3 * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, 2000,
4 * 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009 by Larry Wall
7 * You may distribute under the terms of either the GNU General Public
8 * License or the Artistic License, as specified in the README file.
13 * 'I wonder what the Entish is for "yes" and "no",' he thought.
16 * [p.480 of _The Lord of the Rings_, III/iv: "Treebeard"]
22 * This file contains the code that creates, manipulates and destroys
23 * scalar values (SVs). The other types (AV, HV, GV, etc.) reuse the
24 * structure of an SV, so their creation and destruction is handled
25 * here; higher-level functions are in av.c, hv.c, and so on. Opcode
26 * level functions (eg. substr, split, join) for each of the types are
36 # if __STDC_VERSION__ >= 199901L && !defined(VMS)
47 /* Missing proto on LynxOS */
48 char *gconvert(double, int, int, char *);
51 #ifdef PERL_UTF8_CACHE_ASSERT
52 /* if adding more checks watch out for the following tests:
53 * t/op/index.t t/op/length.t t/op/pat.t t/op/substr.t
54 * lib/utf8.t lib/Unicode/Collate/t/index.t
57 # define ASSERT_UTF8_CACHE(cache) \
58 STMT_START { if (cache) { assert((cache)[0] <= (cache)[1]); \
59 assert((cache)[2] <= (cache)[3]); \
60 assert((cache)[3] <= (cache)[1]);} \
63 # define ASSERT_UTF8_CACHE(cache) NOOP
66 #ifdef PERL_OLD_COPY_ON_WRITE
67 #define SV_COW_NEXT_SV(sv) INT2PTR(SV *,SvUVX(sv))
68 #define SV_COW_NEXT_SV_SET(current,next) SvUV_set(current, PTR2UV(next))
69 /* This is a pessimistic view. Scalar must be purely a read-write PV to copy-
73 /* ============================================================================
75 =head1 Allocation and deallocation of SVs.
77 An SV (or AV, HV, etc.) is allocated in two parts: the head (struct
78 sv, av, hv...) contains type and reference count information, and for
79 many types, a pointer to the body (struct xrv, xpv, xpviv...), which
80 contains fields specific to each type. Some types store all they need
81 in the head, so don't have a body.
83 In all but the most memory-paranoid configurations (ex: PURIFY), heads
84 and bodies are allocated out of arenas, which by default are
85 approximately 4K chunks of memory parcelled up into N heads or bodies.
86 Sv-bodies are allocated by their sv-type, guaranteeing size
87 consistency needed to allocate safely from arrays.
89 For SV-heads, the first slot in each arena is reserved, and holds a
90 link to the next arena, some flags, and a note of the number of slots.
91 Snaked through each arena chain is a linked list of free items; when
92 this becomes empty, an extra arena is allocated and divided up into N
93 items which are threaded into the free list.
95 SV-bodies are similar, but they use arena-sets by default, which
96 separate the link and info from the arena itself, and reclaim the 1st
97 slot in the arena. SV-bodies are further described later.
99 The following global variables are associated with arenas:
101 PL_sv_arenaroot pointer to list of SV arenas
102 PL_sv_root pointer to list of free SV structures
104 PL_body_arenas head of linked-list of body arenas
105 PL_body_roots[] array of pointers to list of free bodies of svtype
106 arrays are indexed by the svtype needed
108 A few special SV heads are not allocated from an arena, but are
109 instead directly created in the interpreter structure, eg PL_sv_undef.
110 The size of arenas can be changed from the default by setting
111 PERL_ARENA_SIZE appropriately at compile time.
113 The SV arena serves the secondary purpose of allowing still-live SVs
114 to be located and destroyed during final cleanup.
116 At the lowest level, the macros new_SV() and del_SV() grab and free
117 an SV head. (If debugging with -DD, del_SV() calls the function S_del_sv()
118 to return the SV to the free list with error checking.) new_SV() calls
119 more_sv() / sv_add_arena() to add an extra arena if the free list is empty.
120 SVs in the free list have their SvTYPE field set to all ones.
122 At the time of very final cleanup, sv_free_arenas() is called from
123 perl_destruct() to physically free all the arenas allocated since the
124 start of the interpreter.
126 The function visit() scans the SV arenas list, and calls a specified
127 function for each SV it finds which is still live - ie which has an SvTYPE
128 other than all 1's, and a non-zero SvREFCNT. visit() is used by the
129 following functions (specified as [function that calls visit()] / [function
130 called by visit() for each SV]):
132 sv_report_used() / do_report_used()
133 dump all remaining SVs (debugging aid)
135 sv_clean_objs() / do_clean_objs(),do_clean_named_objs(),
136 do_clean_named_io_objs()
137 Attempt to free all objects pointed to by RVs,
138 and try to do the same for all objects indirectly
139 referenced by typeglobs too. Called once from
140 perl_destruct(), prior to calling sv_clean_all()
143 sv_clean_all() / do_clean_all()
144 SvREFCNT_dec(sv) each remaining SV, possibly
145 triggering an sv_free(). It also sets the
146 SVf_BREAK flag on the SV to indicate that the
147 refcnt has been artificially lowered, and thus
148 stopping sv_free() from giving spurious warnings
149 about SVs which unexpectedly have a refcnt
150 of zero. called repeatedly from perl_destruct()
151 until there are no SVs left.
153 =head2 Arena allocator API Summary
155 Private API to rest of sv.c
159 new_XPVNV(), del_XPVGV(),
164 sv_report_used(), sv_clean_objs(), sv_clean_all(), sv_free_arenas()
168 * ========================================================================= */
171 * "A time to plant, and a time to uproot what was planted..."
175 # define MEM_LOG_NEW_SV(sv, file, line, func) \
176 Perl_mem_log_new_sv(sv, file, line, func)
177 # define MEM_LOG_DEL_SV(sv, file, line, func) \
178 Perl_mem_log_del_sv(sv, file, line, func)
180 # define MEM_LOG_NEW_SV(sv, file, line, func) NOOP
181 # define MEM_LOG_DEL_SV(sv, file, line, func) NOOP
184 #ifdef DEBUG_LEAKING_SCALARS
185 # define FREE_SV_DEBUG_FILE(sv) Safefree((sv)->sv_debug_file)
186 # define DEBUG_SV_SERIAL(sv) \
187 DEBUG_m(PerlIO_printf(Perl_debug_log, "0x%"UVxf": (%05ld) del_SV\n", \
188 PTR2UV(sv), (long)(sv)->sv_debug_serial))
190 # define FREE_SV_DEBUG_FILE(sv)
191 # define DEBUG_SV_SERIAL(sv) NOOP
195 # define SvARENA_CHAIN(sv) ((sv)->sv_u.svu_rv)
196 # define SvARENA_CHAIN_SET(sv,val) (sv)->sv_u.svu_rv = MUTABLE_SV((val))
197 /* Whilst I'd love to do this, it seems that things like to check on
199 # define POSION_SV_HEAD(sv) PoisonNew(sv, 1, struct STRUCT_SV)
201 # define POSION_SV_HEAD(sv) PoisonNew(&SvANY(sv), 1, void *), \
202 PoisonNew(&SvREFCNT(sv), 1, U32)
204 # define SvARENA_CHAIN(sv) SvANY(sv)
205 # define SvARENA_CHAIN_SET(sv,val) SvANY(sv) = (void *)(val)
206 # define POSION_SV_HEAD(sv)
209 /* Mark an SV head as unused, and add to free list.
211 * If SVf_BREAK is set, skip adding it to the free list, as this SV had
212 * its refcount artificially decremented during global destruction, so
213 * there may be dangling pointers to it. The last thing we want in that
214 * case is for it to be reused. */
216 #define plant_SV(p) \
218 const U32 old_flags = SvFLAGS(p); \
219 MEM_LOG_DEL_SV(p, __FILE__, __LINE__, FUNCTION__); \
220 DEBUG_SV_SERIAL(p); \
221 FREE_SV_DEBUG_FILE(p); \
223 SvFLAGS(p) = SVTYPEMASK; \
224 if (!(old_flags & SVf_BREAK)) { \
225 SvARENA_CHAIN_SET(p, PL_sv_root); \
231 #define uproot_SV(p) \
234 PL_sv_root = MUTABLE_SV(SvARENA_CHAIN(p)); \
239 /* make some more SVs by adding another arena */
246 char *chunk; /* must use New here to match call to */
247 Newx(chunk,PERL_ARENA_SIZE,char); /* Safefree() in sv_free_arenas() */
248 sv_add_arena(chunk, PERL_ARENA_SIZE, 0);
253 /* new_SV(): return a new, empty SV head */
255 #ifdef DEBUG_LEAKING_SCALARS
256 /* provide a real function for a debugger to play with */
258 S_new_SV(pTHX_ const char *file, int line, const char *func)
265 sv = S_more_sv(aTHX);
269 sv->sv_debug_optype = PL_op ? PL_op->op_type : 0;
270 sv->sv_debug_line = (U16) (PL_parser && PL_parser->copline != NOLINE
276 sv->sv_debug_inpad = 0;
277 sv->sv_debug_parent = NULL;
278 sv->sv_debug_file = PL_curcop ? savepv(CopFILE(PL_curcop)): NULL;
280 sv->sv_debug_serial = PL_sv_serial++;
282 MEM_LOG_NEW_SV(sv, file, line, func);
283 DEBUG_m(PerlIO_printf(Perl_debug_log, "0x%"UVxf": (%05ld) new_SV (from %s:%d [%s])\n",
284 PTR2UV(sv), (long)sv->sv_debug_serial, file, line, func));
288 # define new_SV(p) (p)=S_new_SV(aTHX_ __FILE__, __LINE__, FUNCTION__)
296 (p) = S_more_sv(aTHX); \
300 MEM_LOG_NEW_SV(p, __FILE__, __LINE__, FUNCTION__); \
305 /* del_SV(): return an empty SV head to the free list */
318 S_del_sv(pTHX_ SV *p)
322 PERL_ARGS_ASSERT_DEL_SV;
327 for (sva = PL_sv_arenaroot; sva; sva = MUTABLE_SV(SvANY(sva))) {
328 const SV * const sv = sva + 1;
329 const SV * const svend = &sva[SvREFCNT(sva)];
330 if (p >= sv && p < svend) {
336 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL),
337 "Attempt to free non-arena SV: 0x%"UVxf
338 pTHX__FORMAT, PTR2UV(p) pTHX__VALUE);
345 #else /* ! DEBUGGING */
347 #define del_SV(p) plant_SV(p)
349 #endif /* DEBUGGING */
353 =head1 SV Manipulation Functions
355 =for apidoc sv_add_arena
357 Given a chunk of memory, link it to the head of the list of arenas,
358 and split it into a list of free SVs.
364 S_sv_add_arena(pTHX_ char *const ptr, const U32 size, const U32 flags)
367 SV *const sva = MUTABLE_SV(ptr);
371 PERL_ARGS_ASSERT_SV_ADD_ARENA;
373 /* The first SV in an arena isn't an SV. */
374 SvANY(sva) = (void *) PL_sv_arenaroot; /* ptr to next arena */
375 SvREFCNT(sva) = size / sizeof(SV); /* number of SV slots */
376 SvFLAGS(sva) = flags; /* FAKE if not to be freed */
378 PL_sv_arenaroot = sva;
379 PL_sv_root = sva + 1;
381 svend = &sva[SvREFCNT(sva) - 1];
384 SvARENA_CHAIN_SET(sv, (sv + 1));
388 /* Must always set typemask because it's always checked in on cleanup
389 when the arenas are walked looking for objects. */
390 SvFLAGS(sv) = SVTYPEMASK;
393 SvARENA_CHAIN_SET(sv, 0);
397 SvFLAGS(sv) = SVTYPEMASK;
400 /* visit(): call the named function for each non-free SV in the arenas
401 * whose flags field matches the flags/mask args. */
404 S_visit(pTHX_ SVFUNC_t f, const U32 flags, const U32 mask)
410 PERL_ARGS_ASSERT_VISIT;
412 for (sva = PL_sv_arenaroot; sva; sva = MUTABLE_SV(SvANY(sva))) {
413 register const SV * const svend = &sva[SvREFCNT(sva)];
415 for (sv = sva + 1; sv < svend; ++sv) {
416 if (SvTYPE(sv) != SVTYPEMASK
417 && (sv->sv_flags & mask) == flags
430 /* called by sv_report_used() for each live SV */
433 do_report_used(pTHX_ SV *const sv)
435 if (SvTYPE(sv) != SVTYPEMASK) {
436 PerlIO_printf(Perl_debug_log, "****\n");
443 =for apidoc sv_report_used
445 Dump the contents of all SVs not yet freed. (Debugging aid).
451 Perl_sv_report_used(pTHX)
454 visit(do_report_used, 0, 0);
460 /* called by sv_clean_objs() for each live SV */
463 do_clean_objs(pTHX_ SV *const ref)
468 SV * const target = SvRV(ref);
469 if (SvOBJECT(target)) {
470 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning object ref:\n "), sv_dump(ref)));
471 if (SvWEAKREF(ref)) {
472 sv_del_backref(target, ref);
478 SvREFCNT_dec(target);
483 /* XXX Might want to check arrays, etc. */
487 /* clear any slots in a GV which hold objects - except IO;
488 * called by sv_clean_objs() for each live GV */
491 do_clean_named_objs(pTHX_ SV *const sv)
495 assert(SvTYPE(sv) == SVt_PVGV);
496 assert(isGV_with_GP(sv));
500 /* freeing GP entries may indirectly free the current GV;
501 * hold onto it while we mess with the GP slots */
504 if ( ((obj = GvSV(sv) )) && SvOBJECT(obj)) {
505 DEBUG_D((PerlIO_printf(Perl_debug_log,
506 "Cleaning named glob SV object:\n "), sv_dump(obj)));
510 if ( ((obj = MUTABLE_SV(GvAV(sv)) )) && SvOBJECT(obj)) {
511 DEBUG_D((PerlIO_printf(Perl_debug_log,
512 "Cleaning named glob AV object:\n "), sv_dump(obj)));
516 if ( ((obj = MUTABLE_SV(GvHV(sv)) )) && SvOBJECT(obj)) {
517 DEBUG_D((PerlIO_printf(Perl_debug_log,
518 "Cleaning named glob HV object:\n "), sv_dump(obj)));
522 if ( ((obj = MUTABLE_SV(GvCV(sv)) )) && SvOBJECT(obj)) {
523 DEBUG_D((PerlIO_printf(Perl_debug_log,
524 "Cleaning named glob CV object:\n "), sv_dump(obj)));
528 SvREFCNT_dec(sv); /* undo the inc above */
531 /* clear any IO slots in a GV which hold objects (except stderr, defout);
532 * called by sv_clean_objs() for each live GV */
535 do_clean_named_io_objs(pTHX_ SV *const sv)
539 assert(SvTYPE(sv) == SVt_PVGV);
540 assert(isGV_with_GP(sv));
541 if (!GvGP(sv) || sv == (SV*)PL_stderrgv || sv == (SV*)PL_defoutgv)
545 if ( ((obj = MUTABLE_SV(GvIO(sv)) )) && SvOBJECT(obj)) {
546 DEBUG_D((PerlIO_printf(Perl_debug_log,
547 "Cleaning named glob IO object:\n "), sv_dump(obj)));
551 SvREFCNT_dec(sv); /* undo the inc above */
554 /* Void wrapper to pass to visit() */
556 do_curse(pTHX_ SV * const sv) {
557 if ((PL_stderrgv && GvGP(PL_stderrgv) && (SV*)GvIO(PL_stderrgv) == sv)
558 || (PL_defoutgv && GvGP(PL_defoutgv) && (SV*)GvIO(PL_defoutgv) == sv))
564 =for apidoc sv_clean_objs
566 Attempt to destroy all objects not yet freed
572 Perl_sv_clean_objs(pTHX)
576 PL_in_clean_objs = TRUE;
577 visit(do_clean_objs, SVf_ROK, SVf_ROK);
578 /* Some barnacles may yet remain, clinging to typeglobs.
579 * Run the non-IO destructors first: they may want to output
580 * error messages, close files etc */
581 visit(do_clean_named_objs, SVt_PVGV|SVpgv_GP, SVTYPEMASK|SVp_POK|SVpgv_GP);
582 visit(do_clean_named_io_objs, SVt_PVGV|SVpgv_GP, SVTYPEMASK|SVp_POK|SVpgv_GP);
583 /* And if there are some very tenacious barnacles clinging to arrays,
584 closures, or what have you.... */
585 visit(do_curse, SVs_OBJECT, SVs_OBJECT);
586 olddef = PL_defoutgv;
587 PL_defoutgv = NULL; /* disable skip of PL_defoutgv */
588 if (olddef && isGV_with_GP(olddef))
589 do_clean_named_io_objs(aTHX_ MUTABLE_SV(olddef));
590 olderr = PL_stderrgv;
591 PL_stderrgv = NULL; /* disable skip of PL_stderrgv */
592 if (olderr && isGV_with_GP(olderr))
593 do_clean_named_io_objs(aTHX_ MUTABLE_SV(olderr));
594 SvREFCNT_dec(olddef);
595 PL_in_clean_objs = FALSE;
598 /* called by sv_clean_all() for each live SV */
601 do_clean_all(pTHX_ SV *const sv)
604 if (sv == (const SV *) PL_fdpid || sv == (const SV *)PL_strtab) {
605 /* don't clean pid table and strtab */
608 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning loops: SV at 0x%"UVxf"\n", PTR2UV(sv)) ));
609 SvFLAGS(sv) |= SVf_BREAK;
614 =for apidoc sv_clean_all
616 Decrement the refcnt of each remaining SV, possibly triggering a
617 cleanup. This function may have to be called multiple times to free
618 SVs which are in complex self-referential hierarchies.
624 Perl_sv_clean_all(pTHX)
628 PL_in_clean_all = TRUE;
629 cleaned = visit(do_clean_all, 0,0);
634 ARENASETS: a meta-arena implementation which separates arena-info
635 into struct arena_set, which contains an array of struct
636 arena_descs, each holding info for a single arena. By separating
637 the meta-info from the arena, we recover the 1st slot, formerly
638 borrowed for list management. The arena_set is about the size of an
639 arena, avoiding the needless malloc overhead of a naive linked-list.
641 The cost is 1 arena-set malloc per ~320 arena-mallocs, + the unused
642 memory in the last arena-set (1/2 on average). In trade, we get
643 back the 1st slot in each arena (ie 1.7% of a CV-arena, less for
644 smaller types). The recovery of the wasted space allows use of
645 small arenas for large, rare body types, by changing array* fields
646 in body_details_by_type[] below.
649 char *arena; /* the raw storage, allocated aligned */
650 size_t size; /* its size ~4k typ */
651 svtype utype; /* bodytype stored in arena */
656 /* Get the maximum number of elements in set[] such that struct arena_set
657 will fit within PERL_ARENA_SIZE, which is probably just under 4K, and
658 therefore likely to be 1 aligned memory page. */
660 #define ARENAS_PER_SET ((PERL_ARENA_SIZE - sizeof(struct arena_set*) \
661 - 2 * sizeof(int)) / sizeof (struct arena_desc))
664 struct arena_set* next;
665 unsigned int set_size; /* ie ARENAS_PER_SET */
666 unsigned int curr; /* index of next available arena-desc */
667 struct arena_desc set[ARENAS_PER_SET];
671 =for apidoc sv_free_arenas
673 Deallocate the memory used by all arenas. Note that all the individual SV
674 heads and bodies within the arenas must already have been freed.
679 Perl_sv_free_arenas(pTHX)
686 /* Free arenas here, but be careful about fake ones. (We assume
687 contiguity of the fake ones with the corresponding real ones.) */
689 for (sva = PL_sv_arenaroot; sva; sva = svanext) {
690 svanext = MUTABLE_SV(SvANY(sva));
691 while (svanext && SvFAKE(svanext))
692 svanext = MUTABLE_SV(SvANY(svanext));
699 struct arena_set *aroot = (struct arena_set*) PL_body_arenas;
702 struct arena_set *current = aroot;
705 assert(aroot->set[i].arena);
706 Safefree(aroot->set[i].arena);
714 i = PERL_ARENA_ROOTS_SIZE;
716 PL_body_roots[i] = 0;
723 Here are mid-level routines that manage the allocation of bodies out
724 of the various arenas. There are 5 kinds of arenas:
726 1. SV-head arenas, which are discussed and handled above
727 2. regular body arenas
728 3. arenas for reduced-size bodies
731 Arena types 2 & 3 are chained by body-type off an array of
732 arena-root pointers, which is indexed by svtype. Some of the
733 larger/less used body types are malloced singly, since a large
734 unused block of them is wasteful. Also, several svtypes dont have
735 bodies; the data fits into the sv-head itself. The arena-root
736 pointer thus has a few unused root-pointers (which may be hijacked
737 later for arena types 4,5)
739 3 differs from 2 as an optimization; some body types have several
740 unused fields in the front of the structure (which are kept in-place
741 for consistency). These bodies can be allocated in smaller chunks,
742 because the leading fields arent accessed. Pointers to such bodies
743 are decremented to point at the unused 'ghost' memory, knowing that
744 the pointers are used with offsets to the real memory.
747 =head1 SV-Body Allocation
749 Allocation of SV-bodies is similar to SV-heads, differing as follows;
750 the allocation mechanism is used for many body types, so is somewhat
751 more complicated, it uses arena-sets, and has no need for still-live
754 At the outermost level, (new|del)_X*V macros return bodies of the
755 appropriate type. These macros call either (new|del)_body_type or
756 (new|del)_body_allocated macro pairs, depending on specifics of the
757 type. Most body types use the former pair, the latter pair is used to
758 allocate body types with "ghost fields".
760 "ghost fields" are fields that are unused in certain types, and
761 consequently don't need to actually exist. They are declared because
762 they're part of a "base type", which allows use of functions as
763 methods. The simplest examples are AVs and HVs, 2 aggregate types
764 which don't use the fields which support SCALAR semantics.
766 For these types, the arenas are carved up into appropriately sized
767 chunks, we thus avoid wasted memory for those unaccessed members.
768 When bodies are allocated, we adjust the pointer back in memory by the
769 size of the part not allocated, so it's as if we allocated the full
770 structure. (But things will all go boom if you write to the part that
771 is "not there", because you'll be overwriting the last members of the
772 preceding structure in memory.)
774 We calculate the correction using the STRUCT_OFFSET macro on the first
775 member present. If the allocated structure is smaller (no initial NV
776 actually allocated) then the net effect is to subtract the size of the NV
777 from the pointer, to return a new pointer as if an initial NV were actually
778 allocated. (We were using structures named *_allocated for this, but
779 this turned out to be a subtle bug, because a structure without an NV
780 could have a lower alignment constraint, but the compiler is allowed to
781 optimised accesses based on the alignment constraint of the actual pointer
782 to the full structure, for example, using a single 64 bit load instruction
783 because it "knows" that two adjacent 32 bit members will be 8-byte aligned.)
785 This is the same trick as was used for NV and IV bodies. Ironically it
786 doesn't need to be used for NV bodies any more, because NV is now at
787 the start of the structure. IV bodies don't need it either, because
788 they are no longer allocated.
790 In turn, the new_body_* allocators call S_new_body(), which invokes
791 new_body_inline macro, which takes a lock, and takes a body off the
792 linked list at PL_body_roots[sv_type], calling Perl_more_bodies() if
793 necessary to refresh an empty list. Then the lock is released, and
794 the body is returned.
796 Perl_more_bodies allocates a new arena, and carves it up into an array of N
797 bodies, which it strings into a linked list. It looks up arena-size
798 and body-size from the body_details table described below, thus
799 supporting the multiple body-types.
801 If PURIFY is defined, or PERL_ARENA_SIZE=0, arenas are not used, and
802 the (new|del)_X*V macros are mapped directly to malloc/free.
804 For each sv-type, struct body_details bodies_by_type[] carries
805 parameters which control these aspects of SV handling:
807 Arena_size determines whether arenas are used for this body type, and if
808 so, how big they are. PURIFY or PERL_ARENA_SIZE=0 set this field to
809 zero, forcing individual mallocs and frees.
811 Body_size determines how big a body is, and therefore how many fit into
812 each arena. Offset carries the body-pointer adjustment needed for
813 "ghost fields", and is used in *_allocated macros.
815 But its main purpose is to parameterize info needed in
816 Perl_sv_upgrade(). The info here dramatically simplifies the function
817 vs the implementation in 5.8.8, making it table-driven. All fields
818 are used for this, except for arena_size.
820 For the sv-types that have no bodies, arenas are not used, so those
821 PL_body_roots[sv_type] are unused, and can be overloaded. In
822 something of a special case, SVt_NULL is borrowed for HE arenas;
823 PL_body_roots[HE_SVSLOT=SVt_NULL] is filled by S_more_he, but the
824 bodies_by_type[SVt_NULL] slot is not used, as the table is not
829 struct body_details {
830 U8 body_size; /* Size to allocate */
831 U8 copy; /* Size of structure to copy (may be shorter) */
833 unsigned int type : 4; /* We have space for a sanity check. */
834 unsigned int cant_upgrade : 1; /* Cannot upgrade this type */
835 unsigned int zero_nv : 1; /* zero the NV when upgrading from this */
836 unsigned int arena : 1; /* Allocated from an arena */
837 size_t arena_size; /* Size of arena to allocate */
845 /* With -DPURFIY we allocate everything directly, and don't use arenas.
846 This seems a rather elegant way to simplify some of the code below. */
847 #define HASARENA FALSE
849 #define HASARENA TRUE
851 #define NOARENA FALSE
853 /* Size the arenas to exactly fit a given number of bodies. A count
854 of 0 fits the max number bodies into a PERL_ARENA_SIZE.block,
855 simplifying the default. If count > 0, the arena is sized to fit
856 only that many bodies, allowing arenas to be used for large, rare
857 bodies (XPVFM, XPVIO) without undue waste. The arena size is
858 limited by PERL_ARENA_SIZE, so we can safely oversize the
861 #define FIT_ARENA0(body_size) \
862 ((size_t)(PERL_ARENA_SIZE / body_size) * body_size)
863 #define FIT_ARENAn(count,body_size) \
864 ( count * body_size <= PERL_ARENA_SIZE) \
865 ? count * body_size \
866 : FIT_ARENA0 (body_size)
867 #define FIT_ARENA(count,body_size) \
869 ? FIT_ARENAn (count, body_size) \
870 : FIT_ARENA0 (body_size)
872 /* Calculate the length to copy. Specifically work out the length less any
873 final padding the compiler needed to add. See the comment in sv_upgrade
874 for why copying the padding proved to be a bug. */
876 #define copy_length(type, last_member) \
877 STRUCT_OFFSET(type, last_member) \
878 + sizeof (((type*)SvANY((const SV *)0))->last_member)
880 static const struct body_details bodies_by_type[] = {
881 /* HEs use this offset for their arena. */
882 { 0, 0, 0, SVt_NULL, FALSE, NONV, NOARENA, 0 },
884 /* The bind placeholder pretends to be an RV for now.
885 Also it's marked as "can't upgrade" to stop anyone using it before it's
887 { 0, 0, 0, SVt_BIND, TRUE, NONV, NOARENA, 0 },
889 /* IVs are in the head, so the allocation size is 0. */
891 sizeof(IV), /* This is used to copy out the IV body. */
892 STRUCT_OFFSET(XPVIV, xiv_iv), SVt_IV, FALSE, NONV,
893 NOARENA /* IVS don't need an arena */, 0
896 /* 8 bytes on most ILP32 with IEEE doubles */
897 { sizeof(NV), sizeof(NV),
898 STRUCT_OFFSET(XPVNV, xnv_u),
899 SVt_NV, FALSE, HADNV, HASARENA, FIT_ARENA(0, sizeof(NV)) },
901 /* 8 bytes on most ILP32 with IEEE doubles */
902 { sizeof(XPV) - STRUCT_OFFSET(XPV, xpv_cur),
903 copy_length(XPV, xpv_len) - STRUCT_OFFSET(XPV, xpv_cur),
904 + STRUCT_OFFSET(XPV, xpv_cur),
905 SVt_PV, FALSE, NONV, HASARENA,
906 FIT_ARENA(0, sizeof(XPV) - STRUCT_OFFSET(XPV, xpv_cur)) },
909 { sizeof(XPVIV) - STRUCT_OFFSET(XPV, xpv_cur),
910 copy_length(XPVIV, xiv_u) - STRUCT_OFFSET(XPV, xpv_cur),
911 + STRUCT_OFFSET(XPV, xpv_cur),
912 SVt_PVIV, FALSE, NONV, HASARENA,
913 FIT_ARENA(0, sizeof(XPVIV) - STRUCT_OFFSET(XPV, xpv_cur)) },
916 { sizeof(XPVNV) - STRUCT_OFFSET(XPV, xpv_cur),
917 copy_length(XPVNV, xnv_u) - STRUCT_OFFSET(XPV, xpv_cur),
918 + STRUCT_OFFSET(XPV, xpv_cur),
919 SVt_PVNV, FALSE, HADNV, HASARENA,
920 FIT_ARENA(0, sizeof(XPVNV) - STRUCT_OFFSET(XPV, xpv_cur)) },
923 { sizeof(XPVMG), copy_length(XPVMG, xnv_u), 0, SVt_PVMG, FALSE, HADNV,
924 HASARENA, FIT_ARENA(0, sizeof(XPVMG)) },
930 SVt_REGEXP, FALSE, NONV, HASARENA,
931 FIT_ARENA(0, sizeof(regexp))
935 { sizeof(XPVGV), sizeof(XPVGV), 0, SVt_PVGV, TRUE, HADNV,
936 HASARENA, FIT_ARENA(0, sizeof(XPVGV)) },
939 { sizeof(XPVLV), sizeof(XPVLV), 0, SVt_PVLV, TRUE, HADNV,
940 HASARENA, FIT_ARENA(0, sizeof(XPVLV)) },
943 copy_length(XPVAV, xav_alloc),
945 SVt_PVAV, TRUE, NONV, HASARENA,
946 FIT_ARENA(0, sizeof(XPVAV)) },
949 copy_length(XPVHV, xhv_max),
951 SVt_PVHV, TRUE, NONV, HASARENA,
952 FIT_ARENA(0, sizeof(XPVHV)) },
958 SVt_PVCV, TRUE, NONV, HASARENA,
959 FIT_ARENA(0, sizeof(XPVCV)) },
964 SVt_PVFM, TRUE, NONV, NOARENA,
965 FIT_ARENA(20, sizeof(XPVFM)) },
967 /* XPVIO is 84 bytes, fits 48x */
971 SVt_PVIO, TRUE, NONV, HASARENA,
972 FIT_ARENA(24, sizeof(XPVIO)) },
975 #define new_body_allocated(sv_type) \
976 (void *)((char *)S_new_body(aTHX_ sv_type) \
977 - bodies_by_type[sv_type].offset)
979 /* return a thing to the free list */
981 #define del_body(thing, root) \
983 void ** const thing_copy = (void **)thing; \
984 *thing_copy = *root; \
985 *root = (void*)thing_copy; \
990 #define new_XNV() safemalloc(sizeof(XPVNV))
991 #define new_XPVNV() safemalloc(sizeof(XPVNV))
992 #define new_XPVMG() safemalloc(sizeof(XPVMG))
994 #define del_XPVGV(p) safefree(p)
998 #define new_XNV() new_body_allocated(SVt_NV)
999 #define new_XPVNV() new_body_allocated(SVt_PVNV)
1000 #define new_XPVMG() new_body_allocated(SVt_PVMG)
1002 #define del_XPVGV(p) del_body(p + bodies_by_type[SVt_PVGV].offset, \
1003 &PL_body_roots[SVt_PVGV])
1007 /* no arena for you! */
1009 #define new_NOARENA(details) \
1010 safemalloc((details)->body_size + (details)->offset)
1011 #define new_NOARENAZ(details) \
1012 safecalloc((details)->body_size + (details)->offset, 1)
1015 Perl_more_bodies (pTHX_ const svtype sv_type, const size_t body_size,
1016 const size_t arena_size)
1019 void ** const root = &PL_body_roots[sv_type];
1020 struct arena_desc *adesc;
1021 struct arena_set *aroot = (struct arena_set *) PL_body_arenas;
1025 const size_t good_arena_size = Perl_malloc_good_size(arena_size);
1026 #if defined(DEBUGGING) && !defined(PERL_GLOBAL_STRUCT_PRIVATE)
1027 static bool done_sanity_check;
1029 /* PERL_GLOBAL_STRUCT_PRIVATE cannot coexist with global
1030 * variables like done_sanity_check. */
1031 if (!done_sanity_check) {
1032 unsigned int i = SVt_LAST;
1034 done_sanity_check = TRUE;
1037 assert (bodies_by_type[i].type == i);
1043 /* may need new arena-set to hold new arena */
1044 if (!aroot || aroot->curr >= aroot->set_size) {
1045 struct arena_set *newroot;
1046 Newxz(newroot, 1, struct arena_set);
1047 newroot->set_size = ARENAS_PER_SET;
1048 newroot->next = aroot;
1050 PL_body_arenas = (void *) newroot;
1051 DEBUG_m(PerlIO_printf(Perl_debug_log, "new arenaset %p\n", (void*)aroot));
1054 /* ok, now have arena-set with at least 1 empty/available arena-desc */
1055 curr = aroot->curr++;
1056 adesc = &(aroot->set[curr]);
1057 assert(!adesc->arena);
1059 Newx(adesc->arena, good_arena_size, char);
1060 adesc->size = good_arena_size;
1061 adesc->utype = sv_type;
1062 DEBUG_m(PerlIO_printf(Perl_debug_log, "arena %d added: %p size %"UVuf"\n",
1063 curr, (void*)adesc->arena, (UV)good_arena_size));
1065 start = (char *) adesc->arena;
1067 /* Get the address of the byte after the end of the last body we can fit.
1068 Remember, this is integer division: */
1069 end = start + good_arena_size / body_size * body_size;
1071 /* computed count doesn't reflect the 1st slot reservation */
1072 #if defined(MYMALLOC) || defined(HAS_MALLOC_GOOD_SIZE)
1073 DEBUG_m(PerlIO_printf(Perl_debug_log,
1074 "arena %p end %p arena-size %d (from %d) type %d "
1076 (void*)start, (void*)end, (int)good_arena_size,
1077 (int)arena_size, sv_type, (int)body_size,
1078 (int)good_arena_size / (int)body_size));
1080 DEBUG_m(PerlIO_printf(Perl_debug_log,
1081 "arena %p end %p arena-size %d type %d size %d ct %d\n",
1082 (void*)start, (void*)end,
1083 (int)arena_size, sv_type, (int)body_size,
1084 (int)good_arena_size / (int)body_size));
1086 *root = (void *)start;
1089 /* Where the next body would start: */
1090 char * const next = start + body_size;
1093 /* This is the last body: */
1094 assert(next == end);
1096 *(void **)start = 0;
1100 *(void**) start = (void *)next;
1105 /* grab a new thing from the free list, allocating more if necessary.
1106 The inline version is used for speed in hot routines, and the
1107 function using it serves the rest (unless PURIFY).
1109 #define new_body_inline(xpv, sv_type) \
1111 void ** const r3wt = &PL_body_roots[sv_type]; \
1112 xpv = (PTR_TBL_ENT_t*) (*((void **)(r3wt)) \
1113 ? *((void **)(r3wt)) : Perl_more_bodies(aTHX_ sv_type, \
1114 bodies_by_type[sv_type].body_size,\
1115 bodies_by_type[sv_type].arena_size)); \
1116 *(r3wt) = *(void**)(xpv); \
1122 S_new_body(pTHX_ const svtype sv_type)
1126 new_body_inline(xpv, sv_type);
1132 static const struct body_details fake_rv =
1133 { 0, 0, 0, SVt_IV, FALSE, NONV, NOARENA, 0 };
1136 =for apidoc sv_upgrade
1138 Upgrade an SV to a more complex form. Generally adds a new body type to the
1139 SV, then copies across as much information as possible from the old body.
1140 You generally want to use the C<SvUPGRADE> macro wrapper. See also C<svtype>.
1146 Perl_sv_upgrade(pTHX_ register SV *const sv, svtype new_type)
1151 const svtype old_type = SvTYPE(sv);
1152 const struct body_details *new_type_details;
1153 const struct body_details *old_type_details
1154 = bodies_by_type + old_type;
1155 SV *referant = NULL;
1157 PERL_ARGS_ASSERT_SV_UPGRADE;
1159 if (old_type == new_type)
1162 /* This clause was purposefully added ahead of the early return above to
1163 the shared string hackery for (sort {$a <=> $b} keys %hash), with the
1164 inference by Nick I-S that it would fix other troublesome cases. See
1165 changes 7162, 7163 (f130fd4589cf5fbb24149cd4db4137c8326f49c1 and parent)
1167 Given that shared hash key scalars are no longer PVIV, but PV, there is
1168 no longer need to unshare so as to free up the IVX slot for its proper
1169 purpose. So it's safe to move the early return earlier. */
1171 if (new_type != SVt_PV && SvIsCOW(sv)) {
1172 sv_force_normal_flags(sv, 0);
1175 old_body = SvANY(sv);
1177 /* Copying structures onto other structures that have been neatly zeroed
1178 has a subtle gotcha. Consider XPVMG
1180 +------+------+------+------+------+-------+-------+
1181 | NV | CUR | LEN | IV | MAGIC | STASH |
1182 +------+------+------+------+------+-------+-------+
1183 0 4 8 12 16 20 24 28
1185 where NVs are aligned to 8 bytes, so that sizeof that structure is
1186 actually 32 bytes long, with 4 bytes of padding at the end:
1188 +------+------+------+------+------+-------+-------+------+
1189 | NV | CUR | LEN | IV | MAGIC | STASH | ??? |
1190 +------+------+------+------+------+-------+-------+------+
1191 0 4 8 12 16 20 24 28 32
1193 so what happens if you allocate memory for this structure:
1195 +------+------+------+------+------+-------+-------+------+------+...
1196 | NV | CUR | LEN | IV | MAGIC | STASH | GP | NAME |
1197 +------+------+------+------+------+-------+-------+------+------+...
1198 0 4 8 12 16 20 24 28 32 36
1200 zero it, then copy sizeof(XPVMG) bytes on top of it? Not quite what you
1201 expect, because you copy the area marked ??? onto GP. Now, ??? may have
1202 started out as zero once, but it's quite possible that it isn't. So now,
1203 rather than a nicely zeroed GP, you have it pointing somewhere random.
1206 (In fact, GP ends up pointing at a previous GP structure, because the
1207 principle cause of the padding in XPVMG getting garbage is a copy of
1208 sizeof(XPVMG) bytes from a XPVGV structure in sv_unglob. Right now
1209 this happens to be moot because XPVGV has been re-ordered, with GP
1210 no longer after STASH)
1212 So we are careful and work out the size of used parts of all the
1220 referant = SvRV(sv);
1221 old_type_details = &fake_rv;
1222 if (new_type == SVt_NV)
1223 new_type = SVt_PVNV;
1225 if (new_type < SVt_PVIV) {
1226 new_type = (new_type == SVt_NV)
1227 ? SVt_PVNV : SVt_PVIV;
1232 if (new_type < SVt_PVNV) {
1233 new_type = SVt_PVNV;
1237 assert(new_type > SVt_PV);
1238 assert(SVt_IV < SVt_PV);
1239 assert(SVt_NV < SVt_PV);
1246 /* Because the XPVMG of PL_mess_sv isn't allocated from the arena,
1247 there's no way that it can be safely upgraded, because perl.c
1248 expects to Safefree(SvANY(PL_mess_sv)) */
1249 assert(sv != PL_mess_sv);
1250 /* This flag bit is used to mean other things in other scalar types.
1251 Given that it only has meaning inside the pad, it shouldn't be set
1252 on anything that can get upgraded. */
1253 assert(!SvPAD_TYPED(sv));
1256 if (old_type_details->cant_upgrade)
1257 Perl_croak(aTHX_ "Can't upgrade %s (%" UVuf ") to %" UVuf,
1258 sv_reftype(sv, 0), (UV) old_type, (UV) new_type);
1261 if (old_type > new_type)
1262 Perl_croak(aTHX_ "sv_upgrade from type %d down to type %d",
1263 (int)old_type, (int)new_type);
1265 new_type_details = bodies_by_type + new_type;
1267 SvFLAGS(sv) &= ~SVTYPEMASK;
1268 SvFLAGS(sv) |= new_type;
1270 /* This can't happen, as SVt_NULL is <= all values of new_type, so one of
1271 the return statements above will have triggered. */
1272 assert (new_type != SVt_NULL);
1275 assert(old_type == SVt_NULL);
1276 SvANY(sv) = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
1280 assert(old_type == SVt_NULL);
1281 SvANY(sv) = new_XNV();
1286 assert(new_type_details->body_size);
1289 assert(new_type_details->arena);
1290 assert(new_type_details->arena_size);
1291 /* This points to the start of the allocated area. */
1292 new_body_inline(new_body, new_type);
1293 Zero(new_body, new_type_details->body_size, char);
1294 new_body = ((char *)new_body) - new_type_details->offset;
1296 /* We always allocated the full length item with PURIFY. To do this
1297 we fake things so that arena is false for all 16 types.. */
1298 new_body = new_NOARENAZ(new_type_details);
1300 SvANY(sv) = new_body;
1301 if (new_type == SVt_PVAV) {
1305 if (old_type_details->body_size) {
1308 /* It will have been zeroed when the new body was allocated.
1309 Lets not write to it, in case it confuses a write-back
1315 #ifndef NODEFAULT_SHAREKEYS
1316 HvSHAREKEYS_on(sv); /* key-sharing on by default */
1318 HvMAX(sv) = 7; /* (start with 8 buckets) */
1321 /* SVt_NULL isn't the only thing upgraded to AV or HV.
1322 The target created by newSVrv also is, and it can have magic.
1323 However, it never has SvPVX set.
1325 if (old_type == SVt_IV) {
1327 } else if (old_type >= SVt_PV) {
1328 assert(SvPVX_const(sv) == 0);
1331 if (old_type >= SVt_PVMG) {
1332 SvMAGIC_set(sv, ((XPVMG*)old_body)->xmg_u.xmg_magic);
1333 SvSTASH_set(sv, ((XPVMG*)old_body)->xmg_stash);
1335 sv->sv_u.svu_array = NULL; /* or svu_hash */
1341 /* This ensures that SvTHINKFIRST(sv) is true, and hence that
1342 sv_force_normal_flags(sv) is called. */
1345 /* XXX Is this still needed? Was it ever needed? Surely as there is
1346 no route from NV to PVIV, NOK can never be true */
1347 assert(!SvNOKp(sv));
1358 assert(new_type_details->body_size);
1359 /* We always allocated the full length item with PURIFY. To do this
1360 we fake things so that arena is false for all 16 types.. */
1361 if(new_type_details->arena) {
1362 /* This points to the start of the allocated area. */
1363 new_body_inline(new_body, new_type);
1364 Zero(new_body, new_type_details->body_size, char);
1365 new_body = ((char *)new_body) - new_type_details->offset;
1367 new_body = new_NOARENAZ(new_type_details);
1369 SvANY(sv) = new_body;
1371 if (old_type_details->copy) {
1372 /* There is now the potential for an upgrade from something without
1373 an offset (PVNV or PVMG) to something with one (PVCV, PVFM) */
1374 int offset = old_type_details->offset;
1375 int length = old_type_details->copy;
1377 if (new_type_details->offset > old_type_details->offset) {
1378 const int difference
1379 = new_type_details->offset - old_type_details->offset;
1380 offset += difference;
1381 length -= difference;
1383 assert (length >= 0);
1385 Copy((char *)old_body + offset, (char *)new_body + offset, length,
1389 #ifndef NV_ZERO_IS_ALLBITS_ZERO
1390 /* If NV 0.0 is stores as all bits 0 then Zero() already creates a
1391 * correct 0.0 for us. Otherwise, if the old body didn't have an
1392 * NV slot, but the new one does, then we need to initialise the
1393 * freshly created NV slot with whatever the correct bit pattern is
1395 if (old_type_details->zero_nv && !new_type_details->zero_nv
1396 && !isGV_with_GP(sv))
1400 if (new_type == SVt_PVIO) {
1401 IO * const io = MUTABLE_IO(sv);
1402 GV *iogv = gv_fetchpvs("IO::File::", GV_ADD, SVt_PVHV);
1405 /* Clear the stashcache because a new IO could overrule a package
1407 hv_clear(PL_stashcache);
1409 SvSTASH_set(io, MUTABLE_HV(SvREFCNT_inc(GvHV(iogv))));
1410 IoPAGE_LEN(sv) = 60;
1412 if (old_type < SVt_PV) {
1413 /* referant will be NULL unless the old type was SVt_IV emulating
1415 sv->sv_u.svu_rv = referant;
1419 Perl_croak(aTHX_ "panic: sv_upgrade to unknown type %lu",
1420 (unsigned long)new_type);
1423 if (old_type > SVt_IV) {
1427 /* Note that there is an assumption that all bodies of types that
1428 can be upgraded came from arenas. Only the more complex non-
1429 upgradable types are allowed to be directly malloc()ed. */
1430 assert(old_type_details->arena);
1431 del_body((void*)((char*)old_body + old_type_details->offset),
1432 &PL_body_roots[old_type]);
1438 =for apidoc sv_backoff
1440 Remove any string offset. You should normally use the C<SvOOK_off> macro
1447 Perl_sv_backoff(pTHX_ register SV *const sv)
1450 const char * const s = SvPVX_const(sv);
1452 PERL_ARGS_ASSERT_SV_BACKOFF;
1453 PERL_UNUSED_CONTEXT;
1456 assert(SvTYPE(sv) != SVt_PVHV);
1457 assert(SvTYPE(sv) != SVt_PVAV);
1459 SvOOK_offset(sv, delta);
1461 SvLEN_set(sv, SvLEN(sv) + delta);
1462 SvPV_set(sv, SvPVX(sv) - delta);
1463 Move(s, SvPVX(sv), SvCUR(sv)+1, char);
1464 SvFLAGS(sv) &= ~SVf_OOK;
1471 Expands the character buffer in the SV. If necessary, uses C<sv_unref> and
1472 upgrades the SV to C<SVt_PV>. Returns a pointer to the character buffer.
1473 Use the C<SvGROW> wrapper instead.
1479 Perl_sv_grow(pTHX_ register SV *const sv, register STRLEN newlen)
1483 PERL_ARGS_ASSERT_SV_GROW;
1485 if (PL_madskills && newlen >= 0x100000) {
1486 PerlIO_printf(Perl_debug_log,
1487 "Allocation too large: %"UVxf"\n", (UV)newlen);
1489 #ifdef HAS_64K_LIMIT
1490 if (newlen >= 0x10000) {
1491 PerlIO_printf(Perl_debug_log,
1492 "Allocation too large: %"UVxf"\n", (UV)newlen);
1495 #endif /* HAS_64K_LIMIT */
1498 if (SvTYPE(sv) < SVt_PV) {
1499 sv_upgrade(sv, SVt_PV);
1500 s = SvPVX_mutable(sv);
1502 else if (SvOOK(sv)) { /* pv is offset? */
1504 s = SvPVX_mutable(sv);
1505 if (newlen > SvLEN(sv))
1506 newlen += 10 * (newlen - SvCUR(sv)); /* avoid copy each time */
1507 #ifdef HAS_64K_LIMIT
1508 if (newlen >= 0x10000)
1513 s = SvPVX_mutable(sv);
1515 if (newlen > SvLEN(sv)) { /* need more room? */
1516 STRLEN minlen = SvCUR(sv);
1517 minlen += (minlen >> PERL_STRLEN_EXPAND_SHIFT) + 10;
1518 if (newlen < minlen)
1520 #ifndef Perl_safesysmalloc_size
1521 newlen = PERL_STRLEN_ROUNDUP(newlen);
1523 if (SvLEN(sv) && s) {
1524 s = (char*)saferealloc(s, newlen);
1527 s = (char*)safemalloc(newlen);
1528 if (SvPVX_const(sv) && SvCUR(sv)) {
1529 Move(SvPVX_const(sv), s, (newlen < SvCUR(sv)) ? newlen : SvCUR(sv), char);
1533 #ifdef Perl_safesysmalloc_size
1534 /* Do this here, do it once, do it right, and then we will never get
1535 called back into sv_grow() unless there really is some growing
1537 SvLEN_set(sv, Perl_safesysmalloc_size(s));
1539 SvLEN_set(sv, newlen);
1546 =for apidoc sv_setiv
1548 Copies an integer into the given SV, upgrading first if necessary.
1549 Does not handle 'set' magic. See also C<sv_setiv_mg>.
1555 Perl_sv_setiv(pTHX_ register SV *const sv, const IV i)
1559 PERL_ARGS_ASSERT_SV_SETIV;
1561 SV_CHECK_THINKFIRST_COW_DROP(sv);
1562 switch (SvTYPE(sv)) {
1565 sv_upgrade(sv, SVt_IV);
1568 sv_upgrade(sv, SVt_PVIV);
1572 if (!isGV_with_GP(sv))
1579 Perl_croak(aTHX_ "Can't coerce %s to integer in %s", sv_reftype(sv,0),
1583 (void)SvIOK_only(sv); /* validate number */
1589 =for apidoc sv_setiv_mg
1591 Like C<sv_setiv>, but also handles 'set' magic.
1597 Perl_sv_setiv_mg(pTHX_ register SV *const sv, const IV i)
1599 PERL_ARGS_ASSERT_SV_SETIV_MG;
1606 =for apidoc sv_setuv
1608 Copies an unsigned integer into the given SV, upgrading first if necessary.
1609 Does not handle 'set' magic. See also C<sv_setuv_mg>.
1615 Perl_sv_setuv(pTHX_ register SV *const sv, const UV u)
1617 PERL_ARGS_ASSERT_SV_SETUV;
1619 /* With these two if statements:
1620 u=1.49 s=0.52 cu=72.49 cs=10.64 scripts=270 tests=20865
1623 u=1.35 s=0.47 cu=73.45 cs=11.43 scripts=270 tests=20865
1625 If you wish to remove them, please benchmark to see what the effect is
1627 if (u <= (UV)IV_MAX) {
1628 sv_setiv(sv, (IV)u);
1637 =for apidoc sv_setuv_mg
1639 Like C<sv_setuv>, but also handles 'set' magic.
1645 Perl_sv_setuv_mg(pTHX_ register SV *const sv, const UV u)
1647 PERL_ARGS_ASSERT_SV_SETUV_MG;
1654 =for apidoc sv_setnv
1656 Copies a double into the given SV, upgrading first if necessary.
1657 Does not handle 'set' magic. See also C<sv_setnv_mg>.
1663 Perl_sv_setnv(pTHX_ register SV *const sv, const NV num)
1667 PERL_ARGS_ASSERT_SV_SETNV;
1669 SV_CHECK_THINKFIRST_COW_DROP(sv);
1670 switch (SvTYPE(sv)) {
1673 sv_upgrade(sv, SVt_NV);
1677 sv_upgrade(sv, SVt_PVNV);
1681 if (!isGV_with_GP(sv))
1688 Perl_croak(aTHX_ "Can't coerce %s to number in %s", sv_reftype(sv,0),
1693 (void)SvNOK_only(sv); /* validate number */
1698 =for apidoc sv_setnv_mg
1700 Like C<sv_setnv>, but also handles 'set' magic.
1706 Perl_sv_setnv_mg(pTHX_ register SV *const sv, const NV num)
1708 PERL_ARGS_ASSERT_SV_SETNV_MG;
1714 /* Print an "isn't numeric" warning, using a cleaned-up,
1715 * printable version of the offending string
1719 S_not_a_number(pTHX_ SV *const sv)
1726 PERL_ARGS_ASSERT_NOT_A_NUMBER;
1729 dsv = newSVpvs_flags("", SVs_TEMP);
1730 pv = sv_uni_display(dsv, sv, 10, 0);
1733 const char * const limit = tmpbuf + sizeof(tmpbuf) - 8;
1734 /* each *s can expand to 4 chars + "...\0",
1735 i.e. need room for 8 chars */
1737 const char *s = SvPVX_const(sv);
1738 const char * const end = s + SvCUR(sv);
1739 for ( ; s < end && d < limit; s++ ) {
1741 if (ch & 128 && !isPRINT_LC(ch)) {
1750 else if (ch == '\r') {
1754 else if (ch == '\f') {
1758 else if (ch == '\\') {
1762 else if (ch == '\0') {
1766 else if (isPRINT_LC(ch))
1783 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
1784 "Argument \"%s\" isn't numeric in %s", pv,
1787 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
1788 "Argument \"%s\" isn't numeric", pv);
1792 =for apidoc looks_like_number
1794 Test if the content of an SV looks like a number (or is a number).
1795 C<Inf> and C<Infinity> are treated as numbers (so will not issue a
1796 non-numeric warning), even if your atof() doesn't grok them.
1802 Perl_looks_like_number(pTHX_ SV *const sv)
1804 register const char *sbegin;
1807 PERL_ARGS_ASSERT_LOOKS_LIKE_NUMBER;
1810 sbegin = SvPVX_const(sv);
1813 else if (SvPOKp(sv))
1814 sbegin = SvPV_const(sv, len);
1816 return SvFLAGS(sv) & (SVf_NOK|SVp_NOK|SVf_IOK|SVp_IOK);
1817 return grok_number(sbegin, len, NULL);
1821 S_glob_2number(pTHX_ GV * const gv)
1823 const U32 wasfake = SvFLAGS(gv) & SVf_FAKE;
1824 SV *const buffer = sv_newmortal();
1826 PERL_ARGS_ASSERT_GLOB_2NUMBER;
1828 /* FAKE globs can get coerced, so need to turn this off temporarily if it
1831 gv_efullname3(buffer, gv, "*");
1832 SvFLAGS(gv) |= wasfake;
1834 /* We know that all GVs stringify to something that is not-a-number,
1835 so no need to test that. */
1836 if (ckWARN(WARN_NUMERIC))
1837 not_a_number(buffer);
1838 /* We just want something true to return, so that S_sv_2iuv_common
1839 can tail call us and return true. */
1843 /* Actually, ISO C leaves conversion of UV to IV undefined, but
1844 until proven guilty, assume that things are not that bad... */
1849 As 64 bit platforms often have an NV that doesn't preserve all bits of
1850 an IV (an assumption perl has been based on to date) it becomes necessary
1851 to remove the assumption that the NV always carries enough precision to
1852 recreate the IV whenever needed, and that the NV is the canonical form.
1853 Instead, IV/UV and NV need to be given equal rights. So as to not lose
1854 precision as a side effect of conversion (which would lead to insanity
1855 and the dragon(s) in t/op/numconvert.t getting very angry) the intent is
1856 1) to distinguish between IV/UV/NV slots that have cached a valid
1857 conversion where precision was lost and IV/UV/NV slots that have a
1858 valid conversion which has lost no precision
1859 2) to ensure that if a numeric conversion to one form is requested that
1860 would lose precision, the precise conversion (or differently
1861 imprecise conversion) is also performed and cached, to prevent
1862 requests for different numeric formats on the same SV causing
1863 lossy conversion chains. (lossless conversion chains are perfectly
1868 SvIOKp is true if the IV slot contains a valid value
1869 SvIOK is true only if the IV value is accurate (UV if SvIOK_UV true)
1870 SvNOKp is true if the NV slot contains a valid value
1871 SvNOK is true only if the NV value is accurate
1874 while converting from PV to NV, check to see if converting that NV to an
1875 IV(or UV) would lose accuracy over a direct conversion from PV to
1876 IV(or UV). If it would, cache both conversions, return NV, but mark
1877 SV as IOK NOKp (ie not NOK).
1879 While converting from PV to IV, check to see if converting that IV to an
1880 NV would lose accuracy over a direct conversion from PV to NV. If it
1881 would, cache both conversions, flag similarly.
1883 Before, the SV value "3.2" could become NV=3.2 IV=3 NOK, IOK quite
1884 correctly because if IV & NV were set NV *always* overruled.
1885 Now, "3.2" will become NV=3.2 IV=3 NOK, IOKp, because the flag's meaning
1886 changes - now IV and NV together means that the two are interchangeable:
1887 SvIVX == (IV) SvNVX && SvNVX == (NV) SvIVX;
1889 The benefit of this is that operations such as pp_add know that if
1890 SvIOK is true for both left and right operands, then integer addition
1891 can be used instead of floating point (for cases where the result won't
1892 overflow). Before, floating point was always used, which could lead to
1893 loss of precision compared with integer addition.
1895 * making IV and NV equal status should make maths accurate on 64 bit
1897 * may speed up maths somewhat if pp_add and friends start to use
1898 integers when possible instead of fp. (Hopefully the overhead in
1899 looking for SvIOK and checking for overflow will not outweigh the
1900 fp to integer speedup)
1901 * will slow down integer operations (callers of SvIV) on "inaccurate"
1902 values, as the change from SvIOK to SvIOKp will cause a call into
1903 sv_2iv each time rather than a macro access direct to the IV slot
1904 * should speed up number->string conversion on integers as IV is
1905 favoured when IV and NV are equally accurate
1907 ####################################################################
1908 You had better be using SvIOK_notUV if you want an IV for arithmetic:
1909 SvIOK is true if (IV or UV), so you might be getting (IV)SvUV.
1910 On the other hand, SvUOK is true iff UV.
1911 ####################################################################
1913 Your mileage will vary depending your CPU's relative fp to integer
1917 #ifndef NV_PRESERVES_UV
1918 # define IS_NUMBER_UNDERFLOW_IV 1
1919 # define IS_NUMBER_UNDERFLOW_UV 2
1920 # define IS_NUMBER_IV_AND_UV 2
1921 # define IS_NUMBER_OVERFLOW_IV 4
1922 # define IS_NUMBER_OVERFLOW_UV 5
1924 /* sv_2iuv_non_preserve(): private routine for use by sv_2iv() and sv_2uv() */
1926 /* For sv_2nv these three cases are "SvNOK and don't bother casting" */
1928 S_sv_2iuv_non_preserve(pTHX_ register SV *const sv
1936 PERL_ARGS_ASSERT_SV_2IUV_NON_PRESERVE;
1938 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));
1939 if (SvNVX(sv) < (NV)IV_MIN) {
1940 (void)SvIOKp_on(sv);
1942 SvIV_set(sv, IV_MIN);
1943 return IS_NUMBER_UNDERFLOW_IV;
1945 if (SvNVX(sv) > (NV)UV_MAX) {
1946 (void)SvIOKp_on(sv);
1949 SvUV_set(sv, UV_MAX);
1950 return IS_NUMBER_OVERFLOW_UV;
1952 (void)SvIOKp_on(sv);
1954 /* Can't use strtol etc to convert this string. (See truth table in
1956 if (SvNVX(sv) <= (UV)IV_MAX) {
1957 SvIV_set(sv, I_V(SvNVX(sv)));
1958 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
1959 SvIOK_on(sv); /* Integer is precise. NOK, IOK */
1961 /* Integer is imprecise. NOK, IOKp */
1963 return SvNVX(sv) < 0 ? IS_NUMBER_UNDERFLOW_UV : IS_NUMBER_IV_AND_UV;
1966 SvUV_set(sv, U_V(SvNVX(sv)));
1967 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
1968 if (SvUVX(sv) == UV_MAX) {
1969 /* As we know that NVs don't preserve UVs, UV_MAX cannot
1970 possibly be preserved by NV. Hence, it must be overflow.
1972 return IS_NUMBER_OVERFLOW_UV;
1974 SvIOK_on(sv); /* Integer is precise. NOK, UOK */
1976 /* Integer is imprecise. NOK, IOKp */
1978 return IS_NUMBER_OVERFLOW_IV;
1980 #endif /* !NV_PRESERVES_UV*/
1983 S_sv_2iuv_common(pTHX_ SV *const sv)
1987 PERL_ARGS_ASSERT_SV_2IUV_COMMON;
1990 /* erm. not sure. *should* never get NOKp (without NOK) from sv_2nv
1991 * without also getting a cached IV/UV from it at the same time
1992 * (ie PV->NV conversion should detect loss of accuracy and cache
1993 * IV or UV at same time to avoid this. */
1994 /* IV-over-UV optimisation - choose to cache IV if possible */
1996 if (SvTYPE(sv) == SVt_NV)
1997 sv_upgrade(sv, SVt_PVNV);
1999 (void)SvIOKp_on(sv); /* Must do this first, to clear any SvOOK */
2000 /* < not <= as for NV doesn't preserve UV, ((NV)IV_MAX+1) will almost
2001 certainly cast into the IV range at IV_MAX, whereas the correct
2002 answer is the UV IV_MAX +1. Hence < ensures that dodgy boundary
2004 #if defined(NAN_COMPARE_BROKEN) && defined(Perl_isnan)
2005 if (Perl_isnan(SvNVX(sv))) {
2011 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2012 SvIV_set(sv, I_V(SvNVX(sv)));
2013 if (SvNVX(sv) == (NV) SvIVX(sv)
2014 #ifndef NV_PRESERVES_UV
2015 && (((UV)1 << NV_PRESERVES_UV_BITS) >
2016 (UV)(SvIVX(sv) > 0 ? SvIVX(sv) : -SvIVX(sv)))
2017 /* Don't flag it as "accurately an integer" if the number
2018 came from a (by definition imprecise) NV operation, and
2019 we're outside the range of NV integer precision */
2023 SvIOK_on(sv); /* Can this go wrong with rounding? NWC */
2025 /* scalar has trailing garbage, eg "42a" */
2027 DEBUG_c(PerlIO_printf(Perl_debug_log,
2028 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (precise)\n",
2034 /* IV not precise. No need to convert from PV, as NV
2035 conversion would already have cached IV if it detected
2036 that PV->IV would be better than PV->NV->IV
2037 flags already correct - don't set public IOK. */
2038 DEBUG_c(PerlIO_printf(Perl_debug_log,
2039 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (imprecise)\n",
2044 /* Can the above go wrong if SvIVX == IV_MIN and SvNVX < IV_MIN,
2045 but the cast (NV)IV_MIN rounds to a the value less (more
2046 negative) than IV_MIN which happens to be equal to SvNVX ??
2047 Analogous to 0xFFFFFFFFFFFFFFFF rounding up to NV (2**64) and
2048 NV rounding back to 0xFFFFFFFFFFFFFFFF, so UVX == UV(NVX) and
2049 (NV)UVX == NVX are both true, but the values differ. :-(
2050 Hopefully for 2s complement IV_MIN is something like
2051 0x8000000000000000 which will be exact. NWC */
2054 SvUV_set(sv, U_V(SvNVX(sv)));
2056 (SvNVX(sv) == (NV) SvUVX(sv))
2057 #ifndef NV_PRESERVES_UV
2058 /* Make sure it's not 0xFFFFFFFFFFFFFFFF */
2059 /*&& (SvUVX(sv) != UV_MAX) irrelevant with code below */
2060 && (((UV)1 << NV_PRESERVES_UV_BITS) > SvUVX(sv))
2061 /* Don't flag it as "accurately an integer" if the number
2062 came from a (by definition imprecise) NV operation, and
2063 we're outside the range of NV integer precision */
2069 DEBUG_c(PerlIO_printf(Perl_debug_log,
2070 "0x%"UVxf" 2iv(%"UVuf" => %"IVdf") (as unsigned)\n",
2076 else if (SvPOKp(sv) && SvLEN(sv)) {
2078 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2079 /* We want to avoid a possible problem when we cache an IV/ a UV which
2080 may be later translated to an NV, and the resulting NV is not
2081 the same as the direct translation of the initial string
2082 (eg 123.456 can shortcut to the IV 123 with atol(), but we must
2083 be careful to ensure that the value with the .456 is around if the
2084 NV value is requested in the future).
2086 This means that if we cache such an IV/a UV, we need to cache the
2087 NV as well. Moreover, we trade speed for space, and do not
2088 cache the NV if we are sure it's not needed.
2091 /* SVt_PVNV is one higher than SVt_PVIV, hence this order */
2092 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2093 == IS_NUMBER_IN_UV) {
2094 /* It's definitely an integer, only upgrade to PVIV */
2095 if (SvTYPE(sv) < SVt_PVIV)
2096 sv_upgrade(sv, SVt_PVIV);
2098 } else if (SvTYPE(sv) < SVt_PVNV)
2099 sv_upgrade(sv, SVt_PVNV);
2101 /* If NVs preserve UVs then we only use the UV value if we know that
2102 we aren't going to call atof() below. If NVs don't preserve UVs
2103 then the value returned may have more precision than atof() will
2104 return, even though value isn't perfectly accurate. */
2105 if ((numtype & (IS_NUMBER_IN_UV
2106 #ifdef NV_PRESERVES_UV
2109 )) == IS_NUMBER_IN_UV) {
2110 /* This won't turn off the public IOK flag if it was set above */
2111 (void)SvIOKp_on(sv);
2113 if (!(numtype & IS_NUMBER_NEG)) {
2115 if (value <= (UV)IV_MAX) {
2116 SvIV_set(sv, (IV)value);
2118 /* it didn't overflow, and it was positive. */
2119 SvUV_set(sv, value);
2123 /* 2s complement assumption */
2124 if (value <= (UV)IV_MIN) {
2125 SvIV_set(sv, -(IV)value);
2127 /* Too negative for an IV. This is a double upgrade, but
2128 I'm assuming it will be rare. */
2129 if (SvTYPE(sv) < SVt_PVNV)
2130 sv_upgrade(sv, SVt_PVNV);
2134 SvNV_set(sv, -(NV)value);
2135 SvIV_set(sv, IV_MIN);
2139 /* For !NV_PRESERVES_UV and IS_NUMBER_IN_UV and IS_NUMBER_NOT_INT we
2140 will be in the previous block to set the IV slot, and the next
2141 block to set the NV slot. So no else here. */
2143 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2144 != IS_NUMBER_IN_UV) {
2145 /* It wasn't an (integer that doesn't overflow the UV). */
2146 SvNV_set(sv, Atof(SvPVX_const(sv)));
2148 if (! numtype && ckWARN(WARN_NUMERIC))
2151 #if defined(USE_LONG_DOUBLE)
2152 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%" PERL_PRIgldbl ")\n",
2153 PTR2UV(sv), SvNVX(sv)));
2155 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"NVgf")\n",
2156 PTR2UV(sv), SvNVX(sv)));
2159 #ifdef NV_PRESERVES_UV
2160 (void)SvIOKp_on(sv);
2162 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2163 SvIV_set(sv, I_V(SvNVX(sv)));
2164 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
2167 NOOP; /* Integer is imprecise. NOK, IOKp */
2169 /* UV will not work better than IV */
2171 if (SvNVX(sv) > (NV)UV_MAX) {
2173 /* Integer is inaccurate. NOK, IOKp, is UV */
2174 SvUV_set(sv, UV_MAX);
2176 SvUV_set(sv, U_V(SvNVX(sv)));
2177 /* 0xFFFFFFFFFFFFFFFF not an issue in here, NVs
2178 NV preservse UV so can do correct comparison. */
2179 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
2182 NOOP; /* Integer is imprecise. NOK, IOKp, is UV */
2187 #else /* NV_PRESERVES_UV */
2188 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2189 == (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT)) {
2190 /* The IV/UV slot will have been set from value returned by
2191 grok_number above. The NV slot has just been set using
2194 assert (SvIOKp(sv));
2196 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2197 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2198 /* Small enough to preserve all bits. */
2199 (void)SvIOKp_on(sv);
2201 SvIV_set(sv, I_V(SvNVX(sv)));
2202 if ((NV)(SvIVX(sv)) == SvNVX(sv))
2204 /* Assumption: first non-preserved integer is < IV_MAX,
2205 this NV is in the preserved range, therefore: */
2206 if (!(U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))
2208 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);
2212 0 0 already failed to read UV.
2213 0 1 already failed to read UV.
2214 1 0 you won't get here in this case. IV/UV
2215 slot set, public IOK, Atof() unneeded.
2216 1 1 already read UV.
2217 so there's no point in sv_2iuv_non_preserve() attempting
2218 to use atol, strtol, strtoul etc. */
2220 sv_2iuv_non_preserve (sv, numtype);
2222 sv_2iuv_non_preserve (sv);
2226 #endif /* NV_PRESERVES_UV */
2227 /* It might be more code efficient to go through the entire logic above
2228 and conditionally set with SvIOKp_on() rather than SvIOK(), but it
2229 gets complex and potentially buggy, so more programmer efficient
2230 to do it this way, by turning off the public flags: */
2232 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK);
2236 if (isGV_with_GP(sv))
2237 return glob_2number(MUTABLE_GV(sv));
2239 if (!(SvFLAGS(sv) & SVs_PADTMP)) {
2240 if (!PL_localizing && ckWARN(WARN_UNINITIALIZED))
2243 if (SvTYPE(sv) < SVt_IV)
2244 /* Typically the caller expects that sv_any is not NULL now. */
2245 sv_upgrade(sv, SVt_IV);
2246 /* Return 0 from the caller. */
2253 =for apidoc sv_2iv_flags
2255 Return the integer value of an SV, doing any necessary string
2256 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2257 Normally used via the C<SvIV(sv)> and C<SvIVx(sv)> macros.
2263 Perl_sv_2iv_flags(pTHX_ register SV *const sv, const I32 flags)
2268 if (SvGMAGICAL(sv) || (SvTYPE(sv) == SVt_PVGV && SvVALID(sv))) {
2269 /* FBMs use the same flag bit as SVf_IVisUV, so must let them
2270 cache IVs just in case. In practice it seems that they never
2271 actually anywhere accessible by user Perl code, let alone get used
2272 in anything other than a string context. */
2273 if (flags & SV_GMAGIC)
2278 return I_V(SvNVX(sv));
2280 if (SvPOKp(sv) && SvLEN(sv)) {
2283 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2285 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2286 == IS_NUMBER_IN_UV) {
2287 /* It's definitely an integer */
2288 if (numtype & IS_NUMBER_NEG) {
2289 if (value < (UV)IV_MIN)
2292 if (value < (UV)IV_MAX)
2297 if (ckWARN(WARN_NUMERIC))
2300 return I_V(Atof(SvPVX_const(sv)));
2305 assert(SvTYPE(sv) >= SVt_PVMG);
2306 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
2307 } else if (SvTHINKFIRST(sv)) {
2312 if (flags & SV_SKIP_OVERLOAD)
2314 tmpstr = AMG_CALLunary(sv, numer_amg);
2315 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2316 return SvIV(tmpstr);
2319 return PTR2IV(SvRV(sv));
2322 sv_force_normal_flags(sv, 0);
2324 if (SvREADONLY(sv) && !SvOK(sv)) {
2325 if (ckWARN(WARN_UNINITIALIZED))
2331 if (S_sv_2iuv_common(aTHX_ sv))
2334 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"IVdf")\n",
2335 PTR2UV(sv),SvIVX(sv)));
2336 return SvIsUV(sv) ? (IV)SvUVX(sv) : SvIVX(sv);
2340 =for apidoc sv_2uv_flags
2342 Return the unsigned integer value of an SV, doing any necessary string
2343 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2344 Normally used via the C<SvUV(sv)> and C<SvUVx(sv)> macros.
2350 Perl_sv_2uv_flags(pTHX_ register SV *const sv, const I32 flags)
2355 if (SvGMAGICAL(sv) || (SvTYPE(sv) == SVt_PVGV && SvVALID(sv))) {
2356 /* FBMs use the same flag bit as SVf_IVisUV, so must let them
2357 cache IVs just in case. */
2358 if (flags & SV_GMAGIC)
2363 return U_V(SvNVX(sv));
2364 if (SvPOKp(sv) && SvLEN(sv)) {
2367 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2369 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2370 == IS_NUMBER_IN_UV) {
2371 /* It's definitely an integer */
2372 if (!(numtype & IS_NUMBER_NEG))
2376 if (ckWARN(WARN_NUMERIC))
2379 return U_V(Atof(SvPVX_const(sv)));
2384 assert(SvTYPE(sv) >= SVt_PVMG);
2385 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
2386 } else if (SvTHINKFIRST(sv)) {
2391 if (flags & SV_SKIP_OVERLOAD)
2393 tmpstr = AMG_CALLunary(sv, numer_amg);
2394 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2395 return SvUV(tmpstr);
2398 return PTR2UV(SvRV(sv));
2401 sv_force_normal_flags(sv, 0);
2403 if (SvREADONLY(sv) && !SvOK(sv)) {
2404 if (ckWARN(WARN_UNINITIALIZED))
2410 if (S_sv_2iuv_common(aTHX_ sv))
2414 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2uv(%"UVuf")\n",
2415 PTR2UV(sv),SvUVX(sv)));
2416 return SvIsUV(sv) ? SvUVX(sv) : (UV)SvIVX(sv);
2420 =for apidoc sv_2nv_flags
2422 Return the num value of an SV, doing any necessary string or integer
2423 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2424 Normally used via the C<SvNV(sv)> and C<SvNVx(sv)> macros.
2430 Perl_sv_2nv_flags(pTHX_ register SV *const sv, const I32 flags)
2435 if (SvGMAGICAL(sv) || (SvTYPE(sv) == SVt_PVGV && SvVALID(sv))) {
2436 /* FBMs use the same flag bit as SVf_IVisUV, so must let them
2437 cache IVs just in case. */
2438 if (flags & SV_GMAGIC)
2442 if ((SvPOKp(sv) && SvLEN(sv)) && !SvIOKp(sv)) {
2443 if (!SvIOKp(sv) && ckWARN(WARN_NUMERIC) &&
2444 !grok_number(SvPVX_const(sv), SvCUR(sv), NULL))
2446 return Atof(SvPVX_const(sv));
2450 return (NV)SvUVX(sv);
2452 return (NV)SvIVX(sv);
2457 assert(SvTYPE(sv) >= SVt_PVMG);
2458 /* This falls through to the report_uninit near the end of the
2460 } else if (SvTHINKFIRST(sv)) {
2465 if (flags & SV_SKIP_OVERLOAD)
2467 tmpstr = AMG_CALLunary(sv, numer_amg);
2468 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2469 return SvNV(tmpstr);
2472 return PTR2NV(SvRV(sv));
2475 sv_force_normal_flags(sv, 0);
2477 if (SvREADONLY(sv) && !SvOK(sv)) {
2478 if (ckWARN(WARN_UNINITIALIZED))
2483 if (SvTYPE(sv) < SVt_NV) {
2484 /* The logic to use SVt_PVNV if necessary is in sv_upgrade. */
2485 sv_upgrade(sv, SVt_NV);
2486 #ifdef USE_LONG_DOUBLE
2488 STORE_NUMERIC_LOCAL_SET_STANDARD();
2489 PerlIO_printf(Perl_debug_log,
2490 "0x%"UVxf" num(%" PERL_PRIgldbl ")\n",
2491 PTR2UV(sv), SvNVX(sv));
2492 RESTORE_NUMERIC_LOCAL();
2496 STORE_NUMERIC_LOCAL_SET_STANDARD();
2497 PerlIO_printf(Perl_debug_log, "0x%"UVxf" num(%"NVgf")\n",
2498 PTR2UV(sv), SvNVX(sv));
2499 RESTORE_NUMERIC_LOCAL();
2503 else if (SvTYPE(sv) < SVt_PVNV)
2504 sv_upgrade(sv, SVt_PVNV);
2509 SvNV_set(sv, SvIsUV(sv) ? (NV)SvUVX(sv) : (NV)SvIVX(sv));
2510 #ifdef NV_PRESERVES_UV
2516 /* Only set the public NV OK flag if this NV preserves the IV */
2517 /* Check it's not 0xFFFFFFFFFFFFFFFF */
2519 SvIsUV(sv) ? ((SvUVX(sv) != UV_MAX)&&(SvUVX(sv) == U_V(SvNVX(sv))))
2520 : (SvIVX(sv) == I_V(SvNVX(sv))))
2526 else if (SvPOKp(sv) && SvLEN(sv)) {
2528 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2529 if (!SvIOKp(sv) && !numtype && ckWARN(WARN_NUMERIC))
2531 #ifdef NV_PRESERVES_UV
2532 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2533 == IS_NUMBER_IN_UV) {
2534 /* It's definitely an integer */
2535 SvNV_set(sv, (numtype & IS_NUMBER_NEG) ? -(NV)value : (NV)value);
2537 SvNV_set(sv, Atof(SvPVX_const(sv)));
2543 SvNV_set(sv, Atof(SvPVX_const(sv)));
2544 /* Only set the public NV OK flag if this NV preserves the value in
2545 the PV at least as well as an IV/UV would.
2546 Not sure how to do this 100% reliably. */
2547 /* if that shift count is out of range then Configure's test is
2548 wonky. We shouldn't be in here with NV_PRESERVES_UV_BITS ==
2550 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2551 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2552 SvNOK_on(sv); /* Definitely small enough to preserve all bits */
2553 } else if (!(numtype & IS_NUMBER_IN_UV)) {
2554 /* Can't use strtol etc to convert this string, so don't try.
2555 sv_2iv and sv_2uv will use the NV to convert, not the PV. */
2558 /* value has been set. It may not be precise. */
2559 if ((numtype & IS_NUMBER_NEG) && (value > (UV)IV_MIN)) {
2560 /* 2s complement assumption for (UV)IV_MIN */
2561 SvNOK_on(sv); /* Integer is too negative. */
2566 if (numtype & IS_NUMBER_NEG) {
2567 SvIV_set(sv, -(IV)value);
2568 } else if (value <= (UV)IV_MAX) {
2569 SvIV_set(sv, (IV)value);
2571 SvUV_set(sv, value);
2575 if (numtype & IS_NUMBER_NOT_INT) {
2576 /* I believe that even if the original PV had decimals,
2577 they are lost beyond the limit of the FP precision.
2578 However, neither is canonical, so both only get p
2579 flags. NWC, 2000/11/25 */
2580 /* Both already have p flags, so do nothing */
2582 const NV nv = SvNVX(sv);
2583 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2584 if (SvIVX(sv) == I_V(nv)) {
2587 /* It had no "." so it must be integer. */
2591 /* between IV_MAX and NV(UV_MAX).
2592 Could be slightly > UV_MAX */
2594 if (numtype & IS_NUMBER_NOT_INT) {
2595 /* UV and NV both imprecise. */
2597 const UV nv_as_uv = U_V(nv);
2599 if (value == nv_as_uv && SvUVX(sv) != UV_MAX) {
2608 /* It might be more code efficient to go through the entire logic above
2609 and conditionally set with SvNOKp_on() rather than SvNOK(), but it
2610 gets complex and potentially buggy, so more programmer efficient
2611 to do it this way, by turning off the public flags: */
2613 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK);
2614 #endif /* NV_PRESERVES_UV */
2617 if (isGV_with_GP(sv)) {
2618 glob_2number(MUTABLE_GV(sv));
2622 if (!PL_localizing && !(SvFLAGS(sv) & SVs_PADTMP) && ckWARN(WARN_UNINITIALIZED))
2624 assert (SvTYPE(sv) >= SVt_NV);
2625 /* Typically the caller expects that sv_any is not NULL now. */
2626 /* XXX Ilya implies that this is a bug in callers that assume this
2627 and ideally should be fixed. */
2630 #if defined(USE_LONG_DOUBLE)
2632 STORE_NUMERIC_LOCAL_SET_STANDARD();
2633 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2nv(%" PERL_PRIgldbl ")\n",
2634 PTR2UV(sv), SvNVX(sv));
2635 RESTORE_NUMERIC_LOCAL();
2639 STORE_NUMERIC_LOCAL_SET_STANDARD();
2640 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 1nv(%"NVgf")\n",
2641 PTR2UV(sv), SvNVX(sv));
2642 RESTORE_NUMERIC_LOCAL();
2651 Return an SV with the numeric value of the source SV, doing any necessary
2652 reference or overload conversion. You must use the C<SvNUM(sv)> macro to
2653 access this function.
2659 Perl_sv_2num(pTHX_ register SV *const sv)
2661 PERL_ARGS_ASSERT_SV_2NUM;
2666 SV * const tmpsv = AMG_CALLunary(sv, numer_amg);
2667 TAINT_IF(tmpsv && SvTAINTED(tmpsv));
2668 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
2669 return sv_2num(tmpsv);
2671 return sv_2mortal(newSVuv(PTR2UV(SvRV(sv))));
2674 /* uiv_2buf(): private routine for use by sv_2pv_flags(): print an IV or
2675 * UV as a string towards the end of buf, and return pointers to start and
2678 * We assume that buf is at least TYPE_CHARS(UV) long.
2682 S_uiv_2buf(char *const buf, const IV iv, UV uv, const int is_uv, char **const peob)
2684 char *ptr = buf + TYPE_CHARS(UV);
2685 char * const ebuf = ptr;
2688 PERL_ARGS_ASSERT_UIV_2BUF;
2700 *--ptr = '0' + (char)(uv % 10);
2709 =for apidoc sv_2pv_flags
2711 Returns a pointer to the string value of an SV, and sets *lp to its length.
2712 If flags includes SV_GMAGIC, does an mg_get() first. Coerces sv to a string
2714 Normally invoked via the C<SvPV_flags> macro. C<sv_2pv()> and C<sv_2pv_nomg>
2715 usually end up here too.
2721 Perl_sv_2pv_flags(pTHX_ register SV *const sv, STRLEN *const lp, const I32 flags)
2731 if (SvGMAGICAL(sv)) {
2732 if (flags & SV_GMAGIC)
2737 if (flags & SV_MUTABLE_RETURN)
2738 return SvPVX_mutable(sv);
2739 if (flags & SV_CONST_RETURN)
2740 return (char *)SvPVX_const(sv);
2743 if (SvIOKp(sv) || SvNOKp(sv)) {
2744 char tbuf[64]; /* Must fit sprintf/Gconvert of longest IV/NV */
2749 ? my_snprintf(tbuf, sizeof(tbuf), "%"UVuf, (UV)SvUVX(sv))
2750 : my_snprintf(tbuf, sizeof(tbuf), "%"IVdf, (IV)SvIVX(sv));
2751 } else if(SvNVX(sv) == 0.0) {
2756 Gconvert(SvNVX(sv), NV_DIG, 0, tbuf);
2763 SvUPGRADE(sv, SVt_PV);
2766 s = SvGROW_mutable(sv, len + 1);
2769 return (char*)memcpy(s, tbuf, len + 1);
2775 assert(SvTYPE(sv) >= SVt_PVMG);
2776 /* This falls through to the report_uninit near the end of the
2778 } else if (SvTHINKFIRST(sv)) {
2783 if (flags & SV_SKIP_OVERLOAD)
2785 tmpstr = AMG_CALLunary(sv, string_amg);
2786 TAINT_IF(tmpstr && SvTAINTED(tmpstr));
2787 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2789 /* char *pv = lp ? SvPV(tmpstr, *lp) : SvPV_nolen(tmpstr);
2793 if ((SvFLAGS(tmpstr) & (SVf_POK)) == SVf_POK) {
2794 if (flags & SV_CONST_RETURN) {
2795 pv = (char *) SvPVX_const(tmpstr);
2797 pv = (flags & SV_MUTABLE_RETURN)
2798 ? SvPVX_mutable(tmpstr) : SvPVX(tmpstr);
2801 *lp = SvCUR(tmpstr);
2803 pv = sv_2pv_flags(tmpstr, lp, flags);
2816 SV *const referent = SvRV(sv);
2820 retval = buffer = savepvn("NULLREF", len);
2821 } else if (SvTYPE(referent) == SVt_REGEXP) {
2822 REGEXP * const re = (REGEXP *)MUTABLE_PTR(referent);
2827 /* If the regex is UTF-8 we want the containing scalar to
2828 have an UTF-8 flag too */
2834 if ((seen_evals = RX_SEEN_EVALS(re)))
2835 PL_reginterp_cnt += seen_evals;
2838 *lp = RX_WRAPLEN(re);
2840 return RX_WRAPPED(re);
2842 const char *const typestr = sv_reftype(referent, 0);
2843 const STRLEN typelen = strlen(typestr);
2844 UV addr = PTR2UV(referent);
2845 const char *stashname = NULL;
2846 STRLEN stashnamelen = 0; /* hush, gcc */
2847 const char *buffer_end;
2849 if (SvOBJECT(referent)) {
2850 const HEK *const name = HvNAME_HEK(SvSTASH(referent));
2853 stashname = HEK_KEY(name);
2854 stashnamelen = HEK_LEN(name);
2856 if (HEK_UTF8(name)) {
2862 stashname = "__ANON__";
2865 len = stashnamelen + 1 /* = */ + typelen + 3 /* (0x */
2866 + 2 * sizeof(UV) + 2 /* )\0 */;
2868 len = typelen + 3 /* (0x */
2869 + 2 * sizeof(UV) + 2 /* )\0 */;
2872 Newx(buffer, len, char);
2873 buffer_end = retval = buffer + len;
2875 /* Working backwards */
2879 *--retval = PL_hexdigit[addr & 15];
2880 } while (addr >>= 4);
2886 memcpy(retval, typestr, typelen);
2890 retval -= stashnamelen;
2891 memcpy(retval, stashname, stashnamelen);
2893 /* retval may not necessarily have reached the start of the
2895 assert (retval >= buffer);
2897 len = buffer_end - retval - 1; /* -1 for that \0 */
2905 if (SvREADONLY(sv) && !SvOK(sv)) {
2908 if (flags & SV_UNDEF_RETURNS_NULL)
2910 if (ckWARN(WARN_UNINITIALIZED))
2915 if (SvIOK(sv) || ((SvIOKp(sv) && !SvNOKp(sv)))) {
2916 /* I'm assuming that if both IV and NV are equally valid then
2917 converting the IV is going to be more efficient */
2918 const U32 isUIOK = SvIsUV(sv);
2919 char buf[TYPE_CHARS(UV)];
2923 if (SvTYPE(sv) < SVt_PVIV)
2924 sv_upgrade(sv, SVt_PVIV);
2925 ptr = uiv_2buf(buf, SvIVX(sv), SvUVX(sv), isUIOK, &ebuf);
2927 /* inlined from sv_setpvn */
2928 s = SvGROW_mutable(sv, len + 1);
2929 Move(ptr, s, len, char);
2933 else if (SvNOKp(sv)) {
2934 if (SvTYPE(sv) < SVt_PVNV)
2935 sv_upgrade(sv, SVt_PVNV);
2936 if (SvNVX(sv) == 0.0) {
2937 s = SvGROW_mutable(sv, 2);
2942 /* The +20 is pure guesswork. Configure test needed. --jhi */
2943 s = SvGROW_mutable(sv, NV_DIG + 20);
2944 /* some Xenix systems wipe out errno here */
2945 Gconvert(SvNVX(sv), NV_DIG, 0, s);
2955 if (isGV_with_GP(sv)) {
2956 GV *const gv = MUTABLE_GV(sv);
2957 const U32 wasfake = SvFLAGS(gv) & SVf_FAKE;
2958 SV *const buffer = sv_newmortal();
2960 /* FAKE globs can get coerced, so need to turn this off temporarily
2963 gv_efullname3(buffer, gv, "*");
2964 SvFLAGS(gv) |= wasfake;
2966 if (SvPOK(buffer)) {
2968 *lp = SvCUR(buffer);
2970 return SvPVX(buffer);
2981 if (flags & SV_UNDEF_RETURNS_NULL)
2983 if (!PL_localizing && !(SvFLAGS(sv) & SVs_PADTMP) && ckWARN(WARN_UNINITIALIZED))
2985 if (SvTYPE(sv) < SVt_PV)
2986 /* Typically the caller expects that sv_any is not NULL now. */
2987 sv_upgrade(sv, SVt_PV);
2991 const STRLEN len = s - SvPVX_const(sv);
2997 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
2998 PTR2UV(sv),SvPVX_const(sv)));
2999 if (flags & SV_CONST_RETURN)
3000 return (char *)SvPVX_const(sv);
3001 if (flags & SV_MUTABLE_RETURN)
3002 return SvPVX_mutable(sv);
3007 =for apidoc sv_copypv
3009 Copies a stringified representation of the source SV into the
3010 destination SV. Automatically performs any necessary mg_get and
3011 coercion of numeric values into strings. Guaranteed to preserve
3012 UTF8 flag even from overloaded objects. Similar in nature to
3013 sv_2pv[_flags] but operates directly on an SV instead of just the
3014 string. Mostly uses sv_2pv_flags to do its work, except when that
3015 would lose the UTF-8'ness of the PV.
3021 Perl_sv_copypv(pTHX_ SV *const dsv, register SV *const ssv)
3024 const char * const s = SvPV_const(ssv,len);
3026 PERL_ARGS_ASSERT_SV_COPYPV;
3028 sv_setpvn(dsv,s,len);
3036 =for apidoc sv_2pvbyte
3038 Return a pointer to the byte-encoded representation of the SV, and set *lp
3039 to its length. May cause the SV to be downgraded from UTF-8 as a
3042 Usually accessed via the C<SvPVbyte> macro.
3048 Perl_sv_2pvbyte(pTHX_ register SV *const sv, STRLEN *const lp)
3050 PERL_ARGS_ASSERT_SV_2PVBYTE;
3053 sv_utf8_downgrade(sv,0);
3054 return lp ? SvPV_nomg(sv,*lp) : SvPV_nomg_nolen(sv);
3058 =for apidoc sv_2pvutf8
3060 Return a pointer to the UTF-8-encoded representation of the SV, and set *lp
3061 to its length. May cause the SV to be upgraded to UTF-8 as a side-effect.
3063 Usually accessed via the C<SvPVutf8> macro.
3069 Perl_sv_2pvutf8(pTHX_ register SV *const sv, STRLEN *const lp)
3071 PERL_ARGS_ASSERT_SV_2PVUTF8;
3073 sv_utf8_upgrade(sv);
3074 return lp ? SvPV(sv,*lp) : SvPV_nolen(sv);
3079 =for apidoc sv_2bool
3081 This macro is only used by sv_true() or its macro equivalent, and only if
3082 the latter's argument is neither SvPOK, SvIOK nor SvNOK.
3083 It calls sv_2bool_flags with the SV_GMAGIC flag.
3085 =for apidoc sv_2bool_flags
3087 This function is only used by sv_true() and friends, and only if
3088 the latter's argument is neither SvPOK, SvIOK nor SvNOK. If the flags
3089 contain SV_GMAGIC, then it does an mg_get() first.
3096 Perl_sv_2bool_flags(pTHX_ register SV *const sv, const I32 flags)
3100 PERL_ARGS_ASSERT_SV_2BOOL_FLAGS;
3102 if(flags & SV_GMAGIC) SvGETMAGIC(sv);
3108 SV * const tmpsv = AMG_CALLunary(sv, bool__amg);
3109 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
3110 return cBOOL(SvTRUE(tmpsv));
3112 return SvRV(sv) != 0;
3115 register XPV* const Xpvtmp = (XPV*)SvANY(sv);
3117 (*sv->sv_u.svu_pv > '0' ||
3118 Xpvtmp->xpv_cur > 1 ||
3119 (Xpvtmp->xpv_cur && *sv->sv_u.svu_pv != '0')))
3126 return SvIVX(sv) != 0;
3129 return SvNVX(sv) != 0.0;
3131 if (isGV_with_GP(sv))
3141 =for apidoc sv_utf8_upgrade
3143 Converts the PV of an SV to its UTF-8-encoded form.
3144 Forces the SV to string form if it is not already.
3145 Will C<mg_get> on C<sv> if appropriate.
3146 Always sets the SvUTF8 flag to avoid future validity checks even
3147 if the whole string is the same in UTF-8 as not.
3148 Returns the number of bytes in the converted string
3150 This is not as a general purpose byte encoding to Unicode interface:
3151 use the Encode extension for that.
3153 =for apidoc sv_utf8_upgrade_nomg
3155 Like sv_utf8_upgrade, but doesn't do magic on C<sv>
3157 =for apidoc sv_utf8_upgrade_flags
3159 Converts the PV of an SV to its UTF-8-encoded form.
3160 Forces the SV to string form if it is not already.
3161 Always sets the SvUTF8 flag to avoid future validity checks even
3162 if all the bytes are invariant in UTF-8. If C<flags> has C<SV_GMAGIC> bit set,
3163 will C<mg_get> on C<sv> if appropriate, else not.
3164 Returns the number of bytes in the converted string
3165 C<sv_utf8_upgrade> and
3166 C<sv_utf8_upgrade_nomg> are implemented in terms of this function.
3168 This is not as a general purpose byte encoding to Unicode interface:
3169 use the Encode extension for that.
3173 The grow version is currently not externally documented. It adds a parameter,
3174 extra, which is the number of unused bytes the string of 'sv' is guaranteed to
3175 have free after it upon return. This allows the caller to reserve extra space
3176 that it intends to fill, to avoid extra grows.
3178 Also externally undocumented for the moment is the flag SV_FORCE_UTF8_UPGRADE,
3179 which can be used to tell this function to not first check to see if there are
3180 any characters that are different in UTF-8 (variant characters) which would
3181 force it to allocate a new string to sv, but to assume there are. Typically
3182 this flag is used by a routine that has already parsed the string to find that
3183 there are such characters, and passes this information on so that the work
3184 doesn't have to be repeated.
3186 (One might think that the calling routine could pass in the position of the
3187 first such variant, so it wouldn't have to be found again. But that is not the
3188 case, because typically when the caller is likely to use this flag, it won't be
3189 calling this routine unless it finds something that won't fit into a byte.
3190 Otherwise it tries to not upgrade and just use bytes. But some things that
3191 do fit into a byte are variants in utf8, and the caller may not have been
3192 keeping track of these.)
3194 If the routine itself changes the string, it adds a trailing NUL. Such a NUL
3195 isn't guaranteed due to having other routines do the work in some input cases,
3196 or if the input is already flagged as being in utf8.
3198 The speed of this could perhaps be improved for many cases if someone wanted to
3199 write a fast function that counts the number of variant characters in a string,
3200 especially if it could return the position of the first one.
3205 Perl_sv_utf8_upgrade_flags_grow(pTHX_ register SV *const sv, const I32 flags, STRLEN extra)
3209 PERL_ARGS_ASSERT_SV_UTF8_UPGRADE_FLAGS_GROW;
3211 if (sv == &PL_sv_undef)
3215 if (SvREADONLY(sv) && (SvPOKp(sv) || SvIOKp(sv) || SvNOKp(sv))) {
3216 (void) sv_2pv_flags(sv,&len, flags);
3218 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3222 (void) SvPV_force(sv,len);
3227 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3232 sv_force_normal_flags(sv, 0);
3235 if (PL_encoding && !(flags & SV_UTF8_NO_ENCODING)) {
3236 sv_recode_to_utf8(sv, PL_encoding);
3237 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3241 if (SvCUR(sv) == 0) {
3242 if (extra) SvGROW(sv, extra);
3243 } else { /* Assume Latin-1/EBCDIC */
3244 /* This function could be much more efficient if we
3245 * had a FLAG in SVs to signal if there are any variant
3246 * chars in the PV. Given that there isn't such a flag
3247 * make the loop as fast as possible (although there are certainly ways
3248 * to speed this up, eg. through vectorization) */
3249 U8 * s = (U8 *) SvPVX_const(sv);
3250 U8 * e = (U8 *) SvEND(sv);
3252 STRLEN two_byte_count = 0;
3254 if (flags & SV_FORCE_UTF8_UPGRADE) goto must_be_utf8;
3256 /* See if really will need to convert to utf8. We mustn't rely on our
3257 * incoming SV being well formed and having a trailing '\0', as certain
3258 * code in pp_formline can send us partially built SVs. */
3262 if (NATIVE_IS_INVARIANT(ch)) continue;
3264 t--; /* t already incremented; re-point to first variant */
3269 /* utf8 conversion not needed because all are invariants. Mark as
3270 * UTF-8 even if no variant - saves scanning loop */
3276 /* Here, the string should be converted to utf8, either because of an
3277 * input flag (two_byte_count = 0), or because a character that
3278 * requires 2 bytes was found (two_byte_count = 1). t points either to
3279 * the beginning of the string (if we didn't examine anything), or to
3280 * the first variant. In either case, everything from s to t - 1 will
3281 * occupy only 1 byte each on output.
3283 * There are two main ways to convert. One is to create a new string
3284 * and go through the input starting from the beginning, appending each
3285 * converted value onto the new string as we go along. It's probably
3286 * best to allocate enough space in the string for the worst possible
3287 * case rather than possibly running out of space and having to
3288 * reallocate and then copy what we've done so far. Since everything
3289 * from s to t - 1 is invariant, the destination can be initialized
3290 * with these using a fast memory copy
3292 * The other way is to figure out exactly how big the string should be
3293 * by parsing the entire input. Then you don't have to make it big
3294 * enough to handle the worst possible case, and more importantly, if
3295 * the string you already have is large enough, you don't have to
3296 * allocate a new string, you can copy the last character in the input
3297 * string to the final position(s) that will be occupied by the
3298 * converted string and go backwards, stopping at t, since everything
3299 * before that is invariant.
3301 * There are advantages and disadvantages to each method.
3303 * In the first method, we can allocate a new string, do the memory
3304 * copy from the s to t - 1, and then proceed through the rest of the
3305 * string byte-by-byte.
3307 * In the second method, we proceed through the rest of the input
3308 * string just calculating how big the converted string will be. Then
3309 * there are two cases:
3310 * 1) if the string has enough extra space to handle the converted
3311 * value. We go backwards through the string, converting until we
3312 * get to the position we are at now, and then stop. If this
3313 * position is far enough along in the string, this method is
3314 * faster than the other method. If the memory copy were the same
3315 * speed as the byte-by-byte loop, that position would be about
3316 * half-way, as at the half-way mark, parsing to the end and back
3317 * is one complete string's parse, the same amount as starting
3318 * over and going all the way through. Actually, it would be
3319 * somewhat less than half-way, as it's faster to just count bytes
3320 * than to also copy, and we don't have the overhead of allocating
3321 * a new string, changing the scalar to use it, and freeing the
3322 * existing one. But if the memory copy is fast, the break-even
3323 * point is somewhere after half way. The counting loop could be
3324 * sped up by vectorization, etc, to move the break-even point
3325 * further towards the beginning.
3326 * 2) if the string doesn't have enough space to handle the converted
3327 * value. A new string will have to be allocated, and one might
3328 * as well, given that, start from the beginning doing the first
3329 * method. We've spent extra time parsing the string and in
3330 * exchange all we've gotten is that we know precisely how big to
3331 * make the new one. Perl is more optimized for time than space,
3332 * so this case is a loser.
3333 * So what I've decided to do is not use the 2nd method unless it is
3334 * guaranteed that a new string won't have to be allocated, assuming
3335 * the worst case. I also decided not to put any more conditions on it
3336 * than this, for now. It seems likely that, since the worst case is
3337 * twice as big as the unknown portion of the string (plus 1), we won't
3338 * be guaranteed enough space, causing us to go to the first method,
3339 * unless the string is short, or the first variant character is near
3340 * the end of it. In either of these cases, it seems best to use the
3341 * 2nd method. The only circumstance I can think of where this would
3342 * be really slower is if the string had once had much more data in it
3343 * than it does now, but there is still a substantial amount in it */
3346 STRLEN invariant_head = t - s;
3347 STRLEN size = invariant_head + (e - t) * 2 + 1 + extra;
3348 if (SvLEN(sv) < size) {
3350 /* Here, have decided to allocate a new string */
3355 Newx(dst, size, U8);
3357 /* If no known invariants at the beginning of the input string,
3358 * set so starts from there. Otherwise, can use memory copy to
3359 * get up to where we are now, and then start from here */
3361 if (invariant_head <= 0) {
3364 Copy(s, dst, invariant_head, char);
3365 d = dst + invariant_head;
3369 const UV uv = NATIVE8_TO_UNI(*t++);
3370 if (UNI_IS_INVARIANT(uv))
3371 *d++ = (U8)UNI_TO_NATIVE(uv);
3373 *d++ = (U8)UTF8_EIGHT_BIT_HI(uv);
3374 *d++ = (U8)UTF8_EIGHT_BIT_LO(uv);
3378 SvPV_free(sv); /* No longer using pre-existing string */
3379 SvPV_set(sv, (char*)dst);
3380 SvCUR_set(sv, d - dst);
3381 SvLEN_set(sv, size);
3384 /* Here, have decided to get the exact size of the string.
3385 * Currently this happens only when we know that there is
3386 * guaranteed enough space to fit the converted string, so
3387 * don't have to worry about growing. If two_byte_count is 0,
3388 * then t points to the first byte of the string which hasn't
3389 * been examined yet. Otherwise two_byte_count is 1, and t
3390 * points to the first byte in the string that will expand to
3391 * two. Depending on this, start examining at t or 1 after t.
3394 U8 *d = t + two_byte_count;
3397 /* Count up the remaining bytes that expand to two */
3400 const U8 chr = *d++;
3401 if (! NATIVE_IS_INVARIANT(chr)) two_byte_count++;
3404 /* The string will expand by just the number of bytes that
3405 * occupy two positions. But we are one afterwards because of
3406 * the increment just above. This is the place to put the
3407 * trailing NUL, and to set the length before we decrement */
3409 d += two_byte_count;
3410 SvCUR_set(sv, d - s);
3414 /* Having decremented d, it points to the position to put the
3415 * very last byte of the expanded string. Go backwards through
3416 * the string, copying and expanding as we go, stopping when we
3417 * get to the part that is invariant the rest of the way down */
3421 const U8 ch = NATIVE8_TO_UNI(*e--);
3422 if (UNI_IS_INVARIANT(ch)) {
3423 *d-- = UNI_TO_NATIVE(ch);
3425 *d-- = (U8)UTF8_EIGHT_BIT_LO(ch);
3426 *d-- = (U8)UTF8_EIGHT_BIT_HI(ch);
3433 /* Mark as UTF-8 even if no variant - saves scanning loop */
3439 =for apidoc sv_utf8_downgrade
3441 Attempts to convert the PV of an SV from characters to bytes.
3442 If the PV contains a character that cannot fit
3443 in a byte, this conversion will fail;
3444 in this case, either returns false or, if C<fail_ok> is not
3447 This is not as a general purpose Unicode to byte encoding interface:
3448 use the Encode extension for that.
3454 Perl_sv_utf8_downgrade(pTHX_ register SV *const sv, const bool fail_ok)
3458 PERL_ARGS_ASSERT_SV_UTF8_DOWNGRADE;
3460 if (SvPOKp(sv) && SvUTF8(sv)) {
3466 sv_force_normal_flags(sv, 0);
3468 s = (U8 *) SvPV(sv, len);
3469 if (!utf8_to_bytes(s, &len)) {
3474 Perl_croak(aTHX_ "Wide character in %s",
3477 Perl_croak(aTHX_ "Wide character");
3488 =for apidoc sv_utf8_encode
3490 Converts the PV of an SV to UTF-8, but then turns the C<SvUTF8>
3491 flag off so that it looks like octets again.
3497 Perl_sv_utf8_encode(pTHX_ register SV *const sv)
3499 PERL_ARGS_ASSERT_SV_UTF8_ENCODE;
3502 sv_force_normal_flags(sv, 0);
3504 if (SvREADONLY(sv)) {
3505 Perl_croak_no_modify(aTHX);
3507 (void) sv_utf8_upgrade(sv);
3512 =for apidoc sv_utf8_decode
3514 If the PV of the SV is an octet sequence in UTF-8
3515 and contains a multiple-byte character, the C<SvUTF8> flag is turned on
3516 so that it looks like a character. If the PV contains only single-byte
3517 characters, the C<SvUTF8> flag stays being off.
3518 Scans PV for validity and returns false if the PV is invalid UTF-8.
3524 Perl_sv_utf8_decode(pTHX_ register SV *const sv)
3526 PERL_ARGS_ASSERT_SV_UTF8_DECODE;
3532 /* The octets may have got themselves encoded - get them back as
3535 if (!sv_utf8_downgrade(sv, TRUE))
3538 /* it is actually just a matter of turning the utf8 flag on, but
3539 * we want to make sure everything inside is valid utf8 first.
3541 c = (const U8 *) SvPVX_const(sv);
3542 if (!is_utf8_string(c, SvCUR(sv)+1))
3544 e = (const U8 *) SvEND(sv);
3547 if (!UTF8_IS_INVARIANT(ch)) {
3557 =for apidoc sv_setsv
3559 Copies the contents of the source SV C<ssv> into the destination SV
3560 C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3561 function if the source SV needs to be reused. Does not handle 'set' magic.
3562 Loosely speaking, it performs a copy-by-value, obliterating any previous
3563 content of the destination.
3565 You probably want to use one of the assortment of wrappers, such as
3566 C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3567 C<SvSetMagicSV_nosteal>.
3569 =for apidoc sv_setsv_flags
3571 Copies the contents of the source SV C<ssv> into the destination SV
3572 C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3573 function if the source SV needs to be reused. Does not handle 'set' magic.
3574 Loosely speaking, it performs a copy-by-value, obliterating any previous
3575 content of the destination.
3576 If the C<flags> parameter has the C<SV_GMAGIC> bit set, will C<mg_get> on
3577 C<ssv> if appropriate, else not. If the C<flags> parameter has the
3578 C<NOSTEAL> bit set then the buffers of temps will not be stolen. <sv_setsv>
3579 and C<sv_setsv_nomg> are implemented in terms of this function.
3581 You probably want to use one of the assortment of wrappers, such as
3582 C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3583 C<SvSetMagicSV_nosteal>.
3585 This is the primary function for copying scalars, and most other
3586 copy-ish functions and macros use this underneath.
3592 S_glob_assign_glob(pTHX_ SV *const dstr, SV *const sstr, const int dtype)
3594 I32 mro_changes = 0; /* 1 = method, 2 = isa, 3 = recursive isa */
3595 HV *old_stash = NULL;
3597 PERL_ARGS_ASSERT_GLOB_ASSIGN_GLOB;
3599 if (dtype != SVt_PVGV && !isGV_with_GP(dstr)) {
3600 const char * const name = GvNAME(sstr);
3601 const STRLEN len = GvNAMELEN(sstr);
3603 if (dtype >= SVt_PV) {
3609 SvUPGRADE(dstr, SVt_PVGV);
3610 (void)SvOK_off(dstr);
3611 /* FIXME - why are we doing this, then turning it off and on again
3613 isGV_with_GP_on(dstr);
3615 GvSTASH(dstr) = GvSTASH(sstr);
3617 Perl_sv_add_backref(aTHX_ MUTABLE_SV(GvSTASH(dstr)), dstr);
3618 gv_name_set(MUTABLE_GV(dstr), name, len, GV_ADD);
3619 SvFAKE_on(dstr); /* can coerce to non-glob */
3622 if(GvGP(MUTABLE_GV(sstr))) {
3623 /* If source has method cache entry, clear it */
3625 SvREFCNT_dec(GvCV(sstr));
3626 GvCV_set(sstr, NULL);
3629 /* If source has a real method, then a method is
3632 GvCV((const GV *)sstr) && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3638 /* If dest already had a real method, that's a change as well */
3640 !mro_changes && GvGP(MUTABLE_GV(dstr)) && GvCVu((const GV *)dstr)
3641 && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3646 /* We don’t need to check the name of the destination if it was not a
3647 glob to begin with. */
3648 if(dtype == SVt_PVGV) {
3649 const char * const name = GvNAME((const GV *)dstr);
3652 /* The stash may have been detached from the symbol table, so
3654 && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3655 && GvAV((const GV *)sstr)
3659 const STRLEN len = GvNAMELEN(dstr);
3660 if (len > 1 && name[len-2] == ':' && name[len-1] == ':') {
3663 /* Set aside the old stash, so we can reset isa caches on
3665 if((old_stash = GvHV(dstr)))
3666 /* Make sure we do not lose it early. */
3667 SvREFCNT_inc_simple_void_NN(
3668 sv_2mortal((SV *)old_stash)
3674 gp_free(MUTABLE_GV(dstr));
3675 isGV_with_GP_off(dstr);
3676 (void)SvOK_off(dstr);
3677 isGV_with_GP_on(dstr);
3678 GvINTRO_off(dstr); /* one-shot flag */
3679 GvGP_set(dstr, gp_ref(GvGP(sstr)));
3680 if (SvTAINTED(sstr))
3682 if (GvIMPORTED(dstr) != GVf_IMPORTED
3683 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3685 GvIMPORTED_on(dstr);
3688 if(mro_changes == 2) {
3690 SV * const sref = (SV *)GvAV((const GV *)dstr);
3691 if (SvSMAGICAL(sref) && (mg = mg_find(sref, PERL_MAGIC_isa))) {
3692 if (SvTYPE(mg->mg_obj) != SVt_PVAV) {
3693 AV * const ary = newAV();
3694 av_push(ary, mg->mg_obj); /* takes the refcount */
3695 mg->mg_obj = (SV *)ary;
3697 av_push((AV *)mg->mg_obj, SvREFCNT_inc_simple_NN(dstr));
3699 else sv_magic(sref, dstr, PERL_MAGIC_isa, NULL, 0);
3700 mro_isa_changed_in(GvSTASH(dstr));
3702 else if(mro_changes == 3) {
3703 HV * const stash = GvHV(dstr);
3704 if(old_stash ? (HV *)HvENAME_get(old_stash) : stash)
3710 else if(mro_changes) mro_method_changed_in(GvSTASH(dstr));
3715 S_glob_assign_ref(pTHX_ SV *const dstr, SV *const sstr)
3717 SV * const sref = SvREFCNT_inc(SvRV(sstr));
3719 const int intro = GvINTRO(dstr);
3722 const U32 stype = SvTYPE(sref);
3724 PERL_ARGS_ASSERT_GLOB_ASSIGN_REF;
3727 GvINTRO_off(dstr); /* one-shot flag */
3728 GvLINE(dstr) = CopLINE(PL_curcop);
3729 GvEGV(dstr) = MUTABLE_GV(dstr);
3734 location = (SV **) &(GvGP(dstr)->gp_cv); /* XXX bypassing GvCV_set */
3735 import_flag = GVf_IMPORTED_CV;
3738 location = (SV **) &GvHV(dstr);
3739 import_flag = GVf_IMPORTED_HV;
3742 location = (SV **) &GvAV(dstr);
3743 import_flag = GVf_IMPORTED_AV;
3746 location = (SV **) &GvIOp(dstr);
3749 location = (SV **) &GvFORM(dstr);
3752 location = &GvSV(dstr);
3753 import_flag = GVf_IMPORTED_SV;
3756 if (stype == SVt_PVCV) {
3757 /*if (GvCVGEN(dstr) && (GvCV(dstr) != (const CV *)sref || GvCVGEN(dstr))) {*/
3758 if (GvCVGEN(dstr)) {
3759 SvREFCNT_dec(GvCV(dstr));
3760 GvCV_set(dstr, NULL);
3761 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3764 SAVEGENERICSV(*location);
3768 if (stype == SVt_PVCV && (*location != sref || GvCVGEN(dstr))) {
3769 CV* const cv = MUTABLE_CV(*location);
3771 if (!GvCVGEN((const GV *)dstr) &&
3772 (CvROOT(cv) || CvXSUB(cv)))
3774 /* Redefining a sub - warning is mandatory if
3775 it was a const and its value changed. */
3776 if (CvCONST(cv) && CvCONST((const CV *)sref)
3778 == cv_const_sv((const CV *)sref)) {
3780 /* They are 2 constant subroutines generated from
3781 the same constant. This probably means that
3782 they are really the "same" proxy subroutine
3783 instantiated in 2 places. Most likely this is
3784 when a constant is exported twice. Don't warn.
3787 else if (ckWARN(WARN_REDEFINE)
3789 && (!CvCONST((const CV *)sref)
3790 || sv_cmp(cv_const_sv(cv),
3791 cv_const_sv((const CV *)
3793 Perl_warner(aTHX_ packWARN(WARN_REDEFINE),
3796 ? "Constant subroutine %s::%s redefined"
3797 : "Subroutine %s::%s redefined"),
3798 HvNAME_get(GvSTASH((const GV *)dstr)),
3799 GvENAME(MUTABLE_GV(dstr)));
3803 cv_ckproto_len(cv, (const GV *)dstr,
3804 SvPOK(sref) ? SvPVX_const(sref) : NULL,
3805 SvPOK(sref) ? SvCUR(sref) : 0);
3807 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3808 GvASSUMECV_on(dstr);
3809 if(GvSTASH(dstr)) mro_method_changed_in(GvSTASH(dstr)); /* sub foo { 1 } sub bar { 2 } *bar = \&foo */
3812 if (import_flag && !(GvFLAGS(dstr) & import_flag)
3813 && CopSTASH_ne(PL_curcop, GvSTASH(dstr))) {
3814 GvFLAGS(dstr) |= import_flag;
3816 if (stype == SVt_PVHV) {
3817 const char * const name = GvNAME((GV*)dstr);
3818 const STRLEN len = GvNAMELEN(dstr);
3820 len > 1 && name[len-2] == ':' && name[len-1] == ':'
3821 && (!dref || HvENAME_get(dref))
3824 (HV *)sref, (HV *)dref,
3830 stype == SVt_PVAV && sref != dref
3831 && strEQ(GvNAME((GV*)dstr), "ISA")
3832 /* The stash may have been detached from the symbol table, so
3833 check its name before doing anything. */
3834 && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3837 MAGIC * const omg = dref && SvSMAGICAL(dref)
3838 ? mg_find(dref, PERL_MAGIC_isa)
3840 if (SvSMAGICAL(sref) && (mg = mg_find(sref, PERL_MAGIC_isa))) {
3841 if (SvTYPE(mg->mg_obj) != SVt_PVAV) {
3842 AV * const ary = newAV();
3843 av_push(ary, mg->mg_obj); /* takes the refcount */
3844 mg->mg_obj = (SV *)ary;
3847 if (SvTYPE(omg->mg_obj) == SVt_PVAV) {
3848 SV **svp = AvARRAY((AV *)omg->mg_obj);
3849 I32 items = AvFILLp((AV *)omg->mg_obj) + 1;
3853 SvREFCNT_inc_simple_NN(*svp++)
3859 SvREFCNT_inc_simple_NN(omg->mg_obj)
3863 av_push((AV *)mg->mg_obj,SvREFCNT_inc_simple_NN(dstr));
3868 sref, omg ? omg->mg_obj : dstr, PERL_MAGIC_isa, NULL, 0
3870 mg = mg_find(sref, PERL_MAGIC_isa);
3872 /* Since the *ISA assignment could have affected more than
3873 one stash, don’t call mro_isa_changed_in directly, but let
3874 magic_clearisa do it for us, as it already has the logic for
3875 dealing with globs vs arrays of globs. */
3877 Perl_magic_clearisa(aTHX_ NULL, mg);
3882 if (SvTAINTED(sstr))
3888 Perl_sv_setsv_flags(pTHX_ SV *dstr, register SV* sstr, const I32 flags)
3891 register U32 sflags;
3893 register svtype stype;
3895 PERL_ARGS_ASSERT_SV_SETSV_FLAGS;
3900 if (SvIS_FREED(dstr)) {
3901 Perl_croak(aTHX_ "panic: attempt to copy value %" SVf
3902 " to a freed scalar %p", SVfARG(sstr), (void *)dstr);
3904 SV_CHECK_THINKFIRST_COW_DROP(dstr);
3906 sstr = &PL_sv_undef;
3907 if (SvIS_FREED(sstr)) {
3908 Perl_croak(aTHX_ "panic: attempt to copy freed scalar %p to %p",
3909 (void*)sstr, (void*)dstr);
3911 stype = SvTYPE(sstr);
3912 dtype = SvTYPE(dstr);
3914 (void)SvAMAGIC_off(dstr);
3917 /* need to nuke the magic */
3921 /* There's a lot of redundancy below but we're going for speed here */
3926 if (dtype != SVt_PVGV && dtype != SVt_PVLV) {
3927 (void)SvOK_off(dstr);
3935 sv_upgrade(dstr, SVt_IV);
3939 sv_upgrade(dstr, SVt_PVIV);
3943 goto end_of_first_switch;
3945 (void)SvIOK_only(dstr);
3946 SvIV_set(dstr, SvIVX(sstr));
3949 /* SvTAINTED can only be true if the SV has taint magic, which in
3950 turn means that the SV type is PVMG (or greater). This is the
3951 case statement for SVt_IV, so this cannot be true (whatever gcov
3953 assert(!SvTAINTED(sstr));
3958 if (dtype < SVt_PV && dtype != SVt_IV)
3959 sv_upgrade(dstr, SVt_IV);
3967 sv_upgrade(dstr, SVt_NV);
3971 sv_upgrade(dstr, SVt_PVNV);
3975 goto end_of_first_switch;
3977 SvNV_set(dstr, SvNVX(sstr));
3978 (void)SvNOK_only(dstr);
3979 /* SvTAINTED can only be true if the SV has taint magic, which in
3980 turn means that the SV type is PVMG (or greater). This is the
3981 case statement for SVt_NV, so this cannot be true (whatever gcov
3983 assert(!SvTAINTED(sstr));
3989 #ifdef PERL_OLD_COPY_ON_WRITE
3990 if ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS) {
3991 if (dtype < SVt_PVIV)
3992 sv_upgrade(dstr, SVt_PVIV);
3999 sv_upgrade(dstr, SVt_PV);
4002 if (dtype < SVt_PVIV)
4003 sv_upgrade(dstr, SVt_PVIV);
4006 if (dtype < SVt_PVNV)
4007 sv_upgrade(dstr, SVt_PVNV);
4011 const char * const type = sv_reftype(sstr,0);
4013 Perl_croak(aTHX_ "Bizarre copy of %s in %s", type, OP_DESC(PL_op));
4015 Perl_croak(aTHX_ "Bizarre copy of %s", type);
4020 if (dtype < SVt_REGEXP)
4021 sv_upgrade(dstr, SVt_REGEXP);
4024 /* case SVt_BIND: */
4027 /* SvVALID means that this PVGV is playing at being an FBM. */
4030 if (SvGMAGICAL(sstr) && (flags & SV_GMAGIC)) {
4032 if (SvTYPE(sstr) != stype)
4033 stype = SvTYPE(sstr);
4035 if (isGV_with_GP(sstr) && dtype <= SVt_PVLV) {
4036 glob_assign_glob(dstr, sstr, dtype);
4039 if (stype == SVt_PVLV)
4040 SvUPGRADE(dstr, SVt_PVNV);
4042 SvUPGRADE(dstr, (svtype)stype);
4044 end_of_first_switch:
4046 /* dstr may have been upgraded. */
4047 dtype = SvTYPE(dstr);
4048 sflags = SvFLAGS(sstr);
4050 if (dtype == SVt_PVCV || dtype == SVt_PVFM) {
4051 /* Assigning to a subroutine sets the prototype. */
4054 const char *const ptr = SvPV_const(sstr, len);
4056 SvGROW(dstr, len + 1);
4057 Copy(ptr, SvPVX(dstr), len + 1, char);
4058 SvCUR_set(dstr, len);
4060 SvFLAGS(dstr) |= sflags & SVf_UTF8;
4064 } else if (dtype == SVt_PVAV || dtype == SVt_PVHV) {
4065 const char * const type = sv_reftype(dstr,0);
4067 Perl_croak(aTHX_ "Cannot copy to %s in %s", type, OP_DESC(PL_op));
4069 Perl_croak(aTHX_ "Cannot copy to %s", type);
4070 } else if (sflags & SVf_ROK) {
4071 if (isGV_with_GP(dstr)
4072 && SvTYPE(SvRV(sstr)) == SVt_PVGV && isGV_with_GP(SvRV(sstr))) {
4075 if (GvIMPORTED(dstr) != GVf_IMPORTED
4076 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
4078 GvIMPORTED_on(dstr);
4083 glob_assign_glob(dstr, sstr, dtype);
4087 if (dtype >= SVt_PV) {
4088 if (isGV_with_GP(dstr)) {
4089 glob_assign_ref(dstr, sstr);
4092 if (SvPVX_const(dstr)) {
4098 (void)SvOK_off(dstr);
4099 SvRV_set(dstr, SvREFCNT_inc(SvRV(sstr)));
4100 SvFLAGS(dstr) |= sflags & SVf_ROK;
4101 assert(!(sflags & SVp_NOK));
4102 assert(!(sflags & SVp_IOK));
4103 assert(!(sflags & SVf_NOK));
4104 assert(!(sflags & SVf_IOK));
4106 else if (isGV_with_GP(dstr)) {
4107 if (!(sflags & SVf_OK)) {
4108 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
4109 "Undefined value assigned to typeglob");
4112 GV *gv = gv_fetchsv(sstr, GV_ADD, SVt_PVGV);
4113 if (dstr != (const SV *)gv) {
4114 const char * const name = GvNAME((const GV *)dstr);
4115 const STRLEN len = GvNAMELEN(dstr);
4116 HV *old_stash = NULL;
4117 bool reset_isa = FALSE;
4118 if (len > 1 && name[len-2] == ':' && name[len-1] == ':') {
4119 /* Set aside the old stash, so we can reset isa caches
4120 on its subclasses. */
4121 if((old_stash = GvHV(dstr))) {
4122 /* Make sure we do not lose it early. */
4123 SvREFCNT_inc_simple_void_NN(
4124 sv_2mortal((SV *)old_stash)
4131 gp_free(MUTABLE_GV(dstr));
4132 GvGP_set(dstr, gp_ref(GvGP(gv)));
4135 HV * const stash = GvHV(dstr);
4137 old_stash ? (HV *)HvENAME_get(old_stash) : stash
4147 else if (dtype == SVt_REGEXP && stype == SVt_REGEXP) {
4148 reg_temp_copy((REGEXP*)dstr, (REGEXP*)sstr);
4150 else if (sflags & SVp_POK) {
4154 * Check to see if we can just swipe the string. If so, it's a
4155 * possible small lose on short strings, but a big win on long ones.
4156 * It might even be a win on short strings if SvPVX_const(dstr)
4157 * has to be allocated and SvPVX_const(sstr) has to be freed.
4158 * Likewise if we can set up COW rather than doing an actual copy, we
4159 * drop to the else clause, as the swipe code and the COW setup code
4160 * have much in common.
4163 /* Whichever path we take through the next code, we want this true,
4164 and doing it now facilitates the COW check. */
4165 (void)SvPOK_only(dstr);
4168 /* If we're already COW then this clause is not true, and if COW
4169 is allowed then we drop down to the else and make dest COW
4170 with us. If caller hasn't said that we're allowed to COW
4171 shared hash keys then we don't do the COW setup, even if the
4172 source scalar is a shared hash key scalar. */
4173 (((flags & SV_COW_SHARED_HASH_KEYS)
4174 ? (sflags & (SVf_FAKE|SVf_READONLY)) != (SVf_FAKE|SVf_READONLY)
4175 : 1 /* If making a COW copy is forbidden then the behaviour we
4176 desire is as if the source SV isn't actually already
4177 COW, even if it is. So we act as if the source flags
4178 are not COW, rather than actually testing them. */
4180 #ifndef PERL_OLD_COPY_ON_WRITE
4181 /* The change that added SV_COW_SHARED_HASH_KEYS makes the logic
4182 when PERL_OLD_COPY_ON_WRITE is defined a little wrong.
4183 Conceptually PERL_OLD_COPY_ON_WRITE being defined should
4184 override SV_COW_SHARED_HASH_KEYS, because it means "always COW"
4185 but in turn, it's somewhat dead code, never expected to go
4186 live, but more kept as a placeholder on how to do it better
4187 in a newer implementation. */
4188 /* If we are COW and dstr is a suitable target then we drop down
4189 into the else and make dest a COW of us. */
4190 || (SvFLAGS(dstr) & CAN_COW_MASK) != CAN_COW_FLAGS
4195 (sflags & SVs_TEMP) && /* slated for free anyway? */
4196 !(sflags & SVf_OOK) && /* and not involved in OOK hack? */
4197 (!(flags & SV_NOSTEAL)) &&
4198 /* and we're allowed to steal temps */
4199 SvREFCNT(sstr) == 1 && /* and no other references to it? */
4200 SvLEN(sstr)) /* and really is a string */
4201 #ifdef PERL_OLD_COPY_ON_WRITE
4202 && ((flags & SV_COW_SHARED_HASH_KEYS)
4203 ? (!((sflags & CAN_COW_MASK) == CAN_COW_FLAGS
4204 && (SvFLAGS(dstr) & CAN_COW_MASK) == CAN_COW_FLAGS
4205 && SvTYPE(sstr) >= SVt_PVIV && SvTYPE(sstr) != SVt_PVFM))
4209 /* Failed the swipe test, and it's not a shared hash key either.
4210 Have to copy the string. */
4211 STRLEN len = SvCUR(sstr);
4212 SvGROW(dstr, len + 1); /* inlined from sv_setpvn */
4213 Move(SvPVX_const(sstr),SvPVX(dstr),len,char);
4214 SvCUR_set(dstr, len);
4215 *SvEND(dstr) = '\0';
4217 /* If PERL_OLD_COPY_ON_WRITE is not defined, then isSwipe will always
4219 /* Either it's a shared hash key, or it's suitable for
4220 copy-on-write or we can swipe the string. */
4222 PerlIO_printf(Perl_debug_log, "Copy on write: sstr --> dstr\n");
4226 #ifdef PERL_OLD_COPY_ON_WRITE
4228 if ((sflags & (SVf_FAKE | SVf_READONLY))
4229 != (SVf_FAKE | SVf_READONLY)) {
4230 SvREADONLY_on(sstr);
4232 /* Make the source SV into a loop of 1.
4233 (about to become 2) */
4234 SV_COW_NEXT_SV_SET(sstr, sstr);
4238 /* Initial code is common. */
4239 if (SvPVX_const(dstr)) { /* we know that dtype >= SVt_PV */
4244 /* making another shared SV. */
4245 STRLEN cur = SvCUR(sstr);
4246 STRLEN len = SvLEN(sstr);
4247 #ifdef PERL_OLD_COPY_ON_WRITE
4249 assert (SvTYPE(dstr) >= SVt_PVIV);
4250 /* SvIsCOW_normal */
4251 /* splice us in between source and next-after-source. */
4252 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4253 SV_COW_NEXT_SV_SET(sstr, dstr);
4254 SvPV_set(dstr, SvPVX_mutable(sstr));
4258 /* SvIsCOW_shared_hash */
4259 DEBUG_C(PerlIO_printf(Perl_debug_log,
4260 "Copy on write: Sharing hash\n"));
4262 assert (SvTYPE(dstr) >= SVt_PV);
4264 HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)))));
4266 SvLEN_set(dstr, len);
4267 SvCUR_set(dstr, cur);
4268 SvREADONLY_on(dstr);
4272 { /* Passes the swipe test. */
4273 SvPV_set(dstr, SvPVX_mutable(sstr));
4274 SvLEN_set(dstr, SvLEN(sstr));
4275 SvCUR_set(dstr, SvCUR(sstr));
4278 (void)SvOK_off(sstr); /* NOTE: nukes most SvFLAGS on sstr */
4279 SvPV_set(sstr, NULL);
4285 if (sflags & SVp_NOK) {
4286 SvNV_set(dstr, SvNVX(sstr));
4288 if (sflags & SVp_IOK) {
4289 SvIV_set(dstr, SvIVX(sstr));
4290 /* Must do this otherwise some other overloaded use of 0x80000000
4291 gets confused. I guess SVpbm_VALID */
4292 if (sflags & SVf_IVisUV)
4295 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_NOK|SVp_NOK|SVf_UTF8);
4297 const MAGIC * const smg = SvVSTRING_mg(sstr);
4299 sv_magic(dstr, NULL, PERL_MAGIC_vstring,
4300 smg->mg_ptr, smg->mg_len);
4301 SvRMAGICAL_on(dstr);
4305 else if (sflags & (SVp_IOK|SVp_NOK)) {
4306 (void)SvOK_off(dstr);
4307 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_IVisUV|SVf_NOK|SVp_NOK);
4308 if (sflags & SVp_IOK) {
4309 /* XXXX Do we want to set IsUV for IV(ROK)? Be extra safe... */
4310 SvIV_set(dstr, SvIVX(sstr));
4312 if (sflags & SVp_NOK) {
4313 SvNV_set(dstr, SvNVX(sstr));
4317 if (isGV_with_GP(sstr)) {
4318 /* This stringification rule for globs is spread in 3 places.
4319 This feels bad. FIXME. */
4320 const U32 wasfake = sflags & SVf_FAKE;
4322 /* FAKE globs can get coerced, so need to turn this off
4323 temporarily if it is on. */
4325 gv_efullname3(dstr, MUTABLE_GV(sstr), "*");
4326 SvFLAGS(sstr) |= wasfake;
4329 (void)SvOK_off(dstr);
4331 if (SvTAINTED(sstr))
4336 =for apidoc sv_setsv_mg
4338 Like C<sv_setsv>, but also handles 'set' magic.
4344 Perl_sv_setsv_mg(pTHX_ SV *const dstr, register SV *const sstr)
4346 PERL_ARGS_ASSERT_SV_SETSV_MG;
4348 sv_setsv(dstr,sstr);
4352 #ifdef PERL_OLD_COPY_ON_WRITE
4354 Perl_sv_setsv_cow(pTHX_ SV *dstr, SV *sstr)
4356 STRLEN cur = SvCUR(sstr);
4357 STRLEN len = SvLEN(sstr);
4358 register char *new_pv;
4360 PERL_ARGS_ASSERT_SV_SETSV_COW;
4363 PerlIO_printf(Perl_debug_log, "Fast copy on write: %p -> %p\n",
4364 (void*)sstr, (void*)dstr);
4371 if (SvTHINKFIRST(dstr))
4372 sv_force_normal_flags(dstr, SV_COW_DROP_PV);
4373 else if (SvPVX_const(dstr))
4374 Safefree(SvPVX_const(dstr));
4378 SvUPGRADE(dstr, SVt_PVIV);
4380 assert (SvPOK(sstr));
4381 assert (SvPOKp(sstr));
4382 assert (!SvIOK(sstr));
4383 assert (!SvIOKp(sstr));
4384 assert (!SvNOK(sstr));
4385 assert (!SvNOKp(sstr));
4387 if (SvIsCOW(sstr)) {
4389 if (SvLEN(sstr) == 0) {
4390 /* source is a COW shared hash key. */
4391 DEBUG_C(PerlIO_printf(Perl_debug_log,
4392 "Fast copy on write: Sharing hash\n"));
4393 new_pv = HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr))));
4396 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4398 assert ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS);
4399 SvUPGRADE(sstr, SVt_PVIV);
4400 SvREADONLY_on(sstr);
4402 DEBUG_C(PerlIO_printf(Perl_debug_log,
4403 "Fast copy on write: Converting sstr to COW\n"));
4404 SV_COW_NEXT_SV_SET(dstr, sstr);
4406 SV_COW_NEXT_SV_SET(sstr, dstr);
4407 new_pv = SvPVX_mutable(sstr);
4410 SvPV_set(dstr, new_pv);
4411 SvFLAGS(dstr) = (SVt_PVIV|SVf_POK|SVp_POK|SVf_FAKE|SVf_READONLY);
4414 SvLEN_set(dstr, len);
4415 SvCUR_set(dstr, cur);
4424 =for apidoc sv_setpvn
4426 Copies a string into an SV. The C<len> parameter indicates the number of
4427 bytes to be copied. If the C<ptr> argument is NULL the SV will become
4428 undefined. Does not handle 'set' magic. See C<sv_setpvn_mg>.
4434 Perl_sv_setpvn(pTHX_ register SV *const sv, register const char *const ptr, register const STRLEN len)
4437 register char *dptr;
4439 PERL_ARGS_ASSERT_SV_SETPVN;
4441 SV_CHECK_THINKFIRST_COW_DROP(sv);
4447 /* len is STRLEN which is unsigned, need to copy to signed */
4450 Perl_croak(aTHX_ "panic: sv_setpvn called with negative strlen");
4452 SvUPGRADE(sv, SVt_PV);
4454 dptr = SvGROW(sv, len + 1);
4455 Move(ptr,dptr,len,char);
4458 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4463 =for apidoc sv_setpvn_mg
4465 Like C<sv_setpvn>, but also handles 'set' magic.
4471 Perl_sv_setpvn_mg(pTHX_ register SV *const sv, register const char *const ptr, register const STRLEN len)
4473 PERL_ARGS_ASSERT_SV_SETPVN_MG;
4475 sv_setpvn(sv,ptr,len);
4480 =for apidoc sv_setpv
4482 Copies a string into an SV. The string must be null-terminated. Does not
4483 handle 'set' magic. See C<sv_setpv_mg>.
4489 Perl_sv_setpv(pTHX_ register SV *const sv, register const char *const ptr)
4492 register STRLEN len;
4494 PERL_ARGS_ASSERT_SV_SETPV;
4496 SV_CHECK_THINKFIRST_COW_DROP(sv);
4502 SvUPGRADE(sv, SVt_PV);
4504 SvGROW(sv, len + 1);
4505 Move(ptr,SvPVX(sv),len+1,char);
4507 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4512 =for apidoc sv_setpv_mg
4514 Like C<sv_setpv>, but also handles 'set' magic.
4520 Perl_sv_setpv_mg(pTHX_ register SV *const sv, register const char *const ptr)
4522 PERL_ARGS_ASSERT_SV_SETPV_MG;
4529 =for apidoc sv_usepvn_flags
4531 Tells an SV to use C<ptr> to find its string value. Normally the
4532 string is stored inside the SV but sv_usepvn allows the SV to use an
4533 outside string. The C<ptr> should point to memory that was allocated
4534 by C<malloc>. The string length, C<len>, must be supplied. By default
4535 this function will realloc (i.e. move) the memory pointed to by C<ptr>,
4536 so that pointer should not be freed or used by the programmer after
4537 giving it to sv_usepvn, and neither should any pointers from "behind"
4538 that pointer (e.g. ptr + 1) be used.
4540 If C<flags> & SV_SMAGIC is true, will call SvSETMAGIC. If C<flags> &
4541 SV_HAS_TRAILING_NUL is true, then C<ptr[len]> must be NUL, and the realloc
4542 will be skipped. (i.e. the buffer is actually at least 1 byte longer than
4543 C<len>, and already meets the requirements for storing in C<SvPVX>)
4549 Perl_sv_usepvn_flags(pTHX_ SV *const sv, char *ptr, const STRLEN len, const U32 flags)
4554 PERL_ARGS_ASSERT_SV_USEPVN_FLAGS;
4556 SV_CHECK_THINKFIRST_COW_DROP(sv);
4557 SvUPGRADE(sv, SVt_PV);
4560 if (flags & SV_SMAGIC)
4564 if (SvPVX_const(sv))
4568 if (flags & SV_HAS_TRAILING_NUL)
4569 assert(ptr[len] == '\0');
4572 allocate = (flags & SV_HAS_TRAILING_NUL)
4574 #ifdef Perl_safesysmalloc_size
4577 PERL_STRLEN_ROUNDUP(len + 1);
4579 if (flags & SV_HAS_TRAILING_NUL) {
4580 /* It's long enough - do nothing.
4581 Specifically Perl_newCONSTSUB is relying on this. */
4584 /* Force a move to shake out bugs in callers. */
4585 char *new_ptr = (char*)safemalloc(allocate);
4586 Copy(ptr, new_ptr, len, char);
4587 PoisonFree(ptr,len,char);
4591 ptr = (char*) saferealloc (ptr, allocate);
4594 #ifdef Perl_safesysmalloc_size
4595 SvLEN_set(sv, Perl_safesysmalloc_size(ptr));
4597 SvLEN_set(sv, allocate);
4601 if (!(flags & SV_HAS_TRAILING_NUL)) {
4604 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4606 if (flags & SV_SMAGIC)
4610 #ifdef PERL_OLD_COPY_ON_WRITE
4611 /* Need to do this *after* making the SV normal, as we need the buffer
4612 pointer to remain valid until after we've copied it. If we let go too early,
4613 another thread could invalidate it by unsharing last of the same hash key
4614 (which it can do by means other than releasing copy-on-write Svs)
4615 or by changing the other copy-on-write SVs in the loop. */
4617 S_sv_release_COW(pTHX_ register SV *sv, const char *pvx, SV *after)
4619 PERL_ARGS_ASSERT_SV_RELEASE_COW;
4621 { /* this SV was SvIsCOW_normal(sv) */
4622 /* we need to find the SV pointing to us. */
4623 SV *current = SV_COW_NEXT_SV(after);
4625 if (current == sv) {
4626 /* The SV we point to points back to us (there were only two of us
4628 Hence other SV is no longer copy on write either. */
4630 SvREADONLY_off(after);
4632 /* We need to follow the pointers around the loop. */
4634 while ((next = SV_COW_NEXT_SV(current)) != sv) {
4637 /* don't loop forever if the structure is bust, and we have
4638 a pointer into a closed loop. */
4639 assert (current != after);
4640 assert (SvPVX_const(current) == pvx);
4642 /* Make the SV before us point to the SV after us. */
4643 SV_COW_NEXT_SV_SET(current, after);
4649 =for apidoc sv_force_normal_flags
4651 Undo various types of fakery on an SV: if the PV is a shared string, make
4652 a private copy; if we're a ref, stop refing; if we're a glob, downgrade to
4653 an xpvmg; if we're a copy-on-write scalar, this is the on-write time when
4654 we do the copy, and is also used locally. If C<SV_COW_DROP_PV> is set
4655 then a copy-on-write scalar drops its PV buffer (if any) and becomes
4656 SvPOK_off rather than making a copy. (Used where this scalar is about to be
4657 set to some other value.) In addition, the C<flags> parameter gets passed to
4658 C<sv_unref_flags()> when unrefing. C<sv_force_normal> calls this function
4659 with flags set to 0.
4665 Perl_sv_force_normal_flags(pTHX_ register SV *const sv, const U32 flags)
4669 PERL_ARGS_ASSERT_SV_FORCE_NORMAL_FLAGS;
4671 #ifdef PERL_OLD_COPY_ON_WRITE
4672 if (SvREADONLY(sv)) {
4674 const char * const pvx = SvPVX_const(sv);
4675 const STRLEN len = SvLEN(sv);
4676 const STRLEN cur = SvCUR(sv);
4677 /* next COW sv in the loop. If len is 0 then this is a shared-hash
4678 key scalar, so we mustn't attempt to call SV_COW_NEXT_SV(), as
4679 we'll fail an assertion. */
4680 SV * const next = len ? SV_COW_NEXT_SV(sv) : 0;
4683 PerlIO_printf(Perl_debug_log,
4684 "Copy on write: Force normal %ld\n",
4690 /* This SV doesn't own the buffer, so need to Newx() a new one: */
4693 if (flags & SV_COW_DROP_PV) {
4694 /* OK, so we don't need to copy our buffer. */
4697 SvGROW(sv, cur + 1);
4698 Move(pvx,SvPVX(sv),cur,char);
4703 sv_release_COW(sv, pvx, next);
4705 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4711 else if (IN_PERL_RUNTIME)
4712 Perl_croak_no_modify(aTHX);
4715 if (SvREADONLY(sv)) {
4717 const char * const pvx = SvPVX_const(sv);
4718 const STRLEN len = SvCUR(sv);
4723 SvGROW(sv, len + 1);
4724 Move(pvx,SvPVX(sv),len,char);
4726 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4728 else if (IN_PERL_RUNTIME)
4729 Perl_croak_no_modify(aTHX);
4733 sv_unref_flags(sv, flags);
4734 else if (SvFAKE(sv) && isGV_with_GP(sv))
4736 else if (SvFAKE(sv) && SvTYPE(sv) == SVt_REGEXP) {
4737 /* Need to downgrade the REGEXP to a simple(r) scalar. This is analogous
4738 to sv_unglob. We only need it here, so inline it. */
4739 const svtype new_type = SvMAGIC(sv) || SvSTASH(sv) ? SVt_PVMG : SVt_PV;
4740 SV *const temp = newSV_type(new_type);
4741 void *const temp_p = SvANY(sv);
4743 if (new_type == SVt_PVMG) {
4744 SvMAGIC_set(temp, SvMAGIC(sv));
4745 SvMAGIC_set(sv, NULL);
4746 SvSTASH_set(temp, SvSTASH(sv));
4747 SvSTASH_set(sv, NULL);
4749 SvCUR_set(temp, SvCUR(sv));
4750 /* Remember that SvPVX is in the head, not the body. */
4752 SvLEN_set(temp, SvLEN(sv));
4753 /* This signals "buffer is owned by someone else" in sv_clear,
4754 which is the least effort way to stop it freeing the buffer.
4756 SvLEN_set(sv, SvLEN(sv)+1);
4758 /* Their buffer is already owned by someone else. */
4759 SvPVX(sv) = savepvn(SvPVX(sv), SvCUR(sv));
4760 SvLEN_set(temp, SvCUR(sv)+1);
4763 /* Now swap the rest of the bodies. */
4765 SvFLAGS(sv) &= ~(SVf_FAKE|SVTYPEMASK);
4766 SvFLAGS(sv) |= new_type;
4767 SvANY(sv) = SvANY(temp);
4769 SvFLAGS(temp) &= ~(SVTYPEMASK);
4770 SvFLAGS(temp) |= SVt_REGEXP|SVf_FAKE;
4771 SvANY(temp) = temp_p;
4780 Efficient removal of characters from the beginning of the string buffer.
4781 SvPOK(sv) must be true and the C<ptr> must be a pointer to somewhere inside
4782 the string buffer. The C<ptr> becomes the first character of the adjusted
4783 string. Uses the "OOK hack".
4784 Beware: after this function returns, C<ptr> and SvPVX_const(sv) may no longer
4785 refer to the same chunk of data.
4791 Perl_sv_chop(pTHX_ register SV *const sv, register const char *const ptr)
4797 const U8 *real_start;
4801 PERL_ARGS_ASSERT_SV_CHOP;
4803 if (!ptr || !SvPOKp(sv))
4805 delta = ptr - SvPVX_const(sv);
4807 /* Nothing to do. */
4810 /* SvPVX(sv) may move in SV_CHECK_THINKFIRST(sv), but after this line,
4811 nothing uses the value of ptr any more. */
4812 max_delta = SvLEN(sv) ? SvLEN(sv) : SvCUR(sv);
4813 if (ptr <= SvPVX_const(sv))
4814 Perl_croak(aTHX_ "panic: sv_chop ptr=%p, start=%p, end=%p",
4815 ptr, SvPVX_const(sv), SvPVX_const(sv) + max_delta);
4816 SV_CHECK_THINKFIRST(sv);
4817 if (delta > max_delta)
4818 Perl_croak(aTHX_ "panic: sv_chop ptr=%p (was %p), start=%p, end=%p",
4819 SvPVX_const(sv) + delta, ptr, SvPVX_const(sv),
4820 SvPVX_const(sv) + max_delta);
4823 if (!SvLEN(sv)) { /* make copy of shared string */
4824 const char *pvx = SvPVX_const(sv);
4825 const STRLEN len = SvCUR(sv);
4826 SvGROW(sv, len + 1);
4827 Move(pvx,SvPVX(sv),len,char);
4830 SvFLAGS(sv) |= SVf_OOK;
4833 SvOOK_offset(sv, old_delta);
4835 SvLEN_set(sv, SvLEN(sv) - delta);
4836 SvCUR_set(sv, SvCUR(sv) - delta);
4837 SvPV_set(sv, SvPVX(sv) + delta);
4839 p = (U8 *)SvPVX_const(sv);
4844 real_start = p - delta;
4848 if (delta < 0x100) {
4852 p -= sizeof(STRLEN);
4853 Copy((U8*)&delta, p, sizeof(STRLEN), U8);
4857 /* Fill the preceding buffer with sentinals to verify that no-one is
4859 while (p > real_start) {
4867 =for apidoc sv_catpvn
4869 Concatenates the string onto the end of the string which is in the SV. The
4870 C<len> indicates number of bytes to copy. If the SV has the UTF-8
4871 status set, then the bytes appended should be valid UTF-8.
4872 Handles 'get' magic, but not 'set' magic. See C<sv_catpvn_mg>.
4874 =for apidoc sv_catpvn_flags
4876 Concatenates the string onto the end of the string which is in the SV. The
4877 C<len> indicates number of bytes to copy. If the SV has the UTF-8
4878 status set, then the bytes appended should be valid UTF-8.
4879 If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<dsv> if
4880 appropriate, else not. C<sv_catpvn> and C<sv_catpvn_nomg> are implemented
4881 in terms of this function.
4887 Perl_sv_catpvn_flags(pTHX_ register SV *const dsv, register const char *sstr, register const STRLEN slen, const I32 flags)
4891 const char * const dstr = SvPV_force_flags(dsv, dlen, flags);
4893 PERL_ARGS_ASSERT_SV_CATPVN_FLAGS;
4895 SvGROW(dsv, dlen + slen + 1);
4897 sstr = SvPVX_const(dsv);
4898 Move(sstr, SvPVX(dsv) + dlen, slen, char);
4899 SvCUR_set(dsv, SvCUR(dsv) + slen);
4901 (void)SvPOK_only_UTF8(dsv); /* validate pointer */
4903 if (flags & SV_SMAGIC)
4908 =for apidoc sv_catsv
4910 Concatenates the string from SV C<ssv> onto the end of the string in
4911 SV C<dsv>. Modifies C<dsv> but not C<ssv>. Handles 'get' magic, but
4912 not 'set' magic. See C<sv_catsv_mg>.
4914 =for apidoc sv_catsv_flags
4916 Concatenates the string from SV C<ssv> onto the end of the string in
4917 SV C<dsv>. Modifies C<dsv> but not C<ssv>. If C<flags> has C<SV_GMAGIC>
4918 bit set, will C<mg_get> on the SVs if appropriate, else not. C<sv_catsv>
4919 and C<sv_catsv_nomg> are implemented in terms of this function.
4924 Perl_sv_catsv_flags(pTHX_ SV *const dsv, register SV *const ssv, const I32 flags)
4928 PERL_ARGS_ASSERT_SV_CATSV_FLAGS;
4932 const char *spv = SvPV_flags_const(ssv, slen, flags);
4934 /* sutf8 and dutf8 were type bool, but under USE_ITHREADS,
4935 gcc version 2.95.2 20000220 (Debian GNU/Linux) for
4936 Linux xxx 2.2.17 on sparc64 with gcc -O2, we erroneously
4937 get dutf8 = 0x20000000, (i.e. SVf_UTF8) even though
4938 dsv->sv_flags doesn't have that bit set.
4939 Andy Dougherty 12 Oct 2001
4941 const I32 sutf8 = DO_UTF8(ssv);
4944 if (SvGMAGICAL(dsv) && (flags & SV_GMAGIC))
4946 dutf8 = DO_UTF8(dsv);
4948 if (dutf8 != sutf8) {
4950 /* Not modifying source SV, so taking a temporary copy. */
4951 SV* const csv = newSVpvn_flags(spv, slen, SVs_TEMP);
4953 sv_utf8_upgrade(csv);
4954 spv = SvPV_const(csv, slen);
4957 /* Leave enough space for the cat that's about to happen */
4958 sv_utf8_upgrade_flags_grow(dsv, 0, slen);
4960 sv_catpvn_nomg(dsv, spv, slen);
4963 if (flags & SV_SMAGIC)
4968 =for apidoc sv_catpv
4970 Concatenates the string onto the end of the string which is in the SV.
4971 If the SV has the UTF-8 status set, then the bytes appended should be
4972 valid UTF-8. Handles 'get' magic, but not 'set' magic. See C<sv_catpv_mg>.
4977 Perl_sv_catpv(pTHX_ register SV *const sv, register const char *ptr)
4980 register STRLEN len;
4984 PERL_ARGS_ASSERT_SV_CATPV;
4988 junk = SvPV_force(sv, tlen);
4990 SvGROW(sv, tlen + len + 1);
4992 ptr = SvPVX_const(sv);
4993 Move(ptr,SvPVX(sv)+tlen,len+1,char);
4994 SvCUR_set(sv, SvCUR(sv) + len);
4995 (void)SvPOK_only_UTF8(sv); /* validate pointer */
5000 =for apidoc sv_catpv_flags
5002 Concatenates the string onto the end of the string which is in the SV.
5003 If the SV has the UTF-8 status set, then the bytes appended should
5004 be valid UTF-8. If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get>
5005 on the SVs if appropriate, else not.
5011 Perl_sv_catpv_flags(pTHX_ SV *dstr, const char *sstr, const I32 flags)
5013 PERL_ARGS_ASSERT_SV_CATPV_FLAGS;
5014 sv_catpvn_flags(dstr, sstr, strlen(sstr), flags);
5018 =for apidoc sv_catpv_mg
5020 Like C<sv_catpv>, but also handles 'set' magic.
5026 Perl_sv_catpv_mg(pTHX_ register SV *const sv, register const char *const ptr)
5028 PERL_ARGS_ASSERT_SV_CATPV_MG;
5037 Creates a new SV. A non-zero C<len> parameter indicates the number of
5038 bytes of preallocated string space the SV should have. An extra byte for a
5039 trailing NUL is also reserved. (SvPOK is not set for the SV even if string
5040 space is allocated.) The reference count for the new SV is set to 1.
5042 In 5.9.3, newSV() replaces the older NEWSV() API, and drops the first
5043 parameter, I<x>, a debug aid which allowed callers to identify themselves.
5044 This aid has been superseded by a new build option, PERL_MEM_LOG (see
5045 L<perlhack/PERL_MEM_LOG>). The older API is still there for use in XS
5046 modules supporting older perls.
5052 Perl_newSV(pTHX_ const STRLEN len)
5059 sv_upgrade(sv, SVt_PV);
5060 SvGROW(sv, len + 1);
5065 =for apidoc sv_magicext
5067 Adds magic to an SV, upgrading it if necessary. Applies the
5068 supplied vtable and returns a pointer to the magic added.
5070 Note that C<sv_magicext> will allow things that C<sv_magic> will not.
5071 In particular, you can add magic to SvREADONLY SVs, and add more than
5072 one instance of the same 'how'.
5074 If C<namlen> is greater than zero then a C<savepvn> I<copy> of C<name> is
5075 stored, if C<namlen> is zero then C<name> is stored as-is and - as another
5076 special case - if C<(name && namlen == HEf_SVKEY)> then C<name> is assumed
5077 to contain an C<SV*> and is stored as-is with its REFCNT incremented.
5079 (This is now used as a subroutine by C<sv_magic>.)
5084 Perl_sv_magicext(pTHX_ SV *const sv, SV *const obj, const int how,
5085 const MGVTBL *const vtable, const char *const name, const I32 namlen)
5090 PERL_ARGS_ASSERT_SV_MAGICEXT;
5092 SvUPGRADE(sv, SVt_PVMG);
5093 Newxz(mg, 1, MAGIC);
5094 mg->mg_moremagic = SvMAGIC(sv);
5095 SvMAGIC_set(sv, mg);
5097 /* Sometimes a magic contains a reference loop, where the sv and
5098 object refer to each other. To prevent a reference loop that
5099 would prevent such objects being freed, we look for such loops
5100 and if we find one we avoid incrementing the object refcount.
5102 Note we cannot do this to avoid self-tie loops as intervening RV must
5103 have its REFCNT incremented to keep it in existence.
5106 if (!obj || obj == sv ||
5107 how == PERL_MAGIC_arylen ||
5108 how == PERL_MAGIC_symtab ||
5109 (SvTYPE(obj) == SVt_PVGV &&
5110 (GvSV(obj) == sv || GvHV(obj) == (const HV *)sv
5111 || GvAV(obj) == (const AV *)sv || GvCV(obj) == (const CV *)sv
5112 || GvIOp(obj) == (const IO *)sv || GvFORM(obj) == (const CV *)sv)))
5117 mg->mg_obj = SvREFCNT_inc_simple(obj);
5118 mg->mg_flags |= MGf_REFCOUNTED;
5121 /* Normal self-ties simply pass a null object, and instead of
5122 using mg_obj directly, use the SvTIED_obj macro to produce a
5123 new RV as needed. For glob "self-ties", we are tieing the PVIO
5124 with an RV obj pointing to the glob containing the PVIO. In
5125 this case, to avoid a reference loop, we need to weaken the
5129 if (how == PERL_MAGIC_tiedscalar && SvTYPE(sv) == SVt_PVIO &&
5130 obj && SvROK(obj) && GvIO(SvRV(obj)) == (const IO *)sv)
5136 mg->mg_len = namlen;
5139 mg->mg_ptr = savepvn(name, namlen);
5140 else if (namlen == HEf_SVKEY) {
5141 /* Yes, this is casting away const. This is only for the case of
5142 HEf_SVKEY. I think we need to document this aberation of the
5143 constness of the API, rather than making name non-const, as
5144 that change propagating outwards a long way. */
5145 mg->mg_ptr = (char*)SvREFCNT_inc_simple_NN((SV *)name);
5147 mg->mg_ptr = (char *) name;
5149 mg->mg_virtual = (MGVTBL *) vtable;
5153 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
5158 =for apidoc sv_magic
5160 Adds magic to an SV. First upgrades C<sv> to type C<SVt_PVMG> if necessary,
5161 then adds a new magic item of type C<how> to the head of the magic list.
5163 See C<sv_magicext> (which C<sv_magic> now calls) for a description of the
5164 handling of the C<name> and C<namlen> arguments.
5166 You need to use C<sv_magicext> to add magic to SvREADONLY SVs and also
5167 to add more than one instance of the same 'how'.
5173 Perl_sv_magic(pTHX_ register SV *const sv, SV *const obj, const int how,
5174 const char *const name, const I32 namlen)
5177 const MGVTBL *vtable;
5180 PERL_ARGS_ASSERT_SV_MAGIC;
5182 #ifdef PERL_OLD_COPY_ON_WRITE
5184 sv_force_normal_flags(sv, 0);
5186 if (SvREADONLY(sv)) {
5188 /* its okay to attach magic to shared strings; the subsequent
5189 * upgrade to PVMG will unshare the string */
5190 !(SvFAKE(sv) && SvTYPE(sv) < SVt_PVMG)
5193 && how != PERL_MAGIC_regex_global
5194 && how != PERL_MAGIC_bm
5195 && how != PERL_MAGIC_fm
5196 && how != PERL_MAGIC_sv
5197 && how != PERL_MAGIC_backref
5200 Perl_croak_no_modify(aTHX);
5203 if (SvMAGICAL(sv) || (how == PERL_MAGIC_taint && SvTYPE(sv) >= SVt_PVMG)) {
5204 if (SvMAGIC(sv) && (mg = mg_find(sv, how))) {
5205 /* sv_magic() refuses to add a magic of the same 'how' as an
5208 if (how == PERL_MAGIC_taint) {
5210 /* Any scalar which already had taint magic on which someone
5211 (erroneously?) did SvIOK_on() or similar will now be
5212 incorrectly sporting public "OK" flags. */
5213 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
5221 vtable = &PL_vtbl_sv;
5223 case PERL_MAGIC_overload:
5224 vtable = &PL_vtbl_amagic;
5226 case PERL_MAGIC_overload_elem:
5227 vtable = &PL_vtbl_amagicelem;
5229 case PERL_MAGIC_overload_table:
5230 vtable = &PL_vtbl_ovrld;
5233 vtable = &PL_vtbl_bm;
5235 case PERL_MAGIC_regdata:
5236 vtable = &PL_vtbl_regdata;
5238 case PERL_MAGIC_regdatum:
5239 vtable = &PL_vtbl_regdatum;
5241 case PERL_MAGIC_env:
5242 vtable = &PL_vtbl_env;
5245 vtable = &PL_vtbl_fm;
5247 case PERL_MAGIC_envelem:
5248 vtable = &PL_vtbl_envelem;
5250 case PERL_MAGIC_regex_global:
5251 vtable = &PL_vtbl_mglob;
5253 case PERL_MAGIC_isa:
5254 vtable = &PL_vtbl_isa;
5256 case PERL_MAGIC_isaelem:
5257 vtable = &PL_vtbl_isaelem;
5259 case PERL_MAGIC_nkeys:
5260 vtable = &PL_vtbl_nkeys;
5262 case PERL_MAGIC_dbfile:
5265 case PERL_MAGIC_dbline:
5266 vtable = &PL_vtbl_dbline;
5268 #ifdef USE_LOCALE_COLLATE
5269 case PERL_MAGIC_collxfrm:
5270 vtable = &PL_vtbl_collxfrm;
5272 #endif /* USE_LOCALE_COLLATE */
5273 case PERL_MAGIC_tied:
5274 vtable = &PL_vtbl_pack;
5276 case PERL_MAGIC_tiedelem:
5277 case PERL_MAGIC_tiedscalar:
5278 vtable = &PL_vtbl_packelem;
5281 vtable = &PL_vtbl_regexp;
5283 case PERL_MAGIC_sig:
5284 vtable = &PL_vtbl_sig;
5286 case PERL_MAGIC_sigelem:
5287 vtable = &PL_vtbl_sigelem;
5289 case PERL_MAGIC_taint:
5290 vtable = &PL_vtbl_taint;
5292 case PERL_MAGIC_uvar:
5293 vtable = &PL_vtbl_uvar;
5295 case PERL_MAGIC_vec:
5296 vtable = &PL_vtbl_vec;
5298 case PERL_MAGIC_arylen_p:
5299 case PERL_MAGIC_rhash:
5300 case PERL_MAGIC_symtab:
5301 case PERL_MAGIC_vstring:
5302 case PERL_MAGIC_checkcall:
5305 case PERL_MAGIC_utf8:
5306 vtable = &PL_vtbl_utf8;
5308 case PERL_MAGIC_substr:
5309 vtable = &PL_vtbl_substr;
5311 case PERL_MAGIC_defelem:
5312 vtable = &PL_vtbl_defelem;
5314 case PERL_MAGIC_arylen:
5315 vtable = &PL_vtbl_arylen;
5317 case PERL_MAGIC_pos:
5318 vtable = &PL_vtbl_pos;
5320 case PERL_MAGIC_backref:
5321 vtable = &PL_vtbl_backref;
5323 case PERL_MAGIC_hintselem:
5324 vtable = &PL_vtbl_hintselem;
5326 case PERL_MAGIC_hints:
5327 vtable = &PL_vtbl_hints;
5329 case PERL_MAGIC_ext:
5330 /* Reserved for use by extensions not perl internals. */
5331 /* Useful for attaching extension internal data to perl vars. */
5332 /* Note that multiple extensions may clash if magical scalars */
5333 /* etc holding private data from one are passed to another. */
5337 Perl_croak(aTHX_ "Don't know how to handle magic of type \\%o", how);
5340 /* Rest of work is done else where */
5341 mg = sv_magicext(sv,obj,how,vtable,name,namlen);
5344 case PERL_MAGIC_taint:
5347 case PERL_MAGIC_ext:
5348 case PERL_MAGIC_dbfile:
5355 S_sv_unmagicext_flags(pTHX_ SV *const sv, const int type, MGVTBL *vtbl, const U32 flags)
5362 if (SvTYPE(sv) < SVt_PVMG || !SvMAGIC(sv))
5364 mgp = &(((XPVMG*) SvANY(sv))->xmg_u.xmg_magic);
5365 for (mg = *mgp; mg; mg = *mgp) {
5366 const MGVTBL* const virt = mg->mg_virtual;
5367 if (mg->mg_type == type && (!flags || virt == vtbl)) {
5368 *mgp = mg->mg_moremagic;
5369 if (virt && virt->svt_free)
5370 virt->svt_free(aTHX_ sv, mg);
5371 if (mg->mg_ptr && mg->mg_type != PERL_MAGIC_regex_global) {
5373 Safefree(mg->mg_ptr);
5374 else if (mg->mg_len == HEf_SVKEY)
5375 SvREFCNT_dec(MUTABLE_SV(mg->mg_ptr));
5376 else if (mg->mg_type == PERL_MAGIC_utf8)
5377 Safefree(mg->mg_ptr);
5379 if (mg->mg_flags & MGf_REFCOUNTED)
5380 SvREFCNT_dec(mg->mg_obj);
5384 mgp = &mg->mg_moremagic;
5387 if (SvMAGICAL(sv)) /* if we're under save_magic, wait for restore_magic; */
5388 mg_magical(sv); /* else fix the flags now */
5392 SvFLAGS(sv) |= (SvFLAGS(sv) & (SVp_IOK|SVp_NOK|SVp_POK)) >> PRIVSHIFT;
5398 =for apidoc sv_unmagic
5400 Removes all magic of type C<type> from an SV.
5406 Perl_sv_unmagic(pTHX_ SV *const sv, const int type)
5408 PERL_ARGS_ASSERT_SV_UNMAGIC;
5409 return S_sv_unmagicext_flags(aTHX_ sv, type, NULL, 0);
5413 =for apidoc sv_unmagicext
5415 Removes all magic of type C<type> with the specified C<vtbl> from an SV.
5421 Perl_sv_unmagicext(pTHX_ SV *const sv, const int type, MGVTBL *vtbl)
5423 PERL_ARGS_ASSERT_SV_UNMAGICEXT;
5424 return S_sv_unmagicext_flags(aTHX_ sv, type, vtbl, 1);
5428 =for apidoc sv_rvweaken
5430 Weaken a reference: set the C<SvWEAKREF> flag on this RV; give the
5431 referred-to SV C<PERL_MAGIC_backref> magic if it hasn't already; and
5432 push a back-reference to this RV onto the array of backreferences
5433 associated with that magic. If the RV is magical, set magic will be
5434 called after the RV is cleared.
5440 Perl_sv_rvweaken(pTHX_ SV *const sv)
5444 PERL_ARGS_ASSERT_SV_RVWEAKEN;
5446 if (!SvOK(sv)) /* let undefs pass */
5449 Perl_croak(aTHX_ "Can't weaken a nonreference");
5450 else if (SvWEAKREF(sv)) {
5451 Perl_ck_warner(aTHX_ packWARN(WARN_MISC), "Reference is already weak");
5455 Perl_sv_add_backref(aTHX_ tsv, sv);
5461 /* Give tsv backref magic if it hasn't already got it, then push a
5462 * back-reference to sv onto the array associated with the backref magic.
5464 * As an optimisation, if there's only one backref and it's not an AV,
5465 * store it directly in the HvAUX or mg_obj slot, avoiding the need to
5466 * allocate an AV. (Whether the slot holds an AV tells us whether this is
5469 * If an HV's backref is stored in magic, it is moved back to HvAUX.
5472 /* A discussion about the backreferences array and its refcount:
5474 * The AV holding the backreferences is pointed to either as the mg_obj of
5475 * PERL_MAGIC_backref, or in the specific case of a HV that has the hv_aux
5476 * structure, from the xhv_backreferences field. (A HV without hv_aux will
5477 * have the standard magic instead.) The array is created with a refcount
5478 * of 2. This means that if during global destruction the array gets
5479 * picked on before its parent to have its refcount decremented by the
5480 * random zapper, it won't actually be freed, meaning it's still there for
5481 * when its parent gets freed.
5483 * When the parent SV is freed, the extra ref is killed by
5484 * Perl_sv_kill_backrefs. The other ref is killed, in the case of magic,
5485 * by mg_free() / MGf_REFCOUNTED, or for a hash, by Perl_hv_kill_backrefs.
5487 * When a single backref SV is stored directly, it is not reference
5492 Perl_sv_add_backref(pTHX_ SV *const tsv, SV *const sv)
5499 PERL_ARGS_ASSERT_SV_ADD_BACKREF;
5501 /* find slot to store array or singleton backref */
5503 if (SvTYPE(tsv) == SVt_PVHV) {
5504 svp = (SV**)Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5507 if ((mg = mg_find(tsv, PERL_MAGIC_backref))) {
5508 /* Aha. They've got it stowed in magic instead.
5509 * Move it back to xhv_backreferences */
5511 /* Stop mg_free decreasing the reference count. */
5513 /* Stop mg_free even calling the destructor, given that
5514 there's no AV to free up. */
5516 sv_unmagic(tsv, PERL_MAGIC_backref);
5522 (SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL))))
5524 sv_magic(tsv, NULL, PERL_MAGIC_backref, NULL, 0);
5525 mg = mg_find(tsv, PERL_MAGIC_backref);
5527 svp = &(mg->mg_obj);
5530 /* create or retrieve the array */
5532 if ( (!*svp && SvTYPE(sv) == SVt_PVAV)
5533 || (*svp && SvTYPE(*svp) != SVt_PVAV)
5538 SvREFCNT_inc_simple_void(av);
5539 /* av now has a refcnt of 2; see discussion above */
5541 /* move single existing backref to the array */
5543 AvARRAY(av)[++AvFILLp(av)] = *svp; /* av_push() */
5547 mg->mg_flags |= MGf_REFCOUNTED;
5550 av = MUTABLE_AV(*svp);
5553 /* optimisation: store single backref directly in HvAUX or mg_obj */
5557 /* push new backref */
5558 assert(SvTYPE(av) == SVt_PVAV);
5559 if (AvFILLp(av) >= AvMAX(av)) {
5560 av_extend(av, AvFILLp(av)+1);
5562 AvARRAY(av)[++AvFILLp(av)] = sv; /* av_push() */
5565 /* delete a back-reference to ourselves from the backref magic associated
5566 * with the SV we point to.
5570 Perl_sv_del_backref(pTHX_ SV *const tsv, SV *const sv)
5575 PERL_ARGS_ASSERT_SV_DEL_BACKREF;
5577 if (SvTYPE(tsv) == SVt_PVHV && SvOOK(tsv)) {
5578 svp = (SV**)Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5580 if (!svp || !*svp) {
5582 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
5583 svp = mg ? &(mg->mg_obj) : NULL;
5587 Perl_croak(aTHX_ "panic: del_backref");
5589 if (SvTYPE(*svp) == SVt_PVAV) {
5593 AV * const av = (AV*)*svp;
5595 assert(!SvIS_FREED(av));
5599 /* for an SV with N weak references to it, if all those
5600 * weak refs are deleted, then sv_del_backref will be called
5601 * N times and O(N^2) compares will be done within the backref
5602 * array. To ameliorate this potential slowness, we:
5603 * 1) make sure this code is as tight as possible;
5604 * 2) when looking for SV, look for it at both the head and tail of the
5605 * array first before searching the rest, since some create/destroy
5606 * patterns will cause the backrefs to be freed in order.
5613 SV **p = &svp[fill];
5614 SV *const topsv = *p;
5621 /* We weren't the last entry.
5622 An unordered list has this property that you
5623 can take the last element off the end to fill
5624 the hole, and it's still an unordered list :-)
5630 break; /* should only be one */
5637 AvFILLp(av) = fill-1;
5640 /* optimisation: only a single backref, stored directly */
5642 Perl_croak(aTHX_ "panic: del_backref");
5649 Perl_sv_kill_backrefs(pTHX_ SV *const sv, AV *const av)
5655 PERL_ARGS_ASSERT_SV_KILL_BACKREFS;
5660 is_array = (SvTYPE(av) == SVt_PVAV);
5662 assert(!SvIS_FREED(av));
5665 last = svp + AvFILLp(av);
5668 /* optimisation: only a single backref, stored directly */
5674 while (svp <= last) {
5676 SV *const referrer = *svp;
5677 if (SvWEAKREF(referrer)) {
5678 /* XXX Should we check that it hasn't changed? */
5679 assert(SvROK(referrer));
5680 SvRV_set(referrer, 0);
5682 SvWEAKREF_off(referrer);
5683 SvSETMAGIC(referrer);
5684 } else if (SvTYPE(referrer) == SVt_PVGV ||
5685 SvTYPE(referrer) == SVt_PVLV) {
5686 assert(SvTYPE(sv) == SVt_PVHV); /* stash backref */
5687 /* You lookin' at me? */
5688 assert(GvSTASH(referrer));
5689 assert(GvSTASH(referrer) == (const HV *)sv);
5690 GvSTASH(referrer) = 0;
5691 } else if (SvTYPE(referrer) == SVt_PVCV ||
5692 SvTYPE(referrer) == SVt_PVFM) {
5693 if (SvTYPE(sv) == SVt_PVHV) { /* stash backref */
5694 /* You lookin' at me? */
5695 assert(CvSTASH(referrer));
5696 assert(CvSTASH(referrer) == (const HV *)sv);
5697 SvANY(MUTABLE_CV(referrer))->xcv_stash = 0;
5700 assert(SvTYPE(sv) == SVt_PVGV);
5701 /* You lookin' at me? */
5702 assert(CvGV(referrer));
5703 assert(CvGV(referrer) == (const GV *)sv);
5704 anonymise_cv_maybe(MUTABLE_GV(sv),
5705 MUTABLE_CV(referrer));
5710 "panic: magic_killbackrefs (flags=%"UVxf")",
5711 (UV)SvFLAGS(referrer));
5722 SvREFCNT_dec(av); /* remove extra count added by sv_add_backref() */
5728 =for apidoc sv_insert
5730 Inserts a string at the specified offset/length within the SV. Similar to
5731 the Perl substr() function. Handles get magic.
5733 =for apidoc sv_insert_flags
5735 Same as C<sv_insert>, but the extra C<flags> are passed the C<SvPV_force_flags> that applies to C<bigstr>.
5741 Perl_sv_insert_flags(pTHX_ SV *const bigstr, const STRLEN offset, const STRLEN len, const char *const little, const STRLEN littlelen, const U32 flags)
5746 register char *midend;
5747 register char *bigend;
5751 PERL_ARGS_ASSERT_SV_INSERT_FLAGS;
5754 Perl_croak(aTHX_ "Can't modify non-existent substring");
5755 SvPV_force_flags(bigstr, curlen, flags);
5756 (void)SvPOK_only_UTF8(bigstr);
5757 if (offset + len > curlen) {
5758 SvGROW(bigstr, offset+len+1);
5759 Zero(SvPVX(bigstr)+curlen, offset+len-curlen, char);
5760 SvCUR_set(bigstr, offset+len);
5764 i = littlelen - len;
5765 if (i > 0) { /* string might grow */
5766 big = SvGROW(bigstr, SvCUR(bigstr) + i + 1);
5767 mid = big + offset + len;
5768 midend = bigend = big + SvCUR(bigstr);
5771 while (midend > mid) /* shove everything down */
5772 *--bigend = *--midend;
5773 Move(little,big+offset,littlelen,char);
5774 SvCUR_set(bigstr, SvCUR(bigstr) + i);
5779 Move(little,SvPVX(bigstr)+offset,len,char);
5784 big = SvPVX(bigstr);
5787 bigend = big + SvCUR(bigstr);
5789 if (midend > bigend)
5790 Perl_croak(aTHX_ "panic: sv_insert");
5792 if (mid - big > bigend - midend) { /* faster to shorten from end */
5794 Move(little, mid, littlelen,char);
5797 i = bigend - midend;
5799 Move(midend, mid, i,char);
5803 SvCUR_set(bigstr, mid - big);
5805 else if ((i = mid - big)) { /* faster from front */
5806 midend -= littlelen;
5808 Move(big, midend - i, i, char);
5809 sv_chop(bigstr,midend-i);
5811 Move(little, mid, littlelen,char);
5813 else if (littlelen) {
5814 midend -= littlelen;
5815 sv_chop(bigstr,midend);
5816 Move(little,midend,littlelen,char);
5819 sv_chop(bigstr,midend);
5825 =for apidoc sv_replace
5827 Make the first argument a copy of the second, then delete the original.
5828 The target SV physically takes over ownership of the body of the source SV
5829 and inherits its flags; however, the target keeps any magic it owns,
5830 and any magic in the source is discarded.
5831 Note that this is a rather specialist SV copying operation; most of the
5832 time you'll want to use C<sv_setsv> or one of its many macro front-ends.
5838 Perl_sv_replace(pTHX_ register SV *const sv, register SV *const nsv)
5841 const U32 refcnt = SvREFCNT(sv);
5843 PERL_ARGS_ASSERT_SV_REPLACE;
5845 SV_CHECK_THINKFIRST_COW_DROP(sv);
5846 if (SvREFCNT(nsv) != 1) {
5847 Perl_croak(aTHX_ "panic: reference miscount on nsv in sv_replace()"
5848 " (%" UVuf " != 1)", (UV) SvREFCNT(nsv));
5850 if (SvMAGICAL(sv)) {
5854 sv_upgrade(nsv, SVt_PVMG);
5855 SvMAGIC_set(nsv, SvMAGIC(sv));
5856 SvFLAGS(nsv) |= SvMAGICAL(sv);
5858 SvMAGIC_set(sv, NULL);
5862 assert(!SvREFCNT(sv));
5863 #ifdef DEBUG_LEAKING_SCALARS
5864 sv->sv_flags = nsv->sv_flags;
5865 sv->sv_any = nsv->sv_any;
5866 sv->sv_refcnt = nsv->sv_refcnt;
5867 sv->sv_u = nsv->sv_u;
5869 StructCopy(nsv,sv,SV);
5871 if(SvTYPE(sv) == SVt_IV) {
5873 = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
5877 #ifdef PERL_OLD_COPY_ON_WRITE
5878 if (SvIsCOW_normal(nsv)) {
5879 /* We need to follow the pointers around the loop to make the
5880 previous SV point to sv, rather than nsv. */
5883 while ((next = SV_COW_NEXT_SV(current)) != nsv) {
5886 assert(SvPVX_const(current) == SvPVX_const(nsv));
5888 /* Make the SV before us point to the SV after us. */
5890 PerlIO_printf(Perl_debug_log, "previous is\n");
5892 PerlIO_printf(Perl_debug_log,
5893 "move it from 0x%"UVxf" to 0x%"UVxf"\n",
5894 (UV) SV_COW_NEXT_SV(current), (UV) sv);
5896 SV_COW_NEXT_SV_SET(current, sv);
5899 SvREFCNT(sv) = refcnt;
5900 SvFLAGS(nsv) |= SVTYPEMASK; /* Mark as freed */
5905 /* We're about to free a GV which has a CV that refers back to us.
5906 * If that CV will outlive us, make it anonymous (i.e. fix up its CvGV
5910 S_anonymise_cv_maybe(pTHX_ GV *gv, CV* cv)
5916 PERL_ARGS_ASSERT_ANONYMISE_CV_MAYBE;
5919 assert(SvREFCNT(gv) == 0);
5920 assert(isGV(gv) && isGV_with_GP(gv));
5922 assert(!CvANON(cv));
5923 assert(CvGV(cv) == gv);
5925 /* will the CV shortly be freed by gp_free() ? */
5926 if (GvCV(gv) == cv && GvGP(gv)->gp_refcnt < 2 && SvREFCNT(cv) < 2) {
5927 SvANY(cv)->xcv_gv = NULL;
5931 /* if not, anonymise: */
5932 stash = GvSTASH(gv) ? HvNAME(GvSTASH(gv)) : NULL;
5933 gvname = Perl_newSVpvf(aTHX_ "%s::__ANON__",
5934 stash ? stash : "__ANON__");
5935 anongv = gv_fetchsv(gvname, GV_ADDMULTI, SVt_PVCV);
5936 SvREFCNT_dec(gvname);
5940 SvANY(cv)->xcv_gv = MUTABLE_GV(SvREFCNT_inc(anongv));
5945 =for apidoc sv_clear
5947 Clear an SV: call any destructors, free up any memory used by the body,
5948 and free the body itself. The SV's head is I<not> freed, although
5949 its type is set to all 1's so that it won't inadvertently be assumed
5950 to be live during global destruction etc.
5951 This function should only be called when REFCNT is zero. Most of the time
5952 you'll want to call C<sv_free()> (or its macro wrapper C<SvREFCNT_dec>)
5959 Perl_sv_clear(pTHX_ SV *const orig_sv)
5964 const struct body_details *sv_type_details;
5967 register SV *sv = orig_sv;
5969 PERL_ARGS_ASSERT_SV_CLEAR;
5971 /* within this loop, sv is the SV currently being freed, and
5972 * iter_sv is the most recent AV or whatever that's being iterated
5973 * over to provide more SVs */
5979 assert(SvREFCNT(sv) == 0);
5980 assert(SvTYPE(sv) != SVTYPEMASK);
5982 if (type <= SVt_IV) {
5983 /* See the comment in sv.h about the collusion between this
5984 * early return and the overloading of the NULL slots in the
5988 SvFLAGS(sv) &= SVf_BREAK;
5989 SvFLAGS(sv) |= SVTYPEMASK;
5994 if (!curse(sv, 1)) goto get_next_sv;
5996 if (type >= SVt_PVMG) {
5997 if (type == SVt_PVMG && SvPAD_OUR(sv)) {
5998 SvREFCNT_dec(SvOURSTASH(sv));
5999 } else if (SvMAGIC(sv))
6001 if (type == SVt_PVMG && SvPAD_TYPED(sv))
6002 SvREFCNT_dec(SvSTASH(sv));
6005 /* case SVt_BIND: */
6008 IoIFP(sv) != PerlIO_stdin() &&
6009 IoIFP(sv) != PerlIO_stdout() &&
6010 IoIFP(sv) != PerlIO_stderr() &&
6011 !(IoFLAGS(sv) & IOf_FAKE_DIRP))
6013 io_close(MUTABLE_IO(sv), FALSE);
6015 if (IoDIRP(sv) && !(IoFLAGS(sv) & IOf_FAKE_DIRP))
6016 PerlDir_close(IoDIRP(sv));
6017 IoDIRP(sv) = (DIR*)NULL;
6018 Safefree(IoTOP_NAME(sv));
6019 Safefree(IoFMT_NAME(sv));
6020 Safefree(IoBOTTOM_NAME(sv));
6023 /* FIXME for plugins */
6024 pregfree2((REGEXP*) sv);
6028 cv_undef(MUTABLE_CV(sv));
6029 /* If we're in a stash, we don't own a reference to it.
6030 * However it does have a back reference to us, which needs to
6032 if ((stash = CvSTASH(sv)))
6033 sv_del_backref(MUTABLE_SV(stash), sv);
6036 if (PL_last_swash_hv == (const HV *)sv) {
6037 PL_last_swash_hv = NULL;
6039 Perl_hv_kill_backrefs(aTHX_ MUTABLE_HV(sv));
6040 Perl_hv_undef_flags(aTHX_ MUTABLE_HV(sv), HV_NAME_SETALL);
6044 AV* av = MUTABLE_AV(sv);
6045 if (PL_comppad == av) {
6049 if (AvREAL(av) && AvFILLp(av) > -1) {
6050 next_sv = AvARRAY(av)[AvFILLp(av)--];
6051 /* save old iter_sv in top-most slot of AV,
6052 * and pray that it doesn't get wiped in the meantime */
6053 AvARRAY(av)[AvMAX(av)] = iter_sv;
6055 goto get_next_sv; /* process this new sv */
6057 Safefree(AvALLOC(av));
6062 if (LvTYPE(sv) == 'T') { /* for tie: return HE to pool */
6063 SvREFCNT_dec(HeKEY_sv((HE*)LvTARG(sv)));
6064 HeNEXT((HE*)LvTARG(sv)) = PL_hv_fetch_ent_mh;
6065 PL_hv_fetch_ent_mh = (HE*)LvTARG(sv);
6067 else if (LvTYPE(sv) != 't') /* unless tie: unrefcnted fake SV** */
6068 SvREFCNT_dec(LvTARG(sv));
6070 if (isGV_with_GP(sv)) {
6071 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
6072 && HvENAME_get(stash))
6073 mro_method_changed_in(stash);
6074 gp_free(MUTABLE_GV(sv));
6076 unshare_hek(GvNAME_HEK(sv));
6077 /* If we're in a stash, we don't own a reference to it.
6078 * However it does have a back reference to us, which
6079 * needs to be cleared. */
6080 if (!SvVALID(sv) && (stash = GvSTASH(sv)))
6081 sv_del_backref(MUTABLE_SV(stash), sv);
6083 /* FIXME. There are probably more unreferenced pointers to SVs
6084 * in the interpreter struct that we should check and tidy in
6085 * a similar fashion to this: */
6086 if ((const GV *)sv == PL_last_in_gv)
6087 PL_last_in_gv = NULL;
6093 /* Don't bother with SvOOK_off(sv); as we're only going to
6097 SvOOK_offset(sv, offset);
6098 SvPV_set(sv, SvPVX_mutable(sv) - offset);
6099 /* Don't even bother with turning off the OOK flag. */
6104 SV * const target = SvRV(sv);
6106 sv_del_backref(target, sv);
6111 #ifdef PERL_OLD_COPY_ON_WRITE
6112 else if (SvPVX_const(sv)
6113 && !(SvTYPE(sv) == SVt_PVIO
6114 && !(IoFLAGS(sv) & IOf_FAKE_DIRP)))
6118 PerlIO_printf(Perl_debug_log, "Copy on write: clear\n");
6122 sv_release_COW(sv, SvPVX_const(sv), SV_COW_NEXT_SV(sv));
6124 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
6128 } else if (SvLEN(sv)) {
6129 Safefree(SvPVX_const(sv));
6133 else if (SvPVX_const(sv) && SvLEN(sv)
6134 && !(SvTYPE(sv) == SVt_PVIO
6135 && !(IoFLAGS(sv) & IOf_FAKE_DIRP)))
6136 Safefree(SvPVX_mutable(sv));
6137 else if (SvPVX_const(sv) && SvREADONLY(sv) && SvFAKE(sv)) {
6138 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
6149 SvFLAGS(sv) &= SVf_BREAK;
6150 SvFLAGS(sv) |= SVTYPEMASK;
6152 sv_type_details = bodies_by_type + type;
6153 if (sv_type_details->arena) {
6154 del_body(((char *)SvANY(sv) + sv_type_details->offset),
6155 &PL_body_roots[type]);
6157 else if (sv_type_details->body_size) {
6158 safefree(SvANY(sv));
6162 /* caller is responsible for freeing the head of the original sv */
6163 if (sv != orig_sv && !SvREFCNT(sv))
6166 /* grab and free next sv, if any */
6174 else if (!iter_sv) {
6176 } else if (SvTYPE(iter_sv) == SVt_PVAV) {
6177 AV *const av = (AV*)iter_sv;
6178 if (AvFILLp(av) > -1) {
6179 sv = AvARRAY(av)[AvFILLp(av)--];
6181 else { /* no more elements of current AV to free */
6184 /* restore previous value, squirrelled away */
6185 iter_sv = AvARRAY(av)[AvMAX(av)];
6186 Safefree(AvALLOC(av));
6191 /* unrolled SvREFCNT_dec and sv_free2 follows: */
6195 if (!SvREFCNT(sv)) {
6199 if (--(SvREFCNT(sv)))
6203 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEBUGGING),
6204 "Attempt to free temp prematurely: SV 0x%"UVxf
6205 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
6209 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
6210 /* make sure SvREFCNT(sv)==0 happens very seldom */
6211 SvREFCNT(sv) = (~(U32)0)/2;
6220 /* This routine curses the sv itself, not the object referenced by sv. So
6221 sv does not have to be ROK. */
6224 S_curse(pTHX_ SV * const sv, const bool check_refcnt) {
6227 PERL_ARGS_ASSERT_CURSE;
6228 assert(SvOBJECT(sv));
6230 if (PL_defstash && /* Still have a symbol table? */
6237 stash = SvSTASH(sv);
6238 destructor = StashHANDLER(stash,DESTROY);
6240 /* A constant subroutine can have no side effects, so
6241 don't bother calling it. */
6242 && !CvCONST(destructor)
6243 /* Don't bother calling an empty destructor */
6244 && (CvISXSUB(destructor)
6245 || (CvSTART(destructor)
6246 && (CvSTART(destructor)->op_next->op_type
6249 SV* const tmpref = newRV(sv);
6250 SvREADONLY_on(tmpref); /* DESTROY() could be naughty */
6252 PUSHSTACKi(PERLSI_DESTROY);
6257 call_sv(MUTABLE_SV(destructor),
6258 G_DISCARD|G_EVAL|G_KEEPERR|G_VOID);
6262 if(SvREFCNT(tmpref) < 2) {
6263 /* tmpref is not kept alive! */
6265 SvRV_set(tmpref, NULL);
6268 SvREFCNT_dec(tmpref);
6270 } while (SvOBJECT(sv) && SvSTASH(sv) != stash);
6273 if (check_refcnt && SvREFCNT(sv)) {
6274 if (PL_in_clean_objs)
6276 "DESTROY created new reference to dead object '%s'",
6278 /* DESTROY gave object new lease on life */
6284 SvREFCNT_dec(SvSTASH(sv)); /* possibly of changed persuasion */
6285 SvOBJECT_off(sv); /* Curse the object. */
6286 if (SvTYPE(sv) != SVt_PVIO)
6287 --PL_sv_objcount;/* XXX Might want something more general */
6293 =for apidoc sv_newref
6295 Increment an SV's reference count. Use the C<SvREFCNT_inc()> wrapper
6302 Perl_sv_newref(pTHX_ SV *const sv)
6304 PERL_UNUSED_CONTEXT;
6313 Decrement an SV's reference count, and if it drops to zero, call
6314 C<sv_clear> to invoke destructors and free up any memory used by
6315 the body; finally, deallocate the SV's head itself.
6316 Normally called via a wrapper macro C<SvREFCNT_dec>.
6322 Perl_sv_free(pTHX_ SV *const sv)
6327 if (SvREFCNT(sv) == 0) {
6328 if (SvFLAGS(sv) & SVf_BREAK)
6329 /* this SV's refcnt has been artificially decremented to
6330 * trigger cleanup */
6332 if (PL_in_clean_all) /* All is fair */
6334 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
6335 /* make sure SvREFCNT(sv)==0 happens very seldom */
6336 SvREFCNT(sv) = (~(U32)0)/2;
6339 if (ckWARN_d(WARN_INTERNAL)) {
6340 #ifdef DEBUG_LEAKING_SCALARS_FORK_DUMP
6341 Perl_dump_sv_child(aTHX_ sv);
6343 #ifdef DEBUG_LEAKING_SCALARS
6346 #ifdef DEBUG_LEAKING_SCALARS_ABORT
6347 if (PL_warnhook == PERL_WARNHOOK_FATAL
6348 || ckDEAD(packWARN(WARN_INTERNAL))) {
6349 /* Don't let Perl_warner cause us to escape our fate: */
6353 /* This may not return: */
6354 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
6355 "Attempt to free unreferenced scalar: SV 0x%"UVxf
6356 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
6359 #ifdef DEBUG_LEAKING_SCALARS_ABORT
6364 if (--(SvREFCNT(sv)) > 0)
6366 Perl_sv_free2(aTHX_ sv);
6370 Perl_sv_free2(pTHX_ SV *const sv)
6374 PERL_ARGS_ASSERT_SV_FREE2;
6378 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEBUGGING),
6379 "Attempt to free temp prematurely: SV 0x%"UVxf
6380 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
6384 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
6385 /* make sure SvREFCNT(sv)==0 happens very seldom */
6386 SvREFCNT(sv) = (~(U32)0)/2;
6397 Returns the length of the string in the SV. Handles magic and type
6398 coercion. See also C<SvCUR>, which gives raw access to the xpv_cur slot.
6404 Perl_sv_len(pTHX_ register SV *const sv)
6412 len = mg_length(sv);
6414 (void)SvPV_const(sv, len);
6419 =for apidoc sv_len_utf8
6421 Returns the number of characters in the string in an SV, counting wide
6422 UTF-8 bytes as a single character. Handles magic and type coercion.
6428 * The length is cached in PERL_MAGIC_utf8, in the mg_len field. Also the
6429 * mg_ptr is used, by sv_pos_u2b() and sv_pos_b2u() - see the comments below.
6430 * (Note that the mg_len is not the length of the mg_ptr field.
6431 * This allows the cache to store the character length of the string without
6432 * needing to malloc() extra storage to attach to the mg_ptr.)
6437 Perl_sv_len_utf8(pTHX_ register SV *const sv)
6443 return mg_length(sv);
6447 const U8 *s = (U8*)SvPV_const(sv, len);
6451 MAGIC *mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_utf8) : NULL;
6453 if (mg && (mg->mg_len != -1 || mg->mg_ptr)) {
6454 if (mg->mg_len != -1)
6457 /* We can use the offset cache for a headstart.
6458 The longer value is stored in the first pair. */
6459 STRLEN *cache = (STRLEN *) mg->mg_ptr;
6461 ulen = cache[0] + Perl_utf8_length(aTHX_ s + cache[1],
6465 if (PL_utf8cache < 0) {
6466 const STRLEN real = Perl_utf8_length(aTHX_ s, s + len);
6467 assert_uft8_cache_coherent("sv_len_utf8", ulen, real, sv);
6471 ulen = Perl_utf8_length(aTHX_ s, s + len);
6472 utf8_mg_len_cache_update(sv, &mg, ulen);
6476 return Perl_utf8_length(aTHX_ s, s + len);
6480 /* Walk forwards to find the byte corresponding to the passed in UTF-8
6483 S_sv_pos_u2b_forwards(const U8 *const start, const U8 *const send,
6484 STRLEN *const uoffset_p, bool *const at_end)
6486 const U8 *s = start;
6487 STRLEN uoffset = *uoffset_p;
6489 PERL_ARGS_ASSERT_SV_POS_U2B_FORWARDS;
6491 while (s < send && uoffset) {
6498 else if (s > send) {
6500 /* This is the existing behaviour. Possibly it should be a croak, as
6501 it's actually a bounds error */
6504 *uoffset_p -= uoffset;
6508 /* Given the length of the string in both bytes and UTF-8 characters, decide
6509 whether to walk forwards or backwards to find the byte corresponding to
6510 the passed in UTF-8 offset. */
6512 S_sv_pos_u2b_midway(const U8 *const start, const U8 *send,
6513 STRLEN uoffset, const STRLEN uend)
6515 STRLEN backw = uend - uoffset;
6517 PERL_ARGS_ASSERT_SV_POS_U2B_MIDWAY;
6519 if (uoffset < 2 * backw) {
6520 /* The assumption is that going forwards is twice the speed of going
6521 forward (that's where the 2 * backw comes from).
6522 (The real figure of course depends on the UTF-8 data.) */
6523 const U8 *s = start;
6525 while (s < send && uoffset--)
6535 while (UTF8_IS_CONTINUATION(*send))
6538 return send - start;
6541 /* For the string representation of the given scalar, find the byte
6542 corresponding to the passed in UTF-8 offset. uoffset0 and boffset0
6543 give another position in the string, *before* the sought offset, which
6544 (which is always true, as 0, 0 is a valid pair of positions), which should
6545 help reduce the amount of linear searching.
6546 If *mgp is non-NULL, it should point to the UTF-8 cache magic, which
6547 will be used to reduce the amount of linear searching. The cache will be
6548 created if necessary, and the found value offered to it for update. */
6550 S_sv_pos_u2b_cached(pTHX_ SV *const sv, MAGIC **const mgp, const U8 *const start,
6551 const U8 *const send, STRLEN uoffset,
6552 STRLEN uoffset0, STRLEN boffset0)
6554 STRLEN boffset = 0; /* Actually always set, but let's keep gcc happy. */
6556 bool at_end = FALSE;
6558 PERL_ARGS_ASSERT_SV_POS_U2B_CACHED;
6560 assert (uoffset >= uoffset0);
6567 && (*mgp || (SvTYPE(sv) >= SVt_PVMG &&
6568 (*mgp = mg_find(sv, PERL_MAGIC_utf8))))) {
6569 if ((*mgp)->mg_ptr) {
6570 STRLEN *cache = (STRLEN *) (*mgp)->mg_ptr;
6571 if (cache[0] == uoffset) {
6572 /* An exact match. */
6575 if (cache[2] == uoffset) {
6576 /* An exact match. */
6580 if (cache[0] < uoffset) {
6581 /* The cache already knows part of the way. */
6582 if (cache[0] > uoffset0) {
6583 /* The cache knows more than the passed in pair */
6584 uoffset0 = cache[0];
6585 boffset0 = cache[1];
6587 if ((*mgp)->mg_len != -1) {
6588 /* And we know the end too. */
6590 + sv_pos_u2b_midway(start + boffset0, send,
6592 (*mgp)->mg_len - uoffset0);
6594 uoffset -= uoffset0;
6596 + sv_pos_u2b_forwards(start + boffset0,
6597 send, &uoffset, &at_end);
6598 uoffset += uoffset0;
6601 else if (cache[2] < uoffset) {
6602 /* We're between the two cache entries. */
6603 if (cache[2] > uoffset0) {
6604 /* and the cache knows more than the passed in pair */
6605 uoffset0 = cache[2];
6606 boffset0 = cache[3];
6610 + sv_pos_u2b_midway(start + boffset0,
6613 cache[0] - uoffset0);
6616 + sv_pos_u2b_midway(start + boffset0,
6619 cache[2] - uoffset0);
6623 else if ((*mgp)->mg_len != -1) {
6624 /* If we can take advantage of a passed in offset, do so. */
6625 /* In fact, offset0 is either 0, or less than offset, so don't
6626 need to worry about the other possibility. */
6628 + sv_pos_u2b_midway(start + boffset0, send,
6630 (*mgp)->mg_len - uoffset0);
6635 if (!found || PL_utf8cache < 0) {
6636 STRLEN real_boffset;
6637 uoffset -= uoffset0;
6638 real_boffset = boffset0 + sv_pos_u2b_forwards(start + boffset0,
6639 send, &uoffset, &at_end);
6640 uoffset += uoffset0;
6642 if (found && PL_utf8cache < 0)
6643 assert_uft8_cache_coherent("sv_pos_u2b_cache", boffset,
6645 boffset = real_boffset;
6650 utf8_mg_len_cache_update(sv, mgp, uoffset);
6652 utf8_mg_pos_cache_update(sv, mgp, boffset, uoffset, send - start);
6659 =for apidoc sv_pos_u2b_flags
6661 Converts the value pointed to by offsetp from a count of UTF-8 chars from
6662 the start of the string, to a count of the equivalent number of bytes; if
6663 lenp is non-zero, it does the same to lenp, but this time starting from
6664 the offset, rather than from the start of the string. Handles type coercion.
6665 I<flags> is passed to C<SvPV_flags>, and usually should be
6666 C<SV_GMAGIC|SV_CONST_RETURN> to handle magic.
6672 * sv_pos_u2b_flags() uses, like sv_pos_b2u(), the mg_ptr of the potential
6673 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6674 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
6679 Perl_sv_pos_u2b_flags(pTHX_ SV *const sv, STRLEN uoffset, STRLEN *const lenp,
6686 PERL_ARGS_ASSERT_SV_POS_U2B_FLAGS;
6688 start = (U8*)SvPV_flags(sv, len, flags);
6690 const U8 * const send = start + len;
6692 boffset = sv_pos_u2b_cached(sv, &mg, start, send, uoffset, 0, 0);
6695 && *lenp /* don't bother doing work for 0, as its bytes equivalent
6696 is 0, and *lenp is already set to that. */) {
6697 /* Convert the relative offset to absolute. */
6698 const STRLEN uoffset2 = uoffset + *lenp;
6699 const STRLEN boffset2
6700 = sv_pos_u2b_cached(sv, &mg, start, send, uoffset2,
6701 uoffset, boffset) - boffset;
6715 =for apidoc sv_pos_u2b
6717 Converts the value pointed to by offsetp from a count of UTF-8 chars from
6718 the start of the string, to a count of the equivalent number of bytes; if
6719 lenp is non-zero, it does the same to lenp, but this time starting from
6720 the offset, rather than from the start of the string. Handles magic and
6723 Use C<sv_pos_u2b_flags> in preference, which correctly handles strings longer
6730 * sv_pos_u2b() uses, like sv_pos_b2u(), the mg_ptr of the potential
6731 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6732 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
6736 /* This function is subject to size and sign problems */
6739 Perl_sv_pos_u2b(pTHX_ register SV *const sv, I32 *const offsetp, I32 *const lenp)
6741 PERL_ARGS_ASSERT_SV_POS_U2B;
6744 STRLEN ulen = (STRLEN)*lenp;
6745 *offsetp = (I32)sv_pos_u2b_flags(sv, (STRLEN)*offsetp, &ulen,
6746 SV_GMAGIC|SV_CONST_RETURN);
6749 *offsetp = (I32)sv_pos_u2b_flags(sv, (STRLEN)*offsetp, NULL,
6750 SV_GMAGIC|SV_CONST_RETURN);
6755 S_utf8_mg_len_cache_update(pTHX_ SV *const sv, MAGIC **const mgp,
6758 PERL_ARGS_ASSERT_UTF8_MG_LEN_CACHE_UPDATE;
6762 if (!*mgp && (SvTYPE(sv) < SVt_PVMG ||
6763 !(*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
6764 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, &PL_vtbl_utf8, 0, 0);
6768 (*mgp)->mg_len = ulen;
6769 /* For now, treat "overflowed" as "still unknown". See RT #72924. */
6770 if (ulen != (STRLEN) (*mgp)->mg_len)
6771 (*mgp)->mg_len = -1;
6774 /* Create and update the UTF8 magic offset cache, with the proffered utf8/
6775 byte length pairing. The (byte) length of the total SV is passed in too,
6776 as blen, because for some (more esoteric) SVs, the call to SvPV_const()
6777 may not have updated SvCUR, so we can't rely on reading it directly.
6779 The proffered utf8/byte length pairing isn't used if the cache already has
6780 two pairs, and swapping either for the proffered pair would increase the
6781 RMS of the intervals between known byte offsets.
6783 The cache itself consists of 4 STRLEN values
6784 0: larger UTF-8 offset
6785 1: corresponding byte offset
6786 2: smaller UTF-8 offset
6787 3: corresponding byte offset
6789 Unused cache pairs have the value 0, 0.
6790 Keeping the cache "backwards" means that the invariant of
6791 cache[0] >= cache[2] is maintained even with empty slots, which means that
6792 the code that uses it doesn't need to worry if only 1 entry has actually
6793 been set to non-zero. It also makes the "position beyond the end of the
6794 cache" logic much simpler, as the first slot is always the one to start
6798 S_utf8_mg_pos_cache_update(pTHX_ SV *const sv, MAGIC **const mgp, const STRLEN byte,
6799 const STRLEN utf8, const STRLEN blen)
6803 PERL_ARGS_ASSERT_UTF8_MG_POS_CACHE_UPDATE;
6808 if (!*mgp && (SvTYPE(sv) < SVt_PVMG ||
6809 !(*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
6810 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, (MGVTBL*)&PL_vtbl_utf8, 0,
6812 (*mgp)->mg_len = -1;
6816 if (!(cache = (STRLEN *)(*mgp)->mg_ptr)) {
6817 Newxz(cache, PERL_MAGIC_UTF8_CACHESIZE * 2, STRLEN);
6818 (*mgp)->mg_ptr = (char *) cache;
6822 if (PL_utf8cache < 0 && SvPOKp(sv)) {
6823 /* SvPOKp() because it's possible that sv has string overloading, and
6824 therefore is a reference, hence SvPVX() is actually a pointer.
6825 This cures the (very real) symptoms of RT 69422, but I'm not actually
6826 sure whether we should even be caching the results of UTF-8
6827 operations on overloading, given that nothing stops overloading
6828 returning a different value every time it's called. */
6829 const U8 *start = (const U8 *) SvPVX_const(sv);
6830 const STRLEN realutf8 = utf8_length(start, start + byte);
6832 assert_uft8_cache_coherent("utf8_mg_pos_cache_update", utf8, realutf8,
6836 /* Cache is held with the later position first, to simplify the code
6837 that deals with unbounded ends. */
6839 ASSERT_UTF8_CACHE(cache);
6840 if (cache[1] == 0) {
6841 /* Cache is totally empty */
6844 } else if (cache[3] == 0) {
6845 if (byte > cache[1]) {
6846 /* New one is larger, so goes first. */
6847 cache[2] = cache[0];
6848 cache[3] = cache[1];
6856 #define THREEWAY_SQUARE(a,b,c,d) \
6857 ((float)((d) - (c))) * ((float)((d) - (c))) \
6858 + ((float)((c) - (b))) * ((float)((c) - (b))) \
6859 + ((float)((b) - (a))) * ((float)((b) - (a)))
6861 /* Cache has 2 slots in use, and we know three potential pairs.
6862 Keep the two that give the lowest RMS distance. Do the
6863 calculation in bytes simply because we always know the byte
6864 length. squareroot has the same ordering as the positive value,
6865 so don't bother with the actual square root. */
6866 const float existing = THREEWAY_SQUARE(0, cache[3], cache[1], blen);
6867 if (byte > cache[1]) {
6868 /* New position is after the existing pair of pairs. */
6869 const float keep_earlier
6870 = THREEWAY_SQUARE(0, cache[3], byte, blen);
6871 const float keep_later
6872 = THREEWAY_SQUARE(0, cache[1], byte, blen);
6874 if (keep_later < keep_earlier) {
6875 if (keep_later < existing) {
6876 cache[2] = cache[0];
6877 cache[3] = cache[1];
6883 if (keep_earlier < existing) {
6889 else if (byte > cache[3]) {
6890 /* New position is between the existing pair of pairs. */
6891 const float keep_earlier
6892 = THREEWAY_SQUARE(0, cache[3], byte, blen);
6893 const float keep_later
6894 = THREEWAY_SQUARE(0, byte, cache[1], blen);
6896 if (keep_later < keep_earlier) {
6897 if (keep_later < existing) {
6903 if (keep_earlier < existing) {
6910 /* New position is before the existing pair of pairs. */
6911 const float keep_earlier
6912 = THREEWAY_SQUARE(0, byte, cache[3], blen);
6913 const float keep_later
6914 = THREEWAY_SQUARE(0, byte, cache[1], blen);
6916 if (keep_later < keep_earlier) {
6917 if (keep_later < existing) {
6923 if (keep_earlier < existing) {
6924 cache[0] = cache[2];
6925 cache[1] = cache[3];
6932 ASSERT_UTF8_CACHE(cache);
6935 /* We already know all of the way, now we may be able to walk back. The same
6936 assumption is made as in S_sv_pos_u2b_midway(), namely that walking
6937 backward is half the speed of walking forward. */
6939 S_sv_pos_b2u_midway(pTHX_ const U8 *const s, const U8 *const target,
6940 const U8 *end, STRLEN endu)
6942 const STRLEN forw = target - s;
6943 STRLEN backw = end - target;
6945 PERL_ARGS_ASSERT_SV_POS_B2U_MIDWAY;
6947 if (forw < 2 * backw) {
6948 return utf8_length(s, target);
6951 while (end > target) {
6953 while (UTF8_IS_CONTINUATION(*end)) {
6962 =for apidoc sv_pos_b2u
6964 Converts the value pointed to by offsetp from a count of bytes from the
6965 start of the string, to a count of the equivalent number of UTF-8 chars.
6966 Handles magic and type coercion.
6972 * sv_pos_b2u() uses, like sv_pos_u2b(), the mg_ptr of the potential
6973 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6978 Perl_sv_pos_b2u(pTHX_ register SV *const sv, I32 *const offsetp)
6981 const STRLEN byte = *offsetp;
6982 STRLEN len = 0; /* Actually always set, but let's keep gcc happy. */
6988 PERL_ARGS_ASSERT_SV_POS_B2U;
6993 s = (const U8*)SvPV_const(sv, blen);
6996 Perl_croak(aTHX_ "panic: sv_pos_b2u: bad byte offset");
7002 && SvTYPE(sv) >= SVt_PVMG
7003 && (mg = mg_find(sv, PERL_MAGIC_utf8)))
7006 STRLEN * const cache = (STRLEN *) mg->mg_ptr;
7007 if (cache[1] == byte) {
7008 /* An exact match. */
7009 *offsetp = cache[0];
7012 if (cache[3] == byte) {
7013 /* An exact match. */
7014 *offsetp = cache[2];
7018 if (cache[1] < byte) {
7019 /* We already know part of the way. */
7020 if (mg->mg_len != -1) {
7021 /* Actually, we know the end too. */
7023 + S_sv_pos_b2u_midway(aTHX_ s + cache[1], send,
7024 s + blen, mg->mg_len - cache[0]);
7026 len = cache[0] + utf8_length(s + cache[1], send);
7029 else if (cache[3] < byte) {
7030 /* We're between the two cached pairs, so we do the calculation
7031 offset by the byte/utf-8 positions for the earlier pair,
7032 then add the utf-8 characters from the string start to
7034 len = S_sv_pos_b2u_midway(aTHX_ s + cache[3], send,
7035 s + cache[1], cache[0] - cache[2])
7039 else { /* cache[3] > byte */
7040 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + cache[3],
7044 ASSERT_UTF8_CACHE(cache);
7046 } else if (mg->mg_len != -1) {
7047 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + blen, mg->mg_len);
7051 if (!found || PL_utf8cache < 0) {
7052 const STRLEN real_len = utf8_length(s, send);
7054 if (found && PL_utf8cache < 0)
7055 assert_uft8_cache_coherent("sv_pos_b2u", len, real_len, sv);
7062 utf8_mg_len_cache_update(sv, &mg, len);
7064 utf8_mg_pos_cache_update(sv, &mg, byte, len, blen);
7069 S_assert_uft8_cache_coherent(pTHX_ const char *const func, STRLEN from_cache,
7070 STRLEN real, SV *const sv)
7072 PERL_ARGS_ASSERT_ASSERT_UFT8_CACHE_COHERENT;
7074 /* As this is debugging only code, save space by keeping this test here,
7075 rather than inlining it in all the callers. */
7076 if (from_cache == real)
7079 /* Need to turn the assertions off otherwise we may recurse infinitely
7080 while printing error messages. */
7081 SAVEI8(PL_utf8cache);
7083 Perl_croak(aTHX_ "panic: %s cache %"UVuf" real %"UVuf" for %"SVf,
7084 func, (UV) from_cache, (UV) real, SVfARG(sv));
7090 Returns a boolean indicating whether the strings in the two SVs are
7091 identical. Is UTF-8 and 'use bytes' aware, handles get magic, and will
7092 coerce its args to strings if necessary.
7094 =for apidoc sv_eq_flags
7096 Returns a boolean indicating whether the strings in the two SVs are
7097 identical. Is UTF-8 and 'use bytes' aware and coerces its args to strings
7098 if necessary. If the flags include SV_GMAGIC, it handles get-magic, too.
7104 Perl_sv_eq_flags(pTHX_ register SV *sv1, register SV *sv2, const U32 flags)
7113 SV* svrecode = NULL;
7120 /* if pv1 and pv2 are the same, second SvPV_const call may
7121 * invalidate pv1 (if we are handling magic), so we may need to
7123 if (sv1 == sv2 && flags & SV_GMAGIC
7124 && (SvTHINKFIRST(sv1) || SvGMAGICAL(sv1))) {
7125 pv1 = SvPV_const(sv1, cur1);
7126 sv1 = newSVpvn_flags(pv1, cur1, SVs_TEMP | SvUTF8(sv2));
7128 pv1 = SvPV_flags_const(sv1, cur1, flags);
7136 pv2 = SvPV_flags_const(sv2, cur2, flags);
7138 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
7139 /* Differing utf8ness.
7140 * Do not UTF8size the comparands as a side-effect. */
7143 svrecode = newSVpvn(pv2, cur2);
7144 sv_recode_to_utf8(svrecode, PL_encoding);
7145 pv2 = SvPV_const(svrecode, cur2);
7148 svrecode = newSVpvn(pv1, cur1);
7149 sv_recode_to_utf8(svrecode, PL_encoding);
7150 pv1 = SvPV_const(svrecode, cur1);
7152 /* Now both are in UTF-8. */
7154 SvREFCNT_dec(svrecode);
7160 /* sv1 is the UTF-8 one */
7161 return bytes_cmp_utf8((const U8*)pv2, cur2,
7162 (const U8*)pv1, cur1) == 0;
7165 /* sv2 is the UTF-8 one */
7166 return bytes_cmp_utf8((const U8*)pv1, cur1,
7167 (const U8*)pv2, cur2) == 0;
7173 eq = (pv1 == pv2) || memEQ(pv1, pv2, cur1);
7175 SvREFCNT_dec(svrecode);
7185 Compares the strings in two SVs. Returns -1, 0, or 1 indicating whether the
7186 string in C<sv1> is less than, equal to, or greater than the string in
7187 C<sv2>. Is UTF-8 and 'use bytes' aware, handles get magic, and will
7188 coerce its args to strings if necessary. See also C<sv_cmp_locale>.
7190 =for apidoc sv_cmp_flags
7192 Compares the strings in two SVs. Returns -1, 0, or 1 indicating whether the
7193 string in C<sv1> is less than, equal to, or greater than the string in
7194 C<sv2>. Is UTF-8 and 'use bytes' aware and will coerce its args to strings
7195 if necessary. If the flags include SV_GMAGIC, it handles get magic. See
7196 also C<sv_cmp_locale_flags>.
7202 Perl_sv_cmp(pTHX_ register SV *const sv1, register SV *const sv2)
7204 return sv_cmp_flags(sv1, sv2, SV_GMAGIC);
7208 Perl_sv_cmp_flags(pTHX_ register SV *const sv1, register SV *const sv2,
7213 const char *pv1, *pv2;
7216 SV *svrecode = NULL;
7223 pv1 = SvPV_flags_const(sv1, cur1, flags);
7230 pv2 = SvPV_flags_const(sv2, cur2, flags);
7232 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
7233 /* Differing utf8ness.
7234 * Do not UTF8size the comparands as a side-effect. */
7237 svrecode = newSVpvn(pv2, cur2);
7238 sv_recode_to_utf8(svrecode, PL_encoding);
7239 pv2 = SvPV_const(svrecode, cur2);
7242 const int retval = -bytes_cmp_utf8((const U8*)pv2, cur2,
7243 (const U8*)pv1, cur1);
7244 return retval ? retval < 0 ? -1 : +1 : 0;
7249 svrecode = newSVpvn(pv1, cur1);
7250 sv_recode_to_utf8(svrecode, PL_encoding);
7251 pv1 = SvPV_const(svrecode, cur1);
7254 const int retval = bytes_cmp_utf8((const U8*)pv1, cur1,
7255 (const U8*)pv2, cur2);
7256 return retval ? retval < 0 ? -1 : +1 : 0;
7262 cmp = cur2 ? -1 : 0;
7266 const I32 retval = memcmp((const void*)pv1, (const void*)pv2, cur1 < cur2 ? cur1 : cur2);
7269 cmp = retval < 0 ? -1 : 1;
7270 } else if (cur1 == cur2) {
7273 cmp = cur1 < cur2 ? -1 : 1;
7277 SvREFCNT_dec(svrecode);
7285 =for apidoc sv_cmp_locale
7287 Compares the strings in two SVs in a locale-aware manner. Is UTF-8 and
7288 'use bytes' aware, handles get magic, and will coerce its args to strings
7289 if necessary. See also C<sv_cmp>.
7291 =for apidoc sv_cmp_locale_flags
7293 Compares the strings in two SVs in a locale-aware manner. Is UTF-8 and
7294 'use bytes' aware and will coerce its args to strings if necessary. If the
7295 flags contain SV_GMAGIC, it handles get magic. See also C<sv_cmp_flags>.
7301 Perl_sv_cmp_locale(pTHX_ register SV *const sv1, register SV *const sv2)
7303 return sv_cmp_locale_flags(sv1, sv2, SV_GMAGIC);
7307 Perl_sv_cmp_locale_flags(pTHX_ register SV *const sv1, register SV *const sv2,
7311 #ifdef USE_LOCALE_COLLATE
7317 if (PL_collation_standard)
7321 pv1 = sv1 ? sv_collxfrm_flags(sv1, &len1, flags) : (char *) NULL;
7323 pv2 = sv2 ? sv_collxfrm_flags(sv2, &len2, flags) : (char *) NULL;
7325 if (!pv1 || !len1) {
7336 retval = memcmp((void*)pv1, (void*)pv2, len1 < len2 ? len1 : len2);
7339 return retval < 0 ? -1 : 1;
7342 * When the result of collation is equality, that doesn't mean
7343 * that there are no differences -- some locales exclude some
7344 * characters from consideration. So to avoid false equalities,
7345 * we use the raw string as a tiebreaker.
7351 #endif /* USE_LOCALE_COLLATE */
7353 return sv_cmp(sv1, sv2);
7357 #ifdef USE_LOCALE_COLLATE
7360 =for apidoc sv_collxfrm
7362 This calls C<sv_collxfrm_flags> with the SV_GMAGIC flag. See
7363 C<sv_collxfrm_flags>.
7365 =for apidoc sv_collxfrm_flags
7367 Add Collate Transform magic to an SV if it doesn't already have it. If the
7368 flags contain SV_GMAGIC, it handles get-magic.
7370 Any scalar variable may carry PERL_MAGIC_collxfrm magic that contains the
7371 scalar data of the variable, but transformed to such a format that a normal
7372 memory comparison can be used to compare the data according to the locale
7379 Perl_sv_collxfrm_flags(pTHX_ SV *const sv, STRLEN *const nxp, const I32 flags)
7384 PERL_ARGS_ASSERT_SV_COLLXFRM_FLAGS;
7386 mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_collxfrm) : (MAGIC *) NULL;
7387 if (!mg || !mg->mg_ptr || *(U32*)mg->mg_ptr != PL_collation_ix) {
7393 Safefree(mg->mg_ptr);
7394 s = SvPV_flags_const(sv, len, flags);
7395 if ((xf = mem_collxfrm(s, len, &xlen))) {
7397 #ifdef PERL_OLD_COPY_ON_WRITE
7399 sv_force_normal_flags(sv, 0);
7401 mg = sv_magicext(sv, 0, PERL_MAGIC_collxfrm, &PL_vtbl_collxfrm,
7415 if (mg && mg->mg_ptr) {
7417 return mg->mg_ptr + sizeof(PL_collation_ix);
7425 #endif /* USE_LOCALE_COLLATE */
7428 S_sv_gets_append_to_utf8(pTHX_ SV *const sv, PerlIO *const fp, I32 append)
7430 SV * const tsv = newSV(0);
7433 sv_gets(tsv, fp, 0);
7434 sv_utf8_upgrade_nomg(tsv);
7435 SvCUR_set(sv,append);
7438 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7442 S_sv_gets_read_record(pTHX_ SV *const sv, PerlIO *const fp, I32 append)
7445 const U32 recsize = SvUV(SvRV(PL_rs)); /* RsRECORD() guarantees > 0. */
7446 /* Grab the size of the record we're getting */
7447 char *const buffer = SvGROW(sv, (STRLEN)(recsize + append + 1)) + append;
7454 /* VMS wants read instead of fread, because fread doesn't respect */
7455 /* RMS record boundaries. This is not necessarily a good thing to be */
7456 /* doing, but we've got no other real choice - except avoid stdio
7457 as implementation - perhaps write a :vms layer ?
7459 fd = PerlIO_fileno(fp);
7461 bytesread = PerlLIO_read(fd, buffer, recsize);
7463 else /* in-memory file from PerlIO::Scalar */
7466 bytesread = PerlIO_read(fp, buffer, recsize);
7471 SvCUR_set(sv, bytesread + append);
7472 buffer[bytesread] = '\0';
7473 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7479 Get a line from the filehandle and store it into the SV, optionally
7480 appending to the currently-stored string.
7486 Perl_sv_gets(pTHX_ register SV *const sv, register PerlIO *const fp, I32 append)
7491 register STDCHAR rslast;
7492 register STDCHAR *bp;
7497 PERL_ARGS_ASSERT_SV_GETS;
7499 if (SvTHINKFIRST(sv))
7500 sv_force_normal_flags(sv, append ? 0 : SV_COW_DROP_PV);
7501 /* XXX. If you make this PVIV, then copy on write can copy scalars read
7503 However, perlbench says it's slower, because the existing swipe code
7504 is faster than copy on write.
7505 Swings and roundabouts. */
7506 SvUPGRADE(sv, SVt_PV);
7511 if (PerlIO_isutf8(fp)) {
7513 sv_utf8_upgrade_nomg(sv);
7514 sv_pos_u2b(sv,&append,0);
7516 } else if (SvUTF8(sv)) {
7517 return S_sv_gets_append_to_utf8(aTHX_ sv, fp, append);
7525 if (PerlIO_isutf8(fp))
7528 if (IN_PERL_COMPILETIME) {
7529 /* we always read code in line mode */
7533 else if (RsSNARF(PL_rs)) {
7534 /* If it is a regular disk file use size from stat() as estimate
7535 of amount we are going to read -- may result in mallocing
7536 more memory than we really need if the layers below reduce
7537 the size we read (e.g. CRLF or a gzip layer).
7540 if (!PerlLIO_fstat(PerlIO_fileno(fp), &st) && S_ISREG(st.st_mode)) {
7541 const Off_t offset = PerlIO_tell(fp);
7542 if (offset != (Off_t) -1 && st.st_size + append > offset) {
7543 (void) SvGROW(sv, (STRLEN)((st.st_size - offset) + append + 1));
7549 else if (RsRECORD(PL_rs)) {
7550 return S_sv_gets_read_record(aTHX_ sv, fp, append);
7552 else if (RsPARA(PL_rs)) {
7558 /* Get $/ i.e. PL_rs into same encoding as stream wants */
7559 if (PerlIO_isutf8(fp)) {
7560 rsptr = SvPVutf8(PL_rs, rslen);
7563 if (SvUTF8(PL_rs)) {
7564 if (!sv_utf8_downgrade(PL_rs, TRUE)) {
7565 Perl_croak(aTHX_ "Wide character in $/");
7568 rsptr = SvPV_const(PL_rs, rslen);
7572 rslast = rslen ? rsptr[rslen - 1] : '\0';
7574 if (rspara) { /* have to do this both before and after */
7575 do { /* to make sure file boundaries work right */
7578 i = PerlIO_getc(fp);
7582 PerlIO_ungetc(fp,i);
7588 /* See if we know enough about I/O mechanism to cheat it ! */
7590 /* This used to be #ifdef test - it is made run-time test for ease
7591 of abstracting out stdio interface. One call should be cheap
7592 enough here - and may even be a macro allowing compile
7596 if (PerlIO_fast_gets(fp)) {
7599 * We're going to steal some values from the stdio struct
7600 * and put EVERYTHING in the innermost loop into registers.
7602 register STDCHAR *ptr;
7606 #if defined(VMS) && defined(PERLIO_IS_STDIO)
7607 /* An ungetc()d char is handled separately from the regular
7608 * buffer, so we getc() it back out and stuff it in the buffer.
7610 i = PerlIO_getc(fp);
7611 if (i == EOF) return 0;
7612 *(--((*fp)->_ptr)) = (unsigned char) i;
7616 /* Here is some breathtakingly efficient cheating */
7618 cnt = PerlIO_get_cnt(fp); /* get count into register */
7619 /* make sure we have the room */
7620 if ((I32)(SvLEN(sv) - append) <= cnt + 1) {
7621 /* Not room for all of it
7622 if we are looking for a separator and room for some
7624 if (rslen && cnt > 80 && (I32)SvLEN(sv) > append) {
7625 /* just process what we have room for */
7626 shortbuffered = cnt - SvLEN(sv) + append + 1;
7627 cnt -= shortbuffered;
7631 /* remember that cnt can be negative */
7632 SvGROW(sv, (STRLEN)(append + (cnt <= 0 ? 2 : (cnt + 1))));
7637 bp = (STDCHAR*)SvPVX_const(sv) + append; /* move these two too to registers */
7638 ptr = (STDCHAR*)PerlIO_get_ptr(fp);
7639 DEBUG_P(PerlIO_printf(Perl_debug_log,
7640 "Screamer: entering, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7641 DEBUG_P(PerlIO_printf(Perl_debug_log,
7642 "Screamer: entering: PerlIO * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7643 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7644 PTR2UV(PerlIO_has_base(fp) ? PerlIO_get_base(fp) : 0)));
7649 while (cnt > 0) { /* this | eat */
7651 if ((*bp++ = *ptr++) == rslast) /* really | dust */
7652 goto thats_all_folks; /* screams | sed :-) */
7656 Copy(ptr, bp, cnt, char); /* this | eat */
7657 bp += cnt; /* screams | dust */
7658 ptr += cnt; /* louder | sed :-) */
7660 assert (!shortbuffered);
7661 goto cannot_be_shortbuffered;
7665 if (shortbuffered) { /* oh well, must extend */
7666 cnt = shortbuffered;
7668 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
7670 SvGROW(sv, SvLEN(sv) + append + cnt + 2);
7671 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
7675 cannot_be_shortbuffered:
7676 DEBUG_P(PerlIO_printf(Perl_debug_log,
7677 "Screamer: going to getc, ptr=%"UVuf", cnt=%ld\n",
7678 PTR2UV(ptr),(long)cnt));
7679 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* deregisterize cnt and ptr */
7681 DEBUG_Pv(PerlIO_printf(Perl_debug_log,
7682 "Screamer: pre: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7683 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7684 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7686 /* This used to call 'filbuf' in stdio form, but as that behaves like
7687 getc when cnt <= 0 we use PerlIO_getc here to avoid introducing
7688 another abstraction. */
7689 i = PerlIO_getc(fp); /* get more characters */
7691 DEBUG_Pv(PerlIO_printf(Perl_debug_log,
7692 "Screamer: post: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7693 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7694 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7696 cnt = PerlIO_get_cnt(fp);
7697 ptr = (STDCHAR*)PerlIO_get_ptr(fp); /* reregisterize cnt and ptr */
7698 DEBUG_P(PerlIO_printf(Perl_debug_log,
7699 "Screamer: after getc, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7701 if (i == EOF) /* all done for ever? */
7702 goto thats_really_all_folks;
7704 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
7706 SvGROW(sv, bpx + cnt + 2);
7707 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
7709 *bp++ = (STDCHAR)i; /* store character from PerlIO_getc */
7711 if (rslen && (STDCHAR)i == rslast) /* all done for now? */
7712 goto thats_all_folks;
7716 if ((rslen > 1 && (STRLEN)(bp - (STDCHAR*)SvPVX_const(sv)) < rslen) ||
7717 memNE((char*)bp - rslen, rsptr, rslen))
7718 goto screamer; /* go back to the fray */
7719 thats_really_all_folks:
7721 cnt += shortbuffered;
7722 DEBUG_P(PerlIO_printf(Perl_debug_log,
7723 "Screamer: quitting, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7724 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* put these back or we're in trouble */
7725 DEBUG_P(PerlIO_printf(Perl_debug_log,
7726 "Screamer: end: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7727 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7728 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7730 SvCUR_set(sv, bp - (STDCHAR*)SvPVX_const(sv)); /* set length */
7731 DEBUG_P(PerlIO_printf(Perl_debug_log,
7732 "Screamer: done, len=%ld, string=|%.*s|\n",
7733 (long)SvCUR(sv),(int)SvCUR(sv),SvPVX_const(sv)));
7737 /*The big, slow, and stupid way. */
7738 #ifdef USE_HEAP_INSTEAD_OF_STACK /* Even slower way. */
7739 STDCHAR *buf = NULL;
7740 Newx(buf, 8192, STDCHAR);
7748 register const STDCHAR * const bpe = buf + sizeof(buf);
7750 while ((i = PerlIO_getc(fp)) != EOF && (*bp++ = (STDCHAR)i) != rslast && bp < bpe)
7751 ; /* keep reading */
7755 cnt = PerlIO_read(fp,(char*)buf, sizeof(buf));
7756 /* Accommodate broken VAXC compiler, which applies U8 cast to
7757 * both args of ?: operator, causing EOF to change into 255
7760 i = (U8)buf[cnt - 1];
7766 cnt = 0; /* we do need to re-set the sv even when cnt <= 0 */
7768 sv_catpvn(sv, (char *) buf, cnt);
7770 sv_setpvn(sv, (char *) buf, cnt);
7772 if (i != EOF && /* joy */
7774 SvCUR(sv) < rslen ||
7775 memNE(SvPVX_const(sv) + SvCUR(sv) - rslen, rsptr, rslen)))
7779 * If we're reading from a TTY and we get a short read,
7780 * indicating that the user hit his EOF character, we need
7781 * to notice it now, because if we try to read from the TTY
7782 * again, the EOF condition will disappear.
7784 * The comparison of cnt to sizeof(buf) is an optimization
7785 * that prevents unnecessary calls to feof().
7789 if (!(cnt < (I32)sizeof(buf) && PerlIO_eof(fp)))
7793 #ifdef USE_HEAP_INSTEAD_OF_STACK
7798 if (rspara) { /* have to do this both before and after */
7799 while (i != EOF) { /* to make sure file boundaries work right */
7800 i = PerlIO_getc(fp);
7802 PerlIO_ungetc(fp,i);
7808 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7814 Auto-increment of the value in the SV, doing string to numeric conversion
7815 if necessary. Handles 'get' magic and operator overloading.
7821 Perl_sv_inc(pTHX_ register SV *const sv)
7830 =for apidoc sv_inc_nomg
7832 Auto-increment of the value in the SV, doing string to numeric conversion
7833 if necessary. Handles operator overloading. Skips handling 'get' magic.
7839 Perl_sv_inc_nomg(pTHX_ register SV *const sv)
7847 if (SvTHINKFIRST(sv)) {
7849 sv_force_normal_flags(sv, 0);
7850 if (SvREADONLY(sv)) {
7851 if (IN_PERL_RUNTIME)
7852 Perl_croak_no_modify(aTHX);
7856 if (SvAMAGIC(sv) && AMG_CALLunary(sv, inc_amg))
7858 i = PTR2IV(SvRV(sv));
7863 flags = SvFLAGS(sv);
7864 if ((flags & (SVp_NOK|SVp_IOK)) == SVp_NOK) {
7865 /* It's (privately or publicly) a float, but not tested as an
7866 integer, so test it to see. */
7868 flags = SvFLAGS(sv);
7870 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
7871 /* It's publicly an integer, or privately an integer-not-float */
7872 #ifdef PERL_PRESERVE_IVUV
7876 if (SvUVX(sv) == UV_MAX)
7877 sv_setnv(sv, UV_MAX_P1);
7879 (void)SvIOK_only_UV(sv);
7880 SvUV_set(sv, SvUVX(sv) + 1);
7882 if (SvIVX(sv) == IV_MAX)
7883 sv_setuv(sv, (UV)IV_MAX + 1);
7885 (void)SvIOK_only(sv);
7886 SvIV_set(sv, SvIVX(sv) + 1);
7891 if (flags & SVp_NOK) {
7892 const NV was = SvNVX(sv);
7893 if (NV_OVERFLOWS_INTEGERS_AT &&
7894 was >= NV_OVERFLOWS_INTEGERS_AT) {
7895 Perl_ck_warner(aTHX_ packWARN(WARN_IMPRECISION),
7896 "Lost precision when incrementing %" NVff " by 1",
7899 (void)SvNOK_only(sv);
7900 SvNV_set(sv, was + 1.0);
7904 if (!(flags & SVp_POK) || !*SvPVX_const(sv)) {
7905 if ((flags & SVTYPEMASK) < SVt_PVIV)
7906 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV ? SVt_PVIV : SVt_IV));
7907 (void)SvIOK_only(sv);
7912 while (isALPHA(*d)) d++;
7913 while (isDIGIT(*d)) d++;
7914 if (d < SvEND(sv)) {
7915 #ifdef PERL_PRESERVE_IVUV
7916 /* Got to punt this as an integer if needs be, but we don't issue
7917 warnings. Probably ought to make the sv_iv_please() that does
7918 the conversion if possible, and silently. */
7919 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
7920 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
7921 /* Need to try really hard to see if it's an integer.
7922 9.22337203685478e+18 is an integer.
7923 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
7924 so $a="9.22337203685478e+18"; $a+0; $a++
7925 needs to be the same as $a="9.22337203685478e+18"; $a++
7932 /* sv_2iv *should* have made this an NV */
7933 if (flags & SVp_NOK) {
7934 (void)SvNOK_only(sv);
7935 SvNV_set(sv, SvNVX(sv) + 1.0);
7938 /* I don't think we can get here. Maybe I should assert this
7939 And if we do get here I suspect that sv_setnv will croak. NWC
7941 #if defined(USE_LONG_DOUBLE)
7942 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",
7943 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7945 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_inc punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
7946 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7949 #endif /* PERL_PRESERVE_IVUV */
7950 sv_setnv(sv,Atof(SvPVX_const(sv)) + 1.0);
7954 while (d >= SvPVX_const(sv)) {
7962 /* MKS: The original code here died if letters weren't consecutive.
7963 * at least it didn't have to worry about non-C locales. The
7964 * new code assumes that ('z'-'a')==('Z'-'A'), letters are
7965 * arranged in order (although not consecutively) and that only
7966 * [A-Za-z] are accepted by isALPHA in the C locale.
7968 if (*d != 'z' && *d != 'Z') {
7969 do { ++*d; } while (!isALPHA(*d));
7972 *(d--) -= 'z' - 'a';
7977 *(d--) -= 'z' - 'a' + 1;
7981 /* oh,oh, the number grew */
7982 SvGROW(sv, SvCUR(sv) + 2);
7983 SvCUR_set(sv, SvCUR(sv) + 1);
7984 for (d = SvPVX(sv) + SvCUR(sv); d > SvPVX_const(sv); d--)
7995 Auto-decrement of the value in the SV, doing string to numeric conversion
7996 if necessary. Handles 'get' magic and operator overloading.
8002 Perl_sv_dec(pTHX_ register SV *const sv)
8012 =for apidoc sv_dec_nomg
8014 Auto-decrement of the value in the SV, doing string to numeric conversion
8015 if necessary. Handles operator overloading. Skips handling 'get' magic.
8021 Perl_sv_dec_nomg(pTHX_ register SV *const sv)
8028 if (SvTHINKFIRST(sv)) {
8030 sv_force_normal_flags(sv, 0);
8031 if (SvREADONLY(sv)) {
8032 if (IN_PERL_RUNTIME)
8033 Perl_croak_no_modify(aTHX);
8037 if (SvAMAGIC(sv) && AMG_CALLunary(sv, dec_amg))
8039 i = PTR2IV(SvRV(sv));
8044 /* Unlike sv_inc we don't have to worry about string-never-numbers
8045 and keeping them magic. But we mustn't warn on punting */
8046 flags = SvFLAGS(sv);
8047 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
8048 /* It's publicly an integer, or privately an integer-not-float */
8049 #ifdef PERL_PRESERVE_IVUV
8053 if (SvUVX(sv) == 0) {
8054 (void)SvIOK_only(sv);
8058 (void)SvIOK_only_UV(sv);
8059 SvUV_set(sv, SvUVX(sv) - 1);
8062 if (SvIVX(sv) == IV_MIN) {
8063 sv_setnv(sv, (NV)IV_MIN);
8067 (void)SvIOK_only(sv);
8068 SvIV_set(sv, SvIVX(sv) - 1);
8073 if (flags & SVp_NOK) {
8076 const NV was = SvNVX(sv);
8077 if (NV_OVERFLOWS_INTEGERS_AT &&
8078 was <= -NV_OVERFLOWS_INTEGERS_AT) {
8079 Perl_ck_warner(aTHX_ packWARN(WARN_IMPRECISION),
8080 "Lost precision when decrementing %" NVff " by 1",
8083 (void)SvNOK_only(sv);
8084 SvNV_set(sv, was - 1.0);
8088 if (!(flags & SVp_POK)) {
8089 if ((flags & SVTYPEMASK) < SVt_PVIV)
8090 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV) ? SVt_PVIV : SVt_IV);
8092 (void)SvIOK_only(sv);
8095 #ifdef PERL_PRESERVE_IVUV
8097 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
8098 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
8099 /* Need to try really hard to see if it's an integer.
8100 9.22337203685478e+18 is an integer.
8101 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
8102 so $a="9.22337203685478e+18"; $a+0; $a--
8103 needs to be the same as $a="9.22337203685478e+18"; $a--
8110 /* sv_2iv *should* have made this an NV */
8111 if (flags & SVp_NOK) {
8112 (void)SvNOK_only(sv);
8113 SvNV_set(sv, SvNVX(sv) - 1.0);
8116 /* I don't think we can get here. Maybe I should assert this
8117 And if we do get here I suspect that sv_setnv will croak. NWC
8119 #if defined(USE_LONG_DOUBLE)
8120 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",
8121 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
8123 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_dec punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
8124 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
8128 #endif /* PERL_PRESERVE_IVUV */
8129 sv_setnv(sv,Atof(SvPVX_const(sv)) - 1.0); /* punt */
8132 /* this define is used to eliminate a chunk of duplicated but shared logic
8133 * it has the suffix __SV_C to signal that it isnt API, and isnt meant to be
8134 * used anywhere but here - yves
8136 #define PUSH_EXTEND_MORTAL__SV_C(AnSv) \
8139 PL_tmps_stack[++PL_tmps_ix] = (AnSv); \
8143 =for apidoc sv_mortalcopy
8145 Creates a new SV which is a copy of the original SV (using C<sv_setsv>).
8146 The new SV is marked as mortal. It will be destroyed "soon", either by an
8147 explicit call to FREETMPS, or by an implicit call at places such as
8148 statement boundaries. See also C<sv_newmortal> and C<sv_2mortal>.
8153 /* Make a string that will exist for the duration of the expression
8154 * evaluation. Actually, it may have to last longer than that, but
8155 * hopefully we won't free it until it has been assigned to a
8156 * permanent location. */
8159 Perl_sv_mortalcopy(pTHX_ SV *const oldstr)
8165 sv_setsv(sv,oldstr);
8166 PUSH_EXTEND_MORTAL__SV_C(sv);
8172 =for apidoc sv_newmortal
8174 Creates a new null SV which is mortal. The reference count of the SV is
8175 set to 1. It will be destroyed "soon", either by an explicit call to
8176 FREETMPS, or by an implicit call at places such as statement boundaries.
8177 See also C<sv_mortalcopy> and C<sv_2mortal>.
8183 Perl_sv_newmortal(pTHX)
8189 SvFLAGS(sv) = SVs_TEMP;
8190 PUSH_EXTEND_MORTAL__SV_C(sv);
8196 =for apidoc newSVpvn_flags
8198 Creates a new SV and copies a string into it. The reference count for the
8199 SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
8200 string. You are responsible for ensuring that the source string is at least
8201 C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
8202 Currently the only flag bits accepted are C<SVf_UTF8> and C<SVs_TEMP>.
8203 If C<SVs_TEMP> is set, then C<sv_2mortal()> is called on the result before
8204 returning. If C<SVf_UTF8> is set, C<s> is considered to be in UTF-8 and the
8205 C<SVf_UTF8> flag will be set on the new SV.
8206 C<newSVpvn_utf8()> is a convenience wrapper for this function, defined as
8208 #define newSVpvn_utf8(s, len, u) \
8209 newSVpvn_flags((s), (len), (u) ? SVf_UTF8 : 0)
8215 Perl_newSVpvn_flags(pTHX_ const char *const s, const STRLEN len, const U32 flags)
8220 /* All the flags we don't support must be zero.
8221 And we're new code so I'm going to assert this from the start. */
8222 assert(!(flags & ~(SVf_UTF8|SVs_TEMP)));
8224 sv_setpvn(sv,s,len);
8226 /* This code used to a sv_2mortal(), however we now unroll the call to sv_2mortal()
8227 * and do what it does ourselves here.
8228 * Since we have asserted that flags can only have the SVf_UTF8 and/or SVs_TEMP flags
8229 * set above we can use it to enable the sv flags directly (bypassing SvTEMP_on), which
8230 * in turn means we dont need to mask out the SVf_UTF8 flag below, which means that we
8231 * eliminate quite a few steps than it looks - Yves (explaining patch by gfx)
8234 SvFLAGS(sv) |= flags;
8236 if(flags & SVs_TEMP){
8237 PUSH_EXTEND_MORTAL__SV_C(sv);
8244 =for apidoc sv_2mortal
8246 Marks an existing SV as mortal. The SV will be destroyed "soon", either
8247 by an explicit call to FREETMPS, or by an implicit call at places such as
8248 statement boundaries. SvTEMP() is turned on which means that the SV's
8249 string buffer can be "stolen" if this SV is copied. See also C<sv_newmortal>
8250 and C<sv_mortalcopy>.
8256 Perl_sv_2mortal(pTHX_ register SV *const sv)
8261 if (SvREADONLY(sv) && SvIMMORTAL(sv))
8263 PUSH_EXTEND_MORTAL__SV_C(sv);
8271 Creates a new SV and copies a string into it. The reference count for the
8272 SV is set to 1. If C<len> is zero, Perl will compute the length using
8273 strlen(). For efficiency, consider using C<newSVpvn> instead.
8279 Perl_newSVpv(pTHX_ const char *const s, const STRLEN len)
8285 sv_setpvn(sv, s, len || s == NULL ? len : strlen(s));
8290 =for apidoc newSVpvn
8292 Creates a new SV and copies a string into it. The reference count for the
8293 SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
8294 string. You are responsible for ensuring that the source string is at least
8295 C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
8301 Perl_newSVpvn(pTHX_ const char *const s, const STRLEN len)
8307 sv_setpvn(sv,s,len);
8312 =for apidoc newSVhek
8314 Creates a new SV from the hash key structure. It will generate scalars that
8315 point to the shared string table where possible. Returns a new (undefined)
8316 SV if the hek is NULL.
8322 Perl_newSVhek(pTHX_ const HEK *const hek)
8332 if (HEK_LEN(hek) == HEf_SVKEY) {
8333 return newSVsv(*(SV**)HEK_KEY(hek));
8335 const int flags = HEK_FLAGS(hek);
8336 if (flags & HVhek_WASUTF8) {
8338 Andreas would like keys he put in as utf8 to come back as utf8
8340 STRLEN utf8_len = HEK_LEN(hek);
8341 SV * const sv = newSV_type(SVt_PV);
8342 char *as_utf8 = (char *)bytes_to_utf8 ((U8*)HEK_KEY(hek), &utf8_len);
8343 /* bytes_to_utf8() allocates a new string, which we can repurpose: */
8344 sv_usepvn_flags(sv, as_utf8, utf8_len, SV_HAS_TRAILING_NUL);
8347 } else if (flags & (HVhek_REHASH|HVhek_UNSHARED)) {
8348 /* We don't have a pointer to the hv, so we have to replicate the
8349 flag into every HEK. This hv is using custom a hasing
8350 algorithm. Hence we can't return a shared string scalar, as
8351 that would contain the (wrong) hash value, and might get passed
8352 into an hv routine with a regular hash.
8353 Similarly, a hash that isn't using shared hash keys has to have
8354 the flag in every key so that we know not to try to call
8355 share_hek_kek on it. */
8357 SV * const sv = newSVpvn (HEK_KEY(hek), HEK_LEN(hek));
8362 /* This will be overwhelminly the most common case. */
8364 /* Inline most of newSVpvn_share(), because share_hek_hek() is far
8365 more efficient than sharepvn(). */
8369 sv_upgrade(sv, SVt_PV);
8370 SvPV_set(sv, (char *)HEK_KEY(share_hek_hek(hek)));
8371 SvCUR_set(sv, HEK_LEN(hek));
8384 =for apidoc newSVpvn_share
8386 Creates a new SV with its SvPVX_const pointing to a shared string in the string
8387 table. If the string does not already exist in the table, it is created
8388 first. Turns on READONLY and FAKE. If the C<hash> parameter is non-zero, that
8389 value is used; otherwise the hash is computed. The string's hash can be later
8390 be retrieved from the SV with the C<SvSHARED_HASH()> macro. The idea here is
8391 that as the string table is used for shared hash keys these strings will have
8392 SvPVX_const == HeKEY and hash lookup will avoid string compare.
8398 Perl_newSVpvn_share(pTHX_ const char *src, I32 len, U32 hash)
8402 bool is_utf8 = FALSE;
8403 const char *const orig_src = src;
8406 STRLEN tmplen = -len;
8408 /* See the note in hv.c:hv_fetch() --jhi */
8409 src = (char*)bytes_from_utf8((const U8*)src, &tmplen, &is_utf8);
8413 PERL_HASH(hash, src, len);
8415 /* The logic for this is inlined in S_mro_get_linear_isa_dfs(), so if it
8416 changes here, update it there too. */
8417 sv_upgrade(sv, SVt_PV);
8418 SvPV_set(sv, sharepvn(src, is_utf8?-len:len, hash));
8426 if (src != orig_src)
8432 =for apidoc newSVpv_share
8434 Like C<newSVpvn_share>, but takes a nul-terminated string instead of a
8441 Perl_newSVpv_share(pTHX_ const char *src, U32 hash)
8443 return newSVpvn_share(src, strlen(src), hash);
8446 #if defined(PERL_IMPLICIT_CONTEXT)
8448 /* pTHX_ magic can't cope with varargs, so this is a no-context
8449 * version of the main function, (which may itself be aliased to us).
8450 * Don't access this version directly.
8454 Perl_newSVpvf_nocontext(const char *const pat, ...)
8460 PERL_ARGS_ASSERT_NEWSVPVF_NOCONTEXT;
8462 va_start(args, pat);
8463 sv = vnewSVpvf(pat, &args);
8470 =for apidoc newSVpvf
8472 Creates a new SV and initializes it with the string formatted like
8479 Perl_newSVpvf(pTHX_ const char *const pat, ...)
8484 PERL_ARGS_ASSERT_NEWSVPVF;
8486 va_start(args, pat);
8487 sv = vnewSVpvf(pat, &args);
8492 /* backend for newSVpvf() and newSVpvf_nocontext() */
8495 Perl_vnewSVpvf(pTHX_ const char *const pat, va_list *const args)
8500 PERL_ARGS_ASSERT_VNEWSVPVF;
8503 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
8510 Creates a new SV and copies a floating point value into it.
8511 The reference count for the SV is set to 1.
8517 Perl_newSVnv(pTHX_ const NV n)
8530 Creates a new SV and copies an integer into it. The reference count for the
8537 Perl_newSViv(pTHX_ const IV i)
8550 Creates a new SV and copies an unsigned integer into it.
8551 The reference count for the SV is set to 1.
8557 Perl_newSVuv(pTHX_ const UV u)
8568 =for apidoc newSV_type
8570 Creates a new SV, of the type specified. The reference count for the new SV
8577 Perl_newSV_type(pTHX_ const svtype type)
8582 sv_upgrade(sv, type);
8587 =for apidoc newRV_noinc
8589 Creates an RV wrapper for an SV. The reference count for the original
8590 SV is B<not> incremented.
8596 Perl_newRV_noinc(pTHX_ SV *const tmpRef)
8599 register SV *sv = newSV_type(SVt_IV);
8601 PERL_ARGS_ASSERT_NEWRV_NOINC;
8604 SvRV_set(sv, tmpRef);
8609 /* newRV_inc is the official function name to use now.
8610 * newRV_inc is in fact #defined to newRV in sv.h
8614 Perl_newRV(pTHX_ SV *const sv)
8618 PERL_ARGS_ASSERT_NEWRV;
8620 return newRV_noinc(SvREFCNT_inc_simple_NN(sv));
8626 Creates a new SV which is an exact duplicate of the original SV.
8633 Perl_newSVsv(pTHX_ register SV *const old)
8640 if (SvTYPE(old) == SVTYPEMASK) {
8641 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL), "semi-panic: attempt to dup freed string");
8645 /* SV_GMAGIC is the default for sv_setv()
8646 SV_NOSTEAL prevents TEMP buffers being, well, stolen, and saves games
8647 with SvTEMP_off and SvTEMP_on round a call to sv_setsv. */
8648 sv_setsv_flags(sv, old, SV_GMAGIC | SV_NOSTEAL);
8653 =for apidoc sv_reset
8655 Underlying implementation for the C<reset> Perl function.
8656 Note that the perl-level function is vaguely deprecated.
8662 Perl_sv_reset(pTHX_ register const char *s, HV *const stash)
8665 char todo[PERL_UCHAR_MAX+1];
8667 PERL_ARGS_ASSERT_SV_RESET;
8672 if (!*s) { /* reset ?? searches */
8673 MAGIC * const mg = mg_find((const SV *)stash, PERL_MAGIC_symtab);
8675 const U32 count = mg->mg_len / sizeof(PMOP**);
8676 PMOP **pmp = (PMOP**) mg->mg_ptr;
8677 PMOP *const *const end = pmp + count;
8681 SvREADONLY_off(PL_regex_pad[(*pmp)->op_pmoffset]);
8683 (*pmp)->op_pmflags &= ~PMf_USED;
8691 /* reset variables */
8693 if (!HvARRAY(stash))
8696 Zero(todo, 256, char);
8699 I32 i = (unsigned char)*s;
8703 max = (unsigned char)*s++;
8704 for ( ; i <= max; i++) {
8707 for (i = 0; i <= (I32) HvMAX(stash); i++) {
8709 for (entry = HvARRAY(stash)[i];
8711 entry = HeNEXT(entry))
8716 if (!todo[(U8)*HeKEY(entry)])
8718 gv = MUTABLE_GV(HeVAL(entry));
8721 if (SvTHINKFIRST(sv)) {
8722 if (!SvREADONLY(sv) && SvROK(sv))
8724 /* XXX Is this continue a bug? Why should THINKFIRST
8725 exempt us from resetting arrays and hashes? */
8729 if (SvTYPE(sv) >= SVt_PV) {
8731 if (SvPVX_const(sv) != NULL)
8739 if (GvHV(gv) && !HvNAME_get(GvHV(gv))) {
8741 Perl_die(aTHX_ "Can't reset %%ENV on this system");
8744 # if defined(USE_ENVIRON_ARRAY)
8747 # endif /* USE_ENVIRON_ARRAY */
8758 Using various gambits, try to get an IO from an SV: the IO slot if its a
8759 GV; or the recursive result if we're an RV; or the IO slot of the symbol
8760 named after the PV if we're a string.
8766 Perl_sv_2io(pTHX_ SV *const sv)
8771 PERL_ARGS_ASSERT_SV_2IO;
8773 switch (SvTYPE(sv)) {
8775 io = MUTABLE_IO(sv);
8779 if (isGV_with_GP(sv)) {
8780 gv = MUTABLE_GV(sv);
8783 Perl_croak(aTHX_ "Bad filehandle: %s", GvNAME(gv));
8789 Perl_croak(aTHX_ PL_no_usym, "filehandle");
8791 return sv_2io(SvRV(sv));
8792 gv = gv_fetchsv(sv, 0, SVt_PVIO);
8798 Perl_croak(aTHX_ "Bad filehandle: %"SVf, SVfARG(sv));
8807 Using various gambits, try to get a CV from an SV; in addition, try if
8808 possible to set C<*st> and C<*gvp> to the stash and GV associated with it.
8809 The flags in C<lref> are passed to gv_fetchsv.
8815 Perl_sv_2cv(pTHX_ SV *sv, HV **const st, GV **const gvp, const I32 lref)
8821 PERL_ARGS_ASSERT_SV_2CV;
8828 switch (SvTYPE(sv)) {
8832 return MUTABLE_CV(sv);
8839 if (isGV_with_GP(sv)) {
8840 gv = MUTABLE_GV(sv);
8851 sv = amagic_deref_call(sv, to_cv_amg);
8852 /* At this point I'd like to do SPAGAIN, but really I need to
8853 force it upon my callers. Hmmm. This is a mess... */
8856 if (SvTYPE(sv) == SVt_PVCV) {
8857 cv = MUTABLE_CV(sv);
8862 else if(isGV_with_GP(sv))
8863 gv = MUTABLE_GV(sv);
8865 Perl_croak(aTHX_ "Not a subroutine reference");
8867 else if (isGV_with_GP(sv)) {
8869 gv = MUTABLE_GV(sv);
8872 gv = gv_fetchsv(sv, lref, SVt_PVCV); /* Calls get magic */
8878 /* Some flags to gv_fetchsv mean don't really create the GV */
8879 if (!isGV_with_GP(gv)) {
8885 if (lref && !GvCVu(gv)) {
8889 gv_efullname3(tmpsv, gv, NULL);
8890 /* XXX this is probably not what they think they're getting.
8891 * It has the same effect as "sub name;", i.e. just a forward
8893 newSUB(start_subparse(FALSE, 0),
8894 newSVOP(OP_CONST, 0, tmpsv),
8898 Perl_croak(aTHX_ "Unable to create sub named \"%"SVf"\"",
8899 SVfARG(SvOK(sv) ? sv : &PL_sv_no));
8908 Returns true if the SV has a true value by Perl's rules.
8909 Use the C<SvTRUE> macro instead, which may call C<sv_true()> or may
8910 instead use an in-line version.
8916 Perl_sv_true(pTHX_ register SV *const sv)
8921 register const XPV* const tXpv = (XPV*)SvANY(sv);
8923 (tXpv->xpv_cur > 1 ||
8924 (tXpv->xpv_cur && *sv->sv_u.svu_pv != '0')))
8931 return SvIVX(sv) != 0;
8934 return SvNVX(sv) != 0.0;
8936 return sv_2bool(sv);
8942 =for apidoc sv_pvn_force
8944 Get a sensible string out of the SV somehow.
8945 A private implementation of the C<SvPV_force> macro for compilers which
8946 can't cope with complex macro expressions. Always use the macro instead.
8948 =for apidoc sv_pvn_force_flags
8950 Get a sensible string out of the SV somehow.
8951 If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<sv> if
8952 appropriate, else not. C<sv_pvn_force> and C<sv_pvn_force_nomg> are
8953 implemented in terms of this function.
8954 You normally want to use the various wrapper macros instead: see
8955 C<SvPV_force> and C<SvPV_force_nomg>
8961 Perl_sv_pvn_force_flags(pTHX_ SV *const sv, STRLEN *const lp, const I32 flags)
8965 PERL_ARGS_ASSERT_SV_PVN_FORCE_FLAGS;
8967 if (SvTHINKFIRST(sv) && !SvROK(sv))
8968 sv_force_normal_flags(sv, 0);
8978 if (SvREADONLY(sv) && !(flags & SV_MUTABLE_RETURN)) {
8979 const char * const ref = sv_reftype(sv,0);
8981 Perl_croak(aTHX_ "Can't coerce readonly %s to string in %s",
8982 ref, OP_DESC(PL_op));
8984 Perl_croak(aTHX_ "Can't coerce readonly %s to string", ref);
8986 if ((SvTYPE(sv) > SVt_PVLV && SvTYPE(sv) != SVt_PVFM)
8987 || isGV_with_GP(sv))
8988 Perl_croak(aTHX_ "Can't coerce %s to string in %s", sv_reftype(sv,0),
8990 s = sv_2pv_flags(sv, &len, flags);
8994 if (s != SvPVX_const(sv)) { /* Almost, but not quite, sv_setpvn() */
8997 SvUPGRADE(sv, SVt_PV); /* Never FALSE */
8998 SvGROW(sv, len + 1);
8999 Move(s,SvPVX(sv),len,char);
9001 SvPVX(sv)[len] = '\0';
9004 SvPOK_on(sv); /* validate pointer */
9006 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
9007 PTR2UV(sv),SvPVX_const(sv)));
9010 return SvPVX_mutable(sv);
9014 =for apidoc sv_pvbyten_force
9016 The backend for the C<SvPVbytex_force> macro. Always use the macro instead.
9022 Perl_sv_pvbyten_force(pTHX_ SV *const sv, STRLEN *const lp)
9024 PERL_ARGS_ASSERT_SV_PVBYTEN_FORCE;
9026 sv_pvn_force(sv,lp);
9027 sv_utf8_downgrade(sv,0);
9033 =for apidoc sv_pvutf8n_force
9035 The backend for the C<SvPVutf8x_force> macro. Always use the macro instead.
9041 Perl_sv_pvutf8n_force(pTHX_ SV *const sv, STRLEN *const lp)
9043 PERL_ARGS_ASSERT_SV_PVUTF8N_FORCE;
9045 sv_pvn_force(sv,lp);
9046 sv_utf8_upgrade(sv);
9052 =for apidoc sv_reftype
9054 Returns a string describing what the SV is a reference to.
9060 Perl_sv_reftype(pTHX_ const SV *const sv, const int ob)
9062 PERL_ARGS_ASSERT_SV_REFTYPE;
9064 /* The fact that I don't need to downcast to char * everywhere, only in ?:
9065 inside return suggests a const propagation bug in g++. */
9066 if (ob && SvOBJECT(sv)) {
9067 char * const name = HvNAME_get(SvSTASH(sv));
9068 return name ? name : (char *) "__ANON__";
9071 switch (SvTYPE(sv)) {
9086 case SVt_PVLV: return (char *) (SvROK(sv) ? "REF"
9087 /* tied lvalues should appear to be
9088 * scalars for backwards compatibility */
9089 : (LvTYPE(sv) == 't' || LvTYPE(sv) == 'T')
9090 ? "SCALAR" : "LVALUE");
9091 case SVt_PVAV: return "ARRAY";
9092 case SVt_PVHV: return "HASH";
9093 case SVt_PVCV: return "CODE";
9094 case SVt_PVGV: return (char *) (isGV_with_GP(sv)
9095 ? "GLOB" : "SCALAR");
9096 case SVt_PVFM: return "FORMAT";
9097 case SVt_PVIO: return "IO";
9098 case SVt_BIND: return "BIND";
9099 case SVt_REGEXP: return "REGEXP";
9100 default: return "UNKNOWN";
9106 =for apidoc sv_isobject
9108 Returns a boolean indicating whether the SV is an RV pointing to a blessed
9109 object. If the SV is not an RV, or if the object is not blessed, then this
9116 Perl_sv_isobject(pTHX_ SV *sv)
9132 Returns a boolean indicating whether the SV is blessed into the specified
9133 class. This does not check for subtypes; use C<sv_derived_from> to verify
9134 an inheritance relationship.
9140 Perl_sv_isa(pTHX_ SV *sv, const char *const name)
9144 PERL_ARGS_ASSERT_SV_ISA;
9154 hvname = HvNAME_get(SvSTASH(sv));
9158 return strEQ(hvname, name);
9164 Creates a new SV for the RV, C<rv>, to point to. If C<rv> is not an RV then
9165 it will be upgraded to one. If C<classname> is non-null then the new SV will
9166 be blessed in the specified package. The new SV is returned and its
9167 reference count is 1.
9173 Perl_newSVrv(pTHX_ SV *const rv, const char *const classname)
9178 PERL_ARGS_ASSERT_NEWSVRV;
9182 SV_CHECK_THINKFIRST_COW_DROP(rv);
9183 (void)SvAMAGIC_off(rv);
9185 if (SvTYPE(rv) >= SVt_PVMG) {
9186 const U32 refcnt = SvREFCNT(rv);
9190 SvREFCNT(rv) = refcnt;
9192 sv_upgrade(rv, SVt_IV);
9193 } else if (SvROK(rv)) {
9194 SvREFCNT_dec(SvRV(rv));
9196 prepare_SV_for_RV(rv);
9204 HV* const stash = gv_stashpv(classname, GV_ADD);
9205 (void)sv_bless(rv, stash);
9211 =for apidoc sv_setref_pv
9213 Copies a pointer into a new SV, optionally blessing the SV. The C<rv>
9214 argument will be upgraded to an RV. That RV will be modified to point to
9215 the new SV. If the C<pv> argument is NULL then C<PL_sv_undef> will be placed
9216 into the SV. The C<classname> argument indicates the package for the
9217 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9218 will have a reference count of 1, and the RV will be returned.
9220 Do not use with other Perl types such as HV, AV, SV, CV, because those
9221 objects will become corrupted by the pointer copy process.
9223 Note that C<sv_setref_pvn> copies the string while this copies the pointer.
9229 Perl_sv_setref_pv(pTHX_ SV *const rv, const char *const classname, void *const pv)
9233 PERL_ARGS_ASSERT_SV_SETREF_PV;
9236 sv_setsv(rv, &PL_sv_undef);
9240 sv_setiv(newSVrv(rv,classname), PTR2IV(pv));
9245 =for apidoc sv_setref_iv
9247 Copies an integer into a new SV, optionally blessing the SV. The C<rv>
9248 argument will be upgraded to an RV. That RV will be modified to point to
9249 the new SV. The C<classname> argument indicates the package for the
9250 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9251 will have a reference count of 1, and the RV will be returned.
9257 Perl_sv_setref_iv(pTHX_ SV *const rv, const char *const classname, const IV iv)
9259 PERL_ARGS_ASSERT_SV_SETREF_IV;
9261 sv_setiv(newSVrv(rv,classname), iv);
9266 =for apidoc sv_setref_uv
9268 Copies an unsigned integer into a new SV, optionally blessing the SV. The C<rv>
9269 argument will be upgraded to an RV. That RV will be modified to point to
9270 the new SV. The C<classname> argument indicates the package for the
9271 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9272 will have a reference count of 1, and the RV will be returned.
9278 Perl_sv_setref_uv(pTHX_ SV *const rv, const char *const classname, const UV uv)
9280 PERL_ARGS_ASSERT_SV_SETREF_UV;
9282 sv_setuv(newSVrv(rv,classname), uv);
9287 =for apidoc sv_setref_nv
9289 Copies a double into a new SV, optionally blessing the SV. The C<rv>
9290 argument will be upgraded to an RV. That RV will be modified to point to
9291 the new SV. The C<classname> argument indicates the package for the
9292 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9293 will have a reference count of 1, and the RV will be returned.
9299 Perl_sv_setref_nv(pTHX_ SV *const rv, const char *const classname, const NV nv)
9301 PERL_ARGS_ASSERT_SV_SETREF_NV;
9303 sv_setnv(newSVrv(rv,classname), nv);
9308 =for apidoc sv_setref_pvn
9310 Copies a string into a new SV, optionally blessing the SV. The length of the
9311 string must be specified with C<n>. The C<rv> argument will be upgraded to
9312 an RV. That RV will be modified to point to the new SV. The C<classname>
9313 argument indicates the package for the blessing. Set C<classname> to
9314 C<NULL> to avoid the blessing. The new SV will have a reference count
9315 of 1, and the RV will be returned.
9317 Note that C<sv_setref_pv> copies the pointer while this copies the string.
9323 Perl_sv_setref_pvn(pTHX_ SV *const rv, const char *const classname,
9324 const char *const pv, const STRLEN n)
9326 PERL_ARGS_ASSERT_SV_SETREF_PVN;
9328 sv_setpvn(newSVrv(rv,classname), pv, n);
9333 =for apidoc sv_bless
9335 Blesses an SV into a specified package. The SV must be an RV. The package
9336 must be designated by its stash (see C<gv_stashpv()>). The reference count
9337 of the SV is unaffected.
9343 Perl_sv_bless(pTHX_ SV *const sv, HV *const stash)
9348 PERL_ARGS_ASSERT_SV_BLESS;
9351 Perl_croak(aTHX_ "Can't bless non-reference value");
9353 if (SvFLAGS(tmpRef) & (SVs_OBJECT|SVf_READONLY)) {
9354 if (SvIsCOW(tmpRef))
9355 sv_force_normal_flags(tmpRef, 0);
9356 if (SvREADONLY(tmpRef))
9357 Perl_croak_no_modify(aTHX);
9358 if (SvOBJECT(tmpRef)) {
9359 if (SvTYPE(tmpRef) != SVt_PVIO)
9361 SvREFCNT_dec(SvSTASH(tmpRef));
9364 SvOBJECT_on(tmpRef);
9365 if (SvTYPE(tmpRef) != SVt_PVIO)
9367 SvUPGRADE(tmpRef, SVt_PVMG);
9368 SvSTASH_set(tmpRef, MUTABLE_HV(SvREFCNT_inc_simple(stash)));
9373 (void)SvAMAGIC_off(sv);
9375 if(SvSMAGICAL(tmpRef))
9376 if(mg_find(tmpRef, PERL_MAGIC_ext) || mg_find(tmpRef, PERL_MAGIC_uvar))
9384 /* Downgrades a PVGV to a PVMG. If it’s actually a PVLV, we leave the type
9385 * as it is after unglobbing it.
9389 S_sv_unglob(pTHX_ SV *const sv)
9394 SV * const temp = sv_newmortal();
9396 PERL_ARGS_ASSERT_SV_UNGLOB;
9398 assert(SvTYPE(sv) == SVt_PVGV || SvTYPE(sv) == SVt_PVLV);
9400 gv_efullname3(temp, MUTABLE_GV(sv), "*");
9403 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
9404 && HvNAME_get(stash))
9405 mro_method_changed_in(stash);
9406 gp_free(MUTABLE_GV(sv));
9409 sv_del_backref(MUTABLE_SV(GvSTASH(sv)), sv);
9413 if (GvNAME_HEK(sv)) {
9414 unshare_hek(GvNAME_HEK(sv));
9416 isGV_with_GP_off(sv);
9418 if(SvTYPE(sv) == SVt_PVGV) {
9419 /* need to keep SvANY(sv) in the right arena */
9420 xpvmg = new_XPVMG();
9421 StructCopy(SvANY(sv), xpvmg, XPVMG);
9422 del_XPVGV(SvANY(sv));
9425 SvFLAGS(sv) &= ~SVTYPEMASK;
9426 SvFLAGS(sv) |= SVt_PVMG;
9429 /* Intentionally not calling any local SET magic, as this isn't so much a
9430 set operation as merely an internal storage change. */
9431 sv_setsv_flags(sv, temp, 0);
9435 =for apidoc sv_unref_flags
9437 Unsets the RV status of the SV, and decrements the reference count of
9438 whatever was being referenced by the RV. This can almost be thought of
9439 as a reversal of C<newSVrv>. The C<cflags> argument can contain
9440 C<SV_IMMEDIATE_UNREF> to force the reference count to be decremented
9441 (otherwise the decrementing is conditional on the reference count being
9442 different from one or the reference being a readonly SV).
9449 Perl_sv_unref_flags(pTHX_ SV *const ref, const U32 flags)
9451 SV* const target = SvRV(ref);
9453 PERL_ARGS_ASSERT_SV_UNREF_FLAGS;
9455 if (SvWEAKREF(ref)) {
9456 sv_del_backref(target, ref);
9458 SvRV_set(ref, NULL);
9461 SvRV_set(ref, NULL);
9463 /* You can't have a || SvREADONLY(target) here, as $a = $$a, where $a was
9464 assigned to as BEGIN {$a = \"Foo"} will fail. */
9465 if (SvREFCNT(target) != 1 || (flags & SV_IMMEDIATE_UNREF))
9466 SvREFCNT_dec(target);
9467 else /* XXX Hack, but hard to make $a=$a->[1] work otherwise */
9468 sv_2mortal(target); /* Schedule for freeing later */
9472 =for apidoc sv_untaint
9474 Untaint an SV. Use C<SvTAINTED_off> instead.
9479 Perl_sv_untaint(pTHX_ SV *const sv)
9481 PERL_ARGS_ASSERT_SV_UNTAINT;
9483 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
9484 MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
9491 =for apidoc sv_tainted
9493 Test an SV for taintedness. Use C<SvTAINTED> instead.
9498 Perl_sv_tainted(pTHX_ SV *const sv)
9500 PERL_ARGS_ASSERT_SV_TAINTED;
9502 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
9503 const MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
9504 if (mg && (mg->mg_len & 1) )
9511 =for apidoc sv_setpviv
9513 Copies an integer into the given SV, also updating its string value.
9514 Does not handle 'set' magic. See C<sv_setpviv_mg>.
9520 Perl_sv_setpviv(pTHX_ SV *const sv, const IV iv)
9522 char buf[TYPE_CHARS(UV)];
9524 char * const ptr = uiv_2buf(buf, iv, 0, 0, &ebuf);
9526 PERL_ARGS_ASSERT_SV_SETPVIV;
9528 sv_setpvn(sv, ptr, ebuf - ptr);
9532 =for apidoc sv_setpviv_mg
9534 Like C<sv_setpviv>, but also handles 'set' magic.
9540 Perl_sv_setpviv_mg(pTHX_ SV *const sv, const IV iv)
9542 PERL_ARGS_ASSERT_SV_SETPVIV_MG;
9548 #if defined(PERL_IMPLICIT_CONTEXT)
9550 /* pTHX_ magic can't cope with varargs, so this is a no-context
9551 * version of the main function, (which may itself be aliased to us).
9552 * Don't access this version directly.
9556 Perl_sv_setpvf_nocontext(SV *const sv, const char *const pat, ...)
9561 PERL_ARGS_ASSERT_SV_SETPVF_NOCONTEXT;
9563 va_start(args, pat);
9564 sv_vsetpvf(sv, pat, &args);
9568 /* pTHX_ magic can't cope with varargs, so this is a no-context
9569 * version of the main function, (which may itself be aliased to us).
9570 * Don't access this version directly.
9574 Perl_sv_setpvf_mg_nocontext(SV *const sv, const char *const pat, ...)
9579 PERL_ARGS_ASSERT_SV_SETPVF_MG_NOCONTEXT;
9581 va_start(args, pat);
9582 sv_vsetpvf_mg(sv, pat, &args);
9588 =for apidoc sv_setpvf
9590 Works like C<sv_catpvf> but copies the text into the SV instead of
9591 appending it. Does not handle 'set' magic. See C<sv_setpvf_mg>.
9597 Perl_sv_setpvf(pTHX_ SV *const sv, const char *const pat, ...)
9601 PERL_ARGS_ASSERT_SV_SETPVF;
9603 va_start(args, pat);
9604 sv_vsetpvf(sv, pat, &args);
9609 =for apidoc sv_vsetpvf
9611 Works like C<sv_vcatpvf> but copies the text into the SV instead of
9612 appending it. Does not handle 'set' magic. See C<sv_vsetpvf_mg>.
9614 Usually used via its frontend C<sv_setpvf>.
9620 Perl_sv_vsetpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9622 PERL_ARGS_ASSERT_SV_VSETPVF;
9624 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9628 =for apidoc sv_setpvf_mg
9630 Like C<sv_setpvf>, but also handles 'set' magic.
9636 Perl_sv_setpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
9640 PERL_ARGS_ASSERT_SV_SETPVF_MG;
9642 va_start(args, pat);
9643 sv_vsetpvf_mg(sv, pat, &args);
9648 =for apidoc sv_vsetpvf_mg
9650 Like C<sv_vsetpvf>, but also handles 'set' magic.
9652 Usually used via its frontend C<sv_setpvf_mg>.
9658 Perl_sv_vsetpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9660 PERL_ARGS_ASSERT_SV_VSETPVF_MG;
9662 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9666 #if defined(PERL_IMPLICIT_CONTEXT)
9668 /* pTHX_ magic can't cope with varargs, so this is a no-context
9669 * version of the main function, (which may itself be aliased to us).
9670 * Don't access this version directly.
9674 Perl_sv_catpvf_nocontext(SV *const sv, const char *const pat, ...)
9679 PERL_ARGS_ASSERT_SV_CATPVF_NOCONTEXT;
9681 va_start(args, pat);
9682 sv_vcatpvf(sv, pat, &args);
9686 /* pTHX_ magic can't cope with varargs, so this is a no-context
9687 * version of the main function, (which may itself be aliased to us).
9688 * Don't access this version directly.
9692 Perl_sv_catpvf_mg_nocontext(SV *const sv, const char *const pat, ...)
9697 PERL_ARGS_ASSERT_SV_CATPVF_MG_NOCONTEXT;
9699 va_start(args, pat);
9700 sv_vcatpvf_mg(sv, pat, &args);
9706 =for apidoc sv_catpvf
9708 Processes its arguments like C<sprintf> and appends the formatted
9709 output to an SV. If the appended data contains "wide" characters
9710 (including, but not limited to, SVs with a UTF-8 PV formatted with %s,
9711 and characters >255 formatted with %c), the original SV might get
9712 upgraded to UTF-8. Handles 'get' magic, but not 'set' magic. See
9713 C<sv_catpvf_mg>. If the original SV was UTF-8, the pattern should be
9714 valid UTF-8; if the original SV was bytes, the pattern should be too.
9719 Perl_sv_catpvf(pTHX_ SV *const sv, const char *const pat, ...)
9723 PERL_ARGS_ASSERT_SV_CATPVF;
9725 va_start(args, pat);
9726 sv_vcatpvf(sv, pat, &args);
9731 =for apidoc sv_vcatpvf
9733 Processes its arguments like C<vsprintf> and appends the formatted output
9734 to an SV. Does not handle 'set' magic. See C<sv_vcatpvf_mg>.
9736 Usually used via its frontend C<sv_catpvf>.
9742 Perl_sv_vcatpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9744 PERL_ARGS_ASSERT_SV_VCATPVF;
9746 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9750 =for apidoc sv_catpvf_mg
9752 Like C<sv_catpvf>, but also handles 'set' magic.
9758 Perl_sv_catpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
9762 PERL_ARGS_ASSERT_SV_CATPVF_MG;
9764 va_start(args, pat);
9765 sv_vcatpvf_mg(sv, pat, &args);
9770 =for apidoc sv_vcatpvf_mg
9772 Like C<sv_vcatpvf>, but also handles 'set' magic.
9774 Usually used via its frontend C<sv_catpvf_mg>.
9780 Perl_sv_vcatpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9782 PERL_ARGS_ASSERT_SV_VCATPVF_MG;
9784 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9789 =for apidoc sv_vsetpvfn
9791 Works like C<sv_vcatpvfn> but copies the text into the SV instead of
9794 Usually used via one of its frontends C<sv_vsetpvf> and C<sv_vsetpvf_mg>.
9800 Perl_sv_vsetpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
9801 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
9803 PERL_ARGS_ASSERT_SV_VSETPVFN;
9806 sv_vcatpvfn(sv, pat, patlen, args, svargs, svmax, maybe_tainted);
9811 * Warn of missing argument to sprintf, and then return a defined value
9812 * to avoid inappropriate "use of uninit" warnings [perl #71000].
9814 #define WARN_MISSING WARN_UNINITIALIZED /* Not sure we want a new category */
9816 S_vcatpvfn_missing_argument(pTHX) {
9817 if (ckWARN(WARN_MISSING)) {
9818 Perl_warner(aTHX_ packWARN(WARN_MISSING), "Missing argument in %s",
9819 PL_op ? OP_DESC(PL_op) : "sv_vcatpvfn()");
9826 S_expect_number(pTHX_ char **const pattern)
9831 PERL_ARGS_ASSERT_EXPECT_NUMBER;
9833 switch (**pattern) {
9834 case '1': case '2': case '3':
9835 case '4': case '5': case '6':
9836 case '7': case '8': case '9':
9837 var = *(*pattern)++ - '0';
9838 while (isDIGIT(**pattern)) {
9839 const I32 tmp = var * 10 + (*(*pattern)++ - '0');
9841 Perl_croak(aTHX_ "Integer overflow in format string for %s", (PL_op ? OP_DESC(PL_op) : "sv_vcatpvfn"));
9849 S_F0convert(NV nv, char *const endbuf, STRLEN *const len)
9851 const int neg = nv < 0;
9854 PERL_ARGS_ASSERT_F0CONVERT;
9862 if (uv & 1 && uv == nv)
9863 uv--; /* Round to even */
9865 const unsigned dig = uv % 10;
9878 =for apidoc sv_vcatpvfn
9880 Processes its arguments like C<vsprintf> and appends the formatted output
9881 to an SV. Uses an array of SVs if the C style variable argument list is
9882 missing (NULL). When running with taint checks enabled, indicates via
9883 C<maybe_tainted> if results are untrustworthy (often due to the use of
9886 Usually used via one of its frontends C<sv_vcatpvf> and C<sv_vcatpvf_mg>.
9892 #define VECTORIZE_ARGS vecsv = va_arg(*args, SV*);\
9893 vecstr = (U8*)SvPV_const(vecsv,veclen);\
9894 vec_utf8 = DO_UTF8(vecsv);
9896 /* XXX maybe_tainted is never assigned to, so the doc above is lying. */
9899 Perl_sv_vcatpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
9900 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
9908 static const char nullstr[] = "(null)";
9910 bool has_utf8 = DO_UTF8(sv); /* has the result utf8? */
9911 const bool pat_utf8 = has_utf8; /* the pattern is in utf8? */
9913 /* Times 4: a decimal digit takes more than 3 binary digits.
9914 * NV_DIG: mantissa takes than many decimal digits.
9915 * Plus 32: Playing safe. */
9916 char ebuf[IV_DIG * 4 + NV_DIG + 32];
9917 /* large enough for "%#.#f" --chip */
9918 /* what about long double NVs? --jhi */
9920 PERL_ARGS_ASSERT_SV_VCATPVFN;
9921 PERL_UNUSED_ARG(maybe_tainted);
9923 /* no matter what, this is a string now */
9924 (void)SvPV_force(sv, origlen);
9926 /* special-case "", "%s", and "%-p" (SVf - see below) */
9929 if (patlen == 2 && pat[0] == '%' && pat[1] == 's') {
9931 const char * const s = va_arg(*args, char*);
9932 sv_catpv(sv, s ? s : nullstr);
9934 else if (svix < svmax) {
9935 sv_catsv(sv, *svargs);
9938 S_vcatpvfn_missing_argument(aTHX);
9941 if (args && patlen == 3 && pat[0] == '%' &&
9942 pat[1] == '-' && pat[2] == 'p') {
9943 argsv = MUTABLE_SV(va_arg(*args, void*));
9944 sv_catsv(sv, argsv);
9948 #ifndef USE_LONG_DOUBLE
9949 /* special-case "%.<number>[gf]" */
9950 if ( !args && patlen <= 5 && pat[0] == '%' && pat[1] == '.'
9951 && (pat[patlen-1] == 'g' || pat[patlen-1] == 'f') ) {
9952 unsigned digits = 0;
9956 while (*pp >= '0' && *pp <= '9')
9957 digits = 10 * digits + (*pp++ - '0');
9958 if (pp - pat == (int)patlen - 1 && svix < svmax) {
9959 const NV nv = SvNV(*svargs);
9961 /* Add check for digits != 0 because it seems that some
9962 gconverts are buggy in this case, and we don't yet have
9963 a Configure test for this. */
9964 if (digits && digits < sizeof(ebuf) - NV_DIG - 10) {
9965 /* 0, point, slack */
9966 Gconvert(nv, (int)digits, 0, ebuf);
9968 if (*ebuf) /* May return an empty string for digits==0 */
9971 } else if (!digits) {
9974 if ((p = F0convert(nv, ebuf + sizeof ebuf, &l))) {
9975 sv_catpvn(sv, p, l);
9981 #endif /* !USE_LONG_DOUBLE */
9983 if (!args && svix < svmax && DO_UTF8(*svargs))
9986 patend = (char*)pat + patlen;
9987 for (p = (char*)pat; p < patend; p = q) {
9990 bool vectorize = FALSE;
9991 bool vectorarg = FALSE;
9992 bool vec_utf8 = FALSE;
9998 bool has_precis = FALSE;
10000 const I32 osvix = svix;
10001 bool is_utf8 = FALSE; /* is this item utf8? */
10002 #ifdef HAS_LDBL_SPRINTF_BUG
10003 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
10004 with sfio - Allen <allens@cpan.org> */
10005 bool fix_ldbl_sprintf_bug = FALSE;
10009 U8 utf8buf[UTF8_MAXBYTES+1];
10010 STRLEN esignlen = 0;
10012 const char *eptr = NULL;
10013 const char *fmtstart;
10016 const U8 *vecstr = NULL;
10023 /* we need a long double target in case HAS_LONG_DOUBLE but
10024 not USE_LONG_DOUBLE
10026 #if defined(HAS_LONG_DOUBLE) && LONG_DOUBLESIZE > DOUBLESIZE
10034 const char *dotstr = ".";
10035 STRLEN dotstrlen = 1;
10036 I32 efix = 0; /* explicit format parameter index */
10037 I32 ewix = 0; /* explicit width index */
10038 I32 epix = 0; /* explicit precision index */
10039 I32 evix = 0; /* explicit vector index */
10040 bool asterisk = FALSE;
10042 /* echo everything up to the next format specification */
10043 for (q = p; q < patend && *q != '%'; ++q) ;
10045 if (has_utf8 && !pat_utf8)
10046 sv_catpvn_utf8_upgrade(sv, p, q - p, nsv);
10048 sv_catpvn(sv, p, q - p);
10057 We allow format specification elements in this order:
10058 \d+\$ explicit format parameter index
10060 v|\*(\d+\$)?v vector with optional (optionally specified) arg
10061 0 flag (as above): repeated to allow "v02"
10062 \d+|\*(\d+\$)? width using optional (optionally specified) arg
10063 \.(\d*|\*(\d+\$)?) precision using optional (optionally specified) arg
10065 [%bcdefginopsuxDFOUX] format (mandatory)
10070 As of perl5.9.3, printf format checking is on by default.
10071 Internally, perl uses %p formats to provide an escape to
10072 some extended formatting. This block deals with those
10073 extensions: if it does not match, (char*)q is reset and
10074 the normal format processing code is used.
10076 Currently defined extensions are:
10077 %p include pointer address (standard)
10078 %-p (SVf) include an SV (previously %_)
10079 %-<num>p include an SV with precision <num>
10080 %<num>p reserved for future extensions
10082 Robin Barker 2005-07-14
10084 %1p (VDf) removed. RMB 2007-10-19
10091 n = expect_number(&q);
10093 if (sv) { /* SVf */
10098 argsv = MUTABLE_SV(va_arg(*args, void*));
10099 eptr = SvPV_const(argsv, elen);
10100 if (DO_UTF8(argsv))
10105 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL),
10106 "internal %%<num>p might conflict with future printf extensions");
10112 if ( (width = expect_number(&q)) ) {
10127 if (plus == '+' && *q == ' ') /* '+' over ' ' */
10156 if ( (ewix = expect_number(&q)) )
10165 if ((vectorarg = asterisk)) {
10178 width = expect_number(&q);
10184 vecsv = va_arg(*args, SV*);
10186 vecsv = (evix > 0 && evix <= svmax)
10187 ? svargs[evix-1] : S_vcatpvfn_missing_argument(aTHX);
10189 vecsv = svix < svmax
10190 ? svargs[svix++] : S_vcatpvfn_missing_argument(aTHX);
10192 dotstr = SvPV_const(vecsv, dotstrlen);
10193 /* Keep the DO_UTF8 test *after* the SvPV call, else things go
10194 bad with tied or overloaded values that return UTF8. */
10195 if (DO_UTF8(vecsv))
10197 else if (has_utf8) {
10198 vecsv = sv_mortalcopy(vecsv);
10199 sv_utf8_upgrade(vecsv);
10200 dotstr = SvPV_const(vecsv, dotstrlen);
10207 else if (efix ? (efix > 0 && efix <= svmax) : svix < svmax) {
10208 vecsv = svargs[efix ? efix-1 : svix++];
10209 vecstr = (U8*)SvPV_const(vecsv,veclen);
10210 vec_utf8 = DO_UTF8(vecsv);
10212 /* if this is a version object, we need to convert
10213 * back into v-string notation and then let the
10214 * vectorize happen normally
10216 if (sv_derived_from(vecsv, "version")) {
10217 char *version = savesvpv(vecsv);
10218 if ( hv_exists(MUTABLE_HV(SvRV(vecsv)), "alpha", 5 ) ) {
10219 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
10220 "vector argument not supported with alpha versions");
10223 vecsv = sv_newmortal();
10224 scan_vstring(version, version + veclen, vecsv);
10225 vecstr = (U8*)SvPV_const(vecsv, veclen);
10226 vec_utf8 = DO_UTF8(vecsv);
10238 i = va_arg(*args, int);
10240 i = (ewix ? ewix <= svmax : svix < svmax) ?
10241 SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
10243 width = (i < 0) ? -i : i;
10253 if ( ((epix = expect_number(&q))) && (*q++ != '$') )
10255 /* XXX: todo, support specified precision parameter */
10259 i = va_arg(*args, int);
10261 i = (ewix ? ewix <= svmax : svix < svmax)
10262 ? SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
10264 has_precis = !(i < 0);
10268 while (isDIGIT(*q))
10269 precis = precis * 10 + (*q++ - '0');
10278 case 'I': /* Ix, I32x, and I64x */
10280 if (q[1] == '6' && q[2] == '4') {
10286 if (q[1] == '3' && q[2] == '2') {
10296 #if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
10307 #if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
10308 if (*++q == 'l') { /* lld, llf */
10317 if (*++q == 'h') { /* hhd, hhu */
10346 if (!vectorize && !args) {
10348 const I32 i = efix-1;
10349 argsv = (i >= 0 && i < svmax)
10350 ? svargs[i] : S_vcatpvfn_missing_argument(aTHX);
10352 argsv = (svix >= 0 && svix < svmax)
10353 ? svargs[svix++] : S_vcatpvfn_missing_argument(aTHX);
10357 switch (c = *q++) {
10364 uv = (args) ? va_arg(*args, int) : SvIV(argsv);
10366 (!UNI_IS_INVARIANT(uv) && SvUTF8(sv)))
10368 eptr = (char*)utf8buf;
10369 elen = uvchr_to_utf8((U8*)eptr, uv) - utf8buf;
10383 eptr = va_arg(*args, char*);
10385 elen = strlen(eptr);
10387 eptr = (char *)nullstr;
10388 elen = sizeof nullstr - 1;
10392 eptr = SvPV_const(argsv, elen);
10393 if (DO_UTF8(argsv)) {
10394 STRLEN old_precis = precis;
10395 if (has_precis && precis < elen) {
10396 STRLEN ulen = sv_len_utf8(argsv);
10397 I32 p = precis > ulen ? ulen : precis;
10398 sv_pos_u2b(argsv, &p, 0); /* sticks at end */
10401 if (width) { /* fudge width (can't fudge elen) */
10402 if (has_precis && precis < elen)
10403 width += precis - old_precis;
10405 width += elen - sv_len_utf8(argsv);
10412 if (has_precis && precis < elen)
10419 if (alt || vectorize)
10421 uv = PTR2UV(args ? va_arg(*args, void*) : argsv);
10442 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
10451 esignbuf[esignlen++] = plus;
10455 case 'c': iv = (char)va_arg(*args, int); break;
10456 case 'h': iv = (short)va_arg(*args, int); break;
10457 case 'l': iv = va_arg(*args, long); break;
10458 case 'V': iv = va_arg(*args, IV); break;
10459 case 'z': iv = va_arg(*args, SSize_t); break;
10460 case 't': iv = va_arg(*args, ptrdiff_t); break;
10461 default: iv = va_arg(*args, int); break;
10463 case 'j': iv = va_arg(*args, intmax_t); break;
10467 iv = va_arg(*args, Quad_t); break;
10474 IV tiv = SvIV(argsv); /* work around GCC bug #13488 */
10476 case 'c': iv = (char)tiv; break;
10477 case 'h': iv = (short)tiv; break;
10478 case 'l': iv = (long)tiv; break;
10480 default: iv = tiv; break;
10483 iv = (Quad_t)tiv; break;
10489 if ( !vectorize ) /* we already set uv above */
10494 esignbuf[esignlen++] = plus;
10498 esignbuf[esignlen++] = '-';
10542 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
10553 case 'c': uv = (unsigned char)va_arg(*args, unsigned); break;
10554 case 'h': uv = (unsigned short)va_arg(*args, unsigned); break;
10555 case 'l': uv = va_arg(*args, unsigned long); break;
10556 case 'V': uv = va_arg(*args, UV); break;
10557 case 'z': uv = va_arg(*args, Size_t); break;
10558 case 't': uv = va_arg(*args, ptrdiff_t); break; /* will sign extend, but there is no uptrdiff_t, so oh well */
10560 case 'j': uv = va_arg(*args, uintmax_t); break;
10562 default: uv = va_arg(*args, unsigned); break;
10565 uv = va_arg(*args, Uquad_t); break;
10572 UV tuv = SvUV(argsv); /* work around GCC bug #13488 */
10574 case 'c': uv = (unsigned char)tuv; break;
10575 case 'h': uv = (unsigned short)tuv; break;
10576 case 'l': uv = (unsigned long)tuv; break;
10578 default: uv = tuv; break;
10581 uv = (Uquad_t)tuv; break;
10590 char *ptr = ebuf + sizeof ebuf;
10591 bool tempalt = uv ? alt : FALSE; /* Vectors can't change alt */
10597 p = (char *)((c == 'X') ? PL_hexdigit + 16 : PL_hexdigit);
10601 } while (uv >>= 4);
10603 esignbuf[esignlen++] = '0';
10604 esignbuf[esignlen++] = c; /* 'x' or 'X' */
10610 *--ptr = '0' + dig;
10611 } while (uv >>= 3);
10612 if (alt && *ptr != '0')
10618 *--ptr = '0' + dig;
10619 } while (uv >>= 1);
10621 esignbuf[esignlen++] = '0';
10622 esignbuf[esignlen++] = c;
10625 default: /* it had better be ten or less */
10628 *--ptr = '0' + dig;
10629 } while (uv /= base);
10632 elen = (ebuf + sizeof ebuf) - ptr;
10636 zeros = precis - elen;
10637 else if (precis == 0 && elen == 1 && *eptr == '0'
10638 && !(base == 8 && alt)) /* "%#.0o" prints "0" */
10641 /* a precision nullifies the 0 flag. */
10648 /* FLOATING POINT */
10651 c = 'f'; /* maybe %F isn't supported here */
10653 case 'e': case 'E':
10655 case 'g': case 'G':
10659 /* This is evil, but floating point is even more evil */
10661 /* for SV-style calling, we can only get NV
10662 for C-style calling, we assume %f is double;
10663 for simplicity we allow any of %Lf, %llf, %qf for long double
10667 #if defined(USE_LONG_DOUBLE)
10671 /* [perl #20339] - we should accept and ignore %lf rather than die */
10675 #if defined(USE_LONG_DOUBLE)
10676 intsize = args ? 0 : 'q';
10680 #if defined(HAS_LONG_DOUBLE)
10693 /* now we need (long double) if intsize == 'q', else (double) */
10695 #if LONG_DOUBLESIZE > DOUBLESIZE
10697 va_arg(*args, long double) :
10698 va_arg(*args, double)
10700 va_arg(*args, double)
10705 /* nv * 0 will be NaN for NaN, +Inf and -Inf, and 0 for anything
10706 else. frexp() has some unspecified behaviour for those three */
10707 if (c != 'e' && c != 'E' && (nv * 0) == 0) {
10709 /* FIXME: if HAS_LONG_DOUBLE but not USE_LONG_DOUBLE this
10710 will cast our (long double) to (double) */
10711 (void)Perl_frexp(nv, &i);
10712 if (i == PERL_INT_MIN)
10713 Perl_die(aTHX_ "panic: frexp");
10715 need = BIT_DIGITS(i);
10717 need += has_precis ? precis : 6; /* known default */
10722 #ifdef HAS_LDBL_SPRINTF_BUG
10723 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
10724 with sfio - Allen <allens@cpan.org> */
10727 # define MY_DBL_MAX DBL_MAX
10728 # else /* XXX guessing! HUGE_VAL may be defined as infinity, so not using */
10729 # if DOUBLESIZE >= 8
10730 # define MY_DBL_MAX 1.7976931348623157E+308L
10732 # define MY_DBL_MAX 3.40282347E+38L
10736 # ifdef HAS_LDBL_SPRINTF_BUG_LESS1 /* only between -1L & 1L - Allen */
10737 # define MY_DBL_MAX_BUG 1L
10739 # define MY_DBL_MAX_BUG MY_DBL_MAX
10743 # define MY_DBL_MIN DBL_MIN
10744 # else /* XXX guessing! -Allen */
10745 # if DOUBLESIZE >= 8
10746 # define MY_DBL_MIN 2.2250738585072014E-308L
10748 # define MY_DBL_MIN 1.17549435E-38L
10752 if ((intsize == 'q') && (c == 'f') &&
10753 ((nv < MY_DBL_MAX_BUG) && (nv > -MY_DBL_MAX_BUG)) &&
10754 (need < DBL_DIG)) {
10755 /* it's going to be short enough that
10756 * long double precision is not needed */
10758 if ((nv <= 0L) && (nv >= -0L))
10759 fix_ldbl_sprintf_bug = TRUE; /* 0 is 0 - easiest */
10761 /* would use Perl_fp_class as a double-check but not
10762 * functional on IRIX - see perl.h comments */
10764 if ((nv >= MY_DBL_MIN) || (nv <= -MY_DBL_MIN)) {
10765 /* It's within the range that a double can represent */
10766 #if defined(DBL_MAX) && !defined(DBL_MIN)
10767 if ((nv >= ((long double)1/DBL_MAX)) ||
10768 (nv <= (-(long double)1/DBL_MAX)))
10770 fix_ldbl_sprintf_bug = TRUE;
10773 if (fix_ldbl_sprintf_bug == TRUE) {
10783 # undef MY_DBL_MAX_BUG
10786 #endif /* HAS_LDBL_SPRINTF_BUG */
10788 need += 20; /* fudge factor */
10789 if (PL_efloatsize < need) {
10790 Safefree(PL_efloatbuf);
10791 PL_efloatsize = need + 20; /* more fudge */
10792 Newx(PL_efloatbuf, PL_efloatsize, char);
10793 PL_efloatbuf[0] = '\0';
10796 if ( !(width || left || plus || alt) && fill != '0'
10797 && has_precis && intsize != 'q' ) { /* Shortcuts */
10798 /* See earlier comment about buggy Gconvert when digits,
10800 if ( c == 'g' && precis) {
10801 Gconvert((NV)nv, (int)precis, 0, PL_efloatbuf);
10802 /* May return an empty string for digits==0 */
10803 if (*PL_efloatbuf) {
10804 elen = strlen(PL_efloatbuf);
10805 goto float_converted;
10807 } else if ( c == 'f' && !precis) {
10808 if ((eptr = F0convert(nv, ebuf + sizeof ebuf, &elen)))
10813 char *ptr = ebuf + sizeof ebuf;
10816 /* FIXME: what to do if HAS_LONG_DOUBLE but not PERL_PRIfldbl? */
10817 #if defined(HAS_LONG_DOUBLE) && defined(PERL_PRIfldbl)
10818 if (intsize == 'q') {
10819 /* Copy the one or more characters in a long double
10820 * format before the 'base' ([efgEFG]) character to
10821 * the format string. */
10822 static char const prifldbl[] = PERL_PRIfldbl;
10823 char const *p = prifldbl + sizeof(prifldbl) - 3;
10824 while (p >= prifldbl) { *--ptr = *p--; }
10829 do { *--ptr = '0' + (base % 10); } while (base /= 10);
10834 do { *--ptr = '0' + (base % 10); } while (base /= 10);
10846 /* No taint. Otherwise we are in the strange situation
10847 * where printf() taints but print($float) doesn't.
10849 #if defined(HAS_LONG_DOUBLE)
10850 elen = ((intsize == 'q')
10851 ? my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, nv)
10852 : my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, (double)nv));
10854 elen = my_sprintf(PL_efloatbuf, ptr, nv);
10858 eptr = PL_efloatbuf;
10866 i = SvCUR(sv) - origlen;
10869 case 'c': *(va_arg(*args, char*)) = i; break;
10870 case 'h': *(va_arg(*args, short*)) = i; break;
10871 default: *(va_arg(*args, int*)) = i; break;
10872 case 'l': *(va_arg(*args, long*)) = i; break;
10873 case 'V': *(va_arg(*args, IV*)) = i; break;
10874 case 'z': *(va_arg(*args, SSize_t*)) = i; break;
10875 case 't': *(va_arg(*args, ptrdiff_t*)) = i; break;
10877 case 'j': *(va_arg(*args, intmax_t*)) = i; break;
10881 *(va_arg(*args, Quad_t*)) = i; break;
10888 sv_setuv_mg(argsv, (UV)i);
10889 continue; /* not "break" */
10896 && (PL_op->op_type == OP_PRTF || PL_op->op_type == OP_SPRINTF)
10897 && ckWARN(WARN_PRINTF))
10899 SV * const msg = sv_newmortal();
10900 Perl_sv_setpvf(aTHX_ msg, "Invalid conversion in %sprintf: ",
10901 (PL_op->op_type == OP_PRTF) ? "" : "s");
10902 if (fmtstart < patend) {
10903 const char * const fmtend = q < patend ? q : patend;
10905 sv_catpvs(msg, "\"%");
10906 for (f = fmtstart; f < fmtend; f++) {
10908 sv_catpvn(msg, f, 1);
10910 Perl_sv_catpvf(aTHX_ msg,
10911 "\\%03"UVof, (UV)*f & 0xFF);
10914 sv_catpvs(msg, "\"");
10916 sv_catpvs(msg, "end of string");
10918 Perl_warner(aTHX_ packWARN(WARN_PRINTF), "%"SVf, SVfARG(msg)); /* yes, this is reentrant */
10921 /* output mangled stuff ... */
10927 /* ... right here, because formatting flags should not apply */
10928 SvGROW(sv, SvCUR(sv) + elen + 1);
10930 Copy(eptr, p, elen, char);
10933 SvCUR_set(sv, p - SvPVX_const(sv));
10935 continue; /* not "break" */
10938 if (is_utf8 != has_utf8) {
10941 sv_utf8_upgrade(sv);
10944 const STRLEN old_elen = elen;
10945 SV * const nsv = newSVpvn_flags(eptr, elen, SVs_TEMP);
10946 sv_utf8_upgrade(nsv);
10947 eptr = SvPVX_const(nsv);
10950 if (width) { /* fudge width (can't fudge elen) */
10951 width += elen - old_elen;
10957 have = esignlen + zeros + elen;
10959 Perl_croak_nocontext("%s", PL_memory_wrap);
10961 need = (have > width ? have : width);
10964 if (need >= (((STRLEN)~0) - SvCUR(sv) - dotstrlen - 1))
10965 Perl_croak_nocontext("%s", PL_memory_wrap);
10966 SvGROW(sv, SvCUR(sv) + need + dotstrlen + 1);
10968 if (esignlen && fill == '0') {
10970 for (i = 0; i < (int)esignlen; i++)
10971 *p++ = esignbuf[i];
10973 if (gap && !left) {
10974 memset(p, fill, gap);
10977 if (esignlen && fill != '0') {
10979 for (i = 0; i < (int)esignlen; i++)
10980 *p++ = esignbuf[i];
10984 for (i = zeros; i; i--)
10988 Copy(eptr, p, elen, char);
10992 memset(p, ' ', gap);
10997 Copy(dotstr, p, dotstrlen, char);
11001 vectorize = FALSE; /* done iterating over vecstr */
11008 SvCUR_set(sv, p - SvPVX_const(sv));
11017 /* =========================================================================
11019 =head1 Cloning an interpreter
11021 All the macros and functions in this section are for the private use of
11022 the main function, perl_clone().
11024 The foo_dup() functions make an exact copy of an existing foo thingy.
11025 During the course of a cloning, a hash table is used to map old addresses
11026 to new addresses. The table is created and manipulated with the
11027 ptr_table_* functions.
11031 * =========================================================================*/
11034 #if defined(USE_ITHREADS)
11036 /* XXX Remove this so it doesn't have to go thru the macro and return for nothing */
11037 #ifndef GpREFCNT_inc
11038 # define GpREFCNT_inc(gp) ((gp) ? (++(gp)->gp_refcnt, (gp)) : (GP*)NULL)
11042 /* Certain cases in Perl_ss_dup have been merged, by relying on the fact
11043 that currently av_dup, gv_dup and hv_dup are the same as sv_dup.
11044 If this changes, please unmerge ss_dup.
11045 Likewise, sv_dup_inc_multiple() relies on this fact. */
11046 #define sv_dup_inc_NN(s,t) SvREFCNT_inc_NN(sv_dup_inc(s,t))
11047 #define av_dup(s,t) MUTABLE_AV(sv_dup((const SV *)s,t))
11048 #define av_dup_inc(s,t) MUTABLE_AV(sv_dup_inc((const SV *)s,t))
11049 #define hv_dup(s,t) MUTABLE_HV(sv_dup((const SV *)s,t))
11050 #define hv_dup_inc(s,t) MUTABLE_HV(sv_dup_inc((const SV *)s,t))
11051 #define cv_dup(s,t) MUTABLE_CV(sv_dup((const SV *)s,t))
11052 #define cv_dup_inc(s,t) MUTABLE_CV(sv_dup_inc((const SV *)s,t))
11053 #define io_dup(s,t) MUTABLE_IO(sv_dup((const SV *)s,t))
11054 #define io_dup_inc(s,t) MUTABLE_IO(sv_dup_inc((const SV *)s,t))
11055 #define gv_dup(s,t) MUTABLE_GV(sv_dup((const SV *)s,t))
11056 #define gv_dup_inc(s,t) MUTABLE_GV(sv_dup_inc((const SV *)s,t))
11057 #define SAVEPV(p) ((p) ? savepv(p) : NULL)
11058 #define SAVEPVN(p,n) ((p) ? savepvn(p,n) : NULL)
11060 /* clone a parser */
11063 Perl_parser_dup(pTHX_ const yy_parser *const proto, CLONE_PARAMS *const param)
11067 PERL_ARGS_ASSERT_PARSER_DUP;
11072 /* look for it in the table first */
11073 parser = (yy_parser *)ptr_table_fetch(PL_ptr_table, proto);
11077 /* create anew and remember what it is */
11078 Newxz(parser, 1, yy_parser);
11079 ptr_table_store(PL_ptr_table, proto, parser);
11081 /* XXX these not yet duped */
11082 parser->old_parser = NULL;
11083 parser->stack = NULL;
11085 parser->stack_size = 0;
11086 /* XXX parser->stack->state = 0; */
11088 /* XXX eventually, just Copy() most of the parser struct ? */
11090 parser->lex_brackets = proto->lex_brackets;
11091 parser->lex_casemods = proto->lex_casemods;
11092 parser->lex_brackstack = savepvn(proto->lex_brackstack,
11093 (proto->lex_brackets < 120 ? 120 : proto->lex_brackets));
11094 parser->lex_casestack = savepvn(proto->lex_casestack,
11095 (proto->lex_casemods < 12 ? 12 : proto->lex_casemods));
11096 parser->lex_defer = proto->lex_defer;
11097 parser->lex_dojoin = proto->lex_dojoin;
11098 parser->lex_expect = proto->lex_expect;
11099 parser->lex_formbrack = proto->lex_formbrack;
11100 parser->lex_inpat = proto->lex_inpat;
11101 parser->lex_inwhat = proto->lex_inwhat;
11102 parser->lex_op = proto->lex_op;
11103 parser->lex_repl = sv_dup_inc(proto->lex_repl, param);
11104 parser->lex_starts = proto->lex_starts;
11105 parser->lex_stuff = sv_dup_inc(proto->lex_stuff, param);
11106 parser->multi_close = proto->multi_close;
11107 parser->multi_open = proto->multi_open;
11108 parser->multi_start = proto->multi_start;
11109 parser->multi_end = proto->multi_end;
11110 parser->pending_ident = proto->pending_ident;
11111 parser->preambled = proto->preambled;
11112 parser->sublex_info = proto->sublex_info; /* XXX not quite right */
11113 parser->linestr = sv_dup_inc(proto->linestr, param);
11114 parser->expect = proto->expect;
11115 parser->copline = proto->copline;
11116 parser->last_lop_op = proto->last_lop_op;
11117 parser->lex_state = proto->lex_state;
11118 parser->rsfp = fp_dup(proto->rsfp, '<', param);
11119 /* rsfp_filters entries have fake IoDIRP() */
11120 parser->rsfp_filters= av_dup_inc(proto->rsfp_filters, param);
11121 parser->in_my = proto->in_my;
11122 parser->in_my_stash = hv_dup(proto->in_my_stash, param);
11123 parser->error_count = proto->error_count;
11126 parser->linestr = sv_dup_inc(proto->linestr, param);
11129 char * const ols = SvPVX(proto->linestr);
11130 char * const ls = SvPVX(parser->linestr);
11132 parser->bufptr = ls + (proto->bufptr >= ols ?
11133 proto->bufptr - ols : 0);
11134 parser->oldbufptr = ls + (proto->oldbufptr >= ols ?
11135 proto->oldbufptr - ols : 0);
11136 parser->oldoldbufptr= ls + (proto->oldoldbufptr >= ols ?
11137 proto->oldoldbufptr - ols : 0);
11138 parser->linestart = ls + (proto->linestart >= ols ?
11139 proto->linestart - ols : 0);
11140 parser->last_uni = ls + (proto->last_uni >= ols ?
11141 proto->last_uni - ols : 0);
11142 parser->last_lop = ls + (proto->last_lop >= ols ?
11143 proto->last_lop - ols : 0);
11145 parser->bufend = ls + SvCUR(parser->linestr);
11148 Copy(proto->tokenbuf, parser->tokenbuf, 256, char);
11152 parser->endwhite = proto->endwhite;
11153 parser->faketokens = proto->faketokens;
11154 parser->lasttoke = proto->lasttoke;
11155 parser->nextwhite = proto->nextwhite;
11156 parser->realtokenstart = proto->realtokenstart;
11157 parser->skipwhite = proto->skipwhite;
11158 parser->thisclose = proto->thisclose;
11159 parser->thismad = proto->thismad;
11160 parser->thisopen = proto->thisopen;
11161 parser->thisstuff = proto->thisstuff;
11162 parser->thistoken = proto->thistoken;
11163 parser->thiswhite = proto->thiswhite;
11165 Copy(proto->nexttoke, parser->nexttoke, 5, NEXTTOKE);
11166 parser->curforce = proto->curforce;
11168 Copy(proto->nextval, parser->nextval, 5, YYSTYPE);
11169 Copy(proto->nexttype, parser->nexttype, 5, I32);
11170 parser->nexttoke = proto->nexttoke;
11173 /* XXX should clone saved_curcop here, but we aren't passed
11174 * proto_perl; so do it in perl_clone_using instead */
11180 /* duplicate a file handle */
11183 Perl_fp_dup(pTHX_ PerlIO *const fp, const char type, CLONE_PARAMS *const param)
11187 PERL_ARGS_ASSERT_FP_DUP;
11188 PERL_UNUSED_ARG(type);
11191 return (PerlIO*)NULL;
11193 /* look for it in the table first */
11194 ret = (PerlIO*)ptr_table_fetch(PL_ptr_table, fp);
11198 /* create anew and remember what it is */
11199 ret = PerlIO_fdupopen(aTHX_ fp, param, PERLIO_DUP_CLONE);
11200 ptr_table_store(PL_ptr_table, fp, ret);
11204 /* duplicate a directory handle */
11207 Perl_dirp_dup(pTHX_ DIR *const dp, CLONE_PARAMS *const param)
11213 register const Direntry_t *dirent;
11214 char smallbuf[256];
11220 PERL_UNUSED_CONTEXT;
11221 PERL_ARGS_ASSERT_DIRP_DUP;
11226 /* look for it in the table first */
11227 ret = (DIR*)ptr_table_fetch(PL_ptr_table, dp);
11233 PERL_UNUSED_ARG(param);
11237 /* open the current directory (so we can switch back) */
11238 if (!(pwd = PerlDir_open("."))) return (DIR *)NULL;
11240 /* chdir to our dir handle and open the present working directory */
11241 if (fchdir(my_dirfd(dp)) < 0 || !(ret = PerlDir_open("."))) {
11242 PerlDir_close(pwd);
11243 return (DIR *)NULL;
11245 /* Now we should have two dir handles pointing to the same dir. */
11247 /* Be nice to the calling code and chdir back to where we were. */
11248 fchdir(my_dirfd(pwd)); /* If this fails, then what? */
11250 /* We have no need of the pwd handle any more. */
11251 PerlDir_close(pwd);
11254 # define d_namlen(d) (d)->d_namlen
11256 # define d_namlen(d) strlen((d)->d_name)
11258 /* Iterate once through dp, to get the file name at the current posi-
11259 tion. Then step back. */
11260 pos = PerlDir_tell(dp);
11261 if ((dirent = PerlDir_read(dp))) {
11262 len = d_namlen(dirent);
11263 if (len <= sizeof smallbuf) name = smallbuf;
11264 else Newx(name, len, char);
11265 Move(dirent->d_name, name, len, char);
11267 PerlDir_seek(dp, pos);
11269 /* Iterate through the new dir handle, till we find a file with the
11271 if (!dirent) /* just before the end */
11273 pos = PerlDir_tell(ret);
11274 if (PerlDir_read(ret)) continue; /* not there yet */
11275 PerlDir_seek(ret, pos); /* step back */
11279 const long pos0 = PerlDir_tell(ret);
11281 pos = PerlDir_tell(ret);
11282 if ((dirent = PerlDir_read(ret))) {
11283 if (len == d_namlen(dirent)
11284 && memEQ(name, dirent->d_name, len)) {
11286 PerlDir_seek(ret, pos); /* step back */
11289 /* else we are not there yet; keep iterating */
11291 else { /* This is not meant to happen. The best we can do is
11292 reset the iterator to the beginning. */
11293 PerlDir_seek(ret, pos0);
11300 if (name && name != smallbuf)
11305 ret = win32_dirp_dup(dp, param);
11308 /* pop it in the pointer table */
11310 ptr_table_store(PL_ptr_table, dp, ret);
11315 /* duplicate a typeglob */
11318 Perl_gp_dup(pTHX_ GP *const gp, CLONE_PARAMS *const param)
11322 PERL_ARGS_ASSERT_GP_DUP;
11326 /* look for it in the table first */
11327 ret = (GP*)ptr_table_fetch(PL_ptr_table, gp);
11331 /* create anew and remember what it is */
11333 ptr_table_store(PL_ptr_table, gp, ret);
11336 /* ret->gp_refcnt must be 0 before any other dups are called. We're relying
11337 on Newxz() to do this for us. */
11338 ret->gp_sv = sv_dup_inc(gp->gp_sv, param);
11339 ret->gp_io = io_dup_inc(gp->gp_io, param);
11340 ret->gp_form = cv_dup_inc(gp->gp_form, param);
11341 ret->gp_av = av_dup_inc(gp->gp_av, param);
11342 ret->gp_hv = hv_dup_inc(gp->gp_hv, param);
11343 ret->gp_egv = gv_dup(gp->gp_egv, param);/* GvEGV is not refcounted */
11344 ret->gp_cv = cv_dup_inc(gp->gp_cv, param);
11345 ret->gp_cvgen = gp->gp_cvgen;
11346 ret->gp_line = gp->gp_line;
11347 ret->gp_file_hek = hek_dup(gp->gp_file_hek, param);
11351 /* duplicate a chain of magic */
11354 Perl_mg_dup(pTHX_ MAGIC *mg, CLONE_PARAMS *const param)
11356 MAGIC *mgret = NULL;
11357 MAGIC **mgprev_p = &mgret;
11359 PERL_ARGS_ASSERT_MG_DUP;
11361 for (; mg; mg = mg->mg_moremagic) {
11364 if ((param->flags & CLONEf_JOIN_IN)
11365 && mg->mg_type == PERL_MAGIC_backref)
11366 /* when joining, we let the individual SVs add themselves to
11367 * backref as needed. */
11370 Newx(nmg, 1, MAGIC);
11372 mgprev_p = &(nmg->mg_moremagic);
11374 /* There was a comment "XXX copy dynamic vtable?" but as we don't have
11375 dynamic vtables, I'm not sure why Sarathy wrote it. The comment dates
11376 from the original commit adding Perl_mg_dup() - revision 4538.
11377 Similarly there is the annotation "XXX random ptr?" next to the
11378 assignment to nmg->mg_ptr. */
11381 /* FIXME for plugins
11382 if (nmg->mg_type == PERL_MAGIC_qr) {
11383 nmg->mg_obj = MUTABLE_SV(CALLREGDUPE((REGEXP*)nmg->mg_obj, param));
11387 nmg->mg_obj = (nmg->mg_flags & MGf_REFCOUNTED)
11388 ? nmg->mg_type == PERL_MAGIC_backref
11389 /* The backref AV has its reference
11390 * count deliberately bumped by 1 */
11391 ? SvREFCNT_inc(av_dup_inc((const AV *)
11392 nmg->mg_obj, param))
11393 : sv_dup_inc(nmg->mg_obj, param)
11394 : sv_dup(nmg->mg_obj, param);
11396 if (nmg->mg_ptr && nmg->mg_type != PERL_MAGIC_regex_global) {
11397 if (nmg->mg_len > 0) {
11398 nmg->mg_ptr = SAVEPVN(nmg->mg_ptr, nmg->mg_len);
11399 if (nmg->mg_type == PERL_MAGIC_overload_table &&
11400 AMT_AMAGIC((AMT*)nmg->mg_ptr))
11402 AMT * const namtp = (AMT*)nmg->mg_ptr;
11403 sv_dup_inc_multiple((SV**)(namtp->table),
11404 (SV**)(namtp->table), NofAMmeth, param);
11407 else if (nmg->mg_len == HEf_SVKEY)
11408 nmg->mg_ptr = (char*)sv_dup_inc((const SV *)nmg->mg_ptr, param);
11410 if ((nmg->mg_flags & MGf_DUP) && nmg->mg_virtual && nmg->mg_virtual->svt_dup) {
11411 nmg->mg_virtual->svt_dup(aTHX_ nmg, param);
11417 #endif /* USE_ITHREADS */
11419 struct ptr_tbl_arena {
11420 struct ptr_tbl_arena *next;
11421 struct ptr_tbl_ent array[1023/3]; /* as ptr_tbl_ent has 3 pointers. */
11424 /* create a new pointer-mapping table */
11427 Perl_ptr_table_new(pTHX)
11430 PERL_UNUSED_CONTEXT;
11432 Newx(tbl, 1, PTR_TBL_t);
11433 tbl->tbl_max = 511;
11434 tbl->tbl_items = 0;
11435 tbl->tbl_arena = NULL;
11436 tbl->tbl_arena_next = NULL;
11437 tbl->tbl_arena_end = NULL;
11438 Newxz(tbl->tbl_ary, tbl->tbl_max + 1, PTR_TBL_ENT_t*);
11442 #define PTR_TABLE_HASH(ptr) \
11443 ((PTR2UV(ptr) >> 3) ^ (PTR2UV(ptr) >> (3 + 7)) ^ (PTR2UV(ptr) >> (3 + 17)))
11445 /* map an existing pointer using a table */
11447 STATIC PTR_TBL_ENT_t *
11448 S_ptr_table_find(PTR_TBL_t *const tbl, const void *const sv)
11450 PTR_TBL_ENT_t *tblent;
11451 const UV hash = PTR_TABLE_HASH(sv);
11453 PERL_ARGS_ASSERT_PTR_TABLE_FIND;
11455 tblent = tbl->tbl_ary[hash & tbl->tbl_max];
11456 for (; tblent; tblent = tblent->next) {
11457 if (tblent->oldval == sv)
11464 Perl_ptr_table_fetch(pTHX_ PTR_TBL_t *const tbl, const void *const sv)
11466 PTR_TBL_ENT_t const *const tblent = ptr_table_find(tbl, sv);
11468 PERL_ARGS_ASSERT_PTR_TABLE_FETCH;
11469 PERL_UNUSED_CONTEXT;
11471 return tblent ? tblent->newval : NULL;
11474 /* add a new entry to a pointer-mapping table */
11477 Perl_ptr_table_store(pTHX_ PTR_TBL_t *const tbl, const void *const oldsv, void *const newsv)
11479 PTR_TBL_ENT_t *tblent = ptr_table_find(tbl, oldsv);
11481 PERL_ARGS_ASSERT_PTR_TABLE_STORE;
11482 PERL_UNUSED_CONTEXT;
11485 tblent->newval = newsv;
11487 const UV entry = PTR_TABLE_HASH(oldsv) & tbl->tbl_max;
11489 if (tbl->tbl_arena_next == tbl->tbl_arena_end) {
11490 struct ptr_tbl_arena *new_arena;
11492 Newx(new_arena, 1, struct ptr_tbl_arena);
11493 new_arena->next = tbl->tbl_arena;
11494 tbl->tbl_arena = new_arena;
11495 tbl->tbl_arena_next = new_arena->array;
11496 tbl->tbl_arena_end = new_arena->array
11497 + sizeof(new_arena->array) / sizeof(new_arena->array[0]);
11500 tblent = tbl->tbl_arena_next++;
11502 tblent->oldval = oldsv;
11503 tblent->newval = newsv;
11504 tblent->next = tbl->tbl_ary[entry];
11505 tbl->tbl_ary[entry] = tblent;
11507 if (tblent->next && tbl->tbl_items > tbl->tbl_max)
11508 ptr_table_split(tbl);
11512 /* double the hash bucket size of an existing ptr table */
11515 Perl_ptr_table_split(pTHX_ PTR_TBL_t *const tbl)
11517 PTR_TBL_ENT_t **ary = tbl->tbl_ary;
11518 const UV oldsize = tbl->tbl_max + 1;
11519 UV newsize = oldsize * 2;
11522 PERL_ARGS_ASSERT_PTR_TABLE_SPLIT;
11523 PERL_UNUSED_CONTEXT;
11525 Renew(ary, newsize, PTR_TBL_ENT_t*);
11526 Zero(&ary[oldsize], newsize-oldsize, PTR_TBL_ENT_t*);
11527 tbl->tbl_max = --newsize;
11528 tbl->tbl_ary = ary;
11529 for (i=0; i < oldsize; i++, ary++) {
11530 PTR_TBL_ENT_t **entp = ary;
11531 PTR_TBL_ENT_t *ent = *ary;
11532 PTR_TBL_ENT_t **curentp;
11535 curentp = ary + oldsize;
11537 if ((newsize & PTR_TABLE_HASH(ent->oldval)) != i) {
11539 ent->next = *curentp;
11549 /* remove all the entries from a ptr table */
11550 /* Deprecated - will be removed post 5.14 */
11553 Perl_ptr_table_clear(pTHX_ PTR_TBL_t *const tbl)
11555 if (tbl && tbl->tbl_items) {
11556 struct ptr_tbl_arena *arena = tbl->tbl_arena;
11558 Zero(tbl->tbl_ary, tbl->tbl_max + 1, struct ptr_tbl_ent **);
11561 struct ptr_tbl_arena *next = arena->next;
11567 tbl->tbl_items = 0;
11568 tbl->tbl_arena = NULL;
11569 tbl->tbl_arena_next = NULL;
11570 tbl->tbl_arena_end = NULL;
11574 /* clear and free a ptr table */
11577 Perl_ptr_table_free(pTHX_ PTR_TBL_t *const tbl)
11579 struct ptr_tbl_arena *arena;
11585 arena = tbl->tbl_arena;
11588 struct ptr_tbl_arena *next = arena->next;
11594 Safefree(tbl->tbl_ary);
11598 #if defined(USE_ITHREADS)
11601 Perl_rvpv_dup(pTHX_ SV *const dstr, const SV *const sstr, CLONE_PARAMS *const param)
11603 PERL_ARGS_ASSERT_RVPV_DUP;
11606 if (SvWEAKREF(sstr)) {
11607 SvRV_set(dstr, sv_dup(SvRV_const(sstr), param));
11608 if (param->flags & CLONEf_JOIN_IN) {
11609 /* if joining, we add any back references individually rather
11610 * than copying the whole backref array */
11611 Perl_sv_add_backref(aTHX_ SvRV(dstr), dstr);
11615 SvRV_set(dstr, sv_dup_inc(SvRV_const(sstr), param));
11617 else if (SvPVX_const(sstr)) {
11618 /* Has something there */
11620 /* Normal PV - clone whole allocated space */
11621 SvPV_set(dstr, SAVEPVN(SvPVX_const(sstr), SvLEN(sstr)-1));
11622 if (SvREADONLY(sstr) && SvFAKE(sstr)) {
11623 /* Not that normal - actually sstr is copy on write.
11624 But we are a true, independent SV, so: */
11625 SvREADONLY_off(dstr);
11630 /* Special case - not normally malloced for some reason */
11631 if (isGV_with_GP(sstr)) {
11632 /* Don't need to do anything here. */
11634 else if ((SvREADONLY(sstr) && SvFAKE(sstr))) {
11635 /* A "shared" PV - clone it as "shared" PV */
11637 HEK_KEY(hek_dup(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)),
11641 /* Some other special case - random pointer */
11642 SvPV_set(dstr, (char *) SvPVX_const(sstr));
11647 /* Copy the NULL */
11648 SvPV_set(dstr, NULL);
11652 /* duplicate a list of SVs. source and dest may point to the same memory. */
11654 S_sv_dup_inc_multiple(pTHX_ SV *const *source, SV **dest,
11655 SSize_t items, CLONE_PARAMS *const param)
11657 PERL_ARGS_ASSERT_SV_DUP_INC_MULTIPLE;
11659 while (items-- > 0) {
11660 *dest++ = sv_dup_inc(*source++, param);
11666 /* duplicate an SV of any type (including AV, HV etc) */
11669 S_sv_dup_common(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
11674 PERL_ARGS_ASSERT_SV_DUP_COMMON;
11676 if (SvTYPE(sstr) == SVTYPEMASK) {
11677 #ifdef DEBUG_LEAKING_SCALARS_ABORT
11682 /* look for it in the table first */
11683 dstr = MUTABLE_SV(ptr_table_fetch(PL_ptr_table, sstr));
11687 if(param->flags & CLONEf_JOIN_IN) {
11688 /** We are joining here so we don't want do clone
11689 something that is bad **/
11690 if (SvTYPE(sstr) == SVt_PVHV) {
11691 const HEK * const hvname = HvNAME_HEK(sstr);
11693 /** don't clone stashes if they already exist **/
11694 dstr = MUTABLE_SV(gv_stashpvn(HEK_KEY(hvname), HEK_LEN(hvname), 0));
11695 ptr_table_store(PL_ptr_table, sstr, dstr);
11701 /* create anew and remember what it is */
11704 #ifdef DEBUG_LEAKING_SCALARS
11705 dstr->sv_debug_optype = sstr->sv_debug_optype;
11706 dstr->sv_debug_line = sstr->sv_debug_line;
11707 dstr->sv_debug_inpad = sstr->sv_debug_inpad;
11708 dstr->sv_debug_parent = (SV*)sstr;
11709 FREE_SV_DEBUG_FILE(dstr);
11710 dstr->sv_debug_file = savepv(sstr->sv_debug_file);
11713 ptr_table_store(PL_ptr_table, sstr, dstr);
11716 SvFLAGS(dstr) = SvFLAGS(sstr);
11717 SvFLAGS(dstr) &= ~SVf_OOK; /* don't propagate OOK hack */
11718 SvREFCNT(dstr) = 0; /* must be before any other dups! */
11721 if (SvANY(sstr) && PL_watch_pvx && SvPVX_const(sstr) == PL_watch_pvx)
11722 PerlIO_printf(Perl_debug_log, "watch at %p hit, found string \"%s\"\n",
11723 (void*)PL_watch_pvx, SvPVX_const(sstr));
11726 /* don't clone objects whose class has asked us not to */
11727 if (SvOBJECT(sstr) && ! (SvFLAGS(SvSTASH(sstr)) & SVphv_CLONEABLE)) {
11732 switch (SvTYPE(sstr)) {
11734 SvANY(dstr) = NULL;
11737 SvANY(dstr) = (XPVIV*)((char*)&(dstr->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
11739 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
11741 SvIV_set(dstr, SvIVX(sstr));
11745 SvANY(dstr) = new_XNV();
11746 SvNV_set(dstr, SvNVX(sstr));
11748 /* case SVt_BIND: */
11751 /* These are all the types that need complex bodies allocating. */
11753 const svtype sv_type = SvTYPE(sstr);
11754 const struct body_details *const sv_type_details
11755 = bodies_by_type + sv_type;
11759 Perl_croak(aTHX_ "Bizarre SvTYPE [%" IVdf "]", (IV)SvTYPE(sstr));
11774 assert(sv_type_details->body_size);
11775 if (sv_type_details->arena) {
11776 new_body_inline(new_body, sv_type);
11778 = (void*)((char*)new_body - sv_type_details->offset);
11780 new_body = new_NOARENA(sv_type_details);
11784 SvANY(dstr) = new_body;
11787 Copy(((char*)SvANY(sstr)) + sv_type_details->offset,
11788 ((char*)SvANY(dstr)) + sv_type_details->offset,
11789 sv_type_details->copy, char);
11791 Copy(((char*)SvANY(sstr)),
11792 ((char*)SvANY(dstr)),
11793 sv_type_details->body_size + sv_type_details->offset, char);
11796 if (sv_type != SVt_PVAV && sv_type != SVt_PVHV
11797 && !isGV_with_GP(dstr)
11798 && !(sv_type == SVt_PVIO && !(IoFLAGS(dstr) & IOf_FAKE_DIRP)))
11799 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
11801 /* The Copy above means that all the source (unduplicated) pointers
11802 are now in the destination. We can check the flags and the
11803 pointers in either, but it's possible that there's less cache
11804 missing by always going for the destination.
11805 FIXME - instrument and check that assumption */
11806 if (sv_type >= SVt_PVMG) {
11807 if ((sv_type == SVt_PVMG) && SvPAD_OUR(dstr)) {
11808 SvOURSTASH_set(dstr, hv_dup_inc(SvOURSTASH(dstr), param));
11809 } else if (SvMAGIC(dstr))
11810 SvMAGIC_set(dstr, mg_dup(SvMAGIC(dstr), param));
11812 SvSTASH_set(dstr, hv_dup_inc(SvSTASH(dstr), param));
11815 /* The cast silences a GCC warning about unhandled types. */
11816 switch ((int)sv_type) {
11826 /* FIXME for plugins */
11827 re_dup_guts((REGEXP*) sstr, (REGEXP*) dstr, param);
11830 /* XXX LvTARGOFF sometimes holds PMOP* when DEBUGGING */
11831 if (LvTYPE(dstr) == 't') /* for tie: unrefcnted fake (SV**) */
11832 LvTARG(dstr) = dstr;
11833 else if (LvTYPE(dstr) == 'T') /* for tie: fake HE */
11834 LvTARG(dstr) = MUTABLE_SV(he_dup((HE*)LvTARG(dstr), 0, param));
11836 LvTARG(dstr) = sv_dup_inc(LvTARG(dstr), param);
11838 /* non-GP case already handled above */
11839 if(isGV_with_GP(sstr)) {
11840 GvNAME_HEK(dstr) = hek_dup(GvNAME_HEK(dstr), param);
11841 /* Don't call sv_add_backref here as it's going to be
11842 created as part of the magic cloning of the symbol
11843 table--unless this is during a join and the stash
11844 is not actually being cloned. */
11845 /* Danger Will Robinson - GvGP(dstr) isn't initialised
11846 at the point of this comment. */
11847 GvSTASH(dstr) = hv_dup(GvSTASH(dstr), param);
11848 if (param->flags & CLONEf_JOIN_IN)
11849 Perl_sv_add_backref(aTHX_ MUTABLE_SV(GvSTASH(dstr)), dstr);
11850 GvGP_set(dstr, gp_dup(GvGP(sstr), param));
11851 (void)GpREFCNT_inc(GvGP(dstr));
11855 /* PL_parser->rsfp_filters entries have fake IoDIRP() */
11856 if(IoFLAGS(dstr) & IOf_FAKE_DIRP) {
11857 /* I have no idea why fake dirp (rsfps)
11858 should be treated differently but otherwise
11859 we end up with leaks -- sky*/
11860 IoTOP_GV(dstr) = gv_dup_inc(IoTOP_GV(dstr), param);
11861 IoFMT_GV(dstr) = gv_dup_inc(IoFMT_GV(dstr), param);
11862 IoBOTTOM_GV(dstr) = gv_dup_inc(IoBOTTOM_GV(dstr), param);
11864 IoTOP_GV(dstr) = gv_dup(IoTOP_GV(dstr), param);
11865 IoFMT_GV(dstr) = gv_dup(IoFMT_GV(dstr), param);
11866 IoBOTTOM_GV(dstr) = gv_dup(IoBOTTOM_GV(dstr), param);
11867 if (IoDIRP(dstr)) {
11868 IoDIRP(dstr) = dirp_dup(IoDIRP(dstr), param);
11871 /* IoDIRP(dstr) is already a copy of IoDIRP(sstr) */
11873 IoIFP(dstr) = fp_dup(IoIFP(sstr), IoTYPE(dstr), param);
11875 if (IoOFP(dstr) == IoIFP(sstr))
11876 IoOFP(dstr) = IoIFP(dstr);
11878 IoOFP(dstr) = fp_dup(IoOFP(dstr), IoTYPE(dstr), param);
11879 IoTOP_NAME(dstr) = SAVEPV(IoTOP_NAME(dstr));
11880 IoFMT_NAME(dstr) = SAVEPV(IoFMT_NAME(dstr));
11881 IoBOTTOM_NAME(dstr) = SAVEPV(IoBOTTOM_NAME(dstr));
11884 /* avoid cloning an empty array */
11885 if (AvARRAY((const AV *)sstr) && AvFILLp((const AV *)sstr) >= 0) {
11886 SV **dst_ary, **src_ary;
11887 SSize_t items = AvFILLp((const AV *)sstr) + 1;
11889 src_ary = AvARRAY((const AV *)sstr);
11890 Newxz(dst_ary, AvMAX((const AV *)sstr)+1, SV*);
11891 ptr_table_store(PL_ptr_table, src_ary, dst_ary);
11892 AvARRAY(MUTABLE_AV(dstr)) = dst_ary;
11893 AvALLOC((const AV *)dstr) = dst_ary;
11894 if (AvREAL((const AV *)sstr)) {
11895 dst_ary = sv_dup_inc_multiple(src_ary, dst_ary, items,
11899 while (items-- > 0)
11900 *dst_ary++ = sv_dup(*src_ary++, param);
11902 items = AvMAX((const AV *)sstr) - AvFILLp((const AV *)sstr);
11903 while (items-- > 0) {
11904 *dst_ary++ = &PL_sv_undef;
11908 AvARRAY(MUTABLE_AV(dstr)) = NULL;
11909 AvALLOC((const AV *)dstr) = (SV**)NULL;
11910 AvMAX( (const AV *)dstr) = -1;
11911 AvFILLp((const AV *)dstr) = -1;
11915 if (HvARRAY((const HV *)sstr)) {
11917 const bool sharekeys = !!HvSHAREKEYS(sstr);
11918 XPVHV * const dxhv = (XPVHV*)SvANY(dstr);
11919 XPVHV * const sxhv = (XPVHV*)SvANY(sstr);
11921 Newx(darray, PERL_HV_ARRAY_ALLOC_BYTES(dxhv->xhv_max+1)
11922 + (SvOOK(sstr) ? sizeof(struct xpvhv_aux) : 0),
11924 HvARRAY(dstr) = (HE**)darray;
11925 while (i <= sxhv->xhv_max) {
11926 const HE * const source = HvARRAY(sstr)[i];
11927 HvARRAY(dstr)[i] = source
11928 ? he_dup(source, sharekeys, param) : 0;
11932 const struct xpvhv_aux * const saux = HvAUX(sstr);
11933 struct xpvhv_aux * const daux = HvAUX(dstr);
11934 /* This flag isn't copied. */
11935 /* SvOOK_on(hv) attacks the IV flags. */
11936 SvFLAGS(dstr) |= SVf_OOK;
11938 if (saux->xhv_name_count) {
11939 HEK ** const sname = saux->xhv_name_u.xhvnameu_names;
11941 = saux->xhv_name_count < 0
11942 ? -saux->xhv_name_count
11943 : saux->xhv_name_count;
11944 HEK **shekp = sname + count;
11946 Newx(daux->xhv_name_u.xhvnameu_names, count, HEK *);
11947 dhekp = daux->xhv_name_u.xhvnameu_names + count;
11948 while (shekp-- > sname) {
11950 *dhekp = hek_dup(*shekp, param);
11954 daux->xhv_name_u.xhvnameu_name
11955 = hek_dup(saux->xhv_name_u.xhvnameu_name,
11958 daux->xhv_name_count = saux->xhv_name_count;
11960 daux->xhv_riter = saux->xhv_riter;
11961 daux->xhv_eiter = saux->xhv_eiter
11962 ? he_dup(saux->xhv_eiter,
11963 cBOOL(HvSHAREKEYS(sstr)), param) : 0;
11964 /* backref array needs refcnt=2; see sv_add_backref */
11965 daux->xhv_backreferences =
11966 (param->flags & CLONEf_JOIN_IN)
11967 /* when joining, we let the individual GVs and
11968 * CVs add themselves to backref as
11969 * needed. This avoids pulling in stuff
11970 * that isn't required, and simplifies the
11971 * case where stashes aren't cloned back
11972 * if they already exist in the parent
11975 : saux->xhv_backreferences
11976 ? (SvTYPE(saux->xhv_backreferences) == SVt_PVAV)
11977 ? MUTABLE_AV(SvREFCNT_inc(
11978 sv_dup_inc((const SV *)
11979 saux->xhv_backreferences, param)))
11980 : MUTABLE_AV(sv_dup((const SV *)
11981 saux->xhv_backreferences, param))
11984 daux->xhv_mro_meta = saux->xhv_mro_meta
11985 ? mro_meta_dup(saux->xhv_mro_meta, param)
11988 /* Record stashes for possible cloning in Perl_clone(). */
11990 av_push(param->stashes, dstr);
11994 HvARRAY(MUTABLE_HV(dstr)) = NULL;
11997 if (!(param->flags & CLONEf_COPY_STACKS)) {
12002 /* NOTE: not refcounted */
12003 SvANY(MUTABLE_CV(dstr))->xcv_stash =
12004 hv_dup(CvSTASH(dstr), param);
12005 if ((param->flags & CLONEf_JOIN_IN) && CvSTASH(dstr))
12006 Perl_sv_add_backref(aTHX_ MUTABLE_SV(CvSTASH(dstr)), dstr);
12007 if (!CvISXSUB(dstr)) {
12009 CvROOT(dstr) = OpREFCNT_inc(CvROOT(dstr));
12011 CvFILE(dstr) = SAVEPV(CvFILE(dstr));
12012 } else if (CvCONST(dstr)) {
12013 CvXSUBANY(dstr).any_ptr =
12014 sv_dup_inc((const SV *)CvXSUBANY(dstr).any_ptr, param);
12016 /* don't dup if copying back - CvGV isn't refcounted, so the
12017 * duped GV may never be freed. A bit of a hack! DAPM */
12018 SvANY(MUTABLE_CV(dstr))->xcv_gv =
12020 ? gv_dup_inc(CvGV(sstr), param)
12021 : (param->flags & CLONEf_JOIN_IN)
12023 : gv_dup(CvGV(sstr), param);
12025 CvPADLIST(dstr) = padlist_dup(CvPADLIST(sstr), param);
12027 CvWEAKOUTSIDE(sstr)
12028 ? cv_dup( CvOUTSIDE(dstr), param)
12029 : cv_dup_inc(CvOUTSIDE(dstr), param);
12035 if (SvOBJECT(dstr) && SvTYPE(dstr) != SVt_PVIO)
12042 Perl_sv_dup_inc(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
12044 PERL_ARGS_ASSERT_SV_DUP_INC;
12045 return sstr ? SvREFCNT_inc(sv_dup_common(sstr, param)) : NULL;
12049 Perl_sv_dup(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
12051 SV *dstr = sstr ? sv_dup_common(sstr, param) : NULL;
12052 PERL_ARGS_ASSERT_SV_DUP;
12054 /* Track every SV that (at least initially) had a reference count of 0.
12055 We need to do this by holding an actual reference to it in this array.
12056 If we attempt to cheat, turn AvREAL_off(), and store only pointers
12057 (akin to the stashes hash, and the perl stack), we come unstuck if
12058 a weak reference (or other SV legitimately SvREFCNT() == 0 for this
12059 thread) is manipulated in a CLONE method, because CLONE runs before the
12060 unreferenced array is walked to find SVs still with SvREFCNT() == 0
12061 (and fix things up by giving each a reference via the temps stack).
12062 Instead, during CLONE, if the 0-referenced SV has SvREFCNT_inc() and
12063 then SvREFCNT_dec(), it will be cleaned up (and added to the free list)
12064 before the walk of unreferenced happens and a reference to that is SV
12065 added to the temps stack. At which point we have the same SV considered
12066 to be in use, and free to be re-used. Not good.
12068 if (dstr && !(param->flags & CLONEf_COPY_STACKS) && !SvREFCNT(dstr)) {
12069 assert(param->unreferenced);
12070 av_push(param->unreferenced, SvREFCNT_inc(dstr));
12076 /* duplicate a context */
12079 Perl_cx_dup(pTHX_ PERL_CONTEXT *cxs, I32 ix, I32 max, CLONE_PARAMS* param)
12081 PERL_CONTEXT *ncxs;
12083 PERL_ARGS_ASSERT_CX_DUP;
12086 return (PERL_CONTEXT*)NULL;
12088 /* look for it in the table first */
12089 ncxs = (PERL_CONTEXT*)ptr_table_fetch(PL_ptr_table, cxs);
12093 /* create anew and remember what it is */
12094 Newx(ncxs, max + 1, PERL_CONTEXT);
12095 ptr_table_store(PL_ptr_table, cxs, ncxs);
12096 Copy(cxs, ncxs, max + 1, PERL_CONTEXT);
12099 PERL_CONTEXT * const ncx = &ncxs[ix];
12100 if (CxTYPE(ncx) == CXt_SUBST) {
12101 Perl_croak(aTHX_ "Cloning substitution context is unimplemented");
12104 switch (CxTYPE(ncx)) {
12106 ncx->blk_sub.cv = (ncx->blk_sub.olddepth == 0
12107 ? cv_dup_inc(ncx->blk_sub.cv, param)
12108 : cv_dup(ncx->blk_sub.cv,param));
12109 ncx->blk_sub.argarray = (CxHASARGS(ncx)
12110 ? av_dup_inc(ncx->blk_sub.argarray,
12113 ncx->blk_sub.savearray = av_dup_inc(ncx->blk_sub.savearray,
12115 ncx->blk_sub.oldcomppad = (PAD*)ptr_table_fetch(PL_ptr_table,
12116 ncx->blk_sub.oldcomppad);
12119 ncx->blk_eval.old_namesv = sv_dup_inc(ncx->blk_eval.old_namesv,
12121 ncx->blk_eval.cur_text = sv_dup(ncx->blk_eval.cur_text, param);
12123 case CXt_LOOP_LAZYSV:
12124 ncx->blk_loop.state_u.lazysv.end
12125 = sv_dup_inc(ncx->blk_loop.state_u.lazysv.end, param);
12126 /* We are taking advantage of av_dup_inc and sv_dup_inc
12127 actually being the same function, and order equivalence of
12129 We can assert the later [but only at run time :-(] */
12130 assert ((void *) &ncx->blk_loop.state_u.ary.ary ==
12131 (void *) &ncx->blk_loop.state_u.lazysv.cur);
12133 ncx->blk_loop.state_u.ary.ary
12134 = av_dup_inc(ncx->blk_loop.state_u.ary.ary, param);
12135 case CXt_LOOP_LAZYIV:
12136 case CXt_LOOP_PLAIN:
12137 if (CxPADLOOP(ncx)) {
12138 ncx->blk_loop.itervar_u.oldcomppad
12139 = (PAD*)ptr_table_fetch(PL_ptr_table,
12140 ncx->blk_loop.itervar_u.oldcomppad);
12142 ncx->blk_loop.itervar_u.gv
12143 = gv_dup((const GV *)ncx->blk_loop.itervar_u.gv,
12148 ncx->blk_format.cv = cv_dup(ncx->blk_format.cv, param);
12149 ncx->blk_format.gv = gv_dup(ncx->blk_format.gv, param);
12150 ncx->blk_format.dfoutgv = gv_dup_inc(ncx->blk_format.dfoutgv,
12163 /* duplicate a stack info structure */
12166 Perl_si_dup(pTHX_ PERL_SI *si, CLONE_PARAMS* param)
12170 PERL_ARGS_ASSERT_SI_DUP;
12173 return (PERL_SI*)NULL;
12175 /* look for it in the table first */
12176 nsi = (PERL_SI*)ptr_table_fetch(PL_ptr_table, si);
12180 /* create anew and remember what it is */
12181 Newxz(nsi, 1, PERL_SI);
12182 ptr_table_store(PL_ptr_table, si, nsi);
12184 nsi->si_stack = av_dup_inc(si->si_stack, param);
12185 nsi->si_cxix = si->si_cxix;
12186 nsi->si_cxmax = si->si_cxmax;
12187 nsi->si_cxstack = cx_dup(si->si_cxstack, si->si_cxix, si->si_cxmax, param);
12188 nsi->si_type = si->si_type;
12189 nsi->si_prev = si_dup(si->si_prev, param);
12190 nsi->si_next = si_dup(si->si_next, param);
12191 nsi->si_markoff = si->si_markoff;
12196 #define POPINT(ss,ix) ((ss)[--(ix)].any_i32)
12197 #define TOPINT(ss,ix) ((ss)[ix].any_i32)
12198 #define POPLONG(ss,ix) ((ss)[--(ix)].any_long)
12199 #define TOPLONG(ss,ix) ((ss)[ix].any_long)
12200 #define POPIV(ss,ix) ((ss)[--(ix)].any_iv)
12201 #define TOPIV(ss,ix) ((ss)[ix].any_iv)
12202 #define POPUV(ss,ix) ((ss)[--(ix)].any_uv)
12203 #define TOPUV(ss,ix) ((ss)[ix].any_uv)
12204 #define POPBOOL(ss,ix) ((ss)[--(ix)].any_bool)
12205 #define TOPBOOL(ss,ix) ((ss)[ix].any_bool)
12206 #define POPPTR(ss,ix) ((ss)[--(ix)].any_ptr)
12207 #define TOPPTR(ss,ix) ((ss)[ix].any_ptr)
12208 #define POPDPTR(ss,ix) ((ss)[--(ix)].any_dptr)
12209 #define TOPDPTR(ss,ix) ((ss)[ix].any_dptr)
12210 #define POPDXPTR(ss,ix) ((ss)[--(ix)].any_dxptr)
12211 #define TOPDXPTR(ss,ix) ((ss)[ix].any_dxptr)
12214 #define pv_dup_inc(p) SAVEPV(p)
12215 #define pv_dup(p) SAVEPV(p)
12216 #define svp_dup_inc(p,pp) any_dup(p,pp)
12218 /* map any object to the new equivent - either something in the
12219 * ptr table, or something in the interpreter structure
12223 Perl_any_dup(pTHX_ void *v, const PerlInterpreter *proto_perl)
12227 PERL_ARGS_ASSERT_ANY_DUP;
12230 return (void*)NULL;
12232 /* look for it in the table first */
12233 ret = ptr_table_fetch(PL_ptr_table, v);
12237 /* see if it is part of the interpreter structure */
12238 if (v >= (void*)proto_perl && v < (void*)(proto_perl+1))
12239 ret = (void*)(((char*)aTHX) + (((char*)v) - (char*)proto_perl));
12247 /* duplicate the save stack */
12250 Perl_ss_dup(pTHX_ PerlInterpreter *proto_perl, CLONE_PARAMS* param)
12253 ANY * const ss = proto_perl->Isavestack;
12254 const I32 max = proto_perl->Isavestack_max;
12255 I32 ix = proto_perl->Isavestack_ix;
12268 void (*dptr) (void*);
12269 void (*dxptr) (pTHX_ void*);
12271 PERL_ARGS_ASSERT_SS_DUP;
12273 Newxz(nss, max, ANY);
12276 const UV uv = POPUV(ss,ix);
12277 const U8 type = (U8)uv & SAVE_MASK;
12279 TOPUV(nss,ix) = uv;
12281 case SAVEt_CLEARSV:
12283 case SAVEt_HELEM: /* hash element */
12284 sv = (const SV *)POPPTR(ss,ix);
12285 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12287 case SAVEt_ITEM: /* normal string */
12288 case SAVEt_GVSV: /* scalar slot in GV */
12289 case SAVEt_SV: /* scalar reference */
12290 sv = (const SV *)POPPTR(ss,ix);
12291 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12294 case SAVEt_MORTALIZESV:
12295 sv = (const SV *)POPPTR(ss,ix);
12296 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12298 case SAVEt_SHARED_PVREF: /* char* in shared space */
12299 c = (char*)POPPTR(ss,ix);
12300 TOPPTR(nss,ix) = savesharedpv(c);
12301 ptr = POPPTR(ss,ix);
12302 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12304 case SAVEt_GENERIC_SVREF: /* generic sv */
12305 case SAVEt_SVREF: /* scalar reference */
12306 sv = (const SV *)POPPTR(ss,ix);
12307 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12308 ptr = POPPTR(ss,ix);
12309 TOPPTR(nss,ix) = svp_dup_inc((SV**)ptr, proto_perl);/* XXXXX */
12311 case SAVEt_HV: /* hash reference */
12312 case SAVEt_AV: /* array reference */
12313 sv = (const SV *) POPPTR(ss,ix);
12314 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12316 case SAVEt_COMPPAD:
12318 sv = (const SV *) POPPTR(ss,ix);
12319 TOPPTR(nss,ix) = sv_dup(sv, param);
12321 case SAVEt_INT: /* int reference */
12322 ptr = POPPTR(ss,ix);
12323 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12324 intval = (int)POPINT(ss,ix);
12325 TOPINT(nss,ix) = intval;
12327 case SAVEt_LONG: /* long reference */
12328 ptr = POPPTR(ss,ix);
12329 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12330 longval = (long)POPLONG(ss,ix);
12331 TOPLONG(nss,ix) = longval;
12333 case SAVEt_I32: /* I32 reference */
12334 case SAVEt_COP_ARYBASE: /* call CopARYBASE_set */
12335 ptr = POPPTR(ss,ix);
12336 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12338 TOPINT(nss,ix) = i;
12340 case SAVEt_IV: /* IV reference */
12341 ptr = POPPTR(ss,ix);
12342 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12344 TOPIV(nss,ix) = iv;
12346 case SAVEt_HPTR: /* HV* reference */
12347 case SAVEt_APTR: /* AV* reference */
12348 case SAVEt_SPTR: /* SV* reference */
12349 ptr = POPPTR(ss,ix);
12350 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12351 sv = (const SV *)POPPTR(ss,ix);
12352 TOPPTR(nss,ix) = sv_dup(sv, param);
12354 case SAVEt_VPTR: /* random* reference */
12355 ptr = POPPTR(ss,ix);
12356 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12358 case SAVEt_INT_SMALL:
12359 case SAVEt_I32_SMALL:
12360 case SAVEt_I16: /* I16 reference */
12361 case SAVEt_I8: /* I8 reference */
12363 ptr = POPPTR(ss,ix);
12364 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12366 case SAVEt_GENERIC_PVREF: /* generic char* */
12367 case SAVEt_PPTR: /* char* reference */
12368 ptr = POPPTR(ss,ix);
12369 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12370 c = (char*)POPPTR(ss,ix);
12371 TOPPTR(nss,ix) = pv_dup(c);
12373 case SAVEt_GP: /* scalar reference */
12374 gp = (GP*)POPPTR(ss,ix);
12375 TOPPTR(nss,ix) = gp = gp_dup(gp, param);
12376 (void)GpREFCNT_inc(gp);
12377 gv = (const GV *)POPPTR(ss,ix);
12378 TOPPTR(nss,ix) = gv_dup_inc(gv, param);
12381 ptr = POPPTR(ss,ix);
12382 if (ptr && (((OP*)ptr)->op_private & OPpREFCOUNTED)) {
12383 /* these are assumed to be refcounted properly */
12385 switch (((OP*)ptr)->op_type) {
12387 case OP_LEAVESUBLV:
12391 case OP_LEAVEWRITE:
12392 TOPPTR(nss,ix) = ptr;
12395 (void) OpREFCNT_inc(o);
12399 TOPPTR(nss,ix) = NULL;
12404 TOPPTR(nss,ix) = NULL;
12406 case SAVEt_FREECOPHH:
12407 ptr = POPPTR(ss,ix);
12408 TOPPTR(nss,ix) = cophh_copy((COPHH *)ptr);
12411 hv = (const HV *)POPPTR(ss,ix);
12412 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
12414 TOPINT(nss,ix) = i;
12417 c = (char*)POPPTR(ss,ix);
12418 TOPPTR(nss,ix) = pv_dup_inc(c);
12420 case SAVEt_STACK_POS: /* Position on Perl stack */
12422 TOPINT(nss,ix) = i;
12424 case SAVEt_DESTRUCTOR:
12425 ptr = POPPTR(ss,ix);
12426 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
12427 dptr = POPDPTR(ss,ix);
12428 TOPDPTR(nss,ix) = DPTR2FPTR(void (*)(void*),
12429 any_dup(FPTR2DPTR(void *, dptr),
12432 case SAVEt_DESTRUCTOR_X:
12433 ptr = POPPTR(ss,ix);
12434 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
12435 dxptr = POPDXPTR(ss,ix);
12436 TOPDXPTR(nss,ix) = DPTR2FPTR(void (*)(pTHX_ void*),
12437 any_dup(FPTR2DPTR(void *, dxptr),
12440 case SAVEt_REGCONTEXT:
12442 ix -= uv >> SAVE_TIGHT_SHIFT;
12444 case SAVEt_AELEM: /* array element */
12445 sv = (const SV *)POPPTR(ss,ix);
12446 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12448 TOPINT(nss,ix) = i;
12449 av = (const AV *)POPPTR(ss,ix);
12450 TOPPTR(nss,ix) = av_dup_inc(av, param);
12453 ptr = POPPTR(ss,ix);
12454 TOPPTR(nss,ix) = ptr;
12457 ptr = POPPTR(ss,ix);
12458 ptr = cophh_copy((COPHH*)ptr);
12459 TOPPTR(nss,ix) = ptr;
12461 TOPINT(nss,ix) = i;
12462 if (i & HINT_LOCALIZE_HH) {
12463 hv = (const HV *)POPPTR(ss,ix);
12464 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
12467 case SAVEt_PADSV_AND_MORTALIZE:
12468 longval = (long)POPLONG(ss,ix);
12469 TOPLONG(nss,ix) = longval;
12470 ptr = POPPTR(ss,ix);
12471 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12472 sv = (const SV *)POPPTR(ss,ix);
12473 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12475 case SAVEt_SET_SVFLAGS:
12477 TOPINT(nss,ix) = i;
12479 TOPINT(nss,ix) = i;
12480 sv = (const SV *)POPPTR(ss,ix);
12481 TOPPTR(nss,ix) = sv_dup(sv, param);
12483 case SAVEt_RE_STATE:
12485 const struct re_save_state *const old_state
12486 = (struct re_save_state *)
12487 (ss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
12488 struct re_save_state *const new_state
12489 = (struct re_save_state *)
12490 (nss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
12492 Copy(old_state, new_state, 1, struct re_save_state);
12493 ix -= SAVESTACK_ALLOC_FOR_RE_SAVE_STATE;
12495 new_state->re_state_bostr
12496 = pv_dup(old_state->re_state_bostr);
12497 new_state->re_state_reginput
12498 = pv_dup(old_state->re_state_reginput);
12499 new_state->re_state_regeol
12500 = pv_dup(old_state->re_state_regeol);
12501 new_state->re_state_regoffs
12502 = (regexp_paren_pair*)
12503 any_dup(old_state->re_state_regoffs, proto_perl);
12504 new_state->re_state_reglastparen
12505 = (U32*) any_dup(old_state->re_state_reglastparen,
12507 new_state->re_state_reglastcloseparen
12508 = (U32*)any_dup(old_state->re_state_reglastcloseparen,
12510 /* XXX This just has to be broken. The old save_re_context
12511 code did SAVEGENERICPV(PL_reg_start_tmp);
12512 PL_reg_start_tmp is char **.
12513 Look above to what the dup code does for
12514 SAVEt_GENERIC_PVREF
12515 It can never have worked.
12516 So this is merely a faithful copy of the exiting bug: */
12517 new_state->re_state_reg_start_tmp
12518 = (char **) pv_dup((char *)
12519 old_state->re_state_reg_start_tmp);
12520 /* I assume that it only ever "worked" because no-one called
12521 (pseudo)fork while the regexp engine had re-entered itself.
12523 #ifdef PERL_OLD_COPY_ON_WRITE
12524 new_state->re_state_nrs
12525 = sv_dup(old_state->re_state_nrs, param);
12527 new_state->re_state_reg_magic
12528 = (MAGIC*) any_dup(old_state->re_state_reg_magic,
12530 new_state->re_state_reg_oldcurpm
12531 = (PMOP*) any_dup(old_state->re_state_reg_oldcurpm,
12533 new_state->re_state_reg_curpm
12534 = (PMOP*) any_dup(old_state->re_state_reg_curpm,
12536 new_state->re_state_reg_oldsaved
12537 = pv_dup(old_state->re_state_reg_oldsaved);
12538 new_state->re_state_reg_poscache
12539 = pv_dup(old_state->re_state_reg_poscache);
12540 new_state->re_state_reg_starttry
12541 = pv_dup(old_state->re_state_reg_starttry);
12544 case SAVEt_COMPILE_WARNINGS:
12545 ptr = POPPTR(ss,ix);
12546 TOPPTR(nss,ix) = DUP_WARNINGS((STRLEN*)ptr);
12549 ptr = POPPTR(ss,ix);
12550 TOPPTR(nss,ix) = parser_dup((const yy_parser*)ptr, param);
12554 "panic: ss_dup inconsistency (%"IVdf")", (IV) type);
12562 /* if sv is a stash, call $class->CLONE_SKIP(), and set the SVphv_CLONEABLE
12563 * flag to the result. This is done for each stash before cloning starts,
12564 * so we know which stashes want their objects cloned */
12567 do_mark_cloneable_stash(pTHX_ SV *const sv)
12569 const HEK * const hvname = HvNAME_HEK((const HV *)sv);
12571 GV* const cloner = gv_fetchmethod_autoload(MUTABLE_HV(sv), "CLONE_SKIP", 0);
12572 SvFLAGS(sv) |= SVphv_CLONEABLE; /* clone objects by default */
12573 if (cloner && GvCV(cloner)) {
12580 mXPUSHs(newSVhek(hvname));
12582 call_sv(MUTABLE_SV(GvCV(cloner)), G_SCALAR);
12589 SvFLAGS(sv) &= ~SVphv_CLONEABLE;
12597 =for apidoc perl_clone
12599 Create and return a new interpreter by cloning the current one.
12601 perl_clone takes these flags as parameters:
12603 CLONEf_COPY_STACKS - is used to, well, copy the stacks also,
12604 without it we only clone the data and zero the stacks,
12605 with it we copy the stacks and the new perl interpreter is
12606 ready to run at the exact same point as the previous one.
12607 The pseudo-fork code uses COPY_STACKS while the
12608 threads->create doesn't.
12610 CLONEf_KEEP_PTR_TABLE
12611 perl_clone keeps a ptr_table with the pointer of the old
12612 variable as a key and the new variable as a value,
12613 this allows it to check if something has been cloned and not
12614 clone it again but rather just use the value and increase the
12615 refcount. If KEEP_PTR_TABLE is not set then perl_clone will kill
12616 the ptr_table using the function
12617 C<ptr_table_free(PL_ptr_table); PL_ptr_table = NULL;>,
12618 reason to keep it around is if you want to dup some of your own
12619 variable who are outside the graph perl scans, example of this
12620 code is in threads.xs create
12623 This is a win32 thing, it is ignored on unix, it tells perls
12624 win32host code (which is c++) to clone itself, this is needed on
12625 win32 if you want to run two threads at the same time,
12626 if you just want to do some stuff in a separate perl interpreter
12627 and then throw it away and return to the original one,
12628 you don't need to do anything.
12633 /* XXX the above needs expanding by someone who actually understands it ! */
12634 EXTERN_C PerlInterpreter *
12635 perl_clone_host(PerlInterpreter* proto_perl, UV flags);
12638 perl_clone(PerlInterpreter *proto_perl, UV flags)
12641 #ifdef PERL_IMPLICIT_SYS
12643 PERL_ARGS_ASSERT_PERL_CLONE;
12645 /* perlhost.h so we need to call into it
12646 to clone the host, CPerlHost should have a c interface, sky */
12648 if (flags & CLONEf_CLONE_HOST) {
12649 return perl_clone_host(proto_perl,flags);
12651 return perl_clone_using(proto_perl, flags,
12653 proto_perl->IMemShared,
12654 proto_perl->IMemParse,
12656 proto_perl->IStdIO,
12660 proto_perl->IProc);
12664 perl_clone_using(PerlInterpreter *proto_perl, UV flags,
12665 struct IPerlMem* ipM, struct IPerlMem* ipMS,
12666 struct IPerlMem* ipMP, struct IPerlEnv* ipE,
12667 struct IPerlStdIO* ipStd, struct IPerlLIO* ipLIO,
12668 struct IPerlDir* ipD, struct IPerlSock* ipS,
12669 struct IPerlProc* ipP)
12671 /* XXX many of the string copies here can be optimized if they're
12672 * constants; they need to be allocated as common memory and just
12673 * their pointers copied. */
12676 CLONE_PARAMS clone_params;
12677 CLONE_PARAMS* const param = &clone_params;
12679 PerlInterpreter * const my_perl = (PerlInterpreter*)(*ipM->pMalloc)(ipM, sizeof(PerlInterpreter));
12681 PERL_ARGS_ASSERT_PERL_CLONE_USING;
12682 #else /* !PERL_IMPLICIT_SYS */
12684 CLONE_PARAMS clone_params;
12685 CLONE_PARAMS* param = &clone_params;
12686 PerlInterpreter * const my_perl = (PerlInterpreter*)PerlMem_malloc(sizeof(PerlInterpreter));
12688 PERL_ARGS_ASSERT_PERL_CLONE;
12689 #endif /* PERL_IMPLICIT_SYS */
12691 /* for each stash, determine whether its objects should be cloned */
12692 S_visit(proto_perl, do_mark_cloneable_stash, SVt_PVHV, SVTYPEMASK);
12693 PERL_SET_THX(my_perl);
12696 PoisonNew(my_perl, 1, PerlInterpreter);
12701 PL_scopestack_name = 0;
12703 PL_savestack_ix = 0;
12704 PL_savestack_max = -1;
12705 PL_sig_pending = 0;
12707 Zero(&PL_debug_pad, 1, struct perl_debug_pad);
12708 # ifdef DEBUG_LEAKING_SCALARS
12709 PL_sv_serial = (((UV)my_perl >> 2) & 0xfff) * 1000000;
12711 #else /* !DEBUGGING */
12712 Zero(my_perl, 1, PerlInterpreter);
12713 #endif /* DEBUGGING */
12715 #ifdef PERL_IMPLICIT_SYS
12716 /* host pointers */
12718 PL_MemShared = ipMS;
12719 PL_MemParse = ipMP;
12726 #endif /* PERL_IMPLICIT_SYS */
12728 param->flags = flags;
12729 /* Nothing in the core code uses this, but we make it available to
12730 extensions (using mg_dup). */
12731 param->proto_perl = proto_perl;
12732 /* Likely nothing will use this, but it is initialised to be consistent
12733 with Perl_clone_params_new(). */
12734 param->new_perl = my_perl;
12735 param->unreferenced = NULL;
12737 INIT_TRACK_MEMPOOL(my_perl->Imemory_debug_header, my_perl);
12739 PL_body_arenas = NULL;
12740 Zero(&PL_body_roots, 1, PL_body_roots);
12743 PL_sv_objcount = 0;
12745 PL_sv_arenaroot = NULL;
12747 PL_debug = proto_perl->Idebug;
12749 PL_hash_seed = proto_perl->Ihash_seed;
12750 PL_rehash_seed = proto_perl->Irehash_seed;
12752 #ifdef USE_REENTRANT_API
12753 /* XXX: things like -Dm will segfault here in perlio, but doing
12754 * PERL_SET_CONTEXT(proto_perl);
12755 * breaks too many other things
12757 Perl_reentrant_init(aTHX);
12760 /* create SV map for pointer relocation */
12761 PL_ptr_table = ptr_table_new();
12763 /* initialize these special pointers as early as possible */
12764 SvANY(&PL_sv_undef) = NULL;
12765 SvREFCNT(&PL_sv_undef) = (~(U32)0)/2;
12766 SvFLAGS(&PL_sv_undef) = SVf_READONLY|SVt_NULL;
12767 ptr_table_store(PL_ptr_table, &proto_perl->Isv_undef, &PL_sv_undef);
12769 SvANY(&PL_sv_no) = new_XPVNV();
12770 SvREFCNT(&PL_sv_no) = (~(U32)0)/2;
12771 SvFLAGS(&PL_sv_no) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
12772 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
12773 SvPV_set(&PL_sv_no, savepvn(PL_No, 0));
12774 SvCUR_set(&PL_sv_no, 0);
12775 SvLEN_set(&PL_sv_no, 1);
12776 SvIV_set(&PL_sv_no, 0);
12777 SvNV_set(&PL_sv_no, 0);
12778 ptr_table_store(PL_ptr_table, &proto_perl->Isv_no, &PL_sv_no);
12780 SvANY(&PL_sv_yes) = new_XPVNV();
12781 SvREFCNT(&PL_sv_yes) = (~(U32)0)/2;
12782 SvFLAGS(&PL_sv_yes) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
12783 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
12784 SvPV_set(&PL_sv_yes, savepvn(PL_Yes, 1));
12785 SvCUR_set(&PL_sv_yes, 1);
12786 SvLEN_set(&PL_sv_yes, 2);
12787 SvIV_set(&PL_sv_yes, 1);
12788 SvNV_set(&PL_sv_yes, 1);
12789 ptr_table_store(PL_ptr_table, &proto_perl->Isv_yes, &PL_sv_yes);
12791 /* dbargs array probably holds garbage */
12794 /* create (a non-shared!) shared string table */
12795 PL_strtab = newHV();
12796 HvSHAREKEYS_off(PL_strtab);
12797 hv_ksplit(PL_strtab, HvTOTALKEYS(proto_perl->Istrtab));
12798 ptr_table_store(PL_ptr_table, proto_perl->Istrtab, PL_strtab);
12800 PL_compiling = proto_perl->Icompiling;
12802 /* These two PVs will be free'd special way so must set them same way op.c does */
12803 PL_compiling.cop_stashpv = savesharedpv(PL_compiling.cop_stashpv);
12804 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_stashpv, PL_compiling.cop_stashpv);
12806 PL_compiling.cop_file = savesharedpv(PL_compiling.cop_file);
12807 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_file, PL_compiling.cop_file);
12809 ptr_table_store(PL_ptr_table, &proto_perl->Icompiling, &PL_compiling);
12810 PL_compiling.cop_warnings = DUP_WARNINGS(PL_compiling.cop_warnings);
12811 CopHINTHASH_set(&PL_compiling, cophh_copy(CopHINTHASH_get(&PL_compiling)));
12812 PL_curcop = (COP*)any_dup(proto_perl->Icurcop, proto_perl);
12813 #ifdef PERL_DEBUG_READONLY_OPS
12818 /* pseudo environmental stuff */
12819 PL_origargc = proto_perl->Iorigargc;
12820 PL_origargv = proto_perl->Iorigargv;
12822 param->stashes = newAV(); /* Setup array of objects to call clone on */
12823 /* This makes no difference to the implementation, as it always pushes
12824 and shifts pointers to other SVs without changing their reference
12825 count, with the array becoming empty before it is freed. However, it
12826 makes it conceptually clear what is going on, and will avoid some
12827 work inside av.c, filling slots between AvFILL() and AvMAX() with
12828 &PL_sv_undef, and SvREFCNT_dec()ing those. */
12829 AvREAL_off(param->stashes);
12831 if (!(flags & CLONEf_COPY_STACKS)) {
12832 param->unreferenced = newAV();
12835 /* Set tainting stuff before PerlIO_debug can possibly get called */
12836 PL_tainting = proto_perl->Itainting;
12837 PL_taint_warn = proto_perl->Itaint_warn;
12839 #ifdef PERLIO_LAYERS
12840 /* Clone PerlIO tables as soon as we can handle general xx_dup() */
12841 PerlIO_clone(aTHX_ proto_perl, param);
12844 PL_envgv = gv_dup(proto_perl->Ienvgv, param);
12845 PL_incgv = gv_dup(proto_perl->Iincgv, param);
12846 PL_hintgv = gv_dup(proto_perl->Ihintgv, param);
12847 PL_origfilename = SAVEPV(proto_perl->Iorigfilename);
12848 PL_diehook = sv_dup_inc(proto_perl->Idiehook, param);
12849 PL_warnhook = sv_dup_inc(proto_perl->Iwarnhook, param);
12852 PL_minus_c = proto_perl->Iminus_c;
12853 PL_patchlevel = sv_dup_inc(proto_perl->Ipatchlevel, param);
12854 PL_apiversion = sv_dup_inc(proto_perl->Iapiversion, param);
12855 PL_localpatches = proto_perl->Ilocalpatches;
12856 PL_splitstr = proto_perl->Isplitstr;
12857 PL_minus_n = proto_perl->Iminus_n;
12858 PL_minus_p = proto_perl->Iminus_p;
12859 PL_minus_l = proto_perl->Iminus_l;
12860 PL_minus_a = proto_perl->Iminus_a;
12861 PL_minus_E = proto_perl->Iminus_E;
12862 PL_minus_F = proto_perl->Iminus_F;
12863 PL_doswitches = proto_perl->Idoswitches;
12864 PL_dowarn = proto_perl->Idowarn;
12865 PL_sawampersand = proto_perl->Isawampersand;
12866 PL_unsafe = proto_perl->Iunsafe;
12867 PL_inplace = SAVEPV(proto_perl->Iinplace);
12868 PL_e_script = sv_dup_inc(proto_perl->Ie_script, param);
12869 PL_perldb = proto_perl->Iperldb;
12870 PL_perl_destruct_level = proto_perl->Iperl_destruct_level;
12871 PL_exit_flags = proto_perl->Iexit_flags;
12873 /* magical thingies */
12874 /* XXX time(&PL_basetime) when asked for? */
12875 PL_basetime = proto_perl->Ibasetime;
12876 PL_formfeed = sv_dup(proto_perl->Iformfeed, param);
12878 PL_maxsysfd = proto_perl->Imaxsysfd;
12879 PL_statusvalue = proto_perl->Istatusvalue;
12881 PL_statusvalue_vms = proto_perl->Istatusvalue_vms;
12883 PL_statusvalue_posix = proto_perl->Istatusvalue_posix;
12885 PL_encoding = sv_dup(proto_perl->Iencoding, param);
12887 sv_setpvs(PERL_DEBUG_PAD(0), ""); /* For regex debugging. */
12888 sv_setpvs(PERL_DEBUG_PAD(1), ""); /* ext/re needs these */
12889 sv_setpvs(PERL_DEBUG_PAD(2), ""); /* even without DEBUGGING. */
12892 /* RE engine related */
12893 Zero(&PL_reg_state, 1, struct re_save_state);
12894 PL_reginterp_cnt = 0;
12895 PL_regmatch_slab = NULL;
12897 /* Clone the regex array */
12898 /* ORANGE FIXME for plugins, probably in the SV dup code.
12899 newSViv(PTR2IV(CALLREGDUPE(
12900 INT2PTR(REGEXP *, SvIVX(regex)), param))))
12902 PL_regex_padav = av_dup_inc(proto_perl->Iregex_padav, param);
12903 PL_regex_pad = AvARRAY(PL_regex_padav);
12905 /* shortcuts to various I/O objects */
12906 PL_ofsgv = gv_dup_inc(proto_perl->Iofsgv, param);
12907 PL_stdingv = gv_dup(proto_perl->Istdingv, param);
12908 PL_stderrgv = gv_dup(proto_perl->Istderrgv, param);
12909 PL_defgv = gv_dup(proto_perl->Idefgv, param);
12910 PL_argvgv = gv_dup(proto_perl->Iargvgv, param);
12911 PL_argvoutgv = gv_dup(proto_perl->Iargvoutgv, param);
12912 PL_argvout_stack = av_dup_inc(proto_perl->Iargvout_stack, param);
12914 /* shortcuts to regexp stuff */
12915 PL_replgv = gv_dup(proto_perl->Ireplgv, param);
12917 /* shortcuts to misc objects */
12918 PL_errgv = gv_dup(proto_perl->Ierrgv, param);
12920 /* shortcuts to debugging objects */
12921 PL_DBgv = gv_dup(proto_perl->IDBgv, param);
12922 PL_DBline = gv_dup(proto_perl->IDBline, param);
12923 PL_DBsub = gv_dup(proto_perl->IDBsub, param);
12924 PL_DBsingle = sv_dup(proto_perl->IDBsingle, param);
12925 PL_DBtrace = sv_dup(proto_perl->IDBtrace, param);
12926 PL_DBsignal = sv_dup(proto_perl->IDBsignal, param);
12928 /* symbol tables */
12929 PL_defstash = hv_dup_inc(proto_perl->Idefstash, param);
12930 PL_curstash = hv_dup(proto_perl->Icurstash, param);
12931 PL_debstash = hv_dup(proto_perl->Idebstash, param);
12932 PL_globalstash = hv_dup(proto_perl->Iglobalstash, param);
12933 PL_curstname = sv_dup_inc(proto_perl->Icurstname, param);
12935 PL_beginav = av_dup_inc(proto_perl->Ibeginav, param);
12936 PL_beginav_save = av_dup_inc(proto_perl->Ibeginav_save, param);
12937 PL_checkav_save = av_dup_inc(proto_perl->Icheckav_save, param);
12938 PL_unitcheckav = av_dup_inc(proto_perl->Iunitcheckav, param);
12939 PL_unitcheckav_save = av_dup_inc(proto_perl->Iunitcheckav_save, param);
12940 PL_endav = av_dup_inc(proto_perl->Iendav, param);
12941 PL_checkav = av_dup_inc(proto_perl->Icheckav, param);
12942 PL_initav = av_dup_inc(proto_perl->Iinitav, param);
12944 PL_sub_generation = proto_perl->Isub_generation;
12945 PL_isarev = hv_dup_inc(proto_perl->Iisarev, param);
12947 /* funky return mechanisms */
12948 PL_forkprocess = proto_perl->Iforkprocess;
12950 /* subprocess state */
12951 PL_fdpid = av_dup_inc(proto_perl->Ifdpid, param);
12953 /* internal state */
12954 PL_maxo = proto_perl->Imaxo;
12955 if (proto_perl->Iop_mask)
12956 PL_op_mask = SAVEPVN(proto_perl->Iop_mask, PL_maxo);
12959 /* PL_asserting = proto_perl->Iasserting; */
12961 /* current interpreter roots */
12962 PL_main_cv = cv_dup_inc(proto_perl->Imain_cv, param);
12964 PL_main_root = OpREFCNT_inc(proto_perl->Imain_root);
12966 PL_main_start = proto_perl->Imain_start;
12967 PL_eval_root = proto_perl->Ieval_root;
12968 PL_eval_start = proto_perl->Ieval_start;
12970 /* runtime control stuff */
12971 PL_curcopdb = (COP*)any_dup(proto_perl->Icurcopdb, proto_perl);
12973 PL_filemode = proto_perl->Ifilemode;
12974 PL_lastfd = proto_perl->Ilastfd;
12975 PL_oldname = proto_perl->Ioldname; /* XXX not quite right */
12978 PL_gensym = proto_perl->Igensym;
12979 PL_preambleav = av_dup_inc(proto_perl->Ipreambleav, param);
12980 PL_laststatval = proto_perl->Ilaststatval;
12981 PL_laststype = proto_perl->Ilaststype;
12984 PL_ors_sv = sv_dup_inc(proto_perl->Iors_sv, param);
12986 /* interpreter atexit processing */
12987 PL_exitlistlen = proto_perl->Iexitlistlen;
12988 if (PL_exitlistlen) {
12989 Newx(PL_exitlist, PL_exitlistlen, PerlExitListEntry);
12990 Copy(proto_perl->Iexitlist, PL_exitlist, PL_exitlistlen, PerlExitListEntry);
12993 PL_exitlist = (PerlExitListEntry*)NULL;
12995 PL_my_cxt_size = proto_perl->Imy_cxt_size;
12996 if (PL_my_cxt_size) {
12997 Newx(PL_my_cxt_list, PL_my_cxt_size, void *);
12998 Copy(proto_perl->Imy_cxt_list, PL_my_cxt_list, PL_my_cxt_size, void *);
12999 #ifdef PERL_GLOBAL_STRUCT_PRIVATE
13000 Newx(PL_my_cxt_keys, PL_my_cxt_size, const char *);
13001 Copy(proto_perl->Imy_cxt_keys, PL_my_cxt_keys, PL_my_cxt_size, char *);
13005 PL_my_cxt_list = (void**)NULL;
13006 #ifdef PERL_GLOBAL_STRUCT_PRIVATE
13007 PL_my_cxt_keys = (const char**)NULL;
13010 PL_modglobal = hv_dup_inc(proto_perl->Imodglobal, param);
13011 PL_custom_op_names = hv_dup_inc(proto_perl->Icustom_op_names,param);
13012 PL_custom_op_descs = hv_dup_inc(proto_perl->Icustom_op_descs,param);
13013 PL_custom_ops = hv_dup_inc(proto_perl->Icustom_ops, param);
13015 PL_profiledata = NULL;
13017 PL_compcv = cv_dup(proto_perl->Icompcv, param);
13019 PAD_CLONE_VARS(proto_perl, param);
13021 #ifdef HAVE_INTERP_INTERN
13022 sys_intern_dup(&proto_perl->Isys_intern, &PL_sys_intern);
13025 /* more statics moved here */
13026 PL_generation = proto_perl->Igeneration;
13027 PL_DBcv = cv_dup(proto_perl->IDBcv, param);
13029 PL_in_clean_objs = proto_perl->Iin_clean_objs;
13030 PL_in_clean_all = proto_perl->Iin_clean_all;
13032 PL_uid = proto_perl->Iuid;
13033 PL_euid = proto_perl->Ieuid;
13034 PL_gid = proto_perl->Igid;
13035 PL_egid = proto_perl->Iegid;
13036 PL_nomemok = proto_perl->Inomemok;
13037 PL_an = proto_perl->Ian;
13038 PL_evalseq = proto_perl->Ievalseq;
13039 PL_origenviron = proto_perl->Iorigenviron; /* XXX not quite right */
13040 PL_origalen = proto_perl->Iorigalen;
13041 #ifdef PERL_USES_PL_PIDSTATUS
13042 PL_pidstatus = newHV(); /* XXX flag for cloning? */
13044 PL_osname = SAVEPV(proto_perl->Iosname);
13045 PL_sighandlerp = proto_perl->Isighandlerp;
13047 PL_runops = proto_perl->Irunops;
13049 PL_parser = parser_dup(proto_perl->Iparser, param);
13051 /* XXX this only works if the saved cop has already been cloned */
13052 if (proto_perl->Iparser) {
13053 PL_parser->saved_curcop = (COP*)any_dup(
13054 proto_perl->Iparser->saved_curcop,
13058 PL_subline = proto_perl->Isubline;
13059 PL_subname = sv_dup_inc(proto_perl->Isubname, param);
13062 PL_cryptseen = proto_perl->Icryptseen;
13065 PL_hints = proto_perl->Ihints;
13067 PL_amagic_generation = proto_perl->Iamagic_generation;
13069 #ifdef USE_LOCALE_COLLATE
13070 PL_collation_ix = proto_perl->Icollation_ix;
13071 PL_collation_name = SAVEPV(proto_perl->Icollation_name);
13072 PL_collation_standard = proto_perl->Icollation_standard;
13073 PL_collxfrm_base = proto_perl->Icollxfrm_base;
13074 PL_collxfrm_mult = proto_perl->Icollxfrm_mult;
13075 #endif /* USE_LOCALE_COLLATE */
13077 #ifdef USE_LOCALE_NUMERIC
13078 PL_numeric_name = SAVEPV(proto_perl->Inumeric_name);
13079 PL_numeric_standard = proto_perl->Inumeric_standard;
13080 PL_numeric_local = proto_perl->Inumeric_local;
13081 PL_numeric_radix_sv = sv_dup_inc(proto_perl->Inumeric_radix_sv, param);
13082 #endif /* !USE_LOCALE_NUMERIC */
13084 /* utf8 character classes */
13085 PL_utf8_alnum = sv_dup_inc(proto_perl->Iutf8_alnum, param);
13086 PL_utf8_ascii = sv_dup_inc(proto_perl->Iutf8_ascii, param);
13087 PL_utf8_alpha = sv_dup_inc(proto_perl->Iutf8_alpha, param);
13088 PL_utf8_space = sv_dup_inc(proto_perl->Iutf8_space, param);
13089 PL_utf8_cntrl = sv_dup_inc(proto_perl->Iutf8_cntrl, param);
13090 PL_utf8_graph = sv_dup_inc(proto_perl->Iutf8_graph, param);
13091 PL_utf8_digit = sv_dup_inc(proto_perl->Iutf8_digit, param);
13092 PL_utf8_upper = sv_dup_inc(proto_perl->Iutf8_upper, param);
13093 PL_utf8_lower = sv_dup_inc(proto_perl->Iutf8_lower, param);
13094 PL_utf8_print = sv_dup_inc(proto_perl->Iutf8_print, param);
13095 PL_utf8_punct = sv_dup_inc(proto_perl->Iutf8_punct, param);
13096 PL_utf8_xdigit = sv_dup_inc(proto_perl->Iutf8_xdigit, param);
13097 PL_utf8_mark = sv_dup_inc(proto_perl->Iutf8_mark, param);
13098 PL_utf8_X_begin = sv_dup_inc(proto_perl->Iutf8_X_begin, param);
13099 PL_utf8_X_extend = sv_dup_inc(proto_perl->Iutf8_X_extend, param);
13100 PL_utf8_X_prepend = sv_dup_inc(proto_perl->Iutf8_X_prepend, param);
13101 PL_utf8_X_non_hangul = sv_dup_inc(proto_perl->Iutf8_X_non_hangul, param);
13102 PL_utf8_X_L = sv_dup_inc(proto_perl->Iutf8_X_L, param);
13103 PL_utf8_X_LV = sv_dup_inc(proto_perl->Iutf8_X_LV, param);
13104 PL_utf8_X_LVT = sv_dup_inc(proto_perl->Iutf8_X_LVT, param);
13105 PL_utf8_X_T = sv_dup_inc(proto_perl->Iutf8_X_T, param);
13106 PL_utf8_X_V = sv_dup_inc(proto_perl->Iutf8_X_V, param);
13107 PL_utf8_X_LV_LVT_V = sv_dup_inc(proto_perl->Iutf8_X_LV_LVT_V, param);
13108 PL_utf8_toupper = sv_dup_inc(proto_perl->Iutf8_toupper, param);
13109 PL_utf8_totitle = sv_dup_inc(proto_perl->Iutf8_totitle, param);
13110 PL_utf8_tolower = sv_dup_inc(proto_perl->Iutf8_tolower, param);
13111 PL_utf8_tofold = sv_dup_inc(proto_perl->Iutf8_tofold, param);
13112 PL_utf8_idstart = sv_dup_inc(proto_perl->Iutf8_idstart, param);
13113 PL_utf8_idcont = sv_dup_inc(proto_perl->Iutf8_idcont, param);
13115 /* Did the locale setup indicate UTF-8? */
13116 PL_utf8locale = proto_perl->Iutf8locale;
13117 /* Unicode features (see perlrun/-C) */
13118 PL_unicode = proto_perl->Iunicode;
13120 /* Pre-5.8 signals control */
13121 PL_signals = proto_perl->Isignals;
13123 /* times() ticks per second */
13124 PL_clocktick = proto_perl->Iclocktick;
13126 /* Recursion stopper for PerlIO_find_layer */
13127 PL_in_load_module = proto_perl->Iin_load_module;
13129 /* sort() routine */
13130 PL_sort_RealCmp = proto_perl->Isort_RealCmp;
13132 /* Not really needed/useful since the reenrant_retint is "volatile",
13133 * but do it for consistency's sake. */
13134 PL_reentrant_retint = proto_perl->Ireentrant_retint;
13136 /* Hooks to shared SVs and locks. */
13137 PL_sharehook = proto_perl->Isharehook;
13138 PL_lockhook = proto_perl->Ilockhook;
13139 PL_unlockhook = proto_perl->Iunlockhook;
13140 PL_threadhook = proto_perl->Ithreadhook;
13141 PL_destroyhook = proto_perl->Idestroyhook;
13142 PL_signalhook = proto_perl->Isignalhook;
13144 #ifdef THREADS_HAVE_PIDS
13145 PL_ppid = proto_perl->Ippid;
13149 PL_last_swash_hv = NULL; /* reinits on demand */
13150 PL_last_swash_klen = 0;
13151 PL_last_swash_key[0]= '\0';
13152 PL_last_swash_tmps = (U8*)NULL;
13153 PL_last_swash_slen = 0;
13155 PL_glob_index = proto_perl->Iglob_index;
13156 PL_srand_called = proto_perl->Isrand_called;
13158 if (proto_perl->Ipsig_pend) {
13159 Newxz(PL_psig_pend, SIG_SIZE, int);
13162 PL_psig_pend = (int*)NULL;
13165 if (proto_perl->Ipsig_name) {
13166 Newx(PL_psig_name, 2 * SIG_SIZE, SV*);
13167 sv_dup_inc_multiple(proto_perl->Ipsig_name, PL_psig_name, 2 * SIG_SIZE,
13169 PL_psig_ptr = PL_psig_name + SIG_SIZE;
13172 PL_psig_ptr = (SV**)NULL;
13173 PL_psig_name = (SV**)NULL;
13176 /* intrpvar.h stuff */
13178 if (flags & CLONEf_COPY_STACKS) {
13179 /* next allocation will be PL_tmps_stack[PL_tmps_ix+1] */
13180 PL_tmps_ix = proto_perl->Itmps_ix;
13181 PL_tmps_max = proto_perl->Itmps_max;
13182 PL_tmps_floor = proto_perl->Itmps_floor;
13183 Newx(PL_tmps_stack, PL_tmps_max, SV*);
13184 sv_dup_inc_multiple(proto_perl->Itmps_stack, PL_tmps_stack,
13185 PL_tmps_ix+1, param);
13187 /* next PUSHMARK() sets *(PL_markstack_ptr+1) */
13188 i = proto_perl->Imarkstack_max - proto_perl->Imarkstack;
13189 Newxz(PL_markstack, i, I32);
13190 PL_markstack_max = PL_markstack + (proto_perl->Imarkstack_max
13191 - proto_perl->Imarkstack);
13192 PL_markstack_ptr = PL_markstack + (proto_perl->Imarkstack_ptr
13193 - proto_perl->Imarkstack);
13194 Copy(proto_perl->Imarkstack, PL_markstack,
13195 PL_markstack_ptr - PL_markstack + 1, I32);
13197 /* next push_scope()/ENTER sets PL_scopestack[PL_scopestack_ix]
13198 * NOTE: unlike the others! */
13199 PL_scopestack_ix = proto_perl->Iscopestack_ix;
13200 PL_scopestack_max = proto_perl->Iscopestack_max;
13201 Newxz(PL_scopestack, PL_scopestack_max, I32);
13202 Copy(proto_perl->Iscopestack, PL_scopestack, PL_scopestack_ix, I32);
13205 Newxz(PL_scopestack_name, PL_scopestack_max, const char *);
13206 Copy(proto_perl->Iscopestack_name, PL_scopestack_name, PL_scopestack_ix, const char *);
13208 /* NOTE: si_dup() looks at PL_markstack */
13209 PL_curstackinfo = si_dup(proto_perl->Icurstackinfo, param);
13211 /* PL_curstack = PL_curstackinfo->si_stack; */
13212 PL_curstack = av_dup(proto_perl->Icurstack, param);
13213 PL_mainstack = av_dup(proto_perl->Imainstack, param);
13215 /* next PUSHs() etc. set *(PL_stack_sp+1) */
13216 PL_stack_base = AvARRAY(PL_curstack);
13217 PL_stack_sp = PL_stack_base + (proto_perl->Istack_sp
13218 - proto_perl->Istack_base);
13219 PL_stack_max = PL_stack_base + AvMAX(PL_curstack);
13221 /* next SSPUSHFOO() sets PL_savestack[PL_savestack_ix]
13222 * NOTE: unlike the others! */
13223 PL_savestack_ix = proto_perl->Isavestack_ix;
13224 PL_savestack_max = proto_perl->Isavestack_max;
13225 /*Newxz(PL_savestack, PL_savestack_max, ANY);*/
13226 PL_savestack = ss_dup(proto_perl, param);
13230 ENTER; /* perl_destruct() wants to LEAVE; */
13233 PL_start_env = proto_perl->Istart_env; /* XXXXXX */
13234 PL_top_env = &PL_start_env;
13236 PL_op = proto_perl->Iop;
13239 PL_Xpv = (XPV*)NULL;
13240 my_perl->Ina = proto_perl->Ina;
13242 PL_statbuf = proto_perl->Istatbuf;
13243 PL_statcache = proto_perl->Istatcache;
13244 PL_statgv = gv_dup(proto_perl->Istatgv, param);
13245 PL_statname = sv_dup_inc(proto_perl->Istatname, param);
13247 PL_timesbuf = proto_perl->Itimesbuf;
13250 PL_tainted = proto_perl->Itainted;
13251 PL_curpm = proto_perl->Icurpm; /* XXX No PMOP ref count */
13252 PL_rs = sv_dup_inc(proto_perl->Irs, param);
13253 PL_last_in_gv = gv_dup(proto_perl->Ilast_in_gv, param);
13254 PL_defoutgv = gv_dup_inc(proto_perl->Idefoutgv, param);
13255 PL_chopset = proto_perl->Ichopset; /* XXX never deallocated */
13256 PL_toptarget = sv_dup_inc(proto_perl->Itoptarget, param);
13257 PL_bodytarget = sv_dup_inc(proto_perl->Ibodytarget, param);
13258 PL_formtarget = sv_dup(proto_perl->Iformtarget, param);
13260 PL_restartjmpenv = proto_perl->Irestartjmpenv;
13261 PL_restartop = proto_perl->Irestartop;
13262 PL_in_eval = proto_perl->Iin_eval;
13263 PL_delaymagic = proto_perl->Idelaymagic;
13264 PL_phase = proto_perl->Iphase;
13265 PL_localizing = proto_perl->Ilocalizing;
13267 PL_errors = sv_dup_inc(proto_perl->Ierrors, param);
13268 PL_hv_fetch_ent_mh = NULL;
13269 PL_modcount = proto_perl->Imodcount;
13270 PL_lastgotoprobe = NULL;
13271 PL_dumpindent = proto_perl->Idumpindent;
13273 PL_sortcop = (OP*)any_dup(proto_perl->Isortcop, proto_perl);
13274 PL_sortstash = hv_dup(proto_perl->Isortstash, param);
13275 PL_firstgv = gv_dup(proto_perl->Ifirstgv, param);
13276 PL_secondgv = gv_dup(proto_perl->Isecondgv, param);
13277 PL_efloatbuf = NULL; /* reinits on demand */
13278 PL_efloatsize = 0; /* reinits on demand */
13282 PL_screamfirst = NULL;
13283 PL_screamnext = NULL;
13284 PL_maxscream = -1; /* reinits on demand */
13285 PL_lastscream = NULL;
13288 PL_regdummy = proto_perl->Iregdummy;
13289 PL_colorset = 0; /* reinits PL_colors[] */
13290 /*PL_colors[6] = {0,0,0,0,0,0};*/
13294 /* Pluggable optimizer */
13295 PL_peepp = proto_perl->Ipeepp;
13296 PL_rpeepp = proto_perl->Irpeepp;
13297 /* op_free() hook */
13298 PL_opfreehook = proto_perl->Iopfreehook;
13300 PL_stashcache = newHV();
13302 PL_watchaddr = (char **) ptr_table_fetch(PL_ptr_table,
13303 proto_perl->Iwatchaddr);
13304 PL_watchok = PL_watchaddr ? * PL_watchaddr : NULL;
13305 if (PL_debug && PL_watchaddr) {
13306 PerlIO_printf(Perl_debug_log,
13307 "WATCHING: %"UVxf" cloned as %"UVxf" with value %"UVxf"\n",
13308 PTR2UV(proto_perl->Iwatchaddr), PTR2UV(PL_watchaddr),
13309 PTR2UV(PL_watchok));
13312 PL_registered_mros = hv_dup_inc(proto_perl->Iregistered_mros, param);
13313 PL_blockhooks = av_dup_inc(proto_perl->Iblockhooks, param);
13314 PL_utf8_foldclosures = hv_dup_inc(proto_perl->Iutf8_foldclosures, param);
13316 /* Call the ->CLONE method, if it exists, for each of the stashes
13317 identified by sv_dup() above.
13319 while(av_len(param->stashes) != -1) {
13320 HV* const stash = MUTABLE_HV(av_shift(param->stashes));
13321 GV* const cloner = gv_fetchmethod_autoload(stash, "CLONE", 0);
13322 if (cloner && GvCV(cloner)) {
13327 mXPUSHs(newSVhek(HvNAME_HEK(stash)));
13329 call_sv(MUTABLE_SV(GvCV(cloner)), G_DISCARD);
13335 if (!(flags & CLONEf_KEEP_PTR_TABLE)) {
13336 ptr_table_free(PL_ptr_table);
13337 PL_ptr_table = NULL;
13340 if (!(flags & CLONEf_COPY_STACKS)) {
13341 unreferenced_to_tmp_stack(param->unreferenced);
13344 SvREFCNT_dec(param->stashes);
13346 /* orphaned? eg threads->new inside BEGIN or use */
13347 if (PL_compcv && ! SvREFCNT(PL_compcv)) {
13348 SvREFCNT_inc_simple_void(PL_compcv);
13349 SAVEFREESV(PL_compcv);
13356 S_unreferenced_to_tmp_stack(pTHX_ AV *const unreferenced)
13358 PERL_ARGS_ASSERT_UNREFERENCED_TO_TMP_STACK;
13360 if (AvFILLp(unreferenced) > -1) {
13361 SV **svp = AvARRAY(unreferenced);
13362 SV **const last = svp + AvFILLp(unreferenced);
13366 if (SvREFCNT(*svp) == 1)
13368 } while (++svp <= last);
13370 EXTEND_MORTAL(count);
13371 svp = AvARRAY(unreferenced);
13374 if (SvREFCNT(*svp) == 1) {
13375 /* Our reference is the only one to this SV. This means that
13376 in this thread, the scalar effectively has a 0 reference.
13377 That doesn't work (cleanup never happens), so donate our
13378 reference to it onto the save stack. */
13379 PL_tmps_stack[++PL_tmps_ix] = *svp;
13381 /* As an optimisation, because we are already walking the
13382 entire array, instead of above doing either
13383 SvREFCNT_inc(*svp) or *svp = &PL_sv_undef, we can instead
13384 release our reference to the scalar, so that at the end of
13385 the array owns zero references to the scalars it happens to
13386 point to. We are effectively converting the array from
13387 AvREAL() on to AvREAL() off. This saves the av_clear()
13388 (triggered by the SvREFCNT_dec(unreferenced) below) from
13389 walking the array a second time. */
13390 SvREFCNT_dec(*svp);
13393 } while (++svp <= last);
13394 AvREAL_off(unreferenced);
13396 SvREFCNT_dec(unreferenced);
13400 Perl_clone_params_del(CLONE_PARAMS *param)
13402 /* This seemingly funky ordering keeps the build with PERL_GLOBAL_STRUCT
13404 PerlInterpreter *const to = param->new_perl;
13406 PerlInterpreter *const was = PERL_GET_THX;
13408 PERL_ARGS_ASSERT_CLONE_PARAMS_DEL;
13414 SvREFCNT_dec(param->stashes);
13415 if (param->unreferenced)
13416 unreferenced_to_tmp_stack(param->unreferenced);
13426 Perl_clone_params_new(PerlInterpreter *const from, PerlInterpreter *const to)
13429 /* Need to play this game, as newAV() can call safesysmalloc(), and that
13430 does a dTHX; to get the context from thread local storage.
13431 FIXME - under PERL_CORE Newx(), Safefree() and friends should expand to
13432 a version that passes in my_perl. */
13433 PerlInterpreter *const was = PERL_GET_THX;
13434 CLONE_PARAMS *param;
13436 PERL_ARGS_ASSERT_CLONE_PARAMS_NEW;
13442 /* Given that we've set the context, we can do this unshared. */
13443 Newx(param, 1, CLONE_PARAMS);
13446 param->proto_perl = from;
13447 param->new_perl = to;
13448 param->stashes = (AV *)Perl_newSV_type(to, SVt_PVAV);
13449 AvREAL_off(param->stashes);
13450 param->unreferenced = (AV *)Perl_newSV_type(to, SVt_PVAV);
13458 #endif /* USE_ITHREADS */
13461 =head1 Unicode Support
13463 =for apidoc sv_recode_to_utf8
13465 The encoding is assumed to be an Encode object, on entry the PV
13466 of the sv is assumed to be octets in that encoding, and the sv
13467 will be converted into Unicode (and UTF-8).
13469 If the sv already is UTF-8 (or if it is not POK), or if the encoding
13470 is not a reference, nothing is done to the sv. If the encoding is not
13471 an C<Encode::XS> Encoding object, bad things will happen.
13472 (See F<lib/encoding.pm> and L<Encode>).
13474 The PV of the sv is returned.
13479 Perl_sv_recode_to_utf8(pTHX_ SV *sv, SV *encoding)
13483 PERL_ARGS_ASSERT_SV_RECODE_TO_UTF8;
13485 if (SvPOK(sv) && !SvUTF8(sv) && !IN_BYTES && SvROK(encoding)) {
13499 Passing sv_yes is wrong - it needs to be or'ed set of constants
13500 for Encode::XS, while UTf-8 decode (currently) assumes a true value means
13501 remove converted chars from source.
13503 Both will default the value - let them.
13505 XPUSHs(&PL_sv_yes);
13508 call_method("decode", G_SCALAR);
13512 s = SvPV_const(uni, len);
13513 if (s != SvPVX_const(sv)) {
13514 SvGROW(sv, len + 1);
13515 Move(s, SvPVX(sv), len + 1, char);
13516 SvCUR_set(sv, len);
13523 return SvPOKp(sv) ? SvPVX(sv) : NULL;
13527 =for apidoc sv_cat_decode
13529 The encoding is assumed to be an Encode object, the PV of the ssv is
13530 assumed to be octets in that encoding and decoding the input starts
13531 from the position which (PV + *offset) pointed to. The dsv will be
13532 concatenated the decoded UTF-8 string from ssv. Decoding will terminate
13533 when the string tstr appears in decoding output or the input ends on
13534 the PV of the ssv. The value which the offset points will be modified
13535 to the last input position on the ssv.
13537 Returns TRUE if the terminator was found, else returns FALSE.
13542 Perl_sv_cat_decode(pTHX_ SV *dsv, SV *encoding,
13543 SV *ssv, int *offset, char *tstr, int tlen)
13548 PERL_ARGS_ASSERT_SV_CAT_DECODE;
13550 if (SvPOK(ssv) && SvPOK(dsv) && SvROK(encoding) && offset) {
13561 offsv = newSViv(*offset);
13563 mXPUSHp(tstr, tlen);
13565 call_method("cat_decode", G_SCALAR);
13567 ret = SvTRUE(TOPs);
13568 *offset = SvIV(offsv);
13574 Perl_croak(aTHX_ "Invalid argument to sv_cat_decode");
13579 /* ---------------------------------------------------------------------
13581 * support functions for report_uninit()
13584 /* the maxiumum size of array or hash where we will scan looking
13585 * for the undefined element that triggered the warning */
13587 #define FUV_MAX_SEARCH_SIZE 1000
13589 /* Look for an entry in the hash whose value has the same SV as val;
13590 * If so, return a mortal copy of the key. */
13593 S_find_hash_subscript(pTHX_ const HV *const hv, const SV *const val)
13596 register HE **array;
13599 PERL_ARGS_ASSERT_FIND_HASH_SUBSCRIPT;
13601 if (!hv || SvMAGICAL(hv) || !HvARRAY(hv) ||
13602 (HvTOTALKEYS(hv) > FUV_MAX_SEARCH_SIZE))
13605 array = HvARRAY(hv);
13607 for (i=HvMAX(hv); i>0; i--) {
13608 register HE *entry;
13609 for (entry = array[i]; entry; entry = HeNEXT(entry)) {
13610 if (HeVAL(entry) != val)
13612 if ( HeVAL(entry) == &PL_sv_undef ||
13613 HeVAL(entry) == &PL_sv_placeholder)
13617 if (HeKLEN(entry) == HEf_SVKEY)
13618 return sv_mortalcopy(HeKEY_sv(entry));
13619 return sv_2mortal(newSVhek(HeKEY_hek(entry)));
13625 /* Look for an entry in the array whose value has the same SV as val;
13626 * If so, return the index, otherwise return -1. */
13629 S_find_array_subscript(pTHX_ const AV *const av, const SV *const val)
13633 PERL_ARGS_ASSERT_FIND_ARRAY_SUBSCRIPT;
13635 if (!av || SvMAGICAL(av) || !AvARRAY(av) ||
13636 (AvFILLp(av) > FUV_MAX_SEARCH_SIZE))
13639 if (val != &PL_sv_undef) {
13640 SV ** const svp = AvARRAY(av);
13643 for (i=AvFILLp(av); i>=0; i--)
13650 /* S_varname(): return the name of a variable, optionally with a subscript.
13651 * If gv is non-zero, use the name of that global, along with gvtype (one
13652 * of "$", "@", "%"); otherwise use the name of the lexical at pad offset
13653 * targ. Depending on the value of the subscript_type flag, return:
13656 #define FUV_SUBSCRIPT_NONE 1 /* "@foo" */
13657 #define FUV_SUBSCRIPT_ARRAY 2 /* "$foo[aindex]" */
13658 #define FUV_SUBSCRIPT_HASH 3 /* "$foo{keyname}" */
13659 #define FUV_SUBSCRIPT_WITHIN 4 /* "within @foo" */
13662 S_varname(pTHX_ const GV *const gv, const char gvtype, PADOFFSET targ,
13663 const SV *const keyname, I32 aindex, int subscript_type)
13666 SV * const name = sv_newmortal();
13669 buffer[0] = gvtype;
13672 /* as gv_fullname4(), but add literal '^' for $^FOO names */
13674 gv_fullname4(name, gv, buffer, 0);
13676 if ((unsigned int)SvPVX(name)[1] <= 26) {
13678 buffer[1] = SvPVX(name)[1] + 'A' - 1;
13680 /* Swap the 1 unprintable control character for the 2 byte pretty
13681 version - ie substr($name, 1, 1) = $buffer; */
13682 sv_insert(name, 1, 1, buffer, 2);
13686 CV * const cv = find_runcv(NULL);
13690 if (!cv || !CvPADLIST(cv))
13692 av = MUTABLE_AV((*av_fetch(CvPADLIST(cv), 0, FALSE)));
13693 sv = *av_fetch(av, targ, FALSE);
13694 sv_setpvn(name, SvPV_nolen_const(sv), SvCUR(sv));
13697 if (subscript_type == FUV_SUBSCRIPT_HASH) {
13698 SV * const sv = newSV(0);
13699 *SvPVX(name) = '$';
13700 Perl_sv_catpvf(aTHX_ name, "{%s}",
13701 pv_display(sv,SvPVX_const(keyname), SvCUR(keyname), 0, 32));
13704 else if (subscript_type == FUV_SUBSCRIPT_ARRAY) {
13705 *SvPVX(name) = '$';
13706 Perl_sv_catpvf(aTHX_ name, "[%"IVdf"]", (IV)aindex);
13708 else if (subscript_type == FUV_SUBSCRIPT_WITHIN) {
13709 /* We know that name has no magic, so can use 0 instead of SV_GMAGIC */
13710 Perl_sv_insert_flags(aTHX_ name, 0, 0, STR_WITH_LEN("within "), 0);
13718 =for apidoc find_uninit_var
13720 Find the name of the undefined variable (if any) that caused the operator o
13721 to issue a "Use of uninitialized value" warning.
13722 If match is true, only return a name if it's value matches uninit_sv.
13723 So roughly speaking, if a unary operator (such as OP_COS) generates a
13724 warning, then following the direct child of the op may yield an
13725 OP_PADSV or OP_GV that gives the name of the undefined variable. On the
13726 other hand, with OP_ADD there are two branches to follow, so we only print
13727 the variable name if we get an exact match.
13729 The name is returned as a mortal SV.
13731 Assumes that PL_op is the op that originally triggered the error, and that
13732 PL_comppad/PL_curpad points to the currently executing pad.
13738 S_find_uninit_var(pTHX_ const OP *const obase, const SV *const uninit_sv,
13744 const OP *o, *o2, *kid;
13746 if (!obase || (match && (!uninit_sv || uninit_sv == &PL_sv_undef ||
13747 uninit_sv == &PL_sv_placeholder)))
13750 switch (obase->op_type) {
13757 const bool pad = (obase->op_type == OP_PADAV || obase->op_type == OP_PADHV);
13758 const bool hash = (obase->op_type == OP_PADHV || obase->op_type == OP_RV2HV);
13761 int subscript_type = FUV_SUBSCRIPT_WITHIN;
13763 if (pad) { /* @lex, %lex */
13764 sv = PAD_SVl(obase->op_targ);
13768 if (cUNOPx(obase)->op_first->op_type == OP_GV) {
13769 /* @global, %global */
13770 gv = cGVOPx_gv(cUNOPx(obase)->op_first);
13773 sv = hash ? MUTABLE_SV(GvHV(gv)): MUTABLE_SV(GvAV(gv));
13775 else /* @{expr}, %{expr} */
13776 return find_uninit_var(cUNOPx(obase)->op_first,
13780 /* attempt to find a match within the aggregate */
13782 keysv = find_hash_subscript((const HV*)sv, uninit_sv);
13784 subscript_type = FUV_SUBSCRIPT_HASH;
13787 index = find_array_subscript((const AV *)sv, uninit_sv);
13789 subscript_type = FUV_SUBSCRIPT_ARRAY;
13792 if (match && subscript_type == FUV_SUBSCRIPT_WITHIN)
13795 return varname(gv, hash ? '%' : '@', obase->op_targ,
13796 keysv, index, subscript_type);
13800 if (match && PAD_SVl(obase->op_targ) != uninit_sv)
13802 return varname(NULL, '$', obase->op_targ,
13803 NULL, 0, FUV_SUBSCRIPT_NONE);
13806 gv = cGVOPx_gv(obase);
13807 if (!gv || (match && GvSV(gv) != uninit_sv) || !GvSTASH(gv))
13809 return varname(gv, '$', 0, NULL, 0, FUV_SUBSCRIPT_NONE);
13812 if (obase->op_flags & OPf_SPECIAL) { /* lexical array */
13815 AV *av = MUTABLE_AV(PAD_SV(obase->op_targ));
13816 if (!av || SvRMAGICAL(av))
13818 svp = av_fetch(av, (I32)obase->op_private, FALSE);
13819 if (!svp || *svp != uninit_sv)
13822 return varname(NULL, '$', obase->op_targ,
13823 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
13826 gv = cGVOPx_gv(obase);
13831 AV *const av = GvAV(gv);
13832 if (!av || SvRMAGICAL(av))
13834 svp = av_fetch(av, (I32)obase->op_private, FALSE);
13835 if (!svp || *svp != uninit_sv)
13838 return varname(gv, '$', 0,
13839 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
13844 o = cUNOPx(obase)->op_first;
13845 if (!o || o->op_type != OP_NULL ||
13846 ! (o->op_targ == OP_AELEM || o->op_targ == OP_HELEM))
13848 return find_uninit_var(cBINOPo->op_last, uninit_sv, match);
13852 if (PL_op == obase)
13853 /* $a[uninit_expr] or $h{uninit_expr} */
13854 return find_uninit_var(cBINOPx(obase)->op_last, uninit_sv, match);
13857 o = cBINOPx(obase)->op_first;
13858 kid = cBINOPx(obase)->op_last;
13860 /* get the av or hv, and optionally the gv */
13862 if (o->op_type == OP_PADAV || o->op_type == OP_PADHV) {
13863 sv = PAD_SV(o->op_targ);
13865 else if ((o->op_type == OP_RV2AV || o->op_type == OP_RV2HV)
13866 && cUNOPo->op_first->op_type == OP_GV)
13868 gv = cGVOPx_gv(cUNOPo->op_first);
13872 == OP_RV2HV ? MUTABLE_SV(GvHV(gv)) : MUTABLE_SV(GvAV(gv));
13877 if (kid && kid->op_type == OP_CONST && SvOK(cSVOPx_sv(kid))) {
13878 /* index is constant */
13882 if (obase->op_type == OP_HELEM) {
13883 HE* he = hv_fetch_ent(MUTABLE_HV(sv), cSVOPx_sv(kid), 0, 0);
13884 if (!he || HeVAL(he) != uninit_sv)
13888 SV * const * const svp = av_fetch(MUTABLE_AV(sv), SvIV(cSVOPx_sv(kid)), FALSE);
13889 if (!svp || *svp != uninit_sv)
13893 if (obase->op_type == OP_HELEM)
13894 return varname(gv, '%', o->op_targ,
13895 cSVOPx_sv(kid), 0, FUV_SUBSCRIPT_HASH);
13897 return varname(gv, '@', o->op_targ, NULL,
13898 SvIV(cSVOPx_sv(kid)), FUV_SUBSCRIPT_ARRAY);
13901 /* index is an expression;
13902 * attempt to find a match within the aggregate */
13903 if (obase->op_type == OP_HELEM) {
13904 SV * const keysv = find_hash_subscript((const HV*)sv, uninit_sv);
13906 return varname(gv, '%', o->op_targ,
13907 keysv, 0, FUV_SUBSCRIPT_HASH);
13911 = find_array_subscript((const AV *)sv, uninit_sv);
13913 return varname(gv, '@', o->op_targ,
13914 NULL, index, FUV_SUBSCRIPT_ARRAY);
13919 (o->op_type == OP_PADAV || o->op_type == OP_RV2AV)
13921 o->op_targ, NULL, 0, FUV_SUBSCRIPT_WITHIN);
13926 /* only examine RHS */
13927 return find_uninit_var(cBINOPx(obase)->op_first, uninit_sv, match);
13930 o = cUNOPx(obase)->op_first;
13931 if (o->op_type == OP_PUSHMARK)
13934 if (!o->op_sibling) {
13935 /* one-arg version of open is highly magical */
13937 if (o->op_type == OP_GV) { /* open FOO; */
13939 if (match && GvSV(gv) != uninit_sv)
13941 return varname(gv, '$', 0,
13942 NULL, 0, FUV_SUBSCRIPT_NONE);
13944 /* other possibilities not handled are:
13945 * open $x; or open my $x; should return '${*$x}'
13946 * open expr; should return '$'.expr ideally
13952 /* ops where $_ may be an implicit arg */
13956 if ( !(obase->op_flags & OPf_STACKED)) {
13957 if (uninit_sv == ((obase->op_private & OPpTARGET_MY)
13958 ? PAD_SVl(obase->op_targ)
13961 sv = sv_newmortal();
13962 sv_setpvs(sv, "$_");
13971 match = 1; /* print etc can return undef on defined args */
13972 /* skip filehandle as it can't produce 'undef' warning */
13973 o = cUNOPx(obase)->op_first;
13974 if ((obase->op_flags & OPf_STACKED) && o->op_type == OP_PUSHMARK)
13975 o = o->op_sibling->op_sibling;
13979 case OP_ENTEREVAL: /* could be eval $undef or $x='$undef'; eval $x */
13981 case OP_CUSTOM: /* XS or custom code could trigger random warnings */
13983 /* the following ops are capable of returning PL_sv_undef even for
13984 * defined arg(s) */
14003 case OP_GETPEERNAME:
14051 case OP_SMARTMATCH:
14060 /* XXX tmp hack: these two may call an XS sub, and currently
14061 XS subs don't have a SUB entry on the context stack, so CV and
14062 pad determination goes wrong, and BAD things happen. So, just
14063 don't try to determine the value under those circumstances.
14064 Need a better fix at dome point. DAPM 11/2007 */
14070 GV * const gv = gv_fetchpvs(".", GV_NOTQUAL, SVt_PV);
14071 if (gv && GvSV(gv) == uninit_sv)
14072 return newSVpvs_flags("$.", SVs_TEMP);
14077 /* def-ness of rval pos() is independent of the def-ness of its arg */
14078 if ( !(obase->op_flags & OPf_MOD))
14083 if (SvROK(PL_rs) && uninit_sv == SvRV(PL_rs))
14084 return newSVpvs_flags("${$/}", SVs_TEMP);
14089 if (!(obase->op_flags & OPf_KIDS))
14091 o = cUNOPx(obase)->op_first;
14097 /* if all except one arg are constant, or have no side-effects,
14098 * or are optimized away, then it's unambiguous */
14100 for (kid=o; kid; kid = kid->op_sibling) {
14102 const OPCODE type = kid->op_type;
14103 if ( (type == OP_CONST && SvOK(cSVOPx_sv(kid)))
14104 || (type == OP_NULL && ! (kid->op_flags & OPf_KIDS))
14105 || (type == OP_PUSHMARK)
14107 /* @$a and %$a, but not @a or %a */
14108 (type == OP_RV2AV || type == OP_RV2HV)
14109 && cUNOPx(kid)->op_first
14110 && cUNOPx(kid)->op_first->op_type != OP_GV
14115 if (o2) { /* more than one found */
14122 return find_uninit_var(o2, uninit_sv, match);
14124 /* scan all args */
14126 sv = find_uninit_var(o, uninit_sv, 1);
14138 =for apidoc report_uninit
14140 Print appropriate "Use of uninitialized variable" warning
14146 Perl_report_uninit(pTHX_ const SV *uninit_sv)
14150 SV* varname = NULL;
14152 varname = find_uninit_var(PL_op, uninit_sv,0);
14154 sv_insert(varname, 0, 0, " ", 1);
14156 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
14157 varname ? SvPV_nolen_const(varname) : "",
14158 " in ", OP_DESC(PL_op));
14161 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
14167 * c-indentation-style: bsd
14168 * c-basic-offset: 4
14169 * indent-tabs-mode: t
14172 * ex: set ts=8 sts=4 sw=4 noet: