv_utf8_encode(pTHX_ register SV *const sv)
3533 PERL_ARGS_ASSERT_SV_UTF8_ENCODE;
3536 sv_force_normal_flags(sv, 0);
3538 if (SvREADONLY(sv)) {
3539 Perl_croak_no_modify(aTHX);
3541 (void) sv_utf8_upgrade(sv);
3546 =for apidoc sv_utf8_decode
3548 If the PV of the SV is an octet sequence in UTF-8
3549 and contains a multiple-byte character, the C<SvUTF8> flag is turned on
3550 so that it looks like a character. If the PV contains only single-byte
3551 characters, the C<SvUTF8> flag stays off.
3552 Scans PV for validity and returns false if the PV is invalid UTF-8.
3558 Perl_sv_utf8_decode(pTHX_ register SV *const sv)
3560 PERL_ARGS_ASSERT_SV_UTF8_DECODE;
3563 const U8 *start, *c;
3566 /* The octets may have got themselves encoded - get them back as
3569 if (!sv_utf8_downgrade(sv, TRUE))
3572 /* it is actually just a matter of turning the utf8 flag on, but
3573 * we want to make sure everything inside is valid utf8 first.
3575 c = start = (const U8 *) SvPVX_const(sv);
3576 if (!is_utf8_string(c, SvCUR(sv)+1))
3578 e = (const U8 *) SvEND(sv);
3581 if (!UTF8_IS_INVARIANT(ch)) {
3586 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
3587 /* adjust pos to the start of a UTF8 char sequence */
3588 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
3590 I32 pos = mg->mg_len;
3592 for (c = start + pos; c > start; c--) {
3593 if (UTF8_IS_START(*c))
3596 mg->mg_len = c - start;
3599 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
3600 magic_setutf8(sv,mg); /* clear UTF8 cache */
3607 =for apidoc sv_setsv
3609 Copies the contents of the source SV C<ssv> into the destination SV
3610 C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3611 function if the source SV needs to be reused. Does not handle 'set' magic.
3612 Loosely speaking, it performs a copy-by-value, obliterating any previous
3613 content of the destination.
3615 You probably want to use one of the assortment of wrappers, such as
3616 C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3617 C<SvSetMagicSV_nosteal>.
3619 =for apidoc sv_setsv_flags
3621 Copies the contents of the source SV C<ssv> into the destination SV
3622 C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3623 function if the source SV needs to be reused. Does not handle 'set' magic.
3624 Loosely speaking, it performs a copy-by-value, obliterating any previous
3625 content of the destination.
3626 If the C<flags> parameter has the C<SV_GMAGIC> bit set, will C<mg_get> on
3627 C<ssv> if appropriate, else not. If the C<flags> parameter has the
3628 C<NOSTEAL> bit set then the buffers of temps will not be stolen. <sv_setsv>
3629 and C<sv_setsv_nomg> are implemented in terms of this function.
3631 You probably want to use one of the assortment of wrappers, such as
3632 C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3633 C<SvSetMagicSV_nosteal>.
3635 This is the primary function for copying scalars, and most other
3636 copy-ish functions and macros use this underneath.
3642 S_glob_assign_glob(pTHX_ SV *const dstr, SV *const sstr, const int dtype)
3644 I32 mro_changes = 0; /* 1 = method, 2 = isa, 3 = recursive isa */
3645 HV *old_stash = NULL;
3647 PERL_ARGS_ASSERT_GLOB_ASSIGN_GLOB;
3649 if (dtype != SVt_PVGV && !isGV_with_GP(dstr)) {
3650 const char * const name = GvNAME(sstr);
3651 const STRLEN len = GvNAMELEN(sstr);
3653 if (dtype >= SVt_PV) {
3659 SvUPGRADE(dstr, SVt_PVGV);
3660 (void)SvOK_off(dstr);
3661 /* FIXME - why are we doing this, then turning it off and on again
3663 isGV_with_GP_on(dstr);
3665 GvSTASH(dstr) = GvSTASH(sstr);
3667 Perl_sv_add_backref(aTHX_ MUTABLE_SV(GvSTASH(dstr)), dstr);
3668 gv_name_set(MUTABLE_GV(dstr), name, len, GV_ADD);
3669 SvFAKE_on(dstr); /* can coerce to non-glob */
3672 if(GvGP(MUTABLE_GV(sstr))) {
3673 /* If source has method cache entry, clear it */
3675 SvREFCNT_dec(GvCV(sstr));
3676 GvCV_set(sstr, NULL);
3679 /* If source has a real method, then a method is
3682 GvCV((const GV *)sstr) && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3688 /* If dest already had a real method, that's a change as well */
3690 !mro_changes && GvGP(MUTABLE_GV(dstr)) && GvCVu((const GV *)dstr)
3691 && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3696 /* We don’t need to check the name of the destination if it was not a
3697 glob to begin with. */
3698 if(dtype == SVt_PVGV) {
3699 const char * const name = GvNAME((const GV *)dstr);
3702 /* The stash may have been detached from the symbol table, so
3704 && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3705 && GvAV((const GV *)sstr)
3709 const STRLEN len = GvNAMELEN(dstr);
3710 if ((len > 1 && name[len-2] == ':' && name[len-1] == ':')
3711 || (len == 1 && name[0] == ':')) {
3714 /* Set aside the old stash, so we can reset isa caches on
3716 if((old_stash = GvHV(dstr)))
3717 /* Make sure we do not lose it early. */
3718 SvREFCNT_inc_simple_void_NN(
3719 sv_2mortal((SV *)old_stash)
3725 gp_free(MUTABLE_GV(dstr));
3726 isGV_with_GP_off(dstr);
3727 (void)SvOK_off(dstr);
3728 isGV_with_GP_on(dstr);
3729 GvINTRO_off(dstr); /* one-shot flag */
3730 GvGP_set(dstr, gp_ref(GvGP(sstr)));
3731 if (SvTAINTED(sstr))
3733 if (GvIMPORTED(dstr) != GVf_IMPORTED
3734 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3736 GvIMPORTED_on(dstr);
3739 if(mro_changes == 2) {
3741 SV * const sref = (SV *)GvAV((const GV *)dstr);
3742 if (SvSMAGICAL(sref) && (mg = mg_find(sref, PERL_MAGIC_isa))) {
3743 if (SvTYPE(mg->mg_obj) != SVt_PVAV) {
3744 AV * const ary = newAV();
3745 av_push(ary, mg->mg_obj); /* takes the refcount */
3746 mg->mg_obj = (SV *)ary;
3748 av_push((AV *)mg->mg_obj, SvREFCNT_inc_simple_NN(dstr));
3750 else sv_magic(sref, dstr, PERL_MAGIC_isa, NULL, 0);
3751 mro_isa_changed_in(GvSTASH(dstr));
3753 else if(mro_changes == 3) {
3754 HV * const stash = GvHV(dstr);
3755 if(old_stash ? (HV *)HvENAME_get(old_stash) : stash)
3761 else if(mro_changes) mro_method_changed_in(GvSTASH(dstr));
3766 S_glob_assign_ref(pTHX_ SV *const dstr, SV *const sstr)
3768 SV * const sref = SvREFCNT_inc(SvRV(sstr));
3770 const int intro = GvINTRO(dstr);
3773 const U32 stype = SvTYPE(sref);
3775 PERL_ARGS_ASSERT_GLOB_ASSIGN_REF;
3778 GvINTRO_off(dstr); /* one-shot flag */
3779 GvLINE(dstr) = CopLINE(PL_curcop);
3780 GvEGV(dstr) = MUTABLE_GV(dstr);
3785 location = (SV **) &(GvGP(dstr)->gp_cv); /* XXX bypassing GvCV_set */
3786 import_flag = GVf_IMPORTED_CV;
3789 location = (SV **) &GvHV(dstr);
3790 import_flag = GVf_IMPORTED_HV;
3793 location = (SV **) &GvAV(dstr);
3794 import_flag = GVf_IMPORTED_AV;
3797 location = (SV **) &GvIOp(dstr);
3800 location = (SV **) &GvFORM(dstr);
3803 location = &GvSV(dstr);
3804 import_flag = GVf_IMPORTED_SV;
3807 if (stype == SVt_PVCV) {
3808 /*if (GvCVGEN(dstr) && (GvCV(dstr) != (const CV *)sref || GvCVGEN(dstr))) {*/
3809 if (GvCVGEN(dstr)) {
3810 SvREFCNT_dec(GvCV(dstr));
3811 GvCV_set(dstr, NULL);
3812 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3815 SAVEGENERICSV(*location);
3819 if (stype == SVt_PVCV && (*location != sref || GvCVGEN(dstr))) {
3820 CV* const cv = MUTABLE_CV(*location);
3822 if (!GvCVGEN((const GV *)dstr) &&
3823 (CvROOT(cv) || CvXSUB(cv)))
3825 /* Redefining a sub - warning is mandatory if
3826 it was a const and its value changed. */
3827 if (CvCONST(cv) && CvCONST((const CV *)sref)
3829 == cv_const_sv((const CV *)sref)) {
3831 /* They are 2 constant subroutines generated from
3832 the same constant. This probably means that
3833 they are really the "same" proxy subroutine
3834 instantiated in 2 places. Most likely this is
3835 when a constant is exported twice. Don't warn.
3838 else if (ckWARN(WARN_REDEFINE)
3840 && (!CvCONST((const CV *)sref)
3841 || sv_cmp(cv_const_sv(cv),
3842 cv_const_sv((const CV *)
3844 Perl_warner(aTHX_ packWARN(WARN_REDEFINE),
3847 ? "Constant subroutine %s::%s redefined"
3848 : "Subroutine %s::%s redefined"),
3849 HvNAME_get(GvSTASH((const GV *)dstr)),
3850 GvENAME(MUTABLE_GV(dstr)));
3854 cv_ckproto_len(cv, (const GV *)dstr,
3855 SvPOK(sref) ? SvPVX_const(sref) : NULL,
3856 SvPOK(sref) ? SvCUR(sref) : 0);
3858 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3859 GvASSUMECV_on(dstr);
3860 if(GvSTASH(dstr)) mro_method_changed_in(GvSTASH(dstr)); /* sub foo { 1 } sub bar { 2 } *bar = \&foo */
3863 if (import_flag && !(GvFLAGS(dstr) & import_flag)
3864 && CopSTASH_ne(PL_curcop, GvSTASH(dstr))) {
3865 GvFLAGS(dstr) |= import_flag;
3867 if (stype == SVt_PVHV) {
3868 const char * const name = GvNAME((GV*)dstr);
3869 const STRLEN len = GvNAMELEN(dstr);
3872 (len > 1 && name[len-2] == ':' && name[len-1] == ':')
3873 || (len == 1 && name[0] == ':')
3875 && (!dref || HvENAME_get(dref))
3878 (HV *)sref, (HV *)dref,
3884 stype == SVt_PVAV && sref != dref
3885 && strEQ(GvNAME((GV*)dstr), "ISA")
3886 /* The stash may have been detached from the symbol table, so
3887 check its name before doing anything. */
3888 && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3891 MAGIC * const omg = dref && SvSMAGICAL(dref)
3892 ? mg_find(dref, PERL_MAGIC_isa)
3894 if (SvSMAGICAL(sref) && (mg = mg_find(sref, PERL_MAGIC_isa))) {
3895 if (SvTYPE(mg->mg_obj) != SVt_PVAV) {
3896 AV * const ary = newAV();
3897 av_push(ary, mg->mg_obj); /* takes the refcount */
3898 mg->mg_obj = (SV *)ary;
3901 if (SvTYPE(omg->mg_obj) == SVt_PVAV) {
3902 SV **svp = AvARRAY((AV *)omg->mg_obj);
3903 I32 items = AvFILLp((AV *)omg->mg_obj) + 1;
3907 SvREFCNT_inc_simple_NN(*svp++)
3913 SvREFCNT_inc_simple_NN(omg->mg_obj)
3917 av_push((AV *)mg->mg_obj,SvREFCNT_inc_simple_NN(dstr));
3922 sref, omg ? omg->mg_obj : dstr, PERL_MAGIC_isa, NULL, 0
3924 mg = mg_find(sref, PERL_MAGIC_isa);
3926 /* Since the *ISA assignment could have affected more than
3927 one stash, don’t call mro_isa_changed_in directly, but let
3928 magic_clearisa do it for us, as it already has the logic for
3929 dealing with globs vs arrays of globs. */
3931 Perl_magic_clearisa(aTHX_ NULL, mg);
3936 if (SvTAINTED(sstr))
3942 Perl_sv_setsv_flags(pTHX_ SV *dstr, register SV* sstr, const I32 flags)
3945 register U32 sflags;
3947 register svtype stype;
3949 PERL_ARGS_ASSERT_SV_SETSV_FLAGS;
3954 if (SvIS_FREED(dstr)) {
3955 Perl_croak(aTHX_ "panic: attempt to copy value %" SVf
3956 " to a freed scalar %p", SVfARG(sstr), (void *)dstr);
3958 SV_CHECK_THINKFIRST_COW_DROP(dstr);
3960 sstr = &PL_sv_undef;
3961 if (SvIS_FREED(sstr)) {
3962 Perl_croak(aTHX_ "panic: attempt to copy freed scalar %p to %p",
3963 (void*)sstr, (void*)dstr);
3965 stype = SvTYPE(sstr);
3966 dtype = SvTYPE(dstr);
3968 (void)SvAMAGIC_off(dstr);
3971 /* need to nuke the magic */
3975 /* There's a lot of redundancy below but we're going for speed here */
3980 if (dtype != SVt_PVGV && dtype != SVt_PVLV) {
3981 (void)SvOK_off(dstr);
3989 sv_upgrade(dstr, SVt_IV);
3993 sv_upgrade(dstr, SVt_PVIV);
3997 goto end_of_first_switch;
3999 (void)SvIOK_only(dstr);
4000 SvIV_set(dstr, SvIVX(sstr));
4003 /* SvTAINTED can only be true if the SV has taint magic, which in
4004 turn means that the SV type is PVMG (or greater). This is the
4005 case statement for SVt_IV, so this cannot be true (whatever gcov
4007 assert(!SvTAINTED(sstr));
4012 if (dtype < SVt_PV && dtype != SVt_IV)
4013 sv_upgrade(dstr, SVt_IV);
4021 sv_upgrade(dstr, SVt_NV);
4025 sv_upgrade(dstr, SVt_PVNV);
4029 goto end_of_first_switch;
4031 SvNV_set(dstr, SvNVX(sstr));
4032 (void)SvNOK_only(dstr);
4033 /* SvTAINTED can only be true if the SV has taint magic, which in
4034 turn means that the SV type is PVMG (or greater). This is the
4035 case statement for SVt_NV, so this cannot be true (whatever gcov
4037 assert(!SvTAINTED(sstr));
4043 #ifdef PERL_OLD_COPY_ON_WRITE
4044 if ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS) {
4045 if (dtype < SVt_PVIV)
4046 sv_upgrade(dstr, SVt_PVIV);
4053 sv_upgrade(dstr, SVt_PV);
4056 if (dtype < SVt_PVIV)
4057 sv_upgrade(dstr, SVt_PVIV);
4060 if (dtype < SVt_PVNV)
4061 sv_upgrade(dstr, SVt_PVNV);
4065 const char * const type = sv_reftype(sstr,0);
4067 Perl_croak(aTHX_ "Bizarre copy of %s in %s", type, OP_DESC(PL_op));
4069 Perl_croak(aTHX_ "Bizarre copy of %s", type);
4074 if (dtype < SVt_REGEXP)
4075 sv_upgrade(dstr, SVt_REGEXP);
4078 /* case SVt_BIND: */
4082 if (SvGMAGICAL(sstr) && (flags & SV_GMAGIC)) {
4084 if (SvTYPE(sstr) != stype)
4085 stype = SvTYPE(sstr);
4087 if (isGV_with_GP(sstr) && dtype <= SVt_PVLV) {
4088 glob_assign_glob(dstr, sstr, dtype);
4091 if (stype == SVt_PVLV)
4092 SvUPGRADE(dstr, SVt_PVNV);
4094 SvUPGRADE(dstr, (svtype)stype);
4096 end_of_first_switch:
4098 /* dstr may have been upgraded. */
4099 dtype = SvTYPE(dstr);
4100 sflags = SvFLAGS(sstr);
4102 if (dtype == SVt_PVCV || dtype == SVt_PVFM) {
4103 /* Assigning to a subroutine sets the prototype. */
4106 const char *const ptr = SvPV_const(sstr, len);
4108 SvGROW(dstr, len + 1);
4109 Copy(ptr, SvPVX(dstr), len + 1, char);
4110 SvCUR_set(dstr, len);
4112 SvFLAGS(dstr) |= sflags & SVf_UTF8;
4116 } else if (dtype == SVt_PVAV || dtype == SVt_PVHV) {
4117 const char * const type = sv_reftype(dstr,0);
4119 Perl_croak(aTHX_ "Cannot copy to %s in %s", type, OP_DESC(PL_op));
4121 Perl_croak(aTHX_ "Cannot copy to %s", type);
4122 } else if (sflags & SVf_ROK) {
4123 if (isGV_with_GP(dstr)
4124 && SvTYPE(SvRV(sstr)) == SVt_PVGV && isGV_with_GP(SvRV(sstr))) {
4127 if (GvIMPORTED(dstr) != GVf_IMPORTED
4128 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
4130 GvIMPORTED_on(dstr);
4135 glob_assign_glob(dstr, sstr, dtype);
4139 if (dtype >= SVt_PV) {
4140 if (isGV_with_GP(dstr)) {
4141 glob_assign_ref(dstr, sstr);
4144 if (SvPVX_const(dstr)) {
4150 (void)SvOK_off(dstr);
4151 SvRV_set(dstr, SvREFCNT_inc(SvRV(sstr)));
4152 SvFLAGS(dstr) |= sflags & SVf_ROK;
4153 assert(!(sflags & SVp_NOK));
4154 assert(!(sflags & SVp_IOK));
4155 assert(!(sflags & SVf_NOK));
4156 assert(!(sflags & SVf_IOK));
4158 else if (isGV_with_GP(dstr)) {
4159 if (!(sflags & SVf_OK)) {
4160 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
4161 "Undefined value assigned to typeglob");
4164 GV *gv = gv_fetchsv(sstr, GV_ADD, SVt_PVGV);
4165 if (dstr != (const SV *)gv) {
4166 const char * const name = GvNAME((const GV *)dstr);
4167 const STRLEN len = GvNAMELEN(dstr);
4168 HV *old_stash = NULL;
4169 bool reset_isa = FALSE;
4170 if ((len > 1 && name[len-2] == ':' && name[len-1] == ':')
4171 || (len == 1 && name[0] == ':')) {
4172 /* Set aside the old stash, so we can reset isa caches
4173 on its subclasses. */
4174 if((old_stash = GvHV(dstr))) {
4175 /* Make sure we do not lose it early. */
4176 SvREFCNT_inc_simple_void_NN(
4177 sv_2mortal((SV *)old_stash)
4184 gp_free(MUTABLE_GV(dstr));
4185 GvGP_set(dstr, gp_ref(GvGP(gv)));
4188 HV * const stash = GvHV(dstr);
4190 old_stash ? (HV *)HvENAME_get(old_stash) : stash
4200 else if (dtype == SVt_REGEXP && stype == SVt_REGEXP) {
4201 reg_temp_copy((REGEXP*)dstr, (REGEXP*)sstr);
4203 else if (sflags & SVp_POK) {
4207 * Check to see if we can just swipe the string. If so, it's a
4208 * possible small lose on short strings, but a big win on long ones.
4209 * It might even be a win on short strings if SvPVX_const(dstr)
4210 * has to be allocated and SvPVX_const(sstr) has to be freed.
4211 * Likewise if we can set up COW rather than doing an actual copy, we
4212 * drop to the else clause, as the swipe code and the COW setup code
4213 * have much in common.
4216 /* Whichever path we take through the next code, we want this true,
4217 and doing it now facilitates the COW check. */
4218 (void)SvPOK_only(dstr);
4221 /* If we're already COW then this clause is not true, and if COW
4222 is allowed then we drop down to the else and make dest COW
4223 with us. If caller hasn't said that we're allowed to COW
4224 shared hash keys then we don't do the COW setup, even if the
4225 source scalar is a shared hash key scalar. */
4226 (((flags & SV_COW_SHARED_HASH_KEYS)
4227 ? (sflags & (SVf_FAKE|SVf_READONLY)) != (SVf_FAKE|SVf_READONLY)
4228 : 1 /* If making a COW copy is forbidden then the behaviour we
4229 desire is as if the source SV isn't actually already
4230 COW, even if it is. So we act as if the source flags
4231 are not COW, rather than actually testing them. */
4233 #ifndef PERL_OLD_COPY_ON_WRITE
4234 /* The change that added SV_COW_SHARED_HASH_KEYS makes the logic
4235 when PERL_OLD_COPY_ON_WRITE is defined a little wrong.
4236 Conceptually PERL_OLD_COPY_ON_WRITE being defined should
4237 override SV_COW_SHARED_HASH_KEYS, because it means "always COW"
4238 but in turn, it's somewhat dead code, never expected to go
4239 live, but more kept as a placeholder on how to do it better
4240 in a newer implementation. */
4241 /* If we are COW and dstr is a suitable target then we drop down
4242 into the else and make dest a COW of us. */
4243 || (SvFLAGS(dstr) & CAN_COW_MASK) != CAN_COW_FLAGS
4248 (sflags & SVs_TEMP) && /* slated for free anyway? */
4249 !(sflags & SVf_OOK) && /* and not involved in OOK hack? */
4250 (!(flags & SV_NOSTEAL)) &&
4251 /* and we're allowed to steal temps */
4252 SvREFCNT(sstr) == 1 && /* and no other references to it? */
4253 SvLEN(sstr)) /* and really is a string */
4254 #ifdef PERL_OLD_COPY_ON_WRITE
4255 && ((flags & SV_COW_SHARED_HASH_KEYS)
4256 ? (!((sflags & CAN_COW_MASK) == CAN_COW_FLAGS
4257 && (SvFLAGS(dstr) & CAN_COW_MASK) == CAN_COW_FLAGS
4258 && SvTYPE(sstr) >= SVt_PVIV && SvTYPE(sstr) != SVt_PVFM))
4262 /* Failed the swipe test, and it's not a shared hash key either.
4263 Have to copy the string. */
4264 STRLEN len = SvCUR(sstr);
4265 SvGROW(dstr, len + 1); /* inlined from sv_setpvn */
4266 Move(SvPVX_const(sstr),SvPVX(dstr),len,char);
4267 SvCUR_set(dstr, len);
4268 *SvEND(dstr) = '\0';
4270 /* If PERL_OLD_COPY_ON_WRITE is not defined, then isSwipe will always
4272 /* Either it's a shared hash key, or it's suitable for
4273 copy-on-write or we can swipe the string. */
4275 PerlIO_printf(Perl_debug_log, "Copy on write: sstr --> dstr\n");
4279 #ifdef PERL_OLD_COPY_ON_WRITE
4281 if ((sflags & (SVf_FAKE | SVf_READONLY))
4282 != (SVf_FAKE | SVf_READONLY)) {
4283 SvREADONLY_on(sstr);
4285 /* Make the source SV into a loop of 1.
4286 (about to become 2) */
4287 SV_COW_NEXT_SV_SET(sstr, sstr);
4291 /* Initial code is common. */
4292 if (SvPVX_const(dstr)) { /* we know that dtype >= SVt_PV */
4297 /* making another shared SV. */
4298 STRLEN cur = SvCUR(sstr);
4299 STRLEN len = SvLEN(sstr);
4300 #ifdef PERL_OLD_COPY_ON_WRITE
4302 assert (SvTYPE(dstr) >= SVt_PVIV);
4303 /* SvIsCOW_normal */
4304 /* splice us in between source and next-after-source. */
4305 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4306 SV_COW_NEXT_SV_SET(sstr, dstr);
4307 SvPV_set(dstr, SvPVX_mutable(sstr));
4311 /* SvIsCOW_shared_hash */
4312 DEBUG_C(PerlIO_printf(Perl_debug_log,
4313 "Copy on write: Sharing hash\n"));
4315 assert (SvTYPE(dstr) >= SVt_PV);
4317 HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)))));
4319 SvLEN_set(dstr, len);
4320 SvCUR_set(dstr, cur);
4321 SvREADONLY_on(dstr);
4325 { /* Passes the swipe test. */
4326 SvPV_set(dstr, SvPVX_mutable(sstr));
4327 SvLEN_set(dstr, SvLEN(sstr));
4328 SvCUR_set(dstr, SvCUR(sstr));
4331 (void)SvOK_off(sstr); /* NOTE: nukes most SvFLAGS on sstr */
4332 SvPV_set(sstr, NULL);
4338 if (sflags & SVp_NOK) {
4339 SvNV_set(dstr, SvNVX(sstr));
4341 if (sflags & SVp_IOK) {
4342 SvIV_set(dstr, SvIVX(sstr));
4343 /* Must do this otherwise some other overloaded use of 0x80000000
4344 gets confused. I guess SVpbm_VALID */
4345 if (sflags & SVf_IVisUV)
4348 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_NOK|SVp_NOK|SVf_UTF8);
4350 const MAGIC * const smg = SvVSTRING_mg(sstr);
4352 sv_magic(dstr, NULL, PERL_MAGIC_vstring,
4353 smg->mg_ptr, smg->mg_len);
4354 SvRMAGICAL_on(dstr);
4358 else if (sflags & (SVp_IOK|SVp_NOK)) {
4359 (void)SvOK_off(dstr);
4360 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_IVisUV|SVf_NOK|SVp_NOK);
4361 if (sflags & SVp_IOK) {
4362 /* XXXX Do we want to set IsUV for IV(ROK)? Be extra safe... */
4363 SvIV_set(dstr, SvIVX(sstr));
4365 if (sflags & SVp_NOK) {
4366 SvNV_set(dstr, SvNVX(sstr));
4370 if (isGV_with_GP(sstr)) {
4371 /* This stringification rule for globs is spread in 3 places.
4372 This feels bad. FIXME. */
4373 const U32 wasfake = sflags & SVf_FAKE;
4375 /* FAKE globs can get coerced, so need to turn this off
4376 temporarily if it is on. */
4378 gv_efullname3(dstr, MUTABLE_GV(sstr), "*");
4379 SvFLAGS(sstr) |= wasfake;
4382 (void)SvOK_off(dstr);
4384 if (SvTAINTED(sstr))
4389 =for apidoc sv_setsv_mg
4391 Like C<sv_setsv>, but also handles 'set' magic.
4397 Perl_sv_setsv_mg(pTHX_ SV *const dstr, register SV *const sstr)
4399 PERL_ARGS_ASSERT_SV_SETSV_MG;
4401 sv_setsv(dstr,sstr);
4405 #ifdef PERL_OLD_COPY_ON_WRITE
4407 Perl_sv_setsv_cow(pTHX_ SV *dstr, SV *sstr)
4409 STRLEN cur = SvCUR(sstr);
4410 STRLEN len = SvLEN(sstr);
4411 register char *new_pv;
4413 PERL_ARGS_ASSERT_SV_SETSV_COW;
4416 PerlIO_printf(Perl_debug_log, "Fast copy on write: %p -> %p\n",
4417 (void*)sstr, (void*)dstr);
4424 if (SvTHINKFIRST(dstr))
4425 sv_force_normal_flags(dstr, SV_COW_DROP_PV);
4426 else if (SvPVX_const(dstr))
4427 Safefree(SvPVX_const(dstr));
4431 SvUPGRADE(dstr, SVt_PVIV);
4433 assert (SvPOK(sstr));
4434 assert (SvPOKp(sstr));
4435 assert (!SvIOK(sstr));
4436 assert (!SvIOKp(sstr));
4437 assert (!SvNOK(sstr));
4438 assert (!SvNOKp(sstr));
4440 if (SvIsCOW(sstr)) {
4442 if (SvLEN(sstr) == 0) {
4443 /* source is a COW shared hash key. */
4444 DEBUG_C(PerlIO_printf(Perl_debug_log,
4445 "Fast copy on write: Sharing hash\n"));
4446 new_pv = HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr))));
4449 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4451 assert ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS);
4452 SvUPGRADE(sstr, SVt_PVIV);
4453 SvREADONLY_on(sstr);
4455 DEBUG_C(PerlIO_printf(Perl_debug_log,
4456 "Fast copy on write: Converting sstr to COW\n"));
4457 SV_COW_NEXT_SV_SET(dstr, sstr);
4459 SV_COW_NEXT_SV_SET(sstr, dstr);
4460 new_pv = SvPVX_mutable(sstr);
4463 SvPV_set(dstr, new_pv);
4464 SvFLAGS(dstr) = (SVt_PVIV|SVf_POK|SVp_POK|SVf_FAKE|SVf_READONLY);
4467 SvLEN_set(dstr, len);
4468 SvCUR_set(dstr, cur);
4477 =for apidoc sv_setpvn
4479 Copies a string into an SV. The C<len> parameter indicates the number of
4480 bytes to be copied. If the C<ptr> argument is NULL the SV will become
4481 undefined. Does not handle 'set' magic. See C<sv_setpvn_mg>.
4487 Perl_sv_setpvn(pTHX_ register SV *const sv, register const char *const ptr, register const STRLEN len)
4490 register char *dptr;
4492 PERL_ARGS_ASSERT_SV_SETPVN;
4494 SV_CHECK_THINKFIRST_COW_DROP(sv);
4500 /* len is STRLEN which is unsigned, need to copy to signed */
4503 Perl_croak(aTHX_ "panic: sv_setpvn called with negative strlen");
4505 SvUPGRADE(sv, SVt_PV);
4507 dptr = SvGROW(sv, len + 1);
4508 Move(ptr,dptr,len,char);
4511 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4516 =for apidoc sv_setpvn_mg
4518 Like C<sv_setpvn>, but also handles 'set' magic.
4524 Perl_sv_setpvn_mg(pTHX_ register SV *const sv, register const char *const ptr, register const STRLEN len)
4526 PERL_ARGS_ASSERT_SV_SETPVN_MG;
4528 sv_setpvn(sv,ptr,len);
4533 =for apidoc sv_setpv
4535 Copies a string into an SV. The string must be null-terminated. Does not
4536 handle 'set' magic. See C<sv_setpv_mg>.
4542 Perl_sv_setpv(pTHX_ register SV *const sv, register const char *const ptr)
4545 register STRLEN len;
4547 PERL_ARGS_ASSERT_SV_SETPV;
4549 SV_CHECK_THINKFIRST_COW_DROP(sv);
4555 SvUPGRADE(sv, SVt_PV);
4557 SvGROW(sv, len + 1);
4558 Move(ptr,SvPVX(sv),len+1,char);
4560 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4565 =for apidoc sv_setpv_mg
4567 Like C<sv_setpv>, but also handles 'set' magic.
4573 Perl_sv_setpv_mg(pTHX_ register SV *const sv, register const char *const ptr)
4575 PERL_ARGS_ASSERT_SV_SETPV_MG;
4582 =for apidoc sv_usepvn_flags
4584 Tells an SV to use C<ptr> to find its string value. Normally the
4585 string is stored inside the SV but sv_usepvn allows the SV to use an
4586 outside string. The C<ptr> should point to memory that was allocated
4587 by C<malloc>. The string length, C<len>, must be supplied. By default
4588 this function will realloc (i.e. move) the memory pointed to by C<ptr>,
4589 so that pointer should not be freed or used by the programmer after
4590 giving it to sv_usepvn, and neither should any pointers from "behind"
4591 that pointer (e.g. ptr + 1) be used.
4593 If C<flags> & SV_SMAGIC is true, will call SvSETMAGIC. If C<flags> &
4594 SV_HAS_TRAILING_NUL is true, then C<ptr[len]> must be NUL, and the realloc
4595 will be skipped. (i.e. the buffer is actually at least 1 byte longer than
4596 C<len>, and already meets the requirements for storing in C<SvPVX>)
4602 Perl_sv_usepvn_flags(pTHX_ SV *const sv, char *ptr, const STRLEN len, const U32 flags)
4607 PERL_ARGS_ASSERT_SV_USEPVN_FLAGS;
4609 SV_CHECK_THINKFIRST_COW_DROP(sv);
4610 SvUPGRADE(sv, SVt_PV);
4613 if (flags & SV_SMAGIC)
4617 if (SvPVX_const(sv))
4621 if (flags & SV_HAS_TRAILING_NUL)
4622 assert(ptr[len] == '\0');
4625 allocate = (flags & SV_HAS_TRAILING_NUL)
4627 #ifdef Perl_safesysmalloc_size
4630 PERL_STRLEN_ROUNDUP(len + 1);
4632 if (flags & SV_HAS_TRAILING_NUL) {
4633 /* It's long enough - do nothing.
4634 Specifically Perl_newCONSTSUB is relying on this. */
4637 /* Force a move to shake out bugs in callers. */
4638 char *new_ptr = (char*)safemalloc(allocate);
4639 Copy(ptr, new_ptr, len, char);
4640 PoisonFree(ptr,len,char);
4644 ptr = (char*) saferealloc (ptr, allocate);
4647 #ifdef Perl_safesysmalloc_size
4648 SvLEN_set(sv, Perl_safesysmalloc_size(ptr));
4650 SvLEN_set(sv, allocate);
4654 if (!(flags & SV_HAS_TRAILING_NUL)) {
4657 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4659 if (flags & SV_SMAGIC)
4663 #ifdef PERL_OLD_COPY_ON_WRITE
4664 /* Need to do this *after* making the SV normal, as we need the buffer
4665 pointer to remain valid until after we've copied it. If we let go too early,
4666 another thread could invalidate it by unsharing last of the same hash key
4667 (which it can do by means other than releasing copy-on-write Svs)
4668 or by changing the other copy-on-write SVs in the loop. */
4670 S_sv_release_COW(pTHX_ register SV *sv, const char *pvx, SV *after)
4672 PERL_ARGS_ASSERT_SV_RELEASE_COW;
4674 { /* this SV was SvIsCOW_normal(sv) */
4675 /* we need to find the SV pointing to us. */
4676 SV *current = SV_COW_NEXT_SV(after);
4678 if (current == sv) {
4679 /* The SV we point to points back to us (there were only two of us
4681 Hence other SV is no longer copy on write either. */
4683 SvREADONLY_off(after);
4685 /* We need to follow the pointers around the loop. */
4687 while ((next = SV_COW_NEXT_SV(current)) != sv) {
4690 /* don't loop forever if the structure is bust, and we have
4691 a pointer into a closed loop. */
4692 assert (current != after);
4693 assert (SvPVX_const(current) == pvx);
4695 /* Make the SV before us point to the SV after us. */
4696 SV_COW_NEXT_SV_SET(current, after);
4702 =for apidoc sv_force_normal_flags
4704 Undo various types of fakery on an SV: if the PV is a shared string, make
4705 a private copy; if we're a ref, stop refing; if we're a glob, downgrade to
4706 an xpvmg; if we're a copy-on-write scalar, this is the on-write time when
4707 we do the copy, and is also used locally. If C<SV_COW_DROP_PV> is set
4708 then a copy-on-write scalar drops its PV buffer (if any) and becomes
4709 SvPOK_off rather than making a copy. (Used where this scalar is about to be
4710 set to some other value.) In addition, the C<flags> parameter gets passed to
4711 C<sv_unref_flags()> when unreffing. C<sv_force_normal> calls this function
4712 with flags set to 0.
4718 Perl_sv_force_normal_flags(pTHX_ register SV *const sv, const U32 flags)
4722 PERL_ARGS_ASSERT_SV_FORCE_NORMAL_FLAGS;
4724 #ifdef PERL_OLD_COPY_ON_WRITE
4725 if (SvREADONLY(sv)) {
4727 const char * const pvx = SvPVX_const(sv);
4728 const STRLEN len = SvLEN(sv);
4729 const STRLEN cur = SvCUR(sv);
4730 /* next COW sv in the loop. If len is 0 then this is a shared-hash
4731 key scalar, so we mustn't attempt to call SV_COW_NEXT_SV(), as
4732 we'll fail an assertion. */
4733 SV * const next = len ? SV_COW_NEXT_SV(sv) : 0;
4736 PerlIO_printf(Perl_debug_log,
4737 "Copy on write: Force normal %ld\n",
4743 /* This SV doesn't own the buffer, so need to Newx() a new one: */
4746 if (flags & SV_COW_DROP_PV) {
4747 /* OK, so we don't need to copy our buffer. */
4750 SvGROW(sv, cur + 1);
4751 Move(pvx,SvPVX(sv),cur,char);
4756 sv_release_COW(sv, pvx, next);
4758 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4764 else if (IN_PERL_RUNTIME)
4765 Perl_croak_no_modify(aTHX);
4768 if (SvREADONLY(sv)) {
4769 if (SvFAKE(sv) && !isGV_with_GP(sv)) {
4770 const char * const pvx = SvPVX_const(sv);
4771 const STRLEN len = SvCUR(sv);
4776 SvGROW(sv, len + 1);
4777 Move(pvx,SvPVX(sv),len,char);
4779 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4781 else if (IN_PERL_RUNTIME)
4782 Perl_croak_no_modify(aTHX);
4786 sv_unref_flags(sv, flags);
4787 else if (SvFAKE(sv) && isGV_with_GP(sv))
4789 else if (SvFAKE(sv) && SvTYPE(sv) == SVt_REGEXP) {
4790 /* Need to downgrade the REGEXP to a simple(r) scalar. This is analogous
4791 to sv_unglob. We only need it here, so inline it. */
4792 const svtype new_type = SvMAGIC(sv) || SvSTASH(sv) ? SVt_PVMG : SVt_PV;
4793 SV *const temp = newSV_type(new_type);
4794 void *const temp_p = SvANY(sv);
4796 if (new_type == SVt_PVMG) {
4797 SvMAGIC_set(temp, SvMAGIC(sv));
4798 SvMAGIC_set(sv, NULL);
4799 SvSTASH_set(temp, SvSTASH(sv));
4800 SvSTASH_set(sv, NULL);
4802 SvCUR_set(temp, SvCUR(sv));
4803 /* Remember that SvPVX is in the head, not the body. */
4805 SvLEN_set(temp, SvLEN(sv));
4806 /* This signals "buffer is owned by someone else" in sv_clear,
4807 which is the least effort way to stop it freeing the buffer.
4809 SvLEN_set(sv, SvLEN(sv)+1);
4811 /* Their buffer is already owned by someone else. */
4812 SvPVX(sv) = savepvn(SvPVX(sv), SvCUR(sv));
4813 SvLEN_set(temp, SvCUR(sv)+1);
4816 /* Now swap the rest of the bodies. */
4818 SvFLAGS(sv) &= ~(SVf_FAKE|SVTYPEMASK);
4819 SvFLAGS(sv) |= new_type;
4820 SvANY(sv) = SvANY(temp);
4822 SvFLAGS(temp) &= ~(SVTYPEMASK);
4823 SvFLAGS(temp) |= SVt_REGEXP|SVf_FAKE;
4824 SvANY(temp) = temp_p;
4833 Efficient removal of characters from the beginning of the string buffer.
4834 SvPOK(sv) must be true and the C<ptr> must be a pointer to somewhere inside
4835 the string buffer. The C<ptr> becomes the first character of the adjusted
4836 string. Uses the "OOK hack".
4837 Beware: after this function returns, C<ptr> and SvPVX_const(sv) may no longer
4838 refer to the same chunk of data.
4844 Perl_sv_chop(pTHX_ register SV *const sv, register const char *const ptr)
4850 const U8 *real_start;
4854 PERL_ARGS_ASSERT_SV_CHOP;
4856 if (!ptr || !SvPOKp(sv))
4858 delta = ptr - SvPVX_const(sv);
4860 /* Nothing to do. */
4863 /* SvPVX(sv) may move in SV_CHECK_THINKFIRST(sv), but after this line,
4864 nothing uses the value of ptr any more. */
4865 max_delta = SvLEN(sv) ? SvLEN(sv) : SvCUR(sv);
4866 if (ptr <= SvPVX_const(sv))
4867 Perl_croak(aTHX_ "panic: sv_chop ptr=%p, start=%p, end=%p",
4868 ptr, SvPVX_const(sv), SvPVX_const(sv) + max_delta);
4869 SV_CHECK_THINKFIRST(sv);
4870 if (delta > max_delta)
4871 Perl_croak(aTHX_ "panic: sv_chop ptr=%p (was %p), start=%p, end=%p",
4872 SvPVX_const(sv) + delta, ptr, SvPVX_const(sv),
4873 SvPVX_const(sv) + max_delta);
4876 if (!SvLEN(sv)) { /* make copy of shared string */
4877 const char *pvx = SvPVX_const(sv);
4878 const STRLEN len = SvCUR(sv);
4879 SvGROW(sv, len + 1);
4880 Move(pvx,SvPVX(sv),len,char);
4883 SvFLAGS(sv) |= SVf_OOK;
4886 SvOOK_offset(sv, old_delta);
4888 SvLEN_set(sv, SvLEN(sv) - delta);
4889 SvCUR_set(sv, SvCUR(sv) - delta);
4890 SvPV_set(sv, SvPVX(sv) + delta);
4892 p = (U8 *)SvPVX_const(sv);
4897 real_start = p - delta;
4901 if (delta < 0x100) {
4905 p -= sizeof(STRLEN);
4906 Copy((U8*)&delta, p, sizeof(STRLEN), U8);
4910 /* Fill the preceding buffer with sentinals to verify that no-one is
4912 while (p > real_start) {
4920 =for apidoc sv_catpvn
4922 Concatenates the string onto the end of the string which is in the SV. The
4923 C<len> indicates number of bytes to copy. If the SV has the UTF-8
4924 status set, then the bytes appended should be valid UTF-8.
4925 Handles 'get' magic, but not 'set' magic. See C<sv_catpvn_mg>.
4927 =for apidoc sv_catpvn_flags
4929 Concatenates the string onto the end of the string which is in the SV. The
4930 C<len> indicates number of bytes to copy. If the SV has the UTF-8
4931 status set, then the bytes appended should be valid UTF-8.
4932 If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<dsv> if
4933 appropriate, else not. C<sv_catpvn> and C<sv_catpvn_nomg> are implemented
4934 in terms of this function.
4940 Perl_sv_catpvn_flags(pTHX_ register SV *const dsv, register const char *sstr, register const STRLEN slen, const I32 flags)
4944 const char * const dstr = SvPV_force_flags(dsv, dlen, flags);
4946 PERL_ARGS_ASSERT_SV_CATPVN_FLAGS;
4948 SvGROW(dsv, dlen + slen + 1);
4950 sstr = SvPVX_const(dsv);
4951 Move(sstr, SvPVX(dsv) + dlen, slen, char);
4952 SvCUR_set(dsv, SvCUR(dsv) + slen);
4954 (void)SvPOK_only_UTF8(dsv); /* validate pointer */
4956 if (flags & SV_SMAGIC)
4961 =for apidoc sv_catsv
4963 Concatenates the string from SV C<ssv> onto the end of the string in
4964 SV C<dsv>. Modifies C<dsv> but not C<ssv>. Handles 'get' magic, but
4965 not 'set' magic. See C<sv_catsv_mg>.
4967 =for apidoc sv_catsv_flags
4969 Concatenates the string from SV C<ssv> onto the end of the string in
4970 SV C<dsv>. Modifies C<dsv> but not C<ssv>. If C<flags> has C<SV_GMAGIC>
4971 bit set, will C<mg_get> on the SVs if appropriate, else not. C<sv_catsv>
4972 and C<sv_catsv_nomg> are implemented in terms of this function.
4977 Perl_sv_catsv_flags(pTHX_ SV *const dsv, register SV *const ssv, const I32 flags)
4981 PERL_ARGS_ASSERT_SV_CATSV_FLAGS;
4985 const char *spv = SvPV_flags_const(ssv, slen, flags);
4987 /* sutf8 and dutf8 were type bool, but under USE_ITHREADS,
4988 gcc version 2.95.2 20000220 (Debian GNU/Linux) for
4989 Linux xxx 2.2.17 on sparc64 with gcc -O2, we erroneously
4990 get dutf8 = 0x20000000, (i.e. SVf_UTF8) even though
4991 dsv->sv_flags doesn't have that bit set.
4992 Andy Dougherty 12 Oct 2001
4994 const I32 sutf8 = DO_UTF8(ssv);
4997 if (SvGMAGICAL(dsv) && (flags & SV_GMAGIC))
4999 dutf8 = DO_UTF8(dsv);
5001 if (dutf8 != sutf8) {
5003 /* Not modifying source SV, so taking a temporary copy. */
5004 SV* const csv = newSVpvn_flags(spv, slen, SVs_TEMP);
5006 sv_utf8_upgrade(csv);
5007 spv = SvPV_const(csv, slen);
5010 /* Leave enough space for the cat that's about to happen */
5011 sv_utf8_upgrade_flags_grow(dsv, 0, slen);
5013 sv_catpvn_nomg(dsv, spv, slen);
5016 if (flags & SV_SMAGIC)
5021 =for apidoc sv_catpv
5023 Concatenates the string onto the end of the string which is in the SV.
5024 If the SV has the UTF-8 status set, then the bytes appended should be
5025 valid UTF-8. Handles 'get' magic, but not 'set' magic. See C<sv_catpv_mg>.
5030 Perl_sv_catpv(pTHX_ register SV *const sv, register const char *ptr)
5033 register STRLEN len;
5037 PERL_ARGS_ASSERT_SV_CATPV;
5041 junk = SvPV_force(sv, tlen);
5043 SvGROW(sv, tlen + len + 1);
5045 ptr = SvPVX_const(sv);
5046 Move(ptr,SvPVX(sv)+tlen,len+1,char);
5047 SvCUR_set(sv, SvCUR(sv) + len);
5048 (void)SvPOK_only_UTF8(sv); /* validate pointer */
5053 =for apidoc sv_catpv_flags
5055 Concatenates the string onto the end of the string which is in the SV.
5056 If the SV has the UTF-8 status set, then the bytes appended should
5057 be valid UTF-8. If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get>
5058 on the SVs if appropriate, else not.
5064 Perl_sv_catpv_flags(pTHX_ SV *dstr, const char *sstr, const I32 flags)
5066 PERL_ARGS_ASSERT_SV_CATPV_FLAGS;
5067 sv_catpvn_flags(dstr, sstr, strlen(sstr), flags);
5071 =for apidoc sv_catpv_mg
5073 Like C<sv_catpv>, but also handles 'set' magic.
5079 Perl_sv_catpv_mg(pTHX_ register SV *const sv, register const char *const ptr)
5081 PERL_ARGS_ASSERT_SV_CATPV_MG;
5090 Creates a new SV. A non-zero C<len> parameter indicates the number of
5091 bytes of preallocated string space the SV should have. An extra byte for a
5092 trailing NUL is also reserved. (SvPOK is not set for the SV even if string
5093 space is allocated.) The reference count for the new SV is set to 1.
5095 In 5.9.3, newSV() replaces the older NEWSV() API, and drops the first
5096 parameter, I<x>, a debug aid which allowed callers to identify themselves.
5097 This aid has been superseded by a new build option, PERL_MEM_LOG (see
5098 L<perlhacktips/PERL_MEM_LOG>). The older API is still there for use in XS
5099 modules supporting older perls.
5105 Perl_newSV(pTHX_ const STRLEN len)
5112 sv_upgrade(sv, SVt_PV);
5113 SvGROW(sv, len + 1);
5118 =for apidoc sv_magicext
5120 Adds magic to an SV, upgrading it if necessary. Applies the
5121 supplied vtable and returns a pointer to the magic added.
5123 Note that C<sv_magicext> will allow things that C<sv_magic> will not.
5124 In particular, you can add magic to SvREADONLY SVs, and add more than
5125 one instance of the same 'how'.
5127 If C<namlen> is greater than zero then a C<savepvn> I<copy> of C<name> is
5128 stored, if C<namlen> is zero then C<name> is stored as-is and - as another
5129 special case - if C<(name && namlen == HEf_SVKEY)> then C<name> is assumed
5130 to contain an C<SV*> and is stored as-is with its REFCNT incremented.
5132 (This is now used as a subroutine by C<sv_magic>.)
5137 Perl_sv_magicext(pTHX_ SV *const sv, SV *const obj, const int how,
5138 const MGVTBL *const vtable, const char *const name, const I32 namlen)
5143 PERL_ARGS_ASSERT_SV_MAGICEXT;
5145 SvUPGRADE(sv, SVt_PVMG);
5146 Newxz(mg, 1, MAGIC);
5147 mg->mg_moremagic = SvMAGIC(sv);
5148 SvMAGIC_set(sv, mg);
5150 /* Sometimes a magic contains a reference loop, where the sv and
5151 object refer to each other. To prevent a reference loop that
5152 would prevent such objects being freed, we look for such loops
5153 and if we find one we avoid incrementing the object refcount.
5155 Note we cannot do this to avoid self-tie loops as intervening RV must
5156 have its REFCNT incremented to keep it in existence.
5159 if (!obj || obj == sv ||
5160 how == PERL_MAGIC_arylen ||
5161 how == PERL_MAGIC_symtab ||
5162 (SvTYPE(obj) == SVt_PVGV &&
5163 (GvSV(obj) == sv || GvHV(obj) == (const HV *)sv
5164 || GvAV(obj) == (const AV *)sv || GvCV(obj) == (const CV *)sv
5165 || GvIOp(obj) == (const IO *)sv || GvFORM(obj) == (const CV *)sv)))
5170 mg->mg_obj = SvREFCNT_inc_simple(obj);
5171 mg->mg_flags |= MGf_REFCOUNTED;
5174 /* Normal self-ties simply pass a null object, and instead of
5175 using mg_obj directly, use the SvTIED_obj macro to produce a
5176 new RV as needed. For glob "self-ties", we are tieing the PVIO
5177 with an RV obj pointing to the glob containing the PVIO. In
5178 this case, to avoid a reference loop, we need to weaken the
5182 if (how == PERL_MAGIC_tiedscalar && SvTYPE(sv) == SVt_PVIO &&
5183 obj && SvROK(obj) && GvIO(SvRV(obj)) == (const IO *)sv)
5189 mg->mg_len = namlen;
5192 mg->mg_ptr = savepvn(name, namlen);
5193 else if (namlen == HEf_SVKEY) {
5194 /* Yes, this is casting away const. This is only for the case of
5195 HEf_SVKEY. I think we need to document this aberation of the
5196 constness of the API, rather than making name non-const, as
5197 that change propagating outwards a long way. */
5198 mg->mg_ptr = (char*)SvREFCNT_inc_simple_NN((SV *)name);
5200 mg->mg_ptr = (char *) name;
5202 mg->mg_virtual = (MGVTBL *) vtable;
5206 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
5211 =for apidoc sv_magic
5213 Adds magic to an SV. First upgrades C<sv> to type C<SVt_PVMG> if necessary,
5214 then adds a new magic item of type C<how> to the head of the magic list.
5216 See C<sv_magicext> (which C<sv_magic> now calls) for a description of the
5217 handling of the C<name> and C<namlen> arguments.
5219 You need to use C<sv_magicext> to add magic to SvREADONLY SVs and also
5220 to add more than one instance of the same 'how'.
5226 Perl_sv_magic(pTHX_ register SV *const sv, SV *const obj, const int how,
5227 const char *const name, const I32 namlen)
5230 const MGVTBL *vtable;
5233 unsigned int vtable_index;
5235 PERL_ARGS_ASSERT_SV_MAGIC;
5237 if (how < 0 || (unsigned)how > C_ARRAY_LENGTH(PL_magic_data)
5238 || ((flags = PL_magic_data[how]),
5239 (vtable_index = flags & PERL_MAGIC_VTABLE_MASK)
5240 > magic_vtable_max))
5241 Perl_croak(aTHX_ "Don't know how to handle magic of type \\%o", how);
5243 /* PERL_MAGIC_ext is reserved for use by extensions not perl internals.
5244 Useful for attaching extension internal data to perl vars.
5245 Note that multiple extensions may clash if magical scalars
5246 etc holding private data from one are passed to another. */
5248 vtable = (vtable_index == magic_vtable_max)
5249 ? NULL : PL_magic_vtables + vtable_index;
5251 #ifdef PERL_OLD_COPY_ON_WRITE
5253 sv_force_normal_flags(sv, 0);
5255 if (SvREADONLY(sv)) {
5257 /* its okay to attach magic to shared strings; the subsequent
5258 * upgrade to PVMG will unshare the string */
5259 !(SvFAKE(sv) && SvTYPE(sv) < SVt_PVMG)
5262 && !PERL_MAGIC_TYPE_READONLY_ACCEPTABLE(how)
5265 Perl_croak_no_modify(aTHX);
5268 if (SvMAGICAL(sv) || (how == PERL_MAGIC_taint && SvTYPE(sv) >= SVt_PVMG)) {
5269 if (SvMAGIC(sv) && (mg = mg_find(sv, how))) {
5270 /* sv_magic() refuses to add a magic of the same 'how' as an
5273 if (how == PERL_MAGIC_taint) {
5275 /* Any scalar which already had taint magic on which someone
5276 (erroneously?) did SvIOK_on() or similar will now be
5277 incorrectly sporting public "OK" flags. */
5278 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
5284 /* Rest of work is done else where */
5285 mg = sv_magicext(sv,obj,how,vtable,name,namlen);
5288 case PERL_MAGIC_taint:
5291 case PERL_MAGIC_ext:
5292 case PERL_MAGIC_dbfile:
5299 S_sv_unmagicext_flags(pTHX_ SV *const sv, const int type, MGVTBL *vtbl, const U32 flags)
5306 if (SvTYPE(sv) < SVt_PVMG || !SvMAGIC(sv))
5308 mgp = &(((XPVMG*) SvANY(sv))->xmg_u.xmg_magic);
5309 for (mg = *mgp; mg; mg = *mgp) {
5310 const MGVTBL* const virt = mg->mg_virtual;
5311 if (mg->mg_type == type && (!flags || virt == vtbl)) {
5312 *mgp = mg->mg_moremagic;
5313 if (virt && virt->svt_free)
5314 virt->svt_free(aTHX_ sv, mg);
5315 if (mg->mg_ptr && mg->mg_type != PERL_MAGIC_regex_global) {
5317 Safefree(mg->mg_ptr);
5318 else if (mg->mg_len == HEf_SVKEY)
5319 SvREFCNT_dec(MUTABLE_SV(mg->mg_ptr));
5320 else if (mg->mg_type == PERL_MAGIC_utf8)
5321 Safefree(mg->mg_ptr);
5323 if (mg->mg_flags & MGf_REFCOUNTED)
5324 SvREFCNT_dec(mg->mg_obj);
5328 mgp = &mg->mg_moremagic;
5331 if (SvMAGICAL(sv)) /* if we're under save_magic, wait for restore_magic; */
5332 mg_magical(sv); /* else fix the flags now */
5336 SvFLAGS(sv) |= (SvFLAGS(sv) & (SVp_IOK|SVp_NOK|SVp_POK)) >> PRIVSHIFT;
5342 =for apidoc sv_unmagic
5344 Removes all magic of type C<type> from an SV.
5350 Perl_sv_unmagic(pTHX_ SV *const sv, const int type)
5352 PERL_ARGS_ASSERT_SV_UNMAGIC;
5353 return S_sv_unmagicext_flags(aTHX_ sv, type, NULL, 0);
5357 =for apidoc sv_unmagicext
5359 Removes all magic of type C<type> with the specified C<vtbl> from an SV.
5365 Perl_sv_unmagicext(pTHX_ SV *const sv, const int type, MGVTBL *vtbl)
5367 PERL_ARGS_ASSERT_SV_UNMAGICEXT;
5368 return S_sv_unmagicext_flags(aTHX_ sv, type, vtbl, 1);
5372 =for apidoc sv_rvweaken
5374 Weaken a reference: set the C<SvWEAKREF> flag on this RV; give the
5375 referred-to SV C<PERL_MAGIC_backref> magic if it hasn't already; and
5376 push a back-reference to this RV onto the array of backreferences
5377 associated with that magic. If the RV is magical, set magic will be
5378 called after the RV is cleared.
5384 Perl_sv_rvweaken(pTHX_ SV *const sv)
5388 PERL_ARGS_ASSERT_SV_RVWEAKEN;
5390 if (!SvOK(sv)) /* let undefs pass */
5393 Perl_croak(aTHX_ "Can't weaken a nonreference");
5394 else if (SvWEAKREF(sv)) {
5395 Perl_ck_warner(aTHX_ packWARN(WARN_MISC), "Reference is already weak");
5399 Perl_sv_add_backref(aTHX_ tsv, sv);
5405 /* Give tsv backref magic if it hasn't already got it, then push a
5406 * back-reference to sv onto the array associated with the backref magic.
5408 * As an optimisation, if there's only one backref and it's not an AV,
5409 * store it directly in the HvAUX or mg_obj slot, avoiding the need to
5410 * allocate an AV. (Whether the slot holds an AV tells us whether this is
5414 /* A discussion about the backreferences array and its refcount:
5416 * The AV holding the backreferences is pointed to either as the mg_obj of
5417 * PERL_MAGIC_backref, or in the specific case of a HV, from the
5418 * xhv_backreferences field. The array is created with a refcount
5419 * of 2. This means that if during global destruction the array gets
5420 * picked on before its parent to have its refcount decremented by the
5421 * random zapper, it won't actually be freed, meaning it's still there for
5422 * when its parent gets freed.
5424 * When the parent SV is freed, the extra ref is killed by
5425 * Perl_sv_kill_backrefs. The other ref is killed, in the case of magic,
5426 * by mg_free() / MGf_REFCOUNTED, or for a hash, by Perl_hv_kill_backrefs.
5428 * When a single backref SV is stored directly, it is not reference
5433 Perl_sv_add_backref(pTHX_ SV *const tsv, SV *const sv)
5440 PERL_ARGS_ASSERT_SV_ADD_BACKREF;
5442 /* find slot to store array or singleton backref */
5444 if (SvTYPE(tsv) == SVt_PVHV) {
5445 svp = (SV**)Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5448 (SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL))))
5450 sv_magic(tsv, NULL, PERL_MAGIC_backref, NULL, 0);
5451 mg = mg_find(tsv, PERL_MAGIC_backref);
5453 svp = &(mg->mg_obj);
5456 /* create or retrieve the array */
5458 if ( (!*svp && SvTYPE(sv) == SVt_PVAV)
5459 || (*svp && SvTYPE(*svp) != SVt_PVAV)
5464 SvREFCNT_inc_simple_void(av);
5465 /* av now has a refcnt of 2; see discussion above */
5467 /* move single existing backref to the array */
5469 AvARRAY(av)[++AvFILLp(av)] = *svp; /* av_push() */
5473 mg->mg_flags |= MGf_REFCOUNTED;
5476 av = MUTABLE_AV(*svp);
5479 /* optimisation: store single backref directly in HvAUX or mg_obj */
5483 /* push new backref */
5484 assert(SvTYPE(av) == SVt_PVAV);
5485 if (AvFILLp(av) >= AvMAX(av)) {
5486 av_extend(av, AvFILLp(av)+1);
5488 AvARRAY(av)[++AvFILLp(av)] = sv; /* av_push() */
5491 /* delete a back-reference to ourselves from the backref magic associated
5492 * with the SV we point to.
5496 Perl_sv_del_backref(pTHX_ SV *const tsv, SV *const sv)
5501 PERL_ARGS_ASSERT_SV_DEL_BACKREF;
5503 if (SvTYPE(tsv) == SVt_PVHV) {
5505 svp = (SV**)Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5509 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
5510 svp = mg ? &(mg->mg_obj) : NULL;
5514 Perl_croak(aTHX_ "panic: del_backref");
5516 if (SvTYPE(*svp) == SVt_PVAV) {
5520 AV * const av = (AV*)*svp;
5522 assert(!SvIS_FREED(av));
5526 /* for an SV with N weak references to it, if all those
5527 * weak refs are deleted, then sv_del_backref will be called
5528 * N times and O(N^2) compares will be done within the backref
5529 * array. To ameliorate this potential slowness, we:
5530 * 1) make sure this code is as tight as possible;
5531 * 2) when looking for SV, look for it at both the head and tail of the
5532 * array first before searching the rest, since some create/destroy
5533 * patterns will cause the backrefs to be freed in order.
5540 SV **p = &svp[fill];
5541 SV *const topsv = *p;
5548 /* We weren't the last entry.
5549 An unordered list has this property that you
5550 can take the last element off the end to fill
5551 the hole, and it's still an unordered list :-)
5557 break; /* should only be one */
5564 AvFILLp(av) = fill-1;
5567 /* optimisation: only a single backref, stored directly */
5569 Perl_croak(aTHX_ "panic: del_backref");
5576 Perl_sv_kill_backrefs(pTHX_ SV *const sv, AV *const av)
5582 PERL_ARGS_ASSERT_SV_KILL_BACKREFS;
5587 /* after multiple passes through Perl_sv_clean_all() for a thinngy
5588 * that has badly leaked, the backref array may have gotten freed,
5589 * since we only protect it against 1 round of cleanup */
5590 if (SvIS_FREED(av)) {
5591 if (PL_in_clean_all) /* All is fair */
5594 "panic: magic_killbackrefs (freed backref AV/SV)");
5598 is_array = (SvTYPE(av) == SVt_PVAV);
5600 assert(!SvIS_FREED(av));
5603 last = svp + AvFILLp(av);
5606 /* optimisation: only a single backref, stored directly */
5612 while (svp <= last) {
5614 SV *const referrer = *svp;
5615 if (SvWEAKREF(referrer)) {
5616 /* XXX Should we check that it hasn't changed? */
5617 assert(SvROK(referrer));
5618 SvRV_set(referrer, 0);
5620 SvWEAKREF_off(referrer);
5621 SvSETMAGIC(referrer);
5622 } else if (SvTYPE(referrer) == SVt_PVGV ||
5623 SvTYPE(referrer) == SVt_PVLV) {
5624 assert(SvTYPE(sv) == SVt_PVHV); /* stash backref */
5625 /* You lookin' at me? */
5626 assert(GvSTASH(referrer));
5627 assert(GvSTASH(referrer) == (const HV *)sv);
5628 GvSTASH(referrer) = 0;
5629 } else if (SvTYPE(referrer) == SVt_PVCV ||
5630 SvTYPE(referrer) == SVt_PVFM) {
5631 if (SvTYPE(sv) == SVt_PVHV) { /* stash backref */
5632 /* You lookin' at me? */
5633 assert(CvSTASH(referrer));
5634 assert(CvSTASH(referrer) == (const HV *)sv);
5635 SvANY(MUTABLE_CV(referrer))->xcv_stash = 0;
5638 assert(SvTYPE(sv) == SVt_PVGV);
5639 /* You lookin' at me? */
5640 assert(CvGV(referrer));
5641 assert(CvGV(referrer) == (const GV *)sv);
5642 anonymise_cv_maybe(MUTABLE_GV(sv),
5643 MUTABLE_CV(referrer));
5648 "panic: magic_killbackrefs (flags=%"UVxf")",
5649 (UV)SvFLAGS(referrer));
5660 SvREFCNT_dec(av); /* remove extra count added by sv_add_backref() */
5666 =for apidoc sv_insert
5668 Inserts a string at the specified offset/length within the SV. Similar to
5669 the Perl substr() function. Handles get magic.
5671 =for apidoc sv_insert_flags
5673 Same as C<sv_insert>, but the extra C<flags> are passed the C<SvPV_force_flags> that applies to C<bigstr>.
5679 Perl_sv_insert_flags(pTHX_ SV *const bigstr, const STRLEN offset, const STRLEN len, const char *const little, const STRLEN littlelen, const U32 flags)
5684 register char *midend;
5685 register char *bigend;
5689 PERL_ARGS_ASSERT_SV_INSERT_FLAGS;
5692 Perl_croak(aTHX_ "Can't modify non-existent substring");
5693 SvPV_force_flags(bigstr, curlen, flags);
5694 (void)SvPOK_only_UTF8(bigstr);
5695 if (offset + len > curlen) {
5696 SvGROW(bigstr, offset+len+1);
5697 Zero(SvPVX(bigstr)+curlen, offset+len-curlen, char);
5698 SvCUR_set(bigstr, offset+len);
5702 i = littlelen - len;
5703 if (i > 0) { /* string might grow */
5704 big = SvGROW(bigstr, SvCUR(bigstr) + i + 1);
5705 mid = big + offset + len;
5706 midend = bigend = big + SvCUR(bigstr);
5709 while (midend > mid) /* shove everything down */
5710 *--bigend = *--midend;
5711 Move(little,big+offset,littlelen,char);
5712 SvCUR_set(bigstr, SvCUR(bigstr) + i);
5717 Move(little,SvPVX(bigstr)+offset,len,char);
5722 big = SvPVX(bigstr);
5725 bigend = big + SvCUR(bigstr);
5727 if (midend > bigend)
5728 Perl_croak(aTHX_ "panic: sv_insert");
5730 if (mid - big > bigend - midend) { /* faster to shorten from end */
5732 Move(little, mid, littlelen,char);
5735 i = bigend - midend;
5737 Move(midend, mid, i,char);
5741 SvCUR_set(bigstr, mid - big);
5743 else if ((i = mid - big)) { /* faster from front */
5744 midend -= littlelen;
5746 Move(big, midend - i, i, char);
5747 sv_chop(bigstr,midend-i);
5749 Move(little, mid, littlelen,char);
5751 else if (littlelen) {
5752 midend -= littlelen;
5753 sv_chop(bigstr,midend);
5754 Move(little,midend,littlelen,char);
5757 sv_chop(bigstr,midend);
5763 =for apidoc sv_replace
5765 Make the first argument a copy of the second, then delete the original.
5766 The target SV physically takes over ownership of the body of the source SV
5767 and inherits its flags; however, the target keeps any magic it owns,
5768 and any magic in the source is discarded.
5769 Note that this is a rather specialist SV copying operation; most of the
5770 time you'll want to use C<sv_setsv> or one of its many macro front-ends.
5776 Perl_sv_replace(pTHX_ register SV *const sv, register SV *const nsv)
5779 const U32 refcnt = SvREFCNT(sv);
5781 PERL_ARGS_ASSERT_SV_REPLACE;
5783 SV_CHECK_THINKFIRST_COW_DROP(sv);
5784 if (SvREFCNT(nsv) != 1) {
5785 Perl_croak(aTHX_ "panic: reference miscount on nsv in sv_replace()"
5786 " (%" UVuf " != 1)", (UV) SvREFCNT(nsv));
5788 if (SvMAGICAL(sv)) {
5792 sv_upgrade(nsv, SVt_PVMG);
5793 SvMAGIC_set(nsv, SvMAGIC(sv));
5794 SvFLAGS(nsv) |= SvMAGICAL(sv);
5796 SvMAGIC_set(sv, NULL);
5800 assert(!SvREFCNT(sv));
5801 #ifdef DEBUG_LEAKING_SCALARS
5802 sv->sv_flags = nsv->sv_flags;
5803 sv->sv_any = nsv->sv_any;
5804 sv->sv_refcnt = nsv->sv_refcnt;
5805 sv->sv_u = nsv->sv_u;
5807 StructCopy(nsv,sv,SV);
5809 if(SvTYPE(sv) == SVt_IV) {
5811 = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
5815 #ifdef PERL_OLD_COPY_ON_WRITE
5816 if (SvIsCOW_normal(nsv)) {
5817 /* We need to follow the pointers around the loop to make the
5818 previous SV point to sv, rather than nsv. */
5821 while ((next = SV_COW_NEXT_SV(current)) != nsv) {
5824 assert(SvPVX_const(current) == SvPVX_const(nsv));
5826 /* Make the SV before us point to the SV after us. */
5828 PerlIO_printf(Perl_debug_log, "previous is\n");
5830 PerlIO_printf(Perl_debug_log,
5831 "move it from 0x%"UVxf" to 0x%"UVxf"\n",
5832 (UV) SV_COW_NEXT_SV(current), (UV) sv);
5834 SV_COW_NEXT_SV_SET(current, sv);
5837 SvREFCNT(sv) = refcnt;
5838 SvFLAGS(nsv) |= SVTYPEMASK; /* Mark as freed */
5843 /* We're about to free a GV which has a CV that refers back to us.
5844 * If that CV will outlive us, make it anonymous (i.e. fix up its CvGV
5848 S_anonymise_cv_maybe(pTHX_ GV *gv, CV* cv)
5854 PERL_ARGS_ASSERT_ANONYMISE_CV_MAYBE;
5857 assert(SvREFCNT(gv) == 0);
5858 assert(isGV(gv) && isGV_with_GP(gv));
5860 assert(!CvANON(cv));
5861 assert(CvGV(cv) == gv);
5863 /* will the CV shortly be freed by gp_free() ? */
5864 if (GvCV(gv) == cv && GvGP(gv)->gp_refcnt < 2 && SvREFCNT(cv) < 2) {
5865 SvANY(cv)->xcv_gv = NULL;
5869 /* if not, anonymise: */
5870 stash = GvSTASH(gv) && HvNAME(GvSTASH(gv))
5871 ? HvENAME(GvSTASH(gv)) : NULL;
5872 gvname = Perl_newSVpvf(aTHX_ "%s::__ANON__",
5873 stash ? stash : "__ANON__");
5874 anongv = gv_fetchsv(gvname, GV_ADDMULTI, SVt_PVCV);
5875 SvREFCNT_dec(gvname);
5879 SvANY(cv)->xcv_gv = MUTABLE_GV(SvREFCNT_inc(anongv));
5884 =for apidoc sv_clear
5886 Clear an SV: call any destructors, free up any memory used by the body,
5887 and free the body itself. The SV's head is I<not> freed, although
5888 its type is set to all 1's so that it won't inadvertently be assumed
5889 to be live during global destruction etc.
5890 This function should only be called when REFCNT is zero. Most of the time
5891 you'll want to call C<sv_free()> (or its macro wrapper C<SvREFCNT_dec>)
5898 Perl_sv_clear(pTHX_ SV *const orig_sv)
5903 const struct body_details *sv_type_details;
5906 register SV *sv = orig_sv;
5909 PERL_ARGS_ASSERT_SV_CLEAR;
5911 /* within this loop, sv is the SV currently being freed, and
5912 * iter_sv is the most recent AV or whatever that's being iterated
5913 * over to provide more SVs */
5919 assert(SvREFCNT(sv) == 0);
5920 assert(SvTYPE(sv) != SVTYPEMASK);
5922 if (type <= SVt_IV) {
5923 /* See the comment in sv.h about the collusion between this
5924 * early return and the overloading of the NULL slots in the
5928 SvFLAGS(sv) &= SVf_BREAK;
5929 SvFLAGS(sv) |= SVTYPEMASK;
5933 assert(!SvOBJECT(sv) || type >= SVt_PVMG); /* objs are always >= MG */
5935 if (type >= SVt_PVMG) {
5937 if (!curse(sv, 1)) goto get_next_sv;
5938 type = SvTYPE(sv); /* destructor may have changed it */
5940 /* Free back-references before magic, in case the magic calls
5941 * Perl code that has weak references to sv. */
5942 if (type == SVt_PVHV) {
5943 Perl_hv_kill_backrefs(aTHX_ MUTABLE_HV(sv));
5947 else if (type == SVt_PVMG && SvPAD_OUR(sv)) {
5948 SvREFCNT_dec(SvOURSTASH(sv));
5949 } else if (SvMAGIC(sv)) {
5950 /* Free back-references before other types of magic. */
5951 sv_unmagic(sv, PERL_MAGIC_backref);
5954 if (type == SVt_PVMG && SvPAD_TYPED(sv))
5955 SvREFCNT_dec(SvSTASH(sv));
5958 /* case SVt_BIND: */
5961 IoIFP(sv) != PerlIO_stdin() &&
5962 IoIFP(sv) != PerlIO_stdout() &&
5963 IoIFP(sv) != PerlIO_stderr() &&
5964 !(IoFLAGS(sv) & IOf_FAKE_DIRP))
5966 io_close(MUTABLE_IO(sv), FALSE);
5968 if (IoDIRP(sv) && !(IoFLAGS(sv) & IOf_FAKE_DIRP))
5969 PerlDir_close(IoDIRP(sv));
5970 IoDIRP(sv) = (DIR*)NULL;
5971 Safefree(IoTOP_NAME(sv));
5972 Safefree(IoFMT_NAME(sv));
5973 Safefree(IoBOTTOM_NAME(sv));
5976 /* FIXME for plugins */
5977 pregfree2((REGEXP*) sv);
5981 cv_undef(MUTABLE_CV(sv));
5982 /* If we're in a stash, we don't own a reference to it.
5983 * However it does have a back reference to us, which needs to
5985 if ((stash = CvSTASH(sv)))
5986 sv_del_backref(MUTABLE_SV(stash), sv);
5989 if (PL_last_swash_hv == (const HV *)sv) {
5990 PL_last_swash_hv = NULL;
5992 if (HvTOTALKEYS((HV*)sv) > 0) {
5994 /* this statement should match the one at the beginning of
5995 * hv_undef_flags() */
5996 if ( PL_phase != PERL_PHASE_DESTRUCT
5997 && (name = HvNAME((HV*)sv)))
6000 (void)hv_delete(PL_stashcache, name,
6001 HvNAMELEN_get((HV*)sv), G_DISCARD);
6002 hv_name_set((HV*)sv, NULL, 0, 0);
6005 /* save old iter_sv in unused SvSTASH field */
6006 assert(!SvOBJECT(sv));
6007 SvSTASH(sv) = (HV*)iter_sv;
6010 /* XXX ideally we should save the old value of hash_index
6011 * too, but I can't think of any place to hide it. The
6012 * effect of not saving it is that for freeing hashes of
6013 * hashes, we become quadratic in scanning the HvARRAY of
6014 * the top hash looking for new entries to free; but
6015 * hopefully this will be dwarfed by the freeing of all
6016 * the nested hashes. */
6018 next_sv = Perl_hfree_next_entry(aTHX_ (HV*)sv, &hash_index);
6019 goto get_next_sv; /* process this new sv */
6021 /* free empty hash */
6022 Perl_hv_undef_flags(aTHX_ MUTABLE_HV(sv), HV_NAME_SETALL);
6023 assert(!HvARRAY((HV*)sv));
6027 AV* av = MUTABLE_AV(sv);
6028 if (PL_comppad == av) {
6032 if (AvREAL(av) && AvFILLp(av) > -1) {
6033 next_sv = AvARRAY(av)[AvFILLp(av)--];
6034 /* save old iter_sv in top-most slot of AV,
6035 * and pray that it doesn't get wiped in the meantime */
6036 AvARRAY(av)[AvMAX(av)] = iter_sv;
6038 goto get_next_sv; /* process this new sv */
6040 Safefree(AvALLOC(av));
6045 if (LvTYPE(sv) == 'T') { /* for tie: return HE to pool */
6046 SvREFCNT_dec(HeKEY_sv((HE*)LvTARG(sv)));
6047 HeNEXT((HE*)LvTARG(sv)) = PL_hv_fetch_ent_mh;
6048 PL_hv_fetch_ent_mh = (HE*)LvTARG(sv);
6050 else if (LvTYPE(sv) != 't') /* unless tie: unrefcnted fake SV** */
6051 SvREFCNT_dec(LvTARG(sv));
6053 if (isGV_with_GP(sv)) {
6054 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
6055 && HvENAME_get(stash))
6056 mro_method_changed_in(stash);
6057 gp_free(MUTABLE_GV(sv));
6059 unshare_hek(GvNAME_HEK(sv));
6060 /* If we're in a stash, we don't own a reference to it.
6061 * However it does have a back reference to us, which
6062 * needs to be cleared. */
6063 if (!SvVALID(sv) && (stash = GvSTASH(sv)))
6064 sv_del_backref(MUTABLE_SV(stash), sv);
6066 /* FIXME. There are probably more unreferenced pointers to SVs
6067 * in the interpreter struct that we should check and tidy in
6068 * a similar fashion to this: */
6069 if ((const GV *)sv == PL_last_in_gv)
6070 PL_last_in_gv = NULL;
6076 /* Don't bother with SvOOK_off(sv); as we're only going to
6080 SvOOK_offset(sv, offset);
6081 SvPV_set(sv, SvPVX_mutable(sv) - offset);
6082 /* Don't even bother with turning off the OOK flag. */
6087 SV * const target = SvRV(sv);
6089 sv_del_backref(target, sv);
6094 #ifdef PERL_OLD_COPY_ON_WRITE
6095 else if (SvPVX_const(sv)
6096 && !(SvTYPE(sv) == SVt_PVIO
6097 && !(IoFLAGS(sv) & IOf_FAKE_DIRP)))
6101 PerlIO_printf(Perl_debug_log, "Copy on write: clear\n");
6105 sv_release_COW(sv, SvPVX_const(sv), SV_COW_NEXT_SV(sv));
6107 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
6111 } else if (SvLEN(sv)) {
6112 Safefree(SvPVX_const(sv));
6116 else if (SvPVX_const(sv) && SvLEN(sv)
6117 && !(SvTYPE(sv) == SVt_PVIO
6118 && !(IoFLAGS(sv) & IOf_FAKE_DIRP)))
6119 Safefree(SvPVX_mutable(sv));
6120 else if (SvPVX_const(sv) && SvREADONLY(sv) && SvFAKE(sv)) {
6121 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
6132 SvFLAGS(sv) &= SVf_BREAK;
6133 SvFLAGS(sv) |= SVTYPEMASK;
6135 sv_type_details = bodies_by_type + type;
6136 if (sv_type_details->arena) {
6137 del_body(((char *)SvANY(sv) + sv_type_details->offset),
6138 &PL_body_roots[type]);
6140 else if (sv_type_details->body_size) {
6141 safefree(SvANY(sv));
6145 /* caller is responsible for freeing the head of the original sv */
6146 if (sv != orig_sv && !SvREFCNT(sv))
6149 /* grab and free next sv, if any */
6157 else if (!iter_sv) {
6159 } else if (SvTYPE(iter_sv) == SVt_PVAV) {
6160 AV *const av = (AV*)iter_sv;
6161 if (AvFILLp(av) > -1) {
6162 sv = AvARRAY(av)[AvFILLp(av)--];
6164 else { /* no more elements of current AV to free */
6167 /* restore previous value, squirrelled away */
6168 iter_sv = AvARRAY(av)[AvMAX(av)];
6169 Safefree(AvALLOC(av));
6172 } else if (SvTYPE(iter_sv) == SVt_PVHV) {
6173 sv = Perl_hfree_next_entry(aTHX_ (HV*)iter_sv, &hash_index);
6174 if (!sv && !HvTOTALKEYS((HV *)iter_sv)) {
6175 /* no more elements of current HV to free */
6178 /* Restore previous value of iter_sv, squirrelled away */
6179 assert(!SvOBJECT(sv));
6180 iter_sv = (SV*)SvSTASH(sv);
6182 /* ideally we should restore the old hash_index here,
6183 * but we don't currently save the old value */
6186 /* free any remaining detritus from the hash struct */
6187 Perl_hv_undef_flags(aTHX_ MUTABLE_HV(sv), HV_NAME_SETALL);
6188 assert(!HvARRAY((HV*)sv));
6193 /* unrolled SvREFCNT_dec and sv_free2 follows: */
6197 if (!SvREFCNT(sv)) {
6201 if (--(SvREFCNT(sv)))
6205 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEBUGGING),
6206 "Attempt to free temp prematurely: SV 0x%"UVxf
6207 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
6211 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
6212 /* make sure SvREFCNT(sv)==0 happens very seldom */
6213 SvREFCNT(sv) = (~(U32)0)/2;
6222 /* This routine curses the sv itself, not the object referenced by sv. So
6223 sv does not have to be ROK. */
6226 S_curse(pTHX_ SV * const sv, const bool check_refcnt) {
6229 PERL_ARGS_ASSERT_CURSE;
6230 assert(SvOBJECT(sv));
6232 if (PL_defstash && /* Still have a symbol table? */
6239 stash = SvSTASH(sv);
6240 destructor = StashHANDLER(stash,DESTROY);
6242 /* A constant subroutine can have no side effects, so
6243 don't bother calling it. */
6244 && !CvCONST(destructor)
6245 /* Don't bother calling an empty destructor */
6246 && (CvISXSUB(destructor)
6247 || (CvSTART(destructor)
6248 && (CvSTART(destructor)->op_next->op_type
6251 SV* const tmpref = newRV(sv);
6252 SvREADONLY_on(tmpref); /* DESTROY() could be naughty */
6254 PUSHSTACKi(PERLSI_DESTROY);
6259 call_sv(MUTABLE_SV(destructor),
6260 G_DISCARD|G_EVAL|G_KEEPERR|G_VOID);
6264 if(SvREFCNT(tmpref) < 2) {
6265 /* tmpref is not kept alive! */
6267 SvRV_set(tmpref, NULL);
6270 SvREFCNT_dec(tmpref);
6272 } while (SvOBJECT(sv) && SvSTASH(sv) != stash);
6275 if (check_refcnt && SvREFCNT(sv)) {
6276 if (PL_in_clean_objs)
6278 "DESTROY created new reference to dead object '%s'",
6280 /* DESTROY gave object new lease on life */
6286 SvREFCNT_dec(SvSTASH(sv)); /* possibly of changed persuasion */
6287 SvOBJECT_off(sv); /* Curse the object. */
6288 if (SvTYPE(sv) != SVt_PVIO)
6289 --PL_sv_objcount;/* XXX Might want something more general */
6295 =for apidoc sv_newref
6297 Increment an SV's reference count. Use the C<SvREFCNT_inc()> wrapper
6304 Perl_sv_newref(pTHX_ SV *const sv)
6306 PERL_UNUSED_CONTEXT;
6315 Decrement an SV's reference count, and if it drops to zero, call
6316 C<sv_clear> to invoke destructors and free up any memory used by
6317 the body; finally, deallocate the SV's head itself.
6318 Normally called via a wrapper macro C<SvREFCNT_dec>.
6324 Perl_sv_free(pTHX_ SV *const sv)
6329 if (SvREFCNT(sv) == 0) {
6330 if (SvFLAGS(sv) & SVf_BREAK)
6331 /* this SV's refcnt has been artificially decremented to
6332 * trigger cleanup */
6334 if (PL_in_clean_all) /* All is fair */
6336 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
6337 /* make sure SvREFCNT(sv)==0 happens very seldom */
6338 SvREFCNT(sv) = (~(U32)0)/2;
6341 if (ckWARN_d(WARN_INTERNAL)) {
6342 #ifdef DEBUG_LEAKING_SCALARS_FORK_DUMP
6343 Perl_dump_sv_child(aTHX_ sv);
6345 #ifdef DEBUG_LEAKING_SCALARS
6348 #ifdef DEBUG_LEAKING_SCALARS_ABORT
6349 if (PL_warnhook == PERL_WARNHOOK_FATAL
6350 || ckDEAD(packWARN(WARN_INTERNAL))) {
6351 /* Don't let Perl_warner cause us to escape our fate: */
6355 /* This may not return: */
6356 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
6357 "Attempt to free unreferenced scalar: SV 0x%"UVxf
6358 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
6361 #ifdef DEBUG_LEAKING_SCALARS_ABORT
6366 if (--(SvREFCNT(sv)) > 0)
6368 Perl_sv_free2(aTHX_ sv);
6372 Perl_sv_free2(pTHX_ SV *const sv)
6376 PERL_ARGS_ASSERT_SV_FREE2;
6380 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEBUGGING),
6381 "Attempt to free temp prematurely: SV 0x%"UVxf
6382 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
6386 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
6387 /* make sure SvREFCNT(sv)==0 happens very seldom */
6388 SvREFCNT(sv) = (~(U32)0)/2;
6399 Returns the length of the string in the SV. Handles magic and type
6400 coercion. See also C<SvCUR>, which gives raw access to the xpv_cur slot.
6406 Perl_sv_len(pTHX_ register SV *const sv)
6414 len = mg_length(sv);
6416 (void)SvPV_const(sv, len);
6421 =for apidoc sv_len_utf8
6423 Returns the number of characters in the string in an SV, counting wide
6424 UTF-8 bytes as a single character. Handles magic and type coercion.
6430 * The length is cached in PERL_MAGIC_utf8, in the mg_len field. Also the
6431 * mg_ptr is used, by sv_pos_u2b() and sv_pos_b2u() - see the comments below.
6432 * (Note that the mg_len is not the length of the mg_ptr field.
6433 * This allows the cache to store the character length of the string without
6434 * needing to malloc() extra storage to attach to the mg_ptr.)
6439 Perl_sv_len_utf8(pTHX_ register SV *const sv)
6445 return mg_length(sv);
6449 const U8 *s = (U8*)SvPV_const(sv, len);
6453 MAGIC *mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_utf8) : NULL;
6455 if (mg && (mg->mg_len != -1 || mg->mg_ptr)) {
6456 if (mg->mg_len != -1)
6459 /* We can use the offset cache for a headstart.
6460 The longer value is stored in the first pair. */
6461 STRLEN *cache = (STRLEN *) mg->mg_ptr;
6463 ulen = cache[0] + Perl_utf8_length(aTHX_ s + cache[1],
6467 if (PL_utf8cache < 0) {
6468 const STRLEN real = Perl_utf8_length(aTHX_ s, s + len);
6469 assert_uft8_cache_coherent("sv_len_utf8", ulen, real, sv);
6473 ulen = Perl_utf8_length(aTHX_ s, s + len);
6474 utf8_mg_len_cache_update(sv, &mg, ulen);
6478 return Perl_utf8_length(aTHX_ s, s + len);
6482 /* Walk forwards to find the byte corresponding to the passed in UTF-8
6485 S_sv_pos_u2b_forwards(const U8 *const start, const U8 *const send,
6486 STRLEN *const uoffset_p, bool *const at_end)
6488 const U8 *s = start;
6489 STRLEN uoffset = *uoffset_p;
6491 PERL_ARGS_ASSERT_SV_POS_U2B_FORWARDS;
6493 while (s < send && uoffset) {
6500 else if (s > send) {
6502 /* This is the existing behaviour. Possibly it should be a croak, as
6503 it's actually a bounds error */
6506 *uoffset_p -= uoffset;
6510 /* Given the length of the string in both bytes and UTF-8 characters, decide
6511 whether to walk forwards or backwards to find the byte corresponding to
6512 the passed in UTF-8 offset. */
6514 S_sv_pos_u2b_midway(const U8 *const start, const U8 *send,
6515 STRLEN uoffset, const STRLEN uend)
6517 STRLEN backw = uend - uoffset;
6519 PERL_ARGS_ASSERT_SV_POS_U2B_MIDWAY;
6521 if (uoffset < 2 * backw) {
6522 /* The assumption is that going forwards is twice the speed of going
6523 forward (that's where the 2 * backw comes from).
6524 (The real figure of course depends on the UTF-8 data.) */
6525 const U8 *s = start;
6527 while (s < send && uoffset--)
6537 while (UTF8_IS_CONTINUATION(*send))
6540 return send - start;
6543 /* For the string representation of the given scalar, find the byte
6544 corresponding to the passed in UTF-8 offset. uoffset0 and boffset0
6545 give another position in the string, *before* the sought offset, which
6546 (which is always true, as 0, 0 is a valid pair of positions), which should
6547 help reduce the amount of linear searching.
6548 If *mgp is non-NULL, it should point to the UTF-8 cache magic, which
6549 will be used to reduce the amount of linear searching. The cache will be
6550 created if necessary, and the found value offered to it for update. */
6552 S_sv_pos_u2b_cached(pTHX_ SV *const sv, MAGIC **const mgp, const U8 *const start,
6553 const U8 *const send, STRLEN uoffset,
6554 STRLEN uoffset0, STRLEN boffset0)
6556 STRLEN boffset = 0; /* Actually always set, but let's keep gcc happy. */
6558 bool at_end = FALSE;
6560 PERL_ARGS_ASSERT_SV_POS_U2B_CACHED;
6562 assert (uoffset >= uoffset0);
6569 && (*mgp || (SvTYPE(sv) >= SVt_PVMG &&
6570 (*mgp = mg_find(sv, PERL_MAGIC_utf8))))) {
6571 if ((*mgp)->mg_ptr) {
6572 STRLEN *cache = (STRLEN *) (*mgp)->mg_ptr;
6573 if (cache[0] == uoffset) {
6574 /* An exact match. */
6577 if (cache[2] == uoffset) {
6578 /* An exact match. */
6582 if (cache[0] < uoffset) {
6583 /* The cache already knows part of the way. */
6584 if (cache[0] > uoffset0) {
6585 /* The cache knows more than the passed in pair */
6586 uoffset0 = cache[0];
6587 boffset0 = cache[1];
6589 if ((*mgp)->mg_len != -1) {
6590 /* And we know the end too. */
6592 + sv_pos_u2b_midway(start + boffset0, send,
6594 (*mgp)->mg_len - uoffset0);
6596 uoffset -= uoffset0;
6598 + sv_pos_u2b_forwards(start + boffset0,
6599 send, &uoffset, &at_end);
6600 uoffset += uoffset0;
6603 else if (cache[2] < uoffset) {
6604 /* We're between the two cache entries. */
6605 if (cache[2] > uoffset0) {
6606 /* and the cache knows more than the passed in pair */
6607 uoffset0 = cache[2];
6608 boffset0 = cache[3];
6612 + sv_pos_u2b_midway(start + boffset0,
6615 cache[0] - uoffset0);
6618 + sv_pos_u2b_midway(start + boffset0,
6621 cache[2] - uoffset0);
6625 else if ((*mgp)->mg_len != -1) {
6626 /* If we can take advantage of a passed in offset, do so. */
6627 /* In fact, offset0 is either 0, or less than offset, so don't
6628 need to worry about the other possibility. */
6630 + sv_pos_u2b_midway(start + boffset0, send,
6632 (*mgp)->mg_len - uoffset0);
6637 if (!found || PL_utf8cache < 0) {
6638 STRLEN real_boffset;
6639 uoffset -= uoffset0;
6640 real_boffset = boffset0 + sv_pos_u2b_forwards(start + boffset0,
6641 send, &uoffset, &at_end);
6642 uoffset += uoffset0;
6644 if (found && PL_utf8cache < 0)
6645 assert_uft8_cache_coherent("sv_pos_u2b_cache", boffset,
6647 boffset = real_boffset;
6652 utf8_mg_len_cache_update(sv, mgp, uoffset);
6654 utf8_mg_pos_cache_update(sv, mgp, boffset, uoffset, send - start);
6661 =for apidoc sv_pos_u2b_flags
6663 Converts the value pointed to by offsetp from a count of UTF-8 chars from
6664 the start of the string, to a count of the equivalent number of bytes; if
6665 lenp is non-zero, it does the same to lenp, but this time starting from
6666 the offset, rather than from the start of the string. Handles type coercion.
6667 I<flags> is passed to C<SvPV_flags>, and usually should be
6668 C<SV_GMAGIC|SV_CONST_RETURN> to handle magic.
6674 * sv_pos_u2b_flags() uses, like sv_pos_b2u(), the mg_ptr of the potential
6675 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6676 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
6681 Perl_sv_pos_u2b_flags(pTHX_ SV *const sv, STRLEN uoffset, STRLEN *const lenp,
6688 PERL_ARGS_ASSERT_SV_POS_U2B_FLAGS;
6690 start = (U8*)SvPV_flags(sv, len, flags);
6692 const U8 * const send = start + len;
6694 boffset = sv_pos_u2b_cached(sv, &mg, start, send, uoffset, 0, 0);
6697 && *lenp /* don't bother doing work for 0, as its bytes equivalent
6698 is 0, and *lenp is already set to that. */) {
6699 /* Convert the relative offset to absolute. */
6700 const STRLEN uoffset2 = uoffset + *lenp;
6701 const STRLEN boffset2
6702 = sv_pos_u2b_cached(sv, &mg, start, send, uoffset2,
6703 uoffset, boffset) - boffset;
6717 =for apidoc sv_pos_u2b
6719 Converts the value pointed to by offsetp from a count of UTF-8 chars from
6720 the start of the string, to a count of the equivalent number of bytes; if
6721 lenp is non-zero, it does the same to lenp, but this time starting from
6722 the offset, rather than from the start of the string. Handles magic and
6725 Use C<sv_pos_u2b_flags> in preference, which correctly handles strings longer
6732 * sv_pos_u2b() uses, like sv_pos_b2u(), the mg_ptr of the potential
6733 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6734 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
6738 /* This function is subject to size and sign problems */
6741 Perl_sv_pos_u2b(pTHX_ register SV *const sv, I32 *const offsetp, I32 *const lenp)
6743 PERL_ARGS_ASSERT_SV_POS_U2B;
6746 STRLEN ulen = (STRLEN)*lenp;
6747 *offsetp = (I32)sv_pos_u2b_flags(sv, (STRLEN)*offsetp, &ulen,
6748 SV_GMAGIC|SV_CONST_RETURN);
6751 *offsetp = (I32)sv_pos_u2b_flags(sv, (STRLEN)*offsetp, NULL,
6752 SV_GMAGIC|SV_CONST_RETURN);
6757 S_utf8_mg_len_cache_update(pTHX_ SV *const sv, MAGIC **const mgp,
6760 PERL_ARGS_ASSERT_UTF8_MG_LEN_CACHE_UPDATE;
6764 if (!*mgp && (SvTYPE(sv) < SVt_PVMG ||
6765 !(*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
6766 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, &PL_vtbl_utf8, 0, 0);
6770 (*mgp)->mg_len = ulen;
6771 /* For now, treat "overflowed" as "still unknown". See RT #72924. */
6772 if (ulen != (STRLEN) (*mgp)->mg_len)
6773 (*mgp)->mg_len = -1;
6776 /* Create and update the UTF8 magic offset cache, with the proffered utf8/
6777 byte length pairing. The (byte) length of the total SV is passed in too,
6778 as blen, because for some (more esoteric) SVs, the call to SvPV_const()
6779 may not have updated SvCUR, so we can't rely on reading it directly.
6781 The proffered utf8/byte length pairing isn't used if the cache already has
6782 two pairs, and swapping either for the proffered pair would increase the
6783 RMS of the intervals between known byte offsets.
6785 The cache itself consists of 4 STRLEN values
6786 0: larger UTF-8 offset
6787 1: corresponding byte offset
6788 2: smaller UTF-8 offset
6789 3: corresponding byte offset
6791 Unused cache pairs have the value 0, 0.
6792 Keeping the cache "backwards" means that the invariant of
6793 cache[0] >= cache[2] is maintained even with empty slots, which means that
6794 the code that uses it doesn't need to worry if only 1 entry has actually
6795 been set to non-zero. It also makes the "position beyond the end of the
6796 cache" logic much simpler, as the first slot is always the one to start
6800 S_utf8_mg_pos_cache_update(pTHX_ SV *const sv, MAGIC **const mgp, const STRLEN byte,
6801 const STRLEN utf8, const STRLEN blen)
6805 PERL_ARGS_ASSERT_UTF8_MG_POS_CACHE_UPDATE;
6810 if (!*mgp && (SvTYPE(sv) < SVt_PVMG ||
6811 !(*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
6812 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, (MGVTBL*)&PL_vtbl_utf8, 0,
6814 (*mgp)->mg_len = -1;
6818 if (!(cache = (STRLEN *)(*mgp)->mg_ptr)) {
6819 Newxz(cache, PERL_MAGIC_UTF8_CACHESIZE * 2, STRLEN);
6820 (*mgp)->mg_ptr = (char *) cache;
6824 if (PL_utf8cache < 0 && SvPOKp(sv)) {
6825 /* SvPOKp() because it's possible that sv has string overloading, and
6826 therefore is a reference, hence SvPVX() is actually a pointer.
6827 This cures the (very real) symptoms of RT 69422, but I'm not actually
6828 sure whether we should even be caching the results of UTF-8
6829 operations on overloading, given that nothing stops overloading
6830 returning a different value every time it's called. */
6831 const U8 *start = (const U8 *) SvPVX_const(sv);
6832 const STRLEN realutf8 = utf8_length(start, start + byte);
6834 assert_uft8_cache_coherent("utf8_mg_pos_cache_update", utf8, realutf8,
6838 /* Cache is held with the later position first, to simplify the code
6839 that deals with unbounded ends. */
6841 ASSERT_UTF8_CACHE(cache);
6842 if (cache[1] == 0) {
6843 /* Cache is totally empty */
6846 } else if (cache[3] == 0) {
6847 if (byte > cache[1]) {
6848 /* New one is larger, so goes first. */
6849 cache[2] = cache[0];
6850 cache[3] = cache[1];
6858 #define THREEWAY_SQUARE(a,b,c,d) \
6859 ((float)((d) - (c))) * ((float)((d) - (c))) \
6860 + ((float)((c) - (b))) * ((float)((c) - (b))) \
6861 + ((float)((b) - (a))) * ((float)((b) - (a)))
6863 /* Cache has 2 slots in use, and we know three potential pairs.
6864 Keep the two that give the lowest RMS distance. Do the
6865 calculation in bytes simply because we always know the byte
6866 length. squareroot has the same ordering as the positive value,
6867 so don't bother with the actual square root. */
6868 const float existing = THREEWAY_SQUARE(0, cache[3], cache[1], blen);
6869 if (byte > cache[1]) {
6870 /* New position is after the existing pair of pairs. */
6871 const float keep_earlier
6872 = THREEWAY_SQUARE(0, cache[3], byte, blen);
6873 const float keep_later
6874 = THREEWAY_SQUARE(0, cache[1], byte, blen);
6876 if (keep_later < keep_earlier) {
6877 if (keep_later < existing) {
6878 cache[2] = cache[0];
6879 cache[3] = cache[1];
6885 if (keep_earlier < existing) {
6891 else if (byte > cache[3]) {
6892 /* New position is between the existing pair of pairs. */
6893 const float keep_earlier
6894 = THREEWAY_SQUARE(0, cache[3], byte, blen);
6895 const float keep_later
6896 = THREEWAY_SQUARE(0, byte, cache[1], blen);
6898 if (keep_later < keep_earlier) {
6899 if (keep_later < existing) {
6905 if (keep_earlier < existing) {
6912 /* New position is before the existing pair of pairs. */
6913 const float keep_earlier
6914 = THREEWAY_SQUARE(0, byte, cache[3], blen);
6915 const float keep_later
6916 = THREEWAY_SQUARE(0, byte, cache[1], blen);
6918 if (keep_later < keep_earlier) {
6919 if (keep_later < existing) {
6925 if (keep_earlier < existing) {
6926 cache[0] = cache[2];
6927 cache[1] = cache[3];
6934 ASSERT_UTF8_CACHE(cache);
6937 /* We already know all of the way, now we may be able to walk back. The same
6938 assumption is made as in S_sv_pos_u2b_midway(), namely that walking
6939 backward is half the speed of walking forward. */
6941 S_sv_pos_b2u_midway(pTHX_ const U8 *const s, const U8 *const target,
6942 const U8 *end, STRLEN endu)
6944 const STRLEN forw = target - s;
6945 STRLEN backw = end - target;
6947 PERL_ARGS_ASSERT_SV_POS_B2U_MIDWAY;
6949 if (forw < 2 * backw) {
6950 return utf8_length(s, target);
6953 while (end > target) {
6955 while (UTF8_IS_CONTINUATION(*end)) {
6964 =for apidoc sv_pos_b2u
6966 Converts the value pointed to by offsetp from a count of bytes from the
6967 start of the string, to a count of the equivalent number of UTF-8 chars.
6968 Handles magic and type coercion.
6974 * sv_pos_b2u() uses, like sv_pos_u2b(), the mg_ptr of the potential
6975 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6980 Perl_sv_pos_b2u(pTHX_ register SV *const sv, I32 *const offsetp)
6983 const STRLEN byte = *offsetp;
6984 STRLEN len = 0; /* Actually always set, but let's keep gcc happy. */
6990 PERL_ARGS_ASSERT_SV_POS_B2U;
6995 s = (const U8*)SvPV_const(sv, blen);
6998 Perl_croak(aTHX_ "panic: sv_pos_b2u: bad byte offset");
7004 && SvTYPE(sv) >= SVt_PVMG
7005 && (mg = mg_find(sv, PERL_MAGIC_utf8)))
7008 STRLEN * const cache = (STRLEN *) mg->mg_ptr;
7009 if (cache[1] == byte) {
7010 /* An exact match. */
7011 *offsetp = cache[0];
7014 if (cache[3] == byte) {
7015 /* An exact match. */
7016 *offsetp = cache[2];
7020 if (cache[1] < byte) {
7021 /* We already know part of the way. */
7022 if (mg->mg_len != -1) {
7023 /* Actually, we know the end too. */
7025 + S_sv_pos_b2u_midway(aTHX_ s + cache[1], send,
7026 s + blen, mg->mg_len - cache[0]);
7028 len = cache[0] + utf8_length(s + cache[1], send);
7031 else if (cache[3] < byte) {
7032 /* We're between the two cached pairs, so we do the calculation
7033 offset by the byte/utf-8 positions for the earlier pair,
7034 then add the utf-8 characters from the string start to
7036 len = S_sv_pos_b2u_midway(aTHX_ s + cache[3], send,
7037 s + cache[1], cache[0] - cache[2])
7041 else { /* cache[3] > byte */
7042 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + cache[3],
7046 ASSERT_UTF8_CACHE(cache);
7048 } else if (mg->mg_len != -1) {
7049 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + blen, mg->mg_len);
7053 if (!found || PL_utf8cache < 0) {
7054 const STRLEN real_len = utf8_length(s, send);
7056 if (found && PL_utf8cache < 0)
7057 assert_uft8_cache_coherent("sv_pos_b2u", len, real_len, sv);
7064 utf8_mg_len_cache_update(sv, &mg, len);
7066 utf8_mg_pos_cache_update(sv, &mg, byte, len, blen);
7071 S_assert_uft8_cache_coherent(pTHX_ const char *const func, STRLEN from_cache,
7072 STRLEN real, SV *const sv)
7074 PERL_ARGS_ASSERT_ASSERT_UFT8_CACHE_COHERENT;
7076 /* As this is debugging only code, save space by keeping this test here,
7077 rather than inlining it in all the callers. */
7078 if (from_cache == real)
7081 /* Need to turn the assertions off otherwise we may recurse infinitely
7082 while printing error messages. */
7083 SAVEI8(PL_utf8cache);
7085 Perl_croak(aTHX_ "panic: %s cache %"UVuf" real %"UVuf" for %"SVf,
7086 func, (UV) from_cache, (UV) real, SVfARG(sv));
7092 Returns a boolean indicating whether the strings in the two SVs are
7093 identical. Is UTF-8 and 'use bytes' aware, handles get magic, and will
7094 coerce its args to strings if necessary.
7096 =for apidoc sv_eq_flags
7098 Returns a boolean indicating whether the strings in the two SVs are
7099 identical. Is UTF-8 and 'use bytes' aware and coerces its args to strings
7100 if necessary. If the flags include SV_GMAGIC, it handles get-magic, too.
7106 Perl_sv_eq_flags(pTHX_ register SV *sv1, register SV *sv2, const U32 flags)
7115 SV* svrecode = NULL;
7122 /* if pv1 and pv2 are the same, second SvPV_const call may
7123 * invalidate pv1 (if we are handling magic), so we may need to
7125 if (sv1 == sv2 && flags & SV_GMAGIC
7126 && (SvTHINKFIRST(sv1) || SvGMAGICAL(sv1))) {
7127 pv1 = SvPV_const(sv1, cur1);
7128 sv1 = newSVpvn_flags(pv1, cur1, SVs_TEMP | SvUTF8(sv2));
7130 pv1 = SvPV_flags_const(sv1, cur1, flags);
7138 pv2 = SvPV_flags_const(sv2, cur2, flags);
7140 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
7141 /* Differing utf8ness.
7142 * Do not UTF8size the comparands as a side-effect. */
7145 svrecode = newSVpvn(pv2, cur2);
7146 sv_recode_to_utf8(svrecode, PL_encoding);
7147 pv2 = SvPV_const(svrecode, cur2);
7150 svrecode = newSVpvn(pv1, cur1);
7151 sv_recode_to_utf8(svrecode, PL_encoding);
7152 pv1 = SvPV_const(svrecode, cur1);
7154 /* Now both are in UTF-8. */
7156 SvREFCNT_dec(svrecode);
7162 /* sv1 is the UTF-8 one */
7163 return bytes_cmp_utf8((const U8*)pv2, cur2,
7164 (const U8*)pv1, cur1) == 0;
7167 /* sv2 is the UTF-8 one */
7168 return bytes_cmp_utf8((const U8*)pv1, cur1,
7169 (const U8*)pv2, cur2) == 0;
7175 eq = (pv1 == pv2) || memEQ(pv1, pv2, cur1);
7177 SvREFCNT_dec(svrecode);
7187 Compares the strings in two SVs. Returns -1, 0, or 1 indicating whether the
7188 string in C<sv1> is less than, equal to, or greater than the string in
7189 C<sv2>. Is UTF-8 and 'use bytes' aware, handles get magic, and will
7190 coerce its args to strings if necessary. See also C<sv_cmp_locale>.
7192 =for apidoc sv_cmp_flags
7194 Compares the strings in two SVs. Returns -1, 0, or 1 indicating whether the
7195 string in C<sv1> is less than, equal to, or greater than the string in
7196 C<sv2>. Is UTF-8 and 'use bytes' aware and will coerce its args to strings
7197 if necessary. If the flags include SV_GMAGIC, it handles get magic. See
7198 also C<sv_cmp_locale_flags>.
7204 Perl_sv_cmp(pTHX_ register SV *const sv1, register SV *const sv2)
7206 return sv_cmp_flags(sv1, sv2, SV_GMAGIC);
7210 Perl_sv_cmp_flags(pTHX_ register SV *const sv1, register SV *const sv2,
7215 const char *pv1, *pv2;
7218 SV *svrecode = NULL;
7225 pv1 = SvPV_flags_const(sv1, cur1, flags);
7232 pv2 = SvPV_flags_const(sv2, cur2, flags);
7234 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
7235 /* Differing utf8ness.
7236 * Do not UTF8size the comparands as a side-effect. */
7239 svrecode = newSVpvn(pv2, cur2);
7240 sv_recode_to_utf8(svrecode, PL_encoding);
7241 pv2 = SvPV_const(svrecode, cur2);
7244 const int retval = -bytes_cmp_utf8((const U8*)pv2, cur2,
7245 (const U8*)pv1, cur1);
7246 return retval ? retval < 0 ? -1 : +1 : 0;
7251 svrecode = newSVpvn(pv1, cur1);
7252 sv_recode_to_utf8(svrecode, PL_encoding);
7253 pv1 = SvPV_const(svrecode, cur1);
7256 const int retval = bytes_cmp_utf8((const U8*)pv1, cur1,
7257 (const U8*)pv2, cur2);
7258 return retval ? retval < 0 ? -1 : +1 : 0;
7264 cmp = cur2 ? -1 : 0;
7268 const I32 retval = memcmp((const void*)pv1, (const void*)pv2, cur1 < cur2 ? cur1 : cur2);
7271 cmp = retval < 0 ? -1 : 1;
7272 } else if (cur1 == cur2) {
7275 cmp = cur1 < cur2 ? -1 : 1;
7279 SvREFCNT_dec(svrecode);
7287 =for apidoc sv_cmp_locale
7289 Compares the strings in two SVs in a locale-aware manner. Is UTF-8 and
7290 'use bytes' aware, handles get magic, and will coerce its args to strings
7291 if necessary. See also C<sv_cmp>.
7293 =for apidoc sv_cmp_locale_flags
7295 Compares the strings in two SVs in a locale-aware manner. Is UTF-8 and
7296 'use bytes' aware and will coerce its args to strings if necessary. If the
7297 flags contain SV_GMAGIC, it handles get magic. See also C<sv_cmp_flags>.
7303 Perl_sv_cmp_locale(pTHX_ register SV *const sv1, register SV *const sv2)
7305 return sv_cmp_locale_flags(sv1, sv2, SV_GMAGIC);
7309 Perl_sv_cmp_locale_flags(pTHX_ register SV *const sv1, register SV *const sv2,
7313 #ifdef USE_LOCALE_COLLATE
7319 if (PL_collation_standard)
7323 pv1 = sv1 ? sv_collxfrm_flags(sv1, &len1, flags) : (char *) NULL;
7325 pv2 = sv2 ? sv_collxfrm_flags(sv2, &len2, flags) : (char *) NULL;
7327 if (!pv1 || !len1) {
7338 retval = memcmp((void*)pv1, (void*)pv2, len1 < len2 ? len1 : len2);
7341 return retval < 0 ? -1 : 1;
7344 * When the result of collation is equality, that doesn't mean
7345 * that there are no differences -- some locales exclude some
7346 * characters from consideration. So to avoid false equalities,
7347 * we use the raw string as a tiebreaker.
7353 #endif /* USE_LOCALE_COLLATE */
7355 return sv_cmp(sv1, sv2);
7359 #ifdef USE_LOCALE_COLLATE
7362 =for apidoc sv_collxfrm
7364 This calls C<sv_collxfrm_flags> with the SV_GMAGIC flag. See
7365 C<sv_collxfrm_flags>.
7367 =for apidoc sv_collxfrm_flags
7369 Add Collate Transform magic to an SV if it doesn't already have it. If the
7370 flags contain SV_GMAGIC, it handles get-magic.
7372 Any scalar variable may carry PERL_MAGIC_collxfrm magic that contains the
7373 scalar data of the variable, but transformed to such a format that a normal
7374 memory comparison can be used to compare the data according to the locale
7381 Perl_sv_collxfrm_flags(pTHX_ SV *const sv, STRLEN *const nxp, const I32 flags)
7386 PERL_ARGS_ASSERT_SV_COLLXFRM_FLAGS;
7388 mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_collxfrm) : (MAGIC *) NULL;
7389 if (!mg || !mg->mg_ptr || *(U32*)mg->mg_ptr != PL_collation_ix) {
7395 Safefree(mg->mg_ptr);
7396 s = SvPV_flags_const(sv, len, flags);
7397 if ((xf = mem_collxfrm(s, len, &xlen))) {
7399 #ifdef PERL_OLD_COPY_ON_WRITE
7401 sv_force_normal_flags(sv, 0);
7403 mg = sv_magicext(sv, 0, PERL_MAGIC_collxfrm, &PL_vtbl_collxfrm,
7417 if (mg && mg->mg_ptr) {
7419 return mg->mg_ptr + sizeof(PL_collation_ix);
7427 #endif /* USE_LOCALE_COLLATE */
7430 S_sv_gets_append_to_utf8(pTHX_ SV *const sv, PerlIO *const fp, I32 append)
7432 SV * const tsv = newSV(0);
7435 sv_gets(tsv, fp, 0);
7436 sv_utf8_upgrade_nomg(tsv);
7437 SvCUR_set(sv,append);
7440 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7444 S_sv_gets_read_record(pTHX_ SV *const sv, PerlIO *const fp, I32 append)
7447 const U32 recsize = SvUV(SvRV(PL_rs)); /* RsRECORD() guarantees > 0. */
7448 /* Grab the size of the record we're getting */
7449 char *const buffer = SvGROW(sv, (STRLEN)(recsize + append + 1)) + append;
7456 /* VMS wants read instead of fread, because fread doesn't respect */
7457 /* RMS record boundaries. This is not necessarily a good thing to be */
7458 /* doing, but we've got no other real choice - except avoid stdio
7459 as implementation - perhaps write a :vms layer ?
7461 fd = PerlIO_fileno(fp);
7463 bytesread = PerlLIO_read(fd, buffer, recsize);
7465 else /* in-memory file from PerlIO::Scalar */
7468 bytesread = PerlIO_read(fp, buffer, recsize);
7473 SvCUR_set(sv, bytesread + append);
7474 buffer[bytesread] = '\0';
7475 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7481 Get a line from the filehandle and store it into the SV, optionally
7482 appending to the currently-stored string.
7488 Perl_sv_gets(pTHX_ register SV *const sv, register PerlIO *const fp, I32 append)
7493 register STDCHAR rslast;
7494 register STDCHAR *bp;
7499 PERL_ARGS_ASSERT_SV_GETS;
7501 if (SvTHINKFIRST(sv))
7502 sv_force_normal_flags(sv, append ? 0 : SV_COW_DROP_PV);
7503 /* XXX. If you make this PVIV, then copy on write can copy scalars read
7505 However, perlbench says it's slower, because the existing swipe code
7506 is faster than copy on write.
7507 Swings and roundabouts. */
7508 SvUPGRADE(sv, SVt_PV);
7513 if (PerlIO_isutf8(fp)) {
7515 sv_utf8_upgrade_nomg(sv);
7516 sv_pos_u2b(sv,&append,0);
7518 } else if (SvUTF8(sv)) {
7519 return S_sv_gets_append_to_utf8(aTHX_ sv, fp, append);
7527 if (PerlIO_isutf8(fp))
7530 if (IN_PERL_COMPILETIME) {
7531 /* we always read code in line mode */
7535 else if (RsSNARF(PL_rs)) {
7536 /* If it is a regular disk file use size from stat() as estimate
7537 of amount we are going to read -- may result in mallocing
7538 more memory than we really need if the layers below reduce
7539 the size we read (e.g. CRLF or a gzip layer).
7542 if (!PerlLIO_fstat(PerlIO_fileno(fp), &st) && S_ISREG(st.st_mode)) {
7543 const Off_t offset = PerlIO_tell(fp);
7544 if (offset != (Off_t) -1 && st.st_size + append > offset) {
7545 (void) SvGROW(sv, (STRLEN)((st.st_size - offset) + append + 1));
7551 else if (RsRECORD(PL_rs)) {
7552 return S_sv_gets_read_record(aTHX_ sv, fp, append);
7554 else if (RsPARA(PL_rs)) {
7560 /* Get $/ i.e. PL_rs into same encoding as stream wants */
7561 if (PerlIO_isutf8(fp)) {
7562 rsptr = SvPVutf8(PL_rs, rslen);
7565 if (SvUTF8(PL_rs)) {
7566 if (!sv_utf8_downgrade(PL_rs, TRUE)) {
7567 Perl_croak(aTHX_ "Wide character in $/");
7570 rsptr = SvPV_const(PL_rs, rslen);
7574 rslast = rslen ? rsptr[rslen - 1] : '\0';
7576 if (rspara) { /* have to do this both before and after */
7577 do { /* to make sure file boundaries work right */
7580 i = PerlIO_getc(fp);
7584 PerlIO_ungetc(fp,i);
7590 /* See if we know enough about I/O mechanism to cheat it ! */
7592 /* This used to be #ifdef test - it is made run-time test for ease
7593 of abstracting out stdio interface. One call should be cheap
7594 enough here - and may even be a macro allowing compile
7598 if (PerlIO_fast_gets(fp)) {
7601 * We're going to steal some values from the stdio struct
7602 * and put EVERYTHING in the innermost loop into registers.
7604 register STDCHAR *ptr;
7608 #if defined(VMS) && defined(PERLIO_IS_STDIO)
7609 /* An ungetc()d char is handled separately from the regular
7610 * buffer, so we getc() it back out and stuff it in the buffer.
7612 i = PerlIO_getc(fp);
7613 if (i == EOF) return 0;
7614 *(--((*fp)->_ptr)) = (unsigned char) i;
7618 /* Here is some breathtakingly efficient cheating */
7620 cnt = PerlIO_get_cnt(fp); /* get count into register */
7621 /* make sure we have the room */
7622 if ((I32)(SvLEN(sv) - append) <= cnt + 1) {
7623 /* Not room for all of it
7624 if we are looking for a separator and room for some
7626 if (rslen && cnt > 80 && (I32)SvLEN(sv) > append) {
7627 /* just process what we have room for */
7628 shortbuffered = cnt - SvLEN(sv) + append + 1;
7629 cnt -= shortbuffered;
7633 /* remember that cnt can be negative */
7634 SvGROW(sv, (STRLEN)(append + (cnt <= 0 ? 2 : (cnt + 1))));
7639 bp = (STDCHAR*)SvPVX_const(sv) + append; /* move these two too to registers */
7640 ptr = (STDCHAR*)PerlIO_get_ptr(fp);
7641 DEBUG_P(PerlIO_printf(Perl_debug_log,
7642 "Screamer: entering, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7643 DEBUG_P(PerlIO_printf(Perl_debug_log,
7644 "Screamer: entering: PerlIO * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7645 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7646 PTR2UV(PerlIO_has_base(fp) ? PerlIO_get_base(fp) : 0)));
7651 while (cnt > 0) { /* this | eat */
7653 if ((*bp++ = *ptr++) == rslast) /* really | dust */
7654 goto thats_all_folks; /* screams | sed :-) */
7658 Copy(ptr, bp, cnt, char); /* this | eat */
7659 bp += cnt; /* screams | dust */
7660 ptr += cnt; /* louder | sed :-) */
7662 assert (!shortbuffered);
7663 goto cannot_be_shortbuffered;
7667 if (shortbuffered) { /* oh well, must extend */
7668 cnt = shortbuffered;
7670 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
7672 SvGROW(sv, SvLEN(sv) + append + cnt + 2);
7673 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
7677 cannot_be_shortbuffered:
7678 DEBUG_P(PerlIO_printf(Perl_debug_log,
7679 "Screamer: going to getc, ptr=%"UVuf", cnt=%ld\n",
7680 PTR2UV(ptr),(long)cnt));
7681 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* deregisterize cnt and ptr */
7683 DEBUG_Pv(PerlIO_printf(Perl_debug_log,
7684 "Screamer: pre: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7685 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7686 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7688 /* This used to call 'filbuf' in stdio form, but as that behaves like
7689 getc when cnt <= 0 we use PerlIO_getc here to avoid introducing
7690 another abstraction. */
7691 i = PerlIO_getc(fp); /* get more characters */
7693 DEBUG_Pv(PerlIO_printf(Perl_debug_log,
7694 "Screamer: post: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7695 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7696 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7698 cnt = PerlIO_get_cnt(fp);
7699 ptr = (STDCHAR*)PerlIO_get_ptr(fp); /* reregisterize cnt and ptr */
7700 DEBUG_P(PerlIO_printf(Perl_debug_log,
7701 "Screamer: after getc, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7703 if (i == EOF) /* all done for ever? */
7704 goto thats_really_all_folks;
7706 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
7708 SvGROW(sv, bpx + cnt + 2);
7709 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
7711 *bp++ = (STDCHAR)i; /* store character from PerlIO_getc */
7713 if (rslen && (STDCHAR)i == rslast) /* all done for now? */
7714 goto thats_all_folks;
7718 if ((rslen > 1 && (STRLEN)(bp - (STDCHAR*)SvPVX_const(sv)) < rslen) ||
7719 memNE((char*)bp - rslen, rsptr, rslen))
7720 goto screamer; /* go back to the fray */
7721 thats_really_all_folks:
7723 cnt += shortbuffered;
7724 DEBUG_P(PerlIO_printf(Perl_debug_log,
7725 "Screamer: quitting, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7726 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* put these back or we're in trouble */
7727 DEBUG_P(PerlIO_printf(Perl_debug_log,
7728 "Screamer: end: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7729 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7730 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7732 SvCUR_set(sv, bp - (STDCHAR*)SvPVX_const(sv)); /* set length */
7733 DEBUG_P(PerlIO_printf(Perl_debug_log,
7734 "Screamer: done, len=%ld, string=|%.*s|\n",
7735 (long)SvCUR(sv),(int)SvCUR(sv),SvPVX_const(sv)));
7739 /*The big, slow, and stupid way. */
7740 #ifdef USE_HEAP_INSTEAD_OF_STACK /* Even slower way. */
7741 STDCHAR *buf = NULL;
7742 Newx(buf, 8192, STDCHAR);
7750 register const STDCHAR * const bpe = buf + sizeof(buf);
7752 while ((i = PerlIO_getc(fp)) != EOF && (*bp++ = (STDCHAR)i) != rslast && bp < bpe)
7753 ; /* keep reading */
7757 cnt = PerlIO_read(fp,(char*)buf, sizeof(buf));
7758 /* Accommodate broken VAXC compiler, which applies U8 cast to
7759 * both args of ?: operator, causing EOF to change into 255
7762 i = (U8)buf[cnt - 1];
7768 cnt = 0; /* we do need to re-set the sv even when cnt <= 0 */
7770 sv_catpvn(sv, (char *) buf, cnt);
7772 sv_setpvn(sv, (char *) buf, cnt);
7774 if (i != EOF && /* joy */
7776 SvCUR(sv) < rslen ||
7777 memNE(SvPVX_const(sv) + SvCUR(sv) - rslen, rsptr, rslen)))
7781 * If we're reading from a TTY and we get a short read,
7782 * indicating that the user hit his EOF character, we need
7783 * to notice it now, because if we try to read from the TTY
7784 * again, the EOF condition will disappear.
7786 * The comparison of cnt to sizeof(buf) is an optimization
7787 * that prevents unnecessary calls to feof().
7791 if (!(cnt < (I32)sizeof(buf) && PerlIO_eof(fp)))
7795 #ifdef USE_HEAP_INSTEAD_OF_STACK
7800 if (rspara) { /* have to do this both before and after */
7801 while (i != EOF) { /* to make sure file boundaries work right */
7802 i = PerlIO_getc(fp);
7804 PerlIO_ungetc(fp,i);
7810 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7816 Auto-increment of the value in the SV, doing string to numeric conversion
7817 if necessary. Handles 'get' magic and operator overloading.
7823 Perl_sv_inc(pTHX_ register SV *const sv)
7832 =for apidoc sv_inc_nomg
7834 Auto-increment of the value in the SV, doing string to numeric conversion
7835 if necessary. Handles operator overloading. Skips handling 'get' magic.
7841 Perl_sv_inc_nomg(pTHX_ register SV *const sv)
7849 if (SvTHINKFIRST(sv)) {
7851 sv_force_normal_flags(sv, 0);
7852 if (SvREADONLY(sv)) {
7853 if (IN_PERL_RUNTIME)
7854 Perl_croak_no_modify(aTHX);
7858 if (SvAMAGIC(sv) && AMG_CALLunary(sv, inc_amg))
7860 i = PTR2IV(SvRV(sv));
7865 flags = SvFLAGS(sv);
7866 if ((flags & (SVp_NOK|SVp_IOK)) == SVp_NOK) {
7867 /* It's (privately or publicly) a float, but not tested as an
7868 integer, so test it to see. */
7870 flags = SvFLAGS(sv);
7872 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
7873 /* It's publicly an integer, or privately an integer-not-float */
7874 #ifdef PERL_PRESERVE_IVUV
7878 if (SvUVX(sv) == UV_MAX)
7879 sv_setnv(sv, UV_MAX_P1);
7881 (void)SvIOK_only_UV(sv);
7882 SvUV_set(sv, SvUVX(sv) + 1);
7884 if (SvIVX(sv) == IV_MAX)
7885 sv_setuv(sv, (UV)IV_MAX + 1);
7887 (void)SvIOK_only(sv);
7888 SvIV_set(sv, SvIVX(sv) + 1);
7893 if (flags & SVp_NOK) {
7894 const NV was = SvNVX(sv);
7895 if (NV_OVERFLOWS_INTEGERS_AT &&
7896 was >= NV_OVERFLOWS_INTEGERS_AT) {
7897 Perl_ck_warner(aTHX_ packWARN(WARN_IMPRECISION),
7898 "Lost precision when incrementing %" NVff " by 1",
7901 (void)SvNOK_only(sv);
7902 SvNV_set(sv, was + 1.0);
7906 if (!(flags & SVp_POK) || !*SvPVX_const(sv)) {
7907 if ((flags & SVTYPEMASK) < SVt_PVIV)
7908 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV ? SVt_PVIV : SVt_IV));
7909 (void)SvIOK_only(sv);
7914 while (isALPHA(*d)) d++;
7915 while (isDIGIT(*d)) d++;
7916 if (d < SvEND(sv)) {
7917 #ifdef PERL_PRESERVE_IVUV
7918 /* Got to punt this as an integer if needs be, but we don't issue
7919 warnings. Probably ought to make the sv_iv_please() that does
7920 the conversion if possible, and silently. */
7921 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
7922 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
7923 /* Need to try really hard to see if it's an integer.
7924 9.22337203685478e+18 is an integer.
7925 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
7926 so $a="9.22337203685478e+18"; $a+0; $a++
7927 needs to be the same as $a="9.22337203685478e+18"; $a++
7934 /* sv_2iv *should* have made this an NV */
7935 if (flags & SVp_NOK) {
7936 (void)SvNOK_only(sv);
7937 SvNV_set(sv, SvNVX(sv) + 1.0);
7940 /* I don't think we can get here. Maybe I should assert this
7941 And if we do get here I suspect that sv_setnv will croak. NWC
7943 #if defined(USE_LONG_DOUBLE)
7944 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",
7945 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7947 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_inc punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
7948 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7951 #endif /* PERL_PRESERVE_IVUV */
7952 sv_setnv(sv,Atof(SvPVX_const(sv)) + 1.0);
7956 while (d >= SvPVX_const(sv)) {
7964 /* MKS: The original code here died if letters weren't consecutive.
7965 * at least it didn't have to worry about non-C locales. The
7966 * new code assumes that ('z'-'a')==('Z'-'A'), letters are
7967 * arranged in order (although not consecutively) and that only
7968 * [A-Za-z] are accepted by isALPHA in the C locale.
7970 if (*d != 'z' && *d != 'Z') {
7971 do { ++*d; } while (!isALPHA(*d));
7974 *(d--) -= 'z' - 'a';
7979 *(d--) -= 'z' - 'a' + 1;
7983 /* oh,oh, the number grew */
7984 SvGROW(sv, SvCUR(sv) + 2);
7985 SvCUR_set(sv, SvCUR(sv) + 1);
7986 for (d = SvPVX(sv) + SvCUR(sv); d > SvPVX_const(sv); d--)
7997 Auto-decrement of the value in the SV, doing string to numeric conversion
7998 if necessary. Handles 'get' magic and operator overloading.
8004 Perl_sv_dec(pTHX_ register SV *const sv)
8014 =for apidoc sv_dec_nomg
8016 Auto-decrement of the value in the SV, doing string to numeric conversion
8017 if necessary. Handles operator overloading. Skips handling 'get' magic.
8023 Perl_sv_dec_nomg(pTHX_ register SV *const sv)
8030 if (SvTHINKFIRST(sv)) {
8032 sv_force_normal_flags(sv, 0);
8033 if (SvREADONLY(sv)) {
8034 if (IN_PERL_RUNTIME)
8035 Perl_croak_no_modify(aTHX);
8039 if (SvAMAGIC(sv) && AMG_CALLunary(sv, dec_amg))
8041 i = PTR2IV(SvRV(sv));
8046 /* Unlike sv_inc we don't have to worry about string-never-numbers
8047 and keeping them magic. But we mustn't warn on punting */
8048 flags = SvFLAGS(sv);
8049 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
8050 /* It's publicly an integer, or privately an integer-not-float */
8051 #ifdef PERL_PRESERVE_IVUV
8055 if (SvUVX(sv) == 0) {
8056 (void)SvIOK_only(sv);
8060 (void)SvIOK_only_UV(sv);
8061 SvUV_set(sv, SvUVX(sv) - 1);
8064 if (SvIVX(sv) == IV_MIN) {
8065 sv_setnv(sv, (NV)IV_MIN);
8069 (void)SvIOK_only(sv);
8070 SvIV_set(sv, SvIVX(sv) - 1);
8075 if (flags & SVp_NOK) {
8078 const NV was = SvNVX(sv);
8079 if (NV_OVERFLOWS_INTEGERS_AT &&
8080 was <= -NV_OVERFLOWS_INTEGERS_AT) {
8081 Perl_ck_warner(aTHX_ packWARN(WARN_IMPRECISION),
8082 "Lost precision when decrementing %" NVff " by 1",
8085 (void)SvNOK_only(sv);
8086 SvNV_set(sv, was - 1.0);
8090 if (!(flags & SVp_POK)) {
8091 if ((flags & SVTYPEMASK) < SVt_PVIV)
8092 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV) ? SVt_PVIV : SVt_IV);
8094 (void)SvIOK_only(sv);
8097 #ifdef PERL_PRESERVE_IVUV
8099 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
8100 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
8101 /* Need to try really hard to see if it's an integer.
8102 9.22337203685478e+18 is an integer.
8103 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
8104 so $a="9.22337203685478e+18"; $a+0; $a--
8105 needs to be the same as $a="9.22337203685478e+18"; $a--
8112 /* sv_2iv *should* have made this an NV */
8113 if (flags & SVp_NOK) {
8114 (void)SvNOK_only(sv);
8115 SvNV_set(sv, SvNVX(sv) - 1.0);
8118 /* I don't think we can get here. Maybe I should assert this
8119 And if we do get here I suspect that sv_setnv will croak. NWC
8121 #if defined(USE_LONG_DOUBLE)
8122 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",
8123 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
8125 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_dec punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
8126 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
8130 #endif /* PERL_PRESERVE_IVUV */
8131 sv_setnv(sv,Atof(SvPVX_const(sv)) - 1.0); /* punt */
8134 /* this define is used to eliminate a chunk of duplicated but shared logic
8135 * it has the suffix __SV_C to signal that it isnt API, and isnt meant to be
8136 * used anywhere but here - yves
8138 #define PUSH_EXTEND_MORTAL__SV_C(AnSv) \
8141 PL_tmps_stack[++PL_tmps_ix] = (AnSv); \
8145 =for apidoc sv_mortalcopy
8147 Creates a new SV which is a copy of the original SV (using C<sv_setsv>).
8148 The new SV is marked as mortal. It will be destroyed "soon", either by an
8149 explicit call to FREETMPS, or by an implicit call at places such as
8150 statement boundaries. See also C<sv_newmortal> and C<sv_2mortal>.
8155 /* Make a string that will exist for the duration of the expression
8156 * evaluation. Actually, it may have to last longer than that, but
8157 * hopefully we won't free it until it has been assigned to a
8158 * permanent location. */
8161 Perl_sv_mortalcopy(pTHX_ SV *const oldstr)
8167 sv_setsv(sv,oldstr);
8168 PUSH_EXTEND_MORTAL__SV_C(sv);
8174 =for apidoc sv_newmortal
8176 Creates a new null SV which is mortal. The reference count of the SV is
8177 set to 1. It will be destroyed "soon", either by an explicit call to
8178 FREETMPS, or by an implicit call at places such as statement boundaries.
8179 See also C<sv_mortalcopy> and C<sv_2mortal>.
8185 Perl_sv_newmortal(pTHX)
8191 SvFLAGS(sv) = SVs_TEMP;
8192 PUSH_EXTEND_MORTAL__SV_C(sv);
8198 =for apidoc newSVpvn_flags
8200 Creates a new SV and copies a string into it. The reference count for the
8201 SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
8202 string. You are responsible for ensuring that the source string is at least
8203 C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
8204 Currently the only flag bits accepted are C<SVf_UTF8> and C<SVs_TEMP>.
8205 If C<SVs_TEMP> is set, then C<sv_2mortal()> is called on the result before
8206 returning. If C<SVf_UTF8> is set, C<s> is considered to be in UTF-8 and the
8207 C<SVf_UTF8> flag will be set on the new SV.
8208 C<newSVpvn_utf8()> is a convenience wrapper for this function, defined as
8210 #define newSVpvn_utf8(s, len, u) \
8211 newSVpvn_flags((s), (len), (u) ? SVf_UTF8 : 0)
8217 Perl_newSVpvn_flags(pTHX_ const char *const s, const STRLEN len, const U32 flags)
8222 /* All the flags we don't support must be zero.
8223 And we're new code so I'm going to assert this from the start. */
8224 assert(!(flags & ~(SVf_UTF8|SVs_TEMP)));
8226 sv_setpvn(sv,s,len);
8228 /* This code used to a sv_2mortal(), however we now unroll the call to sv_2mortal()
8229 * and do what it does ourselves here.
8230 * Since we have asserted that flags can only have the SVf_UTF8 and/or SVs_TEMP flags
8231 * set above we can use it to enable the sv flags directly (bypassing SvTEMP_on), which
8232 * in turn means we dont need to mask out the SVf_UTF8 flag below, which means that we
8233 * eliminate quite a few steps than it looks - Yves (explaining patch by gfx)
8236 SvFLAGS(sv) |= flags;
8238 if(flags & SVs_TEMP){
8239 PUSH_EXTEND_MORTAL__SV_C(sv);
8246 =for apidoc sv_2mortal
8248 Marks an existing SV as mortal. The SV will be destroyed "soon", either
8249 by an explicit call to FREETMPS, or by an implicit call at places such as
8250 statement boundaries. SvTEMP() is turned on which means that the SV's
8251 string buffer can be "stolen" if this SV is copied. See also C<sv_newmortal>
8252 and C<sv_mortalcopy>.
8258 Perl_sv_2mortal(pTHX_ register SV *const sv)
8263 if (SvREADONLY(sv) && SvIMMORTAL(sv))
8265 PUSH_EXTEND_MORTAL__SV_C(sv);
8273 Creates a new SV and copies a string into it. The reference count for the
8274 SV is set to 1. If C<len> is zero, Perl will compute the length using
8275 strlen(). For efficiency, consider using C<newSVpvn> instead.
8281 Perl_newSVpv(pTHX_ const char *const s, const STRLEN len)
8287 sv_setpvn(sv, s, len || s == NULL ? len : strlen(s));
8292 =for apidoc newSVpvn
8294 Creates a new SV and copies a string into it. The reference count for the
8295 SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
8296 string. You are responsible for ensuring that the source string is at least
8297 C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
8303 Perl_newSVpvn(pTHX_ const char *const s, const STRLEN len)
8309 sv_setpvn(sv,s,len);
8314 =for apidoc newSVhek
8316 Creates a new SV from the hash key structure. It will generate scalars that
8317 point to the shared string table where possible. Returns a new (undefined)
8318 SV if the hek is NULL.
8324 Perl_newSVhek(pTHX_ const HEK *const hek)
8334 if (HEK_LEN(hek) == HEf_SVKEY) {
8335 return newSVsv(*(SV**)HEK_KEY(hek));
8337 const int flags = HEK_FLAGS(hek);
8338 if (flags & HVhek_WASUTF8) {
8340 Andreas would like keys he put in as utf8 to come back as utf8
8342 STRLEN utf8_len = HEK_LEN(hek);
8343 SV * const sv = newSV_type(SVt_PV);
8344 char *as_utf8 = (char *)bytes_to_utf8 ((U8*)HEK_KEY(hek), &utf8_len);
8345 /* bytes_to_utf8() allocates a new string, which we can repurpose: */
8346 sv_usepvn_flags(sv, as_utf8, utf8_len, SV_HAS_TRAILING_NUL);
8349 } else if (flags & (HVhek_REHASH|HVhek_UNSHARED)) {
8350 /* We don't have a pointer to the hv, so we have to replicate the
8351 flag into every HEK. This hv is using custom a hasing
8352 algorithm. Hence we can't return a shared string scalar, as
8353 that would contain the (wrong) hash value, and might get passed
8354 into an hv routine with a regular hash.
8355 Similarly, a hash that isn't using shared hash keys has to have
8356 the flag in every key so that we know not to try to call
8357 share_hek_kek on it. */
8359 SV * const sv = newSVpvn (HEK_KEY(hek), HEK_LEN(hek));
8364 /* This will be overwhelminly the most common case. */
8366 /* Inline most of newSVpvn_share(), because share_hek_hek() is far
8367 more efficient than sharepvn(). */
8371 sv_upgrade(sv, SVt_PV);
8372 SvPV_set(sv, (char *)HEK_KEY(share_hek_hek(hek)));
8373 SvCUR_set(sv, HEK_LEN(hek));
8386 =for apidoc newSVpvn_share
8388 Creates a new SV with its SvPVX_const pointing to a shared string in the string
8389 table. If the string does not already exist in the table, it is created
8390 first. Turns on READONLY and FAKE. If the C<hash> parameter is non-zero, that
8391 value is used; otherwise the hash is computed. The string's hash can be later
8392 be retrieved from the SV with the C<SvSHARED_HASH()> macro. The idea here is
8393 that as the string table is used for shared hash keys these strings will have
8394 SvPVX_const == HeKEY and hash lookup will avoid string compare.
8400 Perl_newSVpvn_share(pTHX_ const char *src, I32 len, U32 hash)
8404 bool is_utf8 = FALSE;
8405 const char *const orig_src = src;
8408 STRLEN tmplen = -len;
8410 /* See the note in hv.c:hv_fetch() --jhi */
8411 src = (char*)bytes_from_utf8((const U8*)src, &tmplen, &is_utf8);
8415 PERL_HASH(hash, src, len);
8417 /* The logic for this is inlined in S_mro_get_linear_isa_dfs(), so if it
8418 changes here, update it there too. */
8419 sv_upgrade(sv, SVt_PV);
8420 SvPV_set(sv, sharepvn(src, is_utf8?-len:len, hash));
8428 if (src != orig_src)
8434 =for apidoc newSVpv_share
8436 Like C<newSVpvn_share>, but takes a nul-terminated string instead of a
8443 Perl_newSVpv_share(pTHX_ const char *src, U32 hash)
8445 return newSVpvn_share(src, strlen(src), hash);
8448 #if defined(PERL_IMPLICIT_CONTEXT)
8450 /* pTHX_ magic can't cope with varargs, so this is a no-context
8451 * version of the main function, (which may itself be aliased to us).
8452 * Don't access this version directly.
8456 Perl_newSVpvf_nocontext(const char *const pat, ...)
8462 PERL_ARGS_ASSERT_NEWSVPVF_NOCONTEXT;
8464 va_start(args, pat);
8465 sv = vnewSVpvf(pat, &args);
8472 =for apidoc newSVpvf
8474 Creates a new SV and initializes it with the string formatted like
8481 Perl_newSVpvf(pTHX_ const char *const pat, ...)
8486 PERL_ARGS_ASSERT_NEWSVPVF;
8488 va_start(args, pat);
8489 sv = vnewSVpvf(pat, &args);
8494 /* backend for newSVpvf() and newSVpvf_nocontext() */
8497 Perl_vnewSVpvf(pTHX_ const char *const pat, va_list *const args)
8502 PERL_ARGS_ASSERT_VNEWSVPVF;
8505 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
8512 Creates a new SV and copies a floating point value into it.
8513 The reference count for the SV is set to 1.
8519 Perl_newSVnv(pTHX_ const NV n)
8532 Creates a new SV and copies an integer into it. The reference count for the
8539 Perl_newSViv(pTHX_ const IV i)
8552 Creates a new SV and copies an unsigned integer into it.
8553 The reference count for the SV is set to 1.
8559 Perl_newSVuv(pTHX_ const UV u)
8570 =for apidoc newSV_type
8572 Creates a new SV, of the type specified. The reference count for the new SV
8579 Perl_newSV_type(pTHX_ const svtype type)
8584 sv_upgrade(sv, type);
8589 =for apidoc newRV_noinc
8591 Creates an RV wrapper for an SV. The reference count for the original
8592 SV is B<not> incremented.
8598 Perl_newRV_noinc(pTHX_ SV *const tmpRef)
8601 register SV *sv = newSV_type(SVt_IV);
8603 PERL_ARGS_ASSERT_NEWRV_NOINC;
8606 SvRV_set(sv, tmpRef);
8611 /* newRV_inc is the official function name to use now.
8612 * newRV_inc is in fact #defined to newRV in sv.h
8616 Perl_newRV(pTHX_ SV *const sv)
8620 PERL_ARGS_ASSERT_NEWRV;
8622 return newRV_noinc(SvREFCNT_inc_simple_NN(sv));
8628 Creates a new SV which is an exact duplicate of the original SV.
8635 Perl_newSVsv(pTHX_ register SV *const old)
8642 if (SvTYPE(old) == SVTYPEMASK) {
8643 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL), "semi-panic: attempt to dup freed string");
8647 /* SV_GMAGIC is the default for sv_setv()
8648 SV_NOSTEAL prevents TEMP buffers being, well, stolen, and saves games
8649 with SvTEMP_off and SvTEMP_on round a call to sv_setsv. */
8650 sv_setsv_flags(sv, old, SV_GMAGIC | SV_NOSTEAL);
8655 =for apidoc sv_reset
8657 Underlying implementation for the C<reset> Perl function.
8658 Note that the perl-level function is vaguely deprecated.
8664 Perl_sv_reset(pTHX_ register const char *s, HV *const stash)
8667 char todo[PERL_UCHAR_MAX+1];
8669 PERL_ARGS_ASSERT_SV_RESET;
8674 if (!*s) { /* reset ?? searches */
8675 MAGIC * const mg = mg_find((const SV *)stash, PERL_MAGIC_symtab);
8677 const U32 count = mg->mg_len / sizeof(PMOP**);
8678 PMOP **pmp = (PMOP**) mg->mg_ptr;
8679 PMOP *const *const end = pmp + count;
8683 SvREADONLY_off(PL_regex_pad[(*pmp)->op_pmoffset]);
8685 (*pmp)->op_pmflags &= ~PMf_USED;
8693 /* reset variables */
8695 if (!HvARRAY(stash))
8698 Zero(todo, 256, char);
8701 I32 i = (unsigned char)*s;
8705 max = (unsigned char)*s++;
8706 for ( ; i <= max; i++) {
8709 for (i = 0; i <= (I32) HvMAX(stash); i++) {
8711 for (entry = HvARRAY(stash)[i];
8713 entry = HeNEXT(entry))
8718 if (!todo[(U8)*HeKEY(entry)])
8720 gv = MUTABLE_GV(HeVAL(entry));
8723 if (SvTHINKFIRST(sv)) {
8724 if (!SvREADONLY(sv) && SvROK(sv))
8726 /* XXX Is this continue a bug? Why should THINKFIRST
8727 exempt us from resetting arrays and hashes? */
8731 if (SvTYPE(sv) >= SVt_PV) {
8733 if (SvPVX_const(sv) != NULL)
8741 if (GvHV(gv) && !HvNAME_get(GvHV(gv))) {
8743 Perl_die(aTHX_ "Can't reset %%ENV on this system");
8746 # if defined(USE_ENVIRON_ARRAY)
8749 # endif /* USE_ENVIRON_ARRAY */
8760 Using various gambits, try to get an IO from an SV: the IO slot if its a
8761 GV; or the recursive result if we're an RV; or the IO slot of the symbol
8762 named after the PV if we're a string.
8768 Perl_sv_2io(pTHX_ SV *const sv)
8773 PERL_ARGS_ASSERT_SV_2IO;
8775 switch (SvTYPE(sv)) {
8777 io = MUTABLE_IO(sv);
8781 if (isGV_with_GP(sv)) {
8782 gv = MUTABLE_GV(sv);
8785 Perl_croak(aTHX_ "Bad filehandle: %s", GvNAME(gv));
8791 Perl_croak(aTHX_ PL_no_usym, "filehandle");
8793 return sv_2io(SvRV(sv));
8794 gv = gv_fetchsv(sv, 0, SVt_PVIO);
8800 Perl_croak(aTHX_ "Bad filehandle: %"SVf, SVfARG(sv));
8809 Using various gambits, try to get a CV from an SV; in addition, try if
8810 possible to set C<*st> and C<*gvp> to the stash and GV associated with it.
8811 The flags in C<lref> are passed to gv_fetchsv.
8817 Perl_sv_2cv(pTHX_ SV *sv, HV **const st, GV **const gvp, const I32 lref)
8823 PERL_ARGS_ASSERT_SV_2CV;
8830 switch (SvTYPE(sv)) {
8834 return MUTABLE_CV(sv);
8841 if (isGV_with_GP(sv)) {
8842 gv = MUTABLE_GV(sv);
8853 sv = amagic_deref_call(sv, to_cv_amg);
8854 /* At this point I'd like to do SPAGAIN, but really I need to
8855 force it upon my callers. Hmmm. This is a mess... */
8858 if (SvTYPE(sv) == SVt_PVCV) {
8859 cv = MUTABLE_CV(sv);
8864 else if(isGV_with_GP(sv))
8865 gv = MUTABLE_GV(sv);
8867 Perl_croak(aTHX_ "Not a subroutine reference");
8869 else if (isGV_with_GP(sv)) {
8871 gv = MUTABLE_GV(sv);
8874 gv = gv_fetchsv(sv, lref, SVt_PVCV); /* Calls get magic */
8880 /* Some flags to gv_fetchsv mean don't really create the GV */
8881 if (!isGV_with_GP(gv)) {
8887 if (lref && !GvCVu(gv)) {
8891 gv_efullname3(tmpsv, gv, NULL);
8892 /* XXX this is probably not what they think they're getting.
8893 * It has the same effect as "sub name;", i.e. just a forward
8895 newSUB(start_subparse(FALSE, 0),
8896 newSVOP(OP_CONST, 0, tmpsv),
8900 Perl_croak(aTHX_ "Unable to create sub named \"%"SVf"\"",
8901 SVfARG(SvOK(sv) ? sv : &PL_sv_no));
8910 Returns true if the SV has a true value by Perl's rules.
8911 Use the C<SvTRUE> macro instead, which may call C<sv_true()> or may
8912 instead use an in-line version.
8918 Perl_sv_true(pTHX_ register SV *const sv)
8923 register const XPV* const tXpv = (XPV*)SvANY(sv);
8925 (tXpv->xpv_cur > 1 ||
8926 (tXpv->xpv_cur && *sv->sv_u.svu_pv != '0')))
8933 return SvIVX(sv) != 0;
8936 return SvNVX(sv) != 0.0;
8938 return sv_2bool(sv);
8944 =for apidoc sv_pvn_force
8946 Get a sensible string out of the SV somehow.
8947 A private implementation of the C<SvPV_force> macro for compilers which
8948 can't cope with complex macro expressions. Always use the macro instead.
8950 =for apidoc sv_pvn_force_flags
8952 Get a sensible string out of the SV somehow.
8953 If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<sv> if
8954 appropriate, else not. C<sv_pvn_force> and C<sv_pvn_force_nomg> are
8955 implemented in terms of this function.
8956 You normally want to use the various wrapper macros instead: see
8957 C<SvPV_force> and C<SvPV_force_nomg>
8963 Perl_sv_pvn_force_flags(pTHX_ SV *const sv, STRLEN *const lp, const I32 flags)
8967 PERL_ARGS_ASSERT_SV_PVN_FORCE_FLAGS;
8969 if (SvTHINKFIRST(sv) && !SvROK(sv))
8970 sv_force_normal_flags(sv, 0);
8980 if (SvREADONLY(sv) && !(flags & SV_MUTABLE_RETURN)) {
8981 const char * const ref = sv_reftype(sv,0);
8983 Perl_croak(aTHX_ "Can't coerce readonly %s to string in %s",
8984 ref, OP_DESC(PL_op));
8986 Perl_croak(aTHX_ "Can't coerce readonly %s to string", ref);
8988 if ((SvTYPE(sv) > SVt_PVLV && SvTYPE(sv) != SVt_PVFM)
8989 || isGV_with_GP(sv))
8990 /* diag_listed_as: Can't coerce %s to %s in %s */
8991 Perl_croak(aTHX_ "Can't coerce %s to string in %s", sv_reftype(sv,0),
8993 s = sv_2pv_flags(sv, &len, flags);
8997 if (s != SvPVX_const(sv)) { /* Almost, but not quite, sv_setpvn() */
9000 SvUPGRADE(sv, SVt_PV); /* Never FALSE */
9001 SvGROW(sv, len + 1);
9002 Move(s,SvPVX(sv),len,char);
9004 SvPVX(sv)[len] = '\0';
9007 SvPOK_on(sv); /* validate pointer */
9009 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
9010 PTR2UV(sv),SvPVX_const(sv)));
9013 return SvPVX_mutable(sv);
9017 =for apidoc sv_pvbyten_force
9019 The backend for the C<SvPVbytex_force> macro. Always use the macro instead.
9025 Perl_sv_pvbyten_force(pTHX_ SV *const sv, STRLEN *const lp)
9027 PERL_ARGS_ASSERT_SV_PVBYTEN_FORCE;
9029 sv_pvn_force(sv,lp);
9030 sv_utf8_downgrade(sv,0);
9036 =for apidoc sv_pvutf8n_force
9038 The backend for the C<SvPVutf8x_force> macro. Always use the macro instead.
9044 Perl_sv_pvutf8n_force(pTHX_ SV *const sv, STRLEN *const lp)
9046 PERL_ARGS_ASSERT_SV_PVUTF8N_FORCE;
9048 sv_pvn_force(sv,lp);
9049 sv_utf8_upgrade(sv);
9055 =for apidoc sv_reftype
9057 Returns a string describing what the SV is a reference to.
9063 Perl_sv_reftype(pTHX_ const SV *const sv, const int ob)
9065 PERL_ARGS_ASSERT_SV_REFTYPE;
9067 /* The fact that I don't need to downcast to char * everywhere, only in ?:
9068 inside return suggests a const propagation bug in g++. */
9069 if (ob && SvOBJECT(sv)) {
9070 char * const name = HvNAME_get(SvSTASH(sv));
9071 return name ? name : (char *) "__ANON__";
9074 switch (SvTYPE(sv)) {
9089 case SVt_PVLV: return (char *) (SvROK(sv) ? "REF"
9090 /* tied lvalues should appear to be
9091 * scalars for backwards compatibility */
9092 : (LvTYPE(sv) == 't' || LvTYPE(sv) == 'T')
9093 ? "SCALAR" : "LVALUE");
9094 case SVt_PVAV: return "ARRAY";
9095 case SVt_PVHV: return "HASH";
9096 case SVt_PVCV: return "CODE";
9097 case SVt_PVGV: return (char *) (isGV_with_GP(sv)
9098 ? "GLOB" : "SCALAR");
9099 case SVt_PVFM: return "FORMAT";
9100 case SVt_PVIO: return "IO";
9101 case SVt_BIND: return "BIND";
9102 case SVt_REGEXP: return "REGEXP";
9103 default: return "UNKNOWN";
9109 =for apidoc sv_isobject
9111 Returns a boolean indicating whether the SV is an RV pointing to a blessed
9112 object. If the SV is not an RV, or if the object is not blessed, then this
9119 Perl_sv_isobject(pTHX_ SV *sv)
9135 Returns a boolean indicating whether the SV is blessed into the specified
9136 class. This does not check for subtypes; use C<sv_derived_from> to verify
9137 an inheritance relationship.
9143 Perl_sv_isa(pTHX_ SV *sv, const char *const name)
9147 PERL_ARGS_ASSERT_SV_ISA;
9157 hvname = HvNAME_get(SvSTASH(sv));
9161 return strEQ(hvname, name);
9167 Creates a new SV for the RV, C<rv>, to point to. If C<rv> is not an RV then
9168 it will be upgraded to one. If C<classname> is non-null then the new SV will
9169 be blessed in the specified package. The new SV is returned and its
9170 reference count is 1.
9176 Perl_newSVrv(pTHX_ SV *const rv, const char *const classname)
9181 PERL_ARGS_ASSERT_NEWSVRV;
9185 SV_CHECK_THINKFIRST_COW_DROP(rv);
9186 (void)SvAMAGIC_off(rv);
9188 if (SvTYPE(rv) >= SVt_PVMG) {
9189 const U32 refcnt = SvREFCNT(rv);
9193 SvREFCNT(rv) = refcnt;
9195 sv_upgrade(rv, SVt_IV);
9196 } else if (SvROK(rv)) {
9197 SvREFCNT_dec(SvRV(rv));
9199 prepare_SV_for_RV(rv);
9207 HV* const stash = gv_stashpv(classname, GV_ADD);
9208 (void)sv_bless(rv, stash);
9214 =for apidoc sv_setref_pv
9216 Copies a pointer into a new SV, optionally blessing the SV. The C<rv>
9217 argument will be upgraded to an RV. That RV will be modified to point to
9218 the new SV. If the C<pv> argument is NULL then C<PL_sv_undef> will be placed
9219 into the SV. The C<classname> argument indicates the package for the
9220 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9221 will have a reference count of 1, and the RV will be returned.
9223 Do not use with other Perl types such as HV, AV, SV, CV, because those
9224 objects will become corrupted by the pointer copy process.
9226 Note that C<sv_setref_pvn> copies the string while this copies the pointer.
9232 Perl_sv_setref_pv(pTHX_ SV *const rv, const char *const classname, void *const pv)
9236 PERL_ARGS_ASSERT_SV_SETREF_PV;
9239 sv_setsv(rv, &PL_sv_undef);
9243 sv_setiv(newSVrv(rv,classname), PTR2IV(pv));
9248 =for apidoc sv_setref_iv
9250 Copies an integer into a new SV, optionally blessing the SV. The C<rv>
9251 argument will be upgraded to an RV. That RV will be modified to point to
9252 the new SV. The C<classname> argument indicates the package for the
9253 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9254 will have a reference count of 1, and the RV will be returned.
9260 Perl_sv_setref_iv(pTHX_ SV *const rv, const char *const classname, const IV iv)
9262 PERL_ARGS_ASSERT_SV_SETREF_IV;
9264 sv_setiv(newSVrv(rv,classname), iv);
9269 =for apidoc sv_setref_uv
9271 Copies an unsigned integer into a new SV, optionally blessing the SV. The C<rv>
9272 argument will be upgraded to an RV. That RV will be modified to point to
9273 the new SV. The C<classname> argument indicates the package for the
9274 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9275 will have a reference count of 1, and the RV will be returned.
9281 Perl_sv_setref_uv(pTHX_ SV *const rv, const char *const classname, const UV uv)
9283 PERL_ARGS_ASSERT_SV_SETREF_UV;
9285 sv_setuv(newSVrv(rv,classname), uv);
9290 =for apidoc sv_setref_nv
9292 Copies a double into a new SV, optionally blessing the SV. The C<rv>
9293 argument will be upgraded to an RV. That RV will be modified to point to
9294 the new SV. The C<classname> argument indicates the package for the
9295 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9296 will have a reference count of 1, and the RV will be returned.
9302 Perl_sv_setref_nv(pTHX_ SV *const rv, const char *const classname, const NV nv)
9304 PERL_ARGS_ASSERT_SV_SETREF_NV;
9306 sv_setnv(newSVrv(rv,classname), nv);
9311 =for apidoc sv_setref_pvn
9313 Copies a string into a new SV, optionally blessing the SV. The length of the
9314 string must be specified with C<n>. The C<rv> argument will be upgraded to
9315 an RV. That RV will be modified to point to the new SV. The C<classname>
9316 argument indicates the package for the blessing. Set C<classname> to
9317 C<NULL> to avoid the blessing. The new SV will have a reference count
9318 of 1, and the RV will be returned.
9320 Note that C<sv_setref_pv> copies the pointer while this copies the string.
9326 Perl_sv_setref_pvn(pTHX_ SV *const rv, const char *const classname,
9327 const char *const pv, const STRLEN n)
9329 PERL_ARGS_ASSERT_SV_SETREF_PVN;
9331 sv_setpvn(newSVrv(rv,classname), pv, n);
9336 =for apidoc sv_bless
9338 Blesses an SV into a specified package. The SV must be an RV. The package
9339 must be designated by its stash (see C<gv_stashpv()>). The reference count
9340 of the SV is unaffected.
9346 Perl_sv_bless(pTHX_ SV *const sv, HV *const stash)
9351 PERL_ARGS_ASSERT_SV_BLESS;
9354 Perl_croak(aTHX_ "Can't bless non-reference value");
9356 if (SvFLAGS(tmpRef) & (SVs_OBJECT|SVf_READONLY)) {
9357 if (SvIsCOW(tmpRef))
9358 sv_force_normal_flags(tmpRef, 0);
9359 if (SvREADONLY(tmpRef))
9360 Perl_croak_no_modify(aTHX);
9361 if (SvOBJECT(tmpRef)) {
9362 if (SvTYPE(tmpRef) != SVt_PVIO)
9364 SvREFCNT_dec(SvSTASH(tmpRef));
9367 SvOBJECT_on(tmpRef);
9368 if (SvTYPE(tmpRef) != SVt_PVIO)
9370 SvUPGRADE(tmpRef, SVt_PVMG);
9371 SvSTASH_set(tmpRef, MUTABLE_HV(SvREFCNT_inc_simple(stash)));
9376 (void)SvAMAGIC_off(sv);
9378 if(SvSMAGICAL(tmpRef))
9379 if(mg_find(tmpRef, PERL_MAGIC_ext) || mg_find(tmpRef, PERL_MAGIC_uvar))
9387 /* Downgrades a PVGV to a PVMG. If it’s actually a PVLV, we leave the type
9388 * as it is after unglobbing it.
9392 S_sv_unglob(pTHX_ SV *const sv)
9397 SV * const temp = sv_newmortal();
9399 PERL_ARGS_ASSERT_SV_UNGLOB;
9401 assert(SvTYPE(sv) == SVt_PVGV || SvTYPE(sv) == SVt_PVLV);
9403 gv_efullname3(temp, MUTABLE_GV(sv), "*");
9406 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
9407 && HvNAME_get(stash))
9408 mro_method_changed_in(stash);
9409 gp_free(MUTABLE_GV(sv));
9412 sv_del_backref(MUTABLE_SV(GvSTASH(sv)), sv);
9416 if (GvNAME_HEK(sv)) {
9417 unshare_hek(GvNAME_HEK(sv));
9419 isGV_with_GP_off(sv);
9421 if(SvTYPE(sv) == SVt_PVGV) {
9422 /* need to keep SvANY(sv) in the right arena */
9423 xpvmg = new_XPVMG();
9424 StructCopy(SvANY(sv), xpvmg, XPVMG);
9425 del_XPVGV(SvANY(sv));
9428 SvFLAGS(sv) &= ~SVTYPEMASK;
9429 SvFLAGS(sv) |= SVt_PVMG;
9432 /* Intentionally not calling any local SET magic, as this isn't so much a
9433 set operation as merely an internal storage change. */
9434 sv_setsv_flags(sv, temp, 0);
9438 =for apidoc sv_unref_flags
9440 Unsets the RV status of the SV, and decrements the reference count of
9441 whatever was being referenced by the RV. This can almost be thought of
9442 as a reversal of C<newSVrv>. The C<cflags> argument can contain
9443 C<SV_IMMEDIATE_UNREF> to force the reference count to be decremented
9444 (otherwise the decrementing is conditional on the reference count being
9445 different from one or the reference being a readonly SV).
9452 Perl_sv_unref_flags(pTHX_ SV *const ref, const U32 flags)
9454 SV* const target = SvRV(ref);
9456 PERL_ARGS_ASSERT_SV_UNREF_FLAGS;
9458 if (SvWEAKREF(ref)) {
9459 sv_del_backref(target, ref);
9461 SvRV_set(ref, NULL);
9464 SvRV_set(ref, NULL);
9466 /* You can't have a || SvREADONLY(target) here, as $a = $$a, where $a was
9467 assigned to as BEGIN {$a = \"Foo"} will fail. */
9468 if (SvREFCNT(target) != 1 || (flags & SV_IMMEDIATE_UNREF))
9469 SvREFCNT_dec(target);
9470 else /* XXX Hack, but hard to make $a=$a->[1] work otherwise */
9471 sv_2mortal(target); /* Schedule for freeing later */
9475 =for apidoc sv_untaint
9477 Untaint an SV. Use C<SvTAINTED_off> instead.
9483 Perl_sv_untaint(pTHX_ SV *const sv)
9485 PERL_ARGS_ASSERT_SV_UNTAINT;
9487 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
9488 MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
9495 =for apidoc sv_tainted
9497 Test an SV for taintedness. Use C<SvTAINTED> instead.
9503 Perl_sv_tainted(pTHX_ SV *const sv)
9505 PERL_ARGS_ASSERT_SV_TAINTED;
9507 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
9508 const MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
9509 if (mg && (mg->mg_len & 1) )
9516 =for apidoc sv_setpviv
9518 Copies an integer into the given SV, also updating its string value.
9519 Does not handle 'set' magic. See C<sv_setpviv_mg>.
9525 Perl_sv_setpviv(pTHX_ SV *const sv, const IV iv)
9527 char buf[TYPE_CHARS(UV)];
9529 char * const ptr = uiv_2buf(buf, iv, 0, 0, &ebuf);
9531 PERL_ARGS_ASSERT_SV_SETPVIV;
9533 sv_setpvn(sv, ptr, ebuf - ptr);
9537 =for apidoc sv_setpviv_mg
9539 Like C<sv_setpviv>, but also handles 'set' magic.
9545 Perl_sv_setpviv_mg(pTHX_ SV *const sv, const IV iv)
9547 PERL_ARGS_ASSERT_SV_SETPVIV_MG;
9553 #if defined(PERL_IMPLICIT_CONTEXT)
9555 /* pTHX_ magic can't cope with varargs, so this is a no-context
9556 * version of the main function, (which may itself be aliased to us).
9557 * Don't access this version directly.
9561 Perl_sv_setpvf_nocontext(SV *const sv, const char *const pat, ...)
9566 PERL_ARGS_ASSERT_SV_SETPVF_NOCONTEXT;
9568 va_start(args, pat);
9569 sv_vsetpvf(sv, pat, &args);
9573 /* pTHX_ magic can't cope with varargs, so this is a no-context
9574 * version of the main function, (which may itself be aliased to us).
9575 * Don't access this version directly.
9579 Perl_sv_setpvf_mg_nocontext(SV *const sv, const char *const pat, ...)
9584 PERL_ARGS_ASSERT_SV_SETPVF_MG_NOCONTEXT;
9586 va_start(args, pat);
9587 sv_vsetpvf_mg(sv, pat, &args);
9593 =for apidoc sv_setpvf
9595 Works like C<sv_catpvf> but copies the text into the SV instead of
9596 appending it. Does not handle 'set' magic. See C<sv_setpvf_mg>.
9602 Perl_sv_setpvf(pTHX_ SV *const sv, const char *const pat, ...)
9606 PERL_ARGS_ASSERT_SV_SETPVF;
9608 va_start(args, pat);
9609 sv_vsetpvf(sv, pat, &args);
9614 =for apidoc sv_vsetpvf
9616 Works like C<sv_vcatpvf> but copies the text into the SV instead of
9617 appending it. Does not handle 'set' magic. See C<sv_vsetpvf_mg>.
9619 Usually used via its frontend C<sv_setpvf>.
9625 Perl_sv_vsetpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9627 PERL_ARGS_ASSERT_SV_VSETPVF;
9629 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9633 =for apidoc sv_setpvf_mg
9635 Like C<sv_setpvf>, but also handles 'set' magic.
9641 Perl_sv_setpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
9645 PERL_ARGS_ASSERT_SV_SETPVF_MG;
9647 va_start(args, pat);
9648 sv_vsetpvf_mg(sv, pat, &args);
9653 =for apidoc sv_vsetpvf_mg
9655 Like C<sv_vsetpvf>, but also handles 'set' magic.
9657 Usually used via its frontend C<sv_setpvf_mg>.
9663 Perl_sv_vsetpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9665 PERL_ARGS_ASSERT_SV_VSETPVF_MG;
9667 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9671 #if defined(PERL_IMPLICIT_CONTEXT)
9673 /* pTHX_ magic can't cope with varargs, so this is a no-context
9674 * version of the main function, (which may itself be aliased to us).
9675 * Don't access this version directly.
9679 Perl_sv_catpvf_nocontext(SV *const sv, const char *const pat, ...)
9684 PERL_ARGS_ASSERT_SV_CATPVF_NOCONTEXT;
9686 va_start(args, pat);
9687 sv_vcatpvf(sv, pat, &args);
9691 /* pTHX_ magic can't cope with varargs, so this is a no-context
9692 * version of the main function, (which may itself be aliased to us).
9693 * Don't access this version directly.
9697 Perl_sv_catpvf_mg_nocontext(SV *const sv, const char *const pat, ...)
9702 PERL_ARGS_ASSERT_SV_CATPVF_MG_NOCONTEXT;
9704 va_start(args, pat);
9705 sv_vcatpvf_mg(sv, pat, &args);
9711 =for apidoc sv_catpvf
9713 Processes its arguments like C<sprintf> and appends the formatted
9714 output to an SV. If the appended data contains "wide" characters
9715 (including, but not limited to, SVs with a UTF-8 PV formatted with %s,
9716 and characters >255 formatted with %c), the original SV might get
9717 upgraded to UTF-8. Handles 'get' magic, but not 'set' magic. See
9718 C<sv_catpvf_mg>. If the original SV was UTF-8, the pattern should be
9719 valid UTF-8; if the original SV was bytes, the pattern should be too.
9724 Perl_sv_catpvf(pTHX_ SV *const sv, const char *const pat, ...)
9728 PERL_ARGS_ASSERT_SV_CATPVF;
9730 va_start(args, pat);
9731 sv_vcatpvf(sv, pat, &args);
9736 =for apidoc sv_vcatpvf
9738 Processes its arguments like C<vsprintf> and appends the formatted output
9739 to an SV. Does not handle 'set' magic. See C<sv_vcatpvf_mg>.
9741 Usually used via its frontend C<sv_catpvf>.
9747 Perl_sv_vcatpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9749 PERL_ARGS_ASSERT_SV_VCATPVF;
9751 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9755 =for apidoc sv_catpvf_mg
9757 Like C<sv_catpvf>, but also handles 'set' magic.
9763 Perl_sv_catpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
9767 PERL_ARGS_ASSERT_SV_CATPVF_MG;
9769 va_start(args, pat);
9770 sv_vcatpvf_mg(sv, pat, &args);
9775 =for apidoc sv_vcatpvf_mg
9777 Like C<sv_vcatpvf>, but also handles 'set' magic.
9779 Usually used via its frontend C<sv_catpvf_mg>.
9785 Perl_sv_vcatpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9787 PERL_ARGS_ASSERT_SV_VCATPVF_MG;
9789 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9794 =for apidoc sv_vsetpvfn
9796 Works like C<sv_vcatpvfn> but copies the text into the SV instead of
9799 Usually used via one of its frontends C<sv_vsetpvf> and C<sv_vsetpvf_mg>.
9805 Perl_sv_vsetpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
9806 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
9808 PERL_ARGS_ASSERT_SV_VSETPVFN;
9811 sv_vcatpvfn(sv, pat, patlen, args, svargs, svmax, maybe_tainted);
9816 * Warn of missing argument to sprintf, and then return a defined value
9817 * to avoid inappropriate "use of uninit" warnings [perl #71000].
9819 #define WARN_MISSING WARN_UNINITIALIZED /* Not sure we want a new category */
9821 S_vcatpvfn_missing_argument(pTHX) {
9822 if (ckWARN(WARN_MISSING)) {
9823 Perl_warner(aTHX_ packWARN(WARN_MISSING), "Missing argument in %s",
9824 PL_op ? OP_DESC(PL_op) : "sv_vcatpvfn()");
9831 S_expect_number(pTHX_ char **const pattern)
9836 PERL_ARGS_ASSERT_EXPECT_NUMBER;
9838 switch (**pattern) {
9839 case '1': case '2': case '3':
9840 case '4': case '5': case '6':
9841 case '7': case '8': case '9':
9842 var = *(*pattern)++ - '0';
9843 while (isDIGIT(**pattern)) {
9844 const I32 tmp = var * 10 + (*(*pattern)++ - '0');
9846 Perl_croak(aTHX_ "Integer overflow in format string for %s", (PL_op ? OP_DESC(PL_op) : "sv_vcatpvfn"));
9854 S_F0convert(NV nv, char *const endbuf, STRLEN *const len)
9856 const int neg = nv < 0;
9859 PERL_ARGS_ASSERT_F0CONVERT;
9867 if (uv & 1 && uv == nv)
9868 uv--; /* Round to even */
9870 const unsigned dig = uv % 10;
9883 =for apidoc sv_vcatpvfn
9885 Processes its arguments like C<vsprintf> and appends the formatted output
9886 to an SV. Uses an array of SVs if the C style variable argument list is
9887 missing (NULL). When running with taint checks enabled, indicates via
9888 C<maybe_tainted> if results are untrustworthy (often due to the use of
9891 Usually used via one of its frontends C<sv_vcatpvf> and C<sv_vcatpvf_mg>.
9897 #define VECTORIZE_ARGS vecsv = va_arg(*args, SV*);\
9898 vecstr = (U8*)SvPV_const(vecsv,veclen);\
9899 vec_utf8 = DO_UTF8(vecsv);
9901 /* XXX maybe_tainted is never assigned to, so the doc above is lying. */
9904 Perl_sv_vcatpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
9905 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
9913 static const char nullstr[] = "(null)";
9915 bool has_utf8 = DO_UTF8(sv); /* has the result utf8? */
9916 const bool pat_utf8 = has_utf8; /* the pattern is in utf8? */
9918 /* Times 4: a decimal digit takes more than 3 binary digits.
9919 * NV_DIG: mantissa takes than many decimal digits.
9920 * Plus 32: Playing safe. */
9921 char ebuf[IV_DIG * 4 + NV_DIG + 32];
9922 /* large enough for "%#.#f" --chip */
9923 /* what about long double NVs? --jhi */
9925 PERL_ARGS_ASSERT_SV_VCATPVFN;
9926 PERL_UNUSED_ARG(maybe_tainted);
9928 /* no matter what, this is a string now */
9929 (void)SvPV_force(sv, origlen);
9931 /* special-case "", "%s", and "%-p" (SVf - see below) */
9934 if (patlen == 2 && pat[0] == '%' && pat[1] == 's') {
9936 const char * const s = va_arg(*args, char*);
9937 sv_catpv(sv, s ? s : nullstr);
9939 else if (svix < svmax) {
9940 sv_catsv(sv, *svargs);
9943 S_vcatpvfn_missing_argument(aTHX);
9946 if (args && patlen == 3 && pat[0] == '%' &&
9947 pat[1] == '-' && pat[2] == 'p') {
9948 argsv = MUTABLE_SV(va_arg(*args, void*));
9949 sv_catsv(sv, argsv);
9953 #ifndef USE_LONG_DOUBLE
9954 /* special-case "%.<number>[gf]" */
9955 if ( !args && patlen <= 5 && pat[0] == '%' && pat[1] == '.'
9956 && (pat[patlen-1] == 'g' || pat[patlen-1] == 'f') ) {
9957 unsigned digits = 0;
9961 while (*pp >= '0' && *pp <= '9')
9962 digits = 10 * digits + (*pp++ - '0');
9963 if (pp - pat == (int)patlen - 1 && svix < svmax) {
9964 const NV nv = SvNV(*svargs);
9966 /* Add check for digits != 0 because it seems that some
9967 gconverts are buggy in this case, and we don't yet have
9968 a Configure test for this. */
9969 if (digits && digits < sizeof(ebuf) - NV_DIG - 10) {
9970 /* 0, point, slack */
9971 Gconvert(nv, (int)digits, 0, ebuf);
9973 if (*ebuf) /* May return an empty string for digits==0 */
9976 } else if (!digits) {
9979 if ((p = F0convert(nv, ebuf + sizeof ebuf, &l))) {
9980 sv_catpvn(sv, p, l);
9986 #endif /* !USE_LONG_DOUBLE */
9988 if (!args && svix < svmax && DO_UTF8(*svargs))
9991 patend = (char*)pat + patlen;
9992 for (p = (char*)pat; p < patend; p = q) {
9995 bool vectorize = FALSE;
9996 bool vectorarg = FALSE;
9997 bool vec_utf8 = FALSE;
10003 bool has_precis = FALSE;
10005 const I32 osvix = svix;
10006 bool is_utf8 = FALSE; /* is this item utf8? */
10007 #ifdef HAS_LDBL_SPRINTF_BUG
10008 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
10009 with sfio - Allen <allens@cpan.org> */
10010 bool fix_ldbl_sprintf_bug = FALSE;
10014 U8 utf8buf[UTF8_MAXBYTES+1];
10015 STRLEN esignlen = 0;
10017 const char *eptr = NULL;
10018 const char *fmtstart;
10021 const U8 *vecstr = NULL;
10028 /* we need a long double target in case HAS_LONG_DOUBLE but
10029 not USE_LONG_DOUBLE
10031 #if defined(HAS_LONG_DOUBLE) && LONG_DOUBLESIZE > DOUBLESIZE
10039 const char *dotstr = ".";
10040 STRLEN dotstrlen = 1;
10041 I32 efix = 0; /* explicit format parameter index */
10042 I32 ewix = 0; /* explicit width index */
10043 I32 epix = 0; /* explicit precision index */
10044 I32 evix = 0; /* explicit vector index */
10045 bool asterisk = FALSE;
10047 /* echo everything up to the next format specification */
10048 for (q = p; q < patend && *q != '%'; ++q) ;
10050 if (has_utf8 && !pat_utf8)
10051 sv_catpvn_utf8_upgrade(sv, p, q - p, nsv);
10053 sv_catpvn(sv, p, q - p);
10062 We allow format specification elements in this order:
10063 \d+\$ explicit format parameter index
10065 v|\*(\d+\$)?v vector with optional (optionally specified) arg
10066 0 flag (as above): repeated to allow "v02"
10067 \d+|\*(\d+\$)? width using optional (optionally specified) arg
10068 \.(\d*|\*(\d+\$)?) precision using optional (optionally specified) arg
10070 [%bcdefginopsuxDFOUX] format (mandatory)
10075 As of perl5.9.3, printf format checking is on by default.
10076 Internally, perl uses %p formats to provide an escape to
10077 some extended formatting. This block deals with those
10078 extensions: if it does not match, (char*)q is reset and
10079 the normal format processing code is used.
10081 Currently defined extensions are:
10082 %p include pointer address (standard)
10083 %-p (SVf) include an SV (previously %_)
10084 %-<num>p include an SV with precision <num>
10085 %<num>p reserved for future extensions
10087 Robin Barker 2005-07-14
10089 %1p (VDf) removed. RMB 2007-10-19
10096 n = expect_number(&q);
10098 if (sv) { /* SVf */
10103 argsv = MUTABLE_SV(va_arg(*args, void*));
10104 eptr = SvPV_const(argsv, elen);
10105 if (DO_UTF8(argsv))
10110 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL),
10111 "internal %%<num>p might conflict with future printf extensions");
10117 if ( (width = expect_number(&q)) ) {
10132 if (plus == '+' && *q == ' ') /* '+' over ' ' */
10161 if ( (ewix = expect_number(&q)) )
10170 if ((vectorarg = asterisk)) {
10183 width = expect_number(&q);
10186 if (vectorize && vectorarg) {
10187 /* vectorizing, but not with the default "." */
10189 vecsv = va_arg(*args, SV*);
10191 vecsv = (evix > 0 && evix <= svmax)
10192 ? svargs[evix-1] : S_vcatpvfn_missing_argument(aTHX);
10194 vecsv = svix < svmax
10195 ? svargs[svix++] : S_vcatpvfn_missing_argument(aTHX);
10197 dotstr = SvPV_const(vecsv, dotstrlen);
10198 /* Keep the DO_UTF8 test *after* the SvPV call, else things go
10199 bad with tied or overloaded values that return UTF8. */
10200 if (DO_UTF8(vecsv))
10202 else if (has_utf8) {
10203 vecsv = sv_mortalcopy(vecsv);
10204 sv_utf8_upgrade(vecsv);
10205 dotstr = SvPV_const(vecsv, dotstrlen);
10212 i = va_arg(*args, int);
10214 i = (ewix ? ewix <= svmax : svix < svmax) ?
10215 SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
10217 width = (i < 0) ? -i : i;
10227 if ( ((epix = expect_number(&q))) && (*q++ != '$') )
10229 /* XXX: todo, support specified precision parameter */
10233 i = va_arg(*args, int);
10235 i = (ewix ? ewix <= svmax : svix < svmax)
10236 ? SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
10238 has_precis = !(i < 0);
10242 while (isDIGIT(*q))
10243 precis = precis * 10 + (*q++ - '0');
10252 else if (efix ? (efix > 0 && efix <= svmax) : svix < svmax) {
10253 vecsv = svargs[efix ? efix-1 : svix++];
10254 vecstr = (U8*)SvPV_const(vecsv,veclen);
10255 vec_utf8 = DO_UTF8(vecsv);
10257 /* if this is a version object, we need to convert
10258 * back into v-string notation and then let the
10259 * vectorize happen normally
10261 if (sv_derived_from(vecsv, "version")) {
10262 char *version = savesvpv(vecsv);
10263 if ( hv_exists(MUTABLE_HV(SvRV(vecsv)), "alpha", 5 ) ) {
10264 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
10265 "vector argument not supported with alpha versions");
10268 vecsv = sv_newmortal();
10269 scan_vstring(version, version + veclen, vecsv);
10270 vecstr = (U8*)SvPV_const(vecsv, veclen);
10271 vec_utf8 = DO_UTF8(vecsv);
10285 case 'I': /* Ix, I32x, and I64x */
10287 if (q[1] == '6' && q[2] == '4') {
10293 if (q[1] == '3' && q[2] == '2') {
10303 #if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
10315 #if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
10316 if (*q == 'l') { /* lld, llf */
10325 if (*++q == 'h') { /* hhd, hhu */
10354 if (!vectorize && !args) {
10356 const I32 i = efix-1;
10357 argsv = (i >= 0 && i < svmax)
10358 ? svargs[i] : S_vcatpvfn_missing_argument(aTHX);
10360 argsv = (svix >= 0 && svix < svmax)
10361 ? svargs[svix++] : S_vcatpvfn_missing_argument(aTHX);
10365 switch (c = *q++) {
10372 uv = (args) ? va_arg(*args, int) : SvIV(argsv);
10374 (!UNI_IS_INVARIANT(uv) && SvUTF8(sv)))
10376 eptr = (char*)utf8buf;
10377 elen = uvchr_to_utf8((U8*)eptr, uv) - utf8buf;
10391 eptr = va_arg(*args, char*);
10393 elen = strlen(eptr);
10395 eptr = (char *)nullstr;
10396 elen = sizeof nullstr - 1;
10400 eptr = SvPV_const(argsv, elen);
10401 if (DO_UTF8(argsv)) {
10402 STRLEN old_precis = precis;
10403 if (has_precis && precis < elen) {
10404 STRLEN ulen = sv_len_utf8(argsv);
10405 I32 p = precis > ulen ? ulen : precis;
10406 sv_pos_u2b(argsv, &p, 0); /* sticks at end */
10409 if (width) { /* fudge width (can't fudge elen) */
10410 if (has_precis && precis < elen)
10411 width += precis - old_precis;
10413 width += elen - sv_len_utf8(argsv);
10420 if (has_precis && precis < elen)
10427 if (alt || vectorize)
10429 uv = PTR2UV(args ? va_arg(*args, void*) : argsv);
10450 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
10459 esignbuf[esignlen++] = plus;
10463 case 'c': iv = (char)va_arg(*args, int); break;
10464 case 'h': iv = (short)va_arg(*args, int); break;
10465 case 'l': iv = va_arg(*args, long); break;
10466 case 'V': iv = va_arg(*args, IV); break;
10467 case 'z': iv = va_arg(*args, SSize_t); break;
10468 case 't': iv = va_arg(*args, ptrdiff_t); break;
10469 default: iv = va_arg(*args, int); break;
10471 case 'j': iv = va_arg(*args, intmax_t); break;
10475 iv = va_arg(*args, Quad_t); break;
10482 IV tiv = SvIV(argsv); /* work around GCC bug #13488 */
10484 case 'c': iv = (char)tiv; break;
10485 case 'h': iv = (short)tiv; break;
10486 case 'l': iv = (long)tiv; break;
10488 default: iv = tiv; break;
10491 iv = (Quad_t)tiv; break;
10497 if ( !vectorize ) /* we already set uv above */
10502 esignbuf[esignlen++] = plus;
10506 esignbuf[esignlen++] = '-';
10550 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
10561 case 'c': uv = (unsigned char)va_arg(*args, unsigned); break;
10562 case 'h': uv = (unsigned short)va_arg(*args, unsigned); break;
10563 case 'l': uv = va_arg(*args, unsigned long); break;
10564 case 'V': uv = va_arg(*args, UV); break;
10565 case 'z': uv = va_arg(*args, Size_t); break;
10566 case 't': uv = va_arg(*args, ptrdiff_t); break; /* will sign extend, but there is no uptrdiff_t, so oh well */
10568 case 'j': uv = va_arg(*args, uintmax_t); break;
10570 default: uv = va_arg(*args, unsigned); break;
10573 uv = va_arg(*args, Uquad_t); break;
10580 UV tuv = SvUV(argsv); /* work around GCC bug #13488 */
10582 case 'c': uv = (unsigned char)tuv; break;
10583 case 'h': uv = (unsigned short)tuv; break;
10584 case 'l': uv = (unsigned long)tuv; break;
10586 default: uv = tuv; break;
10589 uv = (Uquad_t)tuv; break;
10598 char *ptr = ebuf + sizeof ebuf;
10599 bool tempalt = uv ? alt : FALSE; /* Vectors can't change alt */
10605 p = (char *)((c == 'X') ? PL_hexdigit + 16 : PL_hexdigit);
10609 } while (uv >>= 4);
10611 esignbuf[esignlen++] = '0';
10612 esignbuf[esignlen++] = c; /* 'x' or 'X' */
10618 *--ptr = '0' + dig;
10619 } while (uv >>= 3);
10620 if (alt && *ptr != '0')
10626 *--ptr = '0' + dig;
10627 } while (uv >>= 1);
10629 esignbuf[esignlen++] = '0';
10630 esignbuf[esignlen++] = c;
10633 default: /* it had better be ten or less */
10636 *--ptr = '0' + dig;
10637 } while (uv /= base);
10640 elen = (ebuf + sizeof ebuf) - ptr;
10644 zeros = precis - elen;
10645 else if (precis == 0 && elen == 1 && *eptr == '0'
10646 && !(base == 8 && alt)) /* "%#.0o" prints "0" */
10649 /* a precision nullifies the 0 flag. */
10656 /* FLOATING POINT */
10659 c = 'f'; /* maybe %F isn't supported here */
10661 case 'e': case 'E':
10663 case 'g': case 'G':
10667 /* This is evil, but floating point is even more evil */
10669 /* for SV-style calling, we can only get NV
10670 for C-style calling, we assume %f is double;
10671 for simplicity we allow any of %Lf, %llf, %qf for long double
10675 #if defined(USE_LONG_DOUBLE)
10679 /* [perl #20339] - we should accept and ignore %lf rather than die */
10683 #if defined(USE_LONG_DOUBLE)
10684 intsize = args ? 0 : 'q';
10688 #if defined(HAS_LONG_DOUBLE)
10701 /* now we need (long double) if intsize == 'q', else (double) */
10703 #if LONG_DOUBLESIZE > DOUBLESIZE
10705 va_arg(*args, long double) :
10706 va_arg(*args, double)
10708 va_arg(*args, double)
10713 /* nv * 0 will be NaN for NaN, +Inf and -Inf, and 0 for anything
10714 else. frexp() has some unspecified behaviour for those three */
10715 if (c != 'e' && c != 'E' && (nv * 0) == 0) {
10717 /* FIXME: if HAS_LONG_DOUBLE but not USE_LONG_DOUBLE this
10718 will cast our (long double) to (double) */
10719 (void)Perl_frexp(nv, &i);
10720 if (i == PERL_INT_MIN)
10721 Perl_die(aTHX_ "panic: frexp");
10723 need = BIT_DIGITS(i);
10725 need += has_precis ? precis : 6; /* known default */
10730 #ifdef HAS_LDBL_SPRINTF_BUG
10731 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
10732 with sfio - Allen <allens@cpan.org> */
10735 # define MY_DBL_MAX DBL_MAX
10736 # else /* XXX guessing! HUGE_VAL may be defined as infinity, so not using */
10737 # if DOUBLESIZE >= 8
10738 # define MY_DBL_MAX 1.7976931348623157E+308L
10740 # define MY_DBL_MAX 3.40282347E+38L
10744 # ifdef HAS_LDBL_SPRINTF_BUG_LESS1 /* only between -1L & 1L - Allen */
10745 # define MY_DBL_MAX_BUG 1L
10747 # define MY_DBL_MAX_BUG MY_DBL_MAX
10751 # define MY_DBL_MIN DBL_MIN
10752 # else /* XXX guessing! -Allen */
10753 # if DOUBLESIZE >= 8
10754 # define MY_DBL_MIN 2.2250738585072014E-308L
10756 # define MY_DBL_MIN 1.17549435E-38L
10760 if ((intsize == 'q') && (c == 'f') &&
10761 ((nv < MY_DBL_MAX_BUG) && (nv > -MY_DBL_MAX_BUG)) &&
10762 (need < DBL_DIG)) {
10763 /* it's going to be short enough that
10764 * long double precision is not needed */
10766 if ((nv <= 0L) && (nv >= -0L))
10767 fix_ldbl_sprintf_bug = TRUE; /* 0 is 0 - easiest */
10769 /* would use Perl_fp_class as a double-check but not
10770 * functional on IRIX - see perl.h comments */
10772 if ((nv >= MY_DBL_MIN) || (nv <= -MY_DBL_MIN)) {
10773 /* It's within the range that a double can represent */
10774 #if defined(DBL_MAX) && !defined(DBL_MIN)
10775 if ((nv >= ((long double)1/DBL_MAX)) ||
10776 (nv <= (-(long double)1/DBL_MAX)))
10778 fix_ldbl_sprintf_bug = TRUE;
10781 if (fix_ldbl_sprintf_bug == TRUE) {
10791 # undef MY_DBL_MAX_BUG
10794 #endif /* HAS_LDBL_SPRINTF_BUG */
10796 need += 20; /* fudge factor */
10797 if (PL_efloatsize < need) {
10798 Safefree(PL_efloatbuf);
10799 PL_efloatsize = need + 20; /* more fudge */
10800 Newx(PL_efloatbuf, PL_efloatsize, char);
10801 PL_efloatbuf[0] = '\0';
10804 if ( !(width || left || plus || alt) && fill != '0'
10805 && has_precis && intsize != 'q' ) { /* Shortcuts */
10806 /* See earlier comment about buggy Gconvert when digits,
10808 if ( c == 'g' && precis) {
10809 Gconvert((NV)nv, (int)precis, 0, PL_efloatbuf);
10810 /* May return an empty string for digits==0 */
10811 if (*PL_efloatbuf) {
10812 elen = strlen(PL_efloatbuf);
10813 goto float_converted;
10815 } else if ( c == 'f' && !precis) {
10816 if ((eptr = F0convert(nv, ebuf + sizeof ebuf, &elen)))
10821 char *ptr = ebuf + sizeof ebuf;
10824 /* FIXME: what to do if HAS_LONG_DOUBLE but not PERL_PRIfldbl? */
10825 #if defined(HAS_LONG_DOUBLE) && defined(PERL_PRIfldbl)
10826 if (intsize == 'q') {
10827 /* Copy the one or more characters in a long double
10828 * format before the 'base' ([efgEFG]) character to
10829 * the format string. */
10830 static char const prifldbl[] = PERL_PRIfldbl;
10831 char const *p = prifldbl + sizeof(prifldbl) - 3;
10832 while (p >= prifldbl) { *--ptr = *p--; }
10837 do { *--ptr = '0' + (base % 10); } while (base /= 10);
10842 do { *--ptr = '0' + (base % 10); } while (base /= 10);
10854 /* No taint. Otherwise we are in the strange situation
10855 * where printf() taints but print($float) doesn't.
10857 #if defined(HAS_LONG_DOUBLE)
10858 elen = ((intsize == 'q')
10859 ? my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, nv)
10860 : my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, (double)nv));
10862 elen = my_sprintf(PL_efloatbuf, ptr, nv);
10866 eptr = PL_efloatbuf;
10874 i = SvCUR(sv) - origlen;
10877 case 'c': *(va_arg(*args, char*)) = i; break;
10878 case 'h': *(va_arg(*args, short*)) = i; break;
10879 default: *(va_arg(*args, int*)) = i; break;
10880 case 'l': *(va_arg(*args, long*)) = i; break;
10881 case 'V': *(va_arg(*args, IV*)) = i; break;
10882 case 'z': *(va_arg(*args, SSize_t*)) = i; break;
10883 case 't': *(va_arg(*args, ptrdiff_t*)) = i; break;
10885 case 'j': *(va_arg(*args, intmax_t*)) = i; break;
10889 *(va_arg(*args, Quad_t*)) = i; break;
10896 sv_setuv_mg(argsv, (UV)i);
10897 continue; /* not "break" */
10904 && (PL_op->op_type == OP_PRTF || PL_op->op_type == OP_SPRINTF)
10905 && ckWARN(WARN_PRINTF))
10907 SV * const msg = sv_newmortal();
10908 Perl_sv_setpvf(aTHX_ msg, "Invalid conversion in %sprintf: ",
10909 (PL_op->op_type == OP_PRTF) ? "" : "s");
10910 if (fmtstart < patend) {
10911 const char * const fmtend = q < patend ? q : patend;
10913 sv_catpvs(msg, "\"%");
10914 for (f = fmtstart; f < fmtend; f++) {
10916 sv_catpvn(msg, f, 1);
10918 Perl_sv_catpvf(aTHX_ msg,
10919 "\\%03"UVof, (UV)*f & 0xFF);
10922 sv_catpvs(msg, "\"");
10924 sv_catpvs(msg, "end of string");
10926 Perl_warner(aTHX_ packWARN(WARN_PRINTF), "%"SVf, SVfARG(msg)); /* yes, this is reentrant */
10929 /* output mangled stuff ... */
10935 /* ... right here, because formatting flags should not apply */
10936 SvGROW(sv, SvCUR(sv) + elen + 1);
10938 Copy(eptr, p, elen, char);
10941 SvCUR_set(sv, p - SvPVX_const(sv));
10943 continue; /* not "break" */
10946 if (is_utf8 != has_utf8) {
10949 sv_utf8_upgrade(sv);
10952 const STRLEN old_elen = elen;
10953 SV * const nsv = newSVpvn_flags(eptr, elen, SVs_TEMP);
10954 sv_utf8_upgrade(nsv);
10955 eptr = SvPVX_const(nsv);
10958 if (width) { /* fudge width (can't fudge elen) */
10959 width += elen - old_elen;
10965 have = esignlen + zeros + elen;
10967 Perl_croak_nocontext("%s", PL_memory_wrap);
10969 need = (have > width ? have : width);
10972 if (need >= (((STRLEN)~0) - SvCUR(sv) - dotstrlen - 1))
10973 Perl_croak_nocontext("%s", PL_memory_wrap);
10974 SvGROW(sv, SvCUR(sv) + need + dotstrlen + 1);
10976 if (esignlen && fill == '0') {
10978 for (i = 0; i < (int)esignlen; i++)
10979 *p++ = esignbuf[i];
10981 if (gap && !left) {
10982 memset(p, fill, gap);
10985 if (esignlen && fill != '0') {
10987 for (i = 0; i < (int)esignlen; i++)
10988 *p++ = esignbuf[i];
10992 for (i = zeros; i; i--)
10996 Copy(eptr, p, elen, char);
11000 memset(p, ' ', gap);
11005 Copy(dotstr, p, dotstrlen, char);
11009 vectorize = FALSE; /* done iterating over vecstr */
11016 SvCUR_set(sv, p - SvPVX_const(sv));
11025 /* =========================================================================
11027 =head1 Cloning an interpreter
11029 All the macros and functions in this section are for the private use of
11030 the main function, perl_clone().
11032 The foo_dup() functions make an exact copy of an existing foo thingy.
11033 During the course of a cloning, a hash table is used to map old addresses
11034 to new addresses. The table is created and manipulated with the
11035 ptr_table_* functions.
11039 * =========================================================================*/
11042 #if defined(USE_ITHREADS)
11044 /* XXX Remove this so it doesn't have to go thru the macro and return for nothing */
11045 #ifndef GpREFCNT_inc
11046 # define GpREFCNT_inc(gp) ((gp) ? (++(gp)->gp_refcnt, (gp)) : (GP*)NULL)
11050 /* Certain cases in Perl_ss_dup have been merged, by relying on the fact
11051 that currently av_dup, gv_dup and hv_dup are the same as sv_dup.
11052 If this changes, please unmerge ss_dup.
11053 Likewise, sv_dup_inc_multiple() relies on this fact. */
11054 #define sv_dup_inc_NN(s,t) SvREFCNT_inc_NN(sv_dup_inc(s,t))
11055 #define av_dup(s,t) MUTABLE_AV(sv_dup((const SV *)s,t))
11056 #define av_dup_inc(s,t) MUTABLE_AV(sv_dup_inc((const SV *)s,t))
11057 #define hv_dup(s,t) MUTABLE_HV(sv_dup((const SV *)s,t))
11058 #define hv_dup_inc(s,t) MUTABLE_HV(sv_dup_inc((const SV *)s,t))
11059 #define cv_dup(s,t) MUTABLE_CV(sv_dup((const SV *)s,t))
11060 #define cv_dup_inc(s,t) MUTABLE_CV(sv_dup_inc((const SV *)s,t))
11061 #define io_dup(s,t) MUTABLE_IO(sv_dup((const SV *)s,t))
11062 #define io_dup_inc(s,t) MUTABLE_IO(sv_dup_inc((const SV *)s,t))
11063 #define gv_dup(s,t) MUTABLE_GV(sv_dup((const SV *)s,t))
11064 #define gv_dup_inc(s,t) MUTABLE_GV(sv_dup_inc((const SV *)s,t))
11065 #define SAVEPV(p) ((p) ? savepv(p) : NULL)
11066 #define SAVEPVN(p,n) ((p) ? savepvn(p,n) : NULL)
11068 /* clone a parser */
11071 Perl_parser_dup(pTHX_ const yy_parser *const proto, CLONE_PARAMS *const param)
11075 PERL_ARGS_ASSERT_PARSER_DUP;
11080 /* look for it in the table first */
11081 parser = (yy_parser *)ptr_table_fetch(PL_ptr_table, proto);
11085 /* create anew and remember what it is */
11086 Newxz(parser, 1, yy_parser);
11087 ptr_table_store(PL_ptr_table, proto, parser);
11089 /* XXX these not yet duped */
11090 parser->old_parser = NULL;
11091 parser->stack = NULL;
11093 parser->stack_size = 0;
11094 /* XXX parser->stack->state = 0; */
11096 /* XXX eventually, just Copy() most of the parser struct ? */
11098 parser->lex_brackets = proto->lex_brackets;
11099 parser->lex_casemods = proto->lex_casemods;
11100 parser->lex_brackstack = savepvn(proto->lex_brackstack,
11101 (proto->lex_brackets < 120 ? 120 : proto->lex_brackets));
11102 parser->lex_casestack = savepvn(proto->lex_casestack,
11103 (proto->lex_casemods < 12 ? 12 : proto->lex_casemods));
11104 parser->lex_defer = proto->lex_defer;
11105 parser->lex_dojoin = proto->lex_dojoin;
11106 parser->lex_expect = proto->lex_expect;
11107 parser->lex_formbrack = proto->lex_formbrack;
11108 parser->lex_inpat = proto->lex_inpat;
11109 parser->lex_inwhat = proto->lex_inwhat;
11110 parser->lex_op = proto->lex_op;
11111 parser->lex_repl = sv_dup_inc(proto->lex_repl, param);
11112 parser->lex_starts = proto->lex_starts;
11113 parser->lex_stuff = sv_dup_inc(proto->lex_stuff, param);
11114 parser->multi_close = proto->multi_close;
11115 parser->multi_open = proto->multi_open;
11116 parser->multi_start = proto->multi_start;
11117 parser->multi_end = proto->multi_end;
11118 parser->pending_ident = proto->pending_ident;
11119 parser->preambled = proto->preambled;
11120 parser->sublex_info = proto->sublex_info; /* XXX not quite right */
11121 parser->linestr = sv_dup_inc(proto->linestr, param);
11122 parser->expect = proto->expect;
11123 parser->copline = proto->copline;
11124 parser->last_lop_op = proto->last_lop_op;
11125 parser->lex_state = proto->lex_state;
11126 parser->rsfp = fp_dup(proto->rsfp, '<', param);
11127 /* rsfp_filters entries have fake IoDIRP() */
11128 parser->rsfp_filters= av_dup_inc(proto->rsfp_filters, param);
11129 parser->in_my = proto->in_my;
11130 parser->in_my_stash = hv_dup(proto->in_my_stash, param);
11131 parser->error_count = proto->error_count;
11134 parser->linestr = sv_dup_inc(proto->linestr, param);
11137 char * const ols = SvPVX(proto->linestr);
11138 char * const ls = SvPVX(parser->linestr);
11140 parser->bufptr = ls + (proto->bufptr >= ols ?
11141 proto->bufptr - ols : 0);
11142 parser->oldbufptr = ls + (proto->oldbufptr >= ols ?
11143 proto->oldbufptr - ols : 0);
11144 parser->oldoldbufptr= ls + (proto->oldoldbufptr >= ols ?
11145 proto->oldoldbufptr - ols : 0);
11146 parser->linestart = ls + (proto->linestart >= ols ?
11147 proto->linestart - ols : 0);
11148 parser->last_uni = ls + (proto->last_uni >= ols ?
11149 proto->last_uni - ols : 0);
11150 parser->last_lop = ls + (proto->last_lop >= ols ?
11151 proto->last_lop - ols : 0);
11153 parser->bufend = ls + SvCUR(parser->linestr);
11156 Copy(proto->tokenbuf, parser->tokenbuf, 256, char);
11160 parser->endwhite = proto->endwhite;
11161 parser->faketokens = proto->faketokens;
11162 parser->lasttoke = proto->lasttoke;
11163 parser->nextwhite = proto->nextwhite;
11164 parser->realtokenstart = proto->realtokenstart;
11165 parser->skipwhite = proto->skipwhite;
11166 parser->thisclose = proto->thisclose;
11167 parser->thismad = proto->thismad;
11168 parser->thisopen = proto->thisopen;
11169 parser->thisstuff = proto->thisstuff;
11170 parser->thistoken = proto->thistoken;
11171 parser->thiswhite = proto->thiswhite;
11173 Copy(proto->nexttoke, parser->nexttoke, 5, NEXTTOKE);
11174 parser->curforce = proto->curforce;
11176 Copy(proto->nextval, parser->nextval, 5, YYSTYPE);
11177 Copy(proto->nexttype, parser->nexttype, 5, I32);
11178 parser->nexttoke = proto->nexttoke;
11181 /* XXX should clone saved_curcop here, but we aren't passed
11182 * proto_perl; so do it in perl_clone_using instead */
11188 /* duplicate a file handle */
11191 Perl_fp_dup(pTHX_ PerlIO *const fp, const char type, CLONE_PARAMS *const param)
11195 PERL_ARGS_ASSERT_FP_DUP;
11196 PERL_UNUSED_ARG(type);
11199 return (PerlIO*)NULL;
11201 /* look for it in the table first */
11202 ret = (PerlIO*)ptr_table_fetch(PL_ptr_table, fp);
11206 /* create anew and remember what it is */
11207 ret = PerlIO_fdupopen(aTHX_ fp, param, PERLIO_DUP_CLONE);
11208 ptr_table_store(PL_ptr_table, fp, ret);
11212 /* duplicate a directory handle */
11215 Perl_dirp_dup(pTHX_ DIR *const dp, CLONE_PARAMS *const param)
11221 register const Direntry_t *dirent;
11222 char smallbuf[256];
11228 PERL_UNUSED_CONTEXT;
11229 PERL_ARGS_ASSERT_DIRP_DUP;
11234 /* look for it in the table first */
11235 ret = (DIR*)ptr_table_fetch(PL_ptr_table, dp);
11241 PERL_UNUSED_ARG(param);
11245 /* open the current directory (so we can switch back) */
11246 if (!(pwd = PerlDir_open("."))) return (DIR *)NULL;
11248 /* chdir to our dir handle and open the present working directory */
11249 if (fchdir(my_dirfd(dp)) < 0 || !(ret = PerlDir_open("."))) {
11250 PerlDir_close(pwd);
11251 return (DIR *)NULL;
11253 /* Now we should have two dir handles pointing to the same dir. */
11255 /* Be nice to the calling code and chdir back to where we were. */
11256 fchdir(my_dirfd(pwd)); /* If this fails, then what? */
11258 /* We have no need of the pwd handle any more. */
11259 PerlDir_close(pwd);
11262 # define d_namlen(d) (d)->d_namlen
11264 # define d_namlen(d) strlen((d)->d_name)
11266 /* Iterate once through dp, to get the file name at the current posi-
11267 tion. Then step back. */
11268 pos = PerlDir_tell(dp);
11269 if ((dirent = PerlDir_read(dp))) {
11270 len = d_namlen(dirent);
11271 if (len <= sizeof smallbuf) name = smallbuf;
11272 else Newx(name, len, char);
11273 Move(dirent->d_name, name, len, char);
11275 PerlDir_seek(dp, pos);
11277 /* Iterate through the new dir handle, till we find a file with the
11279 if (!dirent) /* just before the end */
11281 pos = PerlDir_tell(ret);
11282 if (PerlDir_read(ret)) continue; /* not there yet */
11283 PerlDir_seek(ret, pos); /* step back */
11287 const long pos0 = PerlDir_tell(ret);
11289 pos = PerlDir_tell(ret);
11290 if ((dirent = PerlDir_read(ret))) {
11291 if (len == d_namlen(dirent)
11292 && memEQ(name, dirent->d_name, len)) {
11294 PerlDir_seek(ret, pos); /* step back */
11297 /* else we are not there yet; keep iterating */
11299 else { /* This is not meant to happen. The best we can do is
11300 reset the iterator to the beginning. */
11301 PerlDir_seek(ret, pos0);
11308 if (name && name != smallbuf)
11313 ret = win32_dirp_dup(dp, param);
11316 /* pop it in the pointer table */
11318 ptr_table_store(PL_ptr_table, dp, ret);
11323 /* duplicate a typeglob */
11326 Perl_gp_dup(pTHX_ GP *const gp, CLONE_PARAMS *const param)
11330 PERL_ARGS_ASSERT_GP_DUP;
11334 /* look for it in the table first */
11335 ret = (GP*)ptr_table_fetch(PL_ptr_table, gp);
11339 /* create anew and remember what it is */
11341 ptr_table_store(PL_ptr_table, gp, ret);
11344 /* ret->gp_refcnt must be 0 before any other dups are called. We're relying
11345 on Newxz() to do this for us. */
11346 ret->gp_sv = sv_dup_inc(gp->gp_sv, param);
11347 ret->gp_io = io_dup_inc(gp->gp_io, param);
11348 ret->gp_form = cv_dup_inc(gp->gp_form, param);
11349 ret->gp_av = av_dup_inc(gp->gp_av, param);
11350 ret->gp_hv = hv_dup_inc(gp->gp_hv, param);
11351 ret->gp_egv = gv_dup(gp->gp_egv, param);/* GvEGV is not refcounted */
11352 ret->gp_cv = cv_dup_inc(gp->gp_cv, param);
11353 ret->gp_cvgen = gp->gp_cvgen;
11354 ret->gp_line = gp->gp_line;
11355 ret->gp_file_hek = hek_dup(gp->gp_file_hek, param);
11359 /* duplicate a chain of magic */
11362 Perl_mg_dup(pTHX_ MAGIC *mg, CLONE_PARAMS *const param)
11364 MAGIC *mgret = NULL;
11365 MAGIC **mgprev_p = &mgret;
11367 PERL_ARGS_ASSERT_MG_DUP;
11369 for (; mg; mg = mg->mg_moremagic) {
11372 if ((param->flags & CLONEf_JOIN_IN)
11373 && mg->mg_type == PERL_MAGIC_backref)
11374 /* when joining, we let the individual SVs add themselves to
11375 * backref as needed. */
11378 Newx(nmg, 1, MAGIC);
11380 mgprev_p = &(nmg->mg_moremagic);
11382 /* There was a comment "XXX copy dynamic vtable?" but as we don't have
11383 dynamic vtables, I'm not sure why Sarathy wrote it. The comment dates
11384 from the original commit adding Perl_mg_dup() - revision 4538.
11385 Similarly there is the annotation "XXX random ptr?" next to the
11386 assignment to nmg->mg_ptr. */
11389 /* FIXME for plugins
11390 if (nmg->mg_type == PERL_MAGIC_qr) {
11391 nmg->mg_obj = MUTABLE_SV(CALLREGDUPE((REGEXP*)nmg->mg_obj, param));
11395 nmg->mg_obj = (nmg->mg_flags & MGf_REFCOUNTED)
11396 ? nmg->mg_type == PERL_MAGIC_backref
11397 /* The backref AV has its reference
11398 * count deliberately bumped by 1 */
11399 ? SvREFCNT_inc(av_dup_inc((const AV *)
11400 nmg->mg_obj, param))
11401 : sv_dup_inc(nmg->mg_obj, param)
11402 : sv_dup(nmg->mg_obj, param);
11404 if (nmg->mg_ptr && nmg->mg_type != PERL_MAGIC_regex_global) {
11405 if (nmg->mg_len > 0) {
11406 nmg->mg_ptr = SAVEPVN(nmg->mg_ptr, nmg->mg_len);
11407 if (nmg->mg_type == PERL_MAGIC_overload_table &&
11408 AMT_AMAGIC((AMT*)nmg->mg_ptr))
11410 AMT * const namtp = (AMT*)nmg->mg_ptr;
11411 sv_dup_inc_multiple((SV**)(namtp->table),
11412 (SV**)(namtp->table), NofAMmeth, param);
11415 else if (nmg->mg_len == HEf_SVKEY)
11416 nmg->mg_ptr = (char*)sv_dup_inc((const SV *)nmg->mg_ptr, param);
11418 if ((nmg->mg_flags & MGf_DUP) && nmg->mg_virtual && nmg->mg_virtual->svt_dup) {
11419 nmg->mg_virtual->svt_dup(aTHX_ nmg, param);
11425 #endif /* USE_ITHREADS */
11427 struct ptr_tbl_arena {
11428 struct ptr_tbl_arena *next;
11429 struct ptr_tbl_ent array[1023/3]; /* as ptr_tbl_ent has 3 pointers. */
11432 /* create a new pointer-mapping table */
11435 Perl_ptr_table_new(pTHX)
11438 PERL_UNUSED_CONTEXT;
11440 Newx(tbl, 1, PTR_TBL_t);
11441 tbl->tbl_max = 511;
11442 tbl->tbl_items = 0;
11443 tbl->tbl_arena = NULL;
11444 tbl->tbl_arena_next = NULL;
11445 tbl->tbl_arena_end = NULL;
11446 Newxz(tbl->tbl_ary, tbl->tbl_max + 1, PTR_TBL_ENT_t*);
11450 #define PTR_TABLE_HASH(ptr) \
11451 ((PTR2UV(ptr) >> 3) ^ (PTR2UV(ptr) >> (3 + 7)) ^ (PTR2UV(ptr) >> (3 + 17)))
11453 /* map an existing pointer using a table */
11455 STATIC PTR_TBL_ENT_t *
11456 S_ptr_table_find(PTR_TBL_t *const tbl, const void *const sv)
11458 PTR_TBL_ENT_t *tblent;
11459 const UV hash = PTR_TABLE_HASH(sv);
11461 PERL_ARGS_ASSERT_PTR_TABLE_FIND;
11463 tblent = tbl->tbl_ary[hash & tbl->tbl_max];
11464 for (; tblent; tblent = tblent->next) {
11465 if (tblent->oldval == sv)
11472 Perl_ptr_table_fetch(pTHX_ PTR_TBL_t *const tbl, const void *const sv)
11474 PTR_TBL_ENT_t const *const tblent = ptr_table_find(tbl, sv);
11476 PERL_ARGS_ASSERT_PTR_TABLE_FETCH;
11477 PERL_UNUSED_CONTEXT;
11479 return tblent ? tblent->newval : NULL;
11482 /* add a new entry to a pointer-mapping table */
11485 Perl_ptr_table_store(pTHX_ PTR_TBL_t *const tbl, const void *const oldsv, void *const newsv)
11487 PTR_TBL_ENT_t *tblent = ptr_table_find(tbl, oldsv);
11489 PERL_ARGS_ASSERT_PTR_TABLE_STORE;
11490 PERL_UNUSED_CONTEXT;
11493 tblent->newval = newsv;
11495 const UV entry = PTR_TABLE_HASH(oldsv) & tbl->tbl_max;
11497 if (tbl->tbl_arena_next == tbl->tbl_arena_end) {
11498 struct ptr_tbl_arena *new_arena;
11500 Newx(new_arena, 1, struct ptr_tbl_arena);
11501 new_arena->next = tbl->tbl_arena;
11502 tbl->tbl_arena = new_arena;
11503 tbl->tbl_arena_next = new_arena->array;
11504 tbl->tbl_arena_end = new_arena->array
11505 + sizeof(new_arena->array) / sizeof(new_arena->array[0]);
11508 tblent = tbl->tbl_arena_next++;
11510 tblent->oldval = oldsv;
11511 tblent->newval = newsv;
11512 tblent->next = tbl->tbl_ary[entry];
11513 tbl->tbl_ary[entry] = tblent;
11515 if (tblent->next && tbl->tbl_items > tbl->tbl_max)
11516 ptr_table_split(tbl);
11520 /* double the hash bucket size of an existing ptr table */
11523 Perl_ptr_table_split(pTHX_ PTR_TBL_t *const tbl)
11525 PTR_TBL_ENT_t **ary = tbl->tbl_ary;
11526 const UV oldsize = tbl->tbl_max + 1;
11527 UV newsize = oldsize * 2;
11530 PERL_ARGS_ASSERT_PTR_TABLE_SPLIT;
11531 PERL_UNUSED_CONTEXT;
11533 Renew(ary, newsize, PTR_TBL_ENT_t*);
11534 Zero(&ary[oldsize], newsize-oldsize, PTR_TBL_ENT_t*);
11535 tbl->tbl_max = --newsize;
11536 tbl->tbl_ary = ary;
11537 for (i=0; i < oldsize; i++, ary++) {
11538 PTR_TBL_ENT_t **entp = ary;
11539 PTR_TBL_ENT_t *ent = *ary;
11540 PTR_TBL_ENT_t **curentp;
11543 curentp = ary + oldsize;
11545 if ((newsize & PTR_TABLE_HASH(ent->oldval)) != i) {
11547 ent->next = *curentp;
11557 /* remove all the entries from a ptr table */
11558 /* Deprecated - will be removed post 5.14 */
11561 Perl_ptr_table_clear(pTHX_ PTR_TBL_t *const tbl)
11563 if (tbl && tbl->tbl_items) {
11564 struct ptr_tbl_arena *arena = tbl->tbl_arena;
11566 Zero(tbl->tbl_ary, tbl->tbl_max + 1, struct ptr_tbl_ent **);
11569 struct ptr_tbl_arena *next = arena->next;
11575 tbl->tbl_items = 0;
11576 tbl->tbl_arena = NULL;
11577 tbl->tbl_arena_next = NULL;
11578 tbl->tbl_arena_end = NULL;
11582 /* clear and free a ptr table */
11585 Perl_ptr_table_free(pTHX_ PTR_TBL_t *const tbl)
11587 struct ptr_tbl_arena *arena;
11593 arena = tbl->tbl_arena;
11596 struct ptr_tbl_arena *next = arena->next;
11602 Safefree(tbl->tbl_ary);
11606 #if defined(USE_ITHREADS)
11609 Perl_rvpv_dup(pTHX_ SV *const dstr, const SV *const sstr, CLONE_PARAMS *const param)
11611 PERL_ARGS_ASSERT_RVPV_DUP;
11614 if (SvWEAKREF(sstr)) {
11615 SvRV_set(dstr, sv_dup(SvRV_const(sstr), param));
11616 if (param->flags & CLONEf_JOIN_IN) {
11617 /* if joining, we add any back references individually rather
11618 * than copying the whole backref array */
11619 Perl_sv_add_backref(aTHX_ SvRV(dstr), dstr);
11623 SvRV_set(dstr, sv_dup_inc(SvRV_const(sstr), param));
11625 else if (SvPVX_const(sstr)) {
11626 /* Has something there */
11628 /* Normal PV - clone whole allocated space */
11629 SvPV_set(dstr, SAVEPVN(SvPVX_const(sstr), SvLEN(sstr)-1));
11630 if (SvREADONLY(sstr) && SvFAKE(sstr)) {
11631 /* Not that normal - actually sstr is copy on write.
11632 But we are a true, independent SV, so: */
11633 SvREADONLY_off(dstr);
11638 /* Special case - not normally malloced for some reason */
11639 if (isGV_with_GP(sstr)) {
11640 /* Don't need to do anything here. */
11642 else if ((SvREADONLY(sstr) && SvFAKE(sstr))) {
11643 /* A "shared" PV - clone it as "shared" PV */
11645 HEK_KEY(hek_dup(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)),
11649 /* Some other special case - random pointer */
11650 SvPV_set(dstr, (char *) SvPVX_const(sstr));
11655 /* Copy the NULL */
11656 SvPV_set(dstr, NULL);
11660 /* duplicate a list of SVs. source and dest may point to the same memory. */
11662 S_sv_dup_inc_multiple(pTHX_ SV *const *source, SV **dest,
11663 SSize_t items, CLONE_PARAMS *const param)
11665 PERL_ARGS_ASSERT_SV_DUP_INC_MULTIPLE;
11667 while (items-- > 0) {
11668 *dest++ = sv_dup_inc(*source++, param);
11674 /* duplicate an SV of any type (including AV, HV etc) */
11677 S_sv_dup_common(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
11682 PERL_ARGS_ASSERT_SV_DUP_COMMON;
11684 if (SvTYPE(sstr) == SVTYPEMASK) {
11685 #ifdef DEBUG_LEAKING_SCALARS_ABORT
11690 /* look for it in the table first */
11691 dstr = MUTABLE_SV(ptr_table_fetch(PL_ptr_table, sstr));
11695 if(param->flags & CLONEf_JOIN_IN) {
11696 /** We are joining here so we don't want do clone
11697 something that is bad **/
11698 if (SvTYPE(sstr) == SVt_PVHV) {
11699 const HEK * const hvname = HvNAME_HEK(sstr);
11701 /** don't clone stashes if they already exist **/
11702 dstr = MUTABLE_SV(gv_stashpvn(HEK_KEY(hvname), HEK_LEN(hvname), 0));
11703 ptr_table_store(PL_ptr_table, sstr, dstr);
11709 /* create anew and remember what it is */
11712 #ifdef DEBUG_LEAKING_SCALARS
11713 dstr->sv_debug_optype = sstr->sv_debug_optype;
11714 dstr->sv_debug_line = sstr->sv_debug_line;
11715 dstr->sv_debug_inpad = sstr->sv_debug_inpad;
11716 dstr->sv_debug_parent = (SV*)sstr;
11717 FREE_SV_DEBUG_FILE(dstr);
11718 dstr->sv_debug_file = savepv(sstr->sv_debug_file);
11721 ptr_table_store(PL_ptr_table, sstr, dstr);
11724 SvFLAGS(dstr) = SvFLAGS(sstr);
11725 SvFLAGS(dstr) &= ~SVf_OOK; /* don't propagate OOK hack */
11726 SvREFCNT(dstr) = 0; /* must be before any other dups! */
11729 if (SvANY(sstr) && PL_watch_pvx && SvPVX_const(sstr) == PL_watch_pvx)
11730 PerlIO_printf(Perl_debug_log, "watch at %p hit, found string \"%s\"\n",
11731 (void*)PL_watch_pvx, SvPVX_const(sstr));
11734 /* don't clone objects whose class has asked us not to */
11735 if (SvOBJECT(sstr) && ! (SvFLAGS(SvSTASH(sstr)) & SVphv_CLONEABLE)) {
11740 switch (SvTYPE(sstr)) {
11742 SvANY(dstr) = NULL;
11745 SvANY(dstr) = (XPVIV*)((char*)&(dstr->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
11747 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
11749 SvIV_set(dstr, SvIVX(sstr));
11753 SvANY(dstr) = new_XNV();
11754 SvNV_set(dstr, SvNVX(sstr));
11756 /* case SVt_BIND: */
11759 /* These are all the types that need complex bodies allocating. */
11761 const svtype sv_type = SvTYPE(sstr);
11762 const struct body_details *const sv_type_details
11763 = bodies_by_type + sv_type;
11767 Perl_croak(aTHX_ "Bizarre SvTYPE [%" IVdf "]", (IV)SvTYPE(sstr));
11782 assert(sv_type_details->body_size);
11783 if (sv_type_details->arena) {
11784 new_body_inline(new_body, sv_type);
11786 = (void*)((char*)new_body - sv_type_details->offset);
11788 new_body = new_NOARENA(sv_type_details);
11792 SvANY(dstr) = new_body;
11795 Copy(((char*)SvANY(sstr)) + sv_type_details->offset,
11796 ((char*)SvANY(dstr)) + sv_type_details->offset,
11797 sv_type_details->copy, char);
11799 Copy(((char*)SvANY(sstr)),
11800 ((char*)SvANY(dstr)),
11801 sv_type_details->body_size + sv_type_details->offset, char);
11804 if (sv_type != SVt_PVAV && sv_type != SVt_PVHV
11805 && !isGV_with_GP(dstr)
11806 && !(sv_type == SVt_PVIO && !(IoFLAGS(dstr) & IOf_FAKE_DIRP)))
11807 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
11809 /* The Copy above means that all the source (unduplicated) pointers
11810 are now in the destination. We can check the flags and the
11811 pointers in either, but it's possible that there's less cache
11812 missing by always going for the destination.
11813 FIXME - instrument and check that assumption */
11814 if (sv_type >= SVt_PVMG) {
11815 if ((sv_type == SVt_PVMG) && SvPAD_OUR(dstr)) {
11816 SvOURSTASH_set(dstr, hv_dup_inc(SvOURSTASH(dstr), param));
11817 } else if (SvMAGIC(dstr))
11818 SvMAGIC_set(dstr, mg_dup(SvMAGIC(dstr), param));
11820 SvSTASH_set(dstr, hv_dup_inc(SvSTASH(dstr), param));
11823 /* The cast silences a GCC warning about unhandled types. */
11824 switch ((int)sv_type) {
11834 /* FIXME for plugins */
11835 re_dup_guts((REGEXP*) sstr, (REGEXP*) dstr, param);
11838 /* XXX LvTARGOFF sometimes holds PMOP* when DEBUGGING */
11839 if (LvTYPE(dstr) == 't') /* for tie: unrefcnted fake (SV**) */
11840 LvTARG(dstr) = dstr;
11841 else if (LvTYPE(dstr) == 'T') /* for tie: fake HE */
11842 LvTARG(dstr) = MUTABLE_SV(he_dup((HE*)LvTARG(dstr), 0, param));
11844 LvTARG(dstr) = sv_dup_inc(LvTARG(dstr), param);
11846 /* non-GP case already handled above */
11847 if(isGV_with_GP(sstr)) {
11848 GvNAME_HEK(dstr) = hek_dup(GvNAME_HEK(dstr), param);
11849 /* Don't call sv_add_backref here as it's going to be
11850 created as part of the magic cloning of the symbol
11851 table--unless this is during a join and the stash
11852 is not actually being cloned. */
11853 /* Danger Will Robinson - GvGP(dstr) isn't initialised
11854 at the point of this comment. */
11855 GvSTASH(dstr) = hv_dup(GvSTASH(dstr), param);
11856 if (param->flags & CLONEf_JOIN_IN)
11857 Perl_sv_add_backref(aTHX_ MUTABLE_SV(GvSTASH(dstr)), dstr);
11858 GvGP_set(dstr, gp_dup(GvGP(sstr), param));
11859 (void)GpREFCNT_inc(GvGP(dstr));
11863 /* PL_parser->rsfp_filters entries have fake IoDIRP() */
11864 if(IoFLAGS(dstr) & IOf_FAKE_DIRP) {
11865 /* I have no idea why fake dirp (rsfps)
11866 should be treated differently but otherwise
11867 we end up with leaks -- sky*/
11868 IoTOP_GV(dstr) = gv_dup_inc(IoTOP_GV(dstr), param);
11869 IoFMT_GV(dstr) = gv_dup_inc(IoFMT_GV(dstr), param);
11870 IoBOTTOM_GV(dstr) = gv_dup_inc(IoBOTTOM_GV(dstr), param);
11872 IoTOP_GV(dstr) = gv_dup(IoTOP_GV(dstr), param);
11873 IoFMT_GV(dstr) = gv_dup(IoFMT_GV(dstr), param);
11874 IoBOTTOM_GV(dstr) = gv_dup(IoBOTTOM_GV(dstr), param);
11875 if (IoDIRP(dstr)) {
11876 IoDIRP(dstr) = dirp_dup(IoDIRP(dstr), param);
11879 /* IoDIRP(dstr) is already a copy of IoDIRP(sstr) */
11881 IoIFP(dstr) = fp_dup(IoIFP(sstr), IoTYPE(dstr), param);
11883 if (IoOFP(dstr) == IoIFP(sstr))
11884 IoOFP(dstr) = IoIFP(dstr);
11886 IoOFP(dstr) = fp_dup(IoOFP(dstr), IoTYPE(dstr), param);
11887 IoTOP_NAME(dstr) = SAVEPV(IoTOP_NAME(dstr));
11888 IoFMT_NAME(dstr) = SAVEPV(IoFMT_NAME(dstr));
11889 IoBOTTOM_NAME(dstr) = SAVEPV(IoBOTTOM_NAME(dstr));
11892 /* avoid cloning an empty array */
11893 if (AvARRAY((const AV *)sstr) && AvFILLp((const AV *)sstr) >= 0) {
11894 SV **dst_ary, **src_ary;
11895 SSize_t items = AvFILLp((const AV *)sstr) + 1;
11897 src_ary = AvARRAY((const AV *)sstr);
11898 Newxz(dst_ary, AvMAX((const AV *)sstr)+1, SV*);
11899 ptr_table_store(PL_ptr_table, src_ary, dst_ary);
11900 AvARRAY(MUTABLE_AV(dstr)) = dst_ary;
11901 AvALLOC((const AV *)dstr) = dst_ary;
11902 if (AvREAL((const AV *)sstr)) {
11903 dst_ary = sv_dup_inc_multiple(src_ary, dst_ary, items,
11907 while (items-- > 0)
11908 *dst_ary++ = sv_dup(*src_ary++, param);
11910 items = AvMAX((const AV *)sstr) - AvFILLp((const AV *)sstr);
11911 while (items-- > 0) {
11912 *dst_ary++ = &PL_sv_undef;
11916 AvARRAY(MUTABLE_AV(dstr)) = NULL;
11917 AvALLOC((const AV *)dstr) = (SV**)NULL;
11918 AvMAX( (const AV *)dstr) = -1;
11919 AvFILLp((const AV *)dstr) = -1;
11923 if (HvARRAY((const HV *)sstr)) {
11925 const bool sharekeys = !!HvSHAREKEYS(sstr);
11926 XPVHV * const dxhv = (XPVHV*)SvANY(dstr);
11927 XPVHV * const sxhv = (XPVHV*)SvANY(sstr);
11929 Newx(darray, PERL_HV_ARRAY_ALLOC_BYTES(dxhv->xhv_max+1)
11930 + (SvOOK(sstr) ? sizeof(struct xpvhv_aux) : 0),
11932 HvARRAY(dstr) = (HE**)darray;
11933 while (i <= sxhv->xhv_max) {
11934 const HE * const source = HvARRAY(sstr)[i];
11935 HvARRAY(dstr)[i] = source
11936 ? he_dup(source, sharekeys, param) : 0;
11940 const struct xpvhv_aux * const saux = HvAUX(sstr);
11941 struct xpvhv_aux * const daux = HvAUX(dstr);
11942 /* This flag isn't copied. */
11943 /* SvOOK_on(hv) attacks the IV flags. */
11944 SvFLAGS(dstr) |= SVf_OOK;
11946 if (saux->xhv_name_count) {
11947 HEK ** const sname = saux->xhv_name_u.xhvnameu_names;
11949 = saux->xhv_name_count < 0
11950 ? -saux->xhv_name_count
11951 : saux->xhv_name_count;
11952 HEK **shekp = sname + count;
11954 Newx(daux->xhv_name_u.xhvnameu_names, count, HEK *);
11955 dhekp = daux->xhv_name_u.xhvnameu_names + count;
11956 while (shekp-- > sname) {
11958 *dhekp = hek_dup(*shekp, param);
11962 daux->xhv_name_u.xhvnameu_name
11963 = hek_dup(saux->xhv_name_u.xhvnameu_name,
11966 daux->xhv_name_count = saux->xhv_name_count;
11968 daux->xhv_riter = saux->xhv_riter;
11969 daux->xhv_eiter = saux->xhv_eiter
11970 ? he_dup(saux->xhv_eiter,
11971 cBOOL(HvSHAREKEYS(sstr)), param) : 0;
11972 /* backref array needs refcnt=2; see sv_add_backref */
11973 daux->xhv_backreferences =
11974 (param->flags & CLONEf_JOIN_IN)
11975 /* when joining, we let the individual GVs and
11976 * CVs add themselves to backref as
11977 * needed. This avoids pulling in stuff
11978 * that isn't required, and simplifies the
11979 * case where stashes aren't cloned back
11980 * if they already exist in the parent
11983 : saux->xhv_backreferences
11984 ? (SvTYPE(saux->xhv_backreferences) == SVt_PVAV)
11985 ? MUTABLE_AV(SvREFCNT_inc(
11986 sv_dup_inc((const SV *)
11987 saux->xhv_backreferences, param)))
11988 : MUTABLE_AV(sv_dup((const SV *)
11989 saux->xhv_backreferences, param))
11992 daux->xhv_mro_meta = saux->xhv_mro_meta
11993 ? mro_meta_dup(saux->xhv_mro_meta, param)
11996 /* Record stashes for possible cloning in Perl_clone(). */
11998 av_push(param->stashes, dstr);
12002 HvARRAY(MUTABLE_HV(dstr)) = NULL;
12005 if (!(param->flags & CLONEf_COPY_STACKS)) {
12010 /* NOTE: not refcounted */
12011 SvANY(MUTABLE_CV(dstr))->xcv_stash =
12012 hv_dup(CvSTASH(dstr), param);
12013 if ((param->flags & CLONEf_JOIN_IN) && CvSTASH(dstr))
12014 Perl_sv_add_backref(aTHX_ MUTABLE_SV(CvSTASH(dstr)), dstr);
12015 if (!CvISXSUB(dstr)) {
12017 CvROOT(dstr) = OpREFCNT_inc(CvROOT(dstr));
12019 CvFILE(dstr) = SAVEPV(CvFILE(dstr));
12020 } else if (CvCONST(dstr)) {
12021 CvXSUBANY(dstr).any_ptr =
12022 sv_dup_inc((const SV *)CvXSUBANY(dstr).any_ptr, param);
12024 /* don't dup if copying back - CvGV isn't refcounted, so the
12025 * duped GV may never be freed. A bit of a hack! DAPM */
12026 SvANY(MUTABLE_CV(dstr))->xcv_gv =
12028 ? gv_dup_inc(CvGV(sstr), param)
12029 : (param->flags & CLONEf_JOIN_IN)
12031 : gv_dup(CvGV(sstr), param);
12033 CvPADLIST(dstr) = padlist_dup(CvPADLIST(sstr), param);
12035 CvWEAKOUTSIDE(sstr)
12036 ? cv_dup( CvOUTSIDE(dstr), param)
12037 : cv_dup_inc(CvOUTSIDE(dstr), param);
12043 if (SvOBJECT(dstr) && SvTYPE(dstr) != SVt_PVIO)
12050 Perl_sv_dup_inc(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
12052 PERL_ARGS_ASSERT_SV_DUP_INC;
12053 return sstr ? SvREFCNT_inc(sv_dup_common(sstr, param)) : NULL;
12057 Perl_sv_dup(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
12059 SV *dstr = sstr ? sv_dup_common(sstr, param) : NULL;
12060 PERL_ARGS_ASSERT_SV_DUP;
12062 /* Track every SV that (at least initially) had a reference count of 0.
12063 We need to do this by holding an actual reference to it in this array.
12064 If we attempt to cheat, turn AvREAL_off(), and store only pointers
12065 (akin to the stashes hash, and the perl stack), we come unstuck if
12066 a weak reference (or other SV legitimately SvREFCNT() == 0 for this
12067 thread) is manipulated in a CLONE method, because CLONE runs before the
12068 unreferenced array is walked to find SVs still with SvREFCNT() == 0
12069 (and fix things up by giving each a reference via the temps stack).
12070 Instead, during CLONE, if the 0-referenced SV has SvREFCNT_inc() and
12071 then SvREFCNT_dec(), it will be cleaned up (and added to the free list)
12072 before the walk of unreferenced happens and a reference to that is SV
12073 added to the temps stack. At which point we have the same SV considered
12074 to be in use, and free to be re-used. Not good.
12076 if (dstr && !(param->flags & CLONEf_COPY_STACKS) && !SvREFCNT(dstr)) {
12077 assert(param->unreferenced);
12078 av_push(param->unreferenced, SvREFCNT_inc(dstr));
12084 /* duplicate a context */
12087 Perl_cx_dup(pTHX_ PERL_CONTEXT *cxs, I32 ix, I32 max, CLONE_PARAMS* param)
12089 PERL_CONTEXT *ncxs;
12091 PERL_ARGS_ASSERT_CX_DUP;
12094 return (PERL_CONTEXT*)NULL;
12096 /* look for it in the table first */
12097 ncxs = (PERL_CONTEXT*)ptr_table_fetch(PL_ptr_table, cxs);
12101 /* create anew and remember what it is */
12102 Newx(ncxs, max + 1, PERL_CONTEXT);
12103 ptr_table_store(PL_ptr_table, cxs, ncxs);
12104 Copy(cxs, ncxs, max + 1, PERL_CONTEXT);
12107 PERL_CONTEXT * const ncx = &ncxs[ix];
12108 if (CxTYPE(ncx) == CXt_SUBST) {
12109 Perl_croak(aTHX_ "Cloning substitution context is unimplemented");
12112 switch (CxTYPE(ncx)) {
12114 ncx->blk_sub.cv = (ncx->blk_sub.olddepth == 0
12115 ? cv_dup_inc(ncx->blk_sub.cv, param)
12116 : cv_dup(ncx->blk_sub.cv,param));
12117 ncx->blk_sub.argarray = (CxHASARGS(ncx)
12118 ? av_dup_inc(ncx->blk_sub.argarray,
12121 ncx->blk_sub.savearray = av_dup_inc(ncx->blk_sub.savearray,
12123 ncx->blk_sub.oldcomppad = (PAD*)ptr_table_fetch(PL_ptr_table,
12124 ncx->blk_sub.oldcomppad);
12127 ncx->blk_eval.old_namesv = sv_dup_inc(ncx->blk_eval.old_namesv,
12129 ncx->blk_eval.cur_text = sv_dup(ncx->blk_eval.cur_text, param);
12131 case CXt_LOOP_LAZYSV:
12132 ncx->blk_loop.state_u.lazysv.end
12133 = sv_dup_inc(ncx->blk_loop.state_u.lazysv.end, param);
12134 /* We are taking advantage of av_dup_inc and sv_dup_inc
12135 actually being the same function, and order equivalence of
12137 We can assert the later [but only at run time :-(] */
12138 assert ((void *) &ncx->blk_loop.state_u.ary.ary ==
12139 (void *) &ncx->blk_loop.state_u.lazysv.cur);
12141 ncx->blk_loop.state_u.ary.ary
12142 = av_dup_inc(ncx->blk_loop.state_u.ary.ary, param);
12143 case CXt_LOOP_LAZYIV:
12144 case CXt_LOOP_PLAIN:
12145 if (CxPADLOOP(ncx)) {
12146 ncx->blk_loop.itervar_u.oldcomppad
12147 = (PAD*)ptr_table_fetch(PL_ptr_table,
12148 ncx->blk_loop.itervar_u.oldcomppad);
12150 ncx->blk_loop.itervar_u.gv
12151 = gv_dup((const GV *)ncx->blk_loop.itervar_u.gv,
12156 ncx->blk_format.cv = cv_dup(ncx->blk_format.cv, param);
12157 ncx->blk_format.gv = gv_dup(ncx->blk_format.gv, param);
12158 ncx->blk_format.dfoutgv = gv_dup_inc(ncx->blk_format.dfoutgv,
12171 /* duplicate a stack info structure */
12174 Perl_si_dup(pTHX_ PERL_SI *si, CLONE_PARAMS* param)
12178 PERL_ARGS_ASSERT_SI_DUP;
12181 return (PERL_SI*)NULL;
12183 /* look for it in the table first */
12184 nsi = (PERL_SI*)ptr_table_fetch(PL_ptr_table, si);
12188 /* create anew and remember what it is */
12189 Newxz(nsi, 1, PERL_SI);
12190 ptr_table_store(PL_ptr_table, si, nsi);
12192 nsi->si_stack = av_dup_inc(si->si_stack, param);
12193 nsi->si_cxix = si->si_cxix;
12194 nsi->si_cxmax = si->si_cxmax;
12195 nsi->si_cxstack = cx_dup(si->si_cxstack, si->si_cxix, si->si_cxmax, param);
12196 nsi->si_type = si->si_type;
12197 nsi->si_prev = si_dup(si->si_prev, param);
12198 nsi->si_next = si_dup(si->si_next, param);
12199 nsi->si_markoff = si->si_markoff;
12204 #define POPINT(ss,ix) ((ss)[--(ix)].any_i32)
12205 #define TOPINT(ss,ix) ((ss)[ix].any_i32)
12206 #define POPLONG(ss,ix) ((ss)[--(ix)].any_long)
12207 #define TOPLONG(ss,ix) ((ss)[ix].any_long)
12208 #define POPIV(ss,ix) ((ss)[--(ix)].any_iv)
12209 #define TOPIV(ss,ix) ((ss)[ix].any_iv)
12210 #define POPUV(ss,ix) ((ss)[--(ix)].any_uv)
12211 #define TOPUV(ss,ix) ((ss)[ix].any_uv)
12212 #define POPBOOL(ss,ix) ((ss)[--(ix)].any_bool)
12213 #define TOPBOOL(ss,ix) ((ss)[ix].any_bool)
12214 #define POPPTR(ss,ix) ((ss)[--(ix)].any_ptr)
12215 #define TOPPTR(ss,ix) ((ss)[ix].any_ptr)
12216 #define POPDPTR(ss,ix) ((ss)[--(ix)].any_dptr)
12217 #define TOPDPTR(ss,ix) ((ss)[ix].any_dptr)
12218 #define POPDXPTR(ss,ix) ((ss)[--(ix)].any_dxptr)
12219 #define TOPDXPTR(ss,ix) ((ss)[ix].any_dxptr)
12222 #define pv_dup_inc(p) SAVEPV(p)
12223 #define pv_dup(p) SAVEPV(p)
12224 #define svp_dup_inc(p,pp) any_dup(p,pp)
12226 /* map any object to the new equivent - either something in the
12227 * ptr table, or something in the interpreter structure
12231 Perl_any_dup(pTHX_ void *v, const PerlInterpreter *proto_perl)
12235 PERL_ARGS_ASSERT_ANY_DUP;
12238 return (void*)NULL;
12240 /* look for it in the table first */
12241 ret = ptr_table_fetch(PL_ptr_table, v);
12245 /* see if it is part of the interpreter structure */
12246 if (v >= (void*)proto_perl && v < (void*)(proto_perl+1))
12247 ret = (void*)(((char*)aTHX) + (((char*)v) - (char*)proto_perl));
12255 /* duplicate the save stack */
12258 Perl_ss_dup(pTHX_ PerlInterpreter *proto_perl, CLONE_PARAMS* param)
12261 ANY * const ss = proto_perl->Isavestack;
12262 const I32 max = proto_perl->Isavestack_max;
12263 I32 ix = proto_perl->Isavestack_ix;
12276 void (*dptr) (void*);
12277 void (*dxptr) (pTHX_ void*);
12279 PERL_ARGS_ASSERT_SS_DUP;
12281 Newxz(nss, max, ANY);
12284 const UV uv = POPUV(ss,ix);
12285 const U8 type = (U8)uv & SAVE_MASK;
12287 TOPUV(nss,ix) = uv;
12289 case SAVEt_CLEARSV:
12291 case SAVEt_HELEM: /* hash element */
12292 sv = (const SV *)POPPTR(ss,ix);
12293 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12295 case SAVEt_ITEM: /* normal string */
12296 case SAVEt_GVSV: /* scalar slot in GV */
12297 case SAVEt_SV: /* scalar reference */
12298 sv = (const SV *)POPPTR(ss,ix);
12299 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12302 case SAVEt_MORTALIZESV:
12303 sv = (const SV *)POPPTR(ss,ix);
12304 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12306 case SAVEt_SHARED_PVREF: /* char* in shared space */
12307 c = (char*)POPPTR(ss,ix);
12308 TOPPTR(nss,ix) = savesharedpv(c);
12309 ptr = POPPTR(ss,ix);
12310 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12312 case SAVEt_GENERIC_SVREF: /* generic sv */
12313 case SAVEt_SVREF: /* scalar reference */
12314 sv = (const SV *)POPPTR(ss,ix);
12315 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12316 ptr = POPPTR(ss,ix);
12317 TOPPTR(nss,ix) = svp_dup_inc((SV**)ptr, proto_perl);/* XXXXX */
12319 case SAVEt_HV: /* hash reference */
12320 case SAVEt_AV: /* array reference */
12321 sv = (const SV *) POPPTR(ss,ix);
12322 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12324 case SAVEt_COMPPAD:
12326 sv = (const SV *) POPPTR(ss,ix);
12327 TOPPTR(nss,ix) = sv_dup(sv, param);
12329 case SAVEt_INT: /* int reference */
12330 ptr = POPPTR(ss,ix);
12331 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12332 intval = (int)POPINT(ss,ix);
12333 TOPINT(nss,ix) = intval;
12335 case SAVEt_LONG: /* long reference */
12336 ptr = POPPTR(ss,ix);
12337 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12338 longval = (long)POPLONG(ss,ix);
12339 TOPLONG(nss,ix) = longval;
12341 case SAVEt_I32: /* I32 reference */
12342 case SAVEt_COP_ARYBASE: /* call CopARYBASE_set */
12343 ptr = POPPTR(ss,ix);
12344 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12346 TOPINT(nss,ix) = i;
12348 case SAVEt_IV: /* IV reference */
12349 ptr = POPPTR(ss,ix);
12350 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12352 TOPIV(nss,ix) = iv;
12354 case SAVEt_HPTR: /* HV* reference */
12355 case SAVEt_APTR: /* AV* reference */
12356 case SAVEt_SPTR: /* SV* reference */
12357 ptr = POPPTR(ss,ix);
12358 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12359 sv = (const SV *)POPPTR(ss,ix);
12360 TOPPTR(nss,ix) = sv_dup(sv, param);
12362 case SAVEt_VPTR: /* random* reference */
12363 ptr = POPPTR(ss,ix);
12364 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12366 case SAVEt_INT_SMALL:
12367 case SAVEt_I32_SMALL:
12368 case SAVEt_I16: /* I16 reference */
12369 case SAVEt_I8: /* I8 reference */
12371 ptr = POPPTR(ss,ix);
12372 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12374 case SAVEt_GENERIC_PVREF: /* generic char* */
12375 case SAVEt_PPTR: /* char* reference */
12376 ptr = POPPTR(ss,ix);
12377 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12378 c = (char*)POPPTR(ss,ix);
12379 TOPPTR(nss,ix) = pv_dup(c);
12381 case SAVEt_GP: /* scalar reference */
12382 gp = (GP*)POPPTR(ss,ix);
12383 TOPPTR(nss,ix) = gp = gp_dup(gp, param);
12384 (void)GpREFCNT_inc(gp);
12385 gv = (const GV *)POPPTR(ss,ix);
12386 TOPPTR(nss,ix) = gv_dup_inc(gv, param);
12389 ptr = POPPTR(ss,ix);
12390 if (ptr && (((OP*)ptr)->op_private & OPpREFCOUNTED)) {
12391 /* these are assumed to be refcounted properly */
12393 switch (((OP*)ptr)->op_type) {
12395 case OP_LEAVESUBLV:
12399 case OP_LEAVEWRITE:
12400 TOPPTR(nss,ix) = ptr;
12403 (void) OpREFCNT_inc(o);
12407 TOPPTR(nss,ix) = NULL;
12412 TOPPTR(nss,ix) = NULL;
12414 case SAVEt_FREECOPHH:
12415 ptr = POPPTR(ss,ix);
12416 TOPPTR(nss,ix) = cophh_copy((COPHH *)ptr);
12419 hv = (const HV *)POPPTR(ss,ix);
12420 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
12422 TOPINT(nss,ix) = i;
12425 c = (char*)POPPTR(ss,ix);
12426 TOPPTR(nss,ix) = pv_dup_inc(c);
12428 case SAVEt_STACK_POS: /* Position on Perl stack */
12430 TOPINT(nss,ix) = i;
12432 case SAVEt_DESTRUCTOR:
12433 ptr = POPPTR(ss,ix);
12434 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
12435 dptr = POPDPTR(ss,ix);
12436 TOPDPTR(nss,ix) = DPTR2FPTR(void (*)(void*),
12437 any_dup(FPTR2DPTR(void *, dptr),
12440 case SAVEt_DESTRUCTOR_X:
12441 ptr = POPPTR(ss,ix);
12442 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
12443 dxptr = POPDXPTR(ss,ix);
12444 TOPDXPTR(nss,ix) = DPTR2FPTR(void (*)(pTHX_ void*),
12445 any_dup(FPTR2DPTR(void *, dxptr),
12448 case SAVEt_REGCONTEXT:
12450 ix -= uv >> SAVE_TIGHT_SHIFT;
12452 case SAVEt_AELEM: /* array element */
12453 sv = (const SV *)POPPTR(ss,ix);
12454 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12456 TOPINT(nss,ix) = i;
12457 av = (const AV *)POPPTR(ss,ix);
12458 TOPPTR(nss,ix) = av_dup_inc(av, param);
12461 ptr = POPPTR(ss,ix);
12462 TOPPTR(nss,ix) = ptr;
12465 ptr = POPPTR(ss,ix);
12466 ptr = cophh_copy((COPHH*)ptr);
12467 TOPPTR(nss,ix) = ptr;
12469 TOPINT(nss,ix) = i;
12470 if (i & HINT_LOCALIZE_HH) {
12471 hv = (const HV *)POPPTR(ss,ix);
12472 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
12475 case SAVEt_PADSV_AND_MORTALIZE:
12476 longval = (long)POPLONG(ss,ix);
12477 TOPLONG(nss,ix) = longval;
12478 ptr = POPPTR(ss,ix);
12479 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12480 sv = (const SV *)POPPTR(ss,ix);
12481 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12483 case SAVEt_SET_SVFLAGS:
12485 TOPINT(nss,ix) = i;
12487 TOPINT(nss,ix) = i;
12488 sv = (const SV *)POPPTR(ss,ix);
12489 TOPPTR(nss,ix) = sv_dup(sv, param);
12491 case SAVEt_RE_STATE:
12493 const struct re_save_state *const old_state
12494 = (struct re_save_state *)
12495 (ss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
12496 struct re_save_state *const new_state
12497 = (struct re_save_state *)
12498 (nss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
12500 Copy(old_state, new_state, 1, struct re_save_state);
12501 ix -= SAVESTACK_ALLOC_FOR_RE_SAVE_STATE;
12503 new_state->re_state_bostr
12504 = pv_dup(old_state->re_state_bostr);
12505 new_state->re_state_reginput
12506 = pv_dup(old_state->re_state_reginput);
12507 new_state->re_state_regeol
12508 = pv_dup(old_state->re_state_regeol);
12509 new_state->re_state_regoffs
12510 = (regexp_paren_pair*)
12511 any_dup(old_state->re_state_regoffs, proto_perl);
12512 new_state->re_state_reglastparen
12513 = (U32*) any_dup(old_state->re_state_reglastparen,
12515 new_state->re_state_reglastcloseparen
12516 = (U32*)any_dup(old_state->re_state_reglastcloseparen,
12518 /* XXX This just has to be broken. The old save_re_context
12519 code did SAVEGENERICPV(PL_reg_start_tmp);
12520 PL_reg_start_tmp is char **.
12521 Look above to what the dup code does for
12522 SAVEt_GENERIC_PVREF
12523 It can never have worked.
12524 So this is merely a faithful copy of the exiting bug: */
12525 new_state->re_state_reg_start_tmp
12526 = (char **) pv_dup((char *)
12527 old_state->re_state_reg_start_tmp);
12528 /* I assume that it only ever "worked" because no-one called
12529 (pseudo)fork while the regexp engine had re-entered itself.
12531 #ifdef PERL_OLD_COPY_ON_WRITE
12532 new_state->re_state_nrs
12533 = sv_dup(old_state->re_state_nrs, param);
12535 new_state->re_state_reg_magic
12536 = (MAGIC*) any_dup(old_state->re_state_reg_magic,
12538 new_state->re_state_reg_oldcurpm
12539 = (PMOP*) any_dup(old_state->re_state_reg_oldcurpm,
12541 new_state->re_state_reg_curpm
12542 = (PMOP*) any_dup(old_state->re_state_reg_curpm,
12544 new_state->re_state_reg_oldsaved
12545 = pv_dup(old_state->re_state_reg_oldsaved);
12546 new_state->re_state_reg_poscache
12547 = pv_dup(old_state->re_state_reg_poscache);
12548 new_state->re_state_reg_starttry
12549 = pv_dup(old_state->re_state_reg_starttry);
12552 case SAVEt_COMPILE_WARNINGS:
12553 ptr = POPPTR(ss,ix);
12554 TOPPTR(nss,ix) = DUP_WARNINGS((STRLEN*)ptr);
12557 ptr = POPPTR(ss,ix);
12558 TOPPTR(nss,ix) = parser_dup((const yy_parser*)ptr, param);
12562 "panic: ss_dup inconsistency (%"IVdf")", (IV) type);
12570 /* if sv is a stash, call $class->CLONE_SKIP(), and set the SVphv_CLONEABLE
12571 * flag to the result. This is done for each stash before cloning starts,
12572 * so we know which stashes want their objects cloned */
12575 do_mark_cloneable_stash(pTHX_ SV *const sv)
12577 const HEK * const hvname = HvNAME_HEK((const HV *)sv);
12579 GV* const cloner = gv_fetchmethod_autoload(MUTABLE_HV(sv), "CLONE_SKIP", 0);
12580 SvFLAGS(sv) |= SVphv_CLONEABLE; /* clone objects by default */
12581 if (cloner && GvCV(cloner)) {
12588 mXPUSHs(newSVhek(hvname));
12590 call_sv(MUTABLE_SV(GvCV(cloner)), G_SCALAR);
12597 SvFLAGS(sv) &= ~SVphv_CLONEABLE;
12605 =for apidoc perl_clone
12607 Create and return a new interpreter by cloning the current one.
12609 perl_clone takes these flags as parameters:
12611 CLONEf_COPY_STACKS - is used to, well, copy the stacks also,
12612 without it we only clone the data and zero the stacks,
12613 with it we copy the stacks and the new perl interpreter is
12614 ready to run at the exact same point as the previous one.
12615 The pseudo-fork code uses COPY_STACKS while the
12616 threads->create doesn't.
12618 CLONEf_KEEP_PTR_TABLE
12619 perl_clone keeps a ptr_table with the pointer of the old
12620 variable as a key and the new variable as a value,
12621 this allows it to check if something has been cloned and not
12622 clone it again but rather just use the value and increase the
12623 refcount. If KEEP_PTR_TABLE is not set then perl_clone will kill
12624 the ptr_table using the function
12625 C<ptr_table_free(PL_ptr_table); PL_ptr_table = NULL;>,
12626 reason to keep it around is if you want to dup some of your own
12627 variable who are outside the graph perl scans, example of this
12628 code is in threads.xs create
12631 This is a win32 thing, it is ignored on unix, it tells perls
12632 win32host code (which is c++) to clone itself, this is needed on
12633 win32 if you want to run two threads at the same time,
12634 if you just want to do some stuff in a separate perl interpreter
12635 and then throw it away and return to the original one,
12636 you don't need to do anything.
12641 /* XXX the above needs expanding by someone who actually understands it ! */
12642 EXTERN_C PerlInterpreter *
12643 perl_clone_host(PerlInterpreter* proto_perl, UV flags);
12646 perl_clone(PerlInterpreter *proto_perl, UV flags)
12649 #ifdef PERL_IMPLICIT_SYS
12651 PERL_ARGS_ASSERT_PERL_CLONE;
12653 /* perlhost.h so we need to call into it
12654 to clone the host, CPerlHost should have a c interface, sky */
12656 if (flags & CLONEf_CLONE_HOST) {
12657 return perl_clone_host(proto_perl,flags);
12659 return perl_clone_using(proto_perl, flags,
12661 proto_perl->IMemShared,
12662 proto_perl->IMemParse,
12664 proto_perl->IStdIO,
12668 proto_perl->IProc);
12672 perl_clone_using(PerlInterpreter *proto_perl, UV flags,
12673 struct IPerlMem* ipM, struct IPerlMem* ipMS,
12674 struct IPerlMem* ipMP, struct IPerlEnv* ipE,
12675 struct IPerlStdIO* ipStd, struct IPerlLIO* ipLIO,
12676 struct IPerlDir* ipD, struct IPerlSock* ipS,
12677 struct IPerlProc* ipP)
12679 /* XXX many of the string copies here can be optimized if they're
12680 * constants; they need to be allocated as common memory and just
12681 * their pointers copied. */
12684 CLONE_PARAMS clone_params;
12685 CLONE_PARAMS* const param = &clone_params;
12687 PerlInterpreter * const my_perl = (PerlInterpreter*)(*ipM->pMalloc)(ipM, sizeof(PerlInterpreter));
12689 PERL_ARGS_ASSERT_PERL_CLONE_USING;
12690 #else /* !PERL_IMPLICIT_SYS */
12692 CLONE_PARAMS clone_params;
12693 CLONE_PARAMS* param = &clone_params;
12694 PerlInterpreter * const my_perl = (PerlInterpreter*)PerlMem_malloc(sizeof(PerlInterpreter));
12696 PERL_ARGS_ASSERT_PERL_CLONE;
12697 #endif /* PERL_IMPLICIT_SYS */
12699 /* for each stash, determine whether its objects should be cloned */
12700 S_visit(proto_perl, do_mark_cloneable_stash, SVt_PVHV, SVTYPEMASK);
12701 PERL_SET_THX(my_perl);
12704 PoisonNew(my_perl, 1, PerlInterpreter);
12707 PL_defstash = NULL; /* may be used by perl malloc() */
12710 PL_scopestack_name = 0;
12712 PL_savestack_ix = 0;
12713 PL_savestack_max = -1;
12714 PL_sig_pending = 0;
12716 Zero(&PL_debug_pad, 1, struct perl_debug_pad);
12717 # ifdef DEBUG_LEAKING_SCALARS
12718 PL_sv_serial = (((UV)my_perl >> 2) & 0xfff) * 1000000;
12720 #else /* !DEBUGGING */
12721 Zero(my_perl, 1, PerlInterpreter);
12722 #endif /* DEBUGGING */
12724 #ifdef PERL_IMPLICIT_SYS
12725 /* host pointers */
12727 PL_MemShared = ipMS;
12728 PL_MemParse = ipMP;
12735 #endif /* PERL_IMPLICIT_SYS */
12737 param->flags = flags;
12738 /* Nothing in the core code uses this, but we make it available to
12739 extensions (using mg_dup). */
12740 param->proto_perl = proto_perl;
12741 /* Likely nothing will use this, but it is initialised to be consistent
12742 with Perl_clone_params_new(). */
12743 param->new_perl = my_perl;
12744 param->unreferenced = NULL;
12746 INIT_TRACK_MEMPOOL(my_perl->Imemory_debug_header, my_perl);
12748 PL_body_arenas = NULL;
12749 Zero(&PL_body_roots, 1, PL_body_roots);
12752 PL_sv_objcount = 0;
12754 PL_sv_arenaroot = NULL;
12756 PL_debug = proto_perl->Idebug;
12758 PL_hash_seed = proto_perl->Ihash_seed;
12759 PL_rehash_seed = proto_perl->Irehash_seed;
12761 SvANY(&PL_sv_undef) = NULL;
12762 SvREFCNT(&PL_sv_undef) = (~(U32)0)/2;
12763 SvFLAGS(&PL_sv_undef) = SVf_READONLY|SVt_NULL;
12764 SvREFCNT(&PL_sv_no) = (~(U32)0)/2;
12765 SvFLAGS(&PL_sv_no) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
12766 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
12768 SvANY(&PL_sv_yes) = new_XPVNV();
12769 SvREFCNT(&PL_sv_yes) = (~(U32)0)/2;
12770 SvFLAGS(&PL_sv_yes) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
12771 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
12773 /* dbargs array probably holds garbage */
12776 PL_compiling = proto_perl->Icompiling;
12778 #ifdef PERL_DEBUG_READONLY_OPS
12783 /* pseudo environmental stuff */
12784 PL_origargc = proto_perl->Iorigargc;
12785 PL_origargv = proto_perl->Iorigargv;
12787 /* Set tainting stuff before PerlIO_debug can possibly get called */
12788 PL_tainting = proto_perl->Itainting;
12789 PL_taint_warn = proto_perl->Itaint_warn;
12791 PL_minus_c = proto_perl->Iminus_c;
12793 PL_localpatches = proto_perl->Ilocalpatches;
12794 PL_splitstr = proto_perl->Isplitstr;
12795 PL_minus_n = proto_perl->Iminus_n;
12796 PL_minus_p = proto_perl->Iminus_p;
12797 PL_minus_l = proto_perl->Iminus_l;
12798 PL_minus_a = proto_perl->Iminus_a;
12799 PL_minus_E = proto_perl->Iminus_E;
12800 PL_minus_F = proto_perl->Iminus_F;
12801 PL_doswitches = proto_perl->Idoswitches;
12802 PL_dowarn = proto_perl->Idowarn;
12803 PL_sawampersand = proto_perl->Isawampersand;
12804 PL_unsafe = proto_perl->Iunsafe;
12805 PL_perldb = proto_perl->Iperldb;
12806 PL_perl_destruct_level = proto_perl->Iperl_destruct_level;
12807 PL_exit_flags = proto_perl->Iexit_flags;
12809 /* XXX time(&PL_basetime) when asked for? */
12810 PL_basetime = proto_perl->Ibasetime;
12812 PL_maxsysfd = proto_perl->Imaxsysfd;
12813 PL_statusvalue = proto_perl->Istatusvalue;
12815 PL_statusvalue_vms = proto_perl->Istatusvalue_vms;
12817 PL_statusvalue_posix = proto_perl->Istatusvalue_posix;
12820 /* RE engine related */
12821 Zero(&PL_reg_state, 1, struct re_save_state);
12822 PL_reginterp_cnt = 0;
12823 PL_regmatch_slab = NULL;
12825 PL_sub_generation = proto_perl->Isub_generation;
12827 /* funky return mechanisms */
12828 PL_forkprocess = proto_perl->Iforkprocess;
12830 /* internal state */
12831 PL_maxo = proto_perl->Imaxo;
12833 PL_main_start = proto_perl->Imain_start;
12834 PL_eval_root = proto_perl->Ieval_root;
12835 PL_eval_start = proto_perl->Ieval_start;
12837 PL_filemode = proto_perl->Ifilemode;
12838 PL_lastfd = proto_perl->Ilastfd;
12839 PL_oldname = proto_perl->Ioldname; /* XXX not quite right */
12842 PL_gensym = proto_perl->Igensym;
12844 PL_laststatval = proto_perl->Ilaststatval;
12845 PL_laststype = proto_perl->Ilaststype;
12848 PL_profiledata = NULL;
12850 PL_generation = proto_perl->Igeneration;
12852 PL_in_clean_objs = proto_perl->Iin_clean_objs;
12853 PL_in_clean_all = proto_perl->Iin_clean_all;
12855 PL_uid = proto_perl->Iuid;
12856 PL_euid = proto_perl->Ieuid;
12857 PL_gid = proto_perl->Igid;
12858 PL_egid = proto_perl->Iegid;
12859 PL_nomemok = proto_perl->Inomemok;
12860 PL_an = proto_perl->Ian;
12861 PL_evalseq = proto_perl->Ievalseq;
12862 PL_origenviron = proto_perl->Iorigenviron; /* XXX not quite right */
12863 PL_origalen = proto_perl->Iorigalen;
12865 PL_sighandlerp = proto_perl->Isighandlerp;
12867 PL_runops = proto_perl->Irunops;
12869 PL_subline = proto_perl->Isubline;
12872 PL_cryptseen = proto_perl->Icryptseen;
12875 PL_hints = proto_perl->Ihints;
12877 PL_amagic_generation = proto_perl->Iamagic_generation;
12879 #ifdef USE_LOCALE_COLLATE
12880 PL_collation_ix = proto_perl->Icollation_ix;
12881 PL_collation_standard = proto_perl->Icollation_standard;
12882 PL_collxfrm_base = proto_perl->Icollxfrm_base;
12883 PL_collxfrm_mult = proto_perl->Icollxfrm_mult;
12884 #endif /* USE_LOCALE_COLLATE */
12886 #ifdef USE_LOCALE_NUMERIC
12887 PL_numeric_standard = proto_perl->Inumeric_standard;
12888 PL_numeric_local = proto_perl->Inumeric_local;
12889 #endif /* !USE_LOCALE_NUMERIC */
12891 /* Did the locale setup indicate UTF-8? */
12892 PL_utf8locale = proto_perl->Iutf8locale;
12893 /* Unicode features (see perlrun/-C) */
12894 PL_unicode = proto_perl->Iunicode;
12896 /* Pre-5.8 signals control */
12897 PL_signals = proto_perl->Isignals;
12899 /* times() ticks per second */
12900 PL_clocktick = proto_perl->Iclocktick;
12902 /* Recursion stopper for PerlIO_find_layer */
12903 PL_in_load_module = proto_perl->Iin_load_module;
12905 /* sort() routine */
12906 PL_sort_RealCmp = proto_perl->Isort_RealCmp;
12908 /* Not really needed/useful since the reenrant_retint is "volatile",
12909 * but do it for consistency's sake. */
12910 PL_reentrant_retint = proto_perl->Ireentrant_retint;
12912 /* Hooks to shared SVs and locks. */
12913 PL_sharehook = proto_perl->Isharehook;
12914 PL_lockhook = proto_perl->Ilockhook;
12915 PL_unlockhook = proto_perl->Iunlockhook;
12916 PL_threadhook = proto_perl->Ithreadhook;
12917 PL_destroyhook = proto_perl->Idestroyhook;
12918 PL_signalhook = proto_perl->Isignalhook;
12920 #ifdef THREADS_HAVE_PIDS
12921 PL_ppid = proto_perl->Ippid;
12925 PL_last_swash_hv = NULL; /* reinits on demand */
12926 PL_last_swash_klen = 0;
12927 PL_last_swash_key[0]= '\0';
12928 PL_last_swash_tmps = (U8*)NULL;
12929 PL_last_swash_slen = 0;
12931 PL_glob_index = proto_perl->Iglob_index;
12932 PL_srand_called = proto_perl->Isrand_called;
12934 if (flags & CLONEf_COPY_STACKS) {
12935 /* next allocation will be PL_tmps_stack[PL_tmps_ix+1] */
12936 PL_tmps_ix = proto_perl->Itmps_ix;
12937 PL_tmps_max = proto_perl->Itmps_max;
12938 PL_tmps_floor = proto_perl->Itmps_floor;
12940 /* next push_scope()/ENTER sets PL_scopestack[PL_scopestack_ix]
12941 * NOTE: unlike the others! */
12942 PL_scopestack_ix = proto_perl->Iscopestack_ix;
12943 PL_scopestack_max = proto_perl->Iscopestack_max;
12945 /* next SSPUSHFOO() sets PL_savestack[PL_savestack_ix]
12946 * NOTE: unlike the others! */
12947 PL_savestack_ix = proto_perl->Isavestack_ix;
12948 PL_savestack_max = proto_perl->Isavestack_max;
12951 PL_start_env = proto_perl->Istart_env; /* XXXXXX */
12952 PL_top_env = &PL_start_env;
12954 PL_op = proto_perl->Iop;
12957 PL_Xpv = (XPV*)NULL;
12958 my_perl->Ina = proto_perl->Ina;
12960 PL_statbuf = proto_perl->Istatbuf;
12961 PL_statcache = proto_perl->Istatcache;
12964 PL_timesbuf = proto_perl->Itimesbuf;
12967 PL_tainted = proto_perl->Itainted;
12968 PL_curpm = proto_perl->Icurpm; /* XXX No PMOP ref count */
12970 PL_chopset = proto_perl->Ichopset; /* XXX never deallocated */
12972 PL_restartjmpenv = proto_perl->Irestartjmpenv;
12973 PL_restartop = proto_perl->Irestartop;
12974 PL_in_eval = proto_perl->Iin_eval;
12975 PL_delaymagic = proto_perl->Idelaymagic;
12976 PL_phase = proto_perl->Iphase;
12977 PL_localizing = proto_perl->Ilocalizing;
12979 PL_hv_fetch_ent_mh = NULL;
12980 PL_modcount = proto_perl->Imodcount;
12981 PL_lastgotoprobe = NULL;
12982 PL_dumpindent = proto_perl->Idumpindent;
12984 PL_efloatbuf = NULL; /* reinits on demand */
12985 PL_efloatsize = 0; /* reinits on demand */
12989 PL_regdummy = proto_perl->Iregdummy;
12990 PL_colorset = 0; /* reinits PL_colors[] */
12991 /*PL_colors[6] = {0,0,0,0,0,0};*/
12993 /* Pluggable optimizer */
12994 PL_peepp = proto_perl->Ipeepp;
12995 PL_rpeepp = proto_perl->Irpeepp;
12996 /* op_free() hook */
12997 PL_opfreehook = proto_perl->Iopfreehook;
12999 #ifdef USE_REENTRANT_API
13000 /* XXX: things like -Dm will segfault here in perlio, but doing
13001 * PERL_SET_CONTEXT(proto_perl);
13002 * breaks too many other things
13004 Perl_reentrant_init(aTHX);
13007 /* create SV map for pointer relocation */
13008 PL_ptr_table = ptr_table_new();
13010 /* initialize these special pointers as early as possible */
13011 ptr_table_store(PL_ptr_table, &proto_perl->Isv_undef, &PL_sv_undef);
13013 SvANY(&PL_sv_no) = new_XPVNV();
13014 SvPV_set(&PL_sv_no, savepvn(PL_No, 0));
13015 SvCUR_set(&PL_sv_no, 0);
13016 SvLEN_set(&PL_sv_no, 1);
13017 SvIV_set(&PL_sv_no, 0);
13018 SvNV_set(&PL_sv_no, 0);
13019 ptr_table_store(PL_ptr_table, &proto_perl->Isv_no, &PL_sv_no);
13021 SvPV_set(&PL_sv_yes, savepvn(PL_Yes, 1));
13022 SvCUR_set(&PL_sv_yes, 1);
13023 SvLEN_set(&PL_sv_yes, 2);
13024 SvIV_set(&PL_sv_yes, 1);
13025 SvNV_set(&PL_sv_yes, 1);
13026 ptr_table_store(PL_ptr_table, &proto_perl->Isv_yes, &PL_sv_yes);
13028 /* create (a non-shared!) shared string table */
13029 PL_strtab = newHV();
13030 HvSHAREKEYS_off(PL_strtab);
13031 hv_ksplit(PL_strtab, HvTOTALKEYS(proto_perl->Istrtab));
13032 ptr_table_store(PL_ptr_table, proto_perl->Istrtab, PL_strtab);
13034 /* These two PVs will be free'd special way so must set them same way op.c does */
13035 PL_compiling.cop_stashpv = savesharedpv(PL_compiling.cop_stashpv);
13036 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_stashpv, PL_compiling.cop_stashpv);
13038 PL_compiling.cop_file = savesharedpv(PL_compiling.cop_file);
13039 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_file, PL_compiling.cop_file);
13041 ptr_table_store(PL_ptr_table, &proto_perl->Icompiling, &PL_compiling);
13042 PL_compiling.cop_warnings = DUP_WARNINGS(PL_compiling.cop_warnings);
13043 CopHINTHASH_set(&PL_compiling, cophh_copy(CopHINTHASH_get(&PL_compiling)));
13044 PL_curcop = (COP*)any_dup(proto_perl->Icurcop, proto_perl);
13046 param->stashes = newAV(); /* Setup array of objects to call clone on */
13047 /* This makes no difference to the implementation, as it always pushes
13048 and shifts pointers to other SVs without changing their reference
13049 count, with the array becoming empty before it is freed. However, it
13050 makes it conceptually clear what is going on, and will avoid some
13051 work inside av.c, filling slots between AvFILL() and AvMAX() with
13052 &PL_sv_undef, and SvREFCNT_dec()ing those. */
13053 AvREAL_off(param->stashes);
13055 if (!(flags & CLONEf_COPY_STACKS)) {
13056 param->unreferenced = newAV();
13059 #ifdef PERLIO_LAYERS
13060 /* Clone PerlIO tables as soon as we can handle general xx_dup() */
13061 PerlIO_clone(aTHX_ proto_perl, param);
13064 PL_envgv = gv_dup(proto_perl->Ienvgv, param);
13065 PL_incgv = gv_dup(proto_perl->Iincgv, param);
13066 PL_hintgv = gv_dup(proto_perl->Ihintgv, param);
13067 PL_origfilename = SAVEPV(proto_perl->Iorigfilename);
13068 PL_diehook = sv_dup_inc(proto_perl->Idiehook, param);
13069 PL_warnhook = sv_dup_inc(proto_perl->Iwarnhook, param);
13072 PL_patchlevel = sv_dup_inc(proto_perl->Ipatchlevel, param);
13073 PL_apiversion = sv_dup_inc(proto_perl->Iapiversion, param);
13074 PL_inplace = SAVEPV(proto_perl->Iinplace);
13075 PL_e_script = sv_dup_inc(proto_perl->Ie_script, param);
13077 /* magical thingies */
13078 PL_formfeed = sv_dup(proto_perl->Iformfeed, param);
13080 PL_encoding = sv_dup(proto_perl->Iencoding, param);
13082 sv_setpvs(PERL_DEBUG_PAD(0), ""); /* For regex debugging. */
13083 sv_setpvs(PERL_DEBUG_PAD(1), ""); /* ext/re needs these */
13084 sv_setpvs(PERL_DEBUG_PAD(2), ""); /* even without DEBUGGING. */
13087 /* Clone the regex array */
13088 /* ORANGE FIXME for plugins, probably in the SV dup code.
13089 newSViv(PTR2IV(CALLREGDUPE(
13090 INT2PTR(REGEXP *, SvIVX(regex)), param))))
13092 PL_regex_padav = av_dup_inc(proto_perl->Iregex_padav, param);
13093 PL_regex_pad = AvARRAY(PL_regex_padav);
13095 /* shortcuts to various I/O objects */
13096 PL_ofsgv = gv_dup_inc(proto_perl->Iofsgv, param);
13097 PL_stdingv = gv_dup(proto_perl->Istdingv, param);
13098 PL_stderrgv = gv_dup(proto_perl->Istderrgv, param);
13099 PL_defgv = gv_dup(proto_perl->Idefgv, param);
13100 PL_argvgv = gv_dup(proto_perl->Iargvgv, param);
13101 PL_argvoutgv = gv_dup(proto_perl->Iargvoutgv, param);
13102 PL_argvout_stack = av_dup_inc(proto_perl->Iargvout_stack, param);
13104 /* shortcuts to regexp stuff */
13105 PL_replgv = gv_dup(proto_perl->Ireplgv, param);
13107 /* shortcuts to misc objects */
13108 PL_errgv = gv_dup(proto_perl->Ierrgv, param);
13110 /* shortcuts to debugging objects */
13111 PL_DBgv = gv_dup(proto_perl->IDBgv, param);
13112 PL_DBline = gv_dup(proto_perl->IDBline, param);
13113 PL_DBsub = gv_dup(proto_perl->IDBsub, param);
13114 PL_DBsingle = sv_dup(proto_perl->IDBsingle, param);
13115 PL_DBtrace = sv_dup(proto_perl->IDBtrace, param);
13116 PL_DBsignal = sv_dup(proto_perl->IDBsignal, param);
13118 /* symbol tables */
13119 PL_defstash = hv_dup_inc(proto_perl->Idefstash, param);
13120 PL_curstash = hv_dup(proto_perl->Icurstash, param);
13121 PL_debstash = hv_dup(proto_perl->Idebstash, param);
13122 PL_globalstash = hv_dup(proto_perl->Iglobalstash, param);
13123 PL_curstname = sv_dup_inc(proto_perl->Icurstname, param);
13125 PL_beginav = av_dup_inc(proto_perl->Ibeginav, param);
13126 PL_beginav_save = av_dup_inc(proto_perl->Ibeginav_save, param);
13127 PL_checkav_save = av_dup_inc(proto_perl->Icheckav_save, param);
13128 PL_unitcheckav = av_dup_inc(proto_perl->Iunitcheckav, param);
13129 PL_unitcheckav_save = av_dup_inc(proto_perl->Iunitcheckav_save, param);
13130 PL_endav = av_dup_inc(proto_perl->Iendav, param);
13131 PL_checkav = av_dup_inc(proto_perl->Icheckav, param);
13132 PL_initav = av_dup_inc(proto_perl->Iinitav, param);
13134 PL_isarev = hv_dup_inc(proto_perl->Iisarev, param);
13136 /* subprocess state */
13137 PL_fdpid = av_dup_inc(proto_perl->Ifdpid, param);
13139 if (proto_perl->Iop_mask)
13140 PL_op_mask = SAVEPVN(proto_perl->Iop_mask, PL_maxo);
13143 /* PL_asserting = proto_perl->Iasserting; */
13145 /* current interpreter roots */
13146 PL_main_cv = cv_dup_inc(proto_perl->Imain_cv, param);
13148 PL_main_root = OpREFCNT_inc(proto_perl->Imain_root);
13151 /* runtime control stuff */
13152 PL_curcopdb = (COP*)any_dup(proto_perl->Icurcopdb, proto_perl);
13154 PL_preambleav = av_dup_inc(proto_perl->Ipreambleav, param);
13156 PL_ors_sv = sv_dup_inc(proto_perl->Iors_sv, param);
13158 /* interpreter atexit processing */
13159 PL_exitlistlen = proto_perl->Iexitlistlen;
13160 if (PL_exitlistlen) {
13161 Newx(PL_exitlist, PL_exitlistlen, PerlExitListEntry);
13162 Copy(proto_perl->Iexitlist, PL_exitlist, PL_exitlistlen, PerlExitListEntry);
13165 PL_exitlist = (PerlExitListEntry*)NULL;
13167 PL_my_cxt_size = proto_perl->Imy_cxt_size;
13168 if (PL_my_cxt_size) {
13169 Newx(PL_my_cxt_list, PL_my_cxt_size, void *);
13170 Copy(proto_perl->Imy_cxt_list, PL_my_cxt_list, PL_my_cxt_size, void *);
13171 #ifdef PERL_GLOBAL_STRUCT_PRIVATE
13172 Newx(PL_my_cxt_keys, PL_my_cxt_size, const char *);
13173 Copy(proto_perl->Imy_cxt_keys, PL_my_cxt_keys, PL_my_cxt_size, char *);
13177 PL_my_cxt_list = (void**)NULL;
13178 #ifdef PERL_GLOBAL_STRUCT_PRIVATE
13179 PL_my_cxt_keys = (const char**)NULL;
13182 PL_modglobal = hv_dup_inc(proto_perl->Imodglobal, param);
13183 PL_custom_op_names = hv_dup_inc(proto_perl->Icustom_op_names,param);
13184 PL_custom_op_descs = hv_dup_inc(proto_perl->Icustom_op_descs,param);
13185 PL_custom_ops = hv_dup_inc(proto_perl->Icustom_ops, param);
13187 PL_compcv = cv_dup(proto_perl->Icompcv, param);
13189 PAD_CLONE_VARS(proto_perl, param);
13191 #ifdef HAVE_INTERP_INTERN
13192 sys_intern_dup(&proto_perl->Isys_intern, &PL_sys_intern);
13195 PL_DBcv = cv_dup(proto_perl->IDBcv, param);
13197 #ifdef PERL_USES_PL_PIDSTATUS
13198 PL_pidstatus = newHV(); /* XXX flag for cloning? */
13200 PL_osname = SAVEPV(proto_perl->Iosname);
13201 PL_parser = parser_dup(proto_perl->Iparser, param);
13203 /* XXX this only works if the saved cop has already been cloned */
13204 if (proto_perl->Iparser) {
13205 PL_parser->saved_curcop = (COP*)any_dup(
13206 proto_perl->Iparser->saved_curcop,
13210 PL_subname = sv_dup_inc(proto_perl->Isubname, param);
13212 #ifdef USE_LOCALE_COLLATE
13213 PL_collation_name = SAVEPV(proto_perl->Icollation_name);
13214 #endif /* USE_LOCALE_COLLATE */
13216 #ifdef USE_LOCALE_NUMERIC
13217 PL_numeric_name = SAVEPV(proto_perl->Inumeric_name);
13218 PL_numeric_radix_sv = sv_dup_inc(proto_perl->Inumeric_radix_sv, param);
13219 #endif /* !USE_LOCALE_NUMERIC */
13221 /* utf8 character classes */
13222 PL_utf8_alnum = sv_dup_inc(proto_perl->Iutf8_alnum, param);
13223 PL_utf8_ascii = sv_dup_inc(proto_perl->Iutf8_ascii, param);
13224 PL_utf8_alpha = sv_dup_inc(proto_perl->Iutf8_alpha, param);
13225 PL_utf8_space = sv_dup_inc(proto_perl->Iutf8_space, param);
13226 PL_utf8_cntrl = sv_dup_inc(proto_perl->Iutf8_cntrl, param);
13227 PL_utf8_graph = sv_dup_inc(proto_perl->Iutf8_graph, param);
13228 PL_utf8_digit = sv_dup_inc(proto_perl->Iutf8_digit, param);
13229 PL_utf8_upper = sv_dup_inc(proto_perl->Iutf8_upper, param);
13230 PL_utf8_lower = sv_dup_inc(proto_perl->Iutf8_lower, param);
13231 PL_utf8_print = sv_dup_inc(proto_perl->Iutf8_print, param);
13232 PL_utf8_punct = sv_dup_inc(proto_perl->Iutf8_punct, param);
13233 PL_utf8_xdigit = sv_dup_inc(proto_perl->Iutf8_xdigit, param);
13234 PL_utf8_mark = sv_dup_inc(proto_perl->Iutf8_mark, param);
13235 PL_utf8_X_begin = sv_dup_inc(proto_perl->Iutf8_X_begin, param);
13236 PL_utf8_X_extend = sv_dup_inc(proto_perl->Iutf8_X_extend, param);
13237 PL_utf8_X_prepend = sv_dup_inc(proto_perl->Iutf8_X_prepend, param);
13238 PL_utf8_X_non_hangul = sv_dup_inc(proto_perl->Iutf8_X_non_hangul, param);
13239 PL_utf8_X_L = sv_dup_inc(proto_perl->Iutf8_X_L, param);
13240 PL_utf8_X_LV = sv_dup_inc(proto_perl->Iutf8_X_LV, param);
13241 PL_utf8_X_LVT = sv_dup_inc(proto_perl->Iutf8_X_LVT, param);
13242 PL_utf8_X_T = sv_dup_inc(proto_perl->Iutf8_X_T, param);
13243 PL_utf8_X_V = sv_dup_inc(proto_perl->Iutf8_X_V, param);
13244 PL_utf8_X_LV_LVT_V = sv_dup_inc(proto_perl->Iutf8_X_LV_LVT_V, param);
13245 PL_utf8_toupper = sv_dup_inc(proto_perl->Iutf8_toupper, param);
13246 PL_utf8_totitle = sv_dup_inc(proto_perl->Iutf8_totitle, param);
13247 PL_utf8_tolower = sv_dup_inc(proto_perl->Iutf8_tolower, param);
13248 PL_utf8_tofold = sv_dup_inc(proto_perl->Iutf8_tofold, param);
13249 PL_utf8_idstart = sv_dup_inc(proto_perl->Iutf8_idstart, param);
13250 PL_utf8_xidstart = sv_dup_inc(proto_perl->Iutf8_xidstart, param);
13251 PL_utf8_idcont = sv_dup_inc(proto_perl->Iutf8_idcont, param);
13252 PL_utf8_xidcont = sv_dup_inc(proto_perl->Iutf8_xidcont, param);
13253 PL_utf8_foldable = sv_dup_inc(proto_perl->Iutf8_foldable, param);
13256 if (proto_perl->Ipsig_pend) {
13257 Newxz(PL_psig_pend, SIG_SIZE, int);
13260 PL_psig_pend = (int*)NULL;
13263 if (proto_perl->Ipsig_name) {
13264 Newx(PL_psig_name, 2 * SIG_SIZE, SV*);
13265 sv_dup_inc_multiple(proto_perl->Ipsig_name, PL_psig_name, 2 * SIG_SIZE,
13267 PL_psig_ptr = PL_psig_name + SIG_SIZE;
13270 PL_psig_ptr = (SV**)NULL;
13271 PL_psig_name = (SV**)NULL;
13274 if (flags & CLONEf_COPY_STACKS) {
13275 Newx(PL_tmps_stack, PL_tmps_max, SV*);
13276 sv_dup_inc_multiple(proto_perl->Itmps_stack, PL_tmps_stack,
13277 PL_tmps_ix+1, param);
13279 /* next PUSHMARK() sets *(PL_markstack_ptr+1) */
13280 i = proto_perl->Imarkstack_max - proto_perl->Imarkstack;
13281 Newxz(PL_markstack, i, I32);
13282 PL_markstack_max = PL_markstack + (proto_perl->Imarkstack_max
13283 - proto_perl->Imarkstack);
13284 PL_markstack_ptr = PL_markstack + (proto_perl->Imarkstack_ptr
13285 - proto_perl->Imarkstack);
13286 Copy(proto_perl->Imarkstack, PL_markstack,
13287 PL_markstack_ptr - PL_markstack + 1, I32);
13289 /* next push_scope()/ENTER sets PL_scopestack[PL_scopestack_ix]
13290 * NOTE: unlike the others! */
13291 Newxz(PL_scopestack, PL_scopestack_max, I32);
13292 Copy(proto_perl->Iscopestack, PL_scopestack, PL_scopestack_ix, I32);
13295 Newxz(PL_scopestack_name, PL_scopestack_max, const char *);
13296 Copy(proto_perl->Iscopestack_name, PL_scopestack_name, PL_scopestack_ix, const char *);
13298 /* NOTE: si_dup() looks at PL_markstack */
13299 PL_curstackinfo = si_dup(proto_perl->Icurstackinfo, param);
13301 /* PL_curstack = PL_curstackinfo->si_stack; */
13302 PL_curstack = av_dup(proto_perl->Icurstack, param);
13303 PL_mainstack = av_dup(proto_perl->Imainstack, param);
13305 /* next PUSHs() etc. set *(PL_stack_sp+1) */
13306 PL_stack_base = AvARRAY(PL_curstack);
13307 PL_stack_sp = PL_stack_base + (proto_perl->Istack_sp
13308 - proto_perl->Istack_base);
13309 PL_stack_max = PL_stack_base + AvMAX(PL_curstack);
13311 /*Newxz(PL_savestack, PL_savestack_max, ANY);*/
13312 PL_savestack = ss_dup(proto_perl, param);
13316 ENTER; /* perl_destruct() wants to LEAVE; */
13319 PL_statgv = gv_dup(proto_perl->Istatgv, param);
13320 PL_statname = sv_dup_inc(proto_perl->Istatname, param);
13322 PL_rs = sv_dup_inc(proto_perl->Irs, param);
13323 PL_last_in_gv = gv_dup(proto_perl->Ilast_in_gv, param);
13324 PL_defoutgv = gv_dup_inc(proto_perl->Idefoutgv, param);
13325 PL_toptarget = sv_dup_inc(proto_perl->Itoptarget, param);
13326 PL_bodytarget = sv_dup_inc(proto_perl->Ibodytarget, param);
13327 PL_formtarget = sv_dup(proto_perl->Iformtarget, param);
13329 PL_errors = sv_dup_inc(proto_perl->Ierrors, param);
13331 PL_sortcop = (OP*)any_dup(proto_perl->Isortcop, proto_perl);
13332 PL_sortstash = hv_dup(proto_perl->Isortstash, param);
13333 PL_firstgv = gv_dup(proto_perl->Ifirstgv, param);
13334 PL_secondgv = gv_dup(proto_perl->Isecondgv, param);
13336 PL_stashcache = newHV();
13338 PL_watchaddr = (char **) ptr_table_fetch(PL_ptr_table,
13339 proto_perl->Iwatchaddr);
13340 PL_watchok = PL_watchaddr ? * PL_watchaddr : NULL;
13341 if (PL_debug && PL_watchaddr) {
13342 PerlIO_printf(Perl_debug_log,
13343 "WATCHING: %"UVxf" cloned as %"UVxf" with value %"UVxf"\n",
13344 PTR2UV(proto_perl->Iwatchaddr), PTR2UV(PL_watchaddr),
13345 PTR2UV(PL_watchok));
13348 PL_registered_mros = hv_dup_inc(proto_perl->Iregistered_mros, param);
13349 PL_blockhooks = av_dup_inc(proto_perl->Iblockhooks, param);
13350 PL_utf8_foldclosures = hv_dup_inc(proto_perl->Iutf8_foldclosures, param);
13352 /* Call the ->CLONE method, if it exists, for each of the stashes
13353 identified by sv_dup() above.
13355 while(av_len(param->stashes) != -1) {
13356 HV* const stash = MUTABLE_HV(av_shift(param->stashes));
13357 GV* const cloner = gv_fetchmethod_autoload(stash, "CLONE", 0);
13358 if (cloner && GvCV(cloner)) {
13363 mXPUSHs(newSVhek(HvNAME_HEK(stash)));
13365 call_sv(MUTABLE_SV(GvCV(cloner)), G_DISCARD);
13371 if (!(flags & CLONEf_KEEP_PTR_TABLE)) {
13372 ptr_table_free(PL_ptr_table);
13373 PL_ptr_table = NULL;
13376 if (!(flags & CLONEf_COPY_STACKS)) {
13377 unreferenced_to_tmp_stack(param->unreferenced);
13380 SvREFCNT_dec(param->stashes);
13382 /* orphaned? eg threads->new inside BEGIN or use */
13383 if (PL_compcv && ! SvREFCNT(PL_compcv)) {
13384 SvREFCNT_inc_simple_void(PL_compcv);
13385 SAVEFREESV(PL_compcv);
13392 S_unreferenced_to_tmp_stack(pTHX_ AV *const unreferenced)
13394 PERL_ARGS_ASSERT_UNREFERENCED_TO_TMP_STACK;
13396 if (AvFILLp(unreferenced) > -1) {
13397 SV **svp = AvARRAY(unreferenced);
13398 SV **const last = svp + AvFILLp(unreferenced);
13402 if (SvREFCNT(*svp) == 1)
13404 } while (++svp <= last);
13406 EXTEND_MORTAL(count);
13407 svp = AvARRAY(unreferenced);
13410 if (SvREFCNT(*svp) == 1) {
13411 /* Our reference is the only one to this SV. This means that
13412 in this thread, the scalar effectively has a 0 reference.
13413 That doesn't work (cleanup never happens), so donate our
13414 reference to it onto the save stack. */
13415 PL_tmps_stack[++PL_tmps_ix] = *svp;
13417 /* As an optimisation, because we are already walking the
13418 entire array, instead of above doing either
13419 SvREFCNT_inc(*svp) or *svp = &PL_sv_undef, we can instead
13420 release our reference to the scalar, so that at the end of
13421 the array owns zero references to the scalars it happens to
13422 point to. We are effectively converting the array from
13423 AvREAL() on to AvREAL() off. This saves the av_clear()
13424 (triggered by the SvREFCNT_dec(unreferenced) below) from
13425 walking the array a second time. */
13426 SvREFCNT_dec(*svp);
13429 } while (++svp <= last);
13430 AvREAL_off(unreferenced);
13432 SvREFCNT_dec(unreferenced);
13436 Perl_clone_params_del(CLONE_PARAMS *param)
13438 /* This seemingly funky ordering keeps the build with PERL_GLOBAL_STRUCT
13440 PerlInterpreter *const to = param->new_perl;
13442 PerlInterpreter *const was = PERL_GET_THX;
13444 PERL_ARGS_ASSERT_CLONE_PARAMS_DEL;
13450 SvREFCNT_dec(param->stashes);
13451 if (param->unreferenced)
13452 unreferenced_to_tmp_stack(param->unreferenced);
13462 Perl_clone_params_new(PerlInterpreter *const from, PerlInterpreter *const to)
13465 /* Need to play this game, as newAV() can call safesysmalloc(), and that
13466 does a dTHX; to get the context from thread local storage.
13467 FIXME - under PERL_CORE Newx(), Safefree() and friends should expand to
13468 a version that passes in my_perl. */
13469 PerlInterpreter *const was = PERL_GET_THX;
13470 CLONE_PARAMS *param;
13472 PERL_ARGS_ASSERT_CLONE_PARAMS_NEW;
13478 /* Given that we've set the context, we can do this unshared. */
13479 Newx(param, 1, CLONE_PARAMS);
13482 param->proto_perl = from;
13483 param->new_perl = to;
13484 param->stashes = (AV *)Perl_newSV_type(to, SVt_PVAV);
13485 AvREAL_off(param->stashes);
13486 param->unreferenced = (AV *)Perl_newSV_type(to, SVt_PVAV);
13494 #endif /* USE_ITHREADS */
13497 =head1 Unicode Support
13499 =for apidoc sv_recode_to_utf8
13501 The encoding is assumed to be an Encode object, on entry the PV
13502 of the sv is assumed to be octets in that encoding, and the sv
13503 will be converted into Unicode (and UTF-8).
13505 If the sv already is UTF-8 (or if it is not POK), or if the encoding
13506 is not a reference, nothing is done to the sv. If the encoding is not
13507 an C<Encode::XS> Encoding object, bad things will happen.
13508 (See F<lib/encoding.pm> and L<Encode>).
13510 The PV of the sv is returned.
13515 Perl_sv_recode_to_utf8(pTHX_ SV *sv, SV *encoding)
13519 PERL_ARGS_ASSERT_SV_RECODE_TO_UTF8;
13521 if (SvPOK(sv) && !SvUTF8(sv) && !IN_BYTES && SvROK(encoding)) {
13535 Passing sv_yes is wrong - it needs to be or'ed set of constants
13536 for Encode::XS, while UTf-8 decode (currently) assumes a true value means
13537 remove converted chars from source.
13539 Both will default the value - let them.
13541 XPUSHs(&PL_sv_yes);
13544 call_method("decode", G_SCALAR);
13548 s = SvPV_const(uni, len);
13549 if (s != SvPVX_const(sv)) {
13550 SvGROW(sv, len + 1);
13551 Move(s, SvPVX(sv), len + 1, char);
13552 SvCUR_set(sv, len);
13556 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
13557 /* clear pos and any utf8 cache */
13558 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
13561 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
13562 magic_setutf8(sv,mg); /* clear UTF8 cache */
13567 return SvPOKp(sv) ? SvPVX(sv) : NULL;
13571 =for apidoc sv_cat_decode
13573 The encoding is assumed to be an Encode object, the PV of the ssv is
13574 assumed to be octets in that encoding and decoding the input starts
13575 from the position which (PV + *offset) pointed to. The dsv will be
13576 concatenated the decoded UTF-8 string from ssv. Decoding will terminate
13577 when the string tstr appears in decoding output or the input ends on
13578 the PV of the ssv. The value which the offset points will be modified
13579 to the last input position on the ssv.
13581 Returns TRUE if the terminator was found, else returns FALSE.
13586 Perl_sv_cat_decode(pTHX_ SV *dsv, SV *encoding,
13587 SV *ssv, int *offset, char *tstr, int tlen)
13592 PERL_ARGS_ASSERT_SV_CAT_DECODE;
13594 if (SvPOK(ssv) && SvPOK(dsv) && SvROK(encoding) && offset) {
13605 offsv = newSViv(*offset);
13607 mXPUSHp(tstr, tlen);
13609 call_method("cat_decode", G_SCALAR);
13611 ret = SvTRUE(TOPs);
13612 *offset = SvIV(offsv);
13618 Perl_croak(aTHX_ "Invalid argument to sv_cat_decode");
13623 /* ---------------------------------------------------------------------
13625 * support functions for report_uninit()
13628 /* the maxiumum size of array or hash where we will scan looking
13629 * for the undefined element that triggered the warning */
13631 #define FUV_MAX_SEARCH_SIZE 1000
13633 /* Look for an entry in the hash whose value has the same SV as val;
13634 * If so, return a mortal copy of the key. */
13637 S_find_hash_subscript(pTHX_ const HV *const hv, const SV *const val)
13640 register HE **array;
13643 PERL_ARGS_ASSERT_FIND_HASH_SUBSCRIPT;
13645 if (!hv || SvMAGICAL(hv) || !HvARRAY(hv) ||
13646 (HvTOTALKEYS(hv) > FUV_MAX_SEARCH_SIZE))
13649 array = HvARRAY(hv);
13651 for (i=HvMAX(hv); i>0; i--) {
13652 register HE *entry;
13653 for (entry = array[i]; entry; entry = HeNEXT(entry)) {
13654 if (HeVAL(entry) != val)
13656 if ( HeVAL(entry) == &PL_sv_undef ||
13657 HeVAL(entry) == &PL_sv_placeholder)
13661 if (HeKLEN(entry) == HEf_SVKEY)
13662 return sv_mortalcopy(HeKEY_sv(entry));
13663 return sv_2mortal(newSVhek(HeKEY_hek(entry)));
13669 /* Look for an entry in the array whose value has the same SV as val;
13670 * If so, return the index, otherwise return -1. */
13673 S_find_array_subscript(pTHX_ const AV *const av, const SV *const val)
13677 PERL_ARGS_ASSERT_FIND_ARRAY_SUBSCRIPT;
13679 if (!av || SvMAGICAL(av) || !AvARRAY(av) ||
13680 (AvFILLp(av) > FUV_MAX_SEARCH_SIZE))
13683 if (val != &PL_sv_undef) {
13684 SV ** const svp = AvARRAY(av);
13687 for (i=AvFILLp(av); i>=0; i--)
13694 /* S_varname(): return the name of a variable, optionally with a subscript.
13695 * If gv is non-zero, use the name of that global, along with gvtype (one
13696 * of "$", "@", "%"); otherwise use the name of the lexical at pad offset
13697 * targ. Depending on the value of the subscript_type flag, return:
13700 #define FUV_SUBSCRIPT_NONE 1 /* "@foo" */
13701 #define FUV_SUBSCRIPT_ARRAY 2 /* "$foo[aindex]" */
13702 #define FUV_SUBSCRIPT_HASH 3 /* "$foo{keyname}" */
13703 #define FUV_SUBSCRIPT_WITHIN 4 /* "within @foo" */
13706 S_varname(pTHX_ const GV *const gv, const char gvtype, PADOFFSET targ,
13707 const SV *const keyname, I32 aindex, int subscript_type)
13710 SV * const name = sv_newmortal();
13713 buffer[0] = gvtype;
13716 /* as gv_fullname4(), but add literal '^' for $^FOO names */
13718 gv_fullname4(name, gv, buffer, 0);
13720 if ((unsigned int)SvPVX(name)[1] <= 26) {
13722 buffer[1] = SvPVX(name)[1] + 'A' - 1;
13724 /* Swap the 1 unprintable control character for the 2 byte pretty
13725 version - ie substr($name, 1, 1) = $buffer; */
13726 sv_insert(name, 1, 1, buffer, 2);
13730 CV * const cv = find_runcv(NULL);
13734 if (!cv || !CvPADLIST(cv))
13736 av = MUTABLE_AV((*av_fetch(CvPADLIST(cv), 0, FALSE)));
13737 sv = *av_fetch(av, targ, FALSE);
13738 sv_setpvn(name, SvPV_nolen_const(sv), SvCUR(sv));
13741 if (subscript_type == FUV_SUBSCRIPT_HASH) {
13742 SV * const sv = newSV(0);
13743 *SvPVX(name) = '$';
13744 Perl_sv_catpvf(aTHX_ name, "{%s}",
13745 pv_display(sv,SvPVX_const(keyname), SvCUR(keyname), 0, 32));
13748 else if (subscript_type == FUV_SUBSCRIPT_ARRAY) {
13749 *SvPVX(name) = '$';
13750 Perl_sv_catpvf(aTHX_ name, "[%"IVdf"]", (IV)aindex);
13752 else if (subscript_type == FUV_SUBSCRIPT_WITHIN) {
13753 /* We know that name has no magic, so can use 0 instead of SV_GMAGIC */
13754 Perl_sv_insert_flags(aTHX_ name, 0, 0, STR_WITH_LEN("within "), 0);
13762 =for apidoc find_uninit_var
13764 Find the name of the undefined variable (if any) that caused the operator o
13765 to issue a "Use of uninitialized value" warning.
13766 If match is true, only return a name if it's value matches uninit_sv.
13767 So roughly speaking, if a unary operator (such as OP_COS) generates a
13768 warning, then following the direct child of the op may yield an
13769 OP_PADSV or OP_GV that gives the name of the undefined variable. On the
13770 other hand, with OP_ADD there are two branches to follow, so we only print
13771 the variable name if we get an exact match.
13773 The name is returned as a mortal SV.
13775 Assumes that PL_op is the op that originally triggered the error, and that
13776 PL_comppad/PL_curpad points to the currently executing pad.
13782 S_find_uninit_var(pTHX_ const OP *const obase, const SV *const uninit_sv,
13788 const OP *o, *o2, *kid;
13790 if (!obase || (match && (!uninit_sv || uninit_sv == &PL_sv_undef ||
13791 uninit_sv == &PL_sv_placeholder)))
13794 switch (obase->op_type) {
13801 const bool pad = (obase->op_type == OP_PADAV || obase->op_type == OP_PADHV);
13802 const bool hash = (obase->op_type == OP_PADHV || obase->op_type == OP_RV2HV);
13805 int subscript_type = FUV_SUBSCRIPT_WITHIN;
13807 if (pad) { /* @lex, %lex */
13808 sv = PAD_SVl(obase->op_targ);
13812 if (cUNOPx(obase)->op_first->op_type == OP_GV) {
13813 /* @global, %global */
13814 gv = cGVOPx_gv(cUNOPx(obase)->op_first);
13817 sv = hash ? MUTABLE_SV(GvHV(gv)): MUTABLE_SV(GvAV(gv));
13819 else /* @{expr}, %{expr} */
13820 return find_uninit_var(cUNOPx(obase)->op_first,
13824 /* attempt to find a match within the aggregate */
13826 keysv = find_hash_subscript((const HV*)sv, uninit_sv);
13828 subscript_type = FUV_SUBSCRIPT_HASH;
13831 index = find_array_subscript((const AV *)sv, uninit_sv);
13833 subscript_type = FUV_SUBSCRIPT_ARRAY;
13836 if (match && subscript_type == FUV_SUBSCRIPT_WITHIN)
13839 return varname(gv, hash ? '%' : '@', obase->op_targ,
13840 keysv, index, subscript_type);
13844 if (match && PAD_SVl(obase->op_targ) != uninit_sv)
13846 return varname(NULL, '$', obase->op_targ,
13847 NULL, 0, FUV_SUBSCRIPT_NONE);
13850 gv = cGVOPx_gv(obase);
13851 if (!gv || (match && GvSV(gv) != uninit_sv) || !GvSTASH(gv))
13853 return varname(gv, '$', 0, NULL, 0, FUV_SUBSCRIPT_NONE);
13855 case OP_AELEMFAST_LEX:
13858 AV *av = MUTABLE_AV(PAD_SV(obase->op_targ));
13859 if (!av || SvRMAGICAL(av))
13861 svp = av_fetch(av, (I32)obase->op_private, FALSE);
13862 if (!svp || *svp != uninit_sv)
13865 return varname(NULL, '$', obase->op_targ,
13866 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
13869 gv = cGVOPx_gv(obase);
13874 AV *const av = GvAV(gv);
13875 if (!av || SvRMAGICAL(av))
13877 svp = av_fetch(av, (I32)obase->op_private, FALSE);
13878 if (!svp || *svp != uninit_sv)
13881 return varname(gv, '$', 0,
13882 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
13887 o = cUNOPx(obase)->op_first;
13888 if (!o || o->op_type != OP_NULL ||
13889 ! (o->op_targ == OP_AELEM || o->op_targ == OP_HELEM))
13891 return find_uninit_var(cBINOPo->op_last, uninit_sv, match);
13895 if (PL_op == obase)
13896 /* $a[uninit_expr] or $h{uninit_expr} */
13897 return find_uninit_var(cBINOPx(obase)->op_last, uninit_sv, match);
13900 o = cBINOPx(obase)->op_first;
13901 kid = cBINOPx(obase)->op_last;
13903 /* get the av or hv, and optionally the gv */
13905 if (o->op_type == OP_PADAV || o->op_type == OP_PADHV) {
13906 sv = PAD_SV(o->op_targ);
13908 else if ((o->op_type == OP_RV2AV || o->op_type == OP_RV2HV)
13909 && cUNOPo->op_first->op_type == OP_GV)
13911 gv = cGVOPx_gv(cUNOPo->op_first);
13915 == OP_RV2HV ? MUTABLE_SV(GvHV(gv)) : MUTABLE_SV(GvAV(gv));
13920 if (kid && kid->op_type == OP_CONST && SvOK(cSVOPx_sv(kid))) {
13921 /* index is constant */
13925 if (obase->op_type == OP_HELEM) {
13926 HE* he = hv_fetch_ent(MUTABLE_HV(sv), cSVOPx_sv(kid), 0, 0);
13927 if (!he || HeVAL(he) != uninit_sv)
13931 SV * const * const svp = av_fetch(MUTABLE_AV(sv), SvIV(cSVOPx_sv(kid)), FALSE);
13932 if (!svp || *svp != uninit_sv)
13936 if (obase->op_type == OP_HELEM)
13937 return varname(gv, '%', o->op_targ,
13938 cSVOPx_sv(kid), 0, FUV_SUBSCRIPT_HASH);
13940 return varname(gv, '@', o->op_targ, NULL,
13941 SvIV(cSVOPx_sv(kid)), FUV_SUBSCRIPT_ARRAY);
13944 /* index is an expression;
13945 * attempt to find a match within the aggregate */
13946 if (obase->op_type == OP_HELEM) {
13947 SV * const keysv = find_hash_subscript((const HV*)sv, uninit_sv);
13949 return varname(gv, '%', o->op_targ,
13950 keysv, 0, FUV_SUBSCRIPT_HASH);
13954 = find_array_subscript((const AV *)sv, uninit_sv);
13956 return varname(gv, '@', o->op_targ,
13957 NULL, index, FUV_SUBSCRIPT_ARRAY);
13962 (o->op_type == OP_PADAV || o->op_type == OP_RV2AV)
13964 o->op_targ, NULL, 0, FUV_SUBSCRIPT_WITHIN);
13969 /* only examine RHS */
13970 return find_uninit_var(cBINOPx(obase)->op_first, uninit_sv, match);
13973 o = cUNOPx(obase)->op_first;
13974 if (o->op_type == OP_PUSHMARK)
13977 if (!o->op_sibling) {
13978 /* one-arg version of open is highly magical */
13980 if (o->op_type == OP_GV) { /* open FOO; */
13982 if (match && GvSV(gv) != uninit_sv)
13984 return varname(gv, '$', 0,
13985 NULL, 0, FUV_SUBSCRIPT_NONE);
13987 /* other possibilities not handled are:
13988 * open $x; or open my $x; should return '${*$x}'
13989 * open expr; should return '$'.expr ideally
13995 /* ops where $_ may be an implicit arg */
13999 if ( !(obase->op_flags & OPf_STACKED)) {
14000 if (uninit_sv == ((obase->op_private & OPpTARGET_MY)
14001 ? PAD_SVl(obase->op_targ)
14004 sv = sv_newmortal();
14005 sv_setpvs(sv, "$_");
14014 match = 1; /* print etc can return undef on defined args */
14015 /* skip filehandle as it can't produce 'undef' warning */
14016 o = cUNOPx(obase)->op_first;
14017 if ((obase->op_flags & OPf_STACKED) && o->op_type == OP_PUSHMARK)
14018 o = o->op_sibling->op_sibling;
14022 case OP_ENTEREVAL: /* could be eval $undef or $x='$undef'; eval $x */
14024 case OP_CUSTOM: /* XS or custom code could trigger random warnings */
14026 /* the following ops are capable of returning PL_sv_undef even for
14027 * defined arg(s) */
14046 case OP_GETPEERNAME:
14094 case OP_SMARTMATCH:
14103 /* XXX tmp hack: these two may call an XS sub, and currently
14104 XS subs don't have a SUB entry on the context stack, so CV and
14105 pad determination goes wrong, and BAD things happen. So, just
14106 don't try to determine the value under those circumstances.
14107 Need a better fix at dome point. DAPM 11/2007 */
14113 GV * const gv = gv_fetchpvs(".", GV_NOTQUAL, SVt_PV);
14114 if (gv && GvSV(gv) == uninit_sv)
14115 return newSVpvs_flags("$.", SVs_TEMP);
14120 /* def-ness of rval pos() is independent of the def-ness of its arg */
14121 if ( !(obase->op_flags & OPf_MOD))
14126 if (SvROK(PL_rs) && uninit_sv == SvRV(PL_rs))
14127 return newSVpvs_flags("${$/}", SVs_TEMP);
14132 if (!(obase->op_flags & OPf_KIDS))
14134 o = cUNOPx(obase)->op_first;
14140 /* if all except one arg are constant, or have no side-effects,
14141 * or are optimized away, then it's unambiguous */
14143 for (kid=o; kid; kid = kid->op_sibling) {
14145 const OPCODE type = kid->op_type;
14146 if ( (type == OP_CONST && SvOK(cSVOPx_sv(kid)))
14147 || (type == OP_NULL && ! (kid->op_flags & OPf_KIDS))
14148 || (type == OP_PUSHMARK)
14150 /* @$a and %$a, but not @a or %a */
14151 (type == OP_RV2AV || type == OP_RV2HV)
14152 && cUNOPx(kid)->op_first
14153 && cUNOPx(kid)->op_first->op_type != OP_GV
14158 if (o2) { /* more than one found */
14165 return find_uninit_var(o2, uninit_sv, match);
14167 /* scan all args */
14169 sv = find_uninit_var(o, uninit_sv, 1);
14181 =for apidoc report_uninit
14183 Print appropriate "Use of uninitialized variable" warning
14189 Perl_report_uninit(pTHX_ const SV *uninit_sv)
14193 SV* varname = NULL;
14195 varname = find_uninit_var(PL_op, uninit_sv,0);
14197 sv_insert(varname, 0, 0, " ", 1);
14199 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
14200 varname ? SvPV_nolen_const(varname) : "",
14201 " in ", OP_DESC(PL_op));
14204 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
14210 * c-indentation-style: bsd
14211 * c-basic-offset: 4
14212 * indent-tabs-mode: t
14215 * ex: set ts=8 sts=4 sw=4 noet: