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 const SV * const svend = &sva[SvREFCNT(sva)];
415 for (sv = sva + 1; sv < svend; ++sv) {
416 if (SvTYPE(sv) != (svtype)SVTYPEMASK
417 && (sv->sv_flags & mask) == flags
430 /* called by sv_report_used() for each live SV */
433 do_report_used(pTHX_ SV *const sv)
435 if (SvTYPE(sv) != (svtype)SVTYPEMASK) {
436 PerlIO_printf(Perl_debug_log, "****\n");
443 =for apidoc sv_report_used
445 Dump the contents of all SVs not yet freed (debugging aid).
451 Perl_sv_report_used(pTHX)
454 visit(do_report_used, 0, 0);
460 /* called by sv_clean_objs() for each live SV */
463 do_clean_objs(pTHX_ SV *const ref)
468 SV * const target = SvRV(ref);
469 if (SvOBJECT(target)) {
470 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning object ref:\n "), sv_dump(ref)));
471 if (SvWEAKREF(ref)) {
472 sv_del_backref(target, ref);
478 SvREFCNT_dec(target);
483 /* XXX Might want to check arrays, etc. */
487 /* clear any slots in a GV which hold objects - except IO;
488 * called by sv_clean_objs() for each live GV */
491 do_clean_named_objs(pTHX_ SV *const sv)
495 assert(SvTYPE(sv) == SVt_PVGV);
496 assert(isGV_with_GP(sv));
500 /* freeing GP entries may indirectly free the current GV;
501 * hold onto it while we mess with the GP slots */
504 if ( ((obj = GvSV(sv) )) && SvOBJECT(obj)) {
505 DEBUG_D((PerlIO_printf(Perl_debug_log,
506 "Cleaning named glob SV object:\n "), sv_dump(obj)));
510 if ( ((obj = MUTABLE_SV(GvAV(sv)) )) && SvOBJECT(obj)) {
511 DEBUG_D((PerlIO_printf(Perl_debug_log,
512 "Cleaning named glob AV object:\n "), sv_dump(obj)));
516 if ( ((obj = MUTABLE_SV(GvHV(sv)) )) && SvOBJECT(obj)) {
517 DEBUG_D((PerlIO_printf(Perl_debug_log,
518 "Cleaning named glob HV object:\n "), sv_dump(obj)));
522 if ( ((obj = MUTABLE_SV(GvCV(sv)) )) && SvOBJECT(obj)) {
523 DEBUG_D((PerlIO_printf(Perl_debug_log,
524 "Cleaning named glob CV object:\n "), sv_dump(obj)));
528 SvREFCNT_dec(sv); /* undo the inc above */
531 /* clear any IO slots in a GV which hold objects (except stderr, defout);
532 * called by sv_clean_objs() for each live GV */
535 do_clean_named_io_objs(pTHX_ SV *const sv)
539 assert(SvTYPE(sv) == SVt_PVGV);
540 assert(isGV_with_GP(sv));
541 if (!GvGP(sv) || sv == (SV*)PL_stderrgv || sv == (SV*)PL_defoutgv)
545 if ( ((obj = MUTABLE_SV(GvIO(sv)) )) && SvOBJECT(obj)) {
546 DEBUG_D((PerlIO_printf(Perl_debug_log,
547 "Cleaning named glob IO object:\n "), sv_dump(obj)));
551 SvREFCNT_dec(sv); /* undo the inc above */
554 /* Void wrapper to pass to visit() */
556 do_curse(pTHX_ SV * const sv) {
557 if ((PL_stderrgv && GvGP(PL_stderrgv) && (SV*)GvIO(PL_stderrgv) == sv)
558 || (PL_defoutgv && GvGP(PL_defoutgv) && (SV*)GvIO(PL_defoutgv) == sv))
564 =for apidoc sv_clean_objs
566 Attempt to destroy all objects not yet freed.
572 Perl_sv_clean_objs(pTHX)
576 PL_in_clean_objs = TRUE;
577 visit(do_clean_objs, SVf_ROK, SVf_ROK);
578 /* Some barnacles may yet remain, clinging to typeglobs.
579 * Run the non-IO destructors first: they may want to output
580 * error messages, close files etc */
581 visit(do_clean_named_objs, SVt_PVGV|SVpgv_GP, SVTYPEMASK|SVp_POK|SVpgv_GP);
582 visit(do_clean_named_io_objs, SVt_PVGV|SVpgv_GP, SVTYPEMASK|SVp_POK|SVpgv_GP);
583 /* And if there are some very tenacious barnacles clinging to arrays,
584 closures, or what have you.... */
585 visit(do_curse, SVs_OBJECT, SVs_OBJECT);
586 olddef = PL_defoutgv;
587 PL_defoutgv = NULL; /* disable skip of PL_defoutgv */
588 if (olddef && isGV_with_GP(olddef))
589 do_clean_named_io_objs(aTHX_ MUTABLE_SV(olddef));
590 olderr = PL_stderrgv;
591 PL_stderrgv = NULL; /* disable skip of PL_stderrgv */
592 if (olderr && isGV_with_GP(olderr))
593 do_clean_named_io_objs(aTHX_ MUTABLE_SV(olderr));
594 SvREFCNT_dec(olddef);
595 PL_in_clean_objs = FALSE;
598 /* called by sv_clean_all() for each live SV */
601 do_clean_all(pTHX_ SV *const sv)
604 if (sv == (const SV *) PL_fdpid || sv == (const SV *)PL_strtab) {
605 /* don't clean pid table and strtab */
608 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning loops: SV at 0x%"UVxf"\n", PTR2UV(sv)) ));
609 SvFLAGS(sv) |= SVf_BREAK;
614 =for apidoc sv_clean_all
616 Decrement the refcnt of each remaining SV, possibly triggering a
617 cleanup. This function may have to be called multiple times to free
618 SVs which are in complex self-referential hierarchies.
624 Perl_sv_clean_all(pTHX)
628 PL_in_clean_all = TRUE;
629 cleaned = visit(do_clean_all, 0,0);
634 ARENASETS: a meta-arena implementation which separates arena-info
635 into struct arena_set, which contains an array of struct
636 arena_descs, each holding info for a single arena. By separating
637 the meta-info from the arena, we recover the 1st slot, formerly
638 borrowed for list management. The arena_set is about the size of an
639 arena, avoiding the needless malloc overhead of a naive linked-list.
641 The cost is 1 arena-set malloc per ~320 arena-mallocs, + the unused
642 memory in the last arena-set (1/2 on average). In trade, we get
643 back the 1st slot in each arena (ie 1.7% of a CV-arena, less for
644 smaller types). The recovery of the wasted space allows use of
645 small arenas for large, rare body types, by changing array* fields
646 in body_details_by_type[] below.
649 char *arena; /* the raw storage, allocated aligned */
650 size_t size; /* its size ~4k typ */
651 svtype utype; /* bodytype stored in arena */
656 /* Get the maximum number of elements in set[] such that struct arena_set
657 will fit within PERL_ARENA_SIZE, which is probably just under 4K, and
658 therefore likely to be 1 aligned memory page. */
660 #define ARENAS_PER_SET ((PERL_ARENA_SIZE - sizeof(struct arena_set*) \
661 - 2 * sizeof(int)) / sizeof (struct arena_desc))
664 struct arena_set* next;
665 unsigned int set_size; /* ie ARENAS_PER_SET */
666 unsigned int curr; /* index of next available arena-desc */
667 struct arena_desc set[ARENAS_PER_SET];
671 =for apidoc sv_free_arenas
673 Deallocate the memory used by all arenas. Note that all the individual SV
674 heads and bodies within the arenas must already have been freed.
679 Perl_sv_free_arenas(pTHX)
686 /* Free arenas here, but be careful about fake ones. (We assume
687 contiguity of the fake ones with the corresponding real ones.) */
689 for (sva = PL_sv_arenaroot; sva; sva = svanext) {
690 svanext = MUTABLE_SV(SvANY(sva));
691 while (svanext && SvFAKE(svanext))
692 svanext = MUTABLE_SV(SvANY(svanext));
699 struct arena_set *aroot = (struct arena_set*) PL_body_arenas;
702 struct arena_set *current = aroot;
705 assert(aroot->set[i].arena);
706 Safefree(aroot->set[i].arena);
714 i = PERL_ARENA_ROOTS_SIZE;
716 PL_body_roots[i] = 0;
723 Here are mid-level routines that manage the allocation of bodies out
724 of the various arenas. There are 5 kinds of arenas:
726 1. SV-head arenas, which are discussed and handled above
727 2. regular body arenas
728 3. arenas for reduced-size bodies
731 Arena types 2 & 3 are chained by body-type off an array of
732 arena-root pointers, which is indexed by svtype. Some of the
733 larger/less used body types are malloced singly, since a large
734 unused block of them is wasteful. Also, several svtypes dont have
735 bodies; the data fits into the sv-head itself. The arena-root
736 pointer thus has a few unused root-pointers (which may be hijacked
737 later for arena types 4,5)
739 3 differs from 2 as an optimization; some body types have several
740 unused fields in the front of the structure (which are kept in-place
741 for consistency). These bodies can be allocated in smaller chunks,
742 because the leading fields arent accessed. Pointers to such bodies
743 are decremented to point at the unused 'ghost' memory, knowing that
744 the pointers are used with offsets to the real memory.
747 =head1 SV-Body Allocation
749 Allocation of SV-bodies is similar to SV-heads, differing as follows;
750 the allocation mechanism is used for many body types, so is somewhat
751 more complicated, it uses arena-sets, and has no need for still-live
754 At the outermost level, (new|del)_X*V macros return bodies of the
755 appropriate type. These macros call either (new|del)_body_type or
756 (new|del)_body_allocated macro pairs, depending on specifics of the
757 type. Most body types use the former pair, the latter pair is used to
758 allocate body types with "ghost fields".
760 "ghost fields" are fields that are unused in certain types, and
761 consequently don't need to actually exist. They are declared because
762 they're part of a "base type", which allows use of functions as
763 methods. The simplest examples are AVs and HVs, 2 aggregate types
764 which don't use the fields which support SCALAR semantics.
766 For these types, the arenas are carved up into appropriately sized
767 chunks, we thus avoid wasted memory for those unaccessed members.
768 When bodies are allocated, we adjust the pointer back in memory by the
769 size of the part not allocated, so it's as if we allocated the full
770 structure. (But things will all go boom if you write to the part that
771 is "not there", because you'll be overwriting the last members of the
772 preceding structure in memory.)
774 We calculate the correction using the STRUCT_OFFSET macro on the first
775 member present. If the allocated structure is smaller (no initial NV
776 actually allocated) then the net effect is to subtract the size of the NV
777 from the pointer, to return a new pointer as if an initial NV were actually
778 allocated. (We were using structures named *_allocated for this, but
779 this turned out to be a subtle bug, because a structure without an NV
780 could have a lower alignment constraint, but the compiler is allowed to
781 optimised accesses based on the alignment constraint of the actual pointer
782 to the full structure, for example, using a single 64 bit load instruction
783 because it "knows" that two adjacent 32 bit members will be 8-byte aligned.)
785 This is the same trick as was used for NV and IV bodies. Ironically it
786 doesn't need to be used for NV bodies any more, because NV is now at
787 the start of the structure. IV bodies don't need it either, because
788 they are no longer allocated.
790 In turn, the new_body_* allocators call S_new_body(), which invokes
791 new_body_inline macro, which takes a lock, and takes a body off the
792 linked list at PL_body_roots[sv_type], calling Perl_more_bodies() if
793 necessary to refresh an empty list. Then the lock is released, and
794 the body is returned.
796 Perl_more_bodies allocates a new arena, and carves it up into an array of N
797 bodies, which it strings into a linked list. It looks up arena-size
798 and body-size from the body_details table described below, thus
799 supporting the multiple body-types.
801 If PURIFY is defined, or PERL_ARENA_SIZE=0, arenas are not used, and
802 the (new|del)_X*V macros are mapped directly to malloc/free.
804 For each sv-type, struct body_details bodies_by_type[] carries
805 parameters which control these aspects of SV handling:
807 Arena_size determines whether arenas are used for this body type, and if
808 so, how big they are. PURIFY or PERL_ARENA_SIZE=0 set this field to
809 zero, forcing individual mallocs and frees.
811 Body_size determines how big a body is, and therefore how many fit into
812 each arena. Offset carries the body-pointer adjustment needed for
813 "ghost fields", and is used in *_allocated macros.
815 But its main purpose is to parameterize info needed in
816 Perl_sv_upgrade(). The info here dramatically simplifies the function
817 vs the implementation in 5.8.8, making it table-driven. All fields
818 are used for this, except for arena_size.
820 For the sv-types that have no bodies, arenas are not used, so those
821 PL_body_roots[sv_type] are unused, and can be overloaded. In
822 something of a special case, SVt_NULL is borrowed for HE arenas;
823 PL_body_roots[HE_SVSLOT=SVt_NULL] is filled by S_more_he, but the
824 bodies_by_type[SVt_NULL] slot is not used, as the table is not
829 struct body_details {
830 U8 body_size; /* Size to allocate */
831 U8 copy; /* Size of structure to copy (may be shorter) */
833 unsigned int type : 4; /* We have space for a sanity check. */
834 unsigned int cant_upgrade : 1; /* Cannot upgrade this type */
835 unsigned int zero_nv : 1; /* zero the NV when upgrading from this */
836 unsigned int arena : 1; /* Allocated from an arena */
837 size_t arena_size; /* Size of arena to allocate */
845 /* With -DPURFIY we allocate everything directly, and don't use arenas.
846 This seems a rather elegant way to simplify some of the code below. */
847 #define HASARENA FALSE
849 #define HASARENA TRUE
851 #define NOARENA FALSE
853 /* Size the arenas to exactly fit a given number of bodies. A count
854 of 0 fits the max number bodies into a PERL_ARENA_SIZE.block,
855 simplifying the default. If count > 0, the arena is sized to fit
856 only that many bodies, allowing arenas to be used for large, rare
857 bodies (XPVFM, XPVIO) without undue waste. The arena size is
858 limited by PERL_ARENA_SIZE, so we can safely oversize the
861 #define FIT_ARENA0(body_size) \
862 ((size_t)(PERL_ARENA_SIZE / body_size) * body_size)
863 #define FIT_ARENAn(count,body_size) \
864 ( count * body_size <= PERL_ARENA_SIZE) \
865 ? count * body_size \
866 : FIT_ARENA0 (body_size)
867 #define FIT_ARENA(count,body_size) \
869 ? FIT_ARENAn (count, body_size) \
870 : FIT_ARENA0 (body_size)
872 /* Calculate the length to copy. Specifically work out the length less any
873 final padding the compiler needed to add. See the comment in sv_upgrade
874 for why copying the padding proved to be a bug. */
876 #define copy_length(type, last_member) \
877 STRUCT_OFFSET(type, last_member) \
878 + sizeof (((type*)SvANY((const SV *)0))->last_member)
880 static const struct body_details bodies_by_type[] = {
881 /* HEs use this offset for their arena. */
882 { 0, 0, 0, SVt_NULL, FALSE, NONV, NOARENA, 0 },
884 /* The bind placeholder pretends to be an RV for now.
885 Also it's marked as "can't upgrade" to stop anyone using it before it's
887 { 0, 0, 0, SVt_BIND, TRUE, NONV, NOARENA, 0 },
889 /* IVs are in the head, so the allocation size is 0. */
891 sizeof(IV), /* This is used to copy out the IV body. */
892 STRUCT_OFFSET(XPVIV, xiv_iv), SVt_IV, FALSE, NONV,
893 NOARENA /* IVS don't need an arena */, 0
896 { sizeof(NV), sizeof(NV),
897 STRUCT_OFFSET(XPVNV, xnv_u),
898 SVt_NV, FALSE, HADNV, HASARENA, FIT_ARENA(0, sizeof(NV)) },
900 { sizeof(XPV) - STRUCT_OFFSET(XPV, xpv_cur),
901 copy_length(XPV, xpv_len) - STRUCT_OFFSET(XPV, xpv_cur),
902 + STRUCT_OFFSET(XPV, xpv_cur),
903 SVt_PV, FALSE, NONV, HASARENA,
904 FIT_ARENA(0, sizeof(XPV) - STRUCT_OFFSET(XPV, xpv_cur)) },
906 { sizeof(XPVIV) - STRUCT_OFFSET(XPV, xpv_cur),
907 copy_length(XPVIV, xiv_u) - STRUCT_OFFSET(XPV, xpv_cur),
908 + STRUCT_OFFSET(XPV, xpv_cur),
909 SVt_PVIV, FALSE, NONV, HASARENA,
910 FIT_ARENA(0, sizeof(XPVIV) - STRUCT_OFFSET(XPV, xpv_cur)) },
912 { sizeof(XPVNV) - STRUCT_OFFSET(XPV, xpv_cur),
913 copy_length(XPVNV, xnv_u) - STRUCT_OFFSET(XPV, xpv_cur),
914 + STRUCT_OFFSET(XPV, xpv_cur),
915 SVt_PVNV, FALSE, HADNV, HASARENA,
916 FIT_ARENA(0, sizeof(XPVNV) - STRUCT_OFFSET(XPV, xpv_cur)) },
918 { sizeof(XPVMG), copy_length(XPVMG, xnv_u), 0, SVt_PVMG, FALSE, HADNV,
919 HASARENA, FIT_ARENA(0, sizeof(XPVMG)) },
924 SVt_REGEXP, FALSE, NONV, HASARENA,
925 FIT_ARENA(0, sizeof(regexp))
928 { sizeof(XPVGV), sizeof(XPVGV), 0, SVt_PVGV, TRUE, HADNV,
929 HASARENA, FIT_ARENA(0, sizeof(XPVGV)) },
931 { sizeof(XPVLV), sizeof(XPVLV), 0, SVt_PVLV, TRUE, HADNV,
932 HASARENA, FIT_ARENA(0, sizeof(XPVLV)) },
935 copy_length(XPVAV, xav_alloc),
937 SVt_PVAV, TRUE, NONV, HASARENA,
938 FIT_ARENA(0, sizeof(XPVAV)) },
941 copy_length(XPVHV, xhv_max),
943 SVt_PVHV, TRUE, NONV, HASARENA,
944 FIT_ARENA(0, sizeof(XPVHV)) },
949 SVt_PVCV, TRUE, NONV, HASARENA,
950 FIT_ARENA(0, sizeof(XPVCV)) },
955 SVt_PVFM, TRUE, NONV, NOARENA,
956 FIT_ARENA(20, sizeof(XPVFM)) },
961 SVt_PVIO, TRUE, NONV, HASARENA,
962 FIT_ARENA(24, sizeof(XPVIO)) },
965 #define new_body_allocated(sv_type) \
966 (void *)((char *)S_new_body(aTHX_ sv_type) \
967 - bodies_by_type[sv_type].offset)
969 /* return a thing to the free list */
971 #define del_body(thing, root) \
973 void ** const thing_copy = (void **)thing; \
974 *thing_copy = *root; \
975 *root = (void*)thing_copy; \
980 #define new_XNV() safemalloc(sizeof(XPVNV))
981 #define new_XPVNV() safemalloc(sizeof(XPVNV))
982 #define new_XPVMG() safemalloc(sizeof(XPVMG))
984 #define del_XPVGV(p) safefree(p)
988 #define new_XNV() new_body_allocated(SVt_NV)
989 #define new_XPVNV() new_body_allocated(SVt_PVNV)
990 #define new_XPVMG() new_body_allocated(SVt_PVMG)
992 #define del_XPVGV(p) del_body(p + bodies_by_type[SVt_PVGV].offset, \
993 &PL_body_roots[SVt_PVGV])
997 /* no arena for you! */
999 #define new_NOARENA(details) \
1000 safemalloc((details)->body_size + (details)->offset)
1001 #define new_NOARENAZ(details) \
1002 safecalloc((details)->body_size + (details)->offset, 1)
1005 Perl_more_bodies (pTHX_ const svtype sv_type, const size_t body_size,
1006 const size_t arena_size)
1009 void ** const root = &PL_body_roots[sv_type];
1010 struct arena_desc *adesc;
1011 struct arena_set *aroot = (struct arena_set *) PL_body_arenas;
1015 const size_t good_arena_size = Perl_malloc_good_size(arena_size);
1016 #if defined(DEBUGGING) && !defined(PERL_GLOBAL_STRUCT_PRIVATE)
1017 static bool done_sanity_check;
1019 /* PERL_GLOBAL_STRUCT_PRIVATE cannot coexist with global
1020 * variables like done_sanity_check. */
1021 if (!done_sanity_check) {
1022 unsigned int i = SVt_LAST;
1024 done_sanity_check = TRUE;
1027 assert (bodies_by_type[i].type == i);
1033 /* may need new arena-set to hold new arena */
1034 if (!aroot || aroot->curr >= aroot->set_size) {
1035 struct arena_set *newroot;
1036 Newxz(newroot, 1, struct arena_set);
1037 newroot->set_size = ARENAS_PER_SET;
1038 newroot->next = aroot;
1040 PL_body_arenas = (void *) newroot;
1041 DEBUG_m(PerlIO_printf(Perl_debug_log, "new arenaset %p\n", (void*)aroot));
1044 /* ok, now have arena-set with at least 1 empty/available arena-desc */
1045 curr = aroot->curr++;
1046 adesc = &(aroot->set[curr]);
1047 assert(!adesc->arena);
1049 Newx(adesc->arena, good_arena_size, char);
1050 adesc->size = good_arena_size;
1051 adesc->utype = sv_type;
1052 DEBUG_m(PerlIO_printf(Perl_debug_log, "arena %d added: %p size %"UVuf"\n",
1053 curr, (void*)adesc->arena, (UV)good_arena_size));
1055 start = (char *) adesc->arena;
1057 /* Get the address of the byte after the end of the last body we can fit.
1058 Remember, this is integer division: */
1059 end = start + good_arena_size / body_size * body_size;
1061 /* computed count doesn't reflect the 1st slot reservation */
1062 #if defined(MYMALLOC) || defined(HAS_MALLOC_GOOD_SIZE)
1063 DEBUG_m(PerlIO_printf(Perl_debug_log,
1064 "arena %p end %p arena-size %d (from %d) type %d "
1066 (void*)start, (void*)end, (int)good_arena_size,
1067 (int)arena_size, sv_type, (int)body_size,
1068 (int)good_arena_size / (int)body_size));
1070 DEBUG_m(PerlIO_printf(Perl_debug_log,
1071 "arena %p end %p arena-size %d type %d size %d ct %d\n",
1072 (void*)start, (void*)end,
1073 (int)arena_size, sv_type, (int)body_size,
1074 (int)good_arena_size / (int)body_size));
1076 *root = (void *)start;
1079 /* Where the next body would start: */
1080 char * const next = start + body_size;
1083 /* This is the last body: */
1084 assert(next == end);
1086 *(void **)start = 0;
1090 *(void**) start = (void *)next;
1095 /* grab a new thing from the free list, allocating more if necessary.
1096 The inline version is used for speed in hot routines, and the
1097 function using it serves the rest (unless PURIFY).
1099 #define new_body_inline(xpv, sv_type) \
1101 void ** const r3wt = &PL_body_roots[sv_type]; \
1102 xpv = (PTR_TBL_ENT_t*) (*((void **)(r3wt)) \
1103 ? *((void **)(r3wt)) : Perl_more_bodies(aTHX_ sv_type, \
1104 bodies_by_type[sv_type].body_size,\
1105 bodies_by_type[sv_type].arena_size)); \
1106 *(r3wt) = *(void**)(xpv); \
1112 S_new_body(pTHX_ const svtype sv_type)
1116 new_body_inline(xpv, sv_type);
1122 static const struct body_details fake_rv =
1123 { 0, 0, 0, SVt_IV, FALSE, NONV, NOARENA, 0 };
1126 =for apidoc sv_upgrade
1128 Upgrade an SV to a more complex form. Generally adds a new body type to the
1129 SV, then copies across as much information as possible from the old body.
1130 It croaks if the SV is already in a more complex form than requested. You
1131 generally want to use the C<SvUPGRADE> macro wrapper, which checks the type
1132 before calling C<sv_upgrade>, and hence does not croak. See also
1139 Perl_sv_upgrade(pTHX_ register SV *const sv, svtype new_type)
1144 const svtype old_type = SvTYPE(sv);
1145 const struct body_details *new_type_details;
1146 const struct body_details *old_type_details
1147 = bodies_by_type + old_type;
1148 SV *referant = NULL;
1150 PERL_ARGS_ASSERT_SV_UPGRADE;
1152 if (old_type == new_type)
1155 /* This clause was purposefully added ahead of the early return above to
1156 the shared string hackery for (sort {$a <=> $b} keys %hash), with the
1157 inference by Nick I-S that it would fix other troublesome cases. See
1158 changes 7162, 7163 (f130fd4589cf5fbb24149cd4db4137c8326f49c1 and parent)
1160 Given that shared hash key scalars are no longer PVIV, but PV, there is
1161 no longer need to unshare so as to free up the IVX slot for its proper
1162 purpose. So it's safe to move the early return earlier. */
1164 if (new_type != SVt_PV && SvIsCOW(sv)) {
1165 sv_force_normal_flags(sv, 0);
1168 old_body = SvANY(sv);
1170 /* Copying structures onto other structures that have been neatly zeroed
1171 has a subtle gotcha. Consider XPVMG
1173 +------+------+------+------+------+-------+-------+
1174 | NV | CUR | LEN | IV | MAGIC | STASH |
1175 +------+------+------+------+------+-------+-------+
1176 0 4 8 12 16 20 24 28
1178 where NVs are aligned to 8 bytes, so that sizeof that structure is
1179 actually 32 bytes long, with 4 bytes of padding at the end:
1181 +------+------+------+------+------+-------+-------+------+
1182 | NV | CUR | LEN | IV | MAGIC | STASH | ??? |
1183 +------+------+------+------+------+-------+-------+------+
1184 0 4 8 12 16 20 24 28 32
1186 so what happens if you allocate memory for this structure:
1188 +------+------+------+------+------+-------+-------+------+------+...
1189 | NV | CUR | LEN | IV | MAGIC | STASH | GP | NAME |
1190 +------+------+------+------+------+-------+-------+------+------+...
1191 0 4 8 12 16 20 24 28 32 36
1193 zero it, then copy sizeof(XPVMG) bytes on top of it? Not quite what you
1194 expect, because you copy the area marked ??? onto GP. Now, ??? may have
1195 started out as zero once, but it's quite possible that it isn't. So now,
1196 rather than a nicely zeroed GP, you have it pointing somewhere random.
1199 (In fact, GP ends up pointing at a previous GP structure, because the
1200 principle cause of the padding in XPVMG getting garbage is a copy of
1201 sizeof(XPVMG) bytes from a XPVGV structure in sv_unglob. Right now
1202 this happens to be moot because XPVGV has been re-ordered, with GP
1203 no longer after STASH)
1205 So we are careful and work out the size of used parts of all the
1213 referant = SvRV(sv);
1214 old_type_details = &fake_rv;
1215 if (new_type == SVt_NV)
1216 new_type = SVt_PVNV;
1218 if (new_type < SVt_PVIV) {
1219 new_type = (new_type == SVt_NV)
1220 ? SVt_PVNV : SVt_PVIV;
1225 if (new_type < SVt_PVNV) {
1226 new_type = SVt_PVNV;
1230 assert(new_type > SVt_PV);
1231 assert(SVt_IV < SVt_PV);
1232 assert(SVt_NV < SVt_PV);
1239 /* Because the XPVMG of PL_mess_sv isn't allocated from the arena,
1240 there's no way that it can be safely upgraded, because perl.c
1241 expects to Safefree(SvANY(PL_mess_sv)) */
1242 assert(sv != PL_mess_sv);
1243 /* This flag bit is used to mean other things in other scalar types.
1244 Given that it only has meaning inside the pad, it shouldn't be set
1245 on anything that can get upgraded. */
1246 assert(!SvPAD_TYPED(sv));
1249 if (old_type_details->cant_upgrade)
1250 Perl_croak(aTHX_ "Can't upgrade %s (%" UVuf ") to %" UVuf,
1251 sv_reftype(sv, 0), (UV) old_type, (UV) new_type);
1254 if (old_type > new_type)
1255 Perl_croak(aTHX_ "sv_upgrade from type %d down to type %d",
1256 (int)old_type, (int)new_type);
1258 new_type_details = bodies_by_type + new_type;
1260 SvFLAGS(sv) &= ~SVTYPEMASK;
1261 SvFLAGS(sv) |= new_type;
1263 /* This can't happen, as SVt_NULL is <= all values of new_type, so one of
1264 the return statements above will have triggered. */
1265 assert (new_type != SVt_NULL);
1268 assert(old_type == SVt_NULL);
1269 SvANY(sv) = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
1273 assert(old_type == SVt_NULL);
1274 SvANY(sv) = new_XNV();
1279 assert(new_type_details->body_size);
1282 assert(new_type_details->arena);
1283 assert(new_type_details->arena_size);
1284 /* This points to the start of the allocated area. */
1285 new_body_inline(new_body, new_type);
1286 Zero(new_body, new_type_details->body_size, char);
1287 new_body = ((char *)new_body) - new_type_details->offset;
1289 /* We always allocated the full length item with PURIFY. To do this
1290 we fake things so that arena is false for all 16 types.. */
1291 new_body = new_NOARENAZ(new_type_details);
1293 SvANY(sv) = new_body;
1294 if (new_type == SVt_PVAV) {
1298 if (old_type_details->body_size) {
1301 /* It will have been zeroed when the new body was allocated.
1302 Lets not write to it, in case it confuses a write-back
1308 #ifndef NODEFAULT_SHAREKEYS
1309 HvSHAREKEYS_on(sv); /* key-sharing on by default */
1311 HvMAX(sv) = 7; /* (start with 8 buckets) */
1314 /* SVt_NULL isn't the only thing upgraded to AV or HV.
1315 The target created by newSVrv also is, and it can have magic.
1316 However, it never has SvPVX set.
1318 if (old_type == SVt_IV) {
1320 } else if (old_type >= SVt_PV) {
1321 assert(SvPVX_const(sv) == 0);
1324 if (old_type >= SVt_PVMG) {
1325 SvMAGIC_set(sv, ((XPVMG*)old_body)->xmg_u.xmg_magic);
1326 SvSTASH_set(sv, ((XPVMG*)old_body)->xmg_stash);
1328 sv->sv_u.svu_array = NULL; /* or svu_hash */
1334 /* This ensures that SvTHINKFIRST(sv) is true, and hence that
1335 sv_force_normal_flags(sv) is called. */
1338 /* XXX Is this still needed? Was it ever needed? Surely as there is
1339 no route from NV to PVIV, NOK can never be true */
1340 assert(!SvNOKp(sv));
1351 assert(new_type_details->body_size);
1352 /* We always allocated the full length item with PURIFY. To do this
1353 we fake things so that arena is false for all 16 types.. */
1354 if(new_type_details->arena) {
1355 /* This points to the start of the allocated area. */
1356 new_body_inline(new_body, new_type);
1357 Zero(new_body, new_type_details->body_size, char);
1358 new_body = ((char *)new_body) - new_type_details->offset;
1360 new_body = new_NOARENAZ(new_type_details);
1362 SvANY(sv) = new_body;
1364 if (old_type_details->copy) {
1365 /* There is now the potential for an upgrade from something without
1366 an offset (PVNV or PVMG) to something with one (PVCV, PVFM) */
1367 int offset = old_type_details->offset;
1368 int length = old_type_details->copy;
1370 if (new_type_details->offset > old_type_details->offset) {
1371 const int difference
1372 = new_type_details->offset - old_type_details->offset;
1373 offset += difference;
1374 length -= difference;
1376 assert (length >= 0);
1378 Copy((char *)old_body + offset, (char *)new_body + offset, length,
1382 #ifndef NV_ZERO_IS_ALLBITS_ZERO
1383 /* If NV 0.0 is stores as all bits 0 then Zero() already creates a
1384 * correct 0.0 for us. Otherwise, if the old body didn't have an
1385 * NV slot, but the new one does, then we need to initialise the
1386 * freshly created NV slot with whatever the correct bit pattern is
1388 if (old_type_details->zero_nv && !new_type_details->zero_nv
1389 && !isGV_with_GP(sv))
1393 if (new_type == SVt_PVIO) {
1394 IO * const io = MUTABLE_IO(sv);
1395 GV *iogv = gv_fetchpvs("IO::File::", GV_ADD, SVt_PVHV);
1398 /* Clear the stashcache because a new IO could overrule a package
1400 hv_clear(PL_stashcache);
1402 SvSTASH_set(io, MUTABLE_HV(SvREFCNT_inc(GvHV(iogv))));
1403 IoPAGE_LEN(sv) = 60;
1405 if (old_type < SVt_PV) {
1406 /* referant will be NULL unless the old type was SVt_IV emulating
1408 sv->sv_u.svu_rv = referant;
1412 Perl_croak(aTHX_ "panic: sv_upgrade to unknown type %lu",
1413 (unsigned long)new_type);
1416 if (old_type > SVt_IV) {
1420 /* Note that there is an assumption that all bodies of types that
1421 can be upgraded came from arenas. Only the more complex non-
1422 upgradable types are allowed to be directly malloc()ed. */
1423 assert(old_type_details->arena);
1424 del_body((void*)((char*)old_body + old_type_details->offset),
1425 &PL_body_roots[old_type]);
1431 =for apidoc sv_backoff
1433 Remove any string offset. You should normally use the C<SvOOK_off> macro
1440 Perl_sv_backoff(pTHX_ register SV *const sv)
1443 const char * const s = SvPVX_const(sv);
1445 PERL_ARGS_ASSERT_SV_BACKOFF;
1446 PERL_UNUSED_CONTEXT;
1449 assert(SvTYPE(sv) != SVt_PVHV);
1450 assert(SvTYPE(sv) != SVt_PVAV);
1452 SvOOK_offset(sv, delta);
1454 SvLEN_set(sv, SvLEN(sv) + delta);
1455 SvPV_set(sv, SvPVX(sv) - delta);
1456 Move(s, SvPVX(sv), SvCUR(sv)+1, char);
1457 SvFLAGS(sv) &= ~SVf_OOK;
1464 Expands the character buffer in the SV. If necessary, uses C<sv_unref> and
1465 upgrades the SV to C<SVt_PV>. Returns a pointer to the character buffer.
1466 Use the C<SvGROW> wrapper instead.
1472 Perl_sv_grow(pTHX_ register SV *const sv, register STRLEN newlen)
1476 PERL_ARGS_ASSERT_SV_GROW;
1478 if (PL_madskills && newlen >= 0x100000) {
1479 PerlIO_printf(Perl_debug_log,
1480 "Allocation too large: %"UVxf"\n", (UV)newlen);
1482 #ifdef HAS_64K_LIMIT
1483 if (newlen >= 0x10000) {
1484 PerlIO_printf(Perl_debug_log,
1485 "Allocation too large: %"UVxf"\n", (UV)newlen);
1488 #endif /* HAS_64K_LIMIT */
1491 if (SvTYPE(sv) < SVt_PV) {
1492 sv_upgrade(sv, SVt_PV);
1493 s = SvPVX_mutable(sv);
1495 else if (SvOOK(sv)) { /* pv is offset? */
1497 s = SvPVX_mutable(sv);
1498 if (newlen > SvLEN(sv))
1499 newlen += 10 * (newlen - SvCUR(sv)); /* avoid copy each time */
1500 #ifdef HAS_64K_LIMIT
1501 if (newlen >= 0x10000)
1506 s = SvPVX_mutable(sv);
1508 if (newlen > SvLEN(sv)) { /* need more room? */
1509 STRLEN minlen = SvCUR(sv);
1510 minlen += (minlen >> PERL_STRLEN_EXPAND_SHIFT) + 10;
1511 if (newlen < minlen)
1513 #ifndef Perl_safesysmalloc_size
1514 newlen = PERL_STRLEN_ROUNDUP(newlen);
1516 if (SvLEN(sv) && s) {
1517 s = (char*)saferealloc(s, newlen);
1520 s = (char*)safemalloc(newlen);
1521 if (SvPVX_const(sv) && SvCUR(sv)) {
1522 Move(SvPVX_const(sv), s, (newlen < SvCUR(sv)) ? newlen : SvCUR(sv), char);
1526 #ifdef Perl_safesysmalloc_size
1527 /* Do this here, do it once, do it right, and then we will never get
1528 called back into sv_grow() unless there really is some growing
1530 SvLEN_set(sv, Perl_safesysmalloc_size(s));
1532 SvLEN_set(sv, newlen);
1539 =for apidoc sv_setiv
1541 Copies an integer into the given SV, upgrading first if necessary.
1542 Does not handle 'set' magic. See also C<sv_setiv_mg>.
1548 Perl_sv_setiv(pTHX_ register SV *const sv, const IV i)
1552 PERL_ARGS_ASSERT_SV_SETIV;
1554 SV_CHECK_THINKFIRST_COW_DROP(sv);
1555 switch (SvTYPE(sv)) {
1558 sv_upgrade(sv, SVt_IV);
1561 sv_upgrade(sv, SVt_PVIV);
1565 if (!isGV_with_GP(sv))
1572 /* diag_listed_as: Can't coerce %s to %s in %s */
1573 Perl_croak(aTHX_ "Can't coerce %s to integer in %s", sv_reftype(sv,0),
1577 (void)SvIOK_only(sv); /* validate number */
1583 =for apidoc sv_setiv_mg
1585 Like C<sv_setiv>, but also handles 'set' magic.
1591 Perl_sv_setiv_mg(pTHX_ register SV *const sv, const IV i)
1593 PERL_ARGS_ASSERT_SV_SETIV_MG;
1600 =for apidoc sv_setuv
1602 Copies an unsigned integer into the given SV, upgrading first if necessary.
1603 Does not handle 'set' magic. See also C<sv_setuv_mg>.
1609 Perl_sv_setuv(pTHX_ register SV *const sv, const UV u)
1611 PERL_ARGS_ASSERT_SV_SETUV;
1613 /* With the if statement to ensure that integers are stored as IVs whenever
1615 u=1.49 s=0.52 cu=72.49 cs=10.64 scripts=270 tests=20865
1618 u=1.35 s=0.47 cu=73.45 cs=11.43 scripts=270 tests=20865
1620 If you wish to remove the following if statement, so that this routine
1621 (and its callers) always return UVs, please benchmark to see what the
1622 effect is. Modern CPUs may be different. Or may not :-)
1624 if (u <= (UV)IV_MAX) {
1625 sv_setiv(sv, (IV)u);
1634 =for apidoc sv_setuv_mg
1636 Like C<sv_setuv>, but also handles 'set' magic.
1642 Perl_sv_setuv_mg(pTHX_ register SV *const sv, const UV u)
1644 PERL_ARGS_ASSERT_SV_SETUV_MG;
1651 =for apidoc sv_setnv
1653 Copies a double into the given SV, upgrading first if necessary.
1654 Does not handle 'set' magic. See also C<sv_setnv_mg>.
1660 Perl_sv_setnv(pTHX_ register SV *const sv, const NV num)
1664 PERL_ARGS_ASSERT_SV_SETNV;
1666 SV_CHECK_THINKFIRST_COW_DROP(sv);
1667 switch (SvTYPE(sv)) {
1670 sv_upgrade(sv, SVt_NV);
1674 sv_upgrade(sv, SVt_PVNV);
1678 if (!isGV_with_GP(sv))
1685 /* diag_listed_as: Can't coerce %s to %s in %s */
1686 Perl_croak(aTHX_ "Can't coerce %s to number in %s", sv_reftype(sv,0),
1691 (void)SvNOK_only(sv); /* validate number */
1696 =for apidoc sv_setnv_mg
1698 Like C<sv_setnv>, but also handles 'set' magic.
1704 Perl_sv_setnv_mg(pTHX_ register SV *const sv, const NV num)
1706 PERL_ARGS_ASSERT_SV_SETNV_MG;
1712 /* Print an "isn't numeric" warning, using a cleaned-up,
1713 * printable version of the offending string
1717 S_not_a_number(pTHX_ SV *const sv)
1724 PERL_ARGS_ASSERT_NOT_A_NUMBER;
1727 dsv = newSVpvs_flags("", SVs_TEMP);
1728 pv = sv_uni_display(dsv, sv, 10, UNI_DISPLAY_ISPRINT);
1731 const char * const limit = tmpbuf + sizeof(tmpbuf) - 8;
1732 /* each *s can expand to 4 chars + "...\0",
1733 i.e. need room for 8 chars */
1735 const char *s = SvPVX_const(sv);
1736 const char * const end = s + SvCUR(sv);
1737 for ( ; s < end && d < limit; s++ ) {
1739 if (ch & 128 && !isPRINT_LC(ch)) {
1748 else if (ch == '\r') {
1752 else if (ch == '\f') {
1756 else if (ch == '\\') {
1760 else if (ch == '\0') {
1764 else if (isPRINT_LC(ch))
1781 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
1782 /* diag_listed_as: Argument "%s" isn't numeric%s */
1783 "Argument \"%s\" isn't numeric in %s", pv,
1786 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
1787 /* diag_listed_as: Argument "%s" isn't numeric%s */
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. Get-magic is
1803 Perl_looks_like_number(pTHX_ SV *const sv)
1808 PERL_ARGS_ASSERT_LOOKS_LIKE_NUMBER;
1810 if (SvPOK(sv) || SvPOKp(sv)) {
1811 sbegin = SvPV_nomg_const(sv, len);
1814 return SvFLAGS(sv) & (SVf_NOK|SVp_NOK|SVf_IOK|SVp_IOK);
1815 return grok_number(sbegin, len, NULL);
1819 S_glob_2number(pTHX_ GV * const gv)
1821 PERL_ARGS_ASSERT_GLOB_2NUMBER;
1823 /* We know that all GVs stringify to something that is not-a-number,
1824 so no need to test that. */
1825 if (ckWARN(WARN_NUMERIC))
1827 SV *const buffer = sv_newmortal();
1828 gv_efullname3(buffer, gv, "*");
1829 not_a_number(buffer);
1831 /* We just want something true to return, so that S_sv_2iuv_common
1832 can tail call us and return true. */
1836 /* Actually, ISO C leaves conversion of UV to IV undefined, but
1837 until proven guilty, assume that things are not that bad... */
1842 As 64 bit platforms often have an NV that doesn't preserve all bits of
1843 an IV (an assumption perl has been based on to date) it becomes necessary
1844 to remove the assumption that the NV always carries enough precision to
1845 recreate the IV whenever needed, and that the NV is the canonical form.
1846 Instead, IV/UV and NV need to be given equal rights. So as to not lose
1847 precision as a side effect of conversion (which would lead to insanity
1848 and the dragon(s) in t/op/numconvert.t getting very angry) the intent is
1849 1) to distinguish between IV/UV/NV slots that have cached a valid
1850 conversion where precision was lost and IV/UV/NV slots that have a
1851 valid conversion which has lost no precision
1852 2) to ensure that if a numeric conversion to one form is requested that
1853 would lose precision, the precise conversion (or differently
1854 imprecise conversion) is also performed and cached, to prevent
1855 requests for different numeric formats on the same SV causing
1856 lossy conversion chains. (lossless conversion chains are perfectly
1861 SvIOKp is true if the IV slot contains a valid value
1862 SvIOK is true only if the IV value is accurate (UV if SvIOK_UV true)
1863 SvNOKp is true if the NV slot contains a valid value
1864 SvNOK is true only if the NV value is accurate
1867 while converting from PV to NV, check to see if converting that NV to an
1868 IV(or UV) would lose accuracy over a direct conversion from PV to
1869 IV(or UV). If it would, cache both conversions, return NV, but mark
1870 SV as IOK NOKp (ie not NOK).
1872 While converting from PV to IV, check to see if converting that IV to an
1873 NV would lose accuracy over a direct conversion from PV to NV. If it
1874 would, cache both conversions, flag similarly.
1876 Before, the SV value "3.2" could become NV=3.2 IV=3 NOK, IOK quite
1877 correctly because if IV & NV were set NV *always* overruled.
1878 Now, "3.2" will become NV=3.2 IV=3 NOK, IOKp, because the flag's meaning
1879 changes - now IV and NV together means that the two are interchangeable:
1880 SvIVX == (IV) SvNVX && SvNVX == (NV) SvIVX;
1882 The benefit of this is that operations such as pp_add know that if
1883 SvIOK is true for both left and right operands, then integer addition
1884 can be used instead of floating point (for cases where the result won't
1885 overflow). Before, floating point was always used, which could lead to
1886 loss of precision compared with integer addition.
1888 * making IV and NV equal status should make maths accurate on 64 bit
1890 * may speed up maths somewhat if pp_add and friends start to use
1891 integers when possible instead of fp. (Hopefully the overhead in
1892 looking for SvIOK and checking for overflow will not outweigh the
1893 fp to integer speedup)
1894 * will slow down integer operations (callers of SvIV) on "inaccurate"
1895 values, as the change from SvIOK to SvIOKp will cause a call into
1896 sv_2iv each time rather than a macro access direct to the IV slot
1897 * should speed up number->string conversion on integers as IV is
1898 favoured when IV and NV are equally accurate
1900 ####################################################################
1901 You had better be using SvIOK_notUV if you want an IV for arithmetic:
1902 SvIOK is true if (IV or UV), so you might be getting (IV)SvUV.
1903 On the other hand, SvUOK is true iff UV.
1904 ####################################################################
1906 Your mileage will vary depending your CPU's relative fp to integer
1910 #ifndef NV_PRESERVES_UV
1911 # define IS_NUMBER_UNDERFLOW_IV 1
1912 # define IS_NUMBER_UNDERFLOW_UV 2
1913 # define IS_NUMBER_IV_AND_UV 2
1914 # define IS_NUMBER_OVERFLOW_IV 4
1915 # define IS_NUMBER_OVERFLOW_UV 5
1917 /* sv_2iuv_non_preserve(): private routine for use by sv_2iv() and sv_2uv() */
1919 /* For sv_2nv these three cases are "SvNOK and don't bother casting" */
1921 S_sv_2iuv_non_preserve(pTHX_ register SV *const sv
1929 PERL_ARGS_ASSERT_SV_2IUV_NON_PRESERVE;
1931 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));
1932 if (SvNVX(sv) < (NV)IV_MIN) {
1933 (void)SvIOKp_on(sv);
1935 SvIV_set(sv, IV_MIN);
1936 return IS_NUMBER_UNDERFLOW_IV;
1938 if (SvNVX(sv) > (NV)UV_MAX) {
1939 (void)SvIOKp_on(sv);
1942 SvUV_set(sv, UV_MAX);
1943 return IS_NUMBER_OVERFLOW_UV;
1945 (void)SvIOKp_on(sv);
1947 /* Can't use strtol etc to convert this string. (See truth table in
1949 if (SvNVX(sv) <= (UV)IV_MAX) {
1950 SvIV_set(sv, I_V(SvNVX(sv)));
1951 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
1952 SvIOK_on(sv); /* Integer is precise. NOK, IOK */
1954 /* Integer is imprecise. NOK, IOKp */
1956 return SvNVX(sv) < 0 ? IS_NUMBER_UNDERFLOW_UV : IS_NUMBER_IV_AND_UV;
1959 SvUV_set(sv, U_V(SvNVX(sv)));
1960 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
1961 if (SvUVX(sv) == UV_MAX) {
1962 /* As we know that NVs don't preserve UVs, UV_MAX cannot
1963 possibly be preserved by NV. Hence, it must be overflow.
1965 return IS_NUMBER_OVERFLOW_UV;
1967 SvIOK_on(sv); /* Integer is precise. NOK, UOK */
1969 /* Integer is imprecise. NOK, IOKp */
1971 return IS_NUMBER_OVERFLOW_IV;
1973 #endif /* !NV_PRESERVES_UV*/
1976 S_sv_2iuv_common(pTHX_ SV *const sv)
1980 PERL_ARGS_ASSERT_SV_2IUV_COMMON;
1983 /* erm. not sure. *should* never get NOKp (without NOK) from sv_2nv
1984 * without also getting a cached IV/UV from it at the same time
1985 * (ie PV->NV conversion should detect loss of accuracy and cache
1986 * IV or UV at same time to avoid this. */
1987 /* IV-over-UV optimisation - choose to cache IV if possible */
1989 if (SvTYPE(sv) == SVt_NV)
1990 sv_upgrade(sv, SVt_PVNV);
1992 (void)SvIOKp_on(sv); /* Must do this first, to clear any SvOOK */
1993 /* < not <= as for NV doesn't preserve UV, ((NV)IV_MAX+1) will almost
1994 certainly cast into the IV range at IV_MAX, whereas the correct
1995 answer is the UV IV_MAX +1. Hence < ensures that dodgy boundary
1997 #if defined(NAN_COMPARE_BROKEN) && defined(Perl_isnan)
1998 if (Perl_isnan(SvNVX(sv))) {
2004 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2005 SvIV_set(sv, I_V(SvNVX(sv)));
2006 if (SvNVX(sv) == (NV) SvIVX(sv)
2007 #ifndef NV_PRESERVES_UV
2008 && (((UV)1 << NV_PRESERVES_UV_BITS) >
2009 (UV)(SvIVX(sv) > 0 ? SvIVX(sv) : -SvIVX(sv)))
2010 /* Don't flag it as "accurately an integer" if the number
2011 came from a (by definition imprecise) NV operation, and
2012 we're outside the range of NV integer precision */
2016 SvIOK_on(sv); /* Can this go wrong with rounding? NWC */
2018 /* scalar has trailing garbage, eg "42a" */
2020 DEBUG_c(PerlIO_printf(Perl_debug_log,
2021 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (precise)\n",
2027 /* IV not precise. No need to convert from PV, as NV
2028 conversion would already have cached IV if it detected
2029 that PV->IV would be better than PV->NV->IV
2030 flags already correct - don't set public IOK. */
2031 DEBUG_c(PerlIO_printf(Perl_debug_log,
2032 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (imprecise)\n",
2037 /* Can the above go wrong if SvIVX == IV_MIN and SvNVX < IV_MIN,
2038 but the cast (NV)IV_MIN rounds to a the value less (more
2039 negative) than IV_MIN which happens to be equal to SvNVX ??
2040 Analogous to 0xFFFFFFFFFFFFFFFF rounding up to NV (2**64) and
2041 NV rounding back to 0xFFFFFFFFFFFFFFFF, so UVX == UV(NVX) and
2042 (NV)UVX == NVX are both true, but the values differ. :-(
2043 Hopefully for 2s complement IV_MIN is something like
2044 0x8000000000000000 which will be exact. NWC */
2047 SvUV_set(sv, U_V(SvNVX(sv)));
2049 (SvNVX(sv) == (NV) SvUVX(sv))
2050 #ifndef NV_PRESERVES_UV
2051 /* Make sure it's not 0xFFFFFFFFFFFFFFFF */
2052 /*&& (SvUVX(sv) != UV_MAX) irrelevant with code below */
2053 && (((UV)1 << NV_PRESERVES_UV_BITS) > SvUVX(sv))
2054 /* Don't flag it as "accurately an integer" if the number
2055 came from a (by definition imprecise) NV operation, and
2056 we're outside the range of NV integer precision */
2062 DEBUG_c(PerlIO_printf(Perl_debug_log,
2063 "0x%"UVxf" 2iv(%"UVuf" => %"IVdf") (as unsigned)\n",
2069 else if (SvPOKp(sv) && SvLEN(sv)) {
2071 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2072 /* We want to avoid a possible problem when we cache an IV/ a UV which
2073 may be later translated to an NV, and the resulting NV is not
2074 the same as the direct translation of the initial string
2075 (eg 123.456 can shortcut to the IV 123 with atol(), but we must
2076 be careful to ensure that the value with the .456 is around if the
2077 NV value is requested in the future).
2079 This means that if we cache such an IV/a UV, we need to cache the
2080 NV as well. Moreover, we trade speed for space, and do not
2081 cache the NV if we are sure it's not needed.
2084 /* SVt_PVNV is one higher than SVt_PVIV, hence this order */
2085 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2086 == IS_NUMBER_IN_UV) {
2087 /* It's definitely an integer, only upgrade to PVIV */
2088 if (SvTYPE(sv) < SVt_PVIV)
2089 sv_upgrade(sv, SVt_PVIV);
2091 } else if (SvTYPE(sv) < SVt_PVNV)
2092 sv_upgrade(sv, SVt_PVNV);
2094 /* If NVs preserve UVs then we only use the UV value if we know that
2095 we aren't going to call atof() below. If NVs don't preserve UVs
2096 then the value returned may have more precision than atof() will
2097 return, even though value isn't perfectly accurate. */
2098 if ((numtype & (IS_NUMBER_IN_UV
2099 #ifdef NV_PRESERVES_UV
2102 )) == IS_NUMBER_IN_UV) {
2103 /* This won't turn off the public IOK flag if it was set above */
2104 (void)SvIOKp_on(sv);
2106 if (!(numtype & IS_NUMBER_NEG)) {
2108 if (value <= (UV)IV_MAX) {
2109 SvIV_set(sv, (IV)value);
2111 /* it didn't overflow, and it was positive. */
2112 SvUV_set(sv, value);
2116 /* 2s complement assumption */
2117 if (value <= (UV)IV_MIN) {
2118 SvIV_set(sv, -(IV)value);
2120 /* Too negative for an IV. This is a double upgrade, but
2121 I'm assuming it will be rare. */
2122 if (SvTYPE(sv) < SVt_PVNV)
2123 sv_upgrade(sv, SVt_PVNV);
2127 SvNV_set(sv, -(NV)value);
2128 SvIV_set(sv, IV_MIN);
2132 /* For !NV_PRESERVES_UV and IS_NUMBER_IN_UV and IS_NUMBER_NOT_INT we
2133 will be in the previous block to set the IV slot, and the next
2134 block to set the NV slot. So no else here. */
2136 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2137 != IS_NUMBER_IN_UV) {
2138 /* It wasn't an (integer that doesn't overflow the UV). */
2139 SvNV_set(sv, Atof(SvPVX_const(sv)));
2141 if (! numtype && ckWARN(WARN_NUMERIC))
2144 #if defined(USE_LONG_DOUBLE)
2145 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%" PERL_PRIgldbl ")\n",
2146 PTR2UV(sv), SvNVX(sv)));
2148 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"NVgf")\n",
2149 PTR2UV(sv), SvNVX(sv)));
2152 #ifdef NV_PRESERVES_UV
2153 (void)SvIOKp_on(sv);
2155 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2156 SvIV_set(sv, I_V(SvNVX(sv)));
2157 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
2160 NOOP; /* Integer is imprecise. NOK, IOKp */
2162 /* UV will not work better than IV */
2164 if (SvNVX(sv) > (NV)UV_MAX) {
2166 /* Integer is inaccurate. NOK, IOKp, is UV */
2167 SvUV_set(sv, UV_MAX);
2169 SvUV_set(sv, U_V(SvNVX(sv)));
2170 /* 0xFFFFFFFFFFFFFFFF not an issue in here, NVs
2171 NV preservse UV so can do correct comparison. */
2172 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
2175 NOOP; /* Integer is imprecise. NOK, IOKp, is UV */
2180 #else /* NV_PRESERVES_UV */
2181 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2182 == (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT)) {
2183 /* The IV/UV slot will have been set from value returned by
2184 grok_number above. The NV slot has just been set using
2187 assert (SvIOKp(sv));
2189 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2190 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2191 /* Small enough to preserve all bits. */
2192 (void)SvIOKp_on(sv);
2194 SvIV_set(sv, I_V(SvNVX(sv)));
2195 if ((NV)(SvIVX(sv)) == SvNVX(sv))
2197 /* Assumption: first non-preserved integer is < IV_MAX,
2198 this NV is in the preserved range, therefore: */
2199 if (!(U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))
2201 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);
2205 0 0 already failed to read UV.
2206 0 1 already failed to read UV.
2207 1 0 you won't get here in this case. IV/UV
2208 slot set, public IOK, Atof() unneeded.
2209 1 1 already read UV.
2210 so there's no point in sv_2iuv_non_preserve() attempting
2211 to use atol, strtol, strtoul etc. */
2213 sv_2iuv_non_preserve (sv, numtype);
2215 sv_2iuv_non_preserve (sv);
2219 #endif /* NV_PRESERVES_UV */
2220 /* It might be more code efficient to go through the entire logic above
2221 and conditionally set with SvIOKp_on() rather than SvIOK(), but it
2222 gets complex and potentially buggy, so more programmer efficient
2223 to do it this way, by turning off the public flags: */
2225 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK);
2229 if (isGV_with_GP(sv))
2230 return glob_2number(MUTABLE_GV(sv));
2232 if (!SvPADTMP(sv)) {
2233 if (!PL_localizing && ckWARN(WARN_UNINITIALIZED))
2236 if (SvTYPE(sv) < SVt_IV)
2237 /* Typically the caller expects that sv_any is not NULL now. */
2238 sv_upgrade(sv, SVt_IV);
2239 /* Return 0 from the caller. */
2246 =for apidoc sv_2iv_flags
2248 Return the integer value of an SV, doing any necessary string
2249 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2250 Normally used via the C<SvIV(sv)> and C<SvIVx(sv)> macros.
2256 Perl_sv_2iv_flags(pTHX_ register SV *const sv, const I32 flags)
2263 if (SvGMAGICAL(sv) && (flags & SV_GMAGIC))
2269 if (flags & SV_SKIP_OVERLOAD)
2271 tmpstr = AMG_CALLunary(sv, numer_amg);
2272 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2273 return SvIV(tmpstr);
2276 return PTR2IV(SvRV(sv));
2280 /* FBMs use the space for SvIVX and SvNVX for other purposes, and use
2281 the same flag bit as SVf_IVisUV, so must not let them cache IVs.
2282 In practice they are extremely unlikely to actually get anywhere
2283 accessible by user Perl code - the only way that I'm aware of is when
2284 a constant subroutine which is used as the second argument to index.
2289 return I_V(SvNVX(sv));
2290 if (SvPOKp(sv) && SvLEN(sv)) {
2293 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2295 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2296 == IS_NUMBER_IN_UV) {
2297 /* It's definitely an integer */
2298 if (numtype & IS_NUMBER_NEG) {
2299 if (value < (UV)IV_MIN)
2302 if (value < (UV)IV_MAX)
2307 if (ckWARN(WARN_NUMERIC))
2310 return I_V(Atof(SvPVX_const(sv)));
2312 if (ckWARN(WARN_UNINITIALIZED))
2317 if (SvTHINKFIRST(sv)) {
2319 sv_force_normal_flags(sv, 0);
2321 if (SvREADONLY(sv) && !SvOK(sv)) {
2322 if (ckWARN(WARN_UNINITIALIZED))
2329 if (S_sv_2iuv_common(aTHX_ sv))
2333 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"IVdf")\n",
2334 PTR2UV(sv),SvIVX(sv)));
2335 return SvIsUV(sv) ? (IV)SvUVX(sv) : SvIVX(sv);
2339 =for apidoc sv_2uv_flags
2341 Return the unsigned integer value of an SV, doing any necessary string
2342 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2343 Normally used via the C<SvUV(sv)> and C<SvUVx(sv)> macros.
2349 Perl_sv_2uv_flags(pTHX_ register SV *const sv, const I32 flags)
2356 if (SvGMAGICAL(sv) && (flags & SV_GMAGIC))
2362 if (flags & SV_SKIP_OVERLOAD)
2364 tmpstr = AMG_CALLunary(sv, numer_amg);
2365 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2366 return SvUV(tmpstr);
2369 return PTR2UV(SvRV(sv));
2373 /* FBMs use the space for SvIVX and SvNVX for other purposes, and use
2374 the same flag bit as SVf_IVisUV, so must not let them cache IVs. */
2378 return U_V(SvNVX(sv));
2379 if (SvPOKp(sv) && SvLEN(sv)) {
2382 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2384 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2385 == IS_NUMBER_IN_UV) {
2386 /* It's definitely an integer */
2387 if (!(numtype & IS_NUMBER_NEG))
2391 if (ckWARN(WARN_NUMERIC))
2394 return U_V(Atof(SvPVX_const(sv)));
2396 if (ckWARN(WARN_UNINITIALIZED))
2401 if (SvTHINKFIRST(sv)) {
2403 sv_force_normal_flags(sv, 0);
2405 if (SvREADONLY(sv) && !SvOK(sv)) {
2406 if (ckWARN(WARN_UNINITIALIZED))
2413 if (S_sv_2iuv_common(aTHX_ sv))
2417 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2uv(%"UVuf")\n",
2418 PTR2UV(sv),SvUVX(sv)));
2419 return SvIsUV(sv) ? SvUVX(sv) : (UV)SvIVX(sv);
2423 =for apidoc sv_2nv_flags
2425 Return the num value of an SV, doing any necessary string or integer
2426 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2427 Normally used via the C<SvNV(sv)> and C<SvNVx(sv)> macros.
2433 Perl_sv_2nv_flags(pTHX_ register SV *const sv, const I32 flags)
2438 if (SvGMAGICAL(sv) || SvVALID(sv)) {
2439 /* FBMs use the space for SvIVX and SvNVX for other purposes, and use
2440 the same flag bit as SVf_IVisUV, so must not let them cache NVs. */
2441 if (flags & SV_GMAGIC)
2445 if ((SvPOKp(sv) && SvLEN(sv)) && !SvIOKp(sv)) {
2446 if (!SvIOKp(sv) && ckWARN(WARN_NUMERIC) &&
2447 !grok_number(SvPVX_const(sv), SvCUR(sv), NULL))
2449 return Atof(SvPVX_const(sv));
2453 return (NV)SvUVX(sv);
2455 return (NV)SvIVX(sv);
2460 assert(SvTYPE(sv) >= SVt_PVMG);
2461 /* This falls through to the report_uninit near the end of the
2463 } else if (SvTHINKFIRST(sv)) {
2468 if (flags & SV_SKIP_OVERLOAD)
2470 tmpstr = AMG_CALLunary(sv, numer_amg);
2471 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2472 return SvNV(tmpstr);
2475 return PTR2NV(SvRV(sv));
2478 sv_force_normal_flags(sv, 0);
2480 if (SvREADONLY(sv) && !SvOK(sv)) {
2481 if (ckWARN(WARN_UNINITIALIZED))
2486 if (SvTYPE(sv) < SVt_NV) {
2487 /* The logic to use SVt_PVNV if necessary is in sv_upgrade. */
2488 sv_upgrade(sv, SVt_NV);
2489 #ifdef USE_LONG_DOUBLE
2491 STORE_NUMERIC_LOCAL_SET_STANDARD();
2492 PerlIO_printf(Perl_debug_log,
2493 "0x%"UVxf" num(%" PERL_PRIgldbl ")\n",
2494 PTR2UV(sv), SvNVX(sv));
2495 RESTORE_NUMERIC_LOCAL();
2499 STORE_NUMERIC_LOCAL_SET_STANDARD();
2500 PerlIO_printf(Perl_debug_log, "0x%"UVxf" num(%"NVgf")\n",
2501 PTR2UV(sv), SvNVX(sv));
2502 RESTORE_NUMERIC_LOCAL();
2506 else if (SvTYPE(sv) < SVt_PVNV)
2507 sv_upgrade(sv, SVt_PVNV);
2512 SvNV_set(sv, SvIsUV(sv) ? (NV)SvUVX(sv) : (NV)SvIVX(sv));
2513 #ifdef NV_PRESERVES_UV
2519 /* Only set the public NV OK flag if this NV preserves the IV */
2520 /* Check it's not 0xFFFFFFFFFFFFFFFF */
2522 SvIsUV(sv) ? ((SvUVX(sv) != UV_MAX)&&(SvUVX(sv) == U_V(SvNVX(sv))))
2523 : (SvIVX(sv) == I_V(SvNVX(sv))))
2529 else if (SvPOKp(sv) && SvLEN(sv)) {
2531 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2532 if (!SvIOKp(sv) && !numtype && ckWARN(WARN_NUMERIC))
2534 #ifdef NV_PRESERVES_UV
2535 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2536 == IS_NUMBER_IN_UV) {
2537 /* It's definitely an integer */
2538 SvNV_set(sv, (numtype & IS_NUMBER_NEG) ? -(NV)value : (NV)value);
2540 SvNV_set(sv, Atof(SvPVX_const(sv)));
2546 SvNV_set(sv, Atof(SvPVX_const(sv)));
2547 /* Only set the public NV OK flag if this NV preserves the value in
2548 the PV at least as well as an IV/UV would.
2549 Not sure how to do this 100% reliably. */
2550 /* if that shift count is out of range then Configure's test is
2551 wonky. We shouldn't be in here with NV_PRESERVES_UV_BITS ==
2553 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2554 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2555 SvNOK_on(sv); /* Definitely small enough to preserve all bits */
2556 } else if (!(numtype & IS_NUMBER_IN_UV)) {
2557 /* Can't use strtol etc to convert this string, so don't try.
2558 sv_2iv and sv_2uv will use the NV to convert, not the PV. */
2561 /* value has been set. It may not be precise. */
2562 if ((numtype & IS_NUMBER_NEG) && (value > (UV)IV_MIN)) {
2563 /* 2s complement assumption for (UV)IV_MIN */
2564 SvNOK_on(sv); /* Integer is too negative. */
2569 if (numtype & IS_NUMBER_NEG) {
2570 SvIV_set(sv, -(IV)value);
2571 } else if (value <= (UV)IV_MAX) {
2572 SvIV_set(sv, (IV)value);
2574 SvUV_set(sv, value);
2578 if (numtype & IS_NUMBER_NOT_INT) {
2579 /* I believe that even if the original PV had decimals,
2580 they are lost beyond the limit of the FP precision.
2581 However, neither is canonical, so both only get p
2582 flags. NWC, 2000/11/25 */
2583 /* Both already have p flags, so do nothing */
2585 const NV nv = SvNVX(sv);
2586 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2587 if (SvIVX(sv) == I_V(nv)) {
2590 /* It had no "." so it must be integer. */
2594 /* between IV_MAX and NV(UV_MAX).
2595 Could be slightly > UV_MAX */
2597 if (numtype & IS_NUMBER_NOT_INT) {
2598 /* UV and NV both imprecise. */
2600 const UV nv_as_uv = U_V(nv);
2602 if (value == nv_as_uv && SvUVX(sv) != UV_MAX) {
2611 /* It might be more code efficient to go through the entire logic above
2612 and conditionally set with SvNOKp_on() rather than SvNOK(), but it
2613 gets complex and potentially buggy, so more programmer efficient
2614 to do it this way, by turning off the public flags: */
2616 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK);
2617 #endif /* NV_PRESERVES_UV */
2620 if (isGV_with_GP(sv)) {
2621 glob_2number(MUTABLE_GV(sv));
2625 if (!PL_localizing && !SvPADTMP(sv) && ckWARN(WARN_UNINITIALIZED))
2627 assert (SvTYPE(sv) >= SVt_NV);
2628 /* Typically the caller expects that sv_any is not NULL now. */
2629 /* XXX Ilya implies that this is a bug in callers that assume this
2630 and ideally should be fixed. */
2633 #if defined(USE_LONG_DOUBLE)
2635 STORE_NUMERIC_LOCAL_SET_STANDARD();
2636 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2nv(%" PERL_PRIgldbl ")\n",
2637 PTR2UV(sv), SvNVX(sv));
2638 RESTORE_NUMERIC_LOCAL();
2642 STORE_NUMERIC_LOCAL_SET_STANDARD();
2643 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 1nv(%"NVgf")\n",
2644 PTR2UV(sv), SvNVX(sv));
2645 RESTORE_NUMERIC_LOCAL();
2654 Return an SV with the numeric value of the source SV, doing any necessary
2655 reference or overload conversion. You must use the C<SvNUM(sv)> macro to
2656 access this function.
2662 Perl_sv_2num(pTHX_ register SV *const sv)
2664 PERL_ARGS_ASSERT_SV_2NUM;
2669 SV * const tmpsv = AMG_CALLunary(sv, numer_amg);
2670 TAINT_IF(tmpsv && SvTAINTED(tmpsv));
2671 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
2672 return sv_2num(tmpsv);
2674 return sv_2mortal(newSVuv(PTR2UV(SvRV(sv))));
2677 /* uiv_2buf(): private routine for use by sv_2pv_flags(): print an IV or
2678 * UV as a string towards the end of buf, and return pointers to start and
2681 * We assume that buf is at least TYPE_CHARS(UV) long.
2685 S_uiv_2buf(char *const buf, const IV iv, UV uv, const int is_uv, char **const peob)
2687 char *ptr = buf + TYPE_CHARS(UV);
2688 char * const ebuf = ptr;
2691 PERL_ARGS_ASSERT_UIV_2BUF;
2703 *--ptr = '0' + (char)(uv % 10);
2712 =for apidoc sv_2pv_flags
2714 Returns a pointer to the string value of an SV, and sets *lp to its length.
2715 If flags includes SV_GMAGIC, does an mg_get() first. Coerces sv to a
2716 string if necessary. Normally invoked via the C<SvPV_flags> macro.
2717 C<sv_2pv()> and C<sv_2pv_nomg> usually end up here too.
2723 Perl_sv_2pv_flags(pTHX_ register SV *const sv, STRLEN *const lp, const I32 flags)
2733 if (SvGMAGICAL(sv) && (flags & SV_GMAGIC))
2738 if (flags & SV_SKIP_OVERLOAD)
2740 tmpstr = AMG_CALLunary(sv, string_amg);
2741 TAINT_IF(tmpstr && SvTAINTED(tmpstr));
2742 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2744 /* char *pv = lp ? SvPV(tmpstr, *lp) : SvPV_nolen(tmpstr);
2748 if ((SvFLAGS(tmpstr) & (SVf_POK)) == SVf_POK) {
2749 if (flags & SV_CONST_RETURN) {
2750 pv = (char *) SvPVX_const(tmpstr);
2752 pv = (flags & SV_MUTABLE_RETURN)
2753 ? SvPVX_mutable(tmpstr) : SvPVX(tmpstr);
2756 *lp = SvCUR(tmpstr);
2758 pv = sv_2pv_flags(tmpstr, lp, flags);
2771 SV *const referent = SvRV(sv);
2775 retval = buffer = savepvn("NULLREF", len);
2776 } else if (SvTYPE(referent) == SVt_REGEXP &&
2777 (!(PL_curcop->cop_hints & HINT_NO_AMAGIC) ||
2778 amagic_is_enabled(string_amg))) {
2779 REGEXP * const re = (REGEXP *)MUTABLE_PTR(referent);
2783 /* If the regex is UTF-8 we want the containing scalar to
2784 have an UTF-8 flag too */
2791 *lp = RX_WRAPLEN(re);
2793 return RX_WRAPPED(re);
2795 const char *const typestr = sv_reftype(referent, 0);
2796 const STRLEN typelen = strlen(typestr);
2797 UV addr = PTR2UV(referent);
2798 const char *stashname = NULL;
2799 STRLEN stashnamelen = 0; /* hush, gcc */
2800 const char *buffer_end;
2802 if (SvOBJECT(referent)) {
2803 const HEK *const name = HvNAME_HEK(SvSTASH(referent));
2806 stashname = HEK_KEY(name);
2807 stashnamelen = HEK_LEN(name);
2809 if (HEK_UTF8(name)) {
2815 stashname = "__ANON__";
2818 len = stashnamelen + 1 /* = */ + typelen + 3 /* (0x */
2819 + 2 * sizeof(UV) + 2 /* )\0 */;
2821 len = typelen + 3 /* (0x */
2822 + 2 * sizeof(UV) + 2 /* )\0 */;
2825 Newx(buffer, len, char);
2826 buffer_end = retval = buffer + len;
2828 /* Working backwards */
2832 *--retval = PL_hexdigit[addr & 15];
2833 } while (addr >>= 4);
2839 memcpy(retval, typestr, typelen);
2843 retval -= stashnamelen;
2844 memcpy(retval, stashname, stashnamelen);
2846 /* retval may not necessarily have reached the start of the
2848 assert (retval >= buffer);
2850 len = buffer_end - retval - 1; /* -1 for that \0 */
2862 if (flags & SV_MUTABLE_RETURN)
2863 return SvPVX_mutable(sv);
2864 if (flags & SV_CONST_RETURN)
2865 return (char *)SvPVX_const(sv);
2870 /* I'm assuming that if both IV and NV are equally valid then
2871 converting the IV is going to be more efficient */
2872 const U32 isUIOK = SvIsUV(sv);
2873 char buf[TYPE_CHARS(UV)];
2877 if (SvTYPE(sv) < SVt_PVIV)
2878 sv_upgrade(sv, SVt_PVIV);
2879 ptr = uiv_2buf(buf, SvIVX(sv), SvUVX(sv), isUIOK, &ebuf);
2881 /* inlined from sv_setpvn */
2882 s = SvGROW_mutable(sv, len + 1);
2883 Move(ptr, s, len, char);
2887 else if (SvNOK(sv)) {
2888 if (SvTYPE(sv) < SVt_PVNV)
2889 sv_upgrade(sv, SVt_PVNV);
2890 if (SvNVX(sv) == 0.0) {
2891 s = SvGROW_mutable(sv, 2);
2896 /* The +20 is pure guesswork. Configure test needed. --jhi */
2897 s = SvGROW_mutable(sv, NV_DIG + 20);
2898 /* some Xenix systems wipe out errno here */
2899 Gconvert(SvNVX(sv), NV_DIG, 0, s);
2908 else if (isGV_with_GP(sv)) {
2909 GV *const gv = MUTABLE_GV(sv);
2910 SV *const buffer = sv_newmortal();
2912 gv_efullname3(buffer, gv, "*");
2914 assert(SvPOK(buffer));
2918 *lp = SvCUR(buffer);
2919 return SvPVX(buffer);
2924 if (flags & SV_UNDEF_RETURNS_NULL)
2926 if (!PL_localizing && !SvPADTMP(sv) && ckWARN(WARN_UNINITIALIZED))
2928 /* Typically the caller expects that sv_any is not NULL now. */
2929 if (!SvREADONLY(sv) && SvTYPE(sv) < SVt_PV)
2930 sv_upgrade(sv, SVt_PV);
2935 const STRLEN len = s - SvPVX_const(sv);
2941 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
2942 PTR2UV(sv),SvPVX_const(sv)));
2943 if (flags & SV_CONST_RETURN)
2944 return (char *)SvPVX_const(sv);
2945 if (flags & SV_MUTABLE_RETURN)
2946 return SvPVX_mutable(sv);
2951 =for apidoc sv_copypv
2953 Copies a stringified representation of the source SV into the
2954 destination SV. Automatically performs any necessary mg_get and
2955 coercion of numeric values into strings. Guaranteed to preserve
2956 UTF8 flag even from overloaded objects. Similar in nature to
2957 sv_2pv[_flags] but operates directly on an SV instead of just the
2958 string. Mostly uses sv_2pv_flags to do its work, except when that
2959 would lose the UTF-8'ness of the PV.
2961 =for apidoc sv_copypv_nomg
2963 Like sv_copypv, but doesn't invoke get magic first.
2965 =for apidoc sv_copypv_flags
2967 Implementation of sv_copypv and sv_copypv_nomg. Calls get magic iff flags
2974 Perl_sv_copypv(pTHX_ SV *const dsv, register SV *const ssv)
2976 PERL_ARGS_ASSERT_SV_COPYPV;
2978 sv_copypv_flags(dsv, ssv, 0);
2982 Perl_sv_copypv_flags(pTHX_ SV *const dsv, register SV *const ssv, const I32 flags)
2987 PERL_ARGS_ASSERT_SV_COPYPV_FLAGS;
2989 if ((flags & SV_GMAGIC) && SvGMAGICAL(ssv))
2991 s = SvPV_nomg_const(ssv,len);
2992 sv_setpvn(dsv,s,len);
3000 =for apidoc sv_2pvbyte
3002 Return a pointer to the byte-encoded representation of the SV, and set *lp
3003 to its length. May cause the SV to be downgraded from UTF-8 as a
3006 Usually accessed via the C<SvPVbyte> macro.
3012 Perl_sv_2pvbyte(pTHX_ register SV *sv, STRLEN *const lp)
3014 PERL_ARGS_ASSERT_SV_2PVBYTE;
3016 if (((SvREADONLY(sv) || SvFAKE(sv)) && !SvIsCOW(sv))
3017 || isGV_with_GP(sv) || SvROK(sv)) {
3018 SV *sv2 = sv_newmortal();
3022 else SvGETMAGIC(sv);
3023 sv_utf8_downgrade(sv,0);
3024 return lp ? SvPV_nomg(sv,*lp) : SvPV_nomg_nolen(sv);
3028 =for apidoc sv_2pvutf8
3030 Return a pointer to the UTF-8-encoded representation of the SV, and set *lp
3031 to its length. May cause the SV to be upgraded to UTF-8 as a side-effect.
3033 Usually accessed via the C<SvPVutf8> macro.
3039 Perl_sv_2pvutf8(pTHX_ register SV *sv, STRLEN *const lp)
3041 PERL_ARGS_ASSERT_SV_2PVUTF8;
3043 if (((SvREADONLY(sv) || SvFAKE(sv)) && !SvIsCOW(sv))
3044 || isGV_with_GP(sv) || SvROK(sv))
3045 sv = sv_mortalcopy(sv);
3048 sv_utf8_upgrade_nomg(sv);
3049 return lp ? SvPV_nomg(sv,*lp) : SvPV_nomg_nolen(sv);
3054 =for apidoc sv_2bool
3056 This macro is only used by sv_true() or its macro equivalent, and only if
3057 the latter's argument is neither SvPOK, SvIOK nor SvNOK.
3058 It calls sv_2bool_flags with the SV_GMAGIC flag.
3060 =for apidoc sv_2bool_flags
3062 This function is only used by sv_true() and friends, and only if
3063 the latter's argument is neither SvPOK, SvIOK nor SvNOK. If the flags
3064 contain SV_GMAGIC, then it does an mg_get() first.
3071 Perl_sv_2bool_flags(pTHX_ register SV *const sv, const I32 flags)
3075 PERL_ARGS_ASSERT_SV_2BOOL_FLAGS;
3077 if(flags & SV_GMAGIC) SvGETMAGIC(sv);
3083 SV * const tmpsv = AMG_CALLunary(sv, bool__amg);
3084 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
3085 return cBOOL(SvTRUE(tmpsv));
3087 return SvRV(sv) != 0;
3089 return SvTRUE_common(sv, isGV_with_GP(sv) ? 1 : 0);
3093 =for apidoc sv_utf8_upgrade
3095 Converts the PV of an SV to its UTF-8-encoded form.
3096 Forces the SV to string form if it is not already.
3097 Will C<mg_get> on C<sv> if appropriate.
3098 Always sets the SvUTF8 flag to avoid future validity checks even
3099 if the whole string is the same in UTF-8 as not.
3100 Returns the number of bytes in the converted string
3102 This is not as a general purpose byte encoding to Unicode interface:
3103 use the Encode extension for that.
3105 =for apidoc sv_utf8_upgrade_nomg
3107 Like sv_utf8_upgrade, but doesn't do magic on C<sv>.
3109 =for apidoc sv_utf8_upgrade_flags
3111 Converts the PV of an SV to its UTF-8-encoded form.
3112 Forces the SV to string form if it is not already.
3113 Always sets the SvUTF8 flag to avoid future validity checks even
3114 if all the bytes are invariant in UTF-8.
3115 If C<flags> has C<SV_GMAGIC> bit set,
3116 will C<mg_get> on C<sv> if appropriate, else not.
3117 Returns the number of bytes in the converted string
3118 C<sv_utf8_upgrade> and
3119 C<sv_utf8_upgrade_nomg> are implemented in terms of this function.
3121 This is not as a general purpose byte encoding to Unicode interface:
3122 use the Encode extension for that.
3126 The grow version is currently not externally documented. It adds a parameter,
3127 extra, which is the number of unused bytes the string of 'sv' is guaranteed to
3128 have free after it upon return. This allows the caller to reserve extra space
3129 that it intends to fill, to avoid extra grows.
3131 Also externally undocumented for the moment is the flag SV_FORCE_UTF8_UPGRADE,
3132 which can be used to tell this function to not first check to see if there are
3133 any characters that are different in UTF-8 (variant characters) which would
3134 force it to allocate a new string to sv, but to assume there are. Typically
3135 this flag is used by a routine that has already parsed the string to find that
3136 there are such characters, and passes this information on so that the work
3137 doesn't have to be repeated.
3139 (One might think that the calling routine could pass in the position of the
3140 first such variant, so it wouldn't have to be found again. But that is not the
3141 case, because typically when the caller is likely to use this flag, it won't be
3142 calling this routine unless it finds something that won't fit into a byte.
3143 Otherwise it tries to not upgrade and just use bytes. But some things that
3144 do fit into a byte are variants in utf8, and the caller may not have been
3145 keeping track of these.)
3147 If the routine itself changes the string, it adds a trailing NUL. Such a NUL
3148 isn't guaranteed due to having other routines do the work in some input cases,
3149 or if the input is already flagged as being in utf8.
3151 The speed of this could perhaps be improved for many cases if someone wanted to
3152 write a fast function that counts the number of variant characters in a string,
3153 especially if it could return the position of the first one.
3158 Perl_sv_utf8_upgrade_flags_grow(pTHX_ register SV *const sv, const I32 flags, STRLEN extra)
3162 PERL_ARGS_ASSERT_SV_UTF8_UPGRADE_FLAGS_GROW;
3164 if (sv == &PL_sv_undef)
3168 if (SvREADONLY(sv) && (SvPOKp(sv) || SvIOKp(sv) || SvNOKp(sv))) {
3169 (void) sv_2pv_flags(sv,&len, flags);
3171 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3175 (void) SvPV_force_flags(sv,len,flags & SV_GMAGIC);
3180 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3185 sv_force_normal_flags(sv, 0);
3188 if (PL_encoding && !(flags & SV_UTF8_NO_ENCODING)) {
3189 sv_recode_to_utf8(sv, PL_encoding);
3190 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3194 if (SvCUR(sv) == 0) {
3195 if (extra) SvGROW(sv, extra);
3196 } else { /* Assume Latin-1/EBCDIC */
3197 /* This function could be much more efficient if we
3198 * had a FLAG in SVs to signal if there are any variant
3199 * chars in the PV. Given that there isn't such a flag
3200 * make the loop as fast as possible (although there are certainly ways
3201 * to speed this up, eg. through vectorization) */
3202 U8 * s = (U8 *) SvPVX_const(sv);
3203 U8 * e = (U8 *) SvEND(sv);
3205 STRLEN two_byte_count = 0;
3207 if (flags & SV_FORCE_UTF8_UPGRADE) goto must_be_utf8;
3209 /* See if really will need to convert to utf8. We mustn't rely on our
3210 * incoming SV being well formed and having a trailing '\0', as certain
3211 * code in pp_formline can send us partially built SVs. */
3215 if (NATIVE_IS_INVARIANT(ch)) continue;
3217 t--; /* t already incremented; re-point to first variant */
3222 /* utf8 conversion not needed because all are invariants. Mark as
3223 * UTF-8 even if no variant - saves scanning loop */
3225 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3230 /* Here, the string should be converted to utf8, either because of an
3231 * input flag (two_byte_count = 0), or because a character that
3232 * requires 2 bytes was found (two_byte_count = 1). t points either to
3233 * the beginning of the string (if we didn't examine anything), or to
3234 * the first variant. In either case, everything from s to t - 1 will
3235 * occupy only 1 byte each on output.
3237 * There are two main ways to convert. One is to create a new string
3238 * and go through the input starting from the beginning, appending each
3239 * converted value onto the new string as we go along. It's probably
3240 * best to allocate enough space in the string for the worst possible
3241 * case rather than possibly running out of space and having to
3242 * reallocate and then copy what we've done so far. Since everything
3243 * from s to t - 1 is invariant, the destination can be initialized
3244 * with these using a fast memory copy
3246 * The other way is to figure out exactly how big the string should be
3247 * by parsing the entire input. Then you don't have to make it big
3248 * enough to handle the worst possible case, and more importantly, if
3249 * the string you already have is large enough, you don't have to
3250 * allocate a new string, you can copy the last character in the input
3251 * string to the final position(s) that will be occupied by the
3252 * converted string and go backwards, stopping at t, since everything
3253 * before that is invariant.
3255 * There are advantages and disadvantages to each method.
3257 * In the first method, we can allocate a new string, do the memory
3258 * copy from the s to t - 1, and then proceed through the rest of the
3259 * string byte-by-byte.
3261 * In the second method, we proceed through the rest of the input
3262 * string just calculating how big the converted string will be. Then
3263 * there are two cases:
3264 * 1) if the string has enough extra space to handle the converted
3265 * value. We go backwards through the string, converting until we
3266 * get to the position we are at now, and then stop. If this
3267 * position is far enough along in the string, this method is
3268 * faster than the other method. If the memory copy were the same
3269 * speed as the byte-by-byte loop, that position would be about
3270 * half-way, as at the half-way mark, parsing to the end and back
3271 * is one complete string's parse, the same amount as starting
3272 * over and going all the way through. Actually, it would be
3273 * somewhat less than half-way, as it's faster to just count bytes
3274 * than to also copy, and we don't have the overhead of allocating
3275 * a new string, changing the scalar to use it, and freeing the
3276 * existing one. But if the memory copy is fast, the break-even
3277 * point is somewhere after half way. The counting loop could be
3278 * sped up by vectorization, etc, to move the break-even point
3279 * further towards the beginning.
3280 * 2) if the string doesn't have enough space to handle the converted
3281 * value. A new string will have to be allocated, and one might
3282 * as well, given that, start from the beginning doing the first
3283 * method. We've spent extra time parsing the string and in
3284 * exchange all we've gotten is that we know precisely how big to
3285 * make the new one. Perl is more optimized for time than space,
3286 * so this case is a loser.
3287 * So what I've decided to do is not use the 2nd method unless it is
3288 * guaranteed that a new string won't have to be allocated, assuming
3289 * the worst case. I also decided not to put any more conditions on it
3290 * than this, for now. It seems likely that, since the worst case is
3291 * twice as big as the unknown portion of the string (plus 1), we won't
3292 * be guaranteed enough space, causing us to go to the first method,
3293 * unless the string is short, or the first variant character is near
3294 * the end of it. In either of these cases, it seems best to use the
3295 * 2nd method. The only circumstance I can think of where this would
3296 * be really slower is if the string had once had much more data in it
3297 * than it does now, but there is still a substantial amount in it */
3300 STRLEN invariant_head = t - s;
3301 STRLEN size = invariant_head + (e - t) * 2 + 1 + extra;
3302 if (SvLEN(sv) < size) {
3304 /* Here, have decided to allocate a new string */
3309 Newx(dst, size, U8);
3311 /* If no known invariants at the beginning of the input string,
3312 * set so starts from there. Otherwise, can use memory copy to
3313 * get up to where we are now, and then start from here */
3315 if (invariant_head <= 0) {
3318 Copy(s, dst, invariant_head, char);
3319 d = dst + invariant_head;
3323 const UV uv = NATIVE8_TO_UNI(*t++);
3324 if (UNI_IS_INVARIANT(uv))
3325 *d++ = (U8)UNI_TO_NATIVE(uv);
3327 *d++ = (U8)UTF8_EIGHT_BIT_HI(uv);
3328 *d++ = (U8)UTF8_EIGHT_BIT_LO(uv);
3332 SvPV_free(sv); /* No longer using pre-existing string */
3333 SvPV_set(sv, (char*)dst);
3334 SvCUR_set(sv, d - dst);
3335 SvLEN_set(sv, size);
3338 /* Here, have decided to get the exact size of the string.
3339 * Currently this happens only when we know that there is
3340 * guaranteed enough space to fit the converted string, so
3341 * don't have to worry about growing. If two_byte_count is 0,
3342 * then t points to the first byte of the string which hasn't
3343 * been examined yet. Otherwise two_byte_count is 1, and t
3344 * points to the first byte in the string that will expand to
3345 * two. Depending on this, start examining at t or 1 after t.
3348 U8 *d = t + two_byte_count;
3351 /* Count up the remaining bytes that expand to two */
3354 const U8 chr = *d++;
3355 if (! NATIVE_IS_INVARIANT(chr)) two_byte_count++;
3358 /* The string will expand by just the number of bytes that
3359 * occupy two positions. But we are one afterwards because of
3360 * the increment just above. This is the place to put the
3361 * trailing NUL, and to set the length before we decrement */
3363 d += two_byte_count;
3364 SvCUR_set(sv, d - s);
3368 /* Having decremented d, it points to the position to put the
3369 * very last byte of the expanded string. Go backwards through
3370 * the string, copying and expanding as we go, stopping when we
3371 * get to the part that is invariant the rest of the way down */
3375 const U8 ch = NATIVE8_TO_UNI(*e--);
3376 if (UNI_IS_INVARIANT(ch)) {
3377 *d-- = UNI_TO_NATIVE(ch);
3379 *d-- = (U8)UTF8_EIGHT_BIT_LO(ch);
3380 *d-- = (U8)UTF8_EIGHT_BIT_HI(ch);
3385 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
3386 /* Update pos. We do it at the end rather than during
3387 * the upgrade, to avoid slowing down the common case
3388 * (upgrade without pos) */
3389 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
3391 I32 pos = mg->mg_len;
3392 if (pos > 0 && (U32)pos > invariant_head) {
3393 U8 *d = (U8*) SvPVX(sv) + invariant_head;
3394 STRLEN n = (U32)pos - invariant_head;
3396 if (UTF8_IS_START(*d))
3401 mg->mg_len = d - (U8*)SvPVX(sv);
3404 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
3405 magic_setutf8(sv,mg); /* clear UTF8 cache */
3410 /* Mark as UTF-8 even if no variant - saves scanning loop */
3416 =for apidoc sv_utf8_downgrade
3418 Attempts to convert the PV of an SV from characters to bytes.
3419 If the PV contains a character that cannot fit
3420 in a byte, this conversion will fail;
3421 in this case, either returns false or, if C<fail_ok> is not
3424 This is not as a general purpose Unicode to byte encoding interface:
3425 use the Encode extension for that.
3431 Perl_sv_utf8_downgrade(pTHX_ register SV *const sv, const bool fail_ok)
3435 PERL_ARGS_ASSERT_SV_UTF8_DOWNGRADE;
3437 if (SvPOKp(sv) && SvUTF8(sv)) {
3441 int mg_flags = SV_GMAGIC;
3444 sv_force_normal_flags(sv, 0);
3446 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
3448 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
3450 I32 pos = mg->mg_len;
3452 sv_pos_b2u(sv, &pos);
3453 mg_flags = 0; /* sv_pos_b2u does get magic */
3457 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
3458 magic_setutf8(sv,mg); /* clear UTF8 cache */
3461 s = (U8 *) SvPV_flags(sv, len, mg_flags);
3463 if (!utf8_to_bytes(s, &len)) {
3468 Perl_croak(aTHX_ "Wide character in %s",
3471 Perl_croak(aTHX_ "Wide character");
3482 =for apidoc sv_utf8_encode
3484 Converts the PV of an SV to UTF-8, but then turns the C<SvUTF8>
3485 flag off so that it looks like octets again.
3491 Perl_sv_utf8_encode(pTHX_ register SV *const sv)
3493 PERL_ARGS_ASSERT_SV_UTF8_ENCODE;
3495 if (SvREADONLY(sv)) {
3496 sv_force_normal_flags(sv, 0);
3498 (void) sv_utf8_upgrade(sv);
3503 =for apidoc sv_utf8_decode
3505 If the PV of the SV is an octet sequence in UTF-8
3506 and contains a multiple-byte character, the C<SvUTF8> flag is turned on
3507 so that it looks like a character. If the PV contains only single-byte
3508 characters, the C<SvUTF8> flag stays off.
3509 Scans PV for validity and returns false if the PV is invalid UTF-8.
3515 Perl_sv_utf8_decode(pTHX_ register SV *const sv)
3517 PERL_ARGS_ASSERT_SV_UTF8_DECODE;
3520 const U8 *start, *c;
3523 /* The octets may have got themselves encoded - get them back as
3526 if (!sv_utf8_downgrade(sv, TRUE))
3529 /* it is actually just a matter of turning the utf8 flag on, but
3530 * we want to make sure everything inside is valid utf8 first.
3532 c = start = (const U8 *) SvPVX_const(sv);
3533 if (!is_utf8_string(c, SvCUR(sv)))
3535 e = (const U8 *) SvEND(sv);
3538 if (!UTF8_IS_INVARIANT(ch)) {
3543 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
3544 /* adjust pos to the start of a UTF8 char sequence */
3545 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
3547 I32 pos = mg->mg_len;
3549 for (c = start + pos; c > start; c--) {
3550 if (UTF8_IS_START(*c))
3553 mg->mg_len = c - start;
3556 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
3557 magic_setutf8(sv,mg); /* clear UTF8 cache */
3564 =for apidoc sv_setsv
3566 Copies the contents of the source SV C<ssv> into the destination SV
3567 C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3568 function if the source SV needs to be reused. Does not handle 'set' magic.
3569 Loosely speaking, it performs a copy-by-value, obliterating any previous
3570 content of the destination.
3572 You probably want to use one of the assortment of wrappers, such as
3573 C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3574 C<SvSetMagicSV_nosteal>.
3576 =for apidoc sv_setsv_flags
3578 Copies the contents of the source SV C<ssv> into the destination SV
3579 C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3580 function if the source SV needs to be reused. Does not handle 'set' magic.
3581 Loosely speaking, it performs a copy-by-value, obliterating any previous
3582 content of the destination.
3583 If the C<flags> parameter has the C<SV_GMAGIC> bit set, will C<mg_get> on
3584 C<ssv> if appropriate, else not. If the C<flags>
3585 parameter has the C<NOSTEAL> bit set then the
3586 buffers of temps will not be stolen. <sv_setsv>
3587 and C<sv_setsv_nomg> are implemented in terms of this function.
3589 You probably want to use one of the assortment of wrappers, such as
3590 C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3591 C<SvSetMagicSV_nosteal>.
3593 This is the primary function for copying scalars, and most other
3594 copy-ish functions and macros use this underneath.
3600 S_glob_assign_glob(pTHX_ SV *const dstr, SV *const sstr, const int dtype)
3602 I32 mro_changes = 0; /* 1 = method, 2 = isa, 3 = recursive isa */
3603 HV *old_stash = NULL;
3605 PERL_ARGS_ASSERT_GLOB_ASSIGN_GLOB;
3607 if (dtype != SVt_PVGV && !isGV_with_GP(dstr)) {
3608 const char * const name = GvNAME(sstr);
3609 const STRLEN len = GvNAMELEN(sstr);
3611 if (dtype >= SVt_PV) {
3617 SvUPGRADE(dstr, SVt_PVGV);
3618 (void)SvOK_off(dstr);
3619 /* We have to turn this on here, even though we turn it off
3620 below, as GvSTASH will fail an assertion otherwise. */
3621 isGV_with_GP_on(dstr);
3623 GvSTASH(dstr) = GvSTASH(sstr);
3625 Perl_sv_add_backref(aTHX_ MUTABLE_SV(GvSTASH(dstr)), dstr);
3626 gv_name_set(MUTABLE_GV(dstr), name, len,
3627 GV_ADD | (GvNAMEUTF8(sstr) ? SVf_UTF8 : 0 ));
3628 SvFAKE_on(dstr); /* can coerce to non-glob */
3631 if(GvGP(MUTABLE_GV(sstr))) {
3632 /* If source has method cache entry, clear it */
3634 SvREFCNT_dec(GvCV(sstr));
3635 GvCV_set(sstr, NULL);
3638 /* If source has a real method, then a method is
3641 GvCV((const GV *)sstr) && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3647 /* If dest already had a real method, that's a change as well */
3649 !mro_changes && GvGP(MUTABLE_GV(dstr)) && GvCVu((const GV *)dstr)
3650 && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3655 /* We don't need to check the name of the destination if it was not a
3656 glob to begin with. */
3657 if(dtype == SVt_PVGV) {
3658 const char * const name = GvNAME((const GV *)dstr);
3661 /* The stash may have been detached from the symbol table, so
3663 && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3667 const STRLEN len = GvNAMELEN(dstr);
3668 if ((len > 1 && name[len-2] == ':' && name[len-1] == ':')
3669 || (len == 1 && name[0] == ':')) {
3672 /* Set aside the old stash, so we can reset isa caches on
3674 if((old_stash = GvHV(dstr)))
3675 /* Make sure we do not lose it early. */
3676 SvREFCNT_inc_simple_void_NN(
3677 sv_2mortal((SV *)old_stash)
3683 gp_free(MUTABLE_GV(dstr));
3684 isGV_with_GP_off(dstr); /* SvOK_off does not like globs. */
3685 (void)SvOK_off(dstr);
3686 isGV_with_GP_on(dstr);
3687 GvINTRO_off(dstr); /* one-shot flag */
3688 GvGP_set(dstr, gp_ref(GvGP(sstr)));
3689 if (SvTAINTED(sstr))
3691 if (GvIMPORTED(dstr) != GVf_IMPORTED
3692 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3694 GvIMPORTED_on(dstr);
3697 if(mro_changes == 2) {
3698 if (GvAV((const GV *)sstr)) {
3700 SV * const sref = (SV *)GvAV((const GV *)dstr);
3701 if (SvSMAGICAL(sref) && (mg = mg_find(sref, PERL_MAGIC_isa))) {
3702 if (SvTYPE(mg->mg_obj) != SVt_PVAV) {
3703 AV * const ary = newAV();
3704 av_push(ary, mg->mg_obj); /* takes the refcount */
3705 mg->mg_obj = (SV *)ary;
3707 av_push((AV *)mg->mg_obj, SvREFCNT_inc_simple_NN(dstr));
3709 else sv_magic(sref, dstr, PERL_MAGIC_isa, NULL, 0);
3711 mro_isa_changed_in(GvSTASH(dstr));
3713 else if(mro_changes == 3) {
3714 HV * const stash = GvHV(dstr);
3715 if(old_stash ? (HV *)HvENAME_get(old_stash) : stash)
3721 else if(mro_changes) mro_method_changed_in(GvSTASH(dstr));
3726 S_glob_assign_ref(pTHX_ SV *const dstr, SV *const sstr)
3728 SV * const sref = SvREFCNT_inc(SvRV(sstr));
3730 const int intro = GvINTRO(dstr);
3733 const U32 stype = SvTYPE(sref);
3735 PERL_ARGS_ASSERT_GLOB_ASSIGN_REF;
3738 GvINTRO_off(dstr); /* one-shot flag */
3739 GvLINE(dstr) = CopLINE(PL_curcop);
3740 GvEGV(dstr) = MUTABLE_GV(dstr);
3745 location = (SV **) &(GvGP(dstr)->gp_cv); /* XXX bypassing GvCV_set */
3746 import_flag = GVf_IMPORTED_CV;
3749 location = (SV **) &GvHV(dstr);
3750 import_flag = GVf_IMPORTED_HV;
3753 location = (SV **) &GvAV(dstr);
3754 import_flag = GVf_IMPORTED_AV;
3757 location = (SV **) &GvIOp(dstr);
3760 location = (SV **) &GvFORM(dstr);
3763 location = &GvSV(dstr);
3764 import_flag = GVf_IMPORTED_SV;
3767 if (stype == SVt_PVCV) {
3768 /*if (GvCVGEN(dstr) && (GvCV(dstr) != (const CV *)sref || GvCVGEN(dstr))) {*/
3769 if (GvCVGEN(dstr)) {
3770 SvREFCNT_dec(GvCV(dstr));
3771 GvCV_set(dstr, NULL);
3772 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3775 SAVEGENERICSV(*location);
3779 if (stype == SVt_PVCV && (*location != sref || GvCVGEN(dstr))) {
3780 CV* const cv = MUTABLE_CV(*location);
3782 if (!GvCVGEN((const GV *)dstr) &&
3783 (CvROOT(cv) || CvXSUB(cv)) &&
3784 /* redundant check that avoids creating the extra SV
3785 most of the time: */
3786 (CvCONST(cv) || ckWARN(WARN_REDEFINE)))
3788 SV * const new_const_sv =
3789 CvCONST((const CV *)sref)
3790 ? cv_const_sv((const CV *)sref)
3792 report_redefined_cv(
3793 sv_2mortal(Perl_newSVpvf(aTHX_
3796 HvNAME_HEK(GvSTASH((const GV *)dstr))
3798 HEKfARG(GvENAME_HEK(MUTABLE_GV(dstr)))
3801 CvCONST((const CV *)sref) ? &new_const_sv : NULL
3805 cv_ckproto_len_flags(cv, (const GV *)dstr,
3806 SvPOK(sref) ? CvPROTO(sref) : NULL,
3807 SvPOK(sref) ? CvPROTOLEN(sref) : 0,
3808 SvPOK(sref) ? SvUTF8(sref) : 0);
3810 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3811 GvASSUMECV_on(dstr);
3812 if(GvSTASH(dstr)) mro_method_changed_in(GvSTASH(dstr)); /* sub foo { 1 } sub bar { 2 } *bar = \&foo */
3815 if (import_flag && !(GvFLAGS(dstr) & import_flag)
3816 && CopSTASH_ne(PL_curcop, GvSTASH(dstr))) {
3817 GvFLAGS(dstr) |= import_flag;
3819 if (stype == SVt_PVHV) {
3820 const char * const name = GvNAME((GV*)dstr);
3821 const STRLEN len = GvNAMELEN(dstr);
3824 (len > 1 && name[len-2] == ':' && name[len-1] == ':')
3825 || (len == 1 && name[0] == ':')
3827 && (!dref || HvENAME_get(dref))
3830 (HV *)sref, (HV *)dref,
3836 stype == SVt_PVAV && sref != dref
3837 && strEQ(GvNAME((GV*)dstr), "ISA")
3838 /* The stash may have been detached from the symbol table, so
3839 check its name before doing anything. */
3840 && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3843 MAGIC * const omg = dref && SvSMAGICAL(dref)
3844 ? mg_find(dref, PERL_MAGIC_isa)
3846 if (SvSMAGICAL(sref) && (mg = mg_find(sref, PERL_MAGIC_isa))) {
3847 if (SvTYPE(mg->mg_obj) != SVt_PVAV) {
3848 AV * const ary = newAV();
3849 av_push(ary, mg->mg_obj); /* takes the refcount */
3850 mg->mg_obj = (SV *)ary;
3853 if (SvTYPE(omg->mg_obj) == SVt_PVAV) {
3854 SV **svp = AvARRAY((AV *)omg->mg_obj);
3855 I32 items = AvFILLp((AV *)omg->mg_obj) + 1;
3859 SvREFCNT_inc_simple_NN(*svp++)
3865 SvREFCNT_inc_simple_NN(omg->mg_obj)
3869 av_push((AV *)mg->mg_obj,SvREFCNT_inc_simple_NN(dstr));
3874 sref, omg ? omg->mg_obj : dstr, PERL_MAGIC_isa, NULL, 0
3876 mg = mg_find(sref, PERL_MAGIC_isa);
3878 /* Since the *ISA assignment could have affected more than
3879 one stash, don't call mro_isa_changed_in directly, but let
3880 magic_clearisa do it for us, as it already has the logic for
3881 dealing with globs vs arrays of globs. */
3883 Perl_magic_clearisa(aTHX_ NULL, mg);
3888 if (SvTAINTED(sstr))
3894 Perl_sv_setsv_flags(pTHX_ SV *dstr, register SV* sstr, const I32 flags)
3901 PERL_ARGS_ASSERT_SV_SETSV_FLAGS;
3906 if (SvIS_FREED(dstr)) {
3907 Perl_croak(aTHX_ "panic: attempt to copy value %" SVf
3908 " to a freed scalar %p", SVfARG(sstr), (void *)dstr);
3910 SV_CHECK_THINKFIRST_COW_DROP(dstr);
3912 sstr = &PL_sv_undef;
3913 if (SvIS_FREED(sstr)) {
3914 Perl_croak(aTHX_ "panic: attempt to copy freed scalar %p to %p",
3915 (void*)sstr, (void*)dstr);
3917 stype = SvTYPE(sstr);
3918 dtype = SvTYPE(dstr);
3920 /* There's a lot of redundancy below but we're going for speed here */
3925 if (dtype != SVt_PVGV && dtype != SVt_PVLV) {
3926 (void)SvOK_off(dstr);
3934 sv_upgrade(dstr, SVt_IV);
3938 sv_upgrade(dstr, SVt_PVIV);
3942 goto end_of_first_switch;
3944 (void)SvIOK_only(dstr);
3945 SvIV_set(dstr, SvIVX(sstr));
3948 /* SvTAINTED can only be true if the SV has taint magic, which in
3949 turn means that the SV type is PVMG (or greater). This is the
3950 case statement for SVt_IV, so this cannot be true (whatever gcov
3952 assert(!SvTAINTED(sstr));
3957 if (dtype < SVt_PV && dtype != SVt_IV)
3958 sv_upgrade(dstr, SVt_IV);
3966 sv_upgrade(dstr, SVt_NV);
3970 sv_upgrade(dstr, SVt_PVNV);
3974 goto end_of_first_switch;
3976 SvNV_set(dstr, SvNVX(sstr));
3977 (void)SvNOK_only(dstr);
3978 /* SvTAINTED can only be true if the SV has taint magic, which in
3979 turn means that the SV type is PVMG (or greater). This is the
3980 case statement for SVt_NV, so this cannot be true (whatever gcov
3982 assert(!SvTAINTED(sstr));
3989 sv_upgrade(dstr, SVt_PV);
3992 if (dtype < SVt_PVIV)
3993 sv_upgrade(dstr, SVt_PVIV);
3996 if (dtype < SVt_PVNV)
3997 sv_upgrade(dstr, SVt_PVNV);
4001 const char * const type = sv_reftype(sstr,0);
4003 /* diag_listed_as: Bizarre copy of %s */
4004 Perl_croak(aTHX_ "Bizarre copy of %s in %s", type, OP_DESC(PL_op));
4006 Perl_croak(aTHX_ "Bizarre copy of %s", type);
4011 if (dtype < SVt_REGEXP)
4012 sv_upgrade(dstr, SVt_REGEXP);
4015 /* case SVt_BIND: */
4019 if (SvGMAGICAL(sstr) && (flags & SV_GMAGIC)) {
4021 if (SvTYPE(sstr) != stype)
4022 stype = SvTYPE(sstr);
4024 if (isGV_with_GP(sstr) && dtype <= SVt_PVLV) {
4025 glob_assign_glob(dstr, sstr, dtype);
4028 if (stype == SVt_PVLV)
4029 SvUPGRADE(dstr, SVt_PVNV);
4031 SvUPGRADE(dstr, (svtype)stype);
4033 end_of_first_switch:
4035 /* dstr may have been upgraded. */
4036 dtype = SvTYPE(dstr);
4037 sflags = SvFLAGS(sstr);
4039 if (dtype == SVt_PVCV) {
4040 /* Assigning to a subroutine sets the prototype. */
4043 const char *const ptr = SvPV_const(sstr, len);
4045 SvGROW(dstr, len + 1);
4046 Copy(ptr, SvPVX(dstr), len + 1, char);
4047 SvCUR_set(dstr, len);
4049 SvFLAGS(dstr) |= sflags & SVf_UTF8;
4050 CvAUTOLOAD_off(dstr);
4055 else if (dtype == SVt_PVAV || dtype == SVt_PVHV || dtype == SVt_PVFM) {
4056 const char * const type = sv_reftype(dstr,0);
4058 /* diag_listed_as: Cannot copy to %s */
4059 Perl_croak(aTHX_ "Cannot copy to %s in %s", type, OP_DESC(PL_op));
4061 Perl_croak(aTHX_ "Cannot copy to %s", type);
4062 } else if (sflags & SVf_ROK) {
4063 if (isGV_with_GP(dstr)
4064 && SvTYPE(SvRV(sstr)) == SVt_PVGV && isGV_with_GP(SvRV(sstr))) {
4067 if (GvIMPORTED(dstr) != GVf_IMPORTED
4068 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
4070 GvIMPORTED_on(dstr);
4075 glob_assign_glob(dstr, sstr, dtype);
4079 if (dtype >= SVt_PV) {
4080 if (isGV_with_GP(dstr)) {
4081 glob_assign_ref(dstr, sstr);
4084 if (SvPVX_const(dstr)) {
4090 (void)SvOK_off(dstr);
4091 SvRV_set(dstr, SvREFCNT_inc(SvRV(sstr)));
4092 SvFLAGS(dstr) |= sflags & SVf_ROK;
4093 assert(!(sflags & SVp_NOK));
4094 assert(!(sflags & SVp_IOK));
4095 assert(!(sflags & SVf_NOK));
4096 assert(!(sflags & SVf_IOK));
4098 else if (isGV_with_GP(dstr)) {
4099 if (!(sflags & SVf_OK)) {
4100 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
4101 "Undefined value assigned to typeglob");
4104 GV *gv = gv_fetchsv_nomg(sstr, GV_ADD, SVt_PVGV);
4105 if (dstr != (const SV *)gv) {
4106 const char * const name = GvNAME((const GV *)dstr);
4107 const STRLEN len = GvNAMELEN(dstr);
4108 HV *old_stash = NULL;
4109 bool reset_isa = FALSE;
4110 if ((len > 1 && name[len-2] == ':' && name[len-1] == ':')
4111 || (len == 1 && name[0] == ':')) {
4112 /* Set aside the old stash, so we can reset isa caches
4113 on its subclasses. */
4114 if((old_stash = GvHV(dstr))) {
4115 /* Make sure we do not lose it early. */
4116 SvREFCNT_inc_simple_void_NN(
4117 sv_2mortal((SV *)old_stash)
4124 gp_free(MUTABLE_GV(dstr));
4125 GvGP_set(dstr, gp_ref(GvGP(gv)));
4128 HV * const stash = GvHV(dstr);
4130 old_stash ? (HV *)HvENAME_get(old_stash) : stash
4140 else if (dtype == SVt_REGEXP && stype == SVt_REGEXP) {
4141 reg_temp_copy((REGEXP*)dstr, (REGEXP*)sstr);
4143 else if (sflags & SVp_POK) {
4147 * Check to see if we can just swipe the string. If so, it's a
4148 * possible small lose on short strings, but a big win on long ones.
4149 * It might even be a win on short strings if SvPVX_const(dstr)
4150 * has to be allocated and SvPVX_const(sstr) has to be freed.
4151 * Likewise if we can set up COW rather than doing an actual copy, we
4152 * drop to the else clause, as the swipe code and the COW setup code
4153 * have much in common.
4156 /* Whichever path we take through the next code, we want this true,
4157 and doing it now facilitates the COW check. */
4158 (void)SvPOK_only(dstr);
4161 /* If we're already COW then this clause is not true, and if COW
4162 is allowed then we drop down to the else and make dest COW
4163 with us. If caller hasn't said that we're allowed to COW
4164 shared hash keys then we don't do the COW setup, even if the
4165 source scalar is a shared hash key scalar. */
4166 (((flags & SV_COW_SHARED_HASH_KEYS)
4167 ? (sflags & (SVf_FAKE|SVf_READONLY)) != (SVf_FAKE|SVf_READONLY)
4168 : 1 /* If making a COW copy is forbidden then the behaviour we
4169 desire is as if the source SV isn't actually already
4170 COW, even if it is. So we act as if the source flags
4171 are not COW, rather than actually testing them. */
4173 #ifndef PERL_OLD_COPY_ON_WRITE
4174 /* The change that added SV_COW_SHARED_HASH_KEYS makes the logic
4175 when PERL_OLD_COPY_ON_WRITE is defined a little wrong.
4176 Conceptually PERL_OLD_COPY_ON_WRITE being defined should
4177 override SV_COW_SHARED_HASH_KEYS, because it means "always COW"
4178 but in turn, it's somewhat dead code, never expected to go
4179 live, but more kept as a placeholder on how to do it better
4180 in a newer implementation. */
4181 /* If we are COW and dstr is a suitable target then we drop down
4182 into the else and make dest a COW of us. */
4183 || (SvFLAGS(dstr) & CAN_COW_MASK) != CAN_COW_FLAGS
4188 (sflags & SVs_TEMP) && /* slated for free anyway? */
4189 !(sflags & SVf_OOK) && /* and not involved in OOK hack? */
4190 (!(flags & SV_NOSTEAL)) &&
4191 /* and we're allowed to steal temps */
4192 SvREFCNT(sstr) == 1 && /* and no other references to it? */
4193 SvLEN(sstr)) /* and really is a string */
4194 #ifdef PERL_OLD_COPY_ON_WRITE
4195 && ((flags & SV_COW_SHARED_HASH_KEYS)
4196 ? (!((sflags & CAN_COW_MASK) == CAN_COW_FLAGS
4197 && (SvFLAGS(dstr) & CAN_COW_MASK) == CAN_COW_FLAGS
4198 && SvTYPE(sstr) >= SVt_PVIV))
4202 /* Failed the swipe test, and it's not a shared hash key either.
4203 Have to copy the string. */
4204 STRLEN len = SvCUR(sstr);
4205 SvGROW(dstr, len + 1); /* inlined from sv_setpvn */
4206 Move(SvPVX_const(sstr),SvPVX(dstr),len,char);
4207 SvCUR_set(dstr, len);
4208 *SvEND(dstr) = '\0';
4210 /* If PERL_OLD_COPY_ON_WRITE is not defined, then isSwipe will always
4212 /* Either it's a shared hash key, or it's suitable for
4213 copy-on-write or we can swipe the string. */
4215 PerlIO_printf(Perl_debug_log, "Copy on write: sstr --> dstr\n");
4219 #ifdef PERL_OLD_COPY_ON_WRITE
4221 if ((sflags & (SVf_FAKE | SVf_READONLY))
4222 != (SVf_FAKE | SVf_READONLY)) {
4223 SvREADONLY_on(sstr);
4225 /* Make the source SV into a loop of 1.
4226 (about to become 2) */
4227 SV_COW_NEXT_SV_SET(sstr, sstr);
4231 /* Initial code is common. */
4232 if (SvPVX_const(dstr)) { /* we know that dtype >= SVt_PV */
4237 /* making another shared SV. */
4238 STRLEN cur = SvCUR(sstr);
4239 STRLEN len = SvLEN(sstr);
4240 #ifdef PERL_OLD_COPY_ON_WRITE
4242 assert (SvTYPE(dstr) >= SVt_PVIV);
4243 /* SvIsCOW_normal */
4244 /* splice us in between source and next-after-source. */
4245 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4246 SV_COW_NEXT_SV_SET(sstr, dstr);
4247 SvPV_set(dstr, SvPVX_mutable(sstr));
4251 /* SvIsCOW_shared_hash */
4252 DEBUG_C(PerlIO_printf(Perl_debug_log,
4253 "Copy on write: Sharing hash\n"));
4255 assert (SvTYPE(dstr) >= SVt_PV);
4257 HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)))));
4259 SvLEN_set(dstr, len);
4260 SvCUR_set(dstr, cur);
4261 SvREADONLY_on(dstr);
4265 { /* Passes the swipe test. */
4266 SvPV_set(dstr, SvPVX_mutable(sstr));
4267 SvLEN_set(dstr, SvLEN(sstr));
4268 SvCUR_set(dstr, SvCUR(sstr));
4271 (void)SvOK_off(sstr); /* NOTE: nukes most SvFLAGS on sstr */
4272 SvPV_set(sstr, NULL);
4278 if (sflags & SVp_NOK) {
4279 SvNV_set(dstr, SvNVX(sstr));
4281 if (sflags & SVp_IOK) {
4282 SvIV_set(dstr, SvIVX(sstr));
4283 /* Must do this otherwise some other overloaded use of 0x80000000
4284 gets confused. I guess SVpbm_VALID */
4285 if (sflags & SVf_IVisUV)
4288 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_NOK|SVp_NOK|SVf_UTF8);
4290 const MAGIC * const smg = SvVSTRING_mg(sstr);
4292 sv_magic(dstr, NULL, PERL_MAGIC_vstring,
4293 smg->mg_ptr, smg->mg_len);
4294 SvRMAGICAL_on(dstr);
4298 else if (sflags & (SVp_IOK|SVp_NOK)) {
4299 (void)SvOK_off(dstr);
4300 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_IVisUV|SVf_NOK|SVp_NOK);
4301 if (sflags & SVp_IOK) {
4302 /* XXXX Do we want to set IsUV for IV(ROK)? Be extra safe... */
4303 SvIV_set(dstr, SvIVX(sstr));
4305 if (sflags & SVp_NOK) {
4306 SvNV_set(dstr, SvNVX(sstr));
4310 if (isGV_with_GP(sstr)) {
4311 gv_efullname3(dstr, MUTABLE_GV(sstr), "*");
4314 (void)SvOK_off(dstr);
4316 if (SvTAINTED(sstr))
4321 =for apidoc sv_setsv_mg
4323 Like C<sv_setsv>, but also handles 'set' magic.
4329 Perl_sv_setsv_mg(pTHX_ SV *const dstr, register SV *const sstr)
4331 PERL_ARGS_ASSERT_SV_SETSV_MG;
4333 sv_setsv(dstr,sstr);
4337 #ifdef PERL_OLD_COPY_ON_WRITE
4339 Perl_sv_setsv_cow(pTHX_ SV *dstr, SV *sstr)
4341 STRLEN cur = SvCUR(sstr);
4342 STRLEN len = SvLEN(sstr);
4345 PERL_ARGS_ASSERT_SV_SETSV_COW;
4348 PerlIO_printf(Perl_debug_log, "Fast copy on write: %p -> %p\n",
4349 (void*)sstr, (void*)dstr);
4356 if (SvTHINKFIRST(dstr))
4357 sv_force_normal_flags(dstr, SV_COW_DROP_PV);
4358 else if (SvPVX_const(dstr))
4359 Safefree(SvPVX_const(dstr));
4363 SvUPGRADE(dstr, SVt_PVIV);
4365 assert (SvPOK(sstr));
4366 assert (SvPOKp(sstr));
4367 assert (!SvIOK(sstr));
4368 assert (!SvIOKp(sstr));
4369 assert (!SvNOK(sstr));
4370 assert (!SvNOKp(sstr));
4372 if (SvIsCOW(sstr)) {
4374 if (SvLEN(sstr) == 0) {
4375 /* source is a COW shared hash key. */
4376 DEBUG_C(PerlIO_printf(Perl_debug_log,
4377 "Fast copy on write: Sharing hash\n"));
4378 new_pv = HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr))));
4381 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4383 assert ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS);
4384 SvUPGRADE(sstr, SVt_PVIV);
4385 SvREADONLY_on(sstr);
4387 DEBUG_C(PerlIO_printf(Perl_debug_log,
4388 "Fast copy on write: Converting sstr to COW\n"));
4389 SV_COW_NEXT_SV_SET(dstr, sstr);
4391 SV_COW_NEXT_SV_SET(sstr, dstr);
4392 new_pv = SvPVX_mutable(sstr);
4395 SvPV_set(dstr, new_pv);
4396 SvFLAGS(dstr) = (SVt_PVIV|SVf_POK|SVp_POK|SVf_FAKE|SVf_READONLY);
4399 SvLEN_set(dstr, len);
4400 SvCUR_set(dstr, cur);
4409 =for apidoc sv_setpvn
4411 Copies a string into an SV. The C<len> parameter indicates the number of
4412 bytes to be copied. If the C<ptr> argument is NULL the SV will become
4413 undefined. Does not handle 'set' magic. See C<sv_setpvn_mg>.
4419 Perl_sv_setpvn(pTHX_ register SV *const sv, register const char *const ptr, register const STRLEN len)
4424 PERL_ARGS_ASSERT_SV_SETPVN;
4426 SV_CHECK_THINKFIRST_COW_DROP(sv);
4432 /* len is STRLEN which is unsigned, need to copy to signed */
4435 Perl_croak(aTHX_ "panic: sv_setpvn called with negative strlen %"
4438 SvUPGRADE(sv, SVt_PV);
4440 dptr = SvGROW(sv, len + 1);
4441 Move(ptr,dptr,len,char);
4444 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4446 if (SvTYPE(sv) == SVt_PVCV) CvAUTOLOAD_off(sv);
4450 =for apidoc sv_setpvn_mg
4452 Like C<sv_setpvn>, but also handles 'set' magic.
4458 Perl_sv_setpvn_mg(pTHX_ register SV *const sv, register const char *const ptr, register const STRLEN len)
4460 PERL_ARGS_ASSERT_SV_SETPVN_MG;
4462 sv_setpvn(sv,ptr,len);
4467 =for apidoc sv_setpv
4469 Copies a string into an SV. The string must be null-terminated. Does not
4470 handle 'set' magic. See C<sv_setpv_mg>.
4476 Perl_sv_setpv(pTHX_ register SV *const sv, register const char *const ptr)
4481 PERL_ARGS_ASSERT_SV_SETPV;
4483 SV_CHECK_THINKFIRST_COW_DROP(sv);
4489 SvUPGRADE(sv, SVt_PV);
4491 SvGROW(sv, len + 1);
4492 Move(ptr,SvPVX(sv),len+1,char);
4494 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4496 if (SvTYPE(sv) == SVt_PVCV) CvAUTOLOAD_off(sv);
4500 =for apidoc sv_setpv_mg
4502 Like C<sv_setpv>, but also handles 'set' magic.
4508 Perl_sv_setpv_mg(pTHX_ register SV *const sv, register const char *const ptr)
4510 PERL_ARGS_ASSERT_SV_SETPV_MG;
4517 Perl_sv_sethek(pTHX_ register SV *const sv, const HEK *const hek)
4521 PERL_ARGS_ASSERT_SV_SETHEK;
4527 if (HEK_LEN(hek) == HEf_SVKEY) {
4528 sv_setsv(sv, *(SV**)HEK_KEY(hek));
4531 const int flags = HEK_FLAGS(hek);
4532 if (flags & HVhek_WASUTF8) {
4533 STRLEN utf8_len = HEK_LEN(hek);
4534 char *as_utf8 = (char *)bytes_to_utf8((U8*)HEK_KEY(hek), &utf8_len);
4535 sv_usepvn_flags(sv, as_utf8, utf8_len, SV_HAS_TRAILING_NUL);
4538 } else if (flags & (HVhek_REHASH|HVhek_UNSHARED)) {
4539 sv_setpvn(sv, HEK_KEY(hek), HEK_LEN(hek));
4542 else SvUTF8_off(sv);
4546 SV_CHECK_THINKFIRST_COW_DROP(sv);
4547 SvUPGRADE(sv, SVt_PV);
4548 Safefree(SvPVX(sv));
4549 SvPV_set(sv,(char *)HEK_KEY(share_hek_hek(hek)));
4550 SvCUR_set(sv, HEK_LEN(hek));
4557 else SvUTF8_off(sv);
4565 =for apidoc sv_usepvn_flags
4567 Tells an SV to use C<ptr> to find its string value. Normally the
4568 string is stored inside the SV but sv_usepvn allows the SV to use an
4569 outside string. The C<ptr> should point to memory that was allocated
4570 by C<malloc>. It must be the start of a mallocked block
4571 of memory, and not a pointer to the middle of it. The
4572 string length, C<len>, must be supplied. By default
4573 this function will realloc (i.e. move) the memory pointed to by C<ptr>,
4574 so that pointer should not be freed or used by the programmer after
4575 giving it to sv_usepvn, and neither should any pointers from "behind"
4576 that pointer (e.g. ptr + 1) be used.
4578 If C<flags> & SV_SMAGIC is true, will call SvSETMAGIC. If C<flags> &
4579 SV_HAS_TRAILING_NUL is true, then C<ptr[len]> must be NUL, and the realloc
4580 will be skipped (i.e. the buffer is actually at least 1 byte longer than
4581 C<len>, and already meets the requirements for storing in C<SvPVX>).
4587 Perl_sv_usepvn_flags(pTHX_ SV *const sv, char *ptr, const STRLEN len, const U32 flags)
4592 PERL_ARGS_ASSERT_SV_USEPVN_FLAGS;
4594 SV_CHECK_THINKFIRST_COW_DROP(sv);
4595 SvUPGRADE(sv, SVt_PV);
4598 if (flags & SV_SMAGIC)
4602 if (SvPVX_const(sv))
4606 if (flags & SV_HAS_TRAILING_NUL)
4607 assert(ptr[len] == '\0');
4610 allocate = (flags & SV_HAS_TRAILING_NUL)
4612 #ifdef Perl_safesysmalloc_size
4615 PERL_STRLEN_ROUNDUP(len + 1);
4617 if (flags & SV_HAS_TRAILING_NUL) {
4618 /* It's long enough - do nothing.
4619 Specifically Perl_newCONSTSUB is relying on this. */
4622 /* Force a move to shake out bugs in callers. */
4623 char *new_ptr = (char*)safemalloc(allocate);
4624 Copy(ptr, new_ptr, len, char);
4625 PoisonFree(ptr,len,char);
4629 ptr = (char*) saferealloc (ptr, allocate);
4632 #ifdef Perl_safesysmalloc_size
4633 SvLEN_set(sv, Perl_safesysmalloc_size(ptr));
4635 SvLEN_set(sv, allocate);
4639 if (!(flags & SV_HAS_TRAILING_NUL)) {
4642 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4644 if (flags & SV_SMAGIC)
4648 #ifdef PERL_OLD_COPY_ON_WRITE
4649 /* Need to do this *after* making the SV normal, as we need the buffer
4650 pointer to remain valid until after we've copied it. If we let go too early,
4651 another thread could invalidate it by unsharing last of the same hash key
4652 (which it can do by means other than releasing copy-on-write Svs)
4653 or by changing the other copy-on-write SVs in the loop. */
4655 S_sv_release_COW(pTHX_ register SV *sv, const char *pvx, SV *after)
4657 PERL_ARGS_ASSERT_SV_RELEASE_COW;
4659 { /* this SV was SvIsCOW_normal(sv) */
4660 /* we need to find the SV pointing to us. */
4661 SV *current = SV_COW_NEXT_SV(after);
4663 if (current == sv) {
4664 /* The SV we point to points back to us (there were only two of us
4666 Hence other SV is no longer copy on write either. */
4668 SvREADONLY_off(after);
4670 /* We need to follow the pointers around the loop. */
4672 while ((next = SV_COW_NEXT_SV(current)) != sv) {
4675 /* don't loop forever if the structure is bust, and we have
4676 a pointer into a closed loop. */
4677 assert (current != after);
4678 assert (SvPVX_const(current) == pvx);
4680 /* Make the SV before us point to the SV after us. */
4681 SV_COW_NEXT_SV_SET(current, after);
4687 =for apidoc sv_force_normal_flags
4689 Undo various types of fakery on an SV, where fakery means
4690 "more than" a string: if the PV is a shared string, make
4691 a private copy; if we're a ref, stop refing; if we're a glob, downgrade to
4692 an xpvmg; if we're a copy-on-write scalar, this is the on-write time when
4693 we do the copy, and is also used locally; if this is a
4694 vstring, drop the vstring magic. If C<SV_COW_DROP_PV> is set
4695 then a copy-on-write scalar drops its PV buffer (if any) and becomes
4696 SvPOK_off rather than making a copy. (Used where this
4697 scalar is about to be set to some other value.) In addition,
4698 the C<flags> parameter gets passed to C<sv_unref_flags()>
4699 when unreffing. C<sv_force_normal> calls this function
4700 with flags set to 0.
4706 Perl_sv_force_normal_flags(pTHX_ register SV *const sv, const U32 flags)
4710 PERL_ARGS_ASSERT_SV_FORCE_NORMAL_FLAGS;
4712 #ifdef PERL_OLD_COPY_ON_WRITE
4713 if (SvREADONLY(sv)) {
4715 const char * const pvx = SvPVX_const(sv);
4716 const STRLEN len = SvLEN(sv);
4717 const STRLEN cur = SvCUR(sv);
4718 /* next COW sv in the loop. If len is 0 then this is a shared-hash
4719 key scalar, so we mustn't attempt to call SV_COW_NEXT_SV(), as
4720 we'll fail an assertion. */
4721 SV * const next = len ? SV_COW_NEXT_SV(sv) : 0;
4724 PerlIO_printf(Perl_debug_log,
4725 "Copy on write: Force normal %ld\n",
4731 /* This SV doesn't own the buffer, so need to Newx() a new one: */
4734 if (flags & SV_COW_DROP_PV) {
4735 /* OK, so we don't need to copy our buffer. */
4738 SvGROW(sv, cur + 1);
4739 Move(pvx,SvPVX(sv),cur,char);
4744 sv_release_COW(sv, pvx, next);
4746 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4752 else if (IN_PERL_RUNTIME)
4753 Perl_croak_no_modify(aTHX);
4756 if (SvREADONLY(sv)) {
4758 const char * const pvx = SvPVX_const(sv);
4759 const STRLEN len = SvCUR(sv);
4764 if (flags & SV_COW_DROP_PV) {
4765 /* OK, so we don't need to copy our buffer. */
4768 SvGROW(sv, len + 1);
4769 Move(pvx,SvPVX(sv),len,char);
4772 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4774 else if (IN_PERL_RUNTIME)
4775 Perl_croak_no_modify(aTHX);
4779 sv_unref_flags(sv, flags);
4780 else if (SvFAKE(sv) && isGV_with_GP(sv))
4781 sv_unglob(sv, flags);
4782 else if (SvFAKE(sv) && SvTYPE(sv) == SVt_REGEXP) {
4783 /* Need to downgrade the REGEXP to a simple(r) scalar. This is analogous
4784 to sv_unglob. We only need it here, so inline it. */
4785 const svtype new_type = SvMAGIC(sv) || SvSTASH(sv) ? SVt_PVMG : SVt_PV;
4786 SV *const temp = newSV_type(new_type);
4787 void *const temp_p = SvANY(sv);
4789 if (new_type == SVt_PVMG) {
4790 SvMAGIC_set(temp, SvMAGIC(sv));
4791 SvMAGIC_set(sv, NULL);
4792 SvSTASH_set(temp, SvSTASH(sv));
4793 SvSTASH_set(sv, NULL);
4795 SvCUR_set(temp, SvCUR(sv));
4796 /* Remember that SvPVX is in the head, not the body. */
4798 SvLEN_set(temp, SvLEN(sv));
4799 /* This signals "buffer is owned by someone else" in sv_clear,
4800 which is the least effort way to stop it freeing the buffer.
4802 SvLEN_set(sv, SvLEN(sv)+1);
4804 /* Their buffer is already owned by someone else. */
4805 SvPVX(sv) = savepvn(SvPVX(sv), SvCUR(sv));
4806 SvLEN_set(temp, SvCUR(sv)+1);
4809 /* Now swap the rest of the bodies. */
4811 SvFLAGS(sv) &= ~(SVf_FAKE|SVTYPEMASK);
4812 SvFLAGS(sv) |= new_type;
4813 SvANY(sv) = SvANY(temp);
4815 SvFLAGS(temp) &= ~(SVTYPEMASK);
4816 SvFLAGS(temp) |= SVt_REGEXP|SVf_FAKE;
4817 SvANY(temp) = temp_p;
4821 else if (SvVOK(sv)) sv_unmagic(sv, PERL_MAGIC_vstring);
4827 Efficient removal of characters from the beginning of the string buffer.
4828 SvPOK(sv), or at least SvPOKp(sv), must be true and the C<ptr> must be a
4829 pointer to somewhere inside the string buffer. The C<ptr> becomes the first
4830 character of the adjusted string. Uses the "OOK hack". On return, only
4831 SvPOK(sv) and SvPOKp(sv) among the OK flags will be true.
4833 Beware: after this function returns, C<ptr> and SvPVX_const(sv) may no longer
4834 refer to the same chunk of data.
4836 The unfortunate similarity of this function's name to that of Perl's C<chop>
4837 operator is strictly coincidental. This function works from the left;
4838 C<chop> works from the right.
4844 Perl_sv_chop(pTHX_ register SV *const sv, register const char *const ptr)
4855 PERL_ARGS_ASSERT_SV_CHOP;
4857 if (!ptr || !SvPOKp(sv))
4859 delta = ptr - SvPVX_const(sv);
4861 /* Nothing to do. */
4864 max_delta = SvLEN(sv) ? SvLEN(sv) : SvCUR(sv);
4865 if (delta > max_delta)
4866 Perl_croak(aTHX_ "panic: sv_chop ptr=%p, start=%p, end=%p",
4867 ptr, SvPVX_const(sv), SvPVX_const(sv) + max_delta);
4868 /* SvPVX(sv) may move in SV_CHECK_THINKFIRST(sv), so don't use ptr any more */
4869 SV_CHECK_THINKFIRST(sv);
4870 SvPOK_only_UTF8(sv);
4873 if (!SvLEN(sv)) { /* make copy of shared string */
4874 const char *pvx = SvPVX_const(sv);
4875 const STRLEN len = SvCUR(sv);
4876 SvGROW(sv, len + 1);
4877 Move(pvx,SvPVX(sv),len,char);
4883 SvOOK_offset(sv, old_delta);
4885 SvLEN_set(sv, SvLEN(sv) - delta);
4886 SvCUR_set(sv, SvCUR(sv) - delta);
4887 SvPV_set(sv, SvPVX(sv) + delta);
4889 p = (U8 *)SvPVX_const(sv);
4892 /* how many bytes were evacuated? we will fill them with sentinel
4893 bytes, except for the part holding the new offset of course. */
4896 evacn += (old_delta < 0x100 ? 1 : 1 + sizeof(STRLEN));
4898 assert(evacn <= delta + old_delta);
4904 if (delta < 0x100) {
4908 p -= sizeof(STRLEN);
4909 Copy((U8*)&delta, p, sizeof(STRLEN), U8);
4913 /* Fill the preceding buffer with sentinals to verify that no-one is
4923 =for apidoc sv_catpvn
4925 Concatenates the string onto the end of the string which is in the SV. The
4926 C<len> indicates number of bytes to copy. If the SV has the UTF-8
4927 status set, then the bytes appended should be valid UTF-8.
4928 Handles 'get' magic, but not 'set' magic. See C<sv_catpvn_mg>.
4930 =for apidoc sv_catpvn_flags
4932 Concatenates the string onto the end of the string which is in the SV. The
4933 C<len> indicates number of bytes to copy. If the SV has the UTF-8
4934 status set, then the bytes appended should be valid UTF-8.
4935 If C<flags> has the C<SV_SMAGIC> bit set, will
4936 C<mg_set> on C<dsv> afterwards if appropriate.
4937 C<sv_catpvn> and C<sv_catpvn_nomg> are implemented
4938 in terms of this function.
4944 Perl_sv_catpvn_flags(pTHX_ register SV *const dsv, register const char *sstr, register const STRLEN slen, const I32 flags)
4948 const char * const dstr = SvPV_force_flags(dsv, dlen, flags);
4950 PERL_ARGS_ASSERT_SV_CATPVN_FLAGS;
4951 assert((flags & (SV_CATBYTES|SV_CATUTF8)) != (SV_CATBYTES|SV_CATUTF8));
4953 if (!(flags & SV_CATBYTES) || !SvUTF8(dsv)) {
4954 if (flags & SV_CATUTF8 && !SvUTF8(dsv)) {
4955 sv_utf8_upgrade_flags_grow(dsv, 0, slen + 1);
4958 else SvGROW(dsv, dlen + slen + 1);
4960 sstr = SvPVX_const(dsv);
4961 Move(sstr, SvPVX(dsv) + dlen, slen, char);
4962 SvCUR_set(dsv, SvCUR(dsv) + slen);
4965 /* We inline bytes_to_utf8, to avoid an extra malloc. */
4966 const char * const send = sstr + slen;
4969 /* Something this code does not account for, which I think is
4970 impossible; it would require the same pv to be treated as
4971 bytes *and* utf8, which would indicate a bug elsewhere. */
4972 assert(sstr != dstr);
4974 SvGROW(dsv, dlen + slen * 2 + 1);
4975 d = (U8 *)SvPVX(dsv) + dlen;
4977 while (sstr < send) {
4978 const UV uv = NATIVE_TO_ASCII((U8)*sstr++);
4979 if (UNI_IS_INVARIANT(uv))
4980 *d++ = (U8)UTF_TO_NATIVE(uv);
4982 *d++ = (U8)UTF8_EIGHT_BIT_HI(uv);
4983 *d++ = (U8)UTF8_EIGHT_BIT_LO(uv);
4986 SvCUR_set(dsv, d-(const U8 *)SvPVX(dsv));
4989 (void)SvPOK_only_UTF8(dsv); /* validate pointer */
4991 if (flags & SV_SMAGIC)
4996 =for apidoc sv_catsv
4998 Concatenates the string from SV C<ssv> onto the end of the string in SV
4999 C<dsv>. If C<ssv> is null, does nothing; otherwise modifies only C<dsv>.
5000 Handles 'get' magic on both SVs, but no 'set' magic. See C<sv_catsv_mg> and
5003 =for apidoc sv_catsv_flags
5005 Concatenates the string from SV C<ssv> onto the end of the string in SV
5006 C<dsv>. If C<ssv> is null, does nothing; otherwise modifies only C<dsv>.
5007 If C<flags> include C<SV_GMAGIC> bit set, will call C<mg_get> on both SVs if
5008 appropriate. If C<flags> include C<SV_SMAGIC>, C<mg_set> will be called on
5009 the modified SV afterward, if appropriate. C<sv_catsv>, C<sv_catsv_nomg>,
5010 and C<sv_catsv_mg> are implemented in terms of this function.
5015 Perl_sv_catsv_flags(pTHX_ SV *const dsv, register SV *const ssv, const I32 flags)
5019 PERL_ARGS_ASSERT_SV_CATSV_FLAGS;
5023 const char *spv = SvPV_flags_const(ssv, slen, flags);
5025 if (flags & SV_GMAGIC)
5027 sv_catpvn_flags(dsv, spv, slen,
5028 DO_UTF8(ssv) ? SV_CATUTF8 : SV_CATBYTES);
5029 if (flags & SV_SMAGIC)
5036 =for apidoc sv_catpv
5038 Concatenates the string onto the end of the string which is in the SV.
5039 If the SV has the UTF-8 status set, then the bytes appended should be
5040 valid UTF-8. Handles 'get' magic, but not 'set' magic. See C<sv_catpv_mg>.
5045 Perl_sv_catpv(pTHX_ register SV *const sv, register const char *ptr)
5052 PERL_ARGS_ASSERT_SV_CATPV;
5056 junk = SvPV_force(sv, tlen);
5058 SvGROW(sv, tlen + len + 1);
5060 ptr = SvPVX_const(sv);
5061 Move(ptr,SvPVX(sv)+tlen,len+1,char);
5062 SvCUR_set(sv, SvCUR(sv) + len);
5063 (void)SvPOK_only_UTF8(sv); /* validate pointer */
5068 =for apidoc sv_catpv_flags
5070 Concatenates the string onto the end of the string which is in the SV.
5071 If the SV has the UTF-8 status set, then the bytes appended should
5072 be valid UTF-8. If C<flags> has the C<SV_SMAGIC> bit set, will C<mg_set>
5073 on the modified SV if appropriate.
5079 Perl_sv_catpv_flags(pTHX_ SV *dstr, const char *sstr, const I32 flags)
5081 PERL_ARGS_ASSERT_SV_CATPV_FLAGS;
5082 sv_catpvn_flags(dstr, sstr, strlen(sstr), flags);
5086 =for apidoc sv_catpv_mg
5088 Like C<sv_catpv>, but also handles 'set' magic.
5094 Perl_sv_catpv_mg(pTHX_ register SV *const sv, register const char *const ptr)
5096 PERL_ARGS_ASSERT_SV_CATPV_MG;
5105 Creates a new SV. A non-zero C<len> parameter indicates the number of
5106 bytes of preallocated string space the SV should have. An extra byte for a
5107 trailing NUL is also reserved. (SvPOK is not set for the SV even if string
5108 space is allocated.) The reference count for the new SV is set to 1.
5110 In 5.9.3, newSV() replaces the older NEWSV() API, and drops the first
5111 parameter, I<x>, a debug aid which allowed callers to identify themselves.
5112 This aid has been superseded by a new build option, PERL_MEM_LOG (see
5113 L<perlhacktips/PERL_MEM_LOG>). The older API is still there for use in XS
5114 modules supporting older perls.
5120 Perl_newSV(pTHX_ const STRLEN len)
5127 sv_upgrade(sv, SVt_PV);
5128 SvGROW(sv, len + 1);
5133 =for apidoc sv_magicext
5135 Adds magic to an SV, upgrading it if necessary. Applies the
5136 supplied vtable and returns a pointer to the magic added.
5138 Note that C<sv_magicext> will allow things that C<sv_magic> will not.
5139 In particular, you can add magic to SvREADONLY SVs, and add more than
5140 one instance of the same 'how'.
5142 If C<namlen> is greater than zero then a C<savepvn> I<copy> of C<name> is
5143 stored, if C<namlen> is zero then C<name> is stored as-is and - as another
5144 special case - if C<(name && namlen == HEf_SVKEY)> then C<name> is assumed
5145 to contain an C<SV*> and is stored as-is with its REFCNT incremented.
5147 (This is now used as a subroutine by C<sv_magic>.)
5152 Perl_sv_magicext(pTHX_ SV *const sv, SV *const obj, const int how,
5153 const MGVTBL *const vtable, const char *const name, const I32 namlen)
5158 PERL_ARGS_ASSERT_SV_MAGICEXT;
5160 SvUPGRADE(sv, SVt_PVMG);
5161 Newxz(mg, 1, MAGIC);
5162 mg->mg_moremagic = SvMAGIC(sv);
5163 SvMAGIC_set(sv, mg);
5165 /* Sometimes a magic contains a reference loop, where the sv and
5166 object refer to each other. To prevent a reference loop that
5167 would prevent such objects being freed, we look for such loops
5168 and if we find one we avoid incrementing the object refcount.
5170 Note we cannot do this to avoid self-tie loops as intervening RV must
5171 have its REFCNT incremented to keep it in existence.
5174 if (!obj || obj == sv ||
5175 how == PERL_MAGIC_arylen ||
5176 how == PERL_MAGIC_symtab ||
5177 (SvTYPE(obj) == SVt_PVGV &&
5178 (GvSV(obj) == sv || GvHV(obj) == (const HV *)sv
5179 || GvAV(obj) == (const AV *)sv || GvCV(obj) == (const CV *)sv
5180 || GvIOp(obj) == (const IO *)sv || GvFORM(obj) == (const CV *)sv)))
5185 mg->mg_obj = SvREFCNT_inc_simple(obj);
5186 mg->mg_flags |= MGf_REFCOUNTED;
5189 /* Normal self-ties simply pass a null object, and instead of
5190 using mg_obj directly, use the SvTIED_obj macro to produce a
5191 new RV as needed. For glob "self-ties", we are tieing the PVIO
5192 with an RV obj pointing to the glob containing the PVIO. In
5193 this case, to avoid a reference loop, we need to weaken the
5197 if (how == PERL_MAGIC_tiedscalar && SvTYPE(sv) == SVt_PVIO &&
5198 obj && SvROK(obj) && GvIO(SvRV(obj)) == (const IO *)sv)
5204 mg->mg_len = namlen;
5207 mg->mg_ptr = savepvn(name, namlen);
5208 else if (namlen == HEf_SVKEY) {
5209 /* Yes, this is casting away const. This is only for the case of
5210 HEf_SVKEY. I think we need to document this aberation of the
5211 constness of the API, rather than making name non-const, as
5212 that change propagating outwards a long way. */
5213 mg->mg_ptr = (char*)SvREFCNT_inc_simple_NN((SV *)name);
5215 mg->mg_ptr = (char *) name;
5217 mg->mg_virtual = (MGVTBL *) vtable;
5224 =for apidoc sv_magic
5226 Adds magic to an SV. First upgrades C<sv> to type C<SVt_PVMG> if
5227 necessary, then adds a new magic item of type C<how> to the head of the
5230 See C<sv_magicext> (which C<sv_magic> now calls) for a description of the
5231 handling of the C<name> and C<namlen> arguments.
5233 You need to use C<sv_magicext> to add magic to SvREADONLY SVs and also
5234 to add more than one instance of the same 'how'.
5240 Perl_sv_magic(pTHX_ register SV *const sv, SV *const obj, const int how,
5241 const char *const name, const I32 namlen)
5244 const MGVTBL *vtable;
5247 unsigned int vtable_index;
5249 PERL_ARGS_ASSERT_SV_MAGIC;
5251 if (how < 0 || (unsigned)how > C_ARRAY_LENGTH(PL_magic_data)
5252 || ((flags = PL_magic_data[how]),
5253 (vtable_index = flags & PERL_MAGIC_VTABLE_MASK)
5254 > magic_vtable_max))
5255 Perl_croak(aTHX_ "Don't know how to handle magic of type \\%o", how);
5257 /* PERL_MAGIC_ext is reserved for use by extensions not perl internals.
5258 Useful for attaching extension internal data to perl vars.
5259 Note that multiple extensions may clash if magical scalars
5260 etc holding private data from one are passed to another. */
5262 vtable = (vtable_index == magic_vtable_max)
5263 ? NULL : PL_magic_vtables + vtable_index;
5265 #ifdef PERL_OLD_COPY_ON_WRITE
5267 sv_force_normal_flags(sv, 0);
5269 if (SvREADONLY(sv)) {
5271 /* its okay to attach magic to shared strings */
5275 && !PERL_MAGIC_TYPE_READONLY_ACCEPTABLE(how)
5278 Perl_croak_no_modify(aTHX);
5281 if (SvMAGICAL(sv) || (how == PERL_MAGIC_taint && SvTYPE(sv) >= SVt_PVMG)) {
5282 if (SvMAGIC(sv) && (mg = mg_find(sv, how))) {
5283 /* sv_magic() refuses to add a magic of the same 'how' as an
5286 if (how == PERL_MAGIC_taint)
5292 /* Rest of work is done else where */
5293 mg = sv_magicext(sv,obj,how,vtable,name,namlen);
5296 case PERL_MAGIC_taint:
5299 case PERL_MAGIC_ext:
5300 case PERL_MAGIC_dbfile:
5307 S_sv_unmagicext_flags(pTHX_ SV *const sv, const int type, MGVTBL *vtbl, const U32 flags)
5314 if (SvTYPE(sv) < SVt_PVMG || !SvMAGIC(sv))
5316 mgp = &(((XPVMG*) SvANY(sv))->xmg_u.xmg_magic);
5317 for (mg = *mgp; mg; mg = *mgp) {
5318 const MGVTBL* const virt = mg->mg_virtual;
5319 if (mg->mg_type == type && (!flags || virt == vtbl)) {
5320 *mgp = mg->mg_moremagic;
5321 if (virt && virt->svt_free)
5322 virt->svt_free(aTHX_ sv, mg);
5323 if (mg->mg_ptr && mg->mg_type != PERL_MAGIC_regex_global) {
5325 Safefree(mg->mg_ptr);
5326 else if (mg->mg_len == HEf_SVKEY)
5327 SvREFCNT_dec(MUTABLE_SV(mg->mg_ptr));
5328 else if (mg->mg_type == PERL_MAGIC_utf8)
5329 Safefree(mg->mg_ptr);
5331 if (mg->mg_flags & MGf_REFCOUNTED)
5332 SvREFCNT_dec(mg->mg_obj);
5336 mgp = &mg->mg_moremagic;
5339 if (SvMAGICAL(sv)) /* if we're under save_magic, wait for restore_magic; */
5340 mg_magical(sv); /* else fix the flags now */
5344 SvFLAGS(sv) |= (SvFLAGS(sv) & (SVp_IOK|SVp_NOK|SVp_POK)) >> PRIVSHIFT;
5350 =for apidoc sv_unmagic
5352 Removes all magic of type C<type> from an SV.
5358 Perl_sv_unmagic(pTHX_ SV *const sv, const int type)
5360 PERL_ARGS_ASSERT_SV_UNMAGIC;
5361 return S_sv_unmagicext_flags(aTHX_ sv, type, NULL, 0);
5365 =for apidoc sv_unmagicext
5367 Removes all magic of type C<type> with the specified C<vtbl> from an SV.
5373 Perl_sv_unmagicext(pTHX_ SV *const sv, const int type, MGVTBL *vtbl)
5375 PERL_ARGS_ASSERT_SV_UNMAGICEXT;
5376 return S_sv_unmagicext_flags(aTHX_ sv, type, vtbl, 1);
5380 =for apidoc sv_rvweaken
5382 Weaken a reference: set the C<SvWEAKREF> flag on this RV; give the
5383 referred-to SV C<PERL_MAGIC_backref> magic if it hasn't already; and
5384 push a back-reference to this RV onto the array of backreferences
5385 associated with that magic. If the RV is magical, set magic will be
5386 called after the RV is cleared.
5392 Perl_sv_rvweaken(pTHX_ SV *const sv)
5396 PERL_ARGS_ASSERT_SV_RVWEAKEN;
5398 if (!SvOK(sv)) /* let undefs pass */
5401 Perl_croak(aTHX_ "Can't weaken a nonreference");
5402 else if (SvWEAKREF(sv)) {
5403 Perl_ck_warner(aTHX_ packWARN(WARN_MISC), "Reference is already weak");
5406 else if (SvREADONLY(sv)) croak_no_modify();
5408 Perl_sv_add_backref(aTHX_ tsv, sv);
5414 /* Give tsv backref magic if it hasn't already got it, then push a
5415 * back-reference to sv onto the array associated with the backref magic.
5417 * As an optimisation, if there's only one backref and it's not an AV,
5418 * store it directly in the HvAUX or mg_obj slot, avoiding the need to
5419 * allocate an AV. (Whether the slot holds an AV tells us whether this is
5423 /* A discussion about the backreferences array and its refcount:
5425 * The AV holding the backreferences is pointed to either as the mg_obj of
5426 * PERL_MAGIC_backref, or in the specific case of a HV, from the
5427 * xhv_backreferences field. The array is created with a refcount
5428 * of 2. This means that if during global destruction the array gets
5429 * picked on before its parent to have its refcount decremented by the
5430 * random zapper, it won't actually be freed, meaning it's still there for
5431 * when its parent gets freed.
5433 * When the parent SV is freed, the extra ref is killed by
5434 * Perl_sv_kill_backrefs. The other ref is killed, in the case of magic,
5435 * by mg_free() / MGf_REFCOUNTED, or for a hash, by Perl_hv_kill_backrefs.
5437 * When a single backref SV is stored directly, it is not reference
5442 Perl_sv_add_backref(pTHX_ SV *const tsv, SV *const sv)
5449 PERL_ARGS_ASSERT_SV_ADD_BACKREF;
5451 /* find slot to store array or singleton backref */
5453 if (SvTYPE(tsv) == SVt_PVHV) {
5454 svp = (SV**)Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5457 (SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL))))
5459 sv_magic(tsv, NULL, PERL_MAGIC_backref, NULL, 0);
5460 mg = mg_find(tsv, PERL_MAGIC_backref);
5462 svp = &(mg->mg_obj);
5465 /* create or retrieve the array */
5467 if ( (!*svp && SvTYPE(sv) == SVt_PVAV)
5468 || (*svp && SvTYPE(*svp) != SVt_PVAV)
5473 SvREFCNT_inc_simple_void(av);
5474 /* av now has a refcnt of 2; see discussion above */
5476 /* move single existing backref to the array */
5478 AvARRAY(av)[++AvFILLp(av)] = *svp; /* av_push() */
5482 mg->mg_flags |= MGf_REFCOUNTED;
5485 av = MUTABLE_AV(*svp);
5488 /* optimisation: store single backref directly in HvAUX or mg_obj */
5492 /* push new backref */
5493 assert(SvTYPE(av) == SVt_PVAV);
5494 if (AvFILLp(av) >= AvMAX(av)) {
5495 av_extend(av, AvFILLp(av)+1);
5497 AvARRAY(av)[++AvFILLp(av)] = sv; /* av_push() */
5500 /* delete a back-reference to ourselves from the backref magic associated
5501 * with the SV we point to.
5505 Perl_sv_del_backref(pTHX_ SV *const tsv, SV *const sv)
5510 PERL_ARGS_ASSERT_SV_DEL_BACKREF;
5512 if (SvTYPE(tsv) == SVt_PVHV) {
5514 svp = (SV**)Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5516 else if (SvIS_FREED(tsv) && PL_phase == PERL_PHASE_DESTRUCT) {
5517 /* It's possible for the the last (strong) reference to tsv to have
5518 become freed *before* the last thing holding a weak reference.
5519 If both survive longer than the backreferences array, then when
5520 the referent's reference count drops to 0 and it is freed, it's
5521 not able to chase the backreferences, so they aren't NULLed.
5523 For example, a CV holds a weak reference to its stash. If both the
5524 CV and the stash survive longer than the backreferences array,
5525 and the CV gets picked for the SvBREAK() treatment first,
5526 *and* it turns out that the stash is only being kept alive because
5527 of an our variable in the pad of the CV, then midway during CV
5528 destruction the stash gets freed, but CvSTASH() isn't set to NULL.
5529 It ends up pointing to the freed HV. Hence it's chased in here, and
5530 if this block wasn't here, it would hit the !svp panic just below.
5532 I don't believe that "better" destruction ordering is going to help
5533 here - during global destruction there's always going to be the
5534 chance that something goes out of order. We've tried to make it
5535 foolproof before, and it only resulted in evolutionary pressure on
5536 fools. Which made us look foolish for our hubris. :-(
5542 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
5543 svp = mg ? &(mg->mg_obj) : NULL;
5547 Perl_croak(aTHX_ "panic: del_backref, svp=0");
5549 /* It's possible that sv is being freed recursively part way through the
5550 freeing of tsv. If this happens, the backreferences array of tsv has
5551 already been freed, and so svp will be NULL. If this is the case,
5552 we should not panic. Instead, nothing needs doing, so return. */
5553 if (PL_phase == PERL_PHASE_DESTRUCT && SvREFCNT(tsv) == 0)
5555 Perl_croak(aTHX_ "panic: del_backref, *svp=%p phase=%s refcnt=%" UVuf,
5556 *svp, PL_phase_names[PL_phase], (UV)SvREFCNT(tsv));
5559 if (SvTYPE(*svp) == SVt_PVAV) {
5563 AV * const av = (AV*)*svp;
5565 assert(!SvIS_FREED(av));
5569 /* for an SV with N weak references to it, if all those
5570 * weak refs are deleted, then sv_del_backref will be called
5571 * N times and O(N^2) compares will be done within the backref
5572 * array. To ameliorate this potential slowness, we:
5573 * 1) make sure this code is as tight as possible;
5574 * 2) when looking for SV, look for it at both the head and tail of the
5575 * array first before searching the rest, since some create/destroy
5576 * patterns will cause the backrefs to be freed in order.
5583 SV **p = &svp[fill];
5584 SV *const topsv = *p;
5591 /* We weren't the last entry.
5592 An unordered list has this property that you
5593 can take the last element off the end to fill
5594 the hole, and it's still an unordered list :-)
5600 break; /* should only be one */
5607 AvFILLp(av) = fill-1;
5609 else if (SvIS_FREED(*svp) && PL_phase == PERL_PHASE_DESTRUCT) {
5610 /* freed AV; skip */
5613 /* optimisation: only a single backref, stored directly */
5615 Perl_croak(aTHX_ "panic: del_backref, *svp=%p, sv=%p", *svp, sv);
5622 Perl_sv_kill_backrefs(pTHX_ SV *const sv, AV *const av)
5628 PERL_ARGS_ASSERT_SV_KILL_BACKREFS;
5633 /* after multiple passes through Perl_sv_clean_all() for a thingy
5634 * that has badly leaked, the backref array may have gotten freed,
5635 * since we only protect it against 1 round of cleanup */
5636 if (SvIS_FREED(av)) {
5637 if (PL_in_clean_all) /* All is fair */
5640 "panic: magic_killbackrefs (freed backref AV/SV)");
5644 is_array = (SvTYPE(av) == SVt_PVAV);
5646 assert(!SvIS_FREED(av));
5649 last = svp + AvFILLp(av);
5652 /* optimisation: only a single backref, stored directly */
5658 while (svp <= last) {
5660 SV *const referrer = *svp;
5661 if (SvWEAKREF(referrer)) {
5662 /* XXX Should we check that it hasn't changed? */
5663 assert(SvROK(referrer));
5664 SvRV_set(referrer, 0);
5666 SvWEAKREF_off(referrer);
5667 SvSETMAGIC(referrer);
5668 } else if (SvTYPE(referrer) == SVt_PVGV ||
5669 SvTYPE(referrer) == SVt_PVLV) {
5670 assert(SvTYPE(sv) == SVt_PVHV); /* stash backref */
5671 /* You lookin' at me? */
5672 assert(GvSTASH(referrer));
5673 assert(GvSTASH(referrer) == (const HV *)sv);
5674 GvSTASH(referrer) = 0;
5675 } else if (SvTYPE(referrer) == SVt_PVCV ||
5676 SvTYPE(referrer) == SVt_PVFM) {
5677 if (SvTYPE(sv) == SVt_PVHV) { /* stash backref */
5678 /* You lookin' at me? */
5679 assert(CvSTASH(referrer));
5680 assert(CvSTASH(referrer) == (const HV *)sv);
5681 SvANY(MUTABLE_CV(referrer))->xcv_stash = 0;
5684 assert(SvTYPE(sv) == SVt_PVGV);
5685 /* You lookin' at me? */
5686 assert(CvGV(referrer));
5687 assert(CvGV(referrer) == (const GV *)sv);
5688 anonymise_cv_maybe(MUTABLE_GV(sv),
5689 MUTABLE_CV(referrer));
5694 "panic: magic_killbackrefs (flags=%"UVxf")",
5695 (UV)SvFLAGS(referrer));
5706 SvREFCNT_dec(av); /* remove extra count added by sv_add_backref() */
5712 =for apidoc sv_insert
5714 Inserts a string at the specified offset/length within the SV. Similar to
5715 the Perl substr() function. Handles get magic.
5717 =for apidoc sv_insert_flags
5719 Same as C<sv_insert>, but the extra C<flags> are passed to the
5720 C<SvPV_force_flags> that applies to C<bigstr>.
5726 Perl_sv_insert_flags(pTHX_ SV *const bigstr, const STRLEN offset, const STRLEN len, const char *const little, const STRLEN littlelen, const U32 flags)
5733 SSize_t i; /* better be sizeof(STRLEN) or bad things happen */
5736 PERL_ARGS_ASSERT_SV_INSERT_FLAGS;
5739 Perl_croak(aTHX_ "Can't modify nonexistent substring");
5740 SvPV_force_flags(bigstr, curlen, flags);
5741 (void)SvPOK_only_UTF8(bigstr);
5742 if (offset + len > curlen) {
5743 SvGROW(bigstr, offset+len+1);
5744 Zero(SvPVX(bigstr)+curlen, offset+len-curlen, char);
5745 SvCUR_set(bigstr, offset+len);
5749 i = littlelen - len;
5750 if (i > 0) { /* string might grow */
5751 big = SvGROW(bigstr, SvCUR(bigstr) + i + 1);
5752 mid = big + offset + len;
5753 midend = bigend = big + SvCUR(bigstr);
5756 while (midend > mid) /* shove everything down */
5757 *--bigend = *--midend;
5758 Move(little,big+offset,littlelen,char);
5759 SvCUR_set(bigstr, SvCUR(bigstr) + i);
5764 Move(little,SvPVX(bigstr)+offset,len,char);
5769 big = SvPVX(bigstr);
5772 bigend = big + SvCUR(bigstr);
5774 if (midend > bigend)
5775 Perl_croak(aTHX_ "panic: sv_insert, midend=%p, bigend=%p",
5778 if (mid - big > bigend - midend) { /* faster to shorten from end */
5780 Move(little, mid, littlelen,char);
5783 i = bigend - midend;
5785 Move(midend, mid, i,char);
5789 SvCUR_set(bigstr, mid - big);
5791 else if ((i = mid - big)) { /* faster from front */
5792 midend -= littlelen;
5794 Move(big, midend - i, i, char);
5795 sv_chop(bigstr,midend-i);
5797 Move(little, mid, littlelen,char);
5799 else if (littlelen) {
5800 midend -= littlelen;
5801 sv_chop(bigstr,midend);
5802 Move(little,midend,littlelen,char);
5805 sv_chop(bigstr,midend);
5811 =for apidoc sv_replace
5813 Make the first argument a copy of the second, then delete the original.
5814 The target SV physically takes over ownership of the body of the source SV
5815 and inherits its flags; however, the target keeps any magic it owns,
5816 and any magic in the source is discarded.
5817 Note that this is a rather specialist SV copying operation; most of the
5818 time you'll want to use C<sv_setsv> or one of its many macro front-ends.
5824 Perl_sv_replace(pTHX_ register SV *const sv, register SV *const nsv)
5827 const U32 refcnt = SvREFCNT(sv);
5829 PERL_ARGS_ASSERT_SV_REPLACE;
5831 SV_CHECK_THINKFIRST_COW_DROP(sv);
5832 if (SvREFCNT(nsv) != 1) {
5833 Perl_croak(aTHX_ "panic: reference miscount on nsv in sv_replace()"
5834 " (%" UVuf " != 1)", (UV) SvREFCNT(nsv));
5836 if (SvMAGICAL(sv)) {
5840 sv_upgrade(nsv, SVt_PVMG);
5841 SvMAGIC_set(nsv, SvMAGIC(sv));
5842 SvFLAGS(nsv) |= SvMAGICAL(sv);
5844 SvMAGIC_set(sv, NULL);
5848 assert(!SvREFCNT(sv));
5849 #ifdef DEBUG_LEAKING_SCALARS
5850 sv->sv_flags = nsv->sv_flags;
5851 sv->sv_any = nsv->sv_any;
5852 sv->sv_refcnt = nsv->sv_refcnt;
5853 sv->sv_u = nsv->sv_u;
5855 StructCopy(nsv,sv,SV);
5857 if(SvTYPE(sv) == SVt_IV) {
5859 = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
5863 #ifdef PERL_OLD_COPY_ON_WRITE
5864 if (SvIsCOW_normal(nsv)) {
5865 /* We need to follow the pointers around the loop to make the
5866 previous SV point to sv, rather than nsv. */
5869 while ((next = SV_COW_NEXT_SV(current)) != nsv) {
5872 assert(SvPVX_const(current) == SvPVX_const(nsv));
5874 /* Make the SV before us point to the SV after us. */
5876 PerlIO_printf(Perl_debug_log, "previous is\n");
5878 PerlIO_printf(Perl_debug_log,
5879 "move it from 0x%"UVxf" to 0x%"UVxf"\n",
5880 (UV) SV_COW_NEXT_SV(current), (UV) sv);
5882 SV_COW_NEXT_SV_SET(current, sv);
5885 SvREFCNT(sv) = refcnt;
5886 SvFLAGS(nsv) |= SVTYPEMASK; /* Mark as freed */
5891 /* We're about to free a GV which has a CV that refers back to us.
5892 * If that CV will outlive us, make it anonymous (i.e. fix up its CvGV
5896 S_anonymise_cv_maybe(pTHX_ GV *gv, CV* cv)
5901 PERL_ARGS_ASSERT_ANONYMISE_CV_MAYBE;
5904 assert(SvREFCNT(gv) == 0);
5905 assert(isGV(gv) && isGV_with_GP(gv));
5907 assert(!CvANON(cv));
5908 assert(CvGV(cv) == gv);
5910 /* will the CV shortly be freed by gp_free() ? */
5911 if (GvCV(gv) == cv && GvGP(gv)->gp_refcnt < 2 && SvREFCNT(cv) < 2) {
5912 SvANY(cv)->xcv_gv = NULL;
5916 /* if not, anonymise: */
5917 gvname = (GvSTASH(gv) && HvNAME(GvSTASH(gv)) && HvENAME(GvSTASH(gv)))
5918 ? newSVhek(HvENAME_HEK(GvSTASH(gv)))
5919 : newSVpvn_flags( "__ANON__", 8, 0 );
5920 sv_catpvs(gvname, "::__ANON__");
5921 anongv = gv_fetchsv(gvname, GV_ADDMULTI, SVt_PVCV);
5922 SvREFCNT_dec(gvname);
5926 SvANY(cv)->xcv_gv = MUTABLE_GV(SvREFCNT_inc(anongv));
5931 =for apidoc sv_clear
5933 Clear an SV: call any destructors, free up any memory used by the body,
5934 and free the body itself. The SV's head is I<not> freed, although
5935 its type is set to all 1's so that it won't inadvertently be assumed
5936 to be live during global destruction etc.
5937 This function should only be called when REFCNT is zero. Most of the time
5938 you'll want to call C<sv_free()> (or its macro wrapper C<SvREFCNT_dec>)
5945 Perl_sv_clear(pTHX_ SV *const orig_sv)
5950 const struct body_details *sv_type_details;
5956 PERL_ARGS_ASSERT_SV_CLEAR;
5958 /* within this loop, sv is the SV currently being freed, and
5959 * iter_sv is the most recent AV or whatever that's being iterated
5960 * over to provide more SVs */
5966 assert(SvREFCNT(sv) == 0);
5967 assert(SvTYPE(sv) != (svtype)SVTYPEMASK);
5969 if (type <= SVt_IV) {
5970 /* See the comment in sv.h about the collusion between this
5971 * early return and the overloading of the NULL slots in the
5975 SvFLAGS(sv) &= SVf_BREAK;
5976 SvFLAGS(sv) |= SVTYPEMASK;
5980 assert(!SvOBJECT(sv) || type >= SVt_PVMG); /* objs are always >= MG */
5982 if (type >= SVt_PVMG) {
5984 if (!curse(sv, 1)) goto get_next_sv;
5985 type = SvTYPE(sv); /* destructor may have changed it */
5987 /* Free back-references before magic, in case the magic calls
5988 * Perl code that has weak references to sv. */
5989 if (type == SVt_PVHV) {
5990 Perl_hv_kill_backrefs(aTHX_ MUTABLE_HV(sv));
5994 else if (type == SVt_PVMG && SvPAD_OUR(sv)) {
5995 SvREFCNT_dec(SvOURSTASH(sv));
5996 } else if (SvMAGIC(sv)) {
5997 /* Free back-references before other types of magic. */
5998 sv_unmagic(sv, PERL_MAGIC_backref);
6002 if (type == SVt_PVMG && SvPAD_TYPED(sv))
6003 SvREFCNT_dec(SvSTASH(sv));
6006 /* case SVt_BIND: */
6009 IoIFP(sv) != PerlIO_stdin() &&
6010 IoIFP(sv) != PerlIO_stdout() &&
6011 IoIFP(sv) != PerlIO_stderr() &&
6012 !(IoFLAGS(sv) & IOf_FAKE_DIRP))
6014 io_close(MUTABLE_IO(sv), FALSE);
6016 if (IoDIRP(sv) && !(IoFLAGS(sv) & IOf_FAKE_DIRP))
6017 PerlDir_close(IoDIRP(sv));
6018 IoDIRP(sv) = (DIR*)NULL;
6019 Safefree(IoTOP_NAME(sv));
6020 Safefree(IoFMT_NAME(sv));
6021 Safefree(IoBOTTOM_NAME(sv));
6022 if ((const GV *)sv == PL_statgv)
6026 /* FIXME for plugins */
6027 pregfree2((REGEXP*) sv);
6031 cv_undef(MUTABLE_CV(sv));
6032 /* If we're in a stash, we don't own a reference to it.
6033 * However it does have a back reference to us, which needs to
6035 if ((stash = CvSTASH(sv)))
6036 sv_del_backref(MUTABLE_SV(stash), sv);
6039 if (PL_last_swash_hv == (const HV *)sv) {
6040 PL_last_swash_hv = NULL;
6042 if (HvTOTALKEYS((HV*)sv) > 0) {
6044 /* this statement should match the one at the beginning of
6045 * hv_undef_flags() */
6046 if ( PL_phase != PERL_PHASE_DESTRUCT
6047 && (name = HvNAME((HV*)sv)))
6050 (void)hv_delete(PL_stashcache, name,
6051 HvNAMEUTF8((HV*)sv) ? -HvNAMELEN_get((HV*)sv) : HvNAMELEN_get((HV*)sv), G_DISCARD);
6052 hv_name_set((HV*)sv, NULL, 0, 0);
6055 /* save old iter_sv in unused SvSTASH field */
6056 assert(!SvOBJECT(sv));
6057 SvSTASH(sv) = (HV*)iter_sv;
6060 /* save old hash_index in unused SvMAGIC field */
6061 assert(!SvMAGICAL(sv));
6062 assert(!SvMAGIC(sv));
6063 ((XPVMG*) SvANY(sv))->xmg_u.xmg_hash_index = hash_index;
6066 next_sv = Perl_hfree_next_entry(aTHX_ (HV*)sv, &hash_index);
6067 goto get_next_sv; /* process this new sv */
6069 /* free empty hash */
6070 Perl_hv_undef_flags(aTHX_ MUTABLE_HV(sv), HV_NAME_SETALL);
6071 assert(!HvARRAY((HV*)sv));
6075 AV* av = MUTABLE_AV(sv);
6076 if (PL_comppad == av) {
6080 if (AvREAL(av) && AvFILLp(av) > -1) {
6081 next_sv = AvARRAY(av)[AvFILLp(av)--];
6082 /* save old iter_sv in top-most slot of AV,
6083 * and pray that it doesn't get wiped in the meantime */
6084 AvARRAY(av)[AvMAX(av)] = iter_sv;
6086 goto get_next_sv; /* process this new sv */
6088 Safefree(AvALLOC(av));
6093 if (LvTYPE(sv) == 'T') { /* for tie: return HE to pool */
6094 SvREFCNT_dec(HeKEY_sv((HE*)LvTARG(sv)));
6095 HeNEXT((HE*)LvTARG(sv)) = PL_hv_fetch_ent_mh;
6096 PL_hv_fetch_ent_mh = (HE*)LvTARG(sv);
6098 else if (LvTYPE(sv) != 't') /* unless tie: unrefcnted fake SV** */
6099 SvREFCNT_dec(LvTARG(sv));
6101 if (isGV_with_GP(sv)) {
6102 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
6103 && HvENAME_get(stash))
6104 mro_method_changed_in(stash);
6105 gp_free(MUTABLE_GV(sv));
6107 unshare_hek(GvNAME_HEK(sv));
6108 /* If we're in a stash, we don't own a reference to it.
6109 * However it does have a back reference to us, which
6110 * needs to be cleared. */
6111 if (!SvVALID(sv) && (stash = GvSTASH(sv)))
6112 sv_del_backref(MUTABLE_SV(stash), sv);
6114 /* FIXME. There are probably more unreferenced pointers to SVs
6115 * in the interpreter struct that we should check and tidy in
6116 * a similar fashion to this: */
6117 /* See also S_sv_unglob, which does the same thing. */
6118 if ((const GV *)sv == PL_last_in_gv)
6119 PL_last_in_gv = NULL;
6120 else if ((const GV *)sv == PL_statgv)
6127 /* Don't bother with SvOOK_off(sv); as we're only going to
6131 SvOOK_offset(sv, offset);
6132 SvPV_set(sv, SvPVX_mutable(sv) - offset);
6133 /* Don't even bother with turning off the OOK flag. */
6138 SV * const target = SvRV(sv);
6140 sv_del_backref(target, sv);
6145 #ifdef PERL_OLD_COPY_ON_WRITE
6146 else if (SvPVX_const(sv)
6147 && !(SvTYPE(sv) == SVt_PVIO
6148 && !(IoFLAGS(sv) & IOf_FAKE_DIRP)))
6152 PerlIO_printf(Perl_debug_log, "Copy on write: clear\n");
6156 sv_release_COW(sv, SvPVX_const(sv), SV_COW_NEXT_SV(sv));
6158 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
6162 } else if (SvLEN(sv)) {
6163 Safefree(SvPVX_const(sv));
6167 else if (SvPVX_const(sv) && SvLEN(sv)
6168 && !(SvTYPE(sv) == SVt_PVIO
6169 && !(IoFLAGS(sv) & IOf_FAKE_DIRP)))
6170 Safefree(SvPVX_mutable(sv));
6171 else if (SvPVX_const(sv) && SvIsCOW(sv)) {
6172 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
6183 SvFLAGS(sv) &= SVf_BREAK;
6184 SvFLAGS(sv) |= SVTYPEMASK;
6186 sv_type_details = bodies_by_type + type;
6187 if (sv_type_details->arena) {
6188 del_body(((char *)SvANY(sv) + sv_type_details->offset),
6189 &PL_body_roots[type]);
6191 else if (sv_type_details->body_size) {
6192 safefree(SvANY(sv));
6196 /* caller is responsible for freeing the head of the original sv */
6197 if (sv != orig_sv && !SvREFCNT(sv))
6200 /* grab and free next sv, if any */
6208 else if (!iter_sv) {
6210 } else if (SvTYPE(iter_sv) == SVt_PVAV) {
6211 AV *const av = (AV*)iter_sv;
6212 if (AvFILLp(av) > -1) {
6213 sv = AvARRAY(av)[AvFILLp(av)--];
6215 else { /* no more elements of current AV to free */
6218 /* restore previous value, squirrelled away */
6219 iter_sv = AvARRAY(av)[AvMAX(av)];
6220 Safefree(AvALLOC(av));
6223 } else if (SvTYPE(iter_sv) == SVt_PVHV) {
6224 sv = Perl_hfree_next_entry(aTHX_ (HV*)iter_sv, &hash_index);
6225 if (!sv && !HvTOTALKEYS((HV *)iter_sv)) {
6226 /* no more elements of current HV to free */
6229 /* Restore previous values of iter_sv and hash_index,
6230 * squirrelled away */
6231 assert(!SvOBJECT(sv));
6232 iter_sv = (SV*)SvSTASH(sv);
6233 assert(!SvMAGICAL(sv));
6234 hash_index = ((XPVMG*) SvANY(sv))->xmg_u.xmg_hash_index;
6236 /* perl -DA does not like rubbish in SvMAGIC. */
6240 /* free any remaining detritus from the hash struct */
6241 Perl_hv_undef_flags(aTHX_ MUTABLE_HV(sv), HV_NAME_SETALL);
6242 assert(!HvARRAY((HV*)sv));
6247 /* unrolled SvREFCNT_dec and sv_free2 follows: */
6251 if (!SvREFCNT(sv)) {
6255 if (--(SvREFCNT(sv)))
6259 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEBUGGING),
6260 "Attempt to free temp prematurely: SV 0x%"UVxf
6261 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
6265 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
6266 /* make sure SvREFCNT(sv)==0 happens very seldom */
6267 SvREFCNT(sv) = (~(U32)0)/2;
6276 /* This routine curses the sv itself, not the object referenced by sv. So
6277 sv does not have to be ROK. */
6280 S_curse(pTHX_ SV * const sv, const bool check_refcnt) {
6283 PERL_ARGS_ASSERT_CURSE;
6284 assert(SvOBJECT(sv));
6286 if (PL_defstash && /* Still have a symbol table? */
6293 stash = SvSTASH(sv);
6294 destructor = StashHANDLER(stash,DESTROY);
6296 /* A constant subroutine can have no side effects, so
6297 don't bother calling it. */
6298 && !CvCONST(destructor)
6299 /* Don't bother calling an empty destructor or one that
6300 returns immediately. */
6301 && (CvISXSUB(destructor)
6302 || (CvSTART(destructor)
6303 && (CvSTART(destructor)->op_next->op_type
6305 && (CvSTART(destructor)->op_next->op_type
6307 || CvSTART(destructor)->op_next->op_next->op_type
6313 SV* const tmpref = newRV(sv);
6314 SvREADONLY_on(tmpref); /* DESTROY() could be naughty */
6316 PUSHSTACKi(PERLSI_DESTROY);
6321 call_sv(MUTABLE_SV(destructor),
6322 G_DISCARD|G_EVAL|G_KEEPERR|G_VOID);
6326 if(SvREFCNT(tmpref) < 2) {
6327 /* tmpref is not kept alive! */
6329 SvRV_set(tmpref, NULL);
6332 SvREFCNT_dec(tmpref);
6334 } while (SvOBJECT(sv) && SvSTASH(sv) != stash);
6337 if (check_refcnt && SvREFCNT(sv)) {
6338 if (PL_in_clean_objs)
6340 "DESTROY created new reference to dead object '%"HEKf"'",
6341 HEKfARG(HvNAME_HEK(stash)));
6342 /* DESTROY gave object new lease on life */
6348 SvREFCNT_dec(SvSTASH(sv)); /* possibly of changed persuasion */
6349 SvOBJECT_off(sv); /* Curse the object. */
6350 if (SvTYPE(sv) != SVt_PVIO)
6351 --PL_sv_objcount;/* XXX Might want something more general */
6357 =for apidoc sv_newref
6359 Increment an SV's reference count. Use the C<SvREFCNT_inc()> wrapper
6366 Perl_sv_newref(pTHX_ SV *const sv)
6368 PERL_UNUSED_CONTEXT;
6377 Decrement an SV's reference count, and if it drops to zero, call
6378 C<sv_clear> to invoke destructors and free up any memory used by
6379 the body; finally, deallocate the SV's head itself.
6380 Normally called via a wrapper macro C<SvREFCNT_dec>.
6386 Perl_sv_free(pTHX_ SV *const sv)
6391 if (SvREFCNT(sv) == 0) {
6392 if (SvFLAGS(sv) & SVf_BREAK)
6393 /* this SV's refcnt has been artificially decremented to
6394 * trigger cleanup */
6396 if (PL_in_clean_all) /* All is fair */
6398 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
6399 /* make sure SvREFCNT(sv)==0 happens very seldom */
6400 SvREFCNT(sv) = (~(U32)0)/2;
6403 if (ckWARN_d(WARN_INTERNAL)) {
6404 #ifdef DEBUG_LEAKING_SCALARS_FORK_DUMP
6405 Perl_dump_sv_child(aTHX_ sv);
6407 #ifdef DEBUG_LEAKING_SCALARS
6410 #ifdef DEBUG_LEAKING_SCALARS_ABORT
6411 if (PL_warnhook == PERL_WARNHOOK_FATAL
6412 || ckDEAD(packWARN(WARN_INTERNAL))) {
6413 /* Don't let Perl_warner cause us to escape our fate: */
6417 /* This may not return: */
6418 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
6419 "Attempt to free unreferenced scalar: SV 0x%"UVxf
6420 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
6423 #ifdef DEBUG_LEAKING_SCALARS_ABORT
6428 if (--(SvREFCNT(sv)) > 0)
6430 Perl_sv_free2(aTHX_ sv);
6434 Perl_sv_free2(pTHX_ SV *const sv)
6438 PERL_ARGS_ASSERT_SV_FREE2;
6442 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEBUGGING),
6443 "Attempt to free temp prematurely: SV 0x%"UVxf
6444 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
6448 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
6449 /* make sure SvREFCNT(sv)==0 happens very seldom */
6450 SvREFCNT(sv) = (~(U32)0)/2;
6461 Returns the length of the string in the SV. Handles magic and type
6462 coercion. See also C<SvCUR>, which gives raw access to the xpv_cur slot.
6468 Perl_sv_len(pTHX_ register SV *const sv)
6476 len = mg_length(sv);
6478 (void)SvPV_const(sv, len);
6483 =for apidoc sv_len_utf8
6485 Returns the number of characters in the string in an SV, counting wide
6486 UTF-8 bytes as a single character. Handles magic and type coercion.
6492 * The length is cached in PERL_MAGIC_utf8, in the mg_len field. Also the
6493 * mg_ptr is used, by sv_pos_u2b() and sv_pos_b2u() - see the comments below.
6494 * (Note that the mg_len is not the length of the mg_ptr field.
6495 * This allows the cache to store the character length of the string without
6496 * needing to malloc() extra storage to attach to the mg_ptr.)
6501 Perl_sv_len_utf8(pTHX_ register SV *const sv)
6507 return mg_length(sv);
6511 return sv_len_utf8_nomg(sv);
6516 Perl_sv_len_utf8_nomg(pTHX_ SV * const sv)
6520 const U8 *s = (U8*)SvPV_nomg_const(sv, len);
6522 PERL_ARGS_ASSERT_SV_LEN_UTF8_NOMG;
6526 MAGIC *mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_utf8) : NULL;
6528 if (mg && (mg->mg_len != -1 || mg->mg_ptr)) {
6529 if (mg->mg_len != -1)
6532 /* We can use the offset cache for a headstart.
6533 The longer value is stored in the first pair. */
6534 STRLEN *cache = (STRLEN *) mg->mg_ptr;
6536 ulen = cache[0] + Perl_utf8_length(aTHX_ s + cache[1],
6540 if (PL_utf8cache < 0) {
6541 const STRLEN real = Perl_utf8_length(aTHX_ s, s + len);
6542 assert_uft8_cache_coherent("sv_len_utf8", ulen, real, sv);
6546 ulen = Perl_utf8_length(aTHX_ s, s + len);
6547 utf8_mg_len_cache_update(sv, &mg, ulen);
6551 return Perl_utf8_length(aTHX_ s, s + len);
6554 /* Walk forwards to find the byte corresponding to the passed in UTF-8
6557 S_sv_pos_u2b_forwards(const U8 *const start, const U8 *const send,
6558 STRLEN *const uoffset_p, bool *const at_end)
6560 const U8 *s = start;
6561 STRLEN uoffset = *uoffset_p;
6563 PERL_ARGS_ASSERT_SV_POS_U2B_FORWARDS;
6565 while (s < send && uoffset) {
6572 else if (s > send) {
6574 /* This is the existing behaviour. Possibly it should be a croak, as
6575 it's actually a bounds error */
6578 *uoffset_p -= uoffset;
6582 /* Given the length of the string in both bytes and UTF-8 characters, decide
6583 whether to walk forwards or backwards to find the byte corresponding to
6584 the passed in UTF-8 offset. */
6586 S_sv_pos_u2b_midway(const U8 *const start, const U8 *send,
6587 STRLEN uoffset, const STRLEN uend)
6589 STRLEN backw = uend - uoffset;
6591 PERL_ARGS_ASSERT_SV_POS_U2B_MIDWAY;
6593 if (uoffset < 2 * backw) {
6594 /* The assumption is that going forwards is twice the speed of going
6595 forward (that's where the 2 * backw comes from).
6596 (The real figure of course depends on the UTF-8 data.) */
6597 const U8 *s = start;
6599 while (s < send && uoffset--)
6609 while (UTF8_IS_CONTINUATION(*send))
6612 return send - start;
6615 /* For the string representation of the given scalar, find the byte
6616 corresponding to the passed in UTF-8 offset. uoffset0 and boffset0
6617 give another position in the string, *before* the sought offset, which
6618 (which is always true, as 0, 0 is a valid pair of positions), which should
6619 help reduce the amount of linear searching.
6620 If *mgp is non-NULL, it should point to the UTF-8 cache magic, which
6621 will be used to reduce the amount of linear searching. The cache will be
6622 created if necessary, and the found value offered to it for update. */
6624 S_sv_pos_u2b_cached(pTHX_ SV *const sv, MAGIC **const mgp, const U8 *const start,
6625 const U8 *const send, STRLEN uoffset,
6626 STRLEN uoffset0, STRLEN boffset0)
6628 STRLEN boffset = 0; /* Actually always set, but let's keep gcc happy. */
6630 bool at_end = FALSE;
6632 PERL_ARGS_ASSERT_SV_POS_U2B_CACHED;
6634 assert (uoffset >= uoffset0);
6641 && (*mgp || (SvTYPE(sv) >= SVt_PVMG &&
6642 (*mgp = mg_find(sv, PERL_MAGIC_utf8))))) {
6643 if ((*mgp)->mg_ptr) {
6644 STRLEN *cache = (STRLEN *) (*mgp)->mg_ptr;
6645 if (cache[0] == uoffset) {
6646 /* An exact match. */
6649 if (cache[2] == uoffset) {
6650 /* An exact match. */
6654 if (cache[0] < uoffset) {
6655 /* The cache already knows part of the way. */
6656 if (cache[0] > uoffset0) {
6657 /* The cache knows more than the passed in pair */
6658 uoffset0 = cache[0];
6659 boffset0 = cache[1];
6661 if ((*mgp)->mg_len != -1) {
6662 /* And we know the end too. */
6664 + sv_pos_u2b_midway(start + boffset0, send,
6666 (*mgp)->mg_len - uoffset0);
6668 uoffset -= uoffset0;
6670 + sv_pos_u2b_forwards(start + boffset0,
6671 send, &uoffset, &at_end);
6672 uoffset += uoffset0;
6675 else if (cache[2] < uoffset) {
6676 /* We're between the two cache entries. */
6677 if (cache[2] > uoffset0) {
6678 /* and the cache knows more than the passed in pair */
6679 uoffset0 = cache[2];
6680 boffset0 = cache[3];
6684 + sv_pos_u2b_midway(start + boffset0,
6687 cache[0] - uoffset0);
6690 + sv_pos_u2b_midway(start + boffset0,
6693 cache[2] - uoffset0);
6697 else if ((*mgp)->mg_len != -1) {
6698 /* If we can take advantage of a passed in offset, do so. */
6699 /* In fact, offset0 is either 0, or less than offset, so don't
6700 need to worry about the other possibility. */
6702 + sv_pos_u2b_midway(start + boffset0, send,
6704 (*mgp)->mg_len - uoffset0);
6709 if (!found || PL_utf8cache < 0) {
6710 STRLEN real_boffset;
6711 uoffset -= uoffset0;
6712 real_boffset = boffset0 + sv_pos_u2b_forwards(start + boffset0,
6713 send, &uoffset, &at_end);
6714 uoffset += uoffset0;
6716 if (found && PL_utf8cache < 0)
6717 assert_uft8_cache_coherent("sv_pos_u2b_cache", boffset,
6719 boffset = real_boffset;
6724 utf8_mg_len_cache_update(sv, mgp, uoffset);
6726 utf8_mg_pos_cache_update(sv, mgp, boffset, uoffset, send - start);
6733 =for apidoc sv_pos_u2b_flags
6735 Converts the value pointed to by offsetp from a count of UTF-8 chars from
6736 the start of the string, to a count of the equivalent number of bytes; if
6737 lenp is non-zero, it does the same to lenp, but this time starting from
6738 the offset, rather than from the start
6739 of the string. Handles type coercion.
6740 I<flags> is passed to C<SvPV_flags>, and usually should be
6741 C<SV_GMAGIC|SV_CONST_RETURN> to handle magic.
6747 * sv_pos_u2b_flags() uses, like sv_pos_b2u(), the mg_ptr of the potential
6748 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6749 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
6754 Perl_sv_pos_u2b_flags(pTHX_ SV *const sv, STRLEN uoffset, STRLEN *const lenp,
6761 PERL_ARGS_ASSERT_SV_POS_U2B_FLAGS;
6763 start = (U8*)SvPV_flags(sv, len, flags);
6765 const U8 * const send = start + len;
6767 boffset = sv_pos_u2b_cached(sv, &mg, start, send, uoffset, 0, 0);
6770 && *lenp /* don't bother doing work for 0, as its bytes equivalent
6771 is 0, and *lenp is already set to that. */) {
6772 /* Convert the relative offset to absolute. */
6773 const STRLEN uoffset2 = uoffset + *lenp;
6774 const STRLEN boffset2
6775 = sv_pos_u2b_cached(sv, &mg, start, send, uoffset2,
6776 uoffset, boffset) - boffset;
6790 =for apidoc sv_pos_u2b
6792 Converts the value pointed to by offsetp from a count of UTF-8 chars from
6793 the start of the string, to a count of the equivalent number of bytes; if
6794 lenp is non-zero, it does the same to lenp, but this time starting from
6795 the offset, rather than from the start of the string. Handles magic and
6798 Use C<sv_pos_u2b_flags> in preference, which correctly handles strings longer
6805 * sv_pos_u2b() uses, like sv_pos_b2u(), the mg_ptr of the potential
6806 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6807 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
6811 /* This function is subject to size and sign problems */
6814 Perl_sv_pos_u2b(pTHX_ register SV *const sv, I32 *const offsetp, I32 *const lenp)
6816 PERL_ARGS_ASSERT_SV_POS_U2B;
6819 STRLEN ulen = (STRLEN)*lenp;
6820 *offsetp = (I32)sv_pos_u2b_flags(sv, (STRLEN)*offsetp, &ulen,
6821 SV_GMAGIC|SV_CONST_RETURN);
6824 *offsetp = (I32)sv_pos_u2b_flags(sv, (STRLEN)*offsetp, NULL,
6825 SV_GMAGIC|SV_CONST_RETURN);
6830 S_utf8_mg_len_cache_update(pTHX_ SV *const sv, MAGIC **const mgp,
6833 PERL_ARGS_ASSERT_UTF8_MG_LEN_CACHE_UPDATE;
6837 if (!*mgp && (SvTYPE(sv) < SVt_PVMG ||
6838 !(*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
6839 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, &PL_vtbl_utf8, 0, 0);
6843 (*mgp)->mg_len = ulen;
6844 /* For now, treat "overflowed" as "still unknown". See RT #72924. */
6845 if (ulen != (STRLEN) (*mgp)->mg_len)
6846 (*mgp)->mg_len = -1;
6849 /* Create and update the UTF8 magic offset cache, with the proffered utf8/
6850 byte length pairing. The (byte) length of the total SV is passed in too,
6851 as blen, because for some (more esoteric) SVs, the call to SvPV_const()
6852 may not have updated SvCUR, so we can't rely on reading it directly.
6854 The proffered utf8/byte length pairing isn't used if the cache already has
6855 two pairs, and swapping either for the proffered pair would increase the
6856 RMS of the intervals between known byte offsets.
6858 The cache itself consists of 4 STRLEN values
6859 0: larger UTF-8 offset
6860 1: corresponding byte offset
6861 2: smaller UTF-8 offset
6862 3: corresponding byte offset
6864 Unused cache pairs have the value 0, 0.
6865 Keeping the cache "backwards" means that the invariant of
6866 cache[0] >= cache[2] is maintained even with empty slots, which means that
6867 the code that uses it doesn't need to worry if only 1 entry has actually
6868 been set to non-zero. It also makes the "position beyond the end of the
6869 cache" logic much simpler, as the first slot is always the one to start
6873 S_utf8_mg_pos_cache_update(pTHX_ SV *const sv, MAGIC **const mgp, const STRLEN byte,
6874 const STRLEN utf8, const STRLEN blen)
6878 PERL_ARGS_ASSERT_UTF8_MG_POS_CACHE_UPDATE;
6883 if (!*mgp && (SvTYPE(sv) < SVt_PVMG ||
6884 !(*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
6885 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, (MGVTBL*)&PL_vtbl_utf8, 0,
6887 (*mgp)->mg_len = -1;
6891 if (!(cache = (STRLEN *)(*mgp)->mg_ptr)) {
6892 Newxz(cache, PERL_MAGIC_UTF8_CACHESIZE * 2, STRLEN);
6893 (*mgp)->mg_ptr = (char *) cache;
6897 if (PL_utf8cache < 0 && SvPOKp(sv)) {
6898 /* SvPOKp() because it's possible that sv has string overloading, and
6899 therefore is a reference, hence SvPVX() is actually a pointer.
6900 This cures the (very real) symptoms of RT 69422, but I'm not actually
6901 sure whether we should even be caching the results of UTF-8
6902 operations on overloading, given that nothing stops overloading
6903 returning a different value every time it's called. */
6904 const U8 *start = (const U8 *) SvPVX_const(sv);
6905 const STRLEN realutf8 = utf8_length(start, start + byte);
6907 assert_uft8_cache_coherent("utf8_mg_pos_cache_update", utf8, realutf8,
6911 /* Cache is held with the later position first, to simplify the code
6912 that deals with unbounded ends. */
6914 ASSERT_UTF8_CACHE(cache);
6915 if (cache[1] == 0) {
6916 /* Cache is totally empty */
6919 } else if (cache[3] == 0) {
6920 if (byte > cache[1]) {
6921 /* New one is larger, so goes first. */
6922 cache[2] = cache[0];
6923 cache[3] = cache[1];
6931 #define THREEWAY_SQUARE(a,b,c,d) \
6932 ((float)((d) - (c))) * ((float)((d) - (c))) \
6933 + ((float)((c) - (b))) * ((float)((c) - (b))) \
6934 + ((float)((b) - (a))) * ((float)((b) - (a)))
6936 /* Cache has 2 slots in use, and we know three potential pairs.
6937 Keep the two that give the lowest RMS distance. Do the
6938 calculation in bytes simply because we always know the byte
6939 length. squareroot has the same ordering as the positive value,
6940 so don't bother with the actual square root. */
6941 if (byte > cache[1]) {
6942 /* New position is after the existing pair of pairs. */
6943 const float keep_earlier
6944 = THREEWAY_SQUARE(0, cache[3], byte, blen);
6945 const float keep_later
6946 = THREEWAY_SQUARE(0, cache[1], byte, blen);
6948 if (keep_later < keep_earlier) {
6949 cache[2] = cache[0];
6950 cache[3] = cache[1];
6959 else if (byte > cache[3]) {
6960 /* New position is between the existing pair of pairs. */
6961 const float keep_earlier
6962 = THREEWAY_SQUARE(0, cache[3], byte, blen);
6963 const float keep_later
6964 = THREEWAY_SQUARE(0, byte, cache[1], blen);
6966 if (keep_later < keep_earlier) {
6976 /* New position is before the existing pair of pairs. */
6977 const float keep_earlier
6978 = THREEWAY_SQUARE(0, byte, cache[3], blen);
6979 const float keep_later
6980 = THREEWAY_SQUARE(0, byte, cache[1], blen);
6982 if (keep_later < keep_earlier) {
6987 cache[0] = cache[2];
6988 cache[1] = cache[3];
6994 ASSERT_UTF8_CACHE(cache);
6997 /* We already know all of the way, now we may be able to walk back. The same
6998 assumption is made as in S_sv_pos_u2b_midway(), namely that walking
6999 backward is half the speed of walking forward. */
7001 S_sv_pos_b2u_midway(pTHX_ const U8 *const s, const U8 *const target,
7002 const U8 *end, STRLEN endu)
7004 const STRLEN forw = target - s;
7005 STRLEN backw = end - target;
7007 PERL_ARGS_ASSERT_SV_POS_B2U_MIDWAY;
7009 if (forw < 2 * backw) {
7010 return utf8_length(s, target);
7013 while (end > target) {
7015 while (UTF8_IS_CONTINUATION(*end)) {
7024 =for apidoc sv_pos_b2u
7026 Converts the value pointed to by offsetp from a count of bytes from the
7027 start of the string, to a count of the equivalent number of UTF-8 chars.
7028 Handles magic and type coercion.
7034 * sv_pos_b2u() uses, like sv_pos_u2b(), the mg_ptr of the potential
7035 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
7040 Perl_sv_pos_b2u(pTHX_ register SV *const sv, I32 *const offsetp)
7043 const STRLEN byte = *offsetp;
7044 STRLEN len = 0; /* Actually always set, but let's keep gcc happy. */
7050 PERL_ARGS_ASSERT_SV_POS_B2U;
7055 s = (const U8*)SvPV_const(sv, blen);
7058 Perl_croak(aTHX_ "panic: sv_pos_b2u: bad byte offset, blen=%"UVuf
7059 ", byte=%"UVuf, (UV)blen, (UV)byte);
7065 && SvTYPE(sv) >= SVt_PVMG
7066 && (mg = mg_find(sv, PERL_MAGIC_utf8)))
7069 STRLEN * const cache = (STRLEN *) mg->mg_ptr;
7070 if (cache[1] == byte) {
7071 /* An exact match. */
7072 *offsetp = cache[0];
7075 if (cache[3] == byte) {
7076 /* An exact match. */
7077 *offsetp = cache[2];
7081 if (cache[1] < byte) {
7082 /* We already know part of the way. */
7083 if (mg->mg_len != -1) {
7084 /* Actually, we know the end too. */
7086 + S_sv_pos_b2u_midway(aTHX_ s + cache[1], send,
7087 s + blen, mg->mg_len - cache[0]);
7089 len = cache[0] + utf8_length(s + cache[1], send);
7092 else if (cache[3] < byte) {
7093 /* We're between the two cached pairs, so we do the calculation
7094 offset by the byte/utf-8 positions for the earlier pair,
7095 then add the utf-8 characters from the string start to
7097 len = S_sv_pos_b2u_midway(aTHX_ s + cache[3], send,
7098 s + cache[1], cache[0] - cache[2])
7102 else { /* cache[3] > byte */
7103 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + cache[3],
7107 ASSERT_UTF8_CACHE(cache);
7109 } else if (mg->mg_len != -1) {
7110 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + blen, mg->mg_len);
7114 if (!found || PL_utf8cache < 0) {
7115 const STRLEN real_len = utf8_length(s, send);
7117 if (found && PL_utf8cache < 0)
7118 assert_uft8_cache_coherent("sv_pos_b2u", len, real_len, sv);
7125 utf8_mg_len_cache_update(sv, &mg, len);
7127 utf8_mg_pos_cache_update(sv, &mg, byte, len, blen);
7132 S_assert_uft8_cache_coherent(pTHX_ const char *const func, STRLEN from_cache,
7133 STRLEN real, SV *const sv)
7135 PERL_ARGS_ASSERT_ASSERT_UFT8_CACHE_COHERENT;
7137 /* As this is debugging only code, save space by keeping this test here,
7138 rather than inlining it in all the callers. */
7139 if (from_cache == real)
7142 /* Need to turn the assertions off otherwise we may recurse infinitely
7143 while printing error messages. */
7144 SAVEI8(PL_utf8cache);
7146 Perl_croak(aTHX_ "panic: %s cache %"UVuf" real %"UVuf" for %"SVf,
7147 func, (UV) from_cache, (UV) real, SVfARG(sv));
7153 Returns a boolean indicating whether the strings in the two SVs are
7154 identical. Is UTF-8 and 'use bytes' aware, handles get magic, and will
7155 coerce its args to strings if necessary.
7157 =for apidoc sv_eq_flags
7159 Returns a boolean indicating whether the strings in the two SVs are
7160 identical. Is UTF-8 and 'use bytes' aware and coerces its args to strings
7161 if necessary. If the flags include SV_GMAGIC, it handles get-magic, too.
7167 Perl_sv_eq_flags(pTHX_ register SV *sv1, register SV *sv2, const U32 flags)
7175 SV* svrecode = NULL;
7182 /* if pv1 and pv2 are the same, second SvPV_const call may
7183 * invalidate pv1 (if we are handling magic), so we may need to
7185 if (sv1 == sv2 && flags & SV_GMAGIC
7186 && (SvTHINKFIRST(sv1) || SvGMAGICAL(sv1))) {
7187 pv1 = SvPV_const(sv1, cur1);
7188 sv1 = newSVpvn_flags(pv1, cur1, SVs_TEMP | SvUTF8(sv2));
7190 pv1 = SvPV_flags_const(sv1, cur1, flags);
7198 pv2 = SvPV_flags_const(sv2, cur2, flags);
7200 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
7201 /* Differing utf8ness.
7202 * Do not UTF8size the comparands as a side-effect. */
7205 svrecode = newSVpvn(pv2, cur2);
7206 sv_recode_to_utf8(svrecode, PL_encoding);
7207 pv2 = SvPV_const(svrecode, cur2);
7210 svrecode = newSVpvn(pv1, cur1);
7211 sv_recode_to_utf8(svrecode, PL_encoding);
7212 pv1 = SvPV_const(svrecode, cur1);
7214 /* Now both are in UTF-8. */
7216 SvREFCNT_dec(svrecode);
7222 /* sv1 is the UTF-8 one */
7223 return bytes_cmp_utf8((const U8*)pv2, cur2,
7224 (const U8*)pv1, cur1) == 0;
7227 /* sv2 is the UTF-8 one */
7228 return bytes_cmp_utf8((const U8*)pv1, cur1,
7229 (const U8*)pv2, cur2) == 0;
7235 eq = (pv1 == pv2) || memEQ(pv1, pv2, cur1);
7237 SvREFCNT_dec(svrecode);
7245 Compares the strings in two SVs. Returns -1, 0, or 1 indicating whether the
7246 string in C<sv1> is less than, equal to, or greater than the string in
7247 C<sv2>. Is UTF-8 and 'use bytes' aware, handles get magic, and will
7248 coerce its args to strings if necessary. See also C<sv_cmp_locale>.
7250 =for apidoc sv_cmp_flags
7252 Compares the strings in two SVs. Returns -1, 0, or 1 indicating whether the
7253 string in C<sv1> is less than, equal to, or greater than the string in
7254 C<sv2>. Is UTF-8 and 'use bytes' aware and will coerce its args to strings
7255 if necessary. If the flags include SV_GMAGIC, it handles get magic. See
7256 also C<sv_cmp_locale_flags>.
7262 Perl_sv_cmp(pTHX_ register SV *const sv1, register SV *const sv2)
7264 return sv_cmp_flags(sv1, sv2, SV_GMAGIC);
7268 Perl_sv_cmp_flags(pTHX_ register SV *const sv1, register SV *const sv2,
7273 const char *pv1, *pv2;
7276 SV *svrecode = NULL;
7283 pv1 = SvPV_flags_const(sv1, cur1, flags);
7290 pv2 = SvPV_flags_const(sv2, cur2, flags);
7292 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
7293 /* Differing utf8ness.
7294 * Do not UTF8size the comparands as a side-effect. */
7297 svrecode = newSVpvn(pv2, cur2);
7298 sv_recode_to_utf8(svrecode, PL_encoding);
7299 pv2 = SvPV_const(svrecode, cur2);
7302 const int retval = -bytes_cmp_utf8((const U8*)pv2, cur2,
7303 (const U8*)pv1, cur1);
7304 return retval ? retval < 0 ? -1 : +1 : 0;
7309 svrecode = newSVpvn(pv1, cur1);
7310 sv_recode_to_utf8(svrecode, PL_encoding);
7311 pv1 = SvPV_const(svrecode, cur1);
7314 const int retval = bytes_cmp_utf8((const U8*)pv1, cur1,
7315 (const U8*)pv2, cur2);
7316 return retval ? retval < 0 ? -1 : +1 : 0;
7322 cmp = cur2 ? -1 : 0;
7326 const I32 retval = memcmp((const void*)pv1, (const void*)pv2, cur1 < cur2 ? cur1 : cur2);
7329 cmp = retval < 0 ? -1 : 1;
7330 } else if (cur1 == cur2) {
7333 cmp = cur1 < cur2 ? -1 : 1;
7337 SvREFCNT_dec(svrecode);
7345 =for apidoc sv_cmp_locale
7347 Compares the strings in two SVs in a locale-aware manner. Is UTF-8 and
7348 'use bytes' aware, handles get magic, and will coerce its args to strings
7349 if necessary. See also C<sv_cmp>.
7351 =for apidoc sv_cmp_locale_flags
7353 Compares the strings in two SVs in a locale-aware manner. Is UTF-8 and
7354 'use bytes' aware and will coerce its args to strings if necessary. If the
7355 flags contain SV_GMAGIC, it handles get magic. See also C<sv_cmp_flags>.
7361 Perl_sv_cmp_locale(pTHX_ register SV *const sv1, register SV *const sv2)
7363 return sv_cmp_locale_flags(sv1, sv2, SV_GMAGIC);
7367 Perl_sv_cmp_locale_flags(pTHX_ register SV *const sv1, register SV *const sv2,
7371 #ifdef USE_LOCALE_COLLATE
7377 if (PL_collation_standard)
7381 pv1 = sv1 ? sv_collxfrm_flags(sv1, &len1, flags) : (char *) NULL;
7383 pv2 = sv2 ? sv_collxfrm_flags(sv2, &len2, flags) : (char *) NULL;
7385 if (!pv1 || !len1) {
7396 retval = memcmp((void*)pv1, (void*)pv2, len1 < len2 ? len1 : len2);
7399 return retval < 0 ? -1 : 1;
7402 * When the result of collation is equality, that doesn't mean
7403 * that there are no differences -- some locales exclude some
7404 * characters from consideration. So to avoid false equalities,
7405 * we use the raw string as a tiebreaker.
7411 #endif /* USE_LOCALE_COLLATE */
7413 return sv_cmp(sv1, sv2);
7417 #ifdef USE_LOCALE_COLLATE
7420 =for apidoc sv_collxfrm
7422 This calls C<sv_collxfrm_flags> with the SV_GMAGIC flag. See
7423 C<sv_collxfrm_flags>.
7425 =for apidoc sv_collxfrm_flags
7427 Add Collate Transform magic to an SV if it doesn't already have it. If the
7428 flags contain SV_GMAGIC, it handles get-magic.
7430 Any scalar variable may carry PERL_MAGIC_collxfrm magic that contains the
7431 scalar data of the variable, but transformed to such a format that a normal
7432 memory comparison can be used to compare the data according to the locale
7439 Perl_sv_collxfrm_flags(pTHX_ SV *const sv, STRLEN *const nxp, const I32 flags)
7444 PERL_ARGS_ASSERT_SV_COLLXFRM_FLAGS;
7446 mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_collxfrm) : (MAGIC *) NULL;
7447 if (!mg || !mg->mg_ptr || *(U32*)mg->mg_ptr != PL_collation_ix) {
7453 Safefree(mg->mg_ptr);
7454 s = SvPV_flags_const(sv, len, flags);
7455 if ((xf = mem_collxfrm(s, len, &xlen))) {
7457 #ifdef PERL_OLD_COPY_ON_WRITE
7459 sv_force_normal_flags(sv, 0);
7461 mg = sv_magicext(sv, 0, PERL_MAGIC_collxfrm, &PL_vtbl_collxfrm,
7475 if (mg && mg->mg_ptr) {
7477 return mg->mg_ptr + sizeof(PL_collation_ix);
7485 #endif /* USE_LOCALE_COLLATE */
7488 S_sv_gets_append_to_utf8(pTHX_ SV *const sv, PerlIO *const fp, I32 append)
7490 SV * const tsv = newSV(0);
7493 sv_gets(tsv, fp, 0);
7494 sv_utf8_upgrade_nomg(tsv);
7495 SvCUR_set(sv,append);
7498 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7502 S_sv_gets_read_record(pTHX_ SV *const sv, PerlIO *const fp, I32 append)
7505 const U32 recsize = SvUV(SvRV(PL_rs)); /* RsRECORD() guarantees > 0. */
7506 /* Grab the size of the record we're getting */
7507 char *const buffer = SvGROW(sv, (STRLEN)(recsize + append + 1)) + append;
7514 /* VMS wants read instead of fread, because fread doesn't respect */
7515 /* RMS record boundaries. This is not necessarily a good thing to be */
7516 /* doing, but we've got no other real choice - except avoid stdio
7517 as implementation - perhaps write a :vms layer ?
7519 fd = PerlIO_fileno(fp);
7521 bytesread = PerlLIO_read(fd, buffer, recsize);
7523 else /* in-memory file from PerlIO::Scalar */
7526 bytesread = PerlIO_read(fp, buffer, recsize);
7531 SvCUR_set(sv, bytesread + append);
7532 buffer[bytesread] = '\0';
7533 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7539 Get a line from the filehandle and store it into the SV, optionally
7540 appending to the currently-stored string. If C<append> is not 0, the
7541 line is appended to the SV instead of overwriting it. C<append> should
7542 be set to the byte offset that the appended string should start at
7543 in the SV (typically, C<SvCUR(sv)> is a suitable choice).
7549 Perl_sv_gets(pTHX_ register SV *const sv, register PerlIO *const fp, I32 append)
7560 PERL_ARGS_ASSERT_SV_GETS;
7562 if (SvTHINKFIRST(sv))
7563 sv_force_normal_flags(sv, append ? 0 : SV_COW_DROP_PV);
7564 /* XXX. If you make this PVIV, then copy on write can copy scalars read
7566 However, perlbench says it's slower, because the existing swipe code
7567 is faster than copy on write.
7568 Swings and roundabouts. */
7569 SvUPGRADE(sv, SVt_PV);
7572 if (PerlIO_isutf8(fp)) {
7574 sv_utf8_upgrade_nomg(sv);
7575 sv_pos_u2b(sv,&append,0);
7577 } else if (SvUTF8(sv)) {
7578 return S_sv_gets_append_to_utf8(aTHX_ sv, fp, append);
7586 if (PerlIO_isutf8(fp))
7589 if (IN_PERL_COMPILETIME) {
7590 /* we always read code in line mode */
7594 else if (RsSNARF(PL_rs)) {
7595 /* If it is a regular disk file use size from stat() as estimate
7596 of amount we are going to read -- may result in mallocing
7597 more memory than we really need if the layers below reduce
7598 the size we read (e.g. CRLF or a gzip layer).
7601 if (!PerlLIO_fstat(PerlIO_fileno(fp), &st) && S_ISREG(st.st_mode)) {
7602 const Off_t offset = PerlIO_tell(fp);
7603 if (offset != (Off_t) -1 && st.st_size + append > offset) {
7604 (void) SvGROW(sv, (STRLEN)((st.st_size - offset) + append + 1));
7610 else if (RsRECORD(PL_rs)) {
7611 return S_sv_gets_read_record(aTHX_ sv, fp, append);
7613 else if (RsPARA(PL_rs)) {
7619 /* Get $/ i.e. PL_rs into same encoding as stream wants */
7620 if (PerlIO_isutf8(fp)) {
7621 rsptr = SvPVutf8(PL_rs, rslen);
7624 if (SvUTF8(PL_rs)) {
7625 if (!sv_utf8_downgrade(PL_rs, TRUE)) {
7626 Perl_croak(aTHX_ "Wide character in $/");
7629 rsptr = SvPV_const(PL_rs, rslen);
7633 rslast = rslen ? rsptr[rslen - 1] : '\0';
7635 if (rspara) { /* have to do this both before and after */
7636 do { /* to make sure file boundaries work right */
7639 i = PerlIO_getc(fp);
7643 PerlIO_ungetc(fp,i);
7649 /* See if we know enough about I/O mechanism to cheat it ! */
7651 /* This used to be #ifdef test - it is made run-time test for ease
7652 of abstracting out stdio interface. One call should be cheap
7653 enough here - and may even be a macro allowing compile
7657 if (PerlIO_fast_gets(fp)) {
7660 * We're going to steal some values from the stdio struct
7661 * and put EVERYTHING in the innermost loop into registers.
7667 #if defined(VMS) && defined(PERLIO_IS_STDIO)
7668 /* An ungetc()d char is handled separately from the regular
7669 * buffer, so we getc() it back out and stuff it in the buffer.
7671 i = PerlIO_getc(fp);
7672 if (i == EOF) return 0;
7673 *(--((*fp)->_ptr)) = (unsigned char) i;
7677 /* Here is some breathtakingly efficient cheating */
7679 cnt = PerlIO_get_cnt(fp); /* get count into register */
7680 /* make sure we have the room */
7681 if ((I32)(SvLEN(sv) - append) <= cnt + 1) {
7682 /* Not room for all of it
7683 if we are looking for a separator and room for some
7685 if (rslen && cnt > 80 && (I32)SvLEN(sv) > append) {
7686 /* just process what we have room for */
7687 shortbuffered = cnt - SvLEN(sv) + append + 1;
7688 cnt -= shortbuffered;
7692 /* remember that cnt can be negative */
7693 SvGROW(sv, (STRLEN)(append + (cnt <= 0 ? 2 : (cnt + 1))));
7698 bp = (STDCHAR*)SvPVX_const(sv) + append; /* move these two too to registers */
7699 ptr = (STDCHAR*)PerlIO_get_ptr(fp);
7700 DEBUG_P(PerlIO_printf(Perl_debug_log,
7701 "Screamer: entering, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7702 DEBUG_P(PerlIO_printf(Perl_debug_log,
7703 "Screamer: entering: PerlIO * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7704 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7705 PTR2UV(PerlIO_has_base(fp) ? PerlIO_get_base(fp) : 0)));
7710 while (cnt > 0) { /* this | eat */
7712 if ((*bp++ = *ptr++) == rslast) /* really | dust */
7713 goto thats_all_folks; /* screams | sed :-) */
7717 Copy(ptr, bp, cnt, char); /* this | eat */
7718 bp += cnt; /* screams | dust */
7719 ptr += cnt; /* louder | sed :-) */
7721 assert (!shortbuffered);
7722 goto cannot_be_shortbuffered;
7726 if (shortbuffered) { /* oh well, must extend */
7727 cnt = shortbuffered;
7729 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
7731 SvGROW(sv, SvLEN(sv) + append + cnt + 2);
7732 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
7736 cannot_be_shortbuffered:
7737 DEBUG_P(PerlIO_printf(Perl_debug_log,
7738 "Screamer: going to getc, ptr=%"UVuf", cnt=%ld\n",
7739 PTR2UV(ptr),(long)cnt));
7740 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* deregisterize cnt and ptr */
7742 DEBUG_Pv(PerlIO_printf(Perl_debug_log,
7743 "Screamer: pre: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7744 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7745 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7747 /* This used to call 'filbuf' in stdio form, but as that behaves like
7748 getc when cnt <= 0 we use PerlIO_getc here to avoid introducing
7749 another abstraction. */
7750 i = PerlIO_getc(fp); /* get more characters */
7752 DEBUG_Pv(PerlIO_printf(Perl_debug_log,
7753 "Screamer: post: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7754 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7755 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7757 cnt = PerlIO_get_cnt(fp);
7758 ptr = (STDCHAR*)PerlIO_get_ptr(fp); /* reregisterize cnt and ptr */
7759 DEBUG_P(PerlIO_printf(Perl_debug_log,
7760 "Screamer: after getc, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7762 if (i == EOF) /* all done for ever? */
7763 goto thats_really_all_folks;
7765 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
7767 SvGROW(sv, bpx + cnt + 2);
7768 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
7770 *bp++ = (STDCHAR)i; /* store character from PerlIO_getc */
7772 if (rslen && (STDCHAR)i == rslast) /* all done for now? */
7773 goto thats_all_folks;
7777 if ((rslen > 1 && (STRLEN)(bp - (STDCHAR*)SvPVX_const(sv)) < rslen) ||
7778 memNE((char*)bp - rslen, rsptr, rslen))
7779 goto screamer; /* go back to the fray */
7780 thats_really_all_folks:
7782 cnt += shortbuffered;
7783 DEBUG_P(PerlIO_printf(Perl_debug_log,
7784 "Screamer: quitting, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7785 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* put these back or we're in trouble */
7786 DEBUG_P(PerlIO_printf(Perl_debug_log,
7787 "Screamer: end: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7788 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7789 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7791 SvCUR_set(sv, bp - (STDCHAR*)SvPVX_const(sv)); /* set length */
7792 DEBUG_P(PerlIO_printf(Perl_debug_log,
7793 "Screamer: done, len=%ld, string=|%.*s|\n",
7794 (long)SvCUR(sv),(int)SvCUR(sv),SvPVX_const(sv)));
7798 /*The big, slow, and stupid way. */
7799 #ifdef USE_HEAP_INSTEAD_OF_STACK /* Even slower way. */
7800 STDCHAR *buf = NULL;
7801 Newx(buf, 8192, STDCHAR);
7809 const STDCHAR * const bpe = buf + sizeof(buf);
7811 while ((i = PerlIO_getc(fp)) != EOF && (*bp++ = (STDCHAR)i) != rslast && bp < bpe)
7812 ; /* keep reading */
7816 cnt = PerlIO_read(fp,(char*)buf, sizeof(buf));
7817 /* Accommodate broken VAXC compiler, which applies U8 cast to
7818 * both args of ?: operator, causing EOF to change into 255
7821 i = (U8)buf[cnt - 1];
7827 cnt = 0; /* we do need to re-set the sv even when cnt <= 0 */
7829 sv_catpvn_nomg(sv, (char *) buf, cnt);
7831 sv_setpvn(sv, (char *) buf, cnt); /* "nomg" is implied */
7833 if (i != EOF && /* joy */
7835 SvCUR(sv) < rslen ||
7836 memNE(SvPVX_const(sv) + SvCUR(sv) - rslen, rsptr, rslen)))
7840 * If we're reading from a TTY and we get a short read,
7841 * indicating that the user hit his EOF character, we need
7842 * to notice it now, because if we try to read from the TTY
7843 * again, the EOF condition will disappear.
7845 * The comparison of cnt to sizeof(buf) is an optimization
7846 * that prevents unnecessary calls to feof().
7850 if (!(cnt < (I32)sizeof(buf) && PerlIO_eof(fp)))
7854 #ifdef USE_HEAP_INSTEAD_OF_STACK
7859 if (rspara) { /* have to do this both before and after */
7860 while (i != EOF) { /* to make sure file boundaries work right */
7861 i = PerlIO_getc(fp);
7863 PerlIO_ungetc(fp,i);
7869 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7875 Auto-increment of the value in the SV, doing string to numeric conversion
7876 if necessary. Handles 'get' magic and operator overloading.
7882 Perl_sv_inc(pTHX_ register SV *const sv)
7891 =for apidoc sv_inc_nomg
7893 Auto-increment of the value in the SV, doing string to numeric conversion
7894 if necessary. Handles operator overloading. Skips handling 'get' magic.
7900 Perl_sv_inc_nomg(pTHX_ register SV *const sv)
7908 if (SvTHINKFIRST(sv)) {
7909 if (SvIsCOW(sv) || isGV_with_GP(sv))
7910 sv_force_normal_flags(sv, 0);
7911 if (SvREADONLY(sv)) {
7912 if (IN_PERL_RUNTIME)
7913 Perl_croak_no_modify(aTHX);
7917 if (SvAMAGIC(sv) && AMG_CALLunary(sv, inc_amg))
7919 i = PTR2IV(SvRV(sv));
7924 flags = SvFLAGS(sv);
7925 if ((flags & (SVp_NOK|SVp_IOK)) == SVp_NOK) {
7926 /* It's (privately or publicly) a float, but not tested as an
7927 integer, so test it to see. */
7929 flags = SvFLAGS(sv);
7931 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
7932 /* It's publicly an integer, or privately an integer-not-float */
7933 #ifdef PERL_PRESERVE_IVUV
7937 if (SvUVX(sv) == UV_MAX)
7938 sv_setnv(sv, UV_MAX_P1);
7940 (void)SvIOK_only_UV(sv);
7941 SvUV_set(sv, SvUVX(sv) + 1);
7943 if (SvIVX(sv) == IV_MAX)
7944 sv_setuv(sv, (UV)IV_MAX + 1);
7946 (void)SvIOK_only(sv);
7947 SvIV_set(sv, SvIVX(sv) + 1);
7952 if (flags & SVp_NOK) {
7953 const NV was = SvNVX(sv);
7954 if (NV_OVERFLOWS_INTEGERS_AT &&
7955 was >= NV_OVERFLOWS_INTEGERS_AT) {
7956 /* diag_listed_as: Lost precision when %s %f by 1 */
7957 Perl_ck_warner(aTHX_ packWARN(WARN_IMPRECISION),
7958 "Lost precision when incrementing %" NVff " by 1",
7961 (void)SvNOK_only(sv);
7962 SvNV_set(sv, was + 1.0);
7966 if (!(flags & SVp_POK) || !*SvPVX_const(sv)) {
7967 if ((flags & SVTYPEMASK) < SVt_PVIV)
7968 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV ? SVt_PVIV : SVt_IV));
7969 (void)SvIOK_only(sv);
7974 while (isALPHA(*d)) d++;
7975 while (isDIGIT(*d)) d++;
7976 if (d < SvEND(sv)) {
7977 #ifdef PERL_PRESERVE_IVUV
7978 /* Got to punt this as an integer if needs be, but we don't issue
7979 warnings. Probably ought to make the sv_iv_please() that does
7980 the conversion if possible, and silently. */
7981 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
7982 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
7983 /* Need to try really hard to see if it's an integer.
7984 9.22337203685478e+18 is an integer.
7985 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
7986 so $a="9.22337203685478e+18"; $a+0; $a++
7987 needs to be the same as $a="9.22337203685478e+18"; $a++
7994 /* sv_2iv *should* have made this an NV */
7995 if (flags & SVp_NOK) {
7996 (void)SvNOK_only(sv);
7997 SvNV_set(sv, SvNVX(sv) + 1.0);
8000 /* I don't think we can get here. Maybe I should assert this
8001 And if we do get here I suspect that sv_setnv will croak. NWC
8003 #if defined(USE_LONG_DOUBLE)
8004 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",
8005 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
8007 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_inc punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
8008 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
8011 #endif /* PERL_PRESERVE_IVUV */
8012 sv_setnv(sv,Atof(SvPVX_const(sv)) + 1.0);
8016 while (d >= SvPVX_const(sv)) {
8024 /* MKS: The original code here died if letters weren't consecutive.
8025 * at least it didn't have to worry about non-C locales. The
8026 * new code assumes that ('z'-'a')==('Z'-'A'), letters are
8027 * arranged in order (although not consecutively) and that only
8028 * [A-Za-z] are accepted by isALPHA in the C locale.
8030 if (*d != 'z' && *d != 'Z') {
8031 do { ++*d; } while (!isALPHA(*d));
8034 *(d--) -= 'z' - 'a';
8039 *(d--) -= 'z' - 'a' + 1;
8043 /* oh,oh, the number grew */
8044 SvGROW(sv, SvCUR(sv) + 2);
8045 SvCUR_set(sv, SvCUR(sv) + 1);
8046 for (d = SvPVX(sv) + SvCUR(sv); d > SvPVX_const(sv); d--)
8057 Auto-decrement of the value in the SV, doing string to numeric conversion
8058 if necessary. Handles 'get' magic and operator overloading.
8064 Perl_sv_dec(pTHX_ register SV *const sv)
8074 =for apidoc sv_dec_nomg
8076 Auto-decrement of the value in the SV, doing string to numeric conversion
8077 if necessary. Handles operator overloading. Skips handling 'get' magic.
8083 Perl_sv_dec_nomg(pTHX_ register SV *const sv)
8090 if (SvTHINKFIRST(sv)) {
8091 if (SvIsCOW(sv) || isGV_with_GP(sv))
8092 sv_force_normal_flags(sv, 0);
8093 if (SvREADONLY(sv)) {
8094 if (IN_PERL_RUNTIME)
8095 Perl_croak_no_modify(aTHX);
8099 if (SvAMAGIC(sv) && AMG_CALLunary(sv, dec_amg))
8101 i = PTR2IV(SvRV(sv));
8106 /* Unlike sv_inc we don't have to worry about string-never-numbers
8107 and keeping them magic. But we mustn't warn on punting */
8108 flags = SvFLAGS(sv);
8109 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
8110 /* It's publicly an integer, or privately an integer-not-float */
8111 #ifdef PERL_PRESERVE_IVUV
8115 if (SvUVX(sv) == 0) {
8116 (void)SvIOK_only(sv);
8120 (void)SvIOK_only_UV(sv);
8121 SvUV_set(sv, SvUVX(sv) - 1);
8124 if (SvIVX(sv) == IV_MIN) {
8125 sv_setnv(sv, (NV)IV_MIN);
8129 (void)SvIOK_only(sv);
8130 SvIV_set(sv, SvIVX(sv) - 1);
8135 if (flags & SVp_NOK) {
8138 const NV was = SvNVX(sv);
8139 if (NV_OVERFLOWS_INTEGERS_AT &&
8140 was <= -NV_OVERFLOWS_INTEGERS_AT) {
8141 /* diag_listed_as: Lost precision when %s %f by 1 */
8142 Perl_ck_warner(aTHX_ packWARN(WARN_IMPRECISION),
8143 "Lost precision when decrementing %" NVff " by 1",
8146 (void)SvNOK_only(sv);
8147 SvNV_set(sv, was - 1.0);
8151 if (!(flags & SVp_POK)) {
8152 if ((flags & SVTYPEMASK) < SVt_PVIV)
8153 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV) ? SVt_PVIV : SVt_IV);
8155 (void)SvIOK_only(sv);
8158 #ifdef PERL_PRESERVE_IVUV
8160 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
8161 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
8162 /* Need to try really hard to see if it's an integer.
8163 9.22337203685478e+18 is an integer.
8164 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
8165 so $a="9.22337203685478e+18"; $a+0; $a--
8166 needs to be the same as $a="9.22337203685478e+18"; $a--
8173 /* sv_2iv *should* have made this an NV */
8174 if (flags & SVp_NOK) {
8175 (void)SvNOK_only(sv);
8176 SvNV_set(sv, SvNVX(sv) - 1.0);
8179 /* I don't think we can get here. Maybe I should assert this
8180 And if we do get here I suspect that sv_setnv will croak. NWC
8182 #if defined(USE_LONG_DOUBLE)
8183 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",
8184 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
8186 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_dec punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
8187 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
8191 #endif /* PERL_PRESERVE_IVUV */
8192 sv_setnv(sv,Atof(SvPVX_const(sv)) - 1.0); /* punt */
8195 /* this define is used to eliminate a chunk of duplicated but shared logic
8196 * it has the suffix __SV_C to signal that it isnt API, and isnt meant to be
8197 * used anywhere but here - yves
8199 #define PUSH_EXTEND_MORTAL__SV_C(AnSv) \
8202 PL_tmps_stack[++PL_tmps_ix] = (AnSv); \
8206 =for apidoc sv_mortalcopy
8208 Creates a new SV which is a copy of the original SV (using C<sv_setsv>).
8209 The new SV is marked as mortal. It will be destroyed "soon", either by an
8210 explicit call to FREETMPS, or by an implicit call at places such as
8211 statement boundaries. See also C<sv_newmortal> and C<sv_2mortal>.
8216 /* Make a string that will exist for the duration of the expression
8217 * evaluation. Actually, it may have to last longer than that, but
8218 * hopefully we won't free it until it has been assigned to a
8219 * permanent location. */
8222 Perl_sv_mortalcopy(pTHX_ SV *const oldstr)
8228 sv_setsv(sv,oldstr);
8229 PUSH_EXTEND_MORTAL__SV_C(sv);
8235 =for apidoc sv_newmortal
8237 Creates a new null SV which is mortal. The reference count of the SV is
8238 set to 1. It will be destroyed "soon", either by an explicit call to
8239 FREETMPS, or by an implicit call at places such as statement boundaries.
8240 See also C<sv_mortalcopy> and C<sv_2mortal>.
8246 Perl_sv_newmortal(pTHX)
8252 SvFLAGS(sv) = SVs_TEMP;
8253 PUSH_EXTEND_MORTAL__SV_C(sv);
8259 =for apidoc newSVpvn_flags
8261 Creates a new SV and copies a string into it. The reference count for the
8262 SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
8263 string. You are responsible for ensuring that the source string is at least
8264 C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
8265 Currently the only flag bits accepted are C<SVf_UTF8> and C<SVs_TEMP>.
8266 If C<SVs_TEMP> is set, then C<sv_2mortal()> is called on the result before
8267 returning. If C<SVf_UTF8> is set, C<s>
8268 is considered to be in UTF-8 and the
8269 C<SVf_UTF8> flag will be set on the new SV.
8270 C<newSVpvn_utf8()> is a convenience wrapper for this function, defined as
8272 #define newSVpvn_utf8(s, len, u) \
8273 newSVpvn_flags((s), (len), (u) ? SVf_UTF8 : 0)
8279 Perl_newSVpvn_flags(pTHX_ const char *const s, const STRLEN len, const U32 flags)
8284 /* All the flags we don't support must be zero.
8285 And we're new code so I'm going to assert this from the start. */
8286 assert(!(flags & ~(SVf_UTF8|SVs_TEMP)));
8288 sv_setpvn(sv,s,len);
8290 /* This code used to a sv_2mortal(), however we now unroll the call to sv_2mortal()
8291 * and do what it does ourselves here.
8292 * Since we have asserted that flags can only have the SVf_UTF8 and/or SVs_TEMP flags
8293 * set above we can use it to enable the sv flags directly (bypassing SvTEMP_on), which
8294 * in turn means we dont need to mask out the SVf_UTF8 flag below, which means that we
8295 * eliminate quite a few steps than it looks - Yves (explaining patch by gfx)
8298 SvFLAGS(sv) |= flags;
8300 if(flags & SVs_TEMP){
8301 PUSH_EXTEND_MORTAL__SV_C(sv);
8308 =for apidoc sv_2mortal
8310 Marks an existing SV as mortal. The SV will be destroyed "soon", either
8311 by an explicit call to FREETMPS, or by an implicit call at places such as
8312 statement boundaries. SvTEMP() is turned on which means that the SV's
8313 string buffer can be "stolen" if this SV is copied. See also C<sv_newmortal>
8314 and C<sv_mortalcopy>.
8320 Perl_sv_2mortal(pTHX_ register SV *const sv)
8325 if (SvREADONLY(sv) && SvIMMORTAL(sv))
8327 PUSH_EXTEND_MORTAL__SV_C(sv);
8335 Creates a new SV and copies a string into it. The reference count for the
8336 SV is set to 1. If C<len> is zero, Perl will compute the length using
8337 strlen(). For efficiency, consider using C<newSVpvn> instead.
8343 Perl_newSVpv(pTHX_ const char *const s, const STRLEN len)
8349 sv_setpvn(sv, s, len || s == NULL ? len : strlen(s));
8354 =for apidoc newSVpvn
8356 Creates a new SV and copies a buffer into it, which may contain NUL characters
8357 (C<\0>) and other binary data. The reference count for the SV is set to 1.
8358 Note that if C<len> is zero, Perl will create a zero length (Perl) string. You
8359 are responsible for ensuring that the source buffer is at least
8360 C<len> bytes long. If the C<buffer> argument is NULL the new SV will be
8367 Perl_newSVpvn(pTHX_ const char *const buffer, const STRLEN len)
8373 sv_setpvn(sv,buffer,len);
8378 =for apidoc newSVhek
8380 Creates a new SV from the hash key structure. It will generate scalars that
8381 point to the shared string table where possible. Returns a new (undefined)
8382 SV if the hek is NULL.
8388 Perl_newSVhek(pTHX_ const HEK *const hek)
8398 if (HEK_LEN(hek) == HEf_SVKEY) {
8399 return newSVsv(*(SV**)HEK_KEY(hek));
8401 const int flags = HEK_FLAGS(hek);
8402 if (flags & HVhek_WASUTF8) {
8404 Andreas would like keys he put in as utf8 to come back as utf8
8406 STRLEN utf8_len = HEK_LEN(hek);
8407 SV * const sv = newSV_type(SVt_PV);
8408 char *as_utf8 = (char *)bytes_to_utf8 ((U8*)HEK_KEY(hek), &utf8_len);
8409 /* bytes_to_utf8() allocates a new string, which we can repurpose: */
8410 sv_usepvn_flags(sv, as_utf8, utf8_len, SV_HAS_TRAILING_NUL);
8413 } else if (flags & (HVhek_REHASH|HVhek_UNSHARED)) {
8414 /* We don't have a pointer to the hv, so we have to replicate the
8415 flag into every HEK. This hv is using custom a hasing
8416 algorithm. Hence we can't return a shared string scalar, as
8417 that would contain the (wrong) hash value, and might get passed
8418 into an hv routine with a regular hash.
8419 Similarly, a hash that isn't using shared hash keys has to have
8420 the flag in every key so that we know not to try to call
8421 share_hek_hek on it. */
8423 SV * const sv = newSVpvn (HEK_KEY(hek), HEK_LEN(hek));
8428 /* This will be overwhelminly the most common case. */
8430 /* Inline most of newSVpvn_share(), because share_hek_hek() is far
8431 more efficient than sharepvn(). */
8435 sv_upgrade(sv, SVt_PV);
8436 SvPV_set(sv, (char *)HEK_KEY(share_hek_hek(hek)));
8437 SvCUR_set(sv, HEK_LEN(hek));
8450 =for apidoc newSVpvn_share
8452 Creates a new SV with its SvPVX_const pointing to a shared string in the string
8453 table. If the string does not already exist in the table, it is
8454 created first. Turns on READONLY and FAKE. If the C<hash> parameter
8455 is non-zero, that value is used; otherwise the hash is computed.
8456 The string's hash can later be retrieved from the SV
8457 with the C<SvSHARED_HASH()> macro. The idea here is
8458 that as the string table is used for shared hash keys these strings will have
8459 SvPVX_const == HeKEY and hash lookup will avoid string compare.
8465 Perl_newSVpvn_share(pTHX_ const char *src, I32 len, U32 hash)
8469 bool is_utf8 = FALSE;
8470 const char *const orig_src = src;
8473 STRLEN tmplen = -len;
8475 /* See the note in hv.c:hv_fetch() --jhi */
8476 src = (char*)bytes_from_utf8((const U8*)src, &tmplen, &is_utf8);
8480 PERL_HASH(hash, src, len);
8482 /* The logic for this is inlined in S_mro_get_linear_isa_dfs(), so if it
8483 changes here, update it there too. */
8484 sv_upgrade(sv, SVt_PV);
8485 SvPV_set(sv, sharepvn(src, is_utf8?-len:len, hash));
8493 if (src != orig_src)
8499 =for apidoc newSVpv_share
8501 Like C<newSVpvn_share>, but takes a nul-terminated string instead of a
8508 Perl_newSVpv_share(pTHX_ const char *src, U32 hash)
8510 return newSVpvn_share(src, strlen(src), hash);
8513 #if defined(PERL_IMPLICIT_CONTEXT)
8515 /* pTHX_ magic can't cope with varargs, so this is a no-context
8516 * version of the main function, (which may itself be aliased to us).
8517 * Don't access this version directly.
8521 Perl_newSVpvf_nocontext(const char *const pat, ...)
8527 PERL_ARGS_ASSERT_NEWSVPVF_NOCONTEXT;
8529 va_start(args, pat);
8530 sv = vnewSVpvf(pat, &args);
8537 =for apidoc newSVpvf
8539 Creates a new SV and initializes it with the string formatted like
8546 Perl_newSVpvf(pTHX_ const char *const pat, ...)
8551 PERL_ARGS_ASSERT_NEWSVPVF;
8553 va_start(args, pat);
8554 sv = vnewSVpvf(pat, &args);
8559 /* backend for newSVpvf() and newSVpvf_nocontext() */
8562 Perl_vnewSVpvf(pTHX_ const char *const pat, va_list *const args)
8567 PERL_ARGS_ASSERT_VNEWSVPVF;
8570 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
8577 Creates a new SV and copies a floating point value into it.
8578 The reference count for the SV is set to 1.
8584 Perl_newSVnv(pTHX_ const NV n)
8597 Creates a new SV and copies an integer into it. The reference count for the
8604 Perl_newSViv(pTHX_ const IV i)
8617 Creates a new SV and copies an unsigned integer into it.
8618 The reference count for the SV is set to 1.
8624 Perl_newSVuv(pTHX_ const UV u)
8635 =for apidoc newSV_type
8637 Creates a new SV, of the type specified. The reference count for the new SV
8644 Perl_newSV_type(pTHX_ const svtype type)
8649 sv_upgrade(sv, type);
8654 =for apidoc newRV_noinc
8656 Creates an RV wrapper for an SV. The reference count for the original
8657 SV is B<not> incremented.
8663 Perl_newRV_noinc(pTHX_ SV *const tmpRef)
8666 SV *sv = newSV_type(SVt_IV);
8668 PERL_ARGS_ASSERT_NEWRV_NOINC;
8671 SvRV_set(sv, tmpRef);
8676 /* newRV_inc is the official function name to use now.
8677 * newRV_inc is in fact #defined to newRV in sv.h
8681 Perl_newRV(pTHX_ SV *const sv)
8685 PERL_ARGS_ASSERT_NEWRV;
8687 return newRV_noinc(SvREFCNT_inc_simple_NN(sv));
8693 Creates a new SV which is an exact duplicate of the original SV.
8700 Perl_newSVsv(pTHX_ register SV *const old)
8707 if (SvTYPE(old) == (svtype)SVTYPEMASK) {
8708 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL), "semi-panic: attempt to dup freed string");
8712 /* SV_GMAGIC is the default for sv_setv()
8713 SV_NOSTEAL prevents TEMP buffers being, well, stolen, and saves games
8714 with SvTEMP_off and SvTEMP_on round a call to sv_setsv. */
8715 sv_setsv_flags(sv, old, SV_GMAGIC | SV_NOSTEAL);
8720 =for apidoc sv_reset
8722 Underlying implementation for the C<reset> Perl function.
8723 Note that the perl-level function is vaguely deprecated.
8729 Perl_sv_reset(pTHX_ register const char *s, HV *const stash)
8732 char todo[PERL_UCHAR_MAX+1];
8734 PERL_ARGS_ASSERT_SV_RESET;
8739 if (!*s) { /* reset ?? searches */
8740 MAGIC * const mg = mg_find((const SV *)stash, PERL_MAGIC_symtab);
8742 const U32 count = mg->mg_len / sizeof(PMOP**);
8743 PMOP **pmp = (PMOP**) mg->mg_ptr;
8744 PMOP *const *const end = pmp + count;
8748 SvREADONLY_off(PL_regex_pad[(*pmp)->op_pmoffset]);
8750 (*pmp)->op_pmflags &= ~PMf_USED;
8758 /* reset variables */
8760 if (!HvARRAY(stash))
8763 Zero(todo, 256, char);
8766 I32 i = (unsigned char)*s;
8770 max = (unsigned char)*s++;
8771 for ( ; i <= max; i++) {
8774 for (i = 0; i <= (I32) HvMAX(stash); i++) {
8776 for (entry = HvARRAY(stash)[i];
8778 entry = HeNEXT(entry))
8783 if (!todo[(U8)*HeKEY(entry)])
8785 gv = MUTABLE_GV(HeVAL(entry));
8788 if (SvTHINKFIRST(sv)) {
8789 if (!SvREADONLY(sv) && SvROK(sv))
8791 /* XXX Is this continue a bug? Why should THINKFIRST
8792 exempt us from resetting arrays and hashes? */
8796 if (SvTYPE(sv) >= SVt_PV) {
8798 if (SvPVX_const(sv) != NULL)
8806 if (GvHV(gv) && !HvNAME_get(GvHV(gv))) {
8808 Perl_die(aTHX_ "Can't reset %%ENV on this system");
8811 # if defined(USE_ENVIRON_ARRAY)
8814 # endif /* USE_ENVIRON_ARRAY */
8825 Using various gambits, try to get an IO from an SV: the IO slot if its a
8826 GV; or the recursive result if we're an RV; or the IO slot of the symbol
8827 named after the PV if we're a string.
8829 'Get' magic is ignored on the sv passed in, but will be called on
8830 C<SvRV(sv)> if sv is an RV.
8836 Perl_sv_2io(pTHX_ SV *const sv)
8841 PERL_ARGS_ASSERT_SV_2IO;
8843 switch (SvTYPE(sv)) {
8845 io = MUTABLE_IO(sv);
8849 if (isGV_with_GP(sv)) {
8850 gv = MUTABLE_GV(sv);
8853 Perl_croak(aTHX_ "Bad filehandle: %"HEKf,
8854 HEKfARG(GvNAME_HEK(gv)));
8860 Perl_croak(aTHX_ PL_no_usym, "filehandle");
8862 SvGETMAGIC(SvRV(sv));
8863 return sv_2io(SvRV(sv));
8865 gv = gv_fetchsv_nomg(sv, 0, SVt_PVIO);
8872 if (SvGMAGICAL(sv)) {
8873 newsv = sv_newmortal();
8874 sv_setsv_nomg(newsv, sv);
8876 Perl_croak(aTHX_ "Bad filehandle: %"SVf, SVfARG(newsv));
8886 Using various gambits, try to get a CV from an SV; in addition, try if
8887 possible to set C<*st> and C<*gvp> to the stash and GV associated with it.
8888 The flags in C<lref> are passed to gv_fetchsv.
8894 Perl_sv_2cv(pTHX_ SV *sv, HV **const st, GV **const gvp, const I32 lref)
8900 PERL_ARGS_ASSERT_SV_2CV;
8907 switch (SvTYPE(sv)) {
8911 return MUTABLE_CV(sv);
8921 sv = amagic_deref_call(sv, to_cv_amg);
8924 if (SvTYPE(sv) == SVt_PVCV) {
8925 cv = MUTABLE_CV(sv);
8930 else if(SvGETMAGIC(sv), isGV_with_GP(sv))
8931 gv = MUTABLE_GV(sv);
8933 Perl_croak(aTHX_ "Not a subroutine reference");
8935 else if (isGV_with_GP(sv)) {
8936 gv = MUTABLE_GV(sv);
8939 gv = gv_fetchsv_nomg(sv, lref, SVt_PVCV);
8946 /* Some flags to gv_fetchsv mean don't really create the GV */
8947 if (!isGV_with_GP(gv)) {
8952 if (lref & ~GV_ADDMG && !GvCVu(gv)) {
8953 /* XXX this is probably not what they think they're getting.
8954 * It has the same effect as "sub name;", i.e. just a forward
8965 Returns true if the SV has a true value by Perl's rules.
8966 Use the C<SvTRUE> macro instead, which may call C<sv_true()> or may
8967 instead use an in-line version.
8973 Perl_sv_true(pTHX_ register SV *const sv)
8978 const XPV* const tXpv = (XPV*)SvANY(sv);
8980 (tXpv->xpv_cur > 1 ||
8981 (tXpv->xpv_cur && *sv->sv_u.svu_pv != '0')))
8988 return SvIVX(sv) != 0;
8991 return SvNVX(sv) != 0.0;
8993 return sv_2bool(sv);
8999 =for apidoc sv_pvn_force
9001 Get a sensible string out of the SV somehow.
9002 A private implementation of the C<SvPV_force> macro for compilers which
9003 can't cope with complex macro expressions. Always use the macro instead.
9005 =for apidoc sv_pvn_force_flags
9007 Get a sensible string out of the SV somehow.
9008 If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<sv> if
9009 appropriate, else not. C<sv_pvn_force> and C<sv_pvn_force_nomg> are
9010 implemented in terms of this function.
9011 You normally want to use the various wrapper macros instead: see
9012 C<SvPV_force> and C<SvPV_force_nomg>
9018 Perl_sv_pvn_force_flags(pTHX_ SV *const sv, STRLEN *const lp, const I32 flags)
9022 PERL_ARGS_ASSERT_SV_PVN_FORCE_FLAGS;
9024 if (flags & SV_GMAGIC) SvGETMAGIC(sv);
9025 if (SvTHINKFIRST(sv) && !SvROK(sv))
9026 sv_force_normal_flags(sv, 0);
9036 if (SvREADONLY(sv) && !(flags & SV_MUTABLE_RETURN)) {
9037 const char * const ref = sv_reftype(sv,0);
9039 Perl_croak(aTHX_ "Can't coerce readonly %s to string in %s",
9040 ref, OP_DESC(PL_op));
9042 Perl_croak(aTHX_ "Can't coerce readonly %s to string", ref);
9044 if (SvTYPE(sv) > SVt_PVLV
9045 || isGV_with_GP(sv))
9046 /* diag_listed_as: Can't coerce %s to %s in %s */
9047 Perl_croak(aTHX_ "Can't coerce %s to string in %s", sv_reftype(sv,0),
9049 s = sv_2pv_flags(sv, &len, flags &~ SV_GMAGIC);
9056 if (s != SvPVX_const(sv)) { /* Almost, but not quite, sv_setpvn() */
9059 SvUPGRADE(sv, SVt_PV); /* Never FALSE */
9060 SvGROW(sv, len + 1);
9061 Move(s,SvPVX(sv),len,char);
9063 SvPVX(sv)[len] = '\0';
9066 SvPOK_on(sv); /* validate pointer */
9068 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
9069 PTR2UV(sv),SvPVX_const(sv)));
9072 (void)SvPOK_only_UTF8(sv);
9073 return SvPVX_mutable(sv);
9077 =for apidoc sv_pvbyten_force
9079 The backend for the C<SvPVbytex_force> macro. Always use the macro
9086 Perl_sv_pvbyten_force(pTHX_ SV *const sv, STRLEN *const lp)
9088 PERL_ARGS_ASSERT_SV_PVBYTEN_FORCE;
9090 sv_pvn_force(sv,lp);
9091 sv_utf8_downgrade(sv,0);
9097 =for apidoc sv_pvutf8n_force
9099 The backend for the C<SvPVutf8x_force> macro. Always use the macro
9106 Perl_sv_pvutf8n_force(pTHX_ SV *const sv, STRLEN *const lp)
9108 PERL_ARGS_ASSERT_SV_PVUTF8N_FORCE;
9110 sv_pvn_force(sv,lp);
9111 sv_utf8_upgrade(sv);
9117 =for apidoc sv_reftype
9119 Returns a string describing what the SV is a reference to.
9125 Perl_sv_reftype(pTHX_ const SV *const sv, const int ob)
9127 PERL_ARGS_ASSERT_SV_REFTYPE;
9128 if (ob && SvOBJECT(sv)) {
9129 return SvPV_nolen_const(sv_ref(NULL, sv, ob));
9132 switch (SvTYPE(sv)) {
9147 case SVt_PVLV: return (char *) (SvROK(sv) ? "REF"
9148 /* tied lvalues should appear to be
9149 * scalars for backwards compatibility */
9150 : (LvTYPE(sv) == 't' || LvTYPE(sv) == 'T')
9151 ? "SCALAR" : "LVALUE");
9152 case SVt_PVAV: return "ARRAY";
9153 case SVt_PVHV: return "HASH";
9154 case SVt_PVCV: return "CODE";
9155 case SVt_PVGV: return (char *) (isGV_with_GP(sv)
9156 ? "GLOB" : "SCALAR");
9157 case SVt_PVFM: return "FORMAT";
9158 case SVt_PVIO: return "IO";
9159 case SVt_BIND: return "BIND";
9160 case SVt_REGEXP: return "REGEXP";
9161 default: return "UNKNOWN";
9169 Returns a SV describing what the SV passed in is a reference to.
9175 Perl_sv_ref(pTHX_ register SV *dst, const SV *const sv, const int ob)
9177 PERL_ARGS_ASSERT_SV_REF;
9180 dst = sv_newmortal();
9182 if (ob && SvOBJECT(sv)) {
9183 HvNAME_get(SvSTASH(sv))
9184 ? sv_sethek(dst, HvNAME_HEK(SvSTASH(sv)))
9185 : sv_setpvn(dst, "__ANON__", 8);
9188 const char * reftype = sv_reftype(sv, 0);
9189 sv_setpv(dst, reftype);
9195 =for apidoc sv_isobject
9197 Returns a boolean indicating whether the SV is an RV pointing to a blessed
9198 object. If the SV is not an RV, or if the object is not blessed, then this
9205 Perl_sv_isobject(pTHX_ SV *sv)
9221 Returns a boolean indicating whether the SV is blessed into the specified
9222 class. This does not check for subtypes; use C<sv_derived_from> to verify
9223 an inheritance relationship.
9229 Perl_sv_isa(pTHX_ SV *sv, const char *const name)
9233 PERL_ARGS_ASSERT_SV_ISA;
9243 hvname = HvNAME_get(SvSTASH(sv));
9247 return strEQ(hvname, name);
9253 Creates a new SV for the RV, C<rv>, to point to. If C<rv> is not an RV then
9254 it will be upgraded to one. If C<classname> is non-null then the new SV will
9255 be blessed in the specified package. The new SV is returned and its
9256 reference count is 1.
9262 Perl_newSVrv(pTHX_ SV *const rv, const char *const classname)
9267 PERL_ARGS_ASSERT_NEWSVRV;
9271 SV_CHECK_THINKFIRST_COW_DROP(rv);
9273 if (SvTYPE(rv) >= SVt_PVMG) {
9274 const U32 refcnt = SvREFCNT(rv);
9278 SvREFCNT(rv) = refcnt;
9280 sv_upgrade(rv, SVt_IV);
9281 } else if (SvROK(rv)) {
9282 SvREFCNT_dec(SvRV(rv));
9284 prepare_SV_for_RV(rv);
9292 HV* const stash = gv_stashpv(classname, GV_ADD);
9293 (void)sv_bless(rv, stash);
9299 =for apidoc sv_setref_pv
9301 Copies a pointer into a new SV, optionally blessing the SV. The C<rv>
9302 argument will be upgraded to an RV. That RV will be modified to point to
9303 the new SV. If the C<pv> argument is NULL then C<PL_sv_undef> will be placed
9304 into the SV. The C<classname> argument indicates the package for the
9305 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9306 will have a reference count of 1, and the RV will be returned.
9308 Do not use with other Perl types such as HV, AV, SV, CV, because those
9309 objects will become corrupted by the pointer copy process.
9311 Note that C<sv_setref_pvn> copies the string while this copies the pointer.
9317 Perl_sv_setref_pv(pTHX_ SV *const rv, const char *const classname, void *const pv)
9321 PERL_ARGS_ASSERT_SV_SETREF_PV;
9324 sv_setsv(rv, &PL_sv_undef);
9328 sv_setiv(newSVrv(rv,classname), PTR2IV(pv));
9333 =for apidoc sv_setref_iv
9335 Copies an integer into a new SV, optionally blessing the SV. The C<rv>
9336 argument will be upgraded to an RV. That RV will be modified to point to
9337 the new SV. The C<classname> argument indicates the package for the
9338 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9339 will have a reference count of 1, and the RV will be returned.
9345 Perl_sv_setref_iv(pTHX_ SV *const rv, const char *const classname, const IV iv)
9347 PERL_ARGS_ASSERT_SV_SETREF_IV;
9349 sv_setiv(newSVrv(rv,classname), iv);
9354 =for apidoc sv_setref_uv
9356 Copies an unsigned integer into a new SV, optionally blessing the SV. The C<rv>
9357 argument will be upgraded to an RV. That RV will be modified to point to
9358 the new SV. The C<classname> argument indicates the package for the
9359 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9360 will have a reference count of 1, and the RV will be returned.
9366 Perl_sv_setref_uv(pTHX_ SV *const rv, const char *const classname, const UV uv)
9368 PERL_ARGS_ASSERT_SV_SETREF_UV;
9370 sv_setuv(newSVrv(rv,classname), uv);
9375 =for apidoc sv_setref_nv
9377 Copies a double into a new SV, optionally blessing the SV. The C<rv>
9378 argument will be upgraded to an RV. That RV will be modified to point to
9379 the new SV. The C<classname> argument indicates the package for the
9380 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9381 will have a reference count of 1, and the RV will be returned.
9387 Perl_sv_setref_nv(pTHX_ SV *const rv, const char *const classname, const NV nv)
9389 PERL_ARGS_ASSERT_SV_SETREF_NV;
9391 sv_setnv(newSVrv(rv,classname), nv);
9396 =for apidoc sv_setref_pvn
9398 Copies a string into a new SV, optionally blessing the SV. The length of the
9399 string must be specified with C<n>. The C<rv> argument will be upgraded to
9400 an RV. That RV will be modified to point to the new SV. The C<classname>
9401 argument indicates the package for the blessing. Set C<classname> to
9402 C<NULL> to avoid the blessing. The new SV will have a reference count
9403 of 1, and the RV will be returned.
9405 Note that C<sv_setref_pv> copies the pointer while this copies the string.
9411 Perl_sv_setref_pvn(pTHX_ SV *const rv, const char *const classname,
9412 const char *const pv, const STRLEN n)
9414 PERL_ARGS_ASSERT_SV_SETREF_PVN;
9416 sv_setpvn(newSVrv(rv,classname), pv, n);
9421 =for apidoc sv_bless
9423 Blesses an SV into a specified package. The SV must be an RV. The package
9424 must be designated by its stash (see C<gv_stashpv()>). The reference count
9425 of the SV is unaffected.
9431 Perl_sv_bless(pTHX_ SV *const sv, HV *const stash)
9436 PERL_ARGS_ASSERT_SV_BLESS;
9439 Perl_croak(aTHX_ "Can't bless non-reference value");
9441 if (SvFLAGS(tmpRef) & (SVs_OBJECT|SVf_READONLY)) {
9442 if (SvIsCOW(tmpRef))
9443 sv_force_normal_flags(tmpRef, 0);
9444 if (SvREADONLY(tmpRef))
9445 Perl_croak_no_modify(aTHX);
9446 if (SvOBJECT(tmpRef)) {
9447 if (SvTYPE(tmpRef) != SVt_PVIO)
9449 SvREFCNT_dec(SvSTASH(tmpRef));
9452 SvOBJECT_on(tmpRef);
9453 if (SvTYPE(tmpRef) != SVt_PVIO)
9455 SvUPGRADE(tmpRef, SVt_PVMG);
9456 SvSTASH_set(tmpRef, MUTABLE_HV(SvREFCNT_inc_simple(stash)));
9458 if(SvSMAGICAL(tmpRef))
9459 if(mg_find(tmpRef, PERL_MAGIC_ext) || mg_find(tmpRef, PERL_MAGIC_uvar))
9467 /* Downgrades a PVGV to a PVMG. If it's actually a PVLV, we leave the type
9468 * as it is after unglobbing it.
9471 PERL_STATIC_INLINE void
9472 S_sv_unglob(pTHX_ SV *const sv, U32 flags)
9477 SV * const temp = flags & SV_COW_DROP_PV ? NULL : sv_newmortal();
9479 PERL_ARGS_ASSERT_SV_UNGLOB;
9481 assert(SvTYPE(sv) == SVt_PVGV || SvTYPE(sv) == SVt_PVLV);
9483 if (!(flags & SV_COW_DROP_PV))
9484 gv_efullname3(temp, MUTABLE_GV(sv), "*");
9487 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
9488 && HvNAME_get(stash))
9489 mro_method_changed_in(stash);
9490 gp_free(MUTABLE_GV(sv));
9493 sv_del_backref(MUTABLE_SV(GvSTASH(sv)), sv);
9497 if (GvNAME_HEK(sv)) {
9498 unshare_hek(GvNAME_HEK(sv));
9500 isGV_with_GP_off(sv);
9502 if(SvTYPE(sv) == SVt_PVGV) {
9503 /* need to keep SvANY(sv) in the right arena */
9504 xpvmg = new_XPVMG();
9505 StructCopy(SvANY(sv), xpvmg, XPVMG);
9506 del_XPVGV(SvANY(sv));
9509 SvFLAGS(sv) &= ~SVTYPEMASK;
9510 SvFLAGS(sv) |= SVt_PVMG;
9513 /* Intentionally not calling any local SET magic, as this isn't so much a
9514 set operation as merely an internal storage change. */
9515 if (flags & SV_COW_DROP_PV) SvOK_off(sv);
9516 else sv_setsv_flags(sv, temp, 0);
9518 if ((const GV *)sv == PL_last_in_gv)
9519 PL_last_in_gv = NULL;
9520 else if ((const GV *)sv == PL_statgv)
9525 =for apidoc sv_unref_flags
9527 Unsets the RV status of the SV, and decrements the reference count of
9528 whatever was being referenced by the RV. This can almost be thought of
9529 as a reversal of C<newSVrv>. The C<cflags> argument can contain
9530 C<SV_IMMEDIATE_UNREF> to force the reference count to be decremented
9531 (otherwise the decrementing is conditional on the reference count being
9532 different from one or the reference being a readonly SV).
9539 Perl_sv_unref_flags(pTHX_ SV *const ref, const U32 flags)
9541 SV* const target = SvRV(ref);
9543 PERL_ARGS_ASSERT_SV_UNREF_FLAGS;
9545 if (SvWEAKREF(ref)) {
9546 sv_del_backref(target, ref);
9548 SvRV_set(ref, NULL);
9551 SvRV_set(ref, NULL);
9553 /* You can't have a || SvREADONLY(target) here, as $a = $$a, where $a was
9554 assigned to as BEGIN {$a = \"Foo"} will fail. */
9555 if (SvREFCNT(target) != 1 || (flags & SV_IMMEDIATE_UNREF))
9556 SvREFCNT_dec(target);
9557 else /* XXX Hack, but hard to make $a=$a->[1] work otherwise */
9558 sv_2mortal(target); /* Schedule for freeing later */
9562 =for apidoc sv_untaint
9564 Untaint an SV. Use C<SvTAINTED_off> instead.
9570 Perl_sv_untaint(pTHX_ SV *const sv)
9572 PERL_ARGS_ASSERT_SV_UNTAINT;
9574 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
9575 MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
9582 =for apidoc sv_tainted
9584 Test an SV for taintedness. Use C<SvTAINTED> instead.
9590 Perl_sv_tainted(pTHX_ SV *const sv)
9592 PERL_ARGS_ASSERT_SV_TAINTED;
9594 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
9595 const MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
9596 if (mg && (mg->mg_len & 1) )
9603 =for apidoc sv_setpviv
9605 Copies an integer into the given SV, also updating its string value.
9606 Does not handle 'set' magic. See C<sv_setpviv_mg>.
9612 Perl_sv_setpviv(pTHX_ SV *const sv, const IV iv)
9614 char buf[TYPE_CHARS(UV)];
9616 char * const ptr = uiv_2buf(buf, iv, 0, 0, &ebuf);
9618 PERL_ARGS_ASSERT_SV_SETPVIV;
9620 sv_setpvn(sv, ptr, ebuf - ptr);
9624 =for apidoc sv_setpviv_mg
9626 Like C<sv_setpviv>, but also handles 'set' magic.
9632 Perl_sv_setpviv_mg(pTHX_ SV *const sv, const IV iv)
9634 PERL_ARGS_ASSERT_SV_SETPVIV_MG;
9640 #if defined(PERL_IMPLICIT_CONTEXT)
9642 /* pTHX_ magic can't cope with varargs, so this is a no-context
9643 * version of the main function, (which may itself be aliased to us).
9644 * Don't access this version directly.
9648 Perl_sv_setpvf_nocontext(SV *const sv, const char *const pat, ...)
9653 PERL_ARGS_ASSERT_SV_SETPVF_NOCONTEXT;
9655 va_start(args, pat);
9656 sv_vsetpvf(sv, pat, &args);
9660 /* pTHX_ magic can't cope with varargs, so this is a no-context
9661 * version of the main function, (which may itself be aliased to us).
9662 * Don't access this version directly.
9666 Perl_sv_setpvf_mg_nocontext(SV *const sv, const char *const pat, ...)
9671 PERL_ARGS_ASSERT_SV_SETPVF_MG_NOCONTEXT;
9673 va_start(args, pat);
9674 sv_vsetpvf_mg(sv, pat, &args);
9680 =for apidoc sv_setpvf
9682 Works like C<sv_catpvf> but copies the text into the SV instead of
9683 appending it. Does not handle 'set' magic. See C<sv_setpvf_mg>.
9689 Perl_sv_setpvf(pTHX_ SV *const sv, const char *const pat, ...)
9693 PERL_ARGS_ASSERT_SV_SETPVF;
9695 va_start(args, pat);
9696 sv_vsetpvf(sv, pat, &args);
9701 =for apidoc sv_vsetpvf
9703 Works like C<sv_vcatpvf> but copies the text into the SV instead of
9704 appending it. Does not handle 'set' magic. See C<sv_vsetpvf_mg>.
9706 Usually used via its frontend C<sv_setpvf>.
9712 Perl_sv_vsetpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9714 PERL_ARGS_ASSERT_SV_VSETPVF;
9716 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9720 =for apidoc sv_setpvf_mg
9722 Like C<sv_setpvf>, but also handles 'set' magic.
9728 Perl_sv_setpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
9732 PERL_ARGS_ASSERT_SV_SETPVF_MG;
9734 va_start(args, pat);
9735 sv_vsetpvf_mg(sv, pat, &args);
9740 =for apidoc sv_vsetpvf_mg
9742 Like C<sv_vsetpvf>, but also handles 'set' magic.
9744 Usually used via its frontend C<sv_setpvf_mg>.
9750 Perl_sv_vsetpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9752 PERL_ARGS_ASSERT_SV_VSETPVF_MG;
9754 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9758 #if defined(PERL_IMPLICIT_CONTEXT)
9760 /* pTHX_ magic can't cope with varargs, so this is a no-context
9761 * version of the main function, (which may itself be aliased to us).
9762 * Don't access this version directly.
9766 Perl_sv_catpvf_nocontext(SV *const sv, const char *const pat, ...)
9771 PERL_ARGS_ASSERT_SV_CATPVF_NOCONTEXT;
9773 va_start(args, pat);
9774 sv_vcatpvf(sv, pat, &args);
9778 /* pTHX_ magic can't cope with varargs, so this is a no-context
9779 * version of the main function, (which may itself be aliased to us).
9780 * Don't access this version directly.
9784 Perl_sv_catpvf_mg_nocontext(SV *const sv, const char *const pat, ...)
9789 PERL_ARGS_ASSERT_SV_CATPVF_MG_NOCONTEXT;
9791 va_start(args, pat);
9792 sv_vcatpvf_mg(sv, pat, &args);
9798 =for apidoc sv_catpvf
9800 Processes its arguments like C<sprintf> and appends the formatted
9801 output to an SV. If the appended data contains "wide" characters
9802 (including, but not limited to, SVs with a UTF-8 PV formatted with %s,
9803 and characters >255 formatted with %c), the original SV might get
9804 upgraded to UTF-8. Handles 'get' magic, but not 'set' magic. See
9805 C<sv_catpvf_mg>. If the original SV was UTF-8, the pattern should be
9806 valid UTF-8; if the original SV was bytes, the pattern should be too.
9811 Perl_sv_catpvf(pTHX_ SV *const sv, const char *const pat, ...)
9815 PERL_ARGS_ASSERT_SV_CATPVF;
9817 va_start(args, pat);
9818 sv_vcatpvf(sv, pat, &args);
9823 =for apidoc sv_vcatpvf
9825 Processes its arguments like C<vsprintf> and appends the formatted output
9826 to an SV. Does not handle 'set' magic. See C<sv_vcatpvf_mg>.
9828 Usually used via its frontend C<sv_catpvf>.
9834 Perl_sv_vcatpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9836 PERL_ARGS_ASSERT_SV_VCATPVF;
9838 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9842 =for apidoc sv_catpvf_mg
9844 Like C<sv_catpvf>, but also handles 'set' magic.
9850 Perl_sv_catpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
9854 PERL_ARGS_ASSERT_SV_CATPVF_MG;
9856 va_start(args, pat);
9857 sv_vcatpvf_mg(sv, pat, &args);
9862 =for apidoc sv_vcatpvf_mg
9864 Like C<sv_vcatpvf>, but also handles 'set' magic.
9866 Usually used via its frontend C<sv_catpvf_mg>.
9872 Perl_sv_vcatpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9874 PERL_ARGS_ASSERT_SV_VCATPVF_MG;
9876 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9881 =for apidoc sv_vsetpvfn
9883 Works like C<sv_vcatpvfn> but copies the text into the SV instead of
9886 Usually used via one of its frontends C<sv_vsetpvf> and C<sv_vsetpvf_mg>.
9892 Perl_sv_vsetpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
9893 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
9895 PERL_ARGS_ASSERT_SV_VSETPVFN;
9898 sv_vcatpvfn_flags(sv, pat, patlen, args, svargs, svmax, maybe_tainted, 0);
9903 * Warn of missing argument to sprintf, and then return a defined value
9904 * to avoid inappropriate "use of uninit" warnings [perl #71000].
9906 #define WARN_MISSING WARN_UNINITIALIZED /* Not sure we want a new category */
9908 S_vcatpvfn_missing_argument(pTHX) {
9909 if (ckWARN(WARN_MISSING)) {
9910 Perl_warner(aTHX_ packWARN(WARN_MISSING), "Missing argument in %s",
9911 PL_op ? OP_DESC(PL_op) : "sv_vcatpvfn()");
9918 S_expect_number(pTHX_ char **const pattern)
9923 PERL_ARGS_ASSERT_EXPECT_NUMBER;
9925 switch (**pattern) {
9926 case '1': case '2': case '3':
9927 case '4': case '5': case '6':
9928 case '7': case '8': case '9':
9929 var = *(*pattern)++ - '0';
9930 while (isDIGIT(**pattern)) {
9931 const I32 tmp = var * 10 + (*(*pattern)++ - '0');
9933 Perl_croak(aTHX_ "Integer overflow in format string for %s", (PL_op ? OP_DESC(PL_op) : "sv_vcatpvfn"));
9941 S_F0convert(NV nv, char *const endbuf, STRLEN *const len)
9943 const int neg = nv < 0;
9946 PERL_ARGS_ASSERT_F0CONVERT;
9954 if (uv & 1 && uv == nv)
9955 uv--; /* Round to even */
9957 const unsigned dig = uv % 10;
9970 =for apidoc sv_vcatpvfn
9972 =for apidoc sv_vcatpvfn_flags
9974 Processes its arguments like C<vsprintf> and appends the formatted output
9975 to an SV. Uses an array of SVs if the C style variable argument list is
9976 missing (NULL). When running with taint checks enabled, indicates via
9977 C<maybe_tainted> if results are untrustworthy (often due to the use of
9980 If called as C<sv_vcatpvfn> or flags include C<SV_GMAGIC>, calls get magic.
9982 Usually used via one of its frontends C<sv_vcatpvf> and C<sv_vcatpvf_mg>.
9987 #define VECTORIZE_ARGS vecsv = va_arg(*args, SV*);\
9988 vecstr = (U8*)SvPV_const(vecsv,veclen);\
9989 vec_utf8 = DO_UTF8(vecsv);
9991 /* XXX maybe_tainted is never assigned to, so the doc above is lying. */
9994 Perl_sv_vcatpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
9995 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
9997 PERL_ARGS_ASSERT_SV_VCATPVFN;
9999 sv_vcatpvfn_flags(sv, pat, patlen, args, svargs, svmax, maybe_tainted, SV_GMAGIC|SV_SMAGIC);
10003 Perl_sv_vcatpvfn_flags(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
10004 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted,
10010 const char *patend;
10013 static const char nullstr[] = "(null)";
10015 bool has_utf8 = DO_UTF8(sv); /* has the result utf8? */
10016 const bool pat_utf8 = has_utf8; /* the pattern is in utf8? */
10018 /* Times 4: a decimal digit takes more than 3 binary digits.
10019 * NV_DIG: mantissa takes than many decimal digits.
10020 * Plus 32: Playing safe. */
10021 char ebuf[IV_DIG * 4 + NV_DIG + 32];
10022 /* large enough for "%#.#f" --chip */
10023 /* what about long double NVs? --jhi */
10025 PERL_ARGS_ASSERT_SV_VCATPVFN_FLAGS;
10026 PERL_UNUSED_ARG(maybe_tainted);
10028 if (flags & SV_GMAGIC)
10031 /* no matter what, this is a string now */
10032 (void)SvPV_force_nomg(sv, origlen);
10034 /* special-case "", "%s", and "%-p" (SVf - see below) */
10037 if (patlen == 2 && pat[0] == '%' && pat[1] == 's') {
10039 const char * const s = va_arg(*args, char*);
10040 sv_catpv_nomg(sv, s ? s : nullstr);
10042 else if (svix < svmax) {
10043 /* we want get magic on the source but not the target. sv_catsv can't do that, though */
10044 SvGETMAGIC(*svargs);
10045 sv_catsv_nomg(sv, *svargs);
10048 S_vcatpvfn_missing_argument(aTHX);
10051 if (args && patlen == 3 && pat[0] == '%' &&
10052 pat[1] == '-' && pat[2] == 'p') {
10053 argsv = MUTABLE_SV(va_arg(*args, void*));
10054 sv_catsv_nomg(sv, argsv);
10058 #ifndef USE_LONG_DOUBLE
10059 /* special-case "%.<number>[gf]" */
10060 if ( !args && patlen <= 5 && pat[0] == '%' && pat[1] == '.'
10061 && (pat[patlen-1] == 'g' || pat[patlen-1] == 'f') ) {
10062 unsigned digits = 0;
10066 while (*pp >= '0' && *pp <= '9')
10067 digits = 10 * digits + (*pp++ - '0');
10068 if (pp - pat == (int)patlen - 1 && svix < svmax) {
10069 const NV nv = SvNV(*svargs);
10071 /* Add check for digits != 0 because it seems that some
10072 gconverts are buggy in this case, and we don't yet have
10073 a Configure test for this. */
10074 if (digits && digits < sizeof(ebuf) - NV_DIG - 10) {
10075 /* 0, point, slack */
10076 Gconvert(nv, (int)digits, 0, ebuf);
10077 sv_catpv_nomg(sv, ebuf);
10078 if (*ebuf) /* May return an empty string for digits==0 */
10081 } else if (!digits) {
10084 if ((p = F0convert(nv, ebuf + sizeof ebuf, &l))) {
10085 sv_catpvn_nomg(sv, p, l);
10091 #endif /* !USE_LONG_DOUBLE */
10093 if (!args && svix < svmax && DO_UTF8(*svargs))
10096 patend = (char*)pat + patlen;
10097 for (p = (char*)pat; p < patend; p = q) {
10100 bool vectorize = FALSE;
10101 bool vectorarg = FALSE;
10102 bool vec_utf8 = FALSE;
10108 bool has_precis = FALSE;
10110 const I32 osvix = svix;
10111 bool is_utf8 = FALSE; /* is this item utf8? */
10112 #ifdef HAS_LDBL_SPRINTF_BUG
10113 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
10114 with sfio - Allen <allens@cpan.org> */
10115 bool fix_ldbl_sprintf_bug = FALSE;
10119 U8 utf8buf[UTF8_MAXBYTES+1];
10120 STRLEN esignlen = 0;
10122 const char *eptr = NULL;
10123 const char *fmtstart;
10126 const U8 *vecstr = NULL;
10133 /* we need a long double target in case HAS_LONG_DOUBLE but
10134 not USE_LONG_DOUBLE
10136 #if defined(HAS_LONG_DOUBLE) && LONG_DOUBLESIZE > DOUBLESIZE
10144 const char *dotstr = ".";
10145 STRLEN dotstrlen = 1;
10146 I32 efix = 0; /* explicit format parameter index */
10147 I32 ewix = 0; /* explicit width index */
10148 I32 epix = 0; /* explicit precision index */
10149 I32 evix = 0; /* explicit vector index */
10150 bool asterisk = FALSE;
10152 /* echo everything up to the next format specification */
10153 for (q = p; q < patend && *q != '%'; ++q) ;
10155 if (has_utf8 && !pat_utf8)
10156 sv_catpvn_nomg_utf8_upgrade(sv, p, q - p, nsv);
10158 sv_catpvn_nomg(sv, p, q - p);
10167 We allow format specification elements in this order:
10168 \d+\$ explicit format parameter index
10170 v|\*(\d+\$)?v vector with optional (optionally specified) arg
10171 0 flag (as above): repeated to allow "v02"
10172 \d+|\*(\d+\$)? width using optional (optionally specified) arg
10173 \.(\d*|\*(\d+\$)?) precision using optional (optionally specified) arg
10175 [%bcdefginopsuxDFOUX] format (mandatory)
10180 As of perl5.9.3, printf format checking is on by default.
10181 Internally, perl uses %p formats to provide an escape to
10182 some extended formatting. This block deals with those
10183 extensions: if it does not match, (char*)q is reset and
10184 the normal format processing code is used.
10186 Currently defined extensions are:
10187 %p include pointer address (standard)
10188 %-p (SVf) include an SV (previously %_)
10189 %-<num>p include an SV with precision <num>
10191 %3p include a HEK with precision of 256
10192 %<num>p (where num != 2 or 3) reserved for future
10195 Robin Barker 2005-07-14 (but modified since)
10197 %1p (VDf) removed. RMB 2007-10-19
10204 n = expect_number(&q);
10206 if (sv) { /* SVf */
10211 argsv = MUTABLE_SV(va_arg(*args, void*));
10212 eptr = SvPV_const(argsv, elen);
10213 if (DO_UTF8(argsv))
10217 else if (n==2 || n==3) { /* HEKf */
10218 HEK * const hek = va_arg(*args, HEK *);
10219 eptr = HEK_KEY(hek);
10220 elen = HEK_LEN(hek);
10221 if (HEK_UTF8(hek)) is_utf8 = TRUE;
10222 if (n==3) precis = 256, has_precis = TRUE;
10226 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL),
10227 "internal %%<num>p might conflict with future printf extensions");
10233 if ( (width = expect_number(&q)) ) {
10248 if (plus == '+' && *q == ' ') /* '+' over ' ' */
10277 if ( (ewix = expect_number(&q)) )
10286 if ((vectorarg = asterisk)) {
10299 width = expect_number(&q);
10302 if (vectorize && vectorarg) {
10303 /* vectorizing, but not with the default "." */
10305 vecsv = va_arg(*args, SV*);
10307 vecsv = (evix > 0 && evix <= svmax)
10308 ? svargs[evix-1] : S_vcatpvfn_missing_argument(aTHX);
10310 vecsv = svix < svmax
10311 ? svargs[svix++] : S_vcatpvfn_missing_argument(aTHX);
10313 dotstr = SvPV_const(vecsv, dotstrlen);
10314 /* Keep the DO_UTF8 test *after* the SvPV call, else things go
10315 bad with tied or overloaded values that return UTF8. */
10316 if (DO_UTF8(vecsv))
10318 else if (has_utf8) {
10319 vecsv = sv_mortalcopy(vecsv);
10320 sv_utf8_upgrade(vecsv);
10321 dotstr = SvPV_const(vecsv, dotstrlen);
10328 i = va_arg(*args, int);
10330 i = (ewix ? ewix <= svmax : svix < svmax) ?
10331 SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
10333 width = (i < 0) ? -i : i;
10343 if ( ((epix = expect_number(&q))) && (*q++ != '$') )
10345 /* XXX: todo, support specified precision parameter */
10349 i = va_arg(*args, int);
10351 i = (ewix ? ewix <= svmax : svix < svmax)
10352 ? SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
10354 has_precis = !(i < 0);
10358 while (isDIGIT(*q))
10359 precis = precis * 10 + (*q++ - '0');
10368 else if (efix ? (efix > 0 && efix <= svmax) : svix < svmax) {
10369 vecsv = svargs[efix ? efix-1 : svix++];
10370 vecstr = (U8*)SvPV_const(vecsv,veclen);
10371 vec_utf8 = DO_UTF8(vecsv);
10373 /* if this is a version object, we need to convert
10374 * back into v-string notation and then let the
10375 * vectorize happen normally
10377 if (sv_isobject(vecsv) && sv_derived_from(vecsv, "version")) {
10378 char *version = savesvpv(vecsv);
10379 if ( hv_exists(MUTABLE_HV(SvRV(vecsv)), "alpha", 5 ) ) {
10380 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
10381 "vector argument not supported with alpha versions");
10384 vecsv = sv_newmortal();
10385 scan_vstring(version, version + veclen, vecsv);
10386 vecstr = (U8*)SvPV_const(vecsv, veclen);
10387 vec_utf8 = DO_UTF8(vecsv);
10401 case 'I': /* Ix, I32x, and I64x */
10402 # ifdef USE_64_BIT_INT
10403 if (q[1] == '6' && q[2] == '4') {
10409 if (q[1] == '3' && q[2] == '2') {
10413 # ifdef USE_64_BIT_INT
10419 #if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
10431 #if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
10432 if (*q == 'l') { /* lld, llf */
10441 if (*++q == 'h') { /* hhd, hhu */
10470 if (!vectorize && !args) {
10472 const I32 i = efix-1;
10473 argsv = (i >= 0 && i < svmax)
10474 ? svargs[i] : S_vcatpvfn_missing_argument(aTHX);
10476 argsv = (svix >= 0 && svix < svmax)
10477 ? svargs[svix++] : S_vcatpvfn_missing_argument(aTHX);
10481 switch (c = *q++) {
10488 uv = (args) ? va_arg(*args, int) : SvIV(argsv);
10490 (!UNI_IS_INVARIANT(uv) && SvUTF8(sv)))
10492 eptr = (char*)utf8buf;
10493 elen = uvchr_to_utf8((U8*)eptr, uv) - utf8buf;
10507 eptr = va_arg(*args, char*);
10509 elen = strlen(eptr);
10511 eptr = (char *)nullstr;
10512 elen = sizeof nullstr - 1;
10516 eptr = SvPV_const(argsv, elen);
10517 if (DO_UTF8(argsv)) {
10518 STRLEN old_precis = precis;
10519 if (has_precis && precis < elen) {
10520 STRLEN ulen = sv_len_utf8_nomg(argsv);
10521 STRLEN p = precis > ulen ? ulen : precis;
10522 precis = sv_pos_u2b_flags(argsv, p, 0, 0);
10523 /* sticks at end */
10525 if (width) { /* fudge width (can't fudge elen) */
10526 if (has_precis && precis < elen)
10527 width += precis - old_precis;
10529 width += elen - sv_len_utf8_nomg(argsv);
10536 if (has_precis && precis < elen)
10543 if (alt || vectorize)
10545 uv = PTR2UV(args ? va_arg(*args, void*) : argsv);
10566 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
10575 esignbuf[esignlen++] = plus;
10579 case 'c': iv = (char)va_arg(*args, int); break;
10580 case 'h': iv = (short)va_arg(*args, int); break;
10581 case 'l': iv = va_arg(*args, long); break;
10582 case 'V': iv = va_arg(*args, IV); break;
10583 case 'z': iv = va_arg(*args, SSize_t); break;
10584 case 't': iv = va_arg(*args, ptrdiff_t); break;
10585 default: iv = va_arg(*args, int); break;
10587 case 'j': iv = va_arg(*args, intmax_t); break;
10591 iv = va_arg(*args, Quad_t); break;
10598 IV tiv = SvIV(argsv); /* work around GCC bug #13488 */
10600 case 'c': iv = (char)tiv; break;
10601 case 'h': iv = (short)tiv; break;
10602 case 'l': iv = (long)tiv; break;
10604 default: iv = tiv; break;
10607 iv = (Quad_t)tiv; break;
10613 if ( !vectorize ) /* we already set uv above */
10618 esignbuf[esignlen++] = plus;
10622 esignbuf[esignlen++] = '-';
10666 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
10677 case 'c': uv = (unsigned char)va_arg(*args, unsigned); break;
10678 case 'h': uv = (unsigned short)va_arg(*args, unsigned); break;
10679 case 'l': uv = va_arg(*args, unsigned long); break;
10680 case 'V': uv = va_arg(*args, UV); break;
10681 case 'z': uv = va_arg(*args, Size_t); break;
10682 case 't': uv = va_arg(*args, ptrdiff_t); break; /* will sign extend, but there is no uptrdiff_t, so oh well */
10684 case 'j': uv = va_arg(*args, uintmax_t); break;
10686 default: uv = va_arg(*args, unsigned); break;
10689 uv = va_arg(*args, Uquad_t); break;
10696 UV tuv = SvUV(argsv); /* work around GCC bug #13488 */
10698 case 'c': uv = (unsigned char)tuv; break;
10699 case 'h': uv = (unsigned short)tuv; break;
10700 case 'l': uv = (unsigned long)tuv; break;
10702 default: uv = tuv; break;
10705 uv = (Uquad_t)tuv; break;
10714 char *ptr = ebuf + sizeof ebuf;
10715 bool tempalt = uv ? alt : FALSE; /* Vectors can't change alt */
10721 p = (char *)((c == 'X') ? PL_hexdigit + 16 : PL_hexdigit);
10725 } while (uv >>= 4);
10727 esignbuf[esignlen++] = '0';
10728 esignbuf[esignlen++] = c; /* 'x' or 'X' */
10734 *--ptr = '0' + dig;
10735 } while (uv >>= 3);
10736 if (alt && *ptr != '0')
10742 *--ptr = '0' + dig;
10743 } while (uv >>= 1);
10745 esignbuf[esignlen++] = '0';
10746 esignbuf[esignlen++] = c;
10749 default: /* it had better be ten or less */
10752 *--ptr = '0' + dig;
10753 } while (uv /= base);
10756 elen = (ebuf + sizeof ebuf) - ptr;
10760 zeros = precis - elen;
10761 else if (precis == 0 && elen == 1 && *eptr == '0'
10762 && !(base == 8 && alt)) /* "%#.0o" prints "0" */
10765 /* a precision nullifies the 0 flag. */
10772 /* FLOATING POINT */
10775 c = 'f'; /* maybe %F isn't supported here */
10777 case 'e': case 'E':
10779 case 'g': case 'G':
10783 /* This is evil, but floating point is even more evil */
10785 /* for SV-style calling, we can only get NV
10786 for C-style calling, we assume %f is double;
10787 for simplicity we allow any of %Lf, %llf, %qf for long double
10791 #if defined(USE_LONG_DOUBLE)
10795 /* [perl #20339] - we should accept and ignore %lf rather than die */
10799 #if defined(USE_LONG_DOUBLE)
10800 intsize = args ? 0 : 'q';
10804 #if defined(HAS_LONG_DOUBLE)
10817 /* now we need (long double) if intsize == 'q', else (double) */
10819 #if LONG_DOUBLESIZE > DOUBLESIZE
10821 va_arg(*args, long double) :
10822 va_arg(*args, double)
10824 va_arg(*args, double)
10829 /* nv * 0 will be NaN for NaN, +Inf and -Inf, and 0 for anything
10830 else. frexp() has some unspecified behaviour for those three */
10831 if (c != 'e' && c != 'E' && (nv * 0) == 0) {
10833 /* FIXME: if HAS_LONG_DOUBLE but not USE_LONG_DOUBLE this
10834 will cast our (long double) to (double) */
10835 (void)Perl_frexp(nv, &i);
10836 if (i == PERL_INT_MIN)
10837 Perl_die(aTHX_ "panic: frexp");
10839 need = BIT_DIGITS(i);
10841 need += has_precis ? precis : 6; /* known default */
10846 #ifdef HAS_LDBL_SPRINTF_BUG
10847 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
10848 with sfio - Allen <allens@cpan.org> */
10851 # define MY_DBL_MAX DBL_MAX
10852 # else /* XXX guessing! HUGE_VAL may be defined as infinity, so not using */
10853 # if DOUBLESIZE >= 8
10854 # define MY_DBL_MAX 1.7976931348623157E+308L
10856 # define MY_DBL_MAX 3.40282347E+38L
10860 # ifdef HAS_LDBL_SPRINTF_BUG_LESS1 /* only between -1L & 1L - Allen */
10861 # define MY_DBL_MAX_BUG 1L
10863 # define MY_DBL_MAX_BUG MY_DBL_MAX
10867 # define MY_DBL_MIN DBL_MIN
10868 # else /* XXX guessing! -Allen */
10869 # if DOUBLESIZE >= 8
10870 # define MY_DBL_MIN 2.2250738585072014E-308L
10872 # define MY_DBL_MIN 1.17549435E-38L
10876 if ((intsize == 'q') && (c == 'f') &&
10877 ((nv < MY_DBL_MAX_BUG) && (nv > -MY_DBL_MAX_BUG)) &&
10878 (need < DBL_DIG)) {
10879 /* it's going to be short enough that
10880 * long double precision is not needed */
10882 if ((nv <= 0L) && (nv >= -0L))
10883 fix_ldbl_sprintf_bug = TRUE; /* 0 is 0 - easiest */
10885 /* would use Perl_fp_class as a double-check but not
10886 * functional on IRIX - see perl.h comments */
10888 if ((nv >= MY_DBL_MIN) || (nv <= -MY_DBL_MIN)) {
10889 /* It's within the range that a double can represent */
10890 #if defined(DBL_MAX) && !defined(DBL_MIN)
10891 if ((nv >= ((long double)1/DBL_MAX)) ||
10892 (nv <= (-(long double)1/DBL_MAX)))
10894 fix_ldbl_sprintf_bug = TRUE;
10897 if (fix_ldbl_sprintf_bug == TRUE) {
10907 # undef MY_DBL_MAX_BUG
10910 #endif /* HAS_LDBL_SPRINTF_BUG */
10912 need += 20; /* fudge factor */
10913 if (PL_efloatsize < need) {
10914 Safefree(PL_efloatbuf);
10915 PL_efloatsize = need + 20; /* more fudge */
10916 Newx(PL_efloatbuf, PL_efloatsize, char);
10917 PL_efloatbuf[0] = '\0';
10920 if ( !(width || left || plus || alt) && fill != '0'
10921 && has_precis && intsize != 'q' ) { /* Shortcuts */
10922 /* See earlier comment about buggy Gconvert when digits,
10924 if ( c == 'g' && precis) {
10925 Gconvert((NV)nv, (int)precis, 0, PL_efloatbuf);
10926 /* May return an empty string for digits==0 */
10927 if (*PL_efloatbuf) {
10928 elen = strlen(PL_efloatbuf);
10929 goto float_converted;
10931 } else if ( c == 'f' && !precis) {
10932 if ((eptr = F0convert(nv, ebuf + sizeof ebuf, &elen)))
10937 char *ptr = ebuf + sizeof ebuf;
10940 /* FIXME: what to do if HAS_LONG_DOUBLE but not PERL_PRIfldbl? */
10941 #if defined(HAS_LONG_DOUBLE) && defined(PERL_PRIfldbl)
10942 if (intsize == 'q') {
10943 /* Copy the one or more characters in a long double
10944 * format before the 'base' ([efgEFG]) character to
10945 * the format string. */
10946 static char const prifldbl[] = PERL_PRIfldbl;
10947 char const *p = prifldbl + sizeof(prifldbl) - 3;
10948 while (p >= prifldbl) { *--ptr = *p--; }
10953 do { *--ptr = '0' + (base % 10); } while (base /= 10);
10958 do { *--ptr = '0' + (base % 10); } while (base /= 10);
10970 /* No taint. Otherwise we are in the strange situation
10971 * where printf() taints but print($float) doesn't.
10973 #if defined(HAS_LONG_DOUBLE)
10974 elen = ((intsize == 'q')
10975 ? my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, nv)
10976 : my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, (double)nv));
10978 elen = my_sprintf(PL_efloatbuf, ptr, nv);
10982 eptr = PL_efloatbuf;
10990 i = SvCUR(sv) - origlen;
10993 case 'c': *(va_arg(*args, char*)) = i; break;
10994 case 'h': *(va_arg(*args, short*)) = i; break;
10995 default: *(va_arg(*args, int*)) = i; break;
10996 case 'l': *(va_arg(*args, long*)) = i; break;
10997 case 'V': *(va_arg(*args, IV*)) = i; break;
10998 case 'z': *(va_arg(*args, SSize_t*)) = i; break;
10999 case 't': *(va_arg(*args, ptrdiff_t*)) = i; break;
11001 case 'j': *(va_arg(*args, intmax_t*)) = i; break;
11005 *(va_arg(*args, Quad_t*)) = i; break;
11012 sv_setuv_mg(argsv, has_utf8 ? (UV)sv_len_utf8(sv) : (UV)i);
11013 continue; /* not "break" */
11020 && (PL_op->op_type == OP_PRTF || PL_op->op_type == OP_SPRINTF)
11021 && ckWARN(WARN_PRINTF))
11023 SV * const msg = sv_newmortal();
11024 Perl_sv_setpvf(aTHX_ msg, "Invalid conversion in %sprintf: ",
11025 (PL_op->op_type == OP_PRTF) ? "" : "s");
11026 if (fmtstart < patend) {
11027 const char * const fmtend = q < patend ? q : patend;
11029 sv_catpvs(msg, "\"%");
11030 for (f = fmtstart; f < fmtend; f++) {
11032 sv_catpvn_nomg(msg, f, 1);
11034 Perl_sv_catpvf(aTHX_ msg,
11035 "\\%03"UVof, (UV)*f & 0xFF);
11038 sv_catpvs(msg, "\"");
11040 sv_catpvs(msg, "end of string");
11042 Perl_warner(aTHX_ packWARN(WARN_PRINTF), "%"SVf, SVfARG(msg)); /* yes, this is reentrant */
11045 /* output mangled stuff ... */
11051 /* ... right here, because formatting flags should not apply */
11052 SvGROW(sv, SvCUR(sv) + elen + 1);
11054 Copy(eptr, p, elen, char);
11057 SvCUR_set(sv, p - SvPVX_const(sv));
11059 continue; /* not "break" */
11062 if (is_utf8 != has_utf8) {
11065 sv_utf8_upgrade(sv);
11068 const STRLEN old_elen = elen;
11069 SV * const nsv = newSVpvn_flags(eptr, elen, SVs_TEMP);
11070 sv_utf8_upgrade(nsv);
11071 eptr = SvPVX_const(nsv);
11074 if (width) { /* fudge width (can't fudge elen) */
11075 width += elen - old_elen;
11081 have = esignlen + zeros + elen;
11083 Perl_croak_nocontext("%s", PL_memory_wrap);
11085 need = (have > width ? have : width);
11088 if (need >= (((STRLEN)~0) - SvCUR(sv) - dotstrlen - 1))
11089 Perl_croak_nocontext("%s", PL_memory_wrap);
11090 SvGROW(sv, SvCUR(sv) + need + dotstrlen + 1);
11092 if (esignlen && fill == '0') {
11094 for (i = 0; i < (int)esignlen; i++)
11095 *p++ = esignbuf[i];
11097 if (gap && !left) {
11098 memset(p, fill, gap);
11101 if (esignlen && fill != '0') {
11103 for (i = 0; i < (int)esignlen; i++)
11104 *p++ = esignbuf[i];
11108 for (i = zeros; i; i--)
11112 Copy(eptr, p, elen, char);
11116 memset(p, ' ', gap);
11121 Copy(dotstr, p, dotstrlen, char);
11125 vectorize = FALSE; /* done iterating over vecstr */
11132 SvCUR_set(sv, p - SvPVX_const(sv));
11141 /* =========================================================================
11143 =head1 Cloning an interpreter
11145 All the macros and functions in this section are for the private use of
11146 the main function, perl_clone().
11148 The foo_dup() functions make an exact copy of an existing foo thingy.
11149 During the course of a cloning, a hash table is used to map old addresses
11150 to new addresses. The table is created and manipulated with the
11151 ptr_table_* functions.
11155 * =========================================================================*/
11158 #if defined(USE_ITHREADS)
11160 /* XXX Remove this so it doesn't have to go thru the macro and return for nothing */
11161 #ifndef GpREFCNT_inc
11162 # define GpREFCNT_inc(gp) ((gp) ? (++(gp)->gp_refcnt, (gp)) : (GP*)NULL)
11166 /* Certain cases in Perl_ss_dup have been merged, by relying on the fact
11167 that currently av_dup, gv_dup and hv_dup are the same as sv_dup.
11168 If this changes, please unmerge ss_dup.
11169 Likewise, sv_dup_inc_multiple() relies on this fact. */
11170 #define sv_dup_inc_NN(s,t) SvREFCNT_inc_NN(sv_dup_inc(s,t))
11171 #define av_dup(s,t) MUTABLE_AV(sv_dup((const SV *)s,t))
11172 #define av_dup_inc(s,t) MUTABLE_AV(sv_dup_inc((const SV *)s,t))
11173 #define hv_dup(s,t) MUTABLE_HV(sv_dup((const SV *)s,t))
11174 #define hv_dup_inc(s,t) MUTABLE_HV(sv_dup_inc((const SV *)s,t))
11175 #define cv_dup(s,t) MUTABLE_CV(sv_dup((const SV *)s,t))
11176 #define cv_dup_inc(s,t) MUTABLE_CV(sv_dup_inc((const SV *)s,t))
11177 #define io_dup(s,t) MUTABLE_IO(sv_dup((const SV *)s,t))
11178 #define io_dup_inc(s,t) MUTABLE_IO(sv_dup_inc((const SV *)s,t))
11179 #define gv_dup(s,t) MUTABLE_GV(sv_dup((const SV *)s,t))
11180 #define gv_dup_inc(s,t) MUTABLE_GV(sv_dup_inc((const SV *)s,t))
11181 #define SAVEPV(p) ((p) ? savepv(p) : NULL)
11182 #define SAVEPVN(p,n) ((p) ? savepvn(p,n) : NULL)
11184 /* clone a parser */
11187 Perl_parser_dup(pTHX_ const yy_parser *const proto, CLONE_PARAMS *const param)
11191 PERL_ARGS_ASSERT_PARSER_DUP;
11196 /* look for it in the table first */
11197 parser = (yy_parser *)ptr_table_fetch(PL_ptr_table, proto);
11201 /* create anew and remember what it is */
11202 Newxz(parser, 1, yy_parser);
11203 ptr_table_store(PL_ptr_table, proto, parser);
11205 /* XXX these not yet duped */
11206 parser->old_parser = NULL;
11207 parser->stack = NULL;
11209 parser->stack_size = 0;
11210 /* XXX parser->stack->state = 0; */
11212 /* XXX eventually, just Copy() most of the parser struct ? */
11214 parser->lex_brackets = proto->lex_brackets;
11215 parser->lex_casemods = proto->lex_casemods;
11216 parser->lex_brackstack = savepvn(proto->lex_brackstack,
11217 (proto->lex_brackets < 120 ? 120 : proto->lex_brackets));
11218 parser->lex_casestack = savepvn(proto->lex_casestack,
11219 (proto->lex_casemods < 12 ? 12 : proto->lex_casemods));
11220 parser->lex_defer = proto->lex_defer;
11221 parser->lex_dojoin = proto->lex_dojoin;
11222 parser->lex_expect = proto->lex_expect;
11223 parser->lex_formbrack = proto->lex_formbrack;
11224 parser->lex_inpat = proto->lex_inpat;
11225 parser->lex_inwhat = proto->lex_inwhat;
11226 parser->lex_op = proto->lex_op;
11227 parser->lex_repl = sv_dup_inc(proto->lex_repl, param);
11228 parser->lex_starts = proto->lex_starts;
11229 parser->lex_stuff = sv_dup_inc(proto->lex_stuff, param);
11230 parser->multi_close = proto->multi_close;
11231 parser->multi_open = proto->multi_open;
11232 parser->multi_start = proto->multi_start;
11233 parser->multi_end = proto->multi_end;
11234 parser->pending_ident = proto->pending_ident;
11235 parser->preambled = proto->preambled;
11236 parser->sublex_info = proto->sublex_info; /* XXX not quite right */
11237 parser->linestr = sv_dup_inc(proto->linestr, param);
11238 parser->expect = proto->expect;
11239 parser->copline = proto->copline;
11240 parser->last_lop_op = proto->last_lop_op;
11241 parser->lex_state = proto->lex_state;
11242 parser->rsfp = fp_dup(proto->rsfp, '<', param);
11243 /* rsfp_filters entries have fake IoDIRP() */
11244 parser->rsfp_filters= av_dup_inc(proto->rsfp_filters, param);
11245 parser->in_my = proto->in_my;
11246 parser->in_my_stash = hv_dup(proto->in_my_stash, param);
11247 parser->error_count = proto->error_count;
11250 parser->linestr = sv_dup_inc(proto->linestr, param);
11253 char * const ols = SvPVX(proto->linestr);
11254 char * const ls = SvPVX(parser->linestr);
11256 parser->bufptr = ls + (proto->bufptr >= ols ?
11257 proto->bufptr - ols : 0);
11258 parser->oldbufptr = ls + (proto->oldbufptr >= ols ?
11259 proto->oldbufptr - ols : 0);
11260 parser->oldoldbufptr= ls + (proto->oldoldbufptr >= ols ?
11261 proto->oldoldbufptr - ols : 0);
11262 parser->linestart = ls + (proto->linestart >= ols ?
11263 proto->linestart - ols : 0);
11264 parser->last_uni = ls + (proto->last_uni >= ols ?
11265 proto->last_uni - ols : 0);
11266 parser->last_lop = ls + (proto->last_lop >= ols ?
11267 proto->last_lop - ols : 0);
11269 parser->bufend = ls + SvCUR(parser->linestr);
11272 Copy(proto->tokenbuf, parser->tokenbuf, 256, char);
11276 parser->endwhite = proto->endwhite;
11277 parser->faketokens = proto->faketokens;
11278 parser->lasttoke = proto->lasttoke;
11279 parser->nextwhite = proto->nextwhite;
11280 parser->realtokenstart = proto->realtokenstart;
11281 parser->skipwhite = proto->skipwhite;
11282 parser->thisclose = proto->thisclose;
11283 parser->thismad = proto->thismad;
11284 parser->thisopen = proto->thisopen;
11285 parser->thisstuff = proto->thisstuff;
11286 parser->thistoken = proto->thistoken;
11287 parser->thiswhite = proto->thiswhite;
11289 Copy(proto->nexttoke, parser->nexttoke, 5, NEXTTOKE);
11290 parser->curforce = proto->curforce;
11292 Copy(proto->nextval, parser->nextval, 5, YYSTYPE);
11293 Copy(proto->nexttype, parser->nexttype, 5, I32);
11294 parser->nexttoke = proto->nexttoke;
11297 /* XXX should clone saved_curcop here, but we aren't passed
11298 * proto_perl; so do it in perl_clone_using instead */
11304 /* duplicate a file handle */
11307 Perl_fp_dup(pTHX_ PerlIO *const fp, const char type, CLONE_PARAMS *const param)
11311 PERL_ARGS_ASSERT_FP_DUP;
11312 PERL_UNUSED_ARG(type);
11315 return (PerlIO*)NULL;
11317 /* look for it in the table first */
11318 ret = (PerlIO*)ptr_table_fetch(PL_ptr_table, fp);
11322 /* create anew and remember what it is */
11323 ret = PerlIO_fdupopen(aTHX_ fp, param, PERLIO_DUP_CLONE);
11324 ptr_table_store(PL_ptr_table, fp, ret);
11328 /* duplicate a directory handle */
11331 Perl_dirp_dup(pTHX_ DIR *const dp, CLONE_PARAMS *const param)
11337 const Direntry_t *dirent;
11338 char smallbuf[256];
11344 PERL_UNUSED_CONTEXT;
11345 PERL_ARGS_ASSERT_DIRP_DUP;
11350 /* look for it in the table first */
11351 ret = (DIR*)ptr_table_fetch(PL_ptr_table, dp);
11357 PERL_UNUSED_ARG(param);
11361 /* open the current directory (so we can switch back) */
11362 if (!(pwd = PerlDir_open("."))) return (DIR *)NULL;
11364 /* chdir to our dir handle and open the present working directory */
11365 if (fchdir(my_dirfd(dp)) < 0 || !(ret = PerlDir_open("."))) {
11366 PerlDir_close(pwd);
11367 return (DIR *)NULL;
11369 /* Now we should have two dir handles pointing to the same dir. */
11371 /* Be nice to the calling code and chdir back to where we were. */
11372 fchdir(my_dirfd(pwd)); /* If this fails, then what? */
11374 /* We have no need of the pwd handle any more. */
11375 PerlDir_close(pwd);
11378 # define d_namlen(d) (d)->d_namlen
11380 # define d_namlen(d) strlen((d)->d_name)
11382 /* Iterate once through dp, to get the file name at the current posi-
11383 tion. Then step back. */
11384 pos = PerlDir_tell(dp);
11385 if ((dirent = PerlDir_read(dp))) {
11386 len = d_namlen(dirent);
11387 if (len <= sizeof smallbuf) name = smallbuf;
11388 else Newx(name, len, char);
11389 Move(dirent->d_name, name, len, char);
11391 PerlDir_seek(dp, pos);
11393 /* Iterate through the new dir handle, till we find a file with the
11395 if (!dirent) /* just before the end */
11397 pos = PerlDir_tell(ret);
11398 if (PerlDir_read(ret)) continue; /* not there yet */
11399 PerlDir_seek(ret, pos); /* step back */
11403 const long pos0 = PerlDir_tell(ret);
11405 pos = PerlDir_tell(ret);
11406 if ((dirent = PerlDir_read(ret))) {
11407 if (len == d_namlen(dirent)
11408 && memEQ(name, dirent->d_name, len)) {
11410 PerlDir_seek(ret, pos); /* step back */
11413 /* else we are not there yet; keep iterating */
11415 else { /* This is not meant to happen. The best we can do is
11416 reset the iterator to the beginning. */
11417 PerlDir_seek(ret, pos0);
11424 if (name && name != smallbuf)
11429 ret = win32_dirp_dup(dp, param);
11432 /* pop it in the pointer table */
11434 ptr_table_store(PL_ptr_table, dp, ret);
11439 /* duplicate a typeglob */
11442 Perl_gp_dup(pTHX_ GP *const gp, CLONE_PARAMS *const param)
11446 PERL_ARGS_ASSERT_GP_DUP;
11450 /* look for it in the table first */
11451 ret = (GP*)ptr_table_fetch(PL_ptr_table, gp);
11455 /* create anew and remember what it is */
11457 ptr_table_store(PL_ptr_table, gp, ret);
11460 /* ret->gp_refcnt must be 0 before any other dups are called. We're relying
11461 on Newxz() to do this for us. */
11462 ret->gp_sv = sv_dup_inc(gp->gp_sv, param);
11463 ret->gp_io = io_dup_inc(gp->gp_io, param);
11464 ret->gp_form = cv_dup_inc(gp->gp_form, param);
11465 ret->gp_av = av_dup_inc(gp->gp_av, param);
11466 ret->gp_hv = hv_dup_inc(gp->gp_hv, param);
11467 ret->gp_egv = gv_dup(gp->gp_egv, param);/* GvEGV is not refcounted */
11468 ret->gp_cv = cv_dup_inc(gp->gp_cv, param);
11469 ret->gp_cvgen = gp->gp_cvgen;
11470 ret->gp_line = gp->gp_line;
11471 ret->gp_file_hek = hek_dup(gp->gp_file_hek, param);
11475 /* duplicate a chain of magic */
11478 Perl_mg_dup(pTHX_ MAGIC *mg, CLONE_PARAMS *const param)
11480 MAGIC *mgret = NULL;
11481 MAGIC **mgprev_p = &mgret;
11483 PERL_ARGS_ASSERT_MG_DUP;
11485 for (; mg; mg = mg->mg_moremagic) {
11488 if ((param->flags & CLONEf_JOIN_IN)
11489 && mg->mg_type == PERL_MAGIC_backref)
11490 /* when joining, we let the individual SVs add themselves to
11491 * backref as needed. */
11494 Newx(nmg, 1, MAGIC);
11496 mgprev_p = &(nmg->mg_moremagic);
11498 /* There was a comment "XXX copy dynamic vtable?" but as we don't have
11499 dynamic vtables, I'm not sure why Sarathy wrote it. The comment dates
11500 from the original commit adding Perl_mg_dup() - revision 4538.
11501 Similarly there is the annotation "XXX random ptr?" next to the
11502 assignment to nmg->mg_ptr. */
11505 /* FIXME for plugins
11506 if (nmg->mg_type == PERL_MAGIC_qr) {
11507 nmg->mg_obj = MUTABLE_SV(CALLREGDUPE((REGEXP*)nmg->mg_obj, param));
11511 nmg->mg_obj = (nmg->mg_flags & MGf_REFCOUNTED)
11512 ? nmg->mg_type == PERL_MAGIC_backref
11513 /* The backref AV has its reference
11514 * count deliberately bumped by 1 */
11515 ? SvREFCNT_inc(av_dup_inc((const AV *)
11516 nmg->mg_obj, param))
11517 : sv_dup_inc(nmg->mg_obj, param)
11518 : sv_dup(nmg->mg_obj, param);
11520 if (nmg->mg_ptr && nmg->mg_type != PERL_MAGIC_regex_global) {
11521 if (nmg->mg_len > 0) {
11522 nmg->mg_ptr = SAVEPVN(nmg->mg_ptr, nmg->mg_len);
11523 if (nmg->mg_type == PERL_MAGIC_overload_table &&
11524 AMT_AMAGIC((AMT*)nmg->mg_ptr))
11526 AMT * const namtp = (AMT*)nmg->mg_ptr;
11527 sv_dup_inc_multiple((SV**)(namtp->table),
11528 (SV**)(namtp->table), NofAMmeth, param);
11531 else if (nmg->mg_len == HEf_SVKEY)
11532 nmg->mg_ptr = (char*)sv_dup_inc((const SV *)nmg->mg_ptr, param);
11534 if ((nmg->mg_flags & MGf_DUP) && nmg->mg_virtual && nmg->mg_virtual->svt_dup) {
11535 nmg->mg_virtual->svt_dup(aTHX_ nmg, param);
11541 #endif /* USE_ITHREADS */
11543 struct ptr_tbl_arena {
11544 struct ptr_tbl_arena *next;
11545 struct ptr_tbl_ent array[1023/3]; /* as ptr_tbl_ent has 3 pointers. */
11548 /* create a new pointer-mapping table */
11551 Perl_ptr_table_new(pTHX)
11554 PERL_UNUSED_CONTEXT;
11556 Newx(tbl, 1, PTR_TBL_t);
11557 tbl->tbl_max = 511;
11558 tbl->tbl_items = 0;
11559 tbl->tbl_arena = NULL;
11560 tbl->tbl_arena_next = NULL;
11561 tbl->tbl_arena_end = NULL;
11562 Newxz(tbl->tbl_ary, tbl->tbl_max + 1, PTR_TBL_ENT_t*);
11566 #define PTR_TABLE_HASH(ptr) \
11567 ((PTR2UV(ptr) >> 3) ^ (PTR2UV(ptr) >> (3 + 7)) ^ (PTR2UV(ptr) >> (3 + 17)))
11569 /* map an existing pointer using a table */
11571 STATIC PTR_TBL_ENT_t *
11572 S_ptr_table_find(PTR_TBL_t *const tbl, const void *const sv)
11574 PTR_TBL_ENT_t *tblent;
11575 const UV hash = PTR_TABLE_HASH(sv);
11577 PERL_ARGS_ASSERT_PTR_TABLE_FIND;
11579 tblent = tbl->tbl_ary[hash & tbl->tbl_max];
11580 for (; tblent; tblent = tblent->next) {
11581 if (tblent->oldval == sv)
11588 Perl_ptr_table_fetch(pTHX_ PTR_TBL_t *const tbl, const void *const sv)
11590 PTR_TBL_ENT_t const *const tblent = ptr_table_find(tbl, sv);
11592 PERL_ARGS_ASSERT_PTR_TABLE_FETCH;
11593 PERL_UNUSED_CONTEXT;
11595 return tblent ? tblent->newval : NULL;
11598 /* add a new entry to a pointer-mapping table */
11601 Perl_ptr_table_store(pTHX_ PTR_TBL_t *const tbl, const void *const oldsv, void *const newsv)
11603 PTR_TBL_ENT_t *tblent = ptr_table_find(tbl, oldsv);
11605 PERL_ARGS_ASSERT_PTR_TABLE_STORE;
11606 PERL_UNUSED_CONTEXT;
11609 tblent->newval = newsv;
11611 const UV entry = PTR_TABLE_HASH(oldsv) & tbl->tbl_max;
11613 if (tbl->tbl_arena_next == tbl->tbl_arena_end) {
11614 struct ptr_tbl_arena *new_arena;
11616 Newx(new_arena, 1, struct ptr_tbl_arena);
11617 new_arena->next = tbl->tbl_arena;
11618 tbl->tbl_arena = new_arena;
11619 tbl->tbl_arena_next = new_arena->array;
11620 tbl->tbl_arena_end = new_arena->array
11621 + sizeof(new_arena->array) / sizeof(new_arena->array[0]);
11624 tblent = tbl->tbl_arena_next++;
11626 tblent->oldval = oldsv;
11627 tblent->newval = newsv;
11628 tblent->next = tbl->tbl_ary[entry];
11629 tbl->tbl_ary[entry] = tblent;
11631 if (tblent->next && tbl->tbl_items > tbl->tbl_max)
11632 ptr_table_split(tbl);
11636 /* double the hash bucket size of an existing ptr table */
11639 Perl_ptr_table_split(pTHX_ PTR_TBL_t *const tbl)
11641 PTR_TBL_ENT_t **ary = tbl->tbl_ary;
11642 const UV oldsize = tbl->tbl_max + 1;
11643 UV newsize = oldsize * 2;
11646 PERL_ARGS_ASSERT_PTR_TABLE_SPLIT;
11647 PERL_UNUSED_CONTEXT;
11649 Renew(ary, newsize, PTR_TBL_ENT_t*);
11650 Zero(&ary[oldsize], newsize-oldsize, PTR_TBL_ENT_t*);
11651 tbl->tbl_max = --newsize;
11652 tbl->tbl_ary = ary;
11653 for (i=0; i < oldsize; i++, ary++) {
11654 PTR_TBL_ENT_t **entp = ary;
11655 PTR_TBL_ENT_t *ent = *ary;
11656 PTR_TBL_ENT_t **curentp;
11659 curentp = ary + oldsize;
11661 if ((newsize & PTR_TABLE_HASH(ent->oldval)) != i) {
11663 ent->next = *curentp;
11673 /* remove all the entries from a ptr table */
11674 /* Deprecated - will be removed post 5.14 */
11677 Perl_ptr_table_clear(pTHX_ PTR_TBL_t *const tbl)
11679 if (tbl && tbl->tbl_items) {
11680 struct ptr_tbl_arena *arena = tbl->tbl_arena;
11682 Zero(tbl->tbl_ary, tbl->tbl_max + 1, struct ptr_tbl_ent **);
11685 struct ptr_tbl_arena *next = arena->next;
11691 tbl->tbl_items = 0;
11692 tbl->tbl_arena = NULL;
11693 tbl->tbl_arena_next = NULL;
11694 tbl->tbl_arena_end = NULL;
11698 /* clear and free a ptr table */
11701 Perl_ptr_table_free(pTHX_ PTR_TBL_t *const tbl)
11703 struct ptr_tbl_arena *arena;
11709 arena = tbl->tbl_arena;
11712 struct ptr_tbl_arena *next = arena->next;
11718 Safefree(tbl->tbl_ary);
11722 #if defined(USE_ITHREADS)
11725 Perl_rvpv_dup(pTHX_ SV *const dstr, const SV *const sstr, CLONE_PARAMS *const param)
11727 PERL_ARGS_ASSERT_RVPV_DUP;
11730 if (SvWEAKREF(sstr)) {
11731 SvRV_set(dstr, sv_dup(SvRV_const(sstr), param));
11732 if (param->flags & CLONEf_JOIN_IN) {
11733 /* if joining, we add any back references individually rather
11734 * than copying the whole backref array */
11735 Perl_sv_add_backref(aTHX_ SvRV(dstr), dstr);
11739 SvRV_set(dstr, sv_dup_inc(SvRV_const(sstr), param));
11741 else if (SvPVX_const(sstr)) {
11742 /* Has something there */
11744 /* Normal PV - clone whole allocated space */
11745 SvPV_set(dstr, SAVEPVN(SvPVX_const(sstr), SvLEN(sstr)-1));
11746 if (SvREADONLY(sstr) && SvFAKE(sstr)) {
11747 /* Not that normal - actually sstr is copy on write.
11748 But we are a true, independent SV, so: */
11749 SvREADONLY_off(dstr);
11754 /* Special case - not normally malloced for some reason */
11755 if (isGV_with_GP(sstr)) {
11756 /* Don't need to do anything here. */
11758 else if ((SvREADONLY(sstr) && SvFAKE(sstr))) {
11759 /* A "shared" PV - clone it as "shared" PV */
11761 HEK_KEY(hek_dup(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)),
11765 /* Some other special case - random pointer */
11766 SvPV_set(dstr, (char *) SvPVX_const(sstr));
11771 /* Copy the NULL */
11772 SvPV_set(dstr, NULL);
11776 /* duplicate a list of SVs. source and dest may point to the same memory. */
11778 S_sv_dup_inc_multiple(pTHX_ SV *const *source, SV **dest,
11779 SSize_t items, CLONE_PARAMS *const param)
11781 PERL_ARGS_ASSERT_SV_DUP_INC_MULTIPLE;
11783 while (items-- > 0) {
11784 *dest++ = sv_dup_inc(*source++, param);
11790 /* duplicate an SV of any type (including AV, HV etc) */
11793 S_sv_dup_common(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
11798 PERL_ARGS_ASSERT_SV_DUP_COMMON;
11800 if (SvTYPE(sstr) == (svtype)SVTYPEMASK) {
11801 #ifdef DEBUG_LEAKING_SCALARS_ABORT
11806 /* look for it in the table first */
11807 dstr = MUTABLE_SV(ptr_table_fetch(PL_ptr_table, sstr));
11811 if(param->flags & CLONEf_JOIN_IN) {
11812 /** We are joining here so we don't want do clone
11813 something that is bad **/
11814 if (SvTYPE(sstr) == SVt_PVHV) {
11815 const HEK * const hvname = HvNAME_HEK(sstr);
11817 /** don't clone stashes if they already exist **/
11818 dstr = MUTABLE_SV(gv_stashpvn(HEK_KEY(hvname), HEK_LEN(hvname),
11819 HEK_UTF8(hvname) ? SVf_UTF8 : 0));
11820 ptr_table_store(PL_ptr_table, sstr, dstr);
11824 else if (SvTYPE(sstr) == SVt_PVGV && !SvFAKE(sstr)) {
11825 HV *stash = GvSTASH(sstr);
11826 const HEK * hvname;
11827 if (stash && (hvname = HvNAME_HEK(stash))) {
11828 /** don't clone GVs if they already exist **/
11830 stash = gv_stashpvn(HEK_KEY(hvname), HEK_LEN(hvname),
11831 HEK_UTF8(hvname) ? SVf_UTF8 : 0);
11833 stash, GvNAME(sstr),
11839 if (svp && *svp && SvTYPE(*svp) == SVt_PVGV) {
11840 ptr_table_store(PL_ptr_table, sstr, *svp);
11847 /* create anew and remember what it is */
11850 #ifdef DEBUG_LEAKING_SCALARS
11851 dstr->sv_debug_optype = sstr->sv_debug_optype;
11852 dstr->sv_debug_line = sstr->sv_debug_line;
11853 dstr->sv_debug_inpad = sstr->sv_debug_inpad;
11854 dstr->sv_debug_parent = (SV*)sstr;
11855 FREE_SV_DEBUG_FILE(dstr);
11856 dstr->sv_debug_file = savepv(sstr->sv_debug_file);
11859 ptr_table_store(PL_ptr_table, sstr, dstr);
11862 SvFLAGS(dstr) = SvFLAGS(sstr);
11863 SvFLAGS(dstr) &= ~SVf_OOK; /* don't propagate OOK hack */
11864 SvREFCNT(dstr) = 0; /* must be before any other dups! */
11867 if (SvANY(sstr) && PL_watch_pvx && SvPVX_const(sstr) == PL_watch_pvx)
11868 PerlIO_printf(Perl_debug_log, "watch at %p hit, found string \"%s\"\n",
11869 (void*)PL_watch_pvx, SvPVX_const(sstr));
11872 /* don't clone objects whose class has asked us not to */
11873 if (SvOBJECT(sstr) && ! (SvFLAGS(SvSTASH(sstr)) & SVphv_CLONEABLE)) {
11878 switch (SvTYPE(sstr)) {
11880 SvANY(dstr) = NULL;
11883 SvANY(dstr) = (XPVIV*)((char*)&(dstr->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
11885 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
11887 SvIV_set(dstr, SvIVX(sstr));
11891 SvANY(dstr) = new_XNV();
11892 SvNV_set(dstr, SvNVX(sstr));
11894 /* case SVt_BIND: */
11897 /* These are all the types that need complex bodies allocating. */
11899 const svtype sv_type = SvTYPE(sstr);
11900 const struct body_details *const sv_type_details
11901 = bodies_by_type + sv_type;
11905 Perl_croak(aTHX_ "Bizarre SvTYPE [%" IVdf "]", (IV)SvTYPE(sstr));
11920 assert(sv_type_details->body_size);
11921 if (sv_type_details->arena) {
11922 new_body_inline(new_body, sv_type);
11924 = (void*)((char*)new_body - sv_type_details->offset);
11926 new_body = new_NOARENA(sv_type_details);
11930 SvANY(dstr) = new_body;
11933 Copy(((char*)SvANY(sstr)) + sv_type_details->offset,
11934 ((char*)SvANY(dstr)) + sv_type_details->offset,
11935 sv_type_details->copy, char);
11937 Copy(((char*)SvANY(sstr)),
11938 ((char*)SvANY(dstr)),
11939 sv_type_details->body_size + sv_type_details->offset, char);
11942 if (sv_type != SVt_PVAV && sv_type != SVt_PVHV
11943 && !isGV_with_GP(dstr)
11944 && !(sv_type == SVt_PVIO && !(IoFLAGS(dstr) & IOf_FAKE_DIRP)))
11945 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
11947 /* The Copy above means that all the source (unduplicated) pointers
11948 are now in the destination. We can check the flags and the
11949 pointers in either, but it's possible that there's less cache
11950 missing by always going for the destination.
11951 FIXME - instrument and check that assumption */
11952 if (sv_type >= SVt_PVMG) {
11953 if ((sv_type == SVt_PVMG) && SvPAD_OUR(dstr)) {
11954 SvOURSTASH_set(dstr, hv_dup_inc(SvOURSTASH(dstr), param));
11955 } else if (SvMAGIC(dstr))
11956 SvMAGIC_set(dstr, mg_dup(SvMAGIC(dstr), param));
11958 SvSTASH_set(dstr, hv_dup_inc(SvSTASH(dstr), param));
11961 /* The cast silences a GCC warning about unhandled types. */
11962 switch ((int)sv_type) {
11972 /* FIXME for plugins */
11973 re_dup_guts((REGEXP*) sstr, (REGEXP*) dstr, param);
11976 /* XXX LvTARGOFF sometimes holds PMOP* when DEBUGGING */
11977 if (LvTYPE(dstr) == 't') /* for tie: unrefcnted fake (SV**) */
11978 LvTARG(dstr) = dstr;
11979 else if (LvTYPE(dstr) == 'T') /* for tie: fake HE */
11980 LvTARG(dstr) = MUTABLE_SV(he_dup((HE*)LvTARG(dstr), 0, param));
11982 LvTARG(dstr) = sv_dup_inc(LvTARG(dstr), param);
11984 /* non-GP case already handled above */
11985 if(isGV_with_GP(sstr)) {
11986 GvNAME_HEK(dstr) = hek_dup(GvNAME_HEK(dstr), param);
11987 /* Don't call sv_add_backref here as it's going to be
11988 created as part of the magic cloning of the symbol
11989 table--unless this is during a join and the stash
11990 is not actually being cloned. */
11991 /* Danger Will Robinson - GvGP(dstr) isn't initialised
11992 at the point of this comment. */
11993 GvSTASH(dstr) = hv_dup(GvSTASH(dstr), param);
11994 if (param->flags & CLONEf_JOIN_IN)
11995 Perl_sv_add_backref(aTHX_ MUTABLE_SV(GvSTASH(dstr)), dstr);
11996 GvGP_set(dstr, gp_dup(GvGP(sstr), param));
11997 (void)GpREFCNT_inc(GvGP(dstr));
12001 /* PL_parser->rsfp_filters entries have fake IoDIRP() */
12002 if(IoFLAGS(dstr) & IOf_FAKE_DIRP) {
12003 /* I have no idea why fake dirp (rsfps)
12004 should be treated differently but otherwise
12005 we end up with leaks -- sky*/
12006 IoTOP_GV(dstr) = gv_dup_inc(IoTOP_GV(dstr), param);
12007 IoFMT_GV(dstr) = gv_dup_inc(IoFMT_GV(dstr), param);
12008 IoBOTTOM_GV(dstr) = gv_dup_inc(IoBOTTOM_GV(dstr), param);
12010 IoTOP_GV(dstr) = gv_dup(IoTOP_GV(dstr), param);
12011 IoFMT_GV(dstr) = gv_dup(IoFMT_GV(dstr), param);
12012 IoBOTTOM_GV(dstr) = gv_dup(IoBOTTOM_GV(dstr), param);
12013 if (IoDIRP(dstr)) {
12014 IoDIRP(dstr) = dirp_dup(IoDIRP(dstr), param);
12017 /* IoDIRP(dstr) is already a copy of IoDIRP(sstr) */
12019 IoIFP(dstr) = fp_dup(IoIFP(sstr), IoTYPE(dstr), param);
12021 if (IoOFP(dstr) == IoIFP(sstr))
12022 IoOFP(dstr) = IoIFP(dstr);
12024 IoOFP(dstr) = fp_dup(IoOFP(dstr), IoTYPE(dstr), param);
12025 IoTOP_NAME(dstr) = SAVEPV(IoTOP_NAME(dstr));
12026 IoFMT_NAME(dstr) = SAVEPV(IoFMT_NAME(dstr));
12027 IoBOTTOM_NAME(dstr) = SAVEPV(IoBOTTOM_NAME(dstr));
12030 /* avoid cloning an empty array */
12031 if (AvARRAY((const AV *)sstr) && AvFILLp((const AV *)sstr) >= 0) {
12032 SV **dst_ary, **src_ary;
12033 SSize_t items = AvFILLp((const AV *)sstr) + 1;
12035 src_ary = AvARRAY((const AV *)sstr);
12036 Newxz(dst_ary, AvMAX((const AV *)sstr)+1, SV*);
12037 ptr_table_store(PL_ptr_table, src_ary, dst_ary);
12038 AvARRAY(MUTABLE_AV(dstr)) = dst_ary;
12039 AvALLOC((const AV *)dstr) = dst_ary;
12040 if (AvREAL((const AV *)sstr)) {
12041 dst_ary = sv_dup_inc_multiple(src_ary, dst_ary, items,
12045 while (items-- > 0)
12046 *dst_ary++ = sv_dup(*src_ary++, param);
12048 items = AvMAX((const AV *)sstr) - AvFILLp((const AV *)sstr);
12049 while (items-- > 0) {
12050 *dst_ary++ = &PL_sv_undef;
12054 AvARRAY(MUTABLE_AV(dstr)) = NULL;
12055 AvALLOC((const AV *)dstr) = (SV**)NULL;
12056 AvMAX( (const AV *)dstr) = -1;
12057 AvFILLp((const AV *)dstr) = -1;
12061 if (HvARRAY((const HV *)sstr)) {
12063 const bool sharekeys = !!HvSHAREKEYS(sstr);
12064 XPVHV * const dxhv = (XPVHV*)SvANY(dstr);
12065 XPVHV * const sxhv = (XPVHV*)SvANY(sstr);
12067 Newx(darray, PERL_HV_ARRAY_ALLOC_BYTES(dxhv->xhv_max+1)
12068 + (SvOOK(sstr) ? sizeof(struct xpvhv_aux) : 0),
12070 HvARRAY(dstr) = (HE**)darray;
12071 while (i <= sxhv->xhv_max) {
12072 const HE * const source = HvARRAY(sstr)[i];
12073 HvARRAY(dstr)[i] = source
12074 ? he_dup(source, sharekeys, param) : 0;
12078 const struct xpvhv_aux * const saux = HvAUX(sstr);
12079 struct xpvhv_aux * const daux = HvAUX(dstr);
12080 /* This flag isn't copied. */
12083 if (saux->xhv_name_count) {
12084 HEK ** const sname = saux->xhv_name_u.xhvnameu_names;
12086 = saux->xhv_name_count < 0
12087 ? -saux->xhv_name_count
12088 : saux->xhv_name_count;
12089 HEK **shekp = sname + count;
12091 Newx(daux->xhv_name_u.xhvnameu_names, count, HEK *);
12092 dhekp = daux->xhv_name_u.xhvnameu_names + count;
12093 while (shekp-- > sname) {
12095 *dhekp = hek_dup(*shekp, param);
12099 daux->xhv_name_u.xhvnameu_name
12100 = hek_dup(saux->xhv_name_u.xhvnameu_name,
12103 daux->xhv_name_count = saux->xhv_name_count;
12105 daux->xhv_riter = saux->xhv_riter;
12106 daux->xhv_eiter = saux->xhv_eiter
12107 ? he_dup(saux->xhv_eiter,
12108 cBOOL(HvSHAREKEYS(sstr)), param) : 0;
12109 /* backref array needs refcnt=2; see sv_add_backref */
12110 daux->xhv_backreferences =
12111 (param->flags & CLONEf_JOIN_IN)
12112 /* when joining, we let the individual GVs and
12113 * CVs add themselves to backref as
12114 * needed. This avoids pulling in stuff
12115 * that isn't required, and simplifies the
12116 * case where stashes aren't cloned back
12117 * if they already exist in the parent
12120 : saux->xhv_backreferences
12121 ? (SvTYPE(saux->xhv_backreferences) == SVt_PVAV)
12122 ? MUTABLE_AV(SvREFCNT_inc(
12123 sv_dup_inc((const SV *)
12124 saux->xhv_backreferences, param)))
12125 : MUTABLE_AV(sv_dup((const SV *)
12126 saux->xhv_backreferences, param))
12129 daux->xhv_mro_meta = saux->xhv_mro_meta
12130 ? mro_meta_dup(saux->xhv_mro_meta, param)
12133 /* Record stashes for possible cloning in Perl_clone(). */
12135 av_push(param->stashes, dstr);
12139 HvARRAY(MUTABLE_HV(dstr)) = NULL;
12142 if (!(param->flags & CLONEf_COPY_STACKS)) {
12147 /* NOTE: not refcounted */
12148 SvANY(MUTABLE_CV(dstr))->xcv_stash =
12149 hv_dup(CvSTASH(dstr), param);
12150 if ((param->flags & CLONEf_JOIN_IN) && CvSTASH(dstr))
12151 Perl_sv_add_backref(aTHX_ MUTABLE_SV(CvSTASH(dstr)), dstr);
12152 if (!CvISXSUB(dstr)) {
12154 CvROOT(dstr) = OpREFCNT_inc(CvROOT(dstr));
12156 CvSLABBED_off(dstr);
12157 } else if (CvCONST(dstr)) {
12158 CvXSUBANY(dstr).any_ptr =
12159 sv_dup_inc((const SV *)CvXSUBANY(dstr).any_ptr, param);
12161 assert(!CvSLABBED(dstr));
12162 if (CvDYNFILE(dstr)) CvFILE(dstr) = SAVEPV(CvFILE(dstr));
12163 /* don't dup if copying back - CvGV isn't refcounted, so the
12164 * duped GV may never be freed. A bit of a hack! DAPM */
12165 SvANY(MUTABLE_CV(dstr))->xcv_gv =
12167 ? gv_dup_inc(CvGV(sstr), param)
12168 : (param->flags & CLONEf_JOIN_IN)
12170 : gv_dup(CvGV(sstr), param);
12172 CvPADLIST(dstr) = padlist_dup(CvPADLIST(sstr), param);
12174 CvWEAKOUTSIDE(sstr)
12175 ? cv_dup( CvOUTSIDE(dstr), param)
12176 : cv_dup_inc(CvOUTSIDE(dstr), param);
12182 if (SvOBJECT(dstr) && SvTYPE(dstr) != SVt_PVIO)
12189 Perl_sv_dup_inc(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
12191 PERL_ARGS_ASSERT_SV_DUP_INC;
12192 return sstr ? SvREFCNT_inc(sv_dup_common(sstr, param)) : NULL;
12196 Perl_sv_dup(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
12198 SV *dstr = sstr ? sv_dup_common(sstr, param) : NULL;
12199 PERL_ARGS_ASSERT_SV_DUP;
12201 /* Track every SV that (at least initially) had a reference count of 0.
12202 We need to do this by holding an actual reference to it in this array.
12203 If we attempt to cheat, turn AvREAL_off(), and store only pointers
12204 (akin to the stashes hash, and the perl stack), we come unstuck if
12205 a weak reference (or other SV legitimately SvREFCNT() == 0 for this
12206 thread) is manipulated in a CLONE method, because CLONE runs before the
12207 unreferenced array is walked to find SVs still with SvREFCNT() == 0
12208 (and fix things up by giving each a reference via the temps stack).
12209 Instead, during CLONE, if the 0-referenced SV has SvREFCNT_inc() and
12210 then SvREFCNT_dec(), it will be cleaned up (and added to the free list)
12211 before the walk of unreferenced happens and a reference to that is SV
12212 added to the temps stack. At which point we have the same SV considered
12213 to be in use, and free to be re-used. Not good.
12215 if (dstr && !(param->flags & CLONEf_COPY_STACKS) && !SvREFCNT(dstr)) {
12216 assert(param->unreferenced);
12217 av_push(param->unreferenced, SvREFCNT_inc(dstr));
12223 /* duplicate a context */
12226 Perl_cx_dup(pTHX_ PERL_CONTEXT *cxs, I32 ix, I32 max, CLONE_PARAMS* param)
12228 PERL_CONTEXT *ncxs;
12230 PERL_ARGS_ASSERT_CX_DUP;
12233 return (PERL_CONTEXT*)NULL;
12235 /* look for it in the table first */
12236 ncxs = (PERL_CONTEXT*)ptr_table_fetch(PL_ptr_table, cxs);
12240 /* create anew and remember what it is */
12241 Newx(ncxs, max + 1, PERL_CONTEXT);
12242 ptr_table_store(PL_ptr_table, cxs, ncxs);
12243 Copy(cxs, ncxs, max + 1, PERL_CONTEXT);
12246 PERL_CONTEXT * const ncx = &ncxs[ix];
12247 if (CxTYPE(ncx) == CXt_SUBST) {
12248 Perl_croak(aTHX_ "Cloning substitution context is unimplemented");
12251 ncx->blk_oldcop = (COP*)any_dup(ncx->blk_oldcop, param->proto_perl);
12252 switch (CxTYPE(ncx)) {
12254 ncx->blk_sub.cv = (ncx->blk_sub.olddepth == 0
12255 ? cv_dup_inc(ncx->blk_sub.cv, param)
12256 : cv_dup(ncx->blk_sub.cv,param));
12257 ncx->blk_sub.argarray = (CxHASARGS(ncx)
12258 ? av_dup_inc(ncx->blk_sub.argarray,
12261 ncx->blk_sub.savearray = av_dup_inc(ncx->blk_sub.savearray,
12263 ncx->blk_sub.oldcomppad = (PAD*)ptr_table_fetch(PL_ptr_table,
12264 ncx->blk_sub.oldcomppad);
12267 ncx->blk_eval.old_namesv = sv_dup_inc(ncx->blk_eval.old_namesv,
12269 ncx->blk_eval.cur_text = sv_dup(ncx->blk_eval.cur_text, param);
12270 ncx->blk_eval.cv = cv_dup(ncx->blk_eval.cv, param);
12272 case CXt_LOOP_LAZYSV:
12273 ncx->blk_loop.state_u.lazysv.end
12274 = sv_dup_inc(ncx->blk_loop.state_u.lazysv.end, param);
12275 /* We are taking advantage of av_dup_inc and sv_dup_inc
12276 actually being the same function, and order equivalence of
12278 We can assert the later [but only at run time :-(] */
12279 assert ((void *) &ncx->blk_loop.state_u.ary.ary ==
12280 (void *) &ncx->blk_loop.state_u.lazysv.cur);
12282 ncx->blk_loop.state_u.ary.ary
12283 = av_dup_inc(ncx->blk_loop.state_u.ary.ary, param);
12284 case CXt_LOOP_LAZYIV:
12285 case CXt_LOOP_PLAIN:
12286 if (CxPADLOOP(ncx)) {
12287 ncx->blk_loop.itervar_u.oldcomppad
12288 = (PAD*)ptr_table_fetch(PL_ptr_table,
12289 ncx->blk_loop.itervar_u.oldcomppad);
12291 ncx->blk_loop.itervar_u.gv
12292 = gv_dup((const GV *)ncx->blk_loop.itervar_u.gv,
12297 ncx->blk_format.cv = cv_dup(ncx->blk_format.cv, param);
12298 ncx->blk_format.gv = gv_dup(ncx->blk_format.gv, param);
12299 ncx->blk_format.dfoutgv = gv_dup_inc(ncx->blk_format.dfoutgv,
12314 /* duplicate a stack info structure */
12317 Perl_si_dup(pTHX_ PERL_SI *si, CLONE_PARAMS* param)
12321 PERL_ARGS_ASSERT_SI_DUP;
12324 return (PERL_SI*)NULL;
12326 /* look for it in the table first */
12327 nsi = (PERL_SI*)ptr_table_fetch(PL_ptr_table, si);
12331 /* create anew and remember what it is */
12332 Newxz(nsi, 1, PERL_SI);
12333 ptr_table_store(PL_ptr_table, si, nsi);
12335 nsi->si_stack = av_dup_inc(si->si_stack, param);
12336 nsi->si_cxix = si->si_cxix;
12337 nsi->si_cxmax = si->si_cxmax;
12338 nsi->si_cxstack = cx_dup(si->si_cxstack, si->si_cxix, si->si_cxmax, param);
12339 nsi->si_type = si->si_type;
12340 nsi->si_prev = si_dup(si->si_prev, param);
12341 nsi->si_next = si_dup(si->si_next, param);
12342 nsi->si_markoff = si->si_markoff;
12347 #define POPINT(ss,ix) ((ss)[--(ix)].any_i32)
12348 #define TOPINT(ss,ix) ((ss)[ix].any_i32)
12349 #define POPLONG(ss,ix) ((ss)[--(ix)].any_long)
12350 #define TOPLONG(ss,ix) ((ss)[ix].any_long)
12351 #define POPIV(ss,ix) ((ss)[--(ix)].any_iv)
12352 #define TOPIV(ss,ix) ((ss)[ix].any_iv)
12353 #define POPUV(ss,ix) ((ss)[--(ix)].any_uv)
12354 #define TOPUV(ss,ix) ((ss)[ix].any_uv)
12355 #define POPBOOL(ss,ix) ((ss)[--(ix)].any_bool)
12356 #define TOPBOOL(ss,ix) ((ss)[ix].any_bool)
12357 #define POPPTR(ss,ix) ((ss)[--(ix)].any_ptr)
12358 #define TOPPTR(ss,ix) ((ss)[ix].any_ptr)
12359 #define POPDPTR(ss,ix) ((ss)[--(ix)].any_dptr)
12360 #define TOPDPTR(ss,ix) ((ss)[ix].any_dptr)
12361 #define POPDXPTR(ss,ix) ((ss)[--(ix)].any_dxptr)
12362 #define TOPDXPTR(ss,ix) ((ss)[ix].any_dxptr)
12365 #define pv_dup_inc(p) SAVEPV(p)
12366 #define pv_dup(p) SAVEPV(p)
12367 #define svp_dup_inc(p,pp) any_dup(p,pp)
12369 /* map any object to the new equivent - either something in the
12370 * ptr table, or something in the interpreter structure
12374 Perl_any_dup(pTHX_ void *v, const PerlInterpreter *proto_perl)
12378 PERL_ARGS_ASSERT_ANY_DUP;
12381 return (void*)NULL;
12383 /* look for it in the table first */
12384 ret = ptr_table_fetch(PL_ptr_table, v);
12388 /* see if it is part of the interpreter structure */
12389 if (v >= (void*)proto_perl && v < (void*)(proto_perl+1))
12390 ret = (void*)(((char*)aTHX) + (((char*)v) - (char*)proto_perl));
12398 /* duplicate the save stack */
12401 Perl_ss_dup(pTHX_ PerlInterpreter *proto_perl, CLONE_PARAMS* param)
12404 ANY * const ss = proto_perl->Isavestack;
12405 const I32 max = proto_perl->Isavestack_max;
12406 I32 ix = proto_perl->Isavestack_ix;
12419 void (*dptr) (void*);
12420 void (*dxptr) (pTHX_ void*);
12422 PERL_ARGS_ASSERT_SS_DUP;
12424 Newxz(nss, max, ANY);
12427 const UV uv = POPUV(ss,ix);
12428 const U8 type = (U8)uv & SAVE_MASK;
12430 TOPUV(nss,ix) = uv;
12432 case SAVEt_CLEARSV:
12434 case SAVEt_HELEM: /* hash element */
12435 sv = (const SV *)POPPTR(ss,ix);
12436 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12438 case SAVEt_ITEM: /* normal string */
12439 case SAVEt_GVSV: /* scalar slot in GV */
12440 case SAVEt_SV: /* scalar reference */
12441 sv = (const SV *)POPPTR(ss,ix);
12442 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12445 case SAVEt_MORTALIZESV:
12446 sv = (const SV *)POPPTR(ss,ix);
12447 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12449 case SAVEt_SHARED_PVREF: /* char* in shared space */
12450 c = (char*)POPPTR(ss,ix);
12451 TOPPTR(nss,ix) = savesharedpv(c);
12452 ptr = POPPTR(ss,ix);
12453 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12455 case SAVEt_GENERIC_SVREF: /* generic sv */
12456 case SAVEt_SVREF: /* scalar reference */
12457 sv = (const SV *)POPPTR(ss,ix);
12458 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12459 ptr = POPPTR(ss,ix);
12460 TOPPTR(nss,ix) = svp_dup_inc((SV**)ptr, proto_perl);/* XXXXX */
12462 case SAVEt_HV: /* hash reference */
12463 case SAVEt_AV: /* array reference */
12464 sv = (const SV *) POPPTR(ss,ix);
12465 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12467 case SAVEt_COMPPAD:
12469 sv = (const SV *) POPPTR(ss,ix);
12470 TOPPTR(nss,ix) = sv_dup(sv, param);
12472 case SAVEt_INT: /* int reference */
12473 ptr = POPPTR(ss,ix);
12474 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12475 intval = (int)POPINT(ss,ix);
12476 TOPINT(nss,ix) = intval;
12478 case SAVEt_LONG: /* long reference */
12479 ptr = POPPTR(ss,ix);
12480 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12481 longval = (long)POPLONG(ss,ix);
12482 TOPLONG(nss,ix) = longval;
12484 case SAVEt_I32: /* I32 reference */
12485 ptr = POPPTR(ss,ix);
12486 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12488 TOPINT(nss,ix) = i;
12490 case SAVEt_IV: /* IV reference */
12491 ptr = POPPTR(ss,ix);
12492 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12494 TOPIV(nss,ix) = iv;
12496 case SAVEt_HPTR: /* HV* reference */
12497 case SAVEt_APTR: /* AV* reference */
12498 case SAVEt_SPTR: /* SV* reference */
12499 ptr = POPPTR(ss,ix);
12500 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12501 sv = (const SV *)POPPTR(ss,ix);
12502 TOPPTR(nss,ix) = sv_dup(sv, param);
12504 case SAVEt_VPTR: /* random* reference */
12505 ptr = POPPTR(ss,ix);
12506 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12508 case SAVEt_INT_SMALL:
12509 case SAVEt_I32_SMALL:
12510 case SAVEt_I16: /* I16 reference */
12511 case SAVEt_I8: /* I8 reference */
12513 ptr = POPPTR(ss,ix);
12514 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12516 case SAVEt_GENERIC_PVREF: /* generic char* */
12517 case SAVEt_PPTR: /* char* reference */
12518 ptr = POPPTR(ss,ix);
12519 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12520 c = (char*)POPPTR(ss,ix);
12521 TOPPTR(nss,ix) = pv_dup(c);
12523 case SAVEt_GP: /* scalar reference */
12524 gp = (GP*)POPPTR(ss,ix);
12525 TOPPTR(nss,ix) = gp = gp_dup(gp, param);
12526 (void)GpREFCNT_inc(gp);
12527 gv = (const GV *)POPPTR(ss,ix);
12528 TOPPTR(nss,ix) = gv_dup_inc(gv, param);
12531 ptr = POPPTR(ss,ix);
12532 if (ptr && (((OP*)ptr)->op_private & OPpREFCOUNTED)) {
12533 /* these are assumed to be refcounted properly */
12535 switch (((OP*)ptr)->op_type) {
12537 case OP_LEAVESUBLV:
12541 case OP_LEAVEWRITE:
12542 TOPPTR(nss,ix) = ptr;
12545 (void) OpREFCNT_inc(o);
12549 TOPPTR(nss,ix) = NULL;
12554 TOPPTR(nss,ix) = NULL;
12556 case SAVEt_FREECOPHH:
12557 ptr = POPPTR(ss,ix);
12558 TOPPTR(nss,ix) = cophh_copy((COPHH *)ptr);
12561 hv = (const HV *)POPPTR(ss,ix);
12562 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
12564 TOPINT(nss,ix) = i;
12567 c = (char*)POPPTR(ss,ix);
12568 TOPPTR(nss,ix) = pv_dup_inc(c);
12570 case SAVEt_STACK_POS: /* Position on Perl stack */
12572 TOPINT(nss,ix) = i;
12574 case SAVEt_DESTRUCTOR:
12575 ptr = POPPTR(ss,ix);
12576 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
12577 dptr = POPDPTR(ss,ix);
12578 TOPDPTR(nss,ix) = DPTR2FPTR(void (*)(void*),
12579 any_dup(FPTR2DPTR(void *, dptr),
12582 case SAVEt_DESTRUCTOR_X:
12583 ptr = POPPTR(ss,ix);
12584 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
12585 dxptr = POPDXPTR(ss,ix);
12586 TOPDXPTR(nss,ix) = DPTR2FPTR(void (*)(pTHX_ void*),
12587 any_dup(FPTR2DPTR(void *, dxptr),
12590 case SAVEt_REGCONTEXT:
12592 ix -= uv >> SAVE_TIGHT_SHIFT;
12594 case SAVEt_AELEM: /* array element */
12595 sv = (const SV *)POPPTR(ss,ix);
12596 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12598 TOPINT(nss,ix) = i;
12599 av = (const AV *)POPPTR(ss,ix);
12600 TOPPTR(nss,ix) = av_dup_inc(av, param);
12603 ptr = POPPTR(ss,ix);
12604 TOPPTR(nss,ix) = ptr;
12607 ptr = POPPTR(ss,ix);
12608 ptr = cophh_copy((COPHH*)ptr);
12609 TOPPTR(nss,ix) = ptr;
12611 TOPINT(nss,ix) = i;
12612 if (i & HINT_LOCALIZE_HH) {
12613 hv = (const HV *)POPPTR(ss,ix);
12614 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
12617 case SAVEt_PADSV_AND_MORTALIZE:
12618 longval = (long)POPLONG(ss,ix);
12619 TOPLONG(nss,ix) = longval;
12620 ptr = POPPTR(ss,ix);
12621 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12622 sv = (const SV *)POPPTR(ss,ix);
12623 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12625 case SAVEt_SET_SVFLAGS:
12627 TOPINT(nss,ix) = i;
12629 TOPINT(nss,ix) = i;
12630 sv = (const SV *)POPPTR(ss,ix);
12631 TOPPTR(nss,ix) = sv_dup(sv, param);
12633 case SAVEt_RE_STATE:
12635 const struct re_save_state *const old_state
12636 = (struct re_save_state *)
12637 (ss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
12638 struct re_save_state *const new_state
12639 = (struct re_save_state *)
12640 (nss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
12642 Copy(old_state, new_state, 1, struct re_save_state);
12643 ix -= SAVESTACK_ALLOC_FOR_RE_SAVE_STATE;
12645 new_state->re_state_bostr
12646 = pv_dup(old_state->re_state_bostr);
12647 new_state->re_state_reginput
12648 = pv_dup(old_state->re_state_reginput);
12649 new_state->re_state_regeol
12650 = pv_dup(old_state->re_state_regeol);
12651 #ifdef PERL_OLD_COPY_ON_WRITE
12652 new_state->re_state_nrs
12653 = sv_dup(old_state->re_state_nrs, param);
12655 new_state->re_state_reg_magic
12656 = (MAGIC*) any_dup(old_state->re_state_reg_magic,
12658 new_state->re_state_reg_oldcurpm
12659 = (PMOP*) any_dup(old_state->re_state_reg_oldcurpm,
12661 new_state->re_state_reg_curpm
12662 = (PMOP*) any_dup(old_state->re_state_reg_curpm,
12664 new_state->re_state_reg_oldsaved
12665 = pv_dup(old_state->re_state_reg_oldsaved);
12666 new_state->re_state_reg_poscache
12667 = pv_dup(old_state->re_state_reg_poscache);
12668 new_state->re_state_reg_starttry
12669 = pv_dup(old_state->re_state_reg_starttry);
12672 case SAVEt_COMPILE_WARNINGS:
12673 ptr = POPPTR(ss,ix);
12674 TOPPTR(nss,ix) = DUP_WARNINGS((STRLEN*)ptr);
12677 ptr = POPPTR(ss,ix);
12678 TOPPTR(nss,ix) = parser_dup((const yy_parser*)ptr, param);
12682 "panic: ss_dup inconsistency (%"IVdf")", (IV) type);
12690 /* if sv is a stash, call $class->CLONE_SKIP(), and set the SVphv_CLONEABLE
12691 * flag to the result. This is done for each stash before cloning starts,
12692 * so we know which stashes want their objects cloned */
12695 do_mark_cloneable_stash(pTHX_ SV *const sv)
12697 const HEK * const hvname = HvNAME_HEK((const HV *)sv);
12699 GV* const cloner = gv_fetchmethod_autoload(MUTABLE_HV(sv), "CLONE_SKIP", 0);
12700 SvFLAGS(sv) |= SVphv_CLONEABLE; /* clone objects by default */
12701 if (cloner && GvCV(cloner)) {
12708 mXPUSHs(newSVhek(hvname));
12710 call_sv(MUTABLE_SV(GvCV(cloner)), G_SCALAR);
12717 SvFLAGS(sv) &= ~SVphv_CLONEABLE;
12725 =for apidoc perl_clone
12727 Create and return a new interpreter by cloning the current one.
12729 perl_clone takes these flags as parameters:
12731 CLONEf_COPY_STACKS - is used to, well, copy the stacks also,
12732 without it we only clone the data and zero the stacks,
12733 with it we copy the stacks and the new perl interpreter is
12734 ready to run at the exact same point as the previous one.
12735 The pseudo-fork code uses COPY_STACKS while the
12736 threads->create doesn't.
12738 CLONEf_KEEP_PTR_TABLE -
12739 perl_clone keeps a ptr_table with the pointer of the old
12740 variable as a key and the new variable as a value,
12741 this allows it to check if something has been cloned and not
12742 clone it again but rather just use the value and increase the
12743 refcount. If KEEP_PTR_TABLE is not set then perl_clone will kill
12744 the ptr_table using the function
12745 C<ptr_table_free(PL_ptr_table); PL_ptr_table = NULL;>,
12746 reason to keep it around is if you want to dup some of your own
12747 variable who are outside the graph perl scans, example of this
12748 code is in threads.xs create.
12750 CLONEf_CLONE_HOST -
12751 This is a win32 thing, it is ignored on unix, it tells perls
12752 win32host code (which is c++) to clone itself, this is needed on
12753 win32 if you want to run two threads at the same time,
12754 if you just want to do some stuff in a separate perl interpreter
12755 and then throw it away and return to the original one,
12756 you don't need to do anything.
12761 /* XXX the above needs expanding by someone who actually understands it ! */
12762 EXTERN_C PerlInterpreter *
12763 perl_clone_host(PerlInterpreter* proto_perl, UV flags);
12766 perl_clone(PerlInterpreter *proto_perl, UV flags)
12769 #ifdef PERL_IMPLICIT_SYS
12771 PERL_ARGS_ASSERT_PERL_CLONE;
12773 /* perlhost.h so we need to call into it
12774 to clone the host, CPerlHost should have a c interface, sky */
12776 if (flags & CLONEf_CLONE_HOST) {
12777 return perl_clone_host(proto_perl,flags);
12779 return perl_clone_using(proto_perl, flags,
12781 proto_perl->IMemShared,
12782 proto_perl->IMemParse,
12784 proto_perl->IStdIO,
12788 proto_perl->IProc);
12792 perl_clone_using(PerlInterpreter *proto_perl, UV flags,
12793 struct IPerlMem* ipM, struct IPerlMem* ipMS,
12794 struct IPerlMem* ipMP, struct IPerlEnv* ipE,
12795 struct IPerlStdIO* ipStd, struct IPerlLIO* ipLIO,
12796 struct IPerlDir* ipD, struct IPerlSock* ipS,
12797 struct IPerlProc* ipP)
12799 /* XXX many of the string copies here can be optimized if they're
12800 * constants; they need to be allocated as common memory and just
12801 * their pointers copied. */
12804 CLONE_PARAMS clone_params;
12805 CLONE_PARAMS* const param = &clone_params;
12807 PerlInterpreter * const my_perl = (PerlInterpreter*)(*ipM->pMalloc)(ipM, sizeof(PerlInterpreter));
12809 PERL_ARGS_ASSERT_PERL_CLONE_USING;
12810 #else /* !PERL_IMPLICIT_SYS */
12812 CLONE_PARAMS clone_params;
12813 CLONE_PARAMS* param = &clone_params;
12814 PerlInterpreter * const my_perl = (PerlInterpreter*)PerlMem_malloc(sizeof(PerlInterpreter));
12816 PERL_ARGS_ASSERT_PERL_CLONE;
12817 #endif /* PERL_IMPLICIT_SYS */
12819 /* for each stash, determine whether its objects should be cloned */
12820 S_visit(proto_perl, do_mark_cloneable_stash, SVt_PVHV, SVTYPEMASK);
12821 PERL_SET_THX(my_perl);
12824 PoisonNew(my_perl, 1, PerlInterpreter);
12827 PL_defstash = NULL; /* may be used by perl malloc() */
12830 PL_scopestack_name = 0;
12832 PL_savestack_ix = 0;
12833 PL_savestack_max = -1;
12834 PL_sig_pending = 0;
12836 Zero(&PL_debug_pad, 1, struct perl_debug_pad);
12837 # ifdef DEBUG_LEAKING_SCALARS
12838 PL_sv_serial = (((UV)my_perl >> 2) & 0xfff) * 1000000;
12840 #else /* !DEBUGGING */
12841 Zero(my_perl, 1, PerlInterpreter);
12842 #endif /* DEBUGGING */
12844 #ifdef PERL_IMPLICIT_SYS
12845 /* host pointers */
12847 PL_MemShared = ipMS;
12848 PL_MemParse = ipMP;
12855 #endif /* PERL_IMPLICIT_SYS */
12857 param->flags = flags;
12858 /* Nothing in the core code uses this, but we make it available to
12859 extensions (using mg_dup). */
12860 param->proto_perl = proto_perl;
12861 /* Likely nothing will use this, but it is initialised to be consistent
12862 with Perl_clone_params_new(). */
12863 param->new_perl = my_perl;
12864 param->unreferenced = NULL;
12866 INIT_TRACK_MEMPOOL(my_perl->Imemory_debug_header, my_perl);
12868 PL_body_arenas = NULL;
12869 Zero(&PL_body_roots, 1, PL_body_roots);
12872 PL_sv_objcount = 0;
12874 PL_sv_arenaroot = NULL;
12876 PL_debug = proto_perl->Idebug;
12878 PL_hash_seed = proto_perl->Ihash_seed;
12879 PL_rehash_seed = proto_perl->Irehash_seed;
12881 /* dbargs array probably holds garbage */
12884 PL_compiling = proto_perl->Icompiling;
12886 /* pseudo environmental stuff */
12887 PL_origargc = proto_perl->Iorigargc;
12888 PL_origargv = proto_perl->Iorigargv;
12890 /* Set tainting stuff before PerlIO_debug can possibly get called */
12891 PL_tainting = proto_perl->Itainting;
12892 PL_taint_warn = proto_perl->Itaint_warn;
12894 PL_minus_c = proto_perl->Iminus_c;
12896 PL_localpatches = proto_perl->Ilocalpatches;
12897 PL_splitstr = proto_perl->Isplitstr;
12898 PL_minus_n = proto_perl->Iminus_n;
12899 PL_minus_p = proto_perl->Iminus_p;
12900 PL_minus_l = proto_perl->Iminus_l;
12901 PL_minus_a = proto_perl->Iminus_a;
12902 PL_minus_E = proto_perl->Iminus_E;
12903 PL_minus_F = proto_perl->Iminus_F;
12904 PL_doswitches = proto_perl->Idoswitches;
12905 PL_dowarn = proto_perl->Idowarn;
12906 PL_sawampersand = proto_perl->Isawampersand;
12907 PL_unsafe = proto_perl->Iunsafe;
12908 PL_perldb = proto_perl->Iperldb;
12909 PL_perl_destruct_level = proto_perl->Iperl_destruct_level;
12910 PL_exit_flags = proto_perl->Iexit_flags;
12912 /* XXX time(&PL_basetime) when asked for? */
12913 PL_basetime = proto_perl->Ibasetime;
12915 PL_maxsysfd = proto_perl->Imaxsysfd;
12916 PL_statusvalue = proto_perl->Istatusvalue;
12918 PL_statusvalue_vms = proto_perl->Istatusvalue_vms;
12920 PL_statusvalue_posix = proto_perl->Istatusvalue_posix;
12923 /* RE engine related */
12924 Zero(&PL_reg_state, 1, struct re_save_state);
12925 PL_regmatch_slab = NULL;
12927 PL_sub_generation = proto_perl->Isub_generation;
12929 /* funky return mechanisms */
12930 PL_forkprocess = proto_perl->Iforkprocess;
12932 /* internal state */
12933 PL_maxo = proto_perl->Imaxo;
12935 PL_main_start = proto_perl->Imain_start;
12936 PL_eval_root = proto_perl->Ieval_root;
12937 PL_eval_start = proto_perl->Ieval_start;
12939 PL_filemode = proto_perl->Ifilemode;
12940 PL_lastfd = proto_perl->Ilastfd;
12941 PL_oldname = proto_perl->Ioldname; /* XXX not quite right */
12944 PL_gensym = proto_perl->Igensym;
12946 PL_laststatval = proto_perl->Ilaststatval;
12947 PL_laststype = proto_perl->Ilaststype;
12950 PL_profiledata = NULL;
12952 PL_generation = proto_perl->Igeneration;
12954 PL_in_clean_objs = proto_perl->Iin_clean_objs;
12955 PL_in_clean_all = proto_perl->Iin_clean_all;
12957 PL_delaymagic_uid = proto_perl->Idelaymagic_uid;
12958 PL_delaymagic_euid = proto_perl->Idelaymagic_euid;
12959 PL_delaymagic_gid = proto_perl->Idelaymagic_gid;
12960 PL_delaymagic_egid = proto_perl->Idelaymagic_egid;
12961 PL_nomemok = proto_perl->Inomemok;
12962 PL_an = proto_perl->Ian;
12963 PL_evalseq = proto_perl->Ievalseq;
12964 PL_origenviron = proto_perl->Iorigenviron; /* XXX not quite right */
12965 PL_origalen = proto_perl->Iorigalen;
12967 PL_sighandlerp = proto_perl->Isighandlerp;
12969 PL_runops = proto_perl->Irunops;
12971 PL_subline = proto_perl->Isubline;
12974 PL_cryptseen = proto_perl->Icryptseen;
12977 PL_hints = proto_perl->Ihints;
12979 #ifdef USE_LOCALE_COLLATE
12980 PL_collation_ix = proto_perl->Icollation_ix;
12981 PL_collation_standard = proto_perl->Icollation_standard;
12982 PL_collxfrm_base = proto_perl->Icollxfrm_base;
12983 PL_collxfrm_mult = proto_perl->Icollxfrm_mult;
12984 #endif /* USE_LOCALE_COLLATE */
12986 #ifdef USE_LOCALE_NUMERIC
12987 PL_numeric_standard = proto_perl->Inumeric_standard;
12988 PL_numeric_local = proto_perl->Inumeric_local;
12989 #endif /* !USE_LOCALE_NUMERIC */
12991 /* Did the locale setup indicate UTF-8? */
12992 PL_utf8locale = proto_perl->Iutf8locale;
12993 /* Unicode features (see perlrun/-C) */
12994 PL_unicode = proto_perl->Iunicode;
12996 /* Pre-5.8 signals control */
12997 PL_signals = proto_perl->Isignals;
12999 /* times() ticks per second */
13000 PL_clocktick = proto_perl->Iclocktick;
13002 /* Recursion stopper for PerlIO_find_layer */
13003 PL_in_load_module = proto_perl->Iin_load_module;
13005 /* sort() routine */
13006 PL_sort_RealCmp = proto_perl->Isort_RealCmp;
13008 /* Not really needed/useful since the reenrant_retint is "volatile",
13009 * but do it for consistency's sake. */
13010 PL_reentrant_retint = proto_perl->Ireentrant_retint;
13012 /* Hooks to shared SVs and locks. */
13013 PL_sharehook = proto_perl->Isharehook;
13014 PL_lockhook = proto_perl->Ilockhook;
13015 PL_unlockhook = proto_perl->Iunlockhook;
13016 PL_threadhook = proto_perl->Ithreadhook;
13017 PL_destroyhook = proto_perl->Idestroyhook;
13018 PL_signalhook = proto_perl->Isignalhook;
13020 PL_globhook = proto_perl->Iglobhook;
13023 PL_last_swash_hv = NULL; /* reinits on demand */
13024 PL_last_swash_klen = 0;
13025 PL_last_swash_key[0]= '\0';
13026 PL_last_swash_tmps = (U8*)NULL;
13027 PL_last_swash_slen = 0;
13029 PL_glob_index = proto_perl->Iglob_index;
13030 PL_srand_called = proto_perl->Isrand_called;
13032 if (flags & CLONEf_COPY_STACKS) {
13033 /* next allocation will be PL_tmps_stack[PL_tmps_ix+1] */
13034 PL_tmps_ix = proto_perl->Itmps_ix;
13035 PL_tmps_max = proto_perl->Itmps_max;
13036 PL_tmps_floor = proto_perl->Itmps_floor;
13038 /* next push_scope()/ENTER sets PL_scopestack[PL_scopestack_ix]
13039 * NOTE: unlike the others! */
13040 PL_scopestack_ix = proto_perl->Iscopestack_ix;
13041 PL_scopestack_max = proto_perl->Iscopestack_max;
13043 /* next SSPUSHFOO() sets PL_savestack[PL_savestack_ix]
13044 * NOTE: unlike the others! */
13045 PL_savestack_ix = proto_perl->Isavestack_ix;
13046 PL_savestack_max = proto_perl->Isavestack_max;
13049 PL_start_env = proto_perl->Istart_env; /* XXXXXX */
13050 PL_top_env = &PL_start_env;
13052 PL_op = proto_perl->Iop;
13055 PL_Xpv = (XPV*)NULL;
13056 my_perl->Ina = proto_perl->Ina;
13058 PL_statbuf = proto_perl->Istatbuf;
13059 PL_statcache = proto_perl->Istatcache;
13062 PL_timesbuf = proto_perl->Itimesbuf;
13065 PL_tainted = proto_perl->Itainted;
13066 PL_curpm = proto_perl->Icurpm; /* XXX No PMOP ref count */
13068 PL_chopset = proto_perl->Ichopset; /* XXX never deallocated */
13070 PL_restartjmpenv = proto_perl->Irestartjmpenv;
13071 PL_restartop = proto_perl->Irestartop;
13072 PL_in_eval = proto_perl->Iin_eval;
13073 PL_delaymagic = proto_perl->Idelaymagic;
13074 PL_phase = proto_perl->Iphase;
13075 PL_localizing = proto_perl->Ilocalizing;
13077 PL_hv_fetch_ent_mh = NULL;
13078 PL_modcount = proto_perl->Imodcount;
13079 PL_lastgotoprobe = NULL;
13080 PL_dumpindent = proto_perl->Idumpindent;
13082 PL_efloatbuf = NULL; /* reinits on demand */
13083 PL_efloatsize = 0; /* reinits on demand */
13087 PL_regdummy = proto_perl->Iregdummy;
13088 PL_colorset = 0; /* reinits PL_colors[] */
13089 /*PL_colors[6] = {0,0,0,0,0,0};*/
13091 /* Pluggable optimizer */
13092 PL_peepp = proto_perl->Ipeepp;
13093 PL_rpeepp = proto_perl->Irpeepp;
13094 /* op_free() hook */
13095 PL_opfreehook = proto_perl->Iopfreehook;
13097 #ifdef USE_REENTRANT_API
13098 /* XXX: things like -Dm will segfault here in perlio, but doing
13099 * PERL_SET_CONTEXT(proto_perl);
13100 * breaks too many other things
13102 Perl_reentrant_init(aTHX);
13105 /* create SV map for pointer relocation */
13106 PL_ptr_table = ptr_table_new();
13108 /* initialize these special pointers as early as possible */
13110 ptr_table_store(PL_ptr_table, &proto_perl->Isv_undef, &PL_sv_undef);
13111 ptr_table_store(PL_ptr_table, &proto_perl->Isv_no, &PL_sv_no);
13112 ptr_table_store(PL_ptr_table, &proto_perl->Isv_yes, &PL_sv_yes);
13114 /* create (a non-shared!) shared string table */
13115 PL_strtab = newHV();
13116 HvSHAREKEYS_off(PL_strtab);
13117 hv_ksplit(PL_strtab, HvTOTALKEYS(proto_perl->Istrtab));
13118 ptr_table_store(PL_ptr_table, proto_perl->Istrtab, PL_strtab);
13120 /* This PV will be free'd special way so must set it same way op.c does */
13121 PL_compiling.cop_file = savesharedpv(PL_compiling.cop_file);
13122 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_file, PL_compiling.cop_file);
13124 ptr_table_store(PL_ptr_table, &proto_perl->Icompiling, &PL_compiling);
13125 PL_compiling.cop_warnings = DUP_WARNINGS(PL_compiling.cop_warnings);
13126 CopHINTHASH_set(&PL_compiling, cophh_copy(CopHINTHASH_get(&PL_compiling)));
13127 PL_curcop = (COP*)any_dup(proto_perl->Icurcop, proto_perl);
13129 param->stashes = newAV(); /* Setup array of objects to call clone on */
13130 /* This makes no difference to the implementation, as it always pushes
13131 and shifts pointers to other SVs without changing their reference
13132 count, with the array becoming empty before it is freed. However, it
13133 makes it conceptually clear what is going on, and will avoid some
13134 work inside av.c, filling slots between AvFILL() and AvMAX() with
13135 &PL_sv_undef, and SvREFCNT_dec()ing those. */
13136 AvREAL_off(param->stashes);
13138 if (!(flags & CLONEf_COPY_STACKS)) {
13139 param->unreferenced = newAV();
13142 #ifdef PERLIO_LAYERS
13143 /* Clone PerlIO tables as soon as we can handle general xx_dup() */
13144 PerlIO_clone(aTHX_ proto_perl, param);
13147 PL_envgv = gv_dup(proto_perl->Ienvgv, param);
13148 PL_incgv = gv_dup(proto_perl->Iincgv, param);
13149 PL_hintgv = gv_dup(proto_perl->Ihintgv, param);
13150 PL_origfilename = SAVEPV(proto_perl->Iorigfilename);
13151 PL_diehook = sv_dup_inc(proto_perl->Idiehook, param);
13152 PL_warnhook = sv_dup_inc(proto_perl->Iwarnhook, param);
13155 PL_patchlevel = sv_dup_inc(proto_perl->Ipatchlevel, param);
13156 PL_apiversion = sv_dup_inc(proto_perl->Iapiversion, param);
13157 PL_inplace = SAVEPV(proto_perl->Iinplace);
13158 PL_e_script = sv_dup_inc(proto_perl->Ie_script, param);
13160 /* magical thingies */
13161 PL_formfeed = sv_dup(proto_perl->Iformfeed, param);
13163 PL_encoding = sv_dup(proto_perl->Iencoding, param);
13165 sv_setpvs(PERL_DEBUG_PAD(0), ""); /* For regex debugging. */
13166 sv_setpvs(PERL_DEBUG_PAD(1), ""); /* ext/re needs these */
13167 sv_setpvs(PERL_DEBUG_PAD(2), ""); /* even without DEBUGGING. */
13170 /* Clone the regex array */
13171 /* ORANGE FIXME for plugins, probably in the SV dup code.
13172 newSViv(PTR2IV(CALLREGDUPE(
13173 INT2PTR(REGEXP *, SvIVX(regex)), param))))
13175 PL_regex_padav = av_dup_inc(proto_perl->Iregex_padav, param);
13176 PL_regex_pad = AvARRAY(PL_regex_padav);
13178 PL_stashpadmax = proto_perl->Istashpadmax;
13179 PL_stashpadix = proto_perl->Istashpadix ;
13180 Newx(PL_stashpad, PL_stashpadmax, HV *);
13183 for (; o < PL_stashpadmax; ++o)
13184 PL_stashpad[o] = hv_dup(proto_perl->Istashpad[o], param);
13187 /* shortcuts to various I/O objects */
13188 PL_ofsgv = gv_dup_inc(proto_perl->Iofsgv, param);
13189 PL_stdingv = gv_dup(proto_perl->Istdingv, param);
13190 PL_stderrgv = gv_dup(proto_perl->Istderrgv, param);
13191 PL_defgv = gv_dup(proto_perl->Idefgv, param);
13192 PL_argvgv = gv_dup(proto_perl->Iargvgv, param);
13193 PL_argvoutgv = gv_dup(proto_perl->Iargvoutgv, param);
13194 PL_argvout_stack = av_dup_inc(proto_perl->Iargvout_stack, param);
13196 /* shortcuts to regexp stuff */
13197 PL_replgv = gv_dup(proto_perl->Ireplgv, param);
13199 /* shortcuts to misc objects */
13200 PL_errgv = gv_dup(proto_perl->Ierrgv, param);
13202 /* shortcuts to debugging objects */
13203 PL_DBgv = gv_dup(proto_perl->IDBgv, param);
13204 PL_DBline = gv_dup(proto_perl->IDBline, param);
13205 PL_DBsub = gv_dup(proto_perl->IDBsub, param);
13206 PL_DBsingle = sv_dup(proto_perl->IDBsingle, param);
13207 PL_DBtrace = sv_dup(proto_perl->IDBtrace, param);
13208 PL_DBsignal = sv_dup(proto_perl->IDBsignal, param);
13210 /* symbol tables */
13211 PL_defstash = hv_dup_inc(proto_perl->Idefstash, param);
13212 PL_curstash = hv_dup_inc(proto_perl->Icurstash, param);
13213 PL_debstash = hv_dup(proto_perl->Idebstash, param);
13214 PL_globalstash = hv_dup(proto_perl->Iglobalstash, param);
13215 PL_curstname = sv_dup_inc(proto_perl->Icurstname, param);
13217 PL_beginav = av_dup_inc(proto_perl->Ibeginav, param);
13218 PL_beginav_save = av_dup_inc(proto_perl->Ibeginav_save, param);
13219 PL_checkav_save = av_dup_inc(proto_perl->Icheckav_save, param);
13220 PL_unitcheckav = av_dup_inc(proto_perl->Iunitcheckav, param);
13221 PL_unitcheckav_save = av_dup_inc(proto_perl->Iunitcheckav_save, param);
13222 PL_endav = av_dup_inc(proto_perl->Iendav, param);
13223 PL_checkav = av_dup_inc(proto_perl->Icheckav, param);
13224 PL_initav = av_dup_inc(proto_perl->Iinitav, param);
13226 PL_isarev = hv_dup_inc(proto_perl->Iisarev, param);
13228 /* subprocess state */
13229 PL_fdpid = av_dup_inc(proto_perl->Ifdpid, param);
13231 if (proto_perl->Iop_mask)
13232 PL_op_mask = SAVEPVN(proto_perl->Iop_mask, PL_maxo);
13235 /* PL_asserting = proto_perl->Iasserting; */
13237 /* current interpreter roots */
13238 PL_main_cv = cv_dup_inc(proto_perl->Imain_cv, param);
13240 PL_main_root = OpREFCNT_inc(proto_perl->Imain_root);
13243 /* runtime control stuff */
13244 PL_curcopdb = (COP*)any_dup(proto_perl->Icurcopdb, proto_perl);
13246 PL_preambleav = av_dup_inc(proto_perl->Ipreambleav, param);
13248 PL_ors_sv = sv_dup_inc(proto_perl->Iors_sv, param);
13250 /* interpreter atexit processing */
13251 PL_exitlistlen = proto_perl->Iexitlistlen;
13252 if (PL_exitlistlen) {
13253 Newx(PL_exitlist, PL_exitlistlen, PerlExitListEntry);
13254 Copy(proto_perl->Iexitlist, PL_exitlist, PL_exitlistlen, PerlExitListEntry);
13257 PL_exitlist = (PerlExitListEntry*)NULL;
13259 PL_my_cxt_size = proto_perl->Imy_cxt_size;
13260 if (PL_my_cxt_size) {
13261 Newx(PL_my_cxt_list, PL_my_cxt_size, void *);
13262 Copy(proto_perl->Imy_cxt_list, PL_my_cxt_list, PL_my_cxt_size, void *);
13263 #ifdef PERL_GLOBAL_STRUCT_PRIVATE
13264 Newx(PL_my_cxt_keys, PL_my_cxt_size, const char *);
13265 Copy(proto_perl->Imy_cxt_keys, PL_my_cxt_keys, PL_my_cxt_size, char *);
13269 PL_my_cxt_list = (void**)NULL;
13270 #ifdef PERL_GLOBAL_STRUCT_PRIVATE
13271 PL_my_cxt_keys = (const char**)NULL;
13274 PL_modglobal = hv_dup_inc(proto_perl->Imodglobal, param);
13275 PL_custom_op_names = hv_dup_inc(proto_perl->Icustom_op_names,param);
13276 PL_custom_op_descs = hv_dup_inc(proto_perl->Icustom_op_descs,param);
13277 PL_custom_ops = hv_dup_inc(proto_perl->Icustom_ops, param);
13279 PL_compcv = cv_dup(proto_perl->Icompcv, param);
13281 PAD_CLONE_VARS(proto_perl, param);
13283 #ifdef HAVE_INTERP_INTERN
13284 sys_intern_dup(&proto_perl->Isys_intern, &PL_sys_intern);
13287 PL_DBcv = cv_dup(proto_perl->IDBcv, param);
13289 #ifdef PERL_USES_PL_PIDSTATUS
13290 PL_pidstatus = newHV(); /* XXX flag for cloning? */
13292 PL_osname = SAVEPV(proto_perl->Iosname);
13293 PL_parser = parser_dup(proto_perl->Iparser, param);
13295 /* XXX this only works if the saved cop has already been cloned */
13296 if (proto_perl->Iparser) {
13297 PL_parser->saved_curcop = (COP*)any_dup(
13298 proto_perl->Iparser->saved_curcop,
13302 PL_subname = sv_dup_inc(proto_perl->Isubname, param);
13304 #ifdef USE_LOCALE_COLLATE
13305 PL_collation_name = SAVEPV(proto_perl->Icollation_name);
13306 #endif /* USE_LOCALE_COLLATE */
13308 #ifdef USE_LOCALE_NUMERIC
13309 PL_numeric_name = SAVEPV(proto_perl->Inumeric_name);
13310 PL_numeric_radix_sv = sv_dup_inc(proto_perl->Inumeric_radix_sv, param);
13311 #endif /* !USE_LOCALE_NUMERIC */
13313 /* Unicode inversion lists */
13314 PL_ASCII = sv_dup_inc(proto_perl->IASCII, param);
13315 PL_Latin1 = sv_dup_inc(proto_perl->ILatin1, param);
13317 PL_PerlSpace = sv_dup_inc(proto_perl->IPerlSpace, param);
13318 PL_XPerlSpace = sv_dup_inc(proto_perl->IXPerlSpace, param);
13320 PL_L1PosixAlnum = sv_dup_inc(proto_perl->IL1PosixAlnum, param);
13321 PL_PosixAlnum = sv_dup_inc(proto_perl->IPosixAlnum, param);
13323 PL_L1PosixAlpha = sv_dup_inc(proto_perl->IL1PosixAlpha, param);
13324 PL_PosixAlpha = sv_dup_inc(proto_perl->IPosixAlpha, param);
13326 PL_PosixBlank = sv_dup_inc(proto_perl->IPosixBlank, param);
13327 PL_XPosixBlank = sv_dup_inc(proto_perl->IXPosixBlank, param);
13329 PL_L1Cased = sv_dup_inc(proto_perl->IL1Cased, param);
13331 PL_PosixCntrl = sv_dup_inc(proto_perl->IPosixCntrl, param);
13332 PL_XPosixCntrl = sv_dup_inc(proto_perl->IXPosixCntrl, param);
13334 PL_PosixDigit = sv_dup_inc(proto_perl->IPosixDigit, param);
13336 PL_L1PosixGraph = sv_dup_inc(proto_perl->IL1PosixGraph, param);
13337 PL_PosixGraph = sv_dup_inc(proto_perl->IPosixGraph, param);
13339 PL_L1PosixLower = sv_dup_inc(proto_perl->IL1PosixLower, param);
13340 PL_PosixLower = sv_dup_inc(proto_perl->IPosixLower, param);
13342 PL_L1PosixPrint = sv_dup_inc(proto_perl->IL1PosixPrint, param);
13343 PL_PosixPrint = sv_dup_inc(proto_perl->IPosixPrint, param);
13345 PL_L1PosixPunct = sv_dup_inc(proto_perl->IL1PosixPunct, param);
13346 PL_PosixPunct = sv_dup_inc(proto_perl->IPosixPunct, param);
13348 PL_PosixSpace = sv_dup_inc(proto_perl->IPosixSpace, param);
13349 PL_XPosixSpace = sv_dup_inc(proto_perl->IXPosixSpace, param);
13351 PL_L1PosixUpper = sv_dup_inc(proto_perl->IL1PosixUpper, param);
13352 PL_PosixUpper = sv_dup_inc(proto_perl->IPosixUpper, param);
13354 PL_L1PosixWord = sv_dup_inc(proto_perl->IL1PosixWord, param);
13355 PL_PosixWord = sv_dup_inc(proto_perl->IPosixWord, param);
13357 PL_PosixXDigit = sv_dup_inc(proto_perl->IPosixXDigit, param);
13358 PL_XPosixXDigit = sv_dup_inc(proto_perl->IXPosixXDigit, param);
13360 PL_VertSpace = sv_dup_inc(proto_perl->IVertSpace, param);
13362 PL_NonL1NonFinalFold = sv_dup_inc(proto_perl->INonL1NonFinalFold, param);
13364 /* utf8 character class swashes */
13365 PL_utf8_alnum = sv_dup_inc(proto_perl->Iutf8_alnum, param);
13366 PL_utf8_alpha = sv_dup_inc(proto_perl->Iutf8_alpha, param);
13367 PL_utf8_blank = sv_dup_inc(proto_perl->Iutf8_blank, param);
13368 PL_utf8_space = sv_dup_inc(proto_perl->Iutf8_space, param);
13369 PL_utf8_graph = sv_dup_inc(proto_perl->Iutf8_graph, param);
13370 PL_utf8_digit = sv_dup_inc(proto_perl->Iutf8_digit, param);
13371 PL_utf8_upper = sv_dup_inc(proto_perl->Iutf8_upper, param);
13372 PL_utf8_lower = sv_dup_inc(proto_perl->Iutf8_lower, param);
13373 PL_utf8_print = sv_dup_inc(proto_perl->Iutf8_print, param);
13374 PL_utf8_punct = sv_dup_inc(proto_perl->Iutf8_punct, param);
13375 PL_utf8_xdigit = sv_dup_inc(proto_perl->Iutf8_xdigit, param);
13376 PL_utf8_mark = sv_dup_inc(proto_perl->Iutf8_mark, param);
13377 PL_utf8_X_regular_begin = sv_dup_inc(proto_perl->Iutf8_X_regular_begin, param);
13378 PL_utf8_X_extend = sv_dup_inc(proto_perl->Iutf8_X_extend, param);
13379 PL_utf8_X_prepend = sv_dup_inc(proto_perl->Iutf8_X_prepend, param);
13380 PL_utf8_X_special_begin = sv_dup_inc(proto_perl->Iutf8_X_special_begin, param);
13381 PL_utf8_X_L = sv_dup_inc(proto_perl->Iutf8_X_L, param);
13382 /*not currently used: PL_utf8_X_LV = sv_dup_inc(proto_perl->Iutf8_X_LV, param);*/
13383 PL_utf8_X_LVT = sv_dup_inc(proto_perl->Iutf8_X_LVT, param);
13384 PL_utf8_X_RI = sv_dup_inc(proto_perl->Iutf8_X_RI, param);
13385 PL_utf8_X_T = sv_dup_inc(proto_perl->Iutf8_X_T, param);
13386 PL_utf8_X_V = sv_dup_inc(proto_perl->Iutf8_X_V, param);
13387 PL_utf8_X_LV_LVT_V = sv_dup_inc(proto_perl->Iutf8_X_LV_LVT_V, param);
13388 PL_utf8_toupper = sv_dup_inc(proto_perl->Iutf8_toupper, param);
13389 PL_utf8_totitle = sv_dup_inc(proto_perl->Iutf8_totitle, param);
13390 PL_utf8_tolower = sv_dup_inc(proto_perl->Iutf8_tolower, param);
13391 PL_utf8_tofold = sv_dup_inc(proto_perl->Iutf8_tofold, param);
13392 PL_utf8_idstart = sv_dup_inc(proto_perl->Iutf8_idstart, param);
13393 PL_utf8_xidstart = sv_dup_inc(proto_perl->Iutf8_xidstart, param);
13394 PL_utf8_perl_idstart = sv_dup_inc(proto_perl->Iutf8_perl_idstart, param);
13395 PL_utf8_idcont = sv_dup_inc(proto_perl->Iutf8_idcont, param);
13396 PL_utf8_xidcont = sv_dup_inc(proto_perl->Iutf8_xidcont, param);
13397 PL_utf8_foldable = sv_dup_inc(proto_perl->Iutf8_foldable, param);
13398 PL_utf8_quotemeta = sv_dup_inc(proto_perl->Iutf8_quotemeta, param);
13399 PL_ASCII = sv_dup_inc(proto_perl->IASCII, param);
13400 PL_AboveLatin1 = sv_dup_inc(proto_perl->IAboveLatin1, param);
13401 PL_Latin1 = sv_dup_inc(proto_perl->ILatin1, param);
13404 if (proto_perl->Ipsig_pend) {
13405 Newxz(PL_psig_pend, SIG_SIZE, int);
13408 PL_psig_pend = (int*)NULL;
13411 if (proto_perl->Ipsig_name) {
13412 Newx(PL_psig_name, 2 * SIG_SIZE, SV*);
13413 sv_dup_inc_multiple(proto_perl->Ipsig_name, PL_psig_name, 2 * SIG_SIZE,
13415 PL_psig_ptr = PL_psig_name + SIG_SIZE;
13418 PL_psig_ptr = (SV**)NULL;
13419 PL_psig_name = (SV**)NULL;
13422 if (flags & CLONEf_COPY_STACKS) {
13423 Newx(PL_tmps_stack, PL_tmps_max, SV*);
13424 sv_dup_inc_multiple(proto_perl->Itmps_stack, PL_tmps_stack,
13425 PL_tmps_ix+1, param);
13427 /* next PUSHMARK() sets *(PL_markstack_ptr+1) */
13428 i = proto_perl->Imarkstack_max - proto_perl->Imarkstack;
13429 Newxz(PL_markstack, i, I32);
13430 PL_markstack_max = PL_markstack + (proto_perl->Imarkstack_max
13431 - proto_perl->Imarkstack);
13432 PL_markstack_ptr = PL_markstack + (proto_perl->Imarkstack_ptr
13433 - proto_perl->Imarkstack);
13434 Copy(proto_perl->Imarkstack, PL_markstack,
13435 PL_markstack_ptr - PL_markstack + 1, I32);
13437 /* next push_scope()/ENTER sets PL_scopestack[PL_scopestack_ix]
13438 * NOTE: unlike the others! */
13439 Newxz(PL_scopestack, PL_scopestack_max, I32);
13440 Copy(proto_perl->Iscopestack, PL_scopestack, PL_scopestack_ix, I32);
13443 Newxz(PL_scopestack_name, PL_scopestack_max, const char *);
13444 Copy(proto_perl->Iscopestack_name, PL_scopestack_name, PL_scopestack_ix, const char *);
13446 /* NOTE: si_dup() looks at PL_markstack */
13447 PL_curstackinfo = si_dup(proto_perl->Icurstackinfo, param);
13449 /* PL_curstack = PL_curstackinfo->si_stack; */
13450 PL_curstack = av_dup(proto_perl->Icurstack, param);
13451 PL_mainstack = av_dup(proto_perl->Imainstack, param);
13453 /* next PUSHs() etc. set *(PL_stack_sp+1) */
13454 PL_stack_base = AvARRAY(PL_curstack);
13455 PL_stack_sp = PL_stack_base + (proto_perl->Istack_sp
13456 - proto_perl->Istack_base);
13457 PL_stack_max = PL_stack_base + AvMAX(PL_curstack);
13459 /*Newxz(PL_savestack, PL_savestack_max, ANY);*/
13460 PL_savestack = ss_dup(proto_perl, param);
13464 ENTER; /* perl_destruct() wants to LEAVE; */
13467 PL_statgv = gv_dup(proto_perl->Istatgv, param);
13468 PL_statname = sv_dup_inc(proto_perl->Istatname, param);
13470 PL_rs = sv_dup_inc(proto_perl->Irs, param);
13471 PL_last_in_gv = gv_dup(proto_perl->Ilast_in_gv, param);
13472 PL_defoutgv = gv_dup_inc(proto_perl->Idefoutgv, param);
13473 PL_toptarget = sv_dup_inc(proto_perl->Itoptarget, param);
13474 PL_bodytarget = sv_dup_inc(proto_perl->Ibodytarget, param);
13475 PL_formtarget = sv_dup(proto_perl->Iformtarget, param);
13477 PL_errors = sv_dup_inc(proto_perl->Ierrors, param);
13479 PL_sortcop = (OP*)any_dup(proto_perl->Isortcop, proto_perl);
13480 PL_sortstash = hv_dup(proto_perl->Isortstash, param);
13481 PL_firstgv = gv_dup(proto_perl->Ifirstgv, param);
13482 PL_secondgv = gv_dup(proto_perl->Isecondgv, param);
13484 PL_stashcache = newHV();
13486 PL_watchaddr = (char **) ptr_table_fetch(PL_ptr_table,
13487 proto_perl->Iwatchaddr);
13488 PL_watchok = PL_watchaddr ? * PL_watchaddr : NULL;
13489 if (PL_debug && PL_watchaddr) {
13490 PerlIO_printf(Perl_debug_log,
13491 "WATCHING: %"UVxf" cloned as %"UVxf" with value %"UVxf"\n",
13492 PTR2UV(proto_perl->Iwatchaddr), PTR2UV(PL_watchaddr),
13493 PTR2UV(PL_watchok));
13496 PL_registered_mros = hv_dup_inc(proto_perl->Iregistered_mros, param);
13497 PL_blockhooks = av_dup_inc(proto_perl->Iblockhooks, param);
13498 PL_utf8_foldclosures = hv_dup_inc(proto_perl->Iutf8_foldclosures, param);
13500 /* Call the ->CLONE method, if it exists, for each of the stashes
13501 identified by sv_dup() above.
13503 while(av_len(param->stashes) != -1) {
13504 HV* const stash = MUTABLE_HV(av_shift(param->stashes));
13505 GV* const cloner = gv_fetchmethod_autoload(stash, "CLONE", 0);
13506 if (cloner && GvCV(cloner)) {
13511 mXPUSHs(newSVhek(HvNAME_HEK(stash)));
13513 call_sv(MUTABLE_SV(GvCV(cloner)), G_DISCARD);
13519 if (!(flags & CLONEf_KEEP_PTR_TABLE)) {
13520 ptr_table_free(PL_ptr_table);
13521 PL_ptr_table = NULL;
13524 if (!(flags & CLONEf_COPY_STACKS)) {
13525 unreferenced_to_tmp_stack(param->unreferenced);
13528 SvREFCNT_dec(param->stashes);
13530 /* orphaned? eg threads->new inside BEGIN or use */
13531 if (PL_compcv && ! SvREFCNT(PL_compcv)) {
13532 SvREFCNT_inc_simple_void(PL_compcv);
13533 SAVEFREESV(PL_compcv);
13540 S_unreferenced_to_tmp_stack(pTHX_ AV *const unreferenced)
13542 PERL_ARGS_ASSERT_UNREFERENCED_TO_TMP_STACK;
13544 if (AvFILLp(unreferenced) > -1) {
13545 SV **svp = AvARRAY(unreferenced);
13546 SV **const last = svp + AvFILLp(unreferenced);
13550 if (SvREFCNT(*svp) == 1)
13552 } while (++svp <= last);
13554 EXTEND_MORTAL(count);
13555 svp = AvARRAY(unreferenced);
13558 if (SvREFCNT(*svp) == 1) {
13559 /* Our reference is the only one to this SV. This means that
13560 in this thread, the scalar effectively has a 0 reference.
13561 That doesn't work (cleanup never happens), so donate our
13562 reference to it onto the save stack. */
13563 PL_tmps_stack[++PL_tmps_ix] = *svp;
13565 /* As an optimisation, because we are already walking the
13566 entire array, instead of above doing either
13567 SvREFCNT_inc(*svp) or *svp = &PL_sv_undef, we can instead
13568 release our reference to the scalar, so that at the end of
13569 the array owns zero references to the scalars it happens to
13570 point to. We are effectively converting the array from
13571 AvREAL() on to AvREAL() off. This saves the av_clear()
13572 (triggered by the SvREFCNT_dec(unreferenced) below) from
13573 walking the array a second time. */
13574 SvREFCNT_dec(*svp);
13577 } while (++svp <= last);
13578 AvREAL_off(unreferenced);
13580 SvREFCNT_dec(unreferenced);
13584 Perl_clone_params_del(CLONE_PARAMS *param)
13586 /* This seemingly funky ordering keeps the build with PERL_GLOBAL_STRUCT
13588 PerlInterpreter *const to = param->new_perl;
13590 PerlInterpreter *const was = PERL_GET_THX;
13592 PERL_ARGS_ASSERT_CLONE_PARAMS_DEL;
13598 SvREFCNT_dec(param->stashes);
13599 if (param->unreferenced)
13600 unreferenced_to_tmp_stack(param->unreferenced);
13610 Perl_clone_params_new(PerlInterpreter *const from, PerlInterpreter *const to)
13613 /* Need to play this game, as newAV() can call safesysmalloc(), and that
13614 does a dTHX; to get the context from thread local storage.
13615 FIXME - under PERL_CORE Newx(), Safefree() and friends should expand to
13616 a version that passes in my_perl. */
13617 PerlInterpreter *const was = PERL_GET_THX;
13618 CLONE_PARAMS *param;
13620 PERL_ARGS_ASSERT_CLONE_PARAMS_NEW;
13626 /* Given that we've set the context, we can do this unshared. */
13627 Newx(param, 1, CLONE_PARAMS);
13630 param->proto_perl = from;
13631 param->new_perl = to;
13632 param->stashes = (AV *)Perl_newSV_type(to, SVt_PVAV);
13633 AvREAL_off(param->stashes);
13634 param->unreferenced = (AV *)Perl_newSV_type(to, SVt_PVAV);
13642 #endif /* USE_ITHREADS */
13645 Perl_init_constants(pTHX)
13647 SvREFCNT(&PL_sv_undef) = (~(U32)0)/2;
13648 SvFLAGS(&PL_sv_undef) = SVf_READONLY|SVt_NULL;
13649 SvANY(&PL_sv_undef) = NULL;
13651 SvANY(&PL_sv_no) = new_XPVNV();
13652 SvREFCNT(&PL_sv_no) = (~(U32)0)/2;
13653 SvFLAGS(&PL_sv_no) = SVt_PVNV|SVf_READONLY
13654 |SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
13657 SvANY(&PL_sv_yes) = new_XPVNV();
13658 SvREFCNT(&PL_sv_yes) = (~(U32)0)/2;
13659 SvFLAGS(&PL_sv_yes) = SVt_PVNV|SVf_READONLY
13660 |SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
13663 SvPV_set(&PL_sv_no, (char*)PL_No);
13664 SvCUR_set(&PL_sv_no, 0);
13665 SvLEN_set(&PL_sv_no, 0);
13666 SvIV_set(&PL_sv_no, 0);
13667 SvNV_set(&PL_sv_no, 0);
13669 SvPV_set(&PL_sv_yes, (char*)PL_Yes);
13670 SvCUR_set(&PL_sv_yes, 1);
13671 SvLEN_set(&PL_sv_yes, 0);
13672 SvIV_set(&PL_sv_yes, 1);
13673 SvNV_set(&PL_sv_yes, 1);
13677 =head1 Unicode Support
13679 =for apidoc sv_recode_to_utf8
13681 The encoding is assumed to be an Encode object, on entry the PV
13682 of the sv is assumed to be octets in that encoding, and the sv
13683 will be converted into Unicode (and UTF-8).
13685 If the sv already is UTF-8 (or if it is not POK), or if the encoding
13686 is not a reference, nothing is done to the sv. If the encoding is not
13687 an C<Encode::XS> Encoding object, bad things will happen.
13688 (See F<lib/encoding.pm> and L<Encode>.)
13690 The PV of the sv is returned.
13695 Perl_sv_recode_to_utf8(pTHX_ SV *sv, SV *encoding)
13699 PERL_ARGS_ASSERT_SV_RECODE_TO_UTF8;
13701 if (SvPOK(sv) && !SvUTF8(sv) && !IN_BYTES && SvROK(encoding)) {
13715 Passing sv_yes is wrong - it needs to be or'ed set of constants
13716 for Encode::XS, while UTf-8 decode (currently) assumes a true value means
13717 remove converted chars from source.
13719 Both will default the value - let them.
13721 XPUSHs(&PL_sv_yes);
13724 call_method("decode", G_SCALAR);
13728 s = SvPV_const(uni, len);
13729 if (s != SvPVX_const(sv)) {
13730 SvGROW(sv, len + 1);
13731 Move(s, SvPVX(sv), len + 1, char);
13732 SvCUR_set(sv, len);
13736 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
13737 /* clear pos and any utf8 cache */
13738 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
13741 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
13742 magic_setutf8(sv,mg); /* clear UTF8 cache */
13747 return SvPOKp(sv) ? SvPVX(sv) : NULL;
13751 =for apidoc sv_cat_decode
13753 The encoding is assumed to be an Encode object, the PV of the ssv is
13754 assumed to be octets in that encoding and decoding the input starts
13755 from the position which (PV + *offset) pointed to. The dsv will be
13756 concatenated the decoded UTF-8 string from ssv. Decoding will terminate
13757 when the string tstr appears in decoding output or the input ends on
13758 the PV of the ssv. The value which the offset points will be modified
13759 to the last input position on the ssv.
13761 Returns TRUE if the terminator was found, else returns FALSE.
13766 Perl_sv_cat_decode(pTHX_ SV *dsv, SV *encoding,
13767 SV *ssv, int *offset, char *tstr, int tlen)
13772 PERL_ARGS_ASSERT_SV_CAT_DECODE;
13774 if (SvPOK(ssv) && SvPOK(dsv) && SvROK(encoding) && offset) {
13785 offsv = newSViv(*offset);
13787 mXPUSHp(tstr, tlen);
13789 call_method("cat_decode", G_SCALAR);
13791 ret = SvTRUE(TOPs);
13792 *offset = SvIV(offsv);
13798 Perl_croak(aTHX_ "Invalid argument to sv_cat_decode");
13803 /* ---------------------------------------------------------------------
13805 * support functions for report_uninit()
13808 /* the maxiumum size of array or hash where we will scan looking
13809 * for the undefined element that triggered the warning */
13811 #define FUV_MAX_SEARCH_SIZE 1000
13813 /* Look for an entry in the hash whose value has the same SV as val;
13814 * If so, return a mortal copy of the key. */
13817 S_find_hash_subscript(pTHX_ const HV *const hv, const SV *const val)
13823 PERL_ARGS_ASSERT_FIND_HASH_SUBSCRIPT;
13825 if (!hv || SvMAGICAL(hv) || !HvARRAY(hv) ||
13826 (HvTOTALKEYS(hv) > FUV_MAX_SEARCH_SIZE))
13829 array = HvARRAY(hv);
13831 for (i=HvMAX(hv); i>0; i--) {
13833 for (entry = array[i]; entry; entry = HeNEXT(entry)) {
13834 if (HeVAL(entry) != val)
13836 if ( HeVAL(entry) == &PL_sv_undef ||
13837 HeVAL(entry) == &PL_sv_placeholder)
13841 if (HeKLEN(entry) == HEf_SVKEY)
13842 return sv_mortalcopy(HeKEY_sv(entry));
13843 return sv_2mortal(newSVhek(HeKEY_hek(entry)));
13849 /* Look for an entry in the array whose value has the same SV as val;
13850 * If so, return the index, otherwise return -1. */
13853 S_find_array_subscript(pTHX_ const AV *const av, const SV *const val)
13857 PERL_ARGS_ASSERT_FIND_ARRAY_SUBSCRIPT;
13859 if (!av || SvMAGICAL(av) || !AvARRAY(av) ||
13860 (AvFILLp(av) > FUV_MAX_SEARCH_SIZE))
13863 if (val != &PL_sv_undef) {
13864 SV ** const svp = AvARRAY(av);
13867 for (i=AvFILLp(av); i>=0; i--)
13874 /* varname(): return the name of a variable, optionally with a subscript.
13875 * If gv is non-zero, use the name of that global, along with gvtype (one
13876 * of "$", "@", "%"); otherwise use the name of the lexical at pad offset
13877 * targ. Depending on the value of the subscript_type flag, return:
13880 #define FUV_SUBSCRIPT_NONE 1 /* "@foo" */
13881 #define FUV_SUBSCRIPT_ARRAY 2 /* "$foo[aindex]" */
13882 #define FUV_SUBSCRIPT_HASH 3 /* "$foo{keyname}" */
13883 #define FUV_SUBSCRIPT_WITHIN 4 /* "within @foo" */
13886 Perl_varname(pTHX_ const GV *const gv, const char gvtype, PADOFFSET targ,
13887 const SV *const keyname, I32 aindex, int subscript_type)
13890 SV * const name = sv_newmortal();
13891 if (gv && isGV(gv)) {
13893 buffer[0] = gvtype;
13896 /* as gv_fullname4(), but add literal '^' for $^FOO names */
13898 gv_fullname4(name, gv, buffer, 0);
13900 if ((unsigned int)SvPVX(name)[1] <= 26) {
13902 buffer[1] = SvPVX(name)[1] + 'A' - 1;
13904 /* Swap the 1 unprintable control character for the 2 byte pretty
13905 version - ie substr($name, 1, 1) = $buffer; */
13906 sv_insert(name, 1, 1, buffer, 2);
13910 CV * const cv = gv ? ((CV *)gv) : find_runcv(NULL);
13914 assert(!cv || SvTYPE(cv) == SVt_PVCV || SvTYPE(cv) == SVt_PVFM);
13916 if (!cv || !CvPADLIST(cv))
13918 av = *PadlistARRAY(CvPADLIST(cv));
13919 sv = *av_fetch(av, targ, FALSE);
13920 sv_setsv(name, sv);
13923 if (subscript_type == FUV_SUBSCRIPT_HASH) {
13924 SV * const sv = newSV(0);
13925 *SvPVX(name) = '$';
13926 Perl_sv_catpvf(aTHX_ name, "{%s}",
13927 pv_pretty(sv, SvPVX_const(keyname), SvCUR(keyname), 32, NULL, NULL,
13928 PERL_PV_PRETTY_DUMP | PERL_PV_ESCAPE_UNI_DETECT ));
13931 else if (subscript_type == FUV_SUBSCRIPT_ARRAY) {
13932 *SvPVX(name) = '$';
13933 Perl_sv_catpvf(aTHX_ name, "[%"IVdf"]", (IV)aindex);
13935 else if (subscript_type == FUV_SUBSCRIPT_WITHIN) {
13936 /* We know that name has no magic, so can use 0 instead of SV_GMAGIC */
13937 Perl_sv_insert_flags(aTHX_ name, 0, 0, STR_WITH_LEN("within "), 0);
13945 =for apidoc find_uninit_var
13947 Find the name of the undefined variable (if any) that caused the operator
13948 to issue a "Use of uninitialized value" warning.
13949 If match is true, only return a name if its value matches uninit_sv.
13950 So roughly speaking, if a unary operator (such as OP_COS) generates a
13951 warning, then following the direct child of the op may yield an
13952 OP_PADSV or OP_GV that gives the name of the undefined variable. On the
13953 other hand, with OP_ADD there are two branches to follow, so we only print
13954 the variable name if we get an exact match.
13956 The name is returned as a mortal SV.
13958 Assumes that PL_op is the op that originally triggered the error, and that
13959 PL_comppad/PL_curpad points to the currently executing pad.
13965 S_find_uninit_var(pTHX_ const OP *const obase, const SV *const uninit_sv,
13971 const OP *o, *o2, *kid;
13973 if (!obase || (match && (!uninit_sv || uninit_sv == &PL_sv_undef ||
13974 uninit_sv == &PL_sv_placeholder)))
13977 switch (obase->op_type) {
13984 const bool pad = (obase->op_type == OP_PADAV || obase->op_type == OP_PADHV);
13985 const bool hash = (obase->op_type == OP_PADHV || obase->op_type == OP_RV2HV);
13988 int subscript_type = FUV_SUBSCRIPT_WITHIN;
13990 if (pad) { /* @lex, %lex */
13991 sv = PAD_SVl(obase->op_targ);
13995 if (cUNOPx(obase)->op_first->op_type == OP_GV) {
13996 /* @global, %global */
13997 gv = cGVOPx_gv(cUNOPx(obase)->op_first);
14000 sv = hash ? MUTABLE_SV(GvHV(gv)): MUTABLE_SV(GvAV(gv));
14002 else if (obase == PL_op) /* @{expr}, %{expr} */
14003 return find_uninit_var(cUNOPx(obase)->op_first,
14005 else /* @{expr}, %{expr} as a sub-expression */
14009 /* attempt to find a match within the aggregate */
14011 keysv = find_hash_subscript((const HV*)sv, uninit_sv);
14013 subscript_type = FUV_SUBSCRIPT_HASH;
14016 index = find_array_subscript((const AV *)sv, uninit_sv);
14018 subscript_type = FUV_SUBSCRIPT_ARRAY;
14021 if (match && subscript_type == FUV_SUBSCRIPT_WITHIN)
14024 return varname(gv, hash ? '%' : '@', obase->op_targ,
14025 keysv, index, subscript_type);
14029 if (cUNOPx(obase)->op_first->op_type == OP_GV) {
14031 gv = cGVOPx_gv(cUNOPx(obase)->op_first);
14032 if (!gv || !GvSTASH(gv))
14034 if (match && (GvSV(gv) != uninit_sv))
14036 return varname(gv, '$', 0, NULL, 0, FUV_SUBSCRIPT_NONE);
14039 return find_uninit_var(cUNOPx(obase)->op_first, uninit_sv, 1);
14042 if (match && PAD_SVl(obase->op_targ) != uninit_sv)
14044 return varname(NULL, '$', obase->op_targ,
14045 NULL, 0, FUV_SUBSCRIPT_NONE);
14048 gv = cGVOPx_gv(obase);
14049 if (!gv || (match && GvSV(gv) != uninit_sv) || !GvSTASH(gv))
14051 return varname(gv, '$', 0, NULL, 0, FUV_SUBSCRIPT_NONE);
14053 case OP_AELEMFAST_LEX:
14056 AV *av = MUTABLE_AV(PAD_SV(obase->op_targ));
14057 if (!av || SvRMAGICAL(av))
14059 svp = av_fetch(av, (I32)obase->op_private, FALSE);
14060 if (!svp || *svp != uninit_sv)
14063 return varname(NULL, '$', obase->op_targ,
14064 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
14067 gv = cGVOPx_gv(obase);
14072 AV *const av = GvAV(gv);
14073 if (!av || SvRMAGICAL(av))
14075 svp = av_fetch(av, (I32)obase->op_private, FALSE);
14076 if (!svp || *svp != uninit_sv)
14079 return varname(gv, '$', 0,
14080 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
14085 o = cUNOPx(obase)->op_first;
14086 if (!o || o->op_type != OP_NULL ||
14087 ! (o->op_targ == OP_AELEM || o->op_targ == OP_HELEM))
14089 return find_uninit_var(cBINOPo->op_last, uninit_sv, match);
14094 bool negate = FALSE;
14096 if (PL_op == obase)
14097 /* $a[uninit_expr] or $h{uninit_expr} */
14098 return find_uninit_var(cBINOPx(obase)->op_last, uninit_sv, match);
14101 o = cBINOPx(obase)->op_first;
14102 kid = cBINOPx(obase)->op_last;
14104 /* get the av or hv, and optionally the gv */
14106 if (o->op_type == OP_PADAV || o->op_type == OP_PADHV) {
14107 sv = PAD_SV(o->op_targ);
14109 else if ((o->op_type == OP_RV2AV || o->op_type == OP_RV2HV)
14110 && cUNOPo->op_first->op_type == OP_GV)
14112 gv = cGVOPx_gv(cUNOPo->op_first);
14116 == OP_RV2HV ? MUTABLE_SV(GvHV(gv)) : MUTABLE_SV(GvAV(gv));
14121 if (kid && kid->op_type == OP_NEGATE) {
14123 kid = cUNOPx(kid)->op_first;
14126 if (kid && kid->op_type == OP_CONST && SvOK(cSVOPx_sv(kid))) {
14127 /* index is constant */
14130 kidsv = sv_2mortal(newSVpvs("-"));
14131 sv_catsv(kidsv, cSVOPx_sv(kid));
14134 kidsv = cSVOPx_sv(kid);
14138 if (obase->op_type == OP_HELEM) {
14139 HE* he = hv_fetch_ent(MUTABLE_HV(sv), kidsv, 0, 0);
14140 if (!he || HeVAL(he) != uninit_sv)
14144 SV * const * const svp = av_fetch(MUTABLE_AV(sv),
14145 negate ? - SvIV(cSVOPx_sv(kid)) : SvIV(cSVOPx_sv(kid)),
14147 if (!svp || *svp != uninit_sv)
14151 if (obase->op_type == OP_HELEM)
14152 return varname(gv, '%', o->op_targ,
14153 kidsv, 0, FUV_SUBSCRIPT_HASH);
14155 return varname(gv, '@', o->op_targ, NULL,
14156 negate ? - SvIV(cSVOPx_sv(kid)) : SvIV(cSVOPx_sv(kid)),
14157 FUV_SUBSCRIPT_ARRAY);
14160 /* index is an expression;
14161 * attempt to find a match within the aggregate */
14162 if (obase->op_type == OP_HELEM) {
14163 SV * const keysv = find_hash_subscript((const HV*)sv, uninit_sv);
14165 return varname(gv, '%', o->op_targ,
14166 keysv, 0, FUV_SUBSCRIPT_HASH);
14170 = find_array_subscript((const AV *)sv, uninit_sv);
14172 return varname(gv, '@', o->op_targ,
14173 NULL, index, FUV_SUBSCRIPT_ARRAY);
14178 (o->op_type == OP_PADAV || o->op_type == OP_RV2AV)
14180 o->op_targ, NULL, 0, FUV_SUBSCRIPT_WITHIN);
14186 /* only examine RHS */
14187 return find_uninit_var(cBINOPx(obase)->op_first, uninit_sv, match);
14190 o = cUNOPx(obase)->op_first;
14191 if (o->op_type == OP_PUSHMARK)
14194 if (!o->op_sibling) {
14195 /* one-arg version of open is highly magical */
14197 if (o->op_type == OP_GV) { /* open FOO; */
14199 if (match && GvSV(gv) != uninit_sv)
14201 return varname(gv, '$', 0,
14202 NULL, 0, FUV_SUBSCRIPT_NONE);
14204 /* other possibilities not handled are:
14205 * open $x; or open my $x; should return '${*$x}'
14206 * open expr; should return '$'.expr ideally
14212 /* ops where $_ may be an implicit arg */
14217 if ( !(obase->op_flags & OPf_STACKED)) {
14218 if (uninit_sv == ((obase->op_private & OPpTARGET_MY)
14219 ? PAD_SVl(obase->op_targ)
14222 sv = sv_newmortal();
14223 sv_setpvs(sv, "$_");
14232 match = 1; /* print etc can return undef on defined args */
14233 /* skip filehandle as it can't produce 'undef' warning */
14234 o = cUNOPx(obase)->op_first;
14235 if ((obase->op_flags & OPf_STACKED) && o->op_type == OP_PUSHMARK)
14236 o = o->op_sibling->op_sibling;
14240 case OP_ENTEREVAL: /* could be eval $undef or $x='$undef'; eval $x */
14241 case OP_CUSTOM: /* XS or custom code could trigger random warnings */
14243 /* the following ops are capable of returning PL_sv_undef even for
14244 * defined arg(s) */
14263 case OP_GETPEERNAME:
14311 case OP_SMARTMATCH:
14320 /* XXX tmp hack: these two may call an XS sub, and currently
14321 XS subs don't have a SUB entry on the context stack, so CV and
14322 pad determination goes wrong, and BAD things happen. So, just
14323 don't try to determine the value under those circumstances.
14324 Need a better fix at dome point. DAPM 11/2007 */
14330 GV * const gv = gv_fetchpvs(".", GV_NOTQUAL, SVt_PV);
14331 if (gv && GvSV(gv) == uninit_sv)
14332 return newSVpvs_flags("$.", SVs_TEMP);
14337 /* def-ness of rval pos() is independent of the def-ness of its arg */
14338 if ( !(obase->op_flags & OPf_MOD))
14343 if (SvROK(PL_rs) && uninit_sv == SvRV(PL_rs))
14344 return newSVpvs_flags("${$/}", SVs_TEMP);
14349 if (!(obase->op_flags & OPf_KIDS))
14351 o = cUNOPx(obase)->op_first;
14357 /* This loop checks all the kid ops, skipping any that cannot pos-
14358 * sibly be responsible for the uninitialized value; i.e., defined
14359 * constants and ops that return nothing. If there is only one op
14360 * left that is not skipped, then we *know* it is responsible for
14361 * the uninitialized value. If there is more than one op left, we
14362 * have to look for an exact match in the while() loop below.
14365 for (kid=o; kid; kid = kid->op_sibling) {
14367 const OPCODE type = kid->op_type;
14368 if ( (type == OP_CONST && SvOK(cSVOPx_sv(kid)))
14369 || (type == OP_NULL && ! (kid->op_flags & OPf_KIDS))
14370 || (type == OP_PUSHMARK)
14374 if (o2) { /* more than one found */
14381 return find_uninit_var(o2, uninit_sv, match);
14383 /* scan all args */
14385 sv = find_uninit_var(o, uninit_sv, 1);
14397 =for apidoc report_uninit
14399 Print appropriate "Use of uninitialized variable" warning.
14405 Perl_report_uninit(pTHX_ const SV *uninit_sv)
14409 SV* varname = NULL;
14410 if (uninit_sv && PL_curpad) {
14411 varname = find_uninit_var(PL_op, uninit_sv,0);
14413 sv_insert(varname, 0, 0, " ", 1);
14415 /* diag_listed_as: Use of uninitialized value%s */
14416 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit_sv,
14417 SVfARG(varname ? varname : &PL_sv_no),
14418 " in ", OP_DESC(PL_op));
14421 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
14427 * c-indentation-style: bsd
14428 * c-basic-offset: 4
14429 * indent-tabs-mode: nil
14432 * ex: set ts=8 sts=4 sw=4 et: