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
38 /* Missing proto on LynxOS */
39 char *gconvert(double, int, int, char *);
42 #ifdef PERL_UTF8_CACHE_ASSERT
43 /* if adding more checks watch out for the following tests:
44 * t/op/index.t t/op/length.t t/op/pat.t t/op/substr.t
45 * lib/utf8.t lib/Unicode/Collate/t/index.t
48 # define ASSERT_UTF8_CACHE(cache) \
49 STMT_START { if (cache) { assert((cache)[0] <= (cache)[1]); \
50 assert((cache)[2] <= (cache)[3]); \
51 assert((cache)[3] <= (cache)[1]);} \
54 # define ASSERT_UTF8_CACHE(cache) NOOP
57 #ifdef PERL_OLD_COPY_ON_WRITE
58 #define SV_COW_NEXT_SV(sv) INT2PTR(SV *,SvUVX(sv))
59 #define SV_COW_NEXT_SV_SET(current,next) SvUV_set(current, PTR2UV(next))
60 /* This is a pessimistic view. Scalar must be purely a read-write PV to copy-
64 /* ============================================================================
66 =head1 Allocation and deallocation of SVs.
68 An SV (or AV, HV, etc.) is allocated in two parts: the head (struct
69 sv, av, hv...) contains type and reference count information, and for
70 many types, a pointer to the body (struct xrv, xpv, xpviv...), which
71 contains fields specific to each type. Some types store all they need
72 in the head, so don't have a body.
74 In all but the most memory-paranoid configuations (ex: PURIFY), heads
75 and bodies are allocated out of arenas, which by default are
76 approximately 4K chunks of memory parcelled up into N heads or bodies.
77 Sv-bodies are allocated by their sv-type, guaranteeing size
78 consistency needed to allocate safely from arrays.
80 For SV-heads, the first slot in each arena is reserved, and holds a
81 link to the next arena, some flags, and a note of the number of slots.
82 Snaked through each arena chain is a linked list of free items; when
83 this becomes empty, an extra arena is allocated and divided up into N
84 items which are threaded into the free list.
86 SV-bodies are similar, but they use arena-sets by default, which
87 separate the link and info from the arena itself, and reclaim the 1st
88 slot in the arena. SV-bodies are further described later.
90 The following global variables are associated with arenas:
92 PL_sv_arenaroot pointer to list of SV arenas
93 PL_sv_root pointer to list of free SV structures
95 PL_body_arenas head of linked-list of body arenas
96 PL_body_roots[] array of pointers to list of free bodies of svtype
97 arrays are indexed by the svtype needed
99 A few special SV heads are not allocated from an arena, but are
100 instead directly created in the interpreter structure, eg PL_sv_undef.
101 The size of arenas can be changed from the default by setting
102 PERL_ARENA_SIZE appropriately at compile time.
104 The SV arena serves the secondary purpose of allowing still-live SVs
105 to be located and destroyed during final cleanup.
107 At the lowest level, the macros new_SV() and del_SV() grab and free
108 an SV head. (If debugging with -DD, del_SV() calls the function S_del_sv()
109 to return the SV to the free list with error checking.) new_SV() calls
110 more_sv() / sv_add_arena() to add an extra arena if the free list is empty.
111 SVs in the free list have their SvTYPE field set to all ones.
113 At the time of very final cleanup, sv_free_arenas() is called from
114 perl_destruct() to physically free all the arenas allocated since the
115 start of the interpreter.
117 The function visit() scans the SV arenas list, and calls a specified
118 function for each SV it finds which is still live - ie which has an SvTYPE
119 other than all 1's, and a non-zero SvREFCNT. visit() is used by the
120 following functions (specified as [function that calls visit()] / [function
121 called by visit() for each SV]):
123 sv_report_used() / do_report_used()
124 dump all remaining SVs (debugging aid)
126 sv_clean_objs() / do_clean_objs(),do_clean_named_objs(),
127 do_clean_named_io_objs()
128 Attempt to free all objects pointed to by RVs,
129 and, unless DISABLE_DESTRUCTOR_KLUDGE is defined,
130 try to do the same for all objects indirectly
131 referenced by typeglobs too. Called once from
132 perl_destruct(), prior to calling sv_clean_all()
135 sv_clean_all() / do_clean_all()
136 SvREFCNT_dec(sv) each remaining SV, possibly
137 triggering an sv_free(). It also sets the
138 SVf_BREAK flag on the SV to indicate that the
139 refcnt has been artificially lowered, and thus
140 stopping sv_free() from giving spurious warnings
141 about SVs which unexpectedly have a refcnt
142 of zero. called repeatedly from perl_destruct()
143 until there are no SVs left.
145 =head2 Arena allocator API Summary
147 Private API to rest of sv.c
151 new_XPVNV(), del_XPVGV(),
156 sv_report_used(), sv_clean_objs(), sv_clean_all(), sv_free_arenas()
160 * ========================================================================= */
163 * "A time to plant, and a time to uproot what was planted..."
167 # define MEM_LOG_NEW_SV(sv, file, line, func) \
168 Perl_mem_log_new_sv(sv, file, line, func)
169 # define MEM_LOG_DEL_SV(sv, file, line, func) \
170 Perl_mem_log_del_sv(sv, file, line, func)
172 # define MEM_LOG_NEW_SV(sv, file, line, func) NOOP
173 # define MEM_LOG_DEL_SV(sv, file, line, func) NOOP
176 #ifdef DEBUG_LEAKING_SCALARS
177 # define FREE_SV_DEBUG_FILE(sv) Safefree((sv)->sv_debug_file)
178 # define DEBUG_SV_SERIAL(sv) \
179 DEBUG_m(PerlIO_printf(Perl_debug_log, "0x%"UVxf": (%05ld) del_SV\n", \
180 PTR2UV(sv), (long)(sv)->sv_debug_serial))
182 # define FREE_SV_DEBUG_FILE(sv)
183 # define DEBUG_SV_SERIAL(sv) NOOP
187 # define SvARENA_CHAIN(sv) ((sv)->sv_u.svu_rv)
188 # define SvARENA_CHAIN_SET(sv,val) (sv)->sv_u.svu_rv = MUTABLE_SV((val))
189 /* Whilst I'd love to do this, it seems that things like to check on
191 # define POSION_SV_HEAD(sv) PoisonNew(sv, 1, struct STRUCT_SV)
193 # define POSION_SV_HEAD(sv) PoisonNew(&SvANY(sv), 1, void *), \
194 PoisonNew(&SvREFCNT(sv), 1, U32)
196 # define SvARENA_CHAIN(sv) SvANY(sv)
197 # define SvARENA_CHAIN_SET(sv,val) SvANY(sv) = (void *)(val)
198 # define POSION_SV_HEAD(sv)
201 /* Mark an SV head as unused, and add to free list.
203 * If SVf_BREAK is set, skip adding it to the free list, as this SV had
204 * its refcount artificially decremented during global destruction, so
205 * there may be dangling pointers to it. The last thing we want in that
206 * case is for it to be reused. */
208 #define plant_SV(p) \
210 const U32 old_flags = SvFLAGS(p); \
211 MEM_LOG_DEL_SV(p, __FILE__, __LINE__, FUNCTION__); \
212 DEBUG_SV_SERIAL(p); \
213 FREE_SV_DEBUG_FILE(p); \
215 SvFLAGS(p) = SVTYPEMASK; \
216 if (!(old_flags & SVf_BREAK)) { \
217 SvARENA_CHAIN_SET(p, PL_sv_root); \
223 #define uproot_SV(p) \
226 PL_sv_root = MUTABLE_SV(SvARENA_CHAIN(p)); \
231 /* make some more SVs by adding another arena */
238 char *chunk; /* must use New here to match call to */
239 Newx(chunk,PERL_ARENA_SIZE,char); /* Safefree() in sv_free_arenas() */
240 sv_add_arena(chunk, PERL_ARENA_SIZE, 0);
245 /* new_SV(): return a new, empty SV head */
247 #ifdef DEBUG_LEAKING_SCALARS
248 /* provide a real function for a debugger to play with */
250 S_new_SV(pTHX_ const char *file, int line, const char *func)
257 sv = S_more_sv(aTHX);
261 sv->sv_debug_optype = PL_op ? PL_op->op_type : 0;
262 sv->sv_debug_line = (U16) (PL_parser && PL_parser->copline != NOLINE
268 sv->sv_debug_inpad = 0;
269 sv->sv_debug_parent = NULL;
270 sv->sv_debug_file = PL_curcop ? savepv(CopFILE(PL_curcop)): NULL;
272 sv->sv_debug_serial = PL_sv_serial++;
274 MEM_LOG_NEW_SV(sv, file, line, func);
275 DEBUG_m(PerlIO_printf(Perl_debug_log, "0x%"UVxf": (%05ld) new_SV (from %s:%d [%s])\n",
276 PTR2UV(sv), (long)sv->sv_debug_serial, file, line, func));
280 # define new_SV(p) (p)=S_new_SV(aTHX_ __FILE__, __LINE__, FUNCTION__)
288 (p) = S_more_sv(aTHX); \
292 MEM_LOG_NEW_SV(p, __FILE__, __LINE__, FUNCTION__); \
297 /* del_SV(): return an empty SV head to the free list */
310 S_del_sv(pTHX_ SV *p)
314 PERL_ARGS_ASSERT_DEL_SV;
319 for (sva = PL_sv_arenaroot; sva; sva = MUTABLE_SV(SvANY(sva))) {
320 const SV * const sv = sva + 1;
321 const SV * const svend = &sva[SvREFCNT(sva)];
322 if (p >= sv && p < svend) {
328 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL),
329 "Attempt to free non-arena SV: 0x%"UVxf
330 pTHX__FORMAT, PTR2UV(p) pTHX__VALUE);
337 #else /* ! DEBUGGING */
339 #define del_SV(p) plant_SV(p)
341 #endif /* DEBUGGING */
345 =head1 SV Manipulation Functions
347 =for apidoc sv_add_arena
349 Given a chunk of memory, link it to the head of the list of arenas,
350 and split it into a list of free SVs.
356 S_sv_add_arena(pTHX_ char *const ptr, const U32 size, const U32 flags)
359 SV *const sva = MUTABLE_SV(ptr);
363 PERL_ARGS_ASSERT_SV_ADD_ARENA;
365 /* The first SV in an arena isn't an SV. */
366 SvANY(sva) = (void *) PL_sv_arenaroot; /* ptr to next arena */
367 SvREFCNT(sva) = size / sizeof(SV); /* number of SV slots */
368 SvFLAGS(sva) = flags; /* FAKE if not to be freed */
370 PL_sv_arenaroot = sva;
371 PL_sv_root = sva + 1;
373 svend = &sva[SvREFCNT(sva) - 1];
376 SvARENA_CHAIN_SET(sv, (sv + 1));
380 /* Must always set typemask because it's always checked in on cleanup
381 when the arenas are walked looking for objects. */
382 SvFLAGS(sv) = SVTYPEMASK;
385 SvARENA_CHAIN_SET(sv, 0);
389 SvFLAGS(sv) = SVTYPEMASK;
392 /* visit(): call the named function for each non-free SV in the arenas
393 * whose flags field matches the flags/mask args. */
396 S_visit(pTHX_ SVFUNC_t f, const U32 flags, const U32 mask)
402 PERL_ARGS_ASSERT_VISIT;
404 for (sva = PL_sv_arenaroot; sva; sva = MUTABLE_SV(SvANY(sva))) {
405 register const SV * const svend = &sva[SvREFCNT(sva)];
407 for (sv = sva + 1; sv < svend; ++sv) {
408 if (SvTYPE(sv) != SVTYPEMASK
409 && (sv->sv_flags & mask) == flags
422 /* called by sv_report_used() for each live SV */
425 do_report_used(pTHX_ SV *const sv)
427 if (SvTYPE(sv) != SVTYPEMASK) {
428 PerlIO_printf(Perl_debug_log, "****\n");
435 =for apidoc sv_report_used
437 Dump the contents of all SVs not yet freed. (Debugging aid).
443 Perl_sv_report_used(pTHX)
446 visit(do_report_used, 0, 0);
452 /* called by sv_clean_objs() for each live SV */
455 do_clean_objs(pTHX_ SV *const ref)
460 SV * const target = SvRV(ref);
461 if (SvOBJECT(target)) {
462 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning object ref:\n "), sv_dump(ref)));
463 if (SvWEAKREF(ref)) {
464 sv_del_backref(target, ref);
470 SvREFCNT_dec(target);
475 /* XXX Might want to check arrays, etc. */
479 #ifndef DISABLE_DESTRUCTOR_KLUDGE
481 /* clear any slots in a GV which hold objects - except IO;
482 * called by sv_clean_objs() for each live GV */
485 do_clean_named_objs(pTHX_ SV *const sv)
489 assert(SvTYPE(sv) == SVt_PVGV);
490 assert(isGV_with_GP(sv));
494 /* freeing GP entries may indirectly free the current GV;
495 * hold onto it while we mess with the GP slots */
498 if ( ((obj = GvSV(sv) )) && SvOBJECT(obj)) {
499 DEBUG_D((PerlIO_printf(Perl_debug_log,
500 "Cleaning named glob SV object:\n "), sv_dump(obj)));
504 if ( ((obj = MUTABLE_SV(GvAV(sv)) )) && SvOBJECT(obj)) {
505 DEBUG_D((PerlIO_printf(Perl_debug_log,
506 "Cleaning named glob AV object:\n "), sv_dump(obj)));
510 if ( ((obj = MUTABLE_SV(GvHV(sv)) )) && SvOBJECT(obj)) {
511 DEBUG_D((PerlIO_printf(Perl_debug_log,
512 "Cleaning named glob HV object:\n "), sv_dump(obj)));
516 if ( ((obj = MUTABLE_SV(GvCV(sv)) )) && SvOBJECT(obj)) {
517 DEBUG_D((PerlIO_printf(Perl_debug_log,
518 "Cleaning named glob CV object:\n "), sv_dump(obj)));
522 SvREFCNT_dec(sv); /* undo the inc above */
525 /* clear any IO slots in a GV which hold objects (except stderr, defout);
526 * called by sv_clean_objs() for each live GV */
529 do_clean_named_io_objs(pTHX_ SV *const sv)
533 assert(SvTYPE(sv) == SVt_PVGV);
534 assert(isGV_with_GP(sv));
535 if (!GvGP(sv) || sv == (SV*)PL_stderrgv || sv == (SV*)PL_defoutgv)
539 if ( ((obj = MUTABLE_SV(GvIO(sv)) )) && SvOBJECT(obj)) {
540 DEBUG_D((PerlIO_printf(Perl_debug_log,
541 "Cleaning named glob IO object:\n "), sv_dump(obj)));
545 SvREFCNT_dec(sv); /* undo the inc above */
550 =for apidoc sv_clean_objs
552 Attempt to destroy all objects not yet freed
558 Perl_sv_clean_objs(pTHX)
562 PL_in_clean_objs = TRUE;
563 visit(do_clean_objs, SVf_ROK, SVf_ROK);
564 #ifndef DISABLE_DESTRUCTOR_KLUDGE
565 /* Some barnacles may yet remain, clinging to typeglobs.
566 * Run the non-IO destructors first: they may want to output
567 * error messages, close files etc */
568 visit(do_clean_named_objs, SVt_PVGV|SVpgv_GP, SVTYPEMASK|SVp_POK|SVpgv_GP);
569 visit(do_clean_named_io_objs, SVt_PVGV|SVpgv_GP, SVTYPEMASK|SVp_POK|SVpgv_GP);
570 olddef = PL_defoutgv;
571 PL_defoutgv = NULL; /* disable skip of PL_defoutgv */
572 if (olddef && isGV_with_GP(olddef))
573 do_clean_named_io_objs(aTHX_ MUTABLE_SV(olddef));
574 olderr = PL_stderrgv;
575 PL_stderrgv = NULL; /* disable skip of PL_stderrgv */
576 if (olderr && isGV_with_GP(olderr))
577 do_clean_named_io_objs(aTHX_ MUTABLE_SV(olderr));
578 SvREFCNT_dec(olddef);
580 PL_in_clean_objs = FALSE;
583 /* called by sv_clean_all() for each live SV */
586 do_clean_all(pTHX_ SV *const sv)
589 if (sv == (const SV *) PL_fdpid || sv == (const SV *)PL_strtab) {
590 /* don't clean pid table and strtab */
593 DEBUG_D((PerlIO_printf(Perl_debug_log, "Cleaning loops: SV at 0x%"UVxf"\n", PTR2UV(sv)) ));
594 SvFLAGS(sv) |= SVf_BREAK;
599 =for apidoc sv_clean_all
601 Decrement the refcnt of each remaining SV, possibly triggering a
602 cleanup. This function may have to be called multiple times to free
603 SVs which are in complex self-referential hierarchies.
609 Perl_sv_clean_all(pTHX)
613 PL_in_clean_all = TRUE;
614 cleaned = visit(do_clean_all, 0,0);
619 ARENASETS: a meta-arena implementation which separates arena-info
620 into struct arena_set, which contains an array of struct
621 arena_descs, each holding info for a single arena. By separating
622 the meta-info from the arena, we recover the 1st slot, formerly
623 borrowed for list management. The arena_set is about the size of an
624 arena, avoiding the needless malloc overhead of a naive linked-list.
626 The cost is 1 arena-set malloc per ~320 arena-mallocs, + the unused
627 memory in the last arena-set (1/2 on average). In trade, we get
628 back the 1st slot in each arena (ie 1.7% of a CV-arena, less for
629 smaller types). The recovery of the wasted space allows use of
630 small arenas for large, rare body types, by changing array* fields
631 in body_details_by_type[] below.
634 char *arena; /* the raw storage, allocated aligned */
635 size_t size; /* its size ~4k typ */
636 svtype utype; /* bodytype stored in arena */
641 /* Get the maximum number of elements in set[] such that struct arena_set
642 will fit within PERL_ARENA_SIZE, which is probably just under 4K, and
643 therefore likely to be 1 aligned memory page. */
645 #define ARENAS_PER_SET ((PERL_ARENA_SIZE - sizeof(struct arena_set*) \
646 - 2 * sizeof(int)) / sizeof (struct arena_desc))
649 struct arena_set* next;
650 unsigned int set_size; /* ie ARENAS_PER_SET */
651 unsigned int curr; /* index of next available arena-desc */
652 struct arena_desc set[ARENAS_PER_SET];
656 =for apidoc sv_free_arenas
658 Deallocate the memory used by all arenas. Note that all the individual SV
659 heads and bodies within the arenas must already have been freed.
664 Perl_sv_free_arenas(pTHX)
671 /* Free arenas here, but be careful about fake ones. (We assume
672 contiguity of the fake ones with the corresponding real ones.) */
674 for (sva = PL_sv_arenaroot; sva; sva = svanext) {
675 svanext = MUTABLE_SV(SvANY(sva));
676 while (svanext && SvFAKE(svanext))
677 svanext = MUTABLE_SV(SvANY(svanext));
684 struct arena_set *aroot = (struct arena_set*) PL_body_arenas;
687 struct arena_set *current = aroot;
690 assert(aroot->set[i].arena);
691 Safefree(aroot->set[i].arena);
699 i = PERL_ARENA_ROOTS_SIZE;
701 PL_body_roots[i] = 0;
708 Here are mid-level routines that manage the allocation of bodies out
709 of the various arenas. There are 5 kinds of arenas:
711 1. SV-head arenas, which are discussed and handled above
712 2. regular body arenas
713 3. arenas for reduced-size bodies
716 Arena types 2 & 3 are chained by body-type off an array of
717 arena-root pointers, which is indexed by svtype. Some of the
718 larger/less used body types are malloced singly, since a large
719 unused block of them is wasteful. Also, several svtypes dont have
720 bodies; the data fits into the sv-head itself. The arena-root
721 pointer thus has a few unused root-pointers (which may be hijacked
722 later for arena types 4,5)
724 3 differs from 2 as an optimization; some body types have several
725 unused fields in the front of the structure (which are kept in-place
726 for consistency). These bodies can be allocated in smaller chunks,
727 because the leading fields arent accessed. Pointers to such bodies
728 are decremented to point at the unused 'ghost' memory, knowing that
729 the pointers are used with offsets to the real memory.
732 =head1 SV-Body Allocation
734 Allocation of SV-bodies is similar to SV-heads, differing as follows;
735 the allocation mechanism is used for many body types, so is somewhat
736 more complicated, it uses arena-sets, and has no need for still-live
739 At the outermost level, (new|del)_X*V macros return bodies of the
740 appropriate type. These macros call either (new|del)_body_type or
741 (new|del)_body_allocated macro pairs, depending on specifics of the
742 type. Most body types use the former pair, the latter pair is used to
743 allocate body types with "ghost fields".
745 "ghost fields" are fields that are unused in certain types, and
746 consequently don't need to actually exist. They are declared because
747 they're part of a "base type", which allows use of functions as
748 methods. The simplest examples are AVs and HVs, 2 aggregate types
749 which don't use the fields which support SCALAR semantics.
751 For these types, the arenas are carved up into appropriately sized
752 chunks, we thus avoid wasted memory for those unaccessed members.
753 When bodies are allocated, we adjust the pointer back in memory by the
754 size of the part not allocated, so it's as if we allocated the full
755 structure. (But things will all go boom if you write to the part that
756 is "not there", because you'll be overwriting the last members of the
757 preceding structure in memory.)
759 We calculate the correction using the STRUCT_OFFSET macro on the first
760 member present. If the allocated structure is smaller (no initial NV
761 actually allocated) then the net effect is to subtract the size of the NV
762 from the pointer, to return a new pointer as if an initial NV were actually
763 allocated. (We were using structures named *_allocated for this, but
764 this turned out to be a subtle bug, because a structure without an NV
765 could have a lower alignment constraint, but the compiler is allowed to
766 optimised accesses based on the alignment constraint of the actual pointer
767 to the full structure, for example, using a single 64 bit load instruction
768 because it "knows" that two adjacent 32 bit members will be 8-byte aligned.)
770 This is the same trick as was used for NV and IV bodies. Ironically it
771 doesn't need to be used for NV bodies any more, because NV is now at
772 the start of the structure. IV bodies don't need it either, because
773 they are no longer allocated.
775 In turn, the new_body_* allocators call S_new_body(), which invokes
776 new_body_inline macro, which takes a lock, and takes a body off the
777 linked list at PL_body_roots[sv_type], calling Perl_more_bodies() if
778 necessary to refresh an empty list. Then the lock is released, and
779 the body is returned.
781 Perl_more_bodies allocates a new arena, and carves it up into an array of N
782 bodies, which it strings into a linked list. It looks up arena-size
783 and body-size from the body_details table described below, thus
784 supporting the multiple body-types.
786 If PURIFY is defined, or PERL_ARENA_SIZE=0, arenas are not used, and
787 the (new|del)_X*V macros are mapped directly to malloc/free.
789 For each sv-type, struct body_details bodies_by_type[] carries
790 parameters which control these aspects of SV handling:
792 Arena_size determines whether arenas are used for this body type, and if
793 so, how big they are. PURIFY or PERL_ARENA_SIZE=0 set this field to
794 zero, forcing individual mallocs and frees.
796 Body_size determines how big a body is, and therefore how many fit into
797 each arena. Offset carries the body-pointer adjustment needed for
798 "ghost fields", and is used in *_allocated macros.
800 But its main purpose is to parameterize info needed in
801 Perl_sv_upgrade(). The info here dramatically simplifies the function
802 vs the implementation in 5.8.8, making it table-driven. All fields
803 are used for this, except for arena_size.
805 For the sv-types that have no bodies, arenas are not used, so those
806 PL_body_roots[sv_type] are unused, and can be overloaded. In
807 something of a special case, SVt_NULL is borrowed for HE arenas;
808 PL_body_roots[HE_SVSLOT=SVt_NULL] is filled by S_more_he, but the
809 bodies_by_type[SVt_NULL] slot is not used, as the table is not
814 struct body_details {
815 U8 body_size; /* Size to allocate */
816 U8 copy; /* Size of structure to copy (may be shorter) */
818 unsigned int type : 4; /* We have space for a sanity check. */
819 unsigned int cant_upgrade : 1; /* Cannot upgrade this type */
820 unsigned int zero_nv : 1; /* zero the NV when upgrading from this */
821 unsigned int arena : 1; /* Allocated from an arena */
822 size_t arena_size; /* Size of arena to allocate */
830 /* With -DPURFIY we allocate everything directly, and don't use arenas.
831 This seems a rather elegant way to simplify some of the code below. */
832 #define HASARENA FALSE
834 #define HASARENA TRUE
836 #define NOARENA FALSE
838 /* Size the arenas to exactly fit a given number of bodies. A count
839 of 0 fits the max number bodies into a PERL_ARENA_SIZE.block,
840 simplifying the default. If count > 0, the arena is sized to fit
841 only that many bodies, allowing arenas to be used for large, rare
842 bodies (XPVFM, XPVIO) without undue waste. The arena size is
843 limited by PERL_ARENA_SIZE, so we can safely oversize the
846 #define FIT_ARENA0(body_size) \
847 ((size_t)(PERL_ARENA_SIZE / body_size) * body_size)
848 #define FIT_ARENAn(count,body_size) \
849 ( count * body_size <= PERL_ARENA_SIZE) \
850 ? count * body_size \
851 : FIT_ARENA0 (body_size)
852 #define FIT_ARENA(count,body_size) \
854 ? FIT_ARENAn (count, body_size) \
855 : FIT_ARENA0 (body_size)
857 /* Calculate the length to copy. Specifically work out the length less any
858 final padding the compiler needed to add. See the comment in sv_upgrade
859 for why copying the padding proved to be a bug. */
861 #define copy_length(type, last_member) \
862 STRUCT_OFFSET(type, last_member) \
863 + sizeof (((type*)SvANY((const SV *)0))->last_member)
865 static const struct body_details bodies_by_type[] = {
866 /* HEs use this offset for their arena. */
867 { 0, 0, 0, SVt_NULL, FALSE, NONV, NOARENA, 0 },
869 /* The bind placeholder pretends to be an RV for now.
870 Also it's marked as "can't upgrade" to stop anyone using it before it's
872 { 0, 0, 0, SVt_BIND, TRUE, NONV, NOARENA, 0 },
874 /* IVs are in the head, so the allocation size is 0. */
876 sizeof(IV), /* This is used to copy out the IV body. */
877 STRUCT_OFFSET(XPVIV, xiv_iv), SVt_IV, FALSE, NONV,
878 NOARENA /* IVS don't need an arena */, 0
881 /* 8 bytes on most ILP32 with IEEE doubles */
882 { sizeof(NV), sizeof(NV),
883 STRUCT_OFFSET(XPVNV, xnv_u),
884 SVt_NV, FALSE, HADNV, HASARENA, FIT_ARENA(0, sizeof(NV)) },
886 /* 8 bytes on most ILP32 with IEEE doubles */
887 { sizeof(XPV) - STRUCT_OFFSET(XPV, xpv_cur),
888 copy_length(XPV, xpv_len) - STRUCT_OFFSET(XPV, xpv_cur),
889 + STRUCT_OFFSET(XPV, xpv_cur),
890 SVt_PV, FALSE, NONV, HASARENA,
891 FIT_ARENA(0, sizeof(XPV) - STRUCT_OFFSET(XPV, xpv_cur)) },
894 { sizeof(XPVIV) - STRUCT_OFFSET(XPV, xpv_cur),
895 copy_length(XPVIV, xiv_u) - STRUCT_OFFSET(XPV, xpv_cur),
896 + STRUCT_OFFSET(XPV, xpv_cur),
897 SVt_PVIV, FALSE, NONV, HASARENA,
898 FIT_ARENA(0, sizeof(XPVIV) - STRUCT_OFFSET(XPV, xpv_cur)) },
901 { sizeof(XPVNV) - STRUCT_OFFSET(XPV, xpv_cur),
902 copy_length(XPVNV, xnv_u) - STRUCT_OFFSET(XPV, xpv_cur),
903 + STRUCT_OFFSET(XPV, xpv_cur),
904 SVt_PVNV, FALSE, HADNV, HASARENA,
905 FIT_ARENA(0, sizeof(XPVNV) - STRUCT_OFFSET(XPV, xpv_cur)) },
908 { sizeof(XPVMG), copy_length(XPVMG, xnv_u), 0, SVt_PVMG, FALSE, HADNV,
909 HASARENA, FIT_ARENA(0, sizeof(XPVMG)) },
915 SVt_REGEXP, FALSE, NONV, HASARENA,
916 FIT_ARENA(0, sizeof(regexp))
920 { sizeof(XPVGV), sizeof(XPVGV), 0, SVt_PVGV, TRUE, HADNV,
921 HASARENA, FIT_ARENA(0, sizeof(XPVGV)) },
924 { sizeof(XPVLV), sizeof(XPVLV), 0, SVt_PVLV, TRUE, HADNV,
925 HASARENA, FIT_ARENA(0, sizeof(XPVLV)) },
928 copy_length(XPVAV, xav_alloc),
930 SVt_PVAV, TRUE, NONV, HASARENA,
931 FIT_ARENA(0, sizeof(XPVAV)) },
934 copy_length(XPVHV, xhv_max),
936 SVt_PVHV, TRUE, NONV, HASARENA,
937 FIT_ARENA(0, sizeof(XPVHV)) },
943 SVt_PVCV, TRUE, NONV, HASARENA,
944 FIT_ARENA(0, sizeof(XPVCV)) },
949 SVt_PVFM, TRUE, NONV, NOARENA,
950 FIT_ARENA(20, sizeof(XPVFM)) },
952 /* XPVIO is 84 bytes, fits 48x */
956 SVt_PVIO, TRUE, NONV, HASARENA,
957 FIT_ARENA(24, sizeof(XPVIO)) },
960 #define new_body_allocated(sv_type) \
961 (void *)((char *)S_new_body(aTHX_ sv_type) \
962 - bodies_by_type[sv_type].offset)
964 /* return a thing to the free list */
966 #define del_body(thing, root) \
968 void ** const thing_copy = (void **)thing; \
969 *thing_copy = *root; \
970 *root = (void*)thing_copy; \
975 #define new_XNV() safemalloc(sizeof(XPVNV))
976 #define new_XPVNV() safemalloc(sizeof(XPVNV))
977 #define new_XPVMG() safemalloc(sizeof(XPVMG))
979 #define del_XPVGV(p) safefree(p)
983 #define new_XNV() new_body_allocated(SVt_NV)
984 #define new_XPVNV() new_body_allocated(SVt_PVNV)
985 #define new_XPVMG() new_body_allocated(SVt_PVMG)
987 #define del_XPVGV(p) del_body(p + bodies_by_type[SVt_PVGV].offset, \
988 &PL_body_roots[SVt_PVGV])
992 /* no arena for you! */
994 #define new_NOARENA(details) \
995 safemalloc((details)->body_size + (details)->offset)
996 #define new_NOARENAZ(details) \
997 safecalloc((details)->body_size + (details)->offset, 1)
1000 Perl_more_bodies (pTHX_ const svtype sv_type, const size_t body_size,
1001 const size_t arena_size)
1004 void ** const root = &PL_body_roots[sv_type];
1005 struct arena_desc *adesc;
1006 struct arena_set *aroot = (struct arena_set *) PL_body_arenas;
1010 const size_t good_arena_size = Perl_malloc_good_size(arena_size);
1011 #if defined(DEBUGGING) && !defined(PERL_GLOBAL_STRUCT_PRIVATE)
1012 static bool done_sanity_check;
1014 /* PERL_GLOBAL_STRUCT_PRIVATE cannot coexist with global
1015 * variables like done_sanity_check. */
1016 if (!done_sanity_check) {
1017 unsigned int i = SVt_LAST;
1019 done_sanity_check = TRUE;
1022 assert (bodies_by_type[i].type == i);
1028 /* may need new arena-set to hold new arena */
1029 if (!aroot || aroot->curr >= aroot->set_size) {
1030 struct arena_set *newroot;
1031 Newxz(newroot, 1, struct arena_set);
1032 newroot->set_size = ARENAS_PER_SET;
1033 newroot->next = aroot;
1035 PL_body_arenas = (void *) newroot;
1036 DEBUG_m(PerlIO_printf(Perl_debug_log, "new arenaset %p\n", (void*)aroot));
1039 /* ok, now have arena-set with at least 1 empty/available arena-desc */
1040 curr = aroot->curr++;
1041 adesc = &(aroot->set[curr]);
1042 assert(!adesc->arena);
1044 Newx(adesc->arena, good_arena_size, char);
1045 adesc->size = good_arena_size;
1046 adesc->utype = sv_type;
1047 DEBUG_m(PerlIO_printf(Perl_debug_log, "arena %d added: %p size %"UVuf"\n",
1048 curr, (void*)adesc->arena, (UV)good_arena_size));
1050 start = (char *) adesc->arena;
1052 /* Get the address of the byte after the end of the last body we can fit.
1053 Remember, this is integer division: */
1054 end = start + good_arena_size / body_size * body_size;
1056 /* computed count doesnt reflect the 1st slot reservation */
1057 #if defined(MYMALLOC) || defined(HAS_MALLOC_GOOD_SIZE)
1058 DEBUG_m(PerlIO_printf(Perl_debug_log,
1059 "arena %p end %p arena-size %d (from %d) type %d "
1061 (void*)start, (void*)end, (int)good_arena_size,
1062 (int)arena_size, sv_type, (int)body_size,
1063 (int)good_arena_size / (int)body_size));
1065 DEBUG_m(PerlIO_printf(Perl_debug_log,
1066 "arena %p end %p arena-size %d type %d size %d ct %d\n",
1067 (void*)start, (void*)end,
1068 (int)arena_size, sv_type, (int)body_size,
1069 (int)good_arena_size / (int)body_size));
1071 *root = (void *)start;
1074 /* Where the next body would start: */
1075 char * const next = start + body_size;
1078 /* This is the last body: */
1079 assert(next == end);
1081 *(void **)start = 0;
1085 *(void**) start = (void *)next;
1090 /* grab a new thing from the free list, allocating more if necessary.
1091 The inline version is used for speed in hot routines, and the
1092 function using it serves the rest (unless PURIFY).
1094 #define new_body_inline(xpv, sv_type) \
1096 void ** const r3wt = &PL_body_roots[sv_type]; \
1097 xpv = (PTR_TBL_ENT_t*) (*((void **)(r3wt)) \
1098 ? *((void **)(r3wt)) : Perl_more_bodies(aTHX_ sv_type, \
1099 bodies_by_type[sv_type].body_size,\
1100 bodies_by_type[sv_type].arena_size)); \
1101 *(r3wt) = *(void**)(xpv); \
1107 S_new_body(pTHX_ const svtype sv_type)
1111 new_body_inline(xpv, sv_type);
1117 static const struct body_details fake_rv =
1118 { 0, 0, 0, SVt_IV, FALSE, NONV, NOARENA, 0 };
1121 =for apidoc sv_upgrade
1123 Upgrade an SV to a more complex form. Generally adds a new body type to the
1124 SV, then copies across as much information as possible from the old body.
1125 You generally want to use the C<SvUPGRADE> macro wrapper. See also C<svtype>.
1131 Perl_sv_upgrade(pTHX_ register SV *const sv, svtype new_type)
1136 const svtype old_type = SvTYPE(sv);
1137 const struct body_details *new_type_details;
1138 const struct body_details *old_type_details
1139 = bodies_by_type + old_type;
1140 SV *referant = NULL;
1142 PERL_ARGS_ASSERT_SV_UPGRADE;
1144 if (old_type == new_type)
1147 /* This clause was purposefully added ahead of the early return above to
1148 the shared string hackery for (sort {$a <=> $b} keys %hash), with the
1149 inference by Nick I-S that it would fix other troublesome cases. See
1150 changes 7162, 7163 (f130fd4589cf5fbb24149cd4db4137c8326f49c1 and parent)
1152 Given that shared hash key scalars are no longer PVIV, but PV, there is
1153 no longer need to unshare so as to free up the IVX slot for its proper
1154 purpose. So it's safe to move the early return earlier. */
1156 if (new_type != SVt_PV && SvIsCOW(sv)) {
1157 sv_force_normal_flags(sv, 0);
1160 old_body = SvANY(sv);
1162 /* Copying structures onto other structures that have been neatly zeroed
1163 has a subtle gotcha. Consider XPVMG
1165 +------+------+------+------+------+-------+-------+
1166 | NV | CUR | LEN | IV | MAGIC | STASH |
1167 +------+------+------+------+------+-------+-------+
1168 0 4 8 12 16 20 24 28
1170 where NVs are aligned to 8 bytes, so that sizeof that structure is
1171 actually 32 bytes long, with 4 bytes of padding at the end:
1173 +------+------+------+------+------+-------+-------+------+
1174 | NV | CUR | LEN | IV | MAGIC | STASH | ??? |
1175 +------+------+------+------+------+-------+-------+------+
1176 0 4 8 12 16 20 24 28 32
1178 so what happens if you allocate memory for this structure:
1180 +------+------+------+------+------+-------+-------+------+------+...
1181 | NV | CUR | LEN | IV | MAGIC | STASH | GP | NAME |
1182 +------+------+------+------+------+-------+-------+------+------+...
1183 0 4 8 12 16 20 24 28 32 36
1185 zero it, then copy sizeof(XPVMG) bytes on top of it? Not quite what you
1186 expect, because you copy the area marked ??? onto GP. Now, ??? may have
1187 started out as zero once, but it's quite possible that it isn't. So now,
1188 rather than a nicely zeroed GP, you have it pointing somewhere random.
1191 (In fact, GP ends up pointing at a previous GP structure, because the
1192 principle cause of the padding in XPVMG getting garbage is a copy of
1193 sizeof(XPVMG) bytes from a XPVGV structure in sv_unglob. Right now
1194 this happens to be moot because XPVGV has been re-ordered, with GP
1195 no longer after STASH)
1197 So we are careful and work out the size of used parts of all the
1205 referant = SvRV(sv);
1206 old_type_details = &fake_rv;
1207 if (new_type == SVt_NV)
1208 new_type = SVt_PVNV;
1210 if (new_type < SVt_PVIV) {
1211 new_type = (new_type == SVt_NV)
1212 ? SVt_PVNV : SVt_PVIV;
1217 if (new_type < SVt_PVNV) {
1218 new_type = SVt_PVNV;
1222 assert(new_type > SVt_PV);
1223 assert(SVt_IV < SVt_PV);
1224 assert(SVt_NV < SVt_PV);
1231 /* Because the XPVMG of PL_mess_sv isn't allocated from the arena,
1232 there's no way that it can be safely upgraded, because perl.c
1233 expects to Safefree(SvANY(PL_mess_sv)) */
1234 assert(sv != PL_mess_sv);
1235 /* This flag bit is used to mean other things in other scalar types.
1236 Given that it only has meaning inside the pad, it shouldn't be set
1237 on anything that can get upgraded. */
1238 assert(!SvPAD_TYPED(sv));
1241 if (old_type_details->cant_upgrade)
1242 Perl_croak(aTHX_ "Can't upgrade %s (%" UVuf ") to %" UVuf,
1243 sv_reftype(sv, 0), (UV) old_type, (UV) new_type);
1246 if (old_type > new_type)
1247 Perl_croak(aTHX_ "sv_upgrade from type %d down to type %d",
1248 (int)old_type, (int)new_type);
1250 new_type_details = bodies_by_type + new_type;
1252 SvFLAGS(sv) &= ~SVTYPEMASK;
1253 SvFLAGS(sv) |= new_type;
1255 /* This can't happen, as SVt_NULL is <= all values of new_type, so one of
1256 the return statements above will have triggered. */
1257 assert (new_type != SVt_NULL);
1260 assert(old_type == SVt_NULL);
1261 SvANY(sv) = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
1265 assert(old_type == SVt_NULL);
1266 SvANY(sv) = new_XNV();
1271 assert(new_type_details->body_size);
1274 assert(new_type_details->arena);
1275 assert(new_type_details->arena_size);
1276 /* This points to the start of the allocated area. */
1277 new_body_inline(new_body, new_type);
1278 Zero(new_body, new_type_details->body_size, char);
1279 new_body = ((char *)new_body) - new_type_details->offset;
1281 /* We always allocated the full length item with PURIFY. To do this
1282 we fake things so that arena is false for all 16 types.. */
1283 new_body = new_NOARENAZ(new_type_details);
1285 SvANY(sv) = new_body;
1286 if (new_type == SVt_PVAV) {
1290 if (old_type_details->body_size) {
1293 /* It will have been zeroed when the new body was allocated.
1294 Lets not write to it, in case it confuses a write-back
1300 #ifndef NODEFAULT_SHAREKEYS
1301 HvSHAREKEYS_on(sv); /* key-sharing on by default */
1303 HvMAX(sv) = 7; /* (start with 8 buckets) */
1306 /* SVt_NULL isn't the only thing upgraded to AV or HV.
1307 The target created by newSVrv also is, and it can have magic.
1308 However, it never has SvPVX set.
1310 if (old_type == SVt_IV) {
1312 } else if (old_type >= SVt_PV) {
1313 assert(SvPVX_const(sv) == 0);
1316 if (old_type >= SVt_PVMG) {
1317 SvMAGIC_set(sv, ((XPVMG*)old_body)->xmg_u.xmg_magic);
1318 SvSTASH_set(sv, ((XPVMG*)old_body)->xmg_stash);
1320 sv->sv_u.svu_array = NULL; /* or svu_hash */
1326 /* This ensures that SvTHINKFIRST(sv) is true, and hence that
1327 sv_force_normal_flags(sv) is called. */
1330 /* XXX Is this still needed? Was it ever needed? Surely as there is
1331 no route from NV to PVIV, NOK can never be true */
1332 assert(!SvNOKp(sv));
1343 assert(new_type_details->body_size);
1344 /* We always allocated the full length item with PURIFY. To do this
1345 we fake things so that arena is false for all 16 types.. */
1346 if(new_type_details->arena) {
1347 /* This points to the start of the allocated area. */
1348 new_body_inline(new_body, new_type);
1349 Zero(new_body, new_type_details->body_size, char);
1350 new_body = ((char *)new_body) - new_type_details->offset;
1352 new_body = new_NOARENAZ(new_type_details);
1354 SvANY(sv) = new_body;
1356 if (old_type_details->copy) {
1357 /* There is now the potential for an upgrade from something without
1358 an offset (PVNV or PVMG) to something with one (PVCV, PVFM) */
1359 int offset = old_type_details->offset;
1360 int length = old_type_details->copy;
1362 if (new_type_details->offset > old_type_details->offset) {
1363 const int difference
1364 = new_type_details->offset - old_type_details->offset;
1365 offset += difference;
1366 length -= difference;
1368 assert (length >= 0);
1370 Copy((char *)old_body + offset, (char *)new_body + offset, length,
1374 #ifndef NV_ZERO_IS_ALLBITS_ZERO
1375 /* If NV 0.0 is stores as all bits 0 then Zero() already creates a
1376 * correct 0.0 for us. Otherwise, if the old body didn't have an
1377 * NV slot, but the new one does, then we need to initialise the
1378 * freshly created NV slot with whatever the correct bit pattern is
1380 if (old_type_details->zero_nv && !new_type_details->zero_nv
1381 && !isGV_with_GP(sv))
1385 if (new_type == SVt_PVIO) {
1386 IO * const io = MUTABLE_IO(sv);
1387 GV *iogv = gv_fetchpvs("IO::File::", GV_ADD, SVt_PVHV);
1390 /* Clear the stashcache because a new IO could overrule a package
1392 hv_clear(PL_stashcache);
1394 SvSTASH_set(io, MUTABLE_HV(SvREFCNT_inc(GvHV(iogv))));
1395 IoPAGE_LEN(sv) = 60;
1397 if (old_type < SVt_PV) {
1398 /* referant will be NULL unless the old type was SVt_IV emulating
1400 sv->sv_u.svu_rv = referant;
1404 Perl_croak(aTHX_ "panic: sv_upgrade to unknown type %lu",
1405 (unsigned long)new_type);
1408 if (old_type > SVt_IV) {
1412 /* Note that there is an assumption that all bodies of types that
1413 can be upgraded came from arenas. Only the more complex non-
1414 upgradable types are allowed to be directly malloc()ed. */
1415 assert(old_type_details->arena);
1416 del_body((void*)((char*)old_body + old_type_details->offset),
1417 &PL_body_roots[old_type]);
1423 =for apidoc sv_backoff
1425 Remove any string offset. You should normally use the C<SvOOK_off> macro
1432 Perl_sv_backoff(pTHX_ register SV *const sv)
1435 const char * const s = SvPVX_const(sv);
1437 PERL_ARGS_ASSERT_SV_BACKOFF;
1438 PERL_UNUSED_CONTEXT;
1441 assert(SvTYPE(sv) != SVt_PVHV);
1442 assert(SvTYPE(sv) != SVt_PVAV);
1444 SvOOK_offset(sv, delta);
1446 SvLEN_set(sv, SvLEN(sv) + delta);
1447 SvPV_set(sv, SvPVX(sv) - delta);
1448 Move(s, SvPVX(sv), SvCUR(sv)+1, char);
1449 SvFLAGS(sv) &= ~SVf_OOK;
1456 Expands the character buffer in the SV. If necessary, uses C<sv_unref> and
1457 upgrades the SV to C<SVt_PV>. Returns a pointer to the character buffer.
1458 Use the C<SvGROW> wrapper instead.
1464 Perl_sv_grow(pTHX_ register SV *const sv, register STRLEN newlen)
1468 PERL_ARGS_ASSERT_SV_GROW;
1470 if (PL_madskills && newlen >= 0x100000) {
1471 PerlIO_printf(Perl_debug_log,
1472 "Allocation too large: %"UVxf"\n", (UV)newlen);
1474 #ifdef HAS_64K_LIMIT
1475 if (newlen >= 0x10000) {
1476 PerlIO_printf(Perl_debug_log,
1477 "Allocation too large: %"UVxf"\n", (UV)newlen);
1480 #endif /* HAS_64K_LIMIT */
1483 if (SvTYPE(sv) < SVt_PV) {
1484 sv_upgrade(sv, SVt_PV);
1485 s = SvPVX_mutable(sv);
1487 else if (SvOOK(sv)) { /* pv is offset? */
1489 s = SvPVX_mutable(sv);
1490 if (newlen > SvLEN(sv))
1491 newlen += 10 * (newlen - SvCUR(sv)); /* avoid copy each time */
1492 #ifdef HAS_64K_LIMIT
1493 if (newlen >= 0x10000)
1498 s = SvPVX_mutable(sv);
1500 if (newlen > SvLEN(sv)) { /* need more room? */
1501 STRLEN minlen = SvCUR(sv);
1502 minlen += (minlen >> PERL_STRLEN_EXPAND_SHIFT) + 10;
1503 if (newlen < minlen)
1505 #ifndef Perl_safesysmalloc_size
1506 newlen = PERL_STRLEN_ROUNDUP(newlen);
1508 if (SvLEN(sv) && s) {
1509 s = (char*)saferealloc(s, newlen);
1512 s = (char*)safemalloc(newlen);
1513 if (SvPVX_const(sv) && SvCUR(sv)) {
1514 Move(SvPVX_const(sv), s, (newlen < SvCUR(sv)) ? newlen : SvCUR(sv), char);
1518 #ifdef Perl_safesysmalloc_size
1519 /* Do this here, do it once, do it right, and then we will never get
1520 called back into sv_grow() unless there really is some growing
1522 SvLEN_set(sv, Perl_safesysmalloc_size(s));
1524 SvLEN_set(sv, newlen);
1531 =for apidoc sv_setiv
1533 Copies an integer into the given SV, upgrading first if necessary.
1534 Does not handle 'set' magic. See also C<sv_setiv_mg>.
1540 Perl_sv_setiv(pTHX_ register SV *const sv, const IV i)
1544 PERL_ARGS_ASSERT_SV_SETIV;
1546 SV_CHECK_THINKFIRST_COW_DROP(sv);
1547 switch (SvTYPE(sv)) {
1550 sv_upgrade(sv, SVt_IV);
1553 sv_upgrade(sv, SVt_PVIV);
1557 if (!isGV_with_GP(sv))
1564 Perl_croak(aTHX_ "Can't coerce %s to integer in %s", sv_reftype(sv,0),
1568 (void)SvIOK_only(sv); /* validate number */
1574 =for apidoc sv_setiv_mg
1576 Like C<sv_setiv>, but also handles 'set' magic.
1582 Perl_sv_setiv_mg(pTHX_ register SV *const sv, const IV i)
1584 PERL_ARGS_ASSERT_SV_SETIV_MG;
1591 =for apidoc sv_setuv
1593 Copies an unsigned integer into the given SV, upgrading first if necessary.
1594 Does not handle 'set' magic. See also C<sv_setuv_mg>.
1600 Perl_sv_setuv(pTHX_ register SV *const sv, const UV u)
1602 PERL_ARGS_ASSERT_SV_SETUV;
1604 /* With these two if statements:
1605 u=1.49 s=0.52 cu=72.49 cs=10.64 scripts=270 tests=20865
1608 u=1.35 s=0.47 cu=73.45 cs=11.43 scripts=270 tests=20865
1610 If you wish to remove them, please benchmark to see what the effect is
1612 if (u <= (UV)IV_MAX) {
1613 sv_setiv(sv, (IV)u);
1622 =for apidoc sv_setuv_mg
1624 Like C<sv_setuv>, but also handles 'set' magic.
1630 Perl_sv_setuv_mg(pTHX_ register SV *const sv, const UV u)
1632 PERL_ARGS_ASSERT_SV_SETUV_MG;
1639 =for apidoc sv_setnv
1641 Copies a double into the given SV, upgrading first if necessary.
1642 Does not handle 'set' magic. See also C<sv_setnv_mg>.
1648 Perl_sv_setnv(pTHX_ register SV *const sv, const NV num)
1652 PERL_ARGS_ASSERT_SV_SETNV;
1654 SV_CHECK_THINKFIRST_COW_DROP(sv);
1655 switch (SvTYPE(sv)) {
1658 sv_upgrade(sv, SVt_NV);
1662 sv_upgrade(sv, SVt_PVNV);
1666 if (!isGV_with_GP(sv))
1673 Perl_croak(aTHX_ "Can't coerce %s to number in %s", sv_reftype(sv,0),
1678 (void)SvNOK_only(sv); /* validate number */
1683 =for apidoc sv_setnv_mg
1685 Like C<sv_setnv>, but also handles 'set' magic.
1691 Perl_sv_setnv_mg(pTHX_ register SV *const sv, const NV num)
1693 PERL_ARGS_ASSERT_SV_SETNV_MG;
1699 /* Print an "isn't numeric" warning, using a cleaned-up,
1700 * printable version of the offending string
1704 S_not_a_number(pTHX_ SV *const sv)
1711 PERL_ARGS_ASSERT_NOT_A_NUMBER;
1714 dsv = newSVpvs_flags("", SVs_TEMP);
1715 pv = sv_uni_display(dsv, sv, 10, 0);
1718 const char * const limit = tmpbuf + sizeof(tmpbuf) - 8;
1719 /* each *s can expand to 4 chars + "...\0",
1720 i.e. need room for 8 chars */
1722 const char *s = SvPVX_const(sv);
1723 const char * const end = s + SvCUR(sv);
1724 for ( ; s < end && d < limit; s++ ) {
1726 if (ch & 128 && !isPRINT_LC(ch)) {
1735 else if (ch == '\r') {
1739 else if (ch == '\f') {
1743 else if (ch == '\\') {
1747 else if (ch == '\0') {
1751 else if (isPRINT_LC(ch))
1768 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
1769 "Argument \"%s\" isn't numeric in %s", pv,
1772 Perl_warner(aTHX_ packWARN(WARN_NUMERIC),
1773 "Argument \"%s\" isn't numeric", pv);
1777 =for apidoc looks_like_number
1779 Test if the content of an SV looks like a number (or is a number).
1780 C<Inf> and C<Infinity> are treated as numbers (so will not issue a
1781 non-numeric warning), even if your atof() doesn't grok them.
1787 Perl_looks_like_number(pTHX_ SV *const sv)
1789 register const char *sbegin;
1792 PERL_ARGS_ASSERT_LOOKS_LIKE_NUMBER;
1795 sbegin = SvPVX_const(sv);
1798 else if (SvPOKp(sv))
1799 sbegin = SvPV_const(sv, len);
1801 return SvFLAGS(sv) & (SVf_NOK|SVp_NOK|SVf_IOK|SVp_IOK);
1802 return grok_number(sbegin, len, NULL);
1806 S_glob_2number(pTHX_ GV * const gv)
1808 const U32 wasfake = SvFLAGS(gv) & SVf_FAKE;
1809 SV *const buffer = sv_newmortal();
1811 PERL_ARGS_ASSERT_GLOB_2NUMBER;
1813 /* FAKE globs can get coerced, so need to turn this off temporarily if it
1816 gv_efullname3(buffer, gv, "*");
1817 SvFLAGS(gv) |= wasfake;
1819 /* We know that all GVs stringify to something that is not-a-number,
1820 so no need to test that. */
1821 if (ckWARN(WARN_NUMERIC))
1822 not_a_number(buffer);
1823 /* We just want something true to return, so that S_sv_2iuv_common
1824 can tail call us and return true. */
1828 /* Actually, ISO C leaves conversion of UV to IV undefined, but
1829 until proven guilty, assume that things are not that bad... */
1834 As 64 bit platforms often have an NV that doesn't preserve all bits of
1835 an IV (an assumption perl has been based on to date) it becomes necessary
1836 to remove the assumption that the NV always carries enough precision to
1837 recreate the IV whenever needed, and that the NV is the canonical form.
1838 Instead, IV/UV and NV need to be given equal rights. So as to not lose
1839 precision as a side effect of conversion (which would lead to insanity
1840 and the dragon(s) in t/op/numconvert.t getting very angry) the intent is
1841 1) to distinguish between IV/UV/NV slots that have cached a valid
1842 conversion where precision was lost and IV/UV/NV slots that have a
1843 valid conversion which has lost no precision
1844 2) to ensure that if a numeric conversion to one form is requested that
1845 would lose precision, the precise conversion (or differently
1846 imprecise conversion) is also performed and cached, to prevent
1847 requests for different numeric formats on the same SV causing
1848 lossy conversion chains. (lossless conversion chains are perfectly
1853 SvIOKp is true if the IV slot contains a valid value
1854 SvIOK is true only if the IV value is accurate (UV if SvIOK_UV true)
1855 SvNOKp is true if the NV slot contains a valid value
1856 SvNOK is true only if the NV value is accurate
1859 while converting from PV to NV, check to see if converting that NV to an
1860 IV(or UV) would lose accuracy over a direct conversion from PV to
1861 IV(or UV). If it would, cache both conversions, return NV, but mark
1862 SV as IOK NOKp (ie not NOK).
1864 While converting from PV to IV, check to see if converting that IV to an
1865 NV would lose accuracy over a direct conversion from PV to NV. If it
1866 would, cache both conversions, flag similarly.
1868 Before, the SV value "3.2" could become NV=3.2 IV=3 NOK, IOK quite
1869 correctly because if IV & NV were set NV *always* overruled.
1870 Now, "3.2" will become NV=3.2 IV=3 NOK, IOKp, because the flag's meaning
1871 changes - now IV and NV together means that the two are interchangeable:
1872 SvIVX == (IV) SvNVX && SvNVX == (NV) SvIVX;
1874 The benefit of this is that operations such as pp_add know that if
1875 SvIOK is true for both left and right operands, then integer addition
1876 can be used instead of floating point (for cases where the result won't
1877 overflow). Before, floating point was always used, which could lead to
1878 loss of precision compared with integer addition.
1880 * making IV and NV equal status should make maths accurate on 64 bit
1882 * may speed up maths somewhat if pp_add and friends start to use
1883 integers when possible instead of fp. (Hopefully the overhead in
1884 looking for SvIOK and checking for overflow will not outweigh the
1885 fp to integer speedup)
1886 * will slow down integer operations (callers of SvIV) on "inaccurate"
1887 values, as the change from SvIOK to SvIOKp will cause a call into
1888 sv_2iv each time rather than a macro access direct to the IV slot
1889 * should speed up number->string conversion on integers as IV is
1890 favoured when IV and NV are equally accurate
1892 ####################################################################
1893 You had better be using SvIOK_notUV if you want an IV for arithmetic:
1894 SvIOK is true if (IV or UV), so you might be getting (IV)SvUV.
1895 On the other hand, SvUOK is true iff UV.
1896 ####################################################################
1898 Your mileage will vary depending your CPU's relative fp to integer
1902 #ifndef NV_PRESERVES_UV
1903 # define IS_NUMBER_UNDERFLOW_IV 1
1904 # define IS_NUMBER_UNDERFLOW_UV 2
1905 # define IS_NUMBER_IV_AND_UV 2
1906 # define IS_NUMBER_OVERFLOW_IV 4
1907 # define IS_NUMBER_OVERFLOW_UV 5
1909 /* sv_2iuv_non_preserve(): private routine for use by sv_2iv() and sv_2uv() */
1911 /* For sv_2nv these three cases are "SvNOK and don't bother casting" */
1913 S_sv_2iuv_non_preserve(pTHX_ register SV *const sv
1921 PERL_ARGS_ASSERT_SV_2IUV_NON_PRESERVE;
1923 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));
1924 if (SvNVX(sv) < (NV)IV_MIN) {
1925 (void)SvIOKp_on(sv);
1927 SvIV_set(sv, IV_MIN);
1928 return IS_NUMBER_UNDERFLOW_IV;
1930 if (SvNVX(sv) > (NV)UV_MAX) {
1931 (void)SvIOKp_on(sv);
1934 SvUV_set(sv, UV_MAX);
1935 return IS_NUMBER_OVERFLOW_UV;
1937 (void)SvIOKp_on(sv);
1939 /* Can't use strtol etc to convert this string. (See truth table in
1941 if (SvNVX(sv) <= (UV)IV_MAX) {
1942 SvIV_set(sv, I_V(SvNVX(sv)));
1943 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
1944 SvIOK_on(sv); /* Integer is precise. NOK, IOK */
1946 /* Integer is imprecise. NOK, IOKp */
1948 return SvNVX(sv) < 0 ? IS_NUMBER_UNDERFLOW_UV : IS_NUMBER_IV_AND_UV;
1951 SvUV_set(sv, U_V(SvNVX(sv)));
1952 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
1953 if (SvUVX(sv) == UV_MAX) {
1954 /* As we know that NVs don't preserve UVs, UV_MAX cannot
1955 possibly be preserved by NV. Hence, it must be overflow.
1957 return IS_NUMBER_OVERFLOW_UV;
1959 SvIOK_on(sv); /* Integer is precise. NOK, UOK */
1961 /* Integer is imprecise. NOK, IOKp */
1963 return IS_NUMBER_OVERFLOW_IV;
1965 #endif /* !NV_PRESERVES_UV*/
1968 S_sv_2iuv_common(pTHX_ SV *const sv)
1972 PERL_ARGS_ASSERT_SV_2IUV_COMMON;
1975 /* erm. not sure. *should* never get NOKp (without NOK) from sv_2nv
1976 * without also getting a cached IV/UV from it at the same time
1977 * (ie PV->NV conversion should detect loss of accuracy and cache
1978 * IV or UV at same time to avoid this. */
1979 /* IV-over-UV optimisation - choose to cache IV if possible */
1981 if (SvTYPE(sv) == SVt_NV)
1982 sv_upgrade(sv, SVt_PVNV);
1984 (void)SvIOKp_on(sv); /* Must do this first, to clear any SvOOK */
1985 /* < not <= as for NV doesn't preserve UV, ((NV)IV_MAX+1) will almost
1986 certainly cast into the IV range at IV_MAX, whereas the correct
1987 answer is the UV IV_MAX +1. Hence < ensures that dodgy boundary
1989 #if defined(NAN_COMPARE_BROKEN) && defined(Perl_isnan)
1990 if (Perl_isnan(SvNVX(sv))) {
1996 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
1997 SvIV_set(sv, I_V(SvNVX(sv)));
1998 if (SvNVX(sv) == (NV) SvIVX(sv)
1999 #ifndef NV_PRESERVES_UV
2000 && (((UV)1 << NV_PRESERVES_UV_BITS) >
2001 (UV)(SvIVX(sv) > 0 ? SvIVX(sv) : -SvIVX(sv)))
2002 /* Don't flag it as "accurately an integer" if the number
2003 came from a (by definition imprecise) NV operation, and
2004 we're outside the range of NV integer precision */
2008 SvIOK_on(sv); /* Can this go wrong with rounding? NWC */
2010 /* scalar has trailing garbage, eg "42a" */
2012 DEBUG_c(PerlIO_printf(Perl_debug_log,
2013 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (precise)\n",
2019 /* IV not precise. No need to convert from PV, as NV
2020 conversion would already have cached IV if it detected
2021 that PV->IV would be better than PV->NV->IV
2022 flags already correct - don't set public IOK. */
2023 DEBUG_c(PerlIO_printf(Perl_debug_log,
2024 "0x%"UVxf" iv(%"NVgf" => %"IVdf") (imprecise)\n",
2029 /* Can the above go wrong if SvIVX == IV_MIN and SvNVX < IV_MIN,
2030 but the cast (NV)IV_MIN rounds to a the value less (more
2031 negative) than IV_MIN which happens to be equal to SvNVX ??
2032 Analogous to 0xFFFFFFFFFFFFFFFF rounding up to NV (2**64) and
2033 NV rounding back to 0xFFFFFFFFFFFFFFFF, so UVX == UV(NVX) and
2034 (NV)UVX == NVX are both true, but the values differ. :-(
2035 Hopefully for 2s complement IV_MIN is something like
2036 0x8000000000000000 which will be exact. NWC */
2039 SvUV_set(sv, U_V(SvNVX(sv)));
2041 (SvNVX(sv) == (NV) SvUVX(sv))
2042 #ifndef NV_PRESERVES_UV
2043 /* Make sure it's not 0xFFFFFFFFFFFFFFFF */
2044 /*&& (SvUVX(sv) != UV_MAX) irrelevant with code below */
2045 && (((UV)1 << NV_PRESERVES_UV_BITS) > SvUVX(sv))
2046 /* Don't flag it as "accurately an integer" if the number
2047 came from a (by definition imprecise) NV operation, and
2048 we're outside the range of NV integer precision */
2054 DEBUG_c(PerlIO_printf(Perl_debug_log,
2055 "0x%"UVxf" 2iv(%"UVuf" => %"IVdf") (as unsigned)\n",
2061 else if (SvPOKp(sv) && SvLEN(sv)) {
2063 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2064 /* We want to avoid a possible problem when we cache an IV/ a UV which
2065 may be later translated to an NV, and the resulting NV is not
2066 the same as the direct translation of the initial string
2067 (eg 123.456 can shortcut to the IV 123 with atol(), but we must
2068 be careful to ensure that the value with the .456 is around if the
2069 NV value is requested in the future).
2071 This means that if we cache such an IV/a UV, we need to cache the
2072 NV as well. Moreover, we trade speed for space, and do not
2073 cache the NV if we are sure it's not needed.
2076 /* SVt_PVNV is one higher than SVt_PVIV, hence this order */
2077 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2078 == IS_NUMBER_IN_UV) {
2079 /* It's definitely an integer, only upgrade to PVIV */
2080 if (SvTYPE(sv) < SVt_PVIV)
2081 sv_upgrade(sv, SVt_PVIV);
2083 } else if (SvTYPE(sv) < SVt_PVNV)
2084 sv_upgrade(sv, SVt_PVNV);
2086 /* If NVs preserve UVs then we only use the UV value if we know that
2087 we aren't going to call atof() below. If NVs don't preserve UVs
2088 then the value returned may have more precision than atof() will
2089 return, even though value isn't perfectly accurate. */
2090 if ((numtype & (IS_NUMBER_IN_UV
2091 #ifdef NV_PRESERVES_UV
2094 )) == IS_NUMBER_IN_UV) {
2095 /* This won't turn off the public IOK flag if it was set above */
2096 (void)SvIOKp_on(sv);
2098 if (!(numtype & IS_NUMBER_NEG)) {
2100 if (value <= (UV)IV_MAX) {
2101 SvIV_set(sv, (IV)value);
2103 /* it didn't overflow, and it was positive. */
2104 SvUV_set(sv, value);
2108 /* 2s complement assumption */
2109 if (value <= (UV)IV_MIN) {
2110 SvIV_set(sv, -(IV)value);
2112 /* Too negative for an IV. This is a double upgrade, but
2113 I'm assuming it will be rare. */
2114 if (SvTYPE(sv) < SVt_PVNV)
2115 sv_upgrade(sv, SVt_PVNV);
2119 SvNV_set(sv, -(NV)value);
2120 SvIV_set(sv, IV_MIN);
2124 /* For !NV_PRESERVES_UV and IS_NUMBER_IN_UV and IS_NUMBER_NOT_INT we
2125 will be in the previous block to set the IV slot, and the next
2126 block to set the NV slot. So no else here. */
2128 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2129 != IS_NUMBER_IN_UV) {
2130 /* It wasn't an (integer that doesn't overflow the UV). */
2131 SvNV_set(sv, Atof(SvPVX_const(sv)));
2133 if (! numtype && ckWARN(WARN_NUMERIC))
2136 #if defined(USE_LONG_DOUBLE)
2137 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%" PERL_PRIgldbl ")\n",
2138 PTR2UV(sv), SvNVX(sv)));
2140 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"NVgf")\n",
2141 PTR2UV(sv), SvNVX(sv)));
2144 #ifdef NV_PRESERVES_UV
2145 (void)SvIOKp_on(sv);
2147 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2148 SvIV_set(sv, I_V(SvNVX(sv)));
2149 if ((NV)(SvIVX(sv)) == SvNVX(sv)) {
2152 NOOP; /* Integer is imprecise. NOK, IOKp */
2154 /* UV will not work better than IV */
2156 if (SvNVX(sv) > (NV)UV_MAX) {
2158 /* Integer is inaccurate. NOK, IOKp, is UV */
2159 SvUV_set(sv, UV_MAX);
2161 SvUV_set(sv, U_V(SvNVX(sv)));
2162 /* 0xFFFFFFFFFFFFFFFF not an issue in here, NVs
2163 NV preservse UV so can do correct comparison. */
2164 if ((NV)(SvUVX(sv)) == SvNVX(sv)) {
2167 NOOP; /* Integer is imprecise. NOK, IOKp, is UV */
2172 #else /* NV_PRESERVES_UV */
2173 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2174 == (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT)) {
2175 /* The IV/UV slot will have been set from value returned by
2176 grok_number above. The NV slot has just been set using
2179 assert (SvIOKp(sv));
2181 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2182 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2183 /* Small enough to preserve all bits. */
2184 (void)SvIOKp_on(sv);
2186 SvIV_set(sv, I_V(SvNVX(sv)));
2187 if ((NV)(SvIVX(sv)) == SvNVX(sv))
2189 /* Assumption: first non-preserved integer is < IV_MAX,
2190 this NV is in the preserved range, therefore: */
2191 if (!(U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))
2193 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);
2197 0 0 already failed to read UV.
2198 0 1 already failed to read UV.
2199 1 0 you won't get here in this case. IV/UV
2200 slot set, public IOK, Atof() unneeded.
2201 1 1 already read UV.
2202 so there's no point in sv_2iuv_non_preserve() attempting
2203 to use atol, strtol, strtoul etc. */
2205 sv_2iuv_non_preserve (sv, numtype);
2207 sv_2iuv_non_preserve (sv);
2211 #endif /* NV_PRESERVES_UV */
2212 /* It might be more code efficient to go through the entire logic above
2213 and conditionally set with SvIOKp_on() rather than SvIOK(), but it
2214 gets complex and potentially buggy, so more programmer efficient
2215 to do it this way, by turning off the public flags: */
2217 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK);
2221 if (isGV_with_GP(sv))
2222 return glob_2number(MUTABLE_GV(sv));
2224 if (!(SvFLAGS(sv) & SVs_PADTMP)) {
2225 if (!PL_localizing && ckWARN(WARN_UNINITIALIZED))
2228 if (SvTYPE(sv) < SVt_IV)
2229 /* Typically the caller expects that sv_any is not NULL now. */
2230 sv_upgrade(sv, SVt_IV);
2231 /* Return 0 from the caller. */
2238 =for apidoc sv_2iv_flags
2240 Return the integer value of an SV, doing any necessary string
2241 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2242 Normally used via the C<SvIV(sv)> and C<SvIVx(sv)> macros.
2248 Perl_sv_2iv_flags(pTHX_ register SV *const sv, const I32 flags)
2253 if (SvGMAGICAL(sv) || (SvTYPE(sv) == SVt_PVGV && SvVALID(sv))) {
2254 /* FBMs use the same flag bit as SVf_IVisUV, so must let them
2255 cache IVs just in case. In practice it seems that they never
2256 actually anywhere accessible by user Perl code, let alone get used
2257 in anything other than a string context. */
2258 if (flags & SV_GMAGIC)
2263 return I_V(SvNVX(sv));
2265 if (SvPOKp(sv) && SvLEN(sv)) {
2268 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2270 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2271 == IS_NUMBER_IN_UV) {
2272 /* It's definitely an integer */
2273 if (numtype & IS_NUMBER_NEG) {
2274 if (value < (UV)IV_MIN)
2277 if (value < (UV)IV_MAX)
2282 if (ckWARN(WARN_NUMERIC))
2285 return I_V(Atof(SvPVX_const(sv)));
2290 assert(SvTYPE(sv) >= SVt_PVMG);
2291 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
2292 } else if (SvTHINKFIRST(sv)) {
2297 if (flags & SV_SKIP_OVERLOAD)
2299 tmpstr=AMG_CALLun(sv,numer);
2300 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2301 return SvIV(tmpstr);
2304 return PTR2IV(SvRV(sv));
2307 sv_force_normal_flags(sv, 0);
2309 if (SvREADONLY(sv) && !SvOK(sv)) {
2310 if (ckWARN(WARN_UNINITIALIZED))
2316 if (S_sv_2iuv_common(aTHX_ sv))
2319 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2iv(%"IVdf")\n",
2320 PTR2UV(sv),SvIVX(sv)));
2321 return SvIsUV(sv) ? (IV)SvUVX(sv) : SvIVX(sv);
2325 =for apidoc sv_2uv_flags
2327 Return the unsigned integer value of an SV, doing any necessary string
2328 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2329 Normally used via the C<SvUV(sv)> and C<SvUVx(sv)> macros.
2335 Perl_sv_2uv_flags(pTHX_ register SV *const sv, const I32 flags)
2340 if (SvGMAGICAL(sv) || (SvTYPE(sv) == SVt_PVGV && SvVALID(sv))) {
2341 /* FBMs use the same flag bit as SVf_IVisUV, so must let them
2342 cache IVs just in case. */
2343 if (flags & SV_GMAGIC)
2348 return U_V(SvNVX(sv));
2349 if (SvPOKp(sv) && SvLEN(sv)) {
2352 = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2354 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2355 == IS_NUMBER_IN_UV) {
2356 /* It's definitely an integer */
2357 if (!(numtype & IS_NUMBER_NEG))
2361 if (ckWARN(WARN_NUMERIC))
2364 return U_V(Atof(SvPVX_const(sv)));
2369 assert(SvTYPE(sv) >= SVt_PVMG);
2370 /* This falls through to the report_uninit inside S_sv_2iuv_common. */
2371 } else if (SvTHINKFIRST(sv)) {
2376 if (flags & SV_SKIP_OVERLOAD)
2378 tmpstr = AMG_CALLun(sv,numer);
2379 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2380 return SvUV(tmpstr);
2383 return PTR2UV(SvRV(sv));
2386 sv_force_normal_flags(sv, 0);
2388 if (SvREADONLY(sv) && !SvOK(sv)) {
2389 if (ckWARN(WARN_UNINITIALIZED))
2395 if (S_sv_2iuv_common(aTHX_ sv))
2399 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2uv(%"UVuf")\n",
2400 PTR2UV(sv),SvUVX(sv)));
2401 return SvIsUV(sv) ? SvUVX(sv) : (UV)SvIVX(sv);
2405 =for apidoc sv_2nv_flags
2407 Return the num value of an SV, doing any necessary string or integer
2408 conversion. If flags includes SV_GMAGIC, does an mg_get() first.
2409 Normally used via the C<SvNV(sv)> and C<SvNVx(sv)> macros.
2415 Perl_sv_2nv_flags(pTHX_ register SV *const sv, const I32 flags)
2420 if (SvGMAGICAL(sv) || (SvTYPE(sv) == SVt_PVGV && SvVALID(sv))) {
2421 /* FBMs use the same flag bit as SVf_IVisUV, so must let them
2422 cache IVs just in case. */
2423 if (flags & SV_GMAGIC)
2427 if ((SvPOKp(sv) && SvLEN(sv)) && !SvIOKp(sv)) {
2428 if (!SvIOKp(sv) && ckWARN(WARN_NUMERIC) &&
2429 !grok_number(SvPVX_const(sv), SvCUR(sv), NULL))
2431 return Atof(SvPVX_const(sv));
2435 return (NV)SvUVX(sv);
2437 return (NV)SvIVX(sv);
2442 assert(SvTYPE(sv) >= SVt_PVMG);
2443 /* This falls through to the report_uninit near the end of the
2445 } else if (SvTHINKFIRST(sv)) {
2450 if (flags & SV_SKIP_OVERLOAD)
2452 tmpstr = AMG_CALLun(sv,numer);
2453 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2454 return SvNV(tmpstr);
2457 return PTR2NV(SvRV(sv));
2460 sv_force_normal_flags(sv, 0);
2462 if (SvREADONLY(sv) && !SvOK(sv)) {
2463 if (ckWARN(WARN_UNINITIALIZED))
2468 if (SvTYPE(sv) < SVt_NV) {
2469 /* The logic to use SVt_PVNV if necessary is in sv_upgrade. */
2470 sv_upgrade(sv, SVt_NV);
2471 #ifdef USE_LONG_DOUBLE
2473 STORE_NUMERIC_LOCAL_SET_STANDARD();
2474 PerlIO_printf(Perl_debug_log,
2475 "0x%"UVxf" num(%" PERL_PRIgldbl ")\n",
2476 PTR2UV(sv), SvNVX(sv));
2477 RESTORE_NUMERIC_LOCAL();
2481 STORE_NUMERIC_LOCAL_SET_STANDARD();
2482 PerlIO_printf(Perl_debug_log, "0x%"UVxf" num(%"NVgf")\n",
2483 PTR2UV(sv), SvNVX(sv));
2484 RESTORE_NUMERIC_LOCAL();
2488 else if (SvTYPE(sv) < SVt_PVNV)
2489 sv_upgrade(sv, SVt_PVNV);
2494 SvNV_set(sv, SvIsUV(sv) ? (NV)SvUVX(sv) : (NV)SvIVX(sv));
2495 #ifdef NV_PRESERVES_UV
2501 /* Only set the public NV OK flag if this NV preserves the IV */
2502 /* Check it's not 0xFFFFFFFFFFFFFFFF */
2504 SvIsUV(sv) ? ((SvUVX(sv) != UV_MAX)&&(SvUVX(sv) == U_V(SvNVX(sv))))
2505 : (SvIVX(sv) == I_V(SvNVX(sv))))
2511 else if (SvPOKp(sv) && SvLEN(sv)) {
2513 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), &value);
2514 if (!SvIOKp(sv) && !numtype && ckWARN(WARN_NUMERIC))
2516 #ifdef NV_PRESERVES_UV
2517 if ((numtype & (IS_NUMBER_IN_UV | IS_NUMBER_NOT_INT))
2518 == IS_NUMBER_IN_UV) {
2519 /* It's definitely an integer */
2520 SvNV_set(sv, (numtype & IS_NUMBER_NEG) ? -(NV)value : (NV)value);
2522 SvNV_set(sv, Atof(SvPVX_const(sv)));
2528 SvNV_set(sv, Atof(SvPVX_const(sv)));
2529 /* Only set the public NV OK flag if this NV preserves the value in
2530 the PV at least as well as an IV/UV would.
2531 Not sure how to do this 100% reliably. */
2532 /* if that shift count is out of range then Configure's test is
2533 wonky. We shouldn't be in here with NV_PRESERVES_UV_BITS ==
2535 if (((UV)1 << NV_PRESERVES_UV_BITS) >
2536 U_V(SvNVX(sv) > 0 ? SvNVX(sv) : -SvNVX(sv))) {
2537 SvNOK_on(sv); /* Definitely small enough to preserve all bits */
2538 } else if (!(numtype & IS_NUMBER_IN_UV)) {
2539 /* Can't use strtol etc to convert this string, so don't try.
2540 sv_2iv and sv_2uv will use the NV to convert, not the PV. */
2543 /* value has been set. It may not be precise. */
2544 if ((numtype & IS_NUMBER_NEG) && (value > (UV)IV_MIN)) {
2545 /* 2s complement assumption for (UV)IV_MIN */
2546 SvNOK_on(sv); /* Integer is too negative. */
2551 if (numtype & IS_NUMBER_NEG) {
2552 SvIV_set(sv, -(IV)value);
2553 } else if (value <= (UV)IV_MAX) {
2554 SvIV_set(sv, (IV)value);
2556 SvUV_set(sv, value);
2560 if (numtype & IS_NUMBER_NOT_INT) {
2561 /* I believe that even if the original PV had decimals,
2562 they are lost beyond the limit of the FP precision.
2563 However, neither is canonical, so both only get p
2564 flags. NWC, 2000/11/25 */
2565 /* Both already have p flags, so do nothing */
2567 const NV nv = SvNVX(sv);
2568 if (SvNVX(sv) < (NV)IV_MAX + 0.5) {
2569 if (SvIVX(sv) == I_V(nv)) {
2572 /* It had no "." so it must be integer. */
2576 /* between IV_MAX and NV(UV_MAX).
2577 Could be slightly > UV_MAX */
2579 if (numtype & IS_NUMBER_NOT_INT) {
2580 /* UV and NV both imprecise. */
2582 const UV nv_as_uv = U_V(nv);
2584 if (value == nv_as_uv && SvUVX(sv) != UV_MAX) {
2593 /* It might be more code efficient to go through the entire logic above
2594 and conditionally set with SvNOKp_on() rather than SvNOK(), but it
2595 gets complex and potentially buggy, so more programmer efficient
2596 to do it this way, by turning off the public flags: */
2598 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK);
2599 #endif /* NV_PRESERVES_UV */
2602 if (isGV_with_GP(sv)) {
2603 glob_2number(MUTABLE_GV(sv));
2607 if (!PL_localizing && !(SvFLAGS(sv) & SVs_PADTMP) && ckWARN(WARN_UNINITIALIZED))
2609 assert (SvTYPE(sv) >= SVt_NV);
2610 /* Typically the caller expects that sv_any is not NULL now. */
2611 /* XXX Ilya implies that this is a bug in callers that assume this
2612 and ideally should be fixed. */
2615 #if defined(USE_LONG_DOUBLE)
2617 STORE_NUMERIC_LOCAL_SET_STANDARD();
2618 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2nv(%" PERL_PRIgldbl ")\n",
2619 PTR2UV(sv), SvNVX(sv));
2620 RESTORE_NUMERIC_LOCAL();
2624 STORE_NUMERIC_LOCAL_SET_STANDARD();
2625 PerlIO_printf(Perl_debug_log, "0x%"UVxf" 1nv(%"NVgf")\n",
2626 PTR2UV(sv), SvNVX(sv));
2627 RESTORE_NUMERIC_LOCAL();
2636 Return an SV with the numeric value of the source SV, doing any necessary
2637 reference or overload conversion. You must use the C<SvNUM(sv)> macro to
2638 access this function.
2644 Perl_sv_2num(pTHX_ register SV *const sv)
2646 PERL_ARGS_ASSERT_SV_2NUM;
2651 SV * const tmpsv = AMG_CALLun(sv,numer);
2652 TAINT_IF(tmpsv && SvTAINTED(tmpsv));
2653 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
2654 return sv_2num(tmpsv);
2656 return sv_2mortal(newSVuv(PTR2UV(SvRV(sv))));
2659 /* uiv_2buf(): private routine for use by sv_2pv_flags(): print an IV or
2660 * UV as a string towards the end of buf, and return pointers to start and
2663 * We assume that buf is at least TYPE_CHARS(UV) long.
2667 S_uiv_2buf(char *const buf, const IV iv, UV uv, const int is_uv, char **const peob)
2669 char *ptr = buf + TYPE_CHARS(UV);
2670 char * const ebuf = ptr;
2673 PERL_ARGS_ASSERT_UIV_2BUF;
2685 *--ptr = '0' + (char)(uv % 10);
2694 =for apidoc sv_2pv_flags
2696 Returns a pointer to the string value of an SV, and sets *lp to its length.
2697 If flags includes SV_GMAGIC, does an mg_get() first. Coerces sv to a string
2699 Normally invoked via the C<SvPV_flags> macro. C<sv_2pv()> and C<sv_2pv_nomg>
2700 usually end up here too.
2706 Perl_sv_2pv_flags(pTHX_ register SV *const sv, STRLEN *const lp, const I32 flags)
2716 if (SvGMAGICAL(sv)) {
2717 if (flags & SV_GMAGIC)
2722 if (flags & SV_MUTABLE_RETURN)
2723 return SvPVX_mutable(sv);
2724 if (flags & SV_CONST_RETURN)
2725 return (char *)SvPVX_const(sv);
2728 if (SvIOKp(sv) || SvNOKp(sv)) {
2729 char tbuf[64]; /* Must fit sprintf/Gconvert of longest IV/NV */
2734 ? my_snprintf(tbuf, sizeof(tbuf), "%"UVuf, (UV)SvUVX(sv))
2735 : my_snprintf(tbuf, sizeof(tbuf), "%"IVdf, (IV)SvIVX(sv));
2737 Gconvert(SvNVX(sv), NV_DIG, 0, tbuf);
2744 if (len == 2 && tbuf[0] == '-' && tbuf[1] == '0') {
2749 SvUPGRADE(sv, SVt_PV);
2752 s = SvGROW_mutable(sv, len + 1);
2755 return (char*)memcpy(s, tbuf, len + 1);
2761 assert(SvTYPE(sv) >= SVt_PVMG);
2762 /* This falls through to the report_uninit near the end of the
2764 } else if (SvTHINKFIRST(sv)) {
2769 if (flags & SV_SKIP_OVERLOAD)
2771 tmpstr = AMG_CALLun(sv,string);
2772 TAINT_IF(tmpstr && SvTAINTED(tmpstr));
2773 if (tmpstr && (!SvROK(tmpstr) || (SvRV(tmpstr) != SvRV(sv)))) {
2775 /* char *pv = lp ? SvPV(tmpstr, *lp) : SvPV_nolen(tmpstr);
2779 if ((SvFLAGS(tmpstr) & (SVf_POK)) == SVf_POK) {
2780 if (flags & SV_CONST_RETURN) {
2781 pv = (char *) SvPVX_const(tmpstr);
2783 pv = (flags & SV_MUTABLE_RETURN)
2784 ? SvPVX_mutable(tmpstr) : SvPVX(tmpstr);
2787 *lp = SvCUR(tmpstr);
2789 pv = sv_2pv_flags(tmpstr, lp, flags);
2802 SV *const referent = SvRV(sv);
2806 retval = buffer = savepvn("NULLREF", len);
2807 } else if (SvTYPE(referent) == SVt_REGEXP) {
2808 REGEXP * const re = (REGEXP *)MUTABLE_PTR(referent);
2813 /* If the regex is UTF-8 we want the containing scalar to
2814 have an UTF-8 flag too */
2820 if ((seen_evals = RX_SEEN_EVALS(re)))
2821 PL_reginterp_cnt += seen_evals;
2824 *lp = RX_WRAPLEN(re);
2826 return RX_WRAPPED(re);
2828 const char *const typestr = sv_reftype(referent, 0);
2829 const STRLEN typelen = strlen(typestr);
2830 UV addr = PTR2UV(referent);
2831 const char *stashname = NULL;
2832 STRLEN stashnamelen = 0; /* hush, gcc */
2833 const char *buffer_end;
2835 if (SvOBJECT(referent)) {
2836 const HEK *const name = HvNAME_HEK(SvSTASH(referent));
2839 stashname = HEK_KEY(name);
2840 stashnamelen = HEK_LEN(name);
2842 if (HEK_UTF8(name)) {
2848 stashname = "__ANON__";
2851 len = stashnamelen + 1 /* = */ + typelen + 3 /* (0x */
2852 + 2 * sizeof(UV) + 2 /* )\0 */;
2854 len = typelen + 3 /* (0x */
2855 + 2 * sizeof(UV) + 2 /* )\0 */;
2858 Newx(buffer, len, char);
2859 buffer_end = retval = buffer + len;
2861 /* Working backwards */
2865 *--retval = PL_hexdigit[addr & 15];
2866 } while (addr >>= 4);
2872 memcpy(retval, typestr, typelen);
2876 retval -= stashnamelen;
2877 memcpy(retval, stashname, stashnamelen);
2879 /* retval may not neccesarily have reached the start of the
2881 assert (retval >= buffer);
2883 len = buffer_end - retval - 1; /* -1 for that \0 */
2891 if (SvREADONLY(sv) && !SvOK(sv)) {
2894 if (flags & SV_UNDEF_RETURNS_NULL)
2896 if (ckWARN(WARN_UNINITIALIZED))
2901 if (SvIOK(sv) || ((SvIOKp(sv) && !SvNOKp(sv)))) {
2902 /* I'm assuming that if both IV and NV are equally valid then
2903 converting the IV is going to be more efficient */
2904 const U32 isUIOK = SvIsUV(sv);
2905 char buf[TYPE_CHARS(UV)];
2909 if (SvTYPE(sv) < SVt_PVIV)
2910 sv_upgrade(sv, SVt_PVIV);
2911 ptr = uiv_2buf(buf, SvIVX(sv), SvUVX(sv), isUIOK, &ebuf);
2913 /* inlined from sv_setpvn */
2914 s = SvGROW_mutable(sv, len + 1);
2915 Move(ptr, s, len, char);
2919 else if (SvNOKp(sv)) {
2921 if (SvTYPE(sv) < SVt_PVNV)
2922 sv_upgrade(sv, SVt_PVNV);
2923 /* The +20 is pure guesswork. Configure test needed. --jhi */
2924 s = SvGROW_mutable(sv, NV_DIG + 20);
2925 /* some Xenix systems wipe out errno here */
2927 if (SvNVX(sv) == 0.0)
2928 my_strlcpy(s, "0", SvLEN(sv));
2932 Gconvert(SvNVX(sv), NV_DIG, 0, s);
2935 if (*s == '-' && s[1] == '0' && !s[2]) {
2946 if (isGV_with_GP(sv)) {
2947 GV *const gv = MUTABLE_GV(sv);
2948 const U32 wasfake = SvFLAGS(gv) & SVf_FAKE;
2949 SV *const buffer = sv_newmortal();
2951 /* FAKE globs can get coerced, so need to turn this off temporarily
2954 gv_efullname3(buffer, gv, "*");
2955 SvFLAGS(gv) |= wasfake;
2957 if (SvPOK(buffer)) {
2959 *lp = SvCUR(buffer);
2961 return SvPVX(buffer);
2972 if (flags & SV_UNDEF_RETURNS_NULL)
2974 if (!PL_localizing && !(SvFLAGS(sv) & SVs_PADTMP) && ckWARN(WARN_UNINITIALIZED))
2976 if (SvTYPE(sv) < SVt_PV)
2977 /* Typically the caller expects that sv_any is not NULL now. */
2978 sv_upgrade(sv, SVt_PV);
2982 const STRLEN len = s - SvPVX_const(sv);
2988 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
2989 PTR2UV(sv),SvPVX_const(sv)));
2990 if (flags & SV_CONST_RETURN)
2991 return (char *)SvPVX_const(sv);
2992 if (flags & SV_MUTABLE_RETURN)
2993 return SvPVX_mutable(sv);
2998 =for apidoc sv_copypv
3000 Copies a stringified representation of the source SV into the
3001 destination SV. Automatically performs any necessary mg_get and
3002 coercion of numeric values into strings. Guaranteed to preserve
3003 UTF8 flag even from overloaded objects. Similar in nature to
3004 sv_2pv[_flags] but operates directly on an SV instead of just the
3005 string. Mostly uses sv_2pv_flags to do its work, except when that
3006 would lose the UTF-8'ness of the PV.
3012 Perl_sv_copypv(pTHX_ SV *const dsv, register SV *const ssv)
3015 const char * const s = SvPV_const(ssv,len);
3017 PERL_ARGS_ASSERT_SV_COPYPV;
3019 sv_setpvn(dsv,s,len);
3027 =for apidoc sv_2pvbyte
3029 Return a pointer to the byte-encoded representation of the SV, and set *lp
3030 to its length. May cause the SV to be downgraded from UTF-8 as a
3033 Usually accessed via the C<SvPVbyte> macro.
3039 Perl_sv_2pvbyte(pTHX_ register SV *const sv, STRLEN *const lp)
3041 PERL_ARGS_ASSERT_SV_2PVBYTE;
3043 sv_utf8_downgrade(sv,0);
3044 return lp ? SvPV(sv,*lp) : SvPV_nolen(sv);
3048 =for apidoc sv_2pvutf8
3050 Return a pointer to the UTF-8-encoded representation of the SV, and set *lp
3051 to its length. May cause the SV to be upgraded to UTF-8 as a side-effect.
3053 Usually accessed via the C<SvPVutf8> macro.
3059 Perl_sv_2pvutf8(pTHX_ register SV *const sv, STRLEN *const lp)
3061 PERL_ARGS_ASSERT_SV_2PVUTF8;
3063 sv_utf8_upgrade(sv);
3064 return lp ? SvPV(sv,*lp) : SvPV_nolen(sv);
3069 =for apidoc sv_2bool
3071 This macro is only used by sv_true() or its macro equivalent, and only if
3072 the latter's argument is neither SvPOK, SvIOK nor SvNOK.
3073 It calls sv_2bool_flags with the SV_GMAGIC flag.
3075 =for apidoc sv_2bool_flags
3077 This function is only used by sv_true() and friends, and only if
3078 the latter's argument is neither SvPOK, SvIOK nor SvNOK. If the flags
3079 contain SV_GMAGIC, then it does an mg_get() first.
3086 Perl_sv_2bool_flags(pTHX_ register SV *const sv, const I32 flags)
3090 PERL_ARGS_ASSERT_SV_2BOOL_FLAGS;
3092 if(flags & SV_GMAGIC) SvGETMAGIC(sv);
3098 SV * const tmpsv = AMG_CALLun(sv,bool_);
3099 if (tmpsv && (!SvROK(tmpsv) || (SvRV(tmpsv) != SvRV(sv))))
3100 return cBOOL(SvTRUE(tmpsv));
3102 return SvRV(sv) != 0;
3105 register XPV* const Xpvtmp = (XPV*)SvANY(sv);
3107 (*sv->sv_u.svu_pv > '0' ||
3108 Xpvtmp->xpv_cur > 1 ||
3109 (Xpvtmp->xpv_cur && *sv->sv_u.svu_pv != '0')))
3116 return SvIVX(sv) != 0;
3119 return SvNVX(sv) != 0.0;
3121 if (isGV_with_GP(sv))
3131 =for apidoc sv_utf8_upgrade
3133 Converts the PV of an SV to its UTF-8-encoded form.
3134 Forces the SV to string form if it is not already.
3135 Will C<mg_get> on C<sv> if appropriate.
3136 Always sets the SvUTF8 flag to avoid future validity checks even
3137 if the whole string is the same in UTF-8 as not.
3138 Returns the number of bytes in the converted string
3140 This is not as a general purpose byte encoding to Unicode interface:
3141 use the Encode extension for that.
3143 =for apidoc sv_utf8_upgrade_nomg
3145 Like sv_utf8_upgrade, but doesn't do magic on C<sv>
3147 =for apidoc sv_utf8_upgrade_flags
3149 Converts the PV of an SV to its UTF-8-encoded form.
3150 Forces the SV to string form if it is not already.
3151 Always sets the SvUTF8 flag to avoid future validity checks even
3152 if all the bytes are invariant in UTF-8. If C<flags> has C<SV_GMAGIC> bit set,
3153 will C<mg_get> on C<sv> if appropriate, else not.
3154 Returns the number of bytes in the converted string
3155 C<sv_utf8_upgrade> and
3156 C<sv_utf8_upgrade_nomg> are implemented in terms of this function.
3158 This is not as a general purpose byte encoding to Unicode interface:
3159 use the Encode extension for that.
3163 The grow version is currently not externally documented. It adds a parameter,
3164 extra, which is the number of unused bytes the string of 'sv' is guaranteed to
3165 have free after it upon return. This allows the caller to reserve extra space
3166 that it intends to fill, to avoid extra grows.
3168 Also externally undocumented for the moment is the flag SV_FORCE_UTF8_UPGRADE,
3169 which can be used to tell this function to not first check to see if there are
3170 any characters that are different in UTF-8 (variant characters) which would
3171 force it to allocate a new string to sv, but to assume there are. Typically
3172 this flag is used by a routine that has already parsed the string to find that
3173 there are such characters, and passes this information on so that the work
3174 doesn't have to be repeated.
3176 (One might think that the calling routine could pass in the position of the
3177 first such variant, so it wouldn't have to be found again. But that is not the
3178 case, because typically when the caller is likely to use this flag, it won't be
3179 calling this routine unless it finds something that won't fit into a byte.
3180 Otherwise it tries to not upgrade and just use bytes. But some things that
3181 do fit into a byte are variants in utf8, and the caller may not have been
3182 keeping track of these.)
3184 If the routine itself changes the string, it adds a trailing NUL. Such a NUL
3185 isn't guaranteed due to having other routines do the work in some input cases,
3186 or if the input is already flagged as being in utf8.
3188 The speed of this could perhaps be improved for many cases if someone wanted to
3189 write a fast function that counts the number of variant characters in a string,
3190 especially if it could return the position of the first one.
3195 Perl_sv_utf8_upgrade_flags_grow(pTHX_ register SV *const sv, const I32 flags, STRLEN extra)
3199 PERL_ARGS_ASSERT_SV_UTF8_UPGRADE_FLAGS_GROW;
3201 if (sv == &PL_sv_undef)
3205 if (SvREADONLY(sv) && (SvPOKp(sv) || SvIOKp(sv) || SvNOKp(sv))) {
3206 (void) sv_2pv_flags(sv,&len, flags);
3208 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3212 (void) SvPV_force(sv,len);
3217 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3222 sv_force_normal_flags(sv, 0);
3225 if (PL_encoding && !(flags & SV_UTF8_NO_ENCODING)) {
3226 sv_recode_to_utf8(sv, PL_encoding);
3227 if (extra) SvGROW(sv, SvCUR(sv) + extra);
3231 if (SvCUR(sv) == 0) {
3232 if (extra) SvGROW(sv, extra);
3233 } else { /* Assume Latin-1/EBCDIC */
3234 /* This function could be much more efficient if we
3235 * had a FLAG in SVs to signal if there are any variant
3236 * chars in the PV. Given that there isn't such a flag
3237 * make the loop as fast as possible (although there are certainly ways
3238 * to speed this up, eg. through vectorization) */
3239 U8 * s = (U8 *) SvPVX_const(sv);
3240 U8 * e = (U8 *) SvEND(sv);
3242 STRLEN two_byte_count = 0;
3244 if (flags & SV_FORCE_UTF8_UPGRADE) goto must_be_utf8;
3246 /* See if really will need to convert to utf8. We mustn't rely on our
3247 * incoming SV being well formed and having a trailing '\0', as certain
3248 * code in pp_formline can send us partially built SVs. */
3252 if (NATIVE_IS_INVARIANT(ch)) continue;
3254 t--; /* t already incremented; re-point to first variant */
3259 /* utf8 conversion not needed because all are invariants. Mark as
3260 * UTF-8 even if no variant - saves scanning loop */
3266 /* Here, the string should be converted to utf8, either because of an
3267 * input flag (two_byte_count = 0), or because a character that
3268 * requires 2 bytes was found (two_byte_count = 1). t points either to
3269 * the beginning of the string (if we didn't examine anything), or to
3270 * the first variant. In either case, everything from s to t - 1 will
3271 * occupy only 1 byte each on output.
3273 * There are two main ways to convert. One is to create a new string
3274 * and go through the input starting from the beginning, appending each
3275 * converted value onto the new string as we go along. It's probably
3276 * best to allocate enough space in the string for the worst possible
3277 * case rather than possibly running out of space and having to
3278 * reallocate and then copy what we've done so far. Since everything
3279 * from s to t - 1 is invariant, the destination can be initialized
3280 * with these using a fast memory copy
3282 * The other way is to figure out exactly how big the string should be
3283 * by parsing the entire input. Then you don't have to make it big
3284 * enough to handle the worst possible case, and more importantly, if
3285 * the string you already have is large enough, you don't have to
3286 * allocate a new string, you can copy the last character in the input
3287 * string to the final position(s) that will be occupied by the
3288 * converted string and go backwards, stopping at t, since everything
3289 * before that is invariant.
3291 * There are advantages and disadvantages to each method.
3293 * In the first method, we can allocate a new string, do the memory
3294 * copy from the s to t - 1, and then proceed through the rest of the
3295 * string byte-by-byte.
3297 * In the second method, we proceed through the rest of the input
3298 * string just calculating how big the converted string will be. Then
3299 * there are two cases:
3300 * 1) if the string has enough extra space to handle the converted
3301 * value. We go backwards through the string, converting until we
3302 * get to the position we are at now, and then stop. If this
3303 * position is far enough along in the string, this method is
3304 * faster than the other method. If the memory copy were the same
3305 * speed as the byte-by-byte loop, that position would be about
3306 * half-way, as at the half-way mark, parsing to the end and back
3307 * is one complete string's parse, the same amount as starting
3308 * over and going all the way through. Actually, it would be
3309 * somewhat less than half-way, as it's faster to just count bytes
3310 * than to also copy, and we don't have the overhead of allocating
3311 * a new string, changing the scalar to use it, and freeing the
3312 * existing one. But if the memory copy is fast, the break-even
3313 * point is somewhere after half way. The counting loop could be
3314 * sped up by vectorization, etc, to move the break-even point
3315 * further towards the beginning.
3316 * 2) if the string doesn't have enough space to handle the converted
3317 * value. A new string will have to be allocated, and one might
3318 * as well, given that, start from the beginning doing the first
3319 * method. We've spent extra time parsing the string and in
3320 * exchange all we've gotten is that we know precisely how big to
3321 * make the new one. Perl is more optimized for time than space,
3322 * so this case is a loser.
3323 * So what I've decided to do is not use the 2nd method unless it is
3324 * guaranteed that a new string won't have to be allocated, assuming
3325 * the worst case. I also decided not to put any more conditions on it
3326 * than this, for now. It seems likely that, since the worst case is
3327 * twice as big as the unknown portion of the string (plus 1), we won't
3328 * be guaranteed enough space, causing us to go to the first method,
3329 * unless the string is short, or the first variant character is near
3330 * the end of it. In either of these cases, it seems best to use the
3331 * 2nd method. The only circumstance I can think of where this would
3332 * be really slower is if the string had once had much more data in it
3333 * than it does now, but there is still a substantial amount in it */
3336 STRLEN invariant_head = t - s;
3337 STRLEN size = invariant_head + (e - t) * 2 + 1 + extra;
3338 if (SvLEN(sv) < size) {
3340 /* Here, have decided to allocate a new string */
3345 Newx(dst, size, U8);
3347 /* If no known invariants at the beginning of the input string,
3348 * set so starts from there. Otherwise, can use memory copy to
3349 * get up to where we are now, and then start from here */
3351 if (invariant_head <= 0) {
3354 Copy(s, dst, invariant_head, char);
3355 d = dst + invariant_head;
3359 const UV uv = NATIVE8_TO_UNI(*t++);
3360 if (UNI_IS_INVARIANT(uv))
3361 *d++ = (U8)UNI_TO_NATIVE(uv);
3363 *d++ = (U8)UTF8_EIGHT_BIT_HI(uv);
3364 *d++ = (U8)UTF8_EIGHT_BIT_LO(uv);
3368 SvPV_free(sv); /* No longer using pre-existing string */
3369 SvPV_set(sv, (char*)dst);
3370 SvCUR_set(sv, d - dst);
3371 SvLEN_set(sv, size);
3374 /* Here, have decided to get the exact size of the string.
3375 * Currently this happens only when we know that there is
3376 * guaranteed enough space to fit the converted string, so
3377 * don't have to worry about growing. If two_byte_count is 0,
3378 * then t points to the first byte of the string which hasn't
3379 * been examined yet. Otherwise two_byte_count is 1, and t
3380 * points to the first byte in the string that will expand to
3381 * two. Depending on this, start examining at t or 1 after t.
3384 U8 *d = t + two_byte_count;
3387 /* Count up the remaining bytes that expand to two */
3390 const U8 chr = *d++;
3391 if (! NATIVE_IS_INVARIANT(chr)) two_byte_count++;
3394 /* The string will expand by just the number of bytes that
3395 * occupy two positions. But we are one afterwards because of
3396 * the increment just above. This is the place to put the
3397 * trailing NUL, and to set the length before we decrement */
3399 d += two_byte_count;
3400 SvCUR_set(sv, d - s);
3404 /* Having decremented d, it points to the position to put the
3405 * very last byte of the expanded string. Go backwards through
3406 * the string, copying and expanding as we go, stopping when we
3407 * get to the part that is invariant the rest of the way down */
3411 const U8 ch = NATIVE8_TO_UNI(*e--);
3412 if (UNI_IS_INVARIANT(ch)) {
3413 *d-- = UNI_TO_NATIVE(ch);
3415 *d-- = (U8)UTF8_EIGHT_BIT_LO(ch);
3416 *d-- = (U8)UTF8_EIGHT_BIT_HI(ch);
3423 /* Mark as UTF-8 even if no variant - saves scanning loop */
3429 =for apidoc sv_utf8_downgrade
3431 Attempts to convert the PV of an SV from characters to bytes.
3432 If the PV contains a character that cannot fit
3433 in a byte, this conversion will fail;
3434 in this case, either returns false or, if C<fail_ok> is not
3437 This is not as a general purpose Unicode to byte encoding interface:
3438 use the Encode extension for that.
3444 Perl_sv_utf8_downgrade(pTHX_ register SV *const sv, const bool fail_ok)
3448 PERL_ARGS_ASSERT_SV_UTF8_DOWNGRADE;
3450 if (SvPOKp(sv) && SvUTF8(sv)) {
3456 sv_force_normal_flags(sv, 0);
3458 s = (U8 *) SvPV(sv, len);
3459 if (!utf8_to_bytes(s, &len)) {
3464 Perl_croak(aTHX_ "Wide character in %s",
3467 Perl_croak(aTHX_ "Wide character");
3478 =for apidoc sv_utf8_encode
3480 Converts the PV of an SV to UTF-8, but then turns the C<SvUTF8>
3481 flag off so that it looks like octets again.
3487 Perl_sv_utf8_encode(pTHX_ register SV *const sv)
3489 PERL_ARGS_ASSERT_SV_UTF8_ENCODE;
3492 sv_force_normal_flags(sv, 0);
3494 if (SvREADONLY(sv)) {
3495 Perl_croak_no_modify(aTHX);
3497 (void) sv_utf8_upgrade(sv);
3502 =for apidoc sv_utf8_decode
3504 If the PV of the SV is an octet sequence in UTF-8
3505 and contains a multiple-byte character, the C<SvUTF8> flag is turned on
3506 so that it looks like a character. If the PV contains only single-byte
3507 characters, the C<SvUTF8> flag stays being off.
3508 Scans PV for validity and returns false if the PV is invalid UTF-8.
3514 Perl_sv_utf8_decode(pTHX_ register SV *const sv)
3516 PERL_ARGS_ASSERT_SV_UTF8_DECODE;
3522 /* The octets may have got themselves encoded - get them back as
3525 if (!sv_utf8_downgrade(sv, TRUE))
3528 /* it is actually just a matter of turning the utf8 flag on, but
3529 * we want to make sure everything inside is valid utf8 first.
3531 c = (const U8 *) SvPVX_const(sv);
3532 if (!is_utf8_string(c, SvCUR(sv)+1))
3534 e = (const U8 *) SvEND(sv);
3537 if (!UTF8_IS_INVARIANT(ch)) {
3547 =for apidoc sv_setsv
3549 Copies the contents of the source SV C<ssv> into the destination SV
3550 C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3551 function if the source SV needs to be reused. Does not handle 'set' magic.
3552 Loosely speaking, it performs a copy-by-value, obliterating any previous
3553 content of the destination.
3555 You probably want to use one of the assortment of wrappers, such as
3556 C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3557 C<SvSetMagicSV_nosteal>.
3559 =for apidoc sv_setsv_flags
3561 Copies the contents of the source SV C<ssv> into the destination SV
3562 C<dsv>. The source SV may be destroyed if it is mortal, so don't use this
3563 function if the source SV needs to be reused. Does not handle 'set' magic.
3564 Loosely speaking, it performs a copy-by-value, obliterating any previous
3565 content of the destination.
3566 If the C<flags> parameter has the C<SV_GMAGIC> bit set, will C<mg_get> on
3567 C<ssv> if appropriate, else not. If the C<flags> parameter has the
3568 C<NOSTEAL> bit set then the buffers of temps will not be stolen. <sv_setsv>
3569 and C<sv_setsv_nomg> are implemented in terms of this function.
3571 You probably want to use one of the assortment of wrappers, such as
3572 C<SvSetSV>, C<SvSetSV_nosteal>, C<SvSetMagicSV> and
3573 C<SvSetMagicSV_nosteal>.
3575 This is the primary function for copying scalars, and most other
3576 copy-ish functions and macros use this underneath.
3582 S_glob_assign_glob(pTHX_ SV *const dstr, SV *const sstr, const int dtype)
3584 I32 mro_changes = 0; /* 1 = method, 2 = isa, 3 = recursive isa */
3585 HV *old_stash = NULL;
3587 PERL_ARGS_ASSERT_GLOB_ASSIGN_GLOB;
3589 if (dtype != SVt_PVGV && !isGV_with_GP(dstr)) {
3590 const char * const name = GvNAME(sstr);
3591 const STRLEN len = GvNAMELEN(sstr);
3593 if (dtype >= SVt_PV) {
3599 SvUPGRADE(dstr, SVt_PVGV);
3600 (void)SvOK_off(dstr);
3601 /* FIXME - why are we doing this, then turning it off and on again
3603 isGV_with_GP_on(dstr);
3605 GvSTASH(dstr) = GvSTASH(sstr);
3607 Perl_sv_add_backref(aTHX_ MUTABLE_SV(GvSTASH(dstr)), dstr);
3608 gv_name_set(MUTABLE_GV(dstr), name, len, GV_ADD);
3609 SvFAKE_on(dstr); /* can coerce to non-glob */
3612 if(GvGP(MUTABLE_GV(sstr))) {
3613 /* If source has method cache entry, clear it */
3615 SvREFCNT_dec(GvCV(sstr));
3619 /* If source has a real method, then a method is
3621 else if(GvCV((const GV *)sstr)) {
3626 /* If dest already had a real method, that's a change as well */
3627 if(!mro_changes && GvGP(MUTABLE_GV(dstr)) && GvCVu((const GV *)dstr)) {
3631 /* We don’t need to check the name of the destination if it was not a
3632 glob to begin with. */
3633 if(dtype == SVt_PVGV) {
3634 const char * const name = GvNAME((const GV *)dstr);
3635 if(strEQ(name,"ISA"))
3638 const STRLEN len = GvNAMELEN(dstr);
3639 if (len > 1 && name[len-2] == ':' && name[len-1] == ':') {
3642 /* Set aside the old stash, so we can reset isa caches on
3644 old_stash = GvHV(dstr);
3649 gp_free(MUTABLE_GV(dstr));
3650 isGV_with_GP_off(dstr);
3651 (void)SvOK_off(dstr);
3652 isGV_with_GP_on(dstr);
3653 GvINTRO_off(dstr); /* one-shot flag */
3654 GvGP(dstr) = gp_ref(GvGP(sstr));
3655 if (SvTAINTED(sstr))
3657 if (GvIMPORTED(dstr) != GVf_IMPORTED
3658 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3660 GvIMPORTED_on(dstr);
3663 if(mro_changes == 2) mro_isa_changed_in(GvSTASH(dstr));
3664 else if(mro_changes == 3) {
3665 const HV * const stash = GvHV(dstr);
3666 if(stash && HvNAME(stash)) mro_package_moved(stash);
3667 if(old_stash && HvNAME(old_stash)) mro_package_moved(old_stash);
3669 else if(mro_changes) mro_method_changed_in(GvSTASH(dstr));
3674 S_glob_assign_ref(pTHX_ SV *const dstr, SV *const sstr)
3676 SV * const sref = SvREFCNT_inc(SvRV(sstr));
3678 const int intro = GvINTRO(dstr);
3681 const U32 stype = SvTYPE(sref);
3683 PERL_ARGS_ASSERT_GLOB_ASSIGN_REF;
3686 GvINTRO_off(dstr); /* one-shot flag */
3687 GvLINE(dstr) = CopLINE(PL_curcop);
3688 GvEGV(dstr) = MUTABLE_GV(dstr);
3693 location = (SV **) &GvCV(dstr);
3694 import_flag = GVf_IMPORTED_CV;
3697 location = (SV **) &GvHV(dstr);
3698 import_flag = GVf_IMPORTED_HV;
3701 location = (SV **) &GvAV(dstr);
3702 import_flag = GVf_IMPORTED_AV;
3705 location = (SV **) &GvIOp(dstr);
3708 location = (SV **) &GvFORM(dstr);
3711 location = &GvSV(dstr);
3712 import_flag = GVf_IMPORTED_SV;
3715 if (stype == SVt_PVCV) {
3716 /*if (GvCVGEN(dstr) && (GvCV(dstr) != (const CV *)sref || GvCVGEN(dstr))) {*/
3717 if (GvCVGEN(dstr)) {
3718 SvREFCNT_dec(GvCV(dstr));
3720 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3723 SAVEGENERICSV(*location);
3727 if (stype == SVt_PVCV && (*location != sref || GvCVGEN(dstr))) {
3728 CV* const cv = MUTABLE_CV(*location);
3730 if (!GvCVGEN((const GV *)dstr) &&
3731 (CvROOT(cv) || CvXSUB(cv)))
3733 /* Redefining a sub - warning is mandatory if
3734 it was a const and its value changed. */
3735 if (CvCONST(cv) && CvCONST((const CV *)sref)
3737 == cv_const_sv((const CV *)sref)) {
3739 /* They are 2 constant subroutines generated from
3740 the same constant. This probably means that
3741 they are really the "same" proxy subroutine
3742 instantiated in 2 places. Most likely this is
3743 when a constant is exported twice. Don't warn.
3746 else if (ckWARN(WARN_REDEFINE)
3748 && (!CvCONST((const CV *)sref)
3749 || sv_cmp(cv_const_sv(cv),
3750 cv_const_sv((const CV *)
3752 Perl_warner(aTHX_ packWARN(WARN_REDEFINE),
3755 ? "Constant subroutine %s::%s redefined"
3756 : "Subroutine %s::%s redefined"),
3757 HvNAME_get(GvSTASH((const GV *)dstr)),
3758 GvENAME(MUTABLE_GV(dstr)));
3762 cv_ckproto_len(cv, (const GV *)dstr,
3763 SvPOK(sref) ? SvPVX_const(sref) : NULL,
3764 SvPOK(sref) ? SvCUR(sref) : 0);
3766 GvCVGEN(dstr) = 0; /* Switch off cacheness. */
3767 GvASSUMECV_on(dstr);
3768 if(GvSTASH(dstr)) mro_method_changed_in(GvSTASH(dstr)); /* sub foo { 1 } sub bar { 2 } *bar = \&foo */
3771 if (import_flag && !(GvFLAGS(dstr) & import_flag)
3772 && CopSTASH_ne(PL_curcop, GvSTASH(dstr))) {
3773 GvFLAGS(dstr) |= import_flag;
3775 if (stype == SVt_PVHV) {
3776 const char * const name = GvNAME((GV*)dstr);
3777 const STRLEN len = GvNAMELEN(dstr);
3778 if (len > 1 && name[len-2] == ':' && name[len-1] == ':') {
3779 if(HvNAME(dref)) mro_package_moved((HV *)dref);
3780 if(HvNAME(sref)) mro_package_moved((HV *)sref);
3783 else if (stype == SVt_PVAV && strEQ(GvNAME((GV*)dstr), "ISA")) {
3784 sv_magic(sref, dstr, PERL_MAGIC_isa, NULL, 0);
3785 mro_isa_changed_in(GvSTASH(dstr));
3790 if (SvTAINTED(sstr))
3796 Perl_sv_setsv_flags(pTHX_ SV *dstr, register SV* sstr, const I32 flags)
3799 register U32 sflags;
3801 register svtype stype;
3803 PERL_ARGS_ASSERT_SV_SETSV_FLAGS;
3808 if (SvIS_FREED(dstr)) {
3809 Perl_croak(aTHX_ "panic: attempt to copy value %" SVf
3810 " to a freed scalar %p", SVfARG(sstr), (void *)dstr);
3812 SV_CHECK_THINKFIRST_COW_DROP(dstr);
3814 sstr = &PL_sv_undef;
3815 if (SvIS_FREED(sstr)) {
3816 Perl_croak(aTHX_ "panic: attempt to copy freed scalar %p to %p",
3817 (void*)sstr, (void*)dstr);
3819 stype = SvTYPE(sstr);
3820 dtype = SvTYPE(dstr);
3822 (void)SvAMAGIC_off(dstr);
3825 /* need to nuke the magic */
3829 /* There's a lot of redundancy below but we're going for speed here */
3834 if (dtype != SVt_PVGV && dtype != SVt_PVLV) {
3835 (void)SvOK_off(dstr);
3843 sv_upgrade(dstr, SVt_IV);
3847 sv_upgrade(dstr, SVt_PVIV);
3851 goto end_of_first_switch;
3853 (void)SvIOK_only(dstr);
3854 SvIV_set(dstr, SvIVX(sstr));
3857 /* SvTAINTED can only be true if the SV has taint magic, which in
3858 turn means that the SV type is PVMG (or greater). This is the
3859 case statement for SVt_IV, so this cannot be true (whatever gcov
3861 assert(!SvTAINTED(sstr));
3866 if (dtype < SVt_PV && dtype != SVt_IV)
3867 sv_upgrade(dstr, SVt_IV);
3875 sv_upgrade(dstr, SVt_NV);
3879 sv_upgrade(dstr, SVt_PVNV);
3883 goto end_of_first_switch;
3885 SvNV_set(dstr, SvNVX(sstr));
3886 (void)SvNOK_only(dstr);
3887 /* SvTAINTED can only be true if the SV has taint magic, which in
3888 turn means that the SV type is PVMG (or greater). This is the
3889 case statement for SVt_NV, so this cannot be true (whatever gcov
3891 assert(!SvTAINTED(sstr));
3897 #ifdef PERL_OLD_COPY_ON_WRITE
3898 if ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS) {
3899 if (dtype < SVt_PVIV)
3900 sv_upgrade(dstr, SVt_PVIV);
3907 sv_upgrade(dstr, SVt_PV);
3910 if (dtype < SVt_PVIV)
3911 sv_upgrade(dstr, SVt_PVIV);
3914 if (dtype < SVt_PVNV)
3915 sv_upgrade(dstr, SVt_PVNV);
3919 const char * const type = sv_reftype(sstr,0);
3921 Perl_croak(aTHX_ "Bizarre copy of %s in %s", type, OP_DESC(PL_op));
3923 Perl_croak(aTHX_ "Bizarre copy of %s", type);
3928 if (dtype < SVt_REGEXP)
3929 sv_upgrade(dstr, SVt_REGEXP);
3932 /* case SVt_BIND: */
3935 if (isGV_with_GP(sstr) && dtype <= SVt_PVLV) {
3936 glob_assign_glob(dstr, sstr, dtype);
3939 /* SvVALID means that this PVGV is playing at being an FBM. */
3943 if (SvGMAGICAL(sstr) && (flags & SV_GMAGIC)) {
3945 if (SvTYPE(sstr) != stype)
3946 stype = SvTYPE(sstr);
3947 if (isGV_with_GP(sstr) && dtype <= SVt_PVLV) {
3948 glob_assign_glob(dstr, sstr, dtype);
3952 if (stype == SVt_PVLV)
3953 SvUPGRADE(dstr, SVt_PVNV);
3955 SvUPGRADE(dstr, (svtype)stype);
3957 end_of_first_switch:
3959 /* dstr may have been upgraded. */
3960 dtype = SvTYPE(dstr);
3961 sflags = SvFLAGS(sstr);
3963 if (dtype == SVt_PVCV || dtype == SVt_PVFM) {
3964 /* Assigning to a subroutine sets the prototype. */
3967 const char *const ptr = SvPV_const(sstr, len);
3969 SvGROW(dstr, len + 1);
3970 Copy(ptr, SvPVX(dstr), len + 1, char);
3971 SvCUR_set(dstr, len);
3973 SvFLAGS(dstr) |= sflags & SVf_UTF8;
3977 } else if (dtype == SVt_PVAV || dtype == SVt_PVHV) {
3978 const char * const type = sv_reftype(dstr,0);
3980 Perl_croak(aTHX_ "Cannot copy to %s in %s", type, OP_DESC(PL_op));
3982 Perl_croak(aTHX_ "Cannot copy to %s", type);
3983 } else if (sflags & SVf_ROK) {
3984 if (isGV_with_GP(dstr)
3985 && SvTYPE(SvRV(sstr)) == SVt_PVGV && isGV_with_GP(SvRV(sstr))) {
3988 if (GvIMPORTED(dstr) != GVf_IMPORTED
3989 && CopSTASH_ne(PL_curcop, GvSTASH(dstr)))
3991 GvIMPORTED_on(dstr);
3996 glob_assign_glob(dstr, sstr, dtype);
4000 if (dtype >= SVt_PV) {
4001 if (isGV_with_GP(dstr)) {
4002 glob_assign_ref(dstr, sstr);
4005 if (SvPVX_const(dstr)) {
4011 (void)SvOK_off(dstr);
4012 SvRV_set(dstr, SvREFCNT_inc(SvRV(sstr)));
4013 SvFLAGS(dstr) |= sflags & SVf_ROK;
4014 assert(!(sflags & SVp_NOK));
4015 assert(!(sflags & SVp_IOK));
4016 assert(!(sflags & SVf_NOK));
4017 assert(!(sflags & SVf_IOK));
4019 else if (isGV_with_GP(dstr)) {
4020 if (!(sflags & SVf_OK)) {
4021 Perl_ck_warner(aTHX_ packWARN(WARN_MISC),
4022 "Undefined value assigned to typeglob");
4025 GV *gv = gv_fetchsv(sstr, GV_ADD, SVt_PVGV);
4026 if (dstr != (const SV *)gv) {
4027 const char * const name = GvNAME((const GV *)dstr);
4028 const STRLEN len = GvNAMELEN(dstr);
4029 HV *old_stash = NULL;
4030 bool reset_isa = FALSE;
4031 if (len > 1 && name[len-2] == ':' && name[len-1] == ':') {
4032 /* Set aside the old stash, so we can reset isa caches
4033 on its subclasses. */
4034 old_stash = GvHV(dstr);
4039 gp_free(MUTABLE_GV(dstr));
4040 GvGP(dstr) = gp_ref(GvGP(gv));
4043 const HV * const stash = GvHV(dstr);
4044 if(stash && HvNAME(stash)) mro_package_moved(stash);
4045 if(old_stash && HvNAME(old_stash))
4046 mro_package_moved(old_stash);
4051 else if (dtype == SVt_REGEXP && stype == SVt_REGEXP) {
4052 reg_temp_copy((REGEXP*)dstr, (REGEXP*)sstr);
4054 else if (sflags & SVp_POK) {
4058 * Check to see if we can just swipe the string. If so, it's a
4059 * possible small lose on short strings, but a big win on long ones.
4060 * It might even be a win on short strings if SvPVX_const(dstr)
4061 * has to be allocated and SvPVX_const(sstr) has to be freed.
4062 * Likewise if we can set up COW rather than doing an actual copy, we
4063 * drop to the else clause, as the swipe code and the COW setup code
4064 * have much in common.
4067 /* Whichever path we take through the next code, we want this true,
4068 and doing it now facilitates the COW check. */
4069 (void)SvPOK_only(dstr);
4072 /* If we're already COW then this clause is not true, and if COW
4073 is allowed then we drop down to the else and make dest COW
4074 with us. If caller hasn't said that we're allowed to COW
4075 shared hash keys then we don't do the COW setup, even if the
4076 source scalar is a shared hash key scalar. */
4077 (((flags & SV_COW_SHARED_HASH_KEYS)
4078 ? (sflags & (SVf_FAKE|SVf_READONLY)) != (SVf_FAKE|SVf_READONLY)
4079 : 1 /* If making a COW copy is forbidden then the behaviour we
4080 desire is as if the source SV isn't actually already
4081 COW, even if it is. So we act as if the source flags
4082 are not COW, rather than actually testing them. */
4084 #ifndef PERL_OLD_COPY_ON_WRITE
4085 /* The change that added SV_COW_SHARED_HASH_KEYS makes the logic
4086 when PERL_OLD_COPY_ON_WRITE is defined a little wrong.
4087 Conceptually PERL_OLD_COPY_ON_WRITE being defined should
4088 override SV_COW_SHARED_HASH_KEYS, because it means "always COW"
4089 but in turn, it's somewhat dead code, never expected to go
4090 live, but more kept as a placeholder on how to do it better
4091 in a newer implementation. */
4092 /* If we are COW and dstr is a suitable target then we drop down
4093 into the else and make dest a COW of us. */
4094 || (SvFLAGS(dstr) & CAN_COW_MASK) != CAN_COW_FLAGS
4099 (sflags & SVs_TEMP) && /* slated for free anyway? */
4100 !(sflags & SVf_OOK) && /* and not involved in OOK hack? */
4101 (!(flags & SV_NOSTEAL)) &&
4102 /* and we're allowed to steal temps */
4103 SvREFCNT(sstr) == 1 && /* and no other references to it? */
4104 SvLEN(sstr)) /* and really is a string */
4105 #ifdef PERL_OLD_COPY_ON_WRITE
4106 && ((flags & SV_COW_SHARED_HASH_KEYS)
4107 ? (!((sflags & CAN_COW_MASK) == CAN_COW_FLAGS
4108 && (SvFLAGS(dstr) & CAN_COW_MASK) == CAN_COW_FLAGS
4109 && SvTYPE(sstr) >= SVt_PVIV && SvTYPE(sstr) != SVt_PVFM))
4113 /* Failed the swipe test, and it's not a shared hash key either.
4114 Have to copy the string. */
4115 STRLEN len = SvCUR(sstr);
4116 SvGROW(dstr, len + 1); /* inlined from sv_setpvn */
4117 Move(SvPVX_const(sstr),SvPVX(dstr),len,char);
4118 SvCUR_set(dstr, len);
4119 *SvEND(dstr) = '\0';
4121 /* If PERL_OLD_COPY_ON_WRITE is not defined, then isSwipe will always
4123 /* Either it's a shared hash key, or it's suitable for
4124 copy-on-write or we can swipe the string. */
4126 PerlIO_printf(Perl_debug_log, "Copy on write: sstr --> dstr\n");
4130 #ifdef PERL_OLD_COPY_ON_WRITE
4132 if ((sflags & (SVf_FAKE | SVf_READONLY))
4133 != (SVf_FAKE | SVf_READONLY)) {
4134 SvREADONLY_on(sstr);
4136 /* Make the source SV into a loop of 1.
4137 (about to become 2) */
4138 SV_COW_NEXT_SV_SET(sstr, sstr);
4142 /* Initial code is common. */
4143 if (SvPVX_const(dstr)) { /* we know that dtype >= SVt_PV */
4148 /* making another shared SV. */
4149 STRLEN cur = SvCUR(sstr);
4150 STRLEN len = SvLEN(sstr);
4151 #ifdef PERL_OLD_COPY_ON_WRITE
4153 assert (SvTYPE(dstr) >= SVt_PVIV);
4154 /* SvIsCOW_normal */
4155 /* splice us in between source and next-after-source. */
4156 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4157 SV_COW_NEXT_SV_SET(sstr, dstr);
4158 SvPV_set(dstr, SvPVX_mutable(sstr));
4162 /* SvIsCOW_shared_hash */
4163 DEBUG_C(PerlIO_printf(Perl_debug_log,
4164 "Copy on write: Sharing hash\n"));
4166 assert (SvTYPE(dstr) >= SVt_PV);
4168 HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)))));
4170 SvLEN_set(dstr, len);
4171 SvCUR_set(dstr, cur);
4172 SvREADONLY_on(dstr);
4176 { /* Passes the swipe test. */
4177 SvPV_set(dstr, SvPVX_mutable(sstr));
4178 SvLEN_set(dstr, SvLEN(sstr));
4179 SvCUR_set(dstr, SvCUR(sstr));
4182 (void)SvOK_off(sstr); /* NOTE: nukes most SvFLAGS on sstr */
4183 SvPV_set(sstr, NULL);
4189 if (sflags & SVp_NOK) {
4190 SvNV_set(dstr, SvNVX(sstr));
4192 if (sflags & SVp_IOK) {
4193 SvIV_set(dstr, SvIVX(sstr));
4194 /* Must do this otherwise some other overloaded use of 0x80000000
4195 gets confused. I guess SVpbm_VALID */
4196 if (sflags & SVf_IVisUV)
4199 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_NOK|SVp_NOK|SVf_UTF8);
4201 const MAGIC * const smg = SvVSTRING_mg(sstr);
4203 sv_magic(dstr, NULL, PERL_MAGIC_vstring,
4204 smg->mg_ptr, smg->mg_len);
4205 SvRMAGICAL_on(dstr);
4209 else if (sflags & (SVp_IOK|SVp_NOK)) {
4210 (void)SvOK_off(dstr);
4211 SvFLAGS(dstr) |= sflags & (SVf_IOK|SVp_IOK|SVf_IVisUV|SVf_NOK|SVp_NOK);
4212 if (sflags & SVp_IOK) {
4213 /* XXXX Do we want to set IsUV for IV(ROK)? Be extra safe... */
4214 SvIV_set(dstr, SvIVX(sstr));
4216 if (sflags & SVp_NOK) {
4217 SvNV_set(dstr, SvNVX(sstr));
4221 if (isGV_with_GP(sstr)) {
4222 /* This stringification rule for globs is spread in 3 places.
4223 This feels bad. FIXME. */
4224 const U32 wasfake = sflags & SVf_FAKE;
4226 /* FAKE globs can get coerced, so need to turn this off
4227 temporarily if it is on. */
4229 gv_efullname3(dstr, MUTABLE_GV(sstr), "*");
4230 SvFLAGS(sstr) |= wasfake;
4233 (void)SvOK_off(dstr);
4235 if (SvTAINTED(sstr))
4240 =for apidoc sv_setsv_mg
4242 Like C<sv_setsv>, but also handles 'set' magic.
4248 Perl_sv_setsv_mg(pTHX_ SV *const dstr, register SV *const sstr)
4250 PERL_ARGS_ASSERT_SV_SETSV_MG;
4252 sv_setsv(dstr,sstr);
4256 #ifdef PERL_OLD_COPY_ON_WRITE
4258 Perl_sv_setsv_cow(pTHX_ SV *dstr, SV *sstr)
4260 STRLEN cur = SvCUR(sstr);
4261 STRLEN len = SvLEN(sstr);
4262 register char *new_pv;
4264 PERL_ARGS_ASSERT_SV_SETSV_COW;
4267 PerlIO_printf(Perl_debug_log, "Fast copy on write: %p -> %p\n",
4268 (void*)sstr, (void*)dstr);
4275 if (SvTHINKFIRST(dstr))
4276 sv_force_normal_flags(dstr, SV_COW_DROP_PV);
4277 else if (SvPVX_const(dstr))
4278 Safefree(SvPVX_const(dstr));
4282 SvUPGRADE(dstr, SVt_PVIV);
4284 assert (SvPOK(sstr));
4285 assert (SvPOKp(sstr));
4286 assert (!SvIOK(sstr));
4287 assert (!SvIOKp(sstr));
4288 assert (!SvNOK(sstr));
4289 assert (!SvNOKp(sstr));
4291 if (SvIsCOW(sstr)) {
4293 if (SvLEN(sstr) == 0) {
4294 /* source is a COW shared hash key. */
4295 DEBUG_C(PerlIO_printf(Perl_debug_log,
4296 "Fast copy on write: Sharing hash\n"));
4297 new_pv = HEK_KEY(share_hek_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr))));
4300 SV_COW_NEXT_SV_SET(dstr, SV_COW_NEXT_SV(sstr));
4302 assert ((SvFLAGS(sstr) & CAN_COW_MASK) == CAN_COW_FLAGS);
4303 SvUPGRADE(sstr, SVt_PVIV);
4304 SvREADONLY_on(sstr);
4306 DEBUG_C(PerlIO_printf(Perl_debug_log,
4307 "Fast copy on write: Converting sstr to COW\n"));
4308 SV_COW_NEXT_SV_SET(dstr, sstr);
4310 SV_COW_NEXT_SV_SET(sstr, dstr);
4311 new_pv = SvPVX_mutable(sstr);
4314 SvPV_set(dstr, new_pv);
4315 SvFLAGS(dstr) = (SVt_PVIV|SVf_POK|SVp_POK|SVf_FAKE|SVf_READONLY);
4318 SvLEN_set(dstr, len);
4319 SvCUR_set(dstr, cur);
4328 =for apidoc sv_setpvn
4330 Copies a string into an SV. The C<len> parameter indicates the number of
4331 bytes to be copied. If the C<ptr> argument is NULL the SV will become
4332 undefined. Does not handle 'set' magic. See C<sv_setpvn_mg>.
4338 Perl_sv_setpvn(pTHX_ register SV *const sv, register const char *const ptr, register const STRLEN len)
4341 register char *dptr;
4343 PERL_ARGS_ASSERT_SV_SETPVN;
4345 SV_CHECK_THINKFIRST_COW_DROP(sv);
4351 /* len is STRLEN which is unsigned, need to copy to signed */
4354 Perl_croak(aTHX_ "panic: sv_setpvn called with negative strlen");
4356 SvUPGRADE(sv, SVt_PV);
4358 dptr = SvGROW(sv, len + 1);
4359 Move(ptr,dptr,len,char);
4362 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4367 =for apidoc sv_setpvn_mg
4369 Like C<sv_setpvn>, but also handles 'set' magic.
4375 Perl_sv_setpvn_mg(pTHX_ register SV *const sv, register const char *const ptr, register const STRLEN len)
4377 PERL_ARGS_ASSERT_SV_SETPVN_MG;
4379 sv_setpvn(sv,ptr,len);
4384 =for apidoc sv_setpv
4386 Copies a string into an SV. The string must be null-terminated. Does not
4387 handle 'set' magic. See C<sv_setpv_mg>.
4393 Perl_sv_setpv(pTHX_ register SV *const sv, register const char *const ptr)
4396 register STRLEN len;
4398 PERL_ARGS_ASSERT_SV_SETPV;
4400 SV_CHECK_THINKFIRST_COW_DROP(sv);
4406 SvUPGRADE(sv, SVt_PV);
4408 SvGROW(sv, len + 1);
4409 Move(ptr,SvPVX(sv),len+1,char);
4411 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4416 =for apidoc sv_setpv_mg
4418 Like C<sv_setpv>, but also handles 'set' magic.
4424 Perl_sv_setpv_mg(pTHX_ register SV *const sv, register const char *const ptr)
4426 PERL_ARGS_ASSERT_SV_SETPV_MG;
4433 =for apidoc sv_usepvn_flags
4435 Tells an SV to use C<ptr> to find its string value. Normally the
4436 string is stored inside the SV but sv_usepvn allows the SV to use an
4437 outside string. The C<ptr> should point to memory that was allocated
4438 by C<malloc>. The string length, C<len>, must be supplied. By default
4439 this function will realloc (i.e. move) the memory pointed to by C<ptr>,
4440 so that pointer should not be freed or used by the programmer after
4441 giving it to sv_usepvn, and neither should any pointers from "behind"
4442 that pointer (e.g. ptr + 1) be used.
4444 If C<flags> & SV_SMAGIC is true, will call SvSETMAGIC. If C<flags> &
4445 SV_HAS_TRAILING_NUL is true, then C<ptr[len]> must be NUL, and the realloc
4446 will be skipped. (i.e. the buffer is actually at least 1 byte longer than
4447 C<len>, and already meets the requirements for storing in C<SvPVX>)
4453 Perl_sv_usepvn_flags(pTHX_ SV *const sv, char *ptr, const STRLEN len, const U32 flags)
4458 PERL_ARGS_ASSERT_SV_USEPVN_FLAGS;
4460 SV_CHECK_THINKFIRST_COW_DROP(sv);
4461 SvUPGRADE(sv, SVt_PV);
4464 if (flags & SV_SMAGIC)
4468 if (SvPVX_const(sv))
4472 if (flags & SV_HAS_TRAILING_NUL)
4473 assert(ptr[len] == '\0');
4476 allocate = (flags & SV_HAS_TRAILING_NUL)
4478 #ifdef Perl_safesysmalloc_size
4481 PERL_STRLEN_ROUNDUP(len + 1);
4483 if (flags & SV_HAS_TRAILING_NUL) {
4484 /* It's long enough - do nothing.
4485 Specfically Perl_newCONSTSUB is relying on this. */
4488 /* Force a move to shake out bugs in callers. */
4489 char *new_ptr = (char*)safemalloc(allocate);
4490 Copy(ptr, new_ptr, len, char);
4491 PoisonFree(ptr,len,char);
4495 ptr = (char*) saferealloc (ptr, allocate);
4498 #ifdef Perl_safesysmalloc_size
4499 SvLEN_set(sv, Perl_safesysmalloc_size(ptr));
4501 SvLEN_set(sv, allocate);
4505 if (!(flags & SV_HAS_TRAILING_NUL)) {
4508 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4510 if (flags & SV_SMAGIC)
4514 #ifdef PERL_OLD_COPY_ON_WRITE
4515 /* Need to do this *after* making the SV normal, as we need the buffer
4516 pointer to remain valid until after we've copied it. If we let go too early,
4517 another thread could invalidate it by unsharing last of the same hash key
4518 (which it can do by means other than releasing copy-on-write Svs)
4519 or by changing the other copy-on-write SVs in the loop. */
4521 S_sv_release_COW(pTHX_ register SV *sv, const char *pvx, SV *after)
4523 PERL_ARGS_ASSERT_SV_RELEASE_COW;
4525 { /* this SV was SvIsCOW_normal(sv) */
4526 /* we need to find the SV pointing to us. */
4527 SV *current = SV_COW_NEXT_SV(after);
4529 if (current == sv) {
4530 /* The SV we point to points back to us (there were only two of us
4532 Hence other SV is no longer copy on write either. */
4534 SvREADONLY_off(after);
4536 /* We need to follow the pointers around the loop. */
4538 while ((next = SV_COW_NEXT_SV(current)) != sv) {
4541 /* don't loop forever if the structure is bust, and we have
4542 a pointer into a closed loop. */
4543 assert (current != after);
4544 assert (SvPVX_const(current) == pvx);
4546 /* Make the SV before us point to the SV after us. */
4547 SV_COW_NEXT_SV_SET(current, after);
4553 =for apidoc sv_force_normal_flags
4555 Undo various types of fakery on an SV: if the PV is a shared string, make
4556 a private copy; if we're a ref, stop refing; if we're a glob, downgrade to
4557 an xpvmg; if we're a copy-on-write scalar, this is the on-write time when
4558 we do the copy, and is also used locally. If C<SV_COW_DROP_PV> is set
4559 then a copy-on-write scalar drops its PV buffer (if any) and becomes
4560 SvPOK_off rather than making a copy. (Used where this scalar is about to be
4561 set to some other value.) In addition, the C<flags> parameter gets passed to
4562 C<sv_unref_flags()> when unrefing. C<sv_force_normal> calls this function
4563 with flags set to 0.
4569 Perl_sv_force_normal_flags(pTHX_ register SV *const sv, const U32 flags)
4573 PERL_ARGS_ASSERT_SV_FORCE_NORMAL_FLAGS;
4575 #ifdef PERL_OLD_COPY_ON_WRITE
4576 if (SvREADONLY(sv)) {
4578 const char * const pvx = SvPVX_const(sv);
4579 const STRLEN len = SvLEN(sv);
4580 const STRLEN cur = SvCUR(sv);
4581 /* next COW sv in the loop. If len is 0 then this is a shared-hash
4582 key scalar, so we mustn't attempt to call SV_COW_NEXT_SV(), as
4583 we'll fail an assertion. */
4584 SV * const next = len ? SV_COW_NEXT_SV(sv) : 0;
4587 PerlIO_printf(Perl_debug_log,
4588 "Copy on write: Force normal %ld\n",
4594 /* This SV doesn't own the buffer, so need to Newx() a new one: */
4597 if (flags & SV_COW_DROP_PV) {
4598 /* OK, so we don't need to copy our buffer. */
4601 SvGROW(sv, cur + 1);
4602 Move(pvx,SvPVX(sv),cur,char);
4607 sv_release_COW(sv, pvx, next);
4609 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4615 else if (IN_PERL_RUNTIME)
4616 Perl_croak_no_modify(aTHX);
4619 if (SvREADONLY(sv)) {
4621 const char * const pvx = SvPVX_const(sv);
4622 const STRLEN len = SvCUR(sv);
4627 SvGROW(sv, len + 1);
4628 Move(pvx,SvPVX(sv),len,char);
4630 unshare_hek(SvSHARED_HEK_FROM_PV(pvx));
4632 else if (IN_PERL_RUNTIME)
4633 Perl_croak_no_modify(aTHX);
4637 sv_unref_flags(sv, flags);
4638 else if (SvFAKE(sv) && isGV_with_GP(sv))
4640 else if (SvFAKE(sv) && SvTYPE(sv) == SVt_REGEXP) {
4641 /* Need to downgrade the REGEXP to a simple(r) scalar. This is analagous
4642 to sv_unglob. We only need it here, so inline it. */
4643 const svtype new_type = SvMAGIC(sv) || SvSTASH(sv) ? SVt_PVMG : SVt_PV;
4644 SV *const temp = newSV_type(new_type);
4645 void *const temp_p = SvANY(sv);
4647 if (new_type == SVt_PVMG) {
4648 SvMAGIC_set(temp, SvMAGIC(sv));
4649 SvMAGIC_set(sv, NULL);
4650 SvSTASH_set(temp, SvSTASH(sv));
4651 SvSTASH_set(sv, NULL);
4653 SvCUR_set(temp, SvCUR(sv));
4654 /* Remember that SvPVX is in the head, not the body. */
4656 SvLEN_set(temp, SvLEN(sv));
4657 /* This signals "buffer is owned by someone else" in sv_clear,
4658 which is the least effort way to stop it freeing the buffer.
4660 SvLEN_set(sv, SvLEN(sv)+1);
4662 /* Their buffer is already owned by someone else. */
4663 SvPVX(sv) = savepvn(SvPVX(sv), SvCUR(sv));
4664 SvLEN_set(temp, SvCUR(sv)+1);
4667 /* Now swap the rest of the bodies. */
4669 SvFLAGS(sv) &= ~(SVf_FAKE|SVTYPEMASK);
4670 SvFLAGS(sv) |= new_type;
4671 SvANY(sv) = SvANY(temp);
4673 SvFLAGS(temp) &= ~(SVTYPEMASK);
4674 SvFLAGS(temp) |= SVt_REGEXP|SVf_FAKE;
4675 SvANY(temp) = temp_p;
4684 Efficient removal of characters from the beginning of the string buffer.
4685 SvPOK(sv) must be true and the C<ptr> must be a pointer to somewhere inside
4686 the string buffer. The C<ptr> becomes the first character of the adjusted
4687 string. Uses the "OOK hack".
4688 Beware: after this function returns, C<ptr> and SvPVX_const(sv) may no longer
4689 refer to the same chunk of data.
4695 Perl_sv_chop(pTHX_ register SV *const sv, register const char *const ptr)
4701 const U8 *real_start;
4705 PERL_ARGS_ASSERT_SV_CHOP;
4707 if (!ptr || !SvPOKp(sv))
4709 delta = ptr - SvPVX_const(sv);
4711 /* Nothing to do. */
4714 /* SvPVX(sv) may move in SV_CHECK_THINKFIRST(sv), but after this line,
4715 nothing uses the value of ptr any more. */
4716 max_delta = SvLEN(sv) ? SvLEN(sv) : SvCUR(sv);
4717 if (ptr <= SvPVX_const(sv))
4718 Perl_croak(aTHX_ "panic: sv_chop ptr=%p, start=%p, end=%p",
4719 ptr, SvPVX_const(sv), SvPVX_const(sv) + max_delta);
4720 SV_CHECK_THINKFIRST(sv);
4721 if (delta > max_delta)
4722 Perl_croak(aTHX_ "panic: sv_chop ptr=%p (was %p), start=%p, end=%p",
4723 SvPVX_const(sv) + delta, ptr, SvPVX_const(sv),
4724 SvPVX_const(sv) + max_delta);
4727 if (!SvLEN(sv)) { /* make copy of shared string */
4728 const char *pvx = SvPVX_const(sv);
4729 const STRLEN len = SvCUR(sv);
4730 SvGROW(sv, len + 1);
4731 Move(pvx,SvPVX(sv),len,char);
4734 SvFLAGS(sv) |= SVf_OOK;
4737 SvOOK_offset(sv, old_delta);
4739 SvLEN_set(sv, SvLEN(sv) - delta);
4740 SvCUR_set(sv, SvCUR(sv) - delta);
4741 SvPV_set(sv, SvPVX(sv) + delta);
4743 p = (U8 *)SvPVX_const(sv);
4748 real_start = p - delta;
4752 if (delta < 0x100) {
4756 p -= sizeof(STRLEN);
4757 Copy((U8*)&delta, p, sizeof(STRLEN), U8);
4761 /* Fill the preceding buffer with sentinals to verify that no-one is
4763 while (p > real_start) {
4771 =for apidoc sv_catpvn
4773 Concatenates the string onto the end of the string which is in the SV. The
4774 C<len> indicates number of bytes to copy. If the SV has the UTF-8
4775 status set, then the bytes appended should be valid UTF-8.
4776 Handles 'get' magic, but not 'set' magic. See C<sv_catpvn_mg>.
4778 =for apidoc sv_catpvn_flags
4780 Concatenates the string onto the end of the string which is in the SV. The
4781 C<len> indicates number of bytes to copy. If the SV has the UTF-8
4782 status set, then the bytes appended should be valid UTF-8.
4783 If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<dsv> if
4784 appropriate, else not. C<sv_catpvn> and C<sv_catpvn_nomg> are implemented
4785 in terms of this function.
4791 Perl_sv_catpvn_flags(pTHX_ register SV *const dsv, register const char *sstr, register const STRLEN slen, const I32 flags)
4795 const char * const dstr = SvPV_force_flags(dsv, dlen, flags);
4797 PERL_ARGS_ASSERT_SV_CATPVN_FLAGS;
4799 SvGROW(dsv, dlen + slen + 1);
4801 sstr = SvPVX_const(dsv);
4802 Move(sstr, SvPVX(dsv) + dlen, slen, char);
4803 SvCUR_set(dsv, SvCUR(dsv) + slen);
4805 (void)SvPOK_only_UTF8(dsv); /* validate pointer */
4807 if (flags & SV_SMAGIC)
4812 =for apidoc sv_catsv
4814 Concatenates the string from SV C<ssv> onto the end of the string in
4815 SV C<dsv>. Modifies C<dsv> but not C<ssv>. Handles 'get' magic, but
4816 not 'set' magic. See C<sv_catsv_mg>.
4818 =for apidoc sv_catsv_flags
4820 Concatenates the string from SV C<ssv> onto the end of the string in
4821 SV C<dsv>. Modifies C<dsv> but not C<ssv>. If C<flags> has C<SV_GMAGIC>
4822 bit set, will C<mg_get> on the SVs if appropriate, else not. C<sv_catsv>
4823 and C<sv_catsv_nomg> are implemented in terms of this function.
4828 Perl_sv_catsv_flags(pTHX_ SV *const dsv, register SV *const ssv, const I32 flags)
4832 PERL_ARGS_ASSERT_SV_CATSV_FLAGS;
4836 const char *spv = SvPV_flags_const(ssv, slen, flags);
4838 /* sutf8 and dutf8 were type bool, but under USE_ITHREADS,
4839 gcc version 2.95.2 20000220 (Debian GNU/Linux) for
4840 Linux xxx 2.2.17 on sparc64 with gcc -O2, we erroneously
4841 get dutf8 = 0x20000000, (i.e. SVf_UTF8) even though
4842 dsv->sv_flags doesn't have that bit set.
4843 Andy Dougherty 12 Oct 2001
4845 const I32 sutf8 = DO_UTF8(ssv);
4848 if (SvGMAGICAL(dsv) && (flags & SV_GMAGIC))
4850 dutf8 = DO_UTF8(dsv);
4852 if (dutf8 != sutf8) {
4854 /* Not modifying source SV, so taking a temporary copy. */
4855 SV* const csv = newSVpvn_flags(spv, slen, SVs_TEMP);
4857 sv_utf8_upgrade(csv);
4858 spv = SvPV_const(csv, slen);
4861 /* Leave enough space for the cat that's about to happen */
4862 sv_utf8_upgrade_flags_grow(dsv, 0, slen);
4864 sv_catpvn_nomg(dsv, spv, slen);
4867 if (flags & SV_SMAGIC)
4872 =for apidoc sv_catpv
4874 Concatenates the string onto the end of the string which is in the SV.
4875 If the SV has the UTF-8 status set, then the bytes appended should be
4876 valid UTF-8. Handles 'get' magic, but not 'set' magic. See C<sv_catpv_mg>.
4881 Perl_sv_catpv(pTHX_ register SV *const sv, register const char *ptr)
4884 register STRLEN len;
4888 PERL_ARGS_ASSERT_SV_CATPV;
4892 junk = SvPV_force(sv, tlen);
4894 SvGROW(sv, tlen + len + 1);
4896 ptr = SvPVX_const(sv);
4897 Move(ptr,SvPVX(sv)+tlen,len+1,char);
4898 SvCUR_set(sv, SvCUR(sv) + len);
4899 (void)SvPOK_only_UTF8(sv); /* validate pointer */
4904 =for apidoc sv_catpv_flags
4906 Concatenates the string onto the end of the string which is in the SV.
4907 If the SV has the UTF-8 status set, then the bytes appended should
4908 be valid UTF-8. If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get>
4909 on the SVs if appropriate, else not.
4915 Perl_sv_catpv_flags(pTHX_ SV *dstr, const char *sstr, I32 flags)
4917 PERL_ARGS_ASSERT_SV_CATPV_FLAGS;
4918 sv_catpvn_flags(dstr, sstr, strlen(sstr), flags);
4922 =for apidoc sv_catpv_mg
4924 Like C<sv_catpv>, but also handles 'set' magic.
4930 Perl_sv_catpv_mg(pTHX_ register SV *const sv, register const char *const ptr)
4932 PERL_ARGS_ASSERT_SV_CATPV_MG;
4941 Creates a new SV. A non-zero C<len> parameter indicates the number of
4942 bytes of preallocated string space the SV should have. An extra byte for a
4943 trailing NUL is also reserved. (SvPOK is not set for the SV even if string
4944 space is allocated.) The reference count for the new SV is set to 1.
4946 In 5.9.3, newSV() replaces the older NEWSV() API, and drops the first
4947 parameter, I<x>, a debug aid which allowed callers to identify themselves.
4948 This aid has been superseded by a new build option, PERL_MEM_LOG (see
4949 L<perlhack/PERL_MEM_LOG>). The older API is still there for use in XS
4950 modules supporting older perls.
4956 Perl_newSV(pTHX_ const STRLEN len)
4963 sv_upgrade(sv, SVt_PV);
4964 SvGROW(sv, len + 1);
4969 =for apidoc sv_magicext
4971 Adds magic to an SV, upgrading it if necessary. Applies the
4972 supplied vtable and returns a pointer to the magic added.
4974 Note that C<sv_magicext> will allow things that C<sv_magic> will not.
4975 In particular, you can add magic to SvREADONLY SVs, and add more than
4976 one instance of the same 'how'.
4978 If C<namlen> is greater than zero then a C<savepvn> I<copy> of C<name> is
4979 stored, if C<namlen> is zero then C<name> is stored as-is and - as another
4980 special case - if C<(name && namlen == HEf_SVKEY)> then C<name> is assumed
4981 to contain an C<SV*> and is stored as-is with its REFCNT incremented.
4983 (This is now used as a subroutine by C<sv_magic>.)
4988 Perl_sv_magicext(pTHX_ SV *const sv, SV *const obj, const int how,
4989 const MGVTBL *const vtable, const char *const name, const I32 namlen)
4994 PERL_ARGS_ASSERT_SV_MAGICEXT;
4996 SvUPGRADE(sv, SVt_PVMG);
4997 Newxz(mg, 1, MAGIC);
4998 mg->mg_moremagic = SvMAGIC(sv);
4999 SvMAGIC_set(sv, mg);
5001 /* Sometimes a magic contains a reference loop, where the sv and
5002 object refer to each other. To prevent a reference loop that
5003 would prevent such objects being freed, we look for such loops
5004 and if we find one we avoid incrementing the object refcount.
5006 Note we cannot do this to avoid self-tie loops as intervening RV must
5007 have its REFCNT incremented to keep it in existence.
5010 if (!obj || obj == sv ||
5011 how == PERL_MAGIC_arylen ||
5012 how == PERL_MAGIC_symtab ||
5013 (SvTYPE(obj) == SVt_PVGV &&
5014 (GvSV(obj) == sv || GvHV(obj) == (const HV *)sv
5015 || GvAV(obj) == (const AV *)sv || GvCV(obj) == (const CV *)sv
5016 || GvIOp(obj) == (const IO *)sv || GvFORM(obj) == (const CV *)sv)))
5021 mg->mg_obj = SvREFCNT_inc_simple(obj);
5022 mg->mg_flags |= MGf_REFCOUNTED;
5025 /* Normal self-ties simply pass a null object, and instead of
5026 using mg_obj directly, use the SvTIED_obj macro to produce a
5027 new RV as needed. For glob "self-ties", we are tieing the PVIO
5028 with an RV obj pointing to the glob containing the PVIO. In
5029 this case, to avoid a reference loop, we need to weaken the
5033 if (how == PERL_MAGIC_tiedscalar && SvTYPE(sv) == SVt_PVIO &&
5034 obj && SvROK(obj) && GvIO(SvRV(obj)) == (const IO *)sv)
5040 mg->mg_len = namlen;
5043 mg->mg_ptr = savepvn(name, namlen);
5044 else if (namlen == HEf_SVKEY) {
5045 /* Yes, this is casting away const. This is only for the case of
5046 HEf_SVKEY. I think we need to document this abberation of the
5047 constness of the API, rather than making name non-const, as
5048 that change propagating outwards a long way. */
5049 mg->mg_ptr = (char*)SvREFCNT_inc_simple_NN((SV *)name);
5051 mg->mg_ptr = (char *) name;
5053 mg->mg_virtual = (MGVTBL *) vtable;
5057 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
5062 =for apidoc sv_magic
5064 Adds magic to an SV. First upgrades C<sv> to type C<SVt_PVMG> if necessary,
5065 then adds a new magic item of type C<how> to the head of the magic list.
5067 See C<sv_magicext> (which C<sv_magic> now calls) for a description of the
5068 handling of the C<name> and C<namlen> arguments.
5070 You need to use C<sv_magicext> to add magic to SvREADONLY SVs and also
5071 to add more than one instance of the same 'how'.
5077 Perl_sv_magic(pTHX_ register SV *const sv, SV *const obj, const int how,
5078 const char *const name, const I32 namlen)
5081 const MGVTBL *vtable;
5084 PERL_ARGS_ASSERT_SV_MAGIC;
5086 #ifdef PERL_OLD_COPY_ON_WRITE
5088 sv_force_normal_flags(sv, 0);
5090 if (SvREADONLY(sv)) {
5092 /* its okay to attach magic to shared strings; the subsequent
5093 * upgrade to PVMG will unshare the string */
5094 !(SvFAKE(sv) && SvTYPE(sv) < SVt_PVMG)
5097 && how != PERL_MAGIC_regex_global
5098 && how != PERL_MAGIC_bm
5099 && how != PERL_MAGIC_fm
5100 && how != PERL_MAGIC_sv
5101 && how != PERL_MAGIC_backref
5104 Perl_croak_no_modify(aTHX);
5107 if (SvMAGICAL(sv) || (how == PERL_MAGIC_taint && SvTYPE(sv) >= SVt_PVMG)) {
5108 if (SvMAGIC(sv) && (mg = mg_find(sv, how))) {
5109 /* sv_magic() refuses to add a magic of the same 'how' as an
5112 if (how == PERL_MAGIC_taint) {
5114 /* Any scalar which already had taint magic on which someone
5115 (erroneously?) did SvIOK_on() or similar will now be
5116 incorrectly sporting public "OK" flags. */
5117 SvFLAGS(sv) &= ~(SVf_IOK|SVf_NOK|SVf_POK);
5125 vtable = &PL_vtbl_sv;
5127 case PERL_MAGIC_overload:
5128 vtable = &PL_vtbl_amagic;
5130 case PERL_MAGIC_overload_elem:
5131 vtable = &PL_vtbl_amagicelem;
5133 case PERL_MAGIC_overload_table:
5134 vtable = &PL_vtbl_ovrld;
5137 vtable = &PL_vtbl_bm;
5139 case PERL_MAGIC_regdata:
5140 vtable = &PL_vtbl_regdata;
5142 case PERL_MAGIC_regdatum:
5143 vtable = &PL_vtbl_regdatum;
5145 case PERL_MAGIC_env:
5146 vtable = &PL_vtbl_env;
5149 vtable = &PL_vtbl_fm;
5151 case PERL_MAGIC_envelem:
5152 vtable = &PL_vtbl_envelem;
5154 case PERL_MAGIC_regex_global:
5155 vtable = &PL_vtbl_mglob;
5157 case PERL_MAGIC_isa:
5158 vtable = &PL_vtbl_isa;
5160 case PERL_MAGIC_isaelem:
5161 vtable = &PL_vtbl_isaelem;
5163 case PERL_MAGIC_nkeys:
5164 vtable = &PL_vtbl_nkeys;
5166 case PERL_MAGIC_dbfile:
5169 case PERL_MAGIC_dbline:
5170 vtable = &PL_vtbl_dbline;
5172 #ifdef USE_LOCALE_COLLATE
5173 case PERL_MAGIC_collxfrm:
5174 vtable = &PL_vtbl_collxfrm;
5176 #endif /* USE_LOCALE_COLLATE */
5177 case PERL_MAGIC_tied:
5178 vtable = &PL_vtbl_pack;
5180 case PERL_MAGIC_tiedelem:
5181 case PERL_MAGIC_tiedscalar:
5182 vtable = &PL_vtbl_packelem;
5185 vtable = &PL_vtbl_regexp;
5187 case PERL_MAGIC_sig:
5188 vtable = &PL_vtbl_sig;
5190 case PERL_MAGIC_sigelem:
5191 vtable = &PL_vtbl_sigelem;
5193 case PERL_MAGIC_taint:
5194 vtable = &PL_vtbl_taint;
5196 case PERL_MAGIC_uvar:
5197 vtable = &PL_vtbl_uvar;
5199 case PERL_MAGIC_vec:
5200 vtable = &PL_vtbl_vec;
5202 case PERL_MAGIC_arylen_p:
5203 case PERL_MAGIC_rhash:
5204 case PERL_MAGIC_symtab:
5205 case PERL_MAGIC_vstring:
5208 case PERL_MAGIC_utf8:
5209 vtable = &PL_vtbl_utf8;
5211 case PERL_MAGIC_substr:
5212 vtable = &PL_vtbl_substr;
5214 case PERL_MAGIC_defelem:
5215 vtable = &PL_vtbl_defelem;
5217 case PERL_MAGIC_arylen:
5218 vtable = &PL_vtbl_arylen;
5220 case PERL_MAGIC_pos:
5221 vtable = &PL_vtbl_pos;
5223 case PERL_MAGIC_backref:
5224 vtable = &PL_vtbl_backref;
5226 case PERL_MAGIC_hintselem:
5227 vtable = &PL_vtbl_hintselem;
5229 case PERL_MAGIC_hints:
5230 vtable = &PL_vtbl_hints;
5232 case PERL_MAGIC_ext:
5233 /* Reserved for use by extensions not perl internals. */
5234 /* Useful for attaching extension internal data to perl vars. */
5235 /* Note that multiple extensions may clash if magical scalars */
5236 /* etc holding private data from one are passed to another. */
5240 Perl_croak(aTHX_ "Don't know how to handle magic of type \\%o", how);
5243 /* Rest of work is done else where */
5244 mg = sv_magicext(sv,obj,how,vtable,name,namlen);
5247 case PERL_MAGIC_taint:
5250 case PERL_MAGIC_ext:
5251 case PERL_MAGIC_dbfile:
5258 =for apidoc sv_unmagic
5260 Removes all magic of type C<type> from an SV.
5266 Perl_sv_unmagic(pTHX_ SV *const sv, const int type)
5271 PERL_ARGS_ASSERT_SV_UNMAGIC;
5273 if (SvTYPE(sv) < SVt_PVMG || !SvMAGIC(sv))
5275 mgp = &(((XPVMG*) SvANY(sv))->xmg_u.xmg_magic);
5276 for (mg = *mgp; mg; mg = *mgp) {
5277 if (mg->mg_type == type) {
5278 const MGVTBL* const vtbl = mg->mg_virtual;
5279 *mgp = mg->mg_moremagic;
5280 if (vtbl && vtbl->svt_free)
5281 vtbl->svt_free(aTHX_ sv, mg);
5282 if (mg->mg_ptr && mg->mg_type != PERL_MAGIC_regex_global) {
5284 Safefree(mg->mg_ptr);
5285 else if (mg->mg_len == HEf_SVKEY)
5286 SvREFCNT_dec(MUTABLE_SV(mg->mg_ptr));
5287 else if (mg->mg_type == PERL_MAGIC_utf8)
5288 Safefree(mg->mg_ptr);
5290 if (mg->mg_flags & MGf_REFCOUNTED)
5291 SvREFCNT_dec(mg->mg_obj);
5295 mgp = &mg->mg_moremagic;
5298 if (SvMAGICAL(sv)) /* if we're under save_magic, wait for restore_magic; */
5299 mg_magical(sv); /* else fix the flags now */
5303 SvFLAGS(sv) |= (SvFLAGS(sv) & (SVp_IOK|SVp_NOK|SVp_POK)) >> PRIVSHIFT;
5309 =for apidoc sv_rvweaken
5311 Weaken a reference: set the C<SvWEAKREF> flag on this RV; give the
5312 referred-to SV C<PERL_MAGIC_backref> magic if it hasn't already; and
5313 push a back-reference to this RV onto the array of backreferences
5314 associated with that magic. If the RV is magical, set magic will be
5315 called after the RV is cleared.
5321 Perl_sv_rvweaken(pTHX_ SV *const sv)
5325 PERL_ARGS_ASSERT_SV_RVWEAKEN;
5327 if (!SvOK(sv)) /* let undefs pass */
5330 Perl_croak(aTHX_ "Can't weaken a nonreference");
5331 else if (SvWEAKREF(sv)) {
5332 Perl_ck_warner(aTHX_ packWARN(WARN_MISC), "Reference is already weak");
5336 Perl_sv_add_backref(aTHX_ tsv, sv);
5342 /* Give tsv backref magic if it hasn't already got it, then push a
5343 * back-reference to sv onto the array associated with the backref magic.
5345 * As an optimisation, if there's only one backref and it's not an AV,
5346 * store it directly in the HvAUX or mg_obj slot, avoiding the need to
5347 * allocate an AV. (Whether the slot holds an AV tells us whether this is
5350 * If an HV's backref is stored in magic, it is moved back to HvAUX.
5353 /* A discussion about the backreferences array and its refcount:
5355 * The AV holding the backreferences is pointed to either as the mg_obj of
5356 * PERL_MAGIC_backref, or in the specific case of a HV that has the hv_aux
5357 * structure, from the xhv_backreferences field. (A HV without hv_aux will
5358 * have the standard magic instead.) The array is created with a refcount
5359 * of 2. This means that if during global destruction the array gets
5360 * picked on before its parent to have its refcount decremented by the
5361 * random zapper, it won't actually be freed, meaning it's still there for
5362 * when its parent gets freed.
5364 * When the parent SV is freed, the extra ref is killed by
5365 * Perl_sv_kill_backrefs. The other ref is killed, in the case of magic,
5366 * by mg_free() / MGf_REFCOUNTED, or for a hash, by Perl_hv_kill_backrefs.
5368 * When a single backref SV is stored directly, it is not reference
5373 Perl_sv_add_backref(pTHX_ SV *const tsv, SV *const sv)
5380 PERL_ARGS_ASSERT_SV_ADD_BACKREF;
5382 /* find slot to store array or singleton backref */
5384 if (SvTYPE(tsv) == SVt_PVHV) {
5385 svp = (SV**)Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5388 if ((mg = mg_find(tsv, PERL_MAGIC_backref))) {
5389 /* Aha. They've got it stowed in magic instead.
5390 * Move it back to xhv_backreferences */
5392 /* Stop mg_free decreasing the reference count. */
5394 /* Stop mg_free even calling the destructor, given that
5395 there's no AV to free up. */
5397 sv_unmagic(tsv, PERL_MAGIC_backref);
5403 (SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL))))
5405 sv_magic(tsv, NULL, PERL_MAGIC_backref, NULL, 0);
5406 mg = mg_find(tsv, PERL_MAGIC_backref);
5408 svp = &(mg->mg_obj);
5411 /* create or retrieve the array */
5413 if ( (!*svp && SvTYPE(sv) == SVt_PVAV)
5414 || (*svp && SvTYPE(*svp) != SVt_PVAV)
5419 SvREFCNT_inc_simple_void(av);
5420 /* av now has a refcnt of 2; see discussion above */
5422 /* move single existing backref to the array */
5424 AvARRAY(av)[++AvFILLp(av)] = *svp; /* av_push() */
5428 mg->mg_flags |= MGf_REFCOUNTED;
5431 av = MUTABLE_AV(*svp);
5434 /* optimisation: store single backref directly in HvAUX or mg_obj */
5438 /* push new backref */
5439 assert(SvTYPE(av) == SVt_PVAV);
5440 if (AvFILLp(av) >= AvMAX(av)) {
5441 av_extend(av, AvFILLp(av)+1);
5443 AvARRAY(av)[++AvFILLp(av)] = sv; /* av_push() */
5446 /* delete a back-reference to ourselves from the backref magic associated
5447 * with the SV we point to.
5451 Perl_sv_del_backref(pTHX_ SV *const tsv, SV *const sv)
5457 PERL_ARGS_ASSERT_SV_DEL_BACKREF;
5459 if (SvTYPE(tsv) == SVt_PVHV && SvOOK(tsv)) {
5460 svp = (SV**)Perl_hv_backreferences_p(aTHX_ MUTABLE_HV(tsv));
5462 if (!svp || !*svp) {
5464 = SvMAGICAL(tsv) ? mg_find(tsv, PERL_MAGIC_backref) : NULL;
5465 svp = mg ? &(mg->mg_obj) : NULL;
5469 Perl_croak(aTHX_ "panic: del_backref");
5471 if (SvTYPE(*svp) == SVt_PVAV) {
5473 AV * const av = (AV*)*svp;
5474 assert(!SvIS_FREED(av));
5476 for (i = AvFILLp(av); i >= 0; i--) {
5478 const SSize_t fill = AvFILLp(av);
5480 /* We weren't the last entry.
5481 An unordered list has this property that you can take the
5482 last element off the end to fill the hole, and it's still
5483 an unordered list :-)
5488 AvFILLp(av) = fill - 1;
5491 break; /* should only be one */
5498 /* optimisation: only a single backref, stored directly */
5500 Perl_croak(aTHX_ "panic: del_backref");
5507 Perl_sv_kill_backrefs(pTHX_ SV *const sv, AV *const av)
5513 PERL_ARGS_ASSERT_SV_KILL_BACKREFS;
5518 is_array = (SvTYPE(av) == SVt_PVAV);
5520 assert(!SvIS_FREED(av));
5523 last = svp + AvFILLp(av);
5526 /* optimisation: only a single backref, stored directly */
5532 while (svp <= last) {
5534 SV *const referrer = *svp;
5535 if (SvWEAKREF(referrer)) {
5536 /* XXX Should we check that it hasn't changed? */
5537 assert(SvROK(referrer));
5538 SvRV_set(referrer, 0);
5540 SvWEAKREF_off(referrer);
5541 SvSETMAGIC(referrer);
5542 } else if (SvTYPE(referrer) == SVt_PVGV ||
5543 SvTYPE(referrer) == SVt_PVLV) {
5544 assert(SvTYPE(sv) == SVt_PVHV); /* stash backref */
5545 /* You lookin' at me? */
5546 assert(GvSTASH(referrer));
5547 assert(GvSTASH(referrer) == (const HV *)sv);
5548 GvSTASH(referrer) = 0;
5549 } else if (SvTYPE(referrer) == SVt_PVCV ||
5550 SvTYPE(referrer) == SVt_PVFM) {
5551 if (SvTYPE(sv) == SVt_PVHV) { /* stash backref */
5552 /* You lookin' at me? */
5553 assert(CvSTASH(referrer));
5554 assert(CvSTASH(referrer) == (const HV *)sv);
5555 CvSTASH(referrer) = 0;
5558 assert(SvTYPE(sv) == SVt_PVGV);
5559 /* You lookin' at me? */
5560 assert(CvGV(referrer));
5561 assert(CvGV(referrer) == (const GV *)sv);
5562 anonymise_cv_maybe(MUTABLE_GV(sv),
5563 MUTABLE_CV(referrer));
5568 "panic: magic_killbackrefs (flags=%"UVxf")",
5569 (UV)SvFLAGS(referrer));
5580 SvREFCNT_dec(av); /* remove extra count added by sv_add_backref() */
5586 =for apidoc sv_insert
5588 Inserts a string at the specified offset/length within the SV. Similar to
5589 the Perl substr() function. Handles get magic.
5591 =for apidoc sv_insert_flags
5593 Same as C<sv_insert>, but the extra C<flags> are passed the C<SvPV_force_flags> that applies to C<bigstr>.
5599 Perl_sv_insert_flags(pTHX_ SV *const bigstr, const STRLEN offset, const STRLEN len, const char *const little, const STRLEN littlelen, const U32 flags)
5604 register char *midend;
5605 register char *bigend;
5609 PERL_ARGS_ASSERT_SV_INSERT_FLAGS;
5612 Perl_croak(aTHX_ "Can't modify non-existent substring");
5613 SvPV_force_flags(bigstr, curlen, flags);
5614 (void)SvPOK_only_UTF8(bigstr);
5615 if (offset + len > curlen) {
5616 SvGROW(bigstr, offset+len+1);
5617 Zero(SvPVX(bigstr)+curlen, offset+len-curlen, char);
5618 SvCUR_set(bigstr, offset+len);
5622 i = littlelen - len;
5623 if (i > 0) { /* string might grow */
5624 big = SvGROW(bigstr, SvCUR(bigstr) + i + 1);
5625 mid = big + offset + len;
5626 midend = bigend = big + SvCUR(bigstr);
5629 while (midend > mid) /* shove everything down */
5630 *--bigend = *--midend;
5631 Move(little,big+offset,littlelen,char);
5632 SvCUR_set(bigstr, SvCUR(bigstr) + i);
5637 Move(little,SvPVX(bigstr)+offset,len,char);
5642 big = SvPVX(bigstr);
5645 bigend = big + SvCUR(bigstr);
5647 if (midend > bigend)
5648 Perl_croak(aTHX_ "panic: sv_insert");
5650 if (mid - big > bigend - midend) { /* faster to shorten from end */
5652 Move(little, mid, littlelen,char);
5655 i = bigend - midend;
5657 Move(midend, mid, i,char);
5661 SvCUR_set(bigstr, mid - big);
5663 else if ((i = mid - big)) { /* faster from front */
5664 midend -= littlelen;
5666 Move(big, midend - i, i, char);
5667 sv_chop(bigstr,midend-i);
5669 Move(little, mid, littlelen,char);
5671 else if (littlelen) {
5672 midend -= littlelen;
5673 sv_chop(bigstr,midend);
5674 Move(little,midend,littlelen,char);
5677 sv_chop(bigstr,midend);
5683 =for apidoc sv_replace
5685 Make the first argument a copy of the second, then delete the original.
5686 The target SV physically takes over ownership of the body of the source SV
5687 and inherits its flags; however, the target keeps any magic it owns,
5688 and any magic in the source is discarded.
5689 Note that this is a rather specialist SV copying operation; most of the
5690 time you'll want to use C<sv_setsv> or one of its many macro front-ends.
5696 Perl_sv_replace(pTHX_ register SV *const sv, register SV *const nsv)
5699 const U32 refcnt = SvREFCNT(sv);
5701 PERL_ARGS_ASSERT_SV_REPLACE;
5703 SV_CHECK_THINKFIRST_COW_DROP(sv);
5704 if (SvREFCNT(nsv) != 1) {
5705 Perl_croak(aTHX_ "panic: reference miscount on nsv in sv_replace()"
5706 " (%" UVuf " != 1)", (UV) SvREFCNT(nsv));
5708 if (SvMAGICAL(sv)) {
5712 sv_upgrade(nsv, SVt_PVMG);
5713 SvMAGIC_set(nsv, SvMAGIC(sv));
5714 SvFLAGS(nsv) |= SvMAGICAL(sv);
5716 SvMAGIC_set(sv, NULL);
5720 assert(!SvREFCNT(sv));
5721 #ifdef DEBUG_LEAKING_SCALARS
5722 sv->sv_flags = nsv->sv_flags;
5723 sv->sv_any = nsv->sv_any;
5724 sv->sv_refcnt = nsv->sv_refcnt;
5725 sv->sv_u = nsv->sv_u;
5727 StructCopy(nsv,sv,SV);
5729 if(SvTYPE(sv) == SVt_IV) {
5731 = (XPVIV*)((char*)&(sv->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
5735 #ifdef PERL_OLD_COPY_ON_WRITE
5736 if (SvIsCOW_normal(nsv)) {
5737 /* We need to follow the pointers around the loop to make the
5738 previous SV point to sv, rather than nsv. */
5741 while ((next = SV_COW_NEXT_SV(current)) != nsv) {
5744 assert(SvPVX_const(current) == SvPVX_const(nsv));
5746 /* Make the SV before us point to the SV after us. */
5748 PerlIO_printf(Perl_debug_log, "previous is\n");
5750 PerlIO_printf(Perl_debug_log,
5751 "move it from 0x%"UVxf" to 0x%"UVxf"\n",
5752 (UV) SV_COW_NEXT_SV(current), (UV) sv);
5754 SV_COW_NEXT_SV_SET(current, sv);
5757 SvREFCNT(sv) = refcnt;
5758 SvFLAGS(nsv) |= SVTYPEMASK; /* Mark as freed */
5763 /* We're about to free a GV which has a CV that refers back to us.
5764 * If that CV will outlive us, make it anonymous (i.e. fix up its CvGV
5768 S_anonymise_cv_maybe(pTHX_ GV *gv, CV* cv)
5774 PERL_ARGS_ASSERT_ANONYMISE_CV_MAYBE;
5777 assert(SvREFCNT(gv) == 0);
5778 assert(isGV(gv) && isGV_with_GP(gv));
5780 assert(!CvANON(cv));
5781 assert(CvGV(cv) == gv);
5783 /* will the CV shortly be freed by gp_free() ? */
5784 if (GvCV(gv) == cv && GvGP(gv)->gp_refcnt < 2 && SvREFCNT(cv) < 2) {
5785 SvANY(cv)->xcv_gv = NULL;
5789 /* if not, anonymise: */
5790 stash = GvSTASH(gv) ? HvNAME(GvSTASH(gv)) : NULL;
5791 gvname = Perl_newSVpvf(aTHX_ "%s::__ANON__",
5792 stash ? stash : "__ANON__");
5793 anongv = gv_fetchsv(gvname, GV_ADDMULTI, SVt_PVCV);
5794 SvREFCNT_dec(gvname);
5798 SvANY(cv)->xcv_gv = MUTABLE_GV(SvREFCNT_inc(anongv));
5803 =for apidoc sv_clear
5805 Clear an SV: call any destructors, free up any memory used by the body,
5806 and free the body itself. The SV's head is I<not> freed, although
5807 its type is set to all 1's so that it won't inadvertently be assumed
5808 to be live during global destruction etc.
5809 This function should only be called when REFCNT is zero. Most of the time
5810 you'll want to call C<sv_free()> (or its macro wrapper C<SvREFCNT_dec>)
5817 Perl_sv_clear(pTHX_ register SV *const sv)
5820 const U32 type = SvTYPE(sv);
5821 const struct body_details *const sv_type_details
5822 = bodies_by_type + type;
5825 PERL_ARGS_ASSERT_SV_CLEAR;
5826 assert(SvREFCNT(sv) == 0);
5827 assert(SvTYPE(sv) != SVTYPEMASK);
5829 if (type <= SVt_IV) {
5830 /* See the comment in sv.h about the collusion between this early
5831 return and the overloading of the NULL slots in the size table. */
5834 SvFLAGS(sv) &= SVf_BREAK;
5835 SvFLAGS(sv) |= SVTYPEMASK;
5840 if (PL_defstash && /* Still have a symbol table? */
5847 stash = SvSTASH(sv);
5848 destructor = StashHANDLER(stash,DESTROY);
5850 /* A constant subroutine can have no side effects, so
5851 don't bother calling it. */
5852 && !CvCONST(destructor)
5853 /* Don't bother calling an empty destructor */
5854 && (CvISXSUB(destructor)
5855 || (CvSTART(destructor)
5856 && (CvSTART(destructor)->op_next->op_type != OP_LEAVESUB))))
5858 SV* const tmpref = newRV(sv);
5859 SvREADONLY_on(tmpref); /* DESTROY() could be naughty */
5861 PUSHSTACKi(PERLSI_DESTROY);
5866 call_sv(MUTABLE_SV(destructor), G_DISCARD|G_EVAL|G_KEEPERR|G_VOID);
5872 if(SvREFCNT(tmpref) < 2) {
5873 /* tmpref is not kept alive! */
5875 SvRV_set(tmpref, NULL);
5878 SvREFCNT_dec(tmpref);
5880 } while (SvOBJECT(sv) && SvSTASH(sv) != stash);
5884 if (PL_in_clean_objs)
5885 Perl_croak(aTHX_ "DESTROY created new reference to dead object '%s'",
5887 /* DESTROY gave object new lease on life */
5893 SvREFCNT_dec(SvSTASH(sv)); /* possibly of changed persuasion */
5894 SvOBJECT_off(sv); /* Curse the object. */
5895 if (type != SVt_PVIO)
5896 --PL_sv_objcount; /* XXX Might want something more general */
5899 if (type >= SVt_PVMG) {
5900 if (type == SVt_PVMG && SvPAD_OUR(sv)) {
5901 SvREFCNT_dec(SvOURSTASH(sv));
5902 } else if (SvMAGIC(sv))
5904 if (type == SVt_PVMG && SvPAD_TYPED(sv))
5905 SvREFCNT_dec(SvSTASH(sv));
5908 /* case SVt_BIND: */
5911 IoIFP(sv) != PerlIO_stdin() &&
5912 IoIFP(sv) != PerlIO_stdout() &&
5913 IoIFP(sv) != PerlIO_stderr() &&
5914 !(IoFLAGS(sv) & IOf_FAKE_DIRP))
5916 io_close(MUTABLE_IO(sv), FALSE);
5918 if (IoDIRP(sv) && !(IoFLAGS(sv) & IOf_FAKE_DIRP))
5919 PerlDir_close(IoDIRP(sv));
5920 IoDIRP(sv) = (DIR*)NULL;
5921 Safefree(IoTOP_NAME(sv));
5922 Safefree(IoFMT_NAME(sv));
5923 Safefree(IoBOTTOM_NAME(sv));
5926 /* FIXME for plugins */
5927 pregfree2((REGEXP*) sv);
5931 cv_undef(MUTABLE_CV(sv));
5932 /* If we're in a stash, we don't own a reference to it. However it does
5933 have a back reference to us, which needs to be cleared. */
5934 if ((stash = CvSTASH(sv)))
5935 sv_del_backref(MUTABLE_SV(stash), sv);
5938 if (PL_last_swash_hv == (const HV *)sv) {
5939 PL_last_swash_hv = NULL;
5941 Perl_hv_kill_backrefs(aTHX_ MUTABLE_HV(sv));
5942 hv_undef(MUTABLE_HV(sv));
5945 if (PL_comppad == MUTABLE_AV(sv)) {
5949 av_undef(MUTABLE_AV(sv));
5952 if (LvTYPE(sv) == 'T') { /* for tie: return HE to pool */
5953 SvREFCNT_dec(HeKEY_sv((HE*)LvTARG(sv)));
5954 HeNEXT((HE*)LvTARG(sv)) = PL_hv_fetch_ent_mh;
5955 PL_hv_fetch_ent_mh = (HE*)LvTARG(sv);
5957 else if (LvTYPE(sv) != 't') /* unless tie: unrefcnted fake SV** */
5958 SvREFCNT_dec(LvTARG(sv));
5960 if (isGV_with_GP(sv)) {
5961 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
5962 && HvNAME_get(stash))
5963 mro_method_changed_in(stash);
5964 gp_free(MUTABLE_GV(sv));
5966 unshare_hek(GvNAME_HEK(sv));
5967 /* If we're in a stash, we don't own a reference to it. However it does
5968 have a back reference to us, which needs to be cleared. */
5969 if (!SvVALID(sv) && (stash = GvSTASH(sv)))
5970 sv_del_backref(MUTABLE_SV(stash), sv);
5972 /* FIXME. There are probably more unreferenced pointers to SVs in the
5973 interpreter struct that we should check and tidy in a similar
5975 if ((const GV *)sv == PL_last_in_gv)
5976 PL_last_in_gv = NULL;
5982 /* Don't bother with SvOOK_off(sv); as we're only going to free it. */
5985 SvOOK_offset(sv, offset);
5986 SvPV_set(sv, SvPVX_mutable(sv) - offset);
5987 /* Don't even bother with turning off the OOK flag. */
5992 SV * const target = SvRV(sv);
5994 sv_del_backref(target, sv);
5996 SvREFCNT_dec(target);
5999 #ifdef PERL_OLD_COPY_ON_WRITE
6000 else if (SvPVX_const(sv)
6001 && !(SvTYPE(sv) == SVt_PVIO && !(IoFLAGS(sv) & IOf_FAKE_DIRP))) {
6004 PerlIO_printf(Perl_debug_log, "Copy on write: clear\n");
6008 sv_release_COW(sv, SvPVX_const(sv), SV_COW_NEXT_SV(sv));
6010 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
6014 } else if (SvLEN(sv)) {
6015 Safefree(SvPVX_const(sv));
6019 else if (SvPVX_const(sv) && SvLEN(sv)
6020 && !(SvTYPE(sv) == SVt_PVIO && !(IoFLAGS(sv) & IOf_FAKE_DIRP)))
6021 Safefree(SvPVX_mutable(sv));
6022 else if (SvPVX_const(sv) && SvREADONLY(sv) && SvFAKE(sv)) {
6023 unshare_hek(SvSHARED_HEK_FROM_PV(SvPVX_const(sv)));
6032 SvFLAGS(sv) &= SVf_BREAK;
6033 SvFLAGS(sv) |= SVTYPEMASK;
6035 if (sv_type_details->arena) {
6036 del_body(((char *)SvANY(sv) + sv_type_details->offset),
6037 &PL_body_roots[type]);
6039 else if (sv_type_details->body_size) {
6040 safefree(SvANY(sv));
6045 =for apidoc sv_newref
6047 Increment an SV's reference count. Use the C<SvREFCNT_inc()> wrapper
6054 Perl_sv_newref(pTHX_ SV *const sv)
6056 PERL_UNUSED_CONTEXT;
6065 Decrement an SV's reference count, and if it drops to zero, call
6066 C<sv_clear> to invoke destructors and free up any memory used by
6067 the body; finally, deallocate the SV's head itself.
6068 Normally called via a wrapper macro C<SvREFCNT_dec>.
6074 Perl_sv_free(pTHX_ SV *const sv)
6079 if (SvREFCNT(sv) == 0) {
6080 if (SvFLAGS(sv) & SVf_BREAK)
6081 /* this SV's refcnt has been artificially decremented to
6082 * trigger cleanup */
6084 if (PL_in_clean_all) /* All is fair */
6086 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
6087 /* make sure SvREFCNT(sv)==0 happens very seldom */
6088 SvREFCNT(sv) = (~(U32)0)/2;
6091 if (ckWARN_d(WARN_INTERNAL)) {
6092 #ifdef DEBUG_LEAKING_SCALARS_FORK_DUMP
6093 Perl_dump_sv_child(aTHX_ sv);
6095 #ifdef DEBUG_LEAKING_SCALARS
6098 #ifdef DEBUG_LEAKING_SCALARS_ABORT
6099 if (PL_warnhook == PERL_WARNHOOK_FATAL
6100 || ckDEAD(packWARN(WARN_INTERNAL))) {
6101 /* Don't let Perl_warner cause us to escape our fate: */
6105 /* This may not return: */
6106 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
6107 "Attempt to free unreferenced scalar: SV 0x%"UVxf
6108 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
6111 #ifdef DEBUG_LEAKING_SCALARS_ABORT
6116 if (--(SvREFCNT(sv)) > 0)
6118 Perl_sv_free2(aTHX_ sv);
6122 Perl_sv_free2(pTHX_ SV *const sv)
6126 PERL_ARGS_ASSERT_SV_FREE2;
6130 Perl_ck_warner_d(aTHX_ packWARN(WARN_DEBUGGING),
6131 "Attempt to free temp prematurely: SV 0x%"UVxf
6132 pTHX__FORMAT, PTR2UV(sv) pTHX__VALUE);
6136 if (SvREADONLY(sv) && SvIMMORTAL(sv)) {
6137 /* make sure SvREFCNT(sv)==0 happens very seldom */
6138 SvREFCNT(sv) = (~(U32)0)/2;
6149 Returns the length of the string in the SV. Handles magic and type
6150 coercion. See also C<SvCUR>, which gives raw access to the xpv_cur slot.
6156 Perl_sv_len(pTHX_ register SV *const sv)
6164 len = mg_length(sv);
6166 (void)SvPV_const(sv, len);
6171 =for apidoc sv_len_utf8
6173 Returns the number of characters in the string in an SV, counting wide
6174 UTF-8 bytes as a single character. Handles magic and type coercion.
6180 * The length is cached in PERL_MAGIC_utf8, in the mg_len field. Also the
6181 * mg_ptr is used, by sv_pos_u2b() and sv_pos_b2u() - see the comments below.
6182 * (Note that the mg_len is not the length of the mg_ptr field.
6183 * This allows the cache to store the character length of the string without
6184 * needing to malloc() extra storage to attach to the mg_ptr.)
6189 Perl_sv_len_utf8(pTHX_ register SV *const sv)
6195 return mg_length(sv);
6199 const U8 *s = (U8*)SvPV_const(sv, len);
6203 MAGIC *mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_utf8) : NULL;
6205 if (mg && (mg->mg_len != -1 || mg->mg_ptr)) {
6206 if (mg->mg_len != -1)
6209 /* We can use the offset cache for a headstart.
6210 The longer value is stored in the first pair. */
6211 STRLEN *cache = (STRLEN *) mg->mg_ptr;
6213 ulen = cache[0] + Perl_utf8_length(aTHX_ s + cache[1],
6217 if (PL_utf8cache < 0) {
6218 const STRLEN real = Perl_utf8_length(aTHX_ s, s + len);
6219 assert_uft8_cache_coherent("sv_len_utf8", ulen, real, sv);
6223 ulen = Perl_utf8_length(aTHX_ s, s + len);
6224 utf8_mg_len_cache_update(sv, &mg, ulen);
6228 return Perl_utf8_length(aTHX_ s, s + len);
6232 /* Walk forwards to find the byte corresponding to the passed in UTF-8
6235 S_sv_pos_u2b_forwards(const U8 *const start, const U8 *const send,
6236 STRLEN *const uoffset_p, bool *const at_end)
6238 const U8 *s = start;
6239 STRLEN uoffset = *uoffset_p;
6241 PERL_ARGS_ASSERT_SV_POS_U2B_FORWARDS;
6243 while (s < send && uoffset) {
6250 else if (s > send) {
6252 /* This is the existing behaviour. Possibly it should be a croak, as
6253 it's actually a bounds error */
6256 *uoffset_p -= uoffset;
6260 /* Given the length of the string in both bytes and UTF-8 characters, decide
6261 whether to walk forwards or backwards to find the byte corresponding to
6262 the passed in UTF-8 offset. */
6264 S_sv_pos_u2b_midway(const U8 *const start, const U8 *send,
6265 STRLEN uoffset, const STRLEN uend)
6267 STRLEN backw = uend - uoffset;
6269 PERL_ARGS_ASSERT_SV_POS_U2B_MIDWAY;
6271 if (uoffset < 2 * backw) {
6272 /* The assumption is that going forwards is twice the speed of going
6273 forward (that's where the 2 * backw comes from).
6274 (The real figure of course depends on the UTF-8 data.) */
6275 const U8 *s = start;
6277 while (s < send && uoffset--)
6287 while (UTF8_IS_CONTINUATION(*send))
6290 return send - start;
6293 /* For the string representation of the given scalar, find the byte
6294 corresponding to the passed in UTF-8 offset. uoffset0 and boffset0
6295 give another position in the string, *before* the sought offset, which
6296 (which is always true, as 0, 0 is a valid pair of positions), which should
6297 help reduce the amount of linear searching.
6298 If *mgp is non-NULL, it should point to the UTF-8 cache magic, which
6299 will be used to reduce the amount of linear searching. The cache will be
6300 created if necessary, and the found value offered to it for update. */
6302 S_sv_pos_u2b_cached(pTHX_ SV *const sv, MAGIC **const mgp, const U8 *const start,
6303 const U8 *const send, STRLEN uoffset,
6304 STRLEN uoffset0, STRLEN boffset0)
6306 STRLEN boffset = 0; /* Actually always set, but let's keep gcc happy. */
6308 bool at_end = FALSE;
6310 PERL_ARGS_ASSERT_SV_POS_U2B_CACHED;
6312 assert (uoffset >= uoffset0);
6319 && (*mgp || (SvTYPE(sv) >= SVt_PVMG &&
6320 (*mgp = mg_find(sv, PERL_MAGIC_utf8))))) {
6321 if ((*mgp)->mg_ptr) {
6322 STRLEN *cache = (STRLEN *) (*mgp)->mg_ptr;
6323 if (cache[0] == uoffset) {
6324 /* An exact match. */
6327 if (cache[2] == uoffset) {
6328 /* An exact match. */
6332 if (cache[0] < uoffset) {
6333 /* The cache already knows part of the way. */
6334 if (cache[0] > uoffset0) {
6335 /* The cache knows more than the passed in pair */
6336 uoffset0 = cache[0];
6337 boffset0 = cache[1];
6339 if ((*mgp)->mg_len != -1) {
6340 /* And we know the end too. */
6342 + sv_pos_u2b_midway(start + boffset0, send,
6344 (*mgp)->mg_len - uoffset0);
6346 uoffset -= uoffset0;
6348 + sv_pos_u2b_forwards(start + boffset0,
6349 send, &uoffset, &at_end);
6350 uoffset += uoffset0;
6353 else if (cache[2] < uoffset) {
6354 /* We're between the two cache entries. */
6355 if (cache[2] > uoffset0) {
6356 /* and the cache knows more than the passed in pair */
6357 uoffset0 = cache[2];
6358 boffset0 = cache[3];
6362 + sv_pos_u2b_midway(start + boffset0,
6365 cache[0] - uoffset0);
6368 + sv_pos_u2b_midway(start + boffset0,
6371 cache[2] - uoffset0);
6375 else if ((*mgp)->mg_len != -1) {
6376 /* If we can take advantage of a passed in offset, do so. */
6377 /* In fact, offset0 is either 0, or less than offset, so don't
6378 need to worry about the other possibility. */
6380 + sv_pos_u2b_midway(start + boffset0, send,
6382 (*mgp)->mg_len - uoffset0);
6387 if (!found || PL_utf8cache < 0) {
6388 STRLEN real_boffset;
6389 uoffset -= uoffset0;
6390 real_boffset = boffset0 + sv_pos_u2b_forwards(start + boffset0,
6391 send, &uoffset, &at_end);
6392 uoffset += uoffset0;
6394 if (found && PL_utf8cache < 0)
6395 assert_uft8_cache_coherent("sv_pos_u2b_cache", boffset,
6397 boffset = real_boffset;
6402 utf8_mg_len_cache_update(sv, mgp, uoffset);
6404 utf8_mg_pos_cache_update(sv, mgp, boffset, uoffset, send - start);
6411 =for apidoc sv_pos_u2b_flags
6413 Converts the value pointed to by offsetp from a count of UTF-8 chars from
6414 the start of the string, to a count of the equivalent number of bytes; if
6415 lenp is non-zero, it does the same to lenp, but this time starting from
6416 the offset, rather than from the start of the string. Handles type coercion.
6417 I<flags> is passed to C<SvPV_flags>, and usually should be
6418 C<SV_GMAGIC|SV_CONST_RETURN> to handle magic.
6424 * sv_pos_u2b_flags() uses, like sv_pos_b2u(), the mg_ptr of the potential
6425 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6426 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
6431 Perl_sv_pos_u2b_flags(pTHX_ SV *const sv, STRLEN uoffset, STRLEN *const lenp,
6438 PERL_ARGS_ASSERT_SV_POS_U2B_FLAGS;
6440 start = (U8*)SvPV_flags(sv, len, flags);
6442 const U8 * const send = start + len;
6444 boffset = sv_pos_u2b_cached(sv, &mg, start, send, uoffset, 0, 0);
6447 && *lenp /* don't bother doing work for 0, as its bytes equivalent
6448 is 0, and *lenp is already set to that. */) {
6449 /* Convert the relative offset to absolute. */
6450 const STRLEN uoffset2 = uoffset + *lenp;
6451 const STRLEN boffset2
6452 = sv_pos_u2b_cached(sv, &mg, start, send, uoffset2,
6453 uoffset, boffset) - boffset;
6467 =for apidoc sv_pos_u2b
6469 Converts the value pointed to by offsetp from a count of UTF-8 chars from
6470 the start of the string, to a count of the equivalent number of bytes; if
6471 lenp is non-zero, it does the same to lenp, but this time starting from
6472 the offset, rather than from the start of the string. Handles magic and
6475 Use C<sv_pos_u2b_flags> in preference, which correctly handles strings longer
6482 * sv_pos_u2b() uses, like sv_pos_b2u(), the mg_ptr of the potential
6483 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6484 * byte offsets. See also the comments of S_utf8_mg_pos_cache_update().
6488 /* This function is subject to size and sign problems */
6491 Perl_sv_pos_u2b(pTHX_ register SV *const sv, I32 *const offsetp, I32 *const lenp)
6493 PERL_ARGS_ASSERT_SV_POS_U2B;
6496 STRLEN ulen = (STRLEN)*lenp;
6497 *offsetp = (I32)sv_pos_u2b_flags(sv, (STRLEN)*offsetp, &ulen,
6498 SV_GMAGIC|SV_CONST_RETURN);
6501 *offsetp = (I32)sv_pos_u2b_flags(sv, (STRLEN)*offsetp, NULL,
6502 SV_GMAGIC|SV_CONST_RETURN);
6507 S_utf8_mg_len_cache_update(pTHX_ SV *const sv, MAGIC **const mgp,
6510 PERL_ARGS_ASSERT_UTF8_MG_LEN_CACHE_UPDATE;
6514 if (!*mgp && (SvTYPE(sv) < SVt_PVMG ||
6515 !(*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
6516 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, &PL_vtbl_utf8, 0, 0);
6520 (*mgp)->mg_len = ulen;
6521 /* For now, treat "overflowed" as "still unknown". See RT #72924. */
6522 if (ulen != (STRLEN) (*mgp)->mg_len)
6523 (*mgp)->mg_len = -1;
6526 /* Create and update the UTF8 magic offset cache, with the proffered utf8/
6527 byte length pairing. The (byte) length of the total SV is passed in too,
6528 as blen, because for some (more esoteric) SVs, the call to SvPV_const()
6529 may not have updated SvCUR, so we can't rely on reading it directly.
6531 The proffered utf8/byte length pairing isn't used if the cache already has
6532 two pairs, and swapping either for the proffered pair would increase the
6533 RMS of the intervals between known byte offsets.
6535 The cache itself consists of 4 STRLEN values
6536 0: larger UTF-8 offset
6537 1: corresponding byte offset
6538 2: smaller UTF-8 offset
6539 3: corresponding byte offset
6541 Unused cache pairs have the value 0, 0.
6542 Keeping the cache "backwards" means that the invariant of
6543 cache[0] >= cache[2] is maintained even with empty slots, which means that
6544 the code that uses it doesn't need to worry if only 1 entry has actually
6545 been set to non-zero. It also makes the "position beyond the end of the
6546 cache" logic much simpler, as the first slot is always the one to start
6550 S_utf8_mg_pos_cache_update(pTHX_ SV *const sv, MAGIC **const mgp, const STRLEN byte,
6551 const STRLEN utf8, const STRLEN blen)
6555 PERL_ARGS_ASSERT_UTF8_MG_POS_CACHE_UPDATE;
6560 if (!*mgp && (SvTYPE(sv) < SVt_PVMG ||
6561 !(*mgp = mg_find(sv, PERL_MAGIC_utf8)))) {
6562 *mgp = sv_magicext(sv, 0, PERL_MAGIC_utf8, (MGVTBL*)&PL_vtbl_utf8, 0,
6564 (*mgp)->mg_len = -1;
6568 if (!(cache = (STRLEN *)(*mgp)->mg_ptr)) {
6569 Newxz(cache, PERL_MAGIC_UTF8_CACHESIZE * 2, STRLEN);
6570 (*mgp)->mg_ptr = (char *) cache;
6574 if (PL_utf8cache < 0 && SvPOKp(sv)) {
6575 /* SvPOKp() because it's possible that sv has string overloading, and
6576 therefore is a reference, hence SvPVX() is actually a pointer.
6577 This cures the (very real) symptoms of RT 69422, but I'm not actually
6578 sure whether we should even be caching the results of UTF-8
6579 operations on overloading, given that nothing stops overloading
6580 returning a different value every time it's called. */
6581 const U8 *start = (const U8 *) SvPVX_const(sv);
6582 const STRLEN realutf8 = utf8_length(start, start + byte);
6584 assert_uft8_cache_coherent("utf8_mg_pos_cache_update", utf8, realutf8,
6588 /* Cache is held with the later position first, to simplify the code
6589 that deals with unbounded ends. */
6591 ASSERT_UTF8_CACHE(cache);
6592 if (cache[1] == 0) {
6593 /* Cache is totally empty */
6596 } else if (cache[3] == 0) {
6597 if (byte > cache[1]) {
6598 /* New one is larger, so goes first. */
6599 cache[2] = cache[0];
6600 cache[3] = cache[1];
6608 #define THREEWAY_SQUARE(a,b,c,d) \
6609 ((float)((d) - (c))) * ((float)((d) - (c))) \
6610 + ((float)((c) - (b))) * ((float)((c) - (b))) \
6611 + ((float)((b) - (a))) * ((float)((b) - (a)))
6613 /* Cache has 2 slots in use, and we know three potential pairs.
6614 Keep the two that give the lowest RMS distance. Do the
6615 calcualation in bytes simply because we always know the byte
6616 length. squareroot has the same ordering as the positive value,
6617 so don't bother with the actual square root. */
6618 const float existing = THREEWAY_SQUARE(0, cache[3], cache[1], blen);
6619 if (byte > cache[1]) {
6620 /* New position is after the existing pair of pairs. */
6621 const float keep_earlier
6622 = THREEWAY_SQUARE(0, cache[3], byte, blen);
6623 const float keep_later
6624 = THREEWAY_SQUARE(0, cache[1], byte, blen);
6626 if (keep_later < keep_earlier) {
6627 if (keep_later < existing) {
6628 cache[2] = cache[0];
6629 cache[3] = cache[1];
6635 if (keep_earlier < existing) {
6641 else if (byte > cache[3]) {
6642 /* New position is between the existing pair of pairs. */
6643 const float keep_earlier
6644 = THREEWAY_SQUARE(0, cache[3], byte, blen);
6645 const float keep_later
6646 = THREEWAY_SQUARE(0, byte, cache[1], blen);
6648 if (keep_later < keep_earlier) {
6649 if (keep_later < existing) {
6655 if (keep_earlier < existing) {
6662 /* New position is before the existing pair of pairs. */
6663 const float keep_earlier
6664 = THREEWAY_SQUARE(0, byte, cache[3], blen);
6665 const float keep_later
6666 = THREEWAY_SQUARE(0, byte, cache[1], blen);
6668 if (keep_later < keep_earlier) {
6669 if (keep_later < existing) {
6675 if (keep_earlier < existing) {
6676 cache[0] = cache[2];
6677 cache[1] = cache[3];
6684 ASSERT_UTF8_CACHE(cache);
6687 /* We already know all of the way, now we may be able to walk back. The same
6688 assumption is made as in S_sv_pos_u2b_midway(), namely that walking
6689 backward is half the speed of walking forward. */
6691 S_sv_pos_b2u_midway(pTHX_ const U8 *const s, const U8 *const target,
6692 const U8 *end, STRLEN endu)
6694 const STRLEN forw = target - s;
6695 STRLEN backw = end - target;
6697 PERL_ARGS_ASSERT_SV_POS_B2U_MIDWAY;
6699 if (forw < 2 * backw) {
6700 return utf8_length(s, target);
6703 while (end > target) {
6705 while (UTF8_IS_CONTINUATION(*end)) {
6714 =for apidoc sv_pos_b2u
6716 Converts the value pointed to by offsetp from a count of bytes from the
6717 start of the string, to a count of the equivalent number of UTF-8 chars.
6718 Handles magic and type coercion.
6724 * sv_pos_b2u() uses, like sv_pos_u2b(), the mg_ptr of the potential
6725 * PERL_MAGIC_utf8 of the sv to store the mapping between UTF-8 and
6730 Perl_sv_pos_b2u(pTHX_ register SV *const sv, I32 *const offsetp)
6733 const STRLEN byte = *offsetp;
6734 STRLEN len = 0; /* Actually always set, but let's keep gcc happy. */
6740 PERL_ARGS_ASSERT_SV_POS_B2U;
6745 s = (const U8*)SvPV_const(sv, blen);
6748 Perl_croak(aTHX_ "panic: sv_pos_b2u: bad byte offset");
6754 && SvTYPE(sv) >= SVt_PVMG
6755 && (mg = mg_find(sv, PERL_MAGIC_utf8)))
6758 STRLEN * const cache = (STRLEN *) mg->mg_ptr;
6759 if (cache[1] == byte) {
6760 /* An exact match. */
6761 *offsetp = cache[0];
6764 if (cache[3] == byte) {
6765 /* An exact match. */
6766 *offsetp = cache[2];
6770 if (cache[1] < byte) {
6771 /* We already know part of the way. */
6772 if (mg->mg_len != -1) {
6773 /* Actually, we know the end too. */
6775 + S_sv_pos_b2u_midway(aTHX_ s + cache[1], send,
6776 s + blen, mg->mg_len - cache[0]);
6778 len = cache[0] + utf8_length(s + cache[1], send);
6781 else if (cache[3] < byte) {
6782 /* We're between the two cached pairs, so we do the calculation
6783 offset by the byte/utf-8 positions for the earlier pair,
6784 then add the utf-8 characters from the string start to
6786 len = S_sv_pos_b2u_midway(aTHX_ s + cache[3], send,
6787 s + cache[1], cache[0] - cache[2])
6791 else { /* cache[3] > byte */
6792 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + cache[3],
6796 ASSERT_UTF8_CACHE(cache);
6798 } else if (mg->mg_len != -1) {
6799 len = S_sv_pos_b2u_midway(aTHX_ s, send, s + blen, mg->mg_len);
6803 if (!found || PL_utf8cache < 0) {
6804 const STRLEN real_len = utf8_length(s, send);
6806 if (found && PL_utf8cache < 0)
6807 assert_uft8_cache_coherent("sv_pos_b2u", len, real_len, sv);
6814 utf8_mg_len_cache_update(sv, &mg, len);
6816 utf8_mg_pos_cache_update(sv, &mg, byte, len, blen);
6821 S_assert_uft8_cache_coherent(pTHX_ const char *const func, STRLEN from_cache,
6822 STRLEN real, SV *const sv)
6824 PERL_ARGS_ASSERT_ASSERT_UFT8_CACHE_COHERENT;
6826 /* As this is debugging only code, save space by keeping this test here,
6827 rather than inlining it in all the callers. */
6828 if (from_cache == real)
6831 /* Need to turn the assertions off otherwise we may recurse infinitely
6832 while printing error messages. */
6833 SAVEI8(PL_utf8cache);
6835 Perl_croak(aTHX_ "panic: %s cache %"UVuf" real %"UVuf" for %"SVf,
6836 func, (UV) from_cache, (UV) real, SVfARG(sv));
6842 Returns a boolean indicating whether the strings in the two SVs are
6843 identical. Is UTF-8 and 'use bytes' aware, handles get magic, and will
6844 coerce its args to strings if necessary.
6846 =for apidoc sv_eq_flags
6848 Returns a boolean indicating whether the strings in the two SVs are
6849 identical. Is UTF-8 and 'use bytes' aware and coerces its args to strings
6850 if necessary. If the flags include SV_GMAGIC, it handles get-magic, too.
6856 Perl_sv_eq_flags(pTHX_ register SV *sv1, register SV *sv2, const I32 flags)
6865 SV* svrecode = NULL;
6872 /* if pv1 and pv2 are the same, second SvPV_const call may
6873 * invalidate pv1 (if we are handling magic), so we may need to
6875 if (sv1 == sv2 && flags & SV_GMAGIC
6876 && (SvTHINKFIRST(sv1) || SvGMAGICAL(sv1))) {
6877 pv1 = SvPV_const(sv1, cur1);
6878 sv1 = newSVpvn_flags(pv1, cur1, SVs_TEMP | SvUTF8(sv2));
6880 pv1 = SvPV_flags_const(sv1, cur1, flags);
6888 pv2 = SvPV_flags_const(sv2, cur2, flags);
6890 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
6891 /* Differing utf8ness.
6892 * Do not UTF8size the comparands as a side-effect. */
6895 svrecode = newSVpvn(pv2, cur2);
6896 sv_recode_to_utf8(svrecode, PL_encoding);
6897 pv2 = SvPV_const(svrecode, cur2);
6900 svrecode = newSVpvn(pv1, cur1);
6901 sv_recode_to_utf8(svrecode, PL_encoding);
6902 pv1 = SvPV_const(svrecode, cur1);
6904 /* Now both are in UTF-8. */
6906 SvREFCNT_dec(svrecode);
6911 bool is_utf8 = TRUE;
6914 /* sv1 is the UTF-8 one,
6915 * if is equal it must be downgrade-able */
6916 char * const pv = (char*)bytes_from_utf8((const U8*)pv1,
6922 /* sv2 is the UTF-8 one,
6923 * if is equal it must be downgrade-able */
6924 char * const pv = (char *)bytes_from_utf8((const U8*)pv2,
6930 /* Downgrade not possible - cannot be eq */
6938 eq = (pv1 == pv2) || memEQ(pv1, pv2, cur1);
6940 SvREFCNT_dec(svrecode);
6950 Compares the strings in two SVs. Returns -1, 0, or 1 indicating whether the
6951 string in C<sv1> is less than, equal to, or greater than the string in
6952 C<sv2>. Is UTF-8 and 'use bytes' aware, handles get magic, and will
6953 coerce its args to strings if necessary. See also C<sv_cmp_locale>.
6955 =for apidoc sv_cmp_flags
6957 Compares the strings in two SVs. Returns -1, 0, or 1 indicating whether the
6958 string in C<sv1> is less than, equal to, or greater than the string in
6959 C<sv2>. Is UTF-8 and 'use bytes' aware and will coerce its args to strings
6960 if necessary. If the flags include SV_GMAGIC, it handles get magic. See
6961 also C<sv_cmp_locale_flags>.
6967 Perl_sv_cmp(pTHX_ register SV *const sv1, register SV *const sv2)
6969 return sv_cmp_flags(sv1, sv2, SV_GMAGIC);
6973 Perl_sv_cmp_flags(pTHX_ register SV *const sv1, register SV *const sv2, const I32 flags)
6977 const char *pv1, *pv2;
6980 SV *svrecode = NULL;
6987 pv1 = SvPV_flags_const(sv1, cur1, flags);
6994 pv2 = SvPV_flags_const(sv2, cur2, flags);
6996 if (cur1 && cur2 && SvUTF8(sv1) != SvUTF8(sv2) && !IN_BYTES) {
6997 /* Differing utf8ness.
6998 * Do not UTF8size the comparands as a side-effect. */
7001 svrecode = newSVpvn(pv2, cur2);
7002 sv_recode_to_utf8(svrecode, PL_encoding);
7003 pv2 = SvPV_const(svrecode, cur2);
7006 pv2 = tpv = (char*)bytes_to_utf8((const U8*)pv2, &cur2);
7011 svrecode = newSVpvn(pv1, cur1);
7012 sv_recode_to_utf8(svrecode, PL_encoding);
7013 pv1 = SvPV_const(svrecode, cur1);
7016 pv1 = tpv = (char*)bytes_to_utf8((const U8*)pv1, &cur1);
7022 cmp = cur2 ? -1 : 0;
7026 const I32 retval = memcmp((const void*)pv1, (const void*)pv2, cur1 < cur2 ? cur1 : cur2);
7029 cmp = retval < 0 ? -1 : 1;
7030 } else if (cur1 == cur2) {
7033 cmp = cur1 < cur2 ? -1 : 1;
7037 SvREFCNT_dec(svrecode);
7045 =for apidoc sv_cmp_locale
7047 Compares the strings in two SVs in a locale-aware manner. Is UTF-8 and
7048 'use bytes' aware, handles get magic, and will coerce its args to strings
7049 if necessary. See also C<sv_cmp>.
7051 =for apidoc sv_cmp_locale_flags
7053 Compares the strings in two SVs in a locale-aware manner. Is UTF-8 and
7054 'use bytes' aware and will coerce its args to strings if necessary. If the
7055 flags contain SV_GMAGIC, it handles get magic. See also C<sv_cmp_flags>.
7061 Perl_sv_cmp_locale(pTHX_ register SV *const sv1, register SV *const sv2)
7063 return sv_cmp_locale_flags(sv1, sv2, SV_GMAGIC);
7067 Perl_sv_cmp_locale_flags(pTHX_ register SV *const sv1, register SV *const sv2, const I32 flags)
7070 #ifdef USE_LOCALE_COLLATE
7076 if (PL_collation_standard)
7080 pv1 = sv1 ? sv_collxfrm_flags(sv1, &len1, flags) : (char *) NULL;
7082 pv2 = sv2 ? sv_collxfrm_flags(sv2, &len2, flags) : (char *) NULL;
7084 if (!pv1 || !len1) {
7095 retval = memcmp((void*)pv1, (void*)pv2, len1 < len2 ? len1 : len2);
7098 return retval < 0 ? -1 : 1;
7101 * When the result of collation is equality, that doesn't mean
7102 * that there are no differences -- some locales exclude some
7103 * characters from consideration. So to avoid false equalities,
7104 * we use the raw string as a tiebreaker.
7110 #endif /* USE_LOCALE_COLLATE */
7112 return sv_cmp(sv1, sv2);
7116 #ifdef USE_LOCALE_COLLATE
7119 =for apidoc sv_collxfrm
7121 This calls C<sv_collxfrm_flags> with the SV_GMAGIC flag. See
7122 C<sv_collxfrm_flags>.
7124 =for apidoc sv_collxfrm_flags
7126 Add Collate Transform magic to an SV if it doesn't already have it. If the
7127 flags contain SV_GMAGIC, it handles get-magic.
7129 Any scalar variable may carry PERL_MAGIC_collxfrm magic that contains the
7130 scalar data of the variable, but transformed to such a format that a normal
7131 memory comparison can be used to compare the data according to the locale
7138 Perl_sv_collxfrm_flags(pTHX_ SV *const sv, STRLEN *const nxp, const I32 flags)
7143 PERL_ARGS_ASSERT_SV_COLLXFRM_FLAGS;
7145 mg = SvMAGICAL(sv) ? mg_find(sv, PERL_MAGIC_collxfrm) : (MAGIC *) NULL;
7146 if (!mg || !mg->mg_ptr || *(U32*)mg->mg_ptr != PL_collation_ix) {
7152 Safefree(mg->mg_ptr);
7153 s = SvPV_flags_const(sv, len, flags);
7154 if ((xf = mem_collxfrm(s, len, &xlen))) {
7156 #ifdef PERL_OLD_COPY_ON_WRITE
7158 sv_force_normal_flags(sv, 0);
7160 mg = sv_magicext(sv, 0, PERL_MAGIC_collxfrm, &PL_vtbl_collxfrm,
7174 if (mg && mg->mg_ptr) {
7176 return mg->mg_ptr + sizeof(PL_collation_ix);
7184 #endif /* USE_LOCALE_COLLATE */
7189 Get a line from the filehandle and store it into the SV, optionally
7190 appending to the currently-stored string.
7196 Perl_sv_gets(pTHX_ register SV *const sv, register PerlIO *const fp, I32 append)
7201 register STDCHAR rslast;
7202 register STDCHAR *bp;
7207 PERL_ARGS_ASSERT_SV_GETS;
7209 if (SvTHINKFIRST(sv))
7210 sv_force_normal_flags(sv, append ? 0 : SV_COW_DROP_PV);
7211 /* XXX. If you make this PVIV, then copy on write can copy scalars read
7213 However, perlbench says it's slower, because the existing swipe code
7214 is faster than copy on write.
7215 Swings and roundabouts. */
7216 SvUPGRADE(sv, SVt_PV);
7221 if (PerlIO_isutf8(fp)) {
7223 sv_utf8_upgrade_nomg(sv);
7224 sv_pos_u2b(sv,&append,0);
7226 } else if (SvUTF8(sv)) {
7227 SV * const tsv = newSV(0);
7228 sv_gets(tsv, fp, 0);
7229 sv_utf8_upgrade_nomg(tsv);
7230 SvCUR_set(sv,append);
7233 goto return_string_or_null;
7241 if (PerlIO_isutf8(fp))
7244 if (IN_PERL_COMPILETIME) {
7245 /* we always read code in line mode */
7249 else if (RsSNARF(PL_rs)) {
7250 /* If it is a regular disk file use size from stat() as estimate
7251 of amount we are going to read -- may result in mallocing
7252 more memory than we really need if the layers below reduce
7253 the size we read (e.g. CRLF or a gzip layer).
7256 if (!PerlLIO_fstat(PerlIO_fileno(fp), &st) && S_ISREG(st.st_mode)) {
7257 const Off_t offset = PerlIO_tell(fp);
7258 if (offset != (Off_t) -1 && st.st_size + append > offset) {
7259 (void) SvGROW(sv, (STRLEN)((st.st_size - offset) + append + 1));
7265 else if (RsRECORD(PL_rs)) {
7273 /* Grab the size of the record we're getting */
7274 recsize = SvUV(SvRV(PL_rs)); /* RsRECORD() guarantees > 0. */
7275 buffer = SvGROW(sv, (STRLEN)(recsize + append + 1)) + append;
7278 /* VMS wants read instead of fread, because fread doesn't respect */
7279 /* RMS record boundaries. This is not necessarily a good thing to be */
7280 /* doing, but we've got no other real choice - except avoid stdio
7281 as implementation - perhaps write a :vms layer ?
7283 fd = PerlIO_fileno(fp);
7284 if (fd == -1) { /* in-memory file from PerlIO::Scalar */
7285 bytesread = PerlIO_read(fp, buffer, recsize);
7288 bytesread = PerlLIO_read(fd, buffer, recsize);
7291 bytesread = PerlIO_read(fp, buffer, recsize);
7295 SvCUR_set(sv, bytesread + append);
7296 buffer[bytesread] = '\0';
7297 goto return_string_or_null;
7299 else if (RsPARA(PL_rs)) {
7305 /* Get $/ i.e. PL_rs into same encoding as stream wants */
7306 if (PerlIO_isutf8(fp)) {
7307 rsptr = SvPVutf8(PL_rs, rslen);
7310 if (SvUTF8(PL_rs)) {
7311 if (!sv_utf8_downgrade(PL_rs, TRUE)) {
7312 Perl_croak(aTHX_ "Wide character in $/");
7315 rsptr = SvPV_const(PL_rs, rslen);
7319 rslast = rslen ? rsptr[rslen - 1] : '\0';
7321 if (rspara) { /* have to do this both before and after */
7322 do { /* to make sure file boundaries work right */
7325 i = PerlIO_getc(fp);
7329 PerlIO_ungetc(fp,i);
7335 /* See if we know enough about I/O mechanism to cheat it ! */
7337 /* This used to be #ifdef test - it is made run-time test for ease
7338 of abstracting out stdio interface. One call should be cheap
7339 enough here - and may even be a macro allowing compile
7343 if (PerlIO_fast_gets(fp)) {
7346 * We're going to steal some values from the stdio struct
7347 * and put EVERYTHING in the innermost loop into registers.
7349 register STDCHAR *ptr;
7353 #if defined(VMS) && defined(PERLIO_IS_STDIO)
7354 /* An ungetc()d char is handled separately from the regular
7355 * buffer, so we getc() it back out and stuff it in the buffer.
7357 i = PerlIO_getc(fp);
7358 if (i == EOF) return 0;
7359 *(--((*fp)->_ptr)) = (unsigned char) i;
7363 /* Here is some breathtakingly efficient cheating */
7365 cnt = PerlIO_get_cnt(fp); /* get count into register */
7366 /* make sure we have the room */
7367 if ((I32)(SvLEN(sv) - append) <= cnt + 1) {
7368 /* Not room for all of it
7369 if we are looking for a separator and room for some
7371 if (rslen && cnt > 80 && (I32)SvLEN(sv) > append) {
7372 /* just process what we have room for */
7373 shortbuffered = cnt - SvLEN(sv) + append + 1;
7374 cnt -= shortbuffered;
7378 /* remember that cnt can be negative */
7379 SvGROW(sv, (STRLEN)(append + (cnt <= 0 ? 2 : (cnt + 1))));
7384 bp = (STDCHAR*)SvPVX_const(sv) + append; /* move these two too to registers */
7385 ptr = (STDCHAR*)PerlIO_get_ptr(fp);
7386 DEBUG_P(PerlIO_printf(Perl_debug_log,
7387 "Screamer: entering, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7388 DEBUG_P(PerlIO_printf(Perl_debug_log,
7389 "Screamer: entering: PerlIO * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7390 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7391 PTR2UV(PerlIO_has_base(fp) ? PerlIO_get_base(fp) : 0)));
7396 while (cnt > 0) { /* this | eat */
7398 if ((*bp++ = *ptr++) == rslast) /* really | dust */
7399 goto thats_all_folks; /* screams | sed :-) */
7403 Copy(ptr, bp, cnt, char); /* this | eat */
7404 bp += cnt; /* screams | dust */
7405 ptr += cnt; /* louder | sed :-) */
7410 if (shortbuffered) { /* oh well, must extend */
7411 cnt = shortbuffered;
7413 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
7415 SvGROW(sv, SvLEN(sv) + append + cnt + 2);
7416 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
7420 DEBUG_P(PerlIO_printf(Perl_debug_log,
7421 "Screamer: going to getc, ptr=%"UVuf", cnt=%ld\n",
7422 PTR2UV(ptr),(long)cnt));
7423 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* deregisterize cnt and ptr */
7425 DEBUG_P(PerlIO_printf(Perl_debug_log,
7426 "Screamer: pre: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7427 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7428 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7430 /* This used to call 'filbuf' in stdio form, but as that behaves like
7431 getc when cnt <= 0 we use PerlIO_getc here to avoid introducing
7432 another abstraction. */
7433 i = PerlIO_getc(fp); /* get more characters */
7435 DEBUG_P(PerlIO_printf(Perl_debug_log,
7436 "Screamer: post: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7437 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7438 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7440 cnt = PerlIO_get_cnt(fp);
7441 ptr = (STDCHAR*)PerlIO_get_ptr(fp); /* reregisterize cnt and ptr */
7442 DEBUG_P(PerlIO_printf(Perl_debug_log,
7443 "Screamer: after getc, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7445 if (i == EOF) /* all done for ever? */
7446 goto thats_really_all_folks;
7448 bpx = bp - (STDCHAR*)SvPVX_const(sv); /* box up before relocation */
7450 SvGROW(sv, bpx + cnt + 2);
7451 bp = (STDCHAR*)SvPVX_const(sv) + bpx; /* unbox after relocation */
7453 *bp++ = (STDCHAR)i; /* store character from PerlIO_getc */
7455 if (rslen && (STDCHAR)i == rslast) /* all done for now? */
7456 goto thats_all_folks;
7460 if ((rslen > 1 && (STRLEN)(bp - (STDCHAR*)SvPVX_const(sv)) < rslen) ||
7461 memNE((char*)bp - rslen, rsptr, rslen))
7462 goto screamer; /* go back to the fray */
7463 thats_really_all_folks:
7465 cnt += shortbuffered;
7466 DEBUG_P(PerlIO_printf(Perl_debug_log,
7467 "Screamer: quitting, ptr=%"UVuf", cnt=%ld\n",PTR2UV(ptr),(long)cnt));
7468 PerlIO_set_ptrcnt(fp, (STDCHAR*)ptr, cnt); /* put these back or we're in trouble */
7469 DEBUG_P(PerlIO_printf(Perl_debug_log,
7470 "Screamer: end: FILE * thinks ptr=%"UVuf", cnt=%ld, base=%"UVuf"\n",
7471 PTR2UV(PerlIO_get_ptr(fp)), (long)PerlIO_get_cnt(fp),
7472 PTR2UV(PerlIO_has_base (fp) ? PerlIO_get_base(fp) : 0)));
7474 SvCUR_set(sv, bp - (STDCHAR*)SvPVX_const(sv)); /* set length */
7475 DEBUG_P(PerlIO_printf(Perl_debug_log,
7476 "Screamer: done, len=%ld, string=|%.*s|\n",
7477 (long)SvCUR(sv),(int)SvCUR(sv),SvPVX_const(sv)));
7481 /*The big, slow, and stupid way. */
7482 #ifdef USE_HEAP_INSTEAD_OF_STACK /* Even slower way. */
7483 STDCHAR *buf = NULL;
7484 Newx(buf, 8192, STDCHAR);
7492 register const STDCHAR * const bpe = buf + sizeof(buf);
7494 while ((i = PerlIO_getc(fp)) != EOF && (*bp++ = (STDCHAR)i) != rslast && bp < bpe)
7495 ; /* keep reading */
7499 cnt = PerlIO_read(fp,(char*)buf, sizeof(buf));
7500 /* Accomodate broken VAXC compiler, which applies U8 cast to
7501 * both args of ?: operator, causing EOF to change into 255
7504 i = (U8)buf[cnt - 1];
7510 cnt = 0; /* we do need to re-set the sv even when cnt <= 0 */
7512 sv_catpvn(sv, (char *) buf, cnt);
7514 sv_setpvn(sv, (char *) buf, cnt);
7516 if (i != EOF && /* joy */
7518 SvCUR(sv) < rslen ||
7519 memNE(SvPVX_const(sv) + SvCUR(sv) - rslen, rsptr, rslen)))
7523 * If we're reading from a TTY and we get a short read,
7524 * indicating that the user hit his EOF character, we need
7525 * to notice it now, because if we try to read from the TTY
7526 * again, the EOF condition will disappear.
7528 * The comparison of cnt to sizeof(buf) is an optimization
7529 * that prevents unnecessary calls to feof().
7533 if (!(cnt < (I32)sizeof(buf) && PerlIO_eof(fp)))
7537 #ifdef USE_HEAP_INSTEAD_OF_STACK
7542 if (rspara) { /* have to do this both before and after */
7543 while (i != EOF) { /* to make sure file boundaries work right */
7544 i = PerlIO_getc(fp);
7546 PerlIO_ungetc(fp,i);
7552 return_string_or_null:
7553 return (SvCUR(sv) - append) ? SvPVX(sv) : NULL;
7559 Auto-increment of the value in the SV, doing string to numeric conversion
7560 if necessary. Handles 'get' magic and operator overloading.
7566 Perl_sv_inc(pTHX_ register SV *const sv)
7575 =for apidoc sv_inc_nomg
7577 Auto-increment of the value in the SV, doing string to numeric conversion
7578 if necessary. Handles operator overloading. Skips handling 'get' magic.
7584 Perl_sv_inc_nomg(pTHX_ register SV *const sv)
7592 if (SvTHINKFIRST(sv)) {
7594 sv_force_normal_flags(sv, 0);
7595 if (SvREADONLY(sv)) {
7596 if (IN_PERL_RUNTIME)
7597 Perl_croak_no_modify(aTHX);
7601 if (SvAMAGIC(sv) && AMG_CALLun(sv,inc))
7603 i = PTR2IV(SvRV(sv));
7608 flags = SvFLAGS(sv);
7609 if ((flags & (SVp_NOK|SVp_IOK)) == SVp_NOK) {
7610 /* It's (privately or publicly) a float, but not tested as an
7611 integer, so test it to see. */
7613 flags = SvFLAGS(sv);
7615 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
7616 /* It's publicly an integer, or privately an integer-not-float */
7617 #ifdef PERL_PRESERVE_IVUV
7621 if (SvUVX(sv) == UV_MAX)
7622 sv_setnv(sv, UV_MAX_P1);
7624 (void)SvIOK_only_UV(sv);
7625 SvUV_set(sv, SvUVX(sv) + 1);
7627 if (SvIVX(sv) == IV_MAX)
7628 sv_setuv(sv, (UV)IV_MAX + 1);
7630 (void)SvIOK_only(sv);
7631 SvIV_set(sv, SvIVX(sv) + 1);
7636 if (flags & SVp_NOK) {
7637 const NV was = SvNVX(sv);
7638 if (NV_OVERFLOWS_INTEGERS_AT &&
7639 was >= NV_OVERFLOWS_INTEGERS_AT) {
7640 Perl_ck_warner(aTHX_ packWARN(WARN_IMPRECISION),
7641 "Lost precision when incrementing %" NVff " by 1",
7644 (void)SvNOK_only(sv);
7645 SvNV_set(sv, was + 1.0);
7649 if (!(flags & SVp_POK) || !*SvPVX_const(sv)) {
7650 if ((flags & SVTYPEMASK) < SVt_PVIV)
7651 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV ? SVt_PVIV : SVt_IV));
7652 (void)SvIOK_only(sv);
7657 while (isALPHA(*d)) d++;
7658 while (isDIGIT(*d)) d++;
7659 if (d < SvEND(sv)) {
7660 #ifdef PERL_PRESERVE_IVUV
7661 /* Got to punt this as an integer if needs be, but we don't issue
7662 warnings. Probably ought to make the sv_iv_please() that does
7663 the conversion if possible, and silently. */
7664 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
7665 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
7666 /* Need to try really hard to see if it's an integer.
7667 9.22337203685478e+18 is an integer.
7668 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
7669 so $a="9.22337203685478e+18"; $a+0; $a++
7670 needs to be the same as $a="9.22337203685478e+18"; $a++
7677 /* sv_2iv *should* have made this an NV */
7678 if (flags & SVp_NOK) {
7679 (void)SvNOK_only(sv);
7680 SvNV_set(sv, SvNVX(sv) + 1.0);
7683 /* I don't think we can get here. Maybe I should assert this
7684 And if we do get here I suspect that sv_setnv will croak. NWC
7686 #if defined(USE_LONG_DOUBLE)
7687 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",
7688 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7690 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_inc punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
7691 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7694 #endif /* PERL_PRESERVE_IVUV */
7695 sv_setnv(sv,Atof(SvPVX_const(sv)) + 1.0);
7699 while (d >= SvPVX_const(sv)) {
7707 /* MKS: The original code here died if letters weren't consecutive.
7708 * at least it didn't have to worry about non-C locales. The
7709 * new code assumes that ('z'-'a')==('Z'-'A'), letters are
7710 * arranged in order (although not consecutively) and that only
7711 * [A-Za-z] are accepted by isALPHA in the C locale.
7713 if (*d != 'z' && *d != 'Z') {
7714 do { ++*d; } while (!isALPHA(*d));
7717 *(d--) -= 'z' - 'a';
7722 *(d--) -= 'z' - 'a' + 1;
7726 /* oh,oh, the number grew */
7727 SvGROW(sv, SvCUR(sv) + 2);
7728 SvCUR_set(sv, SvCUR(sv) + 1);
7729 for (d = SvPVX(sv) + SvCUR(sv); d > SvPVX_const(sv); d--)
7740 Auto-decrement of the value in the SV, doing string to numeric conversion
7741 if necessary. Handles 'get' magic and operator overloading.
7747 Perl_sv_dec(pTHX_ register SV *const sv)
7757 =for apidoc sv_dec_nomg
7759 Auto-decrement of the value in the SV, doing string to numeric conversion
7760 if necessary. Handles operator overloading. Skips handling 'get' magic.
7766 Perl_sv_dec_nomg(pTHX_ register SV *const sv)
7773 if (SvTHINKFIRST(sv)) {
7775 sv_force_normal_flags(sv, 0);
7776 if (SvREADONLY(sv)) {
7777 if (IN_PERL_RUNTIME)
7778 Perl_croak_no_modify(aTHX);
7782 if (SvAMAGIC(sv) && AMG_CALLun(sv,dec))
7784 i = PTR2IV(SvRV(sv));
7789 /* Unlike sv_inc we don't have to worry about string-never-numbers
7790 and keeping them magic. But we mustn't warn on punting */
7791 flags = SvFLAGS(sv);
7792 if ((flags & SVf_IOK) || ((flags & (SVp_IOK | SVp_NOK)) == SVp_IOK)) {
7793 /* It's publicly an integer, or privately an integer-not-float */
7794 #ifdef PERL_PRESERVE_IVUV
7798 if (SvUVX(sv) == 0) {
7799 (void)SvIOK_only(sv);
7803 (void)SvIOK_only_UV(sv);
7804 SvUV_set(sv, SvUVX(sv) - 1);
7807 if (SvIVX(sv) == IV_MIN) {
7808 sv_setnv(sv, (NV)IV_MIN);
7812 (void)SvIOK_only(sv);
7813 SvIV_set(sv, SvIVX(sv) - 1);
7818 if (flags & SVp_NOK) {
7821 const NV was = SvNVX(sv);
7822 if (NV_OVERFLOWS_INTEGERS_AT &&
7823 was <= -NV_OVERFLOWS_INTEGERS_AT) {
7824 Perl_ck_warner(aTHX_ packWARN(WARN_IMPRECISION),
7825 "Lost precision when decrementing %" NVff " by 1",
7828 (void)SvNOK_only(sv);
7829 SvNV_set(sv, was - 1.0);
7833 if (!(flags & SVp_POK)) {
7834 if ((flags & SVTYPEMASK) < SVt_PVIV)
7835 sv_upgrade(sv, ((flags & SVTYPEMASK) > SVt_IV) ? SVt_PVIV : SVt_IV);
7837 (void)SvIOK_only(sv);
7840 #ifdef PERL_PRESERVE_IVUV
7842 const int numtype = grok_number(SvPVX_const(sv), SvCUR(sv), NULL);
7843 if (numtype && !(numtype & IS_NUMBER_INFINITY)) {
7844 /* Need to try really hard to see if it's an integer.
7845 9.22337203685478e+18 is an integer.
7846 but "9.22337203685478e+18" + 0 is UV=9223372036854779904
7847 so $a="9.22337203685478e+18"; $a+0; $a--
7848 needs to be the same as $a="9.22337203685478e+18"; $a--
7855 /* sv_2iv *should* have made this an NV */
7856 if (flags & SVp_NOK) {
7857 (void)SvNOK_only(sv);
7858 SvNV_set(sv, SvNVX(sv) - 1.0);
7861 /* I don't think we can get here. Maybe I should assert this
7862 And if we do get here I suspect that sv_setnv will croak. NWC
7864 #if defined(USE_LONG_DOUBLE)
7865 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",
7866 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7868 DEBUG_c(PerlIO_printf(Perl_debug_log,"sv_dec punt failed to convert '%s' to IOK or NOKp, UV=0x%"UVxf" NV=%"NVgf"\n",
7869 SvPVX_const(sv), SvIVX(sv), SvNVX(sv)));
7873 #endif /* PERL_PRESERVE_IVUV */
7874 sv_setnv(sv,Atof(SvPVX_const(sv)) - 1.0); /* punt */
7877 /* this define is used to eliminate a chunk of duplicated but shared logic
7878 * it has the suffix __SV_C to signal that it isnt API, and isnt meant to be
7879 * used anywhere but here - yves
7881 #define PUSH_EXTEND_MORTAL__SV_C(AnSv) \
7884 PL_tmps_stack[++PL_tmps_ix] = (AnSv); \
7888 =for apidoc sv_mortalcopy
7890 Creates a new SV which is a copy of the original SV (using C<sv_setsv>).
7891 The new SV is marked as mortal. It will be destroyed "soon", either by an
7892 explicit call to FREETMPS, or by an implicit call at places such as
7893 statement boundaries. See also C<sv_newmortal> and C<sv_2mortal>.
7898 /* Make a string that will exist for the duration of the expression
7899 * evaluation. Actually, it may have to last longer than that, but
7900 * hopefully we won't free it until it has been assigned to a
7901 * permanent location. */
7904 Perl_sv_mortalcopy(pTHX_ SV *const oldstr)
7910 sv_setsv(sv,oldstr);
7911 PUSH_EXTEND_MORTAL__SV_C(sv);
7917 =for apidoc sv_newmortal
7919 Creates a new null SV which is mortal. The reference count of the SV is
7920 set to 1. It will be destroyed "soon", either by an explicit call to
7921 FREETMPS, or by an implicit call at places such as statement boundaries.
7922 See also C<sv_mortalcopy> and C<sv_2mortal>.
7928 Perl_sv_newmortal(pTHX)
7934 SvFLAGS(sv) = SVs_TEMP;
7935 PUSH_EXTEND_MORTAL__SV_C(sv);
7941 =for apidoc newSVpvn_flags
7943 Creates a new SV and copies a string into it. The reference count for the
7944 SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
7945 string. You are responsible for ensuring that the source string is at least
7946 C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
7947 Currently the only flag bits accepted are C<SVf_UTF8> and C<SVs_TEMP>.
7948 If C<SVs_TEMP> is set, then C<sv_2mortal()> is called on the result before
7949 returning. If C<SVf_UTF8> is set, C<s> is considered to be in UTF-8 and the
7950 C<SVf_UTF8> flag will be set on the new SV.
7951 C<newSVpvn_utf8()> is a convenience wrapper for this function, defined as
7953 #define newSVpvn_utf8(s, len, u) \
7954 newSVpvn_flags((s), (len), (u) ? SVf_UTF8 : 0)
7960 Perl_newSVpvn_flags(pTHX_ const char *const s, const STRLEN len, const U32 flags)
7965 /* All the flags we don't support must be zero.
7966 And we're new code so I'm going to assert this from the start. */
7967 assert(!(flags & ~(SVf_UTF8|SVs_TEMP)));
7969 sv_setpvn(sv,s,len);
7971 /* This code used to a sv_2mortal(), however we now unroll the call to sv_2mortal()
7972 * and do what it does outselves here.
7973 * Since we have asserted that flags can only have the SVf_UTF8 and/or SVs_TEMP flags
7974 * set above we can use it to enable the sv flags directly (bypassing SvTEMP_on), which
7975 * in turn means we dont need to mask out the SVf_UTF8 flag below, which means that we
7976 * eleminate quite a few steps than it looks - Yves (explaining patch by gfx)
7979 SvFLAGS(sv) |= flags;
7981 if(flags & SVs_TEMP){
7982 PUSH_EXTEND_MORTAL__SV_C(sv);
7989 =for apidoc sv_2mortal
7991 Marks an existing SV as mortal. The SV will be destroyed "soon", either
7992 by an explicit call to FREETMPS, or by an implicit call at places such as
7993 statement boundaries. SvTEMP() is turned on which means that the SV's
7994 string buffer can be "stolen" if this SV is copied. See also C<sv_newmortal>
7995 and C<sv_mortalcopy>.
8001 Perl_sv_2mortal(pTHX_ register SV *const sv)
8006 if (SvREADONLY(sv) && SvIMMORTAL(sv))
8008 PUSH_EXTEND_MORTAL__SV_C(sv);
8016 Creates a new SV and copies a string into it. The reference count for the
8017 SV is set to 1. If C<len> is zero, Perl will compute the length using
8018 strlen(). For efficiency, consider using C<newSVpvn> instead.
8024 Perl_newSVpv(pTHX_ const char *const s, const STRLEN len)
8030 sv_setpvn(sv, s, len || s == NULL ? len : strlen(s));
8035 =for apidoc newSVpvn
8037 Creates a new SV and copies a string into it. The reference count for the
8038 SV is set to 1. Note that if C<len> is zero, Perl will create a zero length
8039 string. You are responsible for ensuring that the source string is at least
8040 C<len> bytes long. If the C<s> argument is NULL the new SV will be undefined.
8046 Perl_newSVpvn(pTHX_ const char *const s, const STRLEN len)
8052 sv_setpvn(sv,s,len);
8057 =for apidoc newSVhek
8059 Creates a new SV from the hash key structure. It will generate scalars that
8060 point to the shared string table where possible. Returns a new (undefined)
8061 SV if the hek is NULL.
8067 Perl_newSVhek(pTHX_ const HEK *const hek)
8077 if (HEK_LEN(hek) == HEf_SVKEY) {
8078 return newSVsv(*(SV**)HEK_KEY(hek));
8080 const int flags = HEK_FLAGS(hek);
8081 if (flags & HVhek_WASUTF8) {
8083 Andreas would like keys he put in as utf8 to come back as utf8
8085 STRLEN utf8_len = HEK_LEN(hek);
8086 const U8 *as_utf8 = bytes_to_utf8 ((U8*)HEK_KEY(hek), &utf8_len);
8087 SV * const sv = newSVpvn ((const char*)as_utf8, utf8_len);
8090 Safefree (as_utf8); /* bytes_to_utf8() allocates a new string */
8092 } else if (flags & (HVhek_REHASH|HVhek_UNSHARED)) {
8093 /* We don't have a pointer to the hv, so we have to replicate the
8094 flag into every HEK. This hv is using custom a hasing
8095 algorithm. Hence we can't return a shared string scalar, as
8096 that would contain the (wrong) hash value, and might get passed
8097 into an hv routine with a regular hash.
8098 Similarly, a hash that isn't using shared hash keys has to have
8099 the flag in every key so that we know not to try to call
8100 share_hek_kek on it. */
8102 SV * const sv = newSVpvn (HEK_KEY(hek), HEK_LEN(hek));
8107 /* This will be overwhelminly the most common case. */
8109 /* Inline most of newSVpvn_share(), because share_hek_hek() is far
8110 more efficient than sharepvn(). */
8114 sv_upgrade(sv, SVt_PV);
8115 SvPV_set(sv, (char *)HEK_KEY(share_hek_hek(hek)));
8116 SvCUR_set(sv, HEK_LEN(hek));
8129 =for apidoc newSVpvn_share
8131 Creates a new SV with its SvPVX_const pointing to a shared string in the string
8132 table. If the string does not already exist in the table, it is created
8133 first. Turns on READONLY and FAKE. If the C<hash> parameter is non-zero, that
8134 value is used; otherwise the hash is computed. The string's hash can be later
8135 be retrieved from the SV with the C<SvSHARED_HASH()> macro. The idea here is
8136 that as the string table is used for shared hash keys these strings will have
8137 SvPVX_const == HeKEY and hash lookup will avoid string compare.
8143 Perl_newSVpvn_share(pTHX_ const char *src, I32 len, U32 hash)
8147 bool is_utf8 = FALSE;
8148 const char *const orig_src = src;
8151 STRLEN tmplen = -len;
8153 /* See the note in hv.c:hv_fetch() --jhi */
8154 src = (char*)bytes_from_utf8((const U8*)src, &tmplen, &is_utf8);
8158 PERL_HASH(hash, src, len);
8160 /* The logic for this is inlined in S_mro_get_linear_isa_dfs(), so if it
8161 changes here, update it there too. */
8162 sv_upgrade(sv, SVt_PV);
8163 SvPV_set(sv, sharepvn(src, is_utf8?-len:len, hash));
8171 if (src != orig_src)
8177 =for apidoc newSVpv_share
8179 Like C<newSVpvn_share>, but takes a nul-terminated string instead of a
8186 Perl_newSVpv_share(pTHX_ const char *src, U32 hash)
8188 return newSVpvn_share(src, strlen(src), hash);
8191 #if defined(PERL_IMPLICIT_CONTEXT)
8193 /* pTHX_ magic can't cope with varargs, so this is a no-context
8194 * version of the main function, (which may itself be aliased to us).
8195 * Don't access this version directly.
8199 Perl_newSVpvf_nocontext(const char *const pat, ...)
8205 PERL_ARGS_ASSERT_NEWSVPVF_NOCONTEXT;
8207 va_start(args, pat);
8208 sv = vnewSVpvf(pat, &args);
8215 =for apidoc newSVpvf
8217 Creates a new SV and initializes it with the string formatted like
8224 Perl_newSVpvf(pTHX_ const char *const pat, ...)
8229 PERL_ARGS_ASSERT_NEWSVPVF;
8231 va_start(args, pat);
8232 sv = vnewSVpvf(pat, &args);
8237 /* backend for newSVpvf() and newSVpvf_nocontext() */
8240 Perl_vnewSVpvf(pTHX_ const char *const pat, va_list *const args)
8245 PERL_ARGS_ASSERT_VNEWSVPVF;
8248 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
8255 Creates a new SV and copies a floating point value into it.
8256 The reference count for the SV is set to 1.
8262 Perl_newSVnv(pTHX_ const NV n)
8275 Creates a new SV and copies an integer into it. The reference count for the
8282 Perl_newSViv(pTHX_ const IV i)
8295 Creates a new SV and copies an unsigned integer into it.
8296 The reference count for the SV is set to 1.
8302 Perl_newSVuv(pTHX_ const UV u)
8313 =for apidoc newSV_type
8315 Creates a new SV, of the type specified. The reference count for the new SV
8322 Perl_newSV_type(pTHX_ const svtype type)
8327 sv_upgrade(sv, type);
8332 =for apidoc newRV_noinc
8334 Creates an RV wrapper for an SV. The reference count for the original
8335 SV is B<not> incremented.
8341 Perl_newRV_noinc(pTHX_ SV *const tmpRef)
8344 register SV *sv = newSV_type(SVt_IV);
8346 PERL_ARGS_ASSERT_NEWRV_NOINC;
8349 SvRV_set(sv, tmpRef);
8354 /* newRV_inc is the official function name to use now.
8355 * newRV_inc is in fact #defined to newRV in sv.h
8359 Perl_newRV(pTHX_ SV *const sv)
8363 PERL_ARGS_ASSERT_NEWRV;
8365 return newRV_noinc(SvREFCNT_inc_simple_NN(sv));
8371 Creates a new SV which is an exact duplicate of the original SV.
8378 Perl_newSVsv(pTHX_ register SV *const old)
8385 if (SvTYPE(old) == SVTYPEMASK) {
8386 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL), "semi-panic: attempt to dup freed string");
8390 /* SV_GMAGIC is the default for sv_setv()
8391 SV_NOSTEAL prevents TEMP buffers being, well, stolen, and saves games
8392 with SvTEMP_off and SvTEMP_on round a call to sv_setsv. */
8393 sv_setsv_flags(sv, old, SV_GMAGIC | SV_NOSTEAL);
8398 =for apidoc sv_reset
8400 Underlying implementation for the C<reset> Perl function.
8401 Note that the perl-level function is vaguely deprecated.
8407 Perl_sv_reset(pTHX_ register const char *s, HV *const stash)
8410 char todo[PERL_UCHAR_MAX+1];
8412 PERL_ARGS_ASSERT_SV_RESET;
8417 if (!*s) { /* reset ?? searches */
8418 MAGIC * const mg = mg_find((const SV *)stash, PERL_MAGIC_symtab);
8420 const U32 count = mg->mg_len / sizeof(PMOP**);
8421 PMOP **pmp = (PMOP**) mg->mg_ptr;
8422 PMOP *const *const end = pmp + count;
8426 SvREADONLY_off(PL_regex_pad[(*pmp)->op_pmoffset]);
8428 (*pmp)->op_pmflags &= ~PMf_USED;
8436 /* reset variables */
8438 if (!HvARRAY(stash))
8441 Zero(todo, 256, char);
8444 I32 i = (unsigned char)*s;
8448 max = (unsigned char)*s++;
8449 for ( ; i <= max; i++) {
8452 for (i = 0; i <= (I32) HvMAX(stash); i++) {
8454 for (entry = HvARRAY(stash)[i];
8456 entry = HeNEXT(entry))
8461 if (!todo[(U8)*HeKEY(entry)])
8463 gv = MUTABLE_GV(HeVAL(entry));
8466 if (SvTHINKFIRST(sv)) {
8467 if (!SvREADONLY(sv) && SvROK(sv))
8469 /* XXX Is this continue a bug? Why should THINKFIRST
8470 exempt us from resetting arrays and hashes? */
8474 if (SvTYPE(sv) >= SVt_PV) {
8476 if (SvPVX_const(sv) != NULL)
8484 if (GvHV(gv) && !HvNAME_get(GvHV(gv))) {
8486 Perl_die(aTHX_ "Can't reset %%ENV on this system");
8489 # if defined(USE_ENVIRON_ARRAY)
8492 # endif /* USE_ENVIRON_ARRAY */
8503 Using various gambits, try to get an IO from an SV: the IO slot if its a
8504 GV; or the recursive result if we're an RV; or the IO slot of the symbol
8505 named after the PV if we're a string.
8511 Perl_sv_2io(pTHX_ SV *const sv)
8516 PERL_ARGS_ASSERT_SV_2IO;
8518 switch (SvTYPE(sv)) {
8520 io = MUTABLE_IO(sv);
8524 if (isGV_with_GP(sv)) {
8525 gv = MUTABLE_GV(sv);
8528 Perl_croak(aTHX_ "Bad filehandle: %s", GvNAME(gv));
8534 Perl_croak(aTHX_ PL_no_usym, "filehandle");
8536 return sv_2io(SvRV(sv));
8537 gv = gv_fetchsv(sv, 0, SVt_PVIO);
8543 Perl_croak(aTHX_ "Bad filehandle: %"SVf, SVfARG(sv));
8552 Using various gambits, try to get a CV from an SV; in addition, try if
8553 possible to set C<*st> and C<*gvp> to the stash and GV associated with it.
8554 The flags in C<lref> are passed to gv_fetchsv.
8560 Perl_sv_2cv(pTHX_ SV *sv, HV **const st, GV **const gvp, const I32 lref)
8566 PERL_ARGS_ASSERT_SV_2CV;
8573 switch (SvTYPE(sv)) {
8577 return MUTABLE_CV(sv);
8584 if (isGV_with_GP(sv)) {
8585 gv = MUTABLE_GV(sv);
8594 SV * const *sp = &sv; /* Used in tryAMAGICunDEREF macro. */
8596 tryAMAGICunDEREF(to_cv);
8599 if (SvTYPE(sv) == SVt_PVCV) {
8600 cv = MUTABLE_CV(sv);
8605 else if(isGV_with_GP(sv))
8606 gv = MUTABLE_GV(sv);
8608 Perl_croak(aTHX_ "Not a subroutine reference");
8610 else if (isGV_with_GP(sv)) {
8612 gv = MUTABLE_GV(sv);
8615 gv = gv_fetchsv(sv, lref, SVt_PVCV); /* Calls get magic */
8621 /* Some flags to gv_fetchsv mean don't really create the GV */
8622 if (!isGV_with_GP(gv)) {
8628 if (lref && !GvCVu(gv)) {
8632 gv_efullname3(tmpsv, gv, NULL);
8633 /* XXX this is probably not what they think they're getting.
8634 * It has the same effect as "sub name;", i.e. just a forward
8636 newSUB(start_subparse(FALSE, 0),
8637 newSVOP(OP_CONST, 0, tmpsv),
8641 Perl_croak(aTHX_ "Unable to create sub named \"%"SVf"\"",
8642 SVfARG(SvOK(sv) ? sv : &PL_sv_no));
8651 Returns true if the SV has a true value by Perl's rules.
8652 Use the C<SvTRUE> macro instead, which may call C<sv_true()> or may
8653 instead use an in-line version.
8659 Perl_sv_true(pTHX_ register SV *const sv)
8664 register const XPV* const tXpv = (XPV*)SvANY(sv);
8666 (tXpv->xpv_cur > 1 ||
8667 (tXpv->xpv_cur && *sv->sv_u.svu_pv != '0')))
8674 return SvIVX(sv) != 0;
8677 return SvNVX(sv) != 0.0;
8679 return sv_2bool(sv);
8685 =for apidoc sv_pvn_force
8687 Get a sensible string out of the SV somehow.
8688 A private implementation of the C<SvPV_force> macro for compilers which
8689 can't cope with complex macro expressions. Always use the macro instead.
8691 =for apidoc sv_pvn_force_flags
8693 Get a sensible string out of the SV somehow.
8694 If C<flags> has C<SV_GMAGIC> bit set, will C<mg_get> on C<sv> if
8695 appropriate, else not. C<sv_pvn_force> and C<sv_pvn_force_nomg> are
8696 implemented in terms of this function.
8697 You normally want to use the various wrapper macros instead: see
8698 C<SvPV_force> and C<SvPV_force_nomg>
8704 Perl_sv_pvn_force_flags(pTHX_ SV *const sv, STRLEN *const lp, const I32 flags)
8708 PERL_ARGS_ASSERT_SV_PVN_FORCE_FLAGS;
8710 if (SvTHINKFIRST(sv) && !SvROK(sv))
8711 sv_force_normal_flags(sv, 0);
8721 if (SvREADONLY(sv) && !(flags & SV_MUTABLE_RETURN)) {
8722 const char * const ref = sv_reftype(sv,0);
8724 Perl_croak(aTHX_ "Can't coerce readonly %s to string in %s",
8725 ref, OP_DESC(PL_op));
8727 Perl_croak(aTHX_ "Can't coerce readonly %s to string", ref);
8729 if ((SvTYPE(sv) > SVt_PVLV && SvTYPE(sv) != SVt_PVFM)
8730 || isGV_with_GP(sv))
8731 Perl_croak(aTHX_ "Can't coerce %s to string in %s", sv_reftype(sv,0),
8733 s = sv_2pv_flags(sv, &len, flags);
8737 if (s != SvPVX_const(sv)) { /* Almost, but not quite, sv_setpvn() */
8740 SvUPGRADE(sv, SVt_PV); /* Never FALSE */
8741 SvGROW(sv, len + 1);
8742 Move(s,SvPVX(sv),len,char);
8744 SvPVX(sv)[len] = '\0';
8747 SvPOK_on(sv); /* validate pointer */
8749 DEBUG_c(PerlIO_printf(Perl_debug_log, "0x%"UVxf" 2pv(%s)\n",
8750 PTR2UV(sv),SvPVX_const(sv)));
8753 return SvPVX_mutable(sv);
8757 =for apidoc sv_pvbyten_force
8759 The backend for the C<SvPVbytex_force> macro. Always use the macro instead.
8765 Perl_sv_pvbyten_force(pTHX_ SV *const sv, STRLEN *const lp)
8767 PERL_ARGS_ASSERT_SV_PVBYTEN_FORCE;
8769 sv_pvn_force(sv,lp);
8770 sv_utf8_downgrade(sv,0);
8776 =for apidoc sv_pvutf8n_force
8778 The backend for the C<SvPVutf8x_force> macro. Always use the macro instead.
8784 Perl_sv_pvutf8n_force(pTHX_ SV *const sv, STRLEN *const lp)
8786 PERL_ARGS_ASSERT_SV_PVUTF8N_FORCE;
8788 sv_pvn_force(sv,lp);
8789 sv_utf8_upgrade(sv);
8795 =for apidoc sv_reftype
8797 Returns a string describing what the SV is a reference to.
8803 Perl_sv_reftype(pTHX_ const SV *const sv, const int ob)
8805 PERL_ARGS_ASSERT_SV_REFTYPE;
8807 /* The fact that I don't need to downcast to char * everywhere, only in ?:
8808 inside return suggests a const propagation bug in g++. */
8809 if (ob && SvOBJECT(sv)) {
8810 char * const name = HvNAME_get(SvSTASH(sv));
8811 return name ? name : (char *) "__ANON__";
8814 switch (SvTYPE(sv)) {
8829 case SVt_PVLV: return (char *) (SvROK(sv) ? "REF"
8830 /* tied lvalues should appear to be
8831 * scalars for backwards compatitbility */
8832 : (LvTYPE(sv) == 't' || LvTYPE(sv) == 'T')
8833 ? "SCALAR" : "LVALUE");
8834 case SVt_PVAV: return "ARRAY";
8835 case SVt_PVHV: return "HASH";
8836 case SVt_PVCV: return "CODE";
8837 case SVt_PVGV: return (char *) (isGV_with_GP(sv)
8838 ? "GLOB" : "SCALAR");
8839 case SVt_PVFM: return "FORMAT";
8840 case SVt_PVIO: return "IO";
8841 case SVt_BIND: return "BIND";
8842 case SVt_REGEXP: return "REGEXP";
8843 default: return "UNKNOWN";
8849 =for apidoc sv_isobject
8851 Returns a boolean indicating whether the SV is an RV pointing to a blessed
8852 object. If the SV is not an RV, or if the object is not blessed, then this
8859 Perl_sv_isobject(pTHX_ SV *sv)
8875 Returns a boolean indicating whether the SV is blessed into the specified
8876 class. This does not check for subtypes; use C<sv_derived_from> to verify
8877 an inheritance relationship.
8883 Perl_sv_isa(pTHX_ SV *sv, const char *const name)
8887 PERL_ARGS_ASSERT_SV_ISA;
8897 hvname = HvNAME_get(SvSTASH(sv));
8901 return strEQ(hvname, name);
8907 Creates a new SV for the RV, C<rv>, to point to. If C<rv> is not an RV then
8908 it will be upgraded to one. If C<classname> is non-null then the new SV will
8909 be blessed in the specified package. The new SV is returned and its
8910 reference count is 1.
8916 Perl_newSVrv(pTHX_ SV *const rv, const char *const classname)
8921 PERL_ARGS_ASSERT_NEWSVRV;
8925 SV_CHECK_THINKFIRST_COW_DROP(rv);
8926 (void)SvAMAGIC_off(rv);
8928 if (SvTYPE(rv) >= SVt_PVMG) {
8929 const U32 refcnt = SvREFCNT(rv);
8933 SvREFCNT(rv) = refcnt;
8935 sv_upgrade(rv, SVt_IV);
8936 } else if (SvROK(rv)) {
8937 SvREFCNT_dec(SvRV(rv));
8939 prepare_SV_for_RV(rv);
8947 HV* const stash = gv_stashpv(classname, GV_ADD);
8948 (void)sv_bless(rv, stash);
8954 =for apidoc sv_setref_pv
8956 Copies a pointer into a new SV, optionally blessing the SV. The C<rv>
8957 argument will be upgraded to an RV. That RV will be modified to point to
8958 the new SV. If the C<pv> argument is NULL then C<PL_sv_undef> will be placed
8959 into the SV. The C<classname> argument indicates the package for the
8960 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
8961 will have a reference count of 1, and the RV will be returned.
8963 Do not use with other Perl types such as HV, AV, SV, CV, because those
8964 objects will become corrupted by the pointer copy process.
8966 Note that C<sv_setref_pvn> copies the string while this copies the pointer.
8972 Perl_sv_setref_pv(pTHX_ SV *const rv, const char *const classname, void *const pv)
8976 PERL_ARGS_ASSERT_SV_SETREF_PV;
8979 sv_setsv(rv, &PL_sv_undef);
8983 sv_setiv(newSVrv(rv,classname), PTR2IV(pv));
8988 =for apidoc sv_setref_iv
8990 Copies an integer into a new SV, optionally blessing the SV. The C<rv>
8991 argument will be upgraded to an RV. That RV will be modified to point to
8992 the new SV. The C<classname> argument indicates the package for the
8993 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
8994 will have a reference count of 1, and the RV will be returned.
9000 Perl_sv_setref_iv(pTHX_ SV *const rv, const char *const classname, const IV iv)
9002 PERL_ARGS_ASSERT_SV_SETREF_IV;
9004 sv_setiv(newSVrv(rv,classname), iv);
9009 =for apidoc sv_setref_uv
9011 Copies an unsigned integer into a new SV, optionally blessing the SV. The C<rv>
9012 argument will be upgraded to an RV. That RV will be modified to point to
9013 the new SV. The C<classname> argument indicates the package for the
9014 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9015 will have a reference count of 1, and the RV will be returned.
9021 Perl_sv_setref_uv(pTHX_ SV *const rv, const char *const classname, const UV uv)
9023 PERL_ARGS_ASSERT_SV_SETREF_UV;
9025 sv_setuv(newSVrv(rv,classname), uv);
9030 =for apidoc sv_setref_nv
9032 Copies a double into a new SV, optionally blessing the SV. The C<rv>
9033 argument will be upgraded to an RV. That RV will be modified to point to
9034 the new SV. The C<classname> argument indicates the package for the
9035 blessing. Set C<classname> to C<NULL> to avoid the blessing. The new SV
9036 will have a reference count of 1, and the RV will be returned.
9042 Perl_sv_setref_nv(pTHX_ SV *const rv, const char *const classname, const NV nv)
9044 PERL_ARGS_ASSERT_SV_SETREF_NV;
9046 sv_setnv(newSVrv(rv,classname), nv);
9051 =for apidoc sv_setref_pvn
9053 Copies a string into a new SV, optionally blessing the SV. The length of the
9054 string must be specified with C<n>. The C<rv> argument will be upgraded to
9055 an RV. That RV will be modified to point to the new SV. The C<classname>
9056 argument indicates the package for the blessing. Set C<classname> to
9057 C<NULL> to avoid the blessing. The new SV will have a reference count
9058 of 1, and the RV will be returned.
9060 Note that C<sv_setref_pv> copies the pointer while this copies the string.
9066 Perl_sv_setref_pvn(pTHX_ SV *const rv, const char *const classname,
9067 const char *const pv, const STRLEN n)
9069 PERL_ARGS_ASSERT_SV_SETREF_PVN;
9071 sv_setpvn(newSVrv(rv,classname), pv, n);
9076 =for apidoc sv_bless
9078 Blesses an SV into a specified package. The SV must be an RV. The package
9079 must be designated by its stash (see C<gv_stashpv()>). The reference count
9080 of the SV is unaffected.
9086 Perl_sv_bless(pTHX_ SV *const sv, HV *const stash)
9091 PERL_ARGS_ASSERT_SV_BLESS;
9094 Perl_croak(aTHX_ "Can't bless non-reference value");
9096 if (SvFLAGS(tmpRef) & (SVs_OBJECT|SVf_READONLY)) {
9097 if (SvIsCOW(tmpRef))
9098 sv_force_normal_flags(tmpRef, 0);
9099 if (SvREADONLY(tmpRef))
9100 Perl_croak_no_modify(aTHX);
9101 if (SvOBJECT(tmpRef)) {
9102 if (SvTYPE(tmpRef) != SVt_PVIO)
9104 SvREFCNT_dec(SvSTASH(tmpRef));
9107 SvOBJECT_on(tmpRef);
9108 if (SvTYPE(tmpRef) != SVt_PVIO)
9110 SvUPGRADE(tmpRef, SVt_PVMG);
9111 SvSTASH_set(tmpRef, MUTABLE_HV(SvREFCNT_inc_simple(stash)));
9116 (void)SvAMAGIC_off(sv);
9118 if(SvSMAGICAL(tmpRef))
9119 if(mg_find(tmpRef, PERL_MAGIC_ext) || mg_find(tmpRef, PERL_MAGIC_uvar))
9127 /* Downgrades a PVGV to a PVMG. If it’s actually a PVLV, we leave the type
9128 * as it is after unglobbing it.
9132 S_sv_unglob(pTHX_ SV *const sv)
9137 SV * const temp = sv_newmortal();
9139 PERL_ARGS_ASSERT_SV_UNGLOB;
9141 assert(SvTYPE(sv) == SVt_PVGV || SvTYPE(sv) == SVt_PVLV);
9143 gv_efullname3(temp, MUTABLE_GV(sv), "*");
9146 if(GvCVu((const GV *)sv) && (stash = GvSTASH(MUTABLE_GV(sv)))
9147 && HvNAME_get(stash))
9148 mro_method_changed_in(stash);
9149 gp_free(MUTABLE_GV(sv));
9152 sv_del_backref(MUTABLE_SV(GvSTASH(sv)), sv);
9156 if (GvNAME_HEK(sv)) {
9157 unshare_hek(GvNAME_HEK(sv));
9159 isGV_with_GP_off(sv);
9161 if(SvTYPE(sv) == SVt_PVGV) {
9162 /* need to keep SvANY(sv) in the right arena */
9163 xpvmg = new_XPVMG();
9164 StructCopy(SvANY(sv), xpvmg, XPVMG);
9165 del_XPVGV(SvANY(sv));
9168 SvFLAGS(sv) &= ~SVTYPEMASK;
9169 SvFLAGS(sv) |= SVt_PVMG;
9172 /* Intentionally not calling any local SET magic, as this isn't so much a
9173 set operation as merely an internal storage change. */
9174 sv_setsv_flags(sv, temp, 0);
9178 =for apidoc sv_unref_flags
9180 Unsets the RV status of the SV, and decrements the reference count of
9181 whatever was being referenced by the RV. This can almost be thought of
9182 as a reversal of C<newSVrv>. The C<cflags> argument can contain
9183 C<SV_IMMEDIATE_UNREF> to force the reference count to be decremented
9184 (otherwise the decrementing is conditional on the reference count being
9185 different from one or the reference being a readonly SV).
9192 Perl_sv_unref_flags(pTHX_ SV *const ref, const U32 flags)
9194 SV* const target = SvRV(ref);
9196 PERL_ARGS_ASSERT_SV_UNREF_FLAGS;
9198 if (SvWEAKREF(ref)) {
9199 sv_del_backref(target, ref);
9201 SvRV_set(ref, NULL);
9204 SvRV_set(ref, NULL);
9206 /* You can't have a || SvREADONLY(target) here, as $a = $$a, where $a was
9207 assigned to as BEGIN {$a = \"Foo"} will fail. */
9208 if (SvREFCNT(target) != 1 || (flags & SV_IMMEDIATE_UNREF))
9209 SvREFCNT_dec(target);
9210 else /* XXX Hack, but hard to make $a=$a->[1] work otherwise */
9211 sv_2mortal(target); /* Schedule for freeing later */
9215 =for apidoc sv_untaint
9217 Untaint an SV. Use C<SvTAINTED_off> instead.
9222 Perl_sv_untaint(pTHX_ SV *const sv)
9224 PERL_ARGS_ASSERT_SV_UNTAINT;
9226 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
9227 MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
9234 =for apidoc sv_tainted
9236 Test an SV for taintedness. Use C<SvTAINTED> instead.
9241 Perl_sv_tainted(pTHX_ SV *const sv)
9243 PERL_ARGS_ASSERT_SV_TAINTED;
9245 if (SvTYPE(sv) >= SVt_PVMG && SvMAGIC(sv)) {
9246 const MAGIC * const mg = mg_find(sv, PERL_MAGIC_taint);
9247 if (mg && (mg->mg_len & 1) )
9254 =for apidoc sv_setpviv
9256 Copies an integer into the given SV, also updating its string value.
9257 Does not handle 'set' magic. See C<sv_setpviv_mg>.
9263 Perl_sv_setpviv(pTHX_ SV *const sv, const IV iv)
9265 char buf[TYPE_CHARS(UV)];
9267 char * const ptr = uiv_2buf(buf, iv, 0, 0, &ebuf);
9269 PERL_ARGS_ASSERT_SV_SETPVIV;
9271 sv_setpvn(sv, ptr, ebuf - ptr);
9275 =for apidoc sv_setpviv_mg
9277 Like C<sv_setpviv>, but also handles 'set' magic.
9283 Perl_sv_setpviv_mg(pTHX_ SV *const sv, const IV iv)
9285 PERL_ARGS_ASSERT_SV_SETPVIV_MG;
9291 #if defined(PERL_IMPLICIT_CONTEXT)
9293 /* pTHX_ magic can't cope with varargs, so this is a no-context
9294 * version of the main function, (which may itself be aliased to us).
9295 * Don't access this version directly.
9299 Perl_sv_setpvf_nocontext(SV *const sv, const char *const pat, ...)
9304 PERL_ARGS_ASSERT_SV_SETPVF_NOCONTEXT;
9306 va_start(args, pat);
9307 sv_vsetpvf(sv, pat, &args);
9311 /* pTHX_ magic can't cope with varargs, so this is a no-context
9312 * version of the main function, (which may itself be aliased to us).
9313 * Don't access this version directly.
9317 Perl_sv_setpvf_mg_nocontext(SV *const sv, const char *const pat, ...)
9322 PERL_ARGS_ASSERT_SV_SETPVF_MG_NOCONTEXT;
9324 va_start(args, pat);
9325 sv_vsetpvf_mg(sv, pat, &args);
9331 =for apidoc sv_setpvf
9333 Works like C<sv_catpvf> but copies the text into the SV instead of
9334 appending it. Does not handle 'set' magic. See C<sv_setpvf_mg>.
9340 Perl_sv_setpvf(pTHX_ SV *const sv, const char *const pat, ...)
9344 PERL_ARGS_ASSERT_SV_SETPVF;
9346 va_start(args, pat);
9347 sv_vsetpvf(sv, pat, &args);
9352 =for apidoc sv_vsetpvf
9354 Works like C<sv_vcatpvf> but copies the text into the SV instead of
9355 appending it. Does not handle 'set' magic. See C<sv_vsetpvf_mg>.
9357 Usually used via its frontend C<sv_setpvf>.
9363 Perl_sv_vsetpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9365 PERL_ARGS_ASSERT_SV_VSETPVF;
9367 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9371 =for apidoc sv_setpvf_mg
9373 Like C<sv_setpvf>, but also handles 'set' magic.
9379 Perl_sv_setpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
9383 PERL_ARGS_ASSERT_SV_SETPVF_MG;
9385 va_start(args, pat);
9386 sv_vsetpvf_mg(sv, pat, &args);
9391 =for apidoc sv_vsetpvf_mg
9393 Like C<sv_vsetpvf>, but also handles 'set' magic.
9395 Usually used via its frontend C<sv_setpvf_mg>.
9401 Perl_sv_vsetpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9403 PERL_ARGS_ASSERT_SV_VSETPVF_MG;
9405 sv_vsetpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9409 #if defined(PERL_IMPLICIT_CONTEXT)
9411 /* pTHX_ magic can't cope with varargs, so this is a no-context
9412 * version of the main function, (which may itself be aliased to us).
9413 * Don't access this version directly.
9417 Perl_sv_catpvf_nocontext(SV *const sv, const char *const pat, ...)
9422 PERL_ARGS_ASSERT_SV_CATPVF_NOCONTEXT;
9424 va_start(args, pat);
9425 sv_vcatpvf(sv, pat, &args);
9429 /* pTHX_ magic can't cope with varargs, so this is a no-context
9430 * version of the main function, (which may itself be aliased to us).
9431 * Don't access this version directly.
9435 Perl_sv_catpvf_mg_nocontext(SV *const sv, const char *const pat, ...)
9440 PERL_ARGS_ASSERT_SV_CATPVF_MG_NOCONTEXT;
9442 va_start(args, pat);
9443 sv_vcatpvf_mg(sv, pat, &args);
9449 =for apidoc sv_catpvf
9451 Processes its arguments like C<sprintf> and appends the formatted
9452 output to an SV. If the appended data contains "wide" characters
9453 (including, but not limited to, SVs with a UTF-8 PV formatted with %s,
9454 and characters >255 formatted with %c), the original SV might get
9455 upgraded to UTF-8. Handles 'get' magic, but not 'set' magic. See
9456 C<sv_catpvf_mg>. If the original SV was UTF-8, the pattern should be
9457 valid UTF-8; if the original SV was bytes, the pattern should be too.
9462 Perl_sv_catpvf(pTHX_ SV *const sv, const char *const pat, ...)
9466 PERL_ARGS_ASSERT_SV_CATPVF;
9468 va_start(args, pat);
9469 sv_vcatpvf(sv, pat, &args);
9474 =for apidoc sv_vcatpvf
9476 Processes its arguments like C<vsprintf> and appends the formatted output
9477 to an SV. Does not handle 'set' magic. See C<sv_vcatpvf_mg>.
9479 Usually used via its frontend C<sv_catpvf>.
9485 Perl_sv_vcatpvf(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9487 PERL_ARGS_ASSERT_SV_VCATPVF;
9489 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9493 =for apidoc sv_catpvf_mg
9495 Like C<sv_catpvf>, but also handles 'set' magic.
9501 Perl_sv_catpvf_mg(pTHX_ SV *const sv, const char *const pat, ...)
9505 PERL_ARGS_ASSERT_SV_CATPVF_MG;
9507 va_start(args, pat);
9508 sv_vcatpvf_mg(sv, pat, &args);
9513 =for apidoc sv_vcatpvf_mg
9515 Like C<sv_vcatpvf>, but also handles 'set' magic.
9517 Usually used via its frontend C<sv_catpvf_mg>.
9523 Perl_sv_vcatpvf_mg(pTHX_ SV *const sv, const char *const pat, va_list *const args)
9525 PERL_ARGS_ASSERT_SV_VCATPVF_MG;
9527 sv_vcatpvfn(sv, pat, strlen(pat), args, NULL, 0, NULL);
9532 =for apidoc sv_vsetpvfn
9534 Works like C<sv_vcatpvfn> but copies the text into the SV instead of
9537 Usually used via one of its frontends C<sv_vsetpvf> and C<sv_vsetpvf_mg>.
9543 Perl_sv_vsetpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
9544 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
9546 PERL_ARGS_ASSERT_SV_VSETPVFN;
9549 sv_vcatpvfn(sv, pat, patlen, args, svargs, svmax, maybe_tainted);
9554 * Warn of missing argument to sprintf, and then return a defined value
9555 * to avoid inappropriate "use of uninit" warnings [perl #71000].
9557 #define WARN_MISSING WARN_UNINITIALIZED /* Not sure we want a new category */
9559 S_vcatpvfn_missing_argument(pTHX) {
9560 if (ckWARN(WARN_MISSING)) {
9561 Perl_warner(aTHX_ packWARN(WARN_MISSING), "Missing argument in %s",
9562 PL_op ? OP_DESC(PL_op) : "sv_vcatpvfn()");
9569 S_expect_number(pTHX_ char **const pattern)
9574 PERL_ARGS_ASSERT_EXPECT_NUMBER;
9576 switch (**pattern) {
9577 case '1': case '2': case '3':
9578 case '4': case '5': case '6':
9579 case '7': case '8': case '9':
9580 var = *(*pattern)++ - '0';
9581 while (isDIGIT(**pattern)) {
9582 const I32 tmp = var * 10 + (*(*pattern)++ - '0');
9584 Perl_croak(aTHX_ "Integer overflow in format string for %s", (PL_op ? OP_DESC(PL_op) : "sv_vcatpvfn"));
9592 S_F0convert(NV nv, char *const endbuf, STRLEN *const len)
9594 const int neg = nv < 0;
9597 PERL_ARGS_ASSERT_F0CONVERT;
9605 if (uv & 1 && uv == nv)
9606 uv--; /* Round to even */
9608 const unsigned dig = uv % 10;
9621 =for apidoc sv_vcatpvfn
9623 Processes its arguments like C<vsprintf> and appends the formatted output
9624 to an SV. Uses an array of SVs if the C style variable argument list is
9625 missing (NULL). When running with taint checks enabled, indicates via
9626 C<maybe_tainted> if results are untrustworthy (often due to the use of
9629 Usually used via one of its frontends C<sv_vcatpvf> and C<sv_vcatpvf_mg>.
9635 #define VECTORIZE_ARGS vecsv = va_arg(*args, SV*);\
9636 vecstr = (U8*)SvPV_const(vecsv,veclen);\
9637 vec_utf8 = DO_UTF8(vecsv);
9639 /* XXX maybe_tainted is never assigned to, so the doc above is lying. */
9642 Perl_sv_vcatpvfn(pTHX_ SV *const sv, const char *const pat, const STRLEN patlen,
9643 va_list *const args, SV **const svargs, const I32 svmax, bool *const maybe_tainted)
9651 static const char nullstr[] = "(null)";
9653 bool has_utf8 = DO_UTF8(sv); /* has the result utf8? */
9654 const bool pat_utf8 = has_utf8; /* the pattern is in utf8? */
9656 /* Times 4: a decimal digit takes more than 3 binary digits.
9657 * NV_DIG: mantissa takes than many decimal digits.
9658 * Plus 32: Playing safe. */
9659 char ebuf[IV_DIG * 4 + NV_DIG + 32];
9660 /* large enough for "%#.#f" --chip */
9661 /* what about long double NVs? --jhi */
9663 PERL_ARGS_ASSERT_SV_VCATPVFN;
9664 PERL_UNUSED_ARG(maybe_tainted);
9666 /* no matter what, this is a string now */
9667 (void)SvPV_force(sv, origlen);
9669 /* special-case "", "%s", and "%-p" (SVf - see below) */
9672 if (patlen == 2 && pat[0] == '%' && pat[1] == 's') {
9674 const char * const s = va_arg(*args, char*);
9675 sv_catpv(sv, s ? s : nullstr);
9677 else if (svix < svmax) {
9678 sv_catsv(sv, *svargs);
9681 S_vcatpvfn_missing_argument(aTHX);
9684 if (args && patlen == 3 && pat[0] == '%' &&
9685 pat[1] == '-' && pat[2] == 'p') {
9686 argsv = MUTABLE_SV(va_arg(*args, void*));
9687 sv_catsv(sv, argsv);
9691 #ifndef USE_LONG_DOUBLE
9692 /* special-case "%.<number>[gf]" */
9693 if ( !args && patlen <= 5 && pat[0] == '%' && pat[1] == '.'
9694 && (pat[patlen-1] == 'g' || pat[patlen-1] == 'f') ) {
9695 unsigned digits = 0;
9699 while (*pp >= '0' && *pp <= '9')
9700 digits = 10 * digits + (*pp++ - '0');
9701 if (pp - pat == (int)patlen - 1 && svix < svmax) {
9702 const NV nv = SvNV(*svargs);
9704 /* Add check for digits != 0 because it seems that some
9705 gconverts are buggy in this case, and we don't yet have
9706 a Configure test for this. */
9707 if (digits && digits < sizeof(ebuf) - NV_DIG - 10) {
9708 /* 0, point, slack */
9709 Gconvert(nv, (int)digits, 0, ebuf);
9711 if (*ebuf) /* May return an empty string for digits==0 */
9714 } else if (!digits) {
9717 if ((p = F0convert(nv, ebuf + sizeof ebuf, &l))) {
9718 sv_catpvn(sv, p, l);
9724 #endif /* !USE_LONG_DOUBLE */
9726 if (!args && svix < svmax && DO_UTF8(*svargs))
9729 patend = (char*)pat + patlen;
9730 for (p = (char*)pat; p < patend; p = q) {
9733 bool vectorize = FALSE;
9734 bool vectorarg = FALSE;
9735 bool vec_utf8 = FALSE;
9741 bool has_precis = FALSE;
9743 const I32 osvix = svix;
9744 bool is_utf8 = FALSE; /* is this item utf8? */
9745 #ifdef HAS_LDBL_SPRINTF_BUG
9746 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
9747 with sfio - Allen <allens@cpan.org> */
9748 bool fix_ldbl_sprintf_bug = FALSE;
9752 U8 utf8buf[UTF8_MAXBYTES+1];
9753 STRLEN esignlen = 0;
9755 const char *eptr = NULL;
9756 const char *fmtstart;
9759 const U8 *vecstr = NULL;
9766 /* we need a long double target in case HAS_LONG_DOUBLE but
9769 #if defined(HAS_LONG_DOUBLE) && LONG_DOUBLESIZE > DOUBLESIZE
9777 const char *dotstr = ".";
9778 STRLEN dotstrlen = 1;
9779 I32 efix = 0; /* explicit format parameter index */
9780 I32 ewix = 0; /* explicit width index */
9781 I32 epix = 0; /* explicit precision index */
9782 I32 evix = 0; /* explicit vector index */
9783 bool asterisk = FALSE;
9785 /* echo everything up to the next format specification */
9786 for (q = p; q < patend && *q != '%'; ++q) ;
9788 if (has_utf8 && !pat_utf8)
9789 sv_catpvn_utf8_upgrade(sv, p, q - p, nsv);
9791 sv_catpvn(sv, p, q - p);
9800 We allow format specification elements in this order:
9801 \d+\$ explicit format parameter index
9803 v|\*(\d+\$)?v vector with optional (optionally specified) arg
9804 0 flag (as above): repeated to allow "v02"
9805 \d+|\*(\d+\$)? width using optional (optionally specified) arg
9806 \.(\d*|\*(\d+\$)?) precision using optional (optionally specified) arg
9808 [%bcdefginopsuxDFOUX] format (mandatory)
9813 As of perl5.9.3, printf format checking is on by default.
9814 Internally, perl uses %p formats to provide an escape to
9815 some extended formatting. This block deals with those
9816 extensions: if it does not match, (char*)q is reset and
9817 the normal format processing code is used.
9819 Currently defined extensions are:
9820 %p include pointer address (standard)
9821 %-p (SVf) include an SV (previously %_)
9822 %-<num>p include an SV with precision <num>
9823 %<num>p reserved for future extensions
9825 Robin Barker 2005-07-14
9827 %1p (VDf) removed. RMB 2007-10-19
9834 n = expect_number(&q);
9841 argsv = MUTABLE_SV(va_arg(*args, void*));
9842 eptr = SvPV_const(argsv, elen);
9848 Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL),
9849 "internal %%<num>p might conflict with future printf extensions");
9855 if ( (width = expect_number(&q)) ) {
9870 if (plus == '+' && *q == ' ') /* '+' over ' ' */
9899 if ( (ewix = expect_number(&q)) )
9908 if ((vectorarg = asterisk)) {
9921 width = expect_number(&q);
9927 vecsv = va_arg(*args, SV*);
9929 vecsv = (evix > 0 && evix <= svmax)
9930 ? svargs[evix-1] : S_vcatpvfn_missing_argument(aTHX);
9932 vecsv = svix < svmax
9933 ? svargs[svix++] : S_vcatpvfn_missing_argument(aTHX);
9935 dotstr = SvPV_const(vecsv, dotstrlen);
9936 /* Keep the DO_UTF8 test *after* the SvPV call, else things go
9937 bad with tied or overloaded values that return UTF8. */
9940 else if (has_utf8) {
9941 vecsv = sv_mortalcopy(vecsv);
9942 sv_utf8_upgrade(vecsv);
9943 dotstr = SvPV_const(vecsv, dotstrlen);
9950 else if (efix ? (efix > 0 && efix <= svmax) : svix < svmax) {
9951 vecsv = svargs[efix ? efix-1 : svix++];
9952 vecstr = (U8*)SvPV_const(vecsv,veclen);
9953 vec_utf8 = DO_UTF8(vecsv);
9955 /* if this is a version object, we need to convert
9956 * back into v-string notation and then let the
9957 * vectorize happen normally
9959 if (sv_derived_from(vecsv, "version")) {
9960 char *version = savesvpv(vecsv);
9961 if ( hv_exists(MUTABLE_HV(SvRV(vecsv)), "alpha", 5 ) ) {
9962 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
9963 "vector argument not supported with alpha versions");
9966 vecsv = sv_newmortal();
9967 scan_vstring(version, version + veclen, vecsv);
9968 vecstr = (U8*)SvPV_const(vecsv, veclen);
9969 vec_utf8 = DO_UTF8(vecsv);
9981 i = va_arg(*args, int);
9983 i = (ewix ? ewix <= svmax : svix < svmax) ?
9984 SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
9986 width = (i < 0) ? -i : i;
9996 if ( ((epix = expect_number(&q))) && (*q++ != '$') )
9998 /* XXX: todo, support specified precision parameter */
10002 i = va_arg(*args, int);
10004 i = (ewix ? ewix <= svmax : svix < svmax)
10005 ? SvIVx(svargs[ewix ? ewix-1 : svix++]) : 0;
10007 has_precis = !(i < 0);
10011 while (isDIGIT(*q))
10012 precis = precis * 10 + (*q++ - '0');
10021 case 'I': /* Ix, I32x, and I64x */
10023 if (q[1] == '6' && q[2] == '4') {
10029 if (q[1] == '3' && q[2] == '2') {
10039 #if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
10050 #if defined(HAS_QUAD) || defined(HAS_LONG_DOUBLE)
10051 if (*(q + 1) == 'l') { /* lld, llf */
10077 if (!vectorize && !args) {
10079 const I32 i = efix-1;
10080 argsv = (i >= 0 && i < svmax)
10081 ? svargs[i] : S_vcatpvfn_missing_argument(aTHX);
10083 argsv = (svix >= 0 && svix < svmax)
10084 ? svargs[svix++] : S_vcatpvfn_missing_argument(aTHX);
10088 switch (c = *q++) {
10095 uv = (args) ? va_arg(*args, int) : SvIV(argsv);
10097 (!UNI_IS_INVARIANT(uv) && SvUTF8(sv)))
10099 eptr = (char*)utf8buf;
10100 elen = uvchr_to_utf8((U8*)eptr, uv) - utf8buf;
10114 eptr = va_arg(*args, char*);
10116 elen = strlen(eptr);
10118 eptr = (char *)nullstr;
10119 elen = sizeof nullstr - 1;
10123 eptr = SvPV_const(argsv, elen);
10124 if (DO_UTF8(argsv)) {
10125 STRLEN old_precis = precis;
10126 if (has_precis && precis < elen) {
10127 STRLEN ulen = sv_len_utf8(argsv);
10128 I32 p = precis > ulen ? ulen : precis;
10129 sv_pos_u2b(argsv, &p, 0); /* sticks at end */
10132 if (width) { /* fudge width (can't fudge elen) */
10133 if (has_precis && precis < elen)
10134 width += precis - old_precis;
10136 width += elen - sv_len_utf8(argsv);
10143 if (has_precis && precis < elen)
10150 if (alt || vectorize)
10152 uv = PTR2UV(args ? va_arg(*args, void*) : argsv);
10173 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
10182 esignbuf[esignlen++] = plus;
10186 case 'h': iv = (short)va_arg(*args, int); break;
10187 case 'l': iv = va_arg(*args, long); break;
10188 case 'V': iv = va_arg(*args, IV); break;
10189 default: iv = va_arg(*args, int); break;
10192 iv = va_arg(*args, Quad_t); break;
10199 IV tiv = SvIV(argsv); /* work around GCC bug #13488 */
10201 case 'h': iv = (short)tiv; break;
10202 case 'l': iv = (long)tiv; break;
10204 default: iv = tiv; break;
10207 iv = (Quad_t)tiv; break;
10213 if ( !vectorize ) /* we already set uv above */
10218 esignbuf[esignlen++] = plus;
10222 esignbuf[esignlen++] = '-';
10266 uv = utf8n_to_uvchr(vecstr, veclen, &ulen,
10277 case 'h': uv = (unsigned short)va_arg(*args, unsigned); break;
10278 case 'l': uv = va_arg(*args, unsigned long); break;
10279 case 'V': uv = va_arg(*args, UV); break;
10280 default: uv = va_arg(*args, unsigned); break;
10283 uv = va_arg(*args, Uquad_t); break;
10290 UV tuv = SvUV(argsv); /* work around GCC bug #13488 */
10292 case 'h': uv = (unsigned short)tuv; break;
10293 case 'l': uv = (unsigned long)tuv; break;
10295 default: uv = tuv; break;
10298 uv = (Uquad_t)tuv; break;
10307 char *ptr = ebuf + sizeof ebuf;
10308 bool tempalt = uv ? alt : FALSE; /* Vectors can't change alt */
10314 p = (char *)((c == 'X') ? PL_hexdigit + 16 : PL_hexdigit);
10318 } while (uv >>= 4);
10320 esignbuf[esignlen++] = '0';
10321 esignbuf[esignlen++] = c; /* 'x' or 'X' */
10327 *--ptr = '0' + dig;
10328 } while (uv >>= 3);
10329 if (alt && *ptr != '0')
10335 *--ptr = '0' + dig;
10336 } while (uv >>= 1);
10338 esignbuf[esignlen++] = '0';
10339 esignbuf[esignlen++] = c;
10342 default: /* it had better be ten or less */
10345 *--ptr = '0' + dig;
10346 } while (uv /= base);
10349 elen = (ebuf + sizeof ebuf) - ptr;
10353 zeros = precis - elen;
10354 else if (precis == 0 && elen == 1 && *eptr == '0'
10355 && !(base == 8 && alt)) /* "%#.0o" prints "0" */
10358 /* a precision nullifies the 0 flag. */
10365 /* FLOATING POINT */
10368 c = 'f'; /* maybe %F isn't supported here */
10370 case 'e': case 'E':
10372 case 'g': case 'G':
10376 /* This is evil, but floating point is even more evil */
10378 /* for SV-style calling, we can only get NV
10379 for C-style calling, we assume %f is double;
10380 for simplicity we allow any of %Lf, %llf, %qf for long double
10384 #if defined(USE_LONG_DOUBLE)
10388 /* [perl #20339] - we should accept and ignore %lf rather than die */
10392 #if defined(USE_LONG_DOUBLE)
10393 intsize = args ? 0 : 'q';
10397 #if defined(HAS_LONG_DOUBLE)
10406 /* now we need (long double) if intsize == 'q', else (double) */
10408 #if LONG_DOUBLESIZE > DOUBLESIZE
10410 va_arg(*args, long double) :
10411 va_arg(*args, double)
10413 va_arg(*args, double)
10418 /* nv * 0 will be NaN for NaN, +Inf and -Inf, and 0 for anything
10419 else. frexp() has some unspecified behaviour for those three */
10420 if (c != 'e' && c != 'E' && (nv * 0) == 0) {
10422 /* FIXME: if HAS_LONG_DOUBLE but not USE_LONG_DOUBLE this
10423 will cast our (long double) to (double) */
10424 (void)Perl_frexp(nv, &i);
10425 if (i == PERL_INT_MIN)
10426 Perl_die(aTHX_ "panic: frexp");
10428 need = BIT_DIGITS(i);
10430 need += has_precis ? precis : 6; /* known default */
10435 #ifdef HAS_LDBL_SPRINTF_BUG
10436 /* This is to try to fix a bug with irix/nonstop-ux/powerux and
10437 with sfio - Allen <allens@cpan.org> */
10440 # define MY_DBL_MAX DBL_MAX
10441 # else /* XXX guessing! HUGE_VAL may be defined as infinity, so not using */
10442 # if DOUBLESIZE >= 8
10443 # define MY_DBL_MAX 1.7976931348623157E+308L
10445 # define MY_DBL_MAX 3.40282347E+38L
10449 # ifdef HAS_LDBL_SPRINTF_BUG_LESS1 /* only between -1L & 1L - Allen */
10450 # define MY_DBL_MAX_BUG 1L
10452 # define MY_DBL_MAX_BUG MY_DBL_MAX
10456 # define MY_DBL_MIN DBL_MIN
10457 # else /* XXX guessing! -Allen */
10458 # if DOUBLESIZE >= 8
10459 # define MY_DBL_MIN 2.2250738585072014E-308L
10461 # define MY_DBL_MIN 1.17549435E-38L
10465 if ((intsize == 'q') && (c == 'f') &&
10466 ((nv < MY_DBL_MAX_BUG) && (nv > -MY_DBL_MAX_BUG)) &&
10467 (need < DBL_DIG)) {
10468 /* it's going to be short enough that
10469 * long double precision is not needed */
10471 if ((nv <= 0L) && (nv >= -0L))
10472 fix_ldbl_sprintf_bug = TRUE; /* 0 is 0 - easiest */
10474 /* would use Perl_fp_class as a double-check but not
10475 * functional on IRIX - see perl.h comments */
10477 if ((nv >= MY_DBL_MIN) || (nv <= -MY_DBL_MIN)) {
10478 /* It's within the range that a double can represent */
10479 #if defined(DBL_MAX) && !defined(DBL_MIN)
10480 if ((nv >= ((long double)1/DBL_MAX)) ||
10481 (nv <= (-(long double)1/DBL_MAX)))
10483 fix_ldbl_sprintf_bug = TRUE;
10486 if (fix_ldbl_sprintf_bug == TRUE) {
10496 # undef MY_DBL_MAX_BUG
10499 #endif /* HAS_LDBL_SPRINTF_BUG */
10501 need += 20; /* fudge factor */
10502 if (PL_efloatsize < need) {
10503 Safefree(PL_efloatbuf);
10504 PL_efloatsize = need + 20; /* more fudge */
10505 Newx(PL_efloatbuf, PL_efloatsize, char);
10506 PL_efloatbuf[0] = '\0';
10509 if ( !(width || left || plus || alt) && fill != '0'
10510 && has_precis && intsize != 'q' ) { /* Shortcuts */
10511 /* See earlier comment about buggy Gconvert when digits,
10513 if ( c == 'g' && precis) {
10514 Gconvert((NV)nv, (int)precis, 0, PL_efloatbuf);
10515 /* May return an empty string for digits==0 */
10516 if (*PL_efloatbuf) {
10517 elen = strlen(PL_efloatbuf);
10518 goto float_converted;
10520 } else if ( c == 'f' && !precis) {
10521 if ((eptr = F0convert(nv, ebuf + sizeof ebuf, &elen)))
10526 char *ptr = ebuf + sizeof ebuf;
10529 /* FIXME: what to do if HAS_LONG_DOUBLE but not PERL_PRIfldbl? */
10530 #if defined(HAS_LONG_DOUBLE) && defined(PERL_PRIfldbl)
10531 if (intsize == 'q') {
10532 /* Copy the one or more characters in a long double
10533 * format before the 'base' ([efgEFG]) character to
10534 * the format string. */
10535 static char const prifldbl[] = PERL_PRIfldbl;
10536 char const *p = prifldbl + sizeof(prifldbl) - 3;
10537 while (p >= prifldbl) { *--ptr = *p--; }
10542 do { *--ptr = '0' + (base % 10); } while (base /= 10);
10547 do { *--ptr = '0' + (base % 10); } while (base /= 10);
10559 /* No taint. Otherwise we are in the strange situation
10560 * where printf() taints but print($float) doesn't.
10562 #if defined(HAS_LONG_DOUBLE)
10563 elen = ((intsize == 'q')
10564 ? my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, nv)
10565 : my_snprintf(PL_efloatbuf, PL_efloatsize, ptr, (double)nv));
10567 elen = my_sprintf(PL_efloatbuf, ptr, nv);
10571 eptr = PL_efloatbuf;
10579 i = SvCUR(sv) - origlen;
10582 case 'h': *(va_arg(*args, short*)) = i; break;
10583 default: *(va_arg(*args, int*)) = i; break;
10584 case 'l': *(va_arg(*args, long*)) = i; break;
10585 case 'V': *(va_arg(*args, IV*)) = i; break;
10588 *(va_arg(*args, Quad_t*)) = i; break;
10595 sv_setuv_mg(argsv, (UV)i);
10596 continue; /* not "break" */
10603 && (PL_op->op_type == OP_PRTF || PL_op->op_type == OP_SPRINTF)
10604 && ckWARN(WARN_PRINTF))
10606 SV * const msg = sv_newmortal();
10607 Perl_sv_setpvf(aTHX_ msg, "Invalid conversion in %sprintf: ",
10608 (PL_op->op_type == OP_PRTF) ? "" : "s");
10609 if (fmtstart < patend) {
10610 const char * const fmtend = q < patend ? q : patend;
10612 sv_catpvs(msg, "\"%");
10613 for (f = fmtstart; f < fmtend; f++) {
10615 sv_catpvn(msg, f, 1);
10617 Perl_sv_catpvf(aTHX_ msg,
10618 "\\%03"UVof, (UV)*f & 0xFF);
10621 sv_catpvs(msg, "\"");
10623 sv_catpvs(msg, "end of string");
10625 Perl_warner(aTHX_ packWARN(WARN_PRINTF), "%"SVf, SVfARG(msg)); /* yes, this is reentrant */
10628 /* output mangled stuff ... */
10634 /* ... right here, because formatting flags should not apply */
10635 SvGROW(sv, SvCUR(sv) + elen + 1);
10637 Copy(eptr, p, elen, char);
10640 SvCUR_set(sv, p - SvPVX_const(sv));
10642 continue; /* not "break" */
10645 if (is_utf8 != has_utf8) {
10648 sv_utf8_upgrade(sv);
10651 const STRLEN old_elen = elen;
10652 SV * const nsv = newSVpvn_flags(eptr, elen, SVs_TEMP);
10653 sv_utf8_upgrade(nsv);
10654 eptr = SvPVX_const(nsv);
10657 if (width) { /* fudge width (can't fudge elen) */
10658 width += elen - old_elen;
10664 have = esignlen + zeros + elen;
10666 Perl_croak_nocontext("%s", PL_memory_wrap);
10668 need = (have > width ? have : width);
10671 if (need >= (((STRLEN)~0) - SvCUR(sv) - dotstrlen - 1))
10672 Perl_croak_nocontext("%s", PL_memory_wrap);
10673 SvGROW(sv, SvCUR(sv) + need + dotstrlen + 1);
10675 if (esignlen && fill == '0') {
10677 for (i = 0; i < (int)esignlen; i++)
10678 *p++ = esignbuf[i];
10680 if (gap && !left) {
10681 memset(p, fill, gap);
10684 if (esignlen && fill != '0') {
10686 for (i = 0; i < (int)esignlen; i++)
10687 *p++ = esignbuf[i];
10691 for (i = zeros; i; i--)
10695 Copy(eptr, p, elen, char);
10699 memset(p, ' ', gap);
10704 Copy(dotstr, p, dotstrlen, char);
10708 vectorize = FALSE; /* done iterating over vecstr */
10715 SvCUR_set(sv, p - SvPVX_const(sv));
10724 /* =========================================================================
10726 =head1 Cloning an interpreter
10728 All the macros and functions in this section are for the private use of
10729 the main function, perl_clone().
10731 The foo_dup() functions make an exact copy of an existing foo thingy.
10732 During the course of a cloning, a hash table is used to map old addresses
10733 to new addresses. The table is created and manipulated with the
10734 ptr_table_* functions.
10738 * =========================================================================*/
10741 #if defined(USE_ITHREADS)
10743 /* XXX Remove this so it doesn't have to go thru the macro and return for nothing */
10744 #ifndef GpREFCNT_inc
10745 # define GpREFCNT_inc(gp) ((gp) ? (++(gp)->gp_refcnt, (gp)) : (GP*)NULL)
10749 /* Certain cases in Perl_ss_dup have been merged, by relying on the fact
10750 that currently av_dup, gv_dup and hv_dup are the same as sv_dup.
10751 If this changes, please unmerge ss_dup.
10752 Likewise, sv_dup_inc_multiple() relies on this fact. */
10753 #define sv_dup_inc_NN(s,t) SvREFCNT_inc_NN(sv_dup_inc(s,t))
10754 #define av_dup(s,t) MUTABLE_AV(sv_dup((const SV *)s,t))
10755 #define av_dup_inc(s,t) MUTABLE_AV(sv_dup_inc((const SV *)s,t))
10756 #define hv_dup(s,t) MUTABLE_HV(sv_dup((const SV *)s,t))
10757 #define hv_dup_inc(s,t) MUTABLE_HV(sv_dup_inc((const SV *)s,t))
10758 #define cv_dup(s,t) MUTABLE_CV(sv_dup((const SV *)s,t))
10759 #define cv_dup_inc(s,t) MUTABLE_CV(sv_dup_inc((const SV *)s,t))
10760 #define io_dup(s,t) MUTABLE_IO(sv_dup((const SV *)s,t))
10761 #define io_dup_inc(s,t) MUTABLE_IO(sv_dup_inc((const SV *)s,t))
10762 #define gv_dup(s,t) MUTABLE_GV(sv_dup((const SV *)s,t))
10763 #define gv_dup_inc(s,t) MUTABLE_GV(sv_dup_inc((const SV *)s,t))
10764 #define SAVEPV(p) ((p) ? savepv(p) : NULL)
10765 #define SAVEPVN(p,n) ((p) ? savepvn(p,n) : NULL)
10767 /* clone a parser */
10770 Perl_parser_dup(pTHX_ const yy_parser *const proto, CLONE_PARAMS *const param)
10774 PERL_ARGS_ASSERT_PARSER_DUP;
10779 /* look for it in the table first */
10780 parser = (yy_parser *)ptr_table_fetch(PL_ptr_table, proto);
10784 /* create anew and remember what it is */
10785 Newxz(parser, 1, yy_parser);
10786 ptr_table_store(PL_ptr_table, proto, parser);
10788 /* XXX these not yet duped */
10789 parser->old_parser = NULL;
10790 parser->stack = NULL;
10792 parser->stack_size = 0;
10793 /* XXX parser->stack->state = 0; */
10795 /* XXX eventually, just Copy() most of the parser struct ? */
10797 parser->lex_brackets = proto->lex_brackets;
10798 parser->lex_casemods = proto->lex_casemods;
10799 parser->lex_brackstack = savepvn(proto->lex_brackstack,
10800 (proto->lex_brackets < 120 ? 120 : proto->lex_brackets));
10801 parser->lex_casestack = savepvn(proto->lex_casestack,
10802 (proto->lex_casemods < 12 ? 12 : proto->lex_casemods));
10803 parser->lex_defer = proto->lex_defer;
10804 parser->lex_dojoin = proto->lex_dojoin;
10805 parser->lex_expect = proto->lex_expect;
10806 parser->lex_formbrack = proto->lex_formbrack;
10807 parser->lex_inpat = proto->lex_inpat;
10808 parser->lex_inwhat = proto->lex_inwhat;
10809 parser->lex_op = proto->lex_op;
10810 parser->lex_repl = sv_dup_inc(proto->lex_repl, param);
10811 parser->lex_starts = proto->lex_starts;
10812 parser->lex_stuff = sv_dup_inc(proto->lex_stuff, param);
10813 parser->multi_close = proto->multi_close;
10814 parser->multi_open = proto->multi_open;
10815 parser->multi_start = proto->multi_start;
10816 parser->multi_end = proto->multi_end;
10817 parser->pending_ident = proto->pending_ident;
10818 parser->preambled = proto->preambled;
10819 parser->sublex_info = proto->sublex_info; /* XXX not quite right */
10820 parser->linestr = sv_dup_inc(proto->linestr, param);
10821 parser->expect = proto->expect;
10822 parser->copline = proto->copline;
10823 parser->last_lop_op = proto->last_lop_op;
10824 parser->lex_state = proto->lex_state;
10825 parser->rsfp = fp_dup(proto->rsfp, '<', param);
10826 /* rsfp_filters entries have fake IoDIRP() */
10827 parser->rsfp_filters= av_dup_inc(proto->rsfp_filters, param);
10828 parser->in_my = proto->in_my;
10829 parser->in_my_stash = hv_dup(proto->in_my_stash, param);
10830 parser->error_count = proto->error_count;
10833 parser->linestr = sv_dup_inc(proto->linestr, param);
10836 char * const ols = SvPVX(proto->linestr);
10837 char * const ls = SvPVX(parser->linestr);
10839 parser->bufptr = ls + (proto->bufptr >= ols ?
10840 proto->bufptr - ols : 0);
10841 parser->oldbufptr = ls + (proto->oldbufptr >= ols ?
10842 proto->oldbufptr - ols : 0);
10843 parser->oldoldbufptr= ls + (proto->oldoldbufptr >= ols ?
10844 proto->oldoldbufptr - ols : 0);
10845 parser->linestart = ls + (proto->linestart >= ols ?
10846 proto->linestart - ols : 0);
10847 parser->last_uni = ls + (proto->last_uni >= ols ?
10848 proto->last_uni - ols : 0);
10849 parser->last_lop = ls + (proto->last_lop >= ols ?
10850 proto->last_lop - ols : 0);
10852 parser->bufend = ls + SvCUR(parser->linestr);
10855 Copy(proto->tokenbuf, parser->tokenbuf, 256, char);
10859 parser->endwhite = proto->endwhite;
10860 parser->faketokens = proto->faketokens;
10861 parser->lasttoke = proto->lasttoke;
10862 parser->nextwhite = proto->nextwhite;
10863 parser->realtokenstart = proto->realtokenstart;
10864 parser->skipwhite = proto->skipwhite;
10865 parser->thisclose = proto->thisclose;
10866 parser->thismad = proto->thismad;
10867 parser->thisopen = proto->thisopen;
10868 parser->thisstuff = proto->thisstuff;
10869 parser->thistoken = proto->thistoken;
10870 parser->thiswhite = proto->thiswhite;
10872 Copy(proto->nexttoke, parser->nexttoke, 5, NEXTTOKE);
10873 parser->curforce = proto->curforce;
10875 Copy(proto->nextval, parser->nextval, 5, YYSTYPE);
10876 Copy(proto->nexttype, parser->nexttype, 5, I32);
10877 parser->nexttoke = proto->nexttoke;
10880 /* XXX should clone saved_curcop here, but we aren't passed
10881 * proto_perl; so do it in perl_clone_using instead */
10887 /* duplicate a file handle */
10890 Perl_fp_dup(pTHX_ PerlIO *const fp, const char type, CLONE_PARAMS *const param)
10894 PERL_ARGS_ASSERT_FP_DUP;
10895 PERL_UNUSED_ARG(type);
10898 return (PerlIO*)NULL;
10900 /* look for it in the table first */
10901 ret = (PerlIO*)ptr_table_fetch(PL_ptr_table, fp);
10905 /* create anew and remember what it is */
10906 ret = PerlIO_fdupopen(aTHX_ fp, param, PERLIO_DUP_CLONE);
10907 ptr_table_store(PL_ptr_table, fp, ret);
10911 /* duplicate a directory handle */
10914 Perl_dirp_dup(pTHX_ DIR *const dp)
10919 register const Direntry_t *dirent;
10920 char smallbuf[256];
10926 PERL_UNUSED_CONTEXT;
10931 /* look for it in the table first */
10932 ret = (DIR*)ptr_table_fetch(PL_ptr_table, dp);
10938 /* open the current directory (so we can switch back) */
10939 if (!(pwd = PerlDir_open("."))) return (DIR *)NULL;
10941 /* chdir to our dir handle and open the present working directory */
10942 if (fchdir(my_dirfd(dp)) < 0 || !(ret = PerlDir_open("."))) {
10943 PerlDir_close(pwd);
10944 return (DIR *)NULL;
10946 /* Now we should have two dir handles pointing to the same dir. */
10948 /* Be nice to the calling code and chdir back to where we were. */
10949 fchdir(my_dirfd(pwd)); /* If this fails, then what? */
10951 /* We have no need of the pwd handle any more. */
10952 PerlDir_close(pwd);
10955 # define d_namlen(d) (d)->d_namlen
10957 # define d_namlen(d) strlen((d)->d_name)
10959 /* Iterate once through dp, to get the file name at the current posi-
10960 tion. Then step back. */
10961 pos = PerlDir_tell(dp);
10962 if ((dirent = PerlDir_read(dp))) {
10963 len = d_namlen(dirent);
10964 if (len <= sizeof smallbuf) name = smallbuf;
10965 else Newx(name, len, char);
10966 Move(dirent->d_name, name, len, char);
10968 PerlDir_seek(dp, pos);
10970 /* Iterate through the new dir handle, till we find a file with the
10972 if (!dirent) /* just before the end */
10974 pos = PerlDir_tell(ret);
10975 if (PerlDir_read(ret)) continue; /* not there yet */
10976 PerlDir_seek(ret, pos); /* step back */
10980 const long pos0 = PerlDir_tell(ret);
10982 pos = PerlDir_tell(ret);
10983 if ((dirent = PerlDir_read(ret))) {
10984 if (len == d_namlen(dirent)
10985 && memEQ(name, dirent->d_name, len)) {
10987 PerlDir_seek(ret, pos); /* step back */
10990 /* else we are not there yet; keep iterating */
10992 else { /* This is not meant to happen. The best we can do is
10993 reset the iterator to the beginning. */
10994 PerlDir_seek(ret, pos0);
11001 if (name && name != smallbuf)
11004 /* pop it in the pointer table */
11005 ptr_table_store(PL_ptr_table, dp, ret);
11013 /* duplicate a typeglob */
11016 Perl_gp_dup(pTHX_ GP *const gp, CLONE_PARAMS *const param)
11020 PERL_ARGS_ASSERT_GP_DUP;
11024 /* look for it in the table first */
11025 ret = (GP*)ptr_table_fetch(PL_ptr_table, gp);
11029 /* create anew and remember what it is */
11031 ptr_table_store(PL_ptr_table, gp, ret);
11034 /* ret->gp_refcnt must be 0 before any other dups are called. We're relying
11035 on Newxz() to do this for us. */
11036 ret->gp_sv = sv_dup_inc(gp->gp_sv, param);
11037 ret->gp_io = io_dup_inc(gp->gp_io, param);
11038 ret->gp_form = cv_dup_inc(gp->gp_form, param);
11039 ret->gp_av = av_dup_inc(gp->gp_av, param);
11040 ret->gp_hv = hv_dup_inc(gp->gp_hv, param);
11041 ret->gp_egv = gv_dup(gp->gp_egv, param);/* GvEGV is not refcounted */
11042 ret->gp_cv = cv_dup_inc(gp->gp_cv, param);
11043 ret->gp_cvgen = gp->gp_cvgen;
11044 ret->gp_line = gp->gp_line;
11045 ret->gp_file_hek = hek_dup(gp->gp_file_hek, param);
11049 /* duplicate a chain of magic */
11052 Perl_mg_dup(pTHX_ MAGIC *mg, CLONE_PARAMS *const param)
11054 MAGIC *mgret = NULL;
11055 MAGIC **mgprev_p = &mgret;
11057 PERL_ARGS_ASSERT_MG_DUP;
11059 for (; mg; mg = mg->mg_moremagic) {
11062 if ((param->flags & CLONEf_JOIN_IN)
11063 && mg->mg_type == PERL_MAGIC_backref)
11064 /* when joining, we let the individual SVs add themselves to
11065 * backref as needed. */
11068 Newx(nmg, 1, MAGIC);
11070 mgprev_p = &(nmg->mg_moremagic);
11072 /* There was a comment "XXX copy dynamic vtable?" but as we don't have
11073 dynamic vtables, I'm not sure why Sarathy wrote it. The comment dates
11074 from the original commit adding Perl_mg_dup() - revision 4538.
11075 Similarly there is the annotation "XXX random ptr?" next to the
11076 assignment to nmg->mg_ptr. */
11079 /* FIXME for plugins
11080 if (nmg->mg_type == PERL_MAGIC_qr) {
11081 nmg->mg_obj = MUTABLE_SV(CALLREGDUPE((REGEXP*)nmg->mg_obj, param));
11085 nmg->mg_obj = (nmg->mg_flags & MGf_REFCOUNTED)
11086 ? nmg->mg_type == PERL_MAGIC_backref
11087 /* The backref AV has its reference
11088 * count deliberately bumped by 1 */
11089 ? SvREFCNT_inc(av_dup_inc((const AV *)
11090 nmg->mg_obj, param))
11091 : sv_dup_inc(nmg->mg_obj, param)
11092 : sv_dup(nmg->mg_obj, param);
11094 if (nmg->mg_ptr && nmg->mg_type != PERL_MAGIC_regex_global) {
11095 if (nmg->mg_len > 0) {
11096 nmg->mg_ptr = SAVEPVN(nmg->mg_ptr, nmg->mg_len);
11097 if (nmg->mg_type == PERL_MAGIC_overload_table &&
11098 AMT_AMAGIC((AMT*)nmg->mg_ptr))
11100 AMT * const namtp = (AMT*)nmg->mg_ptr;
11101 sv_dup_inc_multiple((SV**)(namtp->table),
11102 (SV**)(namtp->table), NofAMmeth, param);
11105 else if (nmg->mg_len == HEf_SVKEY)
11106 nmg->mg_ptr = (char*)sv_dup_inc((const SV *)nmg->mg_ptr, param);
11108 if ((nmg->mg_flags & MGf_DUP) && nmg->mg_virtual && nmg->mg_virtual->svt_dup) {
11109 nmg->mg_virtual->svt_dup(aTHX_ nmg, param);
11115 #endif /* USE_ITHREADS */
11117 struct ptr_tbl_arena {
11118 struct ptr_tbl_arena *next;
11119 struct ptr_tbl_ent array[1023/3]; /* as ptr_tbl_ent has 3 pointers. */
11122 /* create a new pointer-mapping table */
11125 Perl_ptr_table_new(pTHX)
11128 PERL_UNUSED_CONTEXT;
11130 Newx(tbl, 1, PTR_TBL_t);
11131 tbl->tbl_max = 511;
11132 tbl->tbl_items = 0;
11133 tbl->tbl_arena = NULL;
11134 tbl->tbl_arena_next = NULL;
11135 tbl->tbl_arena_end = NULL;
11136 Newxz(tbl->tbl_ary, tbl->tbl_max + 1, PTR_TBL_ENT_t*);
11140 #define PTR_TABLE_HASH(ptr) \
11141 ((PTR2UV(ptr) >> 3) ^ (PTR2UV(ptr) >> (3 + 7)) ^ (PTR2UV(ptr) >> (3 + 17)))
11143 /* map an existing pointer using a table */
11145 STATIC PTR_TBL_ENT_t *
11146 S_ptr_table_find(PTR_TBL_t *const tbl, const void *const sv)
11148 PTR_TBL_ENT_t *tblent;
11149 const UV hash = PTR_TABLE_HASH(sv);
11151 PERL_ARGS_ASSERT_PTR_TABLE_FIND;
11153 tblent = tbl->tbl_ary[hash & tbl->tbl_max];
11154 for (; tblent; tblent = tblent->next) {
11155 if (tblent->oldval == sv)
11162 Perl_ptr_table_fetch(pTHX_ PTR_TBL_t *const tbl, const void *const sv)
11164 PTR_TBL_ENT_t const *const tblent = ptr_table_find(tbl, sv);
11166 PERL_ARGS_ASSERT_PTR_TABLE_FETCH;
11167 PERL_UNUSED_CONTEXT;
11169 return tblent ? tblent->newval : NULL;
11172 /* add a new entry to a pointer-mapping table */
11175 Perl_ptr_table_store(pTHX_ PTR_TBL_t *const tbl, const void *const oldsv, void *const newsv)
11177 PTR_TBL_ENT_t *tblent = ptr_table_find(tbl, oldsv);
11179 PERL_ARGS_ASSERT_PTR_TABLE_STORE;
11180 PERL_UNUSED_CONTEXT;
11183 tblent->newval = newsv;
11185 const UV entry = PTR_TABLE_HASH(oldsv) & tbl->tbl_max;
11187 if (tbl->tbl_arena_next == tbl->tbl_arena_end) {
11188 struct ptr_tbl_arena *new_arena;
11190 Newx(new_arena, 1, struct ptr_tbl_arena);
11191 new_arena->next = tbl->tbl_arena;
11192 tbl->tbl_arena = new_arena;
11193 tbl->tbl_arena_next = new_arena->array;
11194 tbl->tbl_arena_end = new_arena->array
11195 + sizeof(new_arena->array) / sizeof(new_arena->array[0]);
11198 tblent = tbl->tbl_arena_next++;
11200 tblent->oldval = oldsv;
11201 tblent->newval = newsv;
11202 tblent->next = tbl->tbl_ary[entry];
11203 tbl->tbl_ary[entry] = tblent;
11205 if (tblent->next && tbl->tbl_items > tbl->tbl_max)
11206 ptr_table_split(tbl);
11210 /* double the hash bucket size of an existing ptr table */
11213 Perl_ptr_table_split(pTHX_ PTR_TBL_t *const tbl)
11215 PTR_TBL_ENT_t **ary = tbl->tbl_ary;
11216 const UV oldsize = tbl->tbl_max + 1;
11217 UV newsize = oldsize * 2;
11220 PERL_ARGS_ASSERT_PTR_TABLE_SPLIT;
11221 PERL_UNUSED_CONTEXT;
11223 Renew(ary, newsize, PTR_TBL_ENT_t*);
11224 Zero(&ary[oldsize], newsize-oldsize, PTR_TBL_ENT_t*);
11225 tbl->tbl_max = --newsize;
11226 tbl->tbl_ary = ary;
11227 for (i=0; i < oldsize; i++, ary++) {
11228 PTR_TBL_ENT_t **entp = ary;
11229 PTR_TBL_ENT_t *ent = *ary;
11230 PTR_TBL_ENT_t **curentp;
11233 curentp = ary + oldsize;
11235 if ((newsize & PTR_TABLE_HASH(ent->oldval)) != i) {
11237 ent->next = *curentp;
11247 /* remove all the entries from a ptr table */
11248 /* Deprecated - will be removed post 5.14 */
11251 Perl_ptr_table_clear(pTHX_ PTR_TBL_t *const tbl)
11253 if (tbl && tbl->tbl_items) {
11254 struct ptr_tbl_arena *arena = tbl->tbl_arena;
11256 Zero(tbl->tbl_ary, tbl->tbl_max + 1, struct ptr_tbl_ent **);
11259 struct ptr_tbl_arena *next = arena->next;
11265 tbl->tbl_items = 0;
11266 tbl->tbl_arena = NULL;
11267 tbl->tbl_arena_next = NULL;
11268 tbl->tbl_arena_end = NULL;
11272 /* clear and free a ptr table */
11275 Perl_ptr_table_free(pTHX_ PTR_TBL_t *const tbl)
11277 struct ptr_tbl_arena *arena;
11283 arena = tbl->tbl_arena;
11286 struct ptr_tbl_arena *next = arena->next;
11292 Safefree(tbl->tbl_ary);
11296 #if defined(USE_ITHREADS)
11299 Perl_rvpv_dup(pTHX_ SV *const dstr, const SV *const sstr, CLONE_PARAMS *const param)
11301 PERL_ARGS_ASSERT_RVPV_DUP;
11304 if (SvWEAKREF(sstr)) {
11305 SvRV_set(dstr, sv_dup(SvRV_const(sstr), param));
11306 if (param->flags & CLONEf_JOIN_IN) {
11307 /* if joining, we add any back references individually rather
11308 * than copying the whole backref array */
11309 Perl_sv_add_backref(aTHX_ SvRV(dstr), dstr);
11313 SvRV_set(dstr, sv_dup_inc(SvRV_const(sstr), param));
11315 else if (SvPVX_const(sstr)) {
11316 /* Has something there */
11318 /* Normal PV - clone whole allocated space */
11319 SvPV_set(dstr, SAVEPVN(SvPVX_const(sstr), SvLEN(sstr)-1));
11320 if (SvREADONLY(sstr) && SvFAKE(sstr)) {
11321 /* Not that normal - actually sstr is copy on write.
11322 But we are a true, independant SV, so: */
11323 SvREADONLY_off(dstr);
11328 /* Special case - not normally malloced for some reason */
11329 if (isGV_with_GP(sstr)) {
11330 /* Don't need to do anything here. */
11332 else if ((SvREADONLY(sstr) && SvFAKE(sstr))) {
11333 /* A "shared" PV - clone it as "shared" PV */
11335 HEK_KEY(hek_dup(SvSHARED_HEK_FROM_PV(SvPVX_const(sstr)),
11339 /* Some other special case - random pointer */
11340 SvPV_set(dstr, (char *) SvPVX_const(sstr));
11345 /* Copy the NULL */
11346 SvPV_set(dstr, NULL);
11350 /* duplicate a list of SVs. source and dest may point to the same memory. */
11352 S_sv_dup_inc_multiple(pTHX_ SV *const *source, SV **dest,
11353 SSize_t items, CLONE_PARAMS *const param)
11355 PERL_ARGS_ASSERT_SV_DUP_INC_MULTIPLE;
11357 while (items-- > 0) {
11358 *dest++ = sv_dup_inc(*source++, param);
11364 /* duplicate an SV of any type (including AV, HV etc) */
11367 S_sv_dup_common(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
11372 PERL_ARGS_ASSERT_SV_DUP_COMMON;
11374 if (SvTYPE(sstr) == SVTYPEMASK) {
11375 #ifdef DEBUG_LEAKING_SCALARS_ABORT
11380 /* look for it in the table first */
11381 dstr = MUTABLE_SV(ptr_table_fetch(PL_ptr_table, sstr));
11385 if(param->flags & CLONEf_JOIN_IN) {
11386 /** We are joining here so we don't want do clone
11387 something that is bad **/
11388 if (SvTYPE(sstr) == SVt_PVHV) {
11389 const HEK * const hvname = HvNAME_HEK(sstr);
11391 /** don't clone stashes if they already exist **/
11392 dstr = MUTABLE_SV(gv_stashpvn(HEK_KEY(hvname), HEK_LEN(hvname), 0));
11393 ptr_table_store(PL_ptr_table, sstr, dstr);
11399 /* create anew and remember what it is */
11402 #ifdef DEBUG_LEAKING_SCALARS
11403 dstr->sv_debug_optype = sstr->sv_debug_optype;
11404 dstr->sv_debug_line = sstr->sv_debug_line;
11405 dstr->sv_debug_inpad = sstr->sv_debug_inpad;
11406 dstr->sv_debug_parent = (SV*)sstr;
11407 dstr->sv_debug_file = savepv(sstr->sv_debug_file);
11410 ptr_table_store(PL_ptr_table, sstr, dstr);
11413 SvFLAGS(dstr) = SvFLAGS(sstr);
11414 SvFLAGS(dstr) &= ~SVf_OOK; /* don't propagate OOK hack */
11415 SvREFCNT(dstr) = 0; /* must be before any other dups! */
11418 if (SvANY(sstr) && PL_watch_pvx && SvPVX_const(sstr) == PL_watch_pvx)
11419 PerlIO_printf(Perl_debug_log, "watch at %p hit, found string \"%s\"\n",
11420 (void*)PL_watch_pvx, SvPVX_const(sstr));
11423 /* don't clone objects whose class has asked us not to */
11424 if (SvOBJECT(sstr) && ! (SvFLAGS(SvSTASH(sstr)) & SVphv_CLONEABLE)) {
11429 switch (SvTYPE(sstr)) {
11431 SvANY(dstr) = NULL;
11434 SvANY(dstr) = (XPVIV*)((char*)&(dstr->sv_u.svu_iv) - STRUCT_OFFSET(XPVIV, xiv_iv));
11436 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
11438 SvIV_set(dstr, SvIVX(sstr));
11442 SvANY(dstr) = new_XNV();
11443 SvNV_set(dstr, SvNVX(sstr));
11445 /* case SVt_BIND: */
11448 /* These are all the types that need complex bodies allocating. */
11450 const svtype sv_type = SvTYPE(sstr);
11451 const struct body_details *const sv_type_details
11452 = bodies_by_type + sv_type;
11456 Perl_croak(aTHX_ "Bizarre SvTYPE [%" IVdf "]", (IV)SvTYPE(sstr));
11471 assert(sv_type_details->body_size);
11472 if (sv_type_details->arena) {
11473 new_body_inline(new_body, sv_type);
11475 = (void*)((char*)new_body - sv_type_details->offset);
11477 new_body = new_NOARENA(sv_type_details);
11481 SvANY(dstr) = new_body;
11484 Copy(((char*)SvANY(sstr)) + sv_type_details->offset,
11485 ((char*)SvANY(dstr)) + sv_type_details->offset,
11486 sv_type_details->copy, char);
11488 Copy(((char*)SvANY(sstr)),
11489 ((char*)SvANY(dstr)),
11490 sv_type_details->body_size + sv_type_details->offset, char);
11493 if (sv_type != SVt_PVAV && sv_type != SVt_PVHV
11494 && !isGV_with_GP(dstr)
11495 && !(sv_type == SVt_PVIO && !(IoFLAGS(dstr) & IOf_FAKE_DIRP)))
11496 Perl_rvpv_dup(aTHX_ dstr, sstr, param);
11498 /* The Copy above means that all the source (unduplicated) pointers
11499 are now in the destination. We can check the flags and the
11500 pointers in either, but it's possible that there's less cache
11501 missing by always going for the destination.
11502 FIXME - instrument and check that assumption */
11503 if (sv_type >= SVt_PVMG) {
11504 if ((sv_type == SVt_PVMG) && SvPAD_OUR(dstr)) {
11505 SvOURSTASH_set(dstr, hv_dup_inc(SvOURSTASH(dstr), param));
11506 } else if (SvMAGIC(dstr))
11507 SvMAGIC_set(dstr, mg_dup(SvMAGIC(dstr), param));
11509 SvSTASH_set(dstr, hv_dup_inc(SvSTASH(dstr), param));
11512 /* The cast silences a GCC warning about unhandled types. */
11513 switch ((int)sv_type) {
11523 /* FIXME for plugins */
11524 re_dup_guts((REGEXP*) sstr, (REGEXP*) dstr, param);
11527 /* XXX LvTARGOFF sometimes holds PMOP* when DEBUGGING */
11528 if (LvTYPE(dstr) == 't') /* for tie: unrefcnted fake (SV**) */
11529 LvTARG(dstr) = dstr;
11530 else if (LvTYPE(dstr) == 'T') /* for tie: fake HE */
11531 LvTARG(dstr) = MUTABLE_SV(he_dup((HE*)LvTARG(dstr), 0, param));
11533 LvTARG(dstr) = sv_dup_inc(LvTARG(dstr), param);
11535 /* non-GP case already handled above */
11536 if(isGV_with_GP(sstr)) {
11537 GvNAME_HEK(dstr) = hek_dup(GvNAME_HEK(dstr), param);
11538 /* Don't call sv_add_backref here as it's going to be
11539 created as part of the magic cloning of the symbol
11540 table--unless this is during a join and the stash
11541 is not actually being cloned. */
11542 /* Danger Will Robinson - GvGP(dstr) isn't initialised
11543 at the point of this comment. */
11544 GvSTASH(dstr) = hv_dup(GvSTASH(dstr), param);
11545 if (param->flags & CLONEf_JOIN_IN)
11546 Perl_sv_add_backref(aTHX_ MUTABLE_SV(GvSTASH(dstr)), dstr);
11547 GvGP(dstr) = gp_dup(GvGP(sstr), param);
11548 (void)GpREFCNT_inc(GvGP(dstr));
11552 /* PL_parser->rsfp_filters entries have fake IoDIRP() */
11553 if(IoFLAGS(dstr) & IOf_FAKE_DIRP) {
11554 /* I have no idea why fake dirp (rsfps)
11555 should be treated differently but otherwise
11556 we end up with leaks -- sky*/
11557 IoTOP_GV(dstr) = gv_dup_inc(IoTOP_GV(dstr), param);
11558 IoFMT_GV(dstr) = gv_dup_inc(IoFMT_GV(dstr), param);
11559 IoBOTTOM_GV(dstr) = gv_dup_inc(IoBOTTOM_GV(dstr), param);
11561 IoTOP_GV(dstr) = gv_dup(IoTOP_GV(dstr), param);
11562 IoFMT_GV(dstr) = gv_dup(IoFMT_GV(dstr), param);
11563 IoBOTTOM_GV(dstr) = gv_dup(IoBOTTOM_GV(dstr), param);
11564 if (IoDIRP(dstr)) {
11565 IoDIRP(dstr) = dirp_dup(IoDIRP(dstr));
11568 /* IoDIRP(dstr) is already a copy of IoDIRP(sstr) */
11570 IoIFP(dstr) = fp_dup(IoIFP(sstr), IoTYPE(dstr), param);
11572 if (IoOFP(dstr) == IoIFP(sstr))
11573 IoOFP(dstr) = IoIFP(dstr);
11575 IoOFP(dstr) = fp_dup(IoOFP(dstr), IoTYPE(dstr), param);
11576 IoTOP_NAME(dstr) = SAVEPV(IoTOP_NAME(dstr));
11577 IoFMT_NAME(dstr) = SAVEPV(IoFMT_NAME(dstr));
11578 IoBOTTOM_NAME(dstr) = SAVEPV(IoBOTTOM_NAME(dstr));
11581 /* avoid cloning an empty array */
11582 if (AvARRAY((const AV *)sstr) && AvFILLp((const AV *)sstr) >= 0) {
11583 SV **dst_ary, **src_ary;
11584 SSize_t items = AvFILLp((const AV *)sstr) + 1;
11586 src_ary = AvARRAY((const AV *)sstr);
11587 Newxz(dst_ary, AvMAX((const AV *)sstr)+1, SV*);
11588 ptr_table_store(PL_ptr_table, src_ary, dst_ary);
11589 AvARRAY(MUTABLE_AV(dstr)) = dst_ary;
11590 AvALLOC((const AV *)dstr) = dst_ary;
11591 if (AvREAL((const AV *)sstr)) {
11592 dst_ary = sv_dup_inc_multiple(src_ary, dst_ary, items,
11596 while (items-- > 0)
11597 *dst_ary++ = sv_dup(*src_ary++, param);
11599 items = AvMAX((const AV *)sstr) - AvFILLp((const AV *)sstr);
11600 while (items-- > 0) {
11601 *dst_ary++ = &PL_sv_undef;
11605 AvARRAY(MUTABLE_AV(dstr)) = NULL;
11606 AvALLOC((const AV *)dstr) = (SV**)NULL;
11607 AvMAX( (const AV *)dstr) = -1;
11608 AvFILLp((const AV *)dstr) = -1;
11612 if (HvARRAY((const HV *)sstr)) {
11614 const bool sharekeys = !!HvSHAREKEYS(sstr);
11615 XPVHV * const dxhv = (XPVHV*)SvANY(dstr);
11616 XPVHV * const sxhv = (XPVHV*)SvANY(sstr);
11618 Newx(darray, PERL_HV_ARRAY_ALLOC_BYTES(dxhv->xhv_max+1)
11619 + (SvOOK(sstr) ? sizeof(struct xpvhv_aux) : 0),
11621 HvARRAY(dstr) = (HE**)darray;
11622 while (i <= sxhv->xhv_max) {
11623 const HE * const source = HvARRAY(sstr)[i];
11624 HvARRAY(dstr)[i] = source
11625 ? he_dup(source, sharekeys, param) : 0;
11630 const struct xpvhv_aux * const saux = HvAUX(sstr);
11631 struct xpvhv_aux * const daux = HvAUX(dstr);
11632 /* This flag isn't copied. */
11633 /* SvOOK_on(hv) attacks the IV flags. */
11634 SvFLAGS(dstr) |= SVf_OOK;
11636 hvname = saux->xhv_name;
11637 daux->xhv_name = hek_dup(hvname, param);
11639 daux->xhv_riter = saux->xhv_riter;
11640 daux->xhv_eiter = saux->xhv_eiter
11641 ? he_dup(saux->xhv_eiter,
11642 cBOOL(HvSHAREKEYS(sstr)), param) : 0;
11643 /* backref array needs refcnt=2; see sv_add_backref */
11644 daux->xhv_backreferences =
11645 (param->flags & CLONEf_JOIN_IN)
11646 /* when joining, we let the individual GVs and
11647 * CVs add themselves to backref as
11648 * needed. This avoids pulling in stuff
11649 * that isn't required, and simplifies the
11650 * case where stashes aren't cloned back
11651 * if they already exist in the parent
11654 : saux->xhv_backreferences
11655 ? (SvTYPE(saux->xhv_backreferences) == SVt_PVAV)
11656 ? MUTABLE_AV(SvREFCNT_inc(
11657 sv_dup_inc((const SV *)
11658 saux->xhv_backreferences, param)))
11659 : MUTABLE_AV(sv_dup((const SV *)
11660 saux->xhv_backreferences, param))
11663 daux->xhv_mro_meta = saux->xhv_mro_meta
11664 ? mro_meta_dup(saux->xhv_mro_meta, param)
11667 /* Record stashes for possible cloning in Perl_clone(). */
11669 av_push(param->stashes, dstr);
11673 HvARRAY(MUTABLE_HV(dstr)) = NULL;
11676 if (!(param->flags & CLONEf_COPY_STACKS)) {
11681 /* NOTE: not refcounted */
11682 CvSTASH(dstr) = hv_dup(CvSTASH(dstr), param);
11683 if ((param->flags & CLONEf_JOIN_IN) && CvSTASH(dstr))
11684 Perl_sv_add_backref(aTHX_ MUTABLE_SV(CvSTASH(dstr)), dstr);
11686 if (!CvISXSUB(dstr))
11687 CvROOT(dstr) = OpREFCNT_inc(CvROOT(dstr));
11689 if (CvCONST(dstr) && CvISXSUB(dstr)) {
11690 CvXSUBANY(dstr).any_ptr =
11691 sv_dup_inc((const SV *)CvXSUBANY(dstr).any_ptr, param);
11693 /* don't dup if copying back - CvGV isn't refcounted, so the
11694 * duped GV may never be freed. A bit of a hack! DAPM */
11695 SvANY(MUTABLE_CV(dstr))->xcv_gv =
11697 ? gv_dup_inc(CvGV(sstr), param)
11698 : (param->flags & CLONEf_JOIN_IN)
11700 : gv_dup(CvGV(sstr), param);
11702 CvPADLIST(dstr) = padlist_dup(CvPADLIST(sstr), param);
11704 CvWEAKOUTSIDE(sstr)
11705 ? cv_dup( CvOUTSIDE(dstr), param)
11706 : cv_dup_inc(CvOUTSIDE(dstr), param);
11707 if (!CvISXSUB(dstr))
11708 CvFILE(dstr) = SAVEPV(CvFILE(dstr));
11714 if (SvOBJECT(dstr) && SvTYPE(dstr) != SVt_PVIO)
11721 Perl_sv_dup_inc(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
11723 PERL_ARGS_ASSERT_SV_DUP_INC;
11724 return sstr ? SvREFCNT_inc(sv_dup_common(sstr, param)) : NULL;
11728 Perl_sv_dup(pTHX_ const SV *const sstr, CLONE_PARAMS *const param)
11730 SV *dstr = sstr ? sv_dup_common(sstr, param) : NULL;
11731 PERL_ARGS_ASSERT_SV_DUP;
11733 /* Track every SV that (at least initially) had a reference count of 0.
11734 We need to do this by holding an actual reference to it in this array.
11735 If we attempt to cheat, turn AvREAL_off(), and store only pointers
11736 (akin to the stashes hash, and the perl stack), we come unstuck if
11737 a weak reference (or other SV legitimately SvREFCNT() == 0 for this
11738 thread) is manipulated in a CLONE method, because CLONE runs before the
11739 unreferenced array is walked to find SVs still with SvREFCNT() == 0
11740 (and fix things up by giving each a reference via the temps stack).
11741 Instead, during CLONE, if the 0-referenced SV has SvREFCNT_inc() and
11742 then SvREFCNT_dec(), it will be cleaned up (and added to the free list)
11743 before the walk of unreferenced happens and a reference to that is SV
11744 added to the temps stack. At which point we have the same SV considered
11745 to be in use, and free to be re-used. Not good.
11747 if (dstr && !(param->flags & CLONEf_COPY_STACKS) && !SvREFCNT(dstr)) {
11748 assert(param->unreferenced);
11749 av_push(param->unreferenced, SvREFCNT_inc(dstr));
11755 /* duplicate a context */
11758 Perl_cx_dup(pTHX_ PERL_CONTEXT *cxs, I32 ix, I32 max, CLONE_PARAMS* param)
11760 PERL_CONTEXT *ncxs;
11762 PERL_ARGS_ASSERT_CX_DUP;
11765 return (PERL_CONTEXT*)NULL;
11767 /* look for it in the table first */
11768 ncxs = (PERL_CONTEXT*)ptr_table_fetch(PL_ptr_table, cxs);
11772 /* create anew and remember what it is */
11773 Newx(ncxs, max + 1, PERL_CONTEXT);
11774 ptr_table_store(PL_ptr_table, cxs, ncxs);
11775 Copy(cxs, ncxs, max + 1, PERL_CONTEXT);
11778 PERL_CONTEXT * const ncx = &ncxs[ix];
11779 if (CxTYPE(ncx) == CXt_SUBST) {
11780 Perl_croak(aTHX_ "Cloning substitution context is unimplemented");
11783 switch (CxTYPE(ncx)) {
11785 ncx->blk_sub.cv = (ncx->blk_sub.olddepth == 0
11786 ? cv_dup_inc(ncx->blk_sub.cv, param)
11787 : cv_dup(ncx->blk_sub.cv,param));
11788 ncx->blk_sub.argarray = (CxHASARGS(ncx)
11789 ? av_dup_inc(ncx->blk_sub.argarray,
11792 ncx->blk_sub.savearray = av_dup_inc(ncx->blk_sub.savearray,
11794 ncx->blk_sub.oldcomppad = (PAD*)ptr_table_fetch(PL_ptr_table,
11795 ncx->blk_sub.oldcomppad);
11798 ncx->blk_eval.old_namesv = sv_dup_inc(ncx->blk_eval.old_namesv,
11800 ncx->blk_eval.cur_text = sv_dup(ncx->blk_eval.cur_text, param);
11802 case CXt_LOOP_LAZYSV:
11803 ncx->blk_loop.state_u.lazysv.end
11804 = sv_dup_inc(ncx->blk_loop.state_u.lazysv.end, param);
11805 /* We are taking advantage of av_dup_inc and sv_dup_inc
11806 actually being the same function, and order equivalance of
11808 We can assert the later [but only at run time :-(] */
11809 assert ((void *) &ncx->blk_loop.state_u.ary.ary ==
11810 (void *) &ncx->blk_loop.state_u.lazysv.cur);
11812 ncx->blk_loop.state_u.ary.ary
11813 = av_dup_inc(ncx->blk_loop.state_u.ary.ary, param);
11814 case CXt_LOOP_LAZYIV:
11815 case CXt_LOOP_PLAIN:
11816 if (CxPADLOOP(ncx)) {
11817 ncx->blk_loop.itervar_u.oldcomppad
11818 = (PAD*)ptr_table_fetch(PL_ptr_table,
11819 ncx->blk_loop.itervar_u.oldcomppad);
11821 ncx->blk_loop.itervar_u.gv
11822 = gv_dup((const GV *)ncx->blk_loop.itervar_u.gv,
11827 ncx->blk_format.cv = cv_dup(ncx->blk_format.cv, param);
11828 ncx->blk_format.gv = gv_dup(ncx->blk_format.gv, param);
11829 ncx->blk_format.dfoutgv = gv_dup_inc(ncx->blk_format.dfoutgv,
11842 /* duplicate a stack info structure */
11845 Perl_si_dup(pTHX_ PERL_SI *si, CLONE_PARAMS* param)
11849 PERL_ARGS_ASSERT_SI_DUP;
11852 return (PERL_SI*)NULL;
11854 /* look for it in the table first */
11855 nsi = (PERL_SI*)ptr_table_fetch(PL_ptr_table, si);
11859 /* create anew and remember what it is */
11860 Newxz(nsi, 1, PERL_SI);
11861 ptr_table_store(PL_ptr_table, si, nsi);
11863 nsi->si_stack = av_dup_inc(si->si_stack, param);
11864 nsi->si_cxix = si->si_cxix;
11865 nsi->si_cxmax = si->si_cxmax;
11866 nsi->si_cxstack = cx_dup(si->si_cxstack, si->si_cxix, si->si_cxmax, param);
11867 nsi->si_type = si->si_type;
11868 nsi->si_prev = si_dup(si->si_prev, param);
11869 nsi->si_next = si_dup(si->si_next, param);
11870 nsi->si_markoff = si->si_markoff;
11875 #define POPINT(ss,ix) ((ss)[--(ix)].any_i32)
11876 #define TOPINT(ss,ix) ((ss)[ix].any_i32)
11877 #define POPLONG(ss,ix) ((ss)[--(ix)].any_long)
11878 #define TOPLONG(ss,ix) ((ss)[ix].any_long)
11879 #define POPIV(ss,ix) ((ss)[--(ix)].any_iv)
11880 #define TOPIV(ss,ix) ((ss)[ix].any_iv)
11881 #define POPUV(ss,ix) ((ss)[--(ix)].any_uv)
11882 #define TOPUV(ss,ix) ((ss)[ix].any_uv)
11883 #define POPBOOL(ss,ix) ((ss)[--(ix)].any_bool)
11884 #define TOPBOOL(ss,ix) ((ss)[ix].any_bool)
11885 #define POPPTR(ss,ix) ((ss)[--(ix)].any_ptr)
11886 #define TOPPTR(ss,ix) ((ss)[ix].any_ptr)
11887 #define POPDPTR(ss,ix) ((ss)[--(ix)].any_dptr)
11888 #define TOPDPTR(ss,ix) ((ss)[ix].any_dptr)
11889 #define POPDXPTR(ss,ix) ((ss)[--(ix)].any_dxptr)
11890 #define TOPDXPTR(ss,ix) ((ss)[ix].any_dxptr)
11893 #define pv_dup_inc(p) SAVEPV(p)
11894 #define pv_dup(p) SAVEPV(p)
11895 #define svp_dup_inc(p,pp) any_dup(p,pp)
11897 /* map any object to the new equivent - either something in the
11898 * ptr table, or something in the interpreter structure
11902 Perl_any_dup(pTHX_ void *v, const PerlInterpreter *proto_perl)
11906 PERL_ARGS_ASSERT_ANY_DUP;
11909 return (void*)NULL;
11911 /* look for it in the table first */
11912 ret = ptr_table_fetch(PL_ptr_table, v);
11916 /* see if it is part of the interpreter structure */
11917 if (v >= (void*)proto_perl && v < (void*)(proto_perl+1))
11918 ret = (void*)(((char*)aTHX) + (((char*)v) - (char*)proto_perl));
11926 /* duplicate the save stack */
11929 Perl_ss_dup(pTHX_ PerlInterpreter *proto_perl, CLONE_PARAMS* param)
11932 ANY * const ss = proto_perl->Isavestack;
11933 const I32 max = proto_perl->Isavestack_max;
11934 I32 ix = proto_perl->Isavestack_ix;
11947 void (*dptr) (void*);
11948 void (*dxptr) (pTHX_ void*);
11950 PERL_ARGS_ASSERT_SS_DUP;
11952 Newxz(nss, max, ANY);
11955 const UV uv = POPUV(ss,ix);
11956 const U8 type = (U8)uv & SAVE_MASK;
11958 TOPUV(nss,ix) = uv;
11960 case SAVEt_CLEARSV:
11962 case SAVEt_HELEM: /* hash element */
11963 sv = (const SV *)POPPTR(ss,ix);
11964 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11966 case SAVEt_ITEM: /* normal string */
11967 case SAVEt_GVSV: /* scalar slot in GV */
11968 case SAVEt_SV: /* scalar reference */
11969 sv = (const SV *)POPPTR(ss,ix);
11970 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11973 case SAVEt_MORTALIZESV:
11974 sv = (const SV *)POPPTR(ss,ix);
11975 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11977 case SAVEt_SHARED_PVREF: /* char* in shared space */
11978 c = (char*)POPPTR(ss,ix);
11979 TOPPTR(nss,ix) = savesharedpv(c);
11980 ptr = POPPTR(ss,ix);
11981 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
11983 case SAVEt_GENERIC_SVREF: /* generic sv */
11984 case SAVEt_SVREF: /* scalar reference */
11985 sv = (const SV *)POPPTR(ss,ix);
11986 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11987 ptr = POPPTR(ss,ix);
11988 TOPPTR(nss,ix) = svp_dup_inc((SV**)ptr, proto_perl);/* XXXXX */
11990 case SAVEt_HV: /* hash reference */
11991 case SAVEt_AV: /* array reference */
11992 sv = (const SV *) POPPTR(ss,ix);
11993 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
11995 case SAVEt_COMPPAD:
11997 sv = (const SV *) POPPTR(ss,ix);
11998 TOPPTR(nss,ix) = sv_dup(sv, param);
12000 case SAVEt_INT: /* int reference */
12001 ptr = POPPTR(ss,ix);
12002 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12003 intval = (int)POPINT(ss,ix);
12004 TOPINT(nss,ix) = intval;
12006 case SAVEt_LONG: /* long reference */
12007 ptr = POPPTR(ss,ix);
12008 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12009 longval = (long)POPLONG(ss,ix);
12010 TOPLONG(nss,ix) = longval;
12012 case SAVEt_I32: /* I32 reference */
12013 case SAVEt_COP_ARYBASE: /* call CopARYBASE_set */
12014 ptr = POPPTR(ss,ix);
12015 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12017 TOPINT(nss,ix) = i;
12019 case SAVEt_IV: /* IV reference */
12020 ptr = POPPTR(ss,ix);
12021 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12023 TOPIV(nss,ix) = iv;
12025 case SAVEt_HPTR: /* HV* reference */
12026 case SAVEt_APTR: /* AV* reference */
12027 case SAVEt_SPTR: /* SV* reference */
12028 ptr = POPPTR(ss,ix);
12029 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12030 sv = (const SV *)POPPTR(ss,ix);
12031 TOPPTR(nss,ix) = sv_dup(sv, param);
12033 case SAVEt_VPTR: /* random* reference */
12034 ptr = POPPTR(ss,ix);
12035 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12037 case SAVEt_INT_SMALL:
12038 case SAVEt_I32_SMALL:
12039 case SAVEt_I16: /* I16 reference */
12040 case SAVEt_I8: /* I8 reference */
12042 ptr = POPPTR(ss,ix);
12043 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12045 case SAVEt_GENERIC_PVREF: /* generic char* */
12046 case SAVEt_PPTR: /* char* reference */
12047 ptr = POPPTR(ss,ix);
12048 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12049 c = (char*)POPPTR(ss,ix);
12050 TOPPTR(nss,ix) = pv_dup(c);
12052 case SAVEt_GP: /* scalar reference */
12053 gv = (const GV *)POPPTR(ss,ix);
12054 TOPPTR(nss,ix) = gv_dup_inc(gv, param);
12055 gp = (GP*)POPPTR(ss,ix);
12056 TOPPTR(nss,ix) = gp = gp_dup(gp, param);
12057 (void)GpREFCNT_inc(gp);
12059 TOPINT(nss,ix) = i;
12062 ptr = POPPTR(ss,ix);
12063 if (ptr && (((OP*)ptr)->op_private & OPpREFCOUNTED)) {
12064 /* these are assumed to be refcounted properly */
12066 switch (((OP*)ptr)->op_type) {
12068 case OP_LEAVESUBLV:
12072 case OP_LEAVEWRITE:
12073 TOPPTR(nss,ix) = ptr;
12076 (void) OpREFCNT_inc(o);
12080 TOPPTR(nss,ix) = NULL;
12085 TOPPTR(nss,ix) = NULL;
12088 hv = (const HV *)POPPTR(ss,ix);
12089 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
12091 TOPINT(nss,ix) = i;
12094 c = (char*)POPPTR(ss,ix);
12095 TOPPTR(nss,ix) = pv_dup_inc(c);
12097 case SAVEt_STACK_POS: /* Position on Perl stack */
12099 TOPINT(nss,ix) = i;
12101 case SAVEt_DESTRUCTOR:
12102 ptr = POPPTR(ss,ix);
12103 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
12104 dptr = POPDPTR(ss,ix);
12105 TOPDPTR(nss,ix) = DPTR2FPTR(void (*)(void*),
12106 any_dup(FPTR2DPTR(void *, dptr),
12109 case SAVEt_DESTRUCTOR_X:
12110 ptr = POPPTR(ss,ix);
12111 TOPPTR(nss,ix) = any_dup(ptr, proto_perl); /* XXX quite arbitrary */
12112 dxptr = POPDXPTR(ss,ix);
12113 TOPDXPTR(nss,ix) = DPTR2FPTR(void (*)(pTHX_ void*),
12114 any_dup(FPTR2DPTR(void *, dxptr),
12117 case SAVEt_REGCONTEXT:
12119 ix -= uv >> SAVE_TIGHT_SHIFT;
12121 case SAVEt_AELEM: /* array element */
12122 sv = (const SV *)POPPTR(ss,ix);
12123 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12125 TOPINT(nss,ix) = i;
12126 av = (const AV *)POPPTR(ss,ix);
12127 TOPPTR(nss,ix) = av_dup_inc(av, param);
12130 ptr = POPPTR(ss,ix);
12131 TOPPTR(nss,ix) = ptr;
12134 ptr = POPPTR(ss,ix);
12137 ((struct refcounted_he *)ptr)->refcounted_he_refcnt++;
12138 HINTS_REFCNT_UNLOCK;
12140 TOPPTR(nss,ix) = ptr;
12142 TOPINT(nss,ix) = i;
12143 if (i & HINT_LOCALIZE_HH) {
12144 hv = (const HV *)POPPTR(ss,ix);
12145 TOPPTR(nss,ix) = hv_dup_inc(hv, param);
12148 case SAVEt_PADSV_AND_MORTALIZE:
12149 longval = (long)POPLONG(ss,ix);
12150 TOPLONG(nss,ix) = longval;
12151 ptr = POPPTR(ss,ix);
12152 TOPPTR(nss,ix) = any_dup(ptr, proto_perl);
12153 sv = (const SV *)POPPTR(ss,ix);
12154 TOPPTR(nss,ix) = sv_dup_inc(sv, param);
12156 case SAVEt_SET_SVFLAGS:
12158 TOPINT(nss,ix) = i;
12160 TOPINT(nss,ix) = i;
12161 sv = (const SV *)POPPTR(ss,ix);
12162 TOPPTR(nss,ix) = sv_dup(sv, param);
12164 case SAVEt_RE_STATE:
12166 const struct re_save_state *const old_state
12167 = (struct re_save_state *)
12168 (ss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
12169 struct re_save_state *const new_state
12170 = (struct re_save_state *)
12171 (nss + ix - SAVESTACK_ALLOC_FOR_RE_SAVE_STATE);
12173 Copy(old_state, new_state, 1, struct re_save_state);
12174 ix -= SAVESTACK_ALLOC_FOR_RE_SAVE_STATE;
12176 new_state->re_state_bostr
12177 = pv_dup(old_state->re_state_bostr);
12178 new_state->re_state_reginput
12179 = pv_dup(old_state->re_state_reginput);
12180 new_state->re_state_regeol
12181 = pv_dup(old_state->re_state_regeol);
12182 new_state->re_state_regoffs
12183 = (regexp_paren_pair*)
12184 any_dup(old_state->re_state_regoffs, proto_perl);
12185 new_state->re_state_reglastparen
12186 = (U32*) any_dup(old_state->re_state_reglastparen,
12188 new_state->re_state_reglastcloseparen
12189 = (U32*)any_dup(old_state->re_state_reglastcloseparen,
12191 /* XXX This just has to be broken. The old save_re_context
12192 code did SAVEGENERICPV(PL_reg_start_tmp);
12193 PL_reg_start_tmp is char **.
12194 Look above to what the dup code does for
12195 SAVEt_GENERIC_PVREF
12196 It can never have worked.
12197 So this is merely a faithful copy of the exiting bug: */
12198 new_state->re_state_reg_start_tmp
12199 = (char **) pv_dup((char *)
12200 old_state->re_state_reg_start_tmp);
12201 /* I assume that it only ever "worked" because no-one called
12202 (pseudo)fork while the regexp engine had re-entered itself.
12204 #ifdef PERL_OLD_COPY_ON_WRITE
12205 new_state->re_state_nrs
12206 = sv_dup(old_state->re_state_nrs, param);
12208 new_state->re_state_reg_magic
12209 = (MAGIC*) any_dup(old_state->re_state_reg_magic,
12211 new_state->re_state_reg_oldcurpm
12212 = (PMOP*) any_dup(old_state->re_state_reg_oldcurpm,
12214 new_state->re_state_reg_curpm
12215 = (PMOP*) any_dup(old_state->re_state_reg_curpm,
12217 new_state->re_state_reg_oldsaved
12218 = pv_dup(old_state->re_state_reg_oldsaved);
12219 new_state->re_state_reg_poscache
12220 = pv_dup(old_state->re_state_reg_poscache);
12221 new_state->re_state_reg_starttry
12222 = pv_dup(old_state->re_state_reg_starttry);
12225 case SAVEt_COMPILE_WARNINGS:
12226 ptr = POPPTR(ss,ix);
12227 TOPPTR(nss,ix) = DUP_WARNINGS((STRLEN*)ptr);
12230 ptr = POPPTR(ss,ix);
12231 TOPPTR(nss,ix) = parser_dup((const yy_parser*)ptr, param);
12235 "panic: ss_dup inconsistency (%"IVdf")", (IV) type);
12243 /* if sv is a stash, call $class->CLONE_SKIP(), and set the SVphv_CLONEABLE
12244 * flag to the result. This is done for each stash before cloning starts,
12245 * so we know which stashes want their objects cloned */
12248 do_mark_cloneable_stash(pTHX_ SV *const sv)
12250 const HEK * const hvname = HvNAME_HEK((const HV *)sv);
12252 GV* const cloner = gv_fetchmethod_autoload(MUTABLE_HV(sv), "CLONE_SKIP", 0);
12253 SvFLAGS(sv) |= SVphv_CLONEABLE; /* clone objects by default */
12254 if (cloner && GvCV(cloner)) {
12261 mXPUSHs(newSVhek(hvname));
12263 call_sv(MUTABLE_SV(GvCV(cloner)), G_SCALAR);
12270 SvFLAGS(sv) &= ~SVphv_CLONEABLE;
12278 =for apidoc perl_clone
12280 Create and return a new interpreter by cloning the current one.
12282 perl_clone takes these flags as parameters:
12284 CLONEf_COPY_STACKS - is used to, well, copy the stacks also,
12285 without it we only clone the data and zero the stacks,
12286 with it we copy the stacks and the new perl interpreter is
12287 ready to run at the exact same point as the previous one.
12288 The pseudo-fork code uses COPY_STACKS while the
12289 threads->create doesn't.
12291 CLONEf_KEEP_PTR_TABLE
12292 perl_clone keeps a ptr_table with the pointer of the old
12293 variable as a key and the new variable as a value,
12294 this allows it to check if something has been cloned and not
12295 clone it again but rather just use the value and increase the
12296 refcount. If KEEP_PTR_TABLE is not set then perl_clone will kill
12297 the ptr_table using the function
12298 C<ptr_table_free(PL_ptr_table); PL_ptr_table = NULL;>,
12299 reason to keep it around is if you want to dup some of your own
12300 variable who are outside the graph perl scans, example of this
12301 code is in threads.xs create
12304 This is a win32 thing, it is ignored on unix, it tells perls
12305 win32host code (which is c++) to clone itself, this is needed on
12306 win32 if you want to run two threads at the same time,
12307 if you just want to do some stuff in a separate perl interpreter
12308 and then throw it away and return to the original one,
12309 you don't need to do anything.
12314 /* XXX the above needs expanding by someone who actually understands it ! */
12315 EXTERN_C PerlInterpreter *
12316 perl_clone_host(PerlInterpreter* proto_perl, UV flags);
12319 perl_clone(PerlInterpreter *proto_perl, UV flags)
12322 #ifdef PERL_IMPLICIT_SYS
12324 PERL_ARGS_ASSERT_PERL_CLONE;
12326 /* perlhost.h so we need to call into it
12327 to clone the host, CPerlHost should have a c interface, sky */
12329 if (flags & CLONEf_CLONE_HOST) {
12330 return perl_clone_host(proto_perl,flags);
12332 return perl_clone_using(proto_perl, flags,
12334 proto_perl->IMemShared,
12335 proto_perl->IMemParse,
12337 proto_perl->IStdIO,
12341 proto_perl->IProc);
12345 perl_clone_using(PerlInterpreter *proto_perl, UV flags,
12346 struct IPerlMem* ipM, struct IPerlMem* ipMS,
12347 struct IPerlMem* ipMP, struct IPerlEnv* ipE,
12348 struct IPerlStdIO* ipStd, struct IPerlLIO* ipLIO,
12349 struct IPerlDir* ipD, struct IPerlSock* ipS,
12350 struct IPerlProc* ipP)
12352 /* XXX many of the string copies here can be optimized if they're
12353 * constants; they need to be allocated as common memory and just
12354 * their pointers copied. */
12357 CLONE_PARAMS clone_params;
12358 CLONE_PARAMS* const param = &clone_params;
12360 PerlInterpreter * const my_perl = (PerlInterpreter*)(*ipM->pMalloc)(ipM, sizeof(PerlInterpreter));
12362 PERL_ARGS_ASSERT_PERL_CLONE_USING;
12363 #else /* !PERL_IMPLICIT_SYS */
12365 CLONE_PARAMS clone_params;
12366 CLONE_PARAMS* param = &clone_params;
12367 PerlInterpreter * const my_perl = (PerlInterpreter*)PerlMem_malloc(sizeof(PerlInterpreter));
12369 PERL_ARGS_ASSERT_PERL_CLONE;
12370 #endif /* PERL_IMPLICIT_SYS */
12372 /* for each stash, determine whether its objects should be cloned */
12373 S_visit(proto_perl, do_mark_cloneable_stash, SVt_PVHV, SVTYPEMASK);
12374 PERL_SET_THX(my_perl);
12377 PoisonNew(my_perl, 1, PerlInterpreter);
12382 PL_scopestack_name = 0;
12384 PL_savestack_ix = 0;
12385 PL_savestack_max = -1;
12386 PL_sig_pending = 0;
12388 Zero(&PL_debug_pad, 1, struct perl_debug_pad);
12389 # ifdef DEBUG_LEAKING_SCALARS
12390 PL_sv_serial = (((UV)my_perl >> 2) & 0xfff) * 1000000;
12392 #else /* !DEBUGGING */
12393 Zero(my_perl, 1, PerlInterpreter);
12394 #endif /* DEBUGGING */
12396 #ifdef PERL_IMPLICIT_SYS
12397 /* host pointers */
12399 PL_MemShared = ipMS;
12400 PL_MemParse = ipMP;
12407 #endif /* PERL_IMPLICIT_SYS */
12409 param->flags = flags;
12410 /* Nothing in the core code uses this, but we make it available to
12411 extensions (using mg_dup). */
12412 param->proto_perl = proto_perl;
12413 /* Likely nothing will use this, but it is initialised to be consistent
12414 with Perl_clone_params_new(). */
12415 param->proto_perl = my_perl;
12416 param->unreferenced = NULL;
12418 INIT_TRACK_MEMPOOL(my_perl->Imemory_debug_header, my_perl);
12420 PL_body_arenas = NULL;
12421 Zero(&PL_body_roots, 1, PL_body_roots);
12424 PL_sv_objcount = 0;
12426 PL_sv_arenaroot = NULL;
12428 PL_debug = proto_perl->Idebug;
12430 PL_hash_seed = proto_perl->Ihash_seed;
12431 PL_rehash_seed = proto_perl->Irehash_seed;
12433 #ifdef USE_REENTRANT_API
12434 /* XXX: things like -Dm will segfault here in perlio, but doing
12435 * PERL_SET_CONTEXT(proto_perl);
12436 * breaks too many other things
12438 Perl_reentrant_init(aTHX);
12441 /* create SV map for pointer relocation */
12442 PL_ptr_table = ptr_table_new();
12444 /* initialize these special pointers as early as possible */
12445 SvANY(&PL_sv_undef) = NULL;
12446 SvREFCNT(&PL_sv_undef) = (~(U32)0)/2;
12447 SvFLAGS(&PL_sv_undef) = SVf_READONLY|SVt_NULL;
12448 ptr_table_store(PL_ptr_table, &proto_perl->Isv_undef, &PL_sv_undef);
12450 SvANY(&PL_sv_no) = new_XPVNV();
12451 SvREFCNT(&PL_sv_no) = (~(U32)0)/2;
12452 SvFLAGS(&PL_sv_no) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
12453 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
12454 SvPV_set(&PL_sv_no, savepvn(PL_No, 0));
12455 SvCUR_set(&PL_sv_no, 0);
12456 SvLEN_set(&PL_sv_no, 1);
12457 SvIV_set(&PL_sv_no, 0);
12458 SvNV_set(&PL_sv_no, 0);
12459 ptr_table_store(PL_ptr_table, &proto_perl->Isv_no, &PL_sv_no);
12461 SvANY(&PL_sv_yes) = new_XPVNV();
12462 SvREFCNT(&PL_sv_yes) = (~(U32)0)/2;
12463 SvFLAGS(&PL_sv_yes) = SVp_IOK|SVf_IOK|SVp_NOK|SVf_NOK
12464 |SVp_POK|SVf_POK|SVf_READONLY|SVt_PVNV;
12465 SvPV_set(&PL_sv_yes, savepvn(PL_Yes, 1));
12466 SvCUR_set(&PL_sv_yes, 1);
12467 SvLEN_set(&PL_sv_yes, 2);
12468 SvIV_set(&PL_sv_yes, 1);
12469 SvNV_set(&PL_sv_yes, 1);
12470 ptr_table_store(PL_ptr_table, &proto_perl->Isv_yes, &PL_sv_yes);
12472 /* dbargs array probably holds garbage */
12475 /* create (a non-shared!) shared string table */
12476 PL_strtab = newHV();
12477 HvSHAREKEYS_off(PL_strtab);
12478 hv_ksplit(PL_strtab, HvTOTALKEYS(proto_perl->Istrtab));
12479 ptr_table_store(PL_ptr_table, proto_perl->Istrtab, PL_strtab);
12481 PL_compiling = proto_perl->Icompiling;
12483 /* These two PVs will be free'd special way so must set them same way op.c does */
12484 PL_compiling.cop_stashpv = savesharedpv(PL_compiling.cop_stashpv);
12485 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_stashpv, PL_compiling.cop_stashpv);
12487 PL_compiling.cop_file = savesharedpv(PL_compiling.cop_file);
12488 ptr_table_store(PL_ptr_table, proto_perl->Icompiling.cop_file, PL_compiling.cop_file);
12490 ptr_table_store(PL_ptr_table, &proto_perl->Icompiling, &PL_compiling);
12491 PL_compiling.cop_warnings = DUP_WARNINGS(PL_compiling.cop_warnings);
12492 if (PL_compiling.cop_hints_hash) {
12494 PL_compiling.cop_hints_hash->refcounted_he_refcnt++;
12495 HINTS_REFCNT_UNLOCK;
12497 PL_curcop = (COP*)any_dup(proto_perl->Icurcop, proto_perl);
12498 #ifdef PERL_DEBUG_READONLY_OPS
12503 /* pseudo environmental stuff */
12504 PL_origargc = proto_perl->Iorigargc;
12505 PL_origargv = proto_perl->Iorigargv;
12507 param->stashes = newAV(); /* Setup array of objects to call clone on */
12508 /* This makes no difference to the implementation, as it always pushes
12509 and shifts pointers to other SVs without changing their reference
12510 count, with the array becoming empty before it is freed. However, it
12511 makes it conceptually clear what is going on, and will avoid some
12512 work inside av.c, filling slots between AvFILL() and AvMAX() with
12513 &PL_sv_undef, and SvREFCNT_dec()ing those. */
12514 AvREAL_off(param->stashes);
12516 if (!(flags & CLONEf_COPY_STACKS)) {
12517 param->unreferenced = newAV();
12520 /* Set tainting stuff before PerlIO_debug can possibly get called */
12521 PL_tainting = proto_perl->Itainting;
12522 PL_taint_warn = proto_perl->Itaint_warn;
12524 #ifdef PERLIO_LAYERS
12525 /* Clone PerlIO tables as soon as we can handle general xx_dup() */
12526 PerlIO_clone(aTHX_ proto_perl, param);
12529 PL_envgv = gv_dup(proto_perl->Ienvgv, param);
12530 PL_incgv = gv_dup(proto_perl->Iincgv, param);
12531 PL_hintgv = gv_dup(proto_perl->Ihintgv, param);
12532 PL_origfilename = SAVEPV(proto_perl->Iorigfilename);
12533 PL_diehook = sv_dup_inc(proto_perl->Idiehook, param);
12534 PL_warnhook = sv_dup_inc(proto_perl->Iwarnhook, param);
12537 PL_minus_c = proto_perl->Iminus_c;
12538 PL_patchlevel = sv_dup_inc(proto_perl->Ipatchlevel, param);
12539 PL_apiversion = sv_dup_inc(proto_perl->Iapiversion, param);
12540 PL_localpatches = proto_perl->Ilocalpatches;
12541 PL_splitstr = proto_perl->Isplitstr;
12542 PL_minus_n = proto_perl->Iminus_n;
12543 PL_minus_p = proto_perl->Iminus_p;
12544 PL_minus_l = proto_perl->Iminus_l;
12545 PL_minus_a = proto_perl->Iminus_a;
12546 PL_minus_E = proto_perl->Iminus_E;
12547 PL_minus_F = proto_perl->Iminus_F;
12548 PL_doswitches = proto_perl->Idoswitches;
12549 PL_dowarn = proto_perl->Idowarn;
12550 PL_doextract = proto_perl->Idoextract;
12551 PL_sawampersand = proto_perl->Isawampersand;
12552 PL_unsafe = proto_perl->Iunsafe;
12553 PL_inplace = SAVEPV(proto_perl->Iinplace);
12554 PL_e_script = sv_dup_inc(proto_perl->Ie_script, param);
12555 PL_perldb = proto_perl->Iperldb;
12556 PL_perl_destruct_level = proto_perl->Iperl_destruct_level;
12557 PL_exit_flags = proto_perl->Iexit_flags;
12559 /* magical thingies */
12560 /* XXX time(&PL_basetime) when asked for? */
12561 PL_basetime = proto_perl->Ibasetime;
12562 PL_formfeed = sv_dup(proto_perl->Iformfeed, param);
12564 PL_maxsysfd = proto_perl->Imaxsysfd;
12565 PL_statusvalue = proto_perl->Istatusvalue;
12567 PL_statusvalue_vms = proto_perl->Istatusvalue_vms;
12569 PL_statusvalue_posix = proto_perl->Istatusvalue_posix;
12571 PL_encoding = sv_dup(proto_perl->Iencoding, param);
12573 sv_setpvs(PERL_DEBUG_PAD(0), ""); /* For regex debugging. */
12574 sv_setpvs(PERL_DEBUG_PAD(1), ""); /* ext/re needs these */
12575 sv_setpvs(PERL_DEBUG_PAD(2), ""); /* even without DEBUGGING. */
12578 /* RE engine related */
12579 Zero(&PL_reg_state, 1, struct re_save_state);
12580 PL_reginterp_cnt = 0;
12581 PL_regmatch_slab = NULL;
12583 /* Clone the regex array */
12584 /* ORANGE FIXME for plugins, probably in the SV dup code.
12585 newSViv(PTR2IV(CALLREGDUPE(
12586 INT2PTR(REGEXP *, SvIVX(regex)), param))))
12588 PL_regex_padav = av_dup_inc(proto_perl->Iregex_padav, param);
12589 PL_regex_pad = AvARRAY(PL_regex_padav);
12591 /* shortcuts to various I/O objects */
12592 PL_ofsgv = gv_dup_inc(proto_perl->Iofsgv, param);
12593 PL_stdingv = gv_dup(proto_perl->Istdingv, param);
12594 PL_stderrgv = gv_dup(proto_perl->Istderrgv, param);
12595 PL_defgv = gv_dup(proto_perl->Idefgv, param);
12596 PL_argvgv = gv_dup(proto_perl->Iargvgv, param);
12597 PL_argvoutgv = gv_dup(proto_perl->Iargvoutgv, param);
12598 PL_argvout_stack = av_dup_inc(proto_perl->Iargvout_stack, param);
12600 /* shortcuts to regexp stuff */
12601 PL_replgv = gv_dup(proto_perl->Ireplgv, param);
12603 /* shortcuts to misc objects */
12604 PL_errgv = gv_dup(proto_perl->Ierrgv, param);
12606 /* shortcuts to debugging objects */
12607 PL_DBgv = gv_dup(proto_perl->IDBgv, param);
12608 PL_DBline = gv_dup(proto_perl->IDBline, param);
12609 PL_DBsub = gv_dup(proto_perl->IDBsub, param);
12610 PL_DBsingle = sv_dup(proto_perl->IDBsingle, param);
12611 PL_DBtrace = sv_dup(proto_perl->IDBtrace, param);
12612 PL_DBsignal = sv_dup(proto_perl->IDBsignal, param);
12614 /* symbol tables */
12615 PL_defstash = hv_dup_inc(proto_perl->Idefstash, param);
12616 PL_curstash = hv_dup(proto_perl->Icurstash, param);
12617 PL_debstash = hv_dup(proto_perl->Idebstash, param);
12618 PL_globalstash = hv_dup(proto_perl->Iglobalstash, param);
12619 PL_curstname = sv_dup_inc(proto_perl->Icurstname, param);
12621 PL_beginav = av_dup_inc(proto_perl->Ibeginav, param);
12622 PL_beginav_save = av_dup_inc(proto_perl->Ibeginav_save, param);
12623 PL_checkav_save = av_dup_inc(proto_perl->Icheckav_save, param);
12624 PL_unitcheckav = av_dup_inc(proto_perl->Iunitcheckav, param);
12625 PL_unitcheckav_save = av_dup_inc(proto_perl->Iunitcheckav_save, param);
12626 PL_endav = av_dup_inc(proto_perl->Iendav, param);
12627 PL_checkav = av_dup_inc(proto_perl->Icheckav, param);
12628 PL_initav = av_dup_inc(proto_perl->Iinitav, param);
12630 PL_sub_generation = proto_perl->Isub_generation;
12631 PL_isarev = hv_dup_inc(proto_perl->Iisarev, param);
12633 /* funky return mechanisms */
12634 PL_forkprocess = proto_perl->Iforkprocess;
12636 /* subprocess state */
12637 PL_fdpid = av_dup_inc(proto_perl->Ifdpid, param);
12639 /* internal state */
12640 PL_maxo = proto_perl->Imaxo;
12641 if (proto_perl->Iop_mask)
12642 PL_op_mask = SAVEPVN(proto_perl->Iop_mask, PL_maxo);
12645 /* PL_asserting = proto_perl->Iasserting; */
12647 /* current interpreter roots */
12648 PL_main_cv = cv_dup_inc(proto_perl->Imain_cv, param);
12650 PL_main_root = OpREFCNT_inc(proto_perl->Imain_root);
12652 PL_main_start = proto_perl->Imain_start;
12653 PL_eval_root = proto_perl->Ieval_root;
12654 PL_eval_start = proto_perl->Ieval_start;
12656 /* runtime control stuff */
12657 PL_curcopdb = (COP*)any_dup(proto_perl->Icurcopdb, proto_perl);
12659 PL_filemode = proto_perl->Ifilemode;
12660 PL_lastfd = proto_perl->Ilastfd;
12661 PL_oldname = proto_perl->Ioldname; /* XXX not quite right */
12664 PL_gensym = proto_perl->Igensym;
12665 PL_preambleav = av_dup_inc(proto_perl->Ipreambleav, param);
12666 PL_laststatval = proto_perl->Ilaststatval;
12667 PL_laststype = proto_perl->Ilaststype;
12670 PL_ors_sv = sv_dup_inc(proto_perl->Iors_sv, param);
12672 /* interpreter atexit processing */
12673 PL_exitlistlen = proto_perl->Iexitlistlen;
12674 if (PL_exitlistlen) {
12675 Newx(PL_exitlist, PL_exitlistlen, PerlExitListEntry);
12676 Copy(proto_perl->Iexitlist, PL_exitlist, PL_exitlistlen, PerlExitListEntry);
12679 PL_exitlist = (PerlExitListEntry*)NULL;
12681 PL_my_cxt_size = proto_perl->Imy_cxt_size;
12682 if (PL_my_cxt_size) {
12683 Newx(PL_my_cxt_list, PL_my_cxt_size, void *);
12684 Copy(proto_perl->Imy_cxt_list, PL_my_cxt_list, PL_my_cxt_size, void *);
12685 #ifdef PERL_GLOBAL_STRUCT_PRIVATE
12686 Newx(PL_my_cxt_keys, PL_my_cxt_size, const char *);
12687 Copy(proto_perl->Imy_cxt_keys, PL_my_cxt_keys, PL_my_cxt_size, char *);
12691 PL_my_cxt_list = (void**)NULL;
12692 #ifdef PERL_GLOBAL_STRUCT_PRIVATE
12693 PL_my_cxt_keys = (const char**)NULL;
12696 PL_modglobal = hv_dup_inc(proto_perl->Imodglobal, param);
12697 PL_custom_op_names = hv_dup_inc(proto_perl->Icustom_op_names,param);
12698 PL_custom_op_descs = hv_dup_inc(proto_perl->Icustom_op_descs,param);
12700 PL_profiledata = NULL;
12702 PL_compcv = cv_dup(proto_perl->Icompcv, param);
12704 PAD_CLONE_VARS(proto_perl, param);
12706 #ifdef HAVE_INTERP_INTERN
12707 sys_intern_dup(&proto_perl->Isys_intern, &PL_sys_intern);
12710 /* more statics moved here */
12711 PL_generation = proto_perl->Igeneration;
12712 PL_DBcv = cv_dup(proto_perl->IDBcv, param);
12714 PL_in_clean_objs = proto_perl->Iin_clean_objs;
12715 PL_in_clean_all = proto_perl->Iin_clean_all;
12717 PL_uid = proto_perl->Iuid;
12718 PL_euid = proto_perl->Ieuid;
12719 PL_gid = proto_perl->Igid;
12720 PL_egid = proto_perl->Iegid;
12721 PL_nomemok = proto_perl->Inomemok;
12722 PL_an = proto_perl->Ian;
12723 PL_evalseq = proto_perl->Ievalseq;
12724 PL_origenviron = proto_perl->Iorigenviron; /* XXX not quite right */
12725 PL_origalen = proto_perl->Iorigalen;
12726 #ifdef PERL_USES_PL_PIDSTATUS
12727 PL_pidstatus = newHV(); /* XXX flag for cloning? */
12729 PL_osname = SAVEPV(proto_perl->Iosname);
12730 PL_sighandlerp = proto_perl->Isighandlerp;
12732 PL_runops = proto_perl->Irunops;
12734 PL_parser = parser_dup(proto_perl->Iparser, param);
12736 /* XXX this only works if the saved cop has already been cloned */
12737 if (proto_perl->Iparser) {
12738 PL_parser->saved_curcop = (COP*)any_dup(
12739 proto_perl->Iparser->saved_curcop,
12743 PL_subline = proto_perl->Isubline;
12744 PL_subname = sv_dup_inc(proto_perl->Isubname, param);
12747 PL_cryptseen = proto_perl->Icryptseen;
12750 PL_hints = proto_perl->Ihints;
12752 PL_amagic_generation = proto_perl->Iamagic_generation;
12754 #ifdef USE_LOCALE_COLLATE
12755 PL_collation_ix = proto_perl->Icollation_ix;
12756 PL_collation_name = SAVEPV(proto_perl->Icollation_name);
12757 PL_collation_standard = proto_perl->Icollation_standard;
12758 PL_collxfrm_base = proto_perl->Icollxfrm_base;
12759 PL_collxfrm_mult = proto_perl->Icollxfrm_mult;
12760 #endif /* USE_LOCALE_COLLATE */
12762 #ifdef USE_LOCALE_NUMERIC
12763 PL_numeric_name = SAVEPV(proto_perl->Inumeric_name);
12764 PL_numeric_standard = proto_perl->Inumeric_standard;
12765 PL_numeric_local = proto_perl->Inumeric_local;
12766 PL_numeric_radix_sv = sv_dup_inc(proto_perl->Inumeric_radix_sv, param);
12767 #endif /* !USE_LOCALE_NUMERIC */
12769 /* utf8 character classes */
12770 PL_utf8_alnum = sv_dup_inc(proto_perl->Iutf8_alnum, param);
12771 PL_utf8_ascii = sv_dup_inc(proto_perl->Iutf8_ascii, param);
12772 PL_utf8_alpha = sv_dup_inc(proto_perl->Iutf8_alpha, param);
12773 PL_utf8_space = sv_dup_inc(proto_perl->Iutf8_space, param);
12774 PL_utf8_cntrl = sv_dup_inc(proto_perl->Iutf8_cntrl, param);
12775 PL_utf8_graph = sv_dup_inc(proto_perl->Iutf8_graph, param);
12776 PL_utf8_digit = sv_dup_inc(proto_perl->Iutf8_digit, param);
12777 PL_utf8_upper = sv_dup_inc(proto_perl->Iutf8_upper, param);
12778 PL_utf8_lower = sv_dup_inc(proto_perl->Iutf8_lower, param);
12779 PL_utf8_print = sv_dup_inc(proto_perl->Iutf8_print, param);
12780 PL_utf8_punct = sv_dup_inc(proto_perl->Iutf8_punct, param);
12781 PL_utf8_xdigit = sv_dup_inc(proto_perl->Iutf8_xdigit, param);
12782 PL_utf8_mark = sv_dup_inc(proto_perl->Iutf8_mark, param);
12783 PL_utf8_X_begin = sv_dup_inc(proto_perl->Iutf8_X_begin, param);
12784 PL_utf8_X_extend = sv_dup_inc(proto_perl->Iutf8_X_extend, param);
12785 PL_utf8_X_prepend = sv_dup_inc(proto_perl->Iutf8_X_prepend, param);
12786 PL_utf8_X_non_hangul = sv_dup_inc(proto_perl->Iutf8_X_non_hangul, param);
12787 PL_utf8_X_L = sv_dup_inc(proto_perl->Iutf8_X_L, param);
12788 PL_utf8_X_LV = sv_dup_inc(proto_perl->Iutf8_X_LV, param);
12789 PL_utf8_X_LVT = sv_dup_inc(proto_perl->Iutf8_X_LVT, param);
12790 PL_utf8_X_T = sv_dup_inc(proto_perl->Iutf8_X_T, param);
12791 PL_utf8_X_V = sv_dup_inc(proto_perl->Iutf8_X_V, param);
12792 PL_utf8_X_LV_LVT_V = sv_dup_inc(proto_perl->Iutf8_X_LV_LVT_V, param);
12793 PL_utf8_toupper = sv_dup_inc(proto_perl->Iutf8_toupper, param);
12794 PL_utf8_totitle = sv_dup_inc(proto_perl->Iutf8_totitle, param);
12795 PL_utf8_tolower = sv_dup_inc(proto_perl->Iutf8_tolower, param);
12796 PL_utf8_tofold = sv_dup_inc(proto_perl->Iutf8_tofold, param);
12797 PL_utf8_idstart = sv_dup_inc(proto_perl->Iutf8_idstart, param);
12798 PL_utf8_idcont = sv_dup_inc(proto_perl->Iutf8_idcont, param);
12800 /* Did the locale setup indicate UTF-8? */
12801 PL_utf8locale = proto_perl->Iutf8locale;
12802 /* Unicode features (see perlrun/-C) */
12803 PL_unicode = proto_perl->Iunicode;
12805 /* Pre-5.8 signals control */
12806 PL_signals = proto_perl->Isignals;
12808 /* times() ticks per second */
12809 PL_clocktick = proto_perl->Iclocktick;
12811 /* Recursion stopper for PerlIO_find_layer */
12812 PL_in_load_module = proto_perl->Iin_load_module;
12814 /* sort() routine */
12815 PL_sort_RealCmp = proto_perl->Isort_RealCmp;
12817 /* Not really needed/useful since the reenrant_retint is "volatile",
12818 * but do it for consistency's sake. */
12819 PL_reentrant_retint = proto_perl->Ireentrant_retint;
12821 /* Hooks to shared SVs and locks. */
12822 PL_sharehook = proto_perl->Isharehook;
12823 PL_lockhook = proto_perl->Ilockhook;
12824 PL_unlockhook = proto_perl->Iunlockhook;
12825 PL_threadhook = proto_perl->Ithreadhook;
12826 PL_destroyhook = proto_perl->Idestroyhook;
12827 PL_signalhook = proto_perl->Isignalhook;
12829 #ifdef THREADS_HAVE_PIDS
12830 PL_ppid = proto_perl->Ippid;
12834 PL_last_swash_hv = NULL; /* reinits on demand */
12835 PL_last_swash_klen = 0;
12836 PL_last_swash_key[0]= '\0';
12837 PL_last_swash_tmps = (U8*)NULL;
12838 PL_last_swash_slen = 0;
12840 PL_glob_index = proto_perl->Iglob_index;
12841 PL_srand_called = proto_perl->Isrand_called;
12843 if (proto_perl->Ipsig_pend) {
12844 Newxz(PL_psig_pend, SIG_SIZE, int);
12847 PL_psig_pend = (int*)NULL;
12850 if (proto_perl->Ipsig_name) {
12851 Newx(PL_psig_name, 2 * SIG_SIZE, SV*);
12852 sv_dup_inc_multiple(proto_perl->Ipsig_name, PL_psig_name, 2 * SIG_SIZE,
12854 PL_psig_ptr = PL_psig_name + SIG_SIZE;
12857 PL_psig_ptr = (SV**)NULL;
12858 PL_psig_name = (SV**)NULL;
12861 /* intrpvar.h stuff */
12863 if (flags & CLONEf_COPY_STACKS) {
12864 /* next allocation will be PL_tmps_stack[PL_tmps_ix+1] */
12865 PL_tmps_ix = proto_perl->Itmps_ix;
12866 PL_tmps_max = proto_perl->Itmps_max;
12867 PL_tmps_floor = proto_perl->Itmps_floor;
12868 Newx(PL_tmps_stack, PL_tmps_max, SV*);
12869 sv_dup_inc_multiple(proto_perl->Itmps_stack, PL_tmps_stack,
12870 PL_tmps_ix+1, param);
12872 /* next PUSHMARK() sets *(PL_markstack_ptr+1) */
12873 i = proto_perl->Imarkstack_max - proto_perl->Imarkstack;
12874 Newxz(PL_markstack, i, I32);
12875 PL_markstack_max = PL_markstack + (proto_perl->Imarkstack_max
12876 - proto_perl->Imarkstack);
12877 PL_markstack_ptr = PL_markstack + (proto_perl->Imarkstack_ptr
12878 - proto_perl->Imarkstack);
12879 Copy(proto_perl->Imarkstack, PL_markstack,
12880 PL_markstack_ptr - PL_markstack + 1, I32);
12882 /* next push_scope()/ENTER sets PL_scopestack[PL_scopestack_ix]
12883 * NOTE: unlike the others! */
12884 PL_scopestack_ix = proto_perl->Iscopestack_ix;
12885 PL_scopestack_max = proto_perl->Iscopestack_max;
12886 Newxz(PL_scopestack, PL_scopestack_max, I32);
12887 Copy(proto_perl->Iscopestack, PL_scopestack, PL_scopestack_ix, I32);
12890 Newxz(PL_scopestack_name, PL_scopestack_max, const char *);
12891 Copy(proto_perl->Iscopestack_name, PL_scopestack_name, PL_scopestack_ix, const char *);
12893 /* NOTE: si_dup() looks at PL_markstack */
12894 PL_curstackinfo = si_dup(proto_perl->Icurstackinfo, param);
12896 /* PL_curstack = PL_curstackinfo->si_stack; */
12897 PL_curstack = av_dup(proto_perl->Icurstack, param);
12898 PL_mainstack = av_dup(proto_perl->Imainstack, param);
12900 /* next PUSHs() etc. set *(PL_stack_sp+1) */
12901 PL_stack_base = AvARRAY(PL_curstack);
12902 PL_stack_sp = PL_stack_base + (proto_perl->Istack_sp
12903 - proto_perl->Istack_base);
12904 PL_stack_max = PL_stack_base + AvMAX(PL_curstack);
12906 /* next SSPUSHFOO() sets PL_savestack[PL_savestack_ix]
12907 * NOTE: unlike the others! */
12908 PL_savestack_ix = proto_perl->Isavestack_ix;
12909 PL_savestack_max = proto_perl->Isavestack_max;
12910 /*Newxz(PL_savestack, PL_savestack_max, ANY);*/
12911 PL_savestack = ss_dup(proto_perl, param);
12915 ENTER; /* perl_destruct() wants to LEAVE; */
12918 PL_start_env = proto_perl->Istart_env; /* XXXXXX */
12919 PL_top_env = &PL_start_env;
12921 PL_op = proto_perl->Iop;
12924 PL_Xpv = (XPV*)NULL;
12925 my_perl->Ina = proto_perl->Ina;
12927 PL_statbuf = proto_perl->Istatbuf;
12928 PL_statcache = proto_perl->Istatcache;
12929 PL_statgv = gv_dup(proto_perl->Istatgv, param);
12930 PL_statname = sv_dup_inc(proto_perl->Istatname, param);
12932 PL_timesbuf = proto_perl->Itimesbuf;
12935 PL_tainted = proto_perl->Itainted;
12936 PL_curpm = proto_perl->Icurpm; /* XXX No PMOP ref count */
12937 PL_rs = sv_dup_inc(proto_perl->Irs, param);
12938 PL_last_in_gv = gv_dup(proto_perl->Ilast_in_gv, param);
12939 PL_defoutgv = gv_dup_inc(proto_perl->Idefoutgv, param);
12940 PL_chopset = proto_perl->Ichopset; /* XXX never deallocated */
12941 PL_toptarget = sv_dup_inc(proto_perl->Itoptarget, param);
12942 PL_bodytarget = sv_dup_inc(proto_perl->Ibodytarget, param);
12943 PL_formtarget = sv_dup(proto_perl->Iformtarget, param);
12945 PL_restartjmpenv = proto_perl->Irestartjmpenv;
12946 PL_restartop = proto_perl->Irestartop;
12947 PL_in_eval = proto_perl->Iin_eval;
12948 PL_delaymagic = proto_perl->Idelaymagic;
12949 PL_dirty = proto_perl->Idirty;
12950 PL_localizing = proto_perl->Ilocalizing;
12952 PL_errors = sv_dup_inc(proto_perl->Ierrors, param);
12953 PL_hv_fetch_ent_mh = NULL;
12954 PL_modcount = proto_perl->Imodcount;
12955 PL_lastgotoprobe = NULL;
12956 PL_dumpindent = proto_perl->Idumpindent;
12958 PL_sortcop = (OP*)any_dup(proto_perl->Isortcop, proto_perl);
12959 PL_sortstash = hv_dup(proto_perl->Isortstash, param);
12960 PL_firstgv = gv_dup(proto_perl->Ifirstgv, param);
12961 PL_secondgv = gv_dup(proto_perl->Isecondgv, param);
12962 PL_efloatbuf = NULL; /* reinits on demand */
12963 PL_efloatsize = 0; /* reinits on demand */
12967 PL_screamfirst = NULL;
12968 PL_screamnext = NULL;
12969 PL_maxscream = -1; /* reinits on demand */
12970 PL_lastscream = NULL;
12973 PL_regdummy = proto_perl->Iregdummy;
12974 PL_colorset = 0; /* reinits PL_colors[] */
12975 /*PL_colors[6] = {0,0,0,0,0,0};*/
12979 /* Pluggable optimizer */
12980 PL_peepp = proto_perl->Ipeepp;
12981 PL_rpeepp = proto_perl->Irpeepp;
12982 /* op_free() hook */
12983 PL_opfreehook = proto_perl->Iopfreehook;
12985 PL_stashcache = newHV();
12987 PL_watchaddr = (char **) ptr_table_fetch(PL_ptr_table,
12988 proto_perl->Iwatchaddr);
12989 PL_watchok = PL_watchaddr ? * PL_watchaddr : NULL;
12990 if (PL_debug && PL_watchaddr) {
12991 PerlIO_printf(Perl_debug_log,
12992 "WATCHING: %"UVxf" cloned as %"UVxf" with value %"UVxf"\n",
12993 PTR2UV(proto_perl->Iwatchaddr), PTR2UV(PL_watchaddr),
12994 PTR2UV(PL_watchok));
12997 PL_registered_mros = hv_dup_inc(proto_perl->Iregistered_mros, param);
12998 PL_blockhooks = av_dup_inc(proto_perl->Iblockhooks, param);
13000 /* Call the ->CLONE method, if it exists, for each of the stashes
13001 identified by sv_dup() above.
13003 while(av_len(param->stashes) != -1) {
13004 HV* const stash = MUTABLE_HV(av_shift(param->stashes));
13005 GV* const cloner = gv_fetchmethod_autoload(stash, "CLONE", 0);
13006 if (cloner && GvCV(cloner)) {
13011 mXPUSHs(newSVhek(HvNAME_HEK(stash)));
13013 call_sv(MUTABLE_SV(GvCV(cloner)), G_DISCARD);
13019 if (!(flags & CLONEf_KEEP_PTR_TABLE)) {
13020 ptr_table_free(PL_ptr_table);
13021 PL_ptr_table = NULL;
13024 if (!(flags & CLONEf_COPY_STACKS)) {
13025 unreferenced_to_tmp_stack(param->unreferenced);
13028 SvREFCNT_dec(param->stashes);
13030 /* orphaned? eg threads->new inside BEGIN or use */
13031 if (PL_compcv && ! SvREFCNT(PL_compcv)) {
13032 SvREFCNT_inc_simple_void(PL_compcv);
13033 SAVEFREESV(PL_compcv);
13040 S_unreferenced_to_tmp_stack(pTHX_ AV *const unreferenced)
13042 PERL_ARGS_ASSERT_UNREFERENCED_TO_TMP_STACK;
13044 if (AvFILLp(unreferenced) > -1) {
13045 SV **svp = AvARRAY(unreferenced);
13046 SV **const last = svp + AvFILLp(unreferenced);
13050 if (SvREFCNT(*svp) == 1)
13052 } while (++svp <= last);
13054 EXTEND_MORTAL(count);
13055 svp = AvARRAY(unreferenced);
13058 if (SvREFCNT(*svp) == 1) {
13059 /* Our reference is the only one to this SV. This means that
13060 in this thread, the scalar effectively has a 0 reference.
13061 That doesn't work (cleanup never happens), so donate our
13062 reference to it onto the save stack. */
13063 PL_tmps_stack[++PL_tmps_ix] = *svp;
13065 /* As an optimisation, because we are already walking the
13066 entire array, instead of above doing either
13067 SvREFCNT_inc(*svp) or *svp = &PL_sv_undef, we can instead
13068 release our reference to the scalar, so that at the end of
13069 the array owns zero references to the scalars it happens to
13070 point to. We are effectively converting the array from
13071 AvREAL() on to AvREAL() off. This saves the av_clear()
13072 (triggered by the SvREFCNT_dec(unreferenced) below) from
13073 walking the array a second time. */
13074 SvREFCNT_dec(*svp);
13077 } while (++svp <= last);
13078 AvREAL_off(unreferenced);
13080 SvREFCNT_dec(unreferenced);
13084 Perl_clone_params_del(CLONE_PARAMS *param)
13086 /* This seemingly funky ordering keeps the build with PERL_GLOBAL_STRUCT
13088 PerlInterpreter *const to = param->new_perl;
13090 PerlInterpreter *const was = PERL_GET_THX;
13092 PERL_ARGS_ASSERT_CLONE_PARAMS_DEL;
13098 SvREFCNT_dec(param->stashes);
13099 if (param->unreferenced)
13100 unreferenced_to_tmp_stack(param->unreferenced);
13110 Perl_clone_params_new(PerlInterpreter *const from, PerlInterpreter *const to)
13113 /* Need to play this game, as newAV() can call safesysmalloc(), and that
13114 does a dTHX; to get the context from thread local storage.
13115 FIXME - under PERL_CORE Newx(), Safefree() and friends should expand to
13116 a version that passes in my_perl. */
13117 PerlInterpreter *const was = PERL_GET_THX;
13118 CLONE_PARAMS *param;
13120 PERL_ARGS_ASSERT_CLONE_PARAMS_NEW;
13126 /* Given that we've set the context, we can do this unshared. */
13127 Newx(param, 1, CLONE_PARAMS);
13130 param->proto_perl = from;
13131 param->new_perl = to;
13132 param->stashes = (AV *)Perl_newSV_type(to, SVt_PVAV);
13133 AvREAL_off(param->stashes);
13134 param->unreferenced = (AV *)Perl_newSV_type(to, SVt_PVAV);
13142 #endif /* USE_ITHREADS */
13145 =head1 Unicode Support
13147 =for apidoc sv_recode_to_utf8
13149 The encoding is assumed to be an Encode object, on entry the PV
13150 of the sv is assumed to be octets in that encoding, and the sv
13151 will be converted into Unicode (and UTF-8).
13153 If the sv already is UTF-8 (or if it is not POK), or if the encoding
13154 is not a reference, nothing is done to the sv. If the encoding is not
13155 an C<Encode::XS> Encoding object, bad things will happen.
13156 (See F<lib/encoding.pm> and L<Encode>).
13158 The PV of the sv is returned.
13163 Perl_sv_recode_to_utf8(pTHX_ SV *sv, SV *encoding)
13167 PERL_ARGS_ASSERT_SV_RECODE_TO_UTF8;
13169 if (SvPOK(sv) && !SvUTF8(sv) && !IN_BYTES && SvROK(encoding)) {
13183 Passing sv_yes is wrong - it needs to be or'ed set of constants
13184 for Encode::XS, while UTf-8 decode (currently) assumes a true value means
13185 remove converted chars from source.
13187 Both will default the value - let them.
13189 XPUSHs(&PL_sv_yes);
13192 call_method("decode", G_SCALAR);
13196 s = SvPV_const(uni, len);
13197 if (s != SvPVX_const(sv)) {
13198 SvGROW(sv, len + 1);
13199 Move(s, SvPVX(sv), len + 1, char);
13200 SvCUR_set(sv, len);
13207 return SvPOKp(sv) ? SvPVX(sv) : NULL;
13211 =for apidoc sv_cat_decode
13213 The encoding is assumed to be an Encode object, the PV of the ssv is
13214 assumed to be octets in that encoding and decoding the input starts
13215 from the position which (PV + *offset) pointed to. The dsv will be
13216 concatenated the decoded UTF-8 string from ssv. Decoding will terminate
13217 when the string tstr appears in decoding output or the input ends on
13218 the PV of the ssv. The value which the offset points will be modified
13219 to the last input position on the ssv.
13221 Returns TRUE if the terminator was found, else returns FALSE.
13226 Perl_sv_cat_decode(pTHX_ SV *dsv, SV *encoding,
13227 SV *ssv, int *offset, char *tstr, int tlen)
13232 PERL_ARGS_ASSERT_SV_CAT_DECODE;
13234 if (SvPOK(ssv) && SvPOK(dsv) && SvROK(encoding) && offset) {
13245 offsv = newSViv(*offset);
13247 mXPUSHp(tstr, tlen);
13249 call_method("cat_decode", G_SCALAR);
13251 ret = SvTRUE(TOPs);
13252 *offset = SvIV(offsv);
13258 Perl_croak(aTHX_ "Invalid argument to sv_cat_decode");
13263 /* ---------------------------------------------------------------------
13265 * support functions for report_uninit()
13268 /* the maxiumum size of array or hash where we will scan looking
13269 * for the undefined element that triggered the warning */
13271 #define FUV_MAX_SEARCH_SIZE 1000
13273 /* Look for an entry in the hash whose value has the same SV as val;
13274 * If so, return a mortal copy of the key. */
13277 S_find_hash_subscript(pTHX_ const HV *const hv, const SV *const val)
13280 register HE **array;
13283 PERL_ARGS_ASSERT_FIND_HASH_SUBSCRIPT;
13285 if (!hv || SvMAGICAL(hv) || !HvARRAY(hv) ||
13286 (HvTOTALKEYS(hv) > FUV_MAX_SEARCH_SIZE))
13289 array = HvARRAY(hv);
13291 for (i=HvMAX(hv); i>0; i--) {
13292 register HE *entry;
13293 for (entry = array[i]; entry; entry = HeNEXT(entry)) {
13294 if (HeVAL(entry) != val)
13296 if ( HeVAL(entry) == &PL_sv_undef ||
13297 HeVAL(entry) == &PL_sv_placeholder)
13301 if (HeKLEN(entry) == HEf_SVKEY)
13302 return sv_mortalcopy(HeKEY_sv(entry));
13303 return sv_2mortal(newSVhek(HeKEY_hek(entry)));
13309 /* Look for an entry in the array whose value has the same SV as val;
13310 * If so, return the index, otherwise return -1. */
13313 S_find_array_subscript(pTHX_ const AV *const av, const SV *const val)
13317 PERL_ARGS_ASSERT_FIND_ARRAY_SUBSCRIPT;
13319 if (!av || SvMAGICAL(av) || !AvARRAY(av) ||
13320 (AvFILLp(av) > FUV_MAX_SEARCH_SIZE))
13323 if (val != &PL_sv_undef) {
13324 SV ** const svp = AvARRAY(av);
13327 for (i=AvFILLp(av); i>=0; i--)
13334 /* S_varname(): return the name of a variable, optionally with a subscript.
13335 * If gv is non-zero, use the name of that global, along with gvtype (one
13336 * of "$", "@", "%"); otherwise use the name of the lexical at pad offset
13337 * targ. Depending on the value of the subscript_type flag, return:
13340 #define FUV_SUBSCRIPT_NONE 1 /* "@foo" */
13341 #define FUV_SUBSCRIPT_ARRAY 2 /* "$foo[aindex]" */
13342 #define FUV_SUBSCRIPT_HASH 3 /* "$foo{keyname}" */
13343 #define FUV_SUBSCRIPT_WITHIN 4 /* "within @foo" */
13346 S_varname(pTHX_ const GV *const gv, const char gvtype, PADOFFSET targ,
13347 const SV *const keyname, I32 aindex, int subscript_type)
13350 SV * const name = sv_newmortal();
13353 buffer[0] = gvtype;
13356 /* as gv_fullname4(), but add literal '^' for $^FOO names */
13358 gv_fullname4(name, gv, buffer, 0);
13360 if ((unsigned int)SvPVX(name)[1] <= 26) {
13362 buffer[1] = SvPVX(name)[1] + 'A' - 1;
13364 /* Swap the 1 unprintable control character for the 2 byte pretty
13365 version - ie substr($name, 1, 1) = $buffer; */
13366 sv_insert(name, 1, 1, buffer, 2);
13370 CV * const cv = find_runcv(NULL);
13374 if (!cv || !CvPADLIST(cv))
13376 av = MUTABLE_AV((*av_fetch(CvPADLIST(cv), 0, FALSE)));
13377 sv = *av_fetch(av, targ, FALSE);
13378 sv_setpvn(name, SvPV_nolen_const(sv), SvCUR(sv));
13381 if (subscript_type == FUV_SUBSCRIPT_HASH) {
13382 SV * const sv = newSV(0);
13383 *SvPVX(name) = '$';
13384 Perl_sv_catpvf(aTHX_ name, "{%s}",
13385 pv_display(sv,SvPVX_const(keyname), SvCUR(keyname), 0, 32));
13388 else if (subscript_type == FUV_SUBSCRIPT_ARRAY) {
13389 *SvPVX(name) = '$';
13390 Perl_sv_catpvf(aTHX_ name, "[%"IVdf"]", (IV)aindex);
13392 else if (subscript_type == FUV_SUBSCRIPT_WITHIN) {
13393 /* We know that name has no magic, so can use 0 instead of SV_GMAGIC */
13394 Perl_sv_insert_flags(aTHX_ name, 0, 0, STR_WITH_LEN("within "), 0);
13402 =for apidoc find_uninit_var
13404 Find the name of the undefined variable (if any) that caused the operator o
13405 to issue a "Use of uninitialized value" warning.
13406 If match is true, only return a name if it's value matches uninit_sv.
13407 So roughly speaking, if a unary operator (such as OP_COS) generates a
13408 warning, then following the direct child of the op may yield an
13409 OP_PADSV or OP_GV that gives the name of the undefined variable. On the
13410 other hand, with OP_ADD there are two branches to follow, so we only print
13411 the variable name if we get an exact match.
13413 The name is returned as a mortal SV.
13415 Assumes that PL_op is the op that originally triggered the error, and that
13416 PL_comppad/PL_curpad points to the currently executing pad.
13422 S_find_uninit_var(pTHX_ const OP *const obase, const SV *const uninit_sv,
13428 const OP *o, *o2, *kid;
13430 if (!obase || (match && (!uninit_sv || uninit_sv == &PL_sv_undef ||
13431 uninit_sv == &PL_sv_placeholder)))
13434 switch (obase->op_type) {
13441 const bool pad = (obase->op_type == OP_PADAV || obase->op_type == OP_PADHV);
13442 const bool hash = (obase->op_type == OP_PADHV || obase->op_type == OP_RV2HV);
13445 int subscript_type = FUV_SUBSCRIPT_WITHIN;
13447 if (pad) { /* @lex, %lex */
13448 sv = PAD_SVl(obase->op_targ);
13452 if (cUNOPx(obase)->op_first->op_type == OP_GV) {
13453 /* @global, %global */
13454 gv = cGVOPx_gv(cUNOPx(obase)->op_first);
13457 sv = hash ? MUTABLE_SV(GvHV(gv)): MUTABLE_SV(GvAV(gv));
13459 else /* @{expr}, %{expr} */
13460 return find_uninit_var(cUNOPx(obase)->op_first,
13464 /* attempt to find a match within the aggregate */
13466 keysv = find_hash_subscript((const HV*)sv, uninit_sv);
13468 subscript_type = FUV_SUBSCRIPT_HASH;
13471 index = find_array_subscript((const AV *)sv, uninit_sv);
13473 subscript_type = FUV_SUBSCRIPT_ARRAY;
13476 if (match && subscript_type == FUV_SUBSCRIPT_WITHIN)
13479 return varname(gv, hash ? '%' : '@', obase->op_targ,
13480 keysv, index, subscript_type);
13484 if (match && PAD_SVl(obase->op_targ) != uninit_sv)
13486 return varname(NULL, '$', obase->op_targ,
13487 NULL, 0, FUV_SUBSCRIPT_NONE);
13490 gv = cGVOPx_gv(obase);
13491 if (!gv || (match && GvSV(gv) != uninit_sv))
13493 return varname(gv, '$', 0, NULL, 0, FUV_SUBSCRIPT_NONE);
13496 if (obase->op_flags & OPf_SPECIAL) { /* lexical array */
13499 AV *av = MUTABLE_AV(PAD_SV(obase->op_targ));
13500 if (!av || SvRMAGICAL(av))
13502 svp = av_fetch(av, (I32)obase->op_private, FALSE);
13503 if (!svp || *svp != uninit_sv)
13506 return varname(NULL, '$', obase->op_targ,
13507 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
13510 gv = cGVOPx_gv(obase);
13515 AV *const av = GvAV(gv);
13516 if (!av || SvRMAGICAL(av))
13518 svp = av_fetch(av, (I32)obase->op_private, FALSE);
13519 if (!svp || *svp != uninit_sv)
13522 return varname(gv, '$', 0,
13523 NULL, (I32)obase->op_private, FUV_SUBSCRIPT_ARRAY);
13528 o = cUNOPx(obase)->op_first;
13529 if (!o || o->op_type != OP_NULL ||
13530 ! (o->op_targ == OP_AELEM || o->op_targ == OP_HELEM))
13532 return find_uninit_var(cBINOPo->op_last, uninit_sv, match);
13536 if (PL_op == obase)
13537 /* $a[uninit_expr] or $h{uninit_expr} */
13538 return find_uninit_var(cBINOPx(obase)->op_last, uninit_sv, match);
13541 o = cBINOPx(obase)->op_first;
13542 kid = cBINOPx(obase)->op_last;
13544 /* get the av or hv, and optionally the gv */
13546 if (o->op_type == OP_PADAV || o->op_type == OP_PADHV) {
13547 sv = PAD_SV(o->op_targ);
13549 else if ((o->op_type == OP_RV2AV || o->op_type == OP_RV2HV)
13550 && cUNOPo->op_first->op_type == OP_GV)
13552 gv = cGVOPx_gv(cUNOPo->op_first);
13556 == OP_RV2HV ? MUTABLE_SV(GvHV(gv)) : MUTABLE_SV(GvAV(gv));
13561 if (kid && kid->op_type == OP_CONST && SvOK(cSVOPx_sv(kid))) {
13562 /* index is constant */
13566 if (obase->op_type == OP_HELEM) {
13567 HE* he = hv_fetch_ent(MUTABLE_HV(sv), cSVOPx_sv(kid), 0, 0);
13568 if (!he || HeVAL(he) != uninit_sv)
13572 SV * const * const svp = av_fetch(MUTABLE_AV(sv), SvIV(cSVOPx_sv(kid)), FALSE);
13573 if (!svp || *svp != uninit_sv)
13577 if (obase->op_type == OP_HELEM)
13578 return varname(gv, '%', o->op_targ,
13579 cSVOPx_sv(kid), 0, FUV_SUBSCRIPT_HASH);
13581 return varname(gv, '@', o->op_targ, NULL,
13582 SvIV(cSVOPx_sv(kid)), FUV_SUBSCRIPT_ARRAY);
13585 /* index is an expression;
13586 * attempt to find a match within the aggregate */
13587 if (obase->op_type == OP_HELEM) {
13588 SV * const keysv = find_hash_subscript((const HV*)sv, uninit_sv);
13590 return varname(gv, '%', o->op_targ,
13591 keysv, 0, FUV_SUBSCRIPT_HASH);
13595 = find_array_subscript((const AV *)sv, uninit_sv);
13597 return varname(gv, '@', o->op_targ,
13598 NULL, index, FUV_SUBSCRIPT_ARRAY);
13603 (o->op_type == OP_PADAV || o->op_type == OP_RV2AV)
13605 o->op_targ, NULL, 0, FUV_SUBSCRIPT_WITHIN);
13610 /* only examine RHS */
13611 return find_uninit_var(cBINOPx(obase)->op_first, uninit_sv, match);
13614 o = cUNOPx(obase)->op_first;
13615 if (o->op_type == OP_PUSHMARK)
13618 if (!o->op_sibling) {
13619 /* one-arg version of open is highly magical */
13621 if (o->op_type == OP_GV) { /* open FOO; */
13623 if (match && GvSV(gv) != uninit_sv)
13625 return varname(gv, '$', 0,
13626 NULL, 0, FUV_SUBSCRIPT_NONE);
13628 /* other possibilities not handled are:
13629 * open $x; or open my $x; should return '${*$x}'
13630 * open expr; should return '$'.expr ideally
13636 /* ops where $_ may be an implicit arg */
13640 if ( !(obase->op_flags & OPf_STACKED)) {
13641 if (uninit_sv == ((obase->op_private & OPpTARGET_MY)
13642 ? PAD_SVl(obase->op_targ)
13645 sv = sv_newmortal();
13646 sv_setpvs(sv, "$_");
13655 match = 1; /* print etc can return undef on defined args */
13656 /* skip filehandle as it can't produce 'undef' warning */
13657 o = cUNOPx(obase)->op_first;
13658 if ((obase->op_flags & OPf_STACKED) && o->op_type == OP_PUSHMARK)
13659 o = o->op_sibling->op_sibling;
13663 case OP_ENTEREVAL: /* could be eval $undef or $x='$undef'; eval $x */
13665 case OP_CUSTOM: /* XS or custom code could trigger random warnings */
13667 /* the following ops are capable of returning PL_sv_undef even for
13668 * defined arg(s) */
13687 case OP_GETPEERNAME:
13735 case OP_SMARTMATCH:
13744 /* XXX tmp hack: these two may call an XS sub, and currently
13745 XS subs don't have a SUB entry on the context stack, so CV and
13746 pad determination goes wrong, and BAD things happen. So, just
13747 don't try to determine the value under those circumstances.
13748 Need a better fix at dome point. DAPM 11/2007 */
13754 GV * const gv = gv_fetchpvs(".", GV_NOTQUAL, SVt_PV);
13755 if (gv && GvSV(gv) == uninit_sv)
13756 return newSVpvs_flags("$.", SVs_TEMP);
13761 /* def-ness of rval pos() is independent of the def-ness of its arg */
13762 if ( !(obase->op_flags & OPf_MOD))
13767 if (SvROK(PL_rs) && uninit_sv == SvRV(PL_rs))
13768 return newSVpvs_flags("${$/}", SVs_TEMP);
13773 if (!(obase->op_flags & OPf_KIDS))
13775 o = cUNOPx(obase)->op_first;
13781 /* if all except one arg are constant, or have no side-effects,
13782 * or are optimized away, then it's unambiguous */
13784 for (kid=o; kid; kid = kid->op_sibling) {
13786 const OPCODE type = kid->op_type;
13787 if ( (type == OP_CONST && SvOK(cSVOPx_sv(kid)))
13788 || (type == OP_NULL && ! (kid->op_flags & OPf_KIDS))
13789 || (type == OP_PUSHMARK)
13793 if (o2) { /* more than one found */
13800 return find_uninit_var(o2, uninit_sv, match);
13802 /* scan all args */
13804 sv = find_uninit_var(o, uninit_sv, 1);
13816 =for apidoc report_uninit
13818 Print appropriate "Use of uninitialized variable" warning
13824 Perl_report_uninit(pTHX_ const SV *uninit_sv)
13828 SV* varname = NULL;
13830 varname = find_uninit_var(PL_op, uninit_sv,0);
13832 sv_insert(varname, 0, 0, " ", 1);
13834 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
13835 varname ? SvPV_nolen_const(varname) : "",
13836 " in ", OP_DESC(PL_op));
13839 Perl_warner(aTHX_ packWARN(WARN_UNINITIALIZED), PL_warn_uninit,
13845 * c-indentation-style: bsd
13846 * c-basic-offset: 4
13847 * indent-tabs-mode: t
13850 * ex: set ts=8 sts=4 sw=4 noet: