3 * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, 2000,
4 * 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009 by Larry Wall
7 * You may distribute under the terms of either the GNU General Public
8 * License or the Artistic License, as specified in the README file.
13 * 'I wonder what the Entish is for "yes" and "no",' he thought.
16 * [p.480 of _The Lord of the Rings_, III/iv: "Treebeard"]
22 * This file contains the code that creates, manipulates and destroys
23 * scalar values (SVs). The other types (AV, HV, GV, etc.) reuse the
24 * structure of an SV, so their creation and destruction is handled
25 * here; higher-level functions are in av.c, hv.c, and so on. Opcode
26 * level functions (eg. substr, split, join) for each of the types are
36 # if __STDC_VERSION__ >= 199901L && !defined(VMS)
47 /* Missing proto on LynxOS */
48 char *gconvert(double, int, int, char *);
51 #ifdef PERL_UTF8_CACHE_ASSERT
52 /* if adding more checks watch out for the following tests:
53 * t/op/index.t t/op/length.t t/op/pat.t t/op/substr.t
54 * lib/utf8.t lib/Unicode/Collate/t/index.t
57 # define ASSERT_UTF8_CACHE(cache) \
58 STMT_START { if (cache) { assert((cache)[0] <= (cache)[1]); \
59 assert((cache)[2] <= (cache)[3]); \
60 assert((cache)[3] <= (cache)[1]);} \
63 # define ASSERT_UTF8_CACHE(cache) NOOP
66 #ifdef PERL_OLD_COPY_ON_WRITE
67 #define SV_COW_NEXT_SV(sv) INT2PTR(SV *,SvUVX(sv))
68 #define SV_COW_NEXT_SV_SET(current,next) SvUV_set(current, PTR2UV(next))
69 /* This is a pessimistic view. Scalar must be purely a read-write PV to copy-
73 /* ============================================================================
75 =head1 Allocation and deallocation of SVs.
77 An SV (or AV, HV, etc.) is allocated in two parts: the head (struct
78 sv, av, hv...) contains type and reference count information, and for
79 many types, a pointer to the body (struct xrv, xpv, xpviv...), which
80 contains fields specific to each type. Some types store all they need
81 in the head, so don't have a body.
83 In all but the most memory-paranoid configurations (ex: PURIFY), heads
84 and bodies are allocated out of arenas, which by default are
85 approximately 4K chunks of memory parcelled up into N heads or bodies.
86 Sv-bodies are allocated by their sv-type, guaranteeing size
87 consistency needed to allocate safely from arrays.
89 For SV-heads, the first slot in each arena is reserved, and holds a
90 link to the next arena, some flags, and a note of the number of slots.
91 Snaked through each arena chain is a linked list of free items; when
92 this becomes empty, an extra arena is allocated and divided up into N
93 items which are threaded into the free list.
95 SV-bodies are similar, but they use arena-sets by default, which
96 separate the link and info from the arena itself, and reclaim the 1st
97 slot in the arena. SV-bodies are further described later.
99 The following global variables are associated with arenas:
101 PL_sv_arenaroot pointer to list of SV arenas
102 PL_sv_root pointer to list of free SV structures
104 PL_body_arenas head of linked-list of body arenas
105 PL_body_roots[] array of pointers to list of free bodies of svtype
106 arrays are indexed by the svtype needed
108 A few special SV heads are not allocated from an arena, but are
109 instead directly created in the interpreter structure, eg PL_sv_undef.
110 The size of arenas can be changed from the default by setting
111 PERL_ARENA_SIZE appropriately at compile time.
113 The SV arena serves the secondary purpose of allowing still-live SVs
114 to be located and destroyed during final cleanup.
116 At the lowest level, the macros new_SV() and del_SV() grab and free
117 an SV head. (If debugging with -DD, del_SV() calls the function S_del_sv()
118 to return the SV to the free list with error checking.) new_SV() calls
119 more_sv() / sv_add_arena() to add an extra arena if the free list is empty.
120 SVs in the free list have their SvTYPE field set to all ones.
122 At the time of very final cleanup, sv_free_arenas() is called from
123 perl_destruct() to physically free all the arenas allocated since the
124 start of the interpreter.
126 The function visit() scans the SV arenas list, and calls a specified
127 function for each SV it finds which is still live - ie which has an SvTYPE
128 other than all 1's, and a non-zero SvREFCNT. visit() is used by the
129 following functions (specified as [function that calls visit()] / [function
130 called by visit() for each SV]):
132 sv_report_used() / do_report_used()
133 dump all remaining SVs (debugging aid)
135 sv_clean_objs() / do_clean_objs(),do_clean_named_objs(),
136 do_clean_named_io_objs()
137 Attempt to free all objects pointed to by RVs,
138 and try to do the same for all objects indirectly
139 referenced by typeglobs too. Called once from
140 perl_destruct(), prior to calling sv_clean_all()
143 sv_clean_all() / do_clean_all()
144 SvREFCNT_dec(sv) each remaining SV, possibly
145 triggering an sv_free(). It also sets the
146 SVf_BREAK flag on the SV to indicate that the
147 refcnt has been artificially lowered, and thus
148 stopping sv_free() from giving spurious warnings
149 about SVs which unexpectedly have a refcnt
150 of zero. called repeatedly from perl_destruct()
151 until there are no SVs left.
153 =head2 Arena allocator API Summary
155 Private API to rest of sv.c
159 new_XPVNV(), del_XPVGV(),
164 sv_report_used(), sv_clean_objs(), sv_clean_all(), sv_free_arenas()
168 * ========================================================================= */
171 * "A time to plant, and a time to uproot what was planted..."
175 # define MEM_LOG_NEW_SV(sv, file, line, func) \
176 Perl_mem_log_new_sv(sv, file, line, func)
177 # define MEM_LOG_DEL_SV(sv, file, line, func) \
178 Perl_mem_log_del_sv(sv, file, line, func)
180 # define MEM_LOG_NEW_SV(sv, file, line, func) NOOP
181 # define MEM_LOG_DEL_SV(sv, file, line, func) NOOP
184 #ifdef DEBUG_LEAKING_SCALARS
185 # define FREE_SV_DEBUG_FILE(sv) Safefree((sv)->sv_debug_file)
186 # define DEBUG_SV_SERIAL(sv) \
187 DEBUG_m(PerlIO_printf(Perl_debug_log, "0x%"UVxf": (%05ld) del_SV\n", \
188 PTR2UV(sv), (long)(sv)->sv_debug_serial))
190 # define FREE_SV_DEBUG_FILE(sv)
191 # define DEBUG_SV_SERIAL(sv) NOOP
195 # define SvARENA_CHAIN(sv) ((sv)->sv_u.svu_rv)
196 # define SvARENA_CHAIN_SET(sv,val) (sv)->sv_u.svu_rv = MUTABLE_SV((val))
197 /* Whilst I'd love to do this, it seems that things like to check on
199 # define POSION_SV_HEAD(sv) PoisonNew(sv, 1, struct STRUCT_SV)
201 # define POSION_SV_HEAD(sv) PoisonNew(&SvANY(sv), 1, void *), \
202 PoisonNew(&SvREFCNT(sv), 1, U32)
204 # define SvARENA_CHAIN(sv) SvANY(sv)
205 # define SvARENA_CHAIN_SET(sv,val) SvANY(sv) = (void *)(val)
206 # define POSION_SV_HEAD(sv)
209 /* Mark an SV head as unused, and add to free list.
211 * If SVf_BREAK is set, skip adding it to the free list, as this SV had
212 * its refcount artificially decremented during global destruction, so
213 * there may be dangling pointers to it. The last thing we want in that
214 * case is for it to be reused. */
216 #define plant_SV(p) \
218 const U32 old_flags = SvFLAGS(p); \
219 MEM_LOG_DEL_SV(p, __FILE__, __LINE__, FUNCTION__); \
220 DEBUG_SV_SERIAL(p); \
221 FREE_SV_DEBUG_FILE(p); \
223 SvFLAGS(p) = SVTYPEMASK; \
224 if (!(old_flags & SVf_BREAK)) { \
225 SvARENA_CHAIN_SET(p, PL_sv_root); \
231 #define uproot_SV(p) \
234 PL_sv_root = MUTABLE_SV(SvARENA_CHAIN(p)); \
239 /* make some more SVs by adding another arena */
246 char *chunk; /* must use New here to match call to */
247 Newx(chunk,PERL_ARENA_SIZE,char); /* Safefree() in sv_free_arenas() */
248 sv_add_arena(chunk, PERL_ARENA_SIZE, 0);
253 /* new_SV(): return a new, empty SV head */
255 #ifdef DEBUG_LEAKING_SCALARS
256 /* provide a real function for a debugger to play with */
258 S_new_SV(pTHX_ const char *file, int line, const char *func)
265 sv = S_more_sv(aTHX);
269 sv->sv_debug_optype = PL_op ? PL_op->op_type : 0;
270 sv->sv_debug_line = (U16) (PL_parser && PL_parser->copline != NOLINE
276 sv->sv_debug_inpad = 0;
277 sv->sv_debug_parent = NULL;
278 sv->sv_debug_file = PL_curcop ? savepv(CopFILE(PL_curcop)): NULL;
280 sv->sv_debug_serial = PL_sv_serial++;
282 MEM_LOG_NEW_SV(sv, file, line, func);
283 DEBUG_m(PerlIO_printf(Perl_debug_log, "0x%"UVxf": (%05ld) new_SV (from %s:%d [%s])\n",
284 PTR2UV(sv), (long)sv->sv_debug_serial, file, line, func));
288 # define new_SV(p) (p)=S_new_SV(aTHX_ __FILE__, __LINE__, FUNCTION__)
296 (p) = S_more_sv(aTHX); \
300 MEM_LOG_NEW_SV(p, __FILE__, __LINE__, FUNCTION__); \
305 /* del_SV(): return an empty SV head to the free list */
318 S_del_sv(pTHX_ SV *p)
322 PERL_ARGS_ASSERT_DEL_SV;
327 for (sva = PL_sv_arenaroot; sva; sva = MUTABLE_SV(SvANY(sva))) {
328 const SV * const sv = sva + 1;
329 const SV * const svend = &sva[SvREFCNT(sva)];
330 if (p >= sv && p < svend) {
336 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL),
337 "Attempt to free non-arena SV: 0x%"UVxf
338 pTHX__FORMAT, PTR2UV(p) pTHX__VALUE);
345 #else /* ! DEBUGGING */
347 #define del_SV(p) plant_SV(p)
349 #endif /* DEBUGGING */
353 =head1 SV Manipulation Functions
355 =for apidoc sv_add_arena
357 Given a chunk of memory, link it to the head of the list of arenas,
358 and split it into a list of free SVs.
364 S_sv_add_arena(pTHX_ char *const ptr, const U32 size, const U32 flags)
367 SV *const sva = MUTABLE_SV(ptr);
371 PERL_ARGS_ASSERT_SV_ADD_ARENA;
373 /* The first SV in an arena isn't an SV. */
374 SvANY(sva) = (void *) PL_sv_arenaroot; /* ptr to next arena */
375 SvREFCNT(sva) = size / sizeof(SV); /* number of SV slots */
376 SvFLAGS(sva) = flags; /* FAKE if not to be freed */
378 PL_sv_arenaroot = sva;
379 PL_sv_root = sva + 1;
381 svend = &sva[SvREFCNT(sva) - 1];
384 SvARENA_CHAIN_SET(sv, (sv + 1));
388 /* Must always set typemask because it's always checked in on cleanup
389 when the arenas are walked looking for objects. */
390 SvFLAGS(sv) = SVTYPEMASK;
393 SvARENA_CHAIN_SET(sv, 0);
397 SvFLAGS(sv) = SVTYPEMASK;
400 /* visit(): call the named function for each non-free SV in the arenas
401 * whose flags field matches the flags/mask args. */
404 S_visit(pTHX_ SVFUNC_t f, const U32 flags, const U32 mask)
410 PERL_ARGS_ASSERT_VISIT;
412 for (sva = PL_sv_arenaroot; sva; sva = MUTABLE_SV(SvANY(sva))) {
413 register const SV * const svend = &sva[SvREFCNT(sva)];
415 for (sv = sva + 1; sv < svend; ++sv) {
416 if (SvTYPE(sv) != (svtype)SVTYPEMASK
417 && (sv->sv_flags & mask) == flags
430 /* called by sv_report_used() for each live SV */
433 do_report_used(pTHX_ SV *const sv)
435 if (SvTYPE(sv) != (svtype)SVTYPEMASK) {
436 PerlIO_printf(Perl_debug_log, "****\n");
443 =for apidoc sv_report_used
445 Dump the contents of all SVs not yet freed (debugging aid).
451 Perl_sv_report_used(pTHX)
454 visit(do_report_used, 0, 0);
460 /* called by sv_clean_objs() for each live SV */
463 do_clean_objs(pTHX_ SV *const ref)
468 SV * const target = SvRV(ref);
469 if (SvOBJECT(target)) {
470 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning object ref:\n "), sv_dump(ref)));
471 if (SvWEAKREF(ref)) {
472 sv_del_backref(target, ref);
478 SvREFCNT_dec(target);
483 /* XXX Might want to check arrays, etc. */
487 /* clear any slots in a GV which hold objects - except IO;
488 * called by sv_clean_objs() for each live GV */
491 do_clean_named_objs(pTHX_ SV *const sv)
495 assert(SvTYPE(sv) == SVt_PVGV);
496 assert(isGV_with_GP(sv));
500 /* freeing GP entries may indirectly free the current GV;
501 * hold onto it while we mess with the GP slots */
504 if ( ((obj = GvSV(sv) )) && SvOBJECT(obj)) {
505 DEBUG_D((PerlIO_printf(Perl_debug_log,
506 "Cleaning named glob SV object:\n "), sv_dump(obj)));
510 if ( ((obj = MUTABLE_SV(GvAV(sv)) )) && SvOBJECT(obj)) {
511 DEBUG_D((PerlIO_printf(Perl_debug_log,
512 "Cleaning named glob AV object:\n "), sv_dump(obj)));
516 if ( ((obj = MUTABLE_SV(GvHV(sv)) )) && SvOBJECT(obj)) {
517 DEBUG_D((PerlIO_printf(Perl_debug_log,
518 "Cleaning named glob HV object:\n "), sv_dump(obj)));
522 if ( ((obj = MUTABLE_SV(GvCV(sv)) )) && SvOBJECT(obj)) {
523 DEBUG_D((PerlIO_printf(Perl_debug_log,
524 "Cleaning named glob CV object:\n "), sv_dump(obj)));
528 SvREFCNT_dec(sv); /* undo the inc above */
531 /* clear any IO slots in a GV which hold objects (except stderr, defout);
532 * called by sv_clean_objs() for each live GV */
535 do_clean_named_io_objs(pTHX_ SV *const sv)
539 assert(SvTYPE(sv) == SVt_PVGV);
540 assert(isGV_with_GP(sv));
541 if (!GvGP(sv) || sv == (SV*)PL_stderrgv || sv == (SV*)PL_defoutgv)
545 if ( ((obj = MUTABLE_SV(GvIO(sv)) )) && SvOBJECT(obj)) {
546 DEBUG_D((PerlIO_printf(Perl_debug_log,
547 "Cleaning named glob IO object:\n "), sv_dump(obj)));
551 SvREFCNT_dec(sv); /* undo the inc above */
554 /* Void wrapper to pass to visit() */
556 do_curse(pTHX_ SV * const sv) {
557 if ((PL_stderrgv && GvGP(PL_stderrgv) && (SV*)GvIO(PL_stderrgv) == sv)
558 || (PL_defoutgv && GvGP(PL_defoutgv) && (SV*)GvIO(PL_defoutgv) == sv))
564 =for apidoc sv_clean_objs
566 Attempt to destroy all objects not yet freed.
572 Perl_sv_clean_objs(pTHX)
576 PL_in_clean_objs = TRUE;
577 visit(do_clean_objs, SVf_ROK, SVf_ROK);
578 /* Some barnacles may yet remain, clinging to typeglobs.
579 * Run the non-IO destructors first: they may want to output
580 * error messages, close files etc */
581 visit(do_clean_named_objs, SVt_PVGV|SVpgv_GP, SVTYPEMASK|SVp_POK|SVpgv_GP);
582 visit(do_clean_named_io_objs, SVt_PVGV|SVpgv_GP, SVTYPEMASK|SVp_POK|SVpgv_GP);
583 /* And if there are some very tenacious barnacles clinging to arrays,
584 closures, or what have you.... */
585 visit(do_curse, SVs_OBJECT, SVs_OBJECT);
586 olddef = PL_defoutgv;
587 PL_defoutgv = NULL; /* disable skip of PL_defoutgv */
588 if (olddef && isGV_with_GP(olddef))
589 do_clean_named_io_objs(aTHX_ MUTABLE_SV(olddef));
590 olderr = PL_stderrgv;
591 PL_stderrgv = NULL; /* disable skip of PL_stderrgv */
592 if (olderr && isGV_with_GP(olderr))
593 do_clean_named_io_objs(aTHX_ MUTABLE_SV(olderr));
594 SvREFCNT_dec(olddef);
595 PL_in_clean_objs = FALSE;
598 /* called by sv_clean_all() for each live SV */
601 do_clean_all(pTHX_ SV *const sv)
604 if (sv == (const SV *) PL_fdpid || sv == (const SV *)PL_strtab) {
605 /* don't clean pid table and strtab */
608 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning loops: SV at 0x%"UVxf"\n", PTR2UV(sv)) ));
609 SvFLAGS(sv) |= SVf_BREAK;
614 =for apidoc sv_clean_all
616 Decrement the refcnt of each remaining SV, possibly triggering a
617 cleanup. This function may have to be called multiple times to free
618 SVs which are in complex self-referential hierarchies.
624 Perl_sv_clean_all(pTHX)
628 PL_in_clean_all = TRUE;
629 cleaned = visit(do_clean_all, 0,0);
634 ARENASETS: a meta-arena implementation which separates arena-info
635 into struct arena_set, which contains an array of struct
636 arena_descs, each holding info for a single arena. By separating
637 the meta-info from the arena, we recover the 1st slot, formerly
638 borrowed for list management. The arena_set is about the size of an
639 arena, avoiding the needless malloc overhead of a naive linked-list.
641 The cost is 1 arena-set malloc per ~320 arena-mallocs, + the unused
642 memory in the last arena-set (1/2 on average). In trade, we get
643 back the 1st slot in each arena (ie 1.7% of a CV-arena, less for
644 smaller types). The recovery of the wasted space allows use of
645 small arenas for large, rare body types, by changing array* fields
646 in body_details_by_type[] below.
649 char *arena; /* the raw storage, allocated aligned */
650 size_t size; /* its size ~4k typ */
651 svtype utype; /* bodytype stored in arena */
656 /* Get the maximum number of elements in set[] such that struct arena_set
657 will fit within PERL_ARENA_SIZE, which is probably just under 4K, and
658 therefore likely to be 1 aligned memory page. */
660 #define ARENAS_PER_SET ((PERL_ARENA_SIZE - sizeof(struct arena_set*) \
661 - 2 * sizeof(int)) / sizeof (struct arena_desc))
664 struct arena_set* next;
665 unsigned int set_size; /* ie ARENAS_PER_SET */
666 unsigned int curr; /* index of next available arena-desc */
667 struct arena_desc set[ARENAS_PER_SET];
671 =for apidoc sv_free_arenas
673 Deallocate the memory used by all arenas. Note that all the individual SV
674 heads and bodies within the arenas must already have been freed.
679 Perl_sv_free_arenas(pTHX)
686 /* Free arenas here, but be careful about fake ones. (We assume
687 contiguity of the fake ones with the corresponding real ones.) */
689 for (sva = PL_sv_arenaroot; sva; sva = svanext) {
690 svanext = MUTABLE_SV(SvANY(sva));
691 while (svanext && SvFAKE(svanext))
692 svanext = MUTABLE_SV(SvANY(svanext));
699 struct arena_set *aroot = (struct arena_set*) PL_body_arenas;
702 struct arena_set *current = aroot;
705 assert(aroot->set[i].arena);
706 Safefree(aroot->set[i].arena);
714 i = PERL_ARENA_ROOTS_SIZE;
716 PL_body_roots[i] = 0;
723 Here are mid-level routines that manage the allocation of bodies out
724 of the various arenas. There are 5 kinds of arenas:
726 1. SV-head arenas, which are discussed and handled above
727 2. regular body arenas
728 3. arenas for reduced-size bodies
731 Arena types 2 & 3 are chained by body-type off an array of
732 arena-root pointers, which is indexed by svtype. Some of the
733 larger/less used body types are malloced singly, since a large
734 unused block of them is wasteful. Also, several svtypes dont have
735 bodies; the data fits into the sv-head itself. The arena-root
736 pointer thus has a few unused root-pointers (which may be hijacked
737 later for arena types 4,5)
739 3 differs from 2 as an optimization; some body types have several
740 unused fields in the front of the structure (which are kept in-place
741 for consistency). These bodies can be allocated in smaller chunks,
742 because the leading fields arent accessed. Pointers to such bodies
743 are decremented to point at the unused 'ghost' memory, knowing that
744 the pointers are used with offsets to the real memory.
747 =head1 SV-Body Allocation
749 Allocation of SV-bodies is similar to SV-heads, differing as follows;
750 the allocation mechanism is used for many body types, so is somewhat
751 more complicated, it uses arena-sets, and has no need for still-live
754 At the outermost level, (new|del)_X*V macros return bodies of the
755 appropriate type. These macros call either (new|del)_body_type or
756 (new|del)_body_allocated macro pairs, depending on specifics of the
757 type. Most body types use the former pair, the latter pair is used to
758 allocate body types with "ghost fields".
760 "ghost fields" are fields that are unused in certain types, and
761 consequently don't need to actually exist. They are declared because
762 they're part of a "base type", which allows use of functions as
763 methods. The simplest examples are AVs and HVs, 2 aggregate types
764 which don't use the fields which support SCALAR semantics.
766 For these types, the arenas are carved up into appropriately sized
767 chunks, we thus avoid wasted memory for those unaccessed members.
768 When bodies are allocated, we adjust the pointer back in memory by the
769 size of the part not allocated, so it's as if we allocated the full
770 structure. (But things will all go boom if you write to the part that
771 is "not there", because you'll be overwriting the last members of the
772 preceding structure in memory.)
774 We calculate the correction using the STRUCT_OFFSET macro on the first
775 member present. If the allocated structure is smaller (no initial NV
776 actually allocated) then the net effect is to subtract the size of the NV
777 from the pointer, to return a new pointer as if an initial NV were actually
778 allocated. (We were using structures named *_allocated for this, but
779 this turned out to be a subtle bug, because a structure without an NV
780 could have a lower alignment constraint, but the compiler is allowed to
781 optimised accesses based on the alignment constraint of the actual pointer
782 to the full structure, for example, using a single 64 bit load instruction
783 because it "knows" that two adjacent 32 bit members will be 8-byte aligned.)
785 This is the same trick as was used for NV and IV bodies. Ironically it
786 doesn't need to be used for NV bodies any more, because NV is now at
787 the start of the structure. IV bodies don't need it either, because
788 they are no longer allocated.
790 In turn, the new_body_* allocators call S_new_body(), which invokes
791 new_body_inline macro, which takes a lock, and takes a body off the
792 linked list at PL_body_roots[sv_type], calling Perl_more_bodies() if
793 necessary to refresh an empty list. Then the lock is released, and
794 the body is returned.
796 Perl_more_bodies allocates a new arena, and carves it up into an array of N
797 bodies, which it strings into a linked list. It looks up arena-size
798 and body-size from the body_details table described below, thus
799 supporting the multiple body-types.
801 If PURIFY is defined, or PERL_ARENA_SIZE=0, arenas are not used, and
802 the (new|del)_X*V macros are mapped directly to malloc/free.
804 For each sv-type, struct body_details bodies_by_type[] carries
805 parameters which control these aspects of SV handling:
807 Arena_size determines whether arenas are used for this body type, and if
808 so, how big they are. PURIFY or PERL_ARENA_SIZE=0 set this field to
809 zero, forcing individual mallocs and frees.
811 Body_size determines how big a body is, and therefore how many fit into
812 each arena. Offset carries the body-pointer adjustment needed for
813 "ghost fields", and is used in *_allocated macros.
815 But its main purpose is to parameterize info needed in
816 Perl_sv_upgrade(). The info here dramatically simplifies the function
817 vs the implementation in 5.8.8, making it table-driven. All fields
818 are used for this, except for arena_size.
820 For the sv-types that have no bodies, arenas are not used, so those
821 PL_body_roots[sv_type] are unused, and can be overloaded. In
822 something of a special case, SVt_NULL is borrowed for HE arenas;
823 PL_body_roots[HE_SVSLOT=SVt_NULL] is filled by S_more_he, but the
824 bodies_by_type[SVt_NULL] slot is not used, as the table is not
829 struct body_details {
830 U8 body_size; /* Size to allocate */
831 U8 copy; /* Size of structure to copy (may be shorter) */
833 unsigned int type : 4; /* We have space for a sanity check. */
834 unsigned int cant_upgrade : 1; /* Cannot upgrade this type */
835 unsigned int zero_nv : 1; /* zero the NV when upgrading from this */
836 unsigned int arena : 1; /* Allocated from an arena */
837 size_t arena_size; /* Size of arena to allocate */
845 /* With -DPURFIY we allocate everything directly, and don't use arenas.
846 This seems a rather elegant way to simplify some of the code below. */
847 #define HASARENA FALSE
849 #define HASARENA TRUE
851 #define NOARENA FALSE
853 /* Size the arenas to exactly fit a given number of bodies. A count
854 of 0 fits the max number bodies into a PERL_ARENA_SIZE.block,
855 simplifying the default. If count > 0, the arena is sized to fit
856 only that many bodies, allowing arenas to be used for large, rare
857 bodies (XPVFM, XPVIO) without undue waste. The arena size is
858 limited by PERL_ARENA_SIZE, so we can safely oversize the
861 #define FIT_ARENA0(body_size) \
862 ((size_t)(PERL_ARENA_SIZE / body_size) * body_size)
863 #define FIT_ARENAn(count,body_size) \
864 ( count * body_size <= PERL_ARENA_SIZE) \
865 ? count * body_size \
866 : FIT_ARENA0 (body_size)
867 #define FIT_ARENA(count,body_size) \
869 ? FIT_ARENAn (count, body_size) \
870 : FIT_ARENA0 (body_size)
872 /* Calculate the length to copy. Specifically work out the length less any
873 final padding the compiler needed to add. See the comment in sv_upgrade
874 for why copying the padding proved to be a bug. */
876 #define copy_length(type, last_member) \
877 STRUCT_OFFSET(type, last_member) \
878 + sizeof (((type*)SvANY((const SV *)0))->last_member)
880 static const struct body_details bodies_by_type[] = {
881 /* HEs use this offset for their arena. */
882 { 0, 0, 0, SVt_NULL, FALSE, NONV, NOARENA, 0 },
884 /* The bind placeholder pretends to be an RV for now.
885 Also it's marked as "can't upgrade" to stop anyone using it before it's
887 { 0, 0, 0, SVt_BIND, TRUE, NONV, NOARENA, 0 },
889 /* IVs are in the head, so the allocation size is 0. */
891 sizeof(IV), /* This is used to copy out the IV body. */
892 STRUCT_OFFSET(XPVIV, xiv_iv), SVt_IV, FALSE, NONV,
893 NOARENA /* IVS don't need an arena */, 0
896 { sizeof(NV), sizeof(NV),
897 STRUCT_OFFSET(XPVNV, xnv_u),
898 SVt_NV, FALSE, HADNV, HASARENA, FIT_ARENA(0, sizeof(NV)) },
900 { sizeof(XPV) - STRUCT_OFFSET(XPV, xpv_cur),
901 copy_length(XPV, xpv_len) - STRUCT_OFFSET(XPV, xpv_cur),
902 + STRUCT_OFFSET(XPV, xpv_cur),
903 SVt_PV, FALSE, NONV, HASARENA,
904 FIT_ARENA(0, sizeof(XPV) - STRUCT_OFFSET(XPV, xpv_cur)) },
906 { sizeof(XPVIV) - STRUCT_OFFSET(XPV, xpv_cur),
907 copy_length(XPVIV, xiv_u) - STRUCT_OFFSET(XPV, xpv_cur),
908 + STRUCT_OFFSET(XPV, xpv_cur),
909 SVt_PVIV, FALSE, NONV, HASARENA,
910 FIT_ARENA(0, sizeof(XPVIV) - STRUCT_OFFSET(XPV, xpv_cur)) },
912 { sizeof(XPVNV) - STRUCT_OFFSET(XPV, xpv_cur),
913 copy_length(XPVNV, xnv_u) - STRUCT_OFFSET(XPV, xpv_cur),
914 + STRUCT_OFFSET(XPV, xpv_cur),
915 SVt_PVNV, FALSE, HADNV, HASARENA,
916 FIT_ARENA(0, sizeof(XPVNV) - STRUCT_OFFSET(XPV, xpv_cur)) },
918 { sizeof(XPVMG), copy_length(XPVMG, xnv_u), 0, SVt_PVMG, FALSE, HADNV,
919 HASARENA, FIT_ARENA(0, sizeof(XPVMG)) },
924 SVt_REGEXP, FALSE, NONV, HASARENA,
925 FIT_ARENA(0, sizeof(regexp))
928 { sizeof(XPVGV), sizeof(XPVGV), 0, SVt_PVGV, TRUE, HADNV,
929 HASARENA, FIT_ARENA(0, sizeof(XPVGV)) },
931 { sizeof(XPVLV), sizeof(XPVLV), 0, SVt_PVLV, TRUE, HADNV,
932 HASARENA, FIT_ARENA(0, sizeof(XPVLV)) },
935 copy_length(XPVAV, xav_alloc),
937 SVt_PVAV, TRUE, NONV, HASARENA,
938 FIT_ARENA(0, sizeof(XPVAV)) },
941 copy_length(XPVHV, xhv_max),
943 SVt_PVHV, TRUE, NONV, HASARENA,
944 FIT_ARENA(0, sizeof(XPVHV)) },
949 SVt_PVCV, TRUE, NONV, HASARENA,
950 FIT_ARENA(0, sizeof(XPVCV)) },
955 SVt_PVFM, TRUE, NONV, NOARENA,
956 FIT_ARENA(20, sizeof(XPVFM)) },
961 SVt_PVIO, TRUE, NONV, HASARENA,
962 FIT_ARENA(24, sizeof(XPVIO)) },
965 #define new_body_allocated(sv_type) \
966 (void *)((char *)S_new_body(aTHX_ sv_type) \
967 - bodies_by_type[sv_type].offset)
969 /* return a thing to the free list */
971 #define del_body(thing, root) \
973 void ** const thing_copy = (void **)thing; \
974 *thing_copy = *root; \
975 *root = (void*)thing_copy; \
980 #define new_XNV() safemalloc(sizeof(XPVNV))
981 #define new_XPVNV() safemalloc(sizeof(XPVNV))
982 #define new_XPVMG() safemalloc(sizeof(XPVMG))
984 #define del_XPVGV(p) safefree(p)
988 #define new_XNV() new_body_allocated(SVt_NV)
989 #define new_XPVNV() new_body_allocated(SVt_PVNV)
990 #define new_XPVMG() new_body_allocated(SVt_PVMG)
992 #define del_XPVGV(p) del_body(p + bodies_by_type[SVt_PVGV].offset, \
993 &PL_body_roots[SVt_PVGV])
997 /* no arena for you! */
999 #define new_NOARENA(details) \
1000 safemalloc((details)->body_size + (details)->offset)
1001 #define new_NOARENAZ(details) \
1002 safecalloc((details)->body_size + (details)->offset, 1)
1005 Perl_more_bodies (pTHX_ const svtype sv_type, const size_t body_size,
1006 const size_t arena_size)
1009 void ** const root = &PL_body_roots[sv_type];
1010 struct arena_desc *adesc;
1011 struct arena_set *aroot = (struct arena_set *) PL_body_arenas;
1015 const size_t good_arena_size = Perl_malloc_good_size(arena_size);
1016 #if defined(DEBUGGING) && !defined(PERL_GLOBAL_STRUCT_PRIVATE)
1017 static bool done_sanity_check;
1019 /* PERL_GLOBAL_STRUCT_PRIVATE cannot coexist with global
1020 * variables like done_sanity_check. */
1021 if (!done_sanity_check) {
1022 unsigned int i = SVt_LAST;
1024 done_sanity_check = TRUE;
1027 assert (bodies_by_type[i].type == i);
1033 /* may need new arena-set to hold new arena */
1034 if (!aroot || aroot->curr >= aroot->set_size) {
1035 struct arena_set *newroot;
1036 Newxz(newroot, 1, struct arena_set);
1037 newroot->set_size = ARENAS_PER_SET;
1038 newroot->next = aroot;
1040 PL_body_arenas = (void *) newroot;
1041 DEBUG_m(PerlIO_printf(Perl_debug_log, "new arenaset %p\n", (void*)aroot));
1044 /* ok, now have arena-set with at least 1 empty/available arena-desc */
1045 curr = aroot->curr++;
1046 adesc = &(aroot->set[curr]);
1047 assert(!adesc->arena);
1049 Newx(adesc->arena, good_arena_size, char);
1050 adesc->size = good_arena_size;
1051 adesc->utype = sv_type;
1052 DEBUG_m(PerlIO_printf(Perl_debug_log, "arena %d added: %p size %"UVuf"\n",
1053 curr, (void*)adesc->arena, (UV)good_arena_size));
1055 start = (char *) adesc->arena;
1057 /* Get the address of the byte after the end of the last body we can fit.
1058 Remember, this is integer division: */
1059 end = start + good_arena_size / body_size * body_size;
1061 /* computed count doesn't reflect the 1st slot reservation */
1062 #if defined(MYMALLOC) || defined(HAS_MALLOC_GOOD_SIZE)
1063 DEBUG_m(PerlIO_printf(Perl_debug_log,
1064 "arena %p end %p arena-size %d (from %d) type %d "
1066 (void*)start, (void*)end, (int)good_arena_size,
1067 (int)arena_size, sv_type, (int)body_size,
1068 (int)good_arena_size / (int)body_size));
1070 DEBUG_m(PerlIO_printf(Perl_debug_log,
1071 "arena %p end %p arena-size %d type %d size %d ct %d\n",
1072 (void*)start, (void*)end,
1073 (int)arena_size, sv_type, (int)body_size,
1074 (int)good_arena_size / (int)body_size));
1076 *root = (void *)start;
1079 /* Where the next body would start: */
1080 char * const next = start + body_size;
1083 /* This is the last body: */
1084 assert(next == end);
1086 *(void **)start = 0;
1090 *(void**) start = (void *)next;
1095 /* grab a new thing from the free list, allocating more if necessary.
1096 The inline version is used for speed in hot routines, and the
1097 function using it serves the rest (unless PURIFY).
1099 #define new_body_inline(xpv, sv_type) \
1101 void ** const r3wt = &PL_body_roots[sv_type]; \
1102 xpv = (PTR_TBL_ENT_t*) (*((void **)(r3wt)) \
1103 ? *((void **)(r3wt)) : Perl_more_bodies(aTHX_ sv_type, \
1104 bodies_by_type[sv_type].body_size,\
1105 bodies_by_type[sv_type].arena_size)); \
1106 *(r3wt) = *(void**)(xpv); \
1112 S_new_body(pTHX_ const svtype sv_type)
1116 new_body_inline(xpv, sv_type);
1122 static const struct body_details fake_rv =
1123 { 0, 0, 0, SVt_IV, FALSE, NONV, NOARENA, 0 };
1126 =for apidoc sv_upgrade
1128 Upgrade an SV to a more complex form. Generally adds a new body type to the
1129 SV, then copies across as much information as possible from the old body.
1130 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 these two if statements:
1614 u=1.49 s=0.52 cu=72.49 cs=10.64 scripts=270 tests=20865
1617 u=1.35 s=0.47 cu=73.45 cs=11.43 scripts=270 tests=20865
1619 If you wish to remove them, please benchmark to see what the effect is
1621 if (u <= (UV)IV_MAX) {
1622 sv_setiv(sv, (IV)u);
1631 =for apidoc sv_setuv_mg
1633 Like C<sv_setuv>, but also handles 'set' magic.
1639 Perl_sv_setuv_mg(pTHX_ register SV *const sv, const UV u)
1641 PERL_ARGS_ASSERT_SV_SETUV_MG;
1648 =for apidoc sv_setnv
1650 Copies a double into the given SV, upgrading first if necessary.
1651 Does not handle 'set' magic. See also C<sv_setnv_mg>.
1657 Perl_sv_setnv(pTHX_ register SV *const sv, const NV num)
1661 PERL_ARGS_ASSERT_SV_SETNV;
1663 SV_CHECK_THINKFIRST_COW_DROP(sv);
1664 switch (SvTYPE(sv)) {
1667 sv_upgrade(sv, SVt_NV);
1671 sv_upgrade(sv, SVt_PVNV);
1675 if (!isGV_with_GP(sv))
1682 /* diag_listed_as: Can't coerce %s to %s in %s */
1683 Perl_croak(aTHX_ "Can't coerce %s to number in %s", sv_reftype(sv,0),
1688 (void)SvNOK_only(sv); /* validate number */
1693 =for apidoc sv_setnv_mg
1695 Like C<sv_setnv>, but also handles 'set' magic.
1701 Perl_sv_setnv_mg(pTHX_ register SV *const sv, const NV num)
1703 PERL_ARGS_ASSERT_SV_SETNV_MG;
1709 /* Print an "isn't numeric" warning, using a cleaned-up,
1710 * printable version of the offending string
1714 S_not_a_number(pTHX_ SV *const sv)
1721 PERL_ARGS_ASSERT_NOT_A_NUMBER;
1724 dsv = newSVpvs_flags("", SVs_TEMP);
1725 pv = sv_uni_display(dsv, sv, 10, UNI_DISPLAY_ISPRINT);
1728 const char * const limit = tmpbuf + sizeof(tmpbuf) - 8;
1729 /* each *s can expand to 4 chars + "...\0",
1730 i.e. need room for 8 chars */
1732 const char *s = SvPVX_const(sv);
1733 const char * const end = s + SvCUR(sv);
1734 for ( ; s < end && d < limit; s++ ) {
1736 if (ch & 128 && !isPRINT_LC(ch)) {
1745 else if (ch == '\r') {
1749 else if (ch == '\f') {
1753 else if (ch == '\\') {
1757 else if (ch == '\0') {
1761 else if (isPRINT_LC(ch))
1778 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
1779 /* diag_listed_as: Argument "%s" isn't numeric%s */
1780 "Argument \"%s\" isn't numeric in %s", pv,
1783 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
1784 /* diag_listed_as: Argument "%s" isn't numeric%s */
1785 "Argument \"%s\" isn't numeric", pv);
1789 =for apidoc looks_like_number
1791 Test if the content of an SV looks like a number (or is a number).
1792 C<Inf> and C<Infinity> are treated as numbers (so will not issue a
1793 non-numeric warning), even if your atof() doesn't grok them. Get-magic is
1800 Perl_looks_like_number(pTHX_ SV *const sv)
1802 register const char *sbegin;
1805 PERL_ARGS_ASSERT_LOOKS_LIKE_NUMBER;
1807 if (SvPOK(sv) || SvPOKp(sv)) {
1808 sbegin = SvPV_nomg_const(sv, len);
1811 return SvFLAGS(sv) & (SVf_NOK|SVp_NOK|SVf_IOK|SVp_IOK);
1812 return grok_number(sbegin, len, NULL);
1816 S_glob_2number(pTHX_ GV * const gv)
1818 SV *const buffer = sv_newmortal();
1820 PERL_ARGS_ASSERT_GLOB_2NUMBER;
1822 gv_efullname3(buffer, gv, "*");
1824 /* We know that all GVs stringify to something that is not-a-number,
1825 so no need to test that. */
1826 if (ckWARN(WARN_NUMERIC))
1827 not_a_number(buffer);
1828 /* We just want something true to return, so that S_sv_2iuv_common
1829 can tail call us and return true. */
1833 /* Actually, ISO C leaves conversion of UV to IV undefined, but
1834 until proven guilty, assume that things are not that bad... */
1839 As 64 bit platforms often have an NV that doesn't preserve all bits of
1840 an IV (an assumption perl has been based on to date) it becomes necessary
1841 to remove the assumption that the NV always carries enough precision to
1842 recreate the IV whenever needed, and that the NV is the canonical form.
1843 Instead, IV/UV and NV need to be given equal rights. So as to not lose
1844 precision as a side effect of conversion (which would lead to insanity
1845 and the dragon(s) in t/op/numconvert.t getting very angry) the intent is
1846 1) to distinguish between IV/UV/NV slots that have cached a valid
1847 conversion where precision was lost and IV/UV/NV slots that have a
1848 valid conversion which has lost no precision
1849 2) to ensure that if a numeric conversion to one form is requested that
1850 would lose precision, the precise conversion (or differently
1851 imprecise conversion) is also performed and cached, to prevent
1852 requests for different numeric formats on the same SV causing
1853 lossy conversion chains. (lossless conversion chains are perfectly
1858 SvIOKp is true if the IV slot contains a valid value
1859 SvIOK is true only if the IV value is accurate (UV if SvIOK_UV true)
1860 SvNOKp is true if the NV slot contains a valid value
1861 SvNOK is true only if the NV value is accurate
1864 while converting from PV to NV, check to see if converting that NV to an
1865 IV(or UV) would lose accuracy over a direct conversion from PV to
1866 IV(or UV). If it would, cache both conversions, return NV, but mark
1867 SV as IOK NOKp (ie not NOK).
1869 While converting from PV to IV, check to see if converting that IV to an
1870 NV would lose accuracy over a direct conversion from PV to NV. If it
1871 would, cache both conversions, flag similarly.
1873 Before, the SV value "3.2" could become NV=3.2 IV=3 NOK, IOK quite
1874 correctly because if IV & NV were set NV *always* overruled.
1875 Now, "3.2" will become NV=3.2 IV=3 NOK, IOKp, because the flag's meaning
1876 changes - now IV and NV together means that the two are interchangeable:
1877 SvIVX == (IV) SvNVX && SvNVX == (NV) SvIVX;
1879 The benefit of this is that operations such as pp_add know that if
1880 SvIOK is true for both left and right operands, then integer addition
1881 can be used instead of floating point (for cases where the result won't
1882 overflow). Before, floating point was always used, which could lead to
1883 loss of precision compared with integer addition.
1885 * making IV and NV equal status should make maths accurate on 64 bit
1887 * may speed up maths somewhat if pp_add and friends start to use
1888 integers when possible instead of fp. (Hopefully the overhead in
1889 looking for SvIOK and checking for overflow will not outweigh the
1890 fp to integer speedup)
1891 * will slow down integer operations (callers of SvIV) on "inaccurate"
1892 values, as the change from SvIOK to SvIOKp will cause a call into
1893 sv_2iv each time rather than a macro access direct to the IV slot
1894 * should speed up number->string conversion on integers as IV is
1895 favoured when IV and NV are equally accurate
1897 ####################################################################
1898 You had better be using SvIOK_notUV if you want an IV for arithmetic:
1899 SvIOK is true if (IV or UV), so you might be getting (IV)SvUV.
1900 On the other hand, SvUOK is true iff UV.
1901 ####################################################################
1903 Your mileage will vary depending your CPU's relative fp to integer
1907 #ifndef NV_PRESERVES_UV
1908 # define IS_NUMBER_UNDERFLOW_IV 1
1909 # define IS_NUMBER_UNDERFLOW_UV 2
1910 # define IS_NUMBER_IV_AND_UV 2
1911 # define IS_NUMBER_OVERFLOW_IV 4
1912 # define IS_NUMBER_OVERFLOW_UV 5
1914 /* sv_2iuv_non_preserve(): private routine for use by sv_2iv() and sv_2uv() */
1916 /* For sv_2nv these three cases are "SvNOK and don't bother casting" */
1918 S_sv_2iuv_non_preserve(pTHX_ register SV *const sv
1926 PERL_ARGS_ASSERT_SV_2IUV_NON_PRESERVE;
1928 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));
1929 if (SvNVX(sv) < (NV)IV_MIN) {
1930 (void)SvIOKp_on(sv);
1932 SvIV_set(sv, IV_MIN);
1933 return IS_NUMBER_UNDERFLOW_IV;
1935 if (SvNVX(sv) > (NV)UV_MAX) {
1936 (void)SvIOKp_on(sv);
1939 SvUV_set(sv, UV_MAX);
1940 return IS_NUMBER_OVERFLOW_UV;
1942 (void)SvIOKp_on(sv);
1944 /* Can't use strtol etc to convert this string. (See truth table in
1946 if (SvNVX(sv) <= (UV)IV_MAX) {
1947 SvIV_set(sv, I_V(SvNVX(sv)));
1948 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
1949 SvIOK_on(sv); /* Integer is precise. NOK, IOK */
1951 /* Integer is imprecise. NOK, IOKp */
1953 return SvNVX(sv) < 0 ? IS_NUMBER_UNDERFLOW_UV : IS_NUMBER_IV_AND_UV;
1956 SvUV_set(sv, U_V(SvNVX(sv)));
1957 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
1958 if (SvUVX(sv) == UV_MAX) {
1959 /* As we know that NVs don't preserve UVs, UV_MAX cannot
1960 possibly be preserved by NV. Hence, it must be overflow.
1962 return IS_NUMBER_OVERFLOW_UV;
1964 SvIOK_on(sv); /* Integer is precise. NOK, UOK */
1966 /* Integer is imprecise. NOK, IOKp */
1968 return IS_NUMBER_OVERFLOW_IV;
1970 #endif /* !NV_PRESERVES_UV*/
1973 S_sv_2iuv_common(pTHX_ SV *const sv)
1977 PERL_ARGS_ASSERT_SV_2IUV_COMMON;
1980 /* erm. not sure. *should* never get NOKp (without NOK) from sv_2nv
1981 * without also getting a cached IV/UV from it at the same time
1982 * (ie PV->NV conversion should detect loss of accuracy and cache
1983 * IV or UV at same time to avoid this. */
1984 /* IV-over-UV optimisation - choose to cache IV if possible */
1986 if (SvTYPE(sv) == SVt_NV)
1987 sv_upgrade(sv, SVt_PVNV);
1989 (void)SvIOKp_on(sv); /* Must do this first, to clear any SvOOK */
1990 /* < not <= as for NV doesn't preserve UV, ((NV)IV_MAX+1) will almost
1991 certainly cast into the IV range at IV_MAX, whereas the correct
1992 answer is the UV IV_MAX +1. Hence < ensures that dodgy boundary
1994 #if defined(NAN_COMPARE_BROKEN) && defined(Perl_isnan)
1995 if (Perl_isnan(SvNVX(sv))) {
2001 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2002 SvIV_set(sv, I_V(SvNVX(sv)));
2003 if (SvNVX(sv) == (NV) SvIVX(sv)
2004 #ifndef NV_PRESERVES_UV
2005 && (((UV)1 << NV_PRESERVES_UV_BITS) >
2006 (UV)(SvIVX(sv) > 0 ? SvIVX(sv) : -SvIVX(sv)))
2007 /* Don't flag it as "accurately an integer" if the number
2008 came from a (by definition imprecise) NV operation, and
2009 we're outside the range of NV integer precision */
2013 SvIOK_on(sv); /* Can this go wrong with rounding? NWC */
2015 /* scalar has trailing garbage, eg "42a" */
2017 DEBUG_c(PerlIO_printf(Perl_debug_log,
2018 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (precise)\n",
2024 /* IV not precise. No need to convert from PV, as NV
2025 conversion would already have cached IV if it detected
2026 that PV->IV would be better than PV->NV->IV
2027 flags already correct - don't set public IOK. */
2028 DEBUG_c(PerlIO_printf(Perl_debug_log,
2029 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (imprecise)\n",
2034 /* Can the above go wrong if SvIVX == IV_MIN and SvNVX < IV_MIN,
2035 but the cast (NV)IV_MIN rounds to a the value less (more
2036 negative) than IV_MIN which happens to be equal to SvNVX ??
2037 Analogous to 0xFFFFFFFFFFFFFFFF rounding up to NV (2**64) and
2038 NV rounding back to 0xFFFFFFFFFFFFFFFF, so UVX == UV(NVX) and
2039 (NV)UVX == NVX are both true, but the values differ. :-(
2040 Hopefully for 2s complement IV_MIN is something like
2041 0x8000000000000000 which will be exact. NWC */
2044 SvUV_set(sv, U_V(SvNVX(sv)));
2046 (SvNVX(sv) == (NV) SvUVX(sv))
2047 #ifndef NV_PRESERVES_UV
2048 /* Make sure it's not 0xFFFFFFFFFFFFFFFF */
2049 /*&& (SvUVX(sv) != UV_MAX) irrelevant with code below */
2050 && (((UV)1 << NV_PRESERVES_UV_BITS) > SvUVX(sv))
2051 /* Don't flag it as "accurately an integer" if the number
2052 came from a (by definition imprecise) NV operation, and
2053 we're outside the range of NV integer precision */
2059 DEBUG_c(PerlIO_printf(Perl_debug_log,
2060 "0x%"UVxf" 2iv(%"UVuf" => %"IVdf") (as unsigned)\n",
2066 else if (SvPOKp(sv) && SvLEN(sv)) {
2068 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2069 /* We want to avoid a possible problem when we cache an IV/ a UV which
2070 may be later translated to an NV, and the resulting NV is not
2071 the same as the direct translation of the initial string
2072 (eg 123.456 can shortcut to the IV 123 with atol(), but we must
2073 be careful to ensure that the value with the .456 is around if the
2074 NV value is requested in the future).
2076 This means that if we cache such an IV/a UV, we need to cache the
2077 NV as well. Moreover, we trade speed for space, and do not
2078 cache the NV if we are sure it's not needed.
2081 /* SVt_PVNV is one higher than SVt_PVIV, hence this order */
2082 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2083 == IS_NUMBER_IN_UV) {
2084 /* It's definitely an integer, only upgrade to PVIV */
2085 if (SvTYPE(sv) < SVt_PVIV)
2086 sv_upgrade(sv, SVt_PVIV);
2088 } else if (SvTYPE(sv) < SVt_PVNV)
2089 sv_upgrade(sv, SVt_PVNV);
2091 /* If NVs preserve UVs then we only use the UV value if we know that
2092 we aren't going to call atof() below. If NVs don't preserve UVs
2093 then the value returned may have more precision than atof() will
2094 return, even though value isn't perfectly accurate. */
2095 if ((numtype & (IS_NUMBER_IN_UV
2096 #ifdef NV_PRESERVES_UV
2099 )) == IS_NUMBER_IN_UV) {
2100 /* This won't turn off the public IOK flag if it was set above */
2101 (void)SvIOKp_on(sv);
2103 if (!(numtype & IS_NUMBER_NEG)) {
2105 if (value <= (UV)IV_MAX) {
2106 SvIV_set(sv, (IV)value);
2108 /* it didn't overflow, and it was positive. */
2109 SvUV_set(sv, value);
2113 /* 2s complement assumption */
2114 if (value <= (UV)IV_MIN) {
2115 SvIV_set(sv, -(IV)value);
2117 /* Too negative for an IV. This is a double upgrade, but
2118 I'm assuming it will be rare. */
2119 if (SvTYPE(sv) < SVt_PVNV)
2120 sv_upgrade(sv, SVt_PVNV);
2124 SvNV_set(sv, -(NV)value);
2125 SvIV_set(sv, IV_MIN);
2129 /* For !NV_PRESERVES_UV and IS_NUMBER_IN_UV and IS_NUMBER_NOT_INT we
2130 will be in the previous block to set the IV slot, and the next
2131 block to set the NV slot. So no else here. */
2133 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2134 != IS_NUMBER_IN_UV) {
2135 /* It wasn't an (integer that doesn't overflow the UV). */
2136 SvNV_set(sv, Atof(SvPVX_const(sv)));
2138 if (! numtype && ckWARN(WARN_NUMERIC))
2141 #if defined(USE_LONG_DOUBLE)
2142 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%" PERL_PRIgldbl ")\n",
2143 PTR2UV(sv), SvNVX(sv)));
2145 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"NVgf")\n",
2146 PTR2UV(sv), SvNVX(sv)));
2149 #ifdef NV_PRESERVES_UV
2150 (void)SvIOKp_on(sv);
2152 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2153 SvIV_set(sv, I_V(SvNVX(sv)));
2154 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
2157 NOOP; /* Integer is imprecise. NOK, IOKp */
2159 /* UV will not work better than IV */
2161 if (SvNVX(sv) > (NV)UV_MAX) {
2163 /* Integer is inaccurate. NOK, IOKp, is UV */
2164 SvUV_set(sv, UV_MAX);
2166 SvUV_set(sv, U_V(SvNVX(sv)));
2167 /* 0xFFFFFFFFFFFFFFFF not an issue in here, NVs
2168 NV preservse UV so can do correct comparison. */
2169 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
2172 NOOP; /* Integer is imprecise. NOK, IOKp, is UV */
2177 #else /* NV_PRESERVES_UV */
2178 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2179 == (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT)) {
2180 /* The IV/UV slot will have been set from value returned by
2181 grok_number above. The NV slot has just been set using
2184 assert (SvIOKp(sv));
2186 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2187 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2188 /* Small enough to preserve all bits. */
2189 (void)SvIOKp_on(sv);
2191 SvIV_set(sv, I_V(SvNVX(sv)));
2192 if ((NV)(SvIVX(sv)) == SvNVX(sv))
2194 /* Assumption: first non-preserved integer is < IV_MAX,
2195 this NV is in the preserved range, therefore: */
2196 if (!(U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))
2198 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);
2202 0 0 already failed to read UV.
2203 0 1 already failed to read UV.
2204 1 0 you won't get here in this case. IV/UV
2205 slot set, public IOK, Atof() unneeded.
2206 1 1 already read UV.
2207 so there's no point in sv_2iuv_non_preserve() attempting
2208 to use atol, strtol, strtoul etc. */
2210 sv_2iuv_non_preserve (sv, numtype);
2212 sv_2iuv_non_preserve (sv);
2216 #endif /* NV_PRESERVES_UV */
2217 /* It might be more code efficient to go through the entire logic above
2218 and conditionally set with SvIOKp_on() rather than SvIOK(), but it
2219 gets complex and potentially buggy, so more programmer efficient
2220 to do it this way, by turning off the public flags: */
2222 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK);
2226 if (isGV_with_GP(sv))
2227 return glob_2number(MUTABLE_GV(sv));
2229 if (!SvPADTMP(sv)) {
2230 if (!PL_localizing && ckWARN(WARN_UNINITIALIZED))
2233 if (SvTYPE(sv) < SVt_IV)
2234 /* Typically the caller expects that sv_any is not NULL now. */
2235 sv_upgrade(sv, SVt_IV);
2236 /* Return 0 from the caller. */
2243 =for apidoc sv_2iv_flags
2245 Return the integer value of an SV, doing any necessary string
2246 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2247 Normally used via the C<SvIV(sv)> and C<SvIVx(sv)> macros.
2253 Perl_sv_2iv_flags(pTHX_ register SV *const sv, const I32 flags)
2258 if (SvGMAGICAL(sv) || SvVALID(sv)) {
2259 /* FBMs use the space for SvIVX and SvNVX for other purposes, and use
2260 the same flag bit as SVf_IVisUV, so must not let them cache IVs.
2261 In practice they are extremely unlikely to actually get anywhere
2262 accessible by user Perl code - the only way that I'm aware of is when
2263 a constant subroutine which is used as the second argument to index.
2265 if (flags & SV_GMAGIC)
2270 return I_V(SvNVX(sv));
2272 if (SvPOKp(sv) && SvLEN(sv)) {
2275 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2277 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2278 == IS_NUMBER_IN_UV) {
2279 /* It's definitely an integer */
2280 if (numtype & IS_NUMBER_NEG) {
2281 if (value < (UV)IV_MIN)
2284 if (value < (UV)IV_MAX)
2289 if (ckWARN(WARN_NUMERIC))
2292 return I_V(Atof(SvPVX_const(sv)));
2297 assert(SvTYPE(sv) >= SVt_PVMG);
2298 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
2299 } else if (SvTHINKFIRST(sv)) {
2304 if (flags & SV_SKIP_OVERLOAD)
2306 tmpstr = AMG_CALLunary(sv, numer_amg);
2307 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2308 return SvIV(tmpstr);
2311 return PTR2IV(SvRV(sv));
2314 sv_force_normal_flags(sv, 0);
2316 if (SvREADONLY(sv) && !SvOK(sv)) {
2317 if (ckWARN(WARN_UNINITIALIZED))
2323 if (S_sv_2iuv_common(aTHX_ sv))
2326 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"IVdf")\n",
2327 PTR2UV(sv),SvIVX(sv)));
2328 return SvIsUV(sv) ? (IV)SvUVX(sv) : SvIVX(sv);
2332 =for apidoc sv_2uv_flags
2334 Return the unsigned integer value of an SV, doing any necessary string
2335 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2336 Normally used via the C<SvUV(sv)> and C<SvUVx(sv)> macros.
2342 Perl_sv_2uv_flags(pTHX_ register SV *const sv, const I32 flags)
2347 if (SvGMAGICAL(sv) || SvVALID(sv)) {
2348 /* FBMs use the space for SvIVX and SvNVX for other purposes, and use
2349 the same flag bit as SVf_IVisUV, so must not let them cache IVs. */
2350 if (flags & SV_GMAGIC)
2355 return U_V(SvNVX(sv));
2356 if (SvPOKp(sv) && SvLEN(sv)) {
2359 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2361 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2362 == IS_NUMBER_IN_UV) {
2363 /* It's definitely an integer */
2364 if (!(numtype & IS_NUMBER_NEG))
2368 if (ckWARN(WARN_NUMERIC))
2371 return U_V(Atof(SvPVX_const(sv)));
2376 assert(SvTYPE(sv) >= SVt_PVMG);
2377 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
2378 } else if (SvTHINKFIRST(sv)) {
2383 if (flags & SV_SKIP_OVERLOAD)
2385 tmpstr = AMG_CALLunary(sv, numer_amg);
2386 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2387 return SvUV(tmpstr);
2390 return PTR2UV(SvRV(sv));
2393 sv_force_normal_flags(sv, 0);
2395 if (SvREADONLY(sv) && !SvOK(sv)) {
2396 if (ckWARN(WARN_UNINITIALIZED))
2402 if (S_sv_2iuv_common(aTHX_ sv))
2406 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2uv(%"UVuf")\n",
2407 PTR2UV(sv),SvUVX(sv)));
2408 return SvIsUV(sv) ? SvUVX(sv) : (UV)SvIVX(sv);
2412 =for apidoc sv_2nv_flags
2414 Return the num value of an SV, doing any necessary string or integer
2415 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2416 Normally used via the C<SvNV(sv)> and C<SvNVx(sv)> macros.
2422 Perl_sv_2nv_flags(pTHX_ register SV *const sv, const I32 flags)
2427 if (SvGMAGICAL(sv) || SvVALID(sv)) {
2428 /* FBMs use the space for SvIVX and SvNVX for other purposes, and use
2429 the same flag bit as SVf_IVisUV, so must not let them cache NVs. */
2430 if (flags & SV_GMAGIC)
2434 if ((SvPOKp(sv) && SvLEN(sv)) && !SvIOKp(sv)) {
2435 if (!SvIOKp(sv) && ckWARN(WARN_NUMERIC) &&
2436 !grok_number(SvPVX_const(sv), SvCUR(sv), NULL))
2438 return Atof(SvPVX_const(sv));
2442 return (NV)SvUVX(sv);
2444 return (NV)SvIVX(sv);
2449 assert(SvTYPE(sv) >= SVt_PVMG);
2450 /* This falls through to the report_uninit near the end of the
2452 } else if (SvTHINKFIRST(sv)) {
2457 if (flags & SV_SKIP_OVERLOAD)
2459 tmpstr = AMG_CALLunary(sv, numer_amg);
2460 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2461 return SvNV(tmpstr);
2464 return PTR2NV(SvRV(sv));
2467 sv_force_normal_flags(sv, 0);
2469 if (SvREADONLY(sv) && !SvOK(sv)) {
2470 if (ckWARN(WARN_UNINITIALIZED))
2475 if (SvTYPE(sv) < SVt_NV) {
2476 /* The logic to use SVt_PVNV if necessary is in sv_upgrade. */
2477 sv_upgrade(sv, SVt_NV);
2478 #ifdef USE_LONG_DOUBLE
2480 STORE_NUMERIC_LOCAL_SET_STANDARD();
2481 PerlIO_printf(Perl_debug_log,
2482 "0x%"UVxf" num(%" PERL_PRIgldbl ")\n",
2483 PTR2UV(sv), SvNVX(sv));
2484 RESTORE_NUMERIC_LOCAL();
2488 STORE_NUMERIC_LOCAL_SET_STANDARD();
2489 PerlIO_printf(Perl_debug_log, "0x%"UVxf" num(%"NVgf")\n",
2490 PTR2UV(sv), SvNVX(sv));
2491 RESTORE_NUMERIC_LOCAL();
2495 else if (SvTYPE(sv) < SVt_PVNV)
2496 sv_upgrade(sv, SVt_PVNV);
2501 SvNV_set(sv, SvIsUV(sv) ? (NV)SvUVX(sv) : (NV)SvIVX(sv));
2502 #ifdef NV_PRESERVES_UV
2508 /* Only set the public NV OK flag if this NV preserves the IV */
2509 /* Check it's not 0xFFFFFFFFFFFFFFFF */
2511 SvIsUV(sv) ? ((SvUVX(sv) != UV_MAX)&&(SvUVX(sv) == U_V(SvNVX(sv))))
2512 : (SvIVX(sv) == I_V(SvNVX(sv))))
2518 else if (SvPOKp(sv) && SvLEN(sv)) {
2520 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2521 if (!SvIOKp(sv) && !numtype && ckWARN(WARN_NUMERIC))
2523 #ifdef NV_PRESERVES_UV
2524 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2525 == IS_NUMBER_IN_UV) {
2526 /* It's definitely an integer */
2527 SvNV_set(sv, (numtype & IS_NUMBER_NEG) ? -(NV)value : (NV)value);
2529 SvNV_set(sv, Atof(SvPVX_const(sv)));
2535 SvNV_set(sv, Atof(SvPVX_const(sv)));
2536 /* Only set the public NV OK flag if this NV preserves the value in
2537 the PV at least as well as an IV/UV would.
2538 Not sure how to do this 100% reliably. */
2539 /* if that shift count is out of range then Configure's test is
2540 wonky. We shouldn't be in here with NV_PRESERVES_UV_BITS ==
2542 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2543 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2544 SvNOK_on(sv); /* Definitely small enough to preserve all bits */
2545 } else if (!(numtype & IS_NUMBER_IN_UV)) {
2546 /* Can't use strtol etc to convert this string, so don't try.
2547 sv_2iv and sv_2uv will use the NV to convert, not the PV. */
2550 /* value has been set. It may not be precise. */
2551 if ((numtype & IS_NUMBER_NEG) && (value > (UV)IV_MIN)) {
2552 /* 2s complement assumption for (UV)IV_MIN */
2553 SvNOK_on(sv); /* Integer is too negative. */
2558 if (numtype & IS_NUMBER_NEG) {
2559 SvIV_set(sv, -(IV)value);
2560 } else if (value <= (UV)IV_MAX) {
2561 SvIV_set(sv, (IV)value);
2563 SvUV_set(sv, value);
2567 if (numtype & IS_NUMBER_NOT_INT) {
2568 /* I believe that even if the original PV had decimals,
2569 they are lost beyond the limit of the FP precision.
2570 However, neither is canonical, so both only get p
2571 flags. NWC, 2000/11/25 */
2572 /* Both already have p flags, so do nothing */
2574 const NV nv = SvNVX(sv);
2575 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2576 if (SvIVX(sv) == I_V(nv)) {
2579 /* It had no "." so it must be integer. */
2583 /* between IV_MAX and NV(UV_MAX).
2584 Could be slightly > UV_MAX */
2586 if (numtype & IS_NUMBER_NOT_INT) {
2587 /* UV and NV both imprecise. */
2589 const UV nv_as_uv = U_V(nv);
2591 if (value == nv_as_uv && SvUVX(sv) != UV_MAX) {
2600 /* It might be more code efficient to go through the entire logic above
2601 and conditionally set with SvNOKp_on() rather than SvNOK(), but it
2602 gets complex and potentially buggy, so more programmer efficient
2603 to do it this way, by turning off the public flags: */
2605 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK);
2606 #endif /* NV_PRESERVES_UV */
2609 if (isGV_with_GP(sv)) {
2610 glob_2number(MUTABLE_GV(sv));
2614 if (!PL_localizing && !SvPADTMP(sv) && ckWARN(WARN_UNINITIALIZED))
2616 assert (SvTYPE(sv) >= SVt_NV);
2617 /* Typically the caller expects that sv_any is not NULL now. */
2618 /* XXX Ilya implies that this is a bug in callers that assume this
2619 and ideally should be fixed. */
2622 #if defined(USE_LONG_DOUBLE)
2624 STORE_NUMERIC_LOCAL_SET_STANDARD();
2625 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2nv(%" PERL_PRIgldbl ")\n",
2626 PTR2UV(sv), SvNVX(sv));
2627 RESTORE_NUMERIC_LOCAL();
2631 STORE_NUMERIC_LOCAL_SET_STANDARD();
2632 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 1nv(%"NVgf")\n",
2633 PTR2UV(sv), SvNVX(sv));
2634 RESTORE_NUMERIC_LOCAL();
2643 Return an SV with the numeric value of the source SV, doing any necessary
2644 reference or overload conversion. You must use the C<SvNUM(sv)> macro to
2645 access this function.
2651 Perl_sv_2num(pTHX_ register SV *const sv)
2653 PERL_ARGS_ASSERT_SV_2NUM;
2658 SV * const tmpsv = AMG_CALLunary(sv, numer_amg);
2659 TAINT_IF(tmpsv && SvTAINTED(tmpsv));
2660 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
2661 return sv_2num(tmpsv);
2663 return sv_2mortal(newSVuv(PTR2UV(SvRV(sv))));
2666 /* uiv_2buf(): private routine for use by sv_2pv_flags(): print an IV or
2667 * UV as a string towards the end of buf, and return pointers to start and
2670 * We assume that buf is at least TYPE_CHARS(UV) long.
2674 S_uiv_2buf(char *const buf, const IV iv, UV uv, const int is_uv, char **const peob)
2676 char *ptr = buf + TYPE_CHARS(UV);
2677 char * const ebuf = ptr;
2680 PERL_ARGS_ASSERT_UIV_2BUF;
2692 *--ptr = '0' + (char)(uv % 10);
2701 =for apidoc sv_2pv_flags
2703 Returns a pointer to the string value of an SV, and sets *lp to its length.
2704 If flags includes SV_GMAGIC, does an mg_get() first. Coerces sv to a
2705 string if necessary. Normally invoked via the C<SvPV_flags> macro.
2706 C<sv_2pv()> and C<sv_2pv_nomg> usually end up here too.
2712 Perl_sv_2pv_flags(pTHX_ register SV *const sv, STRLEN *const lp, const I32 flags)
2722 if (SvGMAGICAL(sv)) {
2723 if (flags & SV_GMAGIC)
2728 if (flags & SV_MUTABLE_RETURN)
2729 return SvPVX_mutable(sv);
2730 if (flags & SV_CONST_RETURN)
2731 return (char *)SvPVX_const(sv);
2734 if (SvIOKp(sv) || SvNOKp(sv)) {
2735 char tbuf[64]; /* Must fit sprintf/Gconvert of longest IV/NV */
2740 ? my_snprintf(tbuf, sizeof(tbuf), "%"UVuf, (UV)SvUVX(sv))
2741 : my_snprintf(tbuf, sizeof(tbuf), "%"IVdf, (IV)SvIVX(sv));
2742 } else if(SvNVX(sv) == 0.0) {
2747 Gconvert(SvNVX(sv), NV_DIG, 0, tbuf);
2754 SvUPGRADE(sv, SVt_PV);
2757 s = SvGROW_mutable(sv, len + 1);
2760 return (char*)memcpy(s, tbuf, len + 1);
2766 assert(SvTYPE(sv) >= SVt_PVMG);
2767 /* This falls through to the report_uninit near the end of the
2769 } else if (SvTHINKFIRST(sv)) {
2774 if (flags & SV_SKIP_OVERLOAD)
2776 tmpstr = AMG_CALLunary(sv, string_amg);
2777 TAINT_IF(tmpstr && SvTAINTED(tmpstr));
2778 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2780 /* char *pv = lp ? SvPV(tmpstr, *lp) : SvPV_nolen(tmpstr);
2784 if ((SvFLAGS(tmpstr) & (SVf_POK)) == SVf_POK) {
2785 if (flags & SV_CONST_RETURN) {
2786 pv = (char *) SvPVX_const(tmpstr);
2788 pv = (flags & SV_MUTABLE_RETURN)
2789 ? SvPVX_mutable(tmpstr) : SvPVX(tmpstr);
2792 *lp = SvCUR(tmpstr);
2794 pv = sv_2pv_flags(tmpstr, lp, flags);
2807 SV *const referent = SvRV(sv);
2811 retval = buffer = savepvn("NULLREF", len);
2812 } else if (SvTYPE(referent) == SVt_REGEXP) {
2813 REGEXP * const re = (REGEXP *)MUTABLE_PTR(referent);
2818 /* If the regex is UTF-8 we want the containing scalar to
2819 have an UTF-8 flag too */
2825 if ((seen_evals = RX_SEEN_EVALS(re)))
2826 PL_reginterp_cnt += seen_evals;
2829 *lp = RX_WRAPLEN(re);
2831 return RX_WRAPPED(re);
2833 const char *const typestr = sv_reftype(referent, 0);
2834 const STRLEN typelen = strlen(typestr);
2835 UV addr = PTR2UV(referent);
2836 const char *stashname = NULL;
2837 STRLEN stashnamelen = 0; /* hush, gcc */
2838 const char *buffer_end;
2840 if (SvOBJECT(referent)) {
2841 const HEK *const name = HvNAME_HEK(SvSTASH(referent));
2844 stashname = HEK_KEY(name);
2845 stashnamelen = HEK_LEN(name);
2847 if (HEK_UTF8(name)) {
2853 stashname = "__ANON__";
2856 len = stashnamelen + 1 /* = */ + typelen + 3 /* (0x */
2857 + 2 * sizeof(UV) + 2 /* )\0 */;
2859 len = typelen + 3 /* (0x */
2860 + 2 * sizeof(UV) + 2 /* )\0 */;
2863 Newx(buffer, len, char);
2864 buffer_end = retval = buffer + len;
2866 /* Working backwards */
2870 *--retval = PL_hexdigit[addr & 15];
2871 } while (addr >>= 4);
2877 memcpy(retval, typestr, typelen);
2881 retval -= stashnamelen;
2882 memcpy(retval, stashname, stashnamelen);
2884 /* retval may not necessarily have reached the start of the
2886 assert (retval >= buffer);
2888 len = buffer_end - retval - 1; /* -1 for that \0 */
2896 if (SvREADONLY(sv) && !SvOK(sv)) {
2899 if (flags & SV_UNDEF_RETURNS_NULL)
2901 if (ckWARN(WARN_UNINITIALIZED))
2906 if (SvIOK(sv) || ((SvIOKp(sv) && !SvNOKp(sv)))) {
2907 /* I'm assuming that if both IV and NV are equally valid then
2908 converting the IV is going to be more efficient */
2909 const U32 isUIOK = SvIsUV(sv);
2910 char buf[TYPE_CHARS(UV)];
2914 if (SvTYPE(sv) < SVt_PVIV)
2915 sv_upgrade(sv, SVt_PVIV);
2916 ptr = uiv_2buf(buf, SvIVX(sv), SvUVX(sv), isUIOK, &ebuf);
2918 /* inlined from sv_setpvn */
2919 s = SvGROW_mutable(sv, len + 1);
2920 Move(ptr, s, len, char);
2924 else if (SvNOKp(sv)) {
2925 if (SvTYPE(sv) < SVt_PVNV)
2926 sv_upgrade(sv, SVt_PVNV);
2927 if (SvNVX(sv) == 0.0) {
2928 s = SvGROW_mutable(sv, 2);
2933 /* The +20 is pure guesswork. Configure test needed. --jhi */
2934 s = SvGROW_mutable(sv, NV_DIG + 20);
2935 /* some Xenix systems wipe out errno here */
2936 Gconvert(SvNVX(sv), NV_DIG, 0, s);
2946 if (isGV_with_GP(sv)) {
2947 GV *const gv = MUTABLE_GV(sv);
2948 SV *const buffer = sv_newmortal();
2950 gv_efullname3(buffer, gv, "*");
2952 assert(SvPOK(buffer));
2954 *lp = SvCUR(buffer);
2956 if ( SvUTF8(buffer) ) SvUTF8_on(sv);
2957 return SvPVX(buffer);
2962 if (flags & SV_UNDEF_RETURNS_NULL)
2964 if (!PL_localizing && !SvPADTMP(sv) && ckWARN(WARN_UNINITIALIZED))
2966 if (SvTYPE(sv) < SVt_PV)
2967 /* Typically the caller expects that sv_any is not NULL now. */
2968 sv_upgrade(sv, SVt_PV);
2972 const STRLEN len = s - SvPVX_const(sv);
2978 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
2979 PTR2UV(sv),SvPVX_const(sv)));
2980 if (flags & SV_CONST_RETURN)
2981 return (char *)SvPVX_const(sv);
2982 if (flags & SV_MUTABLE_RETURN)
2983 return SvPVX_mutable(sv);
2988 =for apidoc sv_copypv
2990 Copies a stringified representation of the source SV into the
2991 destination SV. Automatically performs any necessary mg_get and
2992 coercion of numeric values into strings. Guaranteed to preserve
2993 UTF8 flag even from overloaded objects. Similar in nature to
2994 sv_2pv[_flags] but operates directly on an SV instead of just the
2995 string. Mostly uses sv_2pv_flags to do its work, except when that
2996 would lose the UTF-8'ness of the PV.
3002 Perl_sv_copypv(pTHX_ SV *const dsv, register SV *const ssv)
3005 const char * const s = SvPV_const(ssv,len);
3007 PERL_ARGS_ASSERT_SV_COPYPV;
3009 sv_setpvn(dsv,s,len);
3017 =for apidoc sv_2pvbyte
3019 Return a pointer to the byte-encoded representation of the SV, and set *lp
3020 to its length. May cause the SV to be downgraded from UTF-8 as a
3023 Usually accessed via the C<SvPVbyte> macro.
3029 Perl_sv_2pvbyte(pTHX_ register SV *const sv, STRLEN *const lp)
3031 PERL_ARGS_ASSERT_SV_2PVBYTE;
3034 sv_utf8_downgrade(sv,0);
3035 return lp ? SvPV_nomg(sv,*lp) : SvPV_nomg_nolen(sv);
3039 =for apidoc sv_2pvutf8
3041 Return a pointer to the UTF-8-encoded representation of the SV, and set *lp
3042 to its length. May cause the SV to be upgraded to UTF-8 as a side-effect.
3044 Usually accessed via the C<SvPVutf8> macro.
3050 Perl_sv_2pvutf8(pTHX_ register SV *const sv, STRLEN *const lp)
3052 PERL_ARGS_ASSERT_SV_2PVUTF8;
3054 sv_utf8_upgrade(sv);
3055 return lp ? SvPV(sv,*lp) : SvPV_nolen(sv);
3060 =for apidoc sv_2bool
3062 This macro is only used by sv_true() or its macro equivalent, and only if
3063 the latter's argument is neither SvPOK, SvIOK nor SvNOK.
3064 It calls sv_2bool_flags with the SV_GMAGIC flag.
3066 =for apidoc sv_2bool_flags
3068 This function is only used by sv_true() and friends, and only if
3069 the latter's argument is neither SvPOK, SvIOK nor SvNOK. If the flags
3070 contain SV_GMAGIC, then it does an mg_get() first.
3077 Perl_sv_2bool_flags(pTHX_ register SV *const sv, const I32 flags)
3081 PERL_ARGS_ASSERT_SV_2BOOL_FLAGS;
3083 if(flags & SV_GMAGIC) SvGETMAGIC(sv);
3089 SV * const tmpsv = AMG_CALLunary(sv, bool__amg);
3090 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
3091 return cBOOL(SvTRUE(tmpsv));
3093 return SvRV(sv) != 0;
3096 register XPV* const Xpvtmp = (XPV*)SvANY(sv);
3098 (*sv->sv_u.svu_pv > '0' ||
3099 Xpvtmp->xpv_cur > 1 ||
3100 (Xpvtmp->xpv_cur && *sv->sv_u.svu_pv != '0')))
3107 return SvIVX(sv) != 0;
3110 return SvNVX(sv) != 0.0;
3112 if (isGV_with_GP(sv))
3122 =for apidoc sv_utf8_upgrade
3124 Converts the PV of an SV to its UTF-8-encoded form.
3125 Forces the SV to string form if it is not already.
3126 Will C<mg_get> on C<sv> if appropriate.
3127 Always sets the SvUTF8 flag to avoid future validity checks even
3128 if the whole string is the same in UTF-8 as not.
3129 Returns the number of bytes in the converted string
3131 This is not as a general purpose byte encoding to Unicode interface:
3132 use the Encode extension for that.
3134 =for apidoc sv_utf8_upgrade_nomg
3136 Like sv_utf8_upgrade, but doesn't do magic on C<sv>.
3138 =for apidoc sv_utf8_upgrade_flags
3140 Converts the PV of an SV to its UTF-8-encoded form.
3141 Forces the SV to string form if it is not already.
3142 Always sets the SvUTF8 flag to avoid future validity checks even
3143 if all the bytes are invariant in UTF-8.
3144 If C<flags> has C<SV_GMAGIC> bit set,
3145 will C<mg_get> on C<sv> if appropriate, else not.
3146 Returns the number of bytes in the converted string
3147 C<sv_utf8_upgrade> and
3148 C<sv_utf8_upgrade_nomg> are implemented in terms of this function.
3150 This is not as a general purpose byte encoding to Unicode interface:
3151 use the Encode extension for that.
3155 The grow version is currently not externally documented. It adds a parameter,
3156 extra, which is the number of unused bytes the string of 'sv' is guaranteed to
3157 have free after it upon return. This allows the caller to reserve extra space
3158 that it intends to fill, to avoid extra grows.
3160 Also externally undocumented for the moment is the flag SV_FORCE_UTF8_UPGRADE,
3161 which can be used to tell this function to not first check to see if there are
3162 any characters that are different in UTF-8 (variant characters) which would
3163 force it to allocate a new string to sv, but to assume there are. Typically
3164 this flag is used by a routine that has already parsed the string to find that
3165 there are such characters, and passes this information on so that the work
3166 doesn't have to be repeated.
3168 (One might think that the calling routine could pass in the position of the
3169 first such variant, so it wouldn't have to be found again. But that is not the
3170 case, because typically when the caller is likely to use this flag, it won't be
3171 calling this routine unless it finds something that won't fit into a byte.
3172 Otherwise it tries to not upgrade and just use bytes. But some things that
3173 do fit into a byte are variants in utf8, and the caller may not have been
3174 keeping track of these.)
3176 If the routine itself changes the string, it adds a trailing NUL. Such a NUL
3177 isn't guaranteed due to having other routines do the work in some input cases,
3178 or if the input is already flagged as being in utf8.
3180 The speed of this could perhaps be improved for many cases if someone wanted to
3181 write a fast function that counts the number of variant characters in a string,
3182 especially if it could return the position of the first one.
3187 Perl_sv_utf8_upgrade_flags_grow(pTHX_ register SV *const sv, const I32 flags, STRLEN extra)
3191 PERL_ARGS_ASSERT_SV_UTF8_UPGRADE_FLAGS_GROW;
3193 if (sv == &PL_sv_undef)
3197 if (SvREADONLY(sv) && (SvPOKp(sv) || SvIOKp(sv) || SvNOKp(sv))) {
3198 (void) sv_2pv_flags(sv,&len, flags);
3200 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3204 (void) SvPV_force_flags(sv,len,flags & SV_GMAGIC);
3209 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3214 sv_force_normal_flags(sv, 0);
3217 if (PL_encoding && !(flags & SV_UTF8_NO_ENCODING)) {
3218 sv_recode_to_utf8(sv, PL_encoding);
3219 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3223 if (SvCUR(sv) == 0) {
3224 if (extra) SvGROW(sv, extra);
3225 } else { /* Assume Latin-1/EBCDIC */
3226 /* This function could be much more efficient if we
3227 * had a FLAG in SVs to signal if there are any variant
3228 * chars in the PV. Given that there isn't such a flag
3229 * make the loop as fast as possible (although there are certainly ways
3230 * to speed this up, eg. through vectorization) */
3231 U8 * s = (U8 *) SvPVX_const(sv);
3232 U8 * e = (U8 *) SvEND(sv);
3234 STRLEN two_byte_count = 0;
3236 if (flags & SV_FORCE_UTF8_UPGRADE) goto must_be_utf8;
3238 /* See if really will need to convert to utf8. We mustn't rely on our
3239 * incoming SV being well formed and having a trailing '\0', as certain
3240 * code in pp_formline can send us partially built SVs. */
3244 if (NATIVE_IS_INVARIANT(ch)) continue;
3246 t--; /* t already incremented; re-point to first variant */
3251 /* utf8 conversion not needed because all are invariants. Mark as
3252 * UTF-8 even if no variant - saves scanning loop */
3254 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3259 /* Here, the string should be converted to utf8, either because of an
3260 * input flag (two_byte_count = 0), or because a character that
3261 * requires 2 bytes was found (two_byte_count = 1). t points either to
3262 * the beginning of the string (if we didn't examine anything), or to
3263 * the first variant. In either case, everything from s to t - 1 will
3264 * occupy only 1 byte each on output.
3266 * There are two main ways to convert. One is to create a new string
3267 * and go through the input starting from the beginning, appending each
3268 * converted value onto the new string as we go along. It's probably
3269 * best to allocate enough space in the string for the worst possible
3270 * case rather than possibly running out of space and having to
3271 * reallocate and then copy what we've done so far. Since everything
3272 * from s to t - 1 is invariant, the destination can be initialized
3273 * with these using a fast memory copy
3275 * The other way is to figure out exactly how big the string should be
3276 * by parsing the entire input. Then you don't have to make it big
3277 * enough to handle the worst possible case, and more importantly, if
3278 * the string you already have is large enough, you don't have to
3279 * allocate a new string, you can copy the last character in the input
3280 * string to the final position(s) that will be occupied by the
3281 * converted string and go backwards, stopping at t, since everything
3282 * before that is invariant.
3284 * There are advantages and disadvantages to each method.
3286 * In the first method, we can allocate a new string, do the memory
3287 * copy from the s to t - 1, and then proceed through the rest of the
3288 * string byte-by-byte.
3290 * In the second method, we proceed through the rest of the input
3291 * string just calculating how big the converted string will be. Then
3292 * there are two cases:
3293 * 1) if the string has enough extra space to handle the converted
3294 * value. We go backwards through the string, converting until we
3295 * get to the position we are at now, and then stop. If this
3296 * position is far enough along in the string, this method is
3297 * faster than the other method. If the memory copy were the same
3298 * speed as the byte-by-byte loop, that position would be about
3299 * half-way, as at the half-way mark, parsing to the end and back
3300 * is one complete string's parse, the same amount as starting
3301 * over and going all the way through. Actually, it would be
3302 * somewhat less than half-way, as it's faster to just count bytes
3303 * than to also copy, and we don't have the overhead of allocating
3304 * a new string, changing the scalar to use it, and freeing the
3305 * existing one. But if the memory copy is fast, the break-even
3306 * point is somewhere after half way. The counting loop could be
3307 * sped up by vectorization, etc, to move the break-even point
3308 * further towards the beginning.
3309 * 2) if the string doesn't have enough space to handle the converted
3310 * value. A new string will have to be allocated, and one might
3311 * as well, given that, start from the beginning doing the first
3312 * method. We've spent extra time parsing the string and in
3313 * exchange all we've gotten is that we know precisely how big to
3314 * make the new one. Perl is more optimized for time than space,
3315 * so this case is a loser.
3316 * So what I've decided to do is not use the 2nd method unless it is
3317 * guaranteed that a new string won't have to be allocated, assuming
3318 * the worst case. I also decided not to put any more conditions on it
3319 * than this, for now. It seems likely that, since the worst case is
3320 * twice as big as the unknown portion of the string (plus 1), we won't
3321 * be guaranteed enough space, causing us to go to the first method,
3322 * unless the string is short, or the first variant character is near
3323 * the end of it. In either of these cases, it seems best to use the
3324 * 2nd method. The only circumstance I can think of where this would
3325 * be really slower is if the string had once had much more data in it
3326 * than it does now, but there is still a substantial amount in it */
3329 STRLEN invariant_head = t - s;
3330 STRLEN size = invariant_head + (e - t) * 2 + 1 + extra;
3331 if (SvLEN(sv) < size) {
3333 /* Here, have decided to allocate a new string */
3338 Newx(dst, size, U8);
3340 /* If no known invariants at the beginning of the input string,
3341 * set so starts from there. Otherwise, can use memory copy to
3342 * get up to where we are now, and then start from here */
3344 if (invariant_head <= 0) {
3347 Copy(s, dst, invariant_head, char);
3348 d = dst + invariant_head;
3352 const UV uv = NATIVE8_TO_UNI(*t++);
3353 if (UNI_IS_INVARIANT(uv))
3354 *d++ = (U8)UNI_TO_NATIVE(uv);
3356 *d++ = (U8)UTF8_EIGHT_BIT_HI(uv);
3357 *d++ = (U8)UTF8_EIGHT_BIT_LO(uv);
3361 SvPV_free(sv); /* No longer using pre-existing string */
3362 SvPV_set(sv, (char*)dst);
3363 SvCUR_set(sv, d - dst);
3364 SvLEN_set(sv, size);
3367 /* Here, have decided to get the exact size of the string.
3368 * Currently this happens only when we know that there is
3369 * guaranteed enough space to fit the converted string, so
3370 * don't have to worry about growing. If two_byte_count is 0,
3371 * then t points to the first byte of the string which hasn't
3372 * been examined yet. Otherwise two_byte_count is 1, and t
3373 * points to the first byte in the string that will expand to
3374 * two. Depending on this, start examining at t or 1 after t.
3377 U8 *d = t + two_byte_count;
3380 /* Count up the remaining bytes that expand to two */
3383 const U8 chr = *d++;
3384 if (! NATIVE_IS_INVARIANT(chr)) two_byte_count++;
3387 /* The string will expand by just the number of bytes that
3388 * occupy two positions. But we are one afterwards because of
3389 * the increment just above. This is the place to put the
3390 * trailing NUL, and to set the length before we decrement */
3392 d += two_byte_count;
3393 SvCUR_set(sv, d - s);
3397 /* Having decremented d, it points to the position to put the
3398 * very last byte of the expanded string. Go backwards through
3399 * the string, copying and expanding as we go, stopping when we
3400 * get to the part that is invariant the rest of the way down */
3404 const U8 ch = NATIVE8_TO_UNI(*e--);
3405 if (UNI_IS_INVARIANT(ch)) {
3406 *d-- = UNI_TO_NATIVE(ch);
3408 *d-- = (U8)UTF8_EIGHT_BIT_LO(ch);
3409 *d-- = (U8)UTF8_EIGHT_BIT_HI(ch);
3414 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
3415 /* Update pos. We do it at the end rather than during
3416 * the upgrade, to avoid slowing down the common case
3417 * (upgrade without pos) */
3418 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
3420 I32 pos = mg->mg_len;
3421 if (pos > 0 && (U32)pos > invariant_head) {
3422 U8 *d = (U8*) SvPVX(sv) + invariant_head;
3423 STRLEN n = (U32)pos - invariant_head;
3425 if (UTF8_IS_START(*d))
3430 mg->mg_len = d - (U8*)SvPVX(sv);
3433 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
3434 magic_setutf8(sv,mg); /* clear UTF8 cache */
3439 /* Mark as UTF-8 even if no variant - saves scanning loop */
3445 =for apidoc sv_utf8_downgrade
3447 Attempts to convert the PV of an SV from characters to bytes.
3448 If the PV contains a character that cannot fit
3449 in a byte, this conversion will fail;
3450 in this case, either returns false or, if C<fail_ok> is not
3453 This is not as a general purpose Unicode to byte encoding interface:
3454 use the Encode extension for that.
3460 Perl_sv_utf8_downgrade(pTHX_ register SV *const sv, const bool fail_ok)
3464 PERL_ARGS_ASSERT_SV_UTF8_DOWNGRADE;
3466 if (SvPOKp(sv) && SvUTF8(sv)) {
3470 int mg_flags = SV_GMAGIC;
3473 sv_force_normal_flags(sv, 0);
3475 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
3477 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
3479 I32 pos = mg->mg_len;
3481 sv_pos_b2u(sv, &pos);
3482 mg_flags = 0; /* sv_pos_b2u does get magic */
3486 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
3487 magic_setutf8(sv,mg); /* clear UTF8 cache */
3490 s = (U8 *) SvPV_flags(sv, len, mg_flags);
3492 if (!utf8_to_bytes(s, &len)) {
3497 Perl_croak(aTHX_ "Wide character in %s",
3500 Perl_croak(aTHX_ "Wide character");
3511 =for apidoc sv_utf8_encode
3513 Converts the PV of an SV to UTF-8, but then turns the C<SvUTF8>
3514 flag off so that it looks like octets again.
3520 Perl_sv_utf8_encode(pTHX_ register SV *const sv)
3522 PERL_ARGS_ASSERT_SV_UTF8_ENCODE;
3525 sv_force_normal_flags(sv, 0);
3527 if (SvREADONLY(sv)) {
3528 Perl_croak_no_modify(aTHX);
3530 (void) sv_utf8_upgrade(sv);
3535 =for apidoc sv_utf8_decode
3537 If the PV of the SV is an octet sequence in UTF-8
3538 and contains a multiple-byte character, the C<SvUTF8> flag is turned on
3539 so that it looks like a character. If the PV contains only single-byte
3540 characters, the C<SvUTF8> flag stays off.
3541 Scans PV for validity and returns false if the PV is invalid UTF-8.
3547 Perl_sv_utf8_decode(pTHX_ register SV *const sv)
3549 PERL_ARGS_ASSERT_SV_UTF8_DECODE;
3552 const U8 *start, *c;
3555 /* The octets may have got themselves encoded - get them back as
3558 if (!sv_utf8_downgrade(sv, TRUE))
3561 /* it is actually just a matter of turning the utf8 flag on, but
3562 * we want to make sure everything inside is valid utf8 first.
3564 c = start = (const U8 *) SvPVX_const(sv);
3565 if (!is_utf8_string(c, SvCUR(sv)+1))
3567 e = (const U8 *) SvEND(sv);
3570 if (!UTF8_IS_INVARIANT(ch)) {
3575 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
3576 /* adjust pos to the start of a UTF8 char sequence */
3577 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
3579 I32 pos = mg->mg_len;
3581 for (c = start + pos; c > start; c--) {
3582 if (UTF8_IS_START(*c))
3585 mg->mg_len = c - start;
3588 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
3589 magic_setutf8(sv,mg); /* clear UTF8 cache */
3596 =for apidoc sv_setsv
3598 Copies the contents of the source SV C<ssv> into the destination SV
3599 C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3600 function if the source SV needs to be reused. Does not handle 'set' magic.
3601 Loosely speaking, it performs a copy-by-value, obliterating any previous
3602 content of the destination.
3604 You probably want to use one of the assortment of wrappers, such as
3605 C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3606 C<SvSetMagicSV_nosteal>.
3608 =for apidoc sv_setsv_flags
3610 Copies the contents of the source SV C<ssv> into the destination SV
3611 C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3612 function if the source SV needs to be reused. Does not handle 'set' magic.
3613 Loosely speaking, it performs a copy-by-value, obliterating any previous
3614 content of the destination.
3615 If the C<flags> parameter has the C<SV_GMAGIC> bit set, will C<mg_get> on
3616 C<ssv> if appropriate, else not. If the C<flags>
3617 parameter has the C<NOSTEAL> bit set then the
3618 buffers of temps will not be stolen. <sv_setsv>
3619 and C<sv_setsv_nomg> are implemented in terms of this function.
3621 You probably want to use one of the assortment of wrappers, such as
3622 C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3623 C<SvSetMagicSV_nosteal>.
3625 This is the primary function for copying scalars, and most other
3626 copy-ish functions and macros use this underneath.
3632 S_glob_assign_glob(pTHX_ SV *const dstr, SV *const sstr, const int dtype)
3634 I32 mro_changes = 0; /* 1 = method, 2 = isa, 3 = recursive isa */
3635 HV *old_stash = NULL;
3637 PERL_ARGS_ASSERT_GLOB_ASSIGN_GLOB;
3639 if (dtype != SVt_PVGV && !isGV_with_GP(dstr)) {
3640 const char * const name = GvNAME(sstr);
3641 const STRLEN len = GvNAMELEN(sstr);
3643 if (dtype >= SVt_PV) {
3649 SvUPGRADE(dstr, SVt_PVGV);
3650 (void)SvOK_off(dstr);
3651 /* We have to turn this on here, even though we turn it off
3652 below, as GvSTASH will fail an assertion otherwise. */
3653 isGV_with_GP_on(dstr);
3655 GvSTASH(dstr) = GvSTASH(sstr);
3657 Perl_sv_add_backref(aTHX_ MUTABLE_SV(GvSTASH(dstr)), dstr);
3658 gv_name_set(MUTABLE_GV(dstr), name, len,
3659 GV_ADD | (GvNAMEUTF8(sstr) ? SVf_UTF8 : 0 ));
3660 SvFAKE_on(dstr); /* can coerce to non-glob */
3663 if(GvGP(MUTABLE_GV(sstr))) {
3664 /* If source has method cache entry, clear it */
3666 SvREFCNT_dec(GvCV(sstr));
3667 GvCV_set(sstr, NULL);
3670 /* If source has a real method, then a method is
3673 GvCV((const GV *)sstr) && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3679 /* If dest already had a real method, that's a change as well */
3681 !mro_changes && GvGP(MUTABLE_GV(dstr)) && GvCVu((const GV *)dstr)
3682 && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3687 /* We don't need to check the name of the destination if it was not a
3688 glob to begin with. */
3689 if(dtype == SVt_PVGV) {
3690 const char * const name = GvNAME((const GV *)dstr);
3693 /* The stash may have been detached from the symbol table, so
3695 && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3696 && GvAV((const GV *)sstr)
3700 const STRLEN len = GvNAMELEN(dstr);
3701 if ((len > 1 && name[len-2] == ':' && name[len-1] == ':')
3702 || (len == 1 && name[0] == ':')) {
3705 /* Set aside the old stash, so we can reset isa caches on
3707 if((old_stash = GvHV(dstr)))
3708 /* Make sure we do not lose it early. */
3709 SvREFCNT_inc_simple_void_NN(
3710 sv_2mortal((SV *)old_stash)
3716 gp_free(MUTABLE_GV(dstr));
3717 isGV_with_GP_off(dstr); /* SvOK_off does not like globs. */
3718 (void)SvOK_off(dstr);
3719 isGV_with_GP_on(dstr);
3720 GvINTRO_off(dstr); /* one-shot flag */
3721 GvGP_set(dstr, gp_ref(GvGP(sstr)));
3722 if (SvTAINTED(sstr))
3724 if (GvIMPORTED(dstr) != GVf_IMPORTED
3725 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3727 GvIMPORTED_on(dstr);
3730 if(mro_changes == 2) {
3732 SV * const sref = (SV *)GvAV((const GV *)dstr);
3733 if (SvSMAGICAL(sref) && (mg = mg_find(sref, PERL_MAGIC_isa))) {
3734 if (SvTYPE(mg->mg_obj) != SVt_PVAV) {
3735 AV * const ary = newAV();
3736 av_push(ary, mg->mg_obj); /* takes the refcount */
3737 mg->mg_obj = (SV *)ary;
3739 av_push((AV *)mg->mg_obj, SvREFCNT_inc_simple_NN(dstr));
3741 else sv_magic(sref, dstr, PERL_MAGIC_isa, NULL, 0);
3742 mro_isa_changed_in(GvSTASH(dstr));
3744 else if(mro_changes == 3) {
3745 HV * const stash = GvHV(dstr);
3746 if(old_stash ? (HV *)HvENAME_get(old_stash) : stash)
3752 else if(mro_changes) mro_method_changed_in(GvSTASH(dstr));
3757 S_glob_assign_ref(pTHX_ SV *const dstr, SV *const sstr)
3759 SV * const sref = SvREFCNT_inc(SvRV(sstr));
3761 const int intro = GvINTRO(dstr);
3764 const U32 stype = SvTYPE(sref);
3766 PERL_ARGS_ASSERT_GLOB_ASSIGN_REF;
3769 GvINTRO_off(dstr); /* one-shot flag */
3770 GvLINE(dstr) = CopLINE(PL_curcop);
3771 GvEGV(dstr) = MUTABLE_GV(dstr);
3776 location = (SV **) &(GvGP(dstr)->gp_cv); /* XXX bypassing GvCV_set */
3777 import_flag = GVf_IMPORTED_CV;
3780 location = (SV **) &GvHV(dstr);
3781 import_flag = GVf_IMPORTED_HV;
3784 location = (SV **) &GvAV(dstr);
3785 import_flag = GVf_IMPORTED_AV;
3788 location = (SV **) &GvIOp(dstr);
3791 location = (SV **) &GvFORM(dstr);
3794 location = &GvSV(dstr);
3795 import_flag = GVf_IMPORTED_SV;
3798 if (stype == SVt_PVCV) {
3799 /*if (GvCVGEN(dstr) && (GvCV(dstr) != (const CV *)sref || GvCVGEN(dstr))) {*/
3800 if (GvCVGEN(dstr)) {
3801 SvREFCNT_dec(GvCV(dstr));
3802 GvCV_set(dstr, NULL);
3803 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3806 SAVEGENERICSV(*location);
3810 if (stype == SVt_PVCV && (*location != sref || GvCVGEN(dstr))) {
3811 CV* const cv = MUTABLE_CV(*location);
3813 if (!GvCVGEN((const GV *)dstr) &&
3814 (CvROOT(cv) || CvXSUB(cv)) &&
3815 /* redundant check that avoids creating the extra SV
3816 most of the time: */
3817 (CvCONST(cv) || ckWARN(WARN_REDEFINE)))
3819 SV * const new_const_sv =
3820 CvCONST((const CV *)sref)
3821 ? cv_const_sv((const CV *)sref)
3823 report_redefined_cv(
3824 sv_2mortal(Perl_newSVpvf(aTHX_
3827 HvNAME_HEK(GvSTASH((const GV *)dstr))
3829 HEKfARG(GvENAME_HEK(MUTABLE_GV(dstr)))
3832 CvCONST((const CV *)sref) ? &new_const_sv : NULL
3836 cv_ckproto_len_flags(cv, (const GV *)dstr,
3837 SvPOK(sref) ? CvPROTO(sref) : NULL,
3838 SvPOK(sref) ? CvPROTOLEN(sref) : 0,
3839 SvPOK(sref) ? SvUTF8(sref) : 0);
3841 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3842 GvASSUMECV_on(dstr);
3843 if(GvSTASH(dstr)) mro_method_changed_in(GvSTASH(dstr)); /* sub foo { 1 } sub bar { 2 } *bar = \&foo */
3846 if (import_flag && !(GvFLAGS(dstr) & import_flag)
3847 && CopSTASH_ne(PL_curcop, GvSTASH(dstr))) {
3848 GvFLAGS(dstr) |= import_flag;
3850 if (stype == SVt_PVHV) {
3851 const char * const name = GvNAME((GV*)dstr);
3852 const STRLEN len = GvNAMELEN(dstr);
3855 (len > 1 && name[len-2] == ':' && name[len-1] == ':')
3856 || (len == 1 && name[0] == ':')
3858 && (!dref || HvENAME_get(dref))
3861 (HV *)sref, (HV *)dref,
3867 stype == SVt_PVAV && sref != dref
3868 && strEQ(GvNAME((GV*)dstr), "ISA")
3869 /* The stash may have been detached from the symbol table, so
3870 check its name before doing anything. */
3871 && GvSTASH(dstr) && HvENAME(GvSTASH(dstr))
3874 MAGIC * const omg = dref && SvSMAGICAL(dref)
3875 ? mg_find(dref, PERL_MAGIC_isa)
3877 if (SvSMAGICAL(sref) && (mg = mg_find(sref, PERL_MAGIC_isa))) {
3878 if (SvTYPE(mg->mg_obj) != SVt_PVAV) {
3879 AV * const ary = newAV();
3880 av_push(ary, mg->mg_obj); /* takes the refcount */
3881 mg->mg_obj = (SV *)ary;
3884 if (SvTYPE(omg->mg_obj) == SVt_PVAV) {
3885 SV **svp = AvARRAY((AV *)omg->mg_obj);
3886 I32 items = AvFILLp((AV *)omg->mg_obj) + 1;
3890 SvREFCNT_inc_simple_NN(*svp++)
3896 SvREFCNT_inc_simple_NN(omg->mg_obj)
3900 av_push((AV *)mg->mg_obj,SvREFCNT_inc_simple_NN(dstr));
3905 sref, omg ? omg->mg_obj : dstr, PERL_MAGIC_isa, NULL, 0
3907 mg = mg_find(sref, PERL_MAGIC_isa);
3909 /* Since the *ISA assignment could have affected more than
3910 one stash, don't call mro_isa_changed_in directly, but let
3911 magic_clearisa do it for us, as it already has the logic for
3912 dealing with globs vs arrays of globs. */
3914 Perl_magic_clearisa(aTHX_ NULL, mg);
3919 if (SvTAINTED(sstr))
3925 Perl_sv_setsv_flags(pTHX_ SV *dstr, register SV* sstr, const I32 flags)
3928 register U32 sflags;
3930 register svtype stype;
3932 PERL_ARGS_ASSERT_SV_SETSV_FLAGS;
3937 if (SvIS_FREED(dstr)) {
3938 Perl_croak(aTHX_ "panic: attempt to copy value %" SVf
3939 " to a freed scalar %p", SVfARG(sstr), (void *)dstr);
3941 SV_CHECK_THINKFIRST_COW_DROP(dstr);
3943 sstr = &PL_sv_undef;
3944 if (SvIS_FREED(sstr)) {
3945 Perl_croak(aTHX_ "panic: attempt to copy freed scalar %p to %p",
3946 (void*)sstr, (void*)dstr);
3948 stype = SvTYPE(sstr);
3949 dtype = SvTYPE(dstr);
3951 (void)SvAMAGIC_off(dstr);
3954 /* need to nuke the magic */
3955 sv_unmagic(dstr, PERL_MAGIC_vstring);
3958 /* There's a lot of redundancy below but we're going for speed here */
3963 if (dtype != SVt_PVGV && dtype != SVt_PVLV) {
3964 (void)SvOK_off(dstr);
3972 sv_upgrade(dstr, SVt_IV);
3976 sv_upgrade(dstr, SVt_PVIV);
3980 goto end_of_first_switch;
3982 (void)SvIOK_only(dstr);
3983 SvIV_set(dstr, SvIVX(sstr));
3986 /* SvTAINTED can only be true if the SV has taint magic, which in
3987 turn means that the SV type is PVMG (or greater). This is the
3988 case statement for SVt_IV, so this cannot be true (whatever gcov
3990 assert(!SvTAINTED(sstr));
3995 if (dtype < SVt_PV && dtype != SVt_IV)
3996 sv_upgrade(dstr, SVt_IV);
4004 sv_upgrade(dstr, SVt_NV);
4008 sv_upgrade(dstr, SVt_PVNV);
4012 goto end_of_first_switch;
4014 SvNV_set(dstr, SvNVX(sstr));
4015 (void)SvNOK_only(dstr);
4016 /* SvTAINTED can only be true if the SV has taint magic, which in
4017 turn means that the SV type is PVMG (or greater). This is the
4018 case statement for SVt_NV, so this cannot be true (whatever gcov
4020 assert(!SvTAINTED(sstr));
4026 #ifdef PERL_OLD_COPY_ON_WRITE
4027 if ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS) {
4028 if (dtype < SVt_PVIV)
4029 sv_upgrade(dstr, SVt_PVIV);
4036 sv_upgrade(dstr, SVt_PV);
4039 if (dtype < SVt_PVIV)
4040 sv_upgrade(dstr, SVt_PVIV);
4043 if (dtype < SVt_PVNV)
4044 sv_upgrade(dstr, SVt_PVNV);
4048 const char * const type = sv_reftype(sstr,0);
4050 /* diag_listed_as: Bizarre copy of %s */
4051 Perl_croak(aTHX_ "Bizarre copy of %s in %s", type, OP_DESC(PL_op));
4053 Perl_croak(aTHX_ "Bizarre copy of %s", type);
4058 if (dtype < SVt_REGEXP)
4059 sv_upgrade(dstr, SVt_REGEXP);
4062 /* case SVt_BIND: */
4066 if (SvGMAGICAL(sstr) && (flags & SV_GMAGIC)) {
4068 if (SvTYPE(sstr) != stype)
4069 stype = SvTYPE(sstr);
4071 if (isGV_with_GP(sstr) && dtype <= SVt_PVLV) {
4072 glob_assign_glob(dstr, sstr, dtype);
4075 if (stype == SVt_PVLV)
4076 SvUPGRADE(dstr, SVt_PVNV);
4078 SvUPGRADE(dstr, (svtype)stype);
4080 end_of_first_switch:
4082 /* dstr may have been upgraded. */
4083 dtype = SvTYPE(dstr);
4084 sflags = SvFLAGS(sstr);
4086 if (dtype == SVt_PVCV || dtype == SVt_PVFM) {
4087 /* Assigning to a subroutine sets the prototype. */
4090 const char *const ptr = SvPV_const(sstr, len);
4092 SvGROW(dstr, len + 1);
4093 Copy(ptr, SvPVX(dstr), len + 1, char);
4094 SvCUR_set(dstr, len);
4096 SvFLAGS(dstr) |= sflags & SVf_UTF8;
4097 CvAUTOLOAD_off(dstr);
4101 } else if (dtype == SVt_PVAV || dtype == SVt_PVHV) {
4102 const char * const type = sv_reftype(dstr,0);
4104 /* diag_listed_as: Cannot copy to %s */
4105 Perl_croak(aTHX_ "Cannot copy to %s in %s", type, OP_DESC(PL_op));
4107 Perl_croak(aTHX_ "Cannot copy to %s", type);
4108 } else if (sflags & SVf_ROK) {
4109 if (isGV_with_GP(dstr)
4110 && SvTYPE(SvRV(sstr)) == SVt_PVGV && isGV_with_GP(SvRV(sstr))) {
4113 if (GvIMPORTED(dstr) != GVf_IMPORTED
4114 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
4116 GvIMPORTED_on(dstr);
4121 glob_assign_glob(dstr, sstr, dtype);
4125 if (dtype >= SVt_PV) {
4126 if (isGV_with_GP(dstr)) {
4127 glob_assign_ref(dstr, sstr);
4130 if (SvPVX_const(dstr)) {
4136 (void)SvOK_off(dstr);
4137 SvRV_set(dstr, SvREFCNT_inc(SvRV(sstr)));
4138 SvFLAGS(dstr) |= sflags & SVf_ROK;
4139 assert(!(sflags & SVp_NOK));
4140 assert(!(sflags & SVp_IOK));
4141 assert(!(sflags & SVf_NOK));
4142 assert(!(sflags & SVf_IOK));
4144 else if (isGV_with_GP(dstr)) {
4145 if (!(sflags & SVf_OK)) {
4146 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
4147 "Undefined value assigned to typeglob");
4150 GV *gv = gv_fetchsv_nomg(sstr, GV_ADD, SVt_PVGV);
4151 if (dstr != (const SV *)gv) {
4152 const char * const name = GvNAME((const GV *)dstr);
4153 const STRLEN len = GvNAMELEN(dstr);
4154 HV *old_stash = NULL;
4155 bool reset_isa = FALSE;
4156 if ((len > 1 && name[len-2] == ':' && name[len-1] == ':')
4157 || (len == 1 && name[0] == ':')) {
4158 /* Set aside the old stash, so we can reset isa caches
4159 on its subclasses. */
4160 if((old_stash = GvHV(dstr))) {
4161 /* Make sure we do not lose it early. */
4162 SvREFCNT_inc_simple_void_NN(
4163 sv_2mortal((SV *)old_stash)
4170 gp_free(MUTABLE_GV(dstr));
4171 GvGP_set(dstr, gp_ref(GvGP(gv)));
4174 HV * const stash = GvHV(dstr);
4176 old_stash ? (HV *)HvENAME_get(old_stash) : stash
4186 else if (dtype == SVt_REGEXP && stype == SVt_REGEXP) {
4187 reg_temp_copy((REGEXP*)dstr, (REGEXP*)sstr);
4189 else if (sflags & SVp_POK) {
4193 * Check to see if we can just swipe the string. If so, it's a
4194 * possible small lose on short strings, but a big win on long ones.
4195 * It might even be a win on short strings if SvPVX_const(dstr)
4196 * has to be allocated and SvPVX_const(sstr) has to be freed.
4197 * Likewise if we can set up COW rather than doing an actual copy, we
4198 * drop to the else clause, as the swipe code and the COW setup code
4199 * have much in common.
4202 /* Whichever path we take through the next code, we want this true,
4203 and doing it now facilitates the COW check. */
4204 (void)SvPOK_only(dstr);
4207 /* If we're already COW then this clause is not true, and if COW
4208 is allowed then we drop down to the else and make dest COW
4209 with us. If caller hasn't said that we're allowed to COW
4210 shared hash keys then we don't do the COW setup, even if the
4211 source scalar is a shared hash key scalar. */
4212 (((flags & SV_COW_SHARED_HASH_KEYS)
4213 ? (sflags & (SVf_FAKE|SVf_READONLY)) != (SVf_FAKE|SVf_READONLY)
4214 : 1 /* If making a COW copy is forbidden then the behaviour we
4215 desire is as if the source SV isn't actually already
4216 COW, even if it is. So we act as if the source flags
4217 are not COW, rather than actually testing them. */
4219 #ifndef PERL_OLD_COPY_ON_WRITE
4220 /* The change that added SV_COW_SHARED_HASH_KEYS makes the logic
4221 when PERL_OLD_COPY_ON_WRITE is defined a little wrong.
4222 Conceptually PERL_OLD_COPY_ON_WRITE being defined should
4223 override SV_COW_SHARED_HASH_KEYS, because it means "always COW"
4224 but in turn, it's somewhat dead code, never expected to go
4225 live, but more kept as a placeholder on how to do it better
4226 in a newer implementation. */
4227 /* If we are COW and dstr is a suitable target then we drop down
4228 into the else and make dest a COW of us. */
4229 || (SvFLAGS(dstr) & CAN_COW_MASK) != CAN_COW_FLAGS
4234 (sflags & SVs_TEMP) && /* slated for free anyway? */
4235 !(sflags & SVf_OOK) && /* and not involved in OOK hack? */
4236 (!(flags & SV_NOSTEAL)) &&
4237 /* and we're allowed to steal temps */
4238 SvREFCNT(sstr) == 1 && /* and no other references to it? */
4239 SvLEN(sstr)) /* and really is a string */
4240 #ifdef PERL_OLD_COPY_ON_WRITE
4241 && ((flags & SV_COW_SHARED_HASH_KEYS)
4242 ? (!((sflags & CAN_COW_MASK) == CAN_COW_FLAGS
4243 && (SvFLAGS(dstr) & CAN_COW_MASK) == CAN_COW_FLAGS
4244 && SvTYPE(sstr) >= SVt_PVIV && SvTYPE(sstr) != SVt_PVFM))
4248 /* Failed the swipe test, and it's not a shared hash key either.
4249 Have to copy the string. */
4250 STRLEN len = SvCUR(sstr);
4251 SvGROW(dstr, len + 1); /* inlined from sv_setpvn */
4252 Move(SvPVX_const(sstr),SvPVX(dstr),len,char);
4253 SvCUR_set(dstr, len);
4254 *SvEND(dstr) = '\0';
4256 /* If PERL_OLD_COPY_ON_WRITE is not defined, then isSwipe will always
4258 /* Either it's a shared hash key, or it's suitable for
4259 copy-on-write or we can swipe the string. */
4261 PerlIO_printf(Perl_debug_log, "Copy on write: sstr --> dstr\n");
4265 #ifdef PERL_OLD_COPY_ON_WRITE
4267 if ((sflags & (SVf_FAKE | SVf_READONLY))
4268 != (SVf_FAKE | SVf_READONLY)) {
4269 SvREADONLY_on(sstr);
4271 /* Make the source SV into a loop of 1.
4272 (about to become 2) */
4273 SV_COW_NEXT_SV_SET(sstr, sstr);
4277 /* Initial code is common. */
4278 if (SvPVX_const(dstr)) { /* we know that dtype >= SVt_PV */
4283 /* making another shared SV. */
4284 STRLEN cur = SvCUR(sstr);
4285 STRLEN len = SvLEN(sstr);
4286 #ifdef PERL_OLD_COPY_ON_WRITE
4288 assert (SvTYPE(dstr) >= SVt_PVIV);
4289 /* SvIsCOW_normal */
4290 /* splice us in between source and next-after-source. */
4291 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4292 SV_COW_NEXT_SV_SET(sstr, dstr);
4293 SvPV_set(dstr, SvPVX_mutable(sstr));
4297 /* SvIsCOW_shared_hash */
4298 DEBUG_C(PerlIO_printf(Perl_debug_log,
4299 "Copy on write: Sharing hash\n"));
4301 assert (SvTYPE(dstr) >= SVt_PV);
4303 HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)))));
4305 SvLEN_set(dstr, len);
4306 SvCUR_set(dstr, cur);
4307 SvREADONLY_on(dstr);
4311 { /* Passes the swipe test. */
4312 SvPV_set(dstr, SvPVX_mutable(sstr));
4313 SvLEN_set(dstr, SvLEN(sstr));
4314 SvCUR_set(dstr, SvCUR(sstr));
4317 (void)SvOK_off(sstr); /* NOTE: nukes most SvFLAGS on sstr */
4318 SvPV_set(sstr, NULL);
4324 if (sflags & SVp_NOK) {
4325 SvNV_set(dstr, SvNVX(sstr));
4327 if (sflags & SVp_IOK) {
4328 SvIV_set(dstr, SvIVX(sstr));
4329 /* Must do this otherwise some other overloaded use of 0x80000000
4330 gets confused. I guess SVpbm_VALID */
4331 if (sflags & SVf_IVisUV)
4334 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_NOK|SVp_NOK|SVf_UTF8);
4336 const MAGIC * const smg = SvVSTRING_mg(sstr);
4338 sv_magic(dstr, NULL, PERL_MAGIC_vstring,
4339 smg->mg_ptr, smg->mg_len);
4340 SvRMAGICAL_on(dstr);
4344 else if (sflags & (SVp_IOK|SVp_NOK)) {
4345 (void)SvOK_off(dstr);
4346 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_IVisUV|SVf_NOK|SVp_NOK);
4347 if (sflags & SVp_IOK) {
4348 /* XXXX Do we want to set IsUV for IV(ROK)? Be extra safe... */
4349 SvIV_set(dstr, SvIVX(sstr));
4351 if (sflags & SVp_NOK) {
4352 SvNV_set(dstr, SvNVX(sstr));
4356 if (isGV_with_GP(sstr)) {
4357 gv_efullname3(dstr, MUTABLE_GV(sstr), "*");
4360 (void)SvOK_off(dstr);
4362 if (SvTAINTED(sstr))
4367 =for apidoc sv_setsv_mg
4369 Like C<sv_setsv>, but also handles 'set' magic.
4375 Perl_sv_setsv_mg(pTHX_ SV *const dstr, register SV *const sstr)
4377 PERL_ARGS_ASSERT_SV_SETSV_MG;
4379 sv_setsv(dstr,sstr);
4383 #ifdef PERL_OLD_COPY_ON_WRITE
4385 Perl_sv_setsv_cow(pTHX_ SV *dstr, SV *sstr)
4387 STRLEN cur = SvCUR(sstr);
4388 STRLEN len = SvLEN(sstr);
4389 register char *new_pv;
4391 PERL_ARGS_ASSERT_SV_SETSV_COW;
4394 PerlIO_printf(Perl_debug_log, "Fast copy on write: %p -> %p\n",
4395 (void*)sstr, (void*)dstr);
4402 if (SvTHINKFIRST(dstr))
4403 sv_force_normal_flags(dstr, SV_COW_DROP_PV);
4404 else if (SvPVX_const(dstr))
4405 Safefree(SvPVX_const(dstr));
4409 SvUPGRADE(dstr, SVt_PVIV);
4411 assert (SvPOK(sstr));
4412 assert (SvPOKp(sstr));
4413 assert (!SvIOK(sstr));
4414 assert (!SvIOKp(sstr));
4415 assert (!SvNOK(sstr));
4416 assert (!SvNOKp(sstr));
4418 if (SvIsCOW(sstr)) {
4420 if (SvLEN(sstr) == 0) {
4421 /* source is a COW shared hash key. */
4422 DEBUG_C(PerlIO_printf(Perl_debug_log,
4423 "Fast copy on write: Sharing hash\n"));
4424 new_pv = HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr))));
4427 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4429 assert ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS);
4430 SvUPGRADE(sstr, SVt_PVIV);
4431 SvREADONLY_on(sstr);
4433 DEBUG_C(PerlIO_printf(Perl_debug_log,
4434 "Fast copy on write: Converting sstr to COW\n"));
4435 SV_COW_NEXT_SV_SET(dstr, sstr);
4437 SV_COW_NEXT_SV_SET(sstr, dstr);
4438 new_pv = SvPVX_mutable(sstr);
4441 SvPV_set(dstr, new_pv);
4442 SvFLAGS(dstr) = (SVt_PVIV|SVf_POK|SVp_POK|SVf_FAKE|SVf_READONLY);
4445 SvLEN_set(dstr, len);
4446 SvCUR_set(dstr, cur);
4455 =for apidoc sv_setpvn
4457 Copies a string into an SV. The C<len> parameter indicates the number of
4458 bytes to be copied. If the C<ptr> argument is NULL the SV will become
4459 undefined. Does not handle 'set' magic. See C<sv_setpvn_mg>.
4465 Perl_sv_setpvn(pTHX_ register SV *const sv, register const char *const ptr, register const STRLEN len)
4468 register char *dptr;
4470 PERL_ARGS_ASSERT_SV_SETPVN;
4472 SV_CHECK_THINKFIRST_COW_DROP(sv);
4478 /* len is STRLEN which is unsigned, need to copy to signed */
4481 Perl_croak(aTHX_ "panic: sv_setpvn called with negative strlen");
4483 SvUPGRADE(sv, SVt_PV);
4485 dptr = SvGROW(sv, len + 1);
4486 Move(ptr,dptr,len,char);
4489 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4491 if (SvTYPE(sv) == SVt_PVCV) CvAUTOLOAD_off(sv);
4495 =for apidoc sv_setpvn_mg
4497 Like C<sv_setpvn>, but also handles 'set' magic.
4503 Perl_sv_setpvn_mg(pTHX_ register SV *const sv, register const char *const ptr, register const STRLEN len)
4505 PERL_ARGS_ASSERT_SV_SETPVN_MG;
4507 sv_setpvn(sv,ptr,len);
4512 =for apidoc sv_setpv
4514 Copies a string into an SV. The string must be null-terminated. Does not
4515 handle 'set' magic. See C<sv_setpv_mg>.
4521 Perl_sv_setpv(pTHX_ register SV *const sv, register const char *const ptr)
4524 register STRLEN len;
4526 PERL_ARGS_ASSERT_SV_SETPV;
4528 SV_CHECK_THINKFIRST_COW_DROP(sv);
4534 SvUPGRADE(sv, SVt_PV);
4536 SvGROW(sv, len + 1);
4537 Move(ptr,SvPVX(sv),len+1,char);
4539 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4541 if (SvTYPE(sv) == SVt_PVCV) CvAUTOLOAD_off(sv);
4545 =for apidoc sv_setpv_mg
4547 Like C<sv_setpv>, but also handles 'set' magic.
4553 Perl_sv_setpv_mg(pTHX_ register SV *const sv, register const char *const ptr)
4555 PERL_ARGS_ASSERT_SV_SETPV_MG;
4562 Perl_sv_sethek(pTHX_ register SV *const sv, const HEK *const hek)
4566 PERL_ARGS_ASSERT_SV_SETHEK;
4572 if (HEK_LEN(hek) == HEf_SVKEY) {
4573 sv_setsv(sv, *(SV**)HEK_KEY(hek));
4576 const int flags = HEK_FLAGS(hek);
4577 if (flags & HVhek_WASUTF8) {
4578 STRLEN utf8_len = HEK_LEN(hek);
4579 char *as_utf8 = (char *)bytes_to_utf8((U8*)HEK_KEY(hek), &utf8_len);
4580 sv_usepvn_flags(sv, as_utf8, utf8_len, SV_HAS_TRAILING_NUL);
4583 } else if (flags & (HVhek_REHASH|HVhek_UNSHARED)) {
4584 sv_setpvn(sv, HEK_KEY(hek), HEK_LEN(hek));
4587 else SvUTF8_off(sv);
4591 SV_CHECK_THINKFIRST_COW_DROP(sv);
4592 SvUPGRADE(sv, SVt_PV);
4593 SvPV_set(sv,(char *)HEK_KEY(share_hek_hek(hek)));
4594 SvCUR_set(sv, HEK_LEN(hek));
4601 else SvUTF8_off(sv);
4609 =for apidoc sv_usepvn_flags
4611 Tells an SV to use C<ptr> to find its string value. Normally the
4612 string is stored inside the SV but sv_usepvn allows the SV to use an
4613 outside string. The C<ptr> should point to memory that was allocated
4614 by C<malloc>. It must be the start of a mallocked block
4615 of memory, and not a pointer to the middle of it. The
4616 string length, C<len>, must be supplied. By default
4617 this function will realloc (i.e. move) the memory pointed to by C<ptr>,
4618 so that pointer should not be freed or used by the programmer after
4619 giving it to sv_usepvn, and neither should any pointers from "behind"
4620 that pointer (e.g. ptr + 1) be used.
4622 If C<flags> & SV_SMAGIC is true, will call SvSETMAGIC. If C<flags> &
4623 SV_HAS_TRAILING_NUL is true, then C<ptr[len]> must be NUL, and the realloc
4624 will be skipped (i.e. the buffer is actually at least 1 byte longer than
4625 C<len>, and already meets the requirements for storing in C<SvPVX>).
4631 Perl_sv_usepvn_flags(pTHX_ SV *const sv, char *ptr, const STRLEN len, const U32 flags)
4636 PERL_ARGS_ASSERT_SV_USEPVN_FLAGS;
4638 SV_CHECK_THINKFIRST_COW_DROP(sv);
4639 SvUPGRADE(sv, SVt_PV);
4642 if (flags & SV_SMAGIC)
4646 if (SvPVX_const(sv))
4650 if (flags & SV_HAS_TRAILING_NUL)
4651 assert(ptr[len] == '\0');
4654 allocate = (flags & SV_HAS_TRAILING_NUL)
4656 #ifdef Perl_safesysmalloc_size
4659 PERL_STRLEN_ROUNDUP(len + 1);
4661 if (flags & SV_HAS_TRAILING_NUL) {
4662 /* It's long enough - do nothing.
4663 Specifically Perl_newCONSTSUB is relying on this. */
4666 /* Force a move to shake out bugs in callers. */
4667 char *new_ptr = (char*)safemalloc(allocate);
4668 Copy(ptr, new_ptr, len, char);
4669 PoisonFree(ptr,len,char);
4673 ptr = (char*) saferealloc (ptr, allocate);
4676 #ifdef Perl_safesysmalloc_size
4677 SvLEN_set(sv, Perl_safesysmalloc_size(ptr));
4679 SvLEN_set(sv, allocate);
4683 if (!(flags & SV_HAS_TRAILING_NUL)) {
4686 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4688 if (flags & SV_SMAGIC)
4692 #ifdef PERL_OLD_COPY_ON_WRITE
4693 /* Need to do this *after* making the SV normal, as we need the buffer
4694 pointer to remain valid until after we've copied it. If we let go too early,
4695 another thread could invalidate it by unsharing last of the same hash key
4696 (which it can do by means other than releasing copy-on-write Svs)
4697 or by changing the other copy-on-write SVs in the loop. */
4699 S_sv_release_COW(pTHX_ register SV *sv, const char *pvx, SV *after)
4701 PERL_ARGS_ASSERT_SV_RELEASE_COW;
4703 { /* this SV was SvIsCOW_normal(sv) */
4704 /* we need to find the SV pointing to us. */
4705 SV *current = SV_COW_NEXT_SV(after);
4707 if (current == sv) {
4708 /* The SV we point to points back to us (there were only two of us
4710 Hence other SV is no longer copy on write either. */
4712 SvREADONLY_off(after);
4714 /* We need to follow the pointers around the loop. */
4716 while ((next = SV_COW_NEXT_SV(current)) != sv) {
4719 /* don't loop forever if the structure is bust, and we have
4720 a pointer into a closed loop. */
4721 assert (current != after);
4722 assert (SvPVX_const(current) == pvx);
4724 /* Make the SV before us point to the SV after us. */
4725 SV_COW_NEXT_SV_SET(current, after);
4731 =for apidoc sv_force_normal_flags
4733 Undo various types of fakery on an SV: if the PV is a shared string, make
4734 a private copy; if we're a ref, stop refing; if we're a glob, downgrade to
4735 an xpvmg; if we're a copy-on-write scalar, this is the on-write time when
4736 we do the copy, and is also used locally. If C<SV_COW_DROP_PV> is set
4737 then a copy-on-write scalar drops its PV buffer (if any) and becomes
4738 SvPOK_off rather than making a copy. (Used where this
4739 scalar is about to be set to some other value.) In addition,
4740 the C<flags> parameter gets passed to C<sv_unref_flags()>
4741 when unreffing. C<sv_force_normal> calls this function
4742 with flags set to 0.
4748 Perl_sv_force_normal_flags(pTHX_ register SV *const sv, const U32 flags)
4752 PERL_ARGS_ASSERT_SV_FORCE_NORMAL_FLAGS;
4754 #ifdef PERL_OLD_COPY_ON_WRITE
4755 if (SvREADONLY(sv)) {
4757 const char * const pvx = SvPVX_const(sv);
4758 const STRLEN len = SvLEN(sv);
4759 const STRLEN cur = SvCUR(sv);
4760 /* next COW sv in the loop. If len is 0 then this is a shared-hash
4761 key scalar, so we mustn't attempt to call SV_COW_NEXT_SV(), as
4762 we'll fail an assertion. */
4763 SV * const next = len ? SV_COW_NEXT_SV(sv) : 0;
4766 PerlIO_printf(Perl_debug_log,
4767 "Copy on write: Force normal %ld\n",
4773 /* This SV doesn't own the buffer, so need to Newx() a new one: */
4776 if (flags & SV_COW_DROP_PV) {
4777 /* OK, so we don't need to copy our buffer. */
4780 SvGROW(sv, cur + 1);
4781 Move(pvx,SvPVX(sv),cur,char);
4786 sv_release_COW(sv, pvx, next);
4788 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4794 else if (IN_PERL_RUNTIME)
4795 Perl_croak_no_modify(aTHX);
4798 if (SvREADONLY(sv)) {
4799 if (SvFAKE(sv) && !isGV_with_GP(sv)) {
4800 const char * const pvx = SvPVX_const(sv);
4801 const STRLEN len = SvCUR(sv);
4806 if (flags & SV_COW_DROP_PV) {
4807 /* OK, so we don't need to copy our buffer. */
4810 SvGROW(sv, len + 1);
4811 Move(pvx,SvPVX(sv),len,char);
4814 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4816 else if (IN_PERL_RUNTIME)
4817 Perl_croak_no_modify(aTHX);
4821 sv_unref_flags(sv, flags);
4822 else if (SvFAKE(sv) && isGV_with_GP(sv))
4823 sv_unglob(sv, flags);
4824 else if (SvFAKE(sv) && SvTYPE(sv) == SVt_REGEXP) {
4825 /* Need to downgrade the REGEXP to a simple(r) scalar. This is analogous
4826 to sv_unglob. We only need it here, so inline it. */
4827 const svtype new_type = SvMAGIC(sv) || SvSTASH(sv) ? SVt_PVMG : SVt_PV;
4828 SV *const temp = newSV_type(new_type);
4829 void *const temp_p = SvANY(sv);
4831 if (new_type == SVt_PVMG) {
4832 SvMAGIC_set(temp, SvMAGIC(sv));
4833 SvMAGIC_set(sv, NULL);
4834 SvSTASH_set(temp, SvSTASH(sv));
4835 SvSTASH_set(sv, NULL);
4837 SvCUR_set(temp, SvCUR(sv));
4838 /* Remember that SvPVX is in the head, not the body. */
4840 SvLEN_set(temp, SvLEN(sv));
4841 /* This signals "buffer is owned by someone else" in sv_clear,
4842 which is the least effort way to stop it freeing the buffer.
4844 SvLEN_set(sv, SvLEN(sv)+1);
4846 /* Their buffer is already owned by someone else. */
4847 SvPVX(sv) = savepvn(SvPVX(sv), SvCUR(sv));
4848 SvLEN_set(temp, SvCUR(sv)+1);
4851 /* Now swap the rest of the bodies. */
4853 SvFLAGS(sv) &= ~(SVf_FAKE|SVTYPEMASK);
4854 SvFLAGS(sv) |= new_type;
4855 SvANY(sv) = SvANY(temp);
4857 SvFLAGS(temp) &= ~(SVTYPEMASK);
4858 SvFLAGS(temp) |= SVt_REGEXP|SVf_FAKE;
4859 SvANY(temp) = temp_p;
4868 Efficient removal of characters from the beginning of the string buffer.
4869 SvPOK(sv) must be true and the C<ptr> must be a pointer to somewhere inside
4870 the string buffer. The C<ptr> becomes the first character of the adjusted
4871 string. Uses the "OOK hack".
4873 Beware: after this function returns, C<ptr> and SvPVX_const(sv) may no longer
4874 refer to the same chunk of data.
4876 The unfortunate similarity of this function's name to that of Perl's C<chop>
4877 operator is strictly coincidental. This function works from the left;
4878 C<chop> works from the right.
4884 Perl_sv_chop(pTHX_ register SV *const sv, register const char *const ptr)
4895 PERL_ARGS_ASSERT_SV_CHOP;
4897 if (!ptr || !SvPOKp(sv))
4899 delta = ptr - SvPVX_const(sv);
4901 /* Nothing to do. */
4904 max_delta = SvLEN(sv) ? SvLEN(sv) : SvCUR(sv);
4905 if (delta > max_delta)
4906 Perl_croak(aTHX_ "panic: sv_chop ptr=%p, start=%p, end=%p",
4907 ptr, SvPVX_const(sv), SvPVX_const(sv) + max_delta);
4908 /* SvPVX(sv) may move in SV_CHECK_THINKFIRST(sv), so don't use ptr any more */
4909 SV_CHECK_THINKFIRST(sv);
4912 if (!SvLEN(sv)) { /* make copy of shared string */
4913 const char *pvx = SvPVX_const(sv);
4914 const STRLEN len = SvCUR(sv);
4915 SvGROW(sv, len + 1);
4916 Move(pvx,SvPVX(sv),len,char);
4922 SvOOK_offset(sv, old_delta);
4924 SvLEN_set(sv, SvLEN(sv) - delta);
4925 SvCUR_set(sv, SvCUR(sv) - delta);
4926 SvPV_set(sv, SvPVX(sv) + delta);
4928 p = (U8 *)SvPVX_const(sv);
4931 /* how many bytes were evacuated? we will fill them with sentinel
4932 bytes, except for the part holding the new offset of course. */
4935 evacn += (old_delta < 0x100 ? 1 : 1 + sizeof(STRLEN));
4937 assert(evacn <= delta + old_delta);
4943 if (delta < 0x100) {
4947 p -= sizeof(STRLEN);
4948 Copy((U8*)&delta, p, sizeof(STRLEN), U8);
4952 /* Fill the preceding buffer with sentinals to verify that no-one is
4962 =for apidoc sv_catpvn
4964 Concatenates the string onto the end of the string which is in the SV. The
4965 C<len> indicates number of bytes to copy. If the SV has the UTF-8
4966 status set, then the bytes appended should be valid UTF-8.
4967 Handles 'get' magic, but not 'set' magic. See C<sv_catpvn_mg>.
4969 =for apidoc sv_catpvn_flags
4971 Concatenates the string onto the end of the string which is in the SV. The
4972 C<len> indicates number of bytes to copy. If the SV has the UTF-8
4973 status set, then the bytes appended should be valid UTF-8.
4974 If C<flags> has the C<SV_SMAGIC> bit set, will
4975 C<mg_set> on C<dsv> afterwards if appropriate.
4976 C<sv_catpvn> and C<sv_catpvn_nomg> are implemented
4977 in terms of this function.
4983 Perl_sv_catpvn_flags(pTHX_ register SV *const dsv, register const char *sstr, register const STRLEN slen, const I32 flags)
4987 const char * const dstr = SvPV_force_flags(dsv, dlen, flags);
4989 PERL_ARGS_ASSERT_SV_CATPVN_FLAGS;
4990 assert((flags & (SV_CATBYTES|SV_CATUTF8)) != (SV_CATBYTES|SV_CATUTF8));
4992 if (!(flags & SV_CATBYTES) || !SvUTF8(dsv)) {
4993 if (flags & SV_CATUTF8 && !SvUTF8(dsv)) {
4994 sv_utf8_upgrade_flags_grow(dsv, 0, slen + 1);
4997 else SvGROW(dsv, dlen + slen + 1);
4999 sstr = SvPVX_const(dsv);
5000 Move(sstr, SvPVX(dsv) + dlen, slen, char);
5001 SvCUR_set(dsv, SvCUR(dsv) + slen);
5004 /* We inline bytes_to_utf8, to avoid an extra malloc. */
5005 const char * const send = sstr + slen;
5008 /* Something this code does not account for, which I think is
5009 impossible; it would require the same pv to be treated as
5010 bytes *and* utf8, which would indicate a bug elsewhere. */
5011 assert(sstr != dstr);
5013 SvGROW(dsv, dlen + slen * 2 + 1);
5014 d = (U8 *)SvPVX(dsv) + dlen;
5016 while (sstr < send) {
5017 const UV uv = NATIVE_TO_ASCII((U8)*sstr++);
5018 if (UNI_IS_INVARIANT(uv))
5019 *d++ = (U8)UTF_TO_NATIVE(uv);
5021 *d++ = (U8)UTF8_EIGHT_BIT_HI(uv);
5022 *d++ = (U8)UTF8_EIGHT_BIT_LO(uv);
5025 SvCUR_set(dsv, d-(const U8 *)SvPVX(dsv));
5028 (void)SvPOK_only_UTF8(dsv); /* validate pointer */
5030 if (flags & SV_SMAGIC)
5035 =for apidoc sv_catsv
5037 Concatenates the string from SV C<ssv> onto the end of the string in
5038 SV C<dsv>. Modifies C<dsv> but not C<ssv>. Handles 'get' magic, but
5039 not 'set' magic. See C<sv_catsv_mg>.
5041 =for apidoc sv_catsv_flags
5043 Concatenates the string from SV C<ssv> onto the end of the string in
5044 SV C<dsv>. Modifies C<dsv> but not C<ssv>. If C<flags> has C<SV_GMAGIC>
5045 bit set, will C<mg_get> on the C<ssv>, if appropriate, before
5046 reading it. If the C<flags> contain C<SV_SMAGIC>, C<mg_set> will be
5047 called on the modified SV afterward, if appropriate. C<sv_catsv>
5048 and C<sv_catsv_nomg> are implemented in terms of this function.
5053 Perl_sv_catsv_flags(pTHX_ SV *const dsv, register SV *const ssv, const I32 flags)
5057 PERL_ARGS_ASSERT_SV_CATSV_FLAGS;
5061 const char *spv = SvPV_flags_const(ssv, slen, flags);
5063 if (SvGMAGICAL(dsv) && (flags & SV_GMAGIC))
5065 sv_catpvn_flags(dsv, spv, slen,
5066 DO_UTF8(ssv) ? SV_CATUTF8 : SV_CATBYTES);
5069 if (flags & SV_SMAGIC)
5074 =for apidoc sv_catpv
5076 Concatenates the string onto the end of the string which is in the SV.
5077 If the SV has the UTF-8 status set, then the bytes appended should be
5078 valid UTF-8. Handles 'get' magic, but not 'set' magic. See C<sv_catpv_mg>.
5083 Perl_sv_catpv(pTHX_ register SV *const sv, register const char *ptr)
5086 register STRLEN len;
5090 PERL_ARGS_ASSERT_SV_CATPV;
5094 junk = SvPV_force(sv, tlen);
5096 SvGROW(sv, tlen + len + 1);
5098 ptr = SvPVX_const(sv);
5099 Move(ptr,SvPVX(sv)+tlen,len+1,char);
5100 SvCUR_set(sv, SvCUR(sv) + len);
5101 (void)SvPOK_only_UTF8(sv); /* validate pointer */
5106 =for apidoc sv_catpv_flags
5108 Concatenates the string onto the end of the string which is in the SV.
5109 If the SV has the UTF-8 status set, then the bytes appended should
5110 be valid UTF-8. If C<flags> has the C<SV_SMAGIC> bit set, will C<mg_set>
5111 on the modified SV if appropriate.
5117 Perl_sv_catpv_flags(pTHX_ SV *dstr, const char *sstr, const I32 flags)
5119 PERL_ARGS_ASSERT_SV_CATPV_FLAGS;
5120 sv_catpvn_flags(dstr, sstr, strlen(sstr), flags);
5124 =for apidoc sv_catpv_mg
5126 Like C<sv_catpv>, but also handles 'set' magic.
5132 Perl_sv_catpv_mg(pTHX_ register SV *const sv, register const char *const ptr)
5134 PERL_ARGS_ASSERT_SV_CATPV_MG;
5143 Creates a new SV. A non-zero C<len> parameter indicates the number of
5144 bytes of preallocated string space the SV should have. An extra byte for a
5145 trailing NUL is also reserved. (SvPOK is not set for the SV even if string
5146 space is allocated.) The reference count for the new SV is set to 1.
5148 In 5.9.3, newSV() replaces the older NEWSV() API, and drops the first
5149 parameter, I<x>, a debug aid which allowed callers to identify themselves.
5150 This aid has been superseded by a new build option, PERL_MEM_LOG (see
5151 L<perlhacktips/PERL_MEM_LOG>). The older API is still there for use in XS
5152 modules supporting older perls.
5158 Perl_newSV(pTHX_ const STRLEN len)
5165 sv_upgrade(sv, SVt_PV);
5166 SvGROW(sv, len + 1);
5171 =for apidoc sv_magicext
5173 Adds magic to an SV, upgrading it if necessary. Applies the
5174 supplied vtable and returns a pointer to the magic added.
5176 Note that C<sv_magicext> will allow things that C<sv_magic> will not.
5177 In particular, you can add magic to SvREADONLY SVs, and add more than
5178 one instance of the same 'how'.
5180 If C<namlen> is greater than zero then a C<savepvn> I<copy> of C<name> is
5181 stored, if C<namlen> is zero then C<name> is stored as-is and - as another
5182 special case - if C<(name && namlen == HEf_SVKEY)> then C<name> is assumed
5183 to contain an C<SV*> and is stored as-is with its REFCNT incremented.
5185 (This is now used as a subroutine by C<sv_magic>.)
5190 Perl_sv_magicext(pTHX_ SV *const sv, SV *const obj, const int how,
5191 const MGVTBL *const vtable, const char *const name, const I32 namlen)
5196 PERL_ARGS_ASSERT_SV_MAGICEXT;
5198 SvUPGRADE(sv, SVt_PVMG);
5199 Newxz(mg, 1, MAGIC);
5200 mg->mg_moremagic = SvMAGIC(sv);
5201 SvMAGIC_set(sv, mg);
5203 /* Sometimes a magic contains a reference loop, where the sv and
5204 object refer to each other. To prevent a reference loop that
5205 would prevent such objects being freed, we look for such loops
5206 and if we find one we avoid incrementing the object refcount.
5208 Note we cannot do this to avoid self-tie loops as intervening RV must
5209 have its REFCNT incremented to keep it in existence.
5212 if (!obj || obj == sv ||
5213 how == PERL_MAGIC_arylen ||
5214 how == PERL_MAGIC_symtab ||
5215 (SvTYPE(obj) == SVt_PVGV &&
5216 (GvSV(obj) == sv || GvHV(obj) == (const HV *)sv
5217 || GvAV(obj) == (const AV *)sv || GvCV(obj) == (const CV *)sv
5218 || GvIOp(obj) == (const IO *)sv || GvFORM(obj) == (const CV *)sv)))
5223 mg->mg_obj = SvREFCNT_inc_simple(obj);
5224 mg->mg_flags |= MGf_REFCOUNTED;
5227 /* Normal self-ties simply pass a null object, and instead of
5228 using mg_obj directly, use the SvTIED_obj macro to produce a
5229 new RV as needed. For glob "self-ties", we are tieing the PVIO
5230 with an RV obj pointing to the glob containing the PVIO. In
5231 this case, to avoid a reference loop, we need to weaken the
5235 if (how == PERL_MAGIC_tiedscalar && SvTYPE(sv) == SVt_PVIO &&
5236 obj && SvROK(obj) && GvIO(SvRV(obj)) == (const IO *)sv)
5242 mg->mg_len = namlen;
5245 mg->mg_ptr = savepvn(name, namlen);
5246 else if (namlen == HEf_SVKEY) {
5247 /* Yes, this is casting away const. This is only for the case of
5248 HEf_SVKEY. I think we need to document this aberation of the
5249 constness of the API, rather than making name non-const, as
5250 that change propagating outwards a long way. */
5251 mg->mg_ptr = (char*)SvREFCNT_inc_simple_NN((SV *)name);
5253 mg->mg_ptr = (char *) name;
5255 mg->mg_virtual = (MGVTBL *) vtable;
5259 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
5264 =for apidoc sv_magic
5266 Adds magic to an SV. First upgrades C<sv> to type C<SVt_PVMG> if
5267 necessary, then adds a new magic item of type C<how> to the head of the
5270 See C<sv_magicext> (which C<sv_magic> now calls) for a description of the
5271 handling of the C<name> and C<namlen> arguments.
5273 You need to use C<sv_magicext> to add magic to SvREADONLY SVs and also
5274 to add more than one instance of the same 'how'.
5280 Perl_sv_magic(pTHX_ register SV *const sv, SV *const obj, const int how,
5281 const char *const name, const I32 namlen)
5284 const MGVTBL *vtable;
5287 unsigned int vtable_index;
5289 PERL_ARGS_ASSERT_SV_MAGIC;
5291 if (how < 0 || (unsigned)how > C_ARRAY_LENGTH(PL_magic_data)
5292 || ((flags = PL_magic_data[how]),
5293 (vtable_index = flags & PERL_MAGIC_VTABLE_MASK)
5294 > magic_vtable_max))
5295 Perl_croak(aTHX_ "Don't know how to handle magic of type \\%o", how);
5297 /* PERL_MAGIC_ext is reserved for use by extensions not perl internals.
5298 Useful for attaching extension internal data to perl vars.
5299 Note that multiple extensions may clash if magical scalars
5300 etc holding private data from one are passed to another. */
5302 vtable = (vtable_index == magic_vtable_max)
5303 ? NULL : PL_magic_vtables + vtable_index;
5305 #ifdef PERL_OLD_COPY_ON_WRITE
5307 sv_force_normal_flags(sv, 0);
5309 if (SvREADONLY(sv)) {
5311 /* its okay to attach magic to shared strings */
5312 (!SvFAKE(sv) || isGV_with_GP(sv))
5315 && !PERL_MAGIC_TYPE_READONLY_ACCEPTABLE(how)
5318 Perl_croak_no_modify(aTHX);
5321 if (SvMAGICAL(sv) || (how == PERL_MAGIC_taint && SvTYPE(sv) >= SVt_PVMG)) {
5322 if (SvMAGIC(sv) && (mg = mg_find(sv, how))) {
5323 /* sv_magic() refuses to add a magic of the same 'how' as an
5326 if (how == PERL_MAGIC_taint) {
5328 /* Any scalar which already had taint magic on which someone
5329 (erroneously?) did SvIOK_on() or similar will now be
5330 incorrectly sporting public "OK" flags. */
5331 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
5337 /* Rest of work is done else where */
5338 mg = sv_magicext(sv,obj,how,vtable,name,namlen);
5341 case PERL_MAGIC_taint:
5344 case PERL_MAGIC_ext:
5345 case PERL_MAGIC_dbfile:
5352 S_sv_unmagicext_flags(pTHX_ SV *const sv, const int type, MGVTBL *vtbl, const U32 flags)
5359 if (SvTYPE(sv) < SVt_PVMG || !SvMAGIC(sv))
5361 mgp = &(((XPVMG*) SvANY(sv))->xmg_u.xmg_magic);
5362 for (mg = *mgp; mg; mg = *mgp) {
5363 const MGVTBL* const virt = mg->mg_virtual;
5364 if (mg->mg_type == type && (!flags || virt == vtbl)) {
5365 *mgp = mg->mg_moremagic;
5366 if (virt && virt->svt_free)
5367 virt->svt_free(aTHX_ sv, mg);
5368 if (mg->mg_ptr && mg->mg_type != PERL_MAGIC_regex_global) {
5370 Safefree(mg->mg_ptr);
5371 else if (mg->mg_len == HEf_SVKEY)
5372 SvREFCNT_dec(MUTABLE_SV(mg->mg_ptr));
5373 else if (mg->mg_type == PERL_MAGIC_utf8)
5374 Safefree(mg->mg_ptr);
5376 if (mg->mg_flags & MGf_REFCOUNTED)
5377 SvREFCNT_dec(mg->mg_obj);
5381 mgp = &mg->mg_moremagic;
5384 if (SvMAGICAL(sv)) /* if we're under save_magic, wait for restore_magic; */
5385 mg_magical(sv); /* else fix the flags now */
5389 SvFLAGS(sv) |= (SvFLAGS(sv) & (SVp_IOK|SVp_NOK|SVp_POK)) >> PRIVSHIFT;
5395 =for apidoc sv_unmagic
5397 Removes all magic of type C<type> from an SV.
5403 Perl_sv_unmagic(pTHX_ SV *const sv, const int type)
5405 PERL_ARGS_ASSERT_SV_UNMAGIC;
5406 return S_sv_unmagicext_flags(aTHX_ sv, type, NULL, 0);
5410 =for apidoc sv_unmagicext
5412 Removes all magic of type C<type> with the specified C<vtbl> from an SV.
5418 Perl_sv_unmagicext(pTHX_ SV *const sv, const int type, MGVTBL *vtbl)
5420 PERL_ARGS_ASSERT_SV_UNMAGICEXT;
5421 return S_sv_unmagicext_flags(aTHX_ sv, type, vtbl, 1);
5425 =for apidoc sv_rvweaken
5427 Weaken a reference: set the C<SvWEAKREF> flag on this RV; give the
5428 referred-to SV C<PERL_MAGIC_backref> magic if it hasn't already; and
5429 push a back-reference to this RV onto the array of backreferences
5430 associated with that magic. If the RV is magical, set magic will be
5431 called after the RV is cleared.
5437 Perl_sv_rvweaken(pTHX_ SV *const sv)
5441 PERL_ARGS_ASSERT_SV_RVWEAKEN;
5443 if (!SvOK(sv)) /* let undefs pass */
5446 Perl_croak(aTHX_ "Can't weaken a nonreference");
5447 else if (SvWEAKREF(sv)) {
5448 Perl_ck_warner(aTHX_ packWARN(WARN_MISC), "Reference is already weak");
5451 else if (SvREADONLY(sv)) croak_no_modify();
5453 Perl_sv_add_backref(aTHX_ tsv, sv);
5459 /* Give tsv backref magic if it hasn't already got it, then push a
5460 * back-reference to sv onto the array associated with the backref magic.
5462 * As an optimisation, if there's only one backref and it's not an AV,
5463 * store it directly in the HvAUX or mg_obj slot, avoiding the need to
5464 * allocate an AV. (Whether the slot holds an AV tells us whether this is
5468 /* A discussion about the backreferences array and its refcount:
5470 * The AV holding the backreferences is pointed to either as the mg_obj of
5471 * PERL_MAGIC_backref, or in the specific case of a HV, from the
5472 * xhv_backreferences field. The array is created with a refcount
5473 * of 2. This means that if during global destruction the array gets
5474 * picked on before its parent to have its refcount decremented by the
5475 * random zapper, it won't actually be freed, meaning it's still there for
5476 * when its parent gets freed.
5478 * When the parent SV is freed, the extra ref is killed by
5479 * Perl_sv_kill_backrefs. The other ref is killed, in the case of magic,
5480 * by mg_free() / MGf_REFCOUNTED, or for a hash, by Perl_hv_kill_backrefs.
5482 * When a single backref SV is stored directly, it is not reference
5487 Perl_sv_add_backref(pTHX_ SV *const tsv, SV *const sv)
5494 PERL_ARGS_ASSERT_SV_ADD_BACKREF;
5496 /* find slot to store array or singleton backref */
5498 if (SvTYPE(tsv) == SVt_PVHV) {
5499 svp = (SV**)Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5502 (SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL))))
5504 sv_magic(tsv, NULL, PERL_MAGIC_backref, NULL, 0);
5505 mg = mg_find(tsv, PERL_MAGIC_backref);
5507 svp = &(mg->mg_obj);
5510 /* create or retrieve the array */
5512 if ( (!*svp && SvTYPE(sv) == SVt_PVAV)
5513 || (*svp && SvTYPE(*svp) != SVt_PVAV)
5518 SvREFCNT_inc_simple_void(av);
5519 /* av now has a refcnt of 2; see discussion above */
5521 /* move single existing backref to the array */
5523 AvARRAY(av)[++AvFILLp(av)] = *svp; /* av_push() */
5527 mg->mg_flags |= MGf_REFCOUNTED;
5530 av = MUTABLE_AV(*svp);
5533 /* optimisation: store single backref directly in HvAUX or mg_obj */
5537 /* push new backref */
5538 assert(SvTYPE(av) == SVt_PVAV);
5539 if (AvFILLp(av) >= AvMAX(av)) {
5540 av_extend(av, AvFILLp(av)+1);
5542 AvARRAY(av)[++AvFILLp(av)] = sv; /* av_push() */
5545 /* delete a back-reference to ourselves from the backref magic associated
5546 * with the SV we point to.
5550 Perl_sv_del_backref(pTHX_ SV *const tsv, SV *const sv)
5555 PERL_ARGS_ASSERT_SV_DEL_BACKREF;
5557 if (SvTYPE(tsv) == SVt_PVHV) {
5559 svp = (SV**)Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5563 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
5564 svp = mg ? &(mg->mg_obj) : NULL;
5568 Perl_croak(aTHX_ "panic: del_backref");
5570 if (SvTYPE(*svp) == SVt_PVAV) {
5574 AV * const av = (AV*)*svp;
5576 assert(!SvIS_FREED(av));
5580 /* for an SV with N weak references to it, if all those
5581 * weak refs are deleted, then sv_del_backref will be called
5582 * N times and O(N^2) compares will be done within the backref
5583 * array. To ameliorate this potential slowness, we:
5584 * 1) make sure this code is as tight as possible;
5585 * 2) when looking for SV, look for it at both the head and tail of the
5586 * array first before searching the rest, since some create/destroy
5587 * patterns will cause the backrefs to be freed in order.
5594 SV **p = &svp[fill];
5595 SV *const topsv = *p;
5602 /* We weren't the last entry.
5603 An unordered list has this property that you
5604 can take the last element off the end to fill
5605 the hole, and it's still an unordered list :-)
5611 break; /* should only be one */
5618 AvFILLp(av) = fill-1;
5621 /* optimisation: only a single backref, stored directly */
5623 Perl_croak(aTHX_ "panic: del_backref");
5630 Perl_sv_kill_backrefs(pTHX_ SV *const sv, AV *const av)
5636 PERL_ARGS_ASSERT_SV_KILL_BACKREFS;
5641 /* after multiple passes through Perl_sv_clean_all() for a thinngy
5642 * that has badly leaked, the backref array may have gotten freed,
5643 * since we only protect it against 1 round of cleanup */
5644 if (SvIS_FREED(av)) {
5645 if (PL_in_clean_all) /* All is fair */
5648 "panic: magic_killbackrefs (freed backref AV/SV)");
5652 is_array = (SvTYPE(av) == SVt_PVAV);
5654 assert(!SvIS_FREED(av));
5657 last = svp + AvFILLp(av);
5660 /* optimisation: only a single backref, stored directly */
5666 while (svp <= last) {
5668 SV *const referrer = *svp;
5669 if (SvWEAKREF(referrer)) {
5670 /* XXX Should we check that it hasn't changed? */
5671 assert(SvROK(referrer));
5672 SvRV_set(referrer, 0);
5674 SvWEAKREF_off(referrer);
5675 SvSETMAGIC(referrer);
5676 } else if (SvTYPE(referrer) == SVt_PVGV ||
5677 SvTYPE(referrer) == SVt_PVLV) {
5678 assert(SvTYPE(sv) == SVt_PVHV); /* stash backref */
5679 /* You lookin' at me? */
5680 assert(GvSTASH(referrer));
5681 assert(GvSTASH(referrer) == (const HV *)sv);
5682 GvSTASH(referrer) = 0;
5683 } else if (SvTYPE(referrer) == SVt_PVCV ||
5684 SvTYPE(referrer) == SVt_PVFM) {
5685 if (SvTYPE(sv) == SVt_PVHV) { /* stash backref */
5686 /* You lookin' at me? */
5687 assert(CvSTASH(referrer));
5688 assert(CvSTASH(referrer) == (const HV *)sv);
5689 SvANY(MUTABLE_CV(referrer))->xcv_stash = 0;
5692 assert(SvTYPE(sv) == SVt_PVGV);
5693 /* You lookin' at me? */
5694 assert(CvGV(referrer));
5695 assert(CvGV(referrer) == (const GV *)sv);
5696 anonymise_cv_maybe(MUTABLE_GV(sv),
5697 MUTABLE_CV(referrer));
5702 "panic: magic_killbackrefs (flags=%"UVxf")",
5703 (UV)SvFLAGS(referrer));
5714 SvREFCNT_dec(av); /* remove extra count added by sv_add_backref() */
5720 =for apidoc sv_insert
5722 Inserts a string at the specified offset/length within the SV. Similar to
5723 the Perl substr() function. Handles get magic.
5725 =for apidoc sv_insert_flags
5727 Same as C<sv_insert>, but the extra C<flags> are passed to the
5728 C<SvPV_force_flags> that applies to C<bigstr>.
5734 Perl_sv_insert_flags(pTHX_ SV *const bigstr, const STRLEN offset, const STRLEN len, const char *const little, const STRLEN littlelen, const U32 flags)
5739 register char *midend;
5740 register char *bigend;
5741 register SSize_t i; /* better be sizeof(STRLEN) or bad things happen */
5744 PERL_ARGS_ASSERT_SV_INSERT_FLAGS;
5747 Perl_croak(aTHX_ "Can't modify nonexistent substring");
5748 SvPV_force_flags(bigstr, curlen, flags);
5749 (void)SvPOK_only_UTF8(bigstr);
5750 if (offset + len > curlen) {
5751 SvGROW(bigstr, offset+len+1);
5752 Zero(SvPVX(bigstr)+curlen, offset+len-curlen, char);
5753 SvCUR_set(bigstr, offset+len);
5757 i = littlelen - len;
5758 if (i > 0) { /* string might grow */
5759 big = SvGROW(bigstr, SvCUR(bigstr) + i + 1);
5760 mid = big + offset + len;
5761 midend = bigend = big + SvCUR(bigstr);
5764 while (midend > mid) /* shove everything down */
5765 *--bigend = *--midend;
5766 Move(little,big+offset,littlelen,char);
5767 SvCUR_set(bigstr, SvCUR(bigstr) + i);
5772 Move(little,SvPVX(bigstr)+offset,len,char);
5777 big = SvPVX(bigstr);
5780 bigend = big + SvCUR(bigstr);
5782 if (midend > bigend)
5783 Perl_croak(aTHX_ "panic: sv_insert");
5785 if (mid - big > bigend - midend) { /* faster to shorten from end */
5787 Move(little, mid, littlelen,char);
5790 i = bigend - midend;
5792 Move(midend, mid, i,char);
5796 SvCUR_set(bigstr, mid - big);
5798 else if ((i = mid - big)) { /* faster from front */
5799 midend -= littlelen;
5801 Move(big, midend - i, i, char);
5802 sv_chop(bigstr,midend-i);
5804 Move(little, mid, littlelen,char);
5806 else if (littlelen) {
5807 midend -= littlelen;
5808 sv_chop(bigstr,midend);
5809 Move(little,midend,littlelen,char);
5812 sv_chop(bigstr,midend);
5818 =for apidoc sv_replace
5820 Make the first argument a copy of the second, then delete the original.
5821 The target SV physically takes over ownership of the body of the source SV
5822 and inherits its flags; however, the target keeps any magic it owns,
5823 and any magic in the source is discarded.
5824 Note that this is a rather specialist SV copying operation; most of the
5825 time you'll want to use C<sv_setsv> or one of its many macro front-ends.
5831 Perl_sv_replace(pTHX_ register SV *const sv, register SV *const nsv)
5834 const U32 refcnt = SvREFCNT(sv);
5836 PERL_ARGS_ASSERT_SV_REPLACE;
5838 SV_CHECK_THINKFIRST_COW_DROP(sv);
5839 if (SvREFCNT(nsv) != 1) {
5840 Perl_croak(aTHX_ "panic: reference miscount on nsv in sv_replace()"
5841 " (%" UVuf " != 1)", (UV) SvREFCNT(nsv));
5843 if (SvMAGICAL(sv)) {
5847 sv_upgrade(nsv, SVt_PVMG);
5848 SvMAGIC_set(nsv, SvMAGIC(sv));
5849 SvFLAGS(nsv) |= SvMAGICAL(sv);
5851 SvMAGIC_set(sv, NULL);
5855 assert(!SvREFCNT(sv));
5856 #ifdef DEBUG_LEAKING_SCALARS
5857 sv->sv_flags = nsv->sv_flags;
5858 sv->sv_any = nsv->sv_any;
5859 sv->sv_refcnt = nsv->sv_refcnt;
5860 sv->sv_u = nsv->sv_u;
5862 StructCopy(nsv,sv,SV);
5864 if(SvTYPE(sv) == SVt_IV) {
5866 = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
5870 #ifdef PERL_OLD_COPY_ON_WRITE
5871 if (SvIsCOW_normal(nsv)) {
5872 /* We need to follow the pointers around the loop to make the
5873 previous SV point to sv, rather than nsv. */
5876 while ((next = SV_COW_NEXT_SV(current)) != nsv) {
5879 assert(SvPVX_const(current) == SvPVX_const(nsv));
5881 /* Make the SV before us point to the SV after us. */
5883 PerlIO_printf(Perl_debug_log, "previous is\n");
5885 PerlIO_printf(Perl_debug_log,
5886 "move it from 0x%"UVxf" to 0x%"UVxf"\n",
5887 (UV) SV_COW_NEXT_SV(current), (UV) sv);
5889 SV_COW_NEXT_SV_SET(current, sv);
5892 SvREFCNT(sv) = refcnt;
5893 SvFLAGS(nsv) |= SVTYPEMASK; /* Mark as freed */
5898 /* We're about to free a GV which has a CV that refers back to us.
5899 * If that CV will outlive us, make it anonymous (i.e. fix up its CvGV
5903 S_anonymise_cv_maybe(pTHX_ GV *gv, CV* cv)
5908 PERL_ARGS_ASSERT_ANONYMISE_CV_MAYBE;
5911 assert(SvREFCNT(gv) == 0);
5912 assert(isGV(gv) && isGV_with_GP(gv));
5914 assert(!CvANON(cv));
5915 assert(CvGV(cv) == gv);
5917 /* will the CV shortly be freed by gp_free() ? */
5918 if (GvCV(gv) == cv && GvGP(gv)->gp_refcnt < 2 && SvREFCNT(cv) < 2) {
5919 SvANY(cv)->xcv_gv = NULL;
5923 /* if not, anonymise: */
5924 gvname = (GvSTASH(gv) && HvNAME(GvSTASH(gv)) && HvENAME(GvSTASH(gv)))
5925 ? newSVhek(HvENAME_HEK(GvSTASH(gv)))
5926 : newSVpvn_flags( "__ANON__", 8, 0 );
5927 sv_catpvs(gvname, "::__ANON__");
5928 anongv = gv_fetchsv(gvname, GV_ADDMULTI, SVt_PVCV);
5929 SvREFCNT_dec(gvname);
5933 SvANY(cv)->xcv_gv = MUTABLE_GV(SvREFCNT_inc(anongv));
5938 =for apidoc sv_clear
5940 Clear an SV: call any destructors, free up any memory used by the body,
5941 and free the body itself. The SV's head is I<not> freed, although
5942 its type is set to all 1's so that it won't inadvertently be assumed
5943 to be live during global destruction etc.
5944 This function should only be called when REFCNT is zero. Most of the time
5945 you'll want to call C<sv_free()> (or its macro wrapper C<SvREFCNT_dec>)
5952 Perl_sv_clear(pTHX_ SV *const orig_sv)
5957 const struct body_details *sv_type_details;
5960 register SV *sv = orig_sv;
5963 PERL_ARGS_ASSERT_SV_CLEAR;
5965 /* within this loop, sv is the SV currently being freed, and
5966 * iter_sv is the most recent AV or whatever that's being iterated
5967 * over to provide more SVs */
5973 assert(SvREFCNT(sv) == 0);
5974 assert(SvTYPE(sv) != (svtype)SVTYPEMASK);
5976 if (type <= SVt_IV) {
5977 /* See the comment in sv.h about the collusion between this
5978 * early return and the overloading of the NULL slots in the
5982 SvFLAGS(sv) &= SVf_BREAK;
5983 SvFLAGS(sv) |= SVTYPEMASK;
5987 assert(!SvOBJECT(sv) || type >= SVt_PVMG); /* objs are always >= MG */
5989 if (type >= SVt_PVMG) {
5991 if (!curse(sv, 1)) goto get_next_sv;
5992 type = SvTYPE(sv); /* destructor may have changed it */
5994 /* Free back-references before magic, in case the magic calls
5995 * Perl code that has weak references to sv. */
5996 if (type == SVt_PVHV) {
5997 Perl_hv_kill_backrefs(aTHX_ MUTABLE_HV(sv));
6001 else if (type == SVt_PVMG && SvPAD_OUR(sv)) {
6002 SvREFCNT_dec(SvOURSTASH(sv));
6003 } else if (SvMAGIC(sv)) {
6004 /* Free back-references before other types of magic. */
6005 sv_unmagic(sv, PERL_MAGIC_backref);
6008 if (type == SVt_PVMG && SvPAD_TYPED(sv))
6009 SvREFCNT_dec(SvSTASH(sv));
6012 /* case SVt_BIND: */
6015 IoIFP(sv) != PerlIO_stdin() &&
6016 IoIFP(sv) != PerlIO_stdout() &&
6017 IoIFP(sv) != PerlIO_stderr() &&
6018 !(IoFLAGS(sv) & IOf_FAKE_DIRP))
6020 io_close(MUTABLE_IO(sv), FALSE);
6022 if (IoDIRP(sv) && !(IoFLAGS(sv) & IOf_FAKE_DIRP))
6023 PerlDir_close(IoDIRP(sv));
6024 IoDIRP(sv) = (DIR*)NULL;
6025 Safefree(IoTOP_NAME(sv));
6026 Safefree(IoFMT_NAME(sv));
6027 Safefree(IoBOTTOM_NAME(sv));
6030 /* FIXME for plugins */
6031 pregfree2((REGEXP*) sv);
6035 cv_undef(MUTABLE_CV(sv));
6036 /* If we're in a stash, we don't own a reference to it.
6037 * However it does have a back reference to us, which needs to
6039 if ((stash = CvSTASH(sv)))
6040 sv_del_backref(MUTABLE_SV(stash), sv);
6043 if (PL_last_swash_hv == (const HV *)sv) {
6044 PL_last_swash_hv = NULL;
6046 if (HvTOTALKEYS((HV*)sv) > 0) {
6048 /* this statement should match the one at the beginning of
6049 * hv_undef_flags() */
6050 if ( PL_phase != PERL_PHASE_DESTRUCT
6051 && (name = HvNAME((HV*)sv)))
6054 (void)hv_delete(PL_stashcache, name,
6055 HvNAMEUTF8((HV*)sv) ? -HvNAMELEN_get((HV*)sv) : HvNAMELEN_get((HV*)sv), G_DISCARD);
6056 hv_name_set((HV*)sv, NULL, 0, 0);
6059 /* save old iter_sv in unused SvSTASH field */
6060 assert(!SvOBJECT(sv));
6061 SvSTASH(sv) = (HV*)iter_sv;
6064 /* XXX ideally we should save the old value of hash_index
6065 * too, but I can't think of any place to hide it. The
6066 * effect of not saving it is that for freeing hashes of
6067 * hashes, we become quadratic in scanning the HvARRAY of
6068 * the top hash looking for new entries to free; but
6069 * hopefully this will be dwarfed by the freeing of all
6070 * the nested hashes. */
6072 next_sv = Perl_hfree_next_entry(aTHX_ (HV*)sv, &hash_index);
6073 goto get_next_sv; /* process this new sv */
6075 /* free empty hash */
6076 Perl_hv_undef_flags(aTHX_ MUTABLE_HV(sv), HV_NAME_SETALL);
6077 assert(!HvARRAY((HV*)sv));
6081 AV* av = MUTABLE_AV(sv);
6082 if (PL_comppad == av) {
6086 if (AvREAL(av) && AvFILLp(av) > -1) {
6087 next_sv = AvARRAY(av)[AvFILLp(av)--];
6088 /* save old iter_sv in top-most slot of AV,
6089 * and pray that it doesn't get wiped in the meantime */
6090 AvARRAY(av)[AvMAX(av)] = iter_sv;
6092 goto get_next_sv; /* process this new sv */
6094 Safefree(AvALLOC(av));
6099 if (LvTYPE(sv) == 'T') { /* for tie: return HE to pool */
6100 SvREFCNT_dec(HeKEY_sv((HE*)LvTARG(sv)));
6101 HeNEXT((HE*)LvTARG(sv)) = PL_hv_fetch_ent_mh;
6102 PL_hv_fetch_ent_mh = (HE*)LvTARG(sv);
6104 else if (LvTYPE(sv) != 't') /* unless tie: unrefcnted fake SV** */
6105 SvREFCNT_dec(LvTARG(sv));
6107 if (isGV_with_GP(sv)) {
6108 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
6109 && HvENAME_get(stash))
6110 mro_method_changed_in(stash);
6111 gp_free(MUTABLE_GV(sv));
6113 unshare_hek(GvNAME_HEK(sv));
6114 /* If we're in a stash, we don't own a reference to it.
6115 * However it does have a back reference to us, which
6116 * needs to be cleared. */
6117 if (!SvVALID(sv) && (stash = GvSTASH(sv)))
6118 sv_del_backref(MUTABLE_SV(stash), sv);
6120 /* FIXME. There are probably more unreferenced pointers to SVs
6121 * in the interpreter struct that we should check and tidy in
6122 * a similar fashion to this: */
6123 /* See also S_sv_unglob, which does the same thing. */
6124 if ((const GV *)sv == PL_last_in_gv)
6125 PL_last_in_gv = NULL;
6131 /* Don't bother with SvOOK_off(sv); as we're only going to
6135 SvOOK_offset(sv, offset);
6136 SvPV_set(sv, SvPVX_mutable(sv) - offset);
6137 /* Don't even bother with turning off the OOK flag. */
6142 SV * const target = SvRV(sv);
6144 sv_del_backref(target, sv);
6149 #ifdef PERL_OLD_COPY_ON_WRITE
6150 else if (SvPVX_const(sv)
6151 && !(SvTYPE(sv) == SVt_PVIO
6152 && !(IoFLAGS(sv) & IOf_FAKE_DIRP)))
6156 PerlIO_printf(Perl_debug_log, "Copy on write: clear\n");
6160 sv_release_COW(sv, SvPVX_const(sv), SV_COW_NEXT_SV(sv));
6162 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
6166 } else if (SvLEN(sv)) {
6167 Safefree(SvPVX_const(sv));
6171 else if (SvPVX_const(sv) && SvLEN(sv)
6172 && !(SvTYPE(sv) == SVt_PVIO
6173 && !(IoFLAGS(sv) & IOf_FAKE_DIRP)))
6174 Safefree(SvPVX_mutable(sv));
6175 else if (SvPVX_const(sv) && SvREADONLY(sv) && SvFAKE(sv)) {
6176 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
6187 SvFLAGS(sv) &= SVf_BREAK;
6188 SvFLAGS(sv) |= SVTYPEMASK;
6190 sv_type_details = bodies_by_type + type;
6191 if (sv_type_details->arena) {
6192 del_body(((char *)SvANY(sv) + sv_type_details->offset),
6193 &PL_body_roots[type]);
6195 else if (sv_type_details->body_size) {
6196 safefree(SvANY(sv));
6200 /* caller is responsible for freeing the head of the original sv */
6201 if (sv != orig_sv && !SvREFCNT(sv))
6204 /* grab and free next sv, if any */
6212 else if (!iter_sv) {
6214 } else if (SvTYPE(iter_sv) == SVt_PVAV) {
6215 AV *const av = (AV*)iter_sv;
6216 if (AvFILLp(av) > -1) {
6217 sv = AvARRAY(av)[AvFILLp(av)--];
6219 else { /* no more elements of current AV to free */
6222 /* restore previous value, squirrelled away */
6223 iter_sv = AvARRAY(av)[AvMAX(av)];
6224 Safefree(AvALLOC(av));
6227 } else if (SvTYPE(iter_sv) == SVt_PVHV) {
6228 sv = Perl_hfree_next_entry(aTHX_ (HV*)iter_sv, &hash_index);
6229 if (!sv && !HvTOTALKEYS((HV *)iter_sv)) {
6230 /* no more elements of current HV to free */
6233 /* Restore previous value of iter_sv, squirrelled away */
6234 assert(!SvOBJECT(sv));
6235 iter_sv = (SV*)SvSTASH(sv);
6237 /* ideally we should restore the old hash_index here,
6238 * but we don't currently save the old value */
6241 /* free any remaining detritus from the hash struct */
6242 Perl_hv_undef_flags(aTHX_ MUTABLE_HV(sv), HV_NAME_SETALL);
6243 assert(!HvARRAY((HV*)sv));
6248 /* unrolled SvREFCNT_dec and sv_free2 follows: */
6252 if (!SvREFCNT(sv)) {
6256 if (--(SvREFCNT(sv)))
6260 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEBUGGING),
6261 "Attempt to free temp prematurely: SV 0x%"UVxf
6262 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
6266 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
6267 /* make sure SvREFCNT(sv)==0 happens very seldom */
6268 SvREFCNT(sv) = (~(U32)0)/2;
6277 /* This routine curses the sv itself, not the object referenced by sv. So
6278 sv does not have to be ROK. */
6281 S_curse(pTHX_ SV * const sv, const bool check_refcnt) {
6284 PERL_ARGS_ASSERT_CURSE;
6285 assert(SvOBJECT(sv));
6287 if (PL_defstash && /* Still have a symbol table? */
6294 stash = SvSTASH(sv);
6295 destructor = StashHANDLER(stash,DESTROY);
6297 /* A constant subroutine can have no side effects, so
6298 don't bother calling it. */
6299 && !CvCONST(destructor)
6300 /* Don't bother calling an empty destructor or one that
6301 returns immediately. */
6302 && (CvISXSUB(destructor)
6303 || (CvSTART(destructor)
6304 && (CvSTART(destructor)->op_next->op_type
6306 && (CvSTART(destructor)->op_next->op_type
6308 || CvSTART(destructor)->op_next->op_next->op_type
6314 SV* const tmpref = newRV(sv);
6315 SvREADONLY_on(tmpref); /* DESTROY() could be naughty */
6317 PUSHSTACKi(PERLSI_DESTROY);
6322 call_sv(MUTABLE_SV(destructor),
6323 G_DISCARD|G_EVAL|G_KEEPERR|G_VOID);
6327 if(SvREFCNT(tmpref) < 2) {
6328 /* tmpref is not kept alive! */
6330 SvRV_set(tmpref, NULL);
6333 SvREFCNT_dec(tmpref);
6335 } while (SvOBJECT(sv) && SvSTASH(sv) != stash);
6338 if (check_refcnt && SvREFCNT(sv)) {
6339 if (PL_in_clean_objs)
6341 "DESTROY created new reference to dead object '%"HEKf"'",
6342 HEKfARG(HvNAME_HEK(stash)));
6343 /* DESTROY gave object new lease on life */
6349 SvREFCNT_dec(SvSTASH(sv)); /* possibly of changed persuasion */
6350 SvOBJECT_off(sv); /* Curse the object. */
6351 if (SvTYPE(sv) != SVt_PVIO)
6352 --PL_sv_objcount;/* XXX Might want something more general */
6358 =for apidoc sv_newref
6360 Increment an SV's reference count. Use the C<SvREFCNT_inc()> wrapper
6367 Perl_sv_newref(pTHX_ SV *const sv)
6369 PERL_UNUSED_CONTEXT;
6378 Decrement an SV's reference count, and if it drops to zero, call
6379 C<sv_clear> to invoke destructors and free up any memory used by
6380 the body; finally, deallocate the SV's head itself.
6381 Normally called via a wrapper macro C<SvREFCNT_dec>.
6387 Perl_sv_free(pTHX_ SV *const sv)
6392 if (SvREFCNT(sv) == 0) {
6393 if (SvFLAGS(sv) & SVf_BREAK)
6394 /* this SV's refcnt has been artificially decremented to
6395 * trigger cleanup */
6397 if (PL_in_clean_all) /* All is fair */
6399 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
6400 /* make sure SvREFCNT(sv)==0 happens very seldom */
6401 SvREFCNT(sv) = (~(U32)0)/2;
6404 if (ckWARN_d(WARN_INTERNAL)) {
6405 #ifdef DEBUG_LEAKING_SCALARS_FORK_DUMP
6406 Perl_dump_sv_child(aTHX_ sv);
6408 #ifdef DEBUG_LEAKING_SCALARS
6411 #ifdef DEBUG_LEAKING_SCALARS_ABORT
6412 if (PL_warnhook == PERL_WARNHOOK_FATAL
6413 || ckDEAD(packWARN(WARN_INTERNAL))) {
6414 /* Don't let Perl_warner cause us to escape our fate: */
6418 /* This may not return: */
6419 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
6420 "Attempt to free unreferenced scalar: SV 0x%"UVxf
6421 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
6424 #ifdef DEBUG_LEAKING_SCALARS_ABORT
6429 if (--(SvREFCNT(sv)) > 0)
6431 Perl_sv_free2(aTHX_ sv);
6435 Perl_sv_free2(pTHX_ SV *const sv)
6439 PERL_ARGS_ASSERT_SV_FREE2;
6443 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEBUGGING),
6444 "Attempt to free temp prematurely: SV 0x%"UVxf
6445 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
6449 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
6450 /* make sure SvREFCNT(sv)==0 happens very seldom */
6451 SvREFCNT(sv) = (~(U32)0)/2;
6462 Returns the length of the string in the SV. Handles magic and type
6463 coercion. See also C<SvCUR>, which gives raw access to the xpv_cur slot.
6469 Perl_sv_len(pTHX_ register SV *const sv)
6477 len = mg_length(sv);
6479 (void)SvPV_const(sv, len);
6484 =for apidoc sv_len_utf8
6486 Returns the number of characters in the string in an SV, counting wide
6487 UTF-8 bytes as a single character. Handles magic and type coercion.
6493 * The length is cached in PERL_MAGIC_utf8, in the mg_len field. Also the
6494 * mg_ptr is used, by sv_pos_u2b() and sv_pos_b2u() - see the comments below.
6495 * (Note that the mg_len is not the length of the mg_ptr field.
6496 * This allows the cache to store the character length of the string without
6497 * needing to malloc() extra storage to attach to the mg_ptr.)
6502 Perl_sv_len_utf8(pTHX_ register SV *const sv)
6508 return mg_length(sv);
6512 const U8 *s = (U8*)SvPV_const(sv, len);
6516 MAGIC *mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_utf8) : NULL;
6518 if (mg && (mg->mg_len != -1 || mg->mg_ptr)) {
6519 if (mg->mg_len != -1)
6522 /* We can use the offset cache for a headstart.
6523 The longer value is stored in the first pair. */
6524 STRLEN *cache = (STRLEN *) mg->mg_ptr;
6526 ulen = cache[0] + Perl_utf8_length(aTHX_ s + cache[1],
6530 if (PL_utf8cache < 0) {
6531 const STRLEN real = Perl_utf8_length(aTHX_ s, s + len);
6532 assert_uft8_cache_coherent("sv_len_utf8", ulen, real, sv);
6536 ulen = Perl_utf8_length(aTHX_ s, s + len);
6537 utf8_mg_len_cache_update(sv, &mg, ulen);
6541 return Perl_utf8_length(aTHX_ s, s + len);
6545 /* Walk forwards to find the byte corresponding to the passed in UTF-8
6548 S_sv_pos_u2b_forwards(const U8 *const start, const U8 *const send,
6549 STRLEN *const uoffset_p, bool *const at_end)
6551 const U8 *s = start;
6552 STRLEN uoffset = *uoffset_p;
6554 PERL_ARGS_ASSERT_SV_POS_U2B_FORWARDS;
6556 while (s < send && uoffset) {
6563 else if (s > send) {
6565 /* This is the existing behaviour. Possibly it should be a croak, as
6566 it's actually a bounds error */
6569 *uoffset_p -= uoffset;
6573 /* Given the length of the string in both bytes and UTF-8 characters, decide
6574 whether to walk forwards or backwards to find the byte corresponding to
6575 the passed in UTF-8 offset. */
6577 S_sv_pos_u2b_midway(const U8 *const start, const U8 *send,
6578 STRLEN uoffset, const STRLEN uend)
6580 STRLEN backw = uend - uoffset;
6582 PERL_ARGS_ASSERT_SV_POS_U2B_MIDWAY;
6584 if (uoffset < 2 * backw) {
6585 /* The assumption is that going forwards is twice the speed of going
6586 forward (that's where the 2 * backw comes from).
6587 (The real figure of course depends on the UTF-8 data.) */
6588 const U8 *s = start;
6590 while (s < send && uoffset--)
6600 while (UTF8_IS_CONTINUATION(*send))
6603 return send - start;
6606 /* For the string representation of the given scalar, find the byte
6607 corresponding to the passed in UTF-8 offset. uoffset0 and boffset0
6608 give another position in the string, *before* the sought offset, which
6609 (which is always true, as 0, 0 is a valid pair of positions), which should
6610 help reduce the amount of linear searching.
6611 If *mgp is non-NULL, it should point to the UTF-8 cache magic, which
6612 will be used to reduce the amount of linear searching. The cache will be
6613 created if necessary, and the found value offered to it for update. */
6615 S_sv_pos_u2b_cached(pTHX_ SV *const sv, MAGIC **const mgp, const U8 *const start,
6616 const U8 *const send, STRLEN uoffset,
6617 STRLEN uoffset0, STRLEN boffset0)
6619 STRLEN boffset = 0; /* Actually always set, but let's keep gcc happy. */
6621 bool at_end = FALSE;
6623 PERL_ARGS_ASSERT_SV_POS_U2B_CACHED;
6625 assert (uoffset >= uoffset0);
6632 && (*mgp || (SvTYPE(sv) >= SVt_PVMG &&
6633 (*mgp = mg_find(sv, PERL_MAGIC_utf8))))) {
6634 if ((*mgp)->mg_ptr) {
6635 STRLEN *cache = (STRLEN *) (*mgp)->mg_ptr;
6636 if (cache[0] == uoffset) {
6637 /* An exact match. */
6640 if (cache[2] == uoffset) {
6641 /* An exact match. */
6645 if (cache[0] < uoffset) {
6646 /* The cache already knows part of the way. */
6647 if (cache[0] > uoffset0) {
6648 /* The cache knows more than the passed in pair */
6649 uoffset0 = cache[0];
6650 boffset0 = cache[1];
6652 if ((*mgp)->mg_len != -1) {
6653 /* And we know the end too. */
6655 + sv_pos_u2b_midway(start + boffset0, send,
6657 (*mgp)->mg_len - uoffset0);
6659 uoffset -= uoffset0;
6661 + sv_pos_u2b_forwards(start + boffset0,
6662 send, &uoffset, &at_end);
6663 uoffset += uoffset0;
6666 else if (cache[2] < uoffset) {
6667 /* We're between the two cache entries. */
6668 if (cache[2] > uoffset0) {
6669 /* and the cache knows more than the passed in pair */
6670 uoffset0 = cache[2];
6671 boffset0 = cache[3];
6675 + sv_pos_u2b_midway(start + boffset0,
6678 cache[0] - uoffset0);
6681 + sv_pos_u2b_midway(start + boffset0,
6684 cache[2] - uoffset0);
6688 else if ((*mgp)->mg_len != -1) {
6689 /* If we can take advantage of a passed in offset, do so. */
6690 /* In fact, offset0 is either 0, or less than offset, so don't
6691 need to worry about the other possibility. */
6693 + sv_pos_u2b_midway(start + boffset0, send,
6695 (*mgp)->mg_len - uoffset0);
6700 if (!found || PL_utf8cache < 0) {
6701 STRLEN real_boffset;
6702 uoffset -= uoffset0;
6703 real_boffset = boffset0 + sv_pos_u2b_forwards(start + boffset0,
6704 send, &uoffset, &at_end);
6705 uoffset += uoffset0;
6707 if (found && PL_utf8cache < 0)
6708 assert_uft8_cache_coherent("sv_pos_u2b_cache", boffset,
6710 boffset = real_boffset;
6715 utf8_mg_len_cache_update(sv, mgp, uoffset);
6717 utf8_mg_pos_cache_update(sv, mgp, boffset, uoffset, send - start);
6724 =for apidoc sv_pos_u2b_flags
6726 Converts the value pointed to by offsetp from a count of UTF-8 chars from
6727 the start of the string, to a count of the equivalent number of bytes; if
6728 lenp is non-zero, it does the same to lenp, but this time starting from
6729 the offset, rather than from the start
6730 of the string. Handles type coercion.
6731 I<flags> is passed to C<SvPV_flags>, and usually should be
6732 C<SV_GMAGIC|SV_CONST_RETURN> to handle magic.
6738 * sv_pos_u2b_flags() uses, like sv_pos_b2u(), the mg_ptr of the potential
6739 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6740 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
6745 Perl_sv_pos_u2b_flags(pTHX_ SV *const sv, STRLEN uoffset, STRLEN *const lenp,
6752 PERL_ARGS_ASSERT_SV_POS_U2B_FLAGS;
6754 start = (U8*)SvPV_flags(sv, len, flags);
6756 const U8 * const send = start + len;
6758 boffset = sv_pos_u2b_cached(sv, &mg, start, send, uoffset, 0, 0);
6761 && *lenp /* don't bother doing work for 0, as its bytes equivalent
6762 is 0, and *lenp is already set to that. */) {
6763 /* Convert the relative offset to absolute. */
6764 const STRLEN uoffset2 = uoffset + *lenp;
6765 const STRLEN boffset2
6766 = sv_pos_u2b_cached(sv, &mg, start, send, uoffset2,
6767 uoffset, boffset) - boffset;
6781 =for apidoc sv_pos_u2b
6783 Converts the value pointed to by offsetp from a count of UTF-8 chars from
6784 the start of the string, to a count of the equivalent number of bytes; if
6785 lenp is non-zero, it does the same to lenp, but this time starting from
6786 the offset, rather than from the start of the string. Handles magic and
6789 Use C<sv_pos_u2b_flags> in preference, which correctly handles strings longer
6796 * sv_pos_u2b() uses, like sv_pos_b2u(), the mg_ptr of the potential
6797 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6798 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
6802 /* This function is subject to size and sign problems */
6805 Perl_sv_pos_u2b(pTHX_ register SV *const sv, I32 *const offsetp, I32 *const lenp)
6807 PERL_ARGS_ASSERT_SV_POS_U2B;
6810 STRLEN ulen = (STRLEN)*lenp;
6811 *offsetp = (I32)sv_pos_u2b_flags(sv, (STRLEN)*offsetp, &ulen,
6812 SV_GMAGIC|SV_CONST_RETURN);
6815 *offsetp = (I32)sv_pos_u2b_flags(sv, (STRLEN)*offsetp, NULL,
6816 SV_GMAGIC|SV_CONST_RETURN);
6821 S_utf8_mg_len_cache_update(pTHX_ SV *const sv, MAGIC **const mgp,
6824 PERL_ARGS_ASSERT_UTF8_MG_LEN_CACHE_UPDATE;
6828 if (!*mgp && (SvTYPE(sv) < SVt_PVMG ||
6829 !(*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
6830 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, &PL_vtbl_utf8, 0, 0);
6834 (*mgp)->mg_len = ulen;
6835 /* For now, treat "overflowed" as "still unknown". See RT #72924. */
6836 if (ulen != (STRLEN) (*mgp)->mg_len)
6837 (*mgp)->mg_len = -1;
6840 /* Create and update the UTF8 magic offset cache, with the proffered utf8/
6841 byte length pairing. The (byte) length of the total SV is passed in too,
6842 as blen, because for some (more esoteric) SVs, the call to SvPV_const()
6843 may not have updated SvCUR, so we can't rely on reading it directly.
6845 The proffered utf8/byte length pairing isn't used if the cache already has
6846 two pairs, and swapping either for the proffered pair would increase the
6847 RMS of the intervals between known byte offsets.
6849 The cache itself consists of 4 STRLEN values
6850 0: larger UTF-8 offset
6851 1: corresponding byte offset
6852 2: smaller UTF-8 offset
6853 3: corresponding byte offset
6855 Unused cache pairs have the value 0, 0.
6856 Keeping the cache "backwards" means that the invariant of
6857 cache[0] >= cache[2] is maintained even with empty slots, which means that
6858 the code that uses it doesn't need to worry if only 1 entry has actually
6859 been set to non-zero. It also makes the "position beyond the end of the
6860 cache" logic much simpler, as the first slot is always the one to start
6864 S_utf8_mg_pos_cache_update(pTHX_ SV *const sv, MAGIC **const mgp, const STRLEN byte,
6865 const STRLEN utf8, const STRLEN blen)
6869 PERL_ARGS_ASSERT_UTF8_MG_POS_CACHE_UPDATE;
6874 if (!*mgp && (SvTYPE(sv) < SVt_PVMG ||
6875 !(*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
6876 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, (MGVTBL*)&PL_vtbl_utf8, 0,
6878 (*mgp)->mg_len = -1;
6882 if (!(cache = (STRLEN *)(*mgp)->mg_ptr)) {
6883 Newxz(cache, PERL_MAGIC_UTF8_CACHESIZE * 2, STRLEN);
6884 (*mgp)->mg_ptr = (char *) cache;
6888 if (PL_utf8cache < 0 && SvPOKp(sv)) {
6889 /* SvPOKp() because it's possible that sv has string overloading, and
6890 therefore is a reference, hence SvPVX() is actually a pointer.
6891 This cures the (very real) symptoms of RT 69422, but I'm not actually
6892 sure whether we should even be caching the results of UTF-8
6893 operations on overloading, given that nothing stops overloading
6894 returning a different value every time it's called. */
6895 const U8 *start = (const U8 *) SvPVX_const(sv);
6896 const STRLEN realutf8 = utf8_length(start, start + byte);
6898 assert_uft8_cache_coherent("utf8_mg_pos_cache_update", utf8, realutf8,
6902 /* Cache is held with the later position first, to simplify the code
6903 that deals with unbounded ends. */
6905 ASSERT_UTF8_CACHE(cache);
6906 if (cache[1] == 0) {
6907 /* Cache is totally empty */
6910 } else if (cache[3] == 0) {
6911 if (byte > cache[1]) {
6912 /* New one is larger, so goes first. */
6913 cache[2] = cache[0];
6914 cache[3] = cache[1];
6922 #define THREEWAY_SQUARE(a,b,c,d) \
6923 ((float)((d) - (c))) * ((float)((d) - (c))) \
6924 + ((float)((c) - (b))) * ((float)((c) - (b))) \
6925 + ((float)((b) - (a))) * ((float)((b) - (a)))
6927 /* Cache has 2 slots in use, and we know three potential pairs.
6928 Keep the two that give the lowest RMS distance. Do the
6929 calculation in bytes simply because we always know the byte
6930 length. squareroot has the same ordering as the positive value,
6931 so don't bother with the actual square root. */
6932 const float existing = THREEWAY_SQUARE(0, cache[3], cache[1], blen);
6933 if (byte > cache[1]) {
6934 /* New position is after the existing pair of pairs. */
6935 const float keep_earlier
6936 = THREEWAY_SQUARE(0, cache[3], byte, blen);
6937 const float keep_later
6938 = THREEWAY_SQUARE(0, cache[1], byte, blen);
6940 if (keep_later < keep_earlier) {
6941 if (keep_later < existing) {
6942 cache[2] = cache[0];
6943 cache[3] = cache[1];
6949 if (keep_earlier < existing) {
6955 else if (byte > cache[3]) {
6956 /* New position is between the existing pair of pairs. */
6957 const float keep_earlier
6958 = THREEWAY_SQUARE(0, cache[3], byte, blen);
6959 const float keep_later
6960 = THREEWAY_SQUARE(0, byte, cache[1], blen);
6962 if (keep_later < keep_earlier) {
6963 if (keep_later < existing) {
6969 if (keep_earlier < existing) {
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) {
6983 if (keep_later < existing) {
6989 if (keep_earlier < existing) {
6990 cache[0] = cache[2];
6991 cache[1] = cache[3];
6998 ASSERT_UTF8_CACHE(cache);
7001 /* We already know all of the way, now we may be able to walk back. The same
7002 assumption is made as in S_sv_pos_u2b_midway(), namely that walking
7003 backward is half the speed of walking forward. */
7005 S_sv_pos_b2u_midway(pTHX_ const U8 *const s, const U8 *const target,
7006 const U8 *end, STRLEN endu)
7008 const STRLEN forw = target - s;
7009 STRLEN backw = end - target;
7011 PERL_ARGS_ASSERT_SV_POS_B2U_MIDWAY;
7013 if (forw < 2 * backw) {
7014 return utf8_length(s, target);
7017 while (end > target) {
7019 while (UTF8_IS_CONTINUATION(*end)) {
7028 =for apidoc sv_pos_b2u
7030 Converts the value pointed to by offsetp from a count of bytes from the
7031 start of the string, to a count of the equivalent number of UTF-8 chars.
7032 Handles magic and type coercion.
7038 * sv_pos_b2u() uses, like sv_pos_u2b(), the mg_ptr of the potential
7039 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
7044 Perl_sv_pos_b2u(pTHX_ register SV *const sv, I32 *const offsetp)
7047 const STRLEN byte = *offsetp;
7048 STRLEN len = 0; /* Actually always set, but let's keep gcc happy. */
7054 PERL_ARGS_ASSERT_SV_POS_B2U;
7059 s = (const U8*)SvPV_const(sv, blen);
7062 Perl_croak(aTHX_ "panic: sv_pos_b2u: bad byte offset");
7068 && SvTYPE(sv) >= SVt_PVMG
7069 && (mg = mg_find(sv, PERL_MAGIC_utf8)))
7072 STRLEN * const cache = (STRLEN *) mg->mg_ptr;
7073 if (cache[1] == byte) {
7074 /* An exact match. */
7075 *offsetp = cache[0];
7078 if (cache[3] == byte) {
7079 /* An exact match. */
7080 *offsetp = cache[2];
7084 if (cache[1] < byte) {
7085 /* We already know part of the way. */
7086 if (mg->mg_len != -1) {
7087 /* Actually, we know the end too. */
7089 + S_sv_pos_b2u_midway(aTHX_ s + cache[1], send,
7090 s + blen, mg->mg_len - cache[0]);
7092 len = cache[0] + utf8_length(s + cache[1], send);
7095 else if (cache[3] < byte) {
7096 /* We're between the two cached pairs, so we do the calculation
7097 offset by the byte/utf-8 positions for the earlier pair,
7098 then add the utf-8 characters from the string start to
7100 len = S_sv_pos_b2u_midway(aTHX_ s + cache[3], send,
7101 s + cache[1], cache[0] - cache[2])
7105 else { /* cache[3] > byte */
7106 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + cache[3],
7110 ASSERT_UTF8_CACHE(cache);
7112 } else if (mg->mg_len != -1) {
7113 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + blen, mg->mg_len);
7117 if (!found || PL_utf8cache < 0) {
7118 const STRLEN real_len = utf8_length(s, send);
7120 if (found && PL_utf8cache < 0)
7121 assert_uft8_cache_coherent("sv_pos_b2u", len, real_len, sv);
7128 utf8_mg_len_cache_update(sv, &mg, len);
7130 utf8_mg_pos_cache_update(sv, &mg, byte, len, blen);
7135 S_assert_uft8_cache_coherent(pTHX_ const char *const func, STRLEN from_cache,
7136 STRLEN real, SV *const sv)
7138 PERL_ARGS_ASSERT_ASSERT_UFT8_CACHE_COHERENT;
7140 /* As this is debugging only code, save space by keeping this test here,
7141 rather than inlining it in all the callers. */
7142 if (from_cache == real)
7145 /* Need to turn the assertions off otherwise we may recurse infinitely
7146 while printing error messages. */
7147 SAVEI8(PL_utf8cache);
7149 Perl_croak(aTHX_ "panic: %s cache %"UVuf" real %"UVuf" for %"SVf,
7150 func, (UV) from_cache, (UV) real, SVfARG(sv));
7156 Returns a boolean indicating whether the strings in the two SVs are
7157 identical. Is UTF-8 and 'use bytes' aware, handles get magic, and will
7158 coerce its args to strings if necessary.
7160 =for apidoc sv_eq_flags
7162 Returns a boolean indicating whether the strings in the two SVs are
7163 identical. Is UTF-8 and 'use bytes' aware and coerces its args to strings
7164 if necessary. If the flags include SV_GMAGIC, it handles get-magic, too.
7170 Perl_sv_eq_flags(pTHX_ register SV *sv1, register SV *sv2, const U32 flags)
7178 SV* svrecode = NULL;
7185 /* if pv1 and pv2 are the same, second SvPV_const call may
7186 * invalidate pv1 (if we are handling magic), so we may need to
7188 if (sv1 == sv2 && flags & SV_GMAGIC
7189 && (SvTHINKFIRST(sv1) || SvGMAGICAL(sv1))) {
7190 pv1 = SvPV_const(sv1, cur1);
7191 sv1 = newSVpvn_flags(pv1, cur1, SVs_TEMP | SvUTF8(sv2));
7193 pv1 = SvPV_flags_const(sv1, cur1, flags);
7201 pv2 = SvPV_flags_const(sv2, cur2, flags);
7203 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
7204 /* Differing utf8ness.
7205 * Do not UTF8size the comparands as a side-effect. */
7208 svrecode = newSVpvn(pv2, cur2);
7209 sv_recode_to_utf8(svrecode, PL_encoding);
7210 pv2 = SvPV_const(svrecode, cur2);
7213 svrecode = newSVpvn(pv1, cur1);
7214 sv_recode_to_utf8(svrecode, PL_encoding);
7215 pv1 = SvPV_const(svrecode, cur1);
7217 /* Now both are in UTF-8. */
7219 SvREFCNT_dec(svrecode);
7225 /* sv1 is the UTF-8 one */
7226 return bytes_cmp_utf8((const U8*)pv2, cur2,
7227 (const U8*)pv1, cur1) == 0;
7230 /* sv2 is the UTF-8 one */
7231 return bytes_cmp_utf8((const U8*)pv1, cur1,
7232 (const U8*)pv2, cur2) == 0;
7238 eq = (pv1 == pv2) || memEQ(pv1, pv2, cur1);
7240 SvREFCNT_dec(svrecode);
7248 Compares the strings in two SVs. Returns -1, 0, or 1 indicating whether the
7249 string in C<sv1> is less than, equal to, or greater than the string in
7250 C<sv2>. Is UTF-8 and 'use bytes' aware, handles get magic, and will
7251 coerce its args to strings if necessary. See also C<sv_cmp_locale>.
7253 =for apidoc sv_cmp_flags
7255 Compares the strings in two SVs. Returns -1, 0, or 1 indicating whether the
7256 string in C<sv1> is less than, equal to, or greater than the string in
7257 C<sv2>. Is UTF-8 and 'use bytes' aware and will coerce its args to strings
7258 if necessary. If the flags include SV_GMAGIC, it handles get magic. See
7259 also C<sv_cmp_locale_flags>.
7265 Perl_sv_cmp(pTHX_ register SV *const sv1, register SV *const sv2)
7267 return sv_cmp_flags(sv1, sv2, SV_GMAGIC);
7271 Perl_sv_cmp_flags(pTHX_ register SV *const sv1, register SV *const sv2,
7276 const char *pv1, *pv2;
7279 SV *svrecode = NULL;
7286 pv1 = SvPV_flags_const(sv1, cur1, flags);
7293 pv2 = SvPV_flags_const(sv2, cur2, flags);
7295 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
7296 /* Differing utf8ness.
7297 * Do not UTF8size the comparands as a side-effect. */
7300 svrecode = newSVpvn(pv2, cur2);
7301 sv_recode_to_utf8(svrecode, PL_encoding);
7302 pv2 = SvPV_const(svrecode, cur2);
7305 const int retval = -bytes_cmp_utf8((const U8*)pv2, cur2,
7306 (const U8*)pv1, cur1);
7307 return retval ? retval < 0 ? -1 : +1 : 0;
7312 svrecode = newSVpvn(pv1, cur1);
7313 sv_recode_to_utf8(svrecode, PL_encoding);
7314 pv1 = SvPV_const(svrecode, cur1);
7317 const int retval = bytes_cmp_utf8((const U8*)pv1, cur1,
7318 (const U8*)pv2, cur2);
7319 return retval ? retval < 0 ? -1 : +1 : 0;
7325 cmp = cur2 ? -1 : 0;
7329 const I32 retval = memcmp((const void*)pv1, (const void*)pv2, cur1 < cur2 ? cur1 : cur2);
7332 cmp = retval < 0 ? -1 : 1;
7333 } else if (cur1 == cur2) {
7336 cmp = cur1 < cur2 ? -1 : 1;
7340 SvREFCNT_dec(svrecode);
7348 =for apidoc sv_cmp_locale
7350 Compares the strings in two SVs in a locale-aware manner. Is UTF-8 and
7351 'use bytes' aware, handles get magic, and will coerce its args to strings
7352 if necessary. See also C<sv_cmp>.
7354 =for apidoc sv_cmp_locale_flags
7356 Compares the strings in two SVs in a locale-aware manner. Is UTF-8 and
7357 'use bytes' aware and will coerce its args to strings if necessary. If the
7358 flags contain SV_GMAGIC, it handles get magic. See also C<sv_cmp_flags>.
7364 Perl_sv_cmp_locale(pTHX_ register SV *const sv1, register SV *const sv2)
7366 return sv_cmp_locale_flags(sv1, sv2, SV_GMAGIC);
7370 Perl_sv_cmp_locale_flags(pTHX_ register SV *const sv1, register SV *const sv2,
7374 #ifdef USE_LOCALE_COLLATE
7380 if (PL_collation_standard)
7384 pv1 = sv1 ? sv_collxfrm_flags(sv1, &len1, flags) : (char *) NULL;
7386 pv2 = sv2 ? sv_collxfrm_flags(sv2, &len2, flags) : (char *) NULL;
7388 if (!pv1 || !len1) {
7399 retval = memcmp((void*)pv1, (void*)pv2, len1 < len2 ? len1 : len2);
7402 return retval < 0 ? -1 : 1;
7405 * When the result of collation is equality, that doesn't mean
7406 * that there are no differences -- some locales exclude some
7407 * characters from consideration. So to avoid false equalities,
7408 * we use the raw string as a tiebreaker.
7414 #endif /* USE_LOCALE_COLLATE */
7416 return sv_cmp(sv1, sv2);
7420 #ifdef USE_LOCALE_COLLATE
7423 =for apidoc sv_collxfrm
7425 This calls C<sv_collxfrm_flags> with the SV_GMAGIC flag. See
7426 C<sv_collxfrm_flags>.
7428 =for apidoc sv_collxfrm_flags
7430 Add Collate Transform magic to an SV if it doesn't already have it. If the
7431 flags contain SV_GMAGIC, it handles get-magic.
7433 Any scalar variable may carry PERL_MAGIC_collxfrm magic that contains the
7434 scalar data of the variable, but transformed to such a format that a normal
7435 memory comparison can be used to compare the data according to the locale
7442 Perl_sv_collxfrm_flags(pTHX_ SV *const sv, STRLEN *const nxp, const I32 flags)
7447 PERL_ARGS_ASSERT_SV_COLLXFRM_FLAGS;
7449 mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_collxfrm) : (MAGIC *) NULL;
7450 if (!mg || !mg->mg_ptr || *(U32*)mg->mg_ptr != PL_collation_ix) {
7456 Safefree(mg->mg_ptr);
7457 s = SvPV_flags_const(sv, len, flags);
7458 if ((xf = mem_collxfrm(s, len, &xlen))) {
7460 #ifdef PERL_OLD_COPY_ON_WRITE
7462 sv_force_normal_flags(sv, 0);
7464 mg = sv_magicext(sv, 0, PERL_MAGIC_collxfrm, &PL_vtbl_collxfrm,
7478 if (mg && mg->mg_ptr) {
7480 return mg->mg_ptr + sizeof(PL_collation_ix);
7488 #endif /* USE_LOCALE_COLLATE */
7491 S_sv_gets_append_to_utf8(pTHX_ SV *const sv, PerlIO *const fp, I32 append)
7493 SV * const tsv = newSV(0);
7496 sv_gets(tsv, fp, 0);
7497 sv_utf8_upgrade_nomg(tsv);
7498 SvCUR_set(sv,append);
7501 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7505 S_sv_gets_read_record(pTHX_ SV *const sv, PerlIO *const fp, I32 append)
7508 const U32 recsize = SvUV(SvRV(PL_rs)); /* RsRECORD() guarantees > 0. */
7509 /* Grab the size of the record we're getting */
7510 char *const buffer = SvGROW(sv, (STRLEN)(recsize + append + 1)) + append;
7517 /* VMS wants read instead of fread, because fread doesn't respect */
7518 /* RMS record boundaries. This is not necessarily a good thing to be */
7519 /* doing, but we've got no other real choice - except avoid stdio
7520 as implementation - perhaps write a :vms layer ?
7522 fd = PerlIO_fileno(fp);
7524 bytesread = PerlLIO_read(fd, buffer, recsize);
7526 else /* in-memory file from PerlIO::Scalar */
7529 bytesread = PerlIO_read(fp, buffer, recsize);
7534 SvCUR_set(sv, bytesread + append);
7535 buffer[bytesread] = '\0';
7536 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7542 Get a line from the filehandle and store it into the SV, optionally
7543 appending to the currently-stored string.
7549 Perl_sv_gets(pTHX_ register SV *const sv, register PerlIO *const fp, I32 append)
7554 register STDCHAR rslast;
7555 register STDCHAR *bp;
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);
7574 if (PerlIO_isutf8(fp)) {
7576 sv_utf8_upgrade_nomg(sv);
7577 sv_pos_u2b(sv,&append,0);
7579 } else if (SvUTF8(sv)) {
7580 return S_sv_gets_append_to_utf8(aTHX_ sv, fp, append);
7588 if (PerlIO_isutf8(fp))
7591 if (IN_PERL_COMPILETIME) {
7592 /* we always read code in line mode */
7596 else if (RsSNARF(PL_rs)) {
7597 /* If it is a regular disk file use size from stat() as estimate
7598 of amount we are going to read -- may result in mallocing
7599 more memory than we really need if the layers below reduce
7600 the size we read (e.g. CRLF or a gzip layer).
7603 if (!PerlLIO_fstat(PerlIO_fileno(fp), &st) && S_ISREG(st.st_mode)) {
7604 const Off_t offset = PerlIO_tell(fp);
7605 if (offset != (Off_t) -1 && st.st_size + append > offset) {
7606 (void) SvGROW(sv, (STRLEN)((st.st_size - offset) + append + 1));
7612 else if (RsRECORD(PL_rs)) {
7613 return S_sv_gets_read_record(aTHX_ sv, fp, append);
7615 else if (RsPARA(PL_rs)) {
7621 /* Get $/ i.e. PL_rs into same encoding as stream wants */
7622 if (PerlIO_isutf8(fp)) {
7623 rsptr = SvPVutf8(PL_rs, rslen);
7626 if (SvUTF8(PL_rs)) {
7627 if (!sv_utf8_downgrade(PL_rs, TRUE)) {
7628 Perl_croak(aTHX_ "Wide character in $/");
7631 rsptr = SvPV_const(PL_rs, rslen);
7635 rslast = rslen ? rsptr[rslen - 1] : '\0';
7637 if (rspara) { /* have to do this both before and after */
7638 do { /* to make sure file boundaries work right */
7641 i = PerlIO_getc(fp);
7645 PerlIO_ungetc(fp,i);
7651 /* See if we know enough about I/O mechanism to cheat it ! */
7653 /* This used to be #ifdef test - it is made run-time test for ease
7654 of abstracting out stdio interface. One call should be cheap
7655 enough here - and may even be a macro allowing compile
7659 if (PerlIO_fast_gets(fp)) {
7662 * We're going to steal some values from the stdio struct
7663 * and put EVERYTHING in the innermost loop into registers.
7665 register STDCHAR *ptr;
7669 #if defined(VMS) && defined(PERLIO_IS_STDIO)
7670 /* An ungetc()d char is handled separately from the regular
7671 * buffer, so we getc() it back out and stuff it in the buffer.
7673 i = PerlIO_getc(fp);
7674 if (i == EOF) return 0;
7675 *(--((*fp)->_ptr)) = (unsigned char) i;
7679 /* Here is some breathtakingly efficient cheating */
7681 cnt = PerlIO_get_cnt(fp); /* get count into register */
7682 /* make sure we have the room */
7683 if ((I32)(SvLEN(sv) - append) <= cnt + 1) {
7684 /* Not room for all of it
7685 if we are looking for a separator and room for some
7687 if (rslen && cnt > 80 && (I32)SvLEN(sv) > append) {
7688 /* just process what we have room for */
7689 shortbuffered = cnt - SvLEN(sv) + append + 1;
7690 cnt -= shortbuffered;
7694 /* remember that cnt can be negative */
7695 SvGROW(sv, (STRLEN)(append + (cnt <= 0 ? 2 : (cnt + 1))));
7700 bp = (STDCHAR*)SvPVX_const(sv) + append; /* move these two too to registers */
7701 ptr = (STDCHAR*)PerlIO_get_ptr(fp);
7702 DEBUG_P(PerlIO_printf(Perl_debug_log,
7703 "Screamer: entering, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7704 DEBUG_P(PerlIO_printf(Perl_debug_log,
7705 "Screamer: entering: PerlIO * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7706 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7707 PTR2UV(PerlIO_has_base(fp) ? PerlIO_get_base(fp) : 0)));
7712 while (cnt > 0) { /* this | eat */
7714 if ((*bp++ = *ptr++) == rslast) /* really | dust */
7715 goto thats_all_folks; /* screams | sed :-) */
7719 Copy(ptr, bp, cnt, char); /* this | eat */
7720 bp += cnt; /* screams | dust */
7721 ptr += cnt; /* louder | sed :-) */
7723 assert (!shortbuffered);
7724 goto cannot_be_shortbuffered;
7728 if (shortbuffered) { /* oh well, must extend */
7729 cnt = shortbuffered;
7731 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
7733 SvGROW(sv, SvLEN(sv) + append + cnt + 2);
7734 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
7738 cannot_be_shortbuffered:
7739 DEBUG_P(PerlIO_printf(Perl_debug_log,
7740 "Screamer: going to getc, ptr=%"UVuf", cnt=%ld\n",
7741 PTR2UV(ptr),(long)cnt));
7742 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* deregisterize cnt and ptr */
7744 DEBUG_Pv(PerlIO_printf(Perl_debug_log,
7745 "Screamer: pre: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7746 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7747 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7749 /* This used to call 'filbuf' in stdio form, but as that behaves like
7750 getc when cnt <= 0 we use PerlIO_getc here to avoid introducing
7751 another abstraction. */
7752 i = PerlIO_getc(fp); /* get more characters */
7754 DEBUG_Pv(PerlIO_printf(Perl_debug_log,
7755 "Screamer: post: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7756 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7757 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7759 cnt = PerlIO_get_cnt(fp);
7760 ptr = (STDCHAR*)PerlIO_get_ptr(fp); /* reregisterize cnt and ptr */
7761 DEBUG_P(PerlIO_printf(Perl_debug_log,
7762 "Screamer: after getc, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7764 if (i == EOF) /* all done for ever? */
7765 goto thats_really_all_folks;
7767 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
7769 SvGROW(sv, bpx + cnt + 2);
7770 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
7772 *bp++ = (STDCHAR)i; /* store character from PerlIO_getc */
7774 if (rslen && (STDCHAR)i == rslast) /* all done for now? */
7775 goto thats_all_folks;
7779 if ((rslen > 1 && (STRLEN)(bp - (STDCHAR*)SvPVX_const(sv)) < rslen) ||
7780 memNE((char*)bp - rslen, rsptr, rslen))
7781 goto screamer; /* go back to the fray */
7782 thats_really_all_folks:
7784 cnt += shortbuffered;
7785 DEBUG_P(PerlIO_printf(Perl_debug_log,
7786 "Screamer: quitting, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7787 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* put these back or we're in trouble */
7788 DEBUG_P(PerlIO_printf(Perl_debug_log,
7789 "Screamer: end: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7790 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7791 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7793 SvCUR_set(sv, bp - (STDCHAR*)SvPVX_const(sv)); /* set length */
7794 DEBUG_P(PerlIO_printf(Perl_debug_log,
7795 "Screamer: done, len=%ld, string=|%.*s|\n",
7796 (long)SvCUR(sv),(int)SvCUR(sv),SvPVX_const(sv)));
7800 /*The big, slow, and stupid way. */
7801 #ifdef USE_HEAP_INSTEAD_OF_STACK /* Even slower way. */
7802 STDCHAR *buf = NULL;
7803 Newx(buf, 8192, STDCHAR);
7811 register const STDCHAR * const bpe = buf + sizeof(buf);
7813 while ((i = PerlIO_getc(fp)) != EOF && (*bp++ = (STDCHAR)i) != rslast && bp < bpe)
7814 ; /* keep reading */
7818 cnt = PerlIO_read(fp,(char*)buf, sizeof(buf));
7819 /* Accommodate broken VAXC compiler, which applies U8 cast to
7820 * both args of ?: operator, causing EOF to change into 255
7823 i = (U8)buf[cnt - 1];
7829 cnt = 0; /* we do need to re-set the sv even when cnt <= 0 */
7831 sv_catpvn(sv, (char *) buf, cnt);
7833 sv_setpvn(sv, (char *) buf, cnt);
7835 if (i != EOF && /* joy */
7837 SvCUR(sv) < rslen ||
7838 memNE(SvPVX_const(sv) + SvCUR(sv) - rslen, rsptr, rslen)))
7842 * If we're reading from a TTY and we get a short read,
7843 * indicating that the user hit his EOF character, we need
7844 * to notice it now, because if we try to read from the TTY
7845 * again, the EOF condition will disappear.
7847 * The comparison of cnt to sizeof(buf) is an optimization
7848 * that prevents unnecessary calls to feof().
7852 if (!(cnt < (I32)sizeof(buf) && PerlIO_eof(fp)))
7856 #ifdef USE_HEAP_INSTEAD_OF_STACK
7861 if (rspara) { /* have to do this both before and after */
7862 while (i != EOF) { /* to make sure file boundaries work right */
7863 i = PerlIO_getc(fp);
7865 PerlIO_ungetc(fp,i);
7871 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7877 Auto-increment of the value in the SV, doing string to numeric conversion
7878 if necessary. Handles 'get' magic and operator overloading.
7884 Perl_sv_inc(pTHX_ register SV *const sv)
7893 =for apidoc sv_inc_nomg
7895 Auto-increment of the value in the SV, doing string to numeric conversion
7896 if necessary. Handles operator overloading. Skips handling 'get' magic.
7902 Perl_sv_inc_nomg(pTHX_ register SV *const sv)
7910 if (SvTHINKFIRST(sv)) {
7911 if (SvIsCOW(sv) || isGV_with_GP(sv))
7912 sv_force_normal_flags(sv, 0);
7913 if (SvREADONLY(sv)) {
7914 if (IN_PERL_RUNTIME)
7915 Perl_croak_no_modify(aTHX);
7919 if (SvAMAGIC(sv) && AMG_CALLunary(sv, inc_amg))
7921 i = PTR2IV(SvRV(sv));
7926 flags = SvFLAGS(sv);
7927 if ((flags & (SVp_NOK|SVp_IOK)) == SVp_NOK) {
7928 /* It's (privately or publicly) a float, but not tested as an
7929 integer, so test it to see. */
7931 flags = SvFLAGS(sv);
7933 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
7934 /* It's publicly an integer, or privately an integer-not-float */
7935 #ifdef PERL_PRESERVE_IVUV
7939 if (SvUVX(sv) == UV_MAX)
7940 sv_setnv(sv, UV_MAX_P1);
7942 (void)SvIOK_only_UV(sv);
7943 SvUV_set(sv, SvUVX(sv) + 1);
7945 if (SvIVX(sv) == IV_MAX)
7946 sv_setuv(sv, (UV)IV_MAX + 1);
7948 (void)SvIOK_only(sv);
7949 SvIV_set(sv, SvIVX(sv) + 1);
7954 if (flags & SVp_NOK) {
7955 const NV was = SvNVX(sv);
7956 if (NV_OVERFLOWS_INTEGERS_AT &&
7957 was >= NV_OVERFLOWS_INTEGERS_AT) {
7958 /* diag_listed_as: Lost precision when %s %f by 1 */
7959 Perl_ck_warner(aTHX_ packWARN(WARN_IMPRECISION),
7960 "Lost precision when incrementing %" NVff " by 1",
7963 (void)SvNOK_only(sv);
7964 SvNV_set(sv, was + 1.0);
7968 if (!(flags & SVp_POK) || !*SvPVX_const(sv)) {
7969 if ((flags & SVTYPEMASK) < SVt_PVIV)
7970 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV ? SVt_PVIV : SVt_IV));
7971 (void)SvIOK_only(sv);
7976 while (isALPHA(*d)) d++;
7977 while (isDIGIT(*d)) d++;
7978 if (d < SvEND(sv)) {
7979 #ifdef PERL_PRESERVE_IVUV
7980 /* Got to punt this as an integer if needs be, but we don't issue
7981 warnings. Probably ought to make the sv_iv_please() that does
7982 the conversion if possible, and silently. */
7983 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
7984 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
7985 /* Need to try really hard to see if it's an integer.
7986 9.22337203685478e+18 is an integer.
7987 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
7988 so $a="9.22337203685478e+18"; $a+0; $a++
7989 needs to be the same as $a="9.22337203685478e+18"; $a++
7996 /* sv_2iv *should* have made this an NV */
7997 if (flags & SVp_NOK) {
7998 (void)SvNOK_only(sv);
7999 SvNV_set(sv, SvNVX(sv) + 1.0);
8002 /* I don't think we can get here. Maybe I should assert this
8003 And if we do get here I suspect that sv_setnv will croak. NWC
8005 #if defined(USE_LONG_DOUBLE)
8006 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",
8007 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
8009 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_inc punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
8010 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
8013 #endif /* PERL_PRESERVE_IVUV */
8014 sv_setnv(sv,Atof(SvPVX_const(sv)) + 1.0);
8018 while (d >= SvPVX_const(sv)) {
8026 /* MKS: The original code here died if letters weren't consecutive.
8027 * at least it didn't have to worry about non-C locales. The
8028 * new code assumes that ('z'-'a')==('Z'-'A'), letters are
8029 * arranged in order (although not consecutively) and that only
8030 * [A-Za-z] are accepted by isALPHA in the C locale.
8032 if (*d != 'z' && *d != 'Z') {
8033 do { ++*d; } while (!isALPHA(*d));
8036 *(d--) -= 'z' - 'a';
8041 *(d--) -= 'z' - 'a' + 1;
8045 /* oh,oh, the number grew */
8046 SvGROW(sv, SvCUR(sv) + 2);
8047 SvCUR_set(sv, SvCUR(sv) + 1);
8048 for (d = SvPVX(sv) + SvCUR(sv); d > SvPVX_const(sv); d--)
8059 Auto-decrement of the value in the SV, doing string to numeric conversion
8060 if necessary. Handles 'get' magic and operator overloading.
8066 Perl_sv_dec(pTHX_ register SV *const sv)
8076 =for apidoc sv_dec_nomg
8078 Auto-decrement of the value in the SV, doing string to numeric conversion
8079 if necessary. Handles operator overloading. Skips handling 'get' magic.
8085 Perl_sv_dec_nomg(pTHX_ register SV *const sv)
8092 if (SvTHINKFIRST(sv)) {
8093 if (SvIsCOW(sv) || isGV_with_GP(sv))
8094 sv_force_normal_flags(sv, 0);
8095 if (SvREADONLY(sv)) {
8096 if (IN_PERL_RUNTIME)
8097 Perl_croak_no_modify(aTHX);
8101 if (SvAMAGIC(sv) && AMG_CALLunary(sv, dec_amg))
8103 i = PTR2IV(SvRV(sv));
8108 /* Unlike sv_inc we don't have to worry about string-never-numbers
8109 and keeping them magic. But we mustn't warn on punting */
8110 flags = SvFLAGS(sv);
8111 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
8112 /* It's publicly an integer, or privately an integer-not-float */
8113 #ifdef PERL_PRESERVE_IVUV
8117 if (SvUVX(sv) == 0) {
8118 (void)SvIOK_only(sv);
8122 (void)SvIOK_only_UV(sv);
8123 SvUV_set(sv, SvUVX(sv) - 1);
8126 if (SvIVX(sv) == IV_MIN) {
8127 sv_setnv(sv, (NV)IV_MIN);
8131 (void)SvIOK_only(sv);
8132 SvIV_set(sv, SvIVX(sv) - 1);
8137 if (flags & SVp_NOK) {
8140 const NV was = SvNVX(sv);
8141 if (NV_OVERFLOWS_INTEGERS_AT &&
8142 was <= -NV_OVERFLOWS_INTEGERS_AT) {
8143 /* diag_listed_as: Lost precision when %s %f by 1 */
8144 Perl_ck_warner(aTHX_ packWARN(WARN_IMPRECISION),
8145 "Lost precision when decrementing %" NVff " by 1",
8148 (void)SvNOK_only(sv);
8149 SvNV_set(sv, was - 1.0);
8153 if (!(flags & SVp_POK)) {
8154 if ((flags & SVTYPEMASK) < SVt_PVIV)
8155 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV) ? SVt_PVIV : SVt_IV);
8157 (void)SvIOK_only(sv);
8160 #ifdef PERL_PRESERVE_IVUV
8162 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
8163 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
8164 /* Need to try really hard to see if it's an integer.
8165 9.22337203685478e+18 is an integer.
8166 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
8167 so $a="9.22337203685478e+18"; $a+0; $a--
8168 needs to be the same as $a="9.22337203685478e+18"; $a--
8175 /* sv_2iv *should* have made this an NV */
8176 if (flags & SVp_NOK) {
8177 (void)SvNOK_only(sv);
8178 SvNV_set(sv, SvNVX(sv) - 1.0);
8181 /* I don't think we can get here. Maybe I should assert this
8182 And if we do get here I suspect that sv_setnv will croak. NWC
8184 #if defined(USE_LONG_DOUBLE)
8185 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",
8186 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
8188 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_dec punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
8189 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
8193 #endif /* PERL_PRESERVE_IVUV */
8194 sv_setnv(sv,Atof(SvPVX_const(sv)) - 1.0); /* punt */
8197 /* this define is used to eliminate a chunk of duplicated but shared logic
8198 * it has the suffix __SV_C to signal that it isnt API, and isnt meant to be
8199 * used anywhere but here - yves
8201 #define PUSH_EXTEND_MORTAL__SV_C(AnSv) \
8204 PL_tmps_stack[++PL_tmps_ix] = (AnSv); \
8208 =for apidoc sv_mortalcopy
8210 Creates a new SV which is a copy of the original SV (using C<sv_setsv>).
8211 The new SV is marked as mortal. It will be destroyed "soon", either by an
8212 explicit call to FREETMPS, or by an implicit call at places such as
8213 statement boundaries. See also C<sv_newmortal> and C<sv_2mortal>.
8218 /* Make a string that will exist for the duration of the expression
8219 * evaluation. Actually, it may have to last longer than that, but
8220 * hopefully we won't free it until it has been assigned to a
8221 * permanent location. */
8224 Perl_sv_mortalcopy(pTHX_ SV *const oldstr)
8230 sv_setsv(sv,oldstr);
8231 PUSH_EXTEND_MORTAL__SV_C(sv);
8237 =for apidoc sv_newmortal
8239 Creates a new null SV which is mortal. The reference count of the SV is
8240 set to 1. It will be destroyed "soon", either by an explicit call to
8241 FREETMPS, or by an implicit call at places such as statement boundaries.
8242 See also C<sv_mortalcopy> and C<sv_2mortal>.
8248 Perl_sv_newmortal(pTHX)
8254 SvFLAGS(sv) = SVs_TEMP;
8255 PUSH_EXTEND_MORTAL__SV_C(sv);
8261 =for apidoc newSVpvn_flags
8263 Creates a new SV and copies a string into it. The reference count for the
8264 SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
8265 string. You are responsible for ensuring that the source string is at least
8266 C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
8267 Currently the only flag bits accepted are C<SVf_UTF8> and C<SVs_TEMP>.
8268 If C<SVs_TEMP> is set, then C<sv_2mortal()> is called on the result before
8269 returning. If C<SVf_UTF8> is set, C<s>
8270 is considered to be in UTF-8 and the
8271 C<SVf_UTF8> flag will be set on the new SV.
8272 C<newSVpvn_utf8()> is a convenience wrapper for this function, defined as
8274 #define newSVpvn_utf8(s, len, u) \
8275 newSVpvn_flags((s), (len), (u) ? SVf_UTF8 : 0)
8281 Perl_newSVpvn_flags(pTHX_ const char *const s, const STRLEN len, const U32 flags)
8286 /* All the flags we don't support must be zero.
8287 And we're new code so I'm going to assert this from the start. */
8288 assert(!(flags & ~(SVf_UTF8|SVs_TEMP)));
8290 sv_setpvn(sv,s,len);
8292 /* This code used to a sv_2mortal(), however we now unroll the call to sv_2mortal()
8293 * and do what it does ourselves here.
8294 * Since we have asserted that flags can only have the SVf_UTF8 and/or SVs_TEMP flags
8295 * set above we can use it to enable the sv flags directly (bypassing SvTEMP_on), which
8296 * in turn means we dont need to mask out the SVf_UTF8 flag below, which means that we
8297 * eliminate quite a few steps than it looks - Yves (explaining patch by gfx)
8300 SvFLAGS(sv) |= flags;
8302 if(flags & SVs_TEMP){
8303 PUSH_EXTEND_MORTAL__SV_C(sv);
8310 =for apidoc sv_2mortal
8312 Marks an existing SV as mortal. The SV will be destroyed "soon", either
8313 by an explicit call to FREETMPS, or by an implicit call at places such as
8314 statement boundaries. SvTEMP() is turned on which means that the SV's
8315 string buffer can be "stolen" if this SV is copied. See also C<sv_newmortal>
8316 and C<sv_mortalcopy>.
8322 Perl_sv_2mortal(pTHX_ register SV *const sv)
8327 if (SvREADONLY(sv) && SvIMMORTAL(sv))
8329 PUSH_EXTEND_MORTAL__SV_C(sv);
8337 Creates a new SV and copies a string into it. The reference count for the
8338 SV is set to 1. If C<len> is zero, Perl will compute the length using
8339 strlen(). For efficiency, consider using C<newSVpvn> instead.
8345 Perl_newSVpv(pTHX_ const char *const s, const STRLEN len)
8351 sv_setpvn(sv, s, len || s == NULL ? len : strlen(s));
8356 =for apidoc newSVpvn
8358 Creates a new SV and copies a string into it. The reference count for the
8359 SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
8360 string. You are responsible for ensuring that the source string is at least
8361 C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
8367 Perl_newSVpvn(pTHX_ const char *const s, const STRLEN len)
8373 sv_setpvn(sv,s,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 register 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)) {
8956 gv_efullname3(tmpsv, gv, NULL);
8957 /* XXX this is probably not what they think they're getting.
8958 * It has the same effect as "sub name;", i.e. just a forward
8960 newSUB(start_subparse(FALSE, 0),
8961 newSVOP(OP_CONST, 0, tmpsv),
8965 Perl_croak(aTHX_ "Unable to create sub named \"%"SVf"\"",
8966 SVfARG(SvOK(sv) ? sv : &PL_sv_no));
8975 Returns true if the SV has a true value by Perl's rules.
8976 Use the C<SvTRUE> macro instead, which may call C<sv_true()> or may
8977 instead use an in-line version.
8983 Perl_sv_true(pTHX_ register SV *const sv)
8988 register const XPV* const tXpv = (XPV*)SvANY(sv);
8990 (tXpv->xpv_cur > 1 ||
8991 (tXpv->xpv_cur && *sv->sv_u.svu_pv != '0')))
8998 return SvIVX(sv) != 0;
9001 return SvNVX(sv) != 0.0;
9003 return sv_2bool(sv);
9009 =for apidoc sv_pvn_force
9011 Get a sensible string out of the SV somehow.
9012 A private implementation of the C<SvPV_force> macro for compilers which
9013 can't cope with complex macro expressions. Always use the macro instead.
9015 =for apidoc sv_pvn_force_flags
9017 Get a sensible string out of the SV somehow.
9018 If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<sv> if
9019 appropriate, else not. C<sv_pvn_force> and C<sv_pvn_force_nomg> are
9020 implemented in terms of this function.
9021 You normally want to use the various wrapper macros instead: see
9022 C<SvPV_force> and C<SvPV_force_nomg>
9028 Perl_sv_pvn_force_flags(pTHX_ SV *const sv, STRLEN *const lp, const I32 flags)
9032 PERL_ARGS_ASSERT_SV_PVN_FORCE_FLAGS;
9034 if (flags & SV_GMAGIC) SvGETMAGIC(sv);
9035 if (SvTHINKFIRST(sv) && !SvROK(sv))
9036 sv_force_normal_flags(sv, 0);
9046 if (SvREADONLY(sv) && !(flags & SV_MUTABLE_RETURN)) {
9047 const char * const ref = sv_reftype(sv,0);
9049 Perl_croak(aTHX_ "Can't coerce readonly %s to string in %s",
9050 ref, OP_DESC(PL_op));
9052 Perl_croak(aTHX_ "Can't coerce readonly %s to string", ref);
9054 if ((SvTYPE(sv) > SVt_PVLV && SvTYPE(sv) != SVt_PVFM)
9055 || isGV_with_GP(sv))
9056 /* diag_listed_as: Can't coerce %s to %s in %s */
9057 Perl_croak(aTHX_ "Can't coerce %s to string in %s", sv_reftype(sv,0),
9059 s = sv_2pv_flags(sv, &len, flags &~ SV_GMAGIC);
9063 if (s != SvPVX_const(sv)) { /* Almost, but not quite, sv_setpvn() */
9066 SvUPGRADE(sv, SVt_PV); /* Never FALSE */
9067 SvGROW(sv, len + 1);
9068 Move(s,SvPVX(sv),len,char);
9070 SvPVX(sv)[len] = '\0';
9073 SvPOK_on(sv); /* validate pointer */
9075 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
9076 PTR2UV(sv),SvPVX_const(sv)));
9079 return SvPVX_mutable(sv);
9083 =for apidoc sv_pvbyten_force
9085 The backend for the C<SvPVbytex_force> macro. Always use the macro
9092 Perl_sv_pvbyten_force(pTHX_ SV *const sv, STRLEN *const lp)
9094 PERL_ARGS_ASSERT_SV_PVBYTEN_FORCE;
9096 sv_pvn_force(sv,lp);
9097 sv_utf8_downgrade(sv,0);
9103 =for apidoc sv_pvutf8n_force
9105 The backend for the C<SvPVutf8x_force> macro. Always use the macro
9112 Perl_sv_pvutf8n_force(pTHX_ SV *const sv, STRLEN *const lp)
9114 PERL_ARGS_ASSERT_SV_PVUTF8N_FORCE;
9116 sv_pvn_force(sv,lp);
9117 sv_utf8_upgrade(sv);
9123 =for apidoc sv_reftype
9125 Returns a string describing what the SV is a reference to.
9131 Perl_sv_reftype(pTHX_ const SV *const sv, const int ob)
9133 PERL_ARGS_ASSERT_SV_REFTYPE;
9134 if (ob && SvOBJECT(sv)) {
9135 return SvPV_nolen_const(sv_ref(NULL, sv, ob));
9138 switch (SvTYPE(sv)) {
9153 case SVt_PVLV: return (char *) (SvROK(sv) ? "REF"
9154 /* tied lvalues should appear to be
9155 * scalars for backwards compatibility */
9156 : (LvTYPE(sv) == 't' || LvTYPE(sv) == 'T')
9157 ? "SCALAR" : "LVALUE");
9158 case SVt_PVAV: return "ARRAY";
9159 case SVt_PVHV: return "HASH";
9160 case SVt_PVCV: return "CODE";
9161 case SVt_PVGV: return (char *) (isGV_with_GP(sv)
9162 ? "GLOB" : "SCALAR");
9163 case SVt_PVFM: return "FORMAT";
9164 case SVt_PVIO: return "IO";
9165 case SVt_BIND: return "BIND";
9166 case SVt_REGEXP: return "REGEXP";
9167 default: return "UNKNOWN";
9175 Returns a SV describing what the SV passed in is a reference to.
9181 Perl_sv_ref(pTHX_ register SV *dst, const SV *const sv, const int ob)
9183 PERL_ARGS_ASSERT_SV_REF;
9186 dst = sv_newmortal();
9188 if (ob && SvOBJECT(sv)) {
9189 HvNAME_get(SvSTASH(sv))
9190 ? sv_sethek(dst, HvNAME_HEK(SvSTASH(sv)))
9191 : sv_setpvn(dst, "__ANON__", 8);
9194 const char * reftype = sv_reftype(sv, 0);
9195 sv_setpv(dst, reftype);
9201 =for apidoc sv_isobject
9203 Returns a boolean indicating whether the SV is an RV pointing to a blessed
9204 object. If the SV is not an RV, or if the object is not blessed, then this
9211 Perl_sv_isobject(pTHX_ SV *sv)
9227 Returns a boolean indicating whether the SV is blessed into the specified
9228 class. This does not check for subtypes; use C<sv_derived_from> to verify
9229 an inheritance relationship.
9235 Perl_sv_isa(pTHX_ SV *sv, const char *const name)
9239 PERL_ARGS_ASSERT_SV_ISA;
9249 hvname = HvNAME_get(SvSTASH(sv));
9253 return strEQ(hvname, name);
9259 Creates a new SV for the RV, C<rv>, to point to. If C<rv> is not an RV then
9260 it will be upgraded to one. If C<classname> is non-null then the new SV will
9261 be blessed in the specified package. The new SV is returned and its
9262 reference count is 1.
9268 Perl_newSVrv(pTHX_ SV *const rv, const char *const classname)
9273 PERL_ARGS_ASSERT_NEWSVRV;
9277 SV_CHECK_THINKFIRST_COW_DROP(rv);
9278 (void)SvAMAGIC_off(rv);
9280 if (SvTYPE(rv) >= SVt_PVMG) {
9281 const U32 refcnt = SvREFCNT(rv);
9285 SvREFCNT(rv) = refcnt;
9287 sv_upgrade(rv, SVt_IV);
9288 } else if (SvROK(rv)) {
9289 SvREFCNT_dec(SvRV(rv));
9291 prepare_SV_for_RV(rv);
9299 HV* const stash = gv_stashpv(classname, GV_ADD);
9300 (void)sv_bless(rv, stash);
9306 =for apidoc sv_setref_pv
9308 Copies a pointer into a new SV, optionally blessing the SV. The C<rv>
9309 argument will be upgraded to an RV. That RV will be modified to point to
9310 the new SV. If the C<pv> argument is NULL then C<PL_sv_undef> will be placed
9311 into the SV. The C<classname> argument indicates the package for the
9312 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9313 will have a reference count of 1, and the RV will be returned.
9315 Do not use with other Perl types such as HV, AV, SV, CV, because those
9316 objects will become corrupted by the pointer copy process.
9318 Note that C<sv_setref_pvn> copies the string while this copies the pointer.
9324 Perl_sv_setref_pv(pTHX_ SV *const rv, const char *const classname, void *const pv)
9328 PERL_ARGS_ASSERT_SV_SETREF_PV;
9331 sv_setsv(rv, &PL_sv_undef);
9335 sv_setiv(newSVrv(rv,classname), PTR2IV(pv));
9340 =for apidoc sv_setref_iv
9342 Copies an integer into a new SV, optionally blessing the SV. The C<rv>
9343 argument will be upgraded to an RV. That RV will be modified to point to
9344 the new SV. The C<classname> argument indicates the package for the
9345 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9346 will have a reference count of 1, and the RV will be returned.
9352 Perl_sv_setref_iv(pTHX_ SV *const rv, const char *const classname, const IV iv)
9354 PERL_ARGS_ASSERT_SV_SETREF_IV;
9356 sv_setiv(newSVrv(rv,classname), iv);
9361 =for apidoc sv_setref_uv
9363 Copies an unsigned integer into a new SV, optionally blessing the SV. The C<rv>
9364 argument will be upgraded to an RV. That RV will be modified to point to
9365 the new SV. The C<classname> argument indicates the package for the
9366 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9367 will have a reference count of 1, and the RV will be returned.
9373 Perl_sv_setref_uv(pTHX_ SV *const rv, const char *const classname, const UV uv)
9375 PERL_ARGS_ASSERT_SV_SETREF_UV;
9377 sv_setuv(newSVrv(rv,classname), uv);
9382 =for apidoc sv_setref_nv
9384 Copies a double into a new SV, optionally blessing the SV. The C<rv>
9385 argument will be upgraded to an RV. That RV will be modified to point to
9386 the new SV. The C<classname> argument indicates the package for the
9387 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9388 will have a reference count of 1, and the RV will be returned.
9394 Perl_sv_setref_nv(pTHX_ SV *const rv, const char *const classname, const NV nv)
9396 PERL_ARGS_ASSERT_SV_SETREF_NV;
9398 sv_setnv(newSVrv(rv,classname), nv);
9403 =for apidoc sv_setref_pvn
9405 Copies a string into a new SV, optionally blessing the SV. The length of the
9406 string must be specified with C<n>. The C<rv> argument will be upgraded to
9407 an RV. That RV will be modified to point to the new SV. The C<classname>
9408 argument indicates the package for the blessing. Set C<classname> to
9409 C<NULL> to avoid the blessing. The new SV will have a reference count
9410 of 1, and the RV will be returned.
9412 Note that C<sv_setref_pv> copies the pointer while this copies the string.
9418 Perl_sv_setref_pvn(pTHX_ SV *const rv, const char *const classname,
9419 const char *const pv, const STRLEN n)
9421 PERL_ARGS_ASSERT_SV_SETREF_PVN;
9423 sv_setpvn(newSVrv(rv,classname), pv, n);
9428 =for apidoc sv_bless
9430 Blesses an SV into a specified package. The SV must be an RV. The package
9431 must be designated by its stash (see C<gv_stashpv()>). The reference count
9432 of the SV is unaffected.
9438 Perl_sv_bless(pTHX_ SV *const sv, HV *const stash)
9443 PERL_ARGS_ASSERT_SV_BLESS;
9446 Perl_croak(aTHX_ "Can't bless non-reference value");
9448 if (SvFLAGS(tmpRef) & (SVs_OBJECT|SVf_READONLY)) {
9449 if (SvIsCOW(tmpRef))
9450 sv_force_normal_flags(tmpRef, 0);
9451 if (SvREADONLY(tmpRef))
9452 Perl_croak_no_modify(aTHX);
9453 if (SvOBJECT(tmpRef)) {
9454 if (SvTYPE(tmpRef) != SVt_PVIO)
9456 SvREFCNT_dec(SvSTASH(tmpRef));
9459 SvOBJECT_on(tmpRef);
9460 if (SvTYPE(tmpRef) != SVt_PVIO)
9462 SvUPGRADE(tmpRef, SVt_PVMG);
9463 SvSTASH_set(tmpRef, MUTABLE_HV(SvREFCNT_inc_simple(stash)));
9468 (void)SvAMAGIC_off(sv);
9470 if(SvSMAGICAL(tmpRef))
9471 if(mg_find(tmpRef, PERL_MAGIC_ext) || mg_find(tmpRef, PERL_MAGIC_uvar))
9479 /* Downgrades a PVGV to a PVMG. If it's actually a PVLV, we leave the type
9480 * as it is after unglobbing it.
9483 PERL_STATIC_INLINE void
9484 S_sv_unglob(pTHX_ SV *const sv, U32 flags)
9489 SV * const temp = flags & SV_COW_DROP_PV ? NULL : sv_newmortal();
9491 PERL_ARGS_ASSERT_SV_UNGLOB;
9493 assert(SvTYPE(sv) == SVt_PVGV || SvTYPE(sv) == SVt_PVLV);
9495 if (!(flags & SV_COW_DROP_PV))
9496 gv_efullname3(temp, MUTABLE_GV(sv), "*");
9499 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
9500 && HvNAME_get(stash))
9501 mro_method_changed_in(stash);
9502 gp_free(MUTABLE_GV(sv));
9505 sv_del_backref(MUTABLE_SV(GvSTASH(sv)), sv);
9509 if (GvNAME_HEK(sv)) {
9510 unshare_hek(GvNAME_HEK(sv));
9512 isGV_with_GP_off(sv);
9514 if(SvTYPE(sv) == SVt_PVGV) {
9515 /* need to keep SvANY(sv) in the right arena */
9516 xpvmg = new_XPVMG();
9517 StructCopy(SvANY(sv), xpvmg, XPVMG);
9518 del_XPVGV(SvANY(sv));
9521 SvFLAGS(sv) &= ~SVTYPEMASK;
9522 SvFLAGS(sv) |= SVt_PVMG;
9525 /* Intentionally not calling any local SET magic, as this isn't so much a
9526 set operation as merely an internal storage change. */
9527 if (flags & SV_COW_DROP_PV) SvOK_off(sv);
9528 else sv_setsv_flags(sv, temp, 0);
9530 if ((const GV *)sv == PL_last_in_gv)
9531 PL_last_in_gv = NULL;
9535 =for apidoc sv_unref_flags
9537 Unsets the RV status of the SV, and decrements the reference count of
9538 whatever was being referenced by the RV. This can almost be thought of
9539 as a reversal of C<newSVrv>. The C<cflags> argument can contain
9540 C<SV_IMMEDIATE_UNREF> to force the reference count to be decremented
9541 (otherwise the decrementing is conditional on the reference count being
9542 different from one or the reference being a readonly SV).
9549 Perl_sv_unref_flags(pTHX_ SV *const ref, const U32 flags)
9551 SV* const target = SvRV(ref);
9553 PERL_ARGS_ASSERT_SV_UNREF_FLAGS;
9555 if (SvWEAKREF(ref)) {
9556 sv_del_backref(target, ref);
9558 SvRV_set(ref, NULL);
9561 SvRV_set(ref, NULL);
9563 /* You can't have a || SvREADONLY(target) here, as $a = $$a, where $a was
9564 assigned to as BEGIN {$a = \"Foo"} will fail. */
9565 if (SvREFCNT(target) != 1 || (flags & SV_IMMEDIATE_UNREF))
9566 SvREFCNT_dec(target);
9567 else /* XXX Hack, but hard to make $a=$a->[1] work otherwise */
9568 sv_2mortal(target); /* Schedule for freeing later */
9572 =for apidoc sv_untaint
9574 Untaint an SV. Use C<SvTAINTED_off> instead.
9580 Perl_sv_untaint(pTHX_ SV *const sv)
9582 PERL_ARGS_ASSERT_SV_UNTAINT;
9584 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
9585 MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
9592 =for apidoc sv_tainted
9594 Test an SV for taintedness. Use C<SvTAINTED> instead.
9600 Perl_sv_tainted(pTHX_ SV *const sv)
9602 PERL_ARGS_ASSERT_SV_TAINTED;
9604 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
9605 const MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
9606 if (mg && (mg->mg_len & 1) )
9613 =for apidoc sv_setpviv
9615 Copies an integer into the given SV, also updating its string value.
9616 Does not handle 'set' magic. See C<sv_setpviv_mg>.
9622 Perl_sv_setpviv(pTHX_ SV *const sv, const IV iv)
9624 char buf[TYPE_CHARS(UV)];
9626 char * const ptr = uiv_2buf(buf, iv, 0, 0, &ebuf);
9628 PERL_ARGS_ASSERT_SV_SETPVIV;
9630 sv_setpvn(sv, ptr, ebuf - ptr);
9634 =for apidoc sv_setpviv_mg
9636 Like C<sv_setpviv>, but also handles 'set' magic.
9642 Perl_sv_setpviv_mg(pTHX_ SV *const sv, const IV iv)
9644 PERL_ARGS_ASSERT_SV_SETPVIV_MG;
9650 #if defined(PERL_IMPLICIT_CONTEXT)
9652 /* pTHX_ magic can't cope with varargs, so this is a no-context
9653 * version of the main function, (which may itself be aliased to us).
9654 * Don't access this version directly.
9658 Perl_sv_setpvf_nocontext(SV *const sv, const char *const pat, ...)
9663 PERL_ARGS_ASSERT_SV_SETPVF_NOCONTEXT;
9665 va_start(args, pat);
9666 sv_vsetpvf(sv, pat, &args);
9670 /* pTHX_ magic can't cope with varargs, so this is a no-context
9671 * version of the main function, (which may itself be aliased to us).
9672 * Don't access this version directly.
9676 Perl_sv_setpvf_mg_nocontext(SV *const sv, const char *const pat, ...)
9681 PERL_ARGS_ASSERT_SV_SETPVF_MG_NOCONTEXT;
9683 va_start(args, pat);
9684 sv_vsetpvf_mg(sv, pat, &args);
9690 =for apidoc sv_setpvf
9692 Works like C<sv_catpvf> but copies the text into the SV instead of
9693 appending it. Does not handle 'set' magic. See C<sv_setpvf_mg>.
9699 Perl_sv_setpvf(pTHX_ SV *const sv, const char *const pat, ...)
9703 PERL_ARGS_ASSERT_SV_SETPVF;
9705 va_start(args, pat);
9706 sv_vsetpvf(sv, pat, &args);
9711 =for apidoc sv_vsetpvf
9713 Works like C<sv_vcatpvf> but copies the text into the SV instead of
9714 appending it. Does not handle 'set' magic. See C<sv_vsetpvf_mg>.
9716 Usually used via its frontend C<sv_setpvf>.
9722 Perl_sv_vsetpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9724 PERL_ARGS_ASSERT_SV_VSETPVF;
9726 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9730 =for apidoc sv_setpvf_mg
9732 Like C<sv_setpvf>, but also handles 'set' magic.
9738 Perl_sv_setpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
9742 PERL_ARGS_ASSERT_SV_SETPVF_MG;
9744 va_start(args, pat);
9745 sv_vsetpvf_mg(sv, pat, &args);
9750 =for apidoc sv_vsetpvf_mg
9752 Like C<sv_vsetpvf>, but also handles 'set' magic.
9754 Usually used via its frontend C<sv_setpvf_mg>.
9760 Perl_sv_vsetpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9762 PERL_ARGS_ASSERT_SV_VSETPVF_MG;
9764 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9768 #if defined(PERL_IMPLICIT_CONTEXT)
9770 /* pTHX_ magic can't cope with varargs, so this is a no-context
9771 * version of the main function, (which may itself be aliased to us).
9772 * Don't access this version directly.
9776 Perl_sv_catpvf_nocontext(SV *const sv, const char *const pat, ...)
9781 PERL_ARGS_ASSERT_SV_CATPVF_NOCONTEXT;
9783 va_start(args, pat);
9784 sv_vcatpvf(sv, pat, &args);
9788 /* pTHX_ magic can't cope with varargs, so this is a no-context
9789 * version of the main function, (which may itself be aliased to us).
9790 * Don't access this version directly.
9794 Perl_sv_catpvf_mg_nocontext(SV *const sv, const char *const pat, ...)
9799 PERL_ARGS_ASSERT_SV_CATPVF_MG_NOCONTEXT;
9801 va_start(args, pat);
9802 sv_vcatpvf_mg(sv, pat, &args);
9808 =for apidoc sv_catpvf
9810 Processes its arguments like C<sprintf> and appends the formatted
9811 output to an SV. If the appended data contains "wide" characters
9812 (including, but not limited to, SVs with a UTF-8 PV formatted with %s,
9813 and characters >255 formatted with %c), the original SV might get
9814 upgraded to UTF-8. Handles 'get' magic, but not 'set' magic. See
9815 C<sv_catpvf_mg>. If the original SV was UTF-8, the pattern should be
9816 valid UTF-8; if the original SV was bytes, the pattern should be too.
9821 Perl_sv_catpvf(pTHX_ SV *const sv, const char *const pat, ...)
9825 PERL_ARGS_ASSERT_SV_CATPVF;
9827 va_start(args, pat);
9828 sv_vcatpvf(sv, pat, &args);
9833 =for apidoc sv_vcatpvf
9835 Processes its arguments like C<vsprintf> and appends the formatted output
9836 to an SV. Does not handle 'set' magic. See C<sv_vcatpvf_mg>.
9838 Usually used via its frontend C<sv_catpvf>.
9844 Perl_sv_vcatpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9846 PERL_ARGS_ASSERT_SV_VCATPVF;
9848 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9852 =for apidoc sv_catpvf_mg
9854 Like C<sv_catpvf>, but also handles 'set' magic.
9860 Perl_sv_catpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
9864 PERL_ARGS_ASSERT_SV_CATPVF_MG;
9866 va_start(args, pat);
9867 sv_vcatpvf_mg(sv, pat, &args);
9872 =for apidoc sv_vcatpvf_mg
9874 Like C<sv_vcatpvf>, but also handles 'set' magic.
9876 Usually used via its frontend C<sv_catpvf_mg>.
9882 Perl_sv_vcatpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9884 PERL_ARGS_ASSERT_SV_VCATPVF_MG;
9886 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9891 =for apidoc sv_vsetpvfn
9893 Works like C<sv_vcatpvfn> but copies the text into the SV instead of
9896 Usually used via one of its frontends C<sv_vsetpvf> and C<sv_vsetpvf_mg>.
9902 Perl_sv_vsetpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
9903 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
9905 PERL_ARGS_ASSERT_SV_VSETPVFN;
9908 sv_vcatpvfn(sv, pat, patlen, args, svargs, svmax, maybe_tainted);
9913 * Warn of missing argument to sprintf, and then return a defined value
9914 * to avoid inappropriate "use of uninit" warnings [perl #71000].
9916 #define WARN_MISSING WARN_UNINITIALIZED /* Not sure we want a new category */
9918 S_vcatpvfn_missing_argument(pTHX) {
9919 if (ckWARN(WARN_MISSING)) {
9920 Perl_warner(aTHX_ packWARN(WARN_MISSING), "Missing argument in %s",
9921 PL_op ? OP_DESC(PL_op) : "sv_vcatpvfn()");
9928 S_expect_number(pTHX_ char **const pattern)
9933 PERL_ARGS_ASSERT_EXPECT_NUMBER;
9935 switch (**pattern) {
9936 case '1': case '2': case '3':
9937 case '4': case '5': case '6':
9938 case '7': case '8': case '9':
9939 var = *(*pattern)++ - '0';
9940 while (isDIGIT(**pattern)) {
9941 const I32 tmp = var * 10 + (*(*pattern)++ - '0');
9943 Perl_croak(aTHX_ "Integer overflow in format string for %s", (PL_op ? OP_DESC(PL_op) : "sv_vcatpvfn"));
9951 S_F0convert(NV nv, char *const endbuf, STRLEN *const len)
9953 const int neg = nv < 0;
9956 PERL_ARGS_ASSERT_F0CONVERT;
9964 if (uv & 1 && uv == nv)
9965 uv--; /* Round to even */
9967 const unsigned dig = uv % 10;
9980 =for apidoc sv_vcatpvfn
9982 Processes its arguments like C<vsprintf> and appends the formatted output
9983 to an SV. Uses an array of SVs if the C style variable argument list is
9984 missing (NULL). When running with taint checks enabled, indicates via
9985 C<maybe_tainted> if results are untrustworthy (often due to the use of
9988 Usually used via one of its frontends C<sv_vcatpvf> and C<sv_vcatpvf_mg>.
9994 #define VECTORIZE_ARGS vecsv = va_arg(*args, SV*);\
9995 vecstr = (U8*)SvPV_const(vecsv,veclen);\
9996 vec_utf8 = DO_UTF8(vecsv);
9998 /* XXX maybe_tainted is never assigned to, so the doc above is lying. */
10001 Perl_sv_vcatpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
10002 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
10007 const char *patend;
10010 static const char nullstr[] = "(null)";
10012 bool has_utf8 = DO_UTF8(sv); /* has the result utf8? */
10013 const bool pat_utf8 = has_utf8; /* the pattern is in utf8? */
10015 /* Times 4: a decimal digit takes more than 3 binary digits.
10016 * NV_DIG: mantissa takes than many decimal digits.
10017 * Plus 32: Playing safe. */
10018 char ebuf[IV_DIG * 4 + NV_DIG + 32];
10019 /* large enough for "%#.#f" --chip */
10020 /* what about long double NVs? --jhi */
10022 PERL_ARGS_ASSERT_SV_VCATPVFN;
10023 PERL_UNUSED_ARG(maybe_tainted);
10025 /* no matter what, this is a string now */
10026 (void)SvPV_force(sv, origlen);
10028 /* special-case "", "%s", and "%-p" (SVf - see below) */
10031 if (patlen == 2 && pat[0] == '%' && pat[1] == 's') {
10033 const char * const s = va_arg(*args, char*);
10034 sv_catpv(sv, s ? s : nullstr);
10036 else if (svix < svmax) {
10037 sv_catsv(sv, *svargs);
10040 S_vcatpvfn_missing_argument(aTHX);
10043 if (args && patlen == 3 && pat[0] == '%' &&
10044 pat[1] == '-' && pat[2] == 'p') {
10045 argsv = MUTABLE_SV(va_arg(*args, void*));
10046 sv_catsv(sv, argsv);
10050 #ifndef USE_LONG_DOUBLE
10051 /* special-case "%.<number>[gf]" */
10052 if ( !args && patlen <= 5 && pat[0] == '%' && pat[1] == '.'
10053 && (pat[patlen-1] == 'g' || pat[patlen-1] == 'f') ) {
10054 unsigned digits = 0;
10058 while (*pp >= '0' && *pp <= '9')
10059 digits = 10 * digits + (*pp++ - '0');
10060 if (pp - pat == (int)patlen - 1 && svix < svmax) {
10061 const NV nv = SvNV(*svargs);
10063 /* Add check for digits != 0 because it seems that some
10064 gconverts are buggy in this case, and we don't yet have
10065 a Configure test for this. */
10066 if (digits && digits < sizeof(ebuf) - NV_DIG - 10) {
10067 /* 0, point, slack */
10068 Gconvert(nv, (int)digits, 0, ebuf);
10069 sv_catpv(sv, ebuf);
10070 if (*ebuf) /* May return an empty string for digits==0 */
10073 } else if (!digits) {
10076 if ((p = F0convert(nv, ebuf + sizeof ebuf, &l))) {
10077 sv_catpvn(sv, p, l);
10083 #endif /* !USE_LONG_DOUBLE */
10085 if (!args && svix < svmax && DO_UTF8(*svargs))
10088 patend = (char*)pat + patlen;
10089 for (p = (char*)pat; p < patend; p = q) {
10092 bool vectorize = FALSE;
10093 bool vectorarg = FALSE;
10094 bool vec_utf8 = FALSE;
10100 bool has_precis = FALSE;
10102 const I32 osvix = svix;
10103 bool is_utf8 = FALSE; /* is this item utf8? */
10104 #ifdef HAS_LDBL_SPRINTF_BUG
10105 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
10106 with sfio - Allen <allens@cpan.org> */
10107 bool fix_ldbl_sprintf_bug = FALSE;
10111 U8 utf8buf[UTF8_MAXBYTES+1];
10112 STRLEN esignlen = 0;
10114 const char *eptr = NULL;
10115 const char *fmtstart;
10118 const U8 *vecstr = NULL;
10125 /* we need a long double target in case HAS_LONG_DOUBLE but
10126 not USE_LONG_DOUBLE
10128 #if defined(HAS_LONG_DOUBLE) && LONG_DOUBLESIZE > DOUBLESIZE
10136 const char *dotstr = ".";
10137 STRLEN dotstrlen = 1;
10138 I32 efix = 0; /* explicit format parameter index */
10139 I32 ewix = 0; /* explicit width index */
10140 I32 epix = 0; /* explicit precision index */
10141 I32 evix = 0; /* explicit vector index */
10142 bool asterisk = FALSE;
10144 /* echo everything up to the next format specification */
10145 for (q = p; q < patend && *q != '%'; ++q) ;
10147 if (has_utf8 && !pat_utf8)
10148 sv_catpvn_utf8_upgrade(sv, p, q - p, nsv);
10150 sv_catpvn(sv, p, q - p);
10159 We allow format specification elements in this order:
10160 \d+\$ explicit format parameter index
10162 v|\*(\d+\$)?v vector with optional (optionally specified) arg
10163 0 flag (as above): repeated to allow "v02"
10164 \d+|\*(\d+\$)? width using optional (optionally specified) arg
10165 \.(\d*|\*(\d+\$)?) precision using optional (optionally specified) arg
10167 [%bcdefginopsuxDFOUX] format (mandatory)
10172 As of perl5.9.3, printf format checking is on by default.
10173 Internally, perl uses %p formats to provide an escape to
10174 some extended formatting. This block deals with those
10175 extensions: if it does not match, (char*)q is reset and
10176 the normal format processing code is used.
10178 Currently defined extensions are:
10179 %p include pointer address (standard)
10180 %-p (SVf) include an SV (previously %_)
10181 %-<num>p include an SV with precision <num>
10183 %3p include a HEK with precision of 256
10184 %<num>p (where num != 2 or 3) reserved for future
10187 Robin Barker 2005-07-14 (but modified since)
10189 %1p (VDf) removed. RMB 2007-10-19
10196 n = expect_number(&q);
10198 if (sv) { /* SVf */
10203 argsv = MUTABLE_SV(va_arg(*args, void*));
10204 eptr = SvPV_const(argsv, elen);
10205 if (DO_UTF8(argsv))
10209 else if (n==2 || n==3) { /* HEKf */
10210 HEK * const hek = va_arg(*args, HEK *);
10211 eptr = HEK_KEY(hek);
10212 elen = HEK_LEN(hek);
10213 if (HEK_UTF8(hek)) is_utf8 = TRUE;
10214 if (n==3) precis = 256, has_precis = TRUE;
10218 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL),
10219 "internal %%<num>p might conflict with future printf extensions");
10225 if ( (width = expect_number(&q)) ) {
10240 if (plus == '+' && *q == ' ') /* '+' over ' ' */
10269 if ( (ewix = expect_number(&q)) )
10278 if ((vectorarg = asterisk)) {
10291 width = expect_number(&q);
10294 if (vectorize && vectorarg) {
10295 /* vectorizing, but not with the default "." */
10297 vecsv = va_arg(*args, SV*);
10299 vecsv = (evix > 0 && evix <= svmax)
10300 ? svargs[evix-1] : S_vcatpvfn_missing_argument(aTHX);
10302 vecsv = svix < svmax
10303 ? svargs[svix++] : S_vcatpvfn_missing_argument(aTHX);
10305 dotstr = SvPV_const(vecsv, dotstrlen);
10306 /* Keep the DO_UTF8 test *after* the SvPV call, else things go
10307 bad with tied or overloaded values that return UTF8. */
10308 if (DO_UTF8(vecsv))
10310 else if (has_utf8) {
10311 vecsv = sv_mortalcopy(vecsv);
10312 sv_utf8_upgrade(vecsv);
10313 dotstr = SvPV_const(vecsv, dotstrlen);
10320 i = va_arg(*args, int);
10322 i = (ewix ? ewix <= svmax : svix < svmax) ?
10323 SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
10325 width = (i < 0) ? -i : i;
10335 if ( ((epix = expect_number(&q))) && (*q++ != '$') )
10337 /* XXX: todo, support specified precision parameter */
10341 i = va_arg(*args, int);
10343 i = (ewix ? ewix <= svmax : svix < svmax)
10344 ? SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
10346 has_precis = !(i < 0);
10350 while (isDIGIT(*q))
10351 precis = precis * 10 + (*q++ - '0');
10360 else if (efix ? (efix > 0 && efix <= svmax) : svix < svmax) {
10361 vecsv = svargs[efix ? efix-1 : svix++];
10362 vecstr = (U8*)SvPV_const(vecsv,veclen);
10363 vec_utf8 = DO_UTF8(vecsv);
10365 /* if this is a version object, we need to convert
10366 * back into v-string notation and then let the
10367 * vectorize happen normally
10369 if (sv_isobject(vecsv) && sv_derived_from(vecsv, "version")) {
10370 char *version = savesvpv(vecsv);
10371 if ( hv_exists(MUTABLE_HV(SvRV(vecsv)), "alpha", 5 ) ) {
10372 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
10373 "vector argument not supported with alpha versions");
10376 vecsv = sv_newmortal();
10377 scan_vstring(version, version + veclen, vecsv);
10378 vecstr = (U8*)SvPV_const(vecsv, veclen);
10379 vec_utf8 = DO_UTF8(vecsv);
10393 case 'I': /* Ix, I32x, and I64x */
10395 if (q[1] == '6' && q[2] == '4') {
10401 if (q[1] == '3' && q[2] == '2') {
10411 #if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
10423 #if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
10424 if (*q == 'l') { /* lld, llf */
10433 if (*++q == 'h') { /* hhd, hhu */
10462 if (!vectorize && !args) {
10464 const I32 i = efix-1;
10465 argsv = (i >= 0 && i < svmax)
10466 ? svargs[i] : S_vcatpvfn_missing_argument(aTHX);
10468 argsv = (svix >= 0 && svix < svmax)
10469 ? svargs[svix++] : S_vcatpvfn_missing_argument(aTHX);
10473 switch (c = *q++) {
10480 uv = (args) ? va_arg(*args, int) : SvIV(argsv);
10482 (!UNI_IS_INVARIANT(uv) && SvUTF8(sv)))
10484 eptr = (char*)utf8buf;
10485 elen = uvchr_to_utf8((U8*)eptr, uv) - utf8buf;
10499 eptr = va_arg(*args, char*);
10501 elen = strlen(eptr);
10503 eptr = (char *)nullstr;
10504 elen = sizeof nullstr - 1;
10508 eptr = SvPV_const(argsv, elen);
10509 if (DO_UTF8(argsv)) {
10510 STRLEN old_precis = precis;
10511 if (has_precis && precis < elen) {
10512 STRLEN ulen = sv_len_utf8(argsv);
10513 I32 p = precis > ulen ? ulen : precis;
10514 sv_pos_u2b(argsv, &p, 0); /* sticks at end */
10517 if (width) { /* fudge width (can't fudge elen) */
10518 if (has_precis && precis < elen)
10519 width += precis - old_precis;
10521 width += elen - sv_len_utf8(argsv);
10528 if (has_precis && precis < elen)
10535 if (alt || vectorize)
10537 uv = PTR2UV(args ? va_arg(*args, void*) : argsv);
10558 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
10567 esignbuf[esignlen++] = plus;
10571 case 'c': iv = (char)va_arg(*args, int); break;
10572 case 'h': iv = (short)va_arg(*args, int); break;
10573 case 'l': iv = va_arg(*args, long); break;
10574 case 'V': iv = va_arg(*args, IV); break;
10575 case 'z': iv = va_arg(*args, SSize_t); break;
10576 case 't': iv = va_arg(*args, ptrdiff_t); break;
10577 default: iv = va_arg(*args, int); break;
10579 case 'j': iv = va_arg(*args, intmax_t); break;
10583 iv = va_arg(*args, Quad_t); break;
10590 IV tiv = SvIV(argsv); /* work around GCC bug #13488 */
10592 case 'c': iv = (char)tiv; break;
10593 case 'h': iv = (short)tiv; break;
10594 case 'l': iv = (long)tiv; break;
10596 default: iv = tiv; break;
10599 iv = (Quad_t)tiv; break;
10605 if ( !vectorize ) /* we already set uv above */
10610 esignbuf[esignlen++] = plus;
10614 esignbuf[esignlen++] = '-';
10658 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
10669 case 'c': uv = (unsigned char)va_arg(*args, unsigned); break;
10670 case 'h': uv = (unsigned short)va_arg(*args, unsigned); break;
10671 case 'l': uv = va_arg(*args, unsigned long); break;
10672 case 'V': uv = va_arg(*args, UV); break;
10673 case 'z': uv = va_arg(*args, Size_t); break;
10674 case 't': uv = va_arg(*args, ptrdiff_t); break; /* will sign extend, but there is no uptrdiff_t, so oh well */
10676 case 'j': uv = va_arg(*args, uintmax_t); break;
10678 default: uv = va_arg(*args, unsigned); break;
10681 uv = va_arg(*args, Uquad_t); break;
10688 UV tuv = SvUV(argsv); /* work around GCC bug #13488 */
10690 case 'c': uv = (unsigned char)tuv; break;
10691 case 'h': uv = (unsigned short)tuv; break;
10692 case 'l': uv = (unsigned long)tuv; break;
10694 default: uv = tuv; break;
10697 uv = (Uquad_t)tuv; break;
10706 char *ptr = ebuf + sizeof ebuf;
10707 bool tempalt = uv ? alt : FALSE; /* Vectors can't change alt */
10713 p = (char *)((c == 'X') ? PL_hexdigit + 16 : PL_hexdigit);
10717 } while (uv >>= 4);
10719 esignbuf[esignlen++] = '0';
10720 esignbuf[esignlen++] = c; /* 'x' or 'X' */
10726 *--ptr = '0' + dig;
10727 } while (uv >>= 3);
10728 if (alt && *ptr != '0')
10734 *--ptr = '0' + dig;
10735 } while (uv >>= 1);
10737 esignbuf[esignlen++] = '0';
10738 esignbuf[esignlen++] = c;
10741 default: /* it had better be ten or less */
10744 *--ptr = '0' + dig;
10745 } while (uv /= base);
10748 elen = (ebuf + sizeof ebuf) - ptr;
10752 zeros = precis - elen;
10753 else if (precis == 0 && elen == 1 && *eptr == '0'
10754 && !(base == 8 && alt)) /* "%#.0o" prints "0" */
10757 /* a precision nullifies the 0 flag. */
10764 /* FLOATING POINT */
10767 c = 'f'; /* maybe %F isn't supported here */
10769 case 'e': case 'E':
10771 case 'g': case 'G':
10775 /* This is evil, but floating point is even more evil */
10777 /* for SV-style calling, we can only get NV
10778 for C-style calling, we assume %f is double;
10779 for simplicity we allow any of %Lf, %llf, %qf for long double
10783 #if defined(USE_LONG_DOUBLE)
10787 /* [perl #20339] - we should accept and ignore %lf rather than die */
10791 #if defined(USE_LONG_DOUBLE)
10792 intsize = args ? 0 : 'q';
10796 #if defined(HAS_LONG_DOUBLE)
10809 /* now we need (long double) if intsize == 'q', else (double) */
10811 #if LONG_DOUBLESIZE > DOUBLESIZE
10813 va_arg(*args, long double) :
10814 va_arg(*args, double)
10816 va_arg(*args, double)
10821 /* nv * 0 will be NaN for NaN, +Inf and -Inf, and 0 for anything
10822 else. frexp() has some unspecified behaviour for those three */
10823 if (c != 'e' && c != 'E' && (nv * 0) == 0) {
10825 /* FIXME: if HAS_LONG_DOUBLE but not USE_LONG_DOUBLE this
10826 will cast our (long double) to (double) */
10827 (void)Perl_frexp(nv, &i);
10828 if (i == PERL_INT_MIN)
10829 Perl_die(aTHX_ "panic: frexp");
10831 need = BIT_DIGITS(i);
10833 need += has_precis ? precis : 6; /* known default */
10838 #ifdef HAS_LDBL_SPRINTF_BUG
10839 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
10840 with sfio - Allen <allens@cpan.org> */
10843 # define MY_DBL_MAX DBL_MAX
10844 # else /* XXX guessing! HUGE_VAL may be defined as infinity, so not using */
10845 # if DOUBLESIZE >= 8
10846 # define MY_DBL_MAX 1.7976931348623157E+308L
10848 # define MY_DBL_MAX 3.40282347E+38L
10852 # ifdef HAS_LDBL_SPRINTF_BUG_LESS1 /* only between -1L & 1L - Allen */
10853 # define MY_DBL_MAX_BUG 1L
10855 # define MY_DBL_MAX_BUG MY_DBL_MAX
10859 # define MY_DBL_MIN DBL_MIN
10860 # else /* XXX guessing! -Allen */
10861 # if DOUBLESIZE >= 8
10862 # define MY_DBL_MIN 2.2250738585072014E-308L
10864 # define MY_DBL_MIN 1.17549435E-38L
10868 if ((intsize == 'q') && (c == 'f') &&
10869 ((nv < MY_DBL_MAX_BUG) && (nv > -MY_DBL_MAX_BUG)) &&
10870 (need < DBL_DIG)) {
10871 /* it's going to be short enough that
10872 * long double precision is not needed */
10874 if ((nv <= 0L) && (nv >= -0L))
10875 fix_ldbl_sprintf_bug = TRUE; /* 0 is 0 - easiest */
10877 /* would use Perl_fp_class as a double-check but not
10878 * functional on IRIX - see perl.h comments */
10880 if ((nv >= MY_DBL_MIN) || (nv <= -MY_DBL_MIN)) {
10881 /* It's within the range that a double can represent */
10882 #if defined(DBL_MAX) && !defined(DBL_MIN)
10883 if ((nv >= ((long double)1/DBL_MAX)) ||
10884 (nv <= (-(long double)1/DBL_MAX)))
10886 fix_ldbl_sprintf_bug = TRUE;
10889 if (fix_ldbl_sprintf_bug == TRUE) {
10899 # undef MY_DBL_MAX_BUG
10902 #endif /* HAS_LDBL_SPRINTF_BUG */
10904 need += 20; /* fudge factor */
10905 if (PL_efloatsize < need) {
10906 Safefree(PL_efloatbuf);
10907 PL_efloatsize = need + 20; /* more fudge */
10908 Newx(PL_efloatbuf, PL_efloatsize, char);
10909 PL_efloatbuf[0] = '\0';
10912 if ( !(width || left || plus || alt) && fill != '0'
10913 && has_precis && intsize != 'q' ) { /* Shortcuts */
10914 /* See earlier comment about buggy Gconvert when digits,
10916 if ( c == 'g' && precis) {
10917 Gconvert((NV)nv, (int)precis, 0, PL_efloatbuf);
10918 /* May return an empty string for digits==0 */
10919 if (*PL_efloatbuf) {
10920 elen = strlen(PL_efloatbuf);
10921 goto float_converted;
10923 } else if ( c == 'f' && !precis) {
10924 if ((eptr = F0convert(nv, ebuf + sizeof ebuf, &elen)))
10929 char *ptr = ebuf + sizeof ebuf;
10932 /* FIXME: what to do if HAS_LONG_DOUBLE but not PERL_PRIfldbl? */
10933 #if defined(HAS_LONG_DOUBLE) && defined(PERL_PRIfldbl)
10934 if (intsize == 'q') {
10935 /* Copy the one or more characters in a long double
10936 * format before the 'base' ([efgEFG]) character to
10937 * the format string. */
10938 static char const prifldbl[] = PERL_PRIfldbl;
10939 char const *p = prifldbl + sizeof(prifldbl) - 3;
10940 while (p >= prifldbl) { *--ptr = *p--; }
10945 do { *--ptr = '0' + (base % 10); } while (base /= 10);
10950 do { *--ptr = '0' + (base % 10); } while (base /= 10);
10962 /* No taint. Otherwise we are in the strange situation
10963 * where printf() taints but print($float) doesn't.
10965 #if defined(HAS_LONG_DOUBLE)
10966 elen = ((intsize == 'q')
10967 ? my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, nv)
10968 : my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, (double)nv));
10970 elen = my_sprintf(PL_efloatbuf, ptr, nv);
10974 eptr = PL_efloatbuf;
10982 i = SvCUR(sv) - origlen;
10985 case 'c': *(va_arg(*args, char*)) = i; break;
10986 case 'h': *(va_arg(*args, short*)) = i; break;
10987 default: *(va_arg(*args, int*)) = i; break;
10988 case 'l': *(va_arg(*args, long*)) = i; break;
10989 case 'V': *(va_arg(*args, IV*)) = i; break;
10990 case 'z': *(va_arg(*args, SSize_t*)) = i; break;
10991 case 't': *(va_arg(*args, ptrdiff_t*)) = i; break;
10993 case 'j': *(va_arg(*args, intmax_t*)) = i; break;
10997 *(va_arg(*args, Quad_t*)) = i; break;
11004 sv_setuv_mg(argsv, has_utf8 ? (UV)sv_len_utf8(sv) : (UV)i);
11005 continue; /* not "break" */
11012 && (PL_op->op_type == OP_PRTF || PL_op->op_type == OP_SPRINTF)
11013 && ckWARN(WARN_PRINTF))
11015 SV * const msg = sv_newmortal();
11016 Perl_sv_setpvf(aTHX_ msg, "Invalid conversion in %sprintf: ",
11017 (PL_op->op_type == OP_PRTF) ? "" : "s");
11018 if (fmtstart < patend) {
11019 const char * const fmtend = q < patend ? q : patend;
11021 sv_catpvs(msg, "\"%");
11022 for (f = fmtstart; f < fmtend; f++) {
11024 sv_catpvn(msg, f, 1);
11026 Perl_sv_catpvf(aTHX_ msg,
11027 "\\%03"UVof, (UV)*f & 0xFF);
11030 sv_catpvs(msg, "\"");
11032 sv_catpvs(msg, "end of string");
11034 Perl_warner(aTHX_ packWARN(WARN_PRINTF), "%"SVf, SVfARG(msg)); /* yes, this is reentrant */
11037 /* output mangled stuff ... */
11043 /* ... right here, because formatting flags should not apply */
11044 SvGROW(sv, SvCUR(sv) + elen + 1);
11046 Copy(eptr, p, elen, char);
11049 SvCUR_set(sv, p - SvPVX_const(sv));
11051 continue; /* not "break" */
11054 if (is_utf8 != has_utf8) {
11057 sv_utf8_upgrade(sv);
11060 const STRLEN old_elen = elen;
11061 SV * const nsv = newSVpvn_flags(eptr, elen, SVs_TEMP);
11062 sv_utf8_upgrade(nsv);
11063 eptr = SvPVX_const(nsv);
11066 if (width) { /* fudge width (can't fudge elen) */
11067 width += elen - old_elen;
11073 have = esignlen + zeros + elen;
11075 Perl_croak_nocontext("%s", PL_memory_wrap);
11077 need = (have > width ? have : width);
11080 if (need >= (((STRLEN)~0) - SvCUR(sv) - dotstrlen - 1))
11081 Perl_croak_nocontext("%s", PL_memory_wrap);
11082 SvGROW(sv, SvCUR(sv) + need + dotstrlen + 1);
11084 if (esignlen && fill == '0') {
11086 for (i = 0; i < (int)esignlen; i++)
11087 *p++ = esignbuf[i];
11089 if (gap && !left) {
11090 memset(p, fill, gap);
11093 if (esignlen && fill != '0') {
11095 for (i = 0; i < (int)esignlen; i++)
11096 *p++ = esignbuf[i];
11100 for (i = zeros; i; i--)
11104 Copy(eptr, p, elen, char);
11108 memset(p, ' ', gap);
11113 Copy(dotstr, p, dotstrlen, char);
11117 vectorize = FALSE; /* done iterating over vecstr */
11124 SvCUR_set(sv, p - SvPVX_const(sv));
11133 /* =========================================================================
11135 =head1 Cloning an interpreter
11137 All the macros and functions in this section are for the private use of
11138 the main function, perl_clone().
11140 The foo_dup() functions make an exact copy of an existing foo thingy.
11141 During the course of a cloning, a hash table is used to map old addresses
11142 to new addresses. The table is created and manipulated with the
11143 ptr_table_* functions.
11147 * =========================================================================*/
11150 #if defined(USE_ITHREADS)
11152 /* XXX Remove this so it doesn't have to go thru the macro and return for nothing */
11153 #ifndef GpREFCNT_inc
11154 # define GpREFCNT_inc(gp) ((gp) ? (++(gp)->gp_refcnt, (gp)) : (GP*)NULL)
11158 /* Certain cases in Perl_ss_dup have been merged, by relying on the fact
11159 that currently av_dup, gv_dup and hv_dup are the same as sv_dup.
11160 If this changes, please unmerge ss_dup.
11161 Likewise, sv_dup_inc_multiple() relies on this fact. */
11162 #define sv_dup_inc_NN(s,t) SvREFCNT_inc_NN(sv_dup_inc(s,t))
11163 #define av_dup(s,t) MUTABLE_AV(sv_dup((const SV *)s,t))
11164 #define av_dup_inc(s,t) MUTABLE_AV(sv_dup_inc((const SV *)s,t))
11165 #define hv_dup(s,t) MUTABLE_HV(sv_dup((const SV *)s,t))
11166 #define hv_dup_inc(s,t) MUTABLE_HV(sv_dup_inc((const SV *)s,t))
11167 #define cv_dup(s,t) MUTABLE_CV(sv_dup((const SV *)s,t))
11168 #define cv_dup_inc(s,t) MUTABLE_CV(sv_dup_inc((const SV *)s,t))
11169 #define io_dup(s,t) MUTABLE_IO(sv_dup((const SV *)s,t))
11170 #define io_dup_inc(s,t) MUTABLE_IO(sv_dup_inc((const SV *)s,t))
11171 #define gv_dup(s,t) MUTABLE_GV(sv_dup((const SV *)s,t))
11172 #define gv_dup_inc(s,t) MUTABLE_GV(sv_dup_inc((const SV *)s,t))
11173 #define SAVEPV(p) ((p) ? savepv(p) : NULL)
11174 #define SAVEPVN(p,n) ((p) ? savepvn(p,n) : NULL)
11176 /* clone a parser */
11179 Perl_parser_dup(pTHX_ const yy_parser *const proto, CLONE_PARAMS *const param)
11183 PERL_ARGS_ASSERT_PARSER_DUP;
11188 /* look for it in the table first */
11189 parser = (yy_parser *)ptr_table_fetch(PL_ptr_table, proto);
11193 /* create anew and remember what it is */
11194 Newxz(parser, 1, yy_parser);
11195 ptr_table_store(PL_ptr_table, proto, parser);
11197 /* XXX these not yet duped */
11198 parser->old_parser = NULL;
11199 parser->stack = NULL;
11201 parser->stack_size = 0;
11202 /* XXX parser->stack->state = 0; */
11204 /* XXX eventually, just Copy() most of the parser struct ? */
11206 parser->lex_brackets = proto->lex_brackets;
11207 parser->lex_casemods = proto->lex_casemods;
11208 parser->lex_brackstack = savepvn(proto->lex_brackstack,
11209 (proto->lex_brackets < 120 ? 120 : proto->lex_brackets));
11210 parser->lex_casestack = savepvn(proto->lex_casestack,
11211 (proto->lex_casemods < 12 ? 12 : proto->lex_casemods));
11212 parser->lex_defer = proto->lex_defer;
11213 parser->lex_dojoin = proto->lex_dojoin;
11214 parser->lex_expect = proto->lex_expect;
11215 parser->lex_formbrack = proto->lex_formbrack;
11216 parser->lex_inpat = proto->lex_inpat;
11217 parser->lex_inwhat = proto->lex_inwhat;
11218 parser->lex_op = proto->lex_op;
11219 parser->lex_repl = sv_dup_inc(proto->lex_repl, param);
11220 parser->lex_starts = proto->lex_starts;
11221 parser->lex_stuff = sv_dup_inc(proto->lex_stuff, param);
11222 parser->multi_close = proto->multi_close;
11223 parser->multi_open = proto->multi_open;
11224 parser->multi_start = proto->multi_start;
11225 parser->multi_end = proto->multi_end;
11226 parser->pending_ident = proto->pending_ident;
11227 parser->preambled = proto->preambled;
11228 parser->sublex_info = proto->sublex_info; /* XXX not quite right */
11229 parser->linestr = sv_dup_inc(proto->linestr, param);
11230 parser->expect = proto->expect;
11231 parser->copline = proto->copline;
11232 parser->last_lop_op = proto->last_lop_op;
11233 parser->lex_state = proto->lex_state;
11234 parser->rsfp = fp_dup(proto->rsfp, '<', param);
11235 /* rsfp_filters entries have fake IoDIRP() */
11236 parser->rsfp_filters= av_dup_inc(proto->rsfp_filters, param);
11237 parser->in_my = proto->in_my;
11238 parser->in_my_stash = hv_dup(proto->in_my_stash, param);
11239 parser->error_count = proto->error_count;
11242 parser->linestr = sv_dup_inc(proto->linestr, param);
11245 char * const ols = SvPVX(proto->linestr);
11246 char * const ls = SvPVX(parser->linestr);
11248 parser->bufptr = ls + (proto->bufptr >= ols ?
11249 proto->bufptr - ols : 0);
11250 parser->oldbufptr = ls + (proto->oldbufptr >= ols ?
11251 proto->oldbufptr - ols : 0);
11252 parser->oldoldbufptr= ls + (proto->oldoldbufptr >= ols ?
11253 proto->oldoldbufptr - ols : 0);
11254 parser->linestart = ls + (proto->linestart >= ols ?
11255 proto->linestart - ols : 0);
11256 parser->last_uni = ls + (proto->last_uni >= ols ?
11257 proto->last_uni - ols : 0);
11258 parser->last_lop = ls + (proto->last_lop >= ols ?
11259 proto->last_lop - ols : 0);
11261 parser->bufend = ls + SvCUR(parser->linestr);
11264 Copy(proto->tokenbuf, parser->tokenbuf, 256, char);
11268 parser->endwhite = proto->endwhite;
11269 parser->faketokens = proto->faketokens;
11270 parser->lasttoke = proto->lasttoke;
11271 parser->nextwhite = proto->nextwhite;
11272 parser->realtokenstart = proto->realtokenstart;
11273 parser->skipwhite = proto->skipwhite;
11274 parser->thisclose = proto->thisclose;
11275 parser->thismad = proto->thismad;
11276 parser->thisopen = proto->thisopen;
11277 parser->thisstuff = proto->thisstuff;
11278 parser->thistoken = proto->thistoken;
11279 parser->thiswhite = proto->thiswhite;
11281 Copy(proto->nexttoke, parser->nexttoke, 5, NEXTTOKE);
11282 parser->curforce = proto->curforce;
11284 Copy(proto->nextval, parser->nextval, 5, YYSTYPE);
11285 Copy(proto->nexttype, parser->nexttype, 5, I32);
11286 parser->nexttoke = proto->nexttoke;
11289 /* XXX should clone saved_curcop here, but we aren't passed
11290 * proto_perl; so do it in perl_clone_using instead */
11296 /* duplicate a file handle */
11299 Perl_fp_dup(pTHX_ PerlIO *const fp, const char type, CLONE_PARAMS *const param)
11303 PERL_ARGS_ASSERT_FP_DUP;
11304 PERL_UNUSED_ARG(type);
11307 return (PerlIO*)NULL;
11309 /* look for it in the table first */
11310 ret = (PerlIO*)ptr_table_fetch(PL_ptr_table, fp);
11314 /* create anew and remember what it is */
11315 ret = PerlIO_fdupopen(aTHX_ fp, param, PERLIO_DUP_CLONE);
11316 ptr_table_store(PL_ptr_table, fp, ret);
11320 /* duplicate a directory handle */
11323 Perl_dirp_dup(pTHX_ DIR *const dp, CLONE_PARAMS *const param)
11329 register const Direntry_t *dirent;
11330 char smallbuf[256];
11336 PERL_UNUSED_CONTEXT;
11337 PERL_ARGS_ASSERT_DIRP_DUP;
11342 /* look for it in the table first */
11343 ret = (DIR*)ptr_table_fetch(PL_ptr_table, dp);
11349 PERL_UNUSED_ARG(param);
11353 /* open the current directory (so we can switch back) */
11354 if (!(pwd = PerlDir_open("."))) return (DIR *)NULL;
11356 /* chdir to our dir handle and open the present working directory */
11357 if (fchdir(my_dirfd(dp)) < 0 || !(ret = PerlDir_open("."))) {
11358 PerlDir_close(pwd);
11359 return (DIR *)NULL;
11361 /* Now we should have two dir handles pointing to the same dir. */
11363 /* Be nice to the calling code and chdir back to where we were. */
11364 fchdir(my_dirfd(pwd)); /* If this fails, then what? */
11366 /* We have no need of the pwd handle any more. */
11367 PerlDir_close(pwd);
11370 # define d_namlen(d) (d)->d_namlen
11372 # define d_namlen(d) strlen((d)->d_name)
11374 /* Iterate once through dp, to get the file name at the current posi-
11375 tion. Then step back. */
11376 pos = PerlDir_tell(dp);
11377 if ((dirent = PerlDir_read(dp))) {
11378 len = d_namlen(dirent);
11379 if (len <= sizeof smallbuf) name = smallbuf;
11380 else Newx(name, len, char);
11381 Move(dirent->d_name, name, len, char);
11383 PerlDir_seek(dp, pos);
11385 /* Iterate through the new dir handle, till we find a file with the
11387 if (!dirent) /* just before the end */
11389 pos = PerlDir_tell(ret);
11390 if (PerlDir_read(ret)) continue; /* not there yet */
11391 PerlDir_seek(ret, pos); /* step back */
11395 const long pos0 = PerlDir_tell(ret);
11397 pos = PerlDir_tell(ret);
11398 if ((dirent = PerlDir_read(ret))) {
11399 if (len == d_namlen(dirent)
11400 && memEQ(name, dirent->d_name, len)) {
11402 PerlDir_seek(ret, pos); /* step back */
11405 /* else we are not there yet; keep iterating */
11407 else { /* This is not meant to happen. The best we can do is
11408 reset the iterator to the beginning. */
11409 PerlDir_seek(ret, pos0);
11416 if (name && name != smallbuf)
11421 ret = win32_dirp_dup(dp, param);
11424 /* pop it in the pointer table */
11426 ptr_table_store(PL_ptr_table, dp, ret);
11431 /* duplicate a typeglob */
11434 Perl_gp_dup(pTHX_ GP *const gp, CLONE_PARAMS *const param)
11438 PERL_ARGS_ASSERT_GP_DUP;
11442 /* look for it in the table first */
11443 ret = (GP*)ptr_table_fetch(PL_ptr_table, gp);
11447 /* create anew and remember what it is */
11449 ptr_table_store(PL_ptr_table, gp, ret);
11452 /* ret->gp_refcnt must be 0 before any other dups are called. We're relying
11453 on Newxz() to do this for us. */
11454 ret->gp_sv = sv_dup_inc(gp->gp_sv, param);
11455 ret->gp_io = io_dup_inc(gp->gp_io, param);
11456 ret->gp_form = cv_dup_inc(gp->gp_form, param);
11457 ret->gp_av = av_dup_inc(gp->gp_av, param);
11458 ret->gp_hv = hv_dup_inc(gp->gp_hv, param);
11459 ret->gp_egv = gv_dup(gp->gp_egv, param);/* GvEGV is not refcounted */
11460 ret->gp_cv = cv_dup_inc(gp->gp_cv, param);
11461 ret->gp_cvgen = gp->gp_cvgen;
11462 ret->gp_line = gp->gp_line;
11463 ret->gp_file_hek = hek_dup(gp->gp_file_hek, param);
11467 /* duplicate a chain of magic */
11470 Perl_mg_dup(pTHX_ MAGIC *mg, CLONE_PARAMS *const param)
11472 MAGIC *mgret = NULL;
11473 MAGIC **mgprev_p = &mgret;
11475 PERL_ARGS_ASSERT_MG_DUP;
11477 for (; mg; mg = mg->mg_moremagic) {
11480 if ((param->flags & CLONEf_JOIN_IN)
11481 && mg->mg_type == PERL_MAGIC_backref)
11482 /* when joining, we let the individual SVs add themselves to
11483 * backref as needed. */
11486 Newx(nmg, 1, MAGIC);
11488 mgprev_p = &(nmg->mg_moremagic);
11490 /* There was a comment "XXX copy dynamic vtable?" but as we don't have
11491 dynamic vtables, I'm not sure why Sarathy wrote it. The comment dates
11492 from the original commit adding Perl_mg_dup() - revision 4538.
11493 Similarly there is the annotation "XXX random ptr?" next to the
11494 assignment to nmg->mg_ptr. */
11497 /* FIXME for plugins
11498 if (nmg->mg_type == PERL_MAGIC_qr) {
11499 nmg->mg_obj = MUTABLE_SV(CALLREGDUPE((REGEXP*)nmg->mg_obj, param));
11503 nmg->mg_obj = (nmg->mg_flags & MGf_REFCOUNTED)
11504 ? nmg->mg_type == PERL_MAGIC_backref
11505 /* The backref AV has its reference
11506 * count deliberately bumped by 1 */
11507 ? SvREFCNT_inc(av_dup_inc((const AV *)
11508 nmg->mg_obj, param))
11509 : sv_dup_inc(nmg->mg_obj, param)
11510 : sv_dup(nmg->mg_obj, param);
11512 if (nmg->mg_ptr && nmg->mg_type != PERL_MAGIC_regex_global) {
11513 if (nmg->mg_len > 0) {
11514 nmg->mg_ptr = SAVEPVN(nmg->mg_ptr, nmg->mg_len);
11515 if (nmg->mg_type == PERL_MAGIC_overload_table &&
11516 AMT_AMAGIC((AMT*)nmg->mg_ptr))
11518 AMT * const namtp = (AMT*)nmg->mg_ptr;
11519 sv_dup_inc_multiple((SV**)(namtp->table),
11520 (SV**)(namtp->table), NofAMmeth, param);
11523 else if (nmg->mg_len == HEf_SVKEY)
11524 nmg->mg_ptr = (char*)sv_dup_inc((const SV *)nmg->mg_ptr, param);
11526 if ((nmg->mg_flags & MGf_DUP) && nmg->mg_virtual && nmg->mg_virtual->svt_dup) {
11527 nmg->mg_virtual->svt_dup(aTHX_ nmg, param);
11533 #endif /* USE_ITHREADS */
11535 struct ptr_tbl_arena {
11536 struct ptr_tbl_arena *next;
11537 struct ptr_tbl_ent array[1023/3]; /* as ptr_tbl_ent has 3 pointers. */
11540 /* create a new pointer-mapping table */
11543 Perl_ptr_table_new(pTHX)
11546 PERL_UNUSED_CONTEXT;
11548 Newx(tbl, 1, PTR_TBL_t);
11549 tbl->tbl_max = 511;
11550 tbl->tbl_items = 0;
11551 tbl->tbl_arena = NULL;
11552 tbl->tbl_arena_next = NULL;
11553 tbl->tbl_arena_end = NULL;
11554 Newxz(tbl->tbl_ary, tbl->tbl_max + 1, PTR_TBL_ENT_t*);
11558 #define PTR_TABLE_HASH(ptr) \
11559 ((PTR2UV(ptr) >> 3) ^ (PTR2UV(ptr) >> (3 + 7)) ^ (PTR2UV(ptr) >> (3 + 17)))
11561 /* map an existing pointer using a table */
11563 STATIC PTR_TBL_ENT_t *
11564 S_ptr_table_find(PTR_TBL_t *const tbl, const void *const sv)
11566 PTR_TBL_ENT_t *tblent;
11567 const UV hash = PTR_TABLE_HASH(sv);
11569 PERL_ARGS_ASSERT_PTR_TABLE_FIND;
11571 tblent = tbl->tbl_ary[hash & tbl->tbl_max];
11572 for (; tblent; tblent = tblent->next) {
11573 if (tblent->oldval == sv)
11580 Perl_ptr_table_fetch(pTHX_ PTR_TBL_t *const tbl, const void *const sv)
11582 PTR_TBL_ENT_t const *const tblent = ptr_table_find(tbl, sv);
11584 PERL_ARGS_ASSERT_PTR_TABLE_FETCH;
11585 PERL_UNUSED_CONTEXT;
11587 return tblent ? tblent->newval : NULL;
11590 /* add a new entry to a pointer-mapping table */
11593 Perl_ptr_table_store(pTHX_ PTR_TBL_t *const tbl, const void *const oldsv, void *const newsv)
11595 PTR_TBL_ENT_t *tblent = ptr_table_find(tbl, oldsv);
11597 PERL_ARGS_ASSERT_PTR_TABLE_STORE;
11598 PERL_UNUSED_CONTEXT;
11601 tblent->newval = newsv;
11603 const UV entry = PTR_TABLE_HASH(oldsv) & tbl->tbl_max;
11605 if (tbl->tbl_arena_next == tbl->tbl_arena_end) {
11606 struct ptr_tbl_arena *new_arena;
11608 Newx(new_arena, 1, struct ptr_tbl_arena);
11609 new_arena->next = tbl->tbl_arena;
11610 tbl->tbl_arena = new_arena;
11611 tbl->tbl_arena_next = new_arena->array;
11612 tbl->tbl_arena_end = new_arena->array
11613 + sizeof(new_arena->array) / sizeof(new_arena->array[0]);
11616 tblent = tbl->tbl_arena_next++;
11618 tblent->oldval = oldsv;
11619 tblent->newval = newsv;
11620 tblent->next = tbl->tbl_ary[entry];
11621 tbl->tbl_ary[entry] = tblent;
11623 if (tblent->next && tbl->tbl_items > tbl->tbl_max)
11624 ptr_table_split(tbl);
11628 /* double the hash bucket size of an existing ptr table */
11631 Perl_ptr_table_split(pTHX_ PTR_TBL_t *const tbl)
11633 PTR_TBL_ENT_t **ary = tbl->tbl_ary;
11634 const UV oldsize = tbl->tbl_max + 1;
11635 UV newsize = oldsize * 2;
11638 PERL_ARGS_ASSERT_PTR_TABLE_SPLIT;
11639 PERL_UNUSED_CONTEXT;
11641 Renew(ary, newsize, PTR_TBL_ENT_t*);
11642 Zero(&ary[oldsize], newsize-oldsize, PTR_TBL_ENT_t*);
11643 tbl->tbl_max = --newsize;
11644 tbl->tbl_ary = ary;
11645 for (i=0; i < oldsize; i++, ary++) {
11646 PTR_TBL_ENT_t **entp = ary;
11647 PTR_TBL_ENT_t *ent = *ary;
11648 PTR_TBL_ENT_t **curentp;
11651 curentp = ary + oldsize;
11653 if ((newsize & PTR_TABLE_HASH(ent->oldval)) != i) {
11655 ent->next = *curentp;
11665 /* remove all the entries from a ptr table */
11666 /* Deprecated - will be removed post 5.14 */
11669 Perl_ptr_table_clear(pTHX_ PTR_TBL_t *const tbl)
11671 if (tbl && tbl->tbl_items) {
11672 struct ptr_tbl_arena *arena = tbl->tbl_arena;
11674 Zero(tbl->tbl_ary, tbl->tbl_max + 1, struct ptr_tbl_ent **);
11677 struct ptr_tbl_arena *next = arena->next;
11683 tbl->tbl_items = 0;
11684 tbl->tbl_arena = NULL;
11685 tbl->tbl_arena_next = NULL;
11686 tbl->tbl_arena_end = NULL;
11690 /* clear and free a ptr table */
11693 Perl_ptr_table_free(pTHX_ PTR_TBL_t *const tbl)
11695 struct ptr_tbl_arena *arena;
11701 arena = tbl->tbl_arena;
11704 struct ptr_tbl_arena *next = arena->next;
11710 Safefree(tbl->tbl_ary);
11714 #if defined(USE_ITHREADS)
11717 Perl_rvpv_dup(pTHX_ SV *const dstr, const SV *const sstr, CLONE_PARAMS *const param)
11719 PERL_ARGS_ASSERT_RVPV_DUP;
11722 if (SvWEAKREF(sstr)) {
11723 SvRV_set(dstr, sv_dup(SvRV_const(sstr), param));
11724 if (param->flags & CLONEf_JOIN_IN) {
11725 /* if joining, we add any back references individually rather
11726 * than copying the whole backref array */
11727 Perl_sv_add_backref(aTHX_ SvRV(dstr), dstr);
11731 SvRV_set(dstr, sv_dup_inc(SvRV_const(sstr), param));
11733 else if (SvPVX_const(sstr)) {
11734 /* Has something there */
11736 /* Normal PV - clone whole allocated space */
11737 SvPV_set(dstr, SAVEPVN(SvPVX_const(sstr), SvLEN(sstr)-1));
11738 if (SvREADONLY(sstr) && SvFAKE(sstr)) {
11739 /* Not that normal - actually sstr is copy on write.
11740 But we are a true, independent SV, so: */
11741 SvREADONLY_off(dstr);
11746 /* Special case - not normally malloced for some reason */
11747 if (isGV_with_GP(sstr)) {
11748 /* Don't need to do anything here. */
11750 else if ((SvREADONLY(sstr) && SvFAKE(sstr))) {
11751 /* A "shared" PV - clone it as "shared" PV */
11753 HEK_KEY(hek_dup(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)),
11757 /* Some other special case - random pointer */
11758 SvPV_set(dstr, (char *) SvPVX_const(sstr));
11763 /* Copy the NULL */
11764 SvPV_set(dstr, NULL);
11768 /* duplicate a list of SVs. source and dest may point to the same memory. */
11770 S_sv_dup_inc_multiple(pTHX_ SV *const *source, SV **dest,
11771 SSize_t items, CLONE_PARAMS *const param)
11773 PERL_ARGS_ASSERT_SV_DUP_INC_MULTIPLE;
11775 while (items-- > 0) {
11776 *dest++ = sv_dup_inc(*source++, param);
11782 /* duplicate an SV of any type (including AV, HV etc) */
11785 S_sv_dup_common(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
11790 PERL_ARGS_ASSERT_SV_DUP_COMMON;
11792 if (SvTYPE(sstr) == (svtype)SVTYPEMASK) {
11793 #ifdef DEBUG_LEAKING_SCALARS_ABORT
11798 /* look for it in the table first */
11799 dstr = MUTABLE_SV(ptr_table_fetch(PL_ptr_table, sstr));
11803 if(param->flags & CLONEf_JOIN_IN) {
11804 /** We are joining here so we don't want do clone
11805 something that is bad **/
11806 if (SvTYPE(sstr) == SVt_PVHV) {
11807 const HEK * const hvname = HvNAME_HEK(sstr);
11809 /** don't clone stashes if they already exist **/
11810 dstr = MUTABLE_SV(gv_stashpvn(HEK_KEY(hvname), HEK_LEN(hvname),
11811 HEK_UTF8(hvname) ? SVf_UTF8 : 0));
11812 ptr_table_store(PL_ptr_table, sstr, dstr);
11816 else if (SvTYPE(sstr) == SVt_PVGV && !SvFAKE(sstr)) {
11817 HV *stash = GvSTASH(sstr);
11818 const HEK * hvname;
11819 if (stash && (hvname = HvNAME_HEK(stash))) {
11820 /** don't clone GVs if they already exist **/
11822 stash = gv_stashpvn(HEK_KEY(hvname), HEK_LEN(hvname),
11823 HEK_UTF8(hvname) ? SVf_UTF8 : 0);
11825 stash, GvNAME(sstr),
11831 if (svp && *svp && SvTYPE(*svp) == SVt_PVGV) {
11832 ptr_table_store(PL_ptr_table, sstr, *svp);
11839 /* create anew and remember what it is */
11842 #ifdef DEBUG_LEAKING_SCALARS
11843 dstr->sv_debug_optype = sstr->sv_debug_optype;
11844 dstr->sv_debug_line = sstr->sv_debug_line;
11845 dstr->sv_debug_inpad = sstr->sv_debug_inpad;
11846 dstr->sv_debug_parent = (SV*)sstr;
11847 FREE_SV_DEBUG_FILE(dstr);
11848 dstr->sv_debug_file = savepv(sstr->sv_debug_file);
11851 ptr_table_store(PL_ptr_table, sstr, dstr);
11854 SvFLAGS(dstr) = SvFLAGS(sstr);
11855 SvFLAGS(dstr) &= ~SVf_OOK; /* don't propagate OOK hack */
11856 SvREFCNT(dstr) = 0; /* must be before any other dups! */
11859 if (SvANY(sstr) && PL_watch_pvx && SvPVX_const(sstr) == PL_watch_pvx)
11860 PerlIO_printf(Perl_debug_log, "watch at %p hit, found string \"%s\"\n",
11861 (void*)PL_watch_pvx, SvPVX_const(sstr));
11864 /* don't clone objects whose class has asked us not to */
11865 if (SvOBJECT(sstr) && ! (SvFLAGS(SvSTASH(sstr)) & SVphv_CLONEABLE)) {
11870 switch (SvTYPE(sstr)) {
11872 SvANY(dstr) = NULL;
11875 SvANY(dstr) = (XPVIV*)((char*)&(dstr->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
11877 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
11879 SvIV_set(dstr, SvIVX(sstr));
11883 SvANY(dstr) = new_XNV();
11884 SvNV_set(dstr, SvNVX(sstr));
11886 /* case SVt_BIND: */
11889 /* These are all the types that need complex bodies allocating. */
11891 const svtype sv_type = SvTYPE(sstr);
11892 const struct body_details *const sv_type_details
11893 = bodies_by_type + sv_type;
11897 Perl_croak(aTHX_ "Bizarre SvTYPE [%" IVdf "]", (IV)SvTYPE(sstr));
11912 assert(sv_type_details->body_size);
11913 if (sv_type_details->arena) {
11914 new_body_inline(new_body, sv_type);
11916 = (void*)((char*)new_body - sv_type_details->offset);
11918 new_body = new_NOARENA(sv_type_details);
11922 SvANY(dstr) = new_body;
11925 Copy(((char*)SvANY(sstr)) + sv_type_details->offset,
11926 ((char*)SvANY(dstr)) + sv_type_details->offset,
11927 sv_type_details->copy, char);
11929 Copy(((char*)SvANY(sstr)),
11930 ((char*)SvANY(dstr)),
11931 sv_type_details->body_size + sv_type_details->offset, char);
11934 if (sv_type != SVt_PVAV && sv_type != SVt_PVHV
11935 && !isGV_with_GP(dstr)
11936 && !(sv_type == SVt_PVIO && !(IoFLAGS(dstr) & IOf_FAKE_DIRP)))
11937 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
11939 /* The Copy above means that all the source (unduplicated) pointers
11940 are now in the destination. We can check the flags and the
11941 pointers in either, but it's possible that there's less cache
11942 missing by always going for the destination.
11943 FIXME - instrument and check that assumption */
11944 if (sv_type >= SVt_PVMG) {
11945 if ((sv_type == SVt_PVMG) && SvPAD_OUR(dstr)) {
11946 SvOURSTASH_set(dstr, hv_dup_inc(SvOURSTASH(dstr), param));
11947 } else if (SvMAGIC(dstr))
11948 SvMAGIC_set(dstr, mg_dup(SvMAGIC(dstr), param));
11950 SvSTASH_set(dstr, hv_dup_inc(SvSTASH(dstr), param));
11953 /* The cast silences a GCC warning about unhandled types. */
11954 switch ((int)sv_type) {
11964 /* FIXME for plugins */
11965 re_dup_guts((REGEXP*) sstr, (REGEXP*) dstr, param);
11968 /* XXX LvTARGOFF sometimes holds PMOP* when DEBUGGING */
11969 if (LvTYPE(dstr) == 't') /* for tie: unrefcnted fake (SV**) */
11970 LvTARG(dstr) = dstr;
11971 else if (LvTYPE(dstr) == 'T') /* for tie: fake HE */
11972 LvTARG(dstr) = MUTABLE_SV(he_dup((HE*)LvTARG(dstr), 0, param));
11974 LvTARG(dstr) = sv_dup_inc(LvTARG(dstr), param);
11976 /* non-GP case already handled above */
11977 if(isGV_with_GP(sstr)) {
11978 GvNAME_HEK(dstr) = hek_dup(GvNAME_HEK(dstr), param);
11979 /* Don't call sv_add_backref here as it's going to be
11980 created as part of the magic cloning of the symbol
11981 table--unless this is during a join and the stash
11982 is not actually being cloned. */
11983 /* Danger Will Robinson - GvGP(dstr) isn't initialised
11984 at the point of this comment. */
11985 GvSTASH(dstr) = hv_dup(GvSTASH(dstr), param);
11986 if (param->flags & CLONEf_JOIN_IN)
11987 Perl_sv_add_backref(aTHX_ MUTABLE_SV(GvSTASH(dstr)), dstr);
11988 GvGP_set(dstr, gp_dup(GvGP(sstr), param));
11989 (void)GpREFCNT_inc(GvGP(dstr));
11993 /* PL_parser->rsfp_filters entries have fake IoDIRP() */
11994 if(IoFLAGS(dstr) & IOf_FAKE_DIRP) {
11995 /* I have no idea why fake dirp (rsfps)
11996 should be treated differently but otherwise
11997 we end up with leaks -- sky*/
11998 IoTOP_GV(dstr) = gv_dup_inc(IoTOP_GV(dstr), param);
11999 IoFMT_GV(dstr) = gv_dup_inc(IoFMT_GV(dstr), param);
12000 IoBOTTOM_GV(dstr) = gv_dup_inc(IoBOTTOM_GV(dstr), param);
12002 IoTOP_GV(dstr) = gv_dup(IoTOP_GV(dstr), param);
12003 IoFMT_GV(dstr) = gv_dup(IoFMT_GV(dstr), param);
12004 IoBOTTOM_GV(dstr) = gv_dup(IoBOTTOM_GV(dstr), param);
12005 if (IoDIRP(dstr)) {
12006 IoDIRP(dstr) = dirp_dup(IoDIRP(dstr), param);
12009 /* IoDIRP(dstr) is already a copy of IoDIRP(sstr) */
12011 IoIFP(dstr) = fp_dup(IoIFP(sstr), IoTYPE(dstr), param);
12013 if (IoOFP(dstr) == IoIFP(sstr))
12014 IoOFP(dstr) = IoIFP(dstr);
12016 IoOFP(dstr) = fp_dup(IoOFP(dstr), IoTYPE(dstr), param);
12017 IoTOP_NAME(dstr) = SAVEPV(IoTOP_NAME(dstr));
12018 IoFMT_NAME(dstr) = SAVEPV(IoFMT_NAME(dstr));
12019 IoBOTTOM_NAME(dstr) = SAVEPV(IoBOTTOM_NAME(dstr));
12022 /* avoid cloning an empty array */
12023 if (AvARRAY((const AV *)sstr) && AvFILLp((const AV *)sstr) >= 0) {
12024 SV **dst_ary, **src_ary;
12025 SSize_t items = AvFILLp((const AV *)sstr) + 1;
12027 src_ary = AvARRAY((const AV *)sstr);
12028 Newxz(dst_ary, AvMAX((const AV *)sstr)+1, SV*);
12029 ptr_table_store(PL_ptr_table, src_ary, dst_ary);
12030 AvARRAY(MUTABLE_AV(dstr)) = dst_ary;
12031 AvALLOC((const AV *)dstr) = dst_ary;
12032 if (AvREAL((const AV *)sstr)) {
12033 dst_ary = sv_dup_inc_multiple(src_ary, dst_ary, items,
12037 while (items-- > 0)
12038 *dst_ary++ = sv_dup(*src_ary++, param);
12040 items = AvMAX((const AV *)sstr) - AvFILLp((const AV *)sstr);
12041 while (items-- > 0) {
12042 *dst_ary++ = &PL_sv_undef;
12046 AvARRAY(MUTABLE_AV(dstr)) = NULL;
12047 AvALLOC((const AV *)dstr) = (SV**)NULL;
12048 AvMAX( (const AV *)dstr) = -1;
12049 AvFILLp((const AV *)dstr) = -1;
12053 if (HvARRAY((const HV *)sstr)) {
12055 const bool sharekeys = !!HvSHAREKEYS(sstr);
12056 XPVHV * const dxhv = (XPVHV*)SvANY(dstr);
12057 XPVHV * const sxhv = (XPVHV*)SvANY(sstr);
12059 Newx(darray, PERL_HV_ARRAY_ALLOC_BYTES(dxhv->xhv_max+1)
12060 + (SvOOK(sstr) ? sizeof(struct xpvhv_aux) : 0),
12062 HvARRAY(dstr) = (HE**)darray;
12063 while (i <= sxhv->xhv_max) {
12064 const HE * const source = HvARRAY(sstr)[i];
12065 HvARRAY(dstr)[i] = source
12066 ? he_dup(source, sharekeys, param) : 0;
12070 const struct xpvhv_aux * const saux = HvAUX(sstr);
12071 struct xpvhv_aux * const daux = HvAUX(dstr);
12072 /* This flag isn't copied. */
12075 if (saux->xhv_name_count) {
12076 HEK ** const sname = saux->xhv_name_u.xhvnameu_names;
12078 = saux->xhv_name_count < 0
12079 ? -saux->xhv_name_count
12080 : saux->xhv_name_count;
12081 HEK **shekp = sname + count;
12083 Newx(daux->xhv_name_u.xhvnameu_names, count, HEK *);
12084 dhekp = daux->xhv_name_u.xhvnameu_names + count;
12085 while (shekp-- > sname) {
12087 *dhekp = hek_dup(*shekp, param);
12091 daux->xhv_name_u.xhvnameu_name
12092 = hek_dup(saux->xhv_name_u.xhvnameu_name,
12095 daux->xhv_name_count = saux->xhv_name_count;
12097 daux->xhv_riter = saux->xhv_riter;
12098 daux->xhv_eiter = saux->xhv_eiter
12099 ? he_dup(saux->xhv_eiter,
12100 cBOOL(HvSHAREKEYS(sstr)), param) : 0;
12101 /* backref array needs refcnt=2; see sv_add_backref */
12102 daux->xhv_backreferences =
12103 (param->flags & CLONEf_JOIN_IN)
12104 /* when joining, we let the individual GVs and
12105 * CVs add themselves to backref as
12106 * needed. This avoids pulling in stuff
12107 * that isn't required, and simplifies the
12108 * case where stashes aren't cloned back
12109 * if they already exist in the parent
12112 : saux->xhv_backreferences
12113 ? (SvTYPE(saux->xhv_backreferences) == SVt_PVAV)
12114 ? MUTABLE_AV(SvREFCNT_inc(
12115 sv_dup_inc((const SV *)
12116 saux->xhv_backreferences, param)))
12117 : MUTABLE_AV(sv_dup((const SV *)
12118 saux->xhv_backreferences, param))
12121 daux->xhv_mro_meta = saux->xhv_mro_meta
12122 ? mro_meta_dup(saux->xhv_mro_meta, param)
12125 /* Record stashes for possible cloning in Perl_clone(). */
12127 av_push(param->stashes, dstr);
12131 HvARRAY(MUTABLE_HV(dstr)) = NULL;
12134 if (!(param->flags & CLONEf_COPY_STACKS)) {
12139 /* NOTE: not refcounted */
12140 SvANY(MUTABLE_CV(dstr))->xcv_stash =
12141 hv_dup(CvSTASH(dstr), param);
12142 if ((param->flags & CLONEf_JOIN_IN) && CvSTASH(dstr))
12143 Perl_sv_add_backref(aTHX_ MUTABLE_SV(CvSTASH(dstr)), dstr);
12144 if (!CvISXSUB(dstr)) {
12146 CvROOT(dstr) = OpREFCNT_inc(CvROOT(dstr));
12148 } else if (CvCONST(dstr)) {
12149 CvXSUBANY(dstr).any_ptr =
12150 sv_dup_inc((const SV *)CvXSUBANY(dstr).any_ptr, param);
12152 if (CvDYNFILE(dstr)) CvFILE(dstr) = SAVEPV(CvFILE(dstr));
12153 /* don't dup if copying back - CvGV isn't refcounted, so the
12154 * duped GV may never be freed. A bit of a hack! DAPM */
12155 SvANY(MUTABLE_CV(dstr))->xcv_gv =
12157 ? gv_dup_inc(CvGV(sstr), param)
12158 : (param->flags & CLONEf_JOIN_IN)
12160 : gv_dup(CvGV(sstr), param);
12162 CvPADLIST(dstr) = padlist_dup(CvPADLIST(sstr), param);
12164 CvWEAKOUTSIDE(sstr)
12165 ? cv_dup( CvOUTSIDE(dstr), param)
12166 : cv_dup_inc(CvOUTSIDE(dstr), param);
12172 if (SvOBJECT(dstr) && SvTYPE(dstr) != SVt_PVIO)
12179 Perl_sv_dup_inc(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
12181 PERL_ARGS_ASSERT_SV_DUP_INC;
12182 return sstr ? SvREFCNT_inc(sv_dup_common(sstr, param)) : NULL;
12186 Perl_sv_dup(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
12188 SV *dstr = sstr ? sv_dup_common(sstr, param) : NULL;
12189 PERL_ARGS_ASSERT_SV_DUP;
12191 /* Track every SV that (at least initially) had a reference count of 0.
12192 We need to do this by holding an actual reference to it in this array.
12193 If we attempt to cheat, turn AvREAL_off(), and store only pointers
12194 (akin to the stashes hash, and the perl stack), we come unstuck if
12195 a weak reference (or other SV legitimately SvREFCNT() == 0 for this
12196 thread) is manipulated in a CLONE method, because CLONE runs before the
12197 unreferenced array is walked to find SVs still with SvREFCNT() == 0
12198 (and fix things up by giving each a reference via the temps stack).
12199 Instead, during CLONE, if the 0-referenced SV has SvREFCNT_inc() and
12200 then SvREFCNT_dec(), it will be cleaned up (and added to the free list)
12201 before the walk of unreferenced happens and a reference to that is SV
12202 added to the temps stack. At which point we have the same SV considered
12203 to be in use, and free to be re-used. Not good.
12205 if (dstr && !(param->flags & CLONEf_COPY_STACKS) && !SvREFCNT(dstr)) {
12206 assert(param->unreferenced);
12207 av_push(param->unreferenced, SvREFCNT_inc(dstr));
12213 /* duplicate a context */
12216 Perl_cx_dup(pTHX_ PERL_CONTEXT *cxs, I32 ix, I32 max, CLONE_PARAMS* param)
12218 PERL_CONTEXT *ncxs;
12220 PERL_ARGS_ASSERT_CX_DUP;
12223 return (PERL_CONTEXT*)NULL;
12225 /* look for it in the table first */
12226 ncxs = (PERL_CONTEXT*)ptr_table_fetch(PL_ptr_table, cxs);
12230 /* create anew and remember what it is */
12231 Newx(ncxs, max + 1, PERL_CONTEXT);
12232 ptr_table_store(PL_ptr_table, cxs, ncxs);
12233 Copy(cxs, ncxs, max + 1, PERL_CONTEXT);
12236 PERL_CONTEXT * const ncx = &ncxs[ix];
12237 if (CxTYPE(ncx) == CXt_SUBST) {
12238 Perl_croak(aTHX_ "Cloning substitution context is unimplemented");
12241 switch (CxTYPE(ncx)) {
12243 ncx->blk_sub.cv = (ncx->blk_sub.olddepth == 0
12244 ? cv_dup_inc(ncx->blk_sub.cv, param)
12245 : cv_dup(ncx->blk_sub.cv,param));
12246 ncx->blk_sub.argarray = (CxHASARGS(ncx)
12247 ? av_dup_inc(ncx->blk_sub.argarray,
12250 ncx->blk_sub.savearray = av_dup_inc(ncx->blk_sub.savearray,
12252 ncx->blk_sub.oldcomppad = (PAD*)ptr_table_fetch(PL_ptr_table,
12253 ncx->blk_sub.oldcomppad);
12256 ncx->blk_eval.old_namesv = sv_dup_inc(ncx->blk_eval.old_namesv,
12258 ncx->blk_eval.cur_text = sv_dup(ncx->blk_eval.cur_text, param);
12260 case CXt_LOOP_LAZYSV:
12261 ncx->blk_loop.state_u.lazysv.end
12262 = sv_dup_inc(ncx->blk_loop.state_u.lazysv.end, param);
12263 /* We are taking advantage of av_dup_inc and sv_dup_inc
12264 actually being the same function, and order equivalence of
12266 We can assert the later [but only at run time :-(] */
12267 assert ((void *) &ncx->blk_loop.state_u.ary.ary ==
12268 (void *) &ncx->blk_loop.state_u.lazysv.cur);
12270 ncx->blk_loop.state_u.ary.ary
12271 = av_dup_inc(ncx->blk_loop.state_u.ary.ary, param);
12272 case CXt_LOOP_LAZYIV:
12273 case CXt_LOOP_PLAIN:
12274 if (CxPADLOOP(ncx)) {
12275 ncx->blk_loop.itervar_u.oldcomppad
12276 = (PAD*)ptr_table_fetch(PL_ptr_table,
12277 ncx->blk_loop.itervar_u.oldcomppad);
12279 ncx->blk_loop.itervar_u.gv
12280 = gv_dup((const GV *)ncx->blk_loop.itervar_u.gv,
12285 ncx->blk_format.cv = cv_dup(ncx->blk_format.cv, param);
12286 ncx->blk_format.gv = gv_dup(ncx->blk_format.gv, param);
12287 ncx->blk_format.dfoutgv = gv_dup_inc(ncx->blk_format.dfoutgv,
12300 /* duplicate a stack info structure */
12303 Perl_si_dup(pTHX_ PERL_SI *si, CLONE_PARAMS* param)
12307 PERL_ARGS_ASSERT_SI_DUP;
12310 return (PERL_SI*)NULL;
12312 /* look for it in the table first */
12313 nsi = (PERL_SI*)ptr_table_fetch(PL_ptr_table, si);
12317 /* create anew and remember what it is */
12318 Newxz(nsi, 1, PERL_SI);
12319 ptr_table_store(PL_ptr_table, si, nsi);
12321 nsi->si_stack = av_dup_inc(si->si_stack, param);
12322 nsi->si_cxix = si->si_cxix;
12323 nsi->si_cxmax = si->si_cxmax;
12324 nsi->si_cxstack = cx_dup(si->si_cxstack, si->si_cxix, si->si_cxmax, param);
12325 nsi->si_type = si->si_type;
12326 nsi->si_prev = si_dup(si->si_prev, param);
12327 nsi->si_next = si_dup(si->si_next, param);
12328 nsi->si_markoff = si->si_markoff;
12333 #define POPINT(ss,ix) ((ss)[--(ix)].any_i32)
12334 #define TOPINT(ss,ix) ((ss)[ix].any_i32)
12335 #define POPLONG(ss,ix) ((ss)[--(ix)].any_long)
12336 #define TOPLONG(ss,ix) ((ss)[ix].any_long)
12337 #define POPIV(ss,ix) ((ss)[--(ix)].any_iv)
12338 #define TOPIV(ss,ix) ((ss)[ix].any_iv)
12339 #define POPUV(ss,ix) ((ss)[--(ix)].any_uv)
12340 #define TOPUV(ss,ix) ((ss)[ix].any_uv)
12341 #define POPBOOL(ss,ix) ((ss)[--(ix)].any_bool)
12342 #define TOPBOOL(ss,ix) ((ss)[ix].any_bool)
12343 #define POPPTR(ss,ix) ((ss)[--(ix)].any_ptr)
12344 #define TOPPTR(ss,ix) ((ss)[ix].any_ptr)
12345 #define POPDPTR(ss,ix) ((ss)[--(ix)].any_dptr)
12346 #define TOPDPTR(ss,ix) ((ss)[ix].any_dptr)
12347 #define POPDXPTR(ss,ix) ((ss)[--(ix)].any_dxptr)
12348 #define TOPDXPTR(ss,ix) ((ss)[ix].any_dxptr)
12351 #define pv_dup_inc(p) SAVEPV(p)
12352 #define pv_dup(p) SAVEPV(p)
12353 #define svp_dup_inc(p,pp) any_dup(p,pp)
12355 /* map any object to the new equivent - either something in the
12356 * ptr table, or something in the interpreter structure
12360 Perl_any_dup(pTHX_ void *v, const PerlInterpreter *proto_perl)
12364 PERL_ARGS_ASSERT_ANY_DUP;
12367 return (void*)NULL;
12369 /* look for it in the table first */
12370 ret = ptr_table_fetch(PL_ptr_table, v);
12374 /* see if it is part of the interpreter structure */
12375 if (v >= (void*)proto_perl && v < (void*)(proto_perl+1))
12376 ret = (void*)(((char*)aTHX) + (((char*)v) - (char*)proto_perl));
12384 /* duplicate the save stack */
12387 Perl_ss_dup(pTHX_ PerlInterpreter *proto_perl, CLONE_PARAMS* param)
12390 ANY * const ss = proto_perl->Isavestack;
12391 const I32 max = proto_perl->Isavestack_max;
12392 I32 ix = proto_perl->Isavestack_ix;
12405 void (*dptr) (void*);
12406 void (*dxptr) (pTHX_ void*);
12408 PERL_ARGS_ASSERT_SS_DUP;
12410 Newxz(nss, max, ANY);
12413 const UV uv = POPUV(ss,ix);
12414 const U8 type = (U8)uv & SAVE_MASK;
12416 TOPUV(nss,ix) = uv;
12418 case SAVEt_CLEARSV:
12420 case SAVEt_HELEM: /* hash element */
12421 sv = (const SV *)POPPTR(ss,ix);
12422 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12424 case SAVEt_ITEM: /* normal string */
12425 case SAVEt_GVSV: /* scalar slot in GV */
12426 case SAVEt_SV: /* scalar reference */
12427 sv = (const SV *)POPPTR(ss,ix);
12428 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12431 case SAVEt_MORTALIZESV:
12432 sv = (const SV *)POPPTR(ss,ix);
12433 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12435 case SAVEt_SHARED_PVREF: /* char* in shared space */
12436 c = (char*)POPPTR(ss,ix);
12437 TOPPTR(nss,ix) = savesharedpv(c);
12438 ptr = POPPTR(ss,ix);
12439 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12441 case SAVEt_GENERIC_SVREF: /* generic sv */
12442 case SAVEt_SVREF: /* scalar reference */
12443 sv = (const SV *)POPPTR(ss,ix);
12444 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12445 ptr = POPPTR(ss,ix);
12446 TOPPTR(nss,ix) = svp_dup_inc((SV**)ptr, proto_perl);/* XXXXX */
12448 case SAVEt_HV: /* hash reference */
12449 case SAVEt_AV: /* array reference */
12450 sv = (const SV *) POPPTR(ss,ix);
12451 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12453 case SAVEt_COMPPAD:
12455 sv = (const SV *) POPPTR(ss,ix);
12456 TOPPTR(nss,ix) = sv_dup(sv, param);
12458 case SAVEt_INT: /* int reference */
12459 ptr = POPPTR(ss,ix);
12460 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12461 intval = (int)POPINT(ss,ix);
12462 TOPINT(nss,ix) = intval;
12464 case SAVEt_LONG: /* long reference */
12465 ptr = POPPTR(ss,ix);
12466 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12467 longval = (long)POPLONG(ss,ix);
12468 TOPLONG(nss,ix) = longval;
12470 case SAVEt_I32: /* I32 reference */
12471 ptr = POPPTR(ss,ix);
12472 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12474 TOPINT(nss,ix) = i;
12476 case SAVEt_IV: /* IV reference */
12477 ptr = POPPTR(ss,ix);
12478 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12480 TOPIV(nss,ix) = iv;
12482 case SAVEt_HPTR: /* HV* reference */
12483 case SAVEt_APTR: /* AV* reference */
12484 case SAVEt_SPTR: /* SV* reference */
12485 ptr = POPPTR(ss,ix);
12486 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12487 sv = (const SV *)POPPTR(ss,ix);
12488 TOPPTR(nss,ix) = sv_dup(sv, param);
12490 case SAVEt_VPTR: /* random* reference */
12491 ptr = POPPTR(ss,ix);
12492 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12494 case SAVEt_INT_SMALL:
12495 case SAVEt_I32_SMALL:
12496 case SAVEt_I16: /* I16 reference */
12497 case SAVEt_I8: /* I8 reference */
12499 ptr = POPPTR(ss,ix);
12500 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12502 case SAVEt_GENERIC_PVREF: /* generic char* */
12503 case SAVEt_PPTR: /* char* reference */
12504 ptr = POPPTR(ss,ix);
12505 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12506 c = (char*)POPPTR(ss,ix);
12507 TOPPTR(nss,ix) = pv_dup(c);
12509 case SAVEt_GP: /* scalar reference */
12510 gp = (GP*)POPPTR(ss,ix);
12511 TOPPTR(nss,ix) = gp = gp_dup(gp, param);
12512 (void)GpREFCNT_inc(gp);
12513 gv = (const GV *)POPPTR(ss,ix);
12514 TOPPTR(nss,ix) = gv_dup_inc(gv, param);
12517 ptr = POPPTR(ss,ix);
12518 if (ptr && (((OP*)ptr)->op_private & OPpREFCOUNTED)) {
12519 /* these are assumed to be refcounted properly */
12521 switch (((OP*)ptr)->op_type) {
12523 case OP_LEAVESUBLV:
12527 case OP_LEAVEWRITE:
12528 TOPPTR(nss,ix) = ptr;
12531 (void) OpREFCNT_inc(o);
12535 TOPPTR(nss,ix) = NULL;
12540 TOPPTR(nss,ix) = NULL;
12542 case SAVEt_FREECOPHH:
12543 ptr = POPPTR(ss,ix);
12544 TOPPTR(nss,ix) = cophh_copy((COPHH *)ptr);
12547 hv = (const HV *)POPPTR(ss,ix);
12548 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
12550 TOPINT(nss,ix) = i;
12553 c = (char*)POPPTR(ss,ix);
12554 TOPPTR(nss,ix) = pv_dup_inc(c);
12556 case SAVEt_STACK_POS: /* Position on Perl stack */
12558 TOPINT(nss,ix) = i;
12560 case SAVEt_DESTRUCTOR:
12561 ptr = POPPTR(ss,ix);
12562 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
12563 dptr = POPDPTR(ss,ix);
12564 TOPDPTR(nss,ix) = DPTR2FPTR(void (*)(void*),
12565 any_dup(FPTR2DPTR(void *, dptr),
12568 case SAVEt_DESTRUCTOR_X:
12569 ptr = POPPTR(ss,ix);
12570 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
12571 dxptr = POPDXPTR(ss,ix);
12572 TOPDXPTR(nss,ix) = DPTR2FPTR(void (*)(pTHX_ void*),
12573 any_dup(FPTR2DPTR(void *, dxptr),
12576 case SAVEt_REGCONTEXT:
12578 ix -= uv >> SAVE_TIGHT_SHIFT;
12580 case SAVEt_AELEM: /* array element */
12581 sv = (const SV *)POPPTR(ss,ix);
12582 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12584 TOPINT(nss,ix) = i;
12585 av = (const AV *)POPPTR(ss,ix);
12586 TOPPTR(nss,ix) = av_dup_inc(av, param);
12589 ptr = POPPTR(ss,ix);
12590 TOPPTR(nss,ix) = ptr;
12593 ptr = POPPTR(ss,ix);
12594 ptr = cophh_copy((COPHH*)ptr);
12595 TOPPTR(nss,ix) = ptr;
12597 TOPINT(nss,ix) = i;
12598 if (i & HINT_LOCALIZE_HH) {
12599 hv = (const HV *)POPPTR(ss,ix);
12600 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
12603 case SAVEt_PADSV_AND_MORTALIZE:
12604 longval = (long)POPLONG(ss,ix);
12605 TOPLONG(nss,ix) = longval;
12606 ptr = POPPTR(ss,ix);
12607 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12608 sv = (const SV *)POPPTR(ss,ix);
12609 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12611 case SAVEt_SET_SVFLAGS:
12613 TOPINT(nss,ix) = i;
12615 TOPINT(nss,ix) = i;
12616 sv = (const SV *)POPPTR(ss,ix);
12617 TOPPTR(nss,ix) = sv_dup(sv, param);
12619 case SAVEt_RE_STATE:
12621 const struct re_save_state *const old_state
12622 = (struct re_save_state *)
12623 (ss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
12624 struct re_save_state *const new_state
12625 = (struct re_save_state *)
12626 (nss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
12628 Copy(old_state, new_state, 1, struct re_save_state);
12629 ix -= SAVESTACK_ALLOC_FOR_RE_SAVE_STATE;
12631 new_state->re_state_bostr
12632 = pv_dup(old_state->re_state_bostr);
12633 new_state->re_state_reginput
12634 = pv_dup(old_state->re_state_reginput);
12635 new_state->re_state_regeol
12636 = pv_dup(old_state->re_state_regeol);
12637 new_state->re_state_regoffs
12638 = (regexp_paren_pair*)
12639 any_dup(old_state->re_state_regoffs, proto_perl);
12640 new_state->re_state_reglastparen
12641 = (U32*) any_dup(old_state->re_state_reglastparen,
12643 new_state->re_state_reglastcloseparen
12644 = (U32*)any_dup(old_state->re_state_reglastcloseparen,
12646 /* XXX This just has to be broken. The old save_re_context
12647 code did SAVEGENERICPV(PL_reg_start_tmp);
12648 PL_reg_start_tmp is char **.
12649 Look above to what the dup code does for
12650 SAVEt_GENERIC_PVREF
12651 It can never have worked.
12652 So this is merely a faithful copy of the exiting bug: */
12653 new_state->re_state_reg_start_tmp
12654 = (char **) pv_dup((char *)
12655 old_state->re_state_reg_start_tmp);
12656 /* I assume that it only ever "worked" because no-one called
12657 (pseudo)fork while the regexp engine had re-entered itself.
12659 #ifdef PERL_OLD_COPY_ON_WRITE
12660 new_state->re_state_nrs
12661 = sv_dup(old_state->re_state_nrs, param);
12663 new_state->re_state_reg_magic
12664 = (MAGIC*) any_dup(old_state->re_state_reg_magic,
12666 new_state->re_state_reg_oldcurpm
12667 = (PMOP*) any_dup(old_state->re_state_reg_oldcurpm,
12669 new_state->re_state_reg_curpm
12670 = (PMOP*) any_dup(old_state->re_state_reg_curpm,
12672 new_state->re_state_reg_oldsaved
12673 = pv_dup(old_state->re_state_reg_oldsaved);
12674 new_state->re_state_reg_poscache
12675 = pv_dup(old_state->re_state_reg_poscache);
12676 new_state->re_state_reg_starttry
12677 = pv_dup(old_state->re_state_reg_starttry);
12680 case SAVEt_COMPILE_WARNINGS:
12681 ptr = POPPTR(ss,ix);
12682 TOPPTR(nss,ix) = DUP_WARNINGS((STRLEN*)ptr);
12685 ptr = POPPTR(ss,ix);
12686 TOPPTR(nss,ix) = parser_dup((const yy_parser*)ptr, param);
12690 "panic: ss_dup inconsistency (%"IVdf")", (IV) type);
12698 /* if sv is a stash, call $class->CLONE_SKIP(), and set the SVphv_CLONEABLE
12699 * flag to the result. This is done for each stash before cloning starts,
12700 * so we know which stashes want their objects cloned */
12703 do_mark_cloneable_stash(pTHX_ SV *const sv)
12705 const HEK * const hvname = HvNAME_HEK((const HV *)sv);
12707 GV* const cloner = gv_fetchmethod_autoload(MUTABLE_HV(sv), "CLONE_SKIP", 0);
12708 SvFLAGS(sv) |= SVphv_CLONEABLE; /* clone objects by default */
12709 if (cloner && GvCV(cloner)) {
12716 mXPUSHs(newSVhek(hvname));
12718 call_sv(MUTABLE_SV(GvCV(cloner)), G_SCALAR);
12725 SvFLAGS(sv) &= ~SVphv_CLONEABLE;
12733 =for apidoc perl_clone
12735 Create and return a new interpreter by cloning the current one.
12737 perl_clone takes these flags as parameters:
12739 CLONEf_COPY_STACKS - is used to, well, copy the stacks also,
12740 without it we only clone the data and zero the stacks,
12741 with it we copy the stacks and the new perl interpreter is
12742 ready to run at the exact same point as the previous one.
12743 The pseudo-fork code uses COPY_STACKS while the
12744 threads->create doesn't.
12746 CLONEf_KEEP_PTR_TABLE -
12747 perl_clone keeps a ptr_table with the pointer of the old
12748 variable as a key and the new variable as a value,
12749 this allows it to check if something has been cloned and not
12750 clone it again but rather just use the value and increase the
12751 refcount. If KEEP_PTR_TABLE is not set then perl_clone will kill
12752 the ptr_table using the function
12753 C<ptr_table_free(PL_ptr_table); PL_ptr_table = NULL;>,
12754 reason to keep it around is if you want to dup some of your own
12755 variable who are outside the graph perl scans, example of this
12756 code is in threads.xs create.
12758 CLONEf_CLONE_HOST -
12759 This is a win32 thing, it is ignored on unix, it tells perls
12760 win32host code (which is c++) to clone itself, this is needed on
12761 win32 if you want to run two threads at the same time,
12762 if you just want to do some stuff in a separate perl interpreter
12763 and then throw it away and return to the original one,
12764 you don't need to do anything.
12769 /* XXX the above needs expanding by someone who actually understands it ! */
12770 EXTERN_C PerlInterpreter *
12771 perl_clone_host(PerlInterpreter* proto_perl, UV flags);
12774 perl_clone(PerlInterpreter *proto_perl, UV flags)
12777 #ifdef PERL_IMPLICIT_SYS
12779 PERL_ARGS_ASSERT_PERL_CLONE;
12781 /* perlhost.h so we need to call into it
12782 to clone the host, CPerlHost should have a c interface, sky */
12784 if (flags & CLONEf_CLONE_HOST) {
12785 return perl_clone_host(proto_perl,flags);
12787 return perl_clone_using(proto_perl, flags,
12789 proto_perl->IMemShared,
12790 proto_perl->IMemParse,
12792 proto_perl->IStdIO,
12796 proto_perl->IProc);
12800 perl_clone_using(PerlInterpreter *proto_perl, UV flags,
12801 struct IPerlMem* ipM, struct IPerlMem* ipMS,
12802 struct IPerlMem* ipMP, struct IPerlEnv* ipE,
12803 struct IPerlStdIO* ipStd, struct IPerlLIO* ipLIO,
12804 struct IPerlDir* ipD, struct IPerlSock* ipS,
12805 struct IPerlProc* ipP)
12807 /* XXX many of the string copies here can be optimized if they're
12808 * constants; they need to be allocated as common memory and just
12809 * their pointers copied. */
12812 CLONE_PARAMS clone_params;
12813 CLONE_PARAMS* const param = &clone_params;
12815 PerlInterpreter * const my_perl = (PerlInterpreter*)(*ipM->pMalloc)(ipM, sizeof(PerlInterpreter));
12817 PERL_ARGS_ASSERT_PERL_CLONE_USING;
12818 #else /* !PERL_IMPLICIT_SYS */
12820 CLONE_PARAMS clone_params;
12821 CLONE_PARAMS* param = &clone_params;
12822 PerlInterpreter * const my_perl = (PerlInterpreter*)PerlMem_malloc(sizeof(PerlInterpreter));
12824 PERL_ARGS_ASSERT_PERL_CLONE;
12825 #endif /* PERL_IMPLICIT_SYS */
12827 /* for each stash, determine whether its objects should be cloned */
12828 S_visit(proto_perl, do_mark_cloneable_stash, SVt_PVHV, SVTYPEMASK);
12829 PERL_SET_THX(my_perl);
12832 PoisonNew(my_perl, 1, PerlInterpreter);
12835 PL_defstash = NULL; /* may be used by perl malloc() */
12838 PL_scopestack_name = 0;
12840 PL_savestack_ix = 0;
12841 PL_savestack_max = -1;
12842 PL_sig_pending = 0;
12844 Zero(&PL_debug_pad, 1, struct perl_debug_pad);
12845 # ifdef DEBUG_LEAKING_SCALARS
12846 PL_sv_serial = (((UV)my_perl >> 2) & 0xfff) * 1000000;
12848 #else /* !DEBUGGING */
12849 Zero(my_perl, 1, PerlInterpreter);
12850 #endif /* DEBUGGING */
12852 #ifdef PERL_IMPLICIT_SYS
12853 /* host pointers */
12855 PL_MemShared = ipMS;
12856 PL_MemParse = ipMP;
12863 #endif /* PERL_IMPLICIT_SYS */
12865 param->flags = flags;
12866 /* Nothing in the core code uses this, but we make it available to
12867 extensions (using mg_dup). */
12868 param->proto_perl = proto_perl;
12869 /* Likely nothing will use this, but it is initialised to be consistent
12870 with Perl_clone_params_new(). */
12871 param->new_perl = my_perl;
12872 param->unreferenced = NULL;
12874 INIT_TRACK_MEMPOOL(my_perl->Imemory_debug_header, my_perl);
12876 PL_body_arenas = NULL;
12877 Zero(&PL_body_roots, 1, PL_body_roots);
12880 PL_sv_objcount = 0;
12882 PL_sv_arenaroot = NULL;
12884 PL_debug = proto_perl->Idebug;
12886 PL_hash_seed = proto_perl->Ihash_seed;
12887 PL_rehash_seed = proto_perl->Irehash_seed;
12889 SvANY(&PL_sv_undef) = NULL;
12890 SvREFCNT(&PL_sv_undef) = (~(U32)0)/2;
12891 SvFLAGS(&PL_sv_undef) = SVf_READONLY|SVt_NULL;
12892 SvREFCNT(&PL_sv_no) = (~(U32)0)/2;
12893 SvFLAGS(&PL_sv_no) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
12894 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
12896 SvANY(&PL_sv_yes) = new_XPVNV();
12897 SvREFCNT(&PL_sv_yes) = (~(U32)0)/2;
12898 SvFLAGS(&PL_sv_yes) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
12899 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
12901 /* dbargs array probably holds garbage */
12904 PL_compiling = proto_perl->Icompiling;
12906 #ifdef PERL_DEBUG_READONLY_OPS
12911 /* pseudo environmental stuff */
12912 PL_origargc = proto_perl->Iorigargc;
12913 PL_origargv = proto_perl->Iorigargv;
12915 /* Set tainting stuff before PerlIO_debug can possibly get called */
12916 PL_tainting = proto_perl->Itainting;
12917 PL_taint_warn = proto_perl->Itaint_warn;
12919 PL_minus_c = proto_perl->Iminus_c;
12921 PL_localpatches = proto_perl->Ilocalpatches;
12922 PL_splitstr = proto_perl->Isplitstr;
12923 PL_minus_n = proto_perl->Iminus_n;
12924 PL_minus_p = proto_perl->Iminus_p;
12925 PL_minus_l = proto_perl->Iminus_l;
12926 PL_minus_a = proto_perl->Iminus_a;
12927 PL_minus_E = proto_perl->Iminus_E;
12928 PL_minus_F = proto_perl->Iminus_F;
12929 PL_doswitches = proto_perl->Idoswitches;
12930 PL_dowarn = proto_perl->Idowarn;
12931 PL_sawampersand = proto_perl->Isawampersand;
12932 PL_unsafe = proto_perl->Iunsafe;
12933 PL_perldb = proto_perl->Iperldb;
12934 PL_perl_destruct_level = proto_perl->Iperl_destruct_level;
12935 PL_exit_flags = proto_perl->Iexit_flags;
12937 /* XXX time(&PL_basetime) when asked for? */
12938 PL_basetime = proto_perl->Ibasetime;
12940 PL_maxsysfd = proto_perl->Imaxsysfd;
12941 PL_statusvalue = proto_perl->Istatusvalue;
12943 PL_statusvalue_vms = proto_perl->Istatusvalue_vms;
12945 PL_statusvalue_posix = proto_perl->Istatusvalue_posix;
12948 /* RE engine related */
12949 Zero(&PL_reg_state, 1, struct re_save_state);
12950 PL_reginterp_cnt = 0;
12951 PL_regmatch_slab = NULL;
12953 PL_sub_generation = proto_perl->Isub_generation;
12955 /* funky return mechanisms */
12956 PL_forkprocess = proto_perl->Iforkprocess;
12958 /* internal state */
12959 PL_maxo = proto_perl->Imaxo;
12961 PL_main_start = proto_perl->Imain_start;
12962 PL_eval_root = proto_perl->Ieval_root;
12963 PL_eval_start = proto_perl->Ieval_start;
12965 PL_filemode = proto_perl->Ifilemode;
12966 PL_lastfd = proto_perl->Ilastfd;
12967 PL_oldname = proto_perl->Ioldname; /* XXX not quite right */
12970 PL_gensym = proto_perl->Igensym;
12972 PL_laststatval = proto_perl->Ilaststatval;
12973 PL_laststype = proto_perl->Ilaststype;
12976 PL_profiledata = NULL;
12978 PL_generation = proto_perl->Igeneration;
12980 PL_in_clean_objs = proto_perl->Iin_clean_objs;
12981 PL_in_clean_all = proto_perl->Iin_clean_all;
12983 PL_uid = proto_perl->Iuid;
12984 PL_euid = proto_perl->Ieuid;
12985 PL_gid = proto_perl->Igid;
12986 PL_egid = proto_perl->Iegid;
12987 PL_nomemok = proto_perl->Inomemok;
12988 PL_an = proto_perl->Ian;
12989 PL_evalseq = proto_perl->Ievalseq;
12990 PL_origenviron = proto_perl->Iorigenviron; /* XXX not quite right */
12991 PL_origalen = proto_perl->Iorigalen;
12993 PL_sighandlerp = proto_perl->Isighandlerp;
12995 PL_runops = proto_perl->Irunops;
12997 PL_subline = proto_perl->Isubline;
13000 PL_cryptseen = proto_perl->Icryptseen;
13003 PL_hints = proto_perl->Ihints;
13005 PL_amagic_generation = proto_perl->Iamagic_generation;
13007 #ifdef USE_LOCALE_COLLATE
13008 PL_collation_ix = proto_perl->Icollation_ix;
13009 PL_collation_standard = proto_perl->Icollation_standard;
13010 PL_collxfrm_base = proto_perl->Icollxfrm_base;
13011 PL_collxfrm_mult = proto_perl->Icollxfrm_mult;
13012 #endif /* USE_LOCALE_COLLATE */
13014 #ifdef USE_LOCALE_NUMERIC
13015 PL_numeric_standard = proto_perl->Inumeric_standard;
13016 PL_numeric_local = proto_perl->Inumeric_local;
13017 #endif /* !USE_LOCALE_NUMERIC */
13019 /* Did the locale setup indicate UTF-8? */
13020 PL_utf8locale = proto_perl->Iutf8locale;
13021 /* Unicode features (see perlrun/-C) */
13022 PL_unicode = proto_perl->Iunicode;
13024 /* Pre-5.8 signals control */
13025 PL_signals = proto_perl->Isignals;
13027 /* times() ticks per second */
13028 PL_clocktick = proto_perl->Iclocktick;
13030 /* Recursion stopper for PerlIO_find_layer */
13031 PL_in_load_module = proto_perl->Iin_load_module;
13033 /* sort() routine */
13034 PL_sort_RealCmp = proto_perl->Isort_RealCmp;
13036 /* Not really needed/useful since the reenrant_retint is "volatile",
13037 * but do it for consistency's sake. */
13038 PL_reentrant_retint = proto_perl->Ireentrant_retint;
13040 /* Hooks to shared SVs and locks. */
13041 PL_sharehook = proto_perl->Isharehook;
13042 PL_lockhook = proto_perl->Ilockhook;
13043 PL_unlockhook = proto_perl->Iunlockhook;
13044 PL_threadhook = proto_perl->Ithreadhook;
13045 PL_destroyhook = proto_perl->Idestroyhook;
13046 PL_signalhook = proto_perl->Isignalhook;
13048 PL_globhook = proto_perl->Iglobhook;
13050 #ifdef THREADS_HAVE_PIDS
13051 PL_ppid = proto_perl->Ippid;
13055 PL_last_swash_hv = NULL; /* reinits on demand */
13056 PL_last_swash_klen = 0;
13057 PL_last_swash_key[0]= '\0';
13058 PL_last_swash_tmps = (U8*)NULL;
13059 PL_last_swash_slen = 0;
13061 PL_glob_index = proto_perl->Iglob_index;
13062 PL_srand_called = proto_perl->Isrand_called;
13064 if (flags & CLONEf_COPY_STACKS) {
13065 /* next allocation will be PL_tmps_stack[PL_tmps_ix+1] */
13066 PL_tmps_ix = proto_perl->Itmps_ix;
13067 PL_tmps_max = proto_perl->Itmps_max;
13068 PL_tmps_floor = proto_perl->Itmps_floor;
13070 /* next push_scope()/ENTER sets PL_scopestack[PL_scopestack_ix]
13071 * NOTE: unlike the others! */
13072 PL_scopestack_ix = proto_perl->Iscopestack_ix;
13073 PL_scopestack_max = proto_perl->Iscopestack_max;
13075 /* next SSPUSHFOO() sets PL_savestack[PL_savestack_ix]
13076 * NOTE: unlike the others! */
13077 PL_savestack_ix = proto_perl->Isavestack_ix;
13078 PL_savestack_max = proto_perl->Isavestack_max;
13081 PL_start_env = proto_perl->Istart_env; /* XXXXXX */
13082 PL_top_env = &PL_start_env;
13084 PL_op = proto_perl->Iop;
13087 PL_Xpv = (XPV*)NULL;
13088 my_perl->Ina = proto_perl->Ina;
13090 PL_statbuf = proto_perl->Istatbuf;
13091 PL_statcache = proto_perl->Istatcache;
13094 PL_timesbuf = proto_perl->Itimesbuf;
13097 PL_tainted = proto_perl->Itainted;
13098 PL_curpm = proto_perl->Icurpm; /* XXX No PMOP ref count */
13100 PL_chopset = proto_perl->Ichopset; /* XXX never deallocated */
13102 PL_restartjmpenv = proto_perl->Irestartjmpenv;
13103 PL_restartop = proto_perl->Irestartop;
13104 PL_in_eval = proto_perl->Iin_eval;
13105 PL_delaymagic = proto_perl->Idelaymagic;
13106 PL_phase = proto_perl->Iphase;
13107 PL_localizing = proto_perl->Ilocalizing;
13109 PL_hv_fetch_ent_mh = NULL;
13110 PL_modcount = proto_perl->Imodcount;
13111 PL_lastgotoprobe = NULL;
13112 PL_dumpindent = proto_perl->Idumpindent;
13114 PL_efloatbuf = NULL; /* reinits on demand */
13115 PL_efloatsize = 0; /* reinits on demand */
13119 PL_regdummy = proto_perl->Iregdummy;
13120 PL_colorset = 0; /* reinits PL_colors[] */
13121 /*PL_colors[6] = {0,0,0,0,0,0};*/
13123 /* Pluggable optimizer */
13124 PL_peepp = proto_perl->Ipeepp;
13125 PL_rpeepp = proto_perl->Irpeepp;
13126 /* op_free() hook */
13127 PL_opfreehook = proto_perl->Iopfreehook;
13129 #ifdef USE_REENTRANT_API
13130 /* XXX: things like -Dm will segfault here in perlio, but doing
13131 * PERL_SET_CONTEXT(proto_perl);
13132 * breaks too many other things
13134 Perl_reentrant_init(aTHX);
13137 /* create SV map for pointer relocation */
13138 PL_ptr_table = ptr_table_new();
13140 /* initialize these special pointers as early as possible */
13141 ptr_table_store(PL_ptr_table, &proto_perl->Isv_undef, &PL_sv_undef);
13143 SvANY(&PL_sv_no) = new_XPVNV();
13144 SvPV_set(&PL_sv_no, savepvn(PL_No, 0));
13145 SvCUR_set(&PL_sv_no, 0);
13146 SvLEN_set(&PL_sv_no, 1);
13147 SvIV_set(&PL_sv_no, 0);
13148 SvNV_set(&PL_sv_no, 0);
13149 ptr_table_store(PL_ptr_table, &proto_perl->Isv_no, &PL_sv_no);
13151 SvPV_set(&PL_sv_yes, savepvn(PL_Yes, 1));
13152 SvCUR_set(&PL_sv_yes, 1);
13153 SvLEN_set(&PL_sv_yes, 2);
13154 SvIV_set(&PL_sv_yes, 1);
13155 SvNV_set(&PL_sv_yes, 1);
13156 ptr_table_store(PL_ptr_table, &proto_perl->Isv_yes, &PL_sv_yes);
13158 /* create (a non-shared!) shared string table */
13159 PL_strtab = newHV();
13160 HvSHAREKEYS_off(PL_strtab);
13161 hv_ksplit(PL_strtab, HvTOTALKEYS(proto_perl->Istrtab));
13162 ptr_table_store(PL_ptr_table, proto_perl->Istrtab, PL_strtab);
13164 /* These two PVs will be free'd special way so must set them same way op.c does */
13165 PL_compiling.cop_stashpv = savesharedpv(PL_compiling.cop_stashpv);
13166 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_stashpv, PL_compiling.cop_stashpv);
13168 PL_compiling.cop_file = savesharedpv(PL_compiling.cop_file);
13169 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_file, PL_compiling.cop_file);
13171 ptr_table_store(PL_ptr_table, &proto_perl->Icompiling, &PL_compiling);
13172 PL_compiling.cop_warnings = DUP_WARNINGS(PL_compiling.cop_warnings);
13173 CopHINTHASH_set(&PL_compiling, cophh_copy(CopHINTHASH_get(&PL_compiling)));
13174 PL_curcop = (COP*)any_dup(proto_perl->Icurcop, proto_perl);
13176 param->stashes = newAV(); /* Setup array of objects to call clone on */
13177 /* This makes no difference to the implementation, as it always pushes
13178 and shifts pointers to other SVs without changing their reference
13179 count, with the array becoming empty before it is freed. However, it
13180 makes it conceptually clear what is going on, and will avoid some
13181 work inside av.c, filling slots between AvFILL() and AvMAX() with
13182 &PL_sv_undef, and SvREFCNT_dec()ing those. */
13183 AvREAL_off(param->stashes);
13185 if (!(flags & CLONEf_COPY_STACKS)) {
13186 param->unreferenced = newAV();
13189 #ifdef PERLIO_LAYERS
13190 /* Clone PerlIO tables as soon as we can handle general xx_dup() */
13191 PerlIO_clone(aTHX_ proto_perl, param);
13194 PL_envgv = gv_dup(proto_perl->Ienvgv, param);
13195 PL_incgv = gv_dup(proto_perl->Iincgv, param);
13196 PL_hintgv = gv_dup(proto_perl->Ihintgv, param);
13197 PL_origfilename = SAVEPV(proto_perl->Iorigfilename);
13198 PL_diehook = sv_dup_inc(proto_perl->Idiehook, param);
13199 PL_warnhook = sv_dup_inc(proto_perl->Iwarnhook, param);
13202 PL_patchlevel = sv_dup_inc(proto_perl->Ipatchlevel, param);
13203 PL_apiversion = sv_dup_inc(proto_perl->Iapiversion, param);
13204 PL_inplace = SAVEPV(proto_perl->Iinplace);
13205 PL_e_script = sv_dup_inc(proto_perl->Ie_script, param);
13207 /* magical thingies */
13208 PL_formfeed = sv_dup(proto_perl->Iformfeed, param);
13210 PL_encoding = sv_dup(proto_perl->Iencoding, param);
13212 sv_setpvs(PERL_DEBUG_PAD(0), ""); /* For regex debugging. */
13213 sv_setpvs(PERL_DEBUG_PAD(1), ""); /* ext/re needs these */
13214 sv_setpvs(PERL_DEBUG_PAD(2), ""); /* even without DEBUGGING. */
13217 /* Clone the regex array */
13218 /* ORANGE FIXME for plugins, probably in the SV dup code.
13219 newSViv(PTR2IV(CALLREGDUPE(
13220 INT2PTR(REGEXP *, SvIVX(regex)), param))))
13222 PL_regex_padav = av_dup_inc(proto_perl->Iregex_padav, param);
13223 PL_regex_pad = AvARRAY(PL_regex_padav);
13225 /* shortcuts to various I/O objects */
13226 PL_ofsgv = gv_dup_inc(proto_perl->Iofsgv, param);
13227 PL_stdingv = gv_dup(proto_perl->Istdingv, param);
13228 PL_stderrgv = gv_dup(proto_perl->Istderrgv, param);
13229 PL_defgv = gv_dup(proto_perl->Idefgv, param);
13230 PL_argvgv = gv_dup(proto_perl->Iargvgv, param);
13231 PL_argvoutgv = gv_dup(proto_perl->Iargvoutgv, param);
13232 PL_argvout_stack = av_dup_inc(proto_perl->Iargvout_stack, param);
13234 /* shortcuts to regexp stuff */
13235 PL_replgv = gv_dup(proto_perl->Ireplgv, param);
13237 /* shortcuts to misc objects */
13238 PL_errgv = gv_dup(proto_perl->Ierrgv, param);
13240 /* shortcuts to debugging objects */
13241 PL_DBgv = gv_dup(proto_perl->IDBgv, param);
13242 PL_DBline = gv_dup(proto_perl->IDBline, param);
13243 PL_DBsub = gv_dup(proto_perl->IDBsub, param);
13244 PL_DBsingle = sv_dup(proto_perl->IDBsingle, param);
13245 PL_DBtrace = sv_dup(proto_perl->IDBtrace, param);
13246 PL_DBsignal = sv_dup(proto_perl->IDBsignal, param);
13248 /* symbol tables */
13249 PL_defstash = hv_dup_inc(proto_perl->Idefstash, param);
13250 PL_curstash = hv_dup_inc(proto_perl->Icurstash, param);
13251 PL_debstash = hv_dup(proto_perl->Idebstash, param);
13252 PL_globalstash = hv_dup(proto_perl->Iglobalstash, param);
13253 PL_curstname = sv_dup_inc(proto_perl->Icurstname, param);
13255 PL_beginav = av_dup_inc(proto_perl->Ibeginav, param);
13256 PL_beginav_save = av_dup_inc(proto_perl->Ibeginav_save, param);
13257 PL_checkav_save = av_dup_inc(proto_perl->Icheckav_save, param);
13258 PL_unitcheckav = av_dup_inc(proto_perl->Iunitcheckav, param);
13259 PL_unitcheckav_save = av_dup_inc(proto_perl->Iunitcheckav_save, param);
13260 PL_endav = av_dup_inc(proto_perl->Iendav, param);
13261 PL_checkav = av_dup_inc(proto_perl->Icheckav, param);
13262 PL_initav = av_dup_inc(proto_perl->Iinitav, param);
13264 PL_isarev = hv_dup_inc(proto_perl->Iisarev, param);
13266 /* subprocess state */
13267 PL_fdpid = av_dup_inc(proto_perl->Ifdpid, param);
13269 if (proto_perl->Iop_mask)
13270 PL_op_mask = SAVEPVN(proto_perl->Iop_mask, PL_maxo);
13273 /* PL_asserting = proto_perl->Iasserting; */
13275 /* current interpreter roots */
13276 PL_main_cv = cv_dup_inc(proto_perl->Imain_cv, param);
13278 PL_main_root = OpREFCNT_inc(proto_perl->Imain_root);
13281 /* runtime control stuff */
13282 PL_curcopdb = (COP*)any_dup(proto_perl->Icurcopdb, proto_perl);
13284 PL_preambleav = av_dup_inc(proto_perl->Ipreambleav, param);
13286 PL_ors_sv = sv_dup_inc(proto_perl->Iors_sv, param);
13288 /* interpreter atexit processing */
13289 PL_exitlistlen = proto_perl->Iexitlistlen;
13290 if (PL_exitlistlen) {
13291 Newx(PL_exitlist, PL_exitlistlen, PerlExitListEntry);
13292 Copy(proto_perl->Iexitlist, PL_exitlist, PL_exitlistlen, PerlExitListEntry);
13295 PL_exitlist = (PerlExitListEntry*)NULL;
13297 PL_my_cxt_size = proto_perl->Imy_cxt_size;
13298 if (PL_my_cxt_size) {
13299 Newx(PL_my_cxt_list, PL_my_cxt_size, void *);
13300 Copy(proto_perl->Imy_cxt_list, PL_my_cxt_list, PL_my_cxt_size, void *);
13301 #ifdef PERL_GLOBAL_STRUCT_PRIVATE
13302 Newx(PL_my_cxt_keys, PL_my_cxt_size, const char *);
13303 Copy(proto_perl->Imy_cxt_keys, PL_my_cxt_keys, PL_my_cxt_size, char *);
13307 PL_my_cxt_list = (void**)NULL;
13308 #ifdef PERL_GLOBAL_STRUCT_PRIVATE
13309 PL_my_cxt_keys = (const char**)NULL;
13312 PL_modglobal = hv_dup_inc(proto_perl->Imodglobal, param);
13313 PL_custom_op_names = hv_dup_inc(proto_perl->Icustom_op_names,param);
13314 PL_custom_op_descs = hv_dup_inc(proto_perl->Icustom_op_descs,param);
13315 PL_custom_ops = hv_dup_inc(proto_perl->Icustom_ops, param);
13317 PL_compcv = cv_dup(proto_perl->Icompcv, param);
13319 PAD_CLONE_VARS(proto_perl, param);
13321 #ifdef HAVE_INTERP_INTERN
13322 sys_intern_dup(&proto_perl->Isys_intern, &PL_sys_intern);
13325 PL_DBcv = cv_dup(proto_perl->IDBcv, param);
13327 #ifdef PERL_USES_PL_PIDSTATUS
13328 PL_pidstatus = newHV(); /* XXX flag for cloning? */
13330 PL_osname = SAVEPV(proto_perl->Iosname);
13331 PL_parser = parser_dup(proto_perl->Iparser, param);
13333 /* XXX this only works if the saved cop has already been cloned */
13334 if (proto_perl->Iparser) {
13335 PL_parser->saved_curcop = (COP*)any_dup(
13336 proto_perl->Iparser->saved_curcop,
13340 PL_subname = sv_dup_inc(proto_perl->Isubname, param);
13342 #ifdef USE_LOCALE_COLLATE
13343 PL_collation_name = SAVEPV(proto_perl->Icollation_name);
13344 #endif /* USE_LOCALE_COLLATE */
13346 #ifdef USE_LOCALE_NUMERIC
13347 PL_numeric_name = SAVEPV(proto_perl->Inumeric_name);
13348 PL_numeric_radix_sv = sv_dup_inc(proto_perl->Inumeric_radix_sv, param);
13349 #endif /* !USE_LOCALE_NUMERIC */
13351 /* utf8 character classes */
13352 PL_utf8_alnum = sv_dup_inc(proto_perl->Iutf8_alnum, param);
13353 PL_utf8_alpha = sv_dup_inc(proto_perl->Iutf8_alpha, param);
13354 PL_utf8_space = sv_dup_inc(proto_perl->Iutf8_space, param);
13355 PL_utf8_graph = sv_dup_inc(proto_perl->Iutf8_graph, param);
13356 PL_utf8_digit = sv_dup_inc(proto_perl->Iutf8_digit, param);
13357 PL_utf8_upper = sv_dup_inc(proto_perl->Iutf8_upper, param);
13358 PL_utf8_lower = sv_dup_inc(proto_perl->Iutf8_lower, param);
13359 PL_utf8_print = sv_dup_inc(proto_perl->Iutf8_print, param);
13360 PL_utf8_punct = sv_dup_inc(proto_perl->Iutf8_punct, param);
13361 PL_utf8_xdigit = sv_dup_inc(proto_perl->Iutf8_xdigit, param);
13362 PL_utf8_mark = sv_dup_inc(proto_perl->Iutf8_mark, param);
13363 PL_utf8_X_begin = sv_dup_inc(proto_perl->Iutf8_X_begin, param);
13364 PL_utf8_X_extend = sv_dup_inc(proto_perl->Iutf8_X_extend, param);
13365 PL_utf8_X_prepend = sv_dup_inc(proto_perl->Iutf8_X_prepend, param);
13366 PL_utf8_X_non_hangul = sv_dup_inc(proto_perl->Iutf8_X_non_hangul, param);
13367 PL_utf8_X_L = sv_dup_inc(proto_perl->Iutf8_X_L, param);
13368 PL_utf8_X_LV = sv_dup_inc(proto_perl->Iutf8_X_LV, param);
13369 PL_utf8_X_LVT = sv_dup_inc(proto_perl->Iutf8_X_LVT, param);
13370 PL_utf8_X_T = sv_dup_inc(proto_perl->Iutf8_X_T, param);
13371 PL_utf8_X_V = sv_dup_inc(proto_perl->Iutf8_X_V, param);
13372 PL_utf8_X_LV_LVT_V = sv_dup_inc(proto_perl->Iutf8_X_LV_LVT_V, param);
13373 PL_utf8_toupper = sv_dup_inc(proto_perl->Iutf8_toupper, param);
13374 PL_utf8_totitle = sv_dup_inc(proto_perl->Iutf8_totitle, param);
13375 PL_utf8_tolower = sv_dup_inc(proto_perl->Iutf8_tolower, param);
13376 PL_utf8_tofold = sv_dup_inc(proto_perl->Iutf8_tofold, param);
13377 PL_utf8_idstart = sv_dup_inc(proto_perl->Iutf8_idstart, param);
13378 PL_utf8_xidstart = sv_dup_inc(proto_perl->Iutf8_xidstart, param);
13379 PL_utf8_perl_idstart = sv_dup_inc(proto_perl->Iutf8_perl_idstart, param);
13380 PL_utf8_idcont = sv_dup_inc(proto_perl->Iutf8_idcont, param);
13381 PL_utf8_xidcont = sv_dup_inc(proto_perl->Iutf8_xidcont, param);
13382 PL_utf8_foldable = sv_dup_inc(proto_perl->Iutf8_foldable, param);
13385 if (proto_perl->Ipsig_pend) {
13386 Newxz(PL_psig_pend, SIG_SIZE, int);
13389 PL_psig_pend = (int*)NULL;
13392 if (proto_perl->Ipsig_name) {
13393 Newx(PL_psig_name, 2 * SIG_SIZE, SV*);
13394 sv_dup_inc_multiple(proto_perl->Ipsig_name, PL_psig_name, 2 * SIG_SIZE,
13396 PL_psig_ptr = PL_psig_name + SIG_SIZE;
13399 PL_psig_ptr = (SV**)NULL;
13400 PL_psig_name = (SV**)NULL;
13403 if (flags & CLONEf_COPY_STACKS) {
13404 Newx(PL_tmps_stack, PL_tmps_max, SV*);
13405 sv_dup_inc_multiple(proto_perl->Itmps_stack, PL_tmps_stack,
13406 PL_tmps_ix+1, param);
13408 /* next PUSHMARK() sets *(PL_markstack_ptr+1) */
13409 i = proto_perl->Imarkstack_max - proto_perl->Imarkstack;
13410 Newxz(PL_markstack, i, I32);
13411 PL_markstack_max = PL_markstack + (proto_perl->Imarkstack_max
13412 - proto_perl->Imarkstack);
13413 PL_markstack_ptr = PL_markstack + (proto_perl->Imarkstack_ptr
13414 - proto_perl->Imarkstack);
13415 Copy(proto_perl->Imarkstack, PL_markstack,
13416 PL_markstack_ptr - PL_markstack + 1, I32);
13418 /* next push_scope()/ENTER sets PL_scopestack[PL_scopestack_ix]
13419 * NOTE: unlike the others! */
13420 Newxz(PL_scopestack, PL_scopestack_max, I32);
13421 Copy(proto_perl->Iscopestack, PL_scopestack, PL_scopestack_ix, I32);
13424 Newxz(PL_scopestack_name, PL_scopestack_max, const char *);
13425 Copy(proto_perl->Iscopestack_name, PL_scopestack_name, PL_scopestack_ix, const char *);
13427 /* NOTE: si_dup() looks at PL_markstack */
13428 PL_curstackinfo = si_dup(proto_perl->Icurstackinfo, param);
13430 /* PL_curstack = PL_curstackinfo->si_stack; */
13431 PL_curstack = av_dup(proto_perl->Icurstack, param);
13432 PL_mainstack = av_dup(proto_perl->Imainstack, param);
13434 /* next PUSHs() etc. set *(PL_stack_sp+1) */
13435 PL_stack_base = AvARRAY(PL_curstack);
13436 PL_stack_sp = PL_stack_base + (proto_perl->Istack_sp
13437 - proto_perl->Istack_base);
13438 PL_stack_max = PL_stack_base + AvMAX(PL_curstack);
13440 /*Newxz(PL_savestack, PL_savestack_max, ANY);*/
13441 PL_savestack = ss_dup(proto_perl, param);
13445 ENTER; /* perl_destruct() wants to LEAVE; */
13448 PL_statgv = gv_dup(proto_perl->Istatgv, param);
13449 PL_statname = sv_dup_inc(proto_perl->Istatname, param);
13451 PL_rs = sv_dup_inc(proto_perl->Irs, param);
13452 PL_last_in_gv = gv_dup(proto_perl->Ilast_in_gv, param);
13453 PL_defoutgv = gv_dup_inc(proto_perl->Idefoutgv, param);
13454 PL_toptarget = sv_dup_inc(proto_perl->Itoptarget, param);
13455 PL_bodytarget = sv_dup_inc(proto_perl->Ibodytarget, param);
13456 PL_formtarget = sv_dup(proto_perl->Iformtarget, param);
13458 PL_errors = sv_dup_inc(proto_perl->Ierrors, param);
13460 PL_sortcop = (OP*)any_dup(proto_perl->Isortcop, proto_perl);
13461 PL_sortstash = hv_dup(proto_perl->Isortstash, param);
13462 PL_firstgv = gv_dup(proto_perl->Ifirstgv, param);
13463 PL_secondgv = gv_dup(proto_perl->Isecondgv, param);
13465 PL_stashcache = newHV();
13467 PL_watchaddr = (char **) ptr_table_fetch(PL_ptr_table,
13468 proto_perl->Iwatchaddr);
13469 PL_watchok = PL_watchaddr ? * PL_watchaddr : NULL;
13470 if (PL_debug && PL_watchaddr) {
13471 PerlIO_printf(Perl_debug_log,
13472 "WATCHING: %"UVxf" cloned as %"UVxf" with value %"UVxf"\n",
13473 PTR2UV(proto_perl->Iwatchaddr), PTR2UV(PL_watchaddr),
13474 PTR2UV(PL_watchok));
13477 PL_registered_mros = hv_dup_inc(proto_perl->Iregistered_mros, param);
13478 PL_blockhooks = av_dup_inc(proto_perl->Iblockhooks, param);
13479 PL_utf8_foldclosures = hv_dup_inc(proto_perl->Iutf8_foldclosures, param);
13481 /* Call the ->CLONE method, if it exists, for each of the stashes
13482 identified by sv_dup() above.
13484 while(av_len(param->stashes) != -1) {
13485 HV* const stash = MUTABLE_HV(av_shift(param->stashes));
13486 GV* const cloner = gv_fetchmethod_autoload(stash, "CLONE", 0);
13487 if (cloner && GvCV(cloner)) {
13492 mXPUSHs(newSVhek(HvNAME_HEK(stash)));
13494 call_sv(MUTABLE_SV(GvCV(cloner)), G_DISCARD);
13500 if (!(flags & CLONEf_KEEP_PTR_TABLE)) {
13501 ptr_table_free(PL_ptr_table);
13502 PL_ptr_table = NULL;
13505 if (!(flags & CLONEf_COPY_STACKS)) {
13506 unreferenced_to_tmp_stack(param->unreferenced);
13509 SvREFCNT_dec(param->stashes);
13511 /* orphaned? eg threads->new inside BEGIN or use */
13512 if (PL_compcv && ! SvREFCNT(PL_compcv)) {
13513 SvREFCNT_inc_simple_void(PL_compcv);
13514 SAVEFREESV(PL_compcv);
13521 S_unreferenced_to_tmp_stack(pTHX_ AV *const unreferenced)
13523 PERL_ARGS_ASSERT_UNREFERENCED_TO_TMP_STACK;
13525 if (AvFILLp(unreferenced) > -1) {
13526 SV **svp = AvARRAY(unreferenced);
13527 SV **const last = svp + AvFILLp(unreferenced);
13531 if (SvREFCNT(*svp) == 1)
13533 } while (++svp <= last);
13535 EXTEND_MORTAL(count);
13536 svp = AvARRAY(unreferenced);
13539 if (SvREFCNT(*svp) == 1) {
13540 /* Our reference is the only one to this SV. This means that
13541 in this thread, the scalar effectively has a 0 reference.
13542 That doesn't work (cleanup never happens), so donate our
13543 reference to it onto the save stack. */
13544 PL_tmps_stack[++PL_tmps_ix] = *svp;
13546 /* As an optimisation, because we are already walking the
13547 entire array, instead of above doing either
13548 SvREFCNT_inc(*svp) or *svp = &PL_sv_undef, we can instead
13549 release our reference to the scalar, so that at the end of
13550 the array owns zero references to the scalars it happens to
13551 point to. We are effectively converting the array from
13552 AvREAL() on to AvREAL() off. This saves the av_clear()
13553 (triggered by the SvREFCNT_dec(unreferenced) below) from
13554 walking the array a second time. */
13555 SvREFCNT_dec(*svp);
13558 } while (++svp <= last);
13559 AvREAL_off(unreferenced);
13561 SvREFCNT_dec(unreferenced);
13565 Perl_clone_params_del(CLONE_PARAMS *param)
13567 /* This seemingly funky ordering keeps the build with PERL_GLOBAL_STRUCT
13569 PerlInterpreter *const to = param->new_perl;
13571 PerlInterpreter *const was = PERL_GET_THX;
13573 PERL_ARGS_ASSERT_CLONE_PARAMS_DEL;
13579 SvREFCNT_dec(param->stashes);
13580 if (param->unreferenced)
13581 unreferenced_to_tmp_stack(param->unreferenced);
13591 Perl_clone_params_new(PerlInterpreter *const from, PerlInterpreter *const to)
13594 /* Need to play this game, as newAV() can call safesysmalloc(), and that
13595 does a dTHX; to get the context from thread local storage.
13596 FIXME - under PERL_CORE Newx(), Safefree() and friends should expand to
13597 a version that passes in my_perl. */
13598 PerlInterpreter *const was = PERL_GET_THX;
13599 CLONE_PARAMS *param;
13601 PERL_ARGS_ASSERT_CLONE_PARAMS_NEW;
13607 /* Given that we've set the context, we can do this unshared. */
13608 Newx(param, 1, CLONE_PARAMS);
13611 param->proto_perl = from;
13612 param->new_perl = to;
13613 param->stashes = (AV *)Perl_newSV_type(to, SVt_PVAV);
13614 AvREAL_off(param->stashes);
13615 param->unreferenced = (AV *)Perl_newSV_type(to, SVt_PVAV);
13623 #endif /* USE_ITHREADS */
13626 =head1 Unicode Support
13628 =for apidoc sv_recode_to_utf8
13630 The encoding is assumed to be an Encode object, on entry the PV
13631 of the sv is assumed to be octets in that encoding, and the sv
13632 will be converted into Unicode (and UTF-8).
13634 If the sv already is UTF-8 (or if it is not POK), or if the encoding
13635 is not a reference, nothing is done to the sv. If the encoding is not
13636 an C<Encode::XS> Encoding object, bad things will happen.
13637 (See F<lib/encoding.pm> and L<Encode>.)
13639 The PV of the sv is returned.
13644 Perl_sv_recode_to_utf8(pTHX_ SV *sv, SV *encoding)
13648 PERL_ARGS_ASSERT_SV_RECODE_TO_UTF8;
13650 if (SvPOK(sv) && !SvUTF8(sv) && !IN_BYTES && SvROK(encoding)) {
13664 Passing sv_yes is wrong - it needs to be or'ed set of constants
13665 for Encode::XS, while UTf-8 decode (currently) assumes a true value means
13666 remove converted chars from source.
13668 Both will default the value - let them.
13670 XPUSHs(&PL_sv_yes);
13673 call_method("decode", G_SCALAR);
13677 s = SvPV_const(uni, len);
13678 if (s != SvPVX_const(sv)) {
13679 SvGROW(sv, len + 1);
13680 Move(s, SvPVX(sv), len + 1, char);
13681 SvCUR_set(sv, len);
13685 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
13686 /* clear pos and any utf8 cache */
13687 MAGIC * mg = mg_find(sv, PERL_MAGIC_regex_global);
13690 if ((mg = mg_find(sv, PERL_MAGIC_utf8)))
13691 magic_setutf8(sv,mg); /* clear UTF8 cache */
13696 return SvPOKp(sv) ? SvPVX(sv) : NULL;
13700 =for apidoc sv_cat_decode
13702 The encoding is assumed to be an Encode object, the PV of the ssv is
13703 assumed to be octets in that encoding and decoding the input starts
13704 from the position which (PV + *offset) pointed to. The dsv will be
13705 concatenated the decoded UTF-8 string from ssv. Decoding will terminate
13706 when the string tstr appears in decoding output or the input ends on
13707 the PV of the ssv. The value which the offset points will be modified
13708 to the last input position on the ssv.
13710 Returns TRUE if the terminator was found, else returns FALSE.
13715 Perl_sv_cat_decode(pTHX_ SV *dsv, SV *encoding,
13716 SV *ssv, int *offset, char *tstr, int tlen)
13721 PERL_ARGS_ASSERT_SV_CAT_DECODE;
13723 if (SvPOK(ssv) && SvPOK(dsv) && SvROK(encoding) && offset) {
13734 offsv = newSViv(*offset);
13736 mXPUSHp(tstr, tlen);
13738 call_method("cat_decode", G_SCALAR);
13740 ret = SvTRUE(TOPs);
13741 *offset = SvIV(offsv);
13747 Perl_croak(aTHX_ "Invalid argument to sv_cat_decode");
13752 /* ---------------------------------------------------------------------
13754 * support functions for report_uninit()
13757 /* the maxiumum size of array or hash where we will scan looking
13758 * for the undefined element that triggered the warning */
13760 #define FUV_MAX_SEARCH_SIZE 1000
13762 /* Look for an entry in the hash whose value has the same SV as val;
13763 * If so, return a mortal copy of the key. */
13766 S_find_hash_subscript(pTHX_ const HV *const hv, const SV *const val)
13769 register HE **array;
13772 PERL_ARGS_ASSERT_FIND_HASH_SUBSCRIPT;
13774 if (!hv || SvMAGICAL(hv) || !HvARRAY(hv) ||
13775 (HvTOTALKEYS(hv) > FUV_MAX_SEARCH_SIZE))
13778 array = HvARRAY(hv);
13780 for (i=HvMAX(hv); i>0; i--) {
13781 register HE *entry;
13782 for (entry = array[i]; entry; entry = HeNEXT(entry)) {
13783 if (HeVAL(entry) != val)
13785 if ( HeVAL(entry) == &PL_sv_undef ||
13786 HeVAL(entry) == &PL_sv_placeholder)
13790 if (HeKLEN(entry) == HEf_SVKEY)
13791 return sv_mortalcopy(HeKEY_sv(entry));
13792 return sv_2mortal(newSVhek(HeKEY_hek(entry)));
13798 /* Look for an entry in the array whose value has the same SV as val;
13799 * If so, return the index, otherwise return -1. */
13802 S_find_array_subscript(pTHX_ const AV *const av, const SV *const val)
13806 PERL_ARGS_ASSERT_FIND_ARRAY_SUBSCRIPT;
13808 if (!av || SvMAGICAL(av) || !AvARRAY(av) ||
13809 (AvFILLp(av) > FUV_MAX_SEARCH_SIZE))
13812 if (val != &PL_sv_undef) {
13813 SV ** const svp = AvARRAY(av);
13816 for (i=AvFILLp(av); i>=0; i--)
13823 /* S_varname(): return the name of a variable, optionally with a subscript.
13824 * If gv is non-zero, use the name of that global, along with gvtype (one
13825 * of "$", "@", "%"); otherwise use the name of the lexical at pad offset
13826 * targ. Depending on the value of the subscript_type flag, return:
13829 #define FUV_SUBSCRIPT_NONE 1 /* "@foo" */
13830 #define FUV_SUBSCRIPT_ARRAY 2 /* "$foo[aindex]" */
13831 #define FUV_SUBSCRIPT_HASH 3 /* "$foo{keyname}" */
13832 #define FUV_SUBSCRIPT_WITHIN 4 /* "within @foo" */
13835 Perl_varname(pTHX_ const GV *const gv, const char gvtype, PADOFFSET targ,
13836 const SV *const keyname, I32 aindex, int subscript_type)
13839 SV * const name = sv_newmortal();
13842 buffer[0] = gvtype;
13845 /* as gv_fullname4(), but add literal '^' for $^FOO names */
13847 gv_fullname4(name, gv, buffer, 0);
13849 if ((unsigned int)SvPVX(name)[1] <= 26) {
13851 buffer[1] = SvPVX(name)[1] + 'A' - 1;
13853 /* Swap the 1 unprintable control character for the 2 byte pretty
13854 version - ie substr($name, 1, 1) = $buffer; */
13855 sv_insert(name, 1, 1, buffer, 2);
13859 CV * const cv = find_runcv(NULL);
13863 if (!cv || !CvPADLIST(cv))
13865 av = MUTABLE_AV((*av_fetch(CvPADLIST(cv), 0, FALSE)));
13866 sv = *av_fetch(av, targ, FALSE);
13867 sv_setsv(name, sv);
13870 if (subscript_type == FUV_SUBSCRIPT_HASH) {
13871 SV * const sv = newSV(0);
13872 *SvPVX(name) = '$';
13873 Perl_sv_catpvf(aTHX_ name, "{%s}",
13874 pv_display(sv,SvPVX_const(keyname), SvCUR(keyname), 0, 32));
13877 else if (subscript_type == FUV_SUBSCRIPT_ARRAY) {
13878 *SvPVX(name) = '$';
13879 Perl_sv_catpvf(aTHX_ name, "[%"IVdf"]", (IV)aindex);
13881 else if (subscript_type == FUV_SUBSCRIPT_WITHIN) {
13882 /* We know that name has no magic, so can use 0 instead of SV_GMAGIC */
13883 Perl_sv_insert_flags(aTHX_ name, 0, 0, STR_WITH_LEN("within "), 0);
13891 =for apidoc find_uninit_var
13893 Find the name of the undefined variable (if any) that caused the operator
13894 to issue a "Use of uninitialized value" warning.
13895 If match is true, only return a name if its value matches uninit_sv.
13896 So roughly speaking, if a unary operator (such as OP_COS) generates a
13897 warning, then following the direct child of the op may yield an
13898 OP_PADSV or OP_GV that gives the name of the undefined variable. On the
13899 other hand, with OP_ADD there are two branches to follow, so we only print
13900 the variable name if we get an exact match.
13902 The name is returned as a mortal SV.
13904 Assumes that PL_op is the op that originally triggered the error, and that
13905 PL_comppad/PL_curpad points to the currently executing pad.
13911 S_find_uninit_var(pTHX_ const OP *const obase, const SV *const uninit_sv,
13917 const OP *o, *o2, *kid;
13919 if (!obase || (match && (!uninit_sv || uninit_sv == &PL_sv_undef ||
13920 uninit_sv == &PL_sv_placeholder)))
13923 switch (obase->op_type) {
13930 const bool pad = (obase->op_type == OP_PADAV || obase->op_type == OP_PADHV);
13931 const bool hash = (obase->op_type == OP_PADHV || obase->op_type == OP_RV2HV);
13934 int subscript_type = FUV_SUBSCRIPT_WITHIN;
13936 if (pad) { /* @lex, %lex */
13937 sv = PAD_SVl(obase->op_targ);
13941 if (cUNOPx(obase)->op_first->op_type == OP_GV) {
13942 /* @global, %global */
13943 gv = cGVOPx_gv(cUNOPx(obase)->op_first);
13946 sv = hash ? MUTABLE_SV(GvHV(gv)): MUTABLE_SV(GvAV(gv));
13948 else if (obase == PL_op) /* @{expr}, %{expr} */
13949 return find_uninit_var(cUNOPx(obase)->op_first,
13951 else /* @{expr}, %{expr} as a sub-expression */
13955 /* attempt to find a match within the aggregate */
13957 keysv = find_hash_subscript((const HV*)sv, uninit_sv);
13959 subscript_type = FUV_SUBSCRIPT_HASH;
13962 index = find_array_subscript((const AV *)sv, uninit_sv);
13964 subscript_type = FUV_SUBSCRIPT_ARRAY;
13967 if (match && subscript_type == FUV_SUBSCRIPT_WITHIN)
13970 return varname(gv, hash ? '%' : '@', obase->op_targ,
13971 keysv, index, subscript_type);
13975 if (cUNOPx(obase)->op_first->op_type == OP_GV) {
13977 gv = cGVOPx_gv(cUNOPx(obase)->op_first);
13978 if (!gv || !GvSTASH(gv))
13980 if (match && (GvSV(gv) != uninit_sv))
13982 return varname(gv, '$', 0, NULL, 0, FUV_SUBSCRIPT_NONE);
13985 return find_uninit_var(cUNOPx(obase)->op_first, uninit_sv, 1);
13988 if (match && PAD_SVl(obase->op_targ) != uninit_sv)
13990 return varname(NULL, '$', obase->op_targ,
13991 NULL, 0, FUV_SUBSCRIPT_NONE);
13994 gv = cGVOPx_gv(obase);
13995 if (!gv || (match && GvSV(gv) != uninit_sv) || !GvSTASH(gv))
13997 return varname(gv, '$', 0, NULL, 0, FUV_SUBSCRIPT_NONE);
13999 case OP_AELEMFAST_LEX:
14002 AV *av = MUTABLE_AV(PAD_SV(obase->op_targ));
14003 if (!av || SvRMAGICAL(av))
14005 svp = av_fetch(av, (I32)obase->op_private, FALSE);
14006 if (!svp || *svp != uninit_sv)
14009 return varname(NULL, '$', obase->op_targ,
14010 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
14013 gv = cGVOPx_gv(obase);
14018 AV *const av = GvAV(gv);
14019 if (!av || SvRMAGICAL(av))
14021 svp = av_fetch(av, (I32)obase->op_private, FALSE);
14022 if (!svp || *svp != uninit_sv)
14025 return varname(gv, '$', 0,
14026 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
14031 o = cUNOPx(obase)->op_first;
14032 if (!o || o->op_type != OP_NULL ||
14033 ! (o->op_targ == OP_AELEM || o->op_targ == OP_HELEM))
14035 return find_uninit_var(cBINOPo->op_last, uninit_sv, match);
14040 bool negate = FALSE;
14042 if (PL_op == obase)
14043 /* $a[uninit_expr] or $h{uninit_expr} */
14044 return find_uninit_var(cBINOPx(obase)->op_last, uninit_sv, match);
14047 o = cBINOPx(obase)->op_first;
14048 kid = cBINOPx(obase)->op_last;
14050 /* get the av or hv, and optionally the gv */
14052 if (o->op_type == OP_PADAV || o->op_type == OP_PADHV) {
14053 sv = PAD_SV(o->op_targ);
14055 else if ((o->op_type == OP_RV2AV || o->op_type == OP_RV2HV)
14056 && cUNOPo->op_first->op_type == OP_GV)
14058 gv = cGVOPx_gv(cUNOPo->op_first);
14062 == OP_RV2HV ? MUTABLE_SV(GvHV(gv)) : MUTABLE_SV(GvAV(gv));
14067 if (kid && kid->op_type == OP_NEGATE) {
14069 kid = cUNOPx(kid)->op_first;
14072 if (kid && kid->op_type == OP_CONST && SvOK(cSVOPx_sv(kid))) {
14073 /* index is constant */
14076 kidsv = sv_2mortal(newSVpvs("-"));
14077 sv_catsv(kidsv, cSVOPx_sv(kid));
14080 kidsv = cSVOPx_sv(kid);
14084 if (obase->op_type == OP_HELEM) {
14085 HE* he = hv_fetch_ent(MUTABLE_HV(sv), kidsv, 0, 0);
14086 if (!he || HeVAL(he) != uninit_sv)
14090 SV * const * const svp = av_fetch(MUTABLE_AV(sv),
14091 negate ? - SvIV(cSVOPx_sv(kid)) : SvIV(cSVOPx_sv(kid)),
14093 if (!svp || *svp != uninit_sv)
14097 if (obase->op_type == OP_HELEM)
14098 return varname(gv, '%', o->op_targ,
14099 kidsv, 0, FUV_SUBSCRIPT_HASH);
14101 return varname(gv, '@', o->op_targ, NULL,
14102 negate ? - SvIV(cSVOPx_sv(kid)) : SvIV(cSVOPx_sv(kid)),
14103 FUV_SUBSCRIPT_ARRAY);
14106 /* index is an expression;
14107 * attempt to find a match within the aggregate */
14108 if (obase->op_type == OP_HELEM) {
14109 SV * const keysv = find_hash_subscript((const HV*)sv, uninit_sv);
14111 return varname(gv, '%', o->op_targ,
14112 keysv, 0, FUV_SUBSCRIPT_HASH);
14116 = find_array_subscript((const AV *)sv, uninit_sv);
14118 return varname(gv, '@', o->op_targ,
14119 NULL, index, FUV_SUBSCRIPT_ARRAY);
14124 (o->op_type == OP_PADAV || o->op_type == OP_RV2AV)
14126 o->op_targ, NULL, 0, FUV_SUBSCRIPT_WITHIN);
14132 /* only examine RHS */
14133 return find_uninit_var(cBINOPx(obase)->op_first, uninit_sv, match);
14136 o = cUNOPx(obase)->op_first;
14137 if (o->op_type == OP_PUSHMARK)
14140 if (!o->op_sibling) {
14141 /* one-arg version of open is highly magical */
14143 if (o->op_type == OP_GV) { /* open FOO; */
14145 if (match && GvSV(gv) != uninit_sv)
14147 return varname(gv, '$', 0,
14148 NULL, 0, FUV_SUBSCRIPT_NONE);
14150 /* other possibilities not handled are:
14151 * open $x; or open my $x; should return '${*$x}'
14152 * open expr; should return '$'.expr ideally
14158 /* ops where $_ may be an implicit arg */
14163 if ( !(obase->op_flags & OPf_STACKED)) {
14164 if (uninit_sv == ((obase->op_private & OPpTARGET_MY)
14165 ? PAD_SVl(obase->op_targ)
14168 sv = sv_newmortal();
14169 sv_setpvs(sv, "$_");
14178 match = 1; /* print etc can return undef on defined args */
14179 /* skip filehandle as it can't produce 'undef' warning */
14180 o = cUNOPx(obase)->op_first;
14181 if ((obase->op_flags & OPf_STACKED) && o->op_type == OP_PUSHMARK)
14182 o = o->op_sibling->op_sibling;
14186 case OP_ENTEREVAL: /* could be eval $undef or $x='$undef'; eval $x */
14187 case OP_CUSTOM: /* XS or custom code could trigger random warnings */
14189 /* the following ops are capable of returning PL_sv_undef even for
14190 * defined arg(s) */
14209 case OP_GETPEERNAME:
14257 case OP_SMARTMATCH:
14266 /* XXX tmp hack: these two may call an XS sub, and currently
14267 XS subs don't have a SUB entry on the context stack, so CV and
14268 pad determination goes wrong, and BAD things happen. So, just
14269 don't try to determine the value under those circumstances.
14270 Need a better fix at dome point. DAPM 11/2007 */
14276 GV * const gv = gv_fetchpvs(".", GV_NOTQUAL, SVt_PV);
14277 if (gv && GvSV(gv) == uninit_sv)
14278 return newSVpvs_flags("$.", SVs_TEMP);
14283 /* def-ness of rval pos() is independent of the def-ness of its arg */
14284 if ( !(obase->op_flags & OPf_MOD))
14289 if (SvROK(PL_rs) && uninit_sv == SvRV(PL_rs))
14290 return newSVpvs_flags("${$/}", SVs_TEMP);
14295 if (!(obase->op_flags & OPf_KIDS))
14297 o = cUNOPx(obase)->op_first;
14303 /* This loop checks all the kid ops, skipping any that cannot pos-
14304 * sibly be responsible for the uninitialized value; i.e., defined
14305 * constants and ops that return nothing. If there is only one op
14306 * left that is not skipped, then we *know* it is responsible for
14307 * the uninitialized value. If there is more than one op left, we
14308 * have to look for an exact match in the while() loop below.
14311 for (kid=o; kid; kid = kid->op_sibling) {
14313 const OPCODE type = kid->op_type;
14314 if ( (type == OP_CONST && SvOK(cSVOPx_sv(kid)))
14315 || (type == OP_NULL && ! (kid->op_flags & OPf_KIDS))
14316 || (type == OP_PUSHMARK)
14320 if (o2) { /* more than one found */
14327 return find_uninit_var(o2, uninit_sv, match);
14329 /* scan all args */
14331 sv = find_uninit_var(o, uninit_sv, 1);
14343 =for apidoc report_uninit
14345 Print appropriate "Use of uninitialized variable" warning.
14351 Perl_report_uninit(pTHX_ const SV *uninit_sv)
14355 SV* varname = NULL;
14356 if (uninit_sv && PL_curpad) {
14357 varname = find_uninit_var(PL_op, uninit_sv,0);
14359 sv_insert(varname, 0, 0, " ", 1);
14361 /* diag_listed_as: Use of uninitialized value%s */
14362 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit_sv,
14363 SVfARG(varname ? varname : &PL_sv_no),
14364 " in ", OP_DESC(PL_op));
14367 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
14373 * c-indentation-style: bsd
14374 * c-basic-offset: 4
14375 * indent-tabs-mode: t
14378 * ex: set ts=8 sts=4 sw=4 noet: